Yihong Liu

Biography

Yihong Liu was born in Shanghai, China. He is a sixth-year Ph.D. candidate in Economics at UC Santa Barbara. He earned his B.A. in Economics from Shanghai University of Finance and Economics and his M.A. in Economics from Duke University.

Research

His research lies primarily in Econometrics, Applied Econometrics, and Environmental and Natural Resource Economics, with particular interests in difference-in-differences, triple-differences, matching methods, and the estimation of dynamic resource models. His research aims to develop estimation and inference methods that are both theoretically valid and practically useful for applied researchers. As the first chapter of his dissertation, his job market paper studies the pitfalls of commonly used covariate adjustment methods in triple-differences designs and proposes a matched triple-differences framework for valid estimation and inference. The second chapter, which is coauthored, studies the common empirical practice of combining matching with two-way fixed-effects estimation in difference-in-differences designs. The paper identifies potential pitfalls of this approach and proposes a method for valid estimation and inference. The third chapter, also coauthored, he develops a GMM method for estimating structural parameters in a class of dynamic resource games. The proposed estimator is straightforward to implement while preserving the full nonlinearity of the equations of motion, an important feature for policy analysis in environmental and natural resource economics.

Job Market Paper Title

"Matched Triple-Differences: A Framework for Covariate Adjustment"

Abstract

This paper studies how covariate adjustment should be conducted in triple-differences designs. The leading empirical approach is linear regression adjustment, which augments the triple-differences regression with covariates. Under the conditional parallel gap assumption, we first derive a covariate balancing condition that a weighting estimator should satisfy to consistently estimate the average treatment effect on the treated (ATT). Expressing estimators from commonly used linear regression adjustments as weighting estimators reveals that they generally fail to satisfy this balancing condition and are therefore inconsistent for the ATT. To address this limitation, we propose a matched triple-differences framework that achieves the required covariate balance through matching. Within this framework, we construct a consistent matched triple-differences estimator, establish its asymptotic normality, and provide a consistent variance estimator. The framework accommodates a general class of matching procedures, including nearest-neighbor matching and kernel matching.

JEL Classification Codes: C14, C21, C23