Résumé:
Public policies are often designed under limited budgets and must reach a minimum number of beneficiaries, while individuals may differ widely in expected benefits and treatment costs. This seminar discusses how data-driven Optimal Policy Learning (OPL), merging causal inference and machine learning, can be used to design beneficiary allocation rules that balance effectiveness, costs, and coverage requirements. Moving beyond simple cost-effectiveness rankings, the talk introduces an intuitive framework that delivers transparent decision rules and helps policymakers understand key trade-offs in real-world program design. The talk will also present an easy-to-use Stata package for applying the proposed approach to real-world policy contexts.