Experimental Design
We conduct our study in an existing sample, which was selected and first contacted
in 2022. The sample includes a total of 145 rural villages, which are grouped in
four samples: A nationally representative sample of rural villages (35 villages);
representative samples in our three focus departments Kaolack, Matam,
and Sedhiou (30 villages each); and a set of 10 villages in Thies and Diourbel that
were targeted by the NGO Eclosio for a separate irrigation intervention as well as
10 matched villages in the same departments who did not benefit from the irrigation
intervention.In total, the sampling frame covered 3,082 out of 3,463
rural villages across all regions and 35 of 41 departments outside of Dakar.2
Villages were randomly sampled within location-based strata, namely
within district for focus departments and within region for the national sample. A
re-randomization algorithm was employed to ensure representativeness with respect to
a set of characteristics for which administrative data is available (logged population
and household counts, average household size, and the share of women at the village
level, and the average age and share of residents recorded as absent at the level of the
commune).
Within sampled villages, all households were contacted in July and August
2022 for completion of a roster, which includes individual migration histories,
followed by a questionnaire eliciting basic household characteristics and a net-
work module.
Within villages households were further randomly sampled for in-depth
surveys at the household and individual level conducted in September and October
2022. In-depth interviews were carried out with men aged 18–40 and the households
of which they are members, because this constitutes the demographic group in rural
Senegal most likely to relocate for work.Within-village household samples were drawn in proportion to village size, with a
minimum target sample size of 15 households and a maximum of 45 households per
village. Within villages selection was stratified based on roster-reported migration
experiences among individuals with whom the in-depth interview was conducted with,
such that those that have migrated either domestically or internationally in the previous
twelve months were undersampled roughly by a factor of two. Within each strata,household selection probabilities were proportional to the number of within-household
target individuals, i.e. the relevant household size metric. Finally, individual for the
in-depth interview was randomly selected within each sampled household.
In total 9,665 household interviews and 4,755 in-depth interviews were completed.For this study, we follow up the 4,775 male individuals aged 18 to 40 and their house-
holds of the 2022 in-depth survey. The survey will again consist of a household-level
section answered by the head or, if not available, a knowledgeable member of the house-
hold, and an individual-level section with the same male respondent of the 2022 survey.
The experiment will be implemented in the individual-level part of the survey.