This paper presents research on South African household expenditure share behaviour. The research examines whether or not a theoretical and empirical model, which has been successful in explaining expenditure shares in Australia, is valid when applied to South African data. The primary conclusion of the research is that expenditure shares in South Africa do not conform to the assumptions set out in the model. Although there are many potential reasons for non-conformity, this paper provides evidence that the estimates produced within the AID System and the MAID System suffer from heteroskedasticity and non-normality. Therefore, in order to improve the understanding of spending behaviour by South African households, models will have to be specifically developed to deal with the idiosyncrasies of South African data.