2026/05/29 by Daniel Wiesner · 1 voice
Social Sciences · #Computational and Text Analysis Methods #Electoral Systems and Political Participation #Qualitative Comparative Analysis Research
paper · doi:10.31235/osf.io/xh9m7_v1
openalex publication_date 2026/05/29 · openalex created_date 2026/05/30 · openalex updated_date 2026/07/14
Effective politics relies on the availability and use of factual information. However, while theoretical and normative discussions often stress the importance of evidence-based politics, empirical studies of how often and under what conditions politicians refer to scientific evidence remain limited. This study aims to fill this gap by analyzing references to scientific evidence in one important political arena: parliamentary debates. Our dataset covers 30,000 speeches delivered in the main legislative chambers across six European democracies (Austria, Czechia, Denmark, Finland, Italy, and the United Kingdom) from 2018 to 2021. Using large language models (LLMs), we identify two key features in each speech: (1) the use of scientific evidence, and (2) the primary policy issue discussed. We then apply multi-level logistic regression models to evaluate which factors are associated with higher or lower likelihoods of references to scientific evidence. We distinguish between three sets of predictors: (a) macro-level country variables, (b) meso-level party and issue factors, and (c) micro-level speaker characteristics. Findings suggest that institutional roles, policy contexts, and organizational routines are more decisive for evidence use in politics than national context or ideology. The findings are discussed in light of potential measures to facilitate evidence use in political processes.