This is from an excellent new paper by de Kadt and Sands in Political Behavior:

This study investigates how political behavior is shaped by one central outcome of segregation—racial isolation—in post-apartheid South Africa. We focus on the relationship between sustained white isolation-the probability of a white person encountering a non-white person in their local context (Massey and Denton 1988)-on white racial voting. South African elections have been marked by some degree of racial voting since the advent of democracy in 1994; white South Africans have typically voted for “white” parties such as the Democratic Party, the New National Party, the Freedom Front Plus, and the recently conglomerated Democratic Alliance. Scholars of the region have gone so far as to refer to South African elections as a “racial census” (Mattes 1995; Johnson and Schlemmer 1996; Lodge 1999; Ferree 2006, 2010). Our central contribution to this literature is to show that racial voting among white South Africans tends to increase as a function of local white isolation.

……. White South Africans living among other whites are less likely to vote across racial lines, measured as voting for the majority-black African National Congress. At the ward level, greater white isolation in post-apartheid South Africa, even conditional on ward demographics and economic status, leads to lower aggregate support for the ANC. At the individual level, whites living in areas of high white isolation, where inter-racial mixing is limited, are less likely to vote for the ANC than whites who live in less segregated areas.