Which countries have the best public schools?

12 Feb 2024

This year, my son will start school in the UK. In the four years he's been alive, he's already lived in four countries. It got me wondering, which countries are doing public school really well?

Doing education well

To measure education, you ideally want to measure outcomes, the knowledge and skills that students have. But what students?

There's two approaches, look at students in a single year level, or at a single age.

The Trends in International Mathematics and Science Study (TIMSS) measures students in Years 4 and 8. This lets them compare students at the same stage of education, but since countries start formal education at differen ages, the average student age will be different across countries.1 Differences you find might be because the average student is older or younger.

The Programme for International Student Assessment (PISA) takes the opposite approach: it measures students at age 15 instead of at a fixed academic year level.2 This has the benefit that everyone is developmentally in the same place, but then differences you see might be due to having more or less average years of formal schooling.

I bias towards the age-based approach, for a reason that was hard to put into words: I feel each year of my son's life is precious. If two education systems get him to the same place, but one a year later, I'd want to know that.

That means we'll use the PISA data. Let's dive in.

PISA proficiency estimates

PISA measures proficiency in math, science and reading of a sample of 15-year-olds, and then turns those into estimates for each country. Below is a ranking made just of public school performance.3

A handful of countries are doing very well, including Japan, Estonia, Canada and South Korea. Other countries, including ones I've lived in like the UK and Australia, are close to the overall OECD average but actually below-average for wealthy countries.4

How big is the gap with private schools?

Countries that are able to provide excellent public education are likely to be countries with higher social mobility and lower inequality. That said, I can't blame anyone for wanting to use what resources they have to help their children grow into full, excellent human beings.

This leads us to consider private education as well, and the inevitable question of how private schools overall compare to the public system in different countries.

Immediately, a few things stand out in the data.

Firstly, private school students do better in most countries, as we'd expect. This is most likely a combination of extra resources, and extra concentration of students from higher socio-economic backgrounds. The size of the effect can be large, half a standard deviation or more in some countries.

Secondly, private schooling might not mean "more parent money for better education" in every country. In countries like Japan and the Netherlands, where private schools did worse on average than public schools, private schools may fulfil a different purpose than simply adding more resources to the public system. For a true outlier like Slovenia, it could be a measurement error or an unusually small sample of private schools.

Thirdly, there's a substantial gap between wealthy and less wealthy OECD countries in terms of educational outcomes in public schools, although private schools close much of that gap. This is a reminder that the resources available to a country are a big factor in the quality of public education.

Looking forward

So, we found some exemplars of public schooling, such as Japan, Estonia, Finland and Canada; they might provide some clues as to what it means to do public schooling well. We also found a clear gap between public and private education in most countries, which is likely to be real and significant.

It's well-established that the resources invested in education significantly influence outcomes

The big question in my mind is how these gaps will change in the light of new technology.

When I went through school, the internet was just emerging and there was barely email. Today, I can hold conversations with AI about topics in economics, clarify misconceptions and quickly learn new things. Wikipedia exists and I can talk to it. My son's schooling could look quite different from my own, particularly in later years.

If the technology is cheap and widespread enough, it could reduce the gap between public and private education, and between more and less wealthy countries, by lifting students everywhere. If it is expensive, it will likely widen the performance gap even more.

Here I'm optimistic: the current tools for early childhood learning are already impressive,5 and costs for tools like ChatGPT will continue to reduce vastly over the next decade, giving us some hope for a bright and equitable future for kids worldwide.


  1. Estimated age of entry and exit from compulsory education in a few countries, shamelessly gleaned from ChatGPT:


    • Age at compulsory entry: 5 or 6
    • Entry year level: 1
    • Age at final year: 17 or 18


    • Age at compulsory entry: 6
    • Entry year level: 0 (Pre-primary)
    • Age at final year: 18 or 19


    • Age at compulsory entry: 3
    • Entry year level: -2 to -1 (École maternelle)
    • Age at final year: 17 or 18


    • Age at compulsory entry: 6
    • Entry year level: 0 (Prep year)
    • Age at final year: 18 or 19


    • Age at compulsory entry: 4 or 5
    • Entry year level: 1
    • Age at final year: 18 or 19


    • Age at compulsory entry: 5
    • Entry year level: 0 (Reception)
    • Age at final year: 17 or 18

    France (3 y.o.) and Finland (6 y.o.) are the apparent outliers in terms of compulsory age of entry, but the story is messier and more nuanced, since something doesn't need to be mandatory to be customary, and norms around early pre-primary care differ a lot between countries, and between socio-economic groups within countries.

  2. PISA estimates the math, science and reading skills of 15 year olds every every three years in participating OECD countries. The data is calculated by testing a sample of students in participating countries, then using statistical models to estimate the average proficiency of the whole student body.

    PISA's models are calibrated against a baseline set of students; the scores above are relative to that baseline, which had an average score of 500 and standard deviation of 100.

    It's not agreed that this is the best way to measure education, but it's one of the best datasets we have for now.

    See: https://www.oecd.org/pisa/

  3. I've summarised math, science and reading scores with an average. I chose the 2018 data so that I can ignore the effects of the pandemic and look at a more normal year, and filtered to look just at public schools.

    I used a geometric mean, a way of summarising these proficiency scores that captures the intuition that these dimensions of knowledge are valuable in how they combine, rather than individually on their own.

  4. The OECD expanded to include non-European countries in 2010, and we can clearly see that countries with fewer resources to invest in education score more poorly in public schooling.

  5. Khan Academy and Duolingo have particularly impressive free offerings for kids already, I'm excited to see how the space will evolve.