In 2001, the Government of South Africalaunched the Dinaledi Schools program, aimed at increasingthe number of African and Indian secondary students enteringand succeeding in mathematics and physical sciences SeniorCertificate exams. The Dinaledi program provides selectedhigh schools with a combination of supplementary inputs,including teachers, training, textbooks and calculators aswell as close monitoring by the National Department ofEducation. This paper estimates the impact of the Dinalediprogram for about 350 schools assigned to the program in thecourse of its expansion in 2005. It exploits administrativepanel data for the years 2005-07 using adifference-in-difference estimation strategy. This iscombined with a propensity score matching approach thatcompares Dinaledi schools with schools that are observablysimilar prior to the implementation of the program.