Outcome and implications Summary

The theory-building case study In this chapter we assume that a theory-building research objective of one of the four types discussed in Chapter 8 has been formulated and that, after it has been assessed that an experiment is not possible, it has been decided that a theory-building case study needs to be designed and conducted. This chapter has the same structure as the preceding ones: ■ 9.1 How to design and conduct a theory-building case study; ■ 9.2 Case Study 5: Building propositions about the kind of company representatives involved in communication with pro- viders of business services by Wendy Van der Valk and Finn Wynstra; ■ 9.3 Methodological reflection on Case Study 5.

9.1 How to design and conduct a theory-building case study

9.1.1 Introduction

This chapter deals with theory-building case study research, which is always a comparative case study. It discusses case selection, the discov- ery of yet unknown concepts, measurement, data presentation, and the manner in which relationships can be discovered in the obtained scores.

9.1.2 Candidate cases

Obviously, candidate cases should be instances from the object of study. The following two principles guide the identification of a smaller set of candidate cases: convenience, and the maximization of the likelihood that an existing relation between the concepts will be discovered. The latter principle requires that one or more sets of can- didate cases be found or constructed which are as similar to each other in all respects apart from the independent and dependent con- cepts from which a proposition should be constructed. A difficulty in achieving this is that in most theory-building research one or both of these concepts are not known. The most efficient way of finding sets of “similar” instances is by identifying small populations in the theoretical domain and selecting cases from such a population.

9.1.3 Case selection

If a small population of similar instances is identified, some instances must be selected from this population in such a way that instances dif- fer as much as possible in terms of the value of the known concept. The number of cases to be selected depends on the type of the con- cepts that are known i.e. concept B in Figure 8.2; concept A in Figure 8.3; concepts A and B in Figure 8.1 and the known variation of the value of these concepts in the list of candidate cases. For instance, if we are interested in discovering as yet unknown success factors for a proj- ect Figure 8.2, it depends on the type of concept B how many cases and what type of cases should be selected. If B is dichotomous pres- ence or absence of success we could begin with just a couple of instances of each possible value e.g. success and lack of success. If B is measured as a rational or interval variable e.g. if success is measured in amounts of money earned, variation in the value of that variable should be maximized. If we are interested in discovering as yet unknown effects of a given condition or intervention A Figure 8.3, it depends on the type of concept A how many cases and what type of cases should be selected. If A is dichotomous, we could begin with just a couple of instances of each possible value e.g. presence or absence of condition A. If A is measured as a rational or interval variable e.g. the size of the workforce or the amount of money spent, variation in the value of that variable should be maximized. If both the concepts A and B are known Figure 8.1, it is recommended to select cases in such a way that the variation in the value of both concepts is maximized.