Sandwich Example: You are in a restaurant. It offers several different sandwiches. Sandwich filling choices are: beef, peanut butter, salami, egg, and banana. Bread choices are: white, whole wheat, and pumper-nickel. Butter choices are: buttered bread and unbuttered bread. What are all the different sandwiches offered by the restaurant?

Driving Forces and Major Uncertainties

Trend Impact Analysis

Article by Theodore Jay Gordon:

Cross Impact Analysis

Article by Theodore Jay Gordon


A note prepared by Mr. Kapil Verma (PGP 2002), HDFC Bank <kapil.verma@hdfcbank.com>

Brief History

The Cross impact method of scenario development was originally developed by Theodore Gordon and Olaf Helmer in the form of a game for Kaiser Aluminum and Chemical company in the mid 1960s called Future. It represented an effort to extend the forecasting techniques of the Delphi method.

What Is It?

Cross-impact analysis is a technique for analysis of complex systems. It concentrates on the ways in which forces on an organization, external or internal (events or trends) , may interact to produce effects bigger than the sum of parts , or to magnify the effect of one force because of feedback loops. It has been used successfully where the dominant forces can be identified, and the modeling mechanism can be used to increase management's understanding of the relative importance of various factors.

The structure used for understanding the impact of events and trends on each other is called a Cross-Impact matrix.

Objective of the Technique

The objective of using a cross-impact technique of scenario development is to visualize the possible future of a subject of interest, for example: the possible scenarios of the Indian telecommunications industry in the future. To visualize such a future, the concept of "events" is used.

An event is something that can happen to modify the current or actual situation of the subject of interest. Events to be utilized in a cross-impact analysis are defined by two properties.

1. They are expected to happen only once in the interval of time under consideration(i.e., nonrecurrence events)
2. They do not have to happen at all (i.e., transient event)

Thus, the cross-impact approach in its most general context is an attempt to arrive at meaningful analyses of a system composed of transient, nonrecurrent events, which may or may not be dependent upon memory.

How To Do It

The procedure of cross-impact method is as follows:

1. Select a team of experts in the subject of interest. For example: the Indian information technology industry. The experts may belong to various fields of study and not necessarily to the subject of interest.
2. The team of experts will identify the most relevant events that can impact the future explored.
3. The team of experts will then assess their best guess about the probability of each event previously recognized (initial probabilities are fixed at this stage.)
4. Then the team of experts will construct the cross-impact matrix for this exercise.(see figure 1)
5. On the basis of the information of the cross-impact matrix, the team of experts will have to assess the initial conditioned probabilities (for those events that impact other ones).
6. After assessing the initial probabilities, simulate the process of event-occurrence using the Monte Carlo technique (taking as input the initial probabilities). This would help in refining the probabilities.
7. Establish the new values for all initial probabilities (the unconditioned ones and the conditioned ones).
8. Repeat the simulation process taken as input the new refined or reviewed probabilities
9. From the results of the latest simulation exercise the team of experts will obtain the probability of occurrence of each possible scenario.

IMPACTS THE EVENT EVENT	E1	E2	E3	E4
E1	_		_
E2	_	_		_
E3		_	_	_
E4	_	_		_
Figure 1: Cross-Impact Matrix

The number of possible scenarios is determined as 2^ No. of events previously identified. (viz. is 'n'). For example, if n=4, then the No. of possible scenarios will be equal to 2^n=2^4=16

For example, suppose a business group is trying to decide whether to build a new business school. First, they identify a number of trends and potential events that are relevant to their decision. They may identify the following trends:

1. The number of students enrolling in business management courses (MBA) is growing at 20 percent per year.(Trend 1)
2. Privately managed business school enrollment is increasing faster than reputed Government/State University run b-schools' enrollment (Trend 2)
3. Average family income is increasing by 10 percent per year.(Trend 3)
4. Average GMAT / CAT Entrance test scores required for admission into private b-schools are increasing faster than the state average (Trend 4)

Next they may identify the following potential events:

1. The Government may impose a ceiling on the acquisition of land to restrict rapid growth of commercial establishments.(Event 1)
2. The Government may implement a voucher system to allow families to use some portion of public funds (to pay B-school tuition fees) at less than market rate of interest.(Event 2)
3. The government imposes an additional tax-surcharge on corporate profits. (Event 3)

Next the experts array these trends and event along both axes of a matrix, and try to determine what affect the occurrence of an event or the continuation of a trend will have on the other trends and events.

For example, if the government imposes an additional tax-surcharge on corporate income, family income would be less likely to continue rising by 10 percent per year. (See Figure 2) This simplified version of cross-impact analysis forces participants to think through the ways in which various future events may interact and thus can help clarify people's conceptions of the future. In its more sophisticated form, cross-impact also tests the consistency of participants' judgment, by asking them to determine the probability of an event occurring under different circumstances.

TRENDS AND EVENTS	TREND 1	TREND 2	TREND 3	TREND 4
EVENT 1	XX	X	XX	NI
EVENT 2	++	++	++	+
EVENT 3	X	XX	++	NI
Figure 2: Interaction of Events and Trends

XX Very low impact / probability of occurrence; ++ Very high impact / probability of occurrence; X Low impact / probability of occurrence; + High impact / probability of occurrence; NI No Impact

Data Required

In the most sophisticated model, seven estimates are required to depict the connection between an event impacting on the probability of another event:

1. Length of time from the occurrence of the impacting event before its effects would be felt first by the impacted event;
2. The degree of change in the probability of the impacted event at that point when the impacting event would have its maximum impact;
3. The length of time from the occurrence of the impacting event until this maximum impact (that is, change in probability) would be achieved;
4. The length of time from the occurrence of the impacting event that this maximum impact level would endure;
5. If the maximum impact might taper off, the change in probability of the impacted event when its new, stable level were reached;
6. The length of time from the occurrence of the impacting event to reach this stable impact level;
7. A judgment as to whether or not these effects had been taken into account when estimating the probability of the impacting and impacted events in the Delphi.

Eight cross-impact factors need to be estimated to describe the hit of an event on a trend. The first seven are the same as those specified above, except that estimates 2 and 5 are not for changes in probability but for changes in the nominal forecasted value of the trend. The eighth estimate specifies whether the changes in the trend values are to be multiplicative or additive.

Strengths of the Method

1. The cross-impact method forces attention to chains of casualty: X affects Y ; Y affects Z . If the input to a cross-impact matrix falls outside acceptable probabilistic bounds, or if the result of a cross-impact run is surprising, then the researcher is forced to reexamine his or her view of expected reality.
2. The disaggregation required by the method is usually illuminating. Inserting a cross-impact matrix into another model often adds power to that model by bringing into its scope future external events that may, in the limit, change the structure of the model.
3. Cross-impact method provides means of testing sensitivity to changes in probabilities of future events and contemplated policies, an important consideration in "planning" studies.

Weaknesses of the Method

The collection of data can be fatiguing and tedious.

1. The cross-impact method assumes that, somehow and in some applications, conditional probabilities are more accurate that estimates of a-piori probabilities. This is unproved.
2. Cross-impact studies focus on interactions between pairs of events.Yet in the real world , the important interactions may involve not only pairs but triplets and higher-order effects. If such interactions were to be included, however, the complexity of judgement collection would grow tremendously.
3. There is no analytical guidance for resolving the fundamental question of what particular events should make up the specified set of dependent and independent events. This procedure is entirely dependent upon the group, which will be supplying the estimates, and the general problem area that is to be examined.
4. The procedure exemplifies the madness of models over substance. Many researchers focused on the mathematics of the technique and forgotten that the real purpose of their efforts is to generate better forecasts!

Conclusion

In recent years, the work on cross-impact has shifted from "pure" methodological development to applications [Vickers, Brent, October 1992]. Questions about the method remain, ofcourse: how best to ask questions about conditional probabilities; is the method really convergent; how to handle non-coherent input from experts; how to integrate with other methods? But there is no doubt that cross impact questions help illuminate perceptions about hidden causalities and feedback loops in pathways to the future.

References

Web Addresses
'Methods and Approaches of Futures Studies' World Future Society. (http://www.wfs.org/newmeth.htm); [29 September 2001]
The Manoa Journal of Fried and Half-Fried Ideas (about the future) http://www.soc.hawaii.edu/future/jrnls.html#top; [29 September, 2001]
Seminar on Futures Techniques http://ag.arizona.edu/futures/tou/sem1.html#Cross%20Impact%20Analysis ; [30 September, 2001]
Articles, Research Papers
Theodore Jay Gordon, (1994). "Cross Impact Method", African Futures Project. UNDP. pp13-14
Murray Turoff, (1972). "An Alternative Approach to Cross-Impact Analysis". Technological Forecasting and Social Change. p.346.
Dr. Jesús E. Arapé M., (April 2000). "The most recognized and applied methodologies in technology foresight". Regional Conference on Technology Foresight for Central and Eastern Europe and Newly Independent States, (No page numbers available).
Book
Gill Ringland, (1998). Scenario planning: Managing for the Future, John Wiley and Sons, p.26 and p.187.