Modeling Dynamic Ideological Behavior in Political Networks

Abstract

In this article, we model and analyze the dynamic behavior of political networks, both at the individual (party member) and ideological community levels. Our study relies on public data covering 15 years of voting sessions of the House of Representatives of two diverse party system, namely, Brazil and the United States. Whereas the former is an example of a highly fragmented party system, the latter illustrates the case of a highly polarized and non-fragmented system. We characterize the ideological communities, their member polarization and how such communities evolve over time. Also, we propose a temporal ideological space model, based on temporal vertice embeddings, which allows us to assess the individual changes in ideological behavior over time, as expressed by the party members' voting patterns. Our results reveal very distinct patterns across the two case studies, in terms of both structural and dynamic properties.