Contact on the Job: An Experiment on Labor Market Discrimination in Brazil

Policy Context

Discrimination in the labor market against underrepresented minorities is a major concern in many low- and middle-income countries. Discrimination in the labor market not only raises concerns regarding equity and inclusion but also impedes productivity and economic growth (Colonnelli et al. 2024; Hjort 2014). In Brazil, the context of this study, existing evidence points to a large degree of discrimination based on skin color across various domains (Cornwell et al. 2017; Firpo et al. 2021; Gerard et al. 2021).

Despite the growing evidence base documenting discrimination both at the hiring stage and in interactions on the job, research that would allow to design effective policies to reduce such discrimination is still limited (Bertrand and Duflo 2017). Adrian and Vlastimil’s project aims to address this gap by evaluating an intervention that has the potential to significantly reduce discrimination in the labor market and inform policies in low- and middle-income contexts. The team conducts a field experiment with over 2,000 job seekers in Brazil that will test the following research questions: under which conditions do interactions between majority and minority groups at the workplace impact productivity and discrimination? To what extent is the heterogeneity in the effect of contact driven by learning about the outgroup’s performance (statistical discrimination), changes in preferences (taste-based discrimination), or learning about the outgroup’s non-work-related attributes?

Study Design

The team will conduct a field experiment in collaboration with the J-PAL Jobs and Opportunity office in Brazil. In the experiment, 2,000 people are recruited for a job training program and subsequently randomly divided into a control group (N=800) and two treatment groups (N=600, each). The control group will participate in the online training/screening session with another participant from their own racial and political background. In the treatment groups, participants will conduct the training/screening session jointly with another participant who is either of another race (N=600) or of a different political affiliation (N=600).

The main treatment variation allows the team to test whether the identity of the assigned partner affects three outcomes of interest: (i) productivity during the job training, (ii) preferences for discrimination in a series of incentivized co-worker choices, elicited through a follow-on survey after the training, and (iii) beliefs about the outgroup’s performance and characteristics.

Results and Policy Lessons

Forthcoming.