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

Neighbourhoods affect people's livelihoods, and therefore drive and mediate intra-urban inequalities and transformations. While the neighbourhood has long been recognized as an important unit of analysis, there is surprisingly little systematic research on different neighbourhood types, especially in the fast-growing cities of the Global South. In this paper we employ k-means clustering, a common machine-learning algorithm, to develop a neighbourhood typology for South Africa's eight largest cities. Using census data, we identify and describe eight neighbourhood types, each with distinct demographic, socio-economic, structural and infrastructural characteristics. This is followed by a relational comparison of the neighbourhood types along key variables, where we demonstrate the persistent and multi-dimensional nature of residential inequalities. In addition to shedding new light on the internal structure of South African cities, the paper makes an important contribution by applying an inductive, data-driven approach to developing neighbourhood typologies that advances a more sophisticated and nuanced understanding of cities in the Global South.