Semiparametric Continuous Difference-in-Differences with Causal Forests: The Case of Charter School Effect Heterogeneity
Nov 8, 2026ยท,
ยท
0 min read
Feng Chen
Douglas N. Harris
Nicholas Lacoste

Abstract
We develop a difference-in-differences causal forest for estimating heterogeneous treatment effects in settings with continuous and/or time-varying treatment. We then apply this method to the case of U.S. charter schools. While prior research suggested charter school effectiveness was most closely associated with student demographics, we find that this mostly reflects low baseline achievement levels. Higher school funding is the only policy variable that seems to improve charter effectiveness, while two dozen other factors are unrelated to charter effects. Our results suggest that our flexible methods better capture the many linear and nonlinear heterogeneity dimensions present in common policy settings.
Type