Optimal policy learning under constraints in Stata

Optimal policy learning (OPL) has emerged as a powerful framework for designing data-driven treatment allocation rules that maximize social welfare while accounting for treatment-effect heterogeneity. In many real-world policy applications, however, decision-makers face operational constraints such as limited budgets, minimum coverage requirements, or combinations of both, making unconstrained policy learning impractical. This presentation introduces a comprehensive Stata implementation for constrained optimal policy learning that enables researchers and practitioners to estimate, evaluate, and compare treatment assignment policies under realistic resource limitations. The framework integrates modern causal machine learning methods for estimating conditional average treatment effects (CATEs) with efficient optimization algorithms that solve budget-constrained, coverage-constrained, and joint-constrained allocation problems. I discuss the underlying theoretical framework, the computational implementation, and the associated Stata commands, illustrating their use through empirical examples and simulation evidence. The proposed tools provide an accessible and reproducible environment for translating heterogeneous treatment-effect estimates into implementable policy recommendations, thereby bridging the gap between causal inference, machine learning, and evidence-based policy design within the Stata ecosystem.