A New Perspective on Rational Expectations

A central problem of rational expectations equilibrium (REE) theory is that a REE may fail to exist. This failure is known since Kreps (1977). The influential papers of Radner (1979) and Allen (1981) were able to establish that a REE exists only generically. We show that this undesirable feature is related to the assumption that individuals are expected utility maximizers (Bayesians). If individuals have ambiguity aversion in the form of the maximin expected utility (MEU) model introduced by Gilboa and Schmeidler (1989), then a REE exists universally and not only generically. To prove this, we provide a suitable generalization of REE, which we call maximin rational expectations equilibrium (MREE). MREE allocations need not to be measurable with respect to the private information of each individual and with respect to the information that the equilibrium prices generate, as it is in the case of the Bayesian REE. We also prove that a MREE is efficient and incentive compatible. These results are false for the Bayesian REE.