About the data

Pisalytics shows how 15-year-olds have performed in PISA since 2000. Every figure on the site is computed from the OECD’s own student data and checked against the OECD’s published results.

Where the data come from

The figures come from the OECD PISA database: the public student data files for each PISA round (2000, 2003, 2006, 2009, 2012, 2015, 2018, 2022 and 2025), published by the OECD under the Creative Commons Attribution 4.0 licence (CC BY 4.0). They cover around 100 countries and economies; not all took part in every round.

Pisalytics is an independent site. It is not affiliated with or endorsed by the OECD, and the figures shown here are our own calculations from the OECD’s data.

Not included: regional samples that some countries draw in addition to the national one (for example Spanish regions, US states or Scotland), and Cyprus in 2022, whose data are not in the public file. Sub-national participants that the OECD reports as their own entity, such as Shanghai (2009, 2012) or Buenos Aires, are included under their own names.

How the figures are calculated

We follow the OECD’s methods (PISA Data Analysis Manual and technical reports):

• All plausible values. Each figure is computed once for every plausible value (5 per subject up to 2012, 10 from 2015) and the results are averaged.
• Weights. Every figure uses the students’ final weights, so it describes all 15-year-olds in school, not just those tested.
• Margins of error. Standard errors combine sampling uncertainty, estimated with PISA’s 80 replicate weights (balanced repeated replication with Fay’s adjustment, k = 0.5), and the uncertainty between plausible values. The margin of error shown is 1.96 standard errors (a 95% confidence interval).
• Low and top performers are the shares of students below proficiency level 2 and at level 5 or above, using the OECD’s fixed cut-off scores for each subject.
• Percentiles are the weighted percentiles of each plausible value, averaged.
• Socio-economic quarters are formed within each country and round from the PISA index of economic, social and cultural status (ESCS), and recomputed for every replicate weight so that the margins of error reflect the uncertainty in the quarter boundaries. For 2000, the ESCS comes from the OECD’s trend file.
• Gaps (for example girls minus boys) are computed for each plausible value and replicate, so their margins of error account for both groups coming from the same sample.
• OECD averages are the plain average of the member countries’ results, with standard errors combined as for independent samples. Trend averages (23 and 35 countries) are shown only for rounds where all their members have comparable results.
• Small samples. Figures based on fewer than 30 students or 5 schools are shown faded, as the OECD does not report them.

How the figures are checked

Every country’s figures are compared with the tables in the OECD’s PISA results reports (2018, 2022 and 2025 editions): about 85,000 published values, including means, percentiles, shares of low and top performers, results by gender, immigrant background, language and socio-economic quarter, and the OECD averages. The data pipeline stops if any value differs from the OECD’s by more than rounding (slightly more for socio-economic quarters, where the OECD splits students at the quarter boundaries in a way it does not publish).

Where the OECD’s published figures cannot be reproduced from the public data, the reason is documented and the site says so. The main cases:

• Austria 2000: the OECD later adjusted Austria’s 2000 results for a sample that under-covered vocational students. The public data give the original results, so Austria 2000 is shown but not connected to the trend line.
• Iceland 2022: the data were revised after the OECD’s first 2022 report; we use the revised data, as the OECD’s later reports do.
• Palestinian Authority 2022: the PISA 2025 report recalculated 2022 for the smaller 2025 coverage; we show the full 2022 sample.
• PISA 2025: the public file was updated after the OECD’s first 2025 report. Its socio-economic index differs from the one the report used, so results by socio-economic quarter for 2025 can differ from the OECD’s by several points in a few countries (and El Salvador’s index is missing). In 14 countries that recorded a third gender option, the public data only separate boys from everyone else, so “girls” there includes those students.
• Immigrant background: Japan’s 2022 and all 2000 classifications are derived from students’ answers on where they and their parents were born, using the OECD’s definition. Albania 2015 has no immigrant background in the main public file.

Comparing over time

Each subject’s results can be compared over time from the first round in which it was the main subject: reading from 2000, mathematics from 2003 and science from 2006. Earlier results in mathematics and science were on different scales and are not shown in the trend charts.

Whether a country’s result in a given round is comparable follows the OECD’s own trend tables. Rounds the OECD excludes (for example Spain 2018, or rounds where a country’s sample did not meet the standards) are shown as separate, hollow points. In 2015 most countries moved from paper to computer-based tests; the OECD linked the two so that results remain comparable. Some definitions changed over time, notably language at home in 2009. Notes under each chart explain what applies to the countries you have chosen.

What PISA can and cannot tell

PISA describes what 15-year-olds in school know and can do. It does not cover young people who have left school, and it measures some skills rather than everything schools teach. Differences between countries or groups describe outcomes; they do not by themselves show what caused them. Small differences, especially within the margins of error, should not be over-interpreted.

Using the figures

Charts and data can be downloaded from the explorer. Please cite the source as OECD PISA database, for example “Source: OECD PISA database; calculations by Pisalytics”. Terms are explained in the glossary.

Data last updated 7 October 2026.