Deep Neural Network Estimation in Panel Data Models

In this paper we study neural networks and their approximating power in panel data models. We provide asymptotic guarantees on deep feed-forward network estimation of the conditional mean, building work Farrell et al. (2021), explore latent patterns cross-section. use proposed estimators to forecast progression new COVID-19 cases across G7 countries during pandemic. find significant forecasting gains over both linear nonlinear time-series Containment or lockdown policies, as instigated at national level by governments, are found have out-of-sample predictive for cases. illustrate how partial derivatives can help open "black box" facilitate semi-structural analysis: school workplace closures been effective policies restricting pandemic countries. But our methods heterogeneity time variation effectiveness specific containment policies.