828. Harrod to J. M. Keynes , 7 September 1938 [a]
[Replies to 806 and 817 , answered by 833 ]
Christ Church, Oxford #
7 September 1938
Many thanks for letting me see this.  It is impossible to deny the cogency of your criticisms, though I think you may under-estimate the thoroughness of the work done behind the scenes, i.e. the number of other hypotheses tested, the statistical impartiality of his work etc.
My main point, however, is that you have got the balance of emphasis quite wrong, tho' you are amply justified by the obscurity of his screed. At our meeting I persuaded them that he should write a new introduction,  which should entirely shift the emphasis and make the volume produce quite a different impression.
He is interested in the fundamental point that if you assume certain lags of adjustment, e.g. between increase of profit and plant expansion, or between increase of income & spending etc. you automatically get a cycle. It is really a lag theory of the cycle. To show how it works you have to make a model. Now in making the model you have to put in certain assumptions, e.g. people spend of income etc. The duration and amplitude of your cycle will vary according to the co-efficients you put in and the variables you assume important.
The object of the statistical enquiry, as I see it, is to get a first approximation to the right figures for the construction of your model.
It does not follow that these co-efficients will be right for future cycles and indeed he does not claim any great precision for them for the cycle examined. He merely wants to get something a little better than purely hypothetical co-efficients. If you dont have any co-efficients at all you cant make your Theory, and if it happens that there is any truth in the lag theory, a strict interdict would disenable us from ever discovering it! The peculiarity of the lag theory is that you must work with some quantities.
Tinbergen claims that he has tried to include all the measurable factors which economists have urged to be important. But it comes out that some influences, unless the variations in the magnitudes were much bigger than they are, cannot have much effect when brought into relation with other influences in the mathematical machine.
With regard to other random non-measurable influences, these are not left out, but come in as "shocks". Naturally they are not taken into account specifically, for if you insisted on doing that you would have to abandon all hopes of a cycle theory and lapse into cycle history.
Indeed his method not only takes the random influence into account but enables you to deal with them in a more systematic way. Given that his lag causes are roughly right, he can show how intense the shocks have to be to keep the cycle going.
It is rather presumptuous my attempting to interpret him, as my understanding is very incomplete. A number of good brains are now tackling this kind of method and I think it is important not to discourage the authorities from keeping them in funds. In particular I have a very high regard for Tinbergen's ability & objectivity.
2. Harrod refers to the "Introduction" to Tinbergen, "A Statistical Test of Business Cycles Theories" (1938) (later published as Statistical Testing of Business Cycles Theories, 1939). See also letter 823 , [jump to page] .
- a. ALS, four pages on two leaves, in JMK CO/11/281-84. Photocopy in HP II-200. Printed in Keynes, Collected Writings, vol. XIV, pp. 303-4.
top of page
Return to index of this section
Go to previous page
Go to next page