How to read Australia’s latest education report card

The headlines will almost write themselves whatever the results of the 2025 Programme for International Student Assessment (PISA) are when they are released next Tuesday. PISA is the report card that OECD and other participating countries use to compare their educational performance, measuring the maths, reading and science skills of 15-year-old students every 3 years.

Australia has previously been a relatively strong performer in PISA, scoring above the OECD average across all 3 domains since 2006. But both Australia’s scores and the OECD average have been in steady decline over the past 15 years.

PISA scores use a scale that is designed to compare performance across countries and over time. For each domain, the scale was originally set so that the average performance of OECD countries was 500 points, with 100 points representing one standard deviation in the base year. Australia’s maths score has fallen from 520 in 2006 to 487 in 2022 – a big drop, but one that still leaves us performing at a level comparable with the UK, Denmark and Belgium.

Other than national pride, how much do these results matter? Potentially quite a bit. Educational performance has been causally linked to improved mental health, lower rates of incarceration, and greater voter participation and trust. Cross-country research has also linked better educational performance to stronger economic growth. Of course, richer countries also tend to have better-performing students for other reasons too, so the precise causal relationship between test scores and future growth is hard to pin down.

Understanding the implications of changes in PISA scores themselves is harder still because the test itself is not static. Several features of the assessment have changed over time, of which 3 are particularly important: who takes the test, how the test is taken and what questions are asked.

Who takes the test?

While the PISA sample is meant to be representative of 15-year-olds enrolled in Australian schools, the group that actually ends up sitting the exam has changed over time in several important ways.

One big change is that the share of students in participating schools who are excluded from the exam has risen from 1.79 per cent in 2009 to 5.14 per cent in 2022, driven by a more than tripling in the rate at which schools are excluding students for intellectual disability.

Of course, it may be reasonable for students with an intellectual disability to not sit PISA exams. But if these excluded students are lower performing, then having fewer of them sit the test each year has likely artificially boosted Australia’s performance. For example, if excluded students would have scored around the 5^th percentile, adding them back would deepen Australia’s decline by 4 points in mathematics and 5 points in both reading and science.

Unfortunately, the OECD provides no information on excluded students to produce a precise estimate of the impact. That also means it has no way of adjusting for the effects of exclusion rates in the results it publishes. Instead, it simply flags countries when the overall exclusion rate exceeds the 5 per cent threshold, which Australia has now done in every PISA cycle since 2015. Exclusion rates are even higher in some countries, reaching around 12 per cent in Denmark, and much lower in others, such as Mexico, which excludes only 1.4 per cent of students.

A second change in who takes the test has likely moved Australia’s measured performance in the opposite direction. The PISA sample is comprised of 15-year-olds rather than a particular year group. Between 2009 and 2022, the share of Year 10 students taking the test rose from 71 per cent to 83 per cent, while the share of Year 11 students fell from 19 per cent to 6 per cent. The share of Year 9 students remained broadly unchanged at around 10 per cent. This shift was largely driven by earlier changes to school entry policies in Queensland and WA.

This shift means that roughly 12.5 per cent of the PISA sample had one less year of education in 2022 than in 2009. Assuming these students would have performed at their year-level average, this shift would have reduced Australia’s PISA performance by more than 3 points across all subjects.

Put another way, part of the decline in PISA performance between 2009 and 2022 was likely not because the education system taught students less, but because the students sitting the test had less education.

How is the test taken?

In 2015, PISA moved from paper-based tests to computers in most countries. Fifty-seven countries took the test on computers, while 15 remained on paper. The computer test also changed how students could navigate the exam, with less scope to go back and revise earlier answers.

The OECD adjusted its methodology to account for the change, but research using students randomly assigned to paper and computer versions suggests some differences remained. After the adjustment, computer-based testing was associated with scores around 5–10 points lower in reading and mathematics in some countries, and as much as 19 points lower in science in Germany.

There is no equivalent estimate for Australia, so we cannot know how much this contributed to Australia’s declining test scores. But it did come precisely when Australia recorded its sharpest fall in PISA performance, with scores dropping by around 10 points on average between 2012 and 2015.

What questions are asked?

Recently, the test has evolved further. In 2018, PISA introduced adaptive testing for reading, where the difficulty of questions is determined by how students respond to earlier questions. This was extended to mathematics in 2022, and to science in 2025 so that adaptive testing is being used across all 3 core domains.

This makes the assessment of students’ capability more precise. An OECD study suggests adaptive testing reduces measurement error by around 4.5 per cent on average and by around 10 per cent for students at the top and bottom of the performance distribution, without systematically raising or lowering average scores. The science scores released on Tuesday will come from an adaptive test for the first time.

How much do we care?

There is also the question as to how much effort students actually put in. PISA is a low-stakes exam, with the results having no direct consequences for students unlike their other assessments. In Australia, students rated the effort they put into PISA 2022 at 7.2 out of 10, compared with the 9.0 out of 10 they would have put in if results counted towards their marks. Previous research has shown that differences in how seriously students treat PISA can affect both scores and country rankings.

If students are so indifferent, then how much should the rest of us care?

PISA remains important because it allows us to compare outcomes across countries (unlike national tests like NAPLAN). And large movements over time provide meaningful information – it surely says something that Australia fell from 9^th among OECD countries in mathematics in 2009 to 24^th in 2018. Or that we held up better during COVID than many peer countries. Or that Australia’s 15-year-olds outperform their American counterparts in maths (on average) but sit way behind Japan’s.

But some of the commentary next Tuesday will probably focus on much smaller movements in the headline numbers. On that front, a bit of perspective is in order. Testing for cognitive capability is hard, the test itself is evolving, and broader factors can subtly influence aggregate results in ways that are not always obvious.

We should care about PISA and its big picture messages but also understand the changing assumptions that sit behind the numbers.