Teacher supply and the quality of schooling in South Africa. Patterns over space and time

The paper addresses policy questions in South Africa’s education system using a newly merged 1999 to 2013 panel of data that includes school enrolments by grade, staff details from the payroll system, examination and test results and the geo-coordinates of schools. This combination of data, which is seldom used, at least in developing countries, permits new and important knowledge about a schooling system to be uncovered. Whilst policy conclusions are South Africa-specific, the methods would be largely transferable to other contexts. It is shown that school data can complement official population data with respect to the monitoring of within-country migration and in determining the rate of urbanisation. An approach for calculating the viability of small schools in a context of migration out of rural areas is presented, using assumptions around maximum distance to be travelled by pupils and the degree to which multi-grade teaching by teachers should be permitted. Cost reductions associated with a reduced presence of small schools, and greater economies of scale associated with larger schools are found to be smaller than what is generally assumed. Correlations between pupil under-performance and the under-staffing of schools are found to be higher at the primary than the secondary level, apparently confirming the greater importance of personal interaction with a teacher for younger pupils. Between-school movements of pupils other than those associated with urbanisation are found to be high, and highly variable across districts. This further complicates the allocation of publicly paid teachers. An approach to gauging whether teachers avoid moving to schools on the other side of provincial boundaries is presented. It is confirmed that movement across provinces, which are the employers of teachers, is restricted, creating further obstacles to efficient teacher allocation. It is confirmed that teachers tend to move to better performing schools, but that the performance signals that influence this movement are often inaccurate and a few years old.