Uneasy Money

Nobel Laureate T. C. Koopmans wrote one of the most famous economics articles of the twentieth century, “Measurement Without Theory,” a devastating review of an important, and in many ways useful and meritorious, study of business cycles by two of the fathers of empirical business-cycles research, Arthur F. Burns and Wesley C. Mitchell, Measuring Business Cycles. Burns was then the head of the National Bureau of Economic Research, which had been founded by Mitchell, Burns’s mentor, to promote the empirical study and measurement of business cycles. Koopmans, who was then head of the Cowles Commission, another foundation dedicated to the study of business cycles, but with a theoretical, rather than a purely empirical, orientation, explained why attempting to gather empirical facts about business cycles and to measure aspects of business cycles in an atheoretical manner would fail to produce knowledge that useful for trying to understand business cycles as a characteristic feature of modern economic activity, or for devising policies mitigate their undesirable effects.

Koopmans’s review, a rhetorical masterpiece, brilliantly deploying his analytical skill and erudition as a Ph.D. physicist, compared the work of Burns and Mitchell to pre-Newtonian attempts to develop theories of planetary motion through the accumulation of data from planetary observations, largely shut down the program of the National Bureau as a center for business-cycle research, though it has continued to function usefully as a sponsor of a variety of important research efforts on business cycles and other fields of study. Whether the shift in emphasis from atheoretical empirics and data accumulation to theoretical research was a positive development is still controversial, but that discussion is not the point of this post.

All I want to do here is to reproduce a single paragraph from Koopmans’s review in which, without mentioning the Phillips Curve, which had still not been discovered (apologies to Irving Fisher who, almost completely unnoticed, actually discovered it about 30 years before Phillips did), beautifully explains why the Phillips Curve cannot be used as a guide to conducting monetary policy or countercyclical policy of any kind.

[Economic] theories are based on evidence of a different kind than the observations embodied in time series: knowledge of the motives and habits of consumers and of the profit-making objectives of business enterprise, based partly on introspection, partly on interview or on inferences from observed actions of individuals–briefly, a more or less systematized knowledge of man’s behavior and its motives. While much in these theories is incomplete and in need of reformulation and elaboration (particularly in regard to behavior over time under conditions of uncertainty), such theory es we have is an indispensable element in understanding in a quantitative way the formation of economic variables. For according to that theory the relevant economic variables are determined by the simultaneous validity of. an equal number of “structural” equations (of behavior, of law or rule, of technology). The very fact that so many relations are simultaneously valid makes the observation of any one of them difficult, and sometimes even impossible. For any observed regularity between simultaneous and/or successive values of certain variables may have to be ascribed to the validity of several structural equations rather than any one of them. The mere observation of regularities in the interrelations of variables then does not permit us to recognize or to identify behavior equations among such regularities. In the absence of experimentation, such identification is possible, if at all, only if the form of each structural equation is specified, i.e., in particular, if we can indicate the set of variables involved in each equation, and perhaps also the manner in which they are to be combined. In each case, a preliminary study of the system of structural equations held applicable is required to decide whether the specifications regarding any particular equation are sufficiently detailed to permit its identification. Without such identification, measurement of the structural equation involved is not possible, and should therefore not be attempted. 

One might object: why should measurement of the behavior equations of consumers, workers, entrepreneurs be necessary? If observed regularities are due to the simultaneous validity of several behavior equations, these regularities will persist as long as each of the underlying (unknown) behavior patterns persists. However, there are important arguments to counter this objection. Sheer scientific curiosity still urges us on to penetrate to the underlying structural equations. This curiosity is reinforced and justified (if you wish) by the awareness that knowledge of the behavior patterns will help in nderstanding or analyzing different situations, for instance, problems of secular trend, or cyclical problems in other countries or periods–in the same way (although one would not expect with the same exactness) in which the law of gravitation explains celestial and terrestrial phenomena alike. This point has particular relevance with regard to the different situations expected to arise in an impending future period of the same country that has been studied. Behavior patterns are subject to change: gradually through changing habits and tastes, urbanization and industrialization; gradually or unevenly through technological change; abruptly through economic policies or the economic effects of political events. While one particular behavior pattern may be deemed fairly stable over a certain period, a much greater risk is involved in assuming that a whole system of structural equations is stable over time. An observed regularity not traced to underlying behavior patterns, institutional rules, and laws of production, is therefore an instrument of unknown reliability. The predictions it yields cannot be qualified with the help even of known trends in behavior or technology. It is of no help whatever in assessing the probable effects of stated economic policies or institutional changes.

There is no sign in the book of an awareness of the problems of determining the identifiability of, and measuring, structural equations as a prerequisite to the practically important types of prediction. Measurable effects of economic actions are scrutinized, to all appearance, in almost complete detachment from any knowledge we may have of the motives of such actions. The movements of economic variables are studied as if they were the eruptions of a mysterious volcano whose boiling caldron can never be penetrated. There is no explicit discussion at all of the problem of prediction, its possibilities and limitations, with or without structural change, although surely the history of the volcano is important primarily as a key to its future activities. There is no discussion whatever as to what bearing the methods used, and the provisional results reached, may have on questions of economic policy.

This, then, is my second argument against the empiricist position: Without resort to theory, in the sense indicated, conclusions relevant to the guidance of economic policies cannot be drawn.

T. C. Koopmans, “Measurement Without Theory,” The Review of Economics and Statistics 29(3): 161-72, pp. 166-67