Sensitivity Analysis in Population Shares

Filippo Palomba

Papers from arXiv.org

Abstract: Empirical work often relies on assumptions on the distribution of random variables, such as potential outcomes or latent variables, to identify some parameter of interest. We introduce a unified framework for assessing the sensitivity of a target estimand to relaxing these distributional assumptions for a given share of the population. We derive bounds on the values the estimand can take as this share varies between 0 and 1, and compute the bounds using tractable finite-dimensional programs. We propose plug-in estimators for the bounds and characterize their large sample properties. We also show that our bounds recover exact nonparametric sharp bounds at no additional computational cost, and introduce summary measures that separate misspecification uncertainty from sampling uncertainty and rank studies by their robustness. We apply our framework to assess the robustness of (i) average treatment effects in randomized experiments with attrition, (ii) counterfactual predictions in structural models, and (iii) formula instruments. We provide R and Python packages implementing all our methods.

Date: 2026-11
References: Add references at CitEc
Citations:

Downloads: (external link)
https://arxiv.org/pdf/2611.05584 Latest version (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:arx:papers:2611.05584

Access Statistics for this paper

More papers in Papers from arXiv.org
Bibliographic data for series maintained by arXiv administrators ().