Back to contents index

Calculating the preliminary estimates

Once the PES fieldwork was completed it was possible to arrive at a preliminary estimate of the extent of undercount in Census ’96, using the PES questionnaires in the way described below.

Capture of the PES data was subcontracted to an external agency. Preliminary estimates to determine the extent of undercount relied on the respondents, and what they said about whether or not the household itself, and each person in the household, had been counted during the census. There were two types of undercount:

• In one type, the respondent could indicate that a member of the household had been missed even though the other members of the household had been counted.

• The other type comprised persons in households which, according to the PES respondent, had been entirely missed during the census.

In order to obtain the preliminary estimate of the size of the population of South Africa, Stats SA did the following:

• It drew a suitable sample of EAs, and then a sample of census questionnaires from each EA, and counted the number of people in these questionnaires.

• It then raised this count to arrive at an estimate of the total census count.

• Finally, it adjusted this estimate with the preliminary findings about undercount in the PES, allowing for different types of settlements (formal and informal urban areas, tribal areas, commercial farms and other types of non-urban areas such as mission stations) in each province. These steps are explained in more detail in Census’96. Preliminary estimates of the size of the population of South Africa (Central Statistical Service, 1997).

In the course of attempting the next stage, i.e., the matching process, which has already been explained in Chapter 3, it became clear that the preliminary estimates of the extent of undercount were too low.

Table 13 indicates the differences between the preliminary and final counts, both unadjusted and adjusted, and the extent of undercount by province.

• Column 1 of the table lists all the provinces.

• Column 2, labeled ‘estimate’, indicates the preliminary estimate of the size of the population in each province and in South Africa as a whole before adjustment by the PES preliminary findings.

• Column 3, the adjusted estimate, shows the preliminary estimates of the population size in each province and in the country as a whole, after adjustment by the preliminary findings of the extent of undercount.

• Column 4 gives the preliminary estimate of the undercount, expressed as a percentage.

• Column 5, the raw count, is the actual number of people enumerated during Census ’96.

• Column 6, the adjusted count, is the count after adjustment by the final PES-based weighting procedure discussed in Chapter 5.

• Column 7 gives the final PES-based estimate of the extent of undercount in each province, expressed as a percentage.

There are two reasons why the preliminary estimates were too low. Firstly, preliminary calculations of undercount relied on answers of an informant in each household during the PES about which members of the household had been reported upon or omitted during the actual census. The informant may have been a different person from the one who gave the original information, and may not have known or reliably remembered what occurred.

• The matching process indicated that informants tended to under-reported those who had been missed.

• The overall, final extent, of undercount, after careful matching, increased from 6,8% to 10,7% (bottom row in Table 13).

• In certain provinces, particularly the more rural and the more sparsely populated ones, it increased more sharply than in the other provinces. In Northern Province, for example, the undercount increased from 3,9% to 11,3% between preliminary and final estimates, whilst in Eastern Cape it increased from 4,8% to 10,6%, and in Northern Cape from 9,0% to 15,6%.

Secondly, the preliminary estimates were based on a sample of questionnaires as these were being prepared for coding and data capture. Subsequently, it became clear that some of the provincial processing centres had not completed and submitted all their administrative documents at the time that the sample for the preliminary estimates was drawn. These questionnaires were, however, brought into account during the process of data capture, and were included in the final tally and the undercount calculations. This increased the final totals.

• The people counted in the sample of questionnaires used for the preliminary estimates indicated a total of 35,3 million before the PES adjustment, on the basis of information available at the time from too few administrative forms (bottom row of Table 13).

• But the final raw count indicates that 36,3 million people were counted in Census ’96, approximately one million more than indicated by the preliminary sample.

• The provinces in which the most additional questionnaires were found, and included in the final estimates, were Northern Province (approximately 400 000 people), KwaZulu-Natal (approximately 300 000) and the North West (approximately 200 000).

The methodology used to arrive at the final estimate of undercount, and the final adjustment factors, is described in the following chapter.