Introduction

Social scientists and policymakers alike rely heavily on statistical analyses to seek patterns in data collected from our changing world. Collecting information about the complexities of the world surrounding us and understanding the links between intricate sources of information can be a daunting task. Hence, over the last few years, numerous actors in the policy sphere have turned to statistics tools to inform their decision-making and provide their decision processes with credibility. Besides regressions, factorial analyses and other multivariate statistical techniques that are widely adopted in the field of political science and policy studies (Pennings et al, 2006), less known statistical tools can provide different and complementary insights. This chapter focuses on cluster analysis, one of the less adopted, yet powerful, statistical techniques that can be applied in policy making processes and the social sciences generally. The utility of this method will be discussed by various examples in the context of policy research.

Compared with typical regression techniques, cluster analysis deals with sorting data and seeing patterns that are data-based and can be less assumption-driven. It is an explorative technique in nature; thus, you can go on a more open search to your questions compared with regression analyses, where the data generation process must be assumed beforehand. Cluster analysis is a bottom-up, grounded analytic form of theory building in that it seeks connections between data through the power of careful statistical analysis rather than through preformed theories. Compared to standard regression analysis, it is an inductive rather than deductive hypothesis-testing approach. The value and primary purpose of the inductive approach is to allow research findings and potentially policy insights to emerge from the frequent, dominant or significant themes inherent in raw data, without the restraints imposed by structured theories or methodologies. With other approaches, there is a danger that key themes are often obscured, reframed or left invisible because of the practices or assumptions of the analyst and the demands of experimental and hypothesis-testing research. So, cluster analysis provides an inductive starting point but from a quantitative and replicable base. It brings into play qualities only normally associated with more bottom-up qualitative analysis.