PISA 2025: some results taking into account years of schooling and educational track

Recently, the official report on PISA 2025 results was released. Many researchers, educational practitioners, policymakers and journalists have started to discuss the results. In general, the results are not optimistic, since many countries demonstrated a decrease in PISA mathematics, reading and science performance. Some media have linked this decline to the active use of LLMs and digitalization.However, many researchers are cautious about drawing conclusions, as a deeper and more detailed analysis of the data is needed to understand how serious the decline really is and what factors might be associated with it.

Unexpectedly (for me, at least), I participated in a discussion regarding PISA 2025 results in a Telegram chat for parents, here in Belgrade. Unexpectedly, because I got used to thinking that  the results of international large-scale assessments (ILSAs) are interesting for policymakers and researchers, but not for parents.

I was wrong.

Today, many parents want to know more about education in general. They are interested in factors influencing educational performance and success, research findings, the specific characteristics of educational systems, etc. It turns out that this is especially important for some immigrant parents who have moved to a new country and want to know what to expect.

I suggested organizing a small discussion about ILSAs, the specific features of PISA compared, for example, with TIMSS, the results for Serbia and some interesting findings obtained from ILSAs. Last week, I gave a lecture, and we had an interesting meeting (I hope!).

I decided to transform some of information from this meeting into a blog post. My main idea is that looking only at average PISA scores and their changes over time is not enough to conclude that the quality of education in a country has declined or improved. A more detailed analysis of the data is needed to understand what is actually happening. It is necessary to take into account the results of different types of comparative studies and studies conducted within countries.

Besides, even when discussing PISA results, comparing different countries and analysing trends, we should take into account many factors and differences between educational systems.

First, as I have been studying the so-called school effect and grade effect (the effect of one year of schooling), it seems obvious to me that when comparing results between countries, we should take into account the distribution of 15-year-old students (the target population in PISA) across grades and educational programmes (general, vocational with direct access to higher education, or vocational without direct access).

So, I think it is useful not only to compare results between, for example, the 2022 and 2025 cycles and between countries, but also to examine the structure of the samples and compare PISA scores for students enrolled in the same programme and having the same number of years of schooling.

PISA Math Scores in selected countries

So, let's have a look at some results. I selected several countries (high-performing countries, the USA, Balkan countries and some European countries for comparison). The analysis was conducted using the Rrepest package, which was created for analysing ILSA data, and the graphs were created using ggplot2.

In this blog post I use only mathematics results for demonstration of approach. In principle, to reliably estimate trends, we should compare PISA cycles in which the domain was the main focus. For example, to examine trends in mathematics performance, we can compare the results from 2012 and 2022. However, official PISA reports also compare results from adjacent cycles, and so do I.

 First, let's look at the mean mathematics results by country (Fig. 1). Asian countries are traditionally the leaders: Singapore, Chinese Taipei, Japan and Korea. Among these leaders, Singapore demonstrated a decrease in mathematics scores.

In the selected European countries (Germany, Austria and France), results are lower than in the leading countries. In France, mathematics performance became worse in 2025.

Among the Balkan countries, Croatia and Slovenia demonstrated results similar to those in Germany, Austria and France, whereas Serbia, Montenegro and Greece had lower performance. Moreover, in Serbia, mathematics results in 2025 were lower than in 2022.

Fig.1. Mean mathematics scores

International grades and educational tracks

It is worth noting that in many countries, 15-year-olds in the PISA sample may have completed different numbers of years of schooling, so they may be distributed across grades 7 to 12. However, students are mainly enrolled in one or two modal grades, usually corresponding to grades 9–11.

In the selected countries, most students in the samples, or even the whole sample, are from grades 9 and 10 (Fig. 2).

Fig. 2. Proportion of students from differents grades

The proportion of students from different grades varies between countries. For example, in Japan (one of the leaders in the country ranking), 100% of 15-year-old students in the PISA sample are in grade 10, while in Croatia the proportion of tenth-graders is close to 20%.

In general, the distribution across grades remains stable across PISA cycles. However, even in this small number of selected countries, we can see several exceptions. First, in the USA, the proportion of ninth-graders in the sample increased in 2025 compared with 2022. At the same time, students from grade 11 were not included in the sample in 2025, although they were represented in 2022.

Another case is observed in Serbia: the proportion of ninth-graders decreased in 2025, while the proportion of tenth-graders increased.

 It is worth noting that here in Serbia, basic education includes eight years of schooling. After grade 8, students may continue their education in a general secondary school (called a gymnasium) or in a vocational school (three or four years of vocational education). The first year of education in a gymnasium or vocational school corresponds to international grade 9.

Years of schooling are important. As PISA reported, one year of schooling is associated with approximately 20 PISA score points (based on PISA 2025 results). In our paper based on PISA 2012 results, we obtained very similar results. So, a higher proportion of tenth-graders in the sample may contribute to a higher average country score.

In addition, another factor that can affect average country results is the distribution across different educational tracks or programmes. Based on the PISA classification, we can identify a general track (more academically oriented, with direct access to higher education) and vocational tracks (with or without direct access to higher education). In Fig. 3, you can see the sample structure by grade and educational track.

The results revealed that, for example, in Japan, most students in the sample are in the general track, while in Serbia and Montenegro, most tenth-graders are in the vocational track.

Fig. 3. Proportions of students from different grades and tracks

The proportion of vocational students is important because, in the selected countries (and not only there), students in vocational programmes often demonstrate lower performance than students in general programmes within the same grade (Fig. 4).

Fig. 4. Math scores for students in general and vocational tracks (Grdae 10)

So, if we want to compare countries, it may be more informative to compare the results of students with the same number of years of schooling and in the same track.

Results fro different grades and tracks

Below, we can see the results of ninth-graders in the general track. For example, ninth-graders in Chinese Taipei demonstrated the highest results among the selected countries in 2022, but their results declined in 2025, while the average country results did not change significantly. In contrast, the average results for Singapore decreased, while the results of ninth-graders did not change significantly.

The average results for Croatia were lower than the average results for Singapore, Korea, Japan or Chinese Taipei. However, the results of ninth-graders were close to those of the leaders. A similar situation is observed in Serbia. While the average results for Serbia are significantly lower than those for Singapore or Chinese Taipei, the results of ninth-graders in the general track are close to those of the top-ranking countries.

In the grade 10 general track, other patterns can be identified.

Chinese Taipei demonstrated the highest results, and they did not change significantly since 2022. In contrast, in Singapore, the results for tenth-graders decreased in 2025, as they did in Slovenia and France.

However, in France, the results of tenth-graders in the general track are closer to those of the leaders than the results of ninth-graders. While ninth-graders in Serbia perform relatively close to students in the leading countries, the gap is much larger among tenth-graders. This suggests that something may happen during grade 10 that contributes to the widening gap in mathematics performance.

Now let's look at the results of tenth-graders in vocational programmes with direct access to higher education (Fig. 6).

In general, students in vocational programmes demonstrate lower mathematics performance than students in general programmes. However, there are substantial differences between countries. Chinese Taipei and Japan demonstrate the highest results among the selected countries, while Greece, Slovenia, France and Montenegro have considerably lower scores.

It is also interesting to look at changes between 2022 and 2025. In Serbia, Slovenia and France, mathematics performance decreased, while in Chinese Taipei and Croatia it increased. In other countries, the changes were relatively small or less clear. In Serbia, the decline is particularly noticeable, and the results of vocational students in 2025 are substantially lower than those of students in the leading countries.

Short summary

In general, these results are, of course, very preliminary. This analysis does not take into account students' socioeconomic background, gender, school characteristics, or differences between countries in many other factors that may be related to educational performance.

However, even such a simple and rather rough analysis shows that trends in PISA performance may differ between students in different grades and educational programmes. When we look only at average country scores and their changes over time, we cannot see these differences. Conclusions about declining or improving performance describe the country as a whole, but they do not tell us what is happening within different groups of students.

So, I think that anyone who is interested in PISA results and wants to understand what is happening with education in a particular country should analyse these data in more detail. National averages are a useful starting point, but discussions about PISA results should go beyond comparing average scores.

Of course, many other questions arise. PISA data provide a large amount of information for researchers, but there are also more general questions about what PISA results actually tell us about education.

One line of research focuses on what PISA measures. Unlike TIMSS, PISA tasks are not directly linked to school curricula. This raises the question of what aspects of educational outcomes PISA scores actually reflect and to what extent they can be used to evaluate the quality of education.

Another line of research examines whether PISA results can predict students' future educational trajectories and achievements, and to what extent they can be considered an indicator of educational success.

There are quite a few studies addressing these questions, and I would like to discuss some of their findings in my next posts.