Patrick Grady
A History of Macroeconomic Modelling
by Ronald G. Bodkin, Lawrence R. Klein, and Kanta Marwah.
Aldershot, England: Edward Elgar Publishing Limited, 1991.
Pp. xvii, 573. Index.
ISBN 1 85278 369 9

Canadian Journal of Economics, XXV, No.1 (February 1992),pp.244-248.

Macro-modellers will find this book to be fascinating reading for its engrossing account of the development of macroeconometric modelling in the 1930s through the 1970s. They will also take heart from the book's vigourous defence of the scientific validity of model-building against its academic critics. An important theme of the book, which is amply documented, is that of the continuous progress that has been made in macroeconometric modelling since its inception a half century ago.

While the book was published under the names of three authors, there were other distinguished macro-modellers as well that contributed chapters on macroeconomic modelling in particular countries: Anton P. Barten on the Netherlands; James Ball and Sean Holly on the United Kingdom; Raymond Courbis on France; Kazuo Sato on Japan; and Abel Beltran-del-Rio on Latin America. A chapter on regional econometric models is provided by Roger Bolton and one on Project Link and multi-country modelling by Bert G. Hickman.

Bodkin, Klein and Marwah themselves contribute an introduction on the origins of macroeconometric models, chapters on the experience in the United States, Canada and India, and a concluding chapter on the lessons from and prospects for macroeconometric modelling as well as a history of computation in econometrics.

One of the book's great strengths is that its authors, particularly Lawrence Klein who may be considered the father of modern macro-modelling, have been key participants themselves in the history of macroeconometric model-building and have first hand knowledge of the developments whence they write. This is also the book's weakness as participants can not be expected to exhibit the same objectivity and detachment required of good historians. It should thus not be surprising that the authors have occasional minor lapses into what can only be described as self promotion. But these peccadillos can be excused in view of depth of knowledge the authors bring to their task.

The section of the book dealing with the origins of model-building is one of the most interesting, particularly the discussion of the controversy between Keynes and Tinbergen over the representation of Keynesian theory in the form of empirically estimated mathematical relationships. Given Keynes' anti-modelling views, it is certainly ironic that his theory has provided the main impetus behind the development of modelling.

The treatment of the experience in the United States is the most complete in the book, covering a three chapters and accounting for almost 100 pages. The pride of place of the Klein-Goldberger model of 1955 in the family tree of U.S. models is appropriately emphasized. The importance of the Brookings model for its size, scope, complexity, team approach, advances in the use of the computer, and influence on modelling in the 1960s is well documented. The flowering of modelling in the 1970s in government agencies such as the Fed and BEA and private consulting firms such as DRI and Wharton is well covered.

Five important themes of the book that are behind the remarkable progress of modelling and are highlighted in the part of the book on American econometric models are: the improvements in computer technology; the team approach to modelling; the development of an institutional framework largely independent of the original model builders; detailed and large-scale modelling; and increasing theoretical sophistication.

Another interesting theme that comes out in the country chapters is the importance of ideology in determining the approach pursued by model-builders. The Keynesian nature of British modelling in contrast with the planning orientation of modelling in France are good examples of this tendency. The development of models to support particular points of view such as the new Cambridge model and the St. Louis model are others.

Among the country chapters, that by Ball and Holly on the United Kingdom stands out for its lucidity. It avoids the excessive description of equations that weigh down some of the other chapters. Instead it presents the key conceptual and institutional developments and issues in a very readable style. It is also current up to the late 1980s, something that unfortunately can not be said for all of the other chapters.

The Canadian chapter by Bodkin and Marwah will be of most interest to Canadian modellers and will spark much controversy in the close-knit Canadian modelling fraternity about the emphasis and interpretation put on particular developments. But there should be no dispute about the accuracy of the facts presented as the authors have been very careful to verify them with leading Canadian modellers.

Every model-builder will, of course, have his own bone to pick. Mine concerns the discussion of the treatment of model-building in the government of Canada. The sections of the chapter dealing with the early years of modelling between 1947 and 1963 in the old departments of Reconstruction and of Trade and Commerce are very interesting and informative. The important contributions of Lawrence Klein, Merrit Brown and Sid May in building Models I through XIV are well chronicled. However, the treatment of modelling in the Department of Finance after the appearance of Model XVI and the departure of John Kuiper in the summer of 1970 could use some further elaboration.

A discussion of the history of modelling in the Department of Finance in the 1970s can in my view begin with the appointment of Simon Reisman as DM in 1970. He brought in a new crew of economists including most notably William Hood as ADM, Cyril Hodgins as Director of Economic Analysis, and Bob Baguley as Chief of Forecasting. It was decided that the old Model XVI, the linear descendent of the Trade and Commerce models, which had never been used for forecasting, was no longer satisfactory for policy analysis. Since there was no desire to invest the resources to bring the model up to scratch and reestimate it, QFM from the University of Toronto was adapted and used for policy analysis.

This approach of using someone else's model worked for a while but then further dissatisfaction set in. A Committee was established with Michael Walker as a consultant to look into alternatives. The result of this was the adoption of a modified version of RDX2 as the principal vehicle for policy analysis in the Department. This work was done in the Fiscal Policy Division under the leadership of John Sargent. Subsequent versions were developed. A modified version of RDX2 redbook was the Department of Finance's entry in a comparative models seminar sponsored by the Fiscal Policy Division in April 1979 at which all of the major modelling groups participated.

Meanwhile, work had been underway in the Economic Analysis Division of the Department of Finance on the development of a forecasting model. The perceived need for a model capable of both forecasting and policy analysis resulted in the establishment of a Committee under the Chairmanship of Phillip Smith to develop a model for the Fiscal Policy and Economic Analysis Branch. The end result was the Quarterly Forecasting and Simulation (QFS) model put together by Krishna Sahay which was first used to produce the official government forecast in the fall of 1982. The QFS represented a bottom-up approach to model-building collecting together the forecasting equations of the sectoral specialists.

A new Economic Forecasting Division was established in 1983 under the direction of Kevin Lynch and took over the QFS model for forecasting. Another modelling project was soon launched to develop another quarterly forecasting model that had a more rigourous top-down theoretical structure and would have desirable medium and longer run growth properties. One aspect of this project was to commission a detailed study of the structure of existing Canadian models as an aid in specifying the structure of the new model. This study was subsequently made public by the Economic Forecasting Division.

The Canadian Economic and Forecasting Model (CEFM) was completed under the direction of Ernie Stokes in the summer of 1986 and first used for forecasting in the September 1986 quarterly forecasting round by a new consolidated Economic Analysis and Forecasting Division. It has been used ever since to produce the government's official budget forecast and the Department of Finance's record in using it for forecasting was analyzed in the March 1991 Quarterly Economic Review. Thus ends my brief account of the history of macroeconomic model-building in the Department of Finance in the 1970s and 1980s.

The importance of model comparisions in the history of modelling is demonstrated in the book with some examples: the comparative simulation studies of U.S. modellers published in the 1974 and 1975 International Economic Review (p.120); the establishment of the Macroeconomic Modelling Bureau in the United Kingdom in 1983 (pp.212-213); and the comparative simulation studies of Canadian modellers published in the May 1979 issue of the Canadian Journal of Economics (p.267) and in the Bank of Canada Technical Report No. 38 (p.268).

Two of the four examples of model comparisions found in the book are Canadian. A cooperative approach with periodic conferences to compare model responses has been a distinctive feature of Canadian modelling. It has been facilitated by close personal relations among modellers and government and/or Bank of Canada sponsorship. Some convergence of model structure and response properties has been an important result.

My greatest regret about the book is that it peters out after 1980. While I can appreciate publication lags, I do not see why the history could not have been brought a little closer to the present in a book published in 1991. Practitioners are most interested in the history of macroeconometric modelling for what it tells about the state of the art and any lessons for their own current work.

The section on prospects for macroeconometric modelling is overly general. If the history had been a little more current, the authors could have been more specific in commenting on future prospects. Much of what they call future prospects could be more accurately described as work in progress. An important gap is a thorough treatment of the implications of the microcomputer revolution for future macroeconomic modelling, although this is touched upon in some of the country chapters. On a more positive note, a particularly inspiring section of the chapter on future prospects contains the authors' rebuttal of the critics of macroeconomic modelling.

This book will long be valued by macroeconomic model- builders as a indispensable reference source on the development of their craft. It eloquently and thoroughly testifies to the dramatic progress that has taken place over the last fifty years.