Local Polynomial Order in Regression Discontinuity Designs

Treatment effect estimates in regression discontinuity (RD) designs are often sensitive to the choice of bandwidth and polynomial order, two important ingredients widely used local methods. While Imbens Kalyanaraman Calonico, Cattaneo, Titiunik provided guidance on bandwidth, sensitivity order still poses a conundrum RD practitioners. It is understood econometric literature that applying argument bias reduction does not help resolve this conundrum, since it would always lead preferring higher orders. We therefore extend frameworks use asymptotic mean squared error estimator as criterion guide selection. show Monte Carlo simulations proposed selection procedure performs well, particularly large sample sizes typically found empirical applications. This extends easily fuzzy kink designs.