Case Studies and Theory Development in the Social Sciences - Alexander L. George [92]
In sum, DSI’s proposal for achieving scientific inference in qualitative research fails to address squarely the need to ensure that observations imputed to a theory achieve quality, validity, and relevance. In a brief discussion of a hypothetical example it is stated that observations “should be used even if they are not the implications of greatest interest.”364 On the other hand, the authors of DSI do make passing reference to the need for valid observations: “there are situations where a single case study (as always containing many observations) is better than a study based on more observations, each one of which is not as detailed or certain.”365 More generally, a serious tension exists between DSI’s emphasis on the desirability of increasing “leverage”—i.e., explaining as much as possible with as little as possible—which encourages listing all possible observations, and the importance of ensuring that the observations imputed to a theory achieve the quality, validity, and relevance needed to assess the theory.366
DSI also misunderstands process-tracing, which it incorrectly represents as simply another way to obtain more observable implications of a theory. DSI’s major interest in briefly discussing process-tracing toward the end of the book is to label it as consistent with the authors’ own approach. “From our perspective,” they state, “process-tracing and other approaches to elaborations of causal mechanisms increase the number of theoretically relevant observations.”367 This overlooks the fact that within-case observations and methods for analyzing them in their particular historical context are different from cross-case comparisons and methods, which necessarily simplify or omit the contexts of the cases studies. In this context DSI refers briefly to the “within-case” approach of Alexander George and Timothy McKeown. However, it is mislabeled as “within-observation explanation” and asserts that it should be regarded as nothing more than “a strategy of redefining the unit of analysis in order to increase the number of observations.”368
Working with this mischaracterization of process-tracing, DSI then concedes that it and related efforts to get at “psychological underpinnings of a hypothesis developed for units at a higher level of aggregation are very valuable approaches.” This is coupled with an insistence that process-tracing and related approaches should be regarded as “extensions of the more fundamental logic of analysis we have been using, not ways of by-passing it.”369
We instead characterize process-tracing as a procedure for identifying steps in a causal process leading to the outcome of a given dependent variable of a particular case in a particular historical context. As Sidney Tarrow pointed out in his commentary on DSI, noting that although it refers to process-tracing favorably, it errs in assimilating it “to their favorite goal of increasing the number of theoretically relevant observations.” That is, DSI errs in regarding each step in a causal process as nothing more than an observable implication that can be attributed to a theory. In process-tracing, as Tarrow correctly notes, the goal is not, as DSI would have it, to aggregate the individual steps in a causal chain “into a larger number of data points but to connect the phases of the policy process and enable the investigator to identify the reasons for the emergence of a particular decision through the dynamic of events.” As Tarrow writes, “process-tracing is different in kind from observation accumulation and is best employed in conjunction with it”—as was indeed the case, for example, in the study by Lisa Martin (1992) that DSI cites so favorably.370
To be sure, the