Vers une macroéconomie fondée sur des agents autonomes et intelligents

This paper discusses a research agenda that has been advocated by Axel Leijonhufvud for studying the coordination process of a decentralized market economy. The research used agent-based modeling, an approach that assumes exogenous rules for adaptive behavior, thereby endowing agents with an autonomy that is impossible in a rational-expectations-equilibrium approach. The paper argues that this approach also endows people with more intelligence than would a standard optimizing approach. The analysis focuses on the role of business firms as coordinating agents who organize trading activities. Computer simulations show that a network of specialist trading enterprises can emerge from autarky when people follow simple myopic rules. The interactions between trading enterprises and households results in a multiplier process which can be shown to mimic time-series behavior of detrended real GDP in the United States. The analysis can also be used to illustrate the possibly destabilizing effects of excessive wage/price flexibility. JEL Classification: B4, D50, D58, D83.