See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/51437577

The Economy Needs Agent-Based Modeling

Article in Nature · August 2009

CITATIONS READS

1,074 5,288

2 authors, including:

Duncan K. Foley
The New School
179 PUBLICATIONS 6,770 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Marx's Theory of Value and Applications View project

Videos of Advanced Microeconomics II: Information and Behavior in Political Economy (Spring 2016) View project

All content following this page was uploaded by Duncan K. Foley on 18 April 2016.

The user has requested enhancement of the downloaded file.

OPINION
The economy needs agent-based modelling
The leaders of the world are flying the economy by the seat of their pants, say J. Doyne Farmer and
Duncan Foley. Agent-based modelling is the way forward.

I
n today’s high-tech age, one naturally pull society out of a recession: that, as

P. Noble/REUTERS
assumes that US President Barack rising prices had historically stimulated
Obama’s economic team and its inter- supply, producers would respond to
national counterparts are using sophis- the rising prices seen under inflation
ticated quantitative computer models by increasing production and hiring
to guide us out of the current economic more workers. But when US policy-
crisis. They are not. makers increased the money supply in
The best models they have are of two an attempt to stimulate employment, it
types, both with fatal flaws. Type one is didn’t work — they ended up with both
econometric: empirical statistical mod- high inflation and high unemploy-
els that are fit to past data. These suc- ment, a miserable state called ‘stagfla-
cessfully forecast a few quarters ahead tion’. Robert Lucas and others argued
as long as things stay more or less the in 1976 that Keynesian models failed
same, but fail in the face of great change. because they neglected the power of
Type two goes by the name of ‘dynamic human learning and adaptation. Firms
stochastic general equilibrium’. These and workers learned that inflation is
models assume a perfect world, and by just inflation, and is not the same as a
their very nature rule out crises of the real rise in prices relative to wages.
type we are experiencing now.
As a result, economic policy-makers Realistic behaviour
are basing their decisions on common The cure for macroeconomic theory,
sense, and on anecdotal analogies to however, may have been worse than the
previous crises such as Japan’s lost disease. During the last quarter of the
decade or the Great Depression (see twentieth century, rational expectations
Nature 457, 957; 2009). The leaders of Agent-based models could help predict and avoid market crashes. emerged as the dominant paradigm in
the world are flying the economy by the economics. This approach assumes
seat of their pants. current optimism about the future, and behav- that humans have perfect access to informa-
This is hard for most non-economists to ioural rules deduced from psychology experi- tion and adapt instantly and rationally to new
believe. Aren’t people on Wall Street using ments. The computer keeps track of the many situations, maximizing their long-run personal
fancy mathematical models? Yes, but for a agent interactions, to see what happens over advantage. Of course real people often act on
completely different purpose: modelling the time. Agent-based simulations can handle a far the basis of overconfidence, fear and peer pres-
potential profit and risk of individual trades. wider range of nonlinear behaviour than con- sure — topics that behavioural economics is
There is no attempt to assemble the pieces and ventional equilibrium models. Policy-makers now addressing.
understand the behaviour of the whole eco- can thus simulate an artificial economy under But there is a still larger problem. Even if
nomic system. different policy scenarios and quantitatively rational expectations are a reasonable model of
There is a better way: agent-based models. explore their consequences. human behaviour, the mathematical machinery
An agent-based model is a computerized simu- Why is this type of modelling not well- is cumbersome and requires drastic simplifica-
lation of a number of decision-makers (agents) developed in economics? Because of his- tions to get tractable results. The equilibrium
and institutions, which interact torical choices made to address the models that were developed, such as those used
through prescribed rules. The agents complexity of the economy and the by the Federal Reserve, by necessity stripped
can be as diverse as needed — from importance of human reasoning and away most of the structure of a real economy.
consumers to policy-makers and Wall adaptability. There are no banks or derivatives, much less
Street professionals — and the institu- The notion that financial econo- sub-prime mortgages or credit default swaps
tional structure can include everything mies are complex systems can be — these introduce too much nonlinearity and
from banks to the government. Such traced at least as far back as Adam complexity for equilibrium methods to handle.
models do not rely on the assumption Smith in the late 1700s. More recently When it comes to setting policy, the predictions
that the economy will move towards John Maynard Keynes and his fol- of these models aren’t even wrong, they are sim-
a predetermined equilibrium state, as other lowers attempted to describe and quantify ply non-existent (see Nature 455, 1181; 2008).
models do. Instead, at any given time, each this complexity based on historical patterns. Agent-based models potentially present
agent acts according to its current situation, the Keynesian economics enjoyed a heyday in the a way to model the financial economy as a
state of the world around it, and the rules gov- decades after the Second World War, but was complex system, as Keynes attempted to do,
erning its behaviour. An individual consumer, forced out of the mainstream after failing a cru- while taking human adaptation and learn-
for example, might decide whether to save or cial test during the mid-seventies. The Keyne- ing into account, as Lucas advocated. Such
spend based on the rate of inflation, his or her sian predictions suggested that inflation could models allow for the creation of a kind of vir-
685
OPINION NATURE|Vol 460|6 August 2009

tual universe, in which many players can act predetermined maximum value. If the price of all the details of a realistic problem can rapidly
in complex — and realistic — ways. In some the stock drops while a fund is fully leveraged, lead to a complicated simulation, where it is dif-
other areas of science, such as epidemiology the fund’s wealth plummets and its leverage ficult to determine what causes what. To make
or traffic control, agent-based models already increases; thus the fund has to sell stock to pay agent-based modelling useful we must proceed
help policy-making. off part of its loan and keep within its leverage systematically, avoiding arbitrary assumptions,
limit, selling into a falling market. carefully grounding and testing each piece of
Promising efforts This agent-based model shows how the the model against reality and introducing
There are some successful agent-based mod- behaviour of the hedge funds amplifies price additional complexity only when it is needed.
els of small portions of the economy. The fluctuations, and in extreme cases causes Done right, the agent-based method can pro-
models of the financial market built by vide an unprecedented understanding

S. Platt/Getty Images
Blake LeBaron of Brandeis University of the emergent properties of interacting
in Waltham, Massachusetts, for exam- parts in complex circumstances where
ple, provide a plausible explanation intuition fails.
for bubbles and crashes, reproducing
liquidity crises and crashes that never Holistic approach
appear in equilibrium models. Rob A thorough attempt to understand the
Axtell of George Mason University whole economy through agent-based
in Fairfax, Virginia, has devised firm modelling will require integrating mod-
dynamics models that simulate how els of financial interactions with those
companies grow and decline as work- of industrial production, real estate,
ers move between them. These replicate government spending, taxes, business
the power–law distribution of company investment, foreign trade and invest-
size that one sees in real life: a very ment, and with consumer behaviour.
few large firms, and a vast number of The resulting simulation could be used
very small ones with only one or two to evaluate the effectiveness of different
employees. Other promising efforts include crashes. The price statistics from this model approaches to economic stimulus, such as tax
the credit sector model of Mauro Gallegati’s look very much like reality. It shows that the reductions versus public spending.
group at the Marche Polytechnic University standard ways banks attempt to reduce their Such economic models should be able to
in Ancona, Italy, and the monetary model own risk can create more risk for the whole provide an alternative tool to give insight into
developed by Robert Clower of the Univer- system. how government policies could affect the
sity of South Carolina in Columbia and Peter Previous models of leverage based on equilib- broad characteristics of economic perform-
Howitt of Brown University in Providence, rium theory showed qualitatively how leverage ance, to quantitatively explore how the econ-
Rhode Island. These models are very useful, can lead to crashes, but they gave no quantita- omy is likely to react under different scenarios.
but their creators would be the first to say that tive information about how this affects the sta- In principle it might even be possible to create
they provide only a tentative first step. tistical properties of prices. The agent approach an agent-based economic model capable of
To see in more detail how an agent-based simulates complex and nonlinear behaviour making useful forecasts of the real economy,
model works, consider the model that one of us that is so far intractable in equilibrium mod- although this is ambitious.
(Farmer) has developed with Stefan Thurner of els. It could be made more realistic by adding Creating a carefully crafted agent-based
the University of Vienna and John Geanakoplos more detailed information about the behaviour model of the whole economy is, like climate
of Yale University to explore of real banks and funds, modelling, a huge undertaking. It requires
how leverage affects fluctua- “Market simulation could and this could shed light close feedback between simulation, testing, data
tions in stock prices (pub- on many important ques- collection and the development of theory. This
lished in a Santa Fe Institute
be used to evaluate the tions. For example, does demands serious computing power and multi-
working paper). Leverage, effectiveness of economic spreading risk across many disciplinary collaboration among economists,
the investment of borrowed stimulus approaches.” financial institutions stabi- computer scientists, psychologists, biologists
funds, is measured as <ok?] lize the financial system, or and physical scientists with experience in large-
the ratio of total assets owned to the wealth of does it increase financial fragility? Better data scale modelling. A few million dollars — much
the borrower; if a house is bought with a 20% on lending between banks and hedge funds less than 0.001% of the US financial stimulus
down-payment the leverage is five. There are would make it possible to model this accurately. package against the recession — would allow a
four types of agents in this model. ‘Noise trad- What if the banks themselves borrow money serious start on such an effort.
ers’, who trade more or less at random, but are and use leverage too, a process that played a key Given the enormity of the stakes, such an
slightly biased toward driving prices towards role in the current crisis? The model could be approach is well worth trying. ■
a fundamental value; hedge funds, which hold used to see how these banks might behave in an J. Doyne Farmer is at the Santa Fe Institute, 1399
a stock when it is under-priced and otherwise alternative regulatory environment. Hyde Park Road, Santa Fe, New Mexico 87501,
hold cash; investors who decide whether to Agent-based models are not a panacea. The USA, and at LUISS Guido Carli in Rome and
invest in a hedge fund; and a bank that can major challenge lies in specifying how the founded the quantitative trading firm Prediction
lend money to the hedge funds, allowing them agents behave, and in particular, in choosing Company. Duncan Foley is Leo Model professor of
to buy more stock. Normally, the presence of the rules they use to make decisions. In many economics at the New School for Social Research,
the hedge funds damps volatility, pushing the cases this is still done by common sense and 6 East 16th Street, New York 10003, USA, and an
stock price towards its fundamental value. But, guesswork, which is only sometimes sufficient external professor at the Santa Fe Institute.
to contain their risk, the banks cap leverage at a to mimic real behaviour. An attempt to model e-mails: [email protected]; [email protected]
686
View publication stats