vol. 58, no. 4 (April), 1992, pp. 1129-131           

 

A History of Macroeconometric Model-Building 
by Ronald G. Bodkin, Lawrence R. Klein and Kanta Marwah 
Brookfield, VT: Edward Elgar Publishing Co., 1991, pp. xvii, 573 

In today's economics profession, interest in the history of the discipline and interest in macroeconometric
model-building rarely occur in combination. This volume, then, which was produced by economists who have built some of those models and made some of that history, is likely to attract attention. 
       The three author/editors contributed a two-chapter "Introduction" and a three-chapter "Summing Up," as well as a chapter-per-decade treatment of macroeconometric model-building in the United States during the 1950s, 1960s and 1970s plus individual chapters on macroeconometric modelling in Canada and of India. All this material accounts for ten of the seventeen chapters and about half of the book. Model-building in other countries and the macro-modelling of other economies comprise five of the remaining seven chapters, which are contributed by six authors: Anton B. Barten (Netherlands), James Ball and Sean Holly (United Kingdom), Raymond Courbis (France), Kazuo Sato (Japan), and Abel Beltran-del-Rio (Latin America). And finally, Roger Bolton and Bert G. Hickman contributed chapters on
regional econometric models and multi-country modelling, respectively. Only two of the chapters (by Barten and Bolton) have previously appeared in print. 
        The introductory chapter attempts to root the parametric estimation of macroeconomic systems in the general equilibrium theories of Walras and Pareto. But apart from the recognition that "everything depends upon everything else" both in pure theory and in parametric estimation, the actual lineage from nineteenth-century general equilibrium theory to twentieth-century macroeconometric model-building is weak. The historical sketch begins to gain in credibility with the account of the contributions of Ragnar Frisch and Michal Kalecki in the early 1930s. Central to the history, of course, is John Maynard Keynes, who, it can be noted, felt a much stronger kinship to Marshall than to Walras. 
        Detailed coverage of the model-builders and their models is not easily presented in readable prose. The coverage begins with a history of people and their ideas (Tinbergen in Holland followed by Klein and Goldberger in the United States) but turns quickly into a history of institutions and their computers (Social Science Research Council, Bureau of Economic Analysis, Data Resources, Inc., and Wharton Econometric Forecasting Associates). The first four of the five trends in the development of macroeconometric model-building as summarized by the authors (p. 119) help to account for the decreasing significance of individual macroeconometricians: (1) improvements in computer technology, (2) increasing use of the team approach to econometric modelling, (3) development of institutional frameworks largely independent of model originators, (4) increasingly detailed and large-scale modelling, and (5) increasing theoretical sophistication. Gains in sophistication came from using longer and more complex lag schemes, splicing macroeconometric models with input-output models, and not linearizing non-linear relationships (p. 88). 
        The authors refer to a genealogy of macroeconomic models, in which "no other model has left such a vast legacy of style and flavour as the Klein-Goldberger Model" (p. 57). Discussion of the direct and indirect relationships of subsequent models to K-G would have been greatly facilitated by a family tree, but none was provided. Also, description of the various models follows a common format in which the number of behavioral equations and identities, the number of endogenous and exogenous variables, the data base identified by frequency and span of years, and the peak value and time profile of the dynamic multiplier are reported. This information could usefully have been summarized in tabular form. 
        Four models of the 1970s are identified as exceptions to the general trend in macroeconometric modelling in that they are more individualistic than institutionalistic. They reflect the "vision" of the principal model-builder. But except for the St. Louis model, which envisions a stable private economy with markets generally clearing and money having no long-run effect on real variables, these individualistic models are not particularly visionary in the sense of Schumpeter. The Hickman-Coen Model features interrelated factor demands; Fair's Forecasting Model couples his "vision of the economic process" with his "philosophy of forecasting" in which unconditional short-run forecasts are considered largely—or at least ideally—a mechanical process. The Liu-Hwa Model uses monthly rather than annual data in order to give play to the detailed lag structure in behavioral relationships. 
        All too frequently, claims, arguments, and counter-arguments remind the reader of how close Klein is to his subject matter. This history of econometric modelling is not a disinterested history. It is claimed that Klein scooped A. W. Phillips by three years as evidenced by the dynamic solution of the 1955 Klein-Goldberger Model, in which price stability and full employment are in conflict (p. 62). The authors defend, on the basis of forecast accuracy as judged by an imagined "disinterested party," the use of large, detailed models in the "Main Stream" tradition against models in the New Classical tradition, time-series analysis, and consensus surveys (pp. 528-9). They report on a 1972 challenge by Robert Basmann, in which the Brookings Model was criticized as a fuzzy theoretical structure whose hypotheses were tested with defective data, and a "spirited defense" of Brookings by Fromm and Klein, in which Basmann is accused of being pedantic and nihilistic (p. 109-10). According to the book's preface, this history of econometric modelling benefits from the first-hand information of the early participants. Unfortunately, the benefits of first-handedness are partially offset by losses in objectivity and perspective. 
        In addition to the dismissive treatment of Basmann's early criticism and the mention without comment of more recent and more general criticisms by David Hendry and by Ed Leamer, the authors deal obtusely with New Classicism and the rational-expectations revolution. "[F]ads come and go, but many ideas that looked extremely promising when they were first introduced have turned out to be a flash in the pan (quite unimportant) or else a modest addition to the macroeconometric model-builder's intellectual capital" (p. 527). The authors do not mention any specific fad, here, claiming that it would be "ungracious to cite particular examples," but the reader can easily make the connection. The rational expectations hypothesis and the related Lucas critique are characterized elsewhere in the volume as a contrived argument, not terribly important empirically, and not rational (p. 552 and p. 556, n. 17). But the rational expectations critique has, according to the authors, "raised the whole question of the optimal incorporation of lags into the modelling process" (p. 534, n. 8). 
        More than a decade ago, David Hendry [1] identified the three golden rules of econometrics: Test, test, and test. The increasing acceptance of these rules as both the hallmark of economic science and a pre-requisite for publication has almost completely crowded out the caveats. The authors do not warn the reader of the many instances in which tests for statistical significance are passed off as tests for economic significance. Nor do they question the significance of significance tests when performed not on sample data but on population data—as is frequently done in macroeconometrics when the object of study is not market economies in general but, say, the U.S. economy in the postwar era. The authors fail to acknowledge that any significance test which cannot be interpreted straightforwardly as quantifying the probability of a sampling error by analogy to the classical urn model has no meaningful alternative interpretation. 
        Although the book ends with an optimistic chapter on the prospects for macroeconometric modelling, the reader may feel that Klein and his colleagues are, in effect, closing the book on the era of growth for large-scale macroeconometric model-building.

                                                                                Roger W. Garrison
                                                                                 Auburn University
Reference

1. Hendry, David F., "Econometrics—Alchemy or Science?" Economica, vol. 47, no. 188 (November), 1980, p. 403