Structural analysis seeks to represent the ‘system’ by highlighting key variables, which (potentially) influence the problem under study, with the help of a cross-impact matrix (also called a structural analysis matrix). In the cross-impact matrix, the variables are placed in rows and columns, in order to work out systematically whether there are any causal relationships between them.
Structural analysis can be used when a problem is so complex that it is important to ensure no key variables (internal variables, external variables or major actors) are overlooked or to create a common culture when the working committee is heterogeneous or lacks in-depth knowledge of the problem. It is also a useful tool when you want the foresight exercise to last a long time for whatever reason (common reflection, communication, etc.).
It ensures a certain level of quality in the results stemming from the matrix once it has been processed. For each variable it gives:
• an influence index, which measures the intensity with which a variable acts upon the system
• a dependency index, which measures the intensity with which each variable is affected by the system
The structural analysis method can be used in a Foresight working committee to create a common culture and to reach a consensus on the variables driving the problem under study. It allows the working committee to build a solid basis to move on to the next stage of the scenario method. Structural analysis is a tool to pool ideas by reducing the influenced of biased opinions and to structure collective ideas.
Sometimes structural analysis is an end in itself. It can give decision makers as much information about how the working committee perceives reality, and therefore about the committee itself, as about the system under observation.
However, it may be not necessary if people have an in-depth knowledge of the subject.
Structural analysis is carried out by a working committee made up of actors and experts from the field under study, but this does not preclude calling on external advisers.
A group of about ten people, who have previously taken part in listing and defining the variables, fills in the cross-impact matrix over a period of two to three days.
A technical committee of 2 to 3 people helps the working committee by designing meetings (proposal of the first list of variables, writing of certain variable ‘index cards’…). If there is a consultant involved, he/she will be a member of the technical committee.
Listing variables: The first step consists of identifying all kinds of variables, which do or may influence the problem under study and its environment (internal as well as external variables). A list should then be drawn up of the internal and external variables noted, with some consistency (e.g. do not mix up run-of-the-mill variables with highly specific ones, and with as accurate a definition as possible).
One option to consider is whether or not to combine the lists of actors and variables. If you have a separate list for actors, you will have to look separately at the influence of each actor on each other actor and at the position of each actor regarding each variable.
Experience shows that the list of all variables does not generally exceed 70 or 80 variables, often classified in different thematic groups within the internal and external categories.
The list can be drafted by one person; however, to avoid excessive subjectivity, it is usually compiled by a multi-disciplinary working committee made up of actors and experts. The list may lead to further documentation, interviews with experts and various other consultations.
For variables that are not well known, we can use variable ‘index cards’, defined by the members of the working group and then validated collectively.
Template of a variable ‘index card’ / Variable ‘Index card’
Name of the variable:
Category (Internal/External, Technical Variable/Environmental Variable…):
Description of the relationships between variables: The second step consists of analysing the relationships between variables, often with the help of a cross-impact matrix in which the variables are placed in rows and columns, in order to work out systematically whether there are any causal relationships between them.
The matrix is completed with qualitative information. For each pair of variables, the following questions are asked: Is there a relationship of direct influence between variable 1 and variable 2? If there is not, it scores 0. If there is a low direct influence, it scores 1; if there is a medium direct influence, it scores 2; if there is a high direct influence, it scores 3; and possibly if there is a potential direct influence, it scores 4.
The filling-in phase helps to pose n X n-1 questions for n variables (approximately 5000 questions if there is 70 variables), some of which would have been evaded if such a systematic and thorough investigation had not been made. This questioning procedure not only enables you to create a common language within the group; it also allows you to redefine the variables and so tends to make analysis of the system more accurate.
A group of about ten people, who have previously taken part in listing and defining the variables, fills in the cross-impact matrix over a period of two to three days. The matrix is filled in on the basis of consensus between the members of the group. Sometimes, it is necessary to use interviews with experts, documentary research or even specialised studies to fill in the matrix.
In general, the matrix is filled in on the paper so that the entire group can see it and one member of the technical committee seizes the data in the computer for data processing.
Direct influence List of Internal Variables 1, 2, 3, 4, 5, 6 List of External Variables 7, 8, 9, 10, 11, 12…
List of Internal Variables 1, 2, 3, 4, 5, 6
List of External Variables 7, 8, 9, 10, 11, 12…
Identification of the key variables: Once the matrix has been filled in and processed, for each variable it provides:
• an influence index, which measures the intensity with which a variable acts upon the system (by adding rows)
• a dependency index, which measures the intensity with which each variable is affected by the system (by adding columns)
The variables can then be represented in an ‘influence-dependency’ graph, which is a quick way of telling which variables are the driving variables in the system being studied.
At the end of this stage, we have a fairly good idea of the key variables and main actors that determine how the system develops. If structural analysis is integrated in the scenario method, we can now move on to the next stage.
Structural analysis is a time-consuming process and several months will be needed (about 6 months). It is carried out by a working committee made up of actors and experts from the field under study, but this does not preclude calling on external advisers. A technical committee of only 2 or 3 persons designs the work of the working committee.
A group of about ten people, who have previously taken part in listing and defining the variables, fills in the cross-impact matrix over a period of two to three days. It is sometimes necessary to draw upon interviews with experts, documentary research or even specialised studies to fill in this type of matrix.
The time required to use this tool depends on the knowledge of the subject under study by the working committee, on the size of the matrix (the number of variables x the number of variables) and on the size of the group which fills in the matrix.
Naturally, the duration of a structural analysis depends also on the working committee work load and the time devoted to the task by each one.
To help use this tool and avoid drawing up the matrix and computing the scores, programs called ‘MICMAC’ and ‘MACTOR’ can be downloaded from the LIPSOR (Laboratory for Investigation in Prospective Strategy and Organization) website [www.cnam.fr/lipsor/].
The output is the identification of key variables, that is to say, those essential to the system’s development and establishing their hierarchy. It gives a representation of the system.
No special skills are required to use the matrix. However, if a lot of calculations are made from the matrix, the results may be more difficult to explain.
Pros and cons
Structural analysis is a structured method which gives a working committee a common culture and approach to the problem being studied. Its main advantage is that it stimulates thought and generates ideas among group members, thus encouraging them to think about counter-intuitive aspects of system behaviour. Participants should not be taken literally but should be made to think. Obviously, there is no single official reading of the influence-dependency graph and it would be preferable that the group forms its own opinion.
On the downside, its limitations relate to the subjective nature of the list of variables drawn up during the first phase, similar to that of the relationship between the variables. This subjectivity comes from the well-known fact that structural analysis is not a reality in itself but a means of looking at reality.
A second classification of variables can be obtained after increasing the power of the matrix and recalculating influence and dependency indexes. These new indexes uncover certain variables which, because of their indirect actions, had remained hidden (see the works by M. Godet).
Structural analysis is often used to carry out the second stage of the scenario method which consists in constructing the system and identifying key variables. In the scenario method, the identification of driving variables with the help of structural analysis allows the working committee to move on to the next stage: gathering data on the driving variables and drafting hypotheses on their evolution. The driving variables will represent the components of the system used in the morphological analysis for the construction of scenarios.
A lot of calculations can be made from the matrix and it can be processed in more sophisticated manners (see the works by P.F. Ténière-Buchot, by J.W. Forrester, by J.F. Lefebvre, by L. Dirn).
This method could be appropriate if you need to answer questions such as:
• What is the objective of the group? To understand a problem? To build scenarios? To communicate on a subject?
• Is the problem complex? Do members of the group know the problem in depth? Do they share a common culture on the problem?
• Do we have enough time and budget to run such a long exercise?
• Will the working group devote enough time to the exercise?