Abstract
<title>Abstract</title> <p>Background: Non-cognitive dispositions are increasingly recognized as important determinants of mathematics achievement, yet their structure and predictive relevance may vary across educational contexts. Using data from the Programme for International Student Assessment (PISA) 2022 across 24 European countries, this study examines the organization of non-cognitive factors associated with mathematics performance and evaluates their cross-national stability. Methods: A hierarchical framework comprising four macro-domains—Confidence, Motivation, Learning Strategies, and Attitudes—was assessed using PISA 2022 indicators. The analysis additionally incorporated the newly available Creative Thinking measure to examine its position within the broader non-cognitive system. To address multicollinearity among highly correlated student-level variables and identify robust predictors of mathematics achievement, Least Absolute Shrinkage and Selection Operator (LASSO) regularization was used to identify robust predictors of mathematics achievement and assess their stability across countries. Results: The findings support the established four-domain structure and position creativity as a factor linking confidence and learning attitudes within the non-cognitive framework. Mathematics self-efficacy emerged as a consistent predictor of achievement across countries, while mathematics anxiety also demonstrated substantial cross-national stability. In contrast, contextual and relational factors, including teacher support and students' sense of belonging, displayed considerable heterogeneity, with some variables showing changes in magnitude and direction across national settings. These patterns indicate that school climate factors are strongly influenced by country-specific educational and pedagogical contexts. Conclusions: The results suggest that core self-beliefs constitute robust and transferable predictors of mathematics achievement across diverse European educational systems, whereas the effects of school climate and relational factors are context dependent. By integrating creativity into a higher-order non-cognitive framework and applying regularized modeling techniques, this study contributes to a more nuanced understanding of which non-cognitive dispositions are structurally stable and which are shaped by local educational environments.</p>