This study estimates the effect of between-class ability grouping on Chilean secondary students' academic achievement. We rely on a structural feature of the school system: A considerable number of students who complete primary school must change schools to start their secondary education. Cardinality matching was performed to account for confounding variables. Sensitivity analysis was performed to address the fact that differences in unobserved variables could bias our findings. Results show that attending a school that groups students by ability causes a reduction in the average scores obtained in 10th grade of 0.07 SD for reading and 0.08 SD for mathematics. Students matched by their observed covariates could differ in their odds of attending a school with or without ability grouping by 15% for mathematics and 10% for reading, without altering our conclusions. Disadvantaged students are the most affected, whereas best-performing students do not benefit from this policy either.