Enhancing Causal Inference in Comparisons

Abstract

The previous chapter showed that cross-case comparisons tend to be plagued by manifest problems connected with the generation of unambiguous causal inferences. Being cognizant of these problems, the small-n literature made several recommendations that aim at improving cross-case comparisons (King et al. 1994, 208-13; Lijphart 1971, 686-90). The present chapter serves to discuss the potential of these and additional instruments for the enhancement of cross-case inferences. It starts with the problem of suboptimal comparisons and relates them to the multidimensional nature of cases. It is shown how one can invoke these dimensions in the construction of comparisons in order to approach as closely as possible the envisaged ideal design. The next sections specifically focus on the size of property space and present four instruments for the improvement of suboptimal and ideal comparisons. If the problem is one of a large property space and few cases, two solutions that suggest themselves are a smaller property space and a larger number of cases. The degree to which more than two cases can diminish the inferential intricacies is elaborated in Section 5.2. The three remaining tools for enhanced cross-case comparisons all aim at a reduction of the property space. 1 In Section 5.3, I discuss the role of the level of measurement aggregation. The level of measurement aggregation pertains to the number of categories on which causes and/or outcomes are measured, the basic distinction being between bi-and multicategorical measurement of causes and/or the outcome. An additional tool for reducing indeterminacy is strong theory. In Section 5.4, I elaborate what strong theory is in qualitative case studies and how it can contribute to causal inference. The fifth and final instrument, discussed in Section 5.5, has been largely neglected so far and pertains to the transformation of causes into scope conditions. 5.1 Units of analysis and time: Comparability v. generalizability Every case can be located on a temporal and substantive dimension as well as on at least one additional dimension (see Section 2.1). In this section, I