Can work contact improve social cohesion between refugees and locals? Evidence from an experiment in Uganda

Our main treatment is work contact which includes direct and indirect contact at the workplace. The different interventions are as follows:

Direct contact: In order to promote direct contact, we run a job placement program that assists displaced populations in finding jobs in Uganda. The program provided a one-week internship to skilled refugees in Ugandan firms that were willing to participate. In order to match refugees with firms, we first tested refugees' skills. The test is an official exam run by the Directorate of Industrial Training, the agency established by the Ministry of Education to be in charge of the vocational education curriculum in Uganda. We also had the support of two large refugee-led NGOs based in Kampala to organize the skills testing. Refugees who passed the test were randomly matched to firms in the same sector as the refugees' occupation. We offered a small subsidy to refugees for the internship that was substantial and equal to about 85% of the monthly median earnings of the refugees. Half of the subsidy was paid upon beginning the internship, the other half upon completing it. The internship lasted one week. Local workers are employees of the firm.
Indirect contact: The indirect contact took place through a video documentary that we shot in Kampala in March 2021. The video is a short 4-minute documentary about relatable and real-life characters from both groups: Elvis Zani, a Ugandan worker from Kampala -- to relate to local workers -- and Paul Kithima, an urban refugee worker in Kampala. We avoid mentioning the nationality of the refugee worker to make the character relatable to all refugees belonging to any nationality. Both workers work together in permaculture. We chose this specific case as we wanted the characters to work in a sector that does not belong to the direct contact treatment, in order to avoid any priming effect.

In the video, both characters talk about their experience working together, what they learned from each other, and what they think about refugees and Ugandans collaborating on the workplace. The video also has a musical background without lyrics that was piloted and it is relatable to all nationalities. Moreover, the video is in English with subtitles in 6 languages (the languages spoken by our respondents) and respondents could decide in which language they wanted the subtitles to be in. The video was piloted with both groups in June 2021 in order to make sure the main message was transmitted, and no other factors were seen as major points.

2021-10-01

2021-10-31

Social cohesion between refugees and local workers.

We use a compound definition of social cohesion that comprises three major indicators: attitudes towards the out-group, implicit (IATs) and explicit biases and behaviours in real and hypothetical scenarios.

For the analysis, we consider 1 treatment group, which comprises respondents that randomly received direct contact, indirect contact, or both. The control group is composed of refugee workers that are not matched to any firm and local workers that are not matched to work together with refugees, and workers that watch the placebo video.

There were 3 randomization stages. The first one randomized refugees and firms into direct contact or in the control group. We randomized pairs of firms and refugees working in the same sector: if the refugee was a hairdresser, she was matched to a beauty saloon, etc. The pair was randomly assigned to direct contact following a specific procedure as described in Loiacono and Silva-Vargas, 2023. The second randomization cross-randomized refugees and local workers into indirect contact or in the control group. Finally, due to the cross-randomization, some respondents received both interventions.

Individual

No

N/A

273 local workers and 377 refugees

Refugees. With the collaboration of refugee leaders and refugee-led organizations, we composed a database of 1,088 skilled refugees who were (i) job seekers, (ii) were not looking for jobs but were interested in applying to one if possible, or (iii) were not in permanent employment. We set an appointment and approached the respondents with two messages: first, to ask some questions regarding their skills and work experience; and second, to explain what the research program was and get consent for it.

The listing was conducted between February and April 2021. From this list, 1,019 refugees agreed to be registered for the program. The first part of the program took place between April 19th and April 24th and consisted of testing refugees on their skills. A final number of 537 refugee workers successfully passed the test of skills. After the skills testing, refugees were invited to participate in the baseline and reminded that some could receive a one-week of internship offer. For our final sample, we had to drop out refugees that never found a match (N=126).\footnote{That is, firms in the sample of \cite{loiaconosilvamatch} were not interested in hiring these refugees} Furthermore, we had an attrition at endline of 24 refugees. Our final sample is composed by 377 refugee workers.

Local workers: In June 2021, we conducted a listing survey with firms in Kampala, active in sectors that match the occupations of refugee workers. Using the Uganda Census of Establishment Data 2010, the team of enumerators was assigned to different parishes daily and was instructed to interview all the firms that fell within a sector of interest. Enumerators were instructed to (i) look for the owner, the manager, or any employee with faculty to make managerial decisions; and (ii) the owner must be a Ugandan national.

Due to a second wave of COVID-19 in the country, the activities stopped and resumed between September-October 2021, when new firms were recruited. A total of 1,196 firms were recruited but only 536 were willing to hire a refugee. To select local workers, the sampling procedure was: (i) if the firm had only one worker, we interviewed that worker; (ii) if the firm had more than one worker, we asked the owner or manager of the firm which workers were most likely to work in close contact with a new employee.

Since not all the firms in the sample had at least one worker, our final sample of local workers is 273. These are the workers present at baseline and endline. If the worker changed between the two surveys, we kept the baseline answers of the baseline worker, but use the endline replies of the new worker. For this reason, our results are representative of all local workers in the firm, and not of the individual local worker.