Apartheid and Urban Segregation Levels in South Africa¶
A. J. Christopher (1990). Apartheid and Urban Segregation Levels in South Africa. Urban Studies, Vol. 27, No. 3. Urban Studies Journal Limited / SAGE.
| Type | journal article |
| Scope | Wider apartheid literature |
| Method | descriptive quantitative |
| Period studied | 1911–1985 |
| Regions | South Africa, Western Cape, Eastern Cape, Natal, Orange Free State, Transvaal, Durban |
| Themes | origins segregation, bantustans removals influx, demography urbanisation, methods data sources, politics ideology elections |
| Citation key | christopher1990 |
Argument¶
Christopher uses the Duncan index of segregation, calculated from census enumeration tracts for 1911-1985, to show that urban racial segregation rose steadily across the twentieth century, with a sharp jump in the 1960s as the Group Areas Act and resettlement programmes took effect. Colonial-era regional differences (a lax Western Cape, strict Orange Free State) persisted until the 1960s, when central policy made the Western Cape as segregated as other regions. Total separation was never achieved, because Black domestic and compound workers remained in White areas, and because enforcement weakened in the 1980s. He concludes that apartheid's spatial project was impossible to implement in full, yet its built legacy will constrain the post-apartheid city.
Evidence and method¶
Enumerators' summary returns from the 1911, 1921, 1936, 1951, 1960, 1970 and 1985 censuses (Central Archives Depot and Central Statistical Service), divided into tracts of about 600-1500 people. Includes all towns over 1000 people with at least three tracts (83 in 1911, 61 in 1921). Also legislative texts, Durban municipal minutes, and secondary literature. Uses the Mann-Whitney U test for regional comparisons and Spearman rank correlation for town size.
Place in the debates¶
- Intent versus reality of apartheid spatial planning (how far the state achieved total residential separation) (see Urban segregation, Group Areas and the apartheid city). Segregation reached very high levels but total separation was unattainable; about 10% of the urban population in 1985 still lived in the 'wrong' group area.
- Colonial legacy versus post-1910 uniformity in segregation policy (see Urban segregation, Group Areas and the apartheid city). Colonial regional differences survived almost intact until the 1960s and persisted in muted form afterwards, while central legislation drove convergence.
- Relationship between city size and segregation levels (see Urban segregation, Group Areas and the apartheid city). Finds no significant relationship in South Africa, in line with Taeuber and Taeuber (1965) and against Farley and Wilger (1987) for US cities.
Findings¶
-
The national median White index of segregation rose from 54.7 in 1911 to 92.8 in 1985, with the largest jump in the 1960s. (p. 4)
1911 29.5 64.4 42.7 76.9 41 .1 54.7 660
-
Segregation rose dramatically between 1960 and 1970, when the Group Areas reorganisation took effect. (p. 9)
The major effect of the reorganisation and resettlement programmes was to boost segregation levels in the 1960s . Thus between 1960 and 1970 segregation levels throughout the country rose dramatically .
-
The Western Cape Coloured index jumped after Group Areas enforcement. (p. 11)
As a result the median Coloured index of segregation in the Western Cape had risen to 86.6 by 1970 and continued climbing to reach 92.0 by 1985 .
-
Coloured families in the Cape bore the brunt of Group Areas removals. (p. 13)
Over half the families resettled under the Group Areas legislation were Coloured families in the Cape of Good Hope.
-
The Group Areas Act could not move people until 1957, when the machinery for forced removals was in place. (p. 8)
The initial Act proved to be defective, in lacking the mechanism for forced population removals, and only in 1957 was the machinery in place
-
Asian segregation in Durban rose to near-total levels after the Group Areas Act. (p. 15)
Thus in 1985 the Asian index of segregation in Durban had reached 95 .9 .
-
Segregation was never complete: a tenth of urban residents were outside their designated group area in 1985, mostly Black workers in White areas. (p. 19)
Some 10 .0 per cent of the urban population of South Africa in 1985 still lived in the `wrong' group area. The majority (820 000) were Black workers in White areas .
-
Black indices in 1985 were lower than for other groups because of continued domestic service employment in White areas. (p. 17)
because of the continued employment of Black domestic servants in the more affluent White residential areas, Black indices of segregation in 1985 were lower throughout the country than for the other groups .
-
Forced Group Areas resettlement virtually ceased after 1982 because of a court ruling, and mixed inner-city areas emerged. (p. 19)
Forced resettlements under the Group Areas Act virtually ceased after 1982, when the courts ruled that the government was required to provide alternative housing for those evicted .
-
Apartheid's spatial infrastructure will outlast the system. (p. 20)
The infrastructure of apartheid cannot be readily erased, and an integrated society cannot be brought into existence as easily as it can be destroyed.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Enumerators' summary returns (enumeration-tract population by race), post-Union censuses 1911, 1921, 1936 (Central Archives Depot, Pretoria) | population census | 1911-1936 | Source of tract-level population by race group used to compute Duncan indices of segregation for towns in 1911, 1921 and 1936. |
| Enumerators' summary returns (enumeration-tract population by race), censuses 1951, 1960, 1970, 1985 (Central Statistical Service, Pretoria) | population census | 1951-1985 | Source of tract-level race-group populations for the 1951-85 segregation indices, the core dataset of the paper. |
| Published census reports: Report of the Census 1911 (U.G. 32-1912); Population Census 1921 (U.G. 15-1923); 1936 (U.G. 21-1938); 1951 (U.G. 42-1955); 1960 (R.P. 62-1963); 1970 (Report 02-05-01); 1985 (Report 02-85-01) | population census | 1911-1985 | Provide census urban totals used to show the study towns cover 74-91% of urban population and to compare enumerators' books with published returns (e.g. reclassifications in 1911). |
| Official data on racial reclassification under the Population Registration Act, 1988 (South Africa, 1989, Parliamentary Debates) | parliamentary debates | 1988 | Table 3 shows 867 people reclassified in 1988 to illustrate the imprecision and arbitrariness of racial classification. |
| Comparison of census-day enumerator classification with published census, 1911 and 1960 (e.g. Beaufort West, Bloemfontein, Kokstad) | population census | 1911, 1951, 1960 | Documents classification inconsistencies and the decision to use the population group in the enumerators' summary books for calculations. |
| Report of the Transvaal Local Government Commission, T.P. 1-1922 | commission or official report | 1922 | Quoted to show the official doctrine that Blacks were 'temporary sojourners' in towns underlying the Natives (Urban Areas) Act. |
| Durban Corporation, Mayor's Minute 1923 | company records or reports | 1923 | Quoted for the stated municipal policy of separating Europeans from Asiatics and Natives in residential areas. |
| Durban, Mayor's Minute 1950 | company records or reports | 1950 | Quoted to show Durban's municipal welcome of the Group Areas Bill as a solution to racial problems. |
| Legislation: Housing Act 1920, Natives (Urban Areas) Act 1923, Slums Act 1934, Trading and Occupation of Land Act 1943, Asiatic Land Tenure Act 1946, Population Registration Act 1950, Group Areas Act 1950 | legislation gazette | 1913-1950s | Legislative overview establishing the intent of segregation policy against which index outcomes are interpreted. |
| Report of the Committee for Constitutional Affairs of the President's Council, P.C. 4-1987, and 1989 Parliamentary Debates | commission or official report | 1987-1989 | Cited for policy reformulation, 'free settlement' areas and the new 'non-ethnic' group. |
| Fick, De Coning and Olivier (1988) Mayfair study; Picard-Cambridge (1988) Greying of Johannesburg; De Coning et al. (1986) pilot study | secondary literature compilation | 1986-1988 | Used for the claim that non-conforming residents are rising and that mixed suburbs may become single-ethnic again. |
| Estimates of residents in the 'wrong' group area, 1985 (820,000 Black workers, 370,000 in compounds and locations; 230,000 Coloured and Indian) | official statistical series | 1985 | Quantifies the gap between planned and actual residential separation to explain failure of total segregation. |
New data¶
The author built a town-level panel of Duncan indices of segregation for White, Coloured, Asian and Black populations for 1911, 1921, 1936, 1951, 1960, 1970 and 1985, computed from enumeration-tract returns (600-1500 people each; 83 towns in 1911, 61 in 1921, about 17,200 tracts in 1985), with regional medians, statistical tests and maps.
Inference¶
Quantification: Statistical analysis.
Time-series and regional comparison of Duncan segregation indices computed from census enumeration-tract data, interpreted against a narrative of legislation; Mann-Whitney U tests for regional differences and Spearman correlations for the size-segregation relationship; Durban used as a detailed case.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| White segregation rose markedly from 1911 to 1985, with regional medians converging, and the main jump occurring in the 1960s. (p. 428) | Regional median White indices (Table 1) computed from enumerators' tract returns across 1911-85 censuses, with Mann-Whitney tests. | Systematic data |
| Colonial regional differences in White segregation persisted until the 1960s; the Western Cape moved from least to no longer significantly less segregated. (p. 428) | Table 1 regional medians (29.5 Western Cape to 76.9 OFS in 1911) and Mann-Whitney tests. | Systematic data |
| There is no significant relationship between town size and White segregation level; small towns enforced segregation as effectively as large ones. (p. 428) | Spearman rank correlations between town population and White index for 1911, 1960, 1985 (Figure 2), values between 0.2 and -0.2. | Systematic data |
| Coloured segregation in the Western Cape rose sharply in the 1960s as a result of Group Areas enforcement and resettlement, reaching 92.0 by 1985. (p. 430) | Table 4 median Coloured indices (64.8 in 1960, 86.6 in 1970, 92.0 in 1985) combined with legislative narrative; claim about causes rests on inference from timing. | Partial data |
| Asian segregation in Durban rose from 58.9 (1936) to 70.9 (1951) and 95.9 (1985) as Group Areas legislation removed Indians from inner areas. (p. 434) | Durban tract-level indices, racial composition maps (Figures 5-6), Mayor's Minutes 1923 and 1950. | Partial data |
| Black segregation indices in Natal are low because of employer housing and exclusion of KwaZulu locations from census returns, not social integration; homeland transfers distort 1970-85 comparisons. (p. 435) | Table 6 and 7, census urban Black counts for Natal (445,000 to 256,000) and author's knowledge of boundary changes in 12 towns. | Partial data |
| Apartheid failed to achieve total separation: about 10% of urban residents in 1985 lived in the wrong group area, mainly Black workers in White areas, though over 90% conformed to the plan. (p. 438) | Unspecified 1985 census-derived estimates of residents by group area, and the index values themselves; no source given for the figures. | Asserted |
The author's caveats¶
Acknowledges the scale problem of the index (values rise as tracts shrink and with town size), so results show broad trends rather than absolutes. The 1921 indices are circumspect because larger tracts and only 61 towns were used. In 1911 records for the eastern Witwatersrand were destroyed, excluding Benoni, Boksburg and Springs. Published and enumerator totals differ because of late additions. Classification was inconsistent (1911 reclassifications, Kokstad 1960). Homeland transfers affected 12 towns for 1970-85, and homeland cities were excluded.
Gaps in the evidence¶
The causal link between specific legislation (Group Areas Act removals) and index changes is inferred from timing rather than tested with removal records, such as Group Areas Board proclamations or numbers of families moved by town. Figures for residents in the 'wrong' group area and for Group Areas resettlement have no cited source. Finer-grained data (street-level or ward-level, municipal valuation rolls, title deed covenants) and the census-boundary changes for homeland incorporation could better test the index comparability; some of these archives existed then and are accessible now.
Replicability: Data described but not public.