Sampling

Target population

There were two components to the Family Transfer Project sample. The main sample targeted a roughly 50% subsample of the MDICP-1 sample (prime-age sample). A smaller survey instrument was then administered to a subsample of these prime-age respondents' parents (parents sample).

 

Prime-age sample

The prime-age sample aimed to interview 750 MDICP-1 women respondents (and their husbands) across the three sites. Because the FTP was fielded at the same time as the MDICP qualitative follow-up, the field directors for the two projects -Alex Weinreb (FTP) and Eliya Zulu (MDICP)- collaborated in the sampling process. The FTP had the greater sampling demands so it essentially took priority in terms of sample selection. That is, starting with the assumption that it would successfully interview close to 95% of the women it targeted, the FTP purposively chose geographically proximate villages until it had sampled about 270 women in each site. The MDICP qualitative project and the FTP-related qualitative project then sampled from the remaining villages.

As it turned out, the initial sampling assumptions were too optimistic. Even with a gifting strategy assigning one bar of soap and 1 kilogram of kitchen salt to every respondent, nonresponse was higher than expected because of the high population mobility. So, while there were very few cases of people claiming to be too busy to be interviewed or overtly refusing, roughly 12% of the sampled women and 13% of the sampled men were not available for interviews. In total 1,257 complete interviews (717 women and 540 men) were carried out.

 

Parent's sample

Respondents in the prime-age sample were asked about their parents' survival and their place(s) of residence. Originally, it was then intended to sample a proportion of these parents from outside the research areas. Budgetary constraints, however, imposed a change of plans. The parents' sample, therefore, only included parents who were resident in the same areas as their children.

The sampling procedure was straightforward. As each prime-age questionnaire was brought in from the field, the checker noted whether any of the parents were still alive and, if so, where they lived. If the surviving parents qualified for the sample list, the checker also noted how many children the surviving mother had ever given birth to. In order to reduce the extent to which the parent's sample was dominated by those who had given birth to large numbers of children - ie. because their adult children were more likely to be represented in the prime-age sample - the sampled parents were chosen in inverse proportion to their completed parity.

In total, 225 women and 121 men were interviewed.

Quick links:

Kenya
Malawi
Survey data
Qualitative Projects
Data quality
People
Papers

last updated September 18, 2003

Please direct comments or questions about this site to sbignami@ssc.upenn.edu