Qualitative cross-impact analysis with time consideration
Introduction
The future can never be accurately or completely known. The purpose of futures research is to systematically explore, create, and test both possible and desirable futures to improve decisions [1]. The use of futures research methods enhances anticipatory consciousness, which in turn improves the foresight to act faster or earlier making the organization or individual more effective in dealing with change [1]. There are a wide variety of methods used for futures research ranging from simplistic to complex, qualitative to quantitative (for a simple taxonomy of the methods of futures research see [2], [3]). Generally, the choice of the method depends on the problem, resources, and levels of sophistication of the planners and users [4]. Among the best-known methods are the Delphi method, cross-impact analysis (CIA), simulation, and scenario writing. The Delphi method and scenario writing were both developed in the 1950s and constitute the roots of scenario planning [5].
Although the future is a result of interactions of many events related through structures that are dynamic and evolving over time, a basic limitation of many futures research methods is that they produce information only in isolation. That is, events and developments are projected without considering their possible influence on each other. To explore the behavior of a system in the future, the set of variables, key to a systematic description of the system, and their interrelationships that will shape the future have to be analyzed. These interrelationships are called “cross-impact” and the most popular method used to analyze them is the CIA. CIA uses a cross-impact matrix for systematic description of all potential modes of interaction between a given set of variables and for the assessment of the strength of these interactions [6]. The cross-impact method was originally developed by Theodore Gordon and Olaf Helmer in 1966 and reported by Gordon and Hayward [7]. Since then, several versions of CIA have been developed by researchers [8], [9], [10], [11], [12], [13], [14], [15], [16], [17]. These can be classified into three groups: quantitative, qualitative, and mixed CIA. In quantitative CIA, a mathematical model relating to the variables is constructed, while in qualitative CIA, experts are asked to provide subjective estimates of the relationships among the variables, usually in the form of a matrix of conditional probabilities or impact values. In this study, a qualitative CIA based on the structural analysis proposed by Duperrin and Godet [18] is adapted. The aim of the qualitative CIA is to reduce the complexity of the system and to identify the important few variables that should be studied first. Being important for a variable means being strongly connected to the system, which is evaluated by the number and intensity of its relationships, and variables having this characteristic are called key variables. As any change in key variables will affect the whole system, they deserve more attention in the future [6]. Thus, estimating the alternative developments of key variables represents an estimation of the system in the future. Different combinations of the alternative developments of the key variables are then used to construct future scenarios. The qualitative CIA comprises the following steps:
- 1.
Problem analysis: Within this step the scope of the analysis, the scenario field, and the modeling work are defined. An alternative classification for the scenario fields is: external, internal, and systems scenarios [19]. This step also includes the collection of initial data and information.
- 2.
Variable definition: The results of the first step are aggregated into variables that represent a measure of the system and the factors. Variables can be categorical or non-categorical.
- 3.
Relationships analysis: There are two types of relationships used for classification of variables: direct and indirect.
- a.
Direct classification: A cross-impact matrix is used to set up all potential impacts between the variables and to assess the strengths of these impacts. A pairwise analysis takes all relationships into account by asking: “if variable A changed, what would be its direct impact on variable B?” [6] For assessing the impacts different scales (e.g. binary, intensity, linguistic etc.) can be used. Finally, the variables are classified according to their influence and dependence values, where the influence value of a variable refers to the sum of its row entries, and the dependence value to the sum of its column entries.
- b.
Indirect classification: Matrix multiplication is applied to the cross-impact matrix “to study the diffusion of impacts through reaction paths and loops” [12]. This multiplication process continues until the cross-impact matrix is raised to a certain power in which the variables' order proves to be stable. (see for detail, the MICMAC method; [12]). As in the direct classification, the variables are classified according to the sum of row and column entries of the resulting matrix.
- a.
- 4.
Chart analysis: An influence–dependence chart [12] is prepared to interpret the results. Each variable is assigned to a unique position on the chart according to its influence and dependence values. This position of the variable reveals its individual role in relation to the system [6]. The chart is prepared for both the direct and indirect classification.
- 5.
Selection of key variables: Considering the direct and indirect classifications and the chart analysis the variables with both high influence and high dependence are selected as key variables.
A critical weakness of the CIA is that it does not incorporate the time impact into the analysis. Since knowing the time relationship between events is no less important than knowing the causal relationship, a more realistic model would introduce a time lag corresponding to each relationship [20]. This paper proposes a complementary approach to the qualitative CIA including the time impact: Cross-impact Analysis with Time Consideration (CIAT). The rest of this paper is organized as follows. First, we provide a quick overview of the concept “time” and review the literature on CIA with emphasis on methods that incorporate the concept of time into the analysis. Next, we present the proposed approach, which starts by estimating time impacts and uses this information to revise and weight the cross-impacts in order to determine the role of each variable. Then the CIAT is illustrated by an example. Finally, the contribution of the proposed approach is summarized and further research directions are mentioned.
Section snippets
Time and cross-impact analysis
The notion of time in economic theory has been divided between so-called causal and historical time ([21], [22] cited in Ref. [23]). Causal time is a theoretical relation of variables, where if all the variables are considered at the same moment of time the static analysis is used and if there are variables at different moments then the dynamic analysis is used. According to Pfeifer [23], the historical time, on the other hand, implies the growth of uncertainty (the law of entropy) and the
The proposed approach
This paper proposes a complementary approach to the qualitative CIA by analyzing the impact of time on the interrelationships between variables. The originality of the proposed approach lies in its ability to consider both causal and time impacts in systems with feedback. It also enables the examination of indirect impacts. We suggest that it is important to consider the time relationships between all pairs of variables rather than searching for a time sequence between a particular starting and
An illustrative example
In order to illustrate the contribution of the proposed approach on analyzing the interrelationships between variables, we consider the case of security equipment market. The data for the illustrative example is taken from a broader scenario planning application made by Polat and Asan [46], which was aimed at constructing scenarios for the security equipment market in Turkey. The scenarios are system scenarios, where both external (non-influenceable) and internal (influenceable) variables are
Conclusion
Cross-impact analysis, as a means of futures research, reveals the characteristic role and importance of a variable in relation to all other variables in the system by examining all potential interactions. These interactions should be analyzed in two dimensions: impact and time. However, traditional CIA does not reflect the time impact and therefore fails to capture this important aspect.
This paper suggests a complementary approach to CIA which improves the understanding of the structural
Acknowledgements
The authors gratefully acknowledge the constructive and helpful comments of two anonymous referees on the earlier versions of the manuscript.
Seyda Serdar Asan is a research assistant in the Department of Industrial Engineering at the Istanbul Technical University and currently pursuing her PhD degree in the Department of Quality Sciences at the Technical University of Berlin. Her main research interests include quality management and supply chain management.
References (46)
How to develop and use scenarios
Long Range Plan.
(1987)- et al.
Helping managers to identify the key strategic issues
Long Range Plan.
(1997) - et al.
Initial experiments with the cross-impact matrix method of forecasting
Futures
(1968) A primer for a new cross-impact language—KSIM
Technol. Forecast. Soc. Change
(1972)An alternative approach to cross-impact analysis
Technol. Forecast. Soc. Change
(1972)Reassessment of cross-impact analysis
Futures
(1981)A note on cross-impact analysis: A calculus for sequence-dependent events
Futures
(1972)Time-dependent event cross-impact analysis: Results from a new model
Technol. Forecast. Soc. Change
(1977)The evaluation of generic cross-impact models
Futures
(1979)- et al.
A qualitative cross-impact approach to find the key technology
Technol. Forecast. Soc. Change
(1997)
A fuzzy approach to qualitative cross-impact analysis
Omega
Scenario management: An approach to develop future potentials
Technol. Forecast. Soc. Change
Seeking convergence between history and futures research
Futures
Subjective conditional probability modeling
Technol. Forecast. Soc. Change
Cross-impact analysis and classical probability: The question of consistency
Futures
A Markovian cross-impact model
Futures
Marking the future: A review of time horizons
Futures
Introduction to the futures research methods series
Technology futures analysis: Toward integration of the field and new methods
Technol. Forecast. Soc. Change
Multiple scenario analysis and business planning
Adv. Strat. Manage.
Delphi and cross-impact techniques: An effective combination for systematic futures analysis
Futures
From Anticipation to Action
Cited by (0)
Seyda Serdar Asan is a research assistant in the Department of Industrial Engineering at the Istanbul Technical University and currently pursuing her PhD degree in the Department of Quality Sciences at the Technical University of Berlin. Her main research interests include quality management and supply chain management.
Umut Asan is a research assistant in the Department of Industrial Engineering at the Istanbul Technical University and currently working on his PhD at the Chair of Marketing at the Technical University of Berlin. His main research interests include scenario planning, competence based management and strategic marketing.