Comparing Political Agendas on Social Media: A Word Embedding Approach to Analyzing Austrian Party Communication of Parties in Government vs. Opposition

Political Party Communication has shifted to social media over the past decade, which gave political actors the chance to adopt to these new communication channels. In Austria these are, as of now, predominantly Facebook and Twitter. One approach to analyzing party communication is the political agenda. It can be defined as “the list of issues to which political actors pay attention” (Walgrave et al., 2008, p. 815) and measured by looking at issues a party talks about. Recently, several studies have demonstrated the useful applicability of computational approaches to study agenda-setting processes (Gilardi, 2022). This study aims to examine whether there are differences in the topics (or agendas) communicated by political parties on social media when they are in government compared to when they are in opposition. Additionally, the study explores whether parties use different languages to communicate on different social media platforms. Data was collected by scraping all posts by the parties on federal and state accounts, as well as from party chairs via CrowdTangle and the TwitterAPI. A model that was pre-trained on Austrian news media articles was used to embed all posts in semantic spaces. The topic modeling strategy proposed by Angelov (2020) was applied to find dense areas denoted as topics. Three research questions were formulated to guide the study. The first question aimed to identify the topics that parties talked about on Facebook and Twitter. One semantic space per party and social media was created, and the 15 largest clusters were identified to assess topical differences between parties and platforms. The second question aimed to determine if parties used the same words when discussing the same topic, independent of the social media platform. This was achieved by looking at the ten words that were located closest to the topics in the semantic space and comparing how they differed between platforms. The third research question looks at how topics change over time, with a special interest in whether the topic structure changes when parties switch between government and opposition. Hypotheses were not included in the study as topic models were used as an inductive strategy (Isoaho et al., 2021). Overall, this study contributes to the growing body of research on social media and political communication and provides insights into the ways in which political parties use different social media platforms to engage with their audiences.