Decentralization and the search for policy solutions

We use mathematical and computational models to analyze the ability of federated organizations to solve difficult problems. Federated organizations may have some inherent advantages for solving difficult problems since they can empower multiple subunits to search for policy solutions in parallel and exploit and combine the mutual information discovered by the subunits to reveal even better solutions. We find that the benefits of these methods of experimentation first increase and then decrease as the problem of finding good policies increases in difficulty.