Clarke / TAKING CONTEXT SERIOUSLY
importance, whether using inductive or deductive methods. Nor are case studies necessarily the best means of unpacking these relationships: In illustrating the operational realities of seemingly formal electoral structures, the authors appear to resort to personality factors and some implicit (and unstated) criteria to identify “strong” and “weak” mayors in their cities rather than the institu- tional capacities usually referenced by those categories (Reese & Rosenfeld, 2001).
Some of these difficulties may be traced to the lack of documentation and the continued refer- ences to the authors’previously published or forthcoming work. But it is a disservice to the authors and the readers to present research aimed at challenging conventional wisdom without providing any of the means for evaluating that challenge. There are references throughout to results from regression analyses, cluster analyses, and an appendix with factor analysis loadings, but the argu- ment is not directly grounded in these empirical findings. As a result, more questions are raised than answered: Did they actually utilize cluster analysis—a case-wise method for analysis—and are their community cases representing distinctive clusters? Or did they use factor analysis—a variable-wise method of analysis that undermines their concern with contextual features?
Nevertheless, Reese and Rosenfeld (2001) raise some interesting issues meriting further discus- sion. Although their critique of extant models is rather familiar, it points to larger issues of theory development and causality that are not often addressed by scholars of local economic develop- ment. If we are aiming for some ever grander causal theory of local economic development policy making, then the weak and widely varying coefficients produced by mostly incommensurable quantitative studies are important. They signify how far we are from that goal—mis-specification is indeed worth worrying about to the extent that it obscures causality. But if we seek to explain, at best, clusters or classes of events, problems, or dynamics, we need analytic frameworks to help us understand those patterns. Rather than assuming a world without multicollinearity and featuring “arrow through the heart” notions of causality, we would seek out anomalous cases, employ counterfactuals and pattern-matching methods, and delve into patterns of contingent causality. As I tell my students, “The world is multicollinear; get over it”—the appropriate goal is developing ana- lytic skills capable of uncovering complex patterns of contingent causality and communicating their meaning effectively to other scholars and practitioners.
Reese and Rosenfeld (2001) seem more comfortable with this contextual goal as well, but com- bined approaches alone are not necessarily the answer. Adding more methods does not always bring in more context. Although their narratives on U.S. and Canadian towns are interesting, it is hard to characterize them as case studies. They illustrate the points being made, but we don’t really know the criteria used in characterizing the different settings and outcomes, nor do we understand the evidentiary logic that makes the arguments credible. There are limits to doing so in an article format, and the larger work from which this work is drawn may provide a better sense of these judg- ments, but we need to consider case studies more seriously if we are to take them seriously. To be persuasive, there needs to be transparency in the case selection and logic of comparisons, as well as systematic, rigorous comparisons across cases or between cases and theoretical expectations. In a recent debate on the characteristics of a good case study (Symposium, 2000), the extent to which there should be compatible orienting concepts employed and the need for explicit frames of refer- ence in comparing cases were key issues. Furthermore, systematic approaches to integrating and synthesizing case studies open up opportunities for refining analytic frameworks or developing theoretical arguments that can extend beyond the cases in question. Ragin’s (1994) qualitative comparative analysis (QCA) strategies, for example, are one of several means of methodically inte- grating and synthesizing multiple cases. Ragin proposed an emphasis on approaching causality in terms of the necessary and sufficient conditions to account for an outcome rather than the probabil- istic and additive notion of causality underlying conventional statistical approaches.
Reese and Rosenfeld’s (2001) turn to a civic culture argument as the solution to the limits of conventional wisdom is promising but problematic. Most notions of political and civic culture remain captive to the tendency to rely on cultural features as residual concepts that shape every- thing and, in that sense, explain nothing. Here, the admonitions to incorporate civic culture into the study of local economic development are reasonable but come with few empirical designs for doing so. As is often the case, it is not clear how to avoid treating cultural features as residual con- cepts without making them into causally prior factors. Fortunately, a number of recent studies offer
. . . the admonitions to incorporate civic culture into the study of local economic development are reasonable but come with few empirical designs for doing so.