Segregation and socio-economic sorting in Gauteng

Segregation research about South African cities prompts us to reflect on the achievements and limits of spatial transformation (Hamann, 2024). Maps and segregation indices illustrate the remarkable transformations in inner cities and in former whites-only suburbs, and they also show how townships remain racially homogenous (Hamann and Ballard 2017). Interpretations of such patterns assume that racial integration increases with the income level of a residential area (Ballard, Hamann, and Mkhize, 2021). Since it has been more than three decades since racial segregation was a requirement of law, enduring racial patterns are largely the result of affordability. In Gauteng, the pace of desegregation of former whites-only suburbs is closely linked to their class-based nature, where access now depends on income more than anything else (Crankshaw, 2022). Affluent suburbs are 'desegregated' because higher higher income job categories have deracialised, while townships are segregated because most low-income earners are black/African.

The August Map of the Month uses a bivariate mapping technique to examine both the racial diversity and income level of each ward (Figure 1). This map of the month goes beyond depicting the degree of racial mixing or segregation alone, and integrates income level into the visualisation. The map highlights the small number of wards in Gauteng where racial diversity and household income are both very high (the darkest shade of blue) alongside the many wards in Gauteng where racial diversity and household income are both very low (the light grey shading). Between these opposite ends of the matrix, the map also highlights interesting spatial variations and clustering of the different relationships between racial diversity and household income. The patterns are discussed in more detail below.

Figure 1: A classification of Gauteng wards based on the relationship between mean household income and racial diversity of the ward.

Method

The map (Figure 1) is the result of various calculations. Ward-level population group and household income data from the GCRO QoL 4, QoL 5, and QoL 6 surveys (GCRO, 2021, 2019, 2016) were used for this analysis. The method of combining responses from different QoL surveys is unconventional, given that dynamics can change over time. However, this method relies on an average indication of demographic composition and household income in the wards, which is less likely to change drastically over a short period of time. Over the course of the QoL surveys, an average of 80 respondents were interviewed in each ward in Gauteng. The original weight variable from each survey was retained in the analysis.

Drawing on the demographic data, an Entropy Index (Parry and van Eeden, 2015) was used to calculate racial diversity in Gauteng. The Entropy Index includes all four major population groups in the calculation of the index (black African, coloured, Indian/Asian, and white residents). The main benefit of the Entropy Index is that it can be distinctly calculated for separate, small administrative units which can then be mapped. The resulting racial diversity value has a maximum of 1.386 (or ln(4) for four population groups) and can therefore be represented as a percentage of the maximum possible diversity.

The household income responses were recoded to the midpoints of the income categories. The midpoints for QoL 4 (done in 2014/15) and QoL 6 (done in 2020/21) were then adjusted for inflation to 2017 prices before calculating the mean household income to correspond with the completion of QoL 5 (done in 2017/18).

After creating the racial diversity and mean household income variables, a bivariate mapping tool in ArcGIS Pro was used to combine the two variables. The result is a classification of wards based on the relationship between racial diversity and mean household income per ward, as illustrated in Figure 2, below. A ward that is shaded in the same light grey colour as ‘block A’ has less than 15% of the maximum possible racial diversity and a mean household income of less than R10 000 per month. On the other side of the matrix, a ward that is shaded in the same dark blue colour as ‘block B’ has more than 60% of the maximum possible racial diversity and a mean household income of more than R45 000 per month.

Figure 2: Matrix classifying wards based on the relationship between the racial diversity and the mean household income of the ward.

Discussion

The areas on Figure 1 that have low racial diversity and low mean household income are mostly in townships in Gauteng. This includes places like Hammanskraal, Mamelodi, Tembisa, Daveyton, Soweto, and Sebokeng. These spaces were historically designated for black African residents, have not desegregated in the post-apartheid era, and remain low-income residential areas on the periphery of the urban core.

The areas on Figure 1 that have high racial diversity and high mean household income include a few suburban areas that are concentrated in a small part of the province. The suburban areas include those in the north-western parts of Johannesburg and some parts in the south of Tshwane. The ribbon of suburban cluster housing development on the north-western edge of Johannesburg stands out as an area with high racial diversity and high mean household income.

These areas broadly resemble the “middle classing in Roodepoort” that Ivor Chipkin (2012) writes about. Cluster housing developments increased dramatically in these areas since 2003 due to the price, location, security, and internal organisation benefits offered by this type of development. However, Chipkin (2012), posed an important question about whether an area such as Roodepoort can be considered as middle-class? The areas are, without a doubt racially mixed, but Figure 1 suggests that the mean household income remains high. From a municipal perspective, it is clear that high-diversity-high-income areas are present in Johannesburg and Tshwane, but high-diversity-high-income areas are less prominent in Tshwane. This suggests that the pace and nature of race and class desegregation are different between municipalities - something that can be explored in future research.

Wards that are categorised in shades of pink have very low racial diversity, but they vary along the income spectrum. For example, a ward in Pretoria East (south of Mamelodi) has very low racial diversity but very high mean household income. This could be an example of residential areas that have not desegregated due to expensive properties. Bryanston, in Johannesburg, would be another example of an area with high income and somewhat lower racial diversity. Similarly, wards that are shaded in light pink have low racial diversity, but have a mean household income of more than R10 000 per month. For example, various wards in and around Soshanguve have this characteristic. This signals the presence of interracial income inequality among black African residents living in the same township.

Wards with a light blue shading have very low mean household incomes, but they vary along the spectrum of racial diversity. Those with higher levels of racial diversity often buffer areas of other extremes. For example, the area north of Soweto going towards Roodepoort or the area south of Mamelodi going towards Silverlakes. In other instances, these low-income-high-diversity areas are on the periphery of the province.

Adding a microscale perspective

The ward-level analysis is a useful starting point to highlight the spatial patterns of race and class mixing in Gauteng. However, dynamics within a ward can vary substantially and a microscale perspective would provide further insight, as argued in the recent GCRO publication about microscale segregation and socio-economic sorting in Gauteng (Hamann, 2024).

Microscale income data is rarely available, but privately sourced Neighbourhood Lifestyle Index (NLI) data can serve as a proxy for the income perspective used in the ward-level analysis above. Population group estimates and NLI calculations are available for hexagons in Gauteng (400-meter corner-to-corner polygons) and provide a valuable microscale perspective.

Figure 3 illustrates that while an entire area has high racial diversity and high mean household income at a ward level, there are important variations at the micro-scale. For example, the gated area of Randpark is categorised as a high-diversity-high-income area at macro and micro-scales. However, on a microscale it is clear that Randpark has different characteristics than Windsor East and West where racial diversity is still relatively high, but NLI is somewhat lower. Similarly, in the adjacent areas where hexagons are shaded in purple, it is clear that NLI remains high, but that racial diversity is lower - signaling slower desegregation in some instances. Figure 3 also illustrates the similarity between the patterns obtained from GCRO QoL data at a ward level and privately sourced population data at a hexagon level. The two data sources validate each other in terms of the patterns for race and class mixing in Gauteng.

Figure 3: Race and class mixing in Johannesburg, viewed at different geographic scales.

Conclusion

This analysis shows that while racial desegregation has taken place in many parts of Gauteng, the legacy of apartheid remains in the patterns of socioeconomic sorting. While many areas with high household income are also racially diverse, there are high-income areas that still do not desegregate. Segregation and desegregation are thus closely related to inequality. The spatial separation of income groups is a defining feature of racial segregation, which is accentuated by patterns of urban growth and the nature of the housing market (in the way it separates residents based on the affordability of areas).

References

Ballard, R. and Hamann, C. (2021). ‘Income inequality and socio-economic segregation in the City of Johannesburg’. In M. van Ham, T. Tammaru, R. Ubarevičienė and H. Janssen (Eds.), Urban socioeconomic segregation and income inequality: A global perspective (pp. 91–109). Springer, Cham. https://doi.org/10.1007/978-3-030-64569-4_5.

Ballard, R., Hamann, C. and Mkhize, T. (2021). ‘Johannesburg: Repetitions and disruptions of spatial patterns’. In A. Lemon, R. Donaldson and G. Visser (Eds.), South African urban change three decades after apartheid: Homes still apart? (pp. 35–55). Springer, Cham. https://doi.org/10.1007/978-3-030-73073-4.

Chipkin, I. (2012). Middle classing in Roodepoort: Capitalism and social change in South Africa. PARI Long Essays No. 2. Johannesburg: Public Affairs Research Institute (PARI). http://www.gapp-tt.org/wp-content/uploads/2020/06/PARI-L.E.-2-middle-classing-in-roodepoortfinal-edited-version-5June20121-1.pdf.

Crankshaw, O. (2022). Urban inequality: Theory, evidence and method in Johannesburg. London: Bloomsbury Publishing. https://www.bloomsbury.com/uk/urban-inequality-9781786998910/.

GCRO (Gauteng City-Region Observatory). (2016). Quality of Life Survey IV (2015/16) [Dataset]. Version 1. Johannesburg and Cape Town: GCRO and DataFirst. https://doi.org/10.25828/w490-a496.

GCRO (Gauteng City-Region Observatory). (2019). Quality of Life Survey V (2017/18) [Dataset]. Version 1.1. Johannesburg and Cape Town: GCRO and DataFirst. https://doi.org/10.25828/8yf7-9261.

GCRO (Gauteng City-Region Observatory). (2021). Quality of Life Survey 6 (2020/21) [Dataset]. Version 1. Johannesburg and Cape Town: GCRO and DataFirst. https://doi.org/10.25828/wemz-vf31.

Hamann, C. (2024). An analysis of microscale segregation and socio-economic sorting in Gauteng. GCRO Occassional Paper, April 2024. https://doi.org/10.36634/DZPT2443.

Parry, K. and van Eeden, A. (2015). ‘Measuring racial residential segregation at different geographic scales in Cape Town and Johannesburg’. South African Geographical Journal, 97(1), 31–49. https://dx.doi.org/10.1080/03736245.2014.924868.

Map design: Jennifer Murray

Edits and input: Richard Ballard and Graeme Götz.

Suggested citation: Hamann, C. (2024). Segregation and socio-economic sorting in Gauteng. Map of the Month. Gauteng City-Region Observatory. August 2024. https://doi.org/10.36634/MIJV2656.