B0273
Title: Deep learning for individual heterogeneity: An automatic inference framework
Authors: Max Farrell - University of Chicago (United States) [presenting]
Abstract: A methodology is developed for estimation and inference using machine learning to enrich economic models. The framework takes a standard economic model and recasts the parameters as fully flexible nonparametric functions, to capture the rich heterogeneity based on potentially high dimensional or complex observable characteristics. These ``parameter functions'' retain all the interpretability, economic meaning, and discipline of classical parameters. We show that deep learning is well-suited to the structured modeling of heterogeneity in economics. First, we show how the network architecture can be easily designed to match the global structure of the economic model, delivering a novel methodology that moves deep learning away from prediction. Second, we prove convergence rates for the estimated parameter functions. These parameter functions are then the key input into the finite-dimensional parameter of inferential interest. We obtain valid inference based on a novel orthogonal score or influence function calculation that covers any second-stage parameter and any machine-learning-enriched model that uses a smooth per-observation loss function. No additional derivations are required, and the score can be taken directly to data, using automatic differentiation if needed to obtain the components. Our framework covers, as special cases, well-known examples such as average treatment effects and partially linear models, but we also seamlessly deliver new results.
