- 지은이: Don Patinkin
- 출처: Econometrica, Vol. 44, No. 6, November 1976
- 제목: Keynes and Econometrics: on the Interaction between the Macroeconomic Revolutions of the Interwar Period.[주1]
* * *
Keynes's general attitude toward mathematical economics and econometrics, respectively, is discussed in Sections 2-3. The remainder of the paper is devoted to a description and analysis of the interaction between the Keynesian revolution of the mid-1930s and the revolution that actually started somewhat earlier with respect to the preparation of current official estimates of national incme. In this connection an attempt is made to explain why Kuznets' work in the U.S. in the early 1930s was immediately integrated into official national income estimates in the U.S., where Colin Clark's work in Britain was notㅡwith the result that official British national income estimates did not begin to appear until almost a decade later.
[주1] Presidential Addres delivered before the Econometric Society meetings in San Francisco in December 1974, with subsequent elaborations.
A preliminary version of this paper was presented in December 1974 before the Economics Department Seminar at the Hebrew University of Jerusalem, and I am grateful to my colleagues for the many helpful and stimulating comments and suggestions which they then provided. ( ... ) Similarly, I have benefited from stimulating discussions with Simon Kuznets to whom I am also indebted for permission to quote from his correspondence with Keynes. ( ... )
* * *
( ... ) The title will lead me to a discussion of the attitudes and practices of John Maynard Keynes with respect to econometrics. And quite apart from the renewed interest of the last few years in the work of Keynes, there is ample justification for discussing this subject on this occasion. [:]
- First, as we shall see, Keynes was indeed concerned with the econometrics of his timeㅡin the broad sense of empirically oriented economic analysis.
- Second, and more significant, Keynes's General Theory of Employment, Interest, and Money ("General Theory") almost 40 years ago defined the framework of research in macroeconomics for many decades which followed; and the relation of this book to econometrics is amply attested by the fact that, for example, the most influential interpretation of itㅡthe IS-LM interpretation which John Hicks presented in his "Mr. Keynes and the Classics"ㅡappeared as an article in ^Econometrica^(1937).
- Furthermore, and most important in the present context, the desire to quantify the General Theory provided the major impetus for the exponentially-growing econometric work that began to be carried out in the later 1930's on the consumption, investment, and liquidity-preference functions individually and, even more notably, on econometric models of the Keynesian system as a whole.
- But in addition to, and as a reflection of, the impetus that he gave to econometric work, Keynes also had some very important formal connections with the Econometric Society. Thus he was one of the thirty economists from all over the world selected by the Council in 1933 to constitute the first group of Fellows of the Society[Econometrica, 1 (1933), p. 445]. And a year later, at the initiative of Ragnar Frisch, Keynes was elected to the Council itself, and remained a member of it until his death.
But my talk also has a subtitle. And the subject described by this subtitle is as broad and complex as it is fundamental, for it deals with the interaction of ideas, and with the subtle and mysterious ways in which such interactions take place. And it also deals with the equally, if not more, complicated question of the interaction between ideas and institutions.
Let me be more specific: for many years now we have become accustomed to using the term "Keynesian Revolution" to denote the dramatic changes which Keynes' ^General Theory^ effected with respect to macroeconomic theory. I feel, however, that we are much less aware than we should be of the no less significant (though quieter) revolution that began to take place even before the ^General Theory^ with respect to macroeconomic measurement or, more specifically, with respect to the measurement of national income, which is the general term I shall for simplicity frequently use to denote the measurement of any one or more of the national aggregates (income, product, expenditure), broken down by their respective components. In Section 5 below I shall justify the use of the term "revolution" in this context; for the moment let me simply note that it is the one associated primarily with the names of Simon Kuznets in the United States and Colin Clark in England. And it is to the interrelationships between these two revolutions that my subtitles refers. ( .... ) What I hope to do, however, is to highlight the major aspects of this question, with particular emphasis on the way the aforementioned interrelationships manifest themselves in the work of Keynes.
2. Keynes and Mathematical Economics
Though this is my main concern, I cannot discuss the subject of Keynes and econometrics without first digressing briefly on two other points that always arise in this context: the first is Keynes' attitude toward mathematical economics, and the second is Keynes's famous 1939 debate with Tinbergen.
On the first point, I can only repeat what I have already said elsewhere: We are all familiar with Keynes' oft-cited criticism in the General Theory of "symbolic pseudo-mathematical methods of formalizing a system of economic analysis ... which allow the author to lose sight of the complexities and interdependencies of the real world in a maze of pretentious and unhelpful symbols" (GT, pp. 297-298). We should not, however, uncritically accept this statement as if it were an expression of Keynes' unchanging and definitive rejection of mathematical methods. First of all, Keynes' own analysis in the Treatise on Money was in fact largely based on fairly mechanical applications of the so-called "fundamental equations" . Furthermore, the Treatise devoted an entire chapter (20) to "An Exercise in the Pure Theory of the Credit Cycle", in which Keynes explored in a very formalistic manner, and under a variety of alternative assumptions, the mathematical properties of his model of the cycle. Indeed, if ever an author "lost sight of the complexities of the real world in a maze of pretentious and unhelpful symbols", that author was Keynes of the Treatise (see Patinikin (1976, Ch. 4-7).
 See Patinkin (1975, pp. 265-266) or (1976, pp. 21-23) from which this and the following three paragraphs are, with minor changes, reproduced.( ... ... )
3. The Keynes-Tinbergen Debate
Keynes' famous debate with Tinbergen began with Keynes' review in September, 1939 Economic Journal of the pioneering statistical study of business cycles which Tinbergen carried out on behalf of the League of Nations in 1939 as a complement to Haberler's earlier study (1937) of business-cycle theory. In order to place this debate in its proper context however, I must first of all emphasize that Keynes' review-article was entitled "Professor Tinbergen's Method" and was devoted not to the much better-known second volume of this study on ^Business Cycles in the United States of America, 1919-1932^, but to the first volume (which was published a few months earlier) on ^A Method and Its Application to Investment Activity^, in which Tinbergen set out and exemplified the principles of multiple-correlation analysis. Accordingly, the criticisms which Keynes presented in this review were levelled not at Tinbergen's ambitious 46-equation model of the United States economy, but at the use of correlation analysis to estimate even a single equation.
Let me also confess that, though not all of Keynes' criticisms were well taken (e.g., his "suspicion that the assumption of linearity rules out cyclical factors" (^JMK^ vol. 14, p. 313; see also Tinbergen's reply, 1940, p. 150)), I find it somewhat depressing to see how many of them are, in practice, still of relevance today. Thus Keynes wrote:
Am I right in thinking that the method of multiple correlation analyses essentially depends on the economist having furnished not merely a list of the significant causes, which is correct so far as it goes, but a ^complete^ list? For example suppose three factors are taken into account, it is not enough that these should be in fact ^verae causae^; there must be no other significant factor. If there is a further factor, not taken account of, then the method is not able to discover the relative quantitative importance of the first three. If so, this means that the method is only applicable where the economist is able to provide beforehand a correct and indubitably complete analysis of the significant factors. (^JMK^ vol. 14, p. 308, italics in original).What could be a better description of specification bias? Or, again,
... Professor Tinbergen is concerned with "sequence analysis"; he is dealing with non-simultaneous events and time lags. What happens if the phenomenon under investigation itself reacts on the factors by which we are explaining it? For example, when he investigates the fluctuations of investment, Professor Tinbergen makes them depend on the fluctuations of profit. But what happens if the fluctuations of profit partly depends (as, indeed, they clearly do) on the fluctuations of investment? (^JMK^ vol. 14, pp. 309-310).In brief, what we now call the simultaneous-equation bias.
In his review article, Keynes also referred to the basic difficulty of measuring expectations (^JMK^, vol. 14, p. 309), to the restrictive nature of the assumption of universal linearity (^JMK^ vol. 14, pp. 311-315), as well as to "the frightful inadequacy of most of the statistics employed" (^JMK^ vol. 14, p. 317). And, in what I interpret as an allusion to his ^Treatise on Probability^, he stated: "thirty years ago I used to be occupied in examining the slippery problem of passing from statistical description to inductive generalization in the case of simple correlation; and today in the era of multiple correlation I do not find that in this respect practce is much improved" (^JMK^ vol. 14, p. 315). 
Keynes concluded his review of Tinbergen's work with the following comment:
I hope that I have not done injustice to a brave pioneer effort. The labor it involved must have been enormous. The book is full of intelligence, ingenuity and candor; and I leave it with sentiments of respect for the author. But it has been a nightmare to live with, and I fancy that other reader will find the same. I have a feeling that Professor Tinbergen may agree with much of my comment, but that his reaction will be to engage another ten computer and drown his sorrows in arithmetic. (^JMK^ vol. 14, p. 318).( ... ... )
4. The Use of National Income Estimates in the General Theory
Let me now return to my main concernㅡthe interaction between the macroeconomic revolutions of the interwar period. Appropriately enough for this purpose, the impact of the revolution that took place with respect to the availability of national income data can be illustrated most dramatically by contrasting Keynes' two famous work on macroeconomics: his Treatise on Money published October 1930, and the General Theory published in February, 1936.
The ultimate purpose of the Treatise was to explain the fluctuations in output that characterized the business cycle or, as Keynes called it in the Treatise, the credit cycle. The vehicles for this explanation were the famous "fundamental equations," in which a crucial role was played by the relation between investment and saving. As is well known, according to Keynes' definitions in the Treatise, these two quantities are not generally equal. Indeed, the excess of investment over saving equals the excess profits of firms, and the existence of such profits leads firms to expand output. Conversely, when savings exceed investment, there are losses, and output declines. This, in a highly oversimplified form, is the cycle theory of the Treatise.
 For details, see Patinkin (1976, Chapters 4-6).Now the Treatise was not only a book on theory; as its title indicates, it was intended to be a comprehensive treatment of monetary economics, dealing with the applied aspects (which is the subject of Volume II of the Treatise) as well as the theoretical ones (the subject of Volume I). Correspondingly, in this second volume, Keynes attempts to presents empirical estimates of the variables that played a key role in the theory presented in the first one.
In this context he presents an index of total output (for the period of 1920-29) which, for lack of anything better (and Keynes bemoans "the present deplorable state of our banking and other statistics"(TM II, p. 78) which leaves no alternative to such estimates), is simply the average of two existing indices of employment and industrial use of raw materials, respectively(TM II, p. 79 last column of table).
The situation is far worse, and Keynes' complaints correspondingly greater ("the relevant statistics ... are few and unsatisfactory. There is no single set of figures which measures accurately what should be capable of quite precise measurement" (TM II, p. 87)) with respect to estimates of total investment in fixed capital. Keynes rejects the use of data on the volume of new issues for this purpose both because these do not reflect residential construction, a major component of such investment was actually carried out. Accordingly, Keynes suffices with a general reference to Wesley Mitchell's summary in his ^Business Cycle^(1927) of various time series connected with different aspects of investment activity, and claims that this summary shows that "the fluctuations [in the rate of investment in fixed capital] are substantial and that they are correlated with the phases of the credit cycle in quite as high a degree as our theory would lead us to expect" (TM II, pp. 88-89). Keynes then uses a variety of a prioristic assumptions in order to derive an estimate of investment in "working capital" (TM II, pp. 92-100). Finally, he combine his estimates of investment in fixed and working capital, as well as of investment abroad, in order to present a table of total net investment in Britain for the period of 1919-1924, which he describes in the following words: "The following calculation is not based on statistical data, but is a not plausible guess as to what may have happenㄷdㅡintended to illustrated my argument rather than to state an historical data" (TM II, p. 101).
In less than 6 years all this is changed. In particular, when in the General Theory Keynes came to deal with the same basic contentionㅡthe crucial role of the fluctuating volume of investment in generating business cyclesㅡhe was able to support his views by citing the estimates of total investment that Colin Clark had presented for Britain for the period 1928-1931 in his book on ^The National Income: 1924-1931^(1932, pp. 117, 138) as well as the preliminary estimates for the United States for the period 1925-1933 which Simon Kuznets had presented in his NBER Bulletin on ^Gross Capital Formation, 1919-1933^(Bulletin 52, November 15, 1934). Keynes also cites these tables in support of his contention that depreciation allowances and the like "normally" bear a high proportion to the value of gross investment; but at the same time he expresses the view that "Mr. Kuznets' method must surely lead to too low an estimate of the annual increase in depreciation" (GT, pp. 102-104).
( ... .... ... )
Let me at this point note that, to the best of my knowledge, this was the first estimate of the marginal propensity to consume that was based on an examination of statistical time series. And in any event, the members of this audience will be particularly interested to note that by virtue of his having derived his estimate of the marginal propensity indirectly from an estimate of the multiplier, Keynes might be said to have been the first person to have essentially made use (even if unintentionally) of something like the reduced-form method of estimation! And all this several years before Trygve Haavelmo's painstaking work (1943, 1944, and 1947) on the methodological need and justification for such a method.
( ... ... )
( ... .... ... )
5. On the Nature of the Statistical Revolution
From the sequence of events that I have described above, it is obvious that the statistical revolution as represented by Clark's and Kuznets' national income estimates preceded the "Keynesian Revolution" as represented by the General Theory. Let us now look into this story in somewhat greater detail.
( ... ... )
I have elaborated on all this in order to bring out the fundamental point that these national product estimates of the early 1930s were not the outcome of idle curiosity satisfying itself by the mechanical collection of dataㅡnot an exercise in measurement for the sake of measurementㅡbut were in varying degrees motivated by the desire to quantify those macroeconomic variables to which the pre-General Theory theories of the business cycle had already attached crucial significance.
At the same time, it must be emphasized that the subsequent appearance of the General Theory, with its revolutionary analysis of the determination of the equilibrium level of output by means of the aggregate demand for consumption and investment goods, gave a further and decisive impetus to the preparation of national income estimates by these categories. For though Colin Clark had first provided such an estimte in 1932 (p. 117, Table XLV), it is only after the Keynesian Revolution that we find national income estimates widely presented in the C+I+G=Y rubric which is the cornerstone of Keynesian economics.
( ... ... )
And as is so frequently the case with technological improvements, on certain fronts the newcomer to the process surged ahead for a time. I say this especially with respect to the development by Meade and Stone(1941) of the broad conceptual framework of social accounting within which the national income estimates were place. But I also say it with reference to the estimation of national income by the final-product C+I+G=Y rubric which Keynes has developed in his General Theory(1936) and applied empirically in his ^How to Pay for the War^(1940). ( ... ... )
 Which, in turn, drew from his earlier article in the Economic Journal on "The Income and Fiscal Potential of Great Britain"(1939); see also his supplementary note a year later on "The Concept of National Income"(1940)( ...