We estimate the impact on objective measures of financial literacy of a 10-hours financial education program among 15-year old students in compulsory secondary schooling. We use a matched sample of students and teachers in Madrid and two different estimation strategies. Firstly, we use reweighting estimators to compare the performance in a test of financial knowledge of students in treatment and control schools. In another specification, we use school fixed-effect estimates of the effect of the course on change in the score in tests of financial knowledge. The program increased treated students' financial knowledge by between one fourth and one third of a standard deviation. We uncover heterogeneous effects, as students in private schools did not increase their knowledge much, possibly due to a less intensive implementation of the program. Secondly, we analyze the bias that arises because the set of schools that participate in financial literacy programs is not random. Such selection bias is estimated as the pre-program performance in financial PISA of students in applicant schools relative to a nationally representative sample of schools. We then study if estimators that condition on school and parental characteristics mitigate selection bias.