结合样本分割与系统方程的稳健性统计检验方法

结合二元样本分割与系统方程方法,提出一种新的稳健性统计检验.首先建立理论模型,分析结构方程模型中条件外生性假设与参数解释的关系,确定备择模型的构建原则;接着提出混淆变量调整与二元样本分割两种备择模型构建方法;最后基于系统方程方法建立假设检验,并针对奇异性问题改进检验统计量.蒙特卡罗实验表明,本文的稳健性统计检验具有良好的小样本表现.以Acemoglu教授关于人口老龄化与经济增长的研究为例,应用本文方法进行检验,发现该研究的部分模型存在误设问题.

Robustness test approach combining binary sample splitting and system of equations methods

This article proposes a novel robustness test method which combines binary sample splitting and system of equations techniques. First, we build a structural model to analyze relationship between the conditional exogeneity assumption and the explanations of structural parameters, thus establish principles of construction of alternative models. Then, we propose two construction methods:Binary sample splitting and confounding variables adjusting. Finally, we propose a robustness test statistic which based on system of equations method. Furthermore, we adjust computation process of the test to avoid singularity issue. Our Monte Carlo experiments verify the finite sample properties of this robustness test. As an application, we analyze the model settings of an article by Professor Acemoglu, which studied the influence of aging on economic growth.