Forms and Problems of Comparisons

Abstract

In principle, comparisons can be made on the cross-case level, the within-case level, and both simultaneously (see Section 1.4). A within-case comparison is appropriate for discerning whether the causal mechanism and causal processes are similar in the two (or more) cases at hand. In a cross-case comparison, on the other hand, one is determining the nature of the causal effect of a given cause. The two generic forms of comparisons are closely intertwined because it is useful to know whether the causal effect is underpinned by a mechanism and, in addition, what causal effect is produced by that mechanism. In this and the next chapter, the focus is on the cross-case level because of the special issues involved in these comparisons. The following chapters expand into process tracing and the analysis of causal mechanisms. 1 Cross-case comparisons have been subject to heated debate for decades now. This chapter builds on this debate and elaborates the logic and problem of comparative case studies by explicit linkage of the three research goals with the correlational and set-relational view on cross-case relationships. The arguments extend to case studies for which the cross-case level is the theoretical end and the empirical means for process tracing. For presenta-tional purposes, I limit the discussion to the purpose of generating cross-case inferences. This is not to deny that sound comparisons are vital for within-case analysis. On the contrary, if one wants to learn how a sufficient condition produces the outcome, one needs to be sure that the condition is sufficient in the first place and not part of a conjunction or an INUS condition. In light of the discussion of case selection in the previous chapter, all the points made in this chapter can be easily extended to case studies having the within-case level as theoretical end. 2 Throughout the discussion of cross-case comparisons, an important distinction concerns the differentiation between ideal , or optimal , and imperfect , or suboptimal , comparisons. A design is optimal if the observable cross-case scores match the pattern that one wanted to create by the purposeful choice of cases. If one wants to compare two cases that differ