Skip to content

Census Geography of South Africa

Literature › Works

Coleman Dube (2005). Census Geography of South Africa. Statistics South Africa (Geography Division), Pretoria.
Type other
Scope Comparative and methods
Method descriptive quantitative
Period studied 1911–2005
Regions South Africa, Bantustans (Transkei, Bophuthatswana, Venda, Ciskei), national municipalities
Themes methods data sources, demography urbanisation, bantustans removals influx
Citation key dube2005

Argument

The paper is a technical note by a Stats SA geography practitioner. It traces how South Africa's census geography frame has changed with administrative boundaries, from the fragmentation caused by apartheid-era Bantustans to the GIS-based 2001 frame. It then sets out how the frame is maintained after municipal boundary changes. Using the 2005 update as a case study, it shows that most enumeration areas (EAs) split by new boundaries did not need reallocation. It argues that an area-percentage rule combined with manual review keeps the frame consistent, but that changes carry costs for survey master samples and for time-series comparability.

Evidence and method

Stats SA's internal GIS datasets: the 2000 and 2005 Municipal Demarcation Board boundary files, the 2001 EA file, and aerial photography used to settle ambiguous cases. Also internal Stats SA documents and census methodology reports, and Coombes (1995) on EA definitions.

Findings

  • Census coverage of all population groups dates from 1911. (p. 2)

    Population censuses that covered all the population groups in South Africa date as far back as 1911.

  • Apartheid's Bantustans and self-governing territories fragmented the spatial framework of census collection and publication. (p. 2)

    has been fragmented by the Bantustans and self-governing territories imposed by Apartheid.

  • EA definitions are census-specific, which hampers the study of change and the study of poverty. (p. 2)

    Coombes (1995) points out that an enumeration area definition is unique to a particular census.

  • Before 1996 census mapping relied on poor and outdated maps. (p. 2)

    Many maps were insufficient and outdated. New developments were not reflected on the maps.

  • The 2001 census was the first to use a GIS for census management and map production. (p. 3)

    Census 2001 was the first census in South Africa in which an established geographical information system was utilised in census management and map-based production.

  • The 2001 frame coded far more place names than there are actual main places, because place names straddle municipalities. (p. 4)

    A total of 3 109 main place names were coded although there are only 2 674 main places

  • Post-2000 municipal structure: the number of metropolitan areas, district councils, local municipalities and district management areas. (p. 5)

    In total, there are six metropolitan areas, 47 district councils, 231 local municipalities and 25 district management areas.

  • Comparing the 2000 and 2005 municipal files identified 1417 changed areas. (p. 6)

    The result showed that 1417 areas had changed in the new municipalities (Figure 3).

  • Intersecting EAs with new municipalities created 9659 additional polygons, mostly duplicates of split EAs. (p. 7)

    The result gave a count of 90446 new EA polygons created. This meant an additional 9659 EA polygons were generated.

  • Only 304 EAs qualified for automatic reallocation, and 31 of the 198 investigated EAs were reassigned. (p. 10)

    A total of 304 EAs satisfied this criteria and were re-assigned to their new respective municipalities

  • Most investigated EAs stayed in their old municipality. (p. 10)

    Of the 198 investigated EAs, 31 of them were reassigned to new municipalities.

  • Boundary changes require adjusting survey master samples, which affects time series comparability. (p. 14)

    This has serious impact on time series analysis research and comparability of data.