Decision Modeling - The Millennium Projectfor editing, and John Young for proofreading. ... Although...

The Millennium Project Futures Research Methods—V3.0

DECISION MODELING

by

The Futures Group International1

I. History of the Method

II. Description of the Method

III. How to Do It

IV. Strengths and Weaknesses of the Method

V. Frontiers of the Method

VI Applications of the Method

VII. Who is Doing It

Bibliography

1 The Futures Group International, http://www.futuresgroup.com


Decision Modeling ii

Acknowledgments

Version 2.0 of this chapter contained descriptions of both substitution analysis and decision

modeling. In this Version 3.0 edition the two methods have been separated into two chapters.

The managing editors wish to thank the reviewers of this paper who made many important

suggestions and contributions: For the original paper, reviewers were Peter Bishop of the

University of Houston, Graham Molitor, president of Public Policy Forecasting Inc. and vice-

president of the World Future Society, Eleonora Masini of the Pontifical Gregorian University,

Stephen Sapirie of the World Health Organization, and Larry Hills of USAID. Jerome C. Glenn

made contributions to alternative approaches in section II of the paper. Ted Gordon updated his

chapter. Special thanks to Elizabeth Florescu and Neda Zawahri for project support, Sheila Harty

for editing, and John Young for proofreading.


Decision Modeling 1

I. HISTORY OF THE METHOD

Decision modeling attempts to replicate the actual behavior of decision making and is based on

the identification of specific criteria and the assessment of how well competing options meet

those criteria. For example, when purchasing a product, such as an automobile, a consumer

might consider price, quality, service, and options. In decision modeling, each attribute is

weighted by its relative importance, and each car model is judged on how well it matches each

criterion. Decision modeling can also explore the market potential of new technologies by

assessing how well the new technology meets criteria already established by the marketplace. In

these applications, decision modeling quantifies the potential of a product or technology to gain

share from products already on the market.

One can trace the beginnings of the field back to the 17th

century when Pascal posed a problem

of decision making under high uncertainty (no matter how small the probability of the existence

of God, since the payoff is eternal, it is worth believing.) In the 18th

century Bernoulli’s work

introduced the notion of risk and utility in decision making. Most of the early work assumed the

rationality of decision makers; more recently, in this century, various authors have challenged

this notion. Tversky and Kahneman (1979), for example, used psychological experimentation to

show the innate irrationality of human decision making.

It emphasized that in actual human (as opposed to normatively correct) decision-making "losses loom larger than gains", people are more focused on changes in their utility states than the states themselves and estimation of subjective probabilities is severely biased by anchoring.1

The literature in this field is quite extensive. A search of the Scopus data base show that since

the year 2000, a search on the term “decision-modeling (and modeling) futures” yields about 25

hits. Most of this literature deals with specific applications of decision modeling techniques,

derivations of utility functions for various groups in various situations, refinement of the

measurement and even the concept of risks, and rules for selecting appropriate decision criteria.

Michel Godet has also made some interesting additions to the technique of decision modeling.

His MULTIPOL method assumes an alternative future environment rather than the single-valued

future of most decisions. Godet implements that view by adjusting the weights involved in the

decision according to the environment that is being forecast. In buying a car, for instance, the

criterion of fuel efficiency would receive a higher weight in a low-energy future than in a high-

energy one. This flexibility makes possible the discussion of relative advantages of different

policies or strategies across a spectrum of alternative futures.

1 Wikipedia, Decision Theory, 2008

http://en.wikipedia.org/wiki/Anchoring


Decision Modeling 2

II. DESCRIPTION OF THE METHOD

The technique of decision modeling is based on the concept of utility theory from the field of

systems analysis. In utility theory, a rational decision maker selects an alternative product,

policy, or action that best meets his or her criteria for success. As an example, consider the

decision a young person must make when deciding which college to attend. When asked what’s

important in a college, the factors mentioned might include: excels in my field, has good

football, is co-ed. The parent might add price to the set of criteria. These four criteria would be

weighted according to their relative importance. Then, using this method, a matrix could be

constructed in which possible colleges would be compared on the basis of these criteria. For

example;

Criteria >>>>

Teaches what

I want Good football Co-ed Cost Score

Weights >>>> 8 4 4 9

College 1 3 5 3 3 83

College 2 4 4 4 4 100

College 3 5 3 5 3 99

The column labeled “score” is the weighted sum for each college. All other things being equal,

the most rational choice in this example is college 2. If, when such an analysis is performed, the

results seem subjectively wrong, it is appropriate to ask if a criterion has been omitted, or the

weights are improper.

The behavior of a large number of systems is determined, to a great extent, by decisions made by

people or groups within those systems. In population systems, the behavior of couples of child-

bearing age determines the dynamics of the system; in market systems, the collective decisions

of consumers constitute market behavior; in industries, such as the electric utility industry,

decisions by corporate executives on generation expansion determine many characteristics of

that system. Thus, in order to understand the behavior of systems, understanding the nature of

decision-making within the system is important.

To illustrate, if instead of colleges a matrix of this sort listed products, and the criteria, weights,

and cell entries were based on consumer research, then the scores might well have a relationship

to the market shares that the products would derive in direct market competition. Research has

shown that the relationship may not be linear, however, but rather s-shaped. In other words a

good product at the top of the heap has to get a lot better to capture much more market share.

In a third application, the process is reversed. Instead of starting with criteria and producing a

decision, decisions are analyzed to infer the criteria that must have been at work in order for that

decision to have been made. Here, decision modeling attempts to develop a model of the

decision process applied by decision-makers, including consumers, to important decisions within

their system. This approach assumes that decision-makers consider a number of different factors


Decision Modeling 3

implicitly when comparing various alternatives and that some of these factors are more important

than others. Although a decision maker may not actually list these decision factors or

consciously weigh them, they are implicit in the perceived value of alternatives. To choose the

"best" alternative, the decision maker must make a judgment about what constitutes high value

and low value. A choice that has a low cost may be considered more valuable than one with

higher cost, while one with higher benefits may be more valuable than one with lower benefits.

How does a high-cost, high-benefit technology compare with a low-cost, low-benefit alternative?

To answer this question, the decision-maker must specify how important cost and benefits are.

The low-cost, low-benefit alternative may be perceived to be the "better" alternative, if cost is

much more important than benefits; or, the reverse may be true, if benefits are more important.

Numerous other approaches have been developed to model or analyze consumer decisions; one

of the most common is conjoint measurement. Conjoint measurement relies on the ability of

respondents to make judgments about their preference for predetermined combinations of

attributes (Churchill 1987). The objective of this approach is to determine the features that

respondents most prefer. The thinking behind conjoint analysis is that, if respondents were asked

directly, they might find it difficult to discuss how they were combining attributes to form an

overall judgment. In effect, the value systems of respondents are inferred from behaviors

reflected in their choices, rather than by their identification of each attribute's importance.

In constructing a conjoint analysis project, a number of considerations are critical to the project's

success: selection of the attributes; levels of each attribute; determination of the combinations;

selection of presentation form to those who make the judgments; the nature of the judgments;

and selection of analysis technique. Some analysts prefer to use a full-profile approach to collect

judgments; that is, each combination contains each attribute. Others simplify the judgment by

using a pairwise procedure that asks subjects to indicate their preference for pairs. In the

pairwise procedure, two attributes are treated at a time, but all possible pairs are considered.

Various methods of decision analysis have also been incorporated into strategic planning as

exemplified by Jerome Glenn's Strategy Analysis Grid (Glenn 1989). This grid is designed to

illustrate the range of general strategic choice. Glenn asserts that strategies fitting in the upper-

left box are easier to implement but are possibly less effective than those in the lower-right

boxes. Strategies in the lower-right box tend to be more difficult to implement (see Table 1).


Decision Modeling 4

Table 1

STRATEGIC ANALYSIS GRID

Degree of

Difficulty.>>>

Change

Within System

Change from

Outside System

Create

New Systems

Approach 1:

Provide

Information

1

Approach 2:

Provide

Positive/Negative

Reinforcement

2

Approach 3:

Provide

Environmental

Change

3

For example, suppose one had as a goal the reduction of noise in the library. Cell 1 would use

information within the system as a solution, e.g., posting a "Be Quiet" sign on the wall. Cell 2

would provide positive or negative reinforcement within the library (e.g., asking noisy

individuals to leave) to achieve the objective. Cell 3 would look for environmental change to

create a new system (e.g., library patrons could use a computer network with remote library

access to eliminate noise in the library or the library could be equipped with extensive noise

deadening material).

Through this grid, one can examine which strategic approaches were used in the past and

identify tradeoffs between more effective strategies and the degree of difficulty.

Another approach to planning, which also relies on the grid format, was developed by Robert

Bundy, formerly of the Educational Policy Research Center at Syracuse University. In this grid,

the needs of individuals (e.g., physiological, safety, esteem, etc.) are arrayed in rows. Trends,

such as urbanization, pluralism, and automation, head each column. The cells are designed to ask

questions, such as how the need for safety will influence urbanization, how urbanization might

satisfy or frustrate this need, and what policies might result.

There are many other similar grid analyses of policy options. For example, Glenn, building on

Maslow’s hierarchy of human needs, in Table 2 illustrates a matrix that helps analyze the

humanizing or dehumanizing effects of policy options. Each cell is completed by asking how

the minimum condition of each policy option might satisfy and/or frustrate a human need.


Decision Modeling 5

Table 2

HUMAN NEEDS/POLICY OPTION GRID

Policy

Options >>>> Policy 1 Policy 2 Policy 3 Policy 4 Etc.

Human Needs

Physical

Safety

Esteem

Love and

Belongingness

Self-Actualization

Aesthetic

III. HOW TO DO IT*

The first step is to list the decision criteria used to judge the alternatives. Weights of relative

importance are then assigned to each criterion. The degree to which each alternative choice

meets each criterion is then estimated. The result is a matrix, such as the one shown in Table 3.

Table 3

MATRIX

Decision

Criteria

Decision

Weights

Alternative Ratings

A1 A2 A3 A4 A5

C1 W1 A11 A21 A31 A41 A51

C2 W2 A12 A22 A32 A42 A52

C3 W3 A13 A23 A33 A43 A53

C4 W4 A14 A24 A34 A44 A54

C1 through C4 represent the decision criteria. W1 through W4 are the weights of importance for

those criteria. A1 through A5 are the available alternatives, and A11 through A54 indicate the

degree to which a particular alternative meets a particular criterion.

Once these entries are specified, the value of each alternative to the decision-maker is

computed by summing the products of the weights and ratings. Thus, Vi= Σn (Wn × Ain)

*As mentioned, decision modeling involves a number of approaches. The method described here was

developed by The Futures Group.


Decision Modeling 6

where: Vi = perceived value of alternative i

Wn = importance weight of criterion n

Ain = degree to which alternative i satisfies criterion n

Since the value Vi calculated in this manner is dependent on the scales used for weighing the

criteria and rating the alternatives, these values are converted to relative numbers by dividing by

the average value.

RVi = Vi/AV

AV = (Σi Vi)/NOAA

where: RVi = relative value of alternative i

AV = average value of all alternatives

NOAA = number of alternatives available

The alternative with the highest relative value is the expected choice for a given decision maker.

When using this method to simulate the outcomes of a large number of decisions, however, the

highest-valued alternative is not likely to be the choice in every case. In this case, the

importance weights and alternative ratings represent average values perceived by the decision

makers. Thus, while the highest-valued alternative should be chosen in more decisions than any

other alternative, it may not be chosen in all decisions. (Because of differences in regional or

individual perceptions, an alternative may have the highest value for an individual, but not for

the group as a whole, when average weights and ratings are used.)

Therefore, the relative perceived values calculated for each alternative must be converted into

market penetrations. This step is accomplished by plotting relative perceived value versus

market penetration for some historical period for which data are available. The result is usually

an S-shaped curve in which historical data can fit using a logistics equation.

Once developed from past data, this relationship can be used with future values of perceived

value to determine market penetration. Since no guarantee exists that the penetrations thus

calculated will sum to 100 percent of the market, they must be normalized to sum to 100.

When new alternatives that were unavailable historically become available to the decision

maker, these alternatives can be introduced into the model by describing them in terms of the

decision criteria. An example is shown in Tables 4 and 5.


Decision Modeling 7

Table 4

ILLUSTRATION OF A MARKET BEFORE A NEW PRODUCT INTRODUCTION

Criterion

Weight

Ford

Toyota

Nissan

VW

Total Price

10

10

6

7

7

Quality

6

3

10

7

8

Service

8

4

10

5

7

Options

2

1

10

4

5

Total

152

220

160

184

Market Share

21%

32%

22%

25%

100%

Table 5

ILLUSTRATION OF A MARKET AFTER A NEW PRODUCT INTRODUCTION

Criterion

Weight

Ford

Toyota

Nissan

VW

New

auto

Total

Price

10

10

6

7

7

10

Quality

6

3

10

7

8

2

Service

8

4

10

5

7

5

Options

2

1

10

4

5

1

Total

152

220

160

184

154

Market

Share

17%

26%

18%

21%

18%

100%

For most of these new alternatives, a learning period is involved. Decision-makers are often

reluctant to choose a new alternative with which they have no experience. They may also be

uncertain of cost and other figures and thus be hesitant to consider fully a new alternative. As

experience with the new alternative accumulates, however, these problems disappear. In order to

take this learning behavior into account, a learning curve adjustment that reduces the perceived

value of a new alternative must be introduced in the first years.

In summary, decision modeling describes the decision process as a choice among competing

alternatives made on the basis of how well each alternative meets several different criteria of

varying importance. These perceptions are by no means static. The importance weights and

alternative ratings can, and usually do, change with time and can be influenced to a greater or

lesser degree by marketing, advertising, and external circumstances. (Remember how automobile


Decision Modeling 8

choice changed as a result of the increase in gasoline prices.) By incorporating this type of

decision model into a description of the entire system surrounding the decision (possibly a

simulation model of the entire system), a better understanding of the behavior of the system

should result.

IV. STRENGTHS AND WEAKNESSES OF THE METHOD

Decision modeling is quite useful in analyzing past or pending decisions as well as capturing the

essential aspects involved by consumers in making decisions.

The many applications of decision modeling all share common weaknesses:

1. Identifying the criteria. Just how can the elements of a decision be known, either for an

individual or a group? Many psychological impediments distort a true and complete articulation

of what’s important.

2. Collection of information which can be used in conjoint analysis or in establishing decision

criteria and their weights. Such market research is often difficult, inaccurate, and costly.

3. Perceptions shift with time and circumstances. To return to the initial example of this chapter,

once in college, our young student may find other things important, and the original criteria

inadequate.

However, one of these weaknesses is a strength as well: the ability to accept market research

data as an input. In a poll, for example, consumers might be asked what other products they

considered when they made their purchase, what factors led to their choice, and how important

the factors were in their selection process. With this kind of data in hand, a decision matrix can

be completed. If the data do not fit an S-shaped curve, then some decision factor may have been

omitted. In addition, once a good model is established, marketers can identify which attributes to

stress or improve in order to increase market share.

For complex decisions which may affect many people for long periods of time, the simple utility

matrix provides a great deal of clarity since it requires answers to the question: what’s important.

It also promotes thinking about what can go wrong. But as the work of Kahneman and Tversky

shows, posing and answering questions that involve subjectivity can be a tricky business.


Decision Modeling 9

V. FRONTIERS OF THE METHOD

Various computer software programs exist to help decision makers. Generally, these approaches

use utility matrices to help select the option which best meets a set of weighted criteria, or they

may use conjoint analysis to identify preferences and tease out criteria and weights. Some

systems focus on assessment of risks and may estimate risk based on Monte Carlo simulations in

a manner similar to Trend Impact Analysis (also in FRM Version 3.0). Examples of some of

these decision support systems follow:

LOGICAL DECISIONS, version 6.1. Fairfax, VA (http://www.logicaldecisions.com/)

They say: “Logical Decisions provides innovative solutions for hard choices. Our state-

of-the art software package -- Logical Decisions® for Windows -- lets you evaluate

choices by considering many variables at once, separating facts from value judgments,

and explaining your choice to others. “

DECISIONTOOLS® Suite 5.0., Palisade Corp., Ithaca, NY.

(http://www.palisade.com/decisiontools_suite/) They say: “The DecisionTools Suite is an

integrated set of programs for risk analysis and decision making under uncertainty that

run in Microsoft Excel. The new DecisionTools Suite 5.0 includes @RISK 5.0 as well as

all-new versions of PrecisionTree 5.0 and TopRank 5.0. In addition, the Suite has been

expanded to add StatTools 5.0, NeuralTools 5.0, and Evolver 5.0 for prediction, data

analysis and optimization. All programs have been rewritten to work together better than

ever before.”

EXPERT CHOICE, version 7.0. Arlington, VA. (http://www.expertchoice.com/) This

software makes use of the Analytic Hierarchy Process (AHP) of Thomas L. Saaty (1990).

Reviews can be found in various computer periodicals and the Journal of Marketing

Research (Carmone, November 1992). Expert Choice says: “The AHP and Expert Choice

software engage decision makers in structuring a decision into smaller parts, proceeding

from the goal to objectives to sub-objectives down to the alternative courses of action.

Decision makers then make simple pairwise comparison judgments throughout the

hierarchy to arrive at overall priorities for the alternatives. The decision problem may

involve social, political, technical, and economic factors. The AHP helps people cope

with the intuitive, the rational and the irrational, and with risk and uncertainty in complex

settings. It can be used to: predict likely outcomes, plan projected and desired futures,

facilitate group decision making, exercise control over changes in the decision making

system, allocate resources, select alternatives, do cost/benefit comparisons, evaluate

employees and locate wage increases.

“Expert Choice is intuitive, graphically based and structured in a user-friendly fashion so as

to be valuable for conceptual and analytical thinkers, novices and category experts. Because

the criteria are presented in a hierarchical structure, decision makers are able to drill down to

their level of expertise, and apply judgments to the objectives deemed important to achieving


Decision Modeling 10

their goals. At the end of the process, decision makers are fully cognizant of how and why

the decision was made, with results that are meaningful, easy to communicate, and

actionable.”

· Vanguard Software Corporation, Cary NC

(http://www.vanguardsw.com/solutions/application/decision-support/): They say: “(We

offer a) broad application of modeling and simulation techniques for improving decision-

making. Vanguard's decision support system software makes it possible for you to apply

decision analysis techniques throughout your organization to problems ranging from

simple projects to enterprise-wide strategic plans.”

· Professor Hossein Arsham, offers Tools for Decision Analysis.

(http://home.ubalt.edu/ntsbarsh/opre640a/partIX.htm). He says: “This site offers a

decision making procedure for solving complex problems step by step. It presents the

decision-analysis process for both public and private decision-making, using different

decision criteria, different types of information, and information of varying quality. It

describes the elements in the analysis of decision alternatives and choices, as well as the

goals and objectives that guide decision-making. The key issues related to a decision-

maker's preferences regarding alternatives, criteria for choice, and choice modes, together

with the risk assessment tools are also presented.”

Of the packages cited above, Expert Choice is the only one that relies on AHP. Expert Choice

organizes the various components in a structure similar to a relevance tree. The structure

consists of a goal, criteria, and identification of alternative levels. In Expert Choice, users make

pairwise (conjoint) comparisons, and the software program allows for weighing the decision

factors.

User-friendly and relatively inexpensive computer software packages for conjoint analysis

appeared on the market in the mid-1990s. Among these are:

Question Pro, Seattle, WA (http://www.questionpro.com/info/contactUs.html)

Sawtooth Software of Sequim, WA

(http://www.sawtoothsoftware.com/education/techpap.shtml)

Bretton-Clark. Conjoint analysis has been explored as a means of segmenting markets

Decision Analysis is a journal of the Institute for Operations Research and the Management

Sciences. (http://www.informs.org/site/DA/) “a quarterly journal dedicated to advancing the

theory, application, and teaching of all aspects of decision analysis. The primary focus of the

journal is to develop and study operational decision-making methods, drawing on all aspects of

decision theory and decision analysis, with the ultimate objective of providing practical guidance

for decision makers. As such, the journal aims to bridge the theory and practice of decision

analysis, facilitating communication and the exchange of knowledge among decision analysts in

academia, business, industry, and government.”

There is a Decision Analysis Society (http://decision-analysis.society.informs.org/) which

“promotes the development and use of logical methods for the improvement of decision-making

in public and private enterprise. Such methods include models for decision-making under

http://home.ubalt.edu/ntsbarsh/index.html



conditions of uncertainty or multiple objectives; techniques of risk analysis and risk assessment;

experimental and descriptive studies of decision-making behavior; economic analysis of

competitive and strategic decisions; techniques for facilitating decision-making by groups; and

computer modeling software and expert systems for decision support.”

Utility matrix decision methods can easily be combined with agent modeling. In agent modeling

simulation individuals or groups interact in computer-based algorithms represented in on-screen

contests and consequences. The rules of behavior of the agents are usually supplied by the

experimenter from observations or through judgment. However, the rules can be determined

within the algorithm by using utility matrices in which the weights can vary as the environment

of the simulation shifts. For example, if one agent is a predator and the other its prey, the weight

it places on finding food can vary as the population of the prey changes.

Decision models can be constructed using Excel. Moore and Westford have written a textbook

on the subject although it has seemed too complex for some users.



VI. APPLICATIONS OF THE METHOD

A Model of Utility Planning for Base-Load Generation*

Electric utilities are in the business of supplying electrical energy to their customers. They can

produce it in a variety of ways. Utilities, therefore, attempt to choose the "best" generating

alternative each time new capacity is planned. Determining the "best" alternative is not always a

simple task, however, because each utility has a number of criteria to meet. Return on

stockholders' investment is an important criterion to private utilities. The achievement of low

electricity prices may be another criterion since it can lead to expansion of demand. The

maintenance of high system reliability is necessary to ensure uninterrupted service to all

customers. Utilities must also consider a variety of other criteria, such as: suitability of available

plant sites, certainty of a steady supply of fuel for the lifetime of a new plant, acceptance by

regulatory commissions of the utility plans, and actions of other utilities within the same region.

Scope of the Problem

In applying decision modeling to a particular problem, the degree of disaggregation of the model

is an important factor. Ideally, the model should be disaggregated sufficiently so that each

decision group is relatively homogeneous, that is, each member of the group makes decisions by

considering the same criteria and weighing scale. Thus, although this methodology could be

applied to the United States as a whole, to do so could miss important regional differences

significant when introducing new technology. The model described here focuses on the

Northeast Power Coordinating Council (NPCC) region and on base-load generating alternatives.

Interviews with Utility Executives

The first step in developing the model is to develop a weighted list of decision criteria. To

accomplish this task, telephone interviews were conducted with electric utility executives in the

NPCC region. The interviews had four major objectives:

* refine understanding of the decision process and confirm that the approach

adopted captured the essential components of that process

* develop the list of decision criteria considered by decision-makers

*

Part of the research reported in this study by The Futures Group was conducted for the Energy Research

and Development Administration under subcontract to TRW Systems Group. A slightly revised version

of the example appeared in "Stover, J. S. The Use of Decision Modeling for Substitution Analysis:

Application to Acceptance of New Electricity-Generating Technologies," Technological Forecasting and

Social Change, Vol. 12, 1978, pp. 337-351.



* elicit estimates on the importance of each criterion in the final decision for both base-load

and peak capacity for past, present, and future time periods

* elicit estimates of the length of the learning process involved in new technology

introduction.

The interview consisted of two major parts:

* Discussion and weighing of base-load decision factors

* Appraisal of the utility's gamble in adapting new generating technologies.

The respondent was first asked whether the factors listed in Table 6 represented a complete

set of factors determining what type of new base-load generating capacity should be installed.

The respondent was then asked to rate the factors as they now appeared in importance. All the

factors were rated on a scale of 1-10, with 1 signifying little importance and 10 the highest

importance. The respondents were told that the same number could be applied to several factors,

if they were of equal importance or if they could not be disconnected logically from one another.

Having finished the ratings in the applied "present" column, the respondents were then asked to

make judgments on the ratings of factors from the decision making point of view in 1960 and

also to conjecture on what the ratings for the decision factors might be in the year 2000.

Table 6

SUGGESTED LIST OF DECISION FACTORS

USED IN TELEPHONE INTERVIEWS

1. Capital Costs

2. Operating and Maintenance Costs

3. Fuel Costs

4. Fuel Availability

5. Environmental Impact

6. Lead-Time

7. Reliability

The consideration of reliability and environmental impact

represents additional concerns beyond the requirement to meet

certain system reliability criteria and environmental regulations.

A technology not meeting either of these requirements would not

even be considered an available alternative.

When the ratings for both the base-load decision factors and the peak-load decision factors were

completed, the following question was put to the respondent: How many years are likely to pass

after the first utility's acceptance of a new technology before your company would consider

adopting it? The answers to this question were sought for both base-load and peak-load capacity.



In general, all respondents agreed that the seven factors selected for weighing were a

comprehensive set by which to describe the selection among alternative types of both base-load

and peak-load generating capacity. In reality, however, particularly in the case of base-load

plants, the specific conditions of a candidate site could weigh in the selection of the type of

generation. For example, location of a nuclear plant close to current load centers is still

problematic. Further, difficulties in obtaining suitable sites for generating plants were often the

dominant consideration in planning for new generation. Nevertheless, while considerations of

siting could discriminate among types of generating plants, such considerations would be too

site-specific from which to generalize. Thus, the factor of site availability was not added to the

list of seven factors to be considered.

Some of the respondents felt that all seven factors had interrelated cost implications and that

rating them independently was difficult. Several respondents expressed the feeling that current

consumer sensitivity had made the price of electricity the highest importance.

Of the seven factors governing the choice of base-load generation, the ratings for capital costs

and fuel costs show that these two factors dominated decision-making in the past and are

expected to dominate it in the future. Expectedly, concern over securing long-term fuel supplies

has added to the choice of fuel for electric generation. The ratings show that this concern may

increase in the future.

In 1960, capital costs and fuel costs were the primary discriminants in economic analyses that

led to coal-fired and oil-fired base-load plants in the NPCC area. In 1978, with concern for fuel

availability and environmental impacts, nuclear plants were the dominant choice of new base-

load generation. For the future, considerations of fuel availability and environmental impact are

expected to remain at least as strong as they are today. While many respondents expressed

anxiety over secure future supplies for both nuclear and fossil fuels, basic economics would

continue to select nuclear base-load plants for this area.

Most of the respondents indicated that three cost items — capital costs, operation and

maintenance, and fuel costs — were not separable. Therefore, these three items were combined

into one cost item by averaging the three weights. The final set of weights prepared in this

manner is shown in Table 7.

Responses to the question about the length of time a utility would wait before adopting a new

technology were weighted according to the size of the utility responding. Figure 1 shows the

cumulative percentage of utilities willing to adopt a new technology, for both base-load and

peak-load, versus the number of years elapsed after the first introduction of the technology.



Table 7

AVERAGE DECISION WEIGHTS

(Higher Weights Represent Greater Importance)

Decision Weights

Base Load

Peak Load

Decision Factor

1960

1975

2000

1960

1975

2000

Cost

Fuel Availability

Environmental Impact

Lead-Time

Reliability

8.8

3.1

3.2

2.8

5.3

8.6

7.1

6.1

3.5

6.5

8.7

8.2

6.7

3.4

6.6

7.8

2.9

3.1

2.8

5.1

8.0

5.0

5.8

3.4

5.6

8.1

6.4

6.2

3.4

5.6

Figure 1. Learning Curves of Utilities Willing to Adopt a New Technology for Base- and

Peak-Load Versus Time Elapsed Since First Introduction

Model Calibration

In order to validate the model, the model must be calibrated using some historical period, then

run for another historical period, and compared with the actual behavior of the industry over that

time period.

The calibration period chosen was 1950-1960. During this period, essentially three major

alternatives existed for base-load generation: oil-, coal-, and gas-fired plants. For each type, it



was necessary to specify the degree to which each met five decision criteria: cost, reliability, fuel

availability, lead-time, and environmental impact.

Costs were estimated by combining national data on capital costs, operation, and maintenance

with estimates of fuel costs in the NPCC region. Estimated capital, operation, and maintenance

cost figures were based on Steam Electric Plant Construction Cost and Annual Production

Expenses (Federal Power Commission). For each type of plant, plots were made of capital,

operation, and maintenance costs versus year of initial operation for base-load plants placed on-

line between 1950 and 1973. These plots represent actual costs. Decisions, however, are based

on perceived, or expected, costs. Thus, when a utility decides to build a particular type of plant,

the actual costs of the plant are not known until five or six years later. Expected costs were

estimated using the lowest costs during each period, producing figures that compared well with

expected costs reported in the literature.

Estimates of the other decision criteria for each alternative were made. The lead-time involved

with oil and gas plants was assumed to be five years, with six years required for more complex

coal plants.

Estimates of the other three criteria (environmental impact, fuel availability, and reliability) were

made on a relative scale of 0-1, with 1 denoting the best alternative. Using this system,

reliability for gas and oil plants was rated at .9. Reliability for coal was slightly lower, at .8.

(Note: These are relative numbers and may not represent plant operating factors.) For

environmental impact, gas was rated the most desirable at 1.0, oil at .7, and coal worst at .3.

Finally, for fuel availability, oil was rated at 1.0 through the 1970s, .8 in 1970, and then .5 in

1975, as the future availability of oil became questionable. Coal was rated at 1.0 through 1965,

slightly lower at .8 for 1965-1970, when strikes and transportation problems concerned the

utility industry, and at .95 in 1975, when large reserves seemed to promise a secure supply. Gas

was rated at .6 in 1950, .8 in 1955 as pipeline networks were expanded within the NPCC region,

at 1.0 in 1960-1965, .7 in 1970, and .1 in 1975, as supplies diminished, diverted to other uses.

The two factors not estimated on a 0-1 scale, lead-time and cost, were converted to such a scale

by dividing the value of the best cost lead-time by the values for each alternative. The figures

were then combined with the other estimated ratings and their decision weights in order to

calculate relative perceived values for each generating alternative from 1950 to 1960.

In order to calibrate the model, these perceived values must be compared with actual market

penetrations for each energy source. Because of the small number of base-load plants built

during the 1950-1960 period, the actual yearly penetration figures can vary from 0 to 100

percent, depending on whether a single plant is built in a given year. In order to smooth these

variations and display the trend of utility decisions (which is what decision modeling is designed

to simulate), the megawatts of capacity installed per year for each alternative were smoothed

over three years. These three-year moving averages were converted to percentages of market

penetration and plotted. Smooth curves were then drawn through these curves. The results of

this process are shown in Figure 2. This figure also includes market penetrations for 1960-1985,

which are used in model validations. These future installations are based on plans reported in



Steam-Electric Plant Factors/1975 (National Gas Association). Currently planned new plants,

which are not simply additions to existing plant locations, were used for these calculations.

Calibration of the model consists of plotting these smoothed penetration rates against the relative

perceived value for each alternative for every year in the calibration period. This plot, which

includes coal, oil, and gas, is shown in Figure 3. The curve drawn between the plotted points is

the conversion function used to convert relative perceived value into market penetration for each

fuel alternative.

Model Validation

The model was calibrated using data from 1950 to 1960, a period of competition between coal-,

oil-, and gas-fired plants. To validate the model, the period 1960-1975 was used. This period

includes the introduction of nuclear power, providing an interesting example of new technology

market penetration. To use the model for this time period, the data on nuclear power must be

specified for the five decision criteria. Since nuclear power was only introduced during the

1960s, no history of costs can be used to estimate perceived costs. Therefore, published

estimates on expected nuclear costs were used. Lead-time was assumed to be six years before

1962, increasing to nine years by 1966. Perceived environmental impact was given the highest

value of 1.0. Reliability was initially .9, dropping to .6 by 1975. Fuel availability was set at 1.0

until 1970, dropping to .9 in 1975. The initial date for nuclear consideration was set at 1960.

Results based on adding this information to the model, and running the model from 1950 to

1975, are shown in Figure 4.

The fit from 1956 to 1966 is based on the calibration period 1950-1960 (decisions made during

the 1950-1960 period affected capacity additions in the 1956-1966 period). After 1966, the fit

between historical data and model output remains good. The crossover point for oil and nuclear

occurs close to the actual crossover. The ascendance of nuclear to almost total dominance and

the decline of oil are captured by the model. Finally, the expected reemergence of coal in the

1980s is also duplicated by the model. Thus, decision modeling, using decision weights

estimated by utility executives, appears capable of accurately and reliably describing the market

penetration of new technology into the electric utility industry.



Figure 4. Comparison of Model Output with Actual Market Penetration Rates Forecasting

the Adoption of New Technologies Using Decision Modeling

A description of new technologies that might become available can demonstrate decision

modeling as a methodology in forecasting future market penetration of new technology, adding

alternatives to the model. In order to add these new technologies, the date when the first plant is

placed in service must be specified as well as the degree to which each technology meets each

decision criterion.

The capital, operation, and maintenance costs expected for seven new technologies applicable to

the NPCC region and tested in the model are shown in Table 8. Expected fuel costs are shown in

Table 9.

In the absence of any information on the reliability of future technologies, reliability of each was

set at .8 for the first ten years after introduction and at .9 after that. The lead-times associated

with the fossil fuel plants were all set at six years; the lead-time for the LMFBR was set at nine

years.

The relative degree of environmental intrusion estimated for each technology is shown in

Table 7. Relative ratings of the fossil fuel plants were based on a comparison of their emission of

SOx, NOx, and particulates. Nuclear plants are rated higher than fossil plants in the expectation

that utilities will continue to perceive them as the least-polluting alternative. This perception

could change, of course, if an adequate waste disposal system is not developed or if harmful

accidental releases of radiation occur.



Table 8

BASE-LOAD NEW TECHNOLOGY PLANTS COST AND EFFICIENCY DATA

(1975 Dollars)

O&M

Mills/

kWh

Unit

Cost

$/kWh

Effi-

ciency

Commer-

cial

Capabilit

y

Perceived

Environmen

tal Impact

Advanced steam plant

with atmospheric

fluidized bed

4.0

302-403

40-42

1985

0.65

Advanced steam plant

with pressurized

fluidized bed

4.0

336-504

40-44

1990

0.75

Open-cycle gas turbine

with integral low Btu

gasifier & steam

bottom cycle

4.0

302-403

40-42

1985

0.75

Magnetohydrodynamic

topping cycle with

steam bottom cycle

4.0

504-638

45-50

1997

0.65

Metal vapor topping

cycle with steam

bottom cycle

4.0

504-638

45-50

1997

0.65

Closed-cycle gas

turbine

4.0

336-504

40-44

1990

0.65

LMFBR

1.0

545-756

--

1993

1.00

Light water reactor

0.9

438-563

--

--

1.00

Conventional coal

plant with FGD

3.0

300-360

34

--

0.30



Table 9

FUEL COSTS

1985

1995

2005

2015

2025

Coal mine-mouth

costs (¢/MMBtu)

High

Low

70.00

53.00

80.00

60.00

100.00

65.00

116.00

71.00

134.00

78.00 Coal transportation

costs (¢/MMBtu)

High

Low

10.50

6.00

13.20

7.60

15.10

8.70

15.80

9.10

16.50

9.50 LWR fuel costs

(mills/kWh)

High

Low

6.48

4.01

9.33

4.78

13.81

5.83

20.43

7.10

30.25

8.66

Fuel availability for both coal and nuclear plants was set at .9. Fuel oil availability was assumed

to decline to .1 by the year 2000.

The estimates of capital costs and fuel costs cover a range of costs. These ambiguous costs lead

to uncertain market penetration for new technologies. Uncertainties also exist in the calculations

of decision weight estimates, other plant characteristics, and dates of availability. In the sample

forecasts developed for this study, only the best-specified uncertainties — capital costs and fuel

costs — were used; however, the technique is the same for including uncertainty from any area.

Uncertainty was calculated by including a Monte Carlo model as a part of the overall model.

The variations in capital costs and fuel costs (including both mine-mouth and transportation

costs for coal) were specified as random numbers evenly distributed across the range of likely

values. For each calculation, capital costs were randomly selected for each technology to affect

the perceived value of each technology (based on the weighted import for each criterion). Market

penetrations were then calculated for each technology based on these perceived values.

Repeating this procedure a large number of times, and computing the variation in the technology

penetrations, produced a measure of the uncertainty in a particular year. These calculations can

then be repeated for as many years of the forecast as desired.

The penetrations forecast by decision modeling for each of the technologies are shown in

Figure 5. By the 1990s, three of the advanced fossil fuel plants [advanced steam plant with

atmospheric fluidized bed (AFB), advanced steam plant with pressurized fluidized bed (PFB),

and open-cycle gas turbine with low-Btu gasifier (OCT)] are projected to displace conventional

coal plants. The AFB plant and the OCT compete favorably with conventional coal on the basis

of lower overall costs (mainly due to higher conversion efficiencies) and reduced emissions. The

PFB technology has somewhat higher costs (due to larger capital costs) but is much better from

an emissions standpoint. The closed-cycle gas turbine magneto-hydro dynamics (MHD) topping

cycle and metal vapor topping cycle technologies do not penetrate markets to any great extent,

mainly because of higher costs and equal or inferior environmental ratings. Oil-fired plants also

capture none of the future market because of high costs and perceived low availability of future

fuel.



The light water reactor loses some of its predominance to advanced fossil fuel plants but remains

the most successful single technology, until the LMFBR comes in around 2000. The breeder

then captures the bulk of the market because of a good environmental rating and low costs.

The uncertainty associated with these market penetration calculations is shown in Figure 6. Note

these uncertainties arise from uncertainties in capital, fuel, and transportation costs.

A nuclear moratorium scenario was also tested with this model. This scenario assumes that no

new nuclear plants are planned. This scenario was tested in the decision model by simply

removing the LWR and LMFBR as planning options after 1976. In this case, the closed-cycle gas

turbine technology also captures a significant portion of the new market, along with the three



major competitors (AFB, PFB, and OCT). The MHD and metal vapor technologies also capture a

small share of the market in the absence of the nuclear option.

In general, decision modeling appears useful in capturing the essential components involved in

new technology adoption. The validity test shown here supports the assumption that such a

model, when calibrated over one period in time, remains valid for a later period, even when new

technologies appear that were not part of the calibration period. This point is crucial in assessing

any methodology designed for forecasting new technology adoption.

VII. WHO IS DOING IT

Decision modeling and analysis is widely used in market research and strategic planning. For a

description of applications of conjoint analysis, refer to "Commercial Use of Conjoint Analysis:

A Survey" (Catlin and Wittink 1982). For a review of current software that can assist in

analyzing decisions, refer to the Frontiers section of this paper. Some recent specific studies are

listed below:

Linkov, Varghese, et.al.(2004) use multi-criteria decision analysis to study “environmental

projects dealing with contaminated and disturbed sites where risk assessment and stakeholder

participation are of crucial concern.“

Brito and de Almeida (2009), focus on the use of decision modeling techniques to study risks

associated with natural gas pipelines in Brazil. Based on various hazard scenarios, they use the

utility matrix approach to incorporate decision maker’s preferences and produce a hierarchal list

of pipelines according to risks. .

Yeh, C.-H. and Chang, Y.-H (2009) have written a paper based on their use of a multi-criterion

decision making approach in which the judgments of decision makers used in the model are

captured by a conjoint analysis. Rather than the more common rating on scales of 1-10, they

used cardinal preference values and assessed the “winning” option by evaluating its distance

from ideal. The application was to aircraft selection.

The Futures Group (TFG) has used decision modeling extensively in its work. In addition to

forecasting for a utility, TFG has employed the technique in numerous applications.

OTC Pain Relief Opportunities

For a major U.S. client, the company investigated new product opportunities in the

over-the-counter analgesic market. As an integral part of this research, two sets of external

interviews were conducted: one with a group of expert clinicians and researchers to explore

leading-edge technical research; and another with industry analysts and marketing experts to

delve into market issues and consumer behavior. These interview results, along with other

research and analysis, formed the bases for making certain interim conclusions about the

dimensions of the analgesic market, expected trends and directions, and other factors important

http://www.scopus.com.scopeesprx.elsevier.com/scopus/search/submit/author.url?author=de+Almeida%2c+A.T.&authorId=23988442500&origin=recordpage






to developing new products, concepts, and strategy. Potential new product ideas generated as a

result of this research were augmented during a brainstorming session with the client. All new

product and concept ideas were then rated, based on their potential match with specific corporate

criteria, e.g., market size, market share, etc. To complete the work, market strategies were

designed for the leading new product/concept contenders.

Future Automotive Product Positioning Study

The client for this study was a motor vehicle manufacturer based in Japan. The objective was to

determine the optimal strategy for positioning a mid-sized sedan aimed at the U.S. market in the

mid-to-late 1990s. The work combined extensive quantitative and qualitative analysis,

synthesized through decision modeling. The client's existing market research data were

reanalyzed using multivariate statistical techniques to identify key product attributes and

customer segments. Structured brainstorming was employed to identify important technological

developments likely to affect the market within the relevant time frame. Changes that would

affect customers were approached through analysis of economic and demographic forecasts and

through qualitative assessment of psychographic trends. Decision modeling analyzed the

position issue by estimating the relative appeal of possible combinations of product attributes in

light of the projected size and preferences of key market segments in the 1995-2000 time period.

Alternative "Worlds" of 2000 and Their Impacts on U.S. Fast Food Businesses.

A set of strategies was developed for a major participant in the Quick Service Restaurant (QSR)

business. Detailed scenarios were constructed that described four different "histories" of the

world within which the QSR industry could be operating in the final decade of the 20th century.

A number of strategic options were identified for each scenario, and a recommended strategy

was presented to the client for each different future. Key elements of each strategy set were

arrayed, and elements common to several worlds were identified and molded into a "core"

strategy set. A model was then constructed to describe the decision process that a typical

consumer uses to select among different eating options in each of the hypothetical worlds.

Long-Term Market Share Forecast for a New Subcategory of Consumer Products

The client company, the market share leader in the consumer product category under

investigation, needed to decide whether to introduce products in a new form. This decision

required a long-term estimate of the share of the total product category that might eventually

shift to all competing brands in the new form. TFG used a variety of methodologies to forecast

the rate at which consumers would adopt the new form and the eventual size of the new

subcategory. These methods included:

· Market share analysis, a long-term, multi-attribute decision model, using data from a

conjoint analysis experiment performed by the client

· Analysis of data on the rate of adoption of analogous product forms in similar product

categories



· Substitution analysis, employing polynomial curve-fitting to model the diffusion of

innovation

· The scaling-up of existing test market data to national projections, allowing for varying

demographic and consumer preference factors and for the diffusion of innovation over

time

· Structural analysis to project segmented population growth and buyer behavior through

the year 2000.

A Market Forecasting Decision Model

A computerized decision modeling and analysis was performed for 60 separate markets for

office equipment systems to estimate total system market shares for the next l0 years. The

systems were then desegregated into their component products, and total product market shares

were estimated.

As a result of this study, the client company revised one of its key product line strategies. This

revision resulted in a product line that will better fit the forecasted customer value system and

better compete with the forecasted new technology of its competitors' products.

Forecast of Telecommunication Needs and the Networks in Use Over the Next 30 Years

The objective of this work, performed for a manufacturer of fiber optic cable, was to forecast

demand for bandwidth likely to be used in various telecommunication services over the next 30

years. Recognizing the growing need for telecommunication services, the work was also to

identify the specific network types likely to carry these services. The study involved data

gathering through on-line literature search as well as interviewing primary sources on both

technological and market-driven factors. In all, some 25 emerging telecommunication services

were evaluated. Forecasting models were built for each of these, and the results were presented

in terms of bit and bit-rate requirements over time in local, intermediate, and long-distance

service. Peak time of day was also identified. Using decision modeling, network types

"competed" to carry each of the telecommunication services on the basis of their unique

attributes. This decision model also included explicit consideration of other factors, such as:

network availability, consumer loyalty, price elasticity, etc. The model was implemented on an

IBM PC (with extended memory) and utilized Lotus 1-2-3 to facilitate data entry and output.

The model produced forecasts of total demand for information as well as the likely share of the

market by each potential carrier in local, intermediate, and long-distance service.



BIBLIOGRAPHY

Averch, H., and Johnson L. "Behavior of the Firm under Regulatory Constraint." The American

Economic Review 52, 1962.

Ayres, R. V. "A Schumpetrian Model of Technological Substitution." Technological

Forecasting and Social Change 27, 1985, pp. 375-383 .

Baughman, Martin. Dynamic Energy System Modeling Interfuel Competition. Massachusetts

Institute of Technology Energy Lab, Cambridge, MA, 1973.

Blackman, A.W., Jr. "A Mathematical Model for Trend Forecasts." Technological Forecasting

and Social Change 3, 1972, pp. 441-452.

Blackman, A. W., Seligman, E. J., and Sogliero, G. C. "An Innovation Index Based on Factor

Analysis." Technological Forecasting and Social Change 4, 1973, pp. 301-316.

Bretchneider, S. I., and Mahajan, V. "Adaptive Technological Substitution Models."

Technological Forecasting and Social Change 18, 1980, pp. 129-139.

Brito, A.J. and, de Almeida, A.T, “.Multi-Attribute Risk Assessment For Risk Ranking Of

Natural Gas Pipelines”, Federal University of Pernambuco UFPE, Cx. Postal 7462, CEP 50.630-

970 Recife, Brazil, 2009

Cameron, H. M., and Metcalfe, J. S. "On the Economics of Technological Substitution."


Carmone, Frank J., Jr. "New Books in Review." In Journal of Marketing Research. Edited by

Donald E. Stern, Jr., November 1992.

Catlin, Philipe, and Wittink, Dick R. "Commercial Use of Conjoint Analysis: A Survey."

Journal of Marketing, Summer 1982.

Churchill, Gilbert A. Marketing Research. Chicago, IL: The Dryden Press, 1987.

Dyer, R. F.; Forman, E. H.; and Jouflas, G. Marketing Decisions Using Export Choice. DSS,

Inc., 1988.

Easingwood, C., Mahajan, V., and Muller, E. "A Nonsymmetric Responding Logistic Model for

Forecasting Technological Substitution." Technological Forecasting and Social Change 20,

1981, pp. 199-213.



Fishbein, Martin. "A Behavior Theory Approach to the Relations between Beliefs about an

Object and the Attitude Toward the Object." Readings in Attitude Theory and Measurement.

M. Fishbein (ed.). New York: John Wiley & Sons, 1967.

Fisher, J. C., and Pry, R. H. "A Simple Substitution Model of Technological Change."


Floyd, A. "A Methodology for Forecasting of Figures of Merit." In Technological Forecasting

for Industry and Government: Methods and Applications. Edited by James R. Bright.

Englewood Cliffs, NJ: Prentice-Hall, 1968.

Glenn, Jerome C. Future Mind: Artificial Intelligence: Merging the Mystical and the

Technological in the 21st Century. Herndon, VA: Acropolis, 1989.

Godet, Michel. The Crisis in Forecasting and the Emergence of the Prospective Approach:

With Case Studies in Energy and Air Transport. Tarrytown, NY: Pergamum, 1979.

Green, Paul E.; Carroll, J. Douglas; and Goldberg, Stephen M. "A General Approach to Product

Design Optimization Via Conjoint Analysis." Journal of Marketing, Summer 1981.

Green, Paul E., and Krieger, Abba M. "Segmenting Markets with Conjoint Analysis." Journal

of Marketing, October 1991.

Green, Paul E., and Srinivasan, V. "Conjoint Analysis in Consumer Research: Issues and

Outlook." Journal of Consumer Research, September 1978.

Lee, J. C., and Lu, K. W. "On a Family of Data Based Transformed Models Useful in

Forecasting Technological Substitutions." Technological Forecasting and Social Change 31,

1987, pp. 61-78.

Kahneman, Daniel, and Amos Tversky (1979) "Prospect Theory: An Analysis of Decision under

Risk", Econometrica, XLVII (1979), 263-291.

Linkov, I, A. Varghese, S. Jamil, T.P. Seager, G. Kiker, and T. Bridges in their paper, “Multi-

Criteria Decision Analysis: A Framework For Structuring Remedial Decisions At Contaminated

Sites, in Linkov, I. and Ramadan, A. eds “Comparative Risk Assessment and Environmental

Decision Making” Kluwer, 2004, p. 15-54.

http://www.environmentalfutures.org/Images/ArmyPaper_Oct31.pdf

Mansfield, E. "Technical Change and the Rate of Imitation." Econometrica 29, October 1961.

Martino, J. P. Technological Forecasting for Decision Making. 2nd ed. New York: North-

Holland, 1983.

Meade, Nigel. "Technological Substitution: A Framework of Stochastic Models."




Jeffrey H. Moore, and Larry R. Weatherford, “Decision Modeling with Microsoft®

Excel” (6th

Edition) (2001) available through Amazon.com.

Rosenberg, Milton J. "Cognitive Structure and Attitudinal Effect." Journal of Abnormal and

Social Psychology 53, 1956, pp. 367-372.

Saaty, Thomas L. Decision Making for Leaders: The Analytic Hierarchy Process for Decisions

in a Complex World. Pittsburgh, PA: RWS Publications, 1990.

Saaty, Thomas L. Multicriteria Decision Making — The Analytic Hierarchy Process: Planning,

Priority Setting, Resource Allocation. Pittsburgh, PA: RWS Publications, 1990.

Saaty, Thomas L., and Vargas, Louis G. Prediction, Projection & Forecasting: Applications of

the Analytic Hierarchy Process in Politics, Economics & Finance. Norwell, MA: Kluwer Ac.,

1990.

Sahal, D. "A Generalized Logistic Model for Technological Forecasting." Technological

Forecasting and Social Change 7, 1975, pp. 81-97.

Sharif, M. N., and Kabir, C.A. "Generalized Model for Forecasting Technological Substitution."


Skiadas, C. H. "Two Generalized Rational Models for Forecasting Innovation Diffusion."


Skiadas, C. H. "Innovation Diffusion Models Expressing Asymmetry and/or Positively or

Negatively Influencing Forces." Technological Forecasting and Social Change 30, 1986, pp.

313-330.

von Neumann, J., and Morgenstern, O. Theory of Games and Economic Behavior. Princeton,

NJ: Princeton University Press, 1944.

Wilkie, William L., and Pessemier, Edgar A. "Issues in Marketing's Use of Multi-Attribute

Attitude Models." Journal of Marketing Research 10, 1973, pp. 428-441.

Xiao Ting-yan, "A Combined Growth Model for Trend Forecasts." Technological Forecasting


Yeh, C.-H.and Chang, Y.-H, “Modeling Subjective Evaluation For Fuzzy Group Multicriteria

Decision Making. European Journal of Operational Research, Volume 194, Issue 2, 16 April

2009, Pages 464-473.

Tingyan, Xiao. "A Combined Growth Model for Trend Forecasts." Technological Forecasting


http://www.amazon.com/exec/obidos/search-handle-url/ref=ntt_athr_dp_sr_1?%5Fencoding=UTF8&search-type=ss&index=books&field-author=Jeffrey%20H.%20Moore

http://www.amazon.com/exec/obidos/search-handle-url/ref=ntt_athr_dp_sr_2?%5Fencoding=UTF8&search-type=ss&index=books&field-author=Larry%20R.%20Weatherford

http://www.scopus.com.scopeesprx.elsevier.com/scopus/search/submit/author.url?author=Chang%2c+Y.-H.&authorId=7501842462&origin=recordpage

http://www.scopus.com.scopeesprx.elsevier.com/scopus/source/sourceInfo.url?sourceId=22489

Decision Modeling - The Millennium Projectfor editing, and John Young for proofreading. ... Although...

Documents

Transcript of Decision Modeling - The Millennium Projectfor editing, and John Young for proofreading. ... Although...