PETS - QSDS in Early Childhood Development 2010

The sampling frame received from UNICEF and the departments involved in the study, was used to determine the sample allocation among different strata and to draw a representative sample of Grade R in public schools and community-based Early Childhood Development (ECD) facilities. The three provinces were considered separately and a sample size of 200 was initially designed and drawn independently for each province, as it was not the intention to analyze the overall sample as a homogenous group. The selected provinces were significantly different in terms of their characteristics, sophistication and diversity of community-based ECD facilities and Grade R at public school facilities, and the sample frames were not compatible in their structure and coverage. Province 1 selected for the sample was a rich province, Province 2 was a moderately poor one and Province 3 was a large and very poor province, thus the survey reflected the broad spectrum of experiences in ECD.

Each province was treated as an independent sample. Multistage stratified sampling technique was used. Within provinces, key stratification variables were district municipality, local municipality, Grade R in public schools and community-based ECD facilities. Further disaggregation was done using quintiles for Grade R in public schools and districts, and the distinction between subsidized and unsubsidized registered community based facilities. Initially only quintile 1, quintile 2 and quintile 3 were sampled in Grade R in public schools as explicit strata (as discussed later, the sample was later expanded to all five quintiles), while district municipality and local municipality were further used as implicit strata. (Hence, each explicit stratum was ordered according to the implicit stratification variables before the sample was drawn to ensure the best possible representative sample.)

The sample of community-based ECD facilities had to be divided according to their registration and subsidy status. This was further subdivided into registered and subsidized, and registered but non-subsidized ECD facilities in each of the three provinces. The sample sought also to be representative of all districts and municipalities within each province. For this reason the aforementioned two categories of community-based ECD facilities and district municipality were used as explicit stratification variables with local municipalities as implicit stratification. In Province 1, suburb was also included in the implicit stratification. When stratum sizes differ substantially, proportional allocation can give rise to unnecessarily large samples from the large strata and very small (not adequately represented) samples from the small strata. Thus in some cases, a power allocation was used, a procedure appropriate for surveys where the stratum population sizes vary considerably and there is a need for precise estimates at each stratum level.

It was agreed that a sample of thirty unregistered and unsubsidized ECD facilities should be drawn per province. As there was no sampling frame for this subgroup, a convenience sampling strategy was used. The major subgroups were: district and local municipality for two provinces, while for another province it was district, district municipality, local municipality and suburb.
After the fieldwork was completed, weights were calculated according to the sample realization and to reflect the distribution of the target population.

The initial survey sample excluded the top two quintiles, as these were not supposed to receive subsidies. It was later agreed that this would prevent an analysis of differences in quality of care between poorer and more affluent schools, thus the sampling issue had to be revisited. As many facilities had by that time already been informed of the impending survey, it was decided not to redraw the sample but to add an additional sample of Grade R schools in the top two quintiles, thus expanding the total sample somewhat, with resultant increases in field costs.

The survey analyzed 381 public schools containing Grade R classes in the three provinces. Of these schools, 107 were in Province 1,136 in Province 2 and 139 in Province 3.