B1501
Title: Many regression discontinuity estimators for panel data
Authors: Likai Chen - Washington University in Saint Louis (United States)
Weining Wang - University of York (United Kingdom)
Georg Keilbar - Humboldt-University of Berlin (Germany) [presenting]
Abstract: Numerous studies use regression discontinuity designs for panel data, which may have clustered errors. The existing literature mainly focuses on estimating parameters, assuming that the treatment effects are uniform across all groups. However, in reality, treatment effects may vary among different groups. Consequently, it is unclear how to test for the significance of treatment effects when errors are clustered and treatments vary across individuals or groups. The aim is to examine the estimation and inference of multiple treatment effects when the errors are not independent and identically distributed and treatment effects vary across individuals or groups. The analytical expression for the variance-covariance structure of the estimator is derived under various dependency situations. Notably, it is found that the covariance is always smaller than the variance, indicating that the covariance can be ignored due to the localized nature of the statistics. This has an important critical value interpretation. Finally, a test to determine the overall significance of the average treatment effect (ATE) is proposed to determine whether all individuals share the same causal effect. The test relies on a high-dimensional Gaussian approximation (GA) result, which holds when the number of groups tends towards infinity.
