Browse By Day
Browse By Time
Browse By Person
Browse By Mini-Conference
Browse By Division
Browse By Session or Event Type
A growing segment of the global population relies on social media to access and consume political news. As its influence has expanded, the technology once heralded as a tool of democratization has come under increasing scrutiny. Numerous studies argue that social media exacerbates political polarization and distorts the political news landscape (see, for example, Pariser 2011; Conover et al. 2011; Sunstein 2018). However, there is limited empirical evidence to inform how and when social media shifts political attitudes.
This paper contributes to the ongoing debate about social media and its role in exacerbating political polarization. Many researchers argue that social media creates echo chambers, thereby aggravating partisan inclinations (Adamic and Glance 2005; Jamieson and Cappella 2008; Sunstein 2009; Yardi and boyd 2010; Conover et al. 2011). These studies find that political information is unlikely to be transmitted on social media if its content is ideologically cross-cutting, whereas hyperpartisan information spreads rapidly. Moreover, they argue that even when cross-ideological interactions do happen, they do not signal moderation (Conover et al. 2011, 95). In certain cases, exposure to opposing views on social media has actually been shown to increase political polarization (Bail et al. 2018). Other scholars remain more skeptical, arguing that social media facilitates exposure to messages from ‘weak ties,’ which can produce a moderating effect (Gentzkow and Shapiro 2011; Prior 2013; Barbera 2014). This body of scholarship also suggests that real-world interactions provide fewer opportunities for cross-ideological exposure than digital ones (Gentzkow and Shapiro 2011).
Current studies on digital politics suffer from numerous inferential limitations. For example, the literature on polarization uses social network analysis (SNA) to measure the degree to which communities of social media users interact in segregated echo chambers. Whilst network analysis affords insights into the structural properties of echo chambers, SNA largely overlooks what individuals within these networks actually say about politics. With network analysis alone, it is not possible to distinguish between changes in users’ political networks and changes in their political attitudes. Although individuals’ social media networks correlate with their political attitudes, the opinions people hold within any given network are still reasonably heterogeneous. As such, the inferential leap made using network analysis is often too large.
To address this problem, we first use network analysis to identify Twitter users as part of one of four political echo chambers—far right, moderate right, moderate left, or far left. Once we sort sampled users by their political networks, we use a supervised machine learning approach to evaluate the content of their tweets. This two-stage process allows us to evaluate expressed political attitudes on social media, conditional on users’ political media networks. Our study evaluates changes in political tweet content around major news events to understand how information consumption in echo chambers affects partisans’ political attitudes during high salience news cycles. We use these data to discern between three possible types of political attitude change: polarization, sorting, and moderation. That is, we discern between (1) the movement of an individual’s stance toward a political or ideological extreme; (2) the homogenization of an individual’s stance such that their issue assessment comes in line with their partisan priors; or (3) the movement of an individual’s stance toward a political or ideological center. We find that individuals’ opinions on a variety of political issues converge on the modal opinion of media elites within their echo chambers following major political events. We use this evidence to suggest that people politically sort—rather than polarize—vis-à-vis social media echo chambers.