Affective Polarization Among Political Elites in Germany – An Analysis of Twitter Activities During the 2021 Bundestag Election

Political polarization manifests in various forms, with affective polarization gaining prominence due to its impact on group dynamics and democratic norms. Rooted in Social Identity Theory, affective polarization involves individuals harboring strong feelings against opposing groups while displaying allegiance to their own (Iyengar et al. 2019). This paper explores the affective polarization of political elites in Germany during the 2021 Bundestag election, investigating how their rhetoric on Twitter corresponds to affective polarization in the public. Affective polarization tends to peak during election years. This comes as elections increase the visibility of political conflict, thereby activating partisan identities and intensifying affective polarization dynamics (Hern ́andez et al. 2021). The potential consequences of affective polarization, especially among elites, are far-reaching. Affective rhetoric and behavior of political elites can influence public sentiment, leading to increased polarization and, in extreme cases, contributing to physical violence. Moreover, social trust tends to decline, posing a threat to democratic quality (Torcal and Thomson 2023). While existing studies primarily focus on the U.S. two-party system, this research addresses Germany’s multiparty system, which offers a nuanced political landscape with overlapping stances. Recognizing the underexplored realm of political elites, this study focuses on the case of the German parliamentary elections in 2021. To quantify affective polarization among elites, we examine the Twitter activities of 1,537 Bundestag candidates across seven major political parties between July 20, 2021, and January 20, 2022. Affectively polarized rhetoric in tweets is identified with a refined coding scheme (Ballard et al. 2022) and GBERT, a transformer-based classification algorithm (Chan et al. 2020) that is fine-tuned to classify tweets as affectively polarized or not with high reliability (F1=0.88). This sheds light on the prevalence of affective elite polarization in Germany’s multiparty system during the election year and its different campaign phases. Logistic regression models reveal significant impacts of the campaign phase, party affiliation, gender, and age on candidates’ propensity to post affectively polarized tweets. Specifically, we find affectively polarized rhetoric to be highest before election day. All things being equal, politicians who use it are typically members of a fringe party and male. Moreover, the propensity to use affectively polarized rhetoric is higher the older candidates are. The findings contribute to understanding the dynamics of affective polarization among political elites in Germany, offering insights into the potential amplification of polarization during election campaigns.