Moving towards cost-effective delivery models of teacher coaching: evidence from field experiments in South Africa

Structured learning programmes have emerged as a promising way to address the low learning levels observed in many developing countries. The delivery model of these programmes matters, and on-site teacher coaching has been recommended to be highly effective, especially in early grade literacy. In this paper, we report on a series of government-led randomised experiments in South African primary schools that build on each other to test various models of teacher training and coaching. We find that sustained on-site coaching is more cost-effective (0.41 SD increase in test scores per USD 100) than either short coaching interventions (no significant impact) or centralized teacher training workshops (0.23 SD increase in test scores per USD 100). The ability to scale on-site coaching, however, is an open question. In the latest experiment, therefore, a virtual coaching programme was compared to on-site coaching. The focus of this paper is on the midline evaluation results from this experiment. After one year of intervention, virtual coaching was no less effective than on-site coaching at improving both the instructional practice of teachers and the targeted literacy outcomes of children. This points to the potential for technological innovations to enable wider rollout of coaching programmes, even in contexts where teachers are not familiar with new technologies.