Examining Opinion Polarization on Social Media Platform Using a Natural Experiment

Abstract

In recent years, there is a heated discussion on opinion polarization on social media platforms. In this paper, we formulate a hierarchical Bayesian learning model to investigate the dynamics of individual opinion formation, particularly opinion polarization, on a social media platform where individuals are exposed to two sources of social influence: preceding peer microblogs in a microblog chain and comments from general public. We leverage a unique natural experiment contained in the data from a leading microblogging website in China on which comment function was shut down for three days. This setting allows us to identify the impact of comments from that of peer microblogs. We find that shutting down comment function reduces the sentiment polarization on the microblogging site. In addition, this effect is more significant for individuals with higher social media participation. Results of this study shed light on how to identify key users and how to conduct successful social media campaigns.

Keywords: opinion polarization, social media, Bayesian learning

JEL Classification: M3