Regression Discontinuity Design with Endogenous Covariates
Vol.24, No.4, December 2013, 320–337
Kyoo il Kim
(Department of Economics, University of Minnesota)
Abstract
Empirical researchers often include observed covariates in the Wald type implementation of Regression Discontinuity Design (RDD) estimators. When those included covariates are endogenous, we find that the resulting RDD estimator suffers from a larger asymptotic bias than the estimator with exogenous covariates but it is still consistent. We further show that the order of bias increase due to the endogeneity is the same order of bias reduction due to the inclusion of relevant endogenous covariates.