Student achievement and self-confidence in mathematics and science: evidence from TIMSS 2023

Abstract This paper examines whether underachieving students systematically overestimate their abilities and explores how confidence patterns vary across different contexts. Drawing on grade eight mathematics and science data from TIMSS 2023, we fit two-level regression models by country. Findings challenge the “unskilled and unaware” narrative: low-performing students typically reported muted, stable confidence, while high performers demonstrated strong positive achievement-confidence associations. Results align more with calibration theory and the Reciprocal Effects Model, which emphasize developmental feedback processes, than with inherent metacognitive deficits that align with the Dunning-Krueger theory. Further, cross-national contrasts underscore cultural variation: East Asian students reported modest confidence despite high achievement, whereas European systems exhibited steep calibration slopes. Gender differences were robust in mathematics, with boys reporting higher confidence net of performance, while immigrant effects were inconsistent. Overall, results highlight that confidence is socially and culturally structured, warranting context-sensitive approaches to motivation and feedback.