Some Remarks on Policy Making under Model Uncertainty

Policy makers must often cope with circumstances where there is no consensual scientific model of the matter under consideration. We formally represent this situation as one of handling a robust optimization problem. This provides a framework to rigorously compare some commonly used approaches to deal with model undertainty (namely: model averaging, robust control, ambiguity aversion, and voting), as these approaches turn out to be specific dual versions of such an optimization problem. The framework also suggests an economically intuitive way to seek and characterize optimal policies, and reveals generic properties of policy decisions