Comparing Measures of Political Similarity

Signorino and Ritter developed and advocated a new measure of the political similarity of states (S). They demonstrated logical flaws associated with the common Kendall's Ï„ b -based alliance similarity measure and showed that using S may yield quite different similarity estimates. But without a broader empirical comparison and analysis using S, the full empirical ramifications of this difference is not known. In this study, a comparison of S versus Ï„ b is conducted over a wide spatial and temporal domain, examining the relationship between S, Ï„ b , expected utility scores, and conflict. Despite significant positive correlations, important differences in the distribution of S and Ï„ b -based measures of alliance similarity are found. It should not be assumed that the measures are substitutable. Reanalysis of an important expected utility theory of war shows a stronger relationship between equilibrium predictions and conflict when S is used versus Ï„ b .