Abstract

Authors
D. Fernández 1
M. Manisera 2
P. Zuccolotto 2
1 Institute for Research and Innovation in Health (IRIS),
Universitat Politècnica de Catalunya - BarcelonaTech (UPC), Spain
2 Big&Open Data Innovation Laboratory (BODaI-Lab), University of Brescia, Italy

Brief description
This talk presents an interpretable statistical framework to study ideological and geographical patterns in European politics at both party and country levels. Using archetypoid analysis combined with regression and spatial methods applied to data from the 2024 Chapel Hill Expert Survey (CHES), we identify ideological extremes, cross-country heterogeneity, and spatial dependence in European party systems. The results provide insights relevant to academia, public policy, and official statistics by offering a data-driven and interpretable representation of ideological diversity and polarization across Europe.

Abstract
Understanding ideological positioning and its geographical structure is essential for analyzing party systems, electoral competition, and political polarization in Europe. This contribution proposes an integrated statistical framework to determine ideological and geographical patterns in European politics at both the party and country levels. Using data from the 2024 Chapel Hill Expert Survey (CHES), we apply Archetypoid Analysis (ADA), an unsupervised learning method that identifies representative ideological extremes directly from observed data. This approach yields three prototype parties corresponding to the left, center, and right poles of the European ideological space, and represents each party as a convex combination of these extremes, ensuring a high degree of interpretability. Building on the party-level results, we aggregate ideological positions using parliamentary seat shares to construct country-level indices capturing alignment with left-, center-, and right-oriented political contexts. Regression and mixed-effects models are then employed to examine how positions on key political issues and European Union-related policies are associated with parties’ proximity to ideological extremes, while accounting for cross-country heterogeneity.
Finally, spatial statistical methods are used to investigate geographical dependence in ideological orientations across countries. The results reveal clear regional patterns, with neighboring countries often exhibiting similar political profiles, and indicate that spatial dependence is driven primarily by social and cultural policy dimensions rather than economic ones. Overall, the proposed framework demonstrates how interpretable statistical learning, combined with classical and spatial modeling, can provide meaningful insights into both ideological diversity and the geographical structure of European politics.