In France, as in many countries, several experiments conducted in recent years have revealed a relatively widespread discriminatory behaviour among employers (see Rich 2014, Baert 2017, Bertrand and Duflo 2016, and Neumark 2018 for literature reviews). Several correspondence studies have revealed that workers in France of ‘North African’ origin are particularly concerned by the phenomenon (Foroni and Cediey 2008, Petit et al. 2015, Berson 2013, and Edo and Jacquemet 2013 among others). With similar characteristics, applicants of French origin are up to three times more likely to receive positive call backs from employers than applicants of North African origin.
Several recent works have studied different actions to prevent discrimination against ethnic minorities. There is at this time no clear evidence that setting up an anonymous CV policy for some vacancies is effective for ethnic minorities (see Aslund and Skans 2012, Krause et al. 2012, and Behaghel et al. 2015). In another study, Edo and Jacquemet (2013) evaluate the impact of a satisfactory level of French on CVs. Their results are quite encouraging because such a signal makes it possible to reduce the differences observed. However, overall, few tools have proven their effectiveness in fighting discrimination in recruitment.
In a recent paper (Berson et al. 2019), we show that the organisation of recruitment in large companies is a tool to fight against hiring discrimination, notably because human resources (HR) professionals are better trained and have more time to select applicants. We compare the intervention of a centralised HR department to the selection made only within the establishment concerned by the position, generally by a manager responsible for recruitment.
An original and rich dataset
In order to study this important question, we rely on original data from a correspondence study aiming at measuring hiring discrimination linked to the North African origin of workers (versus French origin). The experiment was conducted between April and July 2016 in France by ISM-Corum and the Dares (Ministry of Labour). It covered ten occupations, at both employee and manager levels. A total of 1,500 fictitious applications have been sent, equally distributed between female and male young applicants.
Compared to previous correspondence studies, the originality of the data is twofold. First, unlike most experiments of this type in which each company is generally tested only once, this experiment comprised responding to several dozen offers per company, so that the correspondence test focuses on only 40 randomly selected large multi-establishment companies. The second original feature stems from the first one – all companies were subsequently met after the experiment, so that a certain amount of information on the organisation of recruitment of each company has been collected. In particular, for each application, we know whether the selection was made by a HR department at a centralised level of the company or just within the establishment concerned. We observe that for almost two-thirds of the tests, the selection was carried out via a centralised HR service. However, in most cases, this practice varies within companies, depending in particular on the qualification of the job. These features allow us to study the link between the organisation of recruitment and hiring discrimination. According to the literature, no other correspondence studies have taken such a direction.
North African applications are discriminated against
Just over half (50.8%) of the tests received at least a positive reply from the employer (Figure 1). Recruiters expressed interest in both applications in approximately one-third of the cases, and no response was received by either applicant in approximately another third of the cases. The other situations – where the French application was favoured and where the two applications were refused (listed here in order of importance) – represent slightly more than one quarter of the cases. The situation where the North African application alone interested the recruiter is relatively rare, as are the cases where one of the two applications is refused and the other remains unresponsive.
Figure 1 Responses distribution for all tests
The success rates of both applications are clearly different, with 47% of the 1,433 French applications considered in this study of interest to the recruiters against 36.7% for the North African applications (Figure 2). Thus, French applicants receive approximately 1.3 times more positive responses than North African applicants. This result means that French applicants must send two applications to hope for one positive response, and North African applicants must send three applications to hope for one positive response.
Centralised human resources as a tool to reduce discrimination
When we consider the gap corrected for potential composition effects (occupation, age, education, sex, etc.), we notice very few differences overall compared with the raw gaps, either at the aggregate level or in subsamples by sex. This result is valid even when location or the company-related effects are considered. However, splitting the sample between companies recruiting at the establishment level and those using a centralised HR department, the gap between both success rates is significantly different. Whereas 41% of French applicants receive a positive answer from a firm with a local recruitment process, only 25% of North African counterparts are contacted. When we consider firms belonging to a company using centralised HR, the gap remains between both applicants but is much lower (7% as opposed to 16%).
Figure 2 Success rate of French and North African applicants for the entire sample and according to the type of recruitment organisation
This result is mainly explained by the characteristics of HR professionals. They are better trained and more aware than other recruiters about discrimination. They are also less influenced by local constraints that can generate discrimination (e.g. conforming to consumer preferences or seeking to maintain homogeneous teams to facilitate their management). Finally, they have more time to devote to the selection of applications and therefore make their choices on the basis of ethnic stereotypes less often (Chugh 2004).
Using centralised HR department could be endogenous, that is, unobservable characteristics are likely to influence simultaneously the probability that the recruiter has a discriminatory attitude and that a centralised HR service is involved in recruitment. In other words, those companies might be places for which the discriminatory risk is the greatest and the effect of using the centralised HR department is lower. This endogeneity bias leads to an underestimation of the effect by means of an uncorrected estimate. Consequently, we use an instrumental strategy to correct it. Our instrumental variable is the existence (or not) of a franchise network for the company (or brand) to which the establishment concerned by the job offer belongs. This instrument is appropriate for two reasons. First, this variable is highly correlated with the probability that the firm’s establishments operate with some autonomy. Second, this variable does not affect the likelihood that the employer, that is, the parent company or franchise owner, will adopt a discriminatory attitude. We find similar results even after controlling for this bias.
To conclude, we show that the use of centralised HR services in the recruitment of applicants reduces the degree of discrimination for North African applicants. It means that acting on the organisation of recruitment in large companies can be an effective tool to combat discrimination, at least for the first stage of recruitment (i.e. before interviews). This concrete solution can be easily implemented by firms and confirms the importance of training and de-biasing to fight against discrimination.
Aslund, O, and O N Skans (2012), “Do anonymous job application procedures level the playing field?”, Industrial and Labor Relations Review, 65 (1), 82-107.
Baert, S (2017), “Hiring Discrimination: An Overview of (Almost) All Correspondence Experiments Since 2005”, IZA Discussion Paper no. 10738.
Behaghel, L, B Crépon, and T L Barbanchon (2015), “Unintended Effects of Anonymous Resumés”, American Economic Journal: Applied Economics, 7, 1–27.
Berson, C (2013), “Testing: les difficultés de l’interprétation des discriminations l’embauche”, Travail et Emploi, 135, 27–40.
Berson, C, M Laouenan, and E Valat (2018), “Outsourcing Recruitment as a Solution to Prevent Discrimination: A Correspondence Study”, IZA Discussion Paper no. 12132.
Bertrand, M, and E Duflo (2016), “Field Experiments on Discrimination”, CEPR Discussion Paper no. 11123.
Chugh, D (2004), “Societal and Managerial Implications of Implicit Social Cognition: Why Milliseconds Matter”, Social Justice research, 17 (2), 203-222.
Edo, A, and N Jacquemet (2013), “Discrimination l’embauche selon l’origine et le genre : défiance indifférenciée ou ciblée sur certains groupes?” Economie et Statistique, 464, 155–172.
Foroni, F, and E Cédiey (2008), “Résultats du testing sollicité par le groupe Casino”, ISM-CORUM.
Krause, A, U Rinne, and K F Zimmermann (2012), “Anonymous job applications of fresh PhD economists,” Economics Letter, 117 (2), 441–444.
Petit, P, E Duguet, and Y L’Horty (2015), “Discrimination résidentielle et origine ethnique : une étude expérimentale sur les serveurs en Ile-de-France”, Economie & Prévision, 206-207, 55–69.
Neumark, D (2018), “Experimental Research on Labor Market Discrimination”, Journal of Economic Literature, 56 (3), 799-866.
Rich, J (2014), “What Do Field Experiments of Discrimination in Markets Tell Us? A Meta Analysis of Studies Conducted since 2000”, IZA Discussion Paper no. 8584.