Inference for Heterogeneous Treatment Effects with Efficient Instruments and Machine Learning

We introduce a new instrumental variable (IV) estimator for heterogeneous treatment effects in the presence of endogeneity. Our estimator is based on double/debiased machine learning (DML) and uses efficient machine learning instruments (MLIV) and kernel smoothing. We prove consistency and asymptotic normality of our estimator and also construct confidence sets that are more robust towards weak IV. Along the way, we also provide an accessible discussion of the corresponding estimator for the homogeneous treatment effect with efficient machine learning instruments. The methods are evaluated on synthetic and real datasets and an implementation is made available in the R package IVDML.

💡 Research Summary

The paper introduces a novel instrumental variable (IV) estimator designed for heterogeneous treatment effects (HTE) in the presence of endogeneity. Building on the double/debiased machine learning (DML) framework, the authors combine efficient machine‑learning instruments (MLIV) with kernel smoothing to estimate a non‑parametric function β(v) that captures how the causal effect varies with a continuous covariate V. The structural model is Y_i = β(V_i)·D_i + g(X_i) + ε_i, where D_i is endogenous, X_i are high‑dimensional controls, and Z_i is an instrument satisfying relevance and validity.

The estimator proceeds in two stages. First, machine‑learning methods (e.g., random forests, neural nets) are used to estimate the conditional expectations f(Z,X)=E