Analysing time-dependent independent variables requires the use of process-oriented statistical models. Yet social scientists – especially those in poor countries – have often had to use data collected at a single point in time, making their task difficult. Making several assumptions about the covariates, the present study uses survival analysis and other statistical techniques to analyse the 1996 South African population census data and examine the effects of selected independent variables on the timing of parenthood in the country. It was found that the onset of parenthood occurs late in South Africa compared with the pattern in most other African societies. While education plays a role in the postponement of parenthood within racial groups, it fails to explain the differences between African and Coloured women on the one hand, and White and Asian women on the other hand, a finding that suggests the existence of two regimes of family formation in South African society.