政策效果评估的双重差分方法  

DID Model of Policy Effect Evaluation

作　　者：石华军[1,2] 楚尔鸣[2] 

机构地区：[1]湖南工业大学经济与贸易学院,湖南株洲412007 [2]湘潭大学商学院,湖南湘潭411105

出　　处：《统计与决策》2017年第17期80-83,共4页Statistics & Decision

基　　金：国家自然科学基金资助项目(71273221);湖南省社会科学基金资助项目(15YBA135);湖南省教育厅科学研究项目(15C0443)

摘　　要：双重差分方法(DID)及其拓展是目前国内进行政策效果评估应用最广泛的研究工具,但规范运用DID方法必须解决内生性、选择性偏误、分组样本异质性、动态异质性、自然实验的外生性等一系列问题,否则将导致回归结果的偏误。文章根据实验理论并结合国内外典型范文,通过实验前测、平行趋势检验、三重差分、IV、falsification test、连续型DID等多种方法和检验进行全面的稳健性分析,弥补DID方法运用中的缺陷。Difference-in-Difference （DID） model and its expansions are currently the most widely used research tool for the policy effect evaluation in China. However, as for the standard application of DID model, a series of problems such as endogeneity, selection bias, group sample heterogeneity, dynamic heterogeneity, extrogeneity of the natural experiment and so on must be solved first, otherwise, it will lead to bias of the regression result. In view of this, on the basis of the experiment theory and combining with the model papers at home and abroad, this paper conducts a comprehensive and robust analysis by use of various methods and examinations such as pre-experimental measurement, parallel trend test, triple difference, IV, falsification test, continuous DID and so on, so as to make up for the deficiency in the application of the DID model.

关 键 词：双重差分 政策内生性 选择性偏误 稳健性分析 

分 类 号：F202[经济管理—国民经济]