Few studies tried to quantify the relative importance of each determinants of residential segregation. This mainly comes from a reverse causality problem which hampers the identification of the quantity of interest. In this paper, we decompose the whole change in segregation between 2001 and 2011 in South Africa by using segregation curves. We show that, even without an experimental setting (which might be impossible to obtain), identification of the causal effects can still be achieved by using the dynamics of the phenomenon. The provision of basic public services appears to be one of the main explanation of the gap observed, while differences in sociodemographic characteristics play a minor role only for the least segregated neighborhoods. Housing market is responsible for an important part only among neighborhoods intermediately integrated, while past segregation and income influence moderately segregation throughout more than half of the South African neighborhoods.