Money as medium of exchange-an analysis with genetic algorithms

摘要

This paper uses a model of Kiyotaki and Wright to analyse the medium of exchange function of money. An economy is modeled where money comes up because individuals maximize their expected utility. Individuals choose their trading strategies with respect to maximizing utility which depends on the storage costs of goods and the probabilities of finding a suitable trading partner. There exist two types of Markov-Nash-equilibria in the model, where different goods come up as commodity money, depending on the parameters of the model. Then genetic algorithms are used to study how artificially intelligent agents learn to coordinate their strategies. The first question which is analysed is: do individuals learn the optimal strategies which- characterize these equilibria and if so, the second question is: how much time do individuals need to reach the equilibrium strategies. One obtains that individuals only learn the strategies of the Markov-Nash-equilibrium if the equilibrium is the one with the lowest storage cost good as intermediate good, i.e. money. Otherwise the optimal strategies are not learned without further assumptions and individuals prefer also the lowest cost good as medium of exchange. The Nash-equilibrium is only reached, if the difference between additional utility of using the good with the highest costs as intermediate good and its additional costs is sufficiently high.