In 2001, the Government of South Africa launched the Dinaledi Schools program, aimed at increasing the number of African and Indian secondary students entering and succeeding in mathematics and physical sciences Senior Certificate exams. The Dinaledi program provides selected high schools with a combination of supplementary inputs, including teachers, training, textbooks and calculators as well as close monitoring by the National Department of Education. This paper estimates the impact of the Dinaledi program for about 350 schools assigned to the program in the course of its expansion in 2005. It exploits administrative panel data for the years 2005-07 using a difference-in-difference estimation strategy. This is combined with a propensity score matching approach that compares Dinaledi schools with schools that are observably similar prior to the implementation of the program.