The rise of populist attitudes: using supervised machine learning to identify their main determinants

Autor principal:
António Correia (, University of Porto)
Autores:
Patrício Ricardo Soares Costa (CPUP, Faculty of Psychology and Education Sciences, University of Porto)
Programa:
Sesión 7, Sesión 7
Día: viernes, 11 de septiembre de 2026
Hora: 09:00 a 10:45
Lugar:

The rise of populism has recently gained significant attention and is widely regarded as a topic of concern. Populism is characterized as a thin-centered ideology that divides society into two distinct groups and is hostile towards those outside the category of “ordinary people.” The thin-centered approach allows for measuring populism as an attitude and provides a framework for measuring it among citizens. Moreover, a large amount of research has examined the concept of populism, its theories, and its impact on political parties through diverse methods such as qualitative coding, content analysis, and computerized techniques. However, the study of populist attitudes at the individual level remains under-explored. This dissertation investigates the correlation between populist attitudes and social characteristics, developing a machine-learning framework to identify the key determinants of populist attitudes through the Populist Attitudes Scale (POP-AS). The findings of this dissertation indicate that individuals who exhibit higher levels of populist attitudes tend to show low agreement with statements related to the behaviour of government officials. Moreover, higher education levels are associated with lower values of populist attitudes, while men tend to display higher levels of populist attitudes. Additionally, younger participants tend to have lower levels of populist attitudes compared to older individuals. Regarding the machine learning problem, tree-based algorithms outperformed others. For instance in the regression approach, the XGBoost and Gradient Boosting regression algorithms demonstrated the best performance among the models tested. The XGBoost and Random Forest classifier outperformed other algorithms in the classification task. These findings shed light on the relationship between populist attitudes and various social characteristics, contributing to a better understanding of populism at the individual level. The machine-learning framework employed in this dissertation offers valuable insights into the determinants of populist attitudes, allowing for a more nuanced analysis of this complex phenomenon.

Palabras clave: