In this article, the authors aim to identify the causal effect of the use of graded problem sets on academic performance of Spanish students. The identification strategy relies on an experiment in which the authors exploit variation arising from observing the performance of nearly 300 students taking the same class during the same semester and with the same instructors. Academic performance is measured through a multiple choice final exam in which some questions are related to graded problem sets and others are related to non-graded problem sets given through the semester. After accounting for potential biases and selection concerns, the results show that graded problem sets increase test scores by eight percentage points, or close to a letter grade. (Contains 3 tables and 4 notes.)