Fewer Children, Fewer Stars? Top PISA Performers in Europe

Abstract Background PISA results are used in policy related monitoring, however, they are almost always framed around rates and shares, while little attention is given to what these rates and shares mean in absolute terms if applied to entire cohorts. Birth cohorts in Europe have been contracting for two decades, and a stable share of high performers now means fewer high performers. This study asks whether the number of high performers fell and whether this was due to smaller cohorts or to fewer students reaching the top. Methods PISA mathematics data from eight cycles (2003–2025) were analysed for 38 European countries (18 post-socialist, 20 other) with three or more comparable cycles. Numbers of high performers were calculated using final student weights for students scoring Level 5 and above. Variance estimates used plausible values, replicate weights, and 2025 score linking errors. Differences were decomposed into cohort coverage and rate components (Kitagawa method) and resident population, alignment and rate components (logarithmic mean Divisia index). Trends were compared across regions using random effects meta regression. Results Across the 21 countries with equivalent data across all eight cycles, the number of high performers fell by 50% from about 584,000 to 291,000 (95% CI: 195,000–390,000). By 2018, it was already down 21%. The number fell in 31 of the 38 countries, and fell significantly in 23 countries. 93% of the decline in the panel came from declining proficiency rates, and 6% from changes in resident population size. Cohort contraction was the main component of the count change in 10 of 18 post-socialist countries, compared with 3 of 20 other countries. Proficiency rates declined at a slower rate in post-socialist countries. Tertiary attainment rose in most countries over the same period. Conclusions Europe has been producing more credentials while its pool of high performers shrank, and its monitoring frameworks, built on shares, cannot show it. This should be seen as a human-capital loss in its own right, whatever its cause, and one that should be reported as a count, not only a rate.