Developing neighbourhood typologies and understanding urban inequality: a data-driven approach

Figures & data

Figure 1. Selecting k using gap statistic.

Note: The number of Monte Carlo (‘bootstrap’) samples is restricted to 100.

Figure 2. Total within-cluster sum of squares for iterations of k-means clustering.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Table 1. Sensitivity test results.

	% SAL places in the same cluster
Range normalization	89
Lloyd (1981) initialization method	98
MacQueen (1967) initialization method	91
Excluding correlated variables (with a threshold of 0.8 and −0.8)	89
Excluding correlated variables (with a threshold of 0.6 and −0.6)	85

Source: Census 2011 Small Area Layer; authors’ own estimates.

Table 2. Neighbourhood types and selected descriptors.

Neighbourhood type		Selected descriptors^a	Number of SAL places (% of total)
#	Label
1	Places of wealth/suburbia	Predominantly white; highest employment rates; highest average incomes; large freestanding houses	5640 (19.1%)
2	Places of upward mobility/the emerging middle class	Racially diverse; second highest average incomes; young; highest levels of education; apartments; located in the margins of type 1 neighbourhoods	2888 (9.8%)
3	Places of withdrawal/the affluent pensioners	White; older; relatively high average incomes; single-person households; located as pockets in patches of type 1 neighbourhoods	343 (1.2%)
4	Racial enclaves/the township middle class	Racialized grouping of residents; predominantly Coloured Black African and Indian or Asian; relatively high employment rates although lower levels of education; former ‘group areas’	4356 (14.7%)
5	Places of unemployment/large established townships	Black/African; low to medium average incomes; lower levels of education and second lowest employment rates; subdivided dwellings; access to basic services	7019 (23.7%)
6	Places of transition/poor neighbourhoods with backyard dwellings	Black/African; backyard dwellings without access to basic services; domestic and foreign migrants; diverse locations including inner-city and outlying settlements; particularly common in Johannesburg and Tshwane	3638 (12.3%)
7	Places of entrenched poverty/peri-urban traditional settlements	Black/African; very low average incomes; low levels of education; low population densities; poor access to basic services due to relative isolation; former homelands	2362 (8.0%)
8	Places of extreme poverty and vulnerability/informal settlements	Very high unemployment rates; very low average incomes; no durable assets; shack structures; occupy areas not formally designated for residential use; poor access to basic services	2450 (8.3%)

Note: ^aThe selected descriptors represent the relative concentration for each type but do not necessarily imply a majority.

Figure 3. Johannesburg (left) and Ekurhuleni (right).

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 4. Tshwane.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 5. Cape Town.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 6. eThekwini.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 7. Nelson Mandela Bay.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 8. Buffalo City.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 9. Mangaung.

Source: Census 2011 Small Area Layer; authors’ own estimates.

Figure 10. Racial breakdown by neighbourhood type.

Source: Census 2011; authors’ own estimates.

Figure 11. Social status and neighbourhood clusters.

Note: Kernel density distributions.

Source: Census 2011; authors’ own estimates.

Figure 12. Population distribution of neighbourhood types across metros.

Note: JHB, Johannesburg; EKU, Ekurhuleni; TSH, Tshwane; CPT, Cape Town; ETH, eThekwini; NMB, Nelson Mandela Bay; BC, Buffalo City; and MAN, Mangaung.

Source: Census 2011; authors’ own estimates.

Lloyd, S. P. (1981). Least squares quantization in PCM. EEE Transactions on Information Theory, 28(2), 129–137. https://doi.org/10.1109/TIT.1982.1056489

 Web of Science ®Google Scholar

MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations. Berkeley Symposium on Mathematical Statistics and Probability, 5(1), 281–297.

 Google Scholar