Testable restrictions on the equilibrium manifold under random preferences

General equilibrium theory was criticized for its apparent irrefutability, as seen-tingly implied by the Sonnenschein-Mantel-Debreu theorem. This view was challenged by Brown and Matzkin [Econometrica, 64 (1996) 1249], who showed the existence of testable restrictions on the equilibrium manifold. Brown and Matzkin, however, maintain the assumption that individual preferences are invariant (against psychological evidence). I consider the Brown-Matzkin problem under random preferences: for each profile of endowments one observes a distribution of prices; does there exist a probability distribution of preferences that explains the observed distributions of prices via Walrasian equilibria? I argue that even under random utilities general equilibrium theory is falsifiable. (C) 2003 Elsevier B.V. All rights reserved.