|
|
Forecasting Methods: A Selective Literature Review Technological forecasting is a subset of futures research. Futures research is an umbrella term which encompasses "any activity that improves understanding about the future consequences of present developments and choices" (Amara and Salanik, 1972, p. 415). In defining forecasting, the authors offer the following progression. Forecasting is:
Technological forecasting includes "all efforts to project technological capabilities and to predict the invention and spread of technological innovations" (Ascher, 1979, p. 165). Martino (1983) states that a technological forecast includes four elements: the time of the forecast or the future date when the forecast is to be realized, the technology being forecast, the characteristics of the technology or the functional capabilities of the technology, and a statement about probability. Forecasting a technology is a difficult task "beset with hazards" (Ayers, 1969, p. 18). Some of these hazards include: "the uncertainty and unreliability of data, the complexity of 'real world' feedback interactions, the temptation of wishful or emotional thinking, the fatal attraction of ideology, [and] the dangers of forcing soft and somewhat pliable 'facts' into a preconceived pattern" ( p. 18). To offset the inherent ambiguity and uncertainty of forecasting, technological forecasters have developed a set of methodologies to assist them in their endeavor. In general, as a technology moves from the early stages of laboratory development to widespread acceptance in the marketplace, the forecasting methodologies that are most appropriate move from qualitative to quantitative techniques. Since technological forecasting is employed to predict long-term technological developments, the methods utilized are generally qualitative. Listed below is a brief description and discussion of some of the major qualitative methods and techniques which have been developed to forecast technological developments.
The Delphi procedure is designed for the systematic solicitation of expert opinion. There are three characteristics which distinguish it from interpersonal group interaction: anonymity, iteration with controlled feedback, and statistical group response (Martino, 1983). While many variations of the technique have been offered since it was originally developed at the Rand Corporation in 1950's (see Martino, 1983, pp. 20-22), the conventional Delphi study proceeds as follows. A questionnaire designed by a monitor team is sent to a select group of experts. After the responses are summarized, the results are sent back to the respondents who have the opportunity to re-evaluate their original answers, based upon the responses of the group. By incorporating a second and sometimes third round of questionnaires, the respondents have the opportunity to defend their original answers or change their position to agree with the majority of respondents. The Delphi technique, therefore, is a method of obtaining what could be considered an intuitive consensus of group expert opinions. The accuracy of the forecast produced is limited by the quality of opinions provided by the experts, and it should be noted that some authors (such as, Challis and Wills, 1970 and Wise, 1976) have questioned the accuracy of the opinions of specialists. For an example, see The Virtual Environment -- The Future? at http://www.geocities.com/ResearchTriangle/4681/
A forecast can be generated by "observing a change through time in the character of something and projecting or extrapolating that change into the future" (Cornish, 1977, p. 108). In making such a forecast, the focus is on the long-term trend, so short-term fluctuations are disregarded. Trend extrapolations require that the forecaster have an understanding of the factors which contributed to change in the past, and possess confidence in the notion that these factors will continue to influence developments in a similar fashion in the future (Schwarz, Svedin, Wittrock, 1982, p. 20). One commonly employed approach to trend extrapolation involves the use of growth curves (Cornish, 1977, pp. 110-111). Growth curves are loosely based upon the notion that the growth of a technology can be charted in the same way organic growth can be charted. For example, the growth in height and weight of an individual can be charted, and will commonly display a pattern which indicates a leveling off around early adulthood. It is believed that the growth pattern of a technology can also be plotted and charted in a similar fashion. As an illustration, Martino (1983) describes how this particular technique can be utilized in charting and forecasting the growth in, and leveling off, of the number of cable television subscribers. Regarding the accuracy of trend extrapolation as a forecasting technique, Ascher (1978) questions its "objectivity and reliability" (p. 183). Schnaars (1989) goes even further and admonishes forecasters to discount trend extrapolations. He notes that trends and patterns have no life of their own and are susceptible to sudden changes, and that focusing on trends alone "is often a search for the will-o'-the wisp" (p. 152). As an example of a misuse of trend extrapolation, he notes the actions taken by American electronics firms with regard to television manufacturing. Through the 1950s and the 1960s, television sets steadily grew larger. As American firms continued to make large, cabinet-based systems, Japanese firms began to concentrate on making portable sets. While the American firms acted on the belief that the existing trend toward larger sets would continue, the actual trend within the marketplace shifted toward a greater variability in size.
The use of analogy in forecasting involves a "systematic comparison of the technology to be forecast with some earlier technology that is believed to have been similar in all or most important respects" (Martino, 1983, p. 39). Forecasting by analogy is one of the simpler and more common ways to forecast the growth of a new technology, though as a method its accuracy has been questioned on several accounts. Schnaars (1989) notes that the method has limited predictive value as "what happened before in an industry often blinds those already in the industry to developments that come from outside" (p. 153). In his study of home video forecasts, Klopfenstein (1985) found that many erroneous forecasts of videodisc players sales were based on comparison with the previous introduction of color television. He asserts that historical analogy can serve as a useful guide to forecasting a new technology, but great care must be taken in making the comparison. Martino (1983) asserts that when drawing an analogy, consideration must be given to the numerous dimensions which are known to have an effect on technological change (see Martino, 1983, pp. 40-49 for a discussion of these dimensions). The real challenge facing a forecaster, therefore, is the task of identifying a technological innovation which will truly serve as an accurate historical precedent upon which to base a forecast by analogy.
Each of the above forecasting methods has its own advantages and disadvantages. Therefore, in many cases, it is helpful to combine several methods and forecasts into one. Martino (1983) notes that scenario construction is an effective method for combining forecasts and forecasting methodologies into a holistic composite. Cornish (1977) describes a scenario in simple terms: "it is simply a series of events that we imagine happening in the future." In other words, scenario writing is "making up stories about the future" (p. 11). Schwarz, Svedin, and Wittrock (1982) note that the term "scenario" has numerous meanings. It can be used as a description for "a hypothetical, likely or unlikely, development or situation; a development which is described as caused to some extent by the actions and reactions of various actors: a desirable or nondesireable development or situation" (p. 28). Kahn and Wiener (1967) assert that it is a method which can be employed to focus attention on causal processes and crucial decision points. Martino (1983) states that scenarios serve three basic purposes: 1) to display the interactions among several trends and events in order to provide a holistic picture of the future; 2) to help check the internal consistency of the set of forecasts on which they are based; and 3) to depict a future situation in a way readily understandable by the non-specialist in the subject area (p. 148). While noting the unreliability of forecasting methods, Schnaars (1989) is a strong advocate of the use of scenarios. He notes that they do not pretend to predict the future but rather present a set of possible futures. Godet (1983) notes that since the value of a forecast is dependent upon the underlying assumptions, and that quite often several sets of assumptions can be offered upon which several scenarios can be constructed, no forecast should be published "without giving an indication of the estimated probability of the corresponding scenario" (p. 190). By accounting for a range of possibilities, scenarios can be distinguished from the other methods listed above. They do not generate or present the same degree of specificity, and have even been described as an "alternative to forecasting" (Schnaars, 1989).
Amara, R. and Salanik, G. (1972). Forecasting: From conjectural art toward science. Technological Forecasting and Social Change , 3 (3), 415-426. Ascher, W. (1978). Forecasting: An appraisal for policymakers and planners . Baltimore: Johns Hopkins University Press. Ascher, W. (1979). Forecasting: An appraisal for policymakers and planners (rev. ed.). Baltimore: Johns Hopkins University Press. Ayers, R. (1969). Technological forecasting and long-range planning. New York: McGraw-Hill Book Company. Challis, A. & Wills, G. (1970). Technological forecasting. In D. Ashton & L. Simister (Eds.) The role of forecasting in corporate planning, pp. 100- 124. London: Staples Press. Cornish, E. (1977). The study of the future. Washington, D.C.: World Future Society. Klopfenstein, B. (1985). Forecasting the market for home video players: A retrospective analysis. Unpublished doctoral dissertation. The Ohio State University. 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. Klopfenstein, B. (1989). Problems and potential of forecasting the adoption of new media. In J. Salvaggio & J. Bryant (Eds.) Media use in the Information Age: Emerging patterns of adoption and consumer use (pp. 21-41). Hillsdale, NJ: Lawrence Erlbaum Associates. Martino, J. (1983). Technological forecasting for decision making. New York: Elsevier Science Publishing Company. Schnaars, S. (1989). Megamistakes: Forecasting and the myth of rapid technological change. New York: The Free Press. Schwarz, B., Svedin, U. & Wittrock, B. (1982). Methods in future studies. Boulder, Colorado: Westview Press. Wise, G. (1976). The accuracy of technological forecasts: 1890- 1940. Futures, 8 (5), 411-419. See also FORECASTING ASSESSMENT Material See also the annotated bibliography of books concerning the social impact of new media.
|
