Physics Envy, indeed

A recent post over at Gene Expression points out the problems in mainstream economics in a way familiar to most Austrians:

And yet, even then, physicists will be the first to admit that even the most powerful mathematical machinery they are able to bring to bear on a problem can deal successfully with only the very simplest situations, beyond which their equations are useless. Thus, for example, their equations can be solved for the two body problem but not the three body problem in Newtonian mechanics; they can solve the Schr?dinger equation when there is only one proton and one electron interacting, but not when there are even two protons and two electrons, let alone anything more complicated than that.

Furthermore, on those occassions when physicists do make complex predictions -- such as that nuclear fission would occur en mass, before the first atom bomb was tested (to choose an historical example) -- they do so with caution, double checking all their calculations, and hoping that they haven't overlooked something, or might accidentally set the atmosphere on fire. (See Richard Rhodes excellent The Making of the Atomic Bomb for more on this subject). *

Contrast this with the situation in economics. Here the elementary particle, so to speak, is the individual human being, no two of which are alike. What's more, the forces of attraction and repulsion that each individual feels for the goods of this world cannot be measured with any precision at all, much less to an accuracy of 20 decimal places. Furthermore, these forces of attraction and repulsion do not remain constant, even approximately so, over time and place for the same individual, let alone for different individuals, who vary enormously in their likes and dislikes.

Does any of this deter economists from using the machinery of calculus and systems of linear equations and so forth, to try to model the most complex dynamic systems imaginable, involving multiple individuals and multiplicity of goods? Noooooooooooo. They rush right in; and this, my friends, is precisely what physics envy is all about.

Mises complained of the same thing when he wrote "The Ultimate Foundation of Economic Science," noting the areas where economists flush with Physics Envy borrowed terms in the hope of convincing people that a similar certitude (and ability to predict, and thus engineer) lay within their own discipline.

On empiricism:

Auguste Comte postulated an empirical science of sociology which, modeled after the scheme of classical mechanics, should deal with the laws of society and social facts. The many hundreds and thousands of the adepts of Comte call themselves sociologists and the books they are publishing contributions to sociology. In fact, they deal with various hitherto more or less neglected chapters of history and by and large proceed according to the well-tried methods of historical and ethnological research. It is immaterial whether they mention in the title of their books the period and the geographical area with which they are dealing. Their "empirical" studies necessarily always refer to a definite epoch of history and describe phenomena that come into existence, change, and disappear in the flux of time. The methods of the natural sciences cannot be applied to human behavior because this behavior, apart from what qualifies it as human action and is studied by the a priori science of praxeology, lacks the peculiarity that characterizes events in the field of the natural sciences, viz., regularity.

On confirmation and refutability:

In the natural sciences a theory can be maintained only if it is in agreement with experimentally established facts. This agreement was, up to a short time ago, considered as confirmation. Karl Popper, in 1935, in Logik und Forschung pointed out that facts cannot confirm a theory; they can only refute it. Hence a more correct formulation has to declare: A theory cannot be maintained if it is refuted by the data of experience. In this way experience restricts the scientist's discretion in constructing theories. A hypothesis has to be dropped when experiments show that it is incompatible with the established facts of experience.

It is obvious that all this cannot refer in any way to the problems of the sciences of human action. There are in this orbit no such things as experimentally established facts. All experience in this field is, as must be repeated again and again, historical experience, that is, experience of complex phenomena. Such an experience can never produce something having the logical character of what the natural sciences call "facts of experience."

On aping the research methods of the natural sciences:

The popular ideas concerning the methods the economists employ or ought to employ in the pursuit of their studies are fashioned by the belief that the methods of the natural sciences are also adequate for the study of human action. This fable is supported by the usage that mistakes economic history for economics. A historian, whether he deals with what is called general history or with economic history, has to study and to analyze the available records. He must embark upon research. Although the research activities of a historian are epistemologically and methodologically different from those of a physicist or a biologist, there is no harm in employing for all of them the same appellation, viz., research. Research is not only time-consuming. It is also more or less expensive.

But economics is not history. Economics is a branch of praxeology, the aprioristic theory of human action. The economist does not base his theories upon historical research, but upon theoretical thinking like that of the logician or the mathematician. Although history is, like all other sciences, at the background of his studies, he does not learn directly from history. It is, on the contrary, economic history that needs to be interpreted with the aid of the theories developed by economics.

The reason is obvious, as has been pointed out already. The historian can never derive theorems about cause and effect from the analysis of the material available. Historical experience is not laboratory experience. It is experience of complex phenomena, of the outcome of the joint operation of various forces.

This shows why it is wrong to contend that "it is from observation that even deductive economics obtains its ultimate premises." What we can "observe" is always only complex phenomena. What economic history, observation, or experience can tell us is facts like these: Over a definite period of the past the miner John in the coal mines of the X company in the village of Y earned p dollars for a working day of n hours. There is no way that would lead from the assemblage of such and similar data to any theory concerning the factors determining the height of wage rates.

On terminology:

The distinction between macroeconomics and microeconomics is, as far as terminology is concerned, borrowed from modern physics' distinction between microscopic physics, which deals with systems on an atomic scale, and molar physics, which deals with systems on a scale appreciable to man's gross senses. It implies that ideally the microscopic laws alone are sufficient to cover the whole field of physics, the molar laws being merely a convenient adaptation of them to a special, but frequently occurring problem. Molar law appears as a condensed and bowdlerised version of microscopic law. Thus the evolution that led from macroscopic physics to microscopic physics is seen as a progress from a less satisfactory to a more satisfactory method of dealing with the phenomena of reality.

What the authors who introduced the distinction between macroeconomics and microeconomics into the terminology dealing with economic problems have in mind is precisely the opposite. Their doctrine implies that microeconomics is an unsatisfactory way of studying the problems involved and that the substitution of macroeconomics for microeconomics amounts to the elimination of an unsatisfactory method by the adoption of a more satisfactory method.

On the proper place for positivist methods:

In view of these disastrous effects of a beginning excessive reaction against the excrescences of positivism, there is need to repeat again that the experimental methods of the natural sciences are the only ones adequate for the treatment of the problems involved. Without discussing anew the endeavors to discredit the category of causality and determinism, we have to emphasize the fact that what is wrong with positivism is not what it teaches about the methods of the empirical natural sciences, but what it asserts about matters concerning which?up to now at least?the natural sciences have not succeeded in contributing any information. The positivistic principle of verifiability as rectified by Popper[2] is unassailable as an epistemological principle of the natural sciences. But it is meaningless when applied to anything about which the natural sciences cannot supply any information.

(title shamelessly lifted from the original GNXP post)

Share this

Brian, I don't know how to


I don't know how to ask the following questions well, but here goes (I'm halfway through MES, and am confused on the points mentioned in your post).

If Austrian economics is a science that aims to predict or explain human action in the real world, shouldn't we look at the world to see if the predictions or explanations work?

Neoclassical economics creates a model and claims that it matches the real world closely enough to use the model to make predictions about the real world. Is this what Austrian economics does? If not, what does it do?

Perhaps there are other sources that could help me understand?

I had heard my economics

I had heard my economics professor complain before about the extent that an overly exuberant focus on econometrics can cause academics to lose sight of the "big picture," but I can't say that I fully understood what was meant by this. How, I wondered, could a more precise way of looking at complex problems ever take away from a science such as economics?

The above post is an excellent one. Can I safely conclude that, to the extent that it is addressed to the study of econometrics, it is not condemning its use as such as merely questioning its utility in a field that often strays into the realms of social science? And stating that that utility is not as ubiquitous as some would contend?

As it happens, there is a

As it happens, there is a very recent post on the Mises website, concerning a working paper that dissects the falsities of Milton Friedman's neoclassical positivist methodology. This bears strongly on some of the issues raised here

bill, Most neoclassical


Most neoclassical economics sees itself as making a posteriori, empirical generalizations, and attempting to make such generalizations useful for prediction.

Mises saw the laws of economics as derived from a priori axioms of human nature. These laws of human action (praxeology) were self-evident, based on man's nature as an acting being who attempts to maximize utility.

For Mises, the empirical data provided by economic statistics were too complex to be predicted by any empirical generalization. They could, however, be made easier to understand after the fact if explained in terms of the axioms of praxeology.

Bad formalism is bad. But

Bad formalism is bad. But the Austrian response of no formalism is bad also.

People who know econometric techniques know they really do help empirically resolve disputes. And mathematical models have some use -- if they didn't, they wouldn't pass the market test of being used by guys like Morgan Stanley, who's got an army of PhD economists cranking them in their basement.

However, I agree that the low fruit in mathematical economics has been picked. The fields with the most potential going forward are probably institutional, property rights, and law & economics, which can be done well without much formal theory.

Much of the math-heavy theoretical stuff in the journals is garbage. But you've got to know it to criticize it. The way to counter stupid formalism isn't to lecture to PhD economists with tenure at Chicago. They don't care what any Austrian has to say. The way to counter it is get a PhD and compete, by telling more compelling stories and showing that your methodological heat can also produce light.

I have no doubt that Austrians ask the right economic questions. The problem is they don't have the skills to answer those questions in a language mainstream economists can understand. If more non-mathematical economists would start writing smart papers like Steve Levitt at the U of Chicago, the economics profession would start listening, and would improve.

See here for info on Levitt:

"These laws of human action

"These laws of human action (praxeology) were self-evident"

So it is more like philosophy than science, right? (I'm not saying this pejoratively; both are methods that can discover truth about the real world).

As an analogy, mathematics shows that 1+1=2. If I mix a pint of one liquid with a pint of another liquid and it doesn't make two pints, that doesn't change the fact that 1+1=2; something else must be going on.

So, if observations seem to contradict Austrian economic theory, it doesn't matter, any more than an observation of the real world could disprove 1+1=2; something else must be going on, and whatever it is, it is consistent with Austrian economics.

Am I getting this right?

The above blog-entry

The above blog-entry apparently neglects or is unfamiliar with the methodology used by physicists when the number of particles becomes extremely large or diverse, which is that beyond a certain point we physicists simply don't _CARE_ any more about the detailed behavior of individual particles, but only about gross properties of the system, such as conserved quantities and statistical averages. The great success of statistical mechanics was the realization that the gross behavior of "molar" quantities of matter could be sufficiently well predicted for most practical purposes merely by specifying the total energy, momentum, and other conserved quantities relevant to the problem of interest, and then seeking the "macrostate" that could be realized in the largest possible number of different ways, subject to the constraint that each of the relevant conserved quantities still be preserved; as Edwin T. Jaynes was eventually able to show using the relationships between Information Theory, Probability, and Statistics, in the limit of a large number of atoms, this state becomes _OVERWHELMINGLY LIKELY_ in a statistical sense compared to all the other possible states, in accordance with the so-called "Law of Large Numbers." Recent work by statistical physicists and "econophysicists" strongly suggests that this same approach still works when the entities involved are "actors" instead of atoms, where an "actor" can be viewed as an idealized and simplified model of an economic entity or agent (e.g., a producer or consumer) that behaves according to a small number of specified rules (e.g., each actor or agent attempts to "maximize its utility" in some statistical or average sense, subject to the limited information it has about the "macrostate" of the economy, such as prices or aggregate indices). One often finds that the aggregate or average behaviors of such "ensembles of agents" are _largely insensitive_ to the vagarities of individual agents, so that the aggregate or average behaviors of the complex system often converge onto known and well-characterized "universality classes." Also, highly complex systems often exhibit "emergent behaviors" that again fall into well-characterized "universality classes." The math is complex, but it does not appear to be _hopelessly_ intractable --- especially with the assistance of computer models to numerically estimate those quantities that cannot be analytically calculated.

As much as I respect von Mises, I think he simply missed the boat, here. Physicists and economists are both more clever than he was willing to give them credit for --- and in most cases, the apparent "irreducible complexity" of some of these problems also happens to be _largely irrelevant_ to the quantities of interest, such as average prices or average production. Progress in this area is occuring at an accelerating rate, and I would personally not be surprised if a "Predictive Statistical Macroeconomics" closely paralleling Predictive Statistical Mechanics and Thermodynamics does not emerge sometime within the next decade...

Your site is really good.

Your site is really good.