I conducted the list experiment with a nationally representative sample of 2,000 Kenyans. I designed the survey items and implemented them with the survey firm, Synovate, as part of their quarterly Social, Political, Economic, and Cultural Survey (SPEC). The surveywas conducted in Kiswahili by trained enumerators. Sampling points were selected through multistage cluster sampling. Using the 2009 census administrative boundaries, 55 of 69 districts were selected in the first stage. Sampling points—defined as the sub-location, the smallest administrative unit for which census data are available—were then selected using probability-proportion-to-size sampling, using data from the 2009 census.
After arriving at the selected sampling points, interviewers selected respondents in the following manner. First, interviewers identified a landmark at the sampling site (church, school, shop, and so on). Starting from the landmark, the interviewers used the summation method to select the first household to be sampled. This involves summing the last two digits of the date to determine the first household to be sampled. If the interviews took place on the 27th, the interviewers would have selected the 9th household on the left from the landmark. After the first household, the interviewers sampled every fourth household on the left. If an interview attempt was unsuccessful, the interviewers continued to the immediate next household. After selecting a household, interviewers enumerated all occupants of the household over 18 years of age and used the serial number on the questionnaire to select the respondent. If the respondent was not home, the interviewer would return to the household three times to find the respondent. If the respondent was not available after the three attempts, another member of the household was substituted in the following manner: on even dates, the person above the original respondent on the household list was chosen and on odd dates the person below was selected.