PISA 2025 shows large differences in achievement associated with students’ socio-economic status, gender and immigrant background. These gaps do not all move in the same direction, and a narrower gap does not necessarily mean that lower-performing groups improved. The strongest example in 2025 is socio-economic inequality: across OECD countries, the science gap between advantaged and disadvantaged students narrowed, largely because scores fell among advantaged students.
The Main Achievement Gaps in PISA 2025
OECD averages show that the size and direction of a gap depend on which student characteristic and assessment domain is examined.
Socio-economic, gender and immigrant-background gaps use different group definitions. A score gap between the top and bottom quarters of socio-economic status is not directly interchangeable with a gender difference or an immigrant-background difference. Raw subgroup gaps can also differ from gaps calculated after accounting for other student characteristics.
Socio-Economic Background Remains Closely Related to Achievement
PISA measures socio-economic status with the index of economic, social and cultural status, or ESCS. In PISA 2025, socio-economic status accounted for about 12% of the variation in science performance within countries on average across OECD countries. A one-unit increase in ESCS was associated with about 36 additional science score points.
The contrast is even clearer when students are divided into national socio-economic quarters. Students in the top quarter of ESCS scored 85 points higher in science than students in the bottom quarter, on average across OECD countries. The corresponding gaps were 77 points in reading and 83 points in mathematics.
| PISA 2025 measure | Science | Reading | Mathematics |
|---|---|---|---|
| Advantaged minus disadvantaged students | 85 points | 77 points | 83 points |
| Average score difference associated with one ESCS unit | 36 points | 32 points | 35 points |
| Academically resilient disadvantaged students | 12% | 13% | 12% |
Country and economy patterns vary widely. In science, the gap between the national top and bottom quarters of socio-economic status exceeded 100 points in several systems, including Bulgaria, Germany, Hungary, Luxembourg, the Netherlands, Romania, the Slovak Republic and Zambia. At the other end of the distribution, 17 participating systems had gaps below 50 points.
A weaker relationship between socio-economic status and achievement does not automatically imply high achievement. PISA treats the strength of the socio-economic gradient and the overall level of performance as separate dimensions. A system can have relatively small background-related differences while still recording low mean scores.
Why the Socio-Economic Gap Narrowed After 2022
Changes in the size of a gap need to be decomposed into the score movements of the groups on each side of it.
Between PISA 2022 and PISA 2025, the OECD-average science gap between advantaged and disadvantaged students narrowed by 13 score points. The published changes were about +5 points for disadvantaged students and −8 points for advantaged students. OECD analysis therefore attributes most of the narrowing to lower performance among advantaged students rather than a broad rise among disadvantaged students.
Science and Reading Changed Differently Across Socio-Economic Groups
OECD-average score change from PISA 2022 to PISA 2025
Positive values indicate higher mean performance than in 2022; negative values indicate lower performance. The chart shows reported OECD-average changes rather than a causal effect of socio-economic status.
The reading pattern is even clearer. Advantaged students scored about 20 points lower than in 2022, compared with a decline of about four points among disadvantaged students. The reading socio-economic gap consequently narrowed by about 17 points.
In mathematics, the socio-economic gap narrowed by about 11 points on average across the OECD comparison set. Advantaged students declined by about 14 points, while the change among disadvantaged students was not statistically detectable at the OECD-average level.
- A smaller subgroup gap can result from improvement among the lower-scoring group, deterioration among the higher-scoring group, or both.
- In the OECD-average 2022–2025 results, falling scores among advantaged students account for much of the narrowing in socio-economic gaps.
- Gap size therefore needs to be read together with the absolute performance of both groups.
Socio-Economic Disadvantage Is Also Visible Below Level 2
Mean-score gaps are only one part of the distribution. Among OECD countries, 37% of socio-economically disadvantaged students scored below Level 2 in science, compared with 12% of advantaged students. The difference was therefore about 25 percentage points.
This pattern does not mean that low socio-economic status determines an individual student’s outcome. PISA also identifies academically resilient students: students in the bottom quarter of ESCS in their own country or economy who nevertheless score in the top quarter of science performance within that system.
About 12% of disadvantaged students met this definition in science on average across OECD countries. The share was below 10% in a group of participating systems but reached 22% or more in Cambodia, Kenya, Rwanda and Uzbekistan.
A resilient student is compared with students in the same country or economy. The measure should not be read as the percentage of disadvantaged students reaching an internationally fixed high-performance threshold.
Gender Gaps Reverse Direction Across Subjects
PISA 2025 does not show one consistent gender pattern across science, reading and mathematics. At the OECD-average level, girls scored about one point above boys in science, girls scored 30 points above boys in reading, and boys scored 13 points above girls in mathematics.
The Gender Gap Changes Direction by PISA Domain
Mean score difference calculated as boys minus girls, OECD average, PISA 2025
Negative values indicate higher mean scores among girls; positive values indicate higher mean scores among boys.
The near-zero OECD-average difference in science masks variation across individual systems. Boys scored higher in science in 17 countries and economies, while girls scored higher in 35. In other systems, the estimated difference was not statistically distinguishable.
The contrast is much stronger in reading. The OECD-average reading difference was about 30 points in favor of girls. Mathematics moved in the opposite direction, with boys averaging about 13 points more than girls.
| PISA 2025 OECD average | Science | Reading | Mathematics |
|---|---|---|---|
| Girls’ mean score | 483 | 476 | 456 |
| Boys’ mean score | 481 | 446 | 470 |
| Boys minus girls | −1 | −30 | +13 |
Gender Differences Are Larger at Some Parts of the Performance Distribution
Average scores can hide differences in the proportions of students below Level 2 or at Levels 5 and 6. PISA 2025 shows different gender patterns at both ends of the distribution.
| Domain | Boys below Level 2 | Girls below Level 2 | Boys at Levels 5–6 | Girls at Levels 5–6 |
|---|---|---|---|---|
| Science | 27% | 24% | 8% | 6% |
| Reading | 37% | 25% | 5% | 7% |
| Mathematics | 33% | 36% | 10% | 6% |
The reading distribution is especially uneven by gender: 37% of boys and 25% of girls were below Level 2 on average across OECD countries. Mathematics shows the opposite pattern at the upper end, where about 10% of boys and 6% of girls reached Levels 5 or 6.
A small difference in mean scores does not imply that the two groups have identical performance distributions. Science is the clearest example: the OECD-average mean difference between boys and girls is very small, yet the shares below Level 2 and at Levels 5–6 are not identical.
Immigrant-Background Gaps Change After Accounting for Other Factors
PISA defines immigrant students as students whose mother and father were both born outside the country or economy where the student took the assessment. A non-immigrant student has at least one parent born in the country or economy of assessment.
Across 29 OECD countries with the required data, about 18% of students had an immigrant background in PISA 2025. About 10% were second-generation immigrant students and 8% were first-generation immigrant students.
Immigrant background is closely intertwined with socio-economic status in many systems. On average, about 41% of immigrant students were in the disadvantaged socio-economic group, compared with about 21% of non-immigrant students.
Language background adds another layer. Across the same OECD comparison set, 16% of all students reported mainly speaking a different language at home from the language used in the PISA assessment. Among first-generation immigrant students the share was 71%; among second-generation immigrant students it was 47%.
The OECD-average snapshot reports a raw science mean of 497 for non-immigrant students and 455 for immigrant students, a difference of roughly 43 points in favor of non-immigrant students for the countries included in that average. This raw difference should not be interpreted as an isolated immigrant-status effect.
When OECD analyses account for students’ socio-economic status and language spoken at home, immigrant-background differences generally become smaller. Among countries and economies included in the adjusted comparison, immigrant students scored higher in science in 13 systems, non-immigrant students scored higher in 21, and 10 showed no statistically detectable difference after those background variables were taken into account.
The raw gap describes the observed difference between two student groups. The adjusted comparison estimates the remaining difference after accounting for socio-economic status and home language. Neither measure establishes that immigrant status itself causes a score difference.
The Share of Immigrant Students Does Not Explain System Performance by Itself
PISA 2025 does not support the idea that countries with larger immigrant-student populations necessarily have lower average achievement. The OECD reports a positive cross-system relationship between the share of immigrant students and mean science performance in 2025, while also showing wide differences among systems with similar immigrant-population shares.
For example, Estonia and Iceland each had immigrant-student shares of roughly 10%, yet their average science scores were far apart. This is one reason country-level immigrant-background results need to be read together with socio-economic composition, language background and the characteristics of each immigrant population.
Background Gaps and Score Trends Measure Different Things
PISA 2025 illustrates why changes in inequality and changes in overall achievement should be kept separate. Socio-economic gaps narrowed across the OECD in all three core domains between 2022 and 2025, yet this often occurred while scores among advantaged students declined.
Gender patterns also shifted differently by domain. The OECD-average science gender gap was broadly stable, while the mathematics gap widened in favor of boys between 2022 and 2025. Reading remained the domain with the largest average difference in favor of girls.
Immigrant-background trends require another type of caution because countries differ greatly in the size, socio-economic composition, generation status and home-language profile of their immigrant student populations. The OECD therefore presents both raw comparisons and estimates that account for selected background characteristics.
A smaller gap is not automatically an improvement in outcomes
An achievement gap can narrow while both groups decline, or because the higher-scoring group declines more. PISA subgroup comparisons are most informative when the gap, the score level of each group and the direction of change are examined together.
How PISA Defines the Main Student-Background Measures
- ESCS
- The PISA index of economic, social and cultural status. It is the main PISA measure used to compare students’ socio-economic backgrounds.
- Socio-economically disadvantaged student
- A student in the bottom quarter of the national ESCS distribution within the country or economy being assessed.
- Socio-economically advantaged student
- A student in the top quarter of the national ESCS distribution within the country or economy being assessed.
- Academically resilient student
- A socio-economically disadvantaged student who scores in the top quarter of achievement among students in the same country or economy.
- Gender gap
- The difference in performance between boys and girls. The direction of the gap varies across science, reading and mathematics.
- Immigrant student
- A student whose mother and father were both born outside the country or economy in which the student took the PISA assessment.
- First-generation immigrant student
- A student born outside the country or economy of assessment whose parents were also born outside it.
- Second-generation immigrant student
- A student born in the country or economy of assessment whose parent or parents were born outside it.
- Raw immigrant-background gap
- The observed mean-score difference between immigrant and non-immigrant students before other student characteristics are taken into account.
- Adjusted immigrant-background gap
- A model-based comparison after accounting for selected characteristics such as socio-economic status and language spoken at home.
Data Sources and Measurement
The OECD released PISA 2025 Results (Volume I), Future-Ready Students, on September 8, 2026. PISA 2025 assessed more than 760,000 students across 91 participating countries and economies, representing roughly 33 million 15-year-olds.
The figures on socio-economic gradients, advantaged-disadvantaged gaps, academic resilience, gender differences, immigrant background and home language come from the OECD PISA 2025 Database and the Chapter 2 annex tables in Volume I.
Socio-economic results use the I.B1.2b table series, gender results use I.B1.2c, and immigrant-background results use I.B1.2d. OECD-average populations differ between some measures because subgroup data are not available or comparable for every participating system.
Small score differences should not automatically be treated as statistically distinguishable. Country and economy estimates are based on samples and plausible-value methodology, and some systems carry sampling or questionnaire-coverage cautions in the OECD tables.