Automatic Debiased Machine Learning with Generic Machine Learning for Static and Dynamic Causal Parameters

Joint Economic Research Seminar Series 2023-2024

Dr. Vasilis Syrgkanis, Assistant Professor, Department of Management Science and Engineering, Stanford University, will be presenting

 “Automatic Debiased Machine Learning with Generic Machine Learning for Static and Dynamic Causal Parameters.” 

Abstract:
A variety of causal or structural parameters of interest depend on high dimensional or non parametric regressions. Machine learning can be used to estimate such parameters. However estimators based on machine learning algorithms can be severely biased by regularization and/or model selection. Debiased machine learning uses Neyman orthogonal estimating equations to reduce such biases. Debiased machine learning generally requires estimation of unknown Riesz representers. A primary innovation of this work is to provide Riesz regression estimators of Riesz representers that use the definition of the parameter of interest as a black-box, rather than invoking explicit mathematical characterizations and that can employ any machine learning algorithm, including neural nets and random forests. End-to-end algorithms emerge where the researcher chooses the parameter of interest and the debiased estimation approach follows automatically. We find in Monte Carlo examples that automatic debiasing sometimes performs better than debiasing via inverse propensity scores and never worse. Finite sample mean square error bounds for Riesz regression estimators and asymptotic theory are also given.