Rational Expectations vs. Adaptive Behavior in a Hyperinflationary World: Experimental Evidence

Abstract

We study an overlapping generations (OLG) economy where the government finances a fixed real deficit through seignorage, the structure of the economy and the level of deficit is common knowledge, and agents observe past prices. In this model there is a continuum of nonstationary equilibria and two stationary equilibria. When the rational expectations hypothesis is satisfied, a continuum of equilibria have paths converging to the stationary equilibrium with a higher inflation; conversely, when adaptive behavior is shown by agents, a continuum of inflation paths converge to the lower inflation --Pareto superior-- stationary equilibrium (see [11] and [8]). We test these contrasting hypotheses in an experimental environment. We find that inflationary paths lie close to the lower inflation stationary equilibrium and even when agents start at inflation rates that diverge from it, they do not follow rational expectations noninflationary paths, but paths that converge to a neighborhood of the lower inflation stationary equilibrium; furthermore, agents do not react to advance knowledge of future changes in parameters. The data, however, show a bias towards the Pareto optimal path of constant consumption and the corresponding stationary inflation rate. This bias does not seem to be caused by a deviation from perfect competition, i.e., as a noncooperative solution to a market game. Our subjects gain enough experience to be able to forecast prices with great accuracy within a stationary environment, but are not able to foresee the effects of announced changes in parameters, or to specify sophisticated supply schedules. This factor may explain the bias towards constant consumption.