Hostile Communication on Social Media: How Democratic Regimes and Party System Types Shape Elites’ Hostile Rhetoric

The scholarship indicates that political communication on certain social media (e.g. Twitter/x) increasingly centers on hostile, anger-laden, and negative rhetoric that amplifies conflict and division rather than constructive dialogue. In this respect, recent scholarships have stressed the importance of distinguishing between forms of political hostility that are compatible with democracy and those that undermine it, introducing a normative threshold between agonistic and antagonistic communication styles (Dodeigne 2025). While the scholarship has increasingly examined the different types of hostility (Papacharissi 2004; Muddiman 2017; Rossini 2020), we still lack the systematic institutional and political explanations for cross-national variation in these rhetorical styles (Walter 2021; Nai 2021).  In this wake, this paper investigates the structural conditions under which political actors’ communication on social media is more likely to adopt agonistic rather than antagonistic rhetoric. Our theoretical framework covers two sets of macro-level factors – institutional regimes and party-system configurations – that shape leaders’ propensity to foster agonistic or antagonistic communication. First, building on comparative analyses of institutional design (Bernaerts, Blanckaert & Caluwaerts 2023), we argue that consensus-oriented democracies reduce the incentives for antagonist rhetoric by fostering cooperative norms and power-sharing expectations. Majoritarian systems, by contrast, amplify zero-sum competition and political actors are more prone to evolve into antagonist political styles (Gidron, Adams, Horne, 2019). Second, in proportional based representative democracies, characteristics of party systems such as multipartyism, electoral threat structures, and government vs. opposition status, create varying strategic incentives for either confrontational or moderating tones (Mendoza, Nai and Bos, 2024). In this respect, traditional parties may mimic more radical styles (Valentim, Dinas, Ziblatt, 2025) under pressure from challengers (Meguid 2005), while competitive uncertainty can increase the appeal of polarizing communication. Submitting to the section on “Political Actors, Strategies, and Democratic Resilience”, we empirically examine these relationships using the Twitter Parliamentary Database, which offers a comprehensive longitudinal and cross-national corpus of politicians’ communication in 18 countries from September 2017 to October 2019. We use social media data (Twitter/X) as a ‘most likely case’ for identifying antagonist rhetoric, due to its features to foster angry, polarizing and even violent discourses (Ryan 2012; vander Goot et al. 2025). By applying a fine-tuned BERT model, we classify tweets according to their hostile rhetoric types (agonistic or antagonistic). We then model how institutional regimes and party-system characteristics condition the probability that political actors employ different polarizing styles during electoral campaigns. Overall, the paper offers a theory-driven and empirically grounded account of political actors’ communication on social media to mobilize affective divisions, conditioning distinct institutional regimes and party systems features. This allows to understand the rhetorical boundaries of political competition in developing negative rhetoric and its impact on the quality of democracy (Costa, 2025).