Overcoming migration frictions through targeted mentoring, local endorsement, and network interactions in Senegal
How to cite
Ebert, Cara; Beber, Bernd; Riaz, Zara; Singh, Juni. 2023. "Overcoming migration frictions through targeted mentoring, local endorsement, and network interactions in Senegal." AEA RCT Registry. AEARCTR-0011268. https://doi.org/10.1257/rct.11268-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Developing countries are characterized by large productivity gaps between sectors and high specialization in agriculture in rural areas. Although domestic migration is common in these contexts, it is below levels that would equalize productivity gaps. In this project, we study the role of informational and social capital in home villages for migration decisions and outcomes. We test whether an intensive information treatment in form of two mentoring sessions leads to a change in migration perceptions, intent, behavior and experiences of the treated individual as well as untreated individuals in the village community. We use rich village-level network data to study information diffusion and spillovers in migration outcomes.
Intervention
The intervention consists of two one-on-one mentoring sessions that address target population-specific frictions associated with domestic migration in Senegal. The intervention is targeted at men aged 18 to 40 from rural villages across Senegal. Mentors are current migrants in Dakar who are from the same region as the study participants (but not from study villages), who have direct family members in their origin villages that have not migrated, and who maintain ties to the home community through regular visits.
The first mentoring session is conducted in person and consists of information and discussions about migration, with a focus on urban destinations in general and Dakar specifically. The exact content can be found in the appendix of the pre-analysis plan and was structured in nine sections:
1) Introduction to the mentoring session and a summary of the mentor's own migration history;
2) Statistics-based description of work opportunities as the primary reason to migrate within Senegal, and information on the typical earnings of migrants in comparison to non-migrants, together with a note on the typical cost items of migrants in urban areas and their higher prices;
3) Information about migrants' experiences with respect to the journey, work opportunities at destination, friends and social life at destination, and living conditions and accommodation at destination, all grounded in our prior survey estimates, as discussed below;
4) Discussion of the most common concerns about migration of people like the participant, again based on our prior survey results;
5) Advice on how to find housing at destination;
6) Comprehensive suggestions for how to find work at destination, and written specifics for complimentary job search support and coaching services (the session's longest section, providing information applicable to all urban centers as well as precise details about industrial clusters with potential employers, daily labor markets and support services available in Dakar);
7) Information about using bank or mobile money accounts for savings;
8) Tips on how to stay safe at destination; and
9) A recap of the session contents and transfer of a printed summary flyer.
The flyer contains an abbreviated version of the information about how to find housing, work, and coaching and job search assistance. We use icons, in addition to text, that were used during the mentoring session in order to maximize the flyer's usefulness for illiterate participants.
Statistical estimates shown to participants are drawn from an analysis of our survey data collected during the pre-intervention phase of this project. Figures are tailored to the target populations from which we draw representative samples: Rural villages and their residents across Senegal as well as in each of our three focus departments of Kaolack, Matam, and Sedhiou. This means that participants in a focus department received statistical estimates for their particular department, while participants elsewhere received national estimates. Qualitative insights shared with treated participants reflect focus group discussions with mentors, who are migrants in Dakar, as well as discussions with local stakeholders and partners.
Overall the in-person session lasts approximately 45 minutes. All sections of the mentoring session encourage active processing and engage the participant through graphical elements/icons, activities, questions, and discussion points.
The second mentoring session is conducted over the phone by the same mentors that implemented the first session. The second mentoring session does not provide new information but consists of an extensive recap of the contents of the first mentoring session, engagement with the participants' thoughts and questions about these materials, and a discussion of any potential migration intentions or plans of participants. Similar to the initial mentoring interaction, the second session contains a series of questions designed to maximize participants' active engagement with intervention content. The flyer is again shared with participants, this time in a digital format, preferably via WhatsApp or otherwise by way of four short text messages providing a high-level summary.
The mentoring intervention is implemented in two variations. The first variant (treatment 1) is the intervention as described so far. In the second variant (treatment 2), a local discussant/endorser is added to the first in-person mentoring session. The local discussant is someone who has migrated within the previous five years, has expressed his intent to migrate again in the future, is in the same demographic as our target population (male, between 18 and 40 years old), and occupies a non-peripheral place in the village's social network. We identify potential local discussants by drawing randomly from the set of study participants that meet these conditions. Prior to the mentoring session, the local discussant receives a briefing about the mentoring purpose, content, and his role in the meeting; is asked to add evidence and information based on his own migration experiences; and is tasked to be available for follow-up discussions or questions from the participant after the scheduled meeting. The local discussant receives a small monetary compensation for each mentoring session he attends.
Experimental design
First, we randomly assign the 145 study villages to receive the mentoring intervention (treatment 1, 52 villages), receive the mentoring intervention plus a local discussant (treatment 2, 52 villages), or remain as pure control (41 villages). Half of the treatment 2 villages are randomly assigned to have a highly central local discussant (with eigenvector centrality at or above the 75th percentile), while the social network centrality of the local discussant in the other half of the treatment 2 villages is close to the median (with eigenvector centrality at or between the 40th and 60th percentile).
We also randomly vary village-assigned treatment saturation. About three-quarters of the villages in each treatment condition (80 villages in total) are assigned to either a high or low saturation condition (with 80\% of subjects treated in half of these villages and 20\% treated in the other half), with the remaining 24 treated villages placed in a medium saturation condition (50\% of subjects treated).
Assignment of villages is carried out within five strata: one for the national sample, one for each focus department sample (Kaolack, Matam, and Sedhiou), and one for Eclosio-related villages in Diourbel and Thies. We use a re-randomization algorithm to ensure balance across baseline-measured village-level covariates.
Second, we randomly assign sampled individuals within villages to treatment, with an individual's assignment probabilities reflecting his village's treatment type and saturation level. Assignment is done within 104 strata, one for each village. We again use a re-randomization algorithm to establish balance in individual-level covariates at baseline.
Design details
Cited by
No citation data available for this registration yet.