Back to contents index

The counting in October 1996 of all South Africans as equal citizens of a democracy, and the processing of the data collected during Census ’96, has been completed. On the night of 9-10 October 1996, there were 40,6 million people in South Africa. This total has been adjusted for ‘undercount’, using a post-enumeration survey (PES).

In a departure from past practice, Stats SA has not used a demographic model to adjust the census count. Instead, the numbers and percentages presented are based on empirical evidence. Both the actual count (Chapter 2) and the adjustments by the PES (Chapters 3 and 5) were arrived at by visiting households in enumerator areas throughout the country, as well as hostels and institutions, and obtaining information on the people living in them.

By contrast, less than half the households in South Africa were actually visited door-to-door in a controlled fashion during the 1991 census (see Chapter 7). Some large areas were ‘swept’ without previously being divided into enumerator areas (EAs), others were overflown and estimated, and yet others were imputed. Finally, the mix of results was adjusted by fitting the results to match a demographic model reaching back to 1970. Moreover, there are various assumptions upon which such a model is built regarding fertility and mortality rates, and net-migration. These are sensitive procedures, especially when applied over such a long period, and may prove to be wrong.

Not surprisingly, the results of Census ’96 differ from many of the assumptions applied and outcomes reached by the previous demographic modelling, and the corresponding population estimates. Instead, they resemble the findings of the last nationwide census in 1970, before any adjustments were made to it by demographic modelling. The use of actual data from the count, and the implementation of a PES adjustment process rather than a model, allows users to examine for themselves the outcome of how the information was collected. At the same time, it enables demographers and other interested parties to use the information for their own modelling.

The final PES-adjusted results also differ from preliminary estimates issued in June 1997 (Chapter 4). There are two reasons why the preliminary estimates were too low.

• Firstly, the simple analysis of the post-enumeration survey relied on the say-so of an informant in each household as to 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 during the census, and may not have known or reliably remembered what occurred.

• Secondly, the preliminary estimates were based on a sample of questionnaires as these were being readied for coding and capturing onto computers. 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 capturing them into computers, and could thus be included into the improved undercount calculations. This increased the final totals.

Of the total of 40,6 million people, 19,5 million (48,1%) are males and 21,1 million (51,9%) were females. KwaZulu-Natal has the largest population (8,4 million), followed by Gauteng (7,3 million). Northern Cape has the smallest (0,8 million) followed by Free State (2,6 million).

Table 5 gives the number and the percentage of males and females, and the total, in each province. The percentages add up to 100 across each row. For example, looking across the row for Eastern Cape, and then down the various columns to the level of that row, the third column from the left shows that 46,1% of people in the province were males, while the fifth column shows that 53,9% were females. This adds up to a total of 100%, shown in the seventh column.

• The table shows that in only one province, Gauteng, there were proportionately more men (51,0%) than women (49,0%).

• It also shows that Northern Province and Eastern Cape had the highest proportion of women (54,3% and 53,9% respectively) and the lowest proportion of men (45,7% and 46,1%) compared to the other provinces. The internal migration of young men from the more-rural to the more-urban provinces may, in part, explain this finding.

*    All numbers given in this report are adjusted by the PES and rounded to whole numbers.
      The totals may therefore differ slightly.

**  The percentages are rounded to the first decimal place, therefore they may not always add up to exactly 100.

The lower percentage of males compared to females found by the PES-adjusted counts of Census ’96 than would have been expected from demographic modelling raises a number of issues. It may also be questioned whether or not sufficient adjustments were made by the PES for hard-to-reach groups such as single males. These issues are discussed further in Chapter 7.

Table 5 also highlights differential internal migration among males and females.

Of the total population of 40,6 million, 21,8 million (53,7%) were living in urban areas (see the relevant definitions later in this chapter), and 18,8 million (46,3%) were living in non-urban areas.

Table 6 gives the number and percentage of people living in urban and non-urban areas in each province. The percentages add up to 100 across the rows.

• It shows that Gauteng contained the largest proportion of urban dwellers (97,0%), while Northern Province contained the smallest proportion (11,0%).

• Free State (68,6%) and Northern Cape (70,1%), which consist largely of small towns proclaimed as urban areas by the former government, have a larger-than-expected proportion of urban dwellers.

*    All numbers given in this report are adjusted by the PES and rounded to whole numbers.
      The totals may therefore differ slightly.

**  The percentages are rounded to the first decimal place, therefore they may not always add up to exactly 100.

The 1991 demographic model and its projections to 1996 assumed a lower rate of urbanisation in the country than indicated in the PES-adjusted Census ’96 counts. Since it is known that higher urbanisation and a decreasing fertility rate (defined as the number of live births a woman will, on average, have in her lifetime) are associated, the decline in the fertility rate in South Africa suggested by the count (see Chapter 7) may in part be due to a more rapid urbanisation process than modelling previously suggested.

Table 7 gives the number and percentage of people living in each province by population group. The percentages add up to 100 across the rows.

• It shows, in the bottom row, that 31,1 million (76,7%) of the South African population is African. Africans are in the majority in seven of the nine provinces.

• From the first column, one sees that the African majorities are highest at 96,7% in Northern Province and 91,2% in North West.

• By contrast, coloureds are the largest population group amongst those living in Western Cape (54,2%) and Northern Cape (51,8%).

• The province in which the proportion of whites is largest is Gauteng (23,2%); while the proportion of Indians is largest in KwaZulu-Natal (9,4%).

*      All numbers given in this report are adjusted by the PES and rounded to whole numbers.
        The totals may therefore differ slightly.

**    The percentages are rounded to the first decimal place, therefore they may not always add up to exactly 100.

***  Those who classified themselves as Griquas are included in the ‘Other/not stated’ category.
        There are approximately 8 000 of these people, found largely in Northern Cape.

Both the number and the proportions of whites in the country are lower than anticipated by the 1991 demographic model.

• This may be due in part to undeclared emigration (Stats SA finds that emigration figures from foreign governments indicate that two to three times more people emigrate than are indicated by control forms gathered by the Department of Home Affairs). This may also possibly be due to lower fertility and higher mortality amongst whites than modelled.

• Under-enumeration amongst the white population may be higher than estimated, even after the adjustments by the post-enumeration survey. In Census ’96, whites may have been difficult to reach because of suspicion of the process and difficulty in finding sufficient white enumerators. Further discussion on these problems is to be found in Chapter 7.

Table 8 shows the distribution of South Africans by province within each major population group in a different way. The numbers are the same as in Table 7, but the percentages add up to 100 down each column. For example, the third column from the left indicates that among Africans, 17,5% of those counted in Census ’96 live in Eastern Cape, while 7,1% live in Free State.

*       All numbers given in this report are adjusted by the PES and rounded to whole numbers.
         The totals may therefore differ slightly.

**     The percentages are rounded to the first decimal place, therefore they may not always add up to exactly 100.

***   Those who classified themselves as Griquas are included in the ‘Other/not stated’ category.
         There are approximately 8 000,people self-classified as Griqua, found largely in Northern Cape.

The table shows that Africans, and to a lesser extent whites, are more-widely dispersed across the country than coloureds and Indians:

• Relatively large proportions of the African population live in KwaZulu-Natal (22,1%), Eastern Cape (17,5%), Gauteng (16,5%) and Northern Province (15,3%), and there are appreciable proportions in three other provinces. The proportions are small only in Western Cape (2,7%) and Northern Cape (0,9%). This distribution may be a reflection of the labour policy of the past, where work preference was given to coloureds rather than Africans in these two parts of the country.

• Whites tend to live in provinces which are largely urbanised. The largest proportion of whites is found in Gauteng (38,4%), followed by Western Cape (18,5%) and KwaZulu-Natal (12,6%). There are also appreciable proportions in four other provinces.

• The coloured population is largely found in Western Cape (59,6%), but also to a lesser extent in Eastern Cape (13,0%), Northern Cape (12,1%) and Gauteng (7,7%). The proportions are much smaller in the other provinces.

• More than three-quarters of the Indian population is found in KwaZulu-Natal (75,6%), with the second largest proportion (15,4%) being found in Gauteng.

The age distribution of South Africans in five-year intervals is indicated in Table 9. It shows, in the fourth column from the left, that 34,3% of people in South Africa are under the age of 15 years, while 6,9% are aged 60 years or more. This distribution reflects that, in general, the country has a relatively young population. On the other hand, there is also a relatively large proportion of older people, which may be indicative of diverse age distributions within the country not only by gender, but also among the different population groups.

* Excluding unspecified.

When examining gender differences in age distribution, Figure 1 shows that a larger proportion of the male population (35,2%) compared to females (32,8%) is under the age of 15 years, and a smaller proportion of males (3,8%) compared to females (5,7%) is aged 65 years or more. This is indicative of a longer life expectancy among females than males.

The sex ratios (the proportion of males to every 100 females, as depicted in the right-most column of Table 9) are lower than expected from demographic assumptions in certain age categories, for example the 0-4 year age group and among young adults (aged between 20 and 35 years). These may possibly be due to difficulties in enumeration and PES adjustment, for example under-reporting of babies, and young single males being harder to reach than others. But they may partly be due to non-measurement factors, such as declining fertility and young men perhaps suffering higher mortality.

Figures 2 to 5 give the age distribution in five-year intervals for each of the major population groups in the country. They show differences in the age distribution of each population group, depicting various stages in the demographic transition from a younger to an older population.

• Among Africans, the overall trend shows that there is a relatively large proportion of males (37,4%) and females (34,8%) below the age of 15 years, and a relatively small proportion aged 65 years or more (3,2% of males and 5,0% of females).

• Among coloureds, a smaller proportion of the population (34,2% of males and 32,0% females) is under the age of 15 years, compared to Africans. On the other hand, an even smaller proportion of both males (3,1%) and females (4,2%) is aged 65 years or more.

• Among Indians, compared with Africans and coloureds, an appreciably smaller proportion (28,3% of males and 26,5% of females) is below the age of 15 years.

• Among whites, a yet smaller proportion is aged less than 15 years (22,0% of males and 20,1% of females). A far larger proportion compared to the other population groups is aged 65 years or more (8,7% of males and 18,8% of females.

The smaller-than-expected proportion of those aged 0-4 years in the African population could be the result of two main factors, or a combination of them. Firstly, the extent of undercount in this age category could have been underestimated in the PES adjustment. Secondly, the fertility rates of the African population could have been overestimated in demographic models derived from previous censuses. These factors will be discussed in more detail in Chapter 7.

There were 9,1 million households (excluding hostels and institutions) in South Africa on census night. Table 10 indicates the type of dwelling in which these households were living, broken down by population group of the head of household. The percentages add up to 100 down each column: For example, among African-headed households (second column), 44,9% were living in a formal dwelling such as a house or a block of flats, while 24,7% were living in traditional dwellings, and 21,2% in informal dwellings such as shacks.

• The proportions living in formal dwellings were much higher in households headed by coloureds (82,3%), Indians (91,0%) or whites (95,1%), compared with Africans (44,9%).

* All numbers given in this report are adjusted by the PES and rounded to whole numbers.
        The totals may therefore differ slightly.

**    The percentages are rounded to the first decimal place, therefore they may not always add up to exactly 100.

***  Including those who classified themselves as Griquas.

There were four phases in Census ’96.

• In the first phase, pre-enumeration, which included planning and demarcation, Stats SA divided the whole country into small geographic areas with clearly distinguishable boundaries called enumeration areas (EAs). Thereafter, a list was made of dwellings in each EA. In some areas, because of a lack of maps or aerial photographs, these boundaries were not clear, and had to be obtained from traditional authorities. At the same time as this phase, the development and testing of the census questionnaire was completed.

• In the second phase, enumeration, fieldworkers or enumerators visited the households in these EAs (some nine million in all) and ensured that a questionnaire was completed giving information on persons in the household. They also visited hostels and other institutions, including prisons, police cells, hotels and homes for the aged or disabled, as well as the homeless.

• In phase three, processing, the data from the census was coded and then captured on computer.

• In phase four, which is ongoing, information is being analysed and products are being produced and disseminated. This process of analysis, production and dissemination will continue over the next two years.

• A more detailed discussion is given of how the census was done is given in Chapter 2.

In an operation as vast and quick as the enumeration phase of the census, during which Statistics South Africa employed over 100 000 people for a month, drawn predominantly from the unemployed, some degree of incomplete enumeration, or undercount, is inevitable. A post-enumeration survey (PES) is an internationally recognised method for estimating the extent of undercount in a census and then adjusting the totals.

In 1996, Stats SA conducted a PES immediately after the census. A sample of 800 EAs was drawn, spread across the country. In each EA in the sample, the boundaries were checked and new lists of dwellings were compiled. Then a short questionnaire was administered to all the households in the sampled EAs. The households and people in the PES were matched against the census questionnaire counterparts according to the answers which they gave in the PES.

Table 11 gives the unadjusted tallies, the percentage undercount and the final PES-adjusted estimates of the population size and the number of households in each province and in the country as a whole. For example, in Eastern Cape, 5,6 million people were counted. The undercount was estimated from the PES to be 10,6% for the province. When the raw count was adjusted to include those who were missed, the final estimate of the number of people in the province was 6,3 million.

Table 11 is read similarly for households. For instance, it is seen that Gauteng has the largest number of households in the country (2,0 million), even though it does not have the largest population. KwaZulu-Natal has a larger population in rather fewer households (1,7 million), i.e., a greater average household size. Northern Cape has the fewest (0,2 million).

* All numbers given in this report are adjusted by the PES and rounded to whole numbers.
        The totals may therefore differ slightly.

The final undercount of 10,7% is higher than the preliminary estimate of 6,8%, as a result of a much more painstaking methodology. The actual methodology used, and the differences between the final and preliminary methods of undercount calculation, are discussed in Chapter 3.

In the chapters which follow, the methodologies used in conducting the studies which yielded the final population size are given. These include the 1996 population census, and the post-enumeration survey. The ways in which adjustments were made to tallies in previous censuses, particularly the 1991 census, are also described.