[Paper Review] What's Trending in Difference-in-Differences? A Synthesis of the Recent Econometrics Literature

Jonathan Roth, Pedro H. C. Sant’Anna+2 more|arXiv (Cornell University)|Jan 4, 2022

Advanced Causal Inference Techniques169 citations

TL;DR

This paper synthesizes recent DiD econometrics, clarifying canonical assumptions, extensions, and practical guidance for applied researchers.

ABSTRACT

This paper synthesizes recent advances in the econometrics of difference-in-differences (DiD) and provides concrete recommendations for practitioners. We begin by articulating a simple set of ``canonical'' assumptions under which the econometrics of DiD are well-understood. We then argue that recent advances in DiD methods can be broadly classified as relaxing some components of the canonical DiD setup, with a focus on $(i)$ multiple periods and variation in treatment timing, $(ii)$ potential violations of parallel trends, or $(iii)$ alternative frameworks for inference. Our discussion highlights the different ways that the DiD literature has advanced beyond the canonical model, and helps to clarify when each of the papers will be relevant for empirical work. We conclude by discussing some promising areas for future research.

Motivation & Objective

• Articulate the canonical DiD setup and its identifying assumptions (parallel trends and no anticipation) and the ATT.
• Survey recent advances that relax canonical assumptions along three axes: multiple periods/treatment timing, violations of parallel trends, and alternative inference frameworks.
• Provide practical recommendations and a practitioner checklist for implementing DiD analyses.
• Clarify how new DiD methods relate to standard TWFE models and when each is relevant for empirical work.

Proposed method

• Describe the canonical two-period DiD model and its identification via parallel trends and no anticipation.
• Classify recent DiD innovations by relaxing timing (multiple periods), parallel trends, or sampling/inference assumptions.
• Discuss generalized staggered-treatment models with potential outcomes paths and parallel trends extensions.
• Explain why TWFE estimands may differ from intuitive causal parameters under heterogeneity and outline alternative estimators.
• Provide practical guidance, including checklists and references to software packages for implementation.

Experimental results

Research questions

• RQ1 What are the canonical assumptions that enable identification of the ATT in a DiD framework?
• RQ2 How do recent DiD developments relax core assumptions (timing, parallel trends, inference) and what are the practical implications?
• RQ3 When do TWFE estimates fail to have a straightforward causal interpretation in multi-period, staggered settings, and what alternatives mitigate this issue?
• RQ4 What practical guidance and tools can researchers use to implement contemporary DiD methods?

Key findings

• TWFE in multi-period, heterogeneous-treatment settings can produce biased or counterintuitive estimates due to negative weighting and forbidden comparisons.
• New DiD methods isolate clean treated-versus-not-yet-treated comparisons and aggregate them with user-chosen weights to target specific parameters.
• Robust inference in DiD contexts includes permutation/bootstrap methods and design-based approaches that align with the level of treatment assignment.
• Parallel trends extensions and sensitivity analyses are developed to address concerns about violations of the parallel trends assumption.
• The literature emphasizes clarity in assumptions, comparison groups, causal estimands, estimation methods, and robustness checks.