政策效果评估的双重差分方法
发布时间：2018-10-31 13:37
【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。
[Abstract]:The dual difference method (DID) and its extension are the most widely used research tools for policy effectiveness assessment in China, but the standard application of DID method must solve endogenous, selective bias, heterogeneity of grouping samples and dynamic heterogeneity. A series of problems, such as externality of natural experiments, will lead to errors of regression results. Based on the experimental theory and typical domestic and foreign models, this paper makes a comprehensive robust analysis through many methods and tests, such as experimental pre-test, parallel trend test, triple difference, IV,falsification test, continuous DID and so on. To make up for the defects in the application of DID method.
【作者单位】： 湖南工业大学经济与贸易学院;湘潭大学商学院;
【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)
【分类号】：D0351
，
本文编号：2302388
[Abstract]:The dual difference method (DID) and its extension are the most widely used research tools for policy effectiveness assessment in China, but the standard application of DID method must solve endogenous, selective bias, heterogeneity of grouping samples and dynamic heterogeneity. A series of problems, such as externality of natural experiments, will lead to errors of regression results. Based on the experimental theory and typical domestic and foreign models, this paper makes a comprehensive robust analysis through many methods and tests, such as experimental pre-test, parallel trend test, triple difference, IV,falsification test, continuous DID and so on. To make up for the defects in the application of DID method.
【作者单位】： 湖南工业大学经济与贸易学院;湘潭大学商学院;
【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)
【分类号】：D0351
，
本文编号：2302388
本文链接：https://www.wllwen.com/guanlilunwen/zhengwuguanli/2302388.html
