Strategies of Division: A Comparative Analysis of the Linguistic Construction of Affective Polarization in German Election Manifestos

The increase in affective polarization is driven by the digital revolution and poses new challenges for modern societies as well as democratic systems themselves. In the context of political discourse analysis, this article argues that reality, and consequently affective polarization, is constructed and realized through language. This theoretical approach emphasizes power perspectives, making the strategic communication of political elites, in particular their use of polarizing frames and negative campaign rhetoric, the central subject of analysis. Following on from this, this article locates and analyzes populism as a strategic communication strategy with a central function in the development of affective polarization. Furthermore, within the framework of this analysis, it distinguishes between an inclusive communication strategy directed against economic elites and an exclusive communication strategy based on ethnic nationalism in order to do justice to the different facets of the political spectrum: facets that extend to racist exclusion and aim at hostility toward democracy. Against the theoretical backdrop of political discourse analysis and the concept of framing, this article examines how and to what extent German political parties polarize issues in their election manifestos in the run-up to the 2026 state elections and what different patterns of affective polarization they create as a result. The focus is on the state elections in Baden-Württemberg, Rhineland-Palatinate, Saxony-Anhalt, and Mecklenburg-Western Pomerania. In this sense, election manifestos can be interpreted as internally and externally oriented strategy papers that contain issues, narratives, and moral charges as core beliefs in communication strategy. Accordingly, the article uses three different linguistic dimensions for further operationalization: emotional valence, psycholinguistic indicators, and moral anchors/frames. The article empirically analyzes the election manifestos of the seven relevant parties in the four federal states. To answer the underlying research question, different computer-assisted methods are combined to reflect the complexity of the subject matter. In a first step, the underlying election manifestos are prepared for structural topic modeling (STM). This includes part-of-speech tagging to structure the text data for modeling. STM reveals latent thematic structures and allows their prevalence to be statistically modeled as a function of covariates (such as party affiliation and region). This thematic mapping serves as a basis for subsequently analyzing which of the identified topics are particularly emotionally and morally charged and thus act as anchors of affective polarization. The next step involves capturing the semantic and psychological quality of language using BERT/roBERTa and lexicon-based psychometrics to lay the foundation for classifying party patterns and using quantitative methods to identify qualitative differences between parties in terms of affective polarization. This paper contributes to various aspects: First, it theorizes and operationalizes a differentiated understanding of affective polarization, which, by distinguishing between inclusive and exclusive strategies, enables an assessment of its democratic quality. Second, it provides an empirical comparative measurement of the polarization practices of different party political communications in the German multi-level system. Third, it offers a methodological attempt that integrates structural topic modeling and psychometric approaches to accurately capture the affective dimensions of political texts.