Information Frictions, Migration, and Human Capital:: Evidence from Senegal - Immediate Effects

The first treatment group receives information on predicted personalized income changes and predicted income gains for having apprenticeship training when moving from rural to urban areas in Senegal. These estimates are based on a survey conducted in urban areas. The second treatment group receives information on legal migration routes to the United States. We aim to determine whether these treatments alter the migration intentions, behaviors, and aspirations of the treated individuals.

Treatment 1: Enumerators hand out a printed flyer and review its details with respondents. It provides personalized income changes comparing cities and villages for people with similar backgrounds and also information on potential earnings increases for individuals with apprenticeship training who relocate to a city. Additionally, the flyer informs respondents about a government-funded opportunity to learn a trade: the Fonds de Financement de la Formation Professionnelle et Technique, known as 3FPT.

Treatment 2: Enumerators hand out a printed flyer and review its details with respondents. It provides information on the U.S.\ green card lottery, including the application procedure, requirements, and support available from the study team.

2024-09-06

2024-10-31

migration- and education-related aspirations, income-related perceptions, see PAP

See page 10 of the PAP for the detailed description of the variables.

The survey sampling is given by the framework implemented for the Senegal Migration Panel (SMP), a representative rural sample of most of Senegal covering 144 villages. We exclude all urban communes and communes in the Dakar region as well as some areas bordering Gambia, Mali, or Guinea for security reasons (specifically, the departments of Bignona, Bounkiling, Medina Yoro Foulah, Saraya, Bakel, Salemata, and the district of Fongolembi), with claims concerning representativeness conditional on these exclusions.

In this pre-analysis plan, we outline the experimental intervention, the treatments, the survey sampling of the data collection in 2024, the treatment assignment, and the hypotheses and outcome measures for the immediate outcomes. The immediate outcomes pre-defined in this plan refer to intentions and aspirations regarding (international and internal) migration plans as well as regarding human capital and apprenticeship training. In the next wave of data collection in 2025, we plan to run a follow-up survey that will focus on the same outcomes as we have defined here, but additionally will also elicit information on whether the treatment-induced changes in intentions and aspirations have led to changes in migration behavior and human capital and training decisions. Informed by the immediate findings, we will later pre-register the analysis of those medium-run outcomes, including a mediation analysis.

Sampling was carried out based on a list of administrative units obtained from Senegal’s National Agency for Statistics and Demography, ANSD, with villages defined as settlements containing between 40 and 150 households. We exclude all urban communes and communes in the Dakar region as well as some areas bordering Gambia, Mali, or Guinea for security reasons (specifically, the departments of Bignona, Bounkiling, Medina Yoro Foulah, Saraya, Bakel, Salemata, and the district of Fongolembi), with claims concerning representativeness conditional on these exclusions. In total, the sampling frame covers 3,082 out of 3,463 rural villages across all regions and 35 of 41 departments outside of Dakar.
Villages were randomly sampled within location-based strata}, namely within the district for focus departments and within the region for the national sample. We then randomly sample households within villages for in-depth surveys, including only those with at least one male member aged 18–40, which is the demographic group most likely to relocate for work from rural Senegal.

Within-village household samples were drawn in proportion to village size, with target sample sizes constrained to range from 15 to 45 households per village. We stratify based on prior migration by target individuals, undersampling those that had migrated within the previous year. Within strata, household selection probabilities are proportional to the number of within-household target individuals. Finally, we randomly select a target individual within each sampled household.

Randomization by computer

Individuals: respondents will be randomized into the two treatments and control. In addition, we keep some pure control villages, where nobody will be treated.

Yes

4097 individuals in 144 villages

4097 individuals in 144 villages

We collect data on 4,097 men aged 18-40. We assign 1,366 individuals to treatment 1, 1,366 to treatment 2, and 698 to the control group across 118 villages. Additionally, the study includes 668 individuals in 26 randomly assigned pure control villages.

For a fully clustered village-level randomization, we calculate an MDE of 0.16 standard deviations compared to the control mean. For individual-level randomization, the MDE is 0.11 standard deviations.