Description
Political leaders signal their commitment to democratic or authoritarian values through public speeches. Such signals can constitute a breach of democratic norms and indicate intent to undermine democratic institutions. By listening to what leaders say, we are able to capture a hitherto overlooked sign of a crisis of democracy. We develop indicators for illiberal speech that capture the language by leading representatives of political regimes across the globe. Utilizing a machine learning approach, we fine-tune DeBERTa models to score 28,237 speeches by 452 political leaders representing over 400 administrations in 123 countries between 1989 and 2024, yielding over 400,000 semantically coherent paragraphs. The scores place leaders on an illiberal-liberal scale and have been validated with openai’s gpt-5.2 and human coders. We find that illiberal public discourse is part of the multi-dimensional symptom of democratic decline. Our analysis opens up a new research agenda on the role of public rhetoric in the study of political regimes and their transformation.
| Date made available | 15 Jul 2026 |
|---|---|
| Publisher | Harvard Dataverse |
Cite this
- DataSetCite