What ants tell us

Butterfly Economics

一月 15, 1999

The title, Butterfly Economics , is misleading. It reminds you of one of the cliches of chaotic dynamics: tiny differences in initial conditions can have very large long-run consequences, so that an odd beat of a butterfly's wing in the tropics can be the decisive cause of a tornado setting down half a world away.

In fact, Paul Ormerod makes no real use of chaotic dynamics, which is just as well. So far as I know, no one has yet identified such a tornado; a whole lot of analytic machinery would have to intervene to find the connection between the butterfly in Africa and a trailer park in Kansas.

Ormerod also has another, much more apt connotation in mind. He would like to see a less mechanical-physical, more chancy, more biological-epidemiological approach to economics. Anyway, the book is about ants, not butterflies.

I have a good deal of sympathy with Ormerod's intention at this general, programmatic level. In most, though not all, mainstream economics, the individual agents - consumers, families, workers, firms - are assumed to interact only through the mediation of prices. What you and others do affects what I do only if your actions change the prices and market opportunities to which I react. Otherwise we each consult our own preferences, look after our own interests, and make our own decisions, atomistically.

Ormerod points out what anyone would have to acknowledge; that agents in society influence one another directly, through fashion, observation, imitation, advice and leadership (as well as resistance to all those things). He says, quite reasonably, that economic theory ought to take account of those non-market interactions, because it is likely that they will make a difference in the way the whole economic system behaves, and perhaps even introduce wholly new sorts of phenomena that traditional theory cannot explain, or may explain wrongly.

A few years ago, the economist Alan Kirman published a very nice paper on the behaviour of ants. Ants? Well, he had come across some experimental observations on the foraging behaviour of ants. Some regularities were said to be hard to explain in terms of the intellectual apparatus usually applied to such matters. Kirman tried his hand at making a model of ant behaviour in the experimental set-up. The conventional thing for an economist would have been to start by asking: what does an ant optimise?

This is not as foolish as it sounds. I knew an animal ecologist who was studying the grazing and resting patterns of moose inhabiting an island in Lake Superior. He found that their behaviour was almost exactly explained by the simple assumption that they were trying to minimise their net expenditure of energy.

But Kirman did something different; he modelled an ant as making simple decisions as a function of its own state and what it observed nearby ants doing, with a little randomness thrown in. He found that computer simulation of an ant colony produced behaviour patterns very like those observed in the experimental situation. Neat.

I have no idea what ant biologists have made of Kirman's model. But Ormerod thinks that it is what economics has been needing all these years. He "applies" it to the business cycle, the growth of firms, marriage and the family, and other things as well. (I will explain the quotation marks soon.) The book is chock-full of references to what "the ants" can teach us about economics, in contrast to mainstream ideas. "Once again, the basic principles of our ants model offer new insights into a complex economic problem." This is the last sentence of a chapter on business-cycle fluctuations.

Before I begin to express my misgivings about all this, I had better repeat that I am in favour of broadening economic analysis in the direction proposed by Kirman and adopted by Ormerod. The approved motivational and behavioural assumptions of mainstream economics are too narrow, and the narrowness is often centred in the wrong place. So I am not offering a knee-jerk reaction.

My first objection is that Ormerod's exposure to the literature of modern economics is not very thorough. There is in fact quite a lot of work within the mainstream that belies his characterisation of it. Here are just three indications. His index has no entry under "game theory". But game theory has been a hot subject in economics for years, and you cannot do game theory without paying attention to direct strategic interaction among the players.

My colleague Abhijit Banerjee wrote a famous paper on the theory of "herd behaviour", and this paper has resonated and produced a small literature.

Similarly, Steven Durlauf has a long series of papers in which economic behaviour and evolution are governed by nearest-neighbour interactions.

Finally, the National Bureau of Economic Research, not exactly a hotbed of methodological radicalism, has a standing programme on "behavioural macroeconomics", led by George Akerlof and Richard Thaler, names Ormerod should get to know. My children had an alphabet book called Ant and Bee ; the profession is already up to Bee and beyond.

I am not just being petulant on behalf of economists. A new and different idea in academic economics draws a lot of critical attention, even before publication. Other people scour it for holes, look for empirical tests, ask if the same facts could not be explained more straightforwardly. Knowing this, most academics are pretty cautious. Ormerod, however, applies his ants model rather casually. I give just two examples.

He considers briefly the standard economist's approach to the study of criminal behaviour (about which I have doubts too). It rests, as you would expect, on a careful weighing of incentives. Ormerod points out, correctly, that years of careful study of the evidence have yielded only inconclusive results as to whether positive or negative incentives predominate. He does not consider the possibility that this might reflect near-balance in the actual incentives available. Instead he proposes an ant-like theory that begins by dividing the population into three groups: criminals, susceptibles and non susceptibles, and then allows in a general way for contagion effects to change the proportions of the three groups.

He then discusses in general terms how such a model might behave, cites a few broad, qualitative empirical generalisations, and concludes that this "framework of analysis, stark though it may be in outline, has clear positive implications for policy". I am not against this sort of framework; my point is that it has not yet earned the right to be taken seriously.

Here is a second example. Ormerod mentions the finding by Robert Shiller of "excess volatility" in the prices of securities. Share prices fluctuate much more than seems plausible given the stability of the "fundamentals" that should govern the behaviour of efficient markets for securities. As you would expect, this finding has been dissected and discussed to within an inch of its life by the community of financial economists. Vast bodies of data, powerful statistical methods and sophisticated models have been deployed. All Ormerod has to say is that if one kind of ant is labelled "fundamentalists" and the other kind "chartists", then Kirman's ant model will generate a lot of volatility. It is no harder than that. Is it? You would think there are no alternatives to consider.

A quite different sort of casualness arises in what I can only describe as a very low standard of proof. Again, two quick examples will describe the tone. At one point, Ormerod draws a scatter diagram plotting the change in different countries' average growth rates between 1960-73 and 1974-95, against the same countries' change in the share of profits in national income. There is a loose, positive, approximately linear relationship. It is not clear what this has to do with the ants approach to economics, but Ormerod thinks there is some connection. He interprets causality as running from profit share to growth rate. He mentions the possibility of causality in the other direction, but waves it off. There is no mention of other intervening variables, or of the possibility that both profit share and growth rate reflect some unmentioned common cause.

Finally, in a chapter on business cycles, it is shown that quarter-to-quarter growth rates of United Kingdom national income fluctuate irregularly. A simple Kirman-type model adapted in untested ways to macroeconomics produces simulations in which period-to-period aggregate growth rates fluctuate irregularly. "This piece of evidence is itself by no means decisive in terms of validating our model. But it is one more piece of the jigsaw which fits." Well, it would be fairer and more accurate to say that any one of a thousand models that could easily be written down would have that feature, so the evidential value of this finding is just about nil. In the same chapter, it is taken as an argument against conventional macro-models that they depend on unexplained random disturbances; but Ormerod allows himself to depend on unexplained differences among firms playing the role of ants. Actually the two sorts of assumptions are probably near equivalents.

I suppose there are worse things than overplaying the strength of one's hand. But I fear that Ormerod's book conveys a false picture of what economics is and how hard it is to do it right.

To comply with the provisions of the Truth In Reviewing Act, I should say that Ormerod discusses some work of mine in his chapter on economic growth.

He is very polite, and has one or two nice things to say. But he finds that "once againI our model of ant behaviourI gives a better description of reality than does orthodox economics". I have to report that he misunderstands a lot of things in this chapter that I do not want to argue about. What is curious, however, is that his bibliographical note mentions two references on this subject. One is an ancient paper of mine; the other, described as a "balanced account", is a recent empirical paper (by G.

Mankiw, D. Romer and D. Weil) that concludes that the standard model explains the cross-country facts rather well, which is just what Ormerod's chapter denies.

Robert M. Solow, Nobel laureate, is emeritus professor of economics, Massachusetts Institute of Technology, United States.

Butterfly Economics: A New General Theory of Social and Economic Behaviour

Author - Paul Ormerod
ISBN - 0 571 19005 7
Publisher - Faber
Price - £16.99
Pages - 217



  • 注册是免费的,而且十分便捷
  • 注册成功后,您每月可免费阅读3篇文章
  • 订阅我们的邮件
Please 登录 or 注册 to read this article.