Logos or Pathos: How Political Parties Appeal to Voters on Social Media

Emotions play an (increasingly) important role in party competition. In an age of instant news, political parties find themselves in a constant battle for attention. They must seize every opportunity to position themselves at the centre of the debate. Emotional appeals are a particularly effective means to that end: they increase media coverage during election campaigns and impact electoral success. Social media acts as a magnifying glass in this context, since emotionally charged content tends to generate more engagement and is more likely to be shared than purely political content. Pictures, memes, and videos increase the potential for emotionalization on social media and are preferentially amplified by platform algorithms. In this paper, we study differences in the use of emotional content across parties and platforms, as well as the different reactions that emotional and non-emotional content provoke. We distinguish specifically between textual and visual content to better understand their respective roles and the differences in their impact on the emotionalization of debates. Our analysis draws on a large-scale dataset comprising over 550,000 posts published by members of the German Bundestag on Facebook and Instagram during the 20th legislative period (2021–2024). The period was dominated by highly emotional issues ranging from the war in Ukraine, the Hamas attacks in Israel and the subsequent war in Gaza, to the renewed immigration debate after the surge of deadly attacks involving people with migration background during the election campaign. Against this background, we analyse how members of different political parties use emotions in their posts: how often and with which tonality do they connect their political messages with emotional appeals and do these emotional appeals cluster in particular policy issue, and finally, can we identify certain characteristics of emotional appeals that increase their potential for engagement? Methodologically, we combine established approaches from computational text analysis with emerging techniques in automated image analysis Emotionality in text measures will be complemented by topic classification to assess issue-specific variation. For classifying visual content, we propose a multi-dimensional operationalization that integrates automatically extracted visual features (facial presence and expressions, as well as symbolic elements), low-level aesthetic indicators associated with affect, and human-coded assessments of perceived emotionality. Thus, this paper not only offers insights into the different emotional appeals members of parliament use to gain their voters’ attention and communicate their message, but also proposes and tests new methods for studying these emotions in textual and visual political content.