Learning Competitive Equilibrium

Abstract

We consider a pure exchange economy repeated for an indefinite number of periods from a fixed endowment and posit a learning rule which directs convergence to competitive equilibrium. In each period trade converges to an allocation in the contract set, where agents interpret the current (common) normalized utility gradient as a vector of prices to determine the implied wealth redistribution relative to their endowments. Agents who are less wealthy at the new allocation are designated subsidizers, and demand to provide smaller subsidies in subsequent periods of economic activity. Our model is a globally stable alternative to Walras' tatonnement.

Note: A previous version of this abstract can be found at: http://ssrn.com/abstract=671941.

Keywords: Learning, general equilibrium, stability, Scarf's example

JEL Classification: C63, C68, D44, D51, D58, D61