Self-Effects on Social Media and Political Polarization in the 2019 EU Elections

• In Event: New Directions and Emerging Technologies

Abstract

This study investigates whether citizens’ practices of self-expression on social media encourage polarized identity formation in the European Union. Previous research overwhelmingly theorizes that partisan selective exposure induces a “reinforcing spiral” effect on citizens’ online news consumption, which hardens their political opinions over time (Slater, 2017). Departing from this line of research, we argue that polarization is not driven by information exposure but rather, political self-expression. Expressing oneself politically has been posited to exhibit “self-effects” on the expresser (Pingree, 2007), and Shah et al. (2017) suggest that the ease and scope of expression on social media platforms may magnify these self-effects. While Lane et al. (2019) find evidence that self-expression can influence political behavior, no study has investigated the polarization potential of self-effects. Therefore, we ask: Does self-expression on social media leads to the reinforcement of political identities and consequently, polarization?

To answer this research question, we develop a new theoretical model to explain the relationship between online self-expression and polarized identity formation: the Reinforcing Spirals of Expression Model (RSEM). The RSEM model hypothesizes that repeated acts of political self-expression on social media, coupled with the peer-to-peer feedback systems built into these platforms, encourage partisan expression and the hardening of political identities over time. This, in turn, increases polarization, limits consensus-seeking, and leads to the fragmentation of societies.

We test our model empirically with a two-phase, mixed-methods approach. The first phase is computational and longitudinal in scope. We collected Twitter data from users expressing political opinions relating to the 2019 European Parliament elections (n=597,867). To control for linguistic variation across the EU, we translate all posts into English and perform topic modelling, an approach validated by De Vries et al. (2018). Once the main topics of the election have been identified, we isolate each topic and categorize words that signify contrasting opinions about them. Then, we identify users who frequently expressed partisan positions on these issues, and manually code a subset of their tweets to assign the users a “strength of polarization” score. For each user, we further harvest their Twitter timelines for the entire year 2019. This way, we can consider the election as a treatment and examine whether polarized users’ partisan expression during the election diminishes, is sustained, or increases over time.

In the second phase, we contact these users and invite them to participate in an online survey, which is designed to examine the relationship between political self-expression, selective exposure, and polarization. Furthermore, the survey allows us to gather insight into the demographic variables of these polarized Twitter users, such as their party affiliation, voting history, demographics, and location within the EU.

The study contributes to the literature on polarization on several fronts, most notably by testing the political implications of self-expression on social media. In addition, we provide needed empirical evidence for how computational and survey methods can be combined, as well as understanding about social media activity during the historically understudied European Parliament elections.

Authors

• Michael Bossetta, Lund University, Sweden

• Anamaria Dutceac Segesten, Lund University

• Daniel Lane, UC Santa Barbara