School Choice and Class-Size Effects: Unintended Consequences of a Targeted Voucher Program

Résumé

We propose a novel method to estimate education production functions on observational data in a context of school choice. We exploit panel data of schools and estimate heteroge-neous effects, while allowing for unobserved school, student, and teacher characteristics to be correlated with observed inputs. We then use this model to study the channels behind changes in observed test scores following a voucher reform in Chile. After the reform, many students left public schools, leading to a passive decrease in class size. We show that this can explain part of the policy effects as we find large class size effects for several schools, especially those that saw a decrease after the policy change.