This paper uses household level unit record data from South Africa to examine the behavioural and welfare impacts of private and public transfers. We allow for joint endogeneity of resource variables and the expenditure shares. Our results show that crowding out of private transfers as a result of the introduction of public pensions holds only for poor households and not for the non-poor. Both private transfers and public pensions significantly reduce poverty but private transfers have a larger impact on expenditure patterns. The results also reject the hypothesis of income pooling underlying the conventional unitary model by finding that the marginal impact on expenditures are different for public pension received, private transfer received and other resources flowing into the household. The principal conclusions are robust to changes in specification.