Working Paper
Democracy clusters and patterns of inequality

A k-means approach

This paper employs k-means clustering, a multidimensional pattern recognition method, to categorize countries using information from five Varieties of Democracy indices. K-means country clusters are similar but not entirely identical to both k-medians clusters and arbitrary groups formed using only one measure, namely the liberal democracy index.

Simple correlations with different inequality measures yield varying results. Despite having higher absolute inequality, country clusters with extremely high levels of democracy are characterized by lower relative inequality. They also have the lowest ratio of the income share of the richest 10 per cent to that of the poorest 40 per cent. K-means clusters with low- and medium-quality institutions generally do not differ much in terms of inequality outcomes.