交叠的秘密 (The Secret Behind Staggered: A Primer to Staggered DID in the Field of Empirical Economics)

Abstract

Chinese Abstract: ENTER CHINESE ABSTRACT HERE.


English Abstract: Difference in Differences (DID) design is the most important causal identification method in the main contribution of the Nobel Prize in Economics in 2021 - natural experiment. And more and more researchers use the setting of staggered DID, but the recent DID econometric literature shows that the use of two-way fixed effects (TWFE) estimator under time-varying treatment may produce bias, or even get the opposite Causal effect. Based on this, this article reviews the traditional DID design, briefly describes the decomposition of staggered DID estimators, bias diagnosic, the latest robust DID estimator, and then uses a simulation data and two published economic papers (Beck, Levine, and Levkov (2010, JF); Cao Qingfeng (2020, China Industrial Economy)) to illustrate some necessary/best elements of staggered DID in practice.

Note: Downloadable document is in Chinese.

Keywords: Difference in Differences; time-varying treatment; heterogeneous treatment effect; Bacon decomposition; robust estimator