Amid escalating environmental crises, climate change scepticism persists and, in some cases, rises across Europe (Beiser-McGrath & Bernauer, 2021). Case study evidence from a limited number of European countries and the USA suggests that political parties play a critical role in shaping public scepticism by adopting increasingly sceptical stances. These positions often reflect a nuanced scepticism that acknowledges climate change while denying its anthropogenic origins or minimizing its impacts (Küppers, 2022). This study explores whether political parties influence their constituents’ beliefs by reframing narratives around climate science across 10 EU countries.

We rely on automated text analysis of parliamentary speeches to assess party framing on climate scepticism. Party stances are quantified using a Party Climate Scepticism Index (PCSI), developed from the ParlaMint corpus of national parliamentary speeches. The PCSI measures the proportion of climate-sceptical segments within each party’s speeches on climate change. To construct the index, a binary text classification NLP model is trained using synthetic data generated through a zero-shot data augmentation strategy with ChatGPT and refined with human oversight. Active learning iteratively improves the model by selecting and annotating informative examples from unlabelled data, capturing the linguistic and rhetorical complexity of parliamentary debates.

The PCSI is merged with European Social Survey data (2016, 2020) at the party level to assess whether shifts in party positions influence voter scepticism. Using a first-difference estimator and controlling for second-dimension political competition and welfare state systems, the analysis reveals that party discourse shapes EU constituents' climate scepticism across ideological contexts.