联立方程模型的工具变量方法质疑与商榷

Question and Consultation about Instrument Variable Method in Simultaneous Equations Models

本文质疑联立方程模型前定变量的工具变量性质:前定变量并不保证与当期行为解释变量的相关性,由其构建的工作回归元所完成的分阶段最小二乘估计因而并非两阶段最小二乘估计。建议按照简约式方程构建工作回归元,其具有模型数理逻辑支持下的可替代意义。工作回归元的不同导致结构式方程分阶段最小二乘估计的不同结果,之于恰好识别方程则揭示了业内关于间接最小二乘估计方法的一个误区。

This paper questions the predetermined variables for their properties of in- strument variable in simultaneous equations models. Because predetermined variables do not ensure to have statistic correlation with the behavior explanatory variable, the stages least squares based on the work regressor generated from them then would not be TSLS. Different work regressors may certainly lead to different estimation of a structure equation. And to a just-identified equation, it also discovers a example model is given in the paper. misunderstanding about indirect least squares. An