How to integrate disadvantaged workers such as immigrants and refugees into host-country labor markets is a pressing global question. Refugees may be prevented from entering local labor markets because employers have misperceptions or discriminatory attitudes about refugees’ skills and little incentive to gather information to correct these misperceptions or change their attitudes. This has motivated the design of several labor market policies aimed at reducing firms’ cost of gaining information about disadvantaged workers to improve these workers’ chances of employment and, ultimately, labor market efficiency. In this paper, we use a randomized experiment in Uganda – one of the five largest refugee-host countries in the world – to study the short- and longer-run impact on local firms’ willingness to hire refugees after being provided with a skilled refugee worker for free for one week. We find that treated firms hire three times as many refugees than firms in the control group eight months after the experiment. Data collected immediately after the experiment further show, consistent with a simple Bayesian learning model, exposure to a refugee led firm managers to update their beliefs about refugees’ skills in general. Yet, in the short-term, firms’ willingness to hire refugees, proxied by their willingness to offer a short-term job with a (generic) refugee, did not change on average. To investigate mechanisms for why exposure caused some firms to update their beliefs about refugees’ skills, and be willing to hire them, while others became less inclined to do so, we use a causal forest approach to estimate treatment heterogeneity. The algorithm identifies two predictors: employers’ initial attitudes toward refugees and refugee workers’ attitudes toward locals. We use these results to explore the importance of matching attitudes by estimating the variation in the treatment effect across four groups of employer-refugee pairs, distinguished by the attitude of the employer toward refugees and the attitude of the refugee toward locals. In line with a literature in social psychology, we find that positive matches, i.e., firms with a positive attitude toward refugees who were (randomly) matched with a refugee with positive attitudes toward locals, resulted in a substantial increase in firms’ willingness to hire a (generic) refugee worker, while negative matches decrease firms’ willingness to hire. Finally, we show that the treatment heterogeneity documented in the short-run, also helps explain the longer run results in real-world hiring. Our findings have important policy implications. Short-term exposure interventions can result in longer-term increases in employment for disadvantaged groups, but the size of this effect depends on the initial match quality.
