Monitoring Teachers and Changing Teaching Practice: Evidence from a Field Experiment.∗

There is large documented variation in teacher quality, which strongly correlates with teaching activity in the classroom. But how to change teaching practice? We present results of a randomized evaluation of two different programs aimed at improving early-grade reading in poor schools in South Africa. Both programs provide teachers with structured lesson plans and supporting reading materials, but they differ in the mode of implementation. In some schools, teachers receive two two-day training sessions over the course of the year. In other schools, teachers receive monthly visits from specialized reading coaches who monitor teaching activity and provide feedback. The curriculum and expected teaching methods are the same as in the control. We track a cohort of pupils over two years and find that over this period the coaching program improved learning proficiency by 0.25 standard deviations. In contrast, the training program had a far smaller and statistically insignificant impact of 0.11 standard deviations. The impacts are also dramatically larger in urban schools: 0.42 and 0.75 standard deviations in the training and coaching arms respectively. We conclude that, under a variety of different assumptions, coaching is more cost-effective. Moreover, data from detailed lesson observations reveal that teachers in both interventions are more likely, relative to the control, to enact group-guided reading— a difficult teaching technique where children read aloud in small groups sorted by ability. Related to this technique, intervention teachers are able to provide more individual attention and pupils are more likely to read from reading booklets. These changed classroom practices are consistently larger for teachers that received coaching. Results of this paper yield insights into the appropriate combination of support required to shift teaching practice, and which practices plausibly have a greater impact on learning. ∗We are grateful for useful comments and feedback from Servaas van der Berg, and fantastic research and management support from Janeli Kotze and Mpumi Mohohlwane. The International Initiative for Impact Evaluation provided funding for the evaluation. All errors and omissions are our own. †McCourt School of Public Policy, Georgetown University ‡Department of Basic Education, South Africa