Abstract

In convex economic environments, it is a common practice to characterize efficient distributions by means of a social planner program: a benevolent planner maximizes the weighted sum of agents' utilities subject to feasibility constraints; by varying welfare weights, these social optima exhaust the entire Pareto frontier. This method fatally fails in several nonconvex economic environments that naturally arise due to market imperfections, price rigidities, asymmetric information, limited commitment and, in general, externalities. We argue that, subject to minor technical qualifications, the first-order conditions of the conventional planning program still locally characterize robustly efficient distributions even under nonconvexity. Robustness requires the presence of an unambiguous Pareto improvement conditional on a slightly inaccurate knowledge of the primitives of the economy or, alternatively, subject to small trembles in the implementation of a policy. We thus provide a rigorous foundation to the widespread use of a Ramsey-type planning program both in macroeconomics and microeconomics. We illustrate the significance of our characterization through an extended application to macroprudential policies.