A simple approach to identifying the incentives for policy experimentation

We present a simple concavity condition that describes the incentives for a policy maker to pursue policy experiments. Our approach avoids the computational cost of numerical methods that fully characterise optimal experimentation under learning.