# The cognitive foundations of STEM achievement: a cross-national multilevel and SEM investigation

## TL;DR

This study explores the cognitive and non-cognitive factors influencing STEM achievement using data from over 600,000 students across 80 countries.

## Contribution

The study integrates multilevel analysis with structural equation modeling to uncover indirect pathways affecting STEM performance.

## Key findings

- General academic proficiency explains over 40% of the variance in STEM achievement.
- Self-efficacy and metacognitive strategies influence STEM outcomes through cognitive competencies.
- School-level factors like teacher support and resource availability moderate individual achievement.

## Abstract

Excellence in Science, Technology, Engineering, and Mathematics (STEM) education is vital for national competitiveness and sustainable development. Yet, the cognitive foundations of STEM achievement remain insufficiently understood across diverse educational systems.

Drawing on data from the Programme for International Student Assessment (PISA 2022), this study applies a cross-national, multilevel, and structural equation modeling (SEM) approach. Integrated student, school, and cognitive assessment datasets encompassing over 600,000 students from more than 80 countries were analyzed.

The pooled mean mathematics score was 472.7 (SD = 91.1), with comparable central tendencies in science (M = 483.6, SD = 90.3) and reading (M = 476.6, SD = 93.1). General academic proficiency (math, reading, and science literacy) exhibited the strongest predictive power, accounting for more than 40% of the variance in STEM performance. SEM revealed that self-efficacy, metacognitive strategies, and socio-economic resources indirectly shaped STEM achievement through cognitive competencies. At the school level, teacher support, disciplinary climate, and resource availability moderated individual outcomes, with cross-level interactions explaining an additional 12–15% of the variance.

General academic proficiency across PISA domains were strongly associated with STEM outcomes and formed statistical indirect pathways linking school climate, teacher–student interactions, and socio-demographic inequalities to achievement. Fostering cognitive and non-cognitive skills may help address educational inequities, although causal mechanisms cannot be inferred from cross-sectional data. By integrating multilevel analysis with SEM, this study delivers robust, policy-relevant insights for advancing equitable and effective STEM education globally.