Washington, Nov 30 : Researchers at the U.S. Department of Energy's Argonne National Laboratory are trying to create new computer models that will provide policymakers with more realistic pictures of different types of markets so they can better avert future economic catastrophe.
Traditional economic models rely heavily on "equilibrium theory," which holds that markets are influenced by countervailing balanced forces.
Because these models assume away the decision-making processes of individual consumers or investors, they do not represent the market's true internal dynamics, according to Charles Macal, an Argonne systems scientist.
"The traditional models don't represent individuals in the economy, or else they're all represented the same way - as completely rational agents," Macal said.
"Because they ignore many other aspects of behavior that influence how people make decisions in real life, these models can't always accurately predict the dynamics of the market," he added.
Macal and his Argonne colleagues have created a new set of simulations called "agent-based models" to better anticipate how markets behave.
These new models rely on information gleaned in part from surveys that ask respondents about the factors that influence the way they make decisions.
By gaining a more precise understanding of the behavior patterns of individual actors in a market - for example, how willing they are to accept risk, how strongly they value the future or how much time and effort they are able to spend making decisions, researchers and economists can better predict and avoid meltdowns.
Agent-based models separately calculate likely decisions for each individual actor in a model, then take the results of these decisions and see what impact they have on other agents.
By doing so, they have the potential to foresee a panic, a protracted "hot streak," herd mentality or a number of other market phenomena that pure rational-actor models would tend to miss.
The ability to produce such detailed simulations relies on the availability of high-performance computers that can handle the computational challenges of mathematically representing an enormous number of individual actors.
"Just five years ago, we couldn't model more than a couple dozen agents," Macal said. "Now, we can do a couple million," he added.