Version classiqueVersion mobile

Territories and urbanisation in South Africa

Frédéric Giraut
Céline Vacchiani-Marcuzzo

Part I. Composition and structure of the Dysturb GIS database

Texte intégral

Origin of the data

South African censuses

1The official general censuses of the South African population were performed by successive official bodies (predecessors of the present-day Statistics South Africa) for the years 1904 (districts and provinces, no data for localities), 1911, 1921, 1936, 1951, 1960, 1970, 1980, 1985, 1991, 1996 and 2001. Three censuses were removed from the geo-referenced database for reasons of data reliability (1985), incomplete data (1904 and 1985), and difficulties in tracing localities longitudinally (1996). Thus 9 censuses, with inter-census gaps of around ten years, made up the basis of the geo-referenced base with longitudinal follow-up over one century of South African localities (places), urban areas (agglomerations) and territories.

2Where demographic data are available in two modes of enumeration, de jure and de facto, it is systematically the de facto enumeration that is preferred, as being closer to the actual distribution of the population.

Complementary sources (censuses and reports concerning the homelands)

3For the year 1991, the data concerning the “independent” homelands of Bophuthatswana, Ciskei, Transkei and Venda (TBVC) are derived from official statistical sources in these homelands. In addition, several studies (Graff, 1986; HRC, 1992; McCarthy & Bernstein, 1998; Malan & Hattingh, 1976) were used for the former homelands and townships, in particular for the identification of urban agglomerations.

4Finally, the reports issued by Statistics South Africa, by the South African Institute of Race Relations and by the Urban Foundation constitute valuable sources on issues of socio-spatial segregation and habitat that conditioned the distribution of the South African population on various scales.

Administrative and political data (gazettes, toponymic dictionaries)

5Regarding administrative and political data (dates and terms of the creation or alteration of magisterial districts or local capitals), official government or provincial gazettes listing acts, regulations and proclamations were consulted on the basis of indications in Juta's Index to the South African government and provincial gazettes. Concerning the names of localities, the years of their creation and when they obtained local government status (municipality, different local government councils) the main source was Raper (2004).

Places: The basic entity in Dysturb

Definition of a census locality (place) in South Africa in the course of the 20th century

6The localities or places retained in the censuses underwent considerable changes, in number, name and status. The Dysturb database enables follow-up of localities by systematically establishing correspondences despite changes in name or status.

7It should first of all be noted that until 1991 the localities used in the censuses correspond to the official definition of the urban object. That is to say that the official list of localities was drawn up in accordance with the status of these localities, which was itself linked to the origin of the locality and its racial composition, more than to its dimensions. The existence of a local government council was the urban definition criterion until 1970.

Localities with a local government body
A Municipalities (including Boroughs in Natal and Town and City Councils in Transvaal);
B Village Management Boards in the Cape provinces and the Orange Free State Village Councils in Transvaal; Town Boards in Natal;
C Health Committees in Natal and Transvaal: Local Boards in Cape Town;
D Local Areas in Cape Town and Transvaal; Public Health Areas in Natal.

8The table provided in the CD-Rom annexes entitled “Administrative status and functions of localities gives dates of creation and of allocation of functions up to the year 2000 for all localities that were local government seats and/or possessed local government status, even if only elementary (non-urban).

9A certain number of localities disappear as places in the course of the period: these are colonial settlement villages or missionary centres which were at first identified as fully-fledged places, after which their populations of European origin declined (the case with certain missions), or else they were integrated into neighbouring municipalities or into wider suburbs.

10Among the new places appearing, before 1970 these were mainly villages or suburbs where the growth of the population of European origin led to their obtaining local government status, and therefore urban status for the censuses. The increase in the number of urban places was thus considerable between 1931 and 1951 (rising from 492 to 722, corresponding to an increase of around 50 %).

11The censuses also make it possible to identify a certain number of localities that have no official status as an urban place, but that are included in the Dysturb database. Indeed, the first censuses (1911 and 1921) included lists of towns, villages and rural settlements. In addition, up to 1960, the censuses provide lists of suburbs that are not urban localities but a breakdown of the main metropolitan areas. Finally, retrospective tables enable access to data fot places appearing at a later date. In the Dysturb database localities of this sort (non-official in statistical terms, since only urban places are taken into account) are identified as settlements. Their populations are also included in the total for rural areas in the district to which they belong, which gathers all the populations outside the urban places.

12From 1980 the main black, coloured or Indian townships not possessing an elected council but recognised as urban localities were also considered as official census localities (urban places). This was in no way a concession to democracy on the part of the apartheid regime, but a mere accounting strategy for the black urban areas liable to be involved in forced removals towards “homelands” within the ‘grand apartheid’ policy. Thus more than 350 Black townships appear as urban places in the 1980 census.

13In 1991, two so called “independent” homelands, Bophuthatswana and Transkei, performed an exhaustive enumeration of places in their census, i.e. all the population in each district was allocated to a certain place. Thus for these homelands there was no longer a non-differentiated “rural” category gathering all the population that was not considered to belong to an urban place. This precedes the post-apartheid practices present in the 1996 census where an exhaustive gtid of more than 12 000 places was used, and the census of 2001 where 2 674 main places were determined and delineated, subdivided into 15 966 subplaces. The new administrative organisation is indeed based on seven geographical scales, and the new subdivisions cover the whole of the South African territory.

  • 1 Some had a shorr-lived existence in the censuses, appearing in one or two census and then disappear (...)

14For the database, the correspondences between places of 1991 and before and those of 2001 were first of all systematically sought in the subplaces, and then for remaining localities among the main places. Of some 4 000 localities (3 923) enumerated in 1991 or before1, around 1 430 are presented in the database but have no correspondence with a main place or subplace in 2001, and no geo-referencing. For the most part (1 306 out of these 1 430) they are localities noted in Transkei and Bophuthatswana in 1991 alone.

© Celine Vacchiani-Marcuzzo, UMR Géographie-cités, 2004
Category A: Metropolitan Area
Category B: Municipalities (not Metropolitan Areas)
Category C: Disctrict councils
Source: Statistics South Africa, 2001
Figure 1 – Hierarchical organisation of South African territorial subdivision in 2001.

The issue of name changes

15As in any historical database, the follow-up of the basic entities (in this case places) and the units to which they are attached (in this case districts) comes up against the problem of changes in name. Changes can be linked to transcription, or, on the case of bi- or multilingualism, to the way in which one or other language is given precedence in different periods, or again to dual toponymy. Finally changes in name of a political nature can occur following a change in regime (here the apartheid system), with a desire to erase certain historical references belonging to the abolished regime, or conversely to return to past heritage or commemorate a struggle (Giraut et al, 2008).

16In the case of South Africa, changes in names are frequent in the period covered by Dysturb. A table given in the CD-Rom annexes lists all these changes in the names of districts and places. While it is transcription issues that underpin most of the changes occurring up to 1951, issues relating to the translation of proper names into English or Afrikaans are already found before the start of the apartheid regime. The opposition between the English-speaking and Afrikaner communities can indeed be found in the toponymy. Franschhoek (the French corner, by reference to the Huguenots who left their country after the repeal of the Edict of Nantes) was partially anglicised in the censuses of 1921 and 1936 in the form “French Hoek.” However the apartheid period saw a reinforcement of toponymic bilingualism in favour of Afrikaans. From the 1951 census to that of 1991, numerous places (including those in Natal) and magisterial districts carried two names via an almost systematic translation of Bay (baii in Afrikaans), of North, South, East and West, (Noord, Suid, Oos, Wes), of Road (Weg) and sometimes River (Rivier) or Town (Stad). Mining quarters also took on dual denominations (Mine Compound and Myn Kampong). Prince Albert is translated as Prins Albert to obtain a dual appellation, while Franschoek took over as the sole name.

  • 2 KwaMashu is in fact the Zulu form of a reference to a notable in Natal colony, Marshall, so that th (...)
  • 3 The well-known acronym SOWETO standing for South West Townships (which was a complex of townships t (...)

17Toponymy of Bantu origin with the establishment of apartheid and in particular of the ‘grand apartheid’ policies, underwent a subtle interaction between the established powers and their objectives on the one hand, and subversion by way of local usage on the other. The urban apartheid regime inherited a relatively composite toponymy for the African quarters (created during the colonial and then the Union periods under the denomination of location). Locations in some cases had no official name, or else were given names of European origin (Alexandra, Lamontville) or African origin (Langa, KwaMashu).2 Thereafter, the apartheid regime allocated “neutral” place-names (of non-place-names, a minima) to the townships that it formed under the Group Areas Act 1950.3 The official place-names as they appear in population censuses from 1980, with the creation of new census localities of the township type, were often (183 new place-names in 1980) made up of the name of the town followed by the bilingual (Afrikaans-English) acronym SD (Swart Dorpsegebied)/BT (Black Township). In the 1985 census, with the systematic instatement of the 'grand apartheid' policies, the townships recognised in the census were systematically given African names: allonyms or non-official names in most cases. 122 changes occurred in the 1985 census, and 23 in the 1991 census. It can also be noted that there was some hesitation in this process, since over less than 10 years in the Orange Free State certain townships were given two successive African names (for instance Ventersburg SD/BT becomes Tswelagpele and then Mnamhabane).

Using the Dysturb data

The database for places and territories

18The Dysturb database is designed to enable crosswise and longitudinal study of South African localities (places), urban areas (agglomerations) and territories and their populations (figures 1 and 2).

Cross-sectional study

  • 4 The base statistical units are places and magisterial districts. However, in 2001 places may corres (...)

19The database enables study of the situation and the population of places and magisterial districts (the two statistical units that are observable for the whole period 1911 to 20014, at a precise census date. The yearly files concern a single year and are made up of several spreadsheets: one for each type of statistical unit, and one giving meta-data (PL + year, MD + year, meta-data). In the “MD” (Magisterial District) sheets, each district is allocated an identification code for each year. This identification is made up of the letters MD, the reference date, and a (non-significant) ordinal number; for example, MD91_128 stands for the MD of Inanda in 1991. This is followed by the name of the magisterial district in the spelling at the reference date, and a set of data concerning the composition of the population (detailed according to category). Likewise, in the “PL” (place) sheets, each place has, a single identifier for each year. This is made up of the code for the MD to which the place belongs followed by the letters PL and an ordinal number; for instance, MD91_128_PL_1090 for the place Radcliffe in the magisterial district of Inanda. The name of the place is then given in its spelling at the reference date, and a series of data on the composition of its population. This coding system avoids any confusion between years, between types of statistical unit, and between statistical units of a given type. Any unit can at all times be situated in its temporal and geographical context.

Longitudinal study

20The other aspect of the database is its design enabling study over time. One of the main aims was to enable follow-up of the evolution of South African territories and localities. Correspondence for the statistical units over time has been provided for, so that the database can be used for this purpose. Each place and each magisterial district have been allocated a single code over time at all dates so as to facilitate follow-up. The “dictionary” file makes it possible to know the different codes that a place or a magisterial district has carried in the course of time. Thus it is possible to start from any given year and find the locality in any other year if it is recorded in a census.

The "Urban Agglomeration" database

  • 5 The word “agglomeration is used here in the sense of the French usage to refet to urban or built-u (...)

21The database concerning towns and cities established by Celine Vacchiani-Marcuzzo (CVM database) is based on a morpho-functional definition of the urban agglomeration (association of the white town/city and the non-white townships of the apartheid period, inclusion of commuting phenomena).5 The main source is population censuses, and it covers the long term (1911-2001). Tables (in Excel format) have been constructed so as to enable follow-up and visualisation of urban growth throughout the 20th century.

22Each urban agglomeration is characterised by its total population and its component places. In the agglomerations file [les agglomerations (composition, population - 1911-2001).xls], the sheet population agglomeration comprises the following fields: the agglomeration code, a single code through time corresponding to the code for 2001 of the main place, total population for each reference year. If the population of a given urban agglomeration is to be followed (longitudinal study), it is simply selected on this sheet; and if what is required is an image at a precise date, agglomerations with a population at this date are selected. Another file (compo_agglosl911_2001.xls) provides the composition in terms of places for each agglomeration at each date (with the place code for 2001).



1 - A follow-up of administrative boundaries
2 - A vision of the hierarchy of territorial boundaries
3 - A follow-up of each urban or rural place
Table's content for each level: (provinces, magisterial districts, places)
- Code
- Name
- Population
- Legal and administrative status
- Creation date
- Higher administrative level code
- Code previous census
- Changes between two dates (status, name)
- Agglo Code
- Notices



Figure 2 - General organisation of the database.

The "City Generation" database

23This database (figures 1 and 2), (Excel file “generation_villes.xsl”), indicates the period at which places appear as urban agglomerations. These urban agglomerations are characterised by their codes, corresponding to a place in 2001 (it has not been possible to allocate a geo-code to some of them), the name of the town/city, and the period of its appearance (date and numerical code) (figure 3).


24In addition to the province and homeland levels, which correspond to the aggregation of districts, two main levels are systematically mapped for each of the census dates, providing a set of cartographic matrices (maplayers): these are the magisterial districts (polygons) and places (points up to 1991, and polygons in 2001).

Magisterial districts

25It should first of all be noted that the districting implemented for the purpose of census organisation corresponded to the magisterial districts up to 2001, but that the census districts actually used in some instances anticipated the official proclamation of magisterial districts by a few years. In other words, certain census districts used in a given census only became magisterial districts in the following census interval and hence after the census. Only the 1970 census distinguished ethnic subdivisions of magisterial districts (in particular for Natal) that were later for the most part, but not systematically, officialised as new districts.

26Technically, our task was to obtain the largest possible number of available outline maps for the different periods and grids, and then to select the most compatible in terms of projection and quality. Following this, a choice was made in favour of the 2001 and 1996 census outline maps, to which the other outline maps were fitted. For 1991 a digitised outline maps was available at a different level of generalisation. The earlier outline maps were derived from paper maps for the years 1911, 1960, 1970 and 1980, ensuring that the places were in the correct position in relation to administrative boundaries. For the years 1921, 1936 and 1951, as there was no outline maps enabling verification of the accuracy of the magisterial district boundaries, this was checked by available textual information on administrative changes, and a verification of the situation of the places. Thus the boundaries in these instances are less accurate than they are for the other years. It is fully possible to perform spatial analytical processing on these outline maps. In addition, the geographical objects of the MD files possess a code in common with the statistical units in the attribute files, so that it is possible to map census data from 1911 to 2001.

Source: T. Giraud, 2007.
Figure 3 – Organisation of the database.


27For 2001 it is possible to link all the places in the attribute files to GIS files. The largest possible number of places present before 2001 were then manually linked (on the basis of spelling proximity and belonging to a magisterial district). As it was not possible to assess their spatial cover or their evolution between 1911 and 2001, it is preferable to use only the centre points of places for any dates except 2001. These files can also be used for the representation of urban agglomerations and their dates of appearance. The maplayer for the new municipalities, used by Stats SA jointly with that for the magisterial districts for the 2001 census, is also available. This enables identification of municipalities to which places and districts belong, and ensures continuity with the other GIS which use only the municipality grid, such as SA Explorer, the GIS developed by the Municipal Demarcation Board.

28All the Dysturb files can be used in the main GIS, most are in Shp format for Arcgis ESRI©, and some are also available in the Mid/Mif format for MapInfo©. The maplayers are accompanied by a projection file (.prj) detailing the projection and reference parameters used (geodesic system: WGS_1984, projection system Alber's Equal Area Projection).


1 Some had a shorr-lived existence in the censuses, appearing in one or two census and then disappearing before the 1991 census.

2 KwaMashu is in fact the Zulu form of a reference to a notable in Natal colony, Marshall, so that this place name means Marshall's place.

3 The well-known acronym SOWETO standing for South West Townships (which was a complex of townships to the south-west of Johannesburg) illustrates this functionalistic English toponymy. One of the few African names is that of Gugulethu in Cape Town, meaning “our pride” –which was doubtless more the pride of the planners than that of the inhabitants.

4 The base statistical units are places and magisterial districts. However, in 2001 places may correspond to subplaces, or main places made up of subplaces. Main places and subplaces are distinguished by way of their numerical codes: 5 figures for main places, 8 for subplaces. In addition, the places do not strictly correspond to the magisterial districts, but form the new municipalities. We have however allocated the magisterial district code to them, since it covers them most efficiently. We have also indicated the municipality code, and have shown the new subdivisions in force since 2000.

5 The word “agglomeration is used here in the sense of the French usage to refet to urban or built-up areas, which may be made up of several entities of differenr type (central city, suburbs, satellite towns, etc.).

© IRD Éditions, 2009

Conditions d’utilisation :

Cette publication numérique est issue d’un traitement automatique par reconnaissance optique de caractères.
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search