Reproducibility package for Scaling Literacy Instruction In Fragile Settings: Experimental Evidence On Teacher Coaching

A growing body of evidence shows that bundled interventions can produce large gains in foundational literacy in low-income settings. However, many of these programs rely heavily on external coaches and monitoring teams, raising the question of whether their effects can survive delivery through a country’s own, more limited administrative capacity. This paper provides experimental evidence on this question, evaluating a comprehensive early-grade reading intervention in Haiti, a low-income country with limited government capacity, that combined teacher training, coaching, supervision, and school supplies. To directly test whether the model can be delivered without external coaches, we randomly assigned schools to a costlier “Full” version, in which external coaches led teacher coaching, or a “Light” version, approximately 20 percent less costly, in which coaching was fully delegated to school principals. By endline, 18 months after the start of the intervention, we find effects of approximately 0.3 standard deviations (SD) on student reading outcomes for both versions. Critically, the Light version—which replaces external coaches with existing school leadership, the hardest-to-scale input in this class of interventions—achieves gains statistically indistinguishable from the Full version across every dimension of learning measured. These findings suggest that reliance on external coaching may not be a binding constraint on scaling comprehensive literacy programs, and point to a practical, lower-cost pathway for delivering sustainable pedagogical support within existing school leadership structures in fragile settings.