Semiparametric Estimation of a Censored Regression Model Subject to Nonparametric Sample Selection

This study proposes a semiparametric estimation method for a censored regression model subject to nonparametric sample selection without the exclusion restriction. Consistency and asymptotic normality of the proposed estimator are established under mild regularity conditions. A Monte Carlo simulation study indicates that the estimator performs well in various designs and outperforms parametric maximum likelihood estimators. An empirical application to female smoking is provided to illustrate the usefulness of the estimator.

本研究针对无排他性约束（exclusion restriction）的非参数样本选择删失回归模型（censored regression model），提出了一种半参数估计方法。在温和正则条件（mild regularity conditions）下，本文证明了所提估计量的一致性（consistency）与渐近正态性（asymptotic normality）。蒙特卡洛模拟（Monte Carlo simulation）实验结果表明，该估计量在多种设定下均表现出色，且性能优于参数极大似然估计量（parametric maximum likelihood estimators）。最后，本文以女性吸烟行为为研究对象开展实证应用，以此说明该估计量的实用价值。