Propensity Score Matching and Causal Inference: A methodological review

社会科学对于因果关系的关注使得倾向值匹配(propensity score matching)的方法受到越来越多的重视。本文从以下几个方面对这一方法进行述评。首先,本文系统梳理了倾向值匹配方法的历史、发展及其对于社会学调查研究的意义。其次,通过一个具体实例,本文展示了倾向值匹配的具体实施过程。第三,从密尔对因果关系的论述以及统计学的反事实框架(counterfactual framework)出发,本文阐释了倾向值匹配如何通过“控制”混淆变量从而满足非混淆假设(unconfoundedness)来帮助研究者提出因果性结论。第四,本文将倾向值匹配与回归中断设计和工具变量进行了比较。最后,在简要分析倾向值匹配的局限性之后,本文讨论了倾向值匹配的最新发展趋势。

Propensity score matching method has attracted more and more attention as sociological studies begin to pay closer attention to causal relationship． This article reviews the method from the following aspects． First，the author systematically reviews the history and development of propensity score matching as well as its implications for sociological survey research． Second，this research demonstrates how to perform propensity score matching using an example． Third，based on Mill's discussion on causality and statistical counterfactual framework，this article shows how propensity score matching contributes to causal conclusion by controlling for confounding variables and satisfying the unconfoundedness assumption． Fourth，this paper compares propensity score matching with regression discontinuity design and instrumental variable method． Finally，after summarizing the major limitations of propensity score matching，this article discusses the newly development of propensity score matching in recent years．

欢迎阅读《社会学研究》！您是该文第3704位读者！