线性工具变量模型中有效工具变量的识别研究

Research on Valid Instrumental Variable Identification for Linear Instrumental Variable Models

结合带惩罚的极小绝对偏差估计技术,对含有内生协变量的线性工具变量模型,给出一个基于惩罚的有效工具变量识别方法。该方法在数据含有部分异常值的情况下,能够正确地识别出有效的工具变量,进而可以得到模型参数的相合估计。基于所提出的统计推断方法,对中国31个省份对外贸易开放与经济增长的关系进行了实证分析。研究结果表明在处理贸易开放度的内生性问题上,各地区的国外市场接近度是一个行之有效的工具变量,并且发现对外贸易开放对经济增长有着显著的推动作用。

Based on the penalized least absolute deviation estimation technique,a valid instrumental variable identification method is proposed for linear instrumental variable models with endogenous covariates.The proposed identification method is also workable even in the case of data containing some outliers.Then we can obtain a consistent estimation procedure for model parameters.In addition,based on the proposed statistical inference method,we study the relationship between the opening of foreign trade and economic growth of China.The results show that the proximity of foreign markets is an effective tool variable in dealing with the endogenous problem of trade openness,and it is found that the opening of foreign trade has a significant role in promoting economic growth of China.