Technology Category
Machine Learning: Causal LearningSearch and Optimization: Metareasoning and MetaheuristicsIntelligent Robots: State Estimation
A Meta-learner for Heterogeneous Effects in Difference-in-Differences
🤖 AI Summary
This paper addresses the problem of accurately estimating heterogeneous treatment effects—particularly the conditional average treatment effect on the treated (CATT)—in panel data under a conditional parallel trends assumption. We propose the first doubly robust difference-in-differences (DiD) meta-learner satisfying Neyman orthogonality. Our method formulates CATT estimation as a convex risk minimization problem incorporating auxiliary models, enabling flexible estimation across arbitrary subsets of covariates. It further extends to functional estimation under covariate shift and to instrumental-variable DiD settings with noncompliance. By leveraging generic machine learning methods—such as random forests and neural networks—as auxiliary models, our approach achieves superior empirical performance over existing baselines. It exhibits strong robustness—remaining insensitive to auxiliary model misspecification—and high accuracy—delivering stable estimates across diverse heterogeneity and noncompliance configurations.
Technology Category
Application Category
📝 Abstract
We address the problem of estimating heterogeneous treatment effects in panel data, adopting the popular Difference-in-Differences (DiD) framework under the conditional parallel trends assumption. We propose a novel doubly robust meta-learner for the Conditional Average Treatment Effect on the Treated (CATT), reducing the estimation to a convex risk minimization problem involving a set of auxiliary models. Our framework allows for the flexible estimation of the CATT, when conditioning on any subset of variables of interest using generic machine learning. Leveraging Neyman orthogonality, our proposed approach is robust to estimation errors in the auxiliary models. As a generalization to our main result, we develop a meta-learning approach for the estimation of general conditional functionals under covariate shift. We also provide an extension to the instrumented DiD setting with non-compliance. Empirical results demonstrate the superiority of our approach over existing baselines.
Problem
Research questions and friction points this paper is trying to address.
Estimating heterogeneous treatment effects panel data
Proposing doubly robust meta-learner CATT
Flexible estimation using generic machine learning
Innovation
Methods, ideas, or system contributions that make the work stand out.
Doubly robust meta-learner
Convex risk minimization
Neyman orthogonality robustness
🔎 Similar Papers
No similar papers found.
