Keynes and Tinbergen
In 1939 there was a debate, mainly between Keynes and Tinbergen, ostensibly about econometrics but with broader ramifications for the applications of mathematics and notions of uncertainty.
The debate is described and contextualised by:
F. Louca, The Econometric Challenge to Keynes: Arguments and contradictions in the early debates about a late issue European Journal of the History of Economic Thought, Second draft, October 1997.
From Louca’s abstract:
Keynes was deeply hostile at least to use of the current mathematical formalism, and made no secret of that.
[T]he acceptance of the epistemological primacy of a very peculiar brand of a simple mathematical formalism for the macro-theories led to the wiping out of the major theoretical alternatives of the first half of the century.
Keynes was more hostile and certainly more aware of the dangers of the mechanistic and dominant simplistic mode of mathematical expression of the economic models.
The final result was a Phyrric victory – or defeat – of both the Keynesian program and the original intentions of the founders of econometrics, as most of them recognised with sorrow.
This lingering debate is of relevance to the debate about the 2007/8 crash. When people blame mathematics, it is the ‘very peculiar brand of a simple mathematical formalism’ that seems at fault, not Keynes’ more nuanced approach to uncertainty, as in his mathematical Fellowship Treatise.
[T]he econometricians did not care too much about Keynes‘s critique: it was anticipated and summarized as the mere implication of a nasty … and permanently sceptic attitude against mathematics.
This seems odd. Keynes’ critique was a defense of the logically-founded mathematics of Russell and Whitehead (as reflected in his Treatise) against the more traditional views of Tinbergen, an economist with leanings towards the methods of classical Physics. Wasn’t Keynes defending mathematics?
Loucas quotes a letter from Keynes to Lange:
Does not every case to which Tinbergen has applied his method assume that the same formula is valid over a long period of years? [T]his is not merely a casual assumption but one which is intrinsic to the whole way of proceeding.
The same objection could be applied to the (mis) use of econometrics prior to the 2007/8 crash. In effect, such econometricians assume that nothing surprising can happen. And yet it does. Loucas quotes a 1958 response by Frisch:
I have believed that the analytical work will give higher yields … if they become applied in macroeconomic decision models where the line of thought is the following: ‘If this or that policy is made, and these conditions are met in the period under consideration, probably a tendency to go in this or that direction is created’.
This seems to take on board Keynes’ main objections by verbally recognizing the conditionality and residual uncertainty inherent in any predictions. It would surely been an improvement if, prior to 2007/8, econometricians had recognized this and striven to identify the conditions necessary to the continuance of the status quo. But Loucas goes on:
Keynes addressed the problem in a rather different way, since he restrained himself to the short term and to the use of known behavioural relations, even if not completely quantified.
This seems an odd interpretation of Keynes. If a regularity is conditional and may fail probabilistically, then it is true that one can only expect it to hold ‘in the short term’, as emphasised by notions of ‘Black Swans’. But while Keynes regards simplistic econometric thinking as limited to the short-term, he was clearly not averse to using empirical data (including econometrics) to inform theoretical thinking about conditionality, to inform longer-term institutionalizing and policy-making.
Loucas says of the econometricians:
They did not accept … the non-mathematical alternative formulation Keynes was defending, since they deeply shared the conviction that exactness was desirable, possible, attainable, and even indispensable considering the task of economics.
Again, an odd-seeming comment. As a mathematician who came to economics via statistics, Keynes advocated a mathematical approach that recognized the inherent uncertainties, as Frisch (later) did (above). Keynes did seem to prefer not using mathematics to using mathematics inappropriately, heedless of the appropriate theory. But post the 2007/8 crash this may seem reasonable, even for a mathematician. Perhaps the problem was Keynes’ opponents ‘deeply held conviction’ that exactness was possible, in contrast to the arguments in Keynes’ Treatise. Thus econometrics seemed to be based on ideas from Physics that were already outmoded.
From a historical perspective, Keynes’ attack on what became conventional econometrics is more important for what people thought he said than what he actually said. But what he said is also of interest.
Professor Tinbergen’s Method The Economic Journal, Vol. 49, No. 195 (Sep., 1939), pp. 558-577.
[The method] is a means of providing quantitative precision to what … we already know … where the other considerations given below are already satisfied.
This seems similar to Frisch’s 1958 position, save that it makes use of Keynes’ Treatise in identifying conditions under which induction from a description to something more like a prediction is justified, and with what caveats. At the heart is that Tinbergen assumes that economies are only driven by measureable phenomena, something which Keynes contradicted in his General Theory. The econometric approach implicitly assumed that there would be a fixed policy, reacting to the econometrics in a formulaic manner, thus making ‘thinkers’ like Keynes redundant. Keynes doubted if this were risk-free, and following the 2007/8 crash, so might we.
Technically, the debate seems to have been misrepresented and misunderstood. A consensus seems to have emerged on the general approach, but Keynes’ work on uncertainty seems to have gone unrecognized, until recently – and even then only partially. Econometrics can be used to refine, but does not – in itself – identify uncertainty or its causes.
One might think that it would be better to have a fixed, perhaps global, intervention policy, responding to econometrics in a formulaic manner, rather than trust politicians or central bankers, but that would be another issue. Perhaps the issues got mixed up?