Kenya Girls' Primary School Merit Scholarship Program
MIT Poverty Action Lab (Duflo, Dupas, Kremer) · Western Kenya · 2004
Summary
The Kenya Girls' Scholarship experiment found that merit scholarships raised test scores substantially — but the most striking finding was the spillover to boys, who were explicitly ineligible for the incentive. Boys in treatment schools outperformed boys in control schools by 12.5 percentile points, suggesting that the scholarship changed the learning environment in the classroom, not just individual effort. The likely mechanism: teacher behavior changed. Teachers in treatment schools reported spending more time on academic content, called in more substitute teachers when absent, and received higher satisfaction ratings from students. An incentive aimed at girls appears to have shifted the entire classroom norm toward academic effort. The experiment illustrates how individual incentive programs can produce system-level effects that naive evaluations would miss.
Research question
"Does offering merit-based scholarships to the top-performing girls in a school district improve educational outcomes — and do competitive incentives spill over to non-eligible students?"
Methodology
Intervention
Girls in 127 schools in two Kenyan districts were told that those who scored in the top 15% on the government exam at year-end would receive a scholarship covering secondary school fees for two years, plus school supplies. Boys in the same schools received no incentive.
Assignment
School-level randomized controlled trial; 64 schools assigned to scholarship treatment, 63 to control
Sample size
Approximately 7,000 students across 127 schools
Primary outcome
Standardized test scores in reading and mathematics; dropout rates; attendance
Effect estimate
Girls in treatment schools improved test scores by 19.5 percentile points at year-end; boys in treatment schools also improved significantly (12.5 percentile points), despite being ineligible for the scholarship — indicating positive peer spillovers
Decision
Results contributed to evidence base for conditional incentive programs in education; spillover finding influenced design of subsequent programs to consider whether competitive incentives elevate or depress peer performance
Girls in treatment schools improved test scores by 19.5 percentile points at year-end; boys in treatment schools also improved significantly (12.5 percentile points), despite being ineligible for the scholarship — indicating positive peer spillovers