Locally Robust Semiparametric Estimation

Many economic and causal parameters depend on nonparametric or high dimensional first steps. We give a general construction of locally robust/orthogonal moment functions for GMM, where steps have no effect, locally, average functions. Using these orthogonal moments reduces model selection regularization bias, as is important in many applications, especially machine learning Also, associated standard errors are robust to misspecification when there the same number interest. use cross‐fitting construct debiased estimators conditional quantiles dynamic discrete choice with state variables. show that additional needed globally, approach estimating those characterize double robustness variety new doubly simple regularity conditions asymptotic theory.