Finance models need to get better at predicting how people, and thus markets, actually behave in the real world, according to an article in Engineering & Technology magazine that includes comments from SFI Professor Doyne Farmer.
Models of market trends rely on data from standard factors like Gross Domestic Product and unemployment, but some of these models' assumptions -- that every person has equal access to relevant information, for example, and thus makes rational decisions based on it -- are disputed at best.
Further, such models don’t take into account liquidity and housing markets, the two main factors in the 2008 market crash, Farmer points out. He argues that using agent-based modeling, which incorporates the behavior of individual elements of a system, can indicate potential crises better than data-driven modeling. He and an international team have made a proposal to the European Commission for a $7 million research project to scale up efforts of such computation-intensive modeling, according to the article.
Read the E&T magazine article (June 13, 2011)