A0668
Title: Model averaging in semiparametric estimation of quantile treatment effects
Authors: Antonio Galvao - Michigan State University (United States) [presenting]
Sergio Firpo - Sao Paulo School of Economics FGV (Brazil)
Ulrich Hounyo - University at Albany, SUNY (United States)
Li Lu - School of Business - SUNY Geneseo (United States)
Abstract: Model-averaging methods are proposed to estimate quantile treatment effects (QTE and QTT) under treatment selection based on observables. To address propensity score misspecification, two estimators are developed: One averaging QTE/QTT across models and another that averages propensity scores before estimation. Unconfoundedness is required for at least one covariate set or their union. A data-driven covariate selection criterion is introduced, and asymptotic properties are derived for inference. A novel "unconfoundedness signature plot'' is introduced and helps to assess covariate relevance. Simulations show strong finite-sample performance, and the approach is illustrated by estimating the effect of inherited control on firm performance.
