Agincourt Integrated Family Survey 2002

Agincourt 2002:
Sample selection for the Agincourt Integrated Family Survey involved extracting a sample of households, stratified by age-eligibility for the Old Age Pension and nationality. Age eligibility is defined as having either a woman aged 60 and older or a man aged 65 and older in the household.
The frame from which the sampling wasprovided by the 2001 census round of the Agincourt demographic surveillance system.
The sampling design for Agincourt 2002 had the following characteristics:

1. The villages within the population were stratified into two: a set of "Mozambican villages" and a set of "South African" ones.

2. 11 villages were selected, 4 "Mozambican" and 7 "South African" ones. Villages were selected with probability proportional to number of households.

3. The households within each village were assigned to two strata: age-eligible households and non-age-eligible ones.

4. Within each selected Mozambican village twenty households were selected at random, ten from each stratum. In each South African village, thirty household were selected, twelve from the "age-eligible" stratum and eighteen from the "non-age-eligible" stratum.

The final sampling design was:
Strata SA villages Mozambican villages
age-eligible 84 40
not age-eligible 126 40

The project did not explicitly stratify on nationality, but did so implicitly via the villages. This was not only because of limited confidence in the "nationality" identifiers, but also because a clustered design seemed advisable in any event. Clustering was decided on because it would make the logistics of fieldwork easier and because cluster effects might be interesting in their own right. Location is related both to social factors ("Mozambican" villages are somewhat stigmatized) and to geographical ones (there are differences in microclimates and hence crop patterns and also in access to infrastructure). The design ensured that there were at least twenty households in each cluster which should ensure that some of these cluster effects could be taken into account. The final sample size was 290 households.

The sampling frame for the selection of the villages:
The Agincourt Demographic Surveillance System has information on 21 villages. These formed the initial population from which the sample was drawn. Prior to sampling four villages were excluded from consideration:

(1) The RDP Housing project - The reason for excluding this settlement was that the dwelling units are apparently not a "household" in the sense in which the project wanted to investigate. Apparently many households send one or two junior members to occupy the RDP house (to maintain a claim to the property), but most of the household activities remain centered on dwellings elsewhere within the study site. It seemed sensible, therefore, to exclude the housing project from consideration.

(2) Ireagh A, Ireagh B and Kildare A - The project team were requested by Mark Collinson, the fieldwork manager of the Agincourt Health and Population Unit, not to include these villages in the sample, since other studies would be run (or were currently running) in these villages. The grounds for exclusion were that overstressing villages might contaminate the field results from the studies and might make fieldwork in future studies more difficult.

There were no visible differences between the excluded villages and the other "South African" villages that did not fall within the range of sampling variability.

Stratification within villages by age-eligibility and selection of households:
All households within the selected villages were assigned to one of two strata: age-eligible and non-age-eligible. This was done on the basis of the information available in the data base. Age eligibility was calculated as at the 23rd January 2002 (the date on which the sampling was done). Sampling within each of the strata was done by means of simple random sampling, without replacement.