摘要
自从可信度革命以来,因果推断在计量经济学中的地位日益重要,因果推断方法也层出不穷,甚至令实证研究者眼花缭乱.本文系统梳理了因果推断在计量经济学中的历史渊源与发展脉络,涵盖随机实验、自然实验、断点回归、匹配估计、双稳健估计、工具变量法、双重差分法、合成控制法、回归控制法、分位数控制法等主流因果推断方法.与已有的文献综述相比,本文更注重因果推断方法的历史起源,以帮助实证研究者更好地把握其精神实质,同时将所涉文献更新至2024年的最新前沿(尤其是工具变量法与双重差分法的最新进展),并展望未来的发展方向.
Abstract
Since the Credibility Revolution,causal inference has played an increasingly important role in econometrics.New methods of causal inference emerge constantly,which may bewilder empirical researchers.This paper systematically reviews the historical origins and development of causal inference in econometrics,including mainstream methods for causal inference such as randomized experiment,natural experiment,regression discontinuity,matching estimation,doubly robust estimation,instrumental variable regression,difference in differences,synthetic control method,regression control method,and quantile control method.Compared with existing literature reviews,this paper focuses more on the historical origins of causal inference in order to help empirical researchers understand the essence,while updating the literature to the frontier of 2024(especially the latest advances in instrument variable regression and difference in differences)and looking forward to the directions of future development.
万方数据