关键词: 样本选择模型, 半参数估计, 分布差异测度, 非参数sieve估计。

Abstract: This paper proposes a two-step semiparametric estimator for sample selection models, which estimates the choice probabilities based on the Log-Euclidean distribution divergence measure in the first step, and estimate a partial linear model which contains parametric and nonparametric parts using nonparametric sieve method in the second step. Comparing with the existing semiparametric estimators, the implementation is more convenient and the computational burden is smaller of the estimator proposed in this paper. We demonstrate the consistency and asymptotic normality of our estimator, and derive the formula of the asymptotic variance. Monte Carlo results are consistent with the theoretical conclusion.

Key words: Sample Selection Model, Semiparametric Estimation, Distribution Divergence Measure, Nonparametric Sieve Estimation.