Polarization When People Choose Their Peers

Abstract

Processes of polarization have been documented in several applications. Yet most existing theories focus on how herding behavior and the convergence of opinions tend to occur in different contexts. In this paper, we develop a model where agents can correct their heterogeneous initial opinion by averaging the opinions of their neighbors. Our key contribution is letting the network arise endogenously. To do this, we micro–found how individuals optimally choose reference groups and we characterize a dynamic process where the network evolves along with individual opinions. The results show that there are always conditions on the strength of social influence preventing the network from being connected. This causes polarization in the long run. However, polarization can also temporarily appear during the transition to a consensus. We show how each of these cases is tied to a key network statistic: the initial diameter.

Keywords: network formation, naive learning, opinion polarization, echo chambers