Opportunities Across Borders? The Impacts of Remote Work for Refugees

We will analyze data collected from an experiment commissioned by the World Bank and implemented by Na'amal, an NGO that gives remote work training and mentorship to refugees. The experiment has a two-part intervention:

The first part is a 3-week step-by-step online training on how to use remote work platforms such as Upwork. This intervention aims to alleviate informational and logistical barriers to technology adoption. The training teaches participants how to identify and apply for jobs and imparts best practices that can help them win contracts. To overcome logistical hurdles in the adoption of this technology, the training program will include online small-group coaching sessions and office hours where participants receive personalized feedback. The feedback aims to help participants set up and tailor their profiles to their skills and market demands, draft initial job proposals, and communicate with clients.

The second part is a professional mentorship program that will be paired with the above remote work training for a random subset of participants. The program aims to link participants to mentors who will meet with them one-on-one twice per month for 5 months. During that time, mentors function as a commitment device, encourage confidence, promote a growth-mindset, and reinforce professional skills such as time management. Additionally, mentors share norms around work and communication in a global and remote work setting. As such, the mentorship program aims to alleviate behavioral and cultural barriers to remote work adoption and success.

All participants, including the control group participants, will be given access to an asynchronous online training on soft skills needed for remote work.

2024-04-08

2025-04-01

- Take-up outcomes such as attendance and participation in the mentorship sessions, training, chat groups, and other aspects of the intervention.

- Knowledge on remote work platforms, how to set up a profile, general best practices, and expectations on earnings and work.

- Human capital investments and gained skills, including in technical skills, language skills, and soft skills.

- Actual activity on remote work platforms such as having a profile, applying for a job, number of jobs applied to, and time spent on remote work platforms.

- Wages and financial gains from remote work activity.

- Wages and financial gains from non-remote work activity (offline work).

- Mental health outcomes: PHQ-9 depression score

- Social integration within the refugee and host communities (including through a social integration index)

Social integration index (control mean 0, standard deviation 1): from sum of standardized components, with components reverse coded when necessary so that a higher index score is better. Components include:
- Number of [refugee/host] community members from among the 3 people who the respondent socializes with most frequently
- Total number of [refugee/host] community members the respondent socializes with regularly
- Number of [refugee/host] community members from among the 3 people who the respondent goes to or who go to the respondent for advice most frequently
- Total number of [refugee/host] community members that the respondent goes to or come to the respondent for advice
- Number of [refugee/host] community members from among the 3 people who the respondent goes to or who go to the respondent to borrow
- Total number of [refugee/host] community members that the respondent goes to or come to the respondent to borrow
- Indicator for whether the children in the household have any friends who are [native/refugee] children
- Indicator for whether the respondent has been invited to social gatherings such as birthdays or weddings of host country [natives / refugees] in the last 6 months

- Migration decisions and intentions
- Settlement and local investments outcomes
- Professional network and links within the host community
- Additional measures of social integration within the refugee and the host communities
- Additional employment outcomes both online and offline, including earnings, employment time, formalization, job search, job satisfaction, and underemployment.
- Savings and borrowing
- Behavioral outcomes, including self-efficacy
- Diffusion of remote work knowledge and opportunities among the respondent's network
- Additional mental health outcomes, including stress

- Generalized self-efficacy (GSE-6): Romppel et al. 2013’s shortened version of the Shwarzer and Jerusalem (1995) generalized self efficacy scale. Each item is scored on a scale of 1-4 (1=not at all true, 2=hardly true, 3=moderately true, 4=exactly true); the overall index is the sum of answers to all items

- Stress index: Stress index adapted from the Cohen’s Perceived Stress scale, following Hussam et al. (2022)

The design will be as follows: 2000 participants will be randomized across the study arms, stratifying on cohort, gender, region of participants' location, and refugees status. Eligible participants include (1) refugees or displaced individuals from any non-high income country, and (2) Kenyan host community members. Participants will be divided into the following groups:

a- 1400 participants will be chosen randomly to receive the online remote work training (described in the intervention section)

b- 400 of those participants will additionally receive the mentorship program (described in the intervention section)

c- 600 will serve as an untreated comparison group

All participants, including those in the comparison group, will have access to asynchronous online resources on soft skills used for remote work.

Controls will include baseline level of outcomes when available and variables selected through Lasso.
In addition to stratification variables, heterogeneity analysis will be conducted by variables revealed as predictive of treatment effects using machine learning predictions.

Finally, in addition to the main analysis, we will conduct additional analysis on the Kenyan sub-sample, comparing the outcomes of refugees to host community members.

Not available

Randomization done using Stata

Individuals

No

N/A

2000

Training only arm: 1400
Training + Mentorship arm: 400
Control arm: 600

Assuming a 15% attrition rate: power calculations indicate this will enable detecting 0.15 standard deviation impacts of the pooled treatment (a and b) relative to control group (c), and 0.18 standard deviation impacts of training plus mentoring (b) relative to training alone (a), with 80% power and alpha of 0.05. Assuming a 25% attrition rate: power calculations indicate minimum detectable effect sizes of 0.158 standard deviations for the pooled treatment (a and b) relative to control group (c), 0.192 standard deviations impacts of training plus mentoring (b) relative to training alone (a), with 80% power and alpha of 0.05.