EU Permacrisis through National Lenses: A Blended Reading Analysis of Austrian and German Election Manifestos (2002-2025) using BERTopic and LLM

The digital age has produced extensive datasets documenting communication by political actors, the media, and civil society, indicating a shift in political discourse toward digital platforms. Election manifestos, for example, are now primarily distributed online through campaign websites. The accessibility of such data has increased due to advances in algorithm-based methods for collecting, storing, and analysing large volumes of digital empirical data. However, greater data volume brings increased complexity. These trends introduce methodological challenges in collection, pre-processing, and analysis. To address these problems, our paper proposes a human-in-the-loop blended reading workflow combining BERTopic with a large language model (LLM), demonstrated through a case study of German and Austrian national parliamentary election campaigns between 2002 and 2025. During this period, elections took place against the backdrop of multiple crises that threatened the European Union’s internal security and democratic foundations. Scholars frequently employ the term permacrisis to characterise this persistent state of crisis and its associated challenges. With our explorative approach, we trace which issues are addressed by the Austrian and German parties in their communication through national election manifestos. This shall further allow us to detect policy fields and to analyse if EU permacrisis events are displayed.