Using administrative records from a UK university, we present evidence on the effects of class size on students' test scores. We estimate non-linear class size effects controlling for unobserved heterogeneity of students and faculty. We find that: ("i") at the average class size, the effect size...