The recent economic recession has led states to significantly reduce expenditures for higher education (Grapevine, 2012). As spending cuts typically result in reduced course offerings, one response by colleges and universities is to increase class size in order to accommodate additional students. In California’s community colleges, median class size increased from approximately 27 students to 31 students between 2007 and 2010 (Bohn, Reyes, & Johnson, 2013). The effects of class size have been studied considerably in the K-12 literature, most commonly finding an inverse relationship between class size and academic outcomes (Angrist & Lavy, 1999; Chetty et al., 2011; Fredriksson, Ockert, & Oosterbeek, 2013; Krueger & Whitmore, 2001), though doubts remain about the viability of implementing smaller class size at scale (Jepsen & Rivkin, 2009; Sims, 2008). Significantly less research has evaluated the effects of class size in postsecondary settings, though the results to date are consistent with previous findings in the K-12 literature. Bandiera, Larcinese, and Rasul (2010) and Kokkelenberg, Dillon, and Christy (2008), relying primarily on identification via teacher and student fixed effects, find that larger class size is associated with lower course performance, with effects largely concentrated among the smallest (less than thirty) and largest (over 100 students) classes.

This paper uses data from a large California community college to assess the impact of class size on student outcomes. As simple comparisons of large and small classes are likely to be biased (e.g., a teacher with a reputation as an easy grader may draw larger numbers of students), I instrument for class size using fluctuations in course size that occur after the beginning of the semester. More precisely, in previous work I find that class size in community college courses changes dramatically within each semester, generally increasing as teachers allow students to enroll above official course capacity, before subsequently decreasing as many students withdraw from the course. Using a detailed history of every registration attempt over multiple years, I first identify those students who successfully enroll in a course prior to the beginning of the semester, and then determine the number of these students who later the drop the course. Although students who drop are frequently replaced by wait listed students, the number of total drops is found to be inversely correlated with the final class size, conditional on teacher and course fixed effects. In other words, I compare the academic outcomes of early enrollees (in order to avoid potential endogeneity associated with late enrollees who may be shopping for smaller courses) who took the same course from the same teacher, but are in smaller or larger classes based on the number of originally enrolled students who chose not to take the course. This approach allows me to construct causal estimates of the effect of class size on course outcomes, which are independent of unobserved differences across class types. I further explore how class size effects are moderated by student demographics and prior performance.