Types of Intolerance in Elites’ Political Rhetoric: A Proof of Concept for Studying the Targeting of Minorities in Belgium and the U.S.

Over the past fifteen years, Western democracies have witnessed the unfolding of several mediatized episodes of prejudice and violence targeting minorities – migrants dying at Europe’s borders, shootings in LGBTQIA+ venues, family separation policies of migrants in the United States, and forced evictions of Roma communities in France. These events unfolded amidst a political context increasingly structured by polarizing fear, scapegoating, and moralized hostility toward these minorities. Existing scholarship often links this evolution to elites’ intolerant rhetoric and its potential role in normalizing prejudicial and violent behaviour against minorities. Yet, current approaches in political communication tend to treat intolerance as a unitary phenomenon and insufficiently distinguish between its conceptually different discursive forms. Building on social psychological and cognitive linguistics, this article argues that intolerant rhetoric comprises distinct “shades” that may entail different implications for targeted groups and democratic life. We propose a new conceptual framework that differentiates three mechanisms – (1) moral hierarchization, (2) dehumanization, and (3) demonization – through which political elites construct group-based moral hierarchies and legitimate exclusion or harm. Combining interdisciplinary scholarship on moral emotions and political intolerance, we theorize how these mechanisms range from subtle dual-audience rhetoric (“dog whistles”) to explicit discourses legitimizing punitive and violent responses toward minorities. To assess the applicability of our conceptual framework across minority targets, we analyze elites’ political communication targeting migrants and LGBTQIA+ communities on Twitter/X. To establish its external validity, we test it on two “most different cases”: Belgium, characterized by multiparty coalition-based politics in a relatively stable democracy, and the United States, characterized by two-party and highly polarized politics with evidence of democratic erosion. Empirically, we manually code 849 Belgian tweets (2023–2024) and use a manually coded sample of 1,246 U.S. congressional tweets (2017–2023) to fine-tune and validate a supervised NLP classifier subsequently applied to 158,080 minority-related tweets. Our exploratory results provide initial evidence that moral hierarchization, dehumanization, and demonization can be operationalized and identified across different minority targets and political contexts, while revealing substantial variation in their distribution. Migrants are considerably more exposed to intolerant rhetoric than LGBTQIA+ in both cases, while the composition of intolerance differs across countries. Intolerance is also strongly concentrated among specific party actors and becomes more persistent during periods of heightened political and electoral salience. These findings provide initial support for the cross-contextual applicability of the framework and demonstrate the analytical value of distinguishing different forms of intolerance that would remain concealed by a unitary measure of hostile rhetoric.