Abstract |
The quality of data on employment income is explored using Tanzanian and Zambian household survey datasets. The extent of missing and implausible income data is assessed and four different methods are applied to impute missing or implausible values. The four imputation methods are also applied to artificial missing data for Tanzania and Zambia, and—using one approach—for a South Africa dataset. Post-imputation results are assessed. It is argued that the treatment of missing data cannot be generalized, and that tax compliance should also be taken into account when assessing the validity of a microsimulation model’s simulation of direct taxes. |