Overview of Census ’96
Every person present in South Africa on Census Night, 9-10
October 1996, should have been enumerated in Census '96. The enumeration generally took
place over the period of 10 October to 30 October although, in some situations, it was
necessary to continue enumeration through to December to ensure that as many people as
possible were included. In most cases, respondents were given the choice of being
interviewed or of completing the questionnaire themselves.Design of the sample
This sample is a 10% unit level sample of
(a) all households (excluding special institutions and hostels) and
(b) all persons as enumerated in the 1996 Population Census in South Africa.
The household was basically drawn as a 10% systematic sample of households from the census
household file, stratified as specified below. The 10%person level sample was obtained by
including all persons in these households plus the persons drawn in independent 10%
systematic samples of all persons in special institutions and hostels
Stratification and ordering of the household records for sampling
The census household records were explicitly stratified according to Province and
District Council. Within each District Council the records were further implicitly
stratified by local authority (as defined for the purposes of this sample and explained
further below) and EA type. Within each implicit stratum the household records were
ordered according to the unique seven-digit census Enumerator Area number, of which the
first three digits are the (old) magisterial district number
Defining local councils for the 10% sample
Different terms are used for the local authority boundaries in different parts of the
country. There are Transitional Local Councils (TLCs); Transitional Rural Councils (TRCs);
Local Authority Councils (LACs); Metropolitan Sub-Structures (MSSs); Metropolitan Local
Councils (MLCs); Rural Local Councils (RLCs); District Councils (DCs); Transitional
District Councils (TDCs) and Regional Councils (RCs).
To ensure confidentiality within the 10% sample, a local authority had to have a minimum
of 2000 households. As many local authorities had fewer than this number, they had to be
grouped together to ensure that the minimum number of households was met. For this
purpose, hostel dwellers were treated as single person households.
Local authorities with less than 2000 households were pooled with other local authorities
based on the following principles:
All provinces except KwaZulu/Natal and North West.
A TLC with less than 2000 households was grouped with the TRC within which the TLC was
located. In cases where the TRC was big enough to stand on its own but the TLC's within
its boundaries were too small, the sample was drawn in such a way that the TRC can be
analysed either on its own or together with other TLCs within its boundaries. Where a TRC
plus all the TLCs within its boundaries were less than the minimum of 2000 households the
TRC (including the TLCs within its boundaries) was pooled with the adjacent TRC.
In a few cases the required minimum of 2000 households could not be achieved when all the
local authorities within a District Council were pooled together. In such a case no
further implicit stratification within the District Council was done.
KwaZulu/Natal
The equivalent to a District Council in KwaZulu/Natal is known as a Regional Council.
There are no rural councils in KwaZulu/Natal. Smaller local authorities could therefore
not be pooled with the rural council within which boundaries it falls. Where such TLCs
were adjacent to another TLC they were pooled to form one stratum. In two cases three TLCs
were pooled to form one stratum. In all cases the TLCs that were pooled are adjacent to
each other.
North West
The TRCs in the North West do not encompass TLCs as is the case in other provinces.
The area between TLCs/TRCs in the North West is part of the relevant District Council.
Small TLCs/TRCs in the North West were either pooled with adjacent local authorities or
they were pooled with the relevant District Council.
Census questionnaires
Different methods of enumeration were used to accommodate
different situations and a variety of questionnaires were used. The information collected
with each questionnaire differed slightly. This will be noted in the documentation on each
census variable and can be seen in the electronic versions of the questionnaires provided.
The questionnaires used were as follows:
1. Households
Every household living in a private dwelling should have been enumerated on a household questionnaire. This
questionnaire obtained information about the household
and about each person who was
present in the household on Census Night.
2. Hostels
A summary book for hostels
should have been completed for each hostel (that is, a compound for workers provided by
mines, other employers, municipalities or local authorities). This questionnaire obtained
information about the hostel and
also listed all household and/or persons enumerated in the hostel.
Some hostels contain people living in family groups. Where people were living as a
household in a hostel, they were enumerated as such on a household questionnaire (which obtained information about
the household and about each person who was present in the
household on Census Night). On the final census file, they will be listed as for any other
household and not as part of a hostel.
Generally, hostels accommodate mostly individual workers. In these situations, persons
were enumerated on separate personal
questionnaires. These questionnaires obtained the same information on each person as would have been obtained on
the household questionnaire. The persons will appear on the census file as part of a
hostel.
Some hostels were enumerated as special institutions and not on the questionnaires
designed specifically for hostels.
3. Special institutions and homeless persons
An enumerator's book for special
enumeration should have been completed for each institution such as prisons and
hospitals. This questionnaire obtained information on the institution and listed all persons present.
Each person was asked a brief
sub-set of questions – just 7 compared to around 50 on the household and personal
questionnaires.
People in institutions could not be enumerated as households.
Homeless persons were enumerated
during a sweep on census night using a special questionnaire. The results were later
transcribed to standard enumerator's books for special enumeration to facilitate coding
and data entry.
Instructions for interviewers
All interviewers were given an Enumerator's Manual that
included a section on 'Instructions for specific questions'. These instructions covered
every question on the census questionnaire and have been included in the documentation as
an indication of the theoretical basis of the questions. However, it is important to note
that not every interviewer would have been familiar with all the instructions and they do
not necessarily reflect how the questionns were asked by the interviewer. Also, these
instructions were not provided to people who completed the questionnaire themselves and
they would have been required to give their own interpretation to questions.
Household, person and migrant records
Following the structure of the questionnaires described
above, the SuperSTAR database contains records at three different levels:
- Household (this includes hostel and institution information)
- Person, and
- Migrant.
When a table is produced, it can be produced for any of these levels by specifying the
appropriate level in 'Summation options'. Care should be taken that the appropriate record
type is selected corresponding to the variables used. That is, if a table contains
household variables such as 'Household size', then a household level option should be
selected in Summation options.
Dummy households
When the final census database was compiled, in a few
cases mis-entered data meant that the link between households and persons was lost. In
these cases, dummy households were created to allow the persons to be included on the file
for final analysis. As no information is available for these dummy households, they are
listed separately or as part of the not stated codes for most household level variables.
Weighting factors
Both the 10% household sample file and the 10% person
sample file contain a weight variable. This weight variable is the adjustment factor for
undercount (for households or persons as appropriate) multiplied by 10 to inflate the 10%
sample to the population.
Inevitably some households and people were missed during enumeration. A survey was
conducted after Census '96 to estimate the level of undercount. This information was used
to calculate weights that can be applied to the census data to adjust for undercount.
Users can access both weighted and unweighted census data by selecting the appropriate
categories in Summation options. For more information about the adjustments for
undercount, see the publication 'Calculating the undercount in Census '96', Statistics
South Africa Report No. 03-01-18 (1996).
Hierarchies and standard recodes
Often, census questions obtained more detailed information
than is required by many users in producing tables. For example, there are 120 age
categories and hundreds of magisterial districts. A table of Age by Enumeration Area (EA)
would result in an extremely large table. There are a number of ways users can use
SuperSTAR to produce more useful tables. For example, a user could just select the
categories of interest, such as age groups 15 to 29 in 5 EAs.
SuperSTAR provides two other ways to easily create useful tables. Some variables, such as
the geographic data and Occupation, are saved and presented in SuperSTAR in hierarchies.
If a user only needs broad categories, the first category that appears (provinces for
geographic data, 1 digit codes for Occupation) can be selected. If more detailed
information is required, going through the structure of the hierarchy will expose other
levels of detail, down to EA for geographic data and 3 digit codes for Occupation. For
more information, see the SuperSTAR manual.
The other way to access grouped information is through recodes. Users can use SuperSTAR to
create their own recodes of the categories which can be saved as standard recodes for
later use. For more information on creating recodes or accessing standard recodes, see the
SuperSTAR manual.
More details of the hierarchies are presented in a latter section in the metadata.
Abbreviations
A number of abbreviations have been used in this
documentation. They are:
| A |
Section A of the Household questionnaire |
| B |
Section B of the Household questionnaire |
| Certif. |
Certificate |
| C |
Column (refers to the question column on the
special enumeration questionnaire) |
| Dip. |
Diploma |
| EA |
Enumeration Area |
| Exc. |
Excluding |
| GIS |
Geographic Information System |
| Govt. |
Government |
| Hhold |
household |
| Inst. |
Institution |
| Mnfctr. |
Manufacturing |
| N/A |
Not Applicable (that is, no response was
required to the particular question from these households or persons). |
| NEC |
Not elsewhere classified (detailed information
provided by respondent but classification does not include a detailed code for the
response) |
| NFD |
Not further defined (insufficient information
provided by respondent to enable a more detailed code to be allocated) |
| No. |
Number |
| NU |
Non-Urban |
| Q |
Question number |
| Qu |
Question |
| Qus |
Questions |
| Qual. |
Qualification |
There were also some terms used that require further explanation:
Pre-coded Questions where a list of responses was offered for the respondent to choose
Write in Questions where the respondent/interviewer had to write in an exact response.
Often a precise code was allocated at a later stage.
Back to Metadata: Table of Contents |