Sebastian Calonico, Matias Cattaneo and Rocio Titiunik
Citation
Sebastian Calonico, Matias Cattaneo and Rocio Titiunik's "Robust Nonparametric Confidence Intervals for Regression-Discontinuity Designs" advances political methodology with a new approach to estimating causal effects in regression discontinuity (RD) designs. Noting that confidence intervals for the local-linear polynomial estimated average treatment effect at the threshold tend to be invalid (and overly liberal) because of biases arising from common bandwidth selection procedures, Calonic, Cattaneo and Titiunik propose a bias-corrected estimator that not only improves performance asymptotically but also in small-samples (such as those common in small windows around the discontinuities which drive the RD design). The authors also propose an automatic bandwidth selection procedure that produces valid confidence intervals in wide variety of RD design settings. The creativity of the statistical theory combined with the practicality of the fast to compute and closed-form results has led this paper to change the analysis of RD designs even before it was published in Econometrica after presentation at the Political Methodology Summer Meeting of 2014.