Agent-Based Keynesian Economics; Methodological Issues and a Model

Charlotte Bruun - Aalborg University

This contribution conforms to the ideas of Agent-Based Computational Economics (ACE) - an approach to which The Society of Computational Economics has been an important nursery. The present approach has, however, been named "agent-based Keynesian economics" in order to underline the fact that agent-based computational economics is an approach with a methodology in common, but without a specific approach to economics in common. The famous (or infamous) disagreement between economists is not necessarily going to disappear because of a new methodology. To open up a discussion of different approaches to economics within ACE, it is attempted to relate its methodology to the methodology of Keynes. Besides presenting an agent-based Keynesian model, the presented paper will therefore discuss related methodological issues.

In the seventies Keynesian theory was rejected by the mainstream because of its lack of microfoundation; it appeared to be impossible to reach Keynesian conclusions from the study of individual rational agents1. Since then agent-based techniques have provided us with a new way of relating micro and macro, and therefore it appears to apply the technique to Keynes's thought. With the new techniques it becomes possible to let the suboptimality of the Keynesian world arise as a consequence of the way agents interact; a cause of suboptimality that could not be found by more traditional microeconomic techniques. It seems important to reconsider the work of Keynes himself with this possibility in mind rather than attempting to revive the "keynesian theory" that was rejected, since the methodological starting point of Keynes appear to be fundamentally different from that of the so-called "keynesians".

Besides the fact that agent-based computational economics may satisfy the proclaimed need for a microfoundation of Keynesian theory, the two approaches seem to be related through methodological similarities. We shall try to elucidate these similarities with induction as the central concept, and add computability as a concept that enforces the arguments of both approaches. After a common methodological starting point is laid down, the viability of the approach is tested through an agent-based simulation model. Following from the work of Keynes, accounting relations are central to the model, and price-changes are not important to the selforganizing features of the model. In order to capture the ideas of Keynes it has been necessary to include a financial market as well as markets for consumption goods, investment goods and labour.

1 At least not without resigning to different forms of transaction costs or other market imperfections which were not central to the work of Keynes.

Scheduled for Session 1.6 Model Of Bounded Rationality