Keynes’s Critique of Econometrics Is Surprisingly Good
In a recent article we had a brief look at Ragnar Frisch’s (1895–1973) vision of econometric model building. As mentioned, Frisch was the first economist chosen over Mises to win the Nobel Prize in 1969. In fact, there was a second one in the same year. Frisch won the prize jointly with Dutchman Jan Tinbergen (1903–1994), who applied Frischian econometrics for the first time in large-scale macro models by the end of the 1930s.
In the first volume of his investigations into business cycles commissioned by the League of Nations, entitled Statistical Testing of Business Cycle Theories, published in 1939, Tinbergen exonerates the statistician and econometrician from his responsibility and explains:
The part which the statistician can play in this process of analysis must not be misunderstood. The theories which he submits to examination are handed over to him by the economist, and with the economist the responsibility for them must remain; for no statistical test can prove a theory to be correct.
While classical and Austrian economists would agree that an economic theory cannot be proven correct empirically, they would not as easily let the statistician off the hook. Indeed, the econometrician and statistician have some responsibility for the economic theories that come to be accepted, especially if one holds, as Tinbergen does, that those theories can be proven “incorrect, or at least incomplete, by showing that it does not cover a particular set of facts.”
This is an odd claim, since practically any theory is incomplete, but this does not mean that it is incorrect. Obviously there remains a twofold danger: A wrong theory might not be proven wrong, although it could be done in principle, and a true theory might be “proven wrong” mistakenly, because it is incomplete as it does not account for some particular set of facts. The econometrician would of course be responsible for these errors.
In fact, the application of statistical methods for the falsification of economic theories is not unconditionally justified. Depending on the kind of theory, there are a number of necessary conditions, which are usually not sufficiently recognized, or even deliberately ignored by modern econometricians.
Interestingly, in a review published in The Economic Journal in 1939, “The Master” himself, Lord Keynes, gave a comprehensive critique of Tinbergen’s work and provided a list of some of these conditions. Any honest economist must confess that they are not satisfied with the overwhelming majority of economic problems, particularly those that are as complex as the business cycle.
Completeness: Tinbergen attempts to quantify the extent to which various factors contribute to the business cycle. Keynes points out that one has to have a complete list of all relevant factors for this to be actually possible. It is clear that when one factor is missing and not taken into account, changes that are caused by this factor could falsely be interpreted as the results of the factors that are taken into account.
At the very least, the economy has to be “homogenous” over the time span under consideration, that is to say, ignored factors have to remain constant so that they do not alter the empirical findings. Yet, there is no way of achieving this constancy.
Measurability: All relevant factors have to be measurable in principle, and we have to “have adequate statistical knowledge of their measure” in practice. Keynes asks about “political, social and psychological factors, including such things as government policy, the progress of invention and the state of expectation.” These elements certainly have an impact on the business cycle, but are they quantifiable and measurable? And if so, do we have a way to gain reliable statistical information about them? Probably not.
Independence: Next, Keynes points out that the factors have to be independent, or otherwise we are running the risk of finding “spurious” correlations. Tinbergen is certainly aware of this problem and remarks at one point that “one has to be careful.” “But is he?” asks Keynes. Not really. Although Tinbergen mentions the problem himself, he proceeds without discussion of how to solve it.
Linearity: The fourth problem to which Keynes directs our attention is that of linearity. Tinbergen is only considering linear relationships. Indeed, non-linear equations are hard to handle at times and computers were yet to be developed, but this does not change the fact that the econometricians around Tinbergen set out to falsify economic theories under the crude and extremely simplistic assumption that only linear relationships between the relevant variables are possible. As Keynes writes:
It is a very drastic and usually improbable postulate to suppose that all economic forces are of this character, producing independent changes in the phenomenon under investigation which are directly proportional to the changes in themselves; indeed, it is ridiculous.
Arbitrariness: Moreover, Tinbergen adds time-lags and trends to certain variables in his system. This however happens more or less arbitrarily, in a “trial and error” process, “that is to say, he fidgets about until he finds a time-lag which does not fit in too badly with the theory he is testing and with the general presuppositions of his method.” There is no guarantee that these time-lags and trends are not deceiving the econometrician and his audience about serious drawbacks in the rest of the model.
Although one has to admit that Tinbergen is very careful not to claim too much, his overall mission was a futile endeavor. Keynes summarizes:
I have not noticed any passage in which Prof. Tinbergen himself makes any inductive claims whatever. He appears to be solely concerned with statistical description. Yet the ultimate purpose which Mr. Loveday outlines in the preface is surely an inductive one. If the method cannot prove or disprove a qualitative theory, and if it cannot give a quantitative guide to the future, is it worth while? For, assuredly, it is not a very lucid way of describing the past.
Keynes claims that Tinbergen’s approach if thought through logically must lead to “devastating inconsistencies.” He proposes a rather charming comparison:
It becomes like those puzzles for children where you write down your age, multiply, add this and that, subtract something else, and eventually end up with the number of the Beast in Revelation.
Murray Rothbard once wrote: “There is one good thing about Marx: he was not a Keynesian.” There also seems to be one good thing about Keynes: he was not an econometrician in the modern sense.