Case Studies and the Statistical Worldview: Review of King, Keohane, and Verba's Designing Social Inquiry: Scientific Inference in Qualitative Research

Abstract

Gary King, Robert O. Keohane, and Sydney Verba's Designing Social Inquiry exploits the metaphor of researcher-as-statistician to develop guidelines for conducting social scientific research that are allegedly applicable to all empirical investigations. Their approach has sharp and often unflattering implications for case studies and similar research strategies. Because their statistical worldview is unable to make sense of important aspects of case study research or of the importance that is sometimes attached to the findings of a single case, their argument seemingly casts doubt on the wisdom of producing or consuming such studies.I argue that the foundation of classical statistics and the epistemology of Carl Hempel and Karl Popper is an inadequate and misleading basis for a critical evaluation of case studies. I then present examples of research that are not easily accommodated within the authors' framework and sketch the elements of an alternative epistemological framework rooted in a “pattern” model of explanation. The latter is a standpoint that is much more helpful in understanding and criticizing case studies than the framework presented in Designing Social Inquiry.

Regime Type and Diffusion in Comparative Politics Methodology

“…But the central goal of unifying quantitative and qualitative methodologies has since been embraced by scholars with quite diverse theoretical orientations~Tar-row, 1995;Van Evera, 1997;Munck, 1998;Coppedge, 1999;Laitin, 2003;Brady and Collier, 2004!. Just exactly how this reconciliation of quantitative and qualitative methodological approaches should be effected in practice, however, remains highly contentious. Scholars sympathetic to qualitative methods have called into question the "statistical worldview" espoused by KKV, with its underlying assumptions that causation in social life is generally linear and that units of observation can generally be understood as independent and homogeneous for analytic purposes~Ragin, 1997;McKeown, 1999!. This worldview appears to ignore the intrinsic embeddedness of all observable social phenomena in specific geographical and Acknowledgments: *The authors wish to thank…”

The clean development mechanism and community forests in Sub-Saharan Africa: reconsidering Kyoto’s “moral position” on biocarbon sinks in the carbon market

“…A key challenge with case-study research is the possibility of using such research to generalize about socioeconomic and political phenomenon (Denzin and Lincoln 2000;Niaz 2007). But case studies can make important contributions to socioeconomic and political analysis by identifying causal mechanisms (McKeown 1999). Thus, while the reader is cautioned that the results presented here are not readily generalizable, findings do suggest that one explanation for disputes arising in local communities involved in biocarbon projects is a lack of community control.…”

The Neyman— Rubin Model of Causal Inference and Estimation Via Matching Methods

“…explicit) Bayesianism in their discussion of the relative importance of cases (George and McKeown 1985;McKeown 1999). If one only has a few observations, it is more important than otherwise to pay careful attention to the existing state of knowledge when selecting cases and when deciding how informative they are.…”

“…, nonparametric and does not depend on knowing or estimating the propensity score, but the method is improved when a propensity score is incorporated. Diamond and Sekhon (2005) use this algorithm to show that the long running debate between Dehejia and Wahba (2002;1999;Dehejia 2005) and Smith andTodd (2005b,a, 2001) is largely a result of researchers using models which do not produce good balance-even if some of the models get close by chance to the experimental benchmark of interest. They show that Genetic Matching is able to quickly find good balance and reliably recover the experimental benchmark.…”