Using data from a detailed chronic poverty survey of three South African communities, this paper compares the correlations between traditional (i.e. income and expenditure) and wealth-based measures of poverty in ranking households as poor as well as their ability to explain additional qualitative measures of persistent poverty such as household hunger. We find significant locational differences in terms of the composition of household wealth measures and this complicates the derivation of appropriate wealth indices. Traditional money-metric measures of poverty that abstract from location appear to explain short-term measures of deprivation like household hunger relatively well, and consistently capture the bottom and top deciles of the distribution. On their own wealth-based measures appear less suited to explaining household hunger, suggesting more liquid based measures for short-term indicators are more appropriate.