Atlantic Economic Journal, Vol. 7, No. 2 (1979), p. 97
Review of J. Scott Armstrong's Long Range Forecasting (NY: John Wiley, 1978), xviii + 612 pages
Reviewed by Gordon Welty
Wright State University
Dayton, OH 45435 USA
This comprehensive, readable and witty study of long-range forecasting is based on Armstrong's extensive consulting experiences in business and government over the past decade, plus his review of more than l,300 publications in the field.
Part I presents an overview of the systems approach and the strategies of securing commitment of forecasting by the client. Reminiscent of the later work of K. Lewin, Armstrong stresses the need for client "participation" in the various stages of the systems approach. In support of this prescription, he cites the literature on worker participation in job-redesign studies. However, much of this literature confounds the effects of job-redesign itself (which may implicate participation in one direction) with the effects of concurrent layoffs of recalcitrant workers (which implicates it in quite another direction). The only clear-cut conclusions that can be drawn from these studies is that job-redesign effectively reduces variable costs.
Part II discusses disparate forecasting methods ranging from the Delphi technique to spectral analysis. Some of his discussion, e.g., on the Delphi technique, is clearly definitive. Some of his discussion however may slight the topic, e.g., his treatment of the literature on cycles. He refers only to the obscure Dewey and Dalkin. No mention, let alone assessment is made of the contributions to forecasting of William S. Jevons, Pareto, and Schumpeter, to mention some early writers.
Part III presents a schema for the evaluation of the forecasting model. This has four stages where one ascertains whether assumptions are reasonable, the model is logically consistent (deducible), it generates reliable output data, and these predictions are accurate. This part of the book has intriguing resonances with the sound methodological criteria advanced by Jevons [Principles of Science, 1877, p. 511]. Armstrong has an evident familiarity with the recent philosophical literature of economics (e.g. F. Machlup; M. Friedman), has incorporated its rigor into his own work, and then moved beyond the debates themselves.
Part IV contains a chapter on the cost-benefit assessment of forecasting methods, followed by a chapter on evaluating forecasting accuracy. Part V discusses trends in the use of forecasting methods as well as research on forecasting methods. Trends are from Subjective to Objective methods, from Naive to Causal methods, and towards more appropriate applications of Segmentation and Econometric approaches.
The book concludes with a list of more than 160 publications which Armstrong recommends the reader not read, and an annotated list of almost 800 references which are rated by Armstrong in terms of his judgment of the value of each publication for further study of forecasting methods; all these are in the English language. This means that the Continental researchers, e.g., Bertrand de Jouvenal, Jean Fourastie, et al are not represented.
This is a very handsome volume; I found only one typographical error (p. 240). It will be a welcome addition to any applied research library.