You can't expect a revolution to come from the people in charge. This is worth keeping in mind when considering agent-based modeling -- an approach to understanding the economy that most authoritative academics have so far chosen to ignore.
Economists use models to help them consider how various parts of the real world, typically too numerous or complex to keep in our heads, might fit together. Traditionally, their models have made various simplifying assumptions -- for example, that consumers in their vast diversity act as one person making optimal decisions. Although such models can provide meaningful insights, they often conjure a world so removed from reality as to miss the most important phenomena altogether. Standard models, for example, spectacularly failed to envision the mechanisms behind the most recent financial and economic crises.
Enter agent-based modeling. With the computational power now available, researchers can build much better representations of reality, replacing a handful of equations with thousands or millions of virtual people and companies following unique and changeable rules of behavior.
Most economists seem to hate the very idea, and poke fun at it for its supposed lack of mathematical rigor. But here's something really funny: While the purists have been sniggering, the computational approach has quietly been growing in sophistication and gathering momentum. It's going to start changing the way economics gets done, and quite soon.
Consider a model economy recently developed by economists Domenico Delli Gatti, Mauro Gallegati and a diverse group of colleagues. They call it the Mark I, in anticipation of future upgrades. In the model, various firms produce goods of diverse sorts and sell them to consumers. People in turn sell their labor back to firms. Prevailing interest rates influence everyone's decisions.
Set it running, and this economy does just what real ones do. Research shows, for example, that it is prone to abrupt transitions. It can fall from a state of good performance into a recession, with output suddenly plummeting and unemployment rising sharply. It can also flip the other way. Inflation tends to rise when unemployment is low, and deflation is more likely when unemployment is high.
Old-school economists might rightly say we already know this -- not only from history and recent experience, but also because economists for decades have been studying models in which similar things happen, even if they don't always agree why. True. But the promise of these models is their flexibility and potential to probe much deeper.
For instance, preliminary studies suggest that the simple "two state" picture -- high unemployment and low inflation, and vice versa -- is only a very crude approximation. The low-unemployment state, in a slightly more detailed model, turns out to split into several possible states, some looking more or less stable, and others involving persistent oscillations in economic activity.
The complexity makes sense. After all, freeze some ordinary water -- what could be simpler? -- and the resulting solid ice can exist in at least 15 distinct phases that we currently know of, all with slightly different properties. Scientists studying these phases turn to computer simulations of interacting water molecules to aid their intuition and give some clues about what's possible. Agent-based economic models offer the same capability: to discover important subtleties that the unaided human mind would never imagine.
There's one other thing economists might like. In the Mark I, it turns out that a central bank, in following quite standard monetary policy to manage economic stability, can end up creating a kind of instability. Its actions balance the economy in a precarious middle state: good, but not too good. This naturally enhances the chance that a minor accidental shock can trigger a major recession. Recessions, on finer examination, may be inherently impossible to predict -- a finding that could be a relief to economists, who often get criticized for their poor track record in foreseeing such events.
Of course, the model is only a beginning. It needs a lot more funding and smart people to work on it. It's high time the U.S. Federal Reserve and other economic research institutions established ambitious programs to develop agent-based models with their own policy interests in mind. Economists who still rely on their old-school models are, of course, just doing what they were trained to do. This is understandable. Failing to take advantage of new things when they come along isn't.
The agent-based approach is the economics equivalent of what you now see everywhere across science: the use of computation as a tool to amplify our powers of analysis and discovery. It's moved from an experimental phase into a serious study of monetary policy rules, the consequences of quantitative easing or different regulatory regimes aiming for banking stability. It is going to be a big part of the future of economics. Young researchers, in particular, should be paying attention.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To contact the author on this story:
Mark Buchanan at firstname.lastname@example.org
To contact the editor on this story:
Mark Whitehouse at email@example.com