时间断点回归:原理框架、应用演进与估计流程

Regression Discontinuity in Time:Theoretical Framework,Application Evolution and Estimation Flow

文章首先运用归纳法,提出时间断点回归的概念内涵和基本模型,并按"准随机实验"标准分析其存在的原理缺陷。然后通过梳理国外相关前沿成果,总结出时间断点回归有别于传统计量操作的方面:在适用性判断上,强调内生分组问题的理论分析和协变量连续性的提前估计;在稳定性检验上,包含更换变量指标等正向思路、对照组检验等反向思路、时间序列特征处理和其他非断点回归方法。最后,绘制时间断点回归的估计流程图,主要涉及适用性检验、处理效应的初步判定、模型构建与估计和稳健性检验四个部分。

This paper first proposes the concept and basic model of regression discontinuity in time(RDiT)by use of inductive method,analyzes its specific principle drawbacks according to the standard of“quasi randomized experiment”,and then by combing the relevant frontier achievements abroad,summarizes main aspects that distinguish RDiT from traditional econometrics:In terms of applicability judgment,RDiT emphasizes the theoretical analysis of endogenous sorting and the advance estimation of covariate continuity;for robustness test,there are direct methods such as changing indicators,indirect ways like placebo test,time series data processing and other non-RDiT tools.Finally,the paper comes to the estimation flow chart of RDiT,which mainly involves four main parts:applicability test,pre-judgment on treatment effect,model construction and estimation,and robustness test.