Introduction

The term “rational expectations” was introduced in the famous article of Muth (1961), but the equilibrium existence problem posed by such a concept had been recognized and addressed formally seven years earlier. Grunberg and Modigliani (1954) recognized that predictions of economic events, unlike weather forecasts, can affect predicted events. Expectations of a substantial price increase, for example, can trigger investment and production decisions that increase supply and cause the actual price increase to be smaller than expected. Weather forecasting may be difficult in practice, but even the logical possibility of economic forecasting is problematic. Although they did not model the process of price determination explicitly, Grunberg and Modigliani showed that Brouwer's fixed-point theorem implies the existence of a correct forecast, provided that the requisite continuity and boundary conditions are satisfied. Muth (1961) acknowledged the Grunberg and Modigliani article, but focused more on using the assumption of rational expectations to specify expectations in econometric models in which the equilibrium could be obtained by construction.

The dramatic advances in the mathematical theory of general equilibrium during the fifties and early sixties stimulated efforts to include expectations and uncertainty in general equilibrium models during the late sixties and early seventies. Two articles by Radner (1968 and 1972) stand out from the many notable contributions of this period for their explicit focus on the information structure underlying the formation of expectations. Radner's models imposed the strong informational consistency requirement that an agent's actions could not differ across states of the world that could not be distinguished by the agent's information. This condition restricts trade among agents to events that are mutually observable.