On Multi-Cause Causal Inference with Unobserved Confounding: Counterexamples, Impossibility, and Alternatives

TL;DR

This paper challenges recent claims that multi-cause causal inference can overcome unobserved confounding, demonstrating counterexamples and impossibility results, and proposing proxy variables and sensitivity analysis as alternatives.

Contribution

It provides analytical counterexamples and proves nonparametric identification is impossible in multi-cause causal inference with unobserved confounding, offering practical alternatives.

Findings

Counterexamples invalidate existing methods

Nonparametric identification is impossible

Proposes proxy variables and sensitivity analysis as alternatives

Abstract

Unobserved confounding is a central barrier to drawing causal inferences from observational data. Several authors have recently proposed that this barrier can be overcome in the case where one attempts to infer the effects of several variables simultaneously. In this paper, we present two simple, analytical counterexamples that challenge the general claims that are central to these approaches. In addition, we show that nonparametric identification is impossible in this setting. We discuss practical implications, and suggest alternatives to the methods that have been proposed so far in this line of work: using proxy variables and shifting focus to sensitivity analysis.