The survey was administered in Kaloleni, Magarini, and Ganze of Kilifi County, Kenya. These sub-counties are mainly rural, and many households consist of a large homestead with several small huts built on the compound. Often extended families live together on the compound and share daily tasks. In some homesteads, co-wives reside together, although not necessarily in the same dwelling. Many of these households engage in rainfed agriculture and often rely on water collected from multiple sources of water away from the compound for their domestic needs and livestock. The household survey had two components: a household questionnaire and an individual questionnaire. The household-level questionnaire was first administered to the primary respondent who makes the decisions and is the most informed about the household’s water choices. After administering the household-level questionnaire, each adult household member 18 years and older was interviewed privately. The individual questionnaire included survey tools developed as part of MAGNET focus on dimensions of women’s agency: control over assets and decision making.The data collection was supported by the REACH programme funded by UK Aid from the UK Foreign, Commonwealth and Development Office (FCDO) for the benefit of developing countries (Programme Code 201880), and by the Measures for Advancing Gender Equality (MAGNET) initiative funded by the Bill and Melinda Gates Foundation (Programme Code INV-005620).

Excluding the more populated, urban centers, sub-locations in Kilifi 3 sub-counties were selected purposively to target those rural sub-counties with higher prevalence of polygamous households and extended family households. Within each of the sub-counties, two locations were randomly selected: Kayafungo and Mwanamwinga in Kaloleni, Marafa and Bungale in Magarini, and Ganze and Sokoke in Ganze. Within each location, there are multiple sub-locations. The listing, conducted in August 2022, involved first visiting the official offices at the sub-locations for complete list of villages. This resulted in identifying 206 villages. Excluding the urban and peri-urban villages, 99 villages were randomly selected to list. Listing all the households in a village involved calling each village leader of the selected villages and requesting a list of all the households. The team then travelled to that village to confirm the list of households by meeting with the village leader, checking the list with other village members, and through visual observation. The survey sample was pre-established to include 700 households. From the 99 listed villages, 50 villages were randomly selected for the survey. Within these 50 villages, 350 households were randomly selected. For the other 350 households, the team first selected all polygamous households with co-wives that lived together in the same compound in the list that had not yet been selected. Then, the team randomly selected extended family households (i.e. those with mothers-in-law and daughters-in-law) from the list of remaining households. Sampling weights are used in the analysis to adjust for oversampling. To account for non-responses, replacement household lists were also created by random selection. The interviews were administered in person with a tablet (CAPI). In each of these households, the primary respondent and up to eight additional adults were interviews. The raw data has interviews from 717 households. There are a number of households where not all the adults were interviewed (more than 50 percent). We counted an individual as missing if they refused, were working until night but returning home in the evening, or away for just a short period. There are 662 households that we considered “complete.” Even within these, there are a number of households where one or more members were not available because of work or school commitments or disabilities or illness prevented them from being able to participate.