This paper asks whether income mobility in South Africa over the last decade has indeed
been as impressive as currently thought. Using new national panel data (NIDS),
substantial measurement error in reported income data is found, which is further
corroborated by a provincial income data panel (KIDS). By employing an instrumental
variables approach using two different instruments, measurement error can be
quantified. Specifically, self-reported income in the survey data is shown to suffer from
mean-reverting measurement bias, leading to sizable overestimations of income
convergence in both panel data sets. The preferred estimates indicate that previously
published income dynamics may have been largely overestimated by as much as 77% for
the national NIDS panel and 39% for the provincial KIDS panel. Overall, income
mobility appears much smaller than previously thought, while chronic poverty remains
substantial and transitory poverty is still very limited in South Africa.