Policy determinants of school outcomes under model uncertainty: evidence from South Africa

In this paper, we assess the determinants of secondary school outcomes in South Africa. We use Bayesian model averaging techniques to account for uncertainty in the set of underlying factors that are chosen among a very large pool of explanatory variables in order to identify the subset of explanatory variables most strongly associated with the dependent variable. Our analysis indicates that the socio-economic background of pupils, demographic characteristics such as population groups (Black and White), as well as geographical locations account for a significant variation in pupils’ achievement levels. We also find that the most robust policy determinants of pupils' test scores are the availability of a library at school, the use of IT in the classroom as well as school climate.