Administrative areas of Kenya are divided into counties, sub-counties, divisions, locations, and sub-locations. Our sample was selected from 5 counties in three stages using data provided by the Kenya National Bureau of Statistics. We randomly select sub-locations from the universe of sub-locations in our 5 study counties, after excluding sub-locations in the bottom 5% or top 10% of the county-specific density distribution (population per square kilometer). We then randomly select one enumeration area (essentially, a village) within each sampled sub-location, after excluding villages with fewer than 50 households. We then randomly select approximately 30 households per village to form our experimental sample. Households were randomly chosen from an exhaustive list of households in each village which we collected prior to the experiment with the assistance of village leaders. We stratify selection by intended migration, oversampling households who report that they might send a migrant to a city within the next year.
Treatment to Control, Information, Spillover, Group Information, or Guide + Information was assigned at the village level, stratifying by county, the share of households in each village intending to migrate to Nairobi, and average village income. In villages assigned to Spillover, a simple random 2/3 of the household sample was assigned to receive information, while the remaining 1/3 was assigned to receive no information.