Our sampling frame starts with 100 health centers sampled from among 20 provinces (out of a total of 45 provinces in Burkina Faso). Each of these health centers caters to at least 5 villages (and up to 8). We sampled exactly 5 villages in the catchment area of each health center, for a total of 500 villages.
The first randomization is done at the health center level: 50 health centers are randomly selected for the “Vouchers for free contraceptives” intervention. The other 50 health centers will be involved in a “10% discount” voucher intervention. The second randomization is done across villages within each health center stratum. The 5 villages in the stratum are randomly assigned to one of 5 groups: debates, debates with mortality information, village showing of edutainment video, no village-level intervention with medium-saturation household-level interventions, and no village-level intervention with low-saturation household-level interventions
The third randomization is done across households within villages not sampled for any village-level intervention. Households within these villages will be randomized to receive either mortality information, home viewing of the edutainment film, or nothing. The difference between medium and low saturation villages is in the share of households treated. In medium saturation, we assign 30% of households to the mortality info, 30% of households to edutainment, and 40% to be pure controls. In low saturation villages, these will be 15%, 15% and 70%, respectively.