The Effect of Network Structure on Political Conversations on Twitter

Abstract

Twitter has become an increasingly important tool for politicians to communicate with voters, and also a popular way for voters to discuss their views about politicians. In this paper, I examine the effect of ideological segregation on the exposure of Twitter users to like minded and opposing views about political candidates in the context of the 2016 presidential election. While conservatives often reply to Democratic candidates and liberals often reply to Republican candidates, replies originating from cross-type users receive less exposure than replies originating from same-type users. Furthermore, users exposed to a reply tend to be of the same type as the replier. Instead of being exposed to a representative sample of replies, users are exposed to a biased set of replies from ideologically similar users. I present a theoretical model that predicts this effect in the context of homophily and then prove these results empirically.