Machine Learning Instrument Variables for Causal Inference

This is the authors' abstract. We don't add key points for this paper.

A machine learning framework for constructing strong instrumental variables from weak exogenous data achieves semiparametric efficiency with applications to advertising and demand estimation.