A method based on stochastic dynamic programming is developed to derive efficiency frontiers for the trade-off between the long-run average social benefits and price variation. The method is used to quantify the importance of price variation per se as a criterion in U.S. wheat storage policy. The results suggest that a single criterion of maximum expected social benefits, calculated by the traditional surplus measures, is satisfactory because price variation is incidentally reduced enough that further reductions can be attained only a considerable opportunity cost.