The Signal from the Noise: Weighting Cluster Analysis to Distinguish Policies with the Strongest Institutional Complimentarity

Friday, June 24, 2016: 10:45 AM-12:15 PM
233 Dwinelle (Dwinelle Hall)
Jason O. Jensen, McGill University, Montreal, QC, Canada
Angela Kalyta, McGill University, Montreal, QC, Canada
Institutional complementarities are at the center of various typologies of modern societies. Varieties of Capitalism in particular argues that the institutions of ownership, workforce training, industrial relations, and even electoral systems can interact positively to enable certain modes of production and ultimately make nation-states and their firms more economically competitive. Countries whose institutional arrangement do not have such complementarities become uncompetitive and must adjust their institutions or fall behind in economic development as their firms fall behind in competitiveness. Other than Bruno Amable’s (2003) seminal The Diversity of Modern Capitalism, the majority of work on Varieties of Capitalism has  been comparably narrow in scope: either qualitative and focused on a few cases, or quantitative and focused on a few institutions. The qualitative work points to key institutional complementarities, but as countries of different political-economy types share many similarities it is difficult to identify which institutions are the most tightly coupled - and are, thus, forced to change when other policies are changed - and which are only loosely coupled – and can thus withstand more diverse policy mixes. Amable’s broader empirical treatment is impressive, but the Principal Component Analysis he employed leads to results which are difficult to interpret as many institutional indicators are mapped onto a few abstract axes. Furthermore, it is not possible to compare indicator level data, making it impossible to discern which institutions are most important in driving the clusters. We are developing a more directly interpretable method for identifying tightly coupled institutional indicators while retaining interpretability. This will more clearly identify real constraints to policymakers and provide an indication of where the efforts of other Comparative Political Economy researchers may best be directed.  

 The core of our approach is a weighted cluster analysis in which we quantify the significance of each institutional indicator by varying the weight of each in iterated analyses. We apply the simplifying assumption that institutions and institutional characteristics that share the most important complementarities are more likely to be strictly clustered across many cases than are institutions which do not, and thus, the most complimentary institutional characteristics will produce tighter clusters when assigned greater weight. We have preliminarily tested our method on 25 institutional indicators in the realms of education, finance, industrial relations, product market regulations in OECD countries. The results indicate that union membership rate and the level of administrative burdens placed on corporations are particularly complimentary, in accordance with qualitative accounts from the Varieties of Capitalism literature. With improvements, the ability to separate the signal from the noise in cluster analysis can help us identify which institutional characteristics are practically most important in distinguishing between types of capitalism.  Finally, as this is a methodological advance, the approach is in principle applicable to analyses beyond the realm of comparative capitalisms.