因果推断经验研究中的中介效应与调节效应

Mediating Effects and Moderating Effects in Causal Inference

中介效应分析与调节效应分析是国内经济学因果推断经验研究中被广泛采用的两种研究手段,但是存在不同程度的误用。前者的主要问题在于过度使用中介效应逐步法检验,后者的主要问题是对其在因果识别中的作用阐发不足。为此,本文深入讨论了中介效应检验的偏误、如何正确开展中介效应分析,以及如何使用调节效应分析来强化因果关系论证,并针对当前的使用现状提出了相应的操作建议。

Mediating-effects and moderating-effects analyses are common practices in empirical research of causal relationships, both with problems though.The main problem of mediating-effects analysis is the abuse of causal steps approach borrowed from social psychology.The main problem of moderating-effects analysis is a lack of focus on its role of justifying identification.The paper discusses in depth the bias stemmed from causal steps approach, how to do mediating-effects analysis properly, and how to use moderating-effects analysis to strengthen causal argument.Practical guidelines are provided so as to improve the quality of current research conduct.