Classifying Cases in Federal Studies. An Illustration of why Political Scientists should do more Cluster Analysis

Abstract

Typologies are widely used in research on federalism, e.g. to distinguish dual from cooperative or coming-together from holding-together federations. More general, ideal types, archetypes and categories are frequently used in political science research to define concepts and classify cases. As recently as in 2014, Filho et al. pointed out that Cluster Analysis is still hardly used when it comes to developing typologies in political science. Rather, political scientists rely on more intuitive methods or factor analysis. Our paper argues that Cluster Analysis is of great usefulness because it a) focuses on the relationship between cases and not variables and b) draws on empirical data when identifying the clusters. This paper proposes to apply this fruitful approach to the field of federalism to exemplify its major heuristic potential. Furthermore, we emphasize that testing the secondary validity is a crucial step. Our paper provides two original examples from comparative federal politics and public management that illustrate the strength of Cluster Analysis both in testing and generating hypotheses through the establishment of typologies. For both examples, the validity of the Cluster Analysis is tested by checking for correlations between the clusters and the distribution of power. Hence, the typologies established through Cluster Analysis not only define our respective dependent variables related to aspects of intergovernmental coordination within federations and the normative density of evaluation clauses in the Swiss federation, but also offer strong insights in issues of regional autonomy.