The OECD published PISA 2025 this morning. Around 760,000 fifteen-year-olds in 91 countries sat it, making it the largest round of the assessment there has been.
Average reading across OECD countries is down about 28 points on 2015. The OECD puts that at roughly a year and a half of schooling. One in five students is now a low performer in science, mathematics and reading together, against 16 percent in 2022. American reading fell 14 points in three years and sits near a twenty-five-year low.
Why this particular test is hard to wave away
Standardised international comparisons deserve scepticism, and PISA has attracted plenty. It samples rather than censuses. It cannot fully control for which students are in school at 15 in different countries. Its scale is constructed, so a "point" is a statistical artefact rather than a natural unit.
All true, and all constant. The reason a 28-point decline is difficult to explain away is that the same instrument, administered the same way to the same age group, has produced a run of falling numbers across many countries with different school systems, different funding models, different curricula and different politics.
A measurement problem would have to be a measurement problem that arrived everywhere at once and got steadily worse. That is a less parsimonious explanation than the obvious one.
The number that matters more than the average
Not the mean, which moves for uninteresting reasons, but the low-performer share: from 16 percent to 20 percent of all 15-year-olds, in three years, across all three subjects at once.
A low performer in PISA terms is not someone who did badly on a hard paper. It describes a student who cannot reliably extract meaning from a straightforward text or apply basic mathematics to an unfamiliar situation. That is a threshold of functional literacy and numeracy rather than a ranking, and the population below it has grown by a quarter in one cycle.
That group is the one everything downstream depends on. It determines the share of a cohort that can be trained for skilled work, the share that can complete a further-education course, and the share for whom every subsequent intervention has to start further back.
What the numbers will get blamed for, and what they can support
Phones will be blamed, and so will the pandemic, teacher shortages and curriculum reform. Each has advocates and the OECD's own reporting notes teacher shortages and attention among the surrounding factors.
Attribution is genuinely hard here, and this desk is not going to pretend otherwise. The pandemic is the tempting single explanation and it fails a basic test: the decline was already underway in 2015-2018 data, before it, and has continued after the point at which recovery in other indicators arrived. Whatever is happening began earlier and has outlasted the disruption everybody remembers.
What the data does support is narrower and more useful: this is not one country's policy failure. Any explanation has to work in systems that select at 11 and systems that do not, in countries that spend heavily per pupil and countries that do not.
The part that is a national problem rather than a school problem
A 14-point fall in US reading in three years arrives at the same time as a labour market that has been quietly raising its literacy requirements.
This paper has written about employers dropping degree requirements and finding that skills-based hiring holds up, and about adults returning to evening classes at rates not seen in decades. Both of those assume a working population that can be trained. A quarter of a cohort below functional literacy is a constraint on that assumption, and it arrives on the labour market roughly three years from now.
What to watch
Not the league table, which is the least informative part of PISA and the part every country's press reports.
Watch the within-country spread — the gap between the 10th and 90th percentile — in the full data release. A decline concentrated at the bottom is a different problem from a decline across the whole distribution, and it calls for different money in different places. The average tells you almost nothing about which of those is happening, and it is the only figure most of this coverage will contain.
The scale of PISA 2025 at about 760,000 students across 91 countries and economies; the fall of about 28 points in average OECD reading performance between 2015 and 2025 and the OECD's characterisation of that as roughly a year and a half of learning; the rise in the share of 15-year-olds who are low performers across science, mathematics and reading from 16 percent in 2022 to one in five; the 14-point fall in US reading scores from 2022 with 25.7 percent low performers and 13.3 percent top performers; and OECD Secretary-General Mathias Cormann's statement that reversing the declines is urgent are as published by the OECD on 9 September 2026 and reported by Al Jazeera, The Hill, Education Week and K-12 Dive the same day. Attribution of the decline to any single cause is contested and none is asserted here. The analysis is our own.
Topics nationaleducationschools
