1. Undercount in the census

There are a number of reasons why a person or dwelling may have been not reached in Census ’96. Some of these would be inevitable in any exercise of the magnitude of the population census. Other reasons reflect the fact that this was the first census of the new South Africa, covering the whole of the country, conducted with little time to prepare. People may have been missed because:

• They moved around during the period of the census and were difficult to contact.
• They mistakenly thought that they were included by the informant in another household.
• They were not included by the householder completing the questionnaire who may have thought that, for example, young babies need not be included in the census.
• They were concerned about the confidentiality of their data and declined to be interviewed or to fill in the questionnaire.
• They were concerned about security and denied access to enumerators (particularly in some more affluent urban areas).
• They were on a farm where the enumerator encountered difficulty gaining access, particularly in remote areas or where farmers were not co-operative.
• Their dwelling was missed by the enumerator.
• The area they lived in was not demarcated.

Information on the undercount in Census ’96 has two main purposes. First, the data is used to adjust the census count to produce estimates of the population of South Africa at the time of the census. Second, it is used to evaluate the census and identify improvements for the next census, particularly with respect to segments of the population subject to higher levels of undercount.

The final calculation of the undercount of persons, based on a detailed analysis of the PES, including a sophisticated process of matching and imputation, is described below. This indicated that 10,7% of the people in South Africa on the night of 9-10 October 1996 were not enumerated. The preliminary calculation of the undercount using an elementary analysis of the PES, was 6,8% (see Section 3.2).

The table overleaf presents the final undercount rates for each province. The ‘undercount rate’ is defined as the difference between the final estimate and the raw census count expressed as a percentage of the final count. Note that the raw census count used in that obtained after the completion of processing and differs from that estimated for the preliminary population estimates (35 296 000). The change in the raw census count was highest in Northern Province, where administrative problems affected the accuracy of the original estimate.

The table shows that the rate of undercount of persons varied between provinces, being lowest in Western Cape and highest in Northern Cape. In part, undercount was related to the proportion of the population living in urban areas: some provinces with a higher proportion of people living in urban areas had lower undercount rates (Gauteng, Western Cape), and vice versa (Eastern Cape, Northern Province). More information on undercount is presented in later sections. Undercount rates by type of area are presented in Section 3.4. Undercount rates by demographic characteristics are included in Section 4.

Undercount rates were also calculated for households. The census data on households are also important, because the questions asked of households (such as access to electricity, water and telephones) are vital for planning. The undercount rate for households in South Africa was 6,6%. The method and details of calculating undercount rates for households are presented in Section 3.5.

Back to Contents