Heterogeneous treatment effects in the low track: Revisiting the Kenyan primary school experiment

I present results from a partial re-analysis of the Kenyan school tracking experiment first described in Duflo, Dupas and Kremer (2011). My results suggest that, in a developing country school system with state-employed teachers, tracking can reduce short-run test scores of initially low-ability students with high learning potential. The highest scoring students subjected only to the tracking intervention scored well below comparable students in untracked classrooms at the end of the intervention. In contrast, students assigned to tracking under the experimental alternative teacher intervention experienced gains from tracking that increased across the outcome distribution. These alternative teachers were drawn from local areas, exhibited significantly higher effort levels and faced different incentives to produce learning. I conclude that although Pareto-improvements in test scores from tracking are possible, they are not guaranteed.