Evidence from most developing countries suggest that parents have a preference for sons over daughters. This is know as son preference. This paper uses individual level unit record data to test the son preference hypothesis in South Africa. We use an accelerated hazard model to estimate the duration between successive births and our results indicate that son preference exists only for the Indian community in South Africa. Indian households prefer to have a higher duration between children following the birth of a son, irrespective of the number of children they already have. For the rest of the population, there is either no son preference or in some cases a weak preference for daughters. Our results appear to refute the usual arguments for son preference (including support for elderly parents and contribution to household income) and instead suggest the importance of religious beliefs and social customs (dowry system, lineage, familial and kinship ties etc.) in directing parental preferences.
