Interpersonal comparisons of utility and the policy paralysis problem

Several ``Paretian'' welfare rules are equivalent when policymakers know agents' characteristics, e.g., a policy is optimal if (a) any other policy making someone better o ̈ harms some agent, or (b) it is the maximum of some social welfare function. This paper extends these and other rules to environments where policymakers have a probability distribution over a state space of possible models. Under weak conditions, rule (a), which postulates ex ante preferences for agents, recommends some change from almost every status quo policy. Unfortunately, (a) requires a demanding form of interpersonal welfare comparability. Rule (b) labels all policies optimal if the state space obeys a weak diversity condition. Since the probabilities of states are irrelevant for this result, only a small perturbation of a model with no uncertainty generates policy paralysis.