Existing literature using South African censuses reports an increase in both poverty and inequality over the 1996 to 2001 period. This paper assesses the robustness of these results to a number of weaknesses in the personal income variable. We use a sequential regression multiple imputation approach to impute missing values and to explicitly assess the influence of implausible income values and different rules used to convert income that is measured in bands into point incomes. Overall our results for 1996 and 2001 confirm the major findings from the existing literature while generating more reliable confidence intervals for the key parameters of interest than are available elsewhere.