Royal Economic Society Annual Conference 2021 Sargan Lecture

EDITORIAL

Each year, The Econometrics Journal sponsors a plenary lecture at the Annual Conference of the Royal Economic Society. This Sargan Lecture commemorates the fundamental contributions to and profound influence on econometrics by (John) Denis Sargan. It does so by promoting econometric theory and methods with substantive direct or potential value in applications and their actual empirical application. It is chaired by an editor of The Econometrics Journal and published in the journal.¹

Guido Imbens delivered the Sargan Lecture at the Society’s online 2021 Conference. Guido Imbens is The Applied Econometrics Professor and Professor of Economics at the Stanford Graduate School of Business. His early work on nonparametric instrumental variables estimation of heterogeneous treatment effects, and in particular of local average treatment effects, had a major impact on both econometrics and empirical economics, and is routinely taught and applied. He has since added many key contributions to causal inference and the econometrics of treatment effects. Recently, he studied topics like inference on treatment effects on subjects interacting over networks and the use of machine learning to analyse treatment effect heterogeneity. In 2021, Guido Imbens was awarded the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, together with Joshua Angrist (MIT), ‘for their methodological contributions to the analysis of causal relationships’ (they shared the prize with David Card of UC Berkeley, who was honoured ‘for his empirical contributions to labour economics’).

In his Sargan Lecture, Guido Imbens reviewed the literature on causal panel data analysis. His article “Causal models for longitudinal and panel data” with Dmitry Arkhangelsky (CEMFI) in this issue provides a detailed and updated account of the lecture. It takes lessons from the older panel data literature and relates them to the many recent developments in causal panel data analysis. These include new perspectives on two-way fixed effects models and differences-in-differences estimation, novel uses of factor models, and advances in synthetic control methods.² The result is an excellent, broad, and coherent review of this very active and practically useful literature, with some explicit advice for empirical economists.

The review concludes by listing open issues in causal panel data analysis. One such issue is the analysis of dynamic treatments and dynamic treatment effects. The reviewed literature mostly ignores dynamics. This allows it to develop powerful methods for assessing causal effects without employing tightly specified economic models. However, in many applied settings, treatment is dynamic and agents shape their outcomes through active, forward-looking decision making. In these settings, the potential outcomes at the center of the reviewed approaches are reduced forms that are typically not invariant to changes in the mechanism for assigning treatment.³ This suggests the reviewed literature is closing in on the limits of what can be done without explicating the agents’ decision making process. Advances in combining the reduced-form and structural approaches to policy evaluation may help to stretch these limits.⁴

Guido Imbens delivered his Sargan Lecture under the title “Causal panel data models”. At the start of the session, Michael Jansson, a co-editor of The Econometrics Journal, presented the 2018 Denis Sargan Econometrics Prize to Matt Goldman (Facebook) and David M. Kaplan (University of Missouri) and the 2019 one to Artūras Juodis (University of Amsterdam).⁵ I chaired the session and Tobias Klein, the deputy managing editor, moderated questions from the audience. A video of the session and further information are available from the website of The Econometrics Journal at http://ectj.org.

The Econometrics Journal ensures that all articles it publishes, including the Sargan Lecture, are peer reviewed. I would like to warmly thank the referees who provided us with fast and high-quality feedback.

Footnotes