If no concept is known at the beginning of the study as depicted in Figures 8.4 and 8.5, cases cannot be selected on the basis of the varia-
tion of these concepts and must, therefore, be selected more or less randomly.
Box 12 Michael Porter’s case selection
Michael Porter’s theory on The competitive advantage of nations 1990 is based on case study research. Porter and his team wanted to find conditions for a nation’s industries
that could explain the success of a nation’s global competitiveness. The theory focused on the strategies of firms rather than the strategies of nations, as “firms, not nations,
compete in international markets”. The team selected, from ten important trading nations, the companies that were internationally successful the dependent concept.
Then they identified the determinants that could explain the nation’s success the independent concepts.
Porter and his team found four determinants four points of a “diamond” of a nation’s success: 1 the nation’s position in factors of production such as skilled labour or infra-
structure; 2 demand conditions, the home-market demand for the industry’s product or services; 3 related and supporting industries, the presence or absence in the nation
of supplier industries and other related industries that are internationally competitive; and 4 firm strategy, structure, and rivalry, the conditions in the nation governing how
companies are created, organized, and managed, as well as the nature of the domestic rivalry. These four determinants are necessary for achieving and sustaining competitive
success, or as Porter 1990 : 73 puts it: “Advantages throughout the ‘diamond’ are necessary for achieving and sustaining competitive success in the knowledge-intensive
industries that form the backbone of advanced economies”.
Porter’s case selection procedures are problematic for two main reasons. One is that, by not including non-successful companies or nations in his study, Porter is not able to
distinguish between necessary and sufficient conditions on the one hand, or between necessary and trivial conditions on the other hand. If, for instance, the factors found could
exist in any company or sector in an industrialized country, including non-successful ones, this would make the discovered determinants not less “necessary” but it would
make them trivial for policy. Apparently, Porter implicitly relies on his readers’ knowledge about conditions in non-successful companies and nations. The second reason is that
this form of case selection prohibits finding probabilistic relations. If Porter had found only one single instance without the “necessary” determinants, he would not only have
failed to identify the necessary condition but would also not have been able to find another type of relation between determinants and success. Porter’s case selection
procedures, thus, were appropriate only for finding candidate necessary conditions and he was lucky to find them.
9.1.4 Extracting relevant evidence
If the theory-building case study begins with a conceptual model with an unknown concept as in Figures 8.2–8.5, candidate concepts must
be found in the selected cases. If we start with known concepts and only need to find out what type of relation between these concepts
should be formulated in the proposition as in Figure 8.1, this phase can be skipped and the researcher can immediately start measuring
the concepts as described below in 9.1.5.
There is no specific “method” for how candidate concepts should be found in a theory-building case study. In principle “everything goes”,
just as in other types of exploration described in Chapter 3. This exploration can take place in only one case, or in more than one case,
or in all selected cases at the same time. There is one widely known method of discovering concepts through the comparison of data from
multiple cases, Grounded Theory GT. The GT literature, particu- larly the widely used textbook of Strauss and Corbin 1998, describes
in detail how a concept can be discovered by a “coding” data in a procedure that is called “open coding”, and b comparing these
codes between different instances.
The result of this stage of the theory-building case study is a candi- date concept for the initially unknown concept in the conceptual
model with which the study started, as depicted in Figures 9.1 and 9.2. Although the precise process of discovering concepts candidate
causes and effects and its quality criteria cannot always be described clearly in exploratory activities, at some point such concepts emerge as
Candidate concept A
Concept B ???
Independent Dependent
Figure 9.1
Conceptual model with candidate
determinant
Figure 9.2
Conceptual model with candidate
effect
Concept A Candidate
concept B ???
Independent Dependent
an outcome. In our approach to the theory-building case study, the lack of criteria for the quality in this exploration activity is balanced by
an emphasis on quality control after a candidate concept has been “dis- covered”. This is discussed below.
9.1.5 Coding
If theory-building research stopped at the point depicted by Figures 9.1 and 9.2, and if the resulting candidate concepts in a publication were
offered to other researchers for testing, the study would hardly qualify as research and could better be called a form of “intensive exploration”.
As we discussed in Chapter 8, we consider it essential to good theory- building research in contrast to mere “exploration” that the emerging
proposition is proven to be true in the instances studied and that, thus, the candidate concept is measured in a valid and reliable way in these
instances. A first necessary step is that the concept is defined precisely after its “emergence”. This step is not different in principle from how
definitions of concepts are usually developed, i.e. if the researcher wants to define a concept that has not been “discovered” in theory-building
research. The usual criteria such as precision and non-ambiguity apply.
Grounded Theory argues that, if a theory is “discovered”, the defini- tion of the concept should be “grounded” in the data collected in the
study. For instance, Strauss and Corbin 1998 describe how a concept that is discovered in “open coding” can be refined and defined in a
next step of coding which they call “axial coding”. In our view, such a grounding of a definition is not a requirement for good theory-building
research. However, an advantage of axial coding is that, when a concept is defined, its value in the different cases is already validly “measured”
because the GT result consists of a definition of the concept with references to the data in which it was “grounded”.
If a definition of a concept is derived in another not “grounded” way, or if we start with known concepts as in Figure 8.1, a next neces-
sary step in the research is to develop a valid and reliable measurement instrument. Procedures for measurement are discussed in Appendix 1
“Measurement”.
9.1.6 Data presentation
The result of a successful measurement is that the scores of the rele- vant concepts are known for each case. These scores can be presented