PISA 2025 Results: The Learning Decline Predates the Pandemic

Blame the pandemic if you want. The PISA numbers were sliding years before anyone had heard the word lockdown.

When test scores fall, there is always a convenient culprit. For the past few years it has been the pandemic, and not without reason: shuttered schools and screen-only lessons did real damage. The 2025 PISA results carry an inconvenient detail, though. The decline they confirm did not begin in 2020. It began while classrooms were still full.

That distinction sounds academic. It is actually the whole ballgame, because it changes what a fix has to do.

The Short Version

The 2025 PISA scores are the lowest the OECD has ever recorded, but the drop is not a pandemic story. Reading performance across wealthy countries peaked around 2012 and science around 2009, and math was already falling in a dozen OECD nations before 2018. The 2015 round showed declines in all three subjects. COVID deepened a slide that was well underway. That matters because a structural problem does not repair itself the moment schools reopen.

A Slide That Started Before COVID

The tidy version of recent history goes like this: students were doing fine, the pandemic hit, and scores dropped. The data tells a slower, less flattering story.

Reading performance across the OECD peaked around 2012. Science peaked earlier, around 2009. Math had already turned downward before 2018 in a long list of countries, including Australia, Belgium, Canada, Finland, Korea, the Netherlands, New Zealand, and Switzerland. The first real alarm came after the 2012 assessment, when Finland dropped out of the top ten in math for the first time, and it grew louder after 2015, when all three subjects slipped at once. You can trace the arc in the OECD’s long-term trends analysis.

Add it up and the slide has ten to fifteen years behind it, depending on the subject and the country. The pandemic did not open this door. It walked through one that was already ajar.

What the 2025 Numbers Add

The latest round, released in September 2026, tested roughly 760,000 fifteen-year-olds across 91 education systems. Across OECD countries, reading has fallen 28 points since 2015 and math 22 points, which the OECD describes as about a year and a half of lost schooling in reading alone. Both are the lowest averages the assessment has produced.

Science was the exception. It was the featured subject this cycle, and average scores held between 2022 and 2025 after years of drift. A flat line, in this context, counts as good news.

Put the 2025 figures on the same chart as the earlier ones and the pandemic looks less like a cliff and more like a steeper section of a hill people were already walking down. That framing is unpopular because it removes the tidy villain, as coverage from the OECD’s own education team has noted.

Why “Catch-Up” Framing Falls Short

Most of the money and messaging since 2021 has used the language of learning loss and catch-up. That framing assumes a one-time shock you can recover from, the way a student makes up a missed unit after a long absence.

Let me back up, because this is the part that gets lost. Nobody serious is claiming the pandemic did no harm. Closures clearly hurt, especially for younger and lower-income students. The point is narrower: catch-up programs can repair the COVID dip and still leave the longer drift untouched. If reading was sliding in 2013, tutoring away the 2021 gap gets you back to a number that was already falling.

What a Long Decline Actually Changes

A 15-year trend and a two-year shock call for different responses, and conflating them is how systems waste a decade.

• Short-term tutoring is not enough. It addresses the pandemic gap, not the structural one underneath it.
• The suspects are slow-moving. Attention habits, heavy screen use, reading instruction, and teacher shortages built up over years and will take years to shift.
• Trend beats rank. A country can hold its place in the table while its actual scores keep falling, so year-over-year comparisons matter more than the league position.
• Patience is a policy. The systems that recover will be the ones that pick durable changes and hold them across several PISA cycles, not the ones chasing the next report.

Main Takeaways

• PISA 2025 recorded the lowest OECD reading and math averages ever, but the decline predates COVID.
• Reading peaked around 2012, science around 2009, and math was falling before 2018 in many countries.
• The pandemic accelerated an existing slide rather than starting it.
• “Catch-up” spending targets the COVID dip, not the longer structural drift.
• Real recovery depends on sustained change measured across cycles, not single-year fixes.

For more on schools and learning trends, browse the DelightfulBlogs Education section, or follow the wider story through our News desk.

Frequently Asked Questions

Did the pandemic cause the PISA decline?

Not on its own. PISA scores in reading and math were already falling across many OECD countries before 2020. The pandemic deepened and accelerated the decline, but the trend started years earlier.

When did PISA scores start falling?

Reading performance across the OECD peaked around 2012 and science around 2009. Math was already declining before 2018 in a number of countries. The 2015 assessment showed slippage in all three subjects at once.

Which subjects have declined the most?

Reading has fallen the furthest, down 28 points across the OECD since 2015, followed by math at 22 points. Science was more stable this cycle, holding between 2022 and 2025.

Is “learning loss” the same as this long-term decline?

No. “Learning loss” usually refers to the specific setback from pandemic school closures. The PISA data shows a separate, longer decline that began before the pandemic and continues alongside it.

Can the decline be reversed?

Possibly, but not quickly. Because the causes are structural and long-running, most experts expect recovery to require sustained changes over several PISA cycles rather than a single round of catch-up funding.

Final Takeaway

The most useful thing about the 2025 results is the timeline they force into view. A two-year shock invites a two-year fix, and that is roughly what most systems tried. A 15-year drift asks for something harder: a decade of steady, unglamorous work on reading, attention, and teaching, measured patiently against the next few reports rather than the last headline. The countries that internalize that difference now are the ones likely to look better in 2028.