Comparing Subsidies to Solve Coordination Failure

We use experiments to systematically test the performance of subsidies aimed at inducing efficient coordination in a coordination game. We consider two classes of policies: those based on divide-and-conquer (i.e. iterated dominance) and those making the efficient Nash equilibrium of the game risk dominant. Cost-efficient policies from both classes are equally expensive but differ in the distribution of subsidies among agents. Our results show that risk dominance subsidies increase coordination more effectively or at a lower cost than divide-and-conquer subsidies.