Trend-impact analysis is a technique for projecting future trends from information gathered on past behavior. The uniqueness of this method lies in its combination of statistical method and human judgment. If predictions are based on quantitative data alone, they will fail to reflect the impact of unprecedented future events. On the other hand, human judgment provides only subjective insights into the future. Therefore, because both human judgment and statistical extrapolation have their shortcomings, both should be taken into consideration when predicting future trends. In trend-impact analysis (TIA), past history is first extrapolated with the help of a computer. Then the judgment of experts is sought (usually by means of the delphi technique) to specify a set of unique future events that may have a bearing on the phenomenon under study and to indicate how the trend extrapolation may be affected by the occurrence of each of these events. The computer then uses these judgments to modify its trend extrapolation. Finally, the experts review the adjusted extrapolation and modify the inputs in those cases in which an input appears unreasonable.

To illustrate TIA methods, let us consider the case of the average price of a new prescription drug to the year 2005. The events considered relevant include (a) generic dispensing, which increases 20 percent of all prescriptions filled; (b) Medicaid and Medicare prescription reimbursement, which is based on a fixed monthly fee per covered patient (“capitation plan”); and (c) a 50 percent decrease in the average rate of growth in prescription size. Consider the first event, i.e., 20 percent increase in generic dispensing. Expert judgment may show that this event has a 75 percent chance of occurring by 1997. If this event does occur, it is expected that its first impact on the average price of a new prescription will begin right away. The maximum impact, a 3 percent reduction in the average price, will occur after five years. The combination of these events, probabilities, and impacts with the baseline extrapolation leads to a forecast markedly different from the baseline extrapolation. The curve even begins to taper off in the year 2005. The level of uncertainty is indicated by quartiles above and below the mean forecast. At this juncture, it is desirable to determine the sensitivity of these results to the individual estimates upon which they are based. For example, one might raise valid questions about the estimates of event probability, the magnitude of the impacts used, and the lag time associated with these impacts.

Having prepared these data in a disaggregated fashion, one can very easily vary such estimates and view the change in results. It may also be observed that intervention policies, whether they are institutional (such as lobbying, advertising, or new marketing approaches) or technological (such as increased research and development expenditures), can be viewed as a means of influencing event probabilities or impacts. TIA can be used not only to improve forecasts of time series variables but also to study the sensitivity of these forecasts to policy. Of course, any policy under consideration should attempt to influence as many events as possible rather than one, as in this example. Corporate actions often have both beneficial and detrimental effects because they may increase both desirable and undesirable possibilities. The use of TIA can make such uncertainties more clearly visible than can traditional methods.

CROSS-IMPACT ANALYSIS

Cross-impact analysis, as mentioned earlier, is a technique used for examining the impacts of potential future events upon each other. It indicates the relative importance of specific events, identifies groups of reinforcing or inhibiting events, and reveals relationships between events that appear unrelated. In brief, crossimpact analysis provides a future forecast, making due allowance for the effect of interacting forces on the shape of things to come. Essentially, this technique consists of selecting a group of five to ten project participants who are asked to specify critical events having any relationship with the subject of the analysis. For example, in an analysis of a marketing project, events may fall into any of the following categories:

1. Corporate objectives and goals.

2. Corporate strategy.

3. Markets or customers (potential volume, market share, possible strategies of key customers, etc.).

4. Competitors (product, price, promotion, and distribution strategies).

5. Overall competitive strategic posture, whether aggressive or defensive.

6. Internally or externally developed strategies that might affect the project.

7. Legal or regulatory activities having favorable or unfavorable effects.

8. Other social, demographic, or economic events. The initial attempt at specifying critical events presumably will generate a long list of alternatives that should be consolidated into a manageable size (e.g., 25 to 30 events) by means of group discussion, concentrated thinking, elimination of duplications, and refinement of the problem. It is desirable for each event to contain one and only one variable, thus avoiding double counting. Selected events are represented in an n × n matrix for developing the estimated impact of each event on every other event. This is done by assuming that each specific event has already occurred and that it will have an enhancing, an inhibiting, or no effect on other events. If desired, impacts may be weighted. The project coordinator seeks impact estimates from each project participant individually and consolidates the estimates in the matrix form. Individual results, in summary form, are presented to the group. Project participants vote on the impact of each event. If the spread of votes is too wide, the coordinator asks those persons voting at the extremes to justify their positions. The participants are encouraged to discuss differences in the hope of clarifying problems.

Another round of voting takes place. During this second round, opinions usually converge, and the median value of the votes is entered in the appropriate cell in the matrix. This procedure is repeated until the entire matrix is complete. In the process of completing the matrix, a review of occurrences and interactions identifies events that are strong actors and significant reactors and provides a subjective opinion of their relative strengths. This information then serves as an important input in formulating strategy. The use of cross-impact analysis may be illustrated with reference to a study concerning the future of U.S. automobile component suppliers. The following events were set forth in the study:

1. Motor vehicle safety standards that come into effect between 1992 and 1996 will result in an additional 150 pounds of weight for the average-sized U.S. car.

2. The 1993 NOX emissions regulations will be relaxed by the EPA.

3. The retail price of gasoline (regular grade) will be $2 per gallon.

4. U.S. automakers will introduce passenger cars that will achieve at least 40 mpg under average summer driving conditions.

The arrows show the direction of the analysis. For example, the occurrence of Event A would be likely to bring more pressure to bear upon regulatory officials; consequently, Event B would be more likely to occur. An enhancing arrow is therefore placed in the cell where Row A and Column B intersect. Moving to Column C, it is not expected that the occurrence of Event A will have any effect on Event C, so a horizontal line is placed in this cell. It is judged that the occurrence of Event Awould make Event D less likely to occur, and an inhibiting arrow is placed in this cell. If Event B were to occur, the consensus is that Event A would be more likely; hence the enhancing arrow. Event B is not expected to affect Event C but would make Event D more likely. Cells are completed in accordance with these judgments. Similar analyses for Events C and D complete the matrix. The completed matrix shows the direction of the impact of rows (actors) upon columns (reactors). An analysis of the matrix at this point reveals that Reactor C has only one actor (Event D) because there is only one reaction in Column C. If interest is primarily focused on Event D, Column D should be studied for actor events. Then each actor should be examined to determine what degree of influence, if any, it is likely to have on other actors in order to bring about Event D.

Next, impacts should be quantified to show linkage strengths (i.e., to determine how strongly the occurrence or nonoccurrence of one event would influence the occurrence of every other event). To assist in quantifying interactions, a subjective rating scale, such as the one shown on page 307, may be used. Voting Scale Subjective Scale + 8 Critical: essential for success Enhancing + 6 Major: major item for success + 4 Significant: positive and helpful but not essential + 2 Slight: noticeable enhancing effect 0 No effect – 2 Slight: noticeable inhibiting effect Inhibiting – 4 Significant: retarding effect – 6 Major: major obstacle to success – 8 Critical: almost insurmountable hurdle Consider the impact of Event A upon Event B. It is felt that the occurrence of Event A would significantly improve the likelihood of the occurrence of Event B. Event A’s occurrence would make Event D less likely; therefore, the consensus rating is –4. This process continues until all interactions have been evaluated and the matrix is complete. There are a number of variations for quantifying interactions. For example, the subjective scale could be 0 to 10 rather than –8 to +8, as shown in the example above. Another technique for quantifying interactions involves the use of probabilities. If the probability of the occurrence of each event is assessed before the construction of the matrix, then the change in that probability can be assessed for each interaction. Consider the impact of Event Aon the probable occurrence of Event B. It is judged to be an enhancing effect, and the consensus is that the probability of Event B occurring will change from 0.8 to 0.9. The new probability is therefore entered in the appropriate cell. Event A is judged to have no effect upon Event C; therefore, the original probability, 0.5, is unchanged. Event D is inhibited by the occurrence of Event A, and the resulting probability of occurrence is lowered from 0.5 to 0.4.

The occurrence of Event B increases the probability of Event A occurring from 0.7 to 0.8. Event B has no impact upon Event C (0.5, unchanged) and increases the probability of Event D to 0.7. This procedure is followed until all cells are completed. An examination of the matrix at this stage reveals several important relationships. For example, if we wanted Event D to occur, then the most likely actors are Events B and C. We would then examine Columns B and C to determine what actors might be influenced. Influences that bring about desired results at a critical moment are often secondary, tertiary, or beyond. In many instances, the degree of impact is not the only important information to be gathered from a consideration of interactions. Time relationships are often very important and can be shown in a number of ways. It shows that if Event A were to occur, it would have an enhancing effect upon Event B, raising B’s probability of occurrence from 0.8 to 0.9, and that this enhancement would occur immediately. If Event B were to occur, it would raise the probability of the occurrence of Event D from 0.5 to 0.

7. It would also take two years to reach the probable time of occurrence of Event D.