Each SASAS round is designed to yield a representative sample of 7,000 individuals aged 16 and older in households which are geographically spread across the country’s nine provinces. Each SASAS round of interviewing consists of a sample of 500 Population Census enumeration areas (EAs) as primary sampling units, stratified by province, geographical sub-type and majority population group. Fourteen visiting points are selected in each EA. One eligible respondent is randomly chosen per visiting point. A split sample design is used. Two questionnaires are administered to 3,500 target respondents each to allow for increased content.

Sampling procedure

SASAS has been designed to yield a representative sample of 3500 adult South African citizens aged 16 and older (with no upper age limit), in households geographically spread across the country's nine provinces. The sampling frame used for the survey was based on the 2011 census and a set of small area layers (SALs). Estimates of the population numbers for various categories of the census variables were obtained per SAL. In this sampling frame special institutions (such as hospitals, military camps, old age homes, schools and university hostels), recreational areas, industrial areas and vacant SALs were excluded prior to the drawing of the sample.

Small area layers (SALs) were used as primary sampling units and the estimated number of dwelling units (taken as visiting points) in the SALs as secondary sampling units. In the first sampling stage the primary sampling units (SALs) were drawn with probability proportional to size, using the estimated number of dwelling units in an SAL as measure of size. The dwelling units as secondary sampling units were defined as "separate (non-vacant) residential stands, addresses, structures, flats, homesteads, etc." In the second sampling stage a predetermined number of individual dwelling units (or visiting points) were drawn with equal probability in each of the drawn dwelling units. Finally, in the third sampling stage a person was drawn with equal probability from all 16 year and older persons in the drawn dwelling units.

Three explicit stratification variables were used, namely province, geographic type and majority population group. As stated earlier, within each stratum, the allocated number of primary sampling units (which could differ between different strata) was drawn using proportional to size probability sampling with the estimated number of dwelling units in the primary sampling units as measure of size. In each of these drawn primary sampling units, seven dwelling units were drawn. This resulted in a sample of 3500 individuals.

A list of the 500 drawn SALs were given to geographic information specialists (GIS) and maps were then created for each of the 500 areas, indicating certain navigational beacons such as schools, roads churches etc.

Selection of individuals

For each of the SASAS samples interviewers visited each visiting point drawn in the SALs (PSU) and listed all eligible persons for inclusion in the sample, that is all persons currently aged 16 years or older and resident at the selected visiting point. The interviewer then selected one respondent using a random selection procedure based on a Kish grid.

Weighting

The data were weighted to take account of the fact that not all units covered in the survey had the same probability of selection. The weighting reflected the relative selection probabilities of the individual at the three main stages of selection: visiting point (address), household and individual. In order to ensure representivity of smaller groups, i.e. Northern Cape residents or Indian/Asian people, weights needed to be applied. Person and household weights were benchmarked using the SAS CALMAR macro and province, population group, gender and 5 age groups (i.e. 16-24, 25-34, 35-49, 50-59 and 60 and older). These benchmark variables for persons and province and population group of the respondent in the household were selected due to their reliability and validity. The marginal totals for the benchmark variables were obtained from the 2015 mid-year population estimates as published by Statistics South Africa. The estimated South African population was therefore used as the target population.