Job market paper
Good Schools: inferring short- and long-run returns to school quality from observational data
Improving disadvantaged students' access to good schools is the goal of many education policies, but there are still open questions about how much schools can affect student learning. Using observational data on 500,000 student moves between grades 4 and 8 in North Carolina, I capture the returns to school quality on math test scores in a mover design that allows for heterogeneous effects of moving across individuals and over time while accounting for unobserved student heterogeneity. The average effect on standardized math test scores of moving from a school in the bottom 25% to a school in the top 25% of predicted school quality after one year is 0.25 standard deviations, with larger effects for students who moved in earlier grades. In contrast to theoretical models of neighborhood and school sorting that predict school quality and student income will be correlated in equilibrium, school effectiveness is distributed evenly across measures of students' economic background. Importantly, I find the effect of attending a good school fades out unless students stay in a good school.