Beyond Parallel Trends: An Identification-Strategy-Robust Approach to Causal Inference with Panel Data

2025-11-26

AI Summary

The parallel trends assumption in panel data causal inference is often unverifiable and fragile. Method: This paper proposes a novel “neighbor comparison group” approach that dynamically selects control units with similar outcome trajectories to the treated group during the pre-treatment period—replacing reliance on global parallel trends with local trajectory matching. It establishes a time-homogeneous analytical framework integrating multi-period treatment timing and pre-treatment matching. Contribution/Results: Under non-nested identification assumptions—including local parallel trends and conditional independence—the method consistently estimates the average treatment effect on the treated (ATT). Theoretical analysis demonstrates robustness of the identification strategy. Empirical applications show that, relative to conventional difference-in-differences, this approach substantially improves the robustness and credibility of policy effect estimates.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Intelligent Robots: State Estimation

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

In this paper, we propose a new approach to causal inference with panel data. Instead of using panel data to adjust for differences in the distribution of unobserved heterogeneity between the treated and comparison groups, we instead use panel data to search for "close comparison groups" -- groups that are similar to the treated group in terms of pre-treatment outcomes. Then, we compare the outcomes of the treated group to the outcomes of these close comparison groups in post-treatment periods. We show that this approach is often identification-strategy-robust in the sense that our approach recovers the ATT under many different non-nested panel data identification strategies, including difference-in-differences, change-in-changes, or lagged outcome unconfoundedness, among several others. We provide related, though non-nested, results under "time homogeneity", where outcomes do not systematically change over time for any comparison group. Our strategy asks more out of the research design -- namely that there exist close comparison groups or time homogeneity (neither of which is required for most existing panel data approaches to causal inference) -- but, when available, leads to more credible inferences.

Problem

• Proposes a new panel data method using close comparison groups for causal inference.
• Ensures identification-strategy-robustness across various non-nested panel data approaches.
• Requires close comparison groups or time homogeneity for more credible inferences.

Innovation

• Uses panel data to find close comparison groups
• Compares treated group outcomes with similar pre-treatment groups
• Recovers ATT under multiple non-nested identification strategies