The paradox of excellence: Summer learning loss as a critical vulnerability for high-ability math students: A hierarchical Bayesian analysis of PISA Data

Mathematically high-ability students are often assumed resilient to instructional interruptions. This study examines whether initial achievement may become vulnerable when structured learning is suspended. We combine an exploratory longitudinal sample (Morocco, N = 70) with PISA 2022 data (56 countries, N ≈ 600,000).In the Moroccan data, a quadratic relationship emerges between pretest score and summer loss: higher-performing students show disproportionately larger declines (estimated minimum at −1.13 SD, 95% CrI [−1.28, −0.98]). This pattern is sample-specific and consistent with a reversed Matthew effect hypothesis, serving as a hypothesis-generating result requiring replication in larger independent samples . In PISA data, a hierarchical Bayesian model indicates longer summer breaks are associated with modestly lower mathematics performance (μ_β = –0.021 SD/week; 95% CrI [−0.032, −0.009]), conditional on socioeconomic status and test timing. The cross-national association is linear, whereas individual-level analysis suggests non-linearity – a divergence consistent with the possibility that aggregation smooths heterogeneous within-system patterns. We propose a Bayesian sensitivity framework to assess how seasonal variation affects international rankings. Simulations shift country positions by 1–4 ranks while preserving overall hierarchy. The findings are descriptive and correlational; no causal mechanisms are claimed. We provide a diagnostic tool for evaluating temporal invariance in large-scale assessments.