David Warsh offers a long meditation on the subject. Here’s an extract:

Alan Kirman, of the Univerity of Aix-Marseille, France, a much-decorated veteran; and Robert Axtell, of George Mason University, an up-and-comer; laid out the case for a point of view known as “agent-based” modeling. They describe this as a “bottom-up” approach to thinking about economic phenomena, made practicable by modern computers, in which people are heterogeneous and react directly with each other; in which their information is local and their behavior governed by rules of thumb; and in which the aggregate behavior of the system emerges from behavior of individuals rather than a “representative” agent.” Simulations of this sort by Thomas Schelling, which had formally identified the “tipping point” phenomenon (a fairly concrete discovery, after all) and elucidated mechanisms that give rise to residential segregation, had been laboriously worked out on a checkerboard in the 1970s, Axtell noted; with today’s models, 5 million more intricate calculations can be performed in an instant.

Actually, Schelling’s insight is more remarkable even than that. It was originally worked out, not on a checkerboard but with pencil and paper on a long flight. The checkerboard came later. Warsh continues…

Seem sensible? It does to anyone not properly trained in economics, said Kirman. For the last 125 years, however, ever since the views of Leon Walras and other theorists of general equilibrium became encoded in a famous textbook of Alfred Marshall (or, rather, partially encoded), technical economists have viewed the economy as a system in which individuals deal with one another only through the market mechanism, reacting to signals about prices and quantities as if they were determined by some central authority, best thought of as an auctioneer. Individual actors adapt as best they can to market signals which they are powerless to affect, until some sort of balance between supply and demand is achieved. Then things settle down and no individual has any reason to change his behavior, unless some external “shock” to the system occurs.

Warsh has much more. Also, here is a lovely piece (a few years old) by Jonathan Rauch on agent based modelling.

June 5, 2008 5:53am in Everyday economics, Research | Comment