A0616
Title: Optimal policy learning in empirical practice
Authors: Hannah Busshoff - University of St. Gallen (Switzerland) [presenting]
Michael Lechner - University St Gallen (Switzerland)
Abstract: The purpose is to review and extend recent methods in policy learning for heterogeneous treatment effects. Interpretable decision rules (decision trees) are compared with black-box approaches (policy forests and policy neural networks), highlighting trade-offs in model explainability, robustness to misspecification, stability, and computational demands. Using administrative data on active labor market policies, it is demonstrated how these methods can guide individualized treatment assignment. Results provide practical insights into the design of data-driven, transparent, and performant policy rules.
