In many countries, students are selected into higher education institutions based on their achievement at high school, measured by scores given by their teachers and obtained on final national exams. This paper compares these two measures in terms of their ability to predict students' success in higher education. Accounting for sample selection problems, we find that the high school score is a stronger predictor of students' performance at university. However, the score obtained in the mathematics national exam still adds information. We conclude that both measures are complementary, a result that is relevant for university recruitment and selection policies.