One of the motivating factors behind the rising popularity of mixed methods research designs in political science is found in the possibility of achieving both the in-depth focus on causal mechanisms of case studies, and the potential for broad-spanning inference of large-N analyses. However, scepticism surrounds the popular regression-based nested analyses designs suggested by Evan Lieberman. In particular, the selection of cases for case studies in this approach is subject to criticisms, which question the extent to which case study findings can be generalised to the population of cases.
Taking Qualitative Comparative Analysis (QCA) as a point of departure for nested analysis can contribute a viable solution to this problem. However, some questions remain concerning methodological concepts as well as the concrete procedures necessary in order to achieve generalisable conclusions. This paper attempts to address some of these issues by considering: (1) which conception of causal mechanisms is most useful; (2) how the mechanisms should be tested empirically; (3) the utility of negative cases; and (4) which and how many case studies should be performed in order to enhance the credibility of cross-case inference.
The paper argues that wide-spanning conclusions are best achieved by: Constructing the population of cases and the causal mechanisms carefully; including analyses of negative cases; and employing a mixture of congruence analysis and process tracing for the purpose of case study analysis. If these suggestions are adhered to it will be possible to sustain bounded inferences – and determine where their boundaries lie.
