统计研究

• 论文 • 上一篇    

样本选择性偏误、TS2SLS估计与我国代际收入流动性水平

郭建军等   

  • 出版日期:2017-08-15 发布日期:2017-08-25

Sample Selection Bias, Two-sample Two-stage Least Square Method and Inter-generation Income Elasticity in China

Guo Jianjun et al.   

  • Online:2017-08-15 Published:2017-08-25

摘要: 我国公开使用的微观调查数据库追踪时间较短,从单独一个数据库中整理出的父子配对样本存在因共同居住而产生的样本选择性偏误。本文尝试从两个数据库获得双样本来估计代际收入流动性水平,比较分析双样本两阶段最小二乘方法(TS2SLS)的优良性质,并以CHNS数据库作为主样本、CHIP数据库为辅样本进行了实证研究。研究结果表明,TS2SLS估计方法得到父子、母子代际收入弹性系数值分别为0.4937和0.5134,高于文献中通常采用单样本OLS或者IV方法的估计值,有效改善了代际弹性系数估计偏误。

关键词: 代际收入弹性, 样本选择性偏误, 内生性, TS2SLS

Abstract: It is only with a short history in terms of the Chinese household survey data base. There exists a sample selection bias while compiling the father or mother and son pair sample from a separate data base due to whether they or their descendants are still living together or not. This paper studies how to estimate China's inter-generation income elasticity coefficients with two samples from different databases, analyzes the properties of two-sample two-stage least squares estimators (TS2SLS), and conducts an empirical study by using CHNS database as a primary sample and CHIP database as an auxiliary one. It shows that the father-son and mother-son inter-generation income elasticity coefficients are 0.4937 and 0.5134 respectively with TS2SLS estimation, higher than those from OLS or IV method, so that the estimation bias of those coefficients has efficiently been improved.

Key words: Inter-generation Income Elasticity Coefficient, Sample Selection Bias, Endogeneity, Two-sample Two-stage Least Square Method (TS2SLS)