This article utilises a class of poverty measures to determine the potential cost to the fiscus, in 1995 rands, of alleviating poverty in South Africa. The simulations are undertaken for both households and individuals according to the different covariates of poverty. The study found that the commitment required from the state to reduce poverty is fairly modest, albeit within the parameters of very strict assumptions. In addition, the article illustrates that individual and household-level data impart differential poverty information, which is important for policy prescriptions. Finally, it is evident that for state targeting purposes, the nature of household poverty is fairly easily reduced to a small subgroup of labour market-defined household types.