Kantian fractionalization predicts the conflict propensity of the international system

Abstract: The study of complex social and political phenomena with the perspective and methods of network science has proven fruitful in a variety of areas, including applications in political science and more narrowly the field of international relations. We propose a new line of research in the study of international conflict by showing that the multiplex fractionalization of the international system (which we label Kantian fractionalization) is a powerful predictor of the propensity for violent interstate conflict, a key indicator of the system's stability. In so doing, we also demonstrate the first use of multislice modularity for community detection in a multiplex network application. Even after controlling for established system-level conflict indicators, we find that Kantian fractionalization contributes more to model fit for violent interstate conflict than previously established measures. Moreover, evaluating the influence of each of the constituent networks shows that joint democracy plays little, if any, role in predicting system stability, thus challenging a major empirical finding of the international relations literature. Lastly, a series of Granger causal tests shows that the temporal variability of Kantian fractionalization is consistent with a causal relationship with the prevalence of conflict in the international system. This causal relationship has real-world policy implications as changes in Kantian fractionalization could serve as an early warning sign of international instability.