Popinz Guide

PISA 2025: U.S. Schools vs. Canada, Japan and Singapore

September 27, 2026 · Cross-sectional OECD PISA 2025 and NCES data (associations, not causes); two meta-analyses for research context · 6 min

DOI: 10.1787/73451bc5-en

The OECD released the first volume of PISA 2025 results on September 8, 2026. U.S. 15-year-olds averaged 502 points in science, 490 in reading and 463 in math. The OECD averages were 482, 461 and 463. The United States scored above the OECD average in science and reading, and matched it in math.

The OECD attached a data-quality warning to U.S. results. Behind the national averages, the lowest-scoring students fell further behind.

Read the U.S. results with the sampling caveat

The OECD marks U.S. results with an asterisk. Only 45% of sampled U.S. schools took part before replacement schools were added, and 54% after. Among students, 76% responded. In the OECD’s range-of-bias analysis, school non-participation could shift results by up to 15.4% of a standard deviation in the most extreme case, and student non-response by up to 15.9%. The OECD states that it cannot rule out bias or tell which direction it runs.

The student questionnaire has a bigger gap. More than 40% of responding U.S. students did not receive it. State-level decisions explain about half of that gap, and 9 of the 50 states were left out; the OECD expects the other half, such as schools that tested early in the window, to be unrelated to student characteristics. The OECD asks for “particular caution” with U.S. questionnaire results. In this article, that warning applies to the U.S. figures on socio-economic status, classroom climate and phone policies.

Reading: the decline sits at the bottom

U.S. reading fell from 504 in 2022 to 490 in 2025, a 14-point drop that matches the OECD average decline. According to NCES, 490 is the lowest U.S. reading average among the cycles it reports since 2000.

The loss is concentrated among struggling readers. The score at the 10th percentile fell from 356 to 332, while the 90th percentile held near 644, a change NCES does not rate as significant. The gap between those two points reached 312, against 264 across the OECD.

In reading, 13.3% of U.S. students reached Level 5 or above, compared with 5.5% across the OECD, and 25.7% scored below Level 2, against 31.0%.

Math: average, and 20 points below 2003

In math, the U.S. average of 463 sits 20 points below the 2003 result of 483. The share of students below Level 2, the OECD’s baseline, is 35.2%, close to the OECD’s 34.9%. The share at Level 5 or above is 8.0%. At the 10th percentile, U.S. math scores fell from 368 in 2012 to 334 in 2025.

The OECD also estimates how many points students gain in one year of schooling around age 15. Across ten systems, the typical gain is about 20 points. The U.S. estimate is 13.5 points in math and 12.0 in reading. These estimates carry wide margins of error, and the OECD warns against converting score gaps into “years of learning.”

Canada: the neighbor that fell further

Canada, where the 10 provinces and 3 territories run their own school systems, scored 510 in science, 490 in reading and 485 in math. Canada also carries an asterisk: the OECD expects a small upward bias of up to 15 points in science.

In math, Canada has lost more ground than the United States since 2003: 47 points (532 to 485) against 20 (483 to 463). Its reading average fell 44 points from 534 in 2000. The two countries now tie in reading at 490.

Canada still does better on equity. Advantaged Canadian students outscore disadvantaged ones by 59 points in science, compared with 104 points in the United States and 85 across the OECD. Socio-economic status explains 6.0% of the variation in Canadian science scores and 11.0% in the U.S. figures, which carry the questionnaire caveat.

Japan: math depth and orderly classrooms

Japan scored 538 in science, 503 in reading and 525 in math. In math, 22.3% of Japanese students reached Level 5 or above and 16.1% scored below Level 2.

Japan stands out on the OECD’s disciplinary climate index at 1.12, against 0.14 for the OECD average and 0.27 for the United States. The U.S. value relies on the questionnaire data flagged above.

A 2016 meta-analysis by Korpershoek and colleagues pooled 54 controlled studies, randomized and non-randomized, of classroom management programs published from 2003 to 2013. It found small but significant average effects (g = 0.22) on academic, behavioral and social-emotional outcomes. Those studies took place in elementary schools, so the evidence for high school classrooms is weaker.

Singapore: high scores without a small gap

Singapore scored 560 in science, 535 in reading and 563 in math. In math, 37.4% of its students reached Level 5 or above, and 11.2% scored below Level 2.

Singapore undercuts the idea that top systems must have narrow achievement gaps. Its advantaged students lead disadvantaged peers by 91 points in science, and socio-economic status explains 12.5% of score variation, values close to the U.S. figures. Its 90th-to-10th-percentile gap in math is 284 points, wider than the U.S. gap of 260.

The difference shows up among low scorers. Singapore’s 10th-percentile math score is 412, against 334 in the United States and 341 on average across the OECD.

What PISA cannot tell you

PISA tests a new sample of 15-year-olds each cycle. The data are cross-sectional, so they show associations between countries and between students, not causes. The OECD itself writes that no single explanation accounts for the declines. It lists complementary hypotheses: less sustained reading, heavier device use, higher absenteeism, lower family support and after-effects of the pandemic. A ranking cannot tell you that copying Singapore’s curriculum or Japan’s classroom routines would raise U.S. scores.

Takeaways for schools and families

Phones in school. In 2025, 31.3% of U.S. students attended schools where cell phones are not allowed, against 49.5% across the OECD. Across the OECD, 27% of students in schools with phone bans said classmates get distracted by devices, against 32% in schools with no device policy. The OECD finds that link after accounting for students’ and schools’ socio-economic profile. The U.S. share comes from the flagged questionnaire data.

Family conversations about school. Students reported less family support than in 2022 in almost all participating countries and economies, according to the OECD, with the sharpest drops among boys and disadvantaged students. A 2009 meta-analysis by Hill and Tyson pooled correlations from 50 studies of middle school students. It found that parental involvement was linked to higher achievement, with the exception of help with homework. The strongest link came from “academic socialization”: talking about expectations, goals and why school matters. For a school, that points toward family communication about expectations and goals rather than homework monitoring alone. Because the studies measure associations, this is correlational evidence.

References