Regression Discontinuity Designs for High-frequency TSCS Data

• In Event: Models for TSCS, Time, and Duration Data

Abstract

In this paper, we discuss the promises and pitfalls of using regression discontinuity designs (RDDs) to estimate treatment effects for time-series cross-sectional (TSCS) data. We show that RDDs using time as the running variable yield valid estimates for the instant treatment effects on the treated when data are measured frequently. We develop a bias-correction estimator for RDDs with TSCS data when the data are infrequent and modify the bandwidth selection algorithm for classic RDDs to account for the intertemporal correlation among observations. We apply the method to analyze the effect of a policy change in China on environmental regulations.

Authors

• Ye Wang, New York University

• Yiqing Xu, Stanford University

• Junlong Aaron Zhou, New York University