Assessing the quality of the income data used in SAMOD, a South African tax-benefit microsimulation model

In this paper we explore the income data in two surveys that underpin a South African tax-benefit microsimulation model. The simulated taxes and benefits using each dataset are compared with each other and with administrative data for a common time point. We explore discrepancies between the simulated and administrative data on personal income tax, with reference to the distribution of tax payers and the amount of tax simulated. Both surveys suffer from unit missing or item implausible cases for high income individuals. The paper concludes by highlighting the potential for administrative data to further enhance the quality of the model's underpinning dataset(s).