LONG-RANGE FORECASTING From Crystal Ball to Computer J. SCOTT ARMSTRONG Wharton School University of Pennsylvania Second Edition A WILEY-INTERSCIENCE PUBLICATION JOHN WILEY & SONS, New York Chichester Brisbane Toronto Singapore Copyright © 1985 by John Wiley & Sons, Inc. All rights reserved. Published simultaneously in Canada. Reproduction or translation of any part of this work beyond that permitted by Section 107 or 108 of the 1976 United States Copyright Act without the permission of the copyright owner is unlawful. Requests for permission or further information should be addressed to the Permissions Department, John Wiley & Sons, Inc. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold with the understanding that the publisher is not engaged in rendering legal, accounting, or other professional service. If legal advice or other expert assistance is required, the services of a competent professional person should be sought. From a Declaration ofPrinciples jointly adopted by a Committee ofthe American Bar Association and a Committee ofPublishers. Library of Congress Cataloging in Publication Data: Armstrong, Jon Scott, 1937 Long-range forecasting. "A Wiley-Interscience publication." Includes indexes. 1. Forecasting. 2. Business forecasting. I. Title. H61.4.A76 1985 338.5442 85-3292 ISBN 0-471-82360-0 ISBN 0-471-82260-4 (pbk.) Printed in the United States of America 10 9 8 7 6 5 4 3 2 1 Long-Range Forecasting, Second Edition, is the most practical, read- able, and reliable book on forecast- ing in the social and management sciences. like the first edition, which went into seven printings, it's based heavily on empirical studies, uses the same humorous style that won over so many readers, and shows you how to use the latest forecasting techniques to attain higher levels of accuracy. Research on forecasting has been growing rapidly, and this new edition incor- porates it all. It's the best book to answer the questions: "Which fore- casting method is best to use in a given situation?" and "What are the advantages and disadvantages for each method?" Long-Range Forecasting, Second Edition, covers research in econom- ics, sociology, psychology, trans- port ation, education, and management, with references to medicine, weather, and technology. Long-Range Forecasting empha- sizes the highly practical tech- niques, such as judgmental forecasting, that you use most often. It explains, in clear detail: How to structure a forecasting problem The strengths and weaknesses of the available forecasting methods How to evaluate forecasting systems How to get a new forecasting method accepted by your users How managers should use fore- casts Ideas for future researchwith suggestions on the kind of research that will have the biggest payoff. (continued on back flap) (continued from front flap) This revised and expanded edition combines the results from 350 new books and articles with 700 from the first edition. New selections in- clude a discussion on the relation- ship between planning and forecasting, a synthesis of dramatic findings on the use of scenarios, a quantitative review of all empirical studies examining sophisticated ap- proaches to extrapolation, some benchmarks for assessing forecast accuracy, and guidelines for audit- ing the forecasting process in an organization. Long-Range Forecasting, Second Edition is easy and enjoyable to read, with aphorisms, poems, photos, and over 80 illustrations in- terspersed throughout the text. Re- searchers, managers, consultants, and students will appreciate author J. Scott Armstrong's guiding philos- ophy: "Who said scientific books had to be dull?" About the author J. Scott Armstrong is a Professor at the Wharton School, University of Pennsylvania. He is a founder of the International Institute of Fore- casters, Editor of the International Journal of Forecasting, and a con- tributing editor to Interfaces. Dr. Armstrong, who received his PhD from the Sloan School of Manage- ment, has worked for Eastman Kodak, Polaroid, and Xerox. He has taught at the Stockholm School of Economics, the University of Ha- waii, IMEDE (Switzerland), and Chu- lalongkorn University in Bangkok. THE NEED FOR FORECASTS Environment Industry Company Competitors' actions actions Market share E Costs Sales Results Forecasting Audit Checklist Topic Area No Yes FORECASTING METHODS 1. Forecast independent of top management? Forecast used objective methods? Structured techniques used to obtain judgments? Least expensive experts used? More than one method used to obtain forecasts? Users understand the forecasting methods? Forecasts free of judgmental revisions? Separate documents prepared for plans and forecasts'? ASSUMPTIONS AND DATA Ample budget for analysis and presentation of data? Central data bank exists? Least expensive macroeconomic forecasts used? UNCERTAINTY Upper and lower bounds provided? Quantitative analysis of previous accuracy? Forecasts prepared for alternative futures? Arguments listed against each forecast? COSTS Amount spent on forecasting reasonable? LONG-RANGE FORECASTING From Crystal Ball to Computer POSTSCRIPTS WHY ROME FELL 1 n ancient Rome (A.D. 357), the Emperor Constantius made a law forbidding "anyone to consult a soothsayer, a mathematician, or a forecaster . May curiosity to . . foretell the future be silenced for- ever." Theodosian Code, Book Title 16, 9, Section 4 up NEW YORK IS NEXT? ersons pretending to fore- r- cast the future" shall be considered disorderly under Sub- division 3, Section 901 of the Criminal Code and liable to a fine of $250and/or six months in prison. Section899, New York State Code Criminal Procedure (noted in of ASCHER & OVERHOLT 11983]) 687 To Murrl J. Anderson, my father-in-law, who helped me to accept the future esearch on forecasting has been PREFACE ED, IN growing rapidlymore so than in most other areas of the social Things are more like they are now than and management sciences. It they ever were before. seemed appropriate, then, to re- vise Long-Range Forecasting. Dwight David Eisenhower This second edition of Long- Range Forecasting improves upon the first edition in the following manner: First, it indicates where recent research has made significant con- tributions, either supporting or refuting conclusions from the first edition. (This research is inte- grated into the text and described in the updated bibliography.) Second, the updated bibliog- raphy compiles the most impor- tant research on forecasting since the first edition was published seven years ago. It includes over 350 books and papers. References to these works are easily found in the text as they are cross-refer- enced. The LAST NAMES OF THE AUTHORS have been capitalized to help the reader find the latest research. Third, the graphics have been improved. Shaded areas set off de- scriptions of empirical research. Exhibits have been redrawn for better readability. vii