Abstract:Instrumental variable methods have been widely used to identify causal effects in the presence of unmeasured confounding. A key identification condition known as the exclusion restriction states that the instrument cannot have a direct effect on the outcome which is not mediated by the exposure in view. In the health and social sciences, such an assumption is often not credible. To address this concern, we consider identification conditions of the population average treatment effect with an invalid instrumental variable which does not satisfy the exclusion restriction, and derive the efficient influence function targeting the identifying functional under a nonparametric observed data model. We propose a novel multiply robust locally efficient estimator of the average treatment effect that is consistent in the union of multiple parametric nuisance models, as well as a multiply debiased machine learning estimator for which the nuisance parameters are estimated using generic machine learning methods, that effectively exploit various forms of linear or nonlinear structured sparsity in the nuisance parameter space. When one cannot be confident that any of these machine learners is consistent at sufficiently fast rates to ensure $\surd{n}$-consistency for the average treatment effect, we introduce a new criteria for selective machine learning which leverages the multiple robustness property in order to ensure small bias. The proposed methods are illustrated through extensive simulations and a data analysis evaluating the causal effect of 401(k) participation on savings.

Don't (fully) exclude me, it's not necessary! Identification with semi-IVs

Identification with possibly invalid IVs

Bayesian variable selection in linear regression models with instrumental variables

When is IV identification agnostic about outcomes?

Partial identification in nonseparable count data instrumental variable models

Identification and Inference with Invalid Instruments

On the instrumental variable estimation with many weak and invalid instruments

Partial identification in nonseparable binary response models with endogenous regressors

Instrument-residual estimator for multi-valued instruments under full monotonicity

Sequential Underspecified Instrument Selection for Cause-Effect Estimation

Identification in Endogenous Sequential Treatment Regimes

The Instrumental Variable Model with Categorical Instrument, Treatment and Outcome: Characterization, Partial Identification, and Statistical Inference

Identification and Estimation in a Class of Potential Outcomes Models

Relaxing Instrument Exogeneity with Common Confounders

Using negative controls to identify causal effects with invalid instrumental variables

Nonparametric Identification and Estimation with Independent, Discrete Instruments

Instrumental variable analysis in the presence of unmeasured confounding

BudgetIV: Optimal Partial Identification of Causal Effects with Mostly Invalid Instruments

Selective Machine Learning of the Average Treatment Effect with an Invalid Instrumental Variable

On Doubly Robust Estimation with Nonignorable Missing Data Using Instrumental Variables

Decomposing Identification Gains and Evaluating Instrument Identification Power for Partially Identified Average Treatment Effects