Measuring the communication of threat

Most recently, research on populism, polarization, and democratic backlash has indicated that perceptions of threat among the public increase support for populist, radical, and authoritarian actors. In addition, this research indicated that such actors purposefully use threat-intensive communication to increase the levels of (perceived) threat among citizens. However, no systematic research on threat communication has been undertaken so far. This is especially unfortunate as experimental research analyzing the effect of elite communicationon threat perceptions systematically suffers from pre-treatment biases. To measure the actual effects of communicated threats, one would need to link the degree and kind of threat communication received by citizens and whether or not this communication affects their perceptions and – in the longer run – attitudes over time. We argue that this lack of research is because measuring the communication of threat is way more complicated than it initially sounds. The problem arises already if one needs to define what constitutes a threat in the first place. Our paper takes a first step in systematically and automatically measuring the intensity of threat levels in politically relevant communication. In doing so, it proceeds in four steps. First, we sample media coverage from media outlets varying in ideology and degree of respectability. Second, using this sample of media coverage, we ask 200 persons to code to what extent its articles convey threat on a scale ranging from 1 (no threat) to 6 (highly threatful). Third, we apply and compare different modeling paradigms from large language modeling (i.e., fine-tuning, zero-shot inference, and few-shot inference) to develop a model that may automatically code new articles regarding their intensity of communicated threat. Fourth, we aim to use topic modeling, named-entity recognition, and SHAP values to generate insights on policy topics high/low in threat intensity and persons using more / less threat-intensive communication. Our findings will be interesting to several subfields of Political Science, including research on populism, polarization, political psychology, political methodology, and political communication.