Further progress in the desegregation of South African towns and cities, 1996–2001

Abstract

Between 1996 and 2001 South African towns and cities have continued the slow process of residential desegregation which was first noted in the previous inter-census period. The changes have been group-specific and place-specific, and the experiences of different groups and regions have differed substantially. The key African–White index of dissimilarity remains exceptionally high throughout the country, although it has now shown some evidence of decline. The Free State stands out as being significantly more segregated than the rest of the country. KwaZulu-Natal continues to be the most desegregated province and significant changes are evident in this respect for most inter-group indices. Nevertheless, the vast majority of the urban population continues to live in highly segregated suburbs.

1. Introduction

The results of the 2001 population census, the second all-South Africa population census since 1994, offer South Africans the opportunity to examine the continued progress of the country's towns and cities towards a more equitable and integrated future. In the previous inter-census period (1991 − 1996) there had been some small ‘first steps’ towards dismantling the remarkably high levels of residential segregation which had been enforced by the end of the apartheid era (Christopher, 1994, 2001). However, the degree of desegregation evident by 1996 was limited and highly selective in its spatial distribution. In this South Africa followed the experience of the United States since the enactment of the Civil Rights legislation of the 1960s, where progress towards desegregation has been slow and group-specific (Frey & Farley, 1996; Johnston et al., 2003). Further progress towards city integration in South Africa between 1996 and 2001 appears to have been just as hesitant as in the previous five years. Nevertheless the initial, slight, evidence of a reduction in segregation levels between Whites and Africans represents a symbolic break with the apartheid-era experience of ever-greater spatial separation between the two groups.

2. Desegregation

The Abolition of Racially Based Land Measures Act of 1991 removed the legal framework within which the apartheid city had been created and sustained. However, the Act did little to promote the reintegration of South African towns and cities. In 1994 the Government of National Unity, led by the African National Congress, came to power committed to the Reconstruction and Development Programme (RDP) which, among other aims, sought to reintegrate the urban areas through the promotion of major state-sponsored capital projects. However, since 1996 the government has pursued the free-market Growth, Employment and Redistribution strategy (GEAR) (Bond, 2000). State intervention was thus restricted to housing policies intended to assist the poorest sectors of society through subsidies (Lanegran & Lanegran, 2001). In view of the magnitude of the housing crisis and the limited state funding that was available, solutions were adopted which tended to reinforce the segregated patterns of the previous era. The upgrading and extension of the mono-group African peripheral informal settlements was the subject of considerable effort (Abbott & Douglas, 2003; Huchzermeyer, 2003a). Exceptions to this observation were noted in the Cape Town metropolitan area, where newly formed communities were more integrated (Oldfield, 2000). Land invasions which did not conform to broad government land-use zoning plans were opposed and squatters were evicted, in contrast to the early 1990s (Ramutsindela, 2002). At the same time African and Coloured squatters were prevented from invading the former all-White suburbs and enforced city integration plans were effectively shelved (Saff, 2001). Nevertheless, the re-evaluation of the national housing subsidy policy suggests that government policy can be carried out in such a way as to effect change in the inherited urban form (Huchzermeyer, 2003b). However, whether residential integration is a matter of high priority is a matter for debate (Gibson, 2004).

Residential integration in the private sector was largely dependent upon the availability of mortgage finance. The majority of Africans, Coloureds and Asians were unable to access commercial funding owing to a lack of collateral, low incomes and the informal red-lining of many poorer suburbs. Limited integration did take place, as those qualifying for such loans or possessing the necessary capital moved into formerly White group areas (Kotze & Donaldson, 1998). This is effectively an historic return movement, as it was usually from these areas that members of the African, Coloured and Indian communities had been expelled under apartheid. However, the urban land restitution process remained essentially stalled during the period in question and so did not contribute to reintegration. Nevertheless, the rate of change was slow, being dependent upon the relatively low turnover of family housing and the continued presence of the White population. It has been noted that the integration of the former White areas was only possible as Whites moved out (Lemon, 1991). The transformation of Zimbabwean towns could be ascribed to mass White emigration after independence in 1980, but this movement has not been repeated on such a large scale in South Africa (Cumming, 1993). In some cases the transformation did little more than create new ghettos in the inner areas of the major cities, which were subject to ‘white flight’ and abandoned to the incoming poor Africans (Jurgens et al., 2003). Directly countering these trends was the emergence of the gated suburb and housing complex, which sought to preserve residential exclusivity (Landman, 2000). As a result the trends in South Africa's towns and cities have been complex. Within almost any city processes leading to desegregation and others leading to the reinforcement of segregation operated simultaneously (Beall et al., 2002; Haferburg & Ossenbrugge, 2003).

3. Census 2001

The results of the 2001 census were made available by Statistics South Africa on the same basis as those for 1996, thereby making direct comparisons more realistic than on previous occasions. The population was divided into four population groups, which corresponded with those adopted in the apartheid era (Christopher, 2002) (see Table 1). The experiment of including a fifth group, entitled ‘others’, was unsuccessful. Few people did not identify with one of the inherited groups, and those that did not were reclassified by Statistics South Africa to fit the fourfold categorization. The final 2001 census groups were therefore:

	Africans/Blacks, who were defined as being descended from people indigenous to the continent of Africa
	Indians/Asians, who were defined as people originating or descended from people originating in the Indian subcontinent
	Whites, who were legally defined as such under the now repealed Population Registration Act on a test of descent and acceptance, thereby excluding people of mixed descent
	Coloureds, who were effectively everyone not included in the three previous categories.

Table 1:  Urban population of South Africa, 2001

Province	African	Coloured	Asian	White	Total
Western Cape	1 139 712	2 102 422	44 546	774 411	4 061 091
Eastern Cape	1 740 941	418 281	17 968	274 259	2 451 449
Northern Cape	236 020	346 117	2 203	79 449	663 789
Free State	1 737 071	68 772	3 584	206 587	2 016 014
KwaZulu-Natal	2 902 243	133 333	777 350	450 490	4 263 416
North West	1 248 681	43 890	9 027	195 214	1 496 812
Gauteng	6 252 377	334 390	217 133	1 702 564	8 506 464
Mpumalanga	1 026 032	17 646	10 364	168 049	1 222 091
Limpopo	449 612	5 296	7 539	86 811	549 258
South Africa	16 732 689	3 470 147	1 089 714	3 937 834	25 230 384
Source: Compiled by the author from census statistics supplied by Statistics South Africa.

The census data were made available on the basis of magisterial districts. Some adjustment of the data was necessary. Amalgamations of data sets were necessary for meaningful results in two cases. In the first, a number of the major cities had been divided into several districts in order to facilitate the administration of justice and other central government functions. Thus figures for the magisterial districts of Soweto and Randburg, for example, were consolidated with those of Johannesburg. In the second case, the separate sets of African homeland districts, which included the African suburbs of a number of towns, were, where possible, reunited with the districts including the White, Coloured and Asian suburbs. Thus Sheshego, in the former Lebowa homeland, has been consolidated with Pietersburg to form the new city of Polokwane. New magisterial districts had been defined for the North-West province in 1996, in an attempt to remove the former homeland boundaries from the administrative map (Geyer et al., 2000). In KwaZulu-Natal the majority of the former homeland districts had been reincorporated into the provincial districts.

The data were further divided into enumeration areas. The average size of an urban enumeration area increased from 473 to 673 persons between the 1996 and 2001 censuses. Owing to the highly segregated nature of the population this change was not significant for the results of this study. The enumeration areas covered the designated urban areas of the magisterial districts according to the definition adopted for the 1996 census. This overcame the uncertainties of comparability following the local government restructuring carried out in 2000. The problems with this definition were primarily associated with the exclusion of the extensive informal settlements of the peri-urban African areas of the former homelands, which were nominally classified as rural. This resulted in under-representation of the African population in the urban tallies. Other problems involved the failure to identify urban areas. Three districts returning populations in urban areas in 1996 failed to do so in 2001 and so had to be excluded. Finally, the author was not permitted to establish the position of individual 2001 enumeration areas within individual urban areas, as the information was subject to new confidentiality rules of the government. This deficiency precluded any assessment of changes within the urban areas, including the scale of migration into and out of the former White group areas. It should be noted that the next national population census is only due to be held in 2011 and so the sequence of more closely spaced quinquennial censuses between 1980 and 2001, effectively monitoring the formal demise of apartheid, is at an end.

4. Segregation measures

Following the method devised by Duncan & Duncan (1955), inter-group indices of dissimilarity were calculated for the urban areas of the country. The indices are subject to problems of scale and interpretation, but for the sake of ease of calculation they are widely used both for inter-city and temporal comparisons (Massey & Denton, 1988; Iceland et al, 2002). The index (ID) is calculated as follows:

where ID _xy is the index of dissimilarity between the X and Y groups in the population, x _i is the percentage of the X group in the ith enumeration area and y _i is the percentage of the Y group in the ith enumeration area.

The results are calculated on a scale ranging from 0, which represents an identical distribution between the two groups, to 100, indicating total separation. When interpreting the figures, Kantrowitz (1969) advised that figures below 30 could be regarded as integrated populations, as statistically zero or very low values could not be expected owing to the spatial configurations of the enumeration tracts, varying sizes of populations and other essentially random occurrences. The American studies led him to believe that figures of more than 70 indicated a degree of coercive segregation implicit in segregated housing provision and legal enforcement. Figures between 30 and 70 suggested social, but probably not legalised, segregation between the two groups. He also suggested that when making temporal comparisons, only changes of more than five points could be regarded as statistically significant, as a result of the internal variability within the data sets compared. It was also decided that districts with less than five enumeration areas should be excluded, as the results are open to the vagaries associated with boundary drawing. Similarly, groups with fewer than 100 persons in a particular district have been excluded as being too small to present a valid picture of segregation. As a result the majority of towns in the former African homelands have been excluded from the analysis, as they are virtually entirely inhabited by a single population group. Therefore, some 257 urban areas are examined in this study.

5. Segregation in 2001

Segregation levels in South African towns and cities remained extremely high in 2001 (Table 2). The median indices comparing the distribution of the African population with both the White and Asian populations remained over 90. In contrast the median African–Coloured and Asian–White indices had fallen to 74. In the two latter cases this represented a more than 10-point decline in the course of the five years under consideration (Table 3). Clearly different dynamics are operating with regard to the inter-action between the various groups. The relationship between the African and White populations will be examined closely as representing the interaction of the numerically dominant with the formerly politically dominant groups. It is the most symbolic of the indices, and the one which measures the subject of most intense activity on the part of the previous government in the achievement of its segregation goals.

Table 2:  Median indices of inter-group dissimilarity, 2001

Province	African– Coloured	African– Asian	African– White	Coloured– Asian	Coloured– White	Asian– White
Western Cape	75	92	93	69	92	75
Eastern Cape	84	91	93	87	88	52
Northern Cape	73	*	91	*	89	*
Free State	65	93	96	89	92	58
KwaZulu-Natal	77	88	87	64	77	80
North West	64	95	95	89	91	84
Gauteng	80	94	92	82	81	70
Mpumalanga	77	92	91	85	83	79
Limpopo	73	96	92	87	78	78
South Africa	74	92	93	79	88	74
Number of towns	228	110	245	103	224	108
Number of comparable pairs of figures in 1996 and 2001	214	96	238	90	214	94
Number of significant declines 1996–2001	131	6	38	35	51	52
Note: *Under five towns.
Source: Compiled by the author from census statistics supplied by Statistics South Africa.

Table 3:  National median inter-group indices of dissimilarity, 1985–2001

	African– Coloured	African– Asian	African– White	Coloured– Asian	Coloured– White	Asian– White
1985	84	95	93	87	93	96
1991	88	96	95	85	95	95
1996	86	93	95	84	92	89
2001	74	92	93	79	88	74
Source: Compiled by the author from census statistics supplied by Statistics South Africa.

5.1 African–White

It is notable that the African urban population of South Africa has grown by some 24,1 per cent between 1996 and 2001. At the same time the White urban population declined by 2,0 per cent. Thus the numerical extent of the growth of the African population, housed mainly in the peripheral informal townships and through the redevelopment and upgrading of existing townships, has tended to increase the levels of segregation between Africans and Whites. In contrast the decline in White urban numbers has resulted in the opening of the property market in formerly White areas thereby accelerating the pace of integration. An examination of the indices of dissimilarity would suggest these two processes are closely balanced. Thus of the 238 towns for which comparable indices could be calculated in 1996 and 2001, only 38 showed a significant (≥5-point) decline, while 18 showed a significant increase. As a result the national median value only declined from 95 to 93. Although this might at first sight appear to be insignificant, it contrasts with the static median value noted when the 1991 and 1996 censuses are compared.

Research findings in the United States suggest that the processes at work in encouraging inter-group desegregation are also place-specific (Farley & Frey, 1994; Johnston et al., 2003). It is proposed to examine some of the factors which were found to be of significance in the United States, to see if they offer any indicators for the South African experience.

Regionally, there is little difference in segregation levels between the provinces. However, the index values in the Free State are statistically significantly higher than in the remainder of the country, while those in KwaZulu-Natal are significantly lower.1 These findings follow the patterns determined in earlier censuses, and reflect the differing colonial and subsequent approaches to the segregation of the White from the African population, more particularly the separate housing of domestic workers.

It is remarkable that other factors found to be of significance in American studies such as the size and age of the town, rate of growth and proportion of Afro-Americans in the population (in South Africa, this would be ‘Africans’) were not statistically significant in the South African context.2 The impress of centralised planning remains much in evidence with a long-term heritage producing uniformity across these variables. Only the national and provincial capitals were significantly less segregated than the rest, suggesting that the continued levels of government and government-related employment offered significant opportunities for residential mobility. The anomalous position of Ulundi, the joint capital of KwaZulu-Natal, must be excluded from this observation, as the town is mono-African in terms of this study (Maharaj, 2001). In contrast, the former White university towns, which in the United States exhibited a high degree of residential integration, exhibited no significant difference when compared to other towns in South Africa. Towns formerly situated entirely within the homelands were significantly more integrated than others as there had been a longer period without urban zoning, but the numbers of Whites involved were comparatively small. Otherwise the general lack of significant relationships in South African towns and cities suggests that the legacy of the past has reduced them to a degree of uniformity from which it is difficult to escape, despite the initial recognition of signs of a resurgence of diversity within the system (Gervais-Lambony, 1999).

5.2 African–Coloured

In contrast, an examination of the African–Coloured indices suggests that significant changes have taken place in the inter-census period. The national median index value fell from 86 to 74. Furthermore, of the 214 towns for which comparable indices could be calculated for 1996 and 2001, some 131 showed a significant (≥5-point) decline. This represents a major contrast with the patterns of African–White integration. In this respect the statistics may reflect an overlap in the definition of the Coloured and African groups, as well as the higher levels of family mixing between the two groups (Martin, 1998). However, there also appears to be a significant movement of Africans into the former Coloured group areas and an overlap between the two groups in the allocation of new government housing schemes.

On a regional basis the Eastern Cape was significantly more segregated than the remainder of the country, while the Free State and North-West provinces were significantly less segregated. The Free State towns still demonstrate the influence of a relatively recent attempt to segregate the two groups, which took place as late as the 1980s. Prior to the 1984 tri-cameral constitution members of the two groups had lived mainly in the same locations under nineteenth-century legislation, a situation which was little changed under the initial apartheid legislation. In contrast, in the Eastern Cape the two groups had been separated in colonial and subsequent times through strict enforcement of the laws governing the African population and the provision of separate residential areas.

5.3 African–Asian

The African and Asian populations remained highly segregated, with the national median index only declining from 93 to 92. Only six of the 96 towns for which comparable indices could be calculated demonstrated significant declines. The continuing high levels of segregation between the two groups reflect the contrasting economic positions of the two groups and a long history of enforced separation. Significantly, the towns of KwaZulu-Natal were more integrated in this respect, reflecting the different approach to segregation in that province prior to 1950.

5.4 Coloured–White

The Coloured–White indices demonstrated limited integration between the two groups, with the national median value declining from 92 to 88. It should be noted that of the 214 towns for which comparable indices could be calculated some 51 showed significant declines, again reflecting the movement of Coloured households into the former White group areas.

5.5 Coloured–Asian

The integration between the Coloured and Asian populations showed some progress with the national median index value declining from 84 to 79. Thus, of the 90 towns for which comparable indices could be calculated 35 showed significant declines.

5.6 Asian–White

There was a significant decline in the levels of segregation between Asians and Whites between 1996 and 2001. The national median index value declined from 89 to 74, while 52 of the 94 towns for which comparable indices could be calculated registered a significant decline. The most noticeable declines took place in the two major concentrations of the Indian population, KwaZulu-Natal and Gauteng. In Gauteng growth in the Indian population was evidently accommodated through re-integration into the former White group areas.

6. Conclusion

Desegregation is taking place in South African cities, but it is progressing at a very slow pace and is both inter-group and place-specific. In terms of the key African–White index, which may be regarded as a test of the progress of residential transformation, comparatively little change was evident between 1996 and 2001. Segregation levels generally remained still close to the apartheid-era peak, but a slight decline was evident compared with the static situation between the censuses of 1991 and 1996. Particularly noteworthy are the extremely high levels of residential separation between Africans and Whites in the Free State. Indeed, for the vast majority of urban dwellers throughout the country little change is evident in their residential options and experience of segregation. KwaZulu-Natal again emerges as the most integrated province in the country, with significant (≥5-point) declines in all the median index values except that between Africans and Whites. Even that index is significantly lower than the corresponding indexes in the other provinces, reflecting the long-established patterns inherited from the colonial era. The influence of direct and indirect state employment in fostering integration in the capital cities has proved to be a one-off occurrence, immediately post-1994, as the size of the state bureaucracy is likely to be reduced rather than expanded in the future.

It is disappointing for both researcher and reader that this analysis of census data reports what did not take place rather than what did. Ten years after the repeal of the Group Areas Act and other laws enforcing urban segregation, South African towns and cities have made only limited progress towards desegregation. Nevertheless, progress has been made. However, it is noticeable that whereas there was a long-term, well-funded, government programme to create the apartheid city, there has been no equivalent state intervention aimed at its undoing and the creation of a new specifically non-racial city. It would appear, therefore, that although the processes of desegregation will continue to operate, South African cities will remain highly segregated.

The financial assistance of the National Research Foundation towards this research is acknowledged. The opinions expressed in this article and the conclusions arrived at are those of the author and are not necessarily to be attributed to the National Research Foundation.

Notes

¹The Mann–Whitney U-test for significance was adopted for comparisons between the index values in different provinces and for different urban forms.

²Spearman's rank correlation coefficient was calculated for paired urban attributes, and its significance assessed using a z score.