Who Do They Claim to Be (Close To)? Measuring Politicians’ Rhetorical Self-Positioning Towards Groups

Politicians often claim to belong to a certain group, be close to it, or distance themselves from it. With such claims, politicians signal to voters their positioning relative to different groups in society. Existing research largely looks at which groups politicians mention or appeal to in their rhetoric; yet measurement approaches usually do not capture whether i) the speaker is talking about their own relationship with the group, and ii) their expressed closeness or distance to the group. Here, we present an original computational measure to capture these aspects. We first identify and label politicians’ mentions of groups using existing transformer models fine-tuned to detect group mentions in parliamentary speeches. Second, we identify co-occurring instances of group mentions with self-referential claims, measured as the use of personal pronouns within a proximate syntactic ‘window’. Third, we draw stratified random samples of these data for expert hand-coding of group-related relational claims, drawing from Appraisal Theory in linguistics. Using these human-coded labels, we fine-tune an open-source large language model to identify instances, strength and valence of relational claims. We train and test our model to a large cross-national dataset, covering speeches from eight European parliaments plus the U.S. House of Representatives between 1994-2024. We then validate the approach using a set of performance tests, including comparisons of type and size of transformer model. Finally, we present empirical evidence for politicians’ claims to being close or distant to groups of voters. Our method presents a new angle through which to study and understand representatives’ signaling of proximity or distance to different socio-political groups. It puts the relationship between politicians and groups—rather than mentions of groups—at center.