3. Research Design

Each research strategy has different characteristics regarding the kind of research question addressed, the extent of control a researcher has over behavioral events, and the focus on contemporary as opposed to historical events.

A recent example of a study which combines methodologies is undertaken by Schwaninger & Groesser, (2012). They have used a systemic-cybernetic approach to understand an organizational change process and the organization’s path dependency in relation to organizational closure and autopoiesis. By this means, they are able to reap insights “that are more realistic by using methods that penetrate façades” (Starbuck 2010: 1398).

The differentiation between messy problems and stylized problems is much discussed in the operational research literature (Ackoff, 1979; Jackson, 2006; Mingers, 2009; Rosenhead & Mingers, 2001); it is related to the difference between “soft-OR” and “hard-OR” (Lane, 2000; Lane & Oliva, 1998; Paucar-Caceres & Rodriguez-Ulloa, 2006), and to the divide between qualitative and quantitative research (Müller-Merbach, 2007; Srnka & Koeszegi, 2007).

Mingers (2001) distinguishes four types of multimethodologies: (1) methodology combination, (2) methodology enhancement, (3) single-paradigm multimethodology, (4) and multi-paradigm multimethodology. Until today no clear categorization exists if a multimethodology belongs to the single- or the multi-paradigm category. This distinction is not relevant here. It is most fertile for this study when both methodologies are used in the combinatory mode; i.e., the strengths of the methodology of grounded theory and system dynamics modeling are leveraged optimally by using their respective methods most beneficially.

With a hint to Kuhn (1996), one might argue that the methodologies used in this research belong to the same research paradigm. First, both methodologies are able to develop dynamic models and theories. And second, both methodologies explicitly use the nation of iteration for developing their models or theories. A difference might be that theory of the type of system dynamics are formal, quantified, and mathematically closed (Schwaninger & Groesser, 2008), whereas theories generated with the grounded theory methodology are formal, might be even quantified, but lack the characteristic of a mathematical closeness. However, this is one of the aspects where both of the selected methodologies can complement each other.

A methodology which might be used as an alternative to grounded theory is “Soft Systems Methodology” (SSM) (Checkland, 2008). This methodology was not selected for three reasons: first, the effort to conduct a complete analysis with SSM is prohibitively high compared to the resources available. Second, the experts of the research project (more details in Sect.​ 4.​3) participated on a voluntary basis; it is assumed that is would be difficult to convince participants to engage in a full SSM analysis without compensation. And third, the methodology of SSM is a powerful one, which requires a significant level of innovativeness of the participants. In achieving such a degree, other characteristics of the participants which are more relevant to this study’s success might suffer; for instance, it was assumed that it would not be possible to convince professional building owners with a lower degree of innovativeness to participate in the study.

The methodology of grounded theory was originated by Glaser and Strauss in the 1960s (Glaser & Strauss, 1967a). In the preceding years, two different schools of grounded theory have emerged (Punch, 2005): the Glaserian approach (Glaser, 1992) and the Straussian approach (Strauss, 1987); Juliet Corbin has supported Strauss from the early 1990s (Strauss & Corbin, 1990). Glaser’s approach emphasizes that the coding scheme—a coding heuristic—must emerge exclusively from the analysis of empirical evidence. Strauss and Corbin favor the application of a predefined coding scheme with such categories as phenomena, context conditions, causal relations, intervening conditions, actions strategies, and consequences (Corbin & Strauss, 2008; Strauss & Corbin, 1990). In addition, Strauss and Corbin allow for the explicit appreciation of existing theories in advance of the coding and analysis; in Glaser’s version, existing models and theories come in only after the analysis phase. The discussion of the unbiased perception of empirical evidence will not be elaborated here; the interested reader turn to further references (e.g., Glaser, 1978). I thank Prof. Dr. Katja Mruck and Prof. Dr. Günter Mey (both at the University of Berlin, Germany) for pointing out this small but significant difference between the two schools.

Other methods, e.g., writing memos, are not elaborated here. See Strauss and Corbin (1998) for details.

Theoretical sampling is also known as “handy sampling” (Carberry, 1971).

Other purposes of the application of simulation are prediction of future conditions, performance improvement, training of participants, entertainment, education of students, and proof of conceptual elaborations (Axelrod, 2005).

The deterministic nature of the differential equation-based methodologies can be diluted by the introduction of stochastic elements. I have applied this where necessary for model validation and policy analysis.

The reflection about the adequacy of system dynamics as simulation methodology for this research resulted in the consideration of agent-based modeling (Epstein & Axtell, 1996). Agent-based modeling shares many of the strengths of the system dynamics methodology (Rahmandad & Sterman, 2008); but it lacks, among other things, the important characteristics of white-box modeling and the full-capacity of an elaborated validation methodology. Research has attempted to combine both methodologies to overcome their respective shortcomings (Größler, Stotz, & Schieritz, 2003; Schieritz, 2004; Schieritz & Milling, 2003; Scholl, 2001). Given that this research is still methodologically underdeveloped, I have rejected using agent-based modeling here.

Sterman (2000) provides a basic compendium on the system dynamics modeling approach.

These methods could be clustered differently. Behavior-over-time graphs, reference modes, and causal loop diagrams are methods of system dynamics methodology which formalize and represent system content; expert workshops, group model building, and interviews are foremost procedural methods. Because of the marginal additional insights for this study, I do not elaborate these dimensions.

One alternative to the cognitive mapping interview style would be narrative interviews using interview guidelines. I see this method as being unable to deliver the additional benefits of the cognitive mapping interview method compared to the gain in reliability because of the ex-post coding of the material. I thank Dr. Jan Kruse (University of Freiburg, Germany, specialist in interview methods) for pointing out that the mindset of the interviewer in conducting narrative interviews influences the results regardless of the interview method used (personal communication).

I thank Ms. Stephanie Geisshüsler for her support during the cognitive mapping interviews. In addition, I thank the Interfaculty Centre for General Ecology at the University of Bern for financial support that enabled us to conduct these interview situations.

http://​www.​banxia.​com/​dexplore/​index.​html. I thank the Interfaculty Centre for General Ecology at the University of Bern for granting me access to this software.

The summary of the major interviewing principles is provided upon request.

More formalized: δY/δX > 0 (positive link); δY/δX < 0 (negative link).

The method is comprehensively described in Sterman (2000, especially Chap.​ 5).

A technique which is similar to cognitive mapping and causal loop diagrams is the “structural elicitation approach” (Spevacek, 1999). I did not choose this method because the method does not have advantages relative to the methods selected, and has most often been used only in psychology, not in the management sciences.

The interview guidelines for the interviews phase are included in the Appendix (Chap. VII). The original language of the guidelines was German. The guidelines have been translated into English for this book; the interview documents are in German.

I thank the project team which consisted of Dr. Silvia Ulli-Beer, Dr. Susanne Bruppacher, and Prof. Dr. Ruth Kaufmann-Hayoz. The design of each of four workshops was customized to the current state of the research project. The details of the workshops are not provided here, to protect the personal identities of participants. Contact the author for further information.

Already Forrester (1961) has acknowledged the importance of facilitated decision modeling. I thank Prof. Dr. David Lane (London School of Economics, UK) for bringing this to my attention.

Schwaninger (1997) has established a “paradoxes framework” which could also be used to position the book in the field of philosophy of science. He differentiates the dimensions of the type of modeling (from qualitative to quantitative modeling), the level of rationality (from conceptual to communicational rationality), and the Weltanschauung (from objectivistic to subjectivistic worldview).

For reasons of simplification, I consider the book as being one integrated entity for which a position on the four continuums can be defined, even though different methodologies and methods are used: qualitative explorative interviews, statistical data analysis, grounded theory, and simulation modeling.