The Effects of Partisan Geographical Segregation on Online Behavior on Twitter

• In Event: Local Politics and Geography in U.S. Elections

Abstract

Studies about the causes and the consequences of contemporary levels of political polarization have proliferated, both in the US context (Mason, 2018; Iyengar et al. 2012; Abramowitz and Saunders, 2008 and Fiorina and Abrams, 2008) and in a comparative perspective (Gidron and Lupu 2015). Across distinct scholarly and popular accounts, the role of social media in facilitating ties between citizens who express similar political positions, forming the so-called online echo chambers, has often been blamed for many contemporary democratic ills, including the rising levels of political and affective polarization.

While significant literature has focused on measuring and explaining the dynamics of partisan polarization in the digital world (Bail et. al. 2018; Eady et. al. 2019, Guess 2021, Fletcher et. al. 2021), research is yet to fully explore to what extent offline dynamics influence how social media users live their digital lives. For example, while partisan geographical segregation has been long-studied by scholarship and shown to influence a myriad of offline behaviors and attitudes (Baxter-King et. al., 2022; Enos 2017; Bishop 2009), few studies have connected this feature of the American political context to users’ behaviors on social media.

Building upon a data infrastructure connecting online and offline information for over a million Twitter users, our paper investigates the influence of offline geographical partisan segregation on users’ behavior in the digital world. In particular, we are interested in the influence of isolation among ingroup partisans, and exposure to outgroup partisans, on online polarization and outgroup hostility. Offline information is extracted from the L2 voter file for over 180 million registered voters in the United States. With this data, we build individual and aggregate measures of geographical partisan segregation, as suggested by Brown and Enos (2021). Online data comes from unique users identified in the Twitter Decahose API, a 10% random sample of all tweets. We use both data sources to match Twitter users with their voter file information using the method outlined in Hughes et al. 2021, allowing us to build an augmented panel containing both online and offline information about social media users.

Our investigation is governed by two primary research questions. First, we investigate the relationship between offline partisan sorting and online partisan sorting Second, we examine how offline political segregation influences online behavior, including posting toxic speech, being hostile towards the partisan out-group, and sharing misinformation.

To answer our first research question, we propose methods to estimate comparable levels of online and offline segregation. To measure offline partisan segregation, we use the measure proposed in Brown and Enos, 2021, which calculates partisan exposure by aggregating the weighted distance between a voter and out-partisans. Similarly, our online segregation measure uses a weighted measure of interactions with outgroup users. Much of recent scholarly research has provided a more nuanced view of the prevalence of online echo chambers, and our approach provides a baseline to compare the degree to which political homophily on social media users’ online networks differs from their offline interactions with other registered voters.

Research shows that the levels of partisan geographic sorting are high in the United States (Brown and Enos, 2021), while much of the research has shown a low prevalence of online partisan echo chambers (Barberá et al. 2015). Thus, we first expect that most voters experience lower levels of partisan sorting online than they do offline; that is, most voters will be exposed to their partisan outgroup online more than they are exposed to their partisan outgroup offline. This analysis will provide key insights regarding concerns about the role of online echo chambers or filter bubbles on partisan polarization.

To answer our second research question we augment our data by collecting all tweets, likes, and retweets for all registered voters included in the dataset and examine to what degree geographical political segregation influences online behavior. We focus on the effects of offline homophily on key online measures, such as sharing of low-quality content and misinformation rumors, as well as posting and/or liking toxic speech and out-group negativity. We focus on two distinct processes for partisan segregation: isolation and exposure. The former measures geographical exposure to in-group partisans, while the latter measures geographical proximity to outgroup partisan voters. In addition, we discuss interactive effects between levels of online and offline segregation, exploring the effects of online homophily when holding constant the levels of offline segregation.

Authors

• Tiago Augusto da Silva Ventura, Georgetown University

• Megan Brown, University of Michigan

• Joshua A. Tucker, New York University

• Jonathan Nagler, New York University