Economic orthodoxy is going down

… even if most economists don’t yet know it

Over the past few years of blogging, I’ve spent a fair bit of time — too much time — perusing the blogs and research papers of mainstream economists, especially in macroeconomics. One thing that I have almost never seen discussed anywhere is the research potential for agent-based computational economics, i.e. for using computers to build large scale models of economies in which people, firms, banks, governments etc. interact in realistic ways. The rest of sciences now strongly depend on simulations of this sort to gain insight into complex systems of many kinds — from computer operating systems to the Internet to weather to natural ecologies to … well, you name it. How about the properties of water? Yep. This approach is now as common and as necessary as water itself.

So I’ve been shocked that in economics, the few times I have seen simulation mentioned at all by mainstream figures (somewhere in the comments on Steve Williamson’s blog, for example, as I recall), it’s been immediately dismissed as “uninteresting” or portrayed as the domain of a few fringe economists playing with silly toy models. Williamson is by no means alone. This seems to be THE prevailing view. The simulation approach doesn’t have the mathematical rigor, these people disdainfully say, to count as serious economics.

I’m increasingly convinced that all these people are going to be absolutely dumfounded one day, probably quite soon, when agent-based modelling completely overtakes the field. It will be seen by policy makers and regulators and research funding bodies and anyone else interested in understanding economies in a fundamental way as BY FAR the most powerful and flexible way to go. In a forthcoming column at Bloomberg, I look very briefly at one recent macroeconomic agent-based models developed by Domenico Delli Gatti, Mauro Gallegati and a number of colleagues, including some physicists and computer scientists. This model — and it is just one of a handful I might have looked at — illustrates just how sophisticated these things are becoming. Meanwhile, it seems, very few economists are paying attention.

The researchers 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 (very closely akin, in fact, to the phase transitions one finds in physical substances). 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.

This is not surprising, of course. The promise of these models is their flexibility and potential to probe much deeper. I’m not going to go into more detail; those interested can read the paper I just mentioned, which gives quite a bit more detail. But I did want to give some pointers to literature describing comparable models under development elsewhere. To any young economists out there who might be listening — this is going to be the future of your field. Don’t get caught napping!

Other resources:

First, this paper by economists Giorgio Fagiolo and Andrea Roventini gives an excellent introduction and comparison of the agent-based approach to the more common DSGE approach. Are there issues to be resolved with the agent approach? Of course, and these authors deal with them squarely; this is a new field and needs lots of further development. Many problems are almost wide open (such as how to systematically try to reduce the number of parameters in a model to find simpler, more robust models that still account for observed phenomena, thereby distinguishing the variables that matter most from those which matter much less).

But the most valuable part of the paper is their section 5 which surveys macroeconomic agent-based models now under development for use in real policy making. Many economists, I think, still believe that there are almost no such models. Fagiolo and Roventini suggest instead that they are in fact already too numerous to usefully cover in a single review! They cover a variety of models relevant to four macroeconomic policy areas, namely fiscal policy, monetary policy, bank regulation, and central bank independence.

Is the existence of these models even known to the mainstream? I wonder. They really OUGHT to be known to policy makers, because they already offer insight on a par with anything coming from DSGE.

Second, a should of course mention the great web site run by Leigh Tesfatsion, who has been instrumental in helping to further this field. This provides links to a great many agent models under development for application right across economics, and not only macroeconomics.

Another very worthwhile paper is Adaptive Microfoundations for Emergent Macroeconomics by Edoardo Gaffeo and colleagues. This describes the development of a macroeconomic model in considerable detail, and also shows how this model already stacks up against DSGE. On my reading, it’s already superior. And, of course, only going to get much better with time.

Well, this post is already longer than I wanted it to be. Look at the work of Peter Howitt for a number of other great examples of agent-based models being brought to bear with great effect. Or, how about this article in the New York Times discussing the potential for agent-based modelling in studying the stability of financial networks? The basis for the piece was this nice assessment of such models by Rick Bookstaber of the new Treasury Office of Financial Research.

This IS NOT a fringe part of economics any longer, or at least it shouldn’t be, although it seems that those defending DSGE would like it to be. I suspect that a lot of the economists currently trained to do DSGE don’t actually know or understand much about agent-based computational models; how they work, how they can be used, and how they might be developed further. I’m not sure they know much about this at all, except that it’s a threat to their preferred, familiar, comfortable way of doing things. They’re determined to keep doing what they’re currently doing, which is, of course, a very human thing to do.

But young economists, I hope, will think for themselves, read and explore, find out about agent-based models and what they can do, and make up their own minds. The DSGE approach really does look like The Titanic of economic methodology. Its taking on water, its bow already looming high in the air. Don’t let your career go down with it.

Follow The Physics of Finance on Twitter: @Mark_Buchanan

Other recent articles:

How to consistently beat the markets: solid evidence
A new study documents predictable patterns in market movements, patterns which have persisted for 200 years. No, the markets are not efficient

Bio-finance? Why high-frequency trading makes stocks flock like birds

Plus a bonus: enjoy an “awesome” fish video

Taxis 2.0: Streamlining city transport with graph theory

Big city taxi systems could be 40% more efficient with device-enabled taxi sharing

Arrogant physicists — do they think economics is easy?

No. But their ideas can help improve economics, and here’s why

Random oblivious transfer: is quantum secured finance almost a reality?

Why quantum cryptography could soon play a role in our financial lives

The fragile “wisdom” of crowds

Why prediction markets aren’t quite as amazing as economists once thought

A picture makes it clear: the recession IS over

Not that many people will know it for quite some time

What’s the use of “econo-physics”?

Why physics + economics = more than you think

Economics: we find the truth axiomatically

How economists implicitly assume what they claim to know

Economics: How often do we “lose” decades?

According to actual history, much more often than most economists think

Can Big Data predict stock price movements?

Probably not.

How Google’s algorithm could fix the financial system

Computation could make banks’ systemic risks visible to all

“Just so” stories of modern economics

Economists like to tell stories with their theories. They look like the wrong stories…