In an Institute for New Economic Thinking video interview, SFI Professor Doyne Farmer discusses work to create an agent-based model of the U.S. economy that will help scientists, economists, and policy makers better understand past, and future, financial crises.
Watch Farmer's video interview (13 minutes)
Farmer, with SFI External Professors Rob Axtell and John Geanakoplos, are working to create the model with support from an INET grant. Whereas a traditional economic model makes future predictions based on past market behavior and thus fails in unprecedented situations, their agent-based model takes into account the actions of individual decision makers, assigning behavioral rules to each agent in the model. This computer-based method creates models that are more dynamic and lifelike, Farmer says.
“The financial crisis really made me think about doing something that mattered," Farmer said. “(It) made me think that the real nut to crack is to be able to predict things about the macroeconomy.”
|All SFI News||About SFI||Follow SFI||Support SFI||SFI Home|
|Bob Line - Oct. 31, 2011, 8:13 a.m.|
I quite often search for ABM and housing, as it is an approach to understanding housing market systems that I've thought needed to be developed for many years, so was pleased to find this ambitious project. I first looked at IT Game Technology as a possible way forward for understanding the interactions between households and dwellings, and also attempted something simple in MapInfo GIS using its Mapbasic programming language , as housing is inherently spatial. But various ABM systems have developed further since like Repast that Nigel Gilbert at Surrey used.
There are also companies in the UK which gather masses of housing related data - especially Hometrack for its Housing Intelligence System . http://www.hometrack.co.uk/our-services/interactive-intelligence-systems/housing-intelligence-system-his which uses a massive inter-linked database. If this project makes good progress they might well be interested. http://www.blinehousing.info/