The OECD's average PISA reading score falls below 445

Where things stand

• The latest result doesn't count. PISA 2025 results came out on 8 September 2026, before this question opened. The OECD average in reading was 460.94 (Annex B1, Table I.B1.2a.2). It is also too high, and its testing period ended in 2025, before the 16 September 2026 opening date. Later reprints of the 2025 figures don't count either.
• Only PISA 2029 can resolve this. No other PISA cycle tests students between now and the end of 2031. In PISA 2029, reading is the main subject. Field trials run in 2028, the main testing in 2029, and the OECD brochure schedules "publication of results" in 2030. PISA 2025 was released in September, so September or December 2030 is the most likely date.
• Early quarters are close to zero. No qualifying publication can appear before roughly the third quarter of 2030.
• What is needed. The 2029 headline average must fall below 445. That is a drop of more than 15.9 points from 460.94.

Trend evidence (computed from OECD StatLinks stat.link/xgs41b and stat.link/mrq53f)

• Long-run series. The average for the 23 OECD countries with data in every cycle peaked at 501.4 in 2012. It then fell to 497.2 (2015), 492.8 (2018), 481.9 (2022) and 465.8 (2025).
• Pace is accelerating. Per year, the declines were 1.4, 1.5, 2.7 and then 5.4 points. Across all OECD countries, the equal-weight average change per interval was −3.3, −2.8, −10.3 and −14.2. The share of countries declining rose from 0.68 to 0.92.
• It is a common decline, not country noise. Changes for individual countries barely carry over from one interval to the next (correlations of 0.03 and 0.12), but the OECD-wide decline has persisted and grown. Measurement error on the 2022–25 change is small (standard error about 1.3 points).
• A simple momentum model is pessimistic. I fitted a model where each period's yearly rate of change depends on the previous one, pooling reading, maths and science (18 data pairs). It projects a fall of about 15–17 points over 2025–29.
• Signs of stabilisation pull the other way:
  • In the US, the 2026 long-term trend assessment found 13-year-olds flat and 9-year-olds improving, and the testing firm NWEA reported small reading gains (Education Week, September 2026).
  • Norwegian national tests show smaller declines than PISA does.
  • School phone bans and literacy programmes are spreading.
• Signs of further decline: heavier use of AI chatbots, "hasty reading" (fast, wrong answers) nearly doubling since 2018, falling engagement, and continued reading declines among younger pupils in Norway.
• My central assumption for the average change within current members is −9.5 points (standard deviation 8.5).

Membership composition (a structural push downward)

• New members count in the headline average. PISA 2018 included Colombia once invited and before it formally joined. Latvia and Costa Rica were included right after joining.
• The three EU candidates are close to joining:
  • Romania said on 10 October 2026 it is "technically ready," with 24 of 25 committee reviews done.
  • Croatia expects to join by the end of 2026.
  • Bulgaria had completed 18 of 25 reviews in May 2026.
• Their 2025 reading scores are far lower. Using 2025 scores, each joiner lowers the average by: Croatia 0.2, Romania 1.2, Bulgaria 1.9, Peru 1.8, Argentina 1.8, Brazil 1.4, Thailand 1.8, Indonesia 2.5 points. If all eight joined, the average would be 450.2.
• Expected effect is about −5.8 points. This uses accession probabilities of 0.97, 0.97, 0.87, 0.40, 0.35, 0.20, 0.30 and 0.20 respectively. It also includes small effects from new accessibility features that may test more students with special needs, and a 10% chance the US is not included.

Model result

• I simulated 800,000 draws: the change within current members, the chance each candidate joins (with correlation between them), and how each candidate's own score changes.
• Result: P(2029 average < 445 | published) ≈ 0.47, with a median projected average of about 445.7.
• Sensitivity:
  • If only the three EU candidates join: 0.38.
  • If the within-country change averages −6: about 0.30.
  • If it averages −13: about 0.62.
• Final adjustments: about 95% that the results are out by the end of 2031, and about 98% that they are reported on the established reading scale.
• Cumulative probability: about 0.22 by the end of September 2030, about 0.40 by the end of December 2030, and about 0.44 by the end of 2031.

Strongest case against

The 2018–25 collapse may partly reflect post-COVID disengagement that is now easing; US data for the 2029 cohort show stabilisation. Policy responses could slow the decline to about 1–2 points a year, leaving the average near 450. Accessions in Latin America and Asia could also slip. I weighed this by centring the within-country decline well below the momentum model's projection.

Sources

• PISA 2025 Results Vol. I PDF (oecd.org/content/dam/.../73451bc5-en.pdf), StatLinks stat.link/mrq53f and stat.link/xgs41b, 8 September 2026
• PISA 2029 brochure (OECD)
• ACER, 1 April 2026
• NCES PISA schedule
• PISA 2018 Vol. I Reader's Guide (Colombia inclusion)
• Agerpres, 10 October 2026
• Lider/HRT, July 2026
• Economic.bg, May 2026
• Caretas, May 2026
• Nation Thailand, May 2026
• Japan MOFA, June 2026 OECD ministerial meeting
• Education Week, September 2026
• Aftenposten op-ed, September 2026

Question. By the last day of each quarter, has the OECD reported an OECD-average PISA reading score below 445?

Definitions.

• PISA reading score: the mean performance score for the overall reading domain in the OECD Programme for International Student Assessment’s main assessment of 15-year-old students, on the established PISA reading scale. Reading subscales, proficiency-level percentages, adult assessments and PISA for Schools results do not count.
• OECD average: the OECD’s headline, equally country-weighted average for the assessment cycle, using the membership, participating countries and inclusion rules adopted by the OECD for that release. It is not the population-weighted “OECD total,” an average across all participating economies, or a fixed-country subset used only for historical comparisons.
• Below 445: a published point estimate strictly less than 445. Use the most precise OECD-published value available for that release, prioritising the accompanying statistical table or downloadable data over rounded narrative text. If only an integer is published, 444 or lower qualifies; 445 does not. A confidence interval extending below 445 is insufficient.

Resolves YES if an OECD publication reports a qualifying score for an assessment observation period ending after the opening date, in accordance with the Observation period convention. Official preliminary results and subsequent official revisions qualify if they report the required statistic rather than a projection. A revision that first brings an eligible cycle below 445 qualifies on its publication date. Subsequent upward revisions do not undo YES under the Resolution window convention. No attribution of the score to AI is required.

Resolves NO at a quarter-end if no qualifying publication has appeared by then. Forecasts, unofficial calculations from country scores or microdata, scores for individual countries or student subgroups, and historical-cycle revisions whose observation periods ended on or before the opening date do not qualify. An OECD average carrying sampling-quality warnings still counts if the OECD publishes it as its headline average. A replacement metric on a different scale does not count unless the OECD also reports the corresponding score on the established PISA reading scale.

Resolving source. OECD-authored PISA results publications, their accompanying statistical tables and downloadable datasets, official OECD PISA database releases, OECD corrigenda, or OECD news releases explicitly reporting the qualifying statistic. The reference publication family is PISA Results (Volume I), with overall reading performance reported in its country/economy results tables; see PISA 2022 Results (Volume I): The State of Learning and Equity in Education, Chapter 2 and Annex B1, and the OECD PISA programme page for subsequent releases. Only OECD publications resolve this question; third-party reports do not substitute for them. Apply Public confirmation (A2), restricted to OECD-authored sources, Resolution date (A3), the Resolution window convention and the Observation period convention.