Document created: 26 September 03
Air University Review, May-June 1974

Operations Research and 
Technological Forecasting

Dr. Roy K. Frick

Technological forecasting is a relatively new field and is receiving increasingly more attention as a necessary planning tool for management. It is a field that represents a growth area for operations research. Conversely, the field of operations research can benefit from application of some of the methods now being used in technological forecasting.

development of technological forecasting

Knowledge of the future has always been of interest to civilized man, and numerous efforts by noted persons throughout history to predict the future could be cited. Many historical instances of prediction or forecasting have a technological flavor. Included in this category would be science fiction or novels about future societies, and statements and writings by noted experts and prophets. Books by science-fiction authors such as Jules Verne and H. G. Wells have described many events or technological capabilities that have come to pass. Similarly, Aldous Huxley’s Brave New World and George Orwell’s 1984 have alerted many people to some of the dangers of a future technological society. Writings of such noted experts as Charles de Gaulle and Billy Mitchell, foreseeing the role of armor and aircraft in future warfare, could also be cited.

The start of modern systematic technological forecasting in the United States can be traced to the mid-thirties, but its starting point is generally recognized as being in the late fifties and early sixties. In 1962 Mr. Ralph Lenz, one of the chief planners for the U.S. Air Force, published a monograph entitled “Technological Forecasting,” which has become one of the basic references in the field.¹ This report laid the foundation for most of the systematic and analytic forecasting practiced today. A particularly good discussion of the history of technological forecasting is found in Robert Ayres’s Technological Forecasting and Long-Range Planning.²

operations research for quantitative 
planning and forecasting

Planning can be seen in at least two aspects. A reactive type of plan would be based on the view that economic, political, social, technological, and other forces will determine pathways to the future, and thus planning should be a reflection of these forces at work. The other view is that the future can be influenced, at least in how we reach the future. A plan resulting from this view, which could be called a formative type of plan, would be based on determining preferred strategies for reaching the future in some “best” way.

Operations research can be applied with either view of planning in mind. As an example, if market projections anticipated an increased demand for electronic calculators in the next five years, a reactive plan for a manufacturer of calculators could be developed, based on a study using the operations research tools of decision and game theory. This study could account for competitors' intentions in the marketplace, their production capacities, and their projected share of the market. Decision variables could be in terms of which regional markets to compete in. A long-range plan, on the other hand, might be based on estimates of the future numbers of scientists and engineers. Alternative methods of meeting the demand for calculating aids, including peripheral equipment, could be the basis for a study to determine best strategies for a company in meeting the demand. Potential areas of new technology could be identified as a by-product, giving rise to other studies as to how best to exploit the new technologies. This could be a formative plan because it would be influencing the future, perhaps to a significant degree.

Linear programming has been one of the more popular operations research techniques and has been extensively used in product mix studies, studies of warehouse location, transportation, and assignment problems.

The warehousing problem provides a good illustration of how linear programming can be used as a tool for either reactive or formative planning. A plan based on the best use of an existing network of warehouses for a given type of product, with an assumed transportation capability (e.g., a given truck type) would be primarily reactive in nature. On the other hand, such a study could be broadened to consider the storage network and transportation system as decision variables. In this broadened context, promising alternatives to warehouses or trucks could be identified, and thus the plan could be more formative in nature.

As another example, consider a product mix study using linear programming or whatever optimization technique is appropriate. A reactive plan for a refinery, showing the proper mix of additives to minimize costs, would be based on the assumption that petroleum would remain as the prime energy source for transportation. Alternatively, a higher-level study, showing a preferred mix of energy sources to minimize pollution levels, with appropriate constraints in costs and resources, could provide the basis for a formative plan which, in effect, could influence social and technical changes.

Methodology of 
Technological Forecasting

The methodology of technological forecasting can be divided into two broad classifications corresponding to two philosophical viewpoints of the technological process. One, called the ontological view, is that science and technology change in response to scientific and technical opportunities. In the ontological view, technology is seen as a self-generating process, and thus, if one observes this process and gathers appropriate data on its past and present behavior, conclusions can be drawn concerning its future course. The other school of thought, called the teleological view, holds that science and technology change in response to social, economic, political, and other factors in the total environment. This view sees technology more as a servant or by-product of society.

Thus, the forecasting method used depends on whether technological change is seen as being influenced primarily by endogenous variables or determined by exogenous variables. As a result the experts classify technological forecasting methods according to (1) exploratory forecasting corresponding to the ontological view or (2) normative forecasting corresponding to the teleological view. One could also think of exploratory and normative forecasting as analysis according to the “pull” of objectives versus the “push” of opportunities. (Figure 1) Exploratory and normative forecasting can also be viewed as analogous to reactive and formative planning, respectively, which were previously discussed.

exploratory forecasting

Exploratory forecasting treats technological change as being subject to an internal opportunity-oriented law of development. This force is primarily determined by pressures generated by a competitive market. The historical growth of horsepower in automobiles, increased speed of aircraft over time, and growth in substitution of synthetic materials for wool and cotton are all examples of phenomena that could be analyzed with an exploratory forecasting approach. The various exploratory technological forecasting methods generally in use now or under development are listed in Table 1.                                    

Table 1
Exploratory Technological Forecasting Methods

Intuitive forecasting
A. Individual
B. Consensus
     1. Polls
     2. Panels
     3. Delphi
Trend analysis
A. Trend extrapolation
B. Correlation studies
Use of models
A. Analytic analogies
B. Interactive simulations
    1. Feedback models
    2. Cross-impact matrices

One of the more recently and frequently used methods of intuitive forecasting is the Delphi technique.³ This approach was developed at the RAND Corporation and is still the subject of much research. The method basically combines polls and panels in such a way as to benefit from the strong points of each, without the disadvantages of either.

Trend analysis is very commonly used in exploratory forecasting, but this approach can be highly judgmental and ultimately relies on such factors as choice of scale and recognition of appropriate constraints and limits in the process. One of the first requirements for any good trend analysis is recognition that the real world can more often be described as an exponential, rather than a linear, process. This stems from the fact that often the process of technological innovation starts slowly as ideas are formulated and theory is developed. Later, the growth accelerates, and, as maturity sets in, growth changes to virtual linear change. Eventually, the growth exhibits asymptotic behavior as practical limits are reached. Because this pattern is so common, the use of semilog paper for plotting trends is recommended for purposes of forecasting. With such a graph, early and middle growth trends of a process show up as straight lines. As obsolescence or maximum limits in growth are approached, the trend lines change slope and curve in a discernible manner. (Figure 2)

Exploratory forecasting can use mathematical models in several ways. One approach is based on analogies of technological change to other growth processes. One of the better known analogies is based on the Pearl formula, also known as the logistics curve. The name of the formula comes from Raymond Pearl, who formulated it to describe growth of a population in a limited environment, such as fruit flies in a bottle, yeast cells in a fixed nutrient medium, and white rats in a finite space. The formula is 
             
                       P_o
     P =  ———————      (1)
            1 + A exp (—kt)

where P is the population at time t, P_o is the initial population, and A and k are parameters.

Another model involves the notion of cross-impact matrices. (Figure 3) This matrix examines the interactions among three events E₁, E₂, E₃, shown as rows of the matrix. These events have associated with them probabilities P₁, P₂, and P₃ that they will happen in years Y₁, Y₂, and Y₃, with Y₁ < Y₂ < Y_3. The three events also form the columns of the matrix in chronological order from left to right. Each cell of the matrix indicates an interaction between the events in the corresponding row and column. The impacts are shown in terms of mode, strength, and time lag. For example, if E₁ occurs, its mode of impact is to enhance the likelihood of E₂ by 10 percent, and the impact is felt immediately, without a time lag.

Figure 3. Concept of cross-impact matrix                                                                                

The cross-impact matrix technique seems to be almost the exclusive purview of technological forecasting. This approach is similar to a simulation of interactive processes; e.g., a battle or war game. However, it has two features that many other interactive models do not have: (a) nonstationary probabilities (they change with time or are dependent on preceding events) and (b) the time-lag effect of one event on another.

normative forecasting

Normative forecasting treats technological change as responding to outside stimuli such as economic and social demands. With this view, technology is seen as responding to society's needs. To some extent, normative forecasting is the planning of a road map for the future. Possible, or perhaps desirable, pathways to the future are identified, for the process is goal-oriented. Hence, normative forecasting is done with the view that technology responds to the pressures of a unitary, rather than a competitive, market. 

The methodology of normative forecasting can take many forms. Since normative forecasting is the process of determining preferred alternatives for reaching the future, many of the methods of operations research should find application quite easily. Table 2 gives a breakdown of the various methods of normative forecasting cited in most of the literature. Very little has been reported on possible use of other methods; e.g., queueing theory or mathematical programming, thus presenting challenge and opportunity to the operations research profession.                            
 

Table 2

Normative Technological Forecasting Methods

Morphological analysis
    A. Schematic
    B. Matrix
    Relevance trees
    Mission flow diagrams

Morphological analysis is a systematic procedure for collecting, counting, indexing, and identifying all possible alternatives, to achieve some technological capability. Robert Ayres discusses the basis for morphological analysis and illustrates two methods of graphically displaying the combinations of possibilities in achieving a functional capability. A schematic approach and a matrix approach are presented in Figures 4 and 5, respectively. With the schematic technique, one can define most favorable or promising paths of development; and with the matrix approach, one can identify any opportunities that are promising or have been overlooked or show no promise at all.

A model based on a detailed hierarchy of methods of achieving a specific set of goals is called a relevance tree. Such an approach can be used for selecting projects for a research program, designing a budget, or any number of activities. This is one method of mapping out contingencies, prerequisites, and alternatives. To some extent, it could be thought of as the reverse of a decision tree. Instead of outlining all possible paths resulting from a sequence of decisions, a relevance tree maps all possible paths leading to an end result. In some applications, a relevance tree analysis is equivalent to mathematical programming, because the objective is to determine the optimal strategies, or allocation of resources, for a research or technology program within certain well-defined constraints. As an example, a schematic representation of a relevance tree for an automobile technology program is shown in Figure 6.

Mission flow diagrams can be viewed as mapping all possible contingencies. Thus they are similar in many respects to decision trees, mentioned in the discussion on relevance trees. A decision tree maps out possible contingencies resulting from decisions taken in some sequential order. Similarly, a mission flow diagram identifies all paths and forks in the road, so to speak, resulting from a sequence of events. It is a systematic way of identifying all possible future scenarios.

Use of Operations Research Methods
 in Technological Forecasting

Operations research methods used in addition to the more common technological forecasting techniques are certain applications of decision and game theory, optimization and allocation techniques, and models and simulations. Such methods can be useful for embellishing an existing technological forecast based on conventional forecasting approaches.

Wehrner von Braun, writing in Astronautics and Aeronautics, engaged in some “blue-sky” thinking and offered the following technological forecasts:

a. Laser transmission lines carrying thousands of TV channels and billions of telephone conversations simultaneously, with satellites acting as a switchboard network, will be operational. They will enable instant communications between any people anywhere, thereby allowing homes to be communications centers, circumventing the need for offices, banks, retail stores, postal services, newspapers, and even transportation networks.

b. Solar collections will exist in orbit, capturing the sun's energy and beaming this power to enormous antennas on earth, from which the power will be distributed.

c. Manned space stations in orbit will provide the capability of manufacturing products in zero-g environment; e.g., perfectly round ball bearings and new types of alloys and optical glass that cannot be made under the influence of earth's gravity.

d. Long-term prediction (and later control) of the weather through a system of meteorological satellites will be possible. Also, inventory and analysis of earth resources and physical conditions at all times from the same satellites will be accomplished.⁴

We could view each of these forecasts in at least two ways. One view would be that each is the result of an exploratory exercise, and the use of additional analysis techniques could then serve as a “fine-tuning” activity; that is, we could double-check the forecast by using an independent analysis approach, which may result in a recommended modification of the forecast. Another view would be that each of the forecasts represents a stated set of goals and thus provides the basis for subsequent normative forecasting activity. A normative forecast would identify preferred alternatives in achieving these technological capabilities. This would constitute forecasting at a lower level; e.g., projecting requirements for subsystems, support facilities, and the like.

Let us take one of von Braun's forecasts and illustrate the use of some of the more common operations research methods. The forecast we will use is the one pertaining to manned space stations used for manufacturing purposes. Furthermore, let us put a time dimension on this forecast and say that 1990 is the projected date for operation of the space station.

For the first example, a network analysis, perhaps of the PERT type, could be used to determine whether indeed it would be possible to accomplish this capability by 1990. Before the network is laid out, a set of activities and/or necessary events would have to be defined. As a first cut, a rather gross network could be used for determining critical paths, total project time, or areas where further development at the subsystem level is needed. Concerning this last point, once a critical subsystem technology is determined, another network pertaining only to that subsystem can then be devised. This process could be repeated at progressively lower levels until all relevant activities were accounted for.

This process could be used in either an exploratory or normative approach to forecasting, either determining when to expect total project completion or establishing the total set of technologies needed in order to meet a specified completion date. This can also be true of other operations research methods. For example, consider how a queueing analysis could be useful in sharpening our forecast. An elementary queueing analysis, of the type found in textbooks, addresses the problem of the cost trade-offs in server capacities, numbers of servers, and the number of customers waiting in line. Similarly, a queueing analysis could be structured concerning the trade-offs in space station manufacturing capacity, numbers of space stations, and unfilled orders for the manufactured goods. The implications of such an analysis could be significant: it could identify shuttle technology requirements, required command, control, and communication networks for coordinating manufacturing operations, and the necessary system for launch operations and facilities.

For augmenting the forecast, replacement models or renewal theory can be applied to solve the typical problem of when such items as light bulbs or trucks should be replaced, whether before or after wear-out actually occurs. The space station operation would certainly have problems associated with wear-out and replacement.

Optimization techniques and allocation methods could also be applied to this forecast. For example, linear programming could be applied to determine preferred space station fleet sizes and unit designs of both space stations and shuttle systems.

Many of the technological forecasting methods are common to operations research practice; e.g., Delphi, trend extrapolation, mission flow diagrams, and relevance trees. Other methods of technological forecasting represent areas where operations research has not been extensively applied heretofore; e.g., morphological analysis and application of certain types of simulation. Computer simulations of battles or other competitive processes could be applied to technological forecasting problems. Such a simulation would address the possible outcomes for a technological base resulting from competition between two or more technologies. To set up the scenario for such a simulation would involve a morphological analysis to determine all possible technological alternatives. A flow chart of such a simulation for the space station example is shown in Figure 7.

The Challenges

Technological forecasting is a new field that represents a growth area for operations research. It can become an area of activity similar in scope and depth to many other fields that have heavy operations research flavor; e.g., transportation science, inventory control, quality control, econometrics, decision sciences, and management information systems.

Thus, operations research practitioners have several challenges:

• To become more aware of technological change, in terms of both the state of the art itself and the impact it has on society.
• To be increasingly aware of how conventional analysis methods can be applied to improve forecasts.
• To improve and extend the methods of operations research—e.g., simulations and algorithms for cross-impact studies—for application to forecasting studies.

Technological change, with its impact on society and civilization as a whole, is one of the key issues of our time. Technology assessment and technology forecasting have been recognized as study activities and growing fields of endeavor that will be with us as long as the basic values of our present civilization stay as they are. These values are characterized by some as materialistic, but this is an oversimplification. As contrasted to ancient and medieval times, the values of the modern era, starting with the Renaissance, are based on the premise that man has some control over his destiny and that he should use nature and the earth's resources to improve his lot, rather than being a servant of nature. In recent years, this value system has been extended somewhat and now includes the notion that man is a custodian of nature and has responsibilities for conserving the resources of earth. Still, technology can be viewed as the instrument for both improving human conditions and insuring the conservation of nature. As long as this view holds, technological forecasting will be an integral part of the process of change. The challenge to analysts is clearly there.

Aeronautical Systems Division, AFSC

Notes

1. R. C. Lenz, Jr., “Technological Forecasting,” ASD-TDR-62-414, Aeronautical Systems Division, Air Force Systems Command, June 1962 (DDC Number AD 408 085).

2. Robert U. Ayres, Technological Forecasting and Long-Range Planning (New York: McGraw-Hill, 1969).

3. T. J. Gordon and O. Helmer, “Report on a Long-Range Forecasting Study,” RAND Corporation, September 1964.

4. W. von Braun, “Prospective Space Developments,” Astronautics and Aeronautics, April 1972, pp. 26-35.

Other references

Anderson, R. C., and Sproull W.C., “Requirement Analysis, Need Forecasting, and Technology Planning Using the Honeywell PATTERN Technique,” in Cetron, M. J., and Ralph, C. A., Industrial Applications by Technological Forecasting (New York: John Wiley and Sons, Inc., 1971).

Boylan, Edward S., “The Systems Dynamics Approach to Modeling World-Wide Interactions: A Critical Analysis,” Rutgers University, Department of Mathematics, 1972.

Dalkey, Norman C., “An Elementary Cross-Impact Model,” Technological Forecasting and Social Change, vol. 3, 1972.

Fisher, J. C., and Pry, R. H., “A Simple Substitution Model of Technological Change,” Technological Forecasting and Social Change, vol. 3, 1971.

Forrester, Jay W., Industrial Dynamics (Cambridge, Massachusetts: Massachusetts Institute of Technology Press, 1961).

___. Urban Dynamics (Cambridge, Massachusetts: Massachusetts Institute of Technology Press, 1969).

___. World Dynamics (Cambridge, Massachusetts: Wright-Allen Press, 1971). Gordon, T. J., and Hayward, H., editors, “Initial Experiments with the Cross-Impact Method of Forecasting,” Technological Forecasting (Canoga Park, California: Xyzyx Information Corporation, 1970).

Lenz, R. C., Jr., and Lanford, H. W., “The Substitution Phenomenon,” Business Horizons, Graduate School of Business, Indiana University, Bloomington, Indiana, February 1972.

Martino, J. P., Technological Forecasting for Decision-Making (New York: American Elsevier, 1972).

Meadow, D., et al, The Limits to Growth (New York: Universe Books, 1972).

Schmidt, A. W., and Smith, D. F., “Generation and Application of Technological Forecasts for R&D Programming,” Technological Forecasting for Industry and Government, James R. Bright, editor (Englewood Cliffs, New Jersey: Prentice-Hall 1968).

Sigford, J. V., and Parvin, R. H., “Project PATTERN: A Methodology for Determining Relevance in Complex Decision-Making,” IEEE Transactions on Engineering Management, EM-12, March 1965.

Zwicky, F., Collection of papers found in Morphology of Propulsive Power, Society for Morphological Research, Pasadena, California, 1962.

Contributor

Dr. Roy K. Frick (Ph.D., Ohio State University) is Chief, General Purpose and Airlift Division, Office of the Deputy for Development Planning, Aeronautical Systems Division (AFSC). He is also a lecturer in operations research at the Air Force Institute of Technology and serves on several triservice and NATO panels on systems analysis and operations research. Dr. Frick is the author of numerous technical articles and is currently lecturing for several short courses in technological forecasting.

Disclaimer

The conclusions and opinions expressed in this document are those of the author cultivated in the freedom of expression, academic environment of Air University. They do not reflect the official position of the U.S. Government, Department of Defense, the United States Air Force or the Air University.

Air & Space Power Home Page | Feedback? Email the Editor