Regression Discontinuity Design with Potentially Many Covariates

Abstract

This paper studies the case of possibly high-dimensional covariates in the regression discontinuity design (RDD) analysis. In particular, we propose estimation and inference methods for the RDD models with covariate selection which perform stably regardless of the number of covariates. The proposed methods combine the local approach using kernel weights with ℓ 1 -penalization to handle high-dimensional covariates, and the combination is new in the literature. We provide theoretical and numerical results which illustrate the usefulness of the proposed methods. Theoretically, we present risk and coverage properties for our point estimation and inference methods, respectively. Numerically, our simulation experiments and empirical example show the robust behaviors of the proposed methods to the number of covariates in terms of bias and variance for point estimation and coverage probability and interval length for inference.

A guide to regression discontinuity designs in medical applications

“…A principled alternative is to use the robust bias corrected confidence intervals proposed by Calonico et al, 34 and later extended to other settings. 35,[38][39][40][41] Robust bias corrected confidence intervals modify the classical confidence intervals CI in two ways: (i) the point estimator τSRD is debiased by including an estimate of the leading misspecification error (denoted by B), and (ii) the variance estimator V is increased to incorporate the contribution of the bias correction step to the overall variability of the confidence interval (denoted by Ŵ). Thus, the robust bias corrected confidence intervals take the form…”