PISA 2025: what the results measure, and what they do not say

What PISA measures, and what it does not

PISA assesses how well 15-year-olds apply their knowledge to real-life situations, not how well they master national curricula (French Ministry of Education, September 2026). The 2025 cycle focused on science.

In France, 6,501 students took part, in 302 randomly selected schools (French Ministry of Education). Results are collective: no individual scores, no school league tables, no official country ranking. And scores carry confidence intervals: France's 483 in science is statistically indistinguishable from twelve OECD countries' scores; fourteen scored higher (DEPP, briefing note 26.39, September 2026).

France: three average scores, one downward path

France's three scores sit close to the OECD average (482, 463 and 461), but each has its own history. Science is roughly stable since 2022 (down 4 points) and down 12 points since 2015. Mathematics lost 16 points in three years, after 21 points between 2018 and 2022; reading lost 18 points in three years, after 19 between 2018 and 2022 (DEPP, briefing notes 26.39 and 26.40).

The slide extends beyond France: across a constant set of 24 OECD countries, scores fell by 11 points in science, 26 in mathematics and 31 in reading between 2015 and 2025 (ministry press pack). France's decline is steeper in mathematics (16 points against 9 since 2022), with eight places lost in the ministry's tally (press pack).

The distribution tells a harsher story: 33 percent of French students sit below Level 2 in reading, 36 percent in mathematics and 26 percent in science, and top performers in reading fell from 7 to 4 percent in three years (DEPP, briefing notes 26.39 and 26.40).

Japan, Estonia, Korea, Canada: comparing on conditions

Four systems lead the tables, above the OECD average in all three domains (science, mathematics and reading scores; OECD, 8 September 2026):

The OECD groups these four among the systems combining performance, inclusion and a weak grip of social background, with under 10 percent of the variation in science scores explained by socioeconomic status (Volume I). France sits at the other end: a 100-point gap between its most and least advantaged students in science, against 85 points across the OECD (DEPP, briefing note 26.39).

Traps remain, documented by the OECD: these four countries also lost 12 to 17 points in reading since 2022; Canada carries an asterisk for a sampling standard not fully met; and B-S-J-Z (China), at 597 points, covers four provinces, not a country.

• Japan: 538, 525 and 503; 89 percent reach at least Level 2 in science.
• Estonia: 527, 508 and 499; 13 percent are top performers in mathematics.
• Korea: 526, 522 and 501; 23 percent are top performers in mathematics.
• Canada: 510, 485 and 490; 84 percent reach at least Level 2 in science.

What an average does not say

A score is not pure aptitude. The effort thermometer averages 6.4 out of 10 in France, against 7.2 across the OECD, and declared effort barely tracks the score: 450 points for medium effort in reading, 447 for high effort (DEPP, briefing note 26.40).

Context also sits inside the measurement: socioeconomic status explains 14 percent of French science-score variation, against 12 percent across the OECD (OECD country note on France). And the OECD average shifts with the scope: the reading decline since 2015 is 28 points across current members, 31 on a constant set of 24 (OECD Volume I). France itself met every PISA quality standard (OECD country note).

What a school can take from it

PISA describes a system, not a school. Three signals still touch decisions made on site.

Reading is the measurable front: one student in three below Level 2 and, across the OECD, a slump concentrated in complex tasks such as weighing several sources. National assessments and the seconde placement test track these skills locally; the ministry's holiday reading lists, starting with this autumn's break (press pack), offer simple support.

Screens fall under school rules: where phones are banned, reported digital distraction drops by around 13 percent (OECD country note), yet only 27 percent of French students attend such a school, against 49 percent on average. French law extended the ban to upper secondary schools and requires every school plan to spell out its digital-use rules (press pack).

Family contact closes the list: 49 percent of French students say a parent takes a weekly interest in what they are learning, against 58 percent across the OECD, and the OECD links that weekly exchange to a 26-point science advantage, social background held constant (press pack). PISA sets the national frame, with a stated aim of gaining 20 points by 2029 (press pack); the decisions, and the local data behind them, stay at school level.