interview: Hayek on Statistics & Macroeconomic Explanation

Via Taking Hayek Seriously

interview: Hayek on Statistics & Macroeconomic Explanation: "This post was written by Greg Ransom on April 19, 2009
Posted Under: Biography, Explanation, Probability Theory, Statistics

Below I’ve posted a conversation between Friedrich Hayek and Leo Rosen from 1978 in which Hayek presents some of his conclusions about the relationships between statistics, probability theory, and macroeconomic explanation.

Note well when reading the following that Friedrich Hayek’s first job as an economist had Hayek up to his eyeballs in the collection and statistical analysis of economic data. And ditto Hayek’s second job as lead economist for the Austrian Institute of Trade Cycle Research. So statistics analysis and economic data collection were a central part of Hayek’s early life as a working economist.

ROSTEN: Dr. Hayek, I’m interested in your impressions of the empirical work that was being done by American economists. When you came here, it must have struck you rather forcibly — the stuff that was being done at the National Bureau, stuff on business cycles, in which I think you were interested at one point.

HAYEK: Well, I got interested by my visit to the United States. You see, when I came here as a young man in ‘23, I found they had nothing here to learn in economic theory. The American economic theorists had a great reputation at that time, but by the time I arrived, the few who were surviving were old men. And current teaching wasn’treally interesting from a theoretical point of view. I was actually attached to New York University, but I gate-crashed into Columbia [University] . Then I was working in the New York Public Library on the same table with Willard Thorp and other people from the National Bureau. I was drawn into that circle, and I learned a great deal about descriptive statistical work; in fact, I owe part of my later career to the fact that I learned the technique of time-series analysis at that time and was the only person in Austria who knew it. So I became director of that new institute of business-cycle research.

ROSTEN: This was in Vienna?

HAYEK: That was in Vienna, yes. Information about current affairs is very valuable; the expectation that you will learn much for the explanation of events is largely deceptive. You cannot build a theory on the basis of statistical information, because it’s not aggregates and averages which operate upon each other, but individual actions. And you cannot use statistics to explain the extremely complex structures of society. So while I will use statistics as information about current events, I think their scientific value is rather much more limited than the American economists of the last thirty or forty years have believed.

ROSTEN: I’ve left you at one point. If you say that the description of aggregates and the uses of statistics don’t help you much to explain things, and if you say that they help with contemporary events, they cease to be con- temporary very soon.

HAYEK: Oh, yes.

ROSTEN: You have built up a body of data: now, how important are those data?

HAYEK: Well, they give you an indication of what has probably happened in society during the last six months, [laughter].

ROSTEN: Do you see any more optimistic possibility for the application of statistics?

HAYEK: Not really, in economics. Demography, yes. In all fields we have to deal with true mass phenomena, but economics has not to deal with mass phenomena in the strict sense. You know where you have a sufficiently large number of events to apply the theory of probability, and proper statistics begins where you have to deal with probabilities .

ROSTEN: Well, all the sciences begin with that amassing of what might seem to be formless data. Would you tell us a little more about why you think this is not true in economics? Do you really think that most of economics takes place in discrete, isolated events, decisions, judgments?

HAYEK: Well, this leads very deeply into methodological issues; but the model of science — physical science, in the original form–has relatively simple phenomena, where you can explain what you observe as functions of two or three variables only. All the traditional laws of mechanics can be formulated as functions of two or three variables. Now, there is another extreme field, mass phenomena proper, where you know you cannot get the information on the particular events, but you can substitute probabilities for them. But there is, unfortunately, an intermediate [type of] event, where you have to deal with complex phenomena, which, on the one hand, are so complex that you cannot ascertain all the individual events, but, [on the other] , are not sufficiently mass phenomena to be able to substitute probabilities for information on the individual events. In that field I’m afraid we are very limited.

We can build up beautiful theories which would explain everything, if we could fit into the blanks of the formulae the specific information; but we never have all the specific information. Therefore, all we can explain is what I like to call “pattern prediction.” You can predict what sort of pattern will form itself, but the specific manifestation of it depends on the number of specific data, which you can never completely ascertain. Therefore, in that intermediate field — intermediate between the fields where you can ascertain all the data and the fields where you can substitute probabilities for the data–you are very limited in your predictive capacities.

This really leads to the fact, as one of my students once told me, that nearly everything I say about the methodology of economics amounts to a limitation of the possible knowledge. It’s true; I admit it. I have come to the conclusion that we’re in that field which someone has called organized complexity, as distinct from disorganized complexity.

ROSTEN: Warren Weaver.

HAYEK: Yes, exactly. Warren Weaver spoke about this. Our capacity of prediction in a scientific sense is very seriously limited. We must put up with this. We can only understand the principle on which things operate, but these explanations of the principle, as I sometimes call them, do not enable us to make specific predictions on what will happen tomorrow.

I was just listening to the wireless here, where people were speaking about the inevitable depression. Oh, yes, I also know a depression will come, but whether in six months or three years I haven’t the slightest idea. I don’t think anybody has. [laughter]

ROSTEN: Yes, life is a terminal disease. [laughter] But could you give me some examples of questions to which you — I mean about economics, or in economics — questions to which you would like answers, or to which you do not have any satisfactory –

HAYEK : Oh, any price movement of the future. I have no way of predicting them. Well, that’s exaggerating. There are instances where you can form a shrewd idea of what’s likely to happen, but in that case, of course, the price movements which you anticipate, which you expect, are already anticipated in current prices, and they are no longer true. The only interesting things are the unforeseen price movements, and they, by definition, you cannot foresee. [laughter]

ROSTEN : You were expressing your respect for Frank Knight, and once he said with great exasperation that the difference between the physical sciences and the social sciences is that in the physical sciences they don’t care what you say about them, but in the social sciences you affect the subject matter by talking about it. Now, to the degree to which people in government think they can affect economic policy, whether fine-tuning, to use that old phrase, or large-scale changes, by either changes in money supply or attempts to influence credit or so on, do you feel that we know enough to be able to make any of that kind of prediction plausible?

HAYEK: I’m sure not. I don’t think all this fine-tuning — Well, you see, that really comes back to my basic approach to economics: economic mechanism is a process of adaptation to widely dispersed knowledge, which nobody can possess as a whole. And this process of adaptation to knowledge, which people currently acquire in the course of events, must produce results which are unpredictable. The whole economic process is a process of adaption to unforeseen changes which, in a sense, is self-evident, because we could never have planned how we would arrange things once and for all and could just go on with our original plans.

ROSTEN: You mean, if those who knew, really knew, and acted upon what they knew. Are you saying that the social sciences, particularly economics, as an example, are much more complicated than the physical sciences?

HAYEK: Well, not the sciences; it’s the subject that’s much more complicated, simply in the sense that any [economic] theory would have a larger number of data to insert than any physical theory. As I said a moment ago, all the formulae of mechanics have only two or three variables in them. Of course, in real life you can use this to explain an extremely complex phenomenon, but the
underlying theory is of a very simple character. With us, you can’t have a theory of perfect competition without at least having a few hundred participants. And you would have to be informed about all their knowledge in order to arrive at a specific prediction. The very definition of our subject is that it’s built up of a great many distinct units, and it wouldn’t be a subject of that order if the elements weren’t so numerous. You cannot form a theory of competition with only three elements in it.

ROSTEN: You could certainly have a theory.

HAYEK: Well, it would be wrong, because it wouldn’t be competition with only three acting persons in it.

ROSTEN: Well, just explain that. What about four?

HAYEK: No, I don’t think it’s the approach. But you have to have a number where it’s impossible for any one of them to predict the action of the others, and there must be a sufficient number of others for the one to be unable to predict it.

ROSTEN: You say that’s in the order of a hundred, or hundreds, or thousands, and so on.

HAYEK: Yes.

ROSTEN: It’s a startling theory, and I’ve not heard it put quite this way.

HAYEK: But, you know, the whole market is due to the fact that people are aiming at satisfying needs of people whom they do not know, and use for their purposes facilities provided by people of whom they also have no information. It’s a coordination of activities where the individual can, of necessity, be only a small part of it — any individual, not only the participating individuals but even any outsider. The mistaken conception comes from a very curious use of the term data. The economists speak about data, but they never make clear to whom these data are given. They are so unhappy about it that occasionally they speak even in a pleonasm about “given data,” just to reassure themselves that [the data] are really given. But if you ask them to whom they are given, they have no answer. [laughter]

ROSTEN: You mean “revealed”?

HAYEK: They are fictitiously assumed to be given to the explaining theorists. If the data were such and such, then this would follow. But of course the data are not really given either to them or to any one other single person, They are the widely dispersed knowledge of hundreds of thousands of people, which can in no way be unified; so the data are never data.

ROSTEN: It’s almost as if you were talking about nuclear physics and the difficulty, or impossibility, of talking about an atom and how it’s going to behave.

HAYEK: Yes. It’s a different argument. You see, in nuclear physics, up to a point, you can substitute information about individual elements by probability calculations. There the numbers are big enough for the law of large numbers to operate. In economics they are not. They are too big to know them individually and not big enough to be described by probability calculations.

ROSTEN: Do you think that this is a permanent and unbreakable prison?

HAYEK: Yes. I don’t think we can ever get beyond that.

ROSTEN: –because earlier you had said something about the processes of proof and the fact that you couldn’t prove anything. And I was reminded of the work, of which I know very little and which I know you know a great deal about, of Caddel, at Princeton [University].

HAYEK: Yes.

ROSTEN: — on the terrible, to me tragic, built-in trap that he has discovered in the uses of logic, and in what
you earlier had talked about as the uses of reason.

HAYEK: You see, I became aware of all this not by my work in economics but–I don’t know whether you know that I once wrote a book on psychology.

ROSTEN: No, I did not know.

HAYEK: On physiological psychology–a book called The Sensory Order –in which I make an attempt to provide at least a schema for explaining how physiological processes can generate this enormous variety of qualities which our senses represent. [The schema is] called “the sensory order.” [The book] ends up with the proof that while we can give an explanation of the principle on which it operates, we cannot possibly give an explanation of detail, because our brain is, as it were, an apparatus of classifi cation. And every apparatus of classification must be more complex than what it classifies; so it can never classify itself. It’s impossible for a human brain to explain itself in detail …

.. [Work on The Sensory Order] taught me a great deal on the methodology of science, apart from the special subject. What I later wrote on the subject, the theory of complex phenomena, is equally the product of my work in economics and my work in psychology.

ROSTEN: And you had not then been working in statistics.

HAYEK: No, although I’ve nearly all my life had the title of Professor of Economics and Statistics, I’ve never really done any statistical work. I did do practical statistics as the chief of that Austrian Institute of Trade Cycle Research …

ROSTEN: The work that you started on business cycles, I assume, was not unlike the work later done by [Simon] Kuznets and his group at the institute.

HAYEK: Well, again, you see, it was an abstract schema without much empirical work. I had some very elementary data boom there was an excessive development of production of capital goods, much of which afterwards turned out to be mistaken. And I didn’t need many more facts for my purpose to develop a theory which fits this, and which exclusively shows us, [using] other accepted data, that a credit expansion temporarily allows investment to exceed current savings, and that it would lead to the overdevelopment of capital industries. Once you are no longer able to finance a further increase of investment by credit expansion, the thing must break down. It becomes more complicated in conditions when the credit expansion is no longer done for investment by private industry but very largely by government. Then you have to modify the argument, and our present booms and depressions are no longer explicable by my simple scheme.

But the typical nineteenth- and early- twentieth-century [phenomena], I think, are still adequately explained by my theory — but not adequately to the statisticians, because, again, all I can explain is that a certain pattern will appear. I cannot specify how the pattern will look in particular, because that would require much more information than anyone has. So, again, I limit the possible achievement of economics to the explanation of a type – One of my friends has explained it as a purely algebraic theory.

ROSTEN: An algebraic theory?

HAYEK: Yes, you get an algebraic formula without the constants being put in. Just as you have a formula for, say, a hyperbola; if you haven’t got the constants set in, you don’t know what the shape of the hyperbola is — all you know is it’s a hyperbola. So I can say it will be a certain type of pattern, but what specific quantitative dimensions it will have, I cannot predict, because for that I would have to have more information than anybody actually has."