Sampling

The first survey of the Kenya Diffusion and Ideational Change Project (KDICP-1) was carried out in 1994-5, and interviewed 925 ever-married women of childbearing age and 859 men (of which 672 were husbands of the currently married women ). In 1996-7 and in 2000, respectively, the second and third round of the survey (KDICP-2 and KDICP-3) followed-up the same respondents (if eligible), and also interviewed any new spouse.

All rounds of the KDICP were carried out in four sites in Nyanza Province, in south-west Kenya: Kawadghone, Owich, Wakula South and Obisa. When the KDICP began, all four sites lay within a single administrative district called South Nyanza District. Shortly before the second wave, this was divided into three new districts: Wakula South and Owich were put in the new Suba District, and Kawadghone and Obisa were put in the new Rachuonyo District.

The four sublocations were chosen according to a 2x2 research design that maximized variation according to two characteristics: the spatial extension of social networks and the presence of a Community Based Distribution (CBD) program. The sublocations were chosen to be as similar as possible in every other characteristic. Alan Ferguson, who had long experience with the implementation of, and research on, family planning programs in Nyanza, aided in the selection of the sites.

The KDICP-1 used a standard cluster sampling strategy, with clusters defined as villages. To select villages in each sublocation we asked miji kumis (village elders) in the sublocation to list the households in their village along with the names of the adult members of the households. Once the list was made, we asked the elders to draw slips representing the village from a hat until we reached enough villages to give us 250 women in each site and their husbands. In Obisa, the final sample for the KDICP-1 included 13 out of the 18 villages in the sublocation; in Kawadghone, 11 out of the 16 villages in the sublocation; in Owich, 10 out of the 21 villages in the sublocation (half from the lake, half from the hills). In Wakula South, there were only 5 villages, and all were included.

Once the villages were selected we worked with the elders to refine the list of adult members of the households in each village. Within each village, currently married women and their husbands were eligible respondents. The sampling list was de facto, only those living in the village at the time. The survey team found that these lists excluded few de facto residents of the village, because the village elders were likely to list the de jure residents of the village (including those who were away on a short- or long-term basis) and to omit women who headed their own households. In three villages in Obisa and in two villages in Owich men were oversampled to compensate for men who were away at the time of the fielwork.