Forecasting Assessment Literature

Dr. David Donnelly

     There is an important though relatively small body of literature which examines past efforts to forecast new technologies. Klopfenstein (1986) suggests that the size of this body of literature is perhaps due to the fact that technological forecasting is a relatively new discipline and, therefore, few formal forecasts were undertaken before the 1960's. In other words, not enough time has passed to evaluate a large portion of the available forecasts. Moreover, one might add that forecasting is a field which is inclined to look ahead, and not back. Schnaars (1989) rather cynically notes that "previous forecasts, it seems, are best forgotten or left to the criticism of the pesky outsider.... it is as if forecasters are afraid that by looking back they will be turned into the business equivalent of biblical pillars of salt" (p. 2).

     Some authors have analyzed in detail the forecasts of a specific technological innovation. Pool, et. al (1977) examined forecasts concerning the social effects of telephones from 1876 through 1940. In examining the earliest of these forecasts, the authors found that the small group of individuals responsible for the initial development of the telephone were significantly more accurate in their forecasts than outside commentators. In formulating an explanation for this disparity, the authors note that these pioneers "had the inventions, a vision of how the inventions could be used, and they controlled the business that implemented those visions" (p. 129). In other words, the authors surmise that because the early telephone pioneers were both inventors and capitalists, they understood both the technical components as well as the economic and market conditions.

     Based upon the findings of this specific case study of the telephone, Pool (1983) has offered the following generalization. In successful technological forecasts and assessments, "market and technical analyses must be brought to bear simultaneously. Alone either of them fails; together they can produce some very prescient forecasts" (p. 1). While admitting the apparent banality of this conclusion, Pool points out that a large proportion of technology assessments and forecasts make little use of market analysis.

     Klopfenstein (1986) analyzed twenty-nine video player market forecasts made during the period from 1968 through 1984. He found that amongst the forecasts, there was considerable agreement that the Video Cassette Recorder (VCR) was destined to remain a high priced, luxury item, while the VideoDisc Player (VDP) would become a marketing success. Klopfenstein offers two explanations for this overlap. Forecasts were based upon similar assumptions, some of which were erroneous. Secondly, there was a bandwagon effect created; new forecasts cited earlier forecasts. Klopfenstein summarizes the limitations of the forecasts as follows. Most of the forecasts were not critical; consumer adoption was assumed. Moreover, many of the forecasts were "enamored with the technology." In general, the forecasts did not adequately consider media or product substitution effects. Oftentimes, absolute numbers were forecast rather than a range of possible outcomes. And lastly, the forecasters failed to adequately address the social factors which surround the adoption process.

     There are several important works which provide a broader, more general survey of past forecasting attempts and encompass numerous innovations and subject areas. Schnaars (1989) examines a large number of diverse forecasts which appeared primarily in the popular business press. To a lesser extent, his study also examined forecasts which appeared in trade journals and popular books on the future, while consciously avoiding "sensational sources that published extreme forecasts" (p. 5). In evaluating the accuracy of the forecasts studied, Schnaars asserts that the criteria are "strict." "Simply put, forecasters were held to their own words" (p. 5). "Vaguely stated" forecasts and forecasts "cast in contingencies" were not included in the study. Based upon consistent patterns he identifies among both the successful and unsuccessful forecasts, Schnaars offers three general guidelines which he hopes can help a forecaster avoid the mistakes of the past and improve the overall accuracy of growth market forecasting.

     First, Schnaars advises that forecasters "avoid technological wonder." Many erroneous forecasts have failed because they were enamored with the underlying technology, while they ignored the fundamental market considerations. "It is the applications of technology that move markets, not the technical aspects of the products" (p. 144). This notion leads to Schnaars second guideline: a forecaster or an analyst of a forecast should focus on fundamental marketing issues. Many forecasts have failed because they misread the market they were intended to serve. Lastly, the most important question to ask of a growth market relates to cost-benefit analysis; "whether or not the product upon which it is based provides customers with something special at a price that both the customer and manufacturer will accept" (p. 147).

     Based upon his examination of past forecasts and the findings of other forecasters, Schnaars offers the following specific methodological guidelines intended to improve the overall accuracy of forecasting. He advises forecasters to discount extrapolations. He notes that forecasters should not simply add or subtract existing trends and assume that the future will be a logical progression from the present, and they should "avoid extrapolating the issues of the day" by assuming that the future will be driven by concerns of the present. He asserts that forecasters should also downplay historical precedent: "what happened before in an industry often blinds those already in the industry to developments that come from the outside" (p. 153). In growth marketing forecasts, it is essential that forecasters attend to developments in other industries and to distinguish fads from true growth markets. Schnaars suggests that forecasters try to employ multiple methods in predicting or evaluating a growth market. Lastly, he adds that "the most important advice for improving the accuracy of growth market forecasting is to challenge the assumptions that underlie forecasts" (p. 160). While Schnaars asserts that these guidelines can help avoid many of the forecasting mistakes made in the past, he reminds readers that they cannot guarantee success. He goes on to offer several strategic alternatives to forecasting (see pp. 161-185).

     Wise (1976) analyzed 1,556 predictions publicly made by Americans from 1890 through 1940. He examined two types of predictions: predictions of technological changes that were to occur and predictions of the social, economic and political effects which were expected to accompany these technological changes. His study is an attempt to identify and measure the "correlates of predictive accuracy." Wise found that less than half of the predictions examined have been fulfilled or are in the process of being fulfilled. His findings indicate that the accuracy of the predictions is weakly related to technical expertise as there is little difference in accuracy between expert and nonexpert predictions. Moreover, predictions of the continuation of the status quo are not significantly more accurate than predictions of change, and predictions of the effects of technology are significantly less accurate than predictions of technological change.

     Ascher (1979) compiled over 165 past technological forecasts in six subject areas: population forecasts, economic forecasts, energy forecasts, transportation forecasts and non-ferrous metals forecasts. Ascher asserts that accuracy is the most important criteria upon which to judge and appraise a forecast, and after a review of the relative accuracy of these forecasts, he provides some general comments regarding the correlates of forecasting accuracy. He concludes that the most important determinants of accuracy are the core assumptions underlying a forecast. These assumptions "represent the forecaster's basic outlook on the context within which the specific forecasted trend develops" (p. 199). Other findings include: the notion that the time horizon of the forecast is significant, "the more distant the forecast target date, the less accurate the forecast is expected to be"; different institutional sites generate different resulting forecasts; methodology choice is not generally linked with accuracy; the least accurate forecasts tend to rest on out-dated information, or have been affected by events unanticipated by the forecaster.

     Although Martino (1983) argues that accuracy is not the proper criterion upon which to evaluate a forecast, he admits that "until better records of decision making are kept and made available to scholars, it may be the best we can manage" (p. 226). In reviewing the accuracy of past forecasts, he identifies three major causes of errors. First, he notes that many forecasts have missed their mark because they have underestimated or omitted important environmental factors that could have an impact on the technology forecast. In other words, the forecasters have placed too much emphasis on the technology and have not adequately identified and incorporated major environmental influences, such as technological, political, cultural or social dimensions. Secondly, there are personal factors, factors internal to the forecaster, that affect and, in some cases, invalidate a forecast. Martino asserts that if a forecaster is aware of their own human frailties, and if they take into account the possibility of being afflicted by them, they have a much better chance of avoiding them and can thereby increase the usefulness of their forecasts. Lastly, he concurs with Ascher (1979) and Schnaars (1989) concerning the importance of core assumptions in forecasting and asserts that they may exert a greater influence on the outcome than the actual methodology selected. He suggests that forecasters "recognize that their core assumptions will dominate the outcome of a forecast" (p. 246). They should, therefore, try to identify and verify these assumptions through discussions with other experts in order to improve their accuracy.

     From a decisionmaker's point of view, Martino describes a procedure for evaluating forecasts that centers on how well the forecast employs information about the past and about the nature of change in the specific subject area. This evaluation procedure, which Martino calls "the Interrogation Model" involves four steps: interrogation for need; interrogation for underlying cause; interrogation for relevance; and interrogation for reliability (see pp. 250-260).

     Moyer (1984) offers his own explanations for errors in long-range forecasts. His "reasons for error" include the following. Forecasters may err when they analyze and measure only surface factors while ignoring important underlying forces. Long-range predictions often do not pay enough attention to substitution effects. There is a wide range of expenditure-substitution possibilities, and the more "advanced a society economically, technologically and in educational attainment," the greater the number of substitution possibilities because "these developments increase the number of branches in each person's lifetime decision tree" (p. 69). Moyer also asserts the importance of assumptions in forecasting. He points out the problematic tendency for several independent forecasters to rely on the same assumptions, or to employ outmoded assumptions, even when current data are available, a phenomenon Ascher has dubbed "assumption drag."

     Moyer also points out that time factors help contribute to errors in forecasting. For example, forecasters sometimes do not account for time lags such as when a fully developed technology may be delayed due to political, economic or social forces. Moreover, the process of phasing out existing facilities is sometimes more time-consuming than anticipated. Another important time variable concerns the forecast's time horizon: the longer the time period between the date the forecast is made and the target date, the greater the possibility of change in major trend-determining factors. Therefore, "the length of the forecast and their degree of error are almost always directly related" (p. 69). Because forecasts "are summations of predicted values for several constituents of the phenomenon" that is the subject of the forecast, forecasting accuracy depends both upon the proper selection of appropriate components and upon "accurate prediction of the components' future values" (p. 69). Finally, Moyer suggests that a bias might contribute to errors in forecasting.

     While Moyer points out that the inaccuracy of past long-range forecasts might lead one to question their value, he asserts that they remain an integral part of planning. He offers analysts several ways to account for forecasting errors in their planning, and offers forecasters suggestions on how to improve their level of accuracy. First, Moyer suggests that bias be recognized; for once it is recognized, adjustments can be made. Second, checking the validity of underlying assumptions concerning external factors which relate to the phenomenon being forecast may help to reduce forecasting errors. Once the underlying assumptions have been identified, they must be monitored continually. Lastly, Moyer suggests that forecasters can reduce errors by remembering that a "phenomenon moving in one direction sets in motion forces that will modify its course" (p. 71). Recognizing the strong tendency of phenomena to move toward equilibrium or homeostasis can lead to more accurate forecasts.

     Godet (1983) identifies three factors which contribute to errors in forecasting: "inaccurate data coupled with unstable models; lack of a global, qualitative approach; explanation of the future in terms of the past" (p. 182). He calls for the increased use of prospective analysis in forecasting. Prospective analysis, which rests upon the generation of scenarios, incorporates qualitative parameters and accepts that there is a multiplicity of futures at any one moment, and that the "actual future will be the outcome of the interplay between these various protagonists in a given situation and their respective intentions" (p. 183).

     Upon examining the literature which assesses past forecasting efforts, perhaps the most significant area of disagreement that appears concerns the value and appropriateness of accuracy as a criterion for forecast evaluation. Some writers, for example, Ascher (1979) and Schnaars (1989), argue that the value of a forecast is dependent upon its accuracy. Others disagree. Martino (1983) asserts that "whether or not a forecast comes true is hardly the proper criterion by which it should be judged" (p. 226). Klopfenstein (1989) points out that "the utility of a forecast may be more important than its absolute accuracy" (p. 37). Also Godet (1983) notes that

given the many different uses of forecasts, general rules for deciding what leads to a 'good' forecast are not easily constructed; being proven right by events may or may not be relevant; stimulating imagination or making a political point may be useful criteria against which to judge some exercise (p. 187).

     In examining this body of work, one can also readily identify several major points of overlap and agreement. For example, several writers have noted that the relationship between various forecasts of the same phenomenon or innovation are important. Klopfenstein (1986) notes that often "new forecasts cited earlier forecasts creating a 'bandwagon effect'" (p. 18). Also, the notion of an author's bias influencing the shape of a forecast was commonly noted (see for example, Moyer, 1984 and Martino, 1983).

     Numerous writers agree that many past forecasting errors are due to an overemphasis on the technology. Pool (1983) asserts that an accurate forecast must involve an equal consideration of economic and market forces which will help to counterbalance the tendency to overemphasize technical aspects. Schnaars (1989) concurs and notes that many forecasts err because "the forecasters fall in love with the technology...and ignore the market the technology is supposed to serve" (p. 9).

BIBLIOGRAPHY

Ascher, W. (1979). Forecasting: An appraisal for policymakers and planners (rev. ed.). Baltimore: Johns Hopkins University Press.

Godet, M. (1983). Reducing the blunders in forecasting. Futures , 15 (3), 181-192.

Klopfenstein, B. (1986). Forecasting the market for new communication technology: The home video player experience. Paper presented at the annual meeting of the Broadcast Education Association.

Martino, J. (1983). Technological forecasting for decision making. New York: Elsevier Science Publishing Company.

Moyer, R. (1984). The futility of forecasting. Long Range Planning, 17 (1), 65-72.

Pool, I., et. al. (1977). Forecasting and hindsight: The case of the telephone. In I. Pool (Ed.) The social impact of the telephone (pp. 127-159). Cambridge, MA: The MIT Press.

Pool, I. (1983). Forecasting the telephone: A retrospective technology assessment of the telephone. Norwood, NJ: Ablex.

Schnaars, S. (1989). Megamistakes: Forecasting and the myth of rapid technological change. New York: The Free Press.

Wise, G. (1976). The accuracy of technological forecasts: 1890- 1940. Futures, 8 (5), 411-419.

(See also Final thoughts on the future and forecasting.)