Federal Reserve Bank of Minneapolis
Quarterly Review Spring 1990
Business Cycles: Real Facts and a Monetary Myth *
Finn E. KydlandProfessor of EconomicsGraduate School of Industrial AdministrationCarnegie-Mellon UniversityEdward C. PrescottAdviserResearch DepartmentFederal Reserve Bank of Minneapolis and Professor of EconomicsUniversity of MinnesotaEver since Koopmans (1947) criticized Burns and Mitchell’s (1946) book on Measuring Business Cycles
as being “measurement without theory, ” the reporting of business cycle facts has been taboo in economics. In his essay, Koopmans presents two basic criticisms of Burns and Mitchell’s study. The first is that it provides no systematic discussion of the theoretical reasons for including particular variables over others in their empirical investigation. Before variables can be selected, Koopmans argues, some notion is needed of the theory that generates the economic fluctuations. With this first criticism we completely agree: Theory is crucial in selecting which facts to report.Koopmans’ second criticism is that Burns and Mitchell’s study lacks explicit assumptions about the probability distribution of the variables. That is, their study lacks “assumptions expressing and specifying how random disturbances operate on the economy through the economic relationships between the variables” (Koopmans 1947, p. 172). What Koopmans has in mind as such relationships is clear when he concludes an overview of Burns and Mitchell’s so-called measures with this sentence: “Not a single demand or supply schedule or other equation expressing the behavior of men [i.e., people] or the technical laws of production is employed explicitly in the book, and the cases of implicit use are few and far between” (p. 163). Koopmans repeatedly stresses this need for using a structural system of equations as an organizing principle (pp. 169-70). Economists, he argues, should first hypothesize that the aggregate time series under consideration are generated by some probability model, which the economists must then estimate and test. Koopmans convinced the economics profession that to do otherwise is unscientific. On this point we strongly disagree with Koopmans: We think he did economics a grave disservice, because the reporting of facts—without assuming the data are generated by some probability model—is an important scientific activity. We see no reason for economics to be an exception.As a spectacular example of facts influencing the development of economic theory, we refer to the growth facts that came out of the empirical work of Kuznets and others. According to Solow (1970, p. 2), these facts were instrumental in the development of his own neoclassical growth model, which has since become the most important organizing structure in macroeconomics, whether the issue is one of growth or fluctuations or public finance. Loosely paraphrased, the key growth facts that Solow lists (on pp. 2-3) are
|• Real output per worker (or per worker-hour) grows at a roughly constant rate over extended time periods.|
|• The stock of real capital, crudely measured, grows at a roughly constant rate which exceeds the growth rate of labor input.|
* The authors thank the National Science Foundation for financial support.