Francis Fukuyama reviews Bruce Caldwell’s *Hayek’s Challenge*, an intellectual biography of Friedrich Hayek.

As Caldwell notes, Hayek initially thought the dividing line between possible and impossible positivism lay in the distinction between natural sciences and social sciences, but by the 1950s he had come to understand that the issue was really one of complexity. A positivist, predictive science is possible only for phenomena, whether human or natural, that are relatively simple—particle physics, for example. One can never fully model and predict complex phenomena such as the spontaneous orders produced by the interactions of simpler agents. These orders include the human brain, whose higher functions cannot possibly be inferred from its physical substratum, as well as ecosystems and, of course, markets, cultures, and other human institutions.

…Thus, the highly mathematical and ahistorical turn that academic economics has taken in recent years would have been, for Hayek, as much an abuse of reason as the socialist planning of earlier generations.

In discussing the role of math and econometrics, I think that the fundamental issue is how to evaluate an economic argument. How do we decide that on e person’s paper is valid and worth publishing, while another person’s is not?

If we believe that mathematics can establish the logical validity of an argument, and if we believe that econometrics can establish the empirical validity of an argument, then we have an approach to evaluating papers in economics. Otherwise, we do not appear to have clear criteria.

I believe that economic papers ought to be logical and scientific in spirit, with arguments expressed as precisely as possible. They ought to make falsifiable predictions. However, I believe that math and econometrics are neither necessary nor sufficient for establishing the validity of an economic argument. Thus, to borrow a Hayekian phrase, there is a “fatal conceit” to the sort of technical emphasis that was dominant when I was in graduate school.

*For Discussion*. If the mathematical approach to economics were eliminated, what would take its place?

## READER COMMENTS

## jvn

## May 10 2004 at 11:31am

Surely, the same thing that is used by journals now: Judgment.

Having served as a referee for many journals and as part of some editorial groups, I can assure you that all math does is ratchet up the minimum standards for acceptable publication. It does not eliminate the need for choice (not all sound models are published).

The problem is that good papers without the necessary modelling are not published. And useless models are tacked onto good papers just to get published. It is the latter cynicism that is wrong.

Further, as physicists understand, but many economists do not, rigor is not the same as mathematical formalism. Most formalism in econ is derived from math dept values not science. It has little substantive justification at any level.

## Walker

## May 10 2004 at 3:26pm

Steve Keen puts it nicely in

Debunking Economics…“Many critics of economics have laid the blame for its manifest failures at the feet of mathematics. Mathematics, they claim, has led to an excessive formalism in economics, which has obscured the inherently social nature of the subject.

While it is undeniable that an inordinate love of mathematical formalism has contributed to some of the intellectual excesses in economics, generally this reaction is as erroneous as blaming the piano for the discordant notes of a bad piano player. If anything should be shot, it is the pianist, not the piano. Though mathematics has definite limitations, properly used, it is a logical tool that should illuminate, rather than obscure. Economists have obscured reality using mathematics because they have practised mathematics badly, and because they have not realised the limits of mathematics…”

What form would economics take if the current “formal” approach is cast aside? Tough to say, but hopefully it would take a form that stresses emperical validity and applicability.

## Chris

## May 10 2004 at 11:57pm

The problem is that in order to validate something empirically we at least need some sort of minimal mathematical model. Even in order to get a regression to be useful we need to make a myriad of mathematical assumptions about disturbance terms. Waving hands is an unacceptable substitute.

Fixed point theorems do seem pretty uninteresting, I’ll grant you that. But verbal philosophy with five-syllable German words isn’t any more illuminating. My personal preference is for English spiced with math where it actually helps make a point.

## jvn

## May 11 2004 at 3:55am

Chris misses the point that the number of “fixed point” style papers has exploded in the journals, while good mathematics at the level of the JPE in 1964 are unpublishable in many top journals. English spiced with math is more and more considered the barest minimum at a weak journal.

The point is not to go to German philosophy. The point is that the increase in mathematical sophistication from 1964 to 2004 has not been accompanied by a requisite improvement in a) logic b) applicability c) empirical sophistication or d) useful predictions.

The fact is that neither of Coase’s seminal articles would be accepted to the American Econ Review today. And as Tyler points out, Friedman and Schwarz might not even make it as a Chicago PhD thesis today.

I say that except for the J of Math Econ or Econometrica, no major journal should have math more formal than that typical of the AER and JPE of 1964.

## Lawrance George Lux

## May 11 2004 at 3:23pm

Resort to parallelism, or the fancy verbal drawing of pictures; wait, is that not what We do now, simply inputting numbers every time We cannot think of a correct word. lgl

## Jonathan

## May 11 2004 at 4:19pm

The purpose of mathematical formalism is to cause the writer to not only be logically tight, but also specific enough to avoid arguments over semantics. While this may seem unnecessarily abstract, it is crucial to the advancement of science (incl economics). Steering economics to some sort of non-quantitative basis will only lead to unanswerable questions and unfalsifiable conclusions. I think physics would be the best model for how to apply math to a subject. The fact that a physicist works with real objects allows him to cut away a lot of the more cumbersome parts of mathematical formalism, but in such a way that the predications still have the sharpness of mathematics.

## jvn

## May 12 2004 at 2:31am

I absolutely agree that physics is a good model for this. But given the lack of clear tests, math economists have taken to using pure math as the standard, hence the excessive formalism.

I can assure you, most physicists I know are astonished when they see the degree of formalism in many economics journals.

There should also be a place for verbal speculation that can only be partially mathematized. Given our lack of power laws and true dynamic equations of motion it is imperative that room be given for good ideas that might be properly formalized after people have had time to digest and discuss the ideas.

## Sean Hackbarth

## May 17 2004 at 3:06pm

The argument Hayek made to differentiate physics with economics is the latter deals with complex phenomena. In an economy, there’s too many feedback loops, chaotic attractors, and other elements to make accuarate predictions possible. The best we can hope for is to explain the economic process taking place and describe the direction it’s headed.