A Sensitivity Approach to Causal Inference Under Limited Overlap
This work addresses bias in causal inference for researchers dealing with observational data where treatment groups have limited overlap, offering a method to contextualize findings rather than providing a direct solution.
The paper tackles the challenge of limited overlap between treated and control groups in observational analysis by proposing a sensitivity framework to assess how irregular the outcome function must be to invalidate findings, demonstrating it protects against spurious results by quantifying uncertainty in low-overlap regions.
Limited overlap between treated and control groups is a key challenge in observational analysis. Standard approaches like trimming importance weights can reduce variance but introduce a fundamental bias. We propose a sensitivity framework for contextualizing findings under limited overlap, where we assess how irregular the outcome function has to be in order for the main finding to be invalidated. Our approach is based on worst-case confidence bounds on the bias introduced by standard trimming practices, under explicit assumptions necessary to extrapolate counterfactual estimates from regions of overlap to those without. Empirically, we demonstrate how our sensitivity framework protects against spurious findings by quantifying uncertainty in regions with limited overlap.