Global education quality in 4 charts

Our World in Data presents the empirical evidence on global development in entries dedicated to specific topics. You can read more about this topic in our entries on Education Quality, The Global Rise of Education, and Financing Education.

Over the last century countries around the world have made huge progress in the expansion of schooling: While in 1870 there were just a few countries in which the average person had more than 4 years of education, today there are few countries in which the population has less education than that. This expansion continues until today – just in the last two decades the number of out-of-school children almost halved globally.

This expansion in schooling has been well documented, and it is undoubtedly a great achievement. But schooling is not the same as learning. So how are global learning outcomes changing over time? What progress have we made on education quality, and what are the challenges going forward?

Here are four charts that give us a global perspective. The data comes from a new research paper by Nadir Altinok, Noam Angrist and Harry Patrinos. At the bottom of this post we explain their underlying sources and approach.

Chart 1: A comparison of learning outcomes, country by country

Chart 1 shows that learning outcomes tend to be much higher in richer countries; but differences across countries are very large, even among countries with similar income per capita.

Here incomes are measured by GDP per capita (after adjusting for differences in prices across countries and time), and learning outcomes are measured by average student test scores (after homogenizing and pooling international and regional student assessments across education levels and subjects). Each bubble in this chart is a country, where colours represent regions and bubble sizes denote population.

This way of visualizing the data shows that average learning outcomes in poorer countries are lower than in richer countries: The top-performing country in Sub-Saharan Africa has a lower average score than the lowest-performing country in Western Europe.

Asian countries seem to outperform countries from other regions, followed by North America and Europe; Latin America and the Caribbean and Northern Africa are the next best performers, and Sub-Saharan Africa ranks last.

Low achievement can arise for many reasons. Absent and unqualified teachers, combined with poor complementary inputs, such as unhealthy environments and poor nutrition, can all play a role.

Additionally, often some of the most capable students in poor countries migrate. This selective aspect of migration can lead to lower scores in poor countries, while increasing them in richer countries where these talented students end up going to school.1

Chart 2: The evolution of learning outcomes over time

In the second chart we plot again average learning outcomes, but now we focus on changes over time. Specifically, this scatter plot compares outcomes in 1985 and 2015 (or the closest years with available data).

Among these countries we see a broad positive trend: Most bubbles are above the diagonal line, which means the majority of countries have seen improvements in learning outcomes over the last couple of decades. This is a great accomplishment! It shows that policies matter and learning outcomes can, and often do improve.

The error margin on these differences is often large, so small deviations from the diagonal line are not significant.

But it is worrying that many low-performing countries are substantially below the diagonal line. Consider the comparison between Chile and Burkina Faso in the center of the chart: Both countries had similar average scores a couple of decades ago, but while Chile has improved, Burkina Faso has regressed.

You can check country by country trends over time in this line chart.

Chart 3: Student achievement beyond average scores

In the third chart we plot the share of students who achieve minimum proficiency (i.e. the proportion who pass a global benchmark for minimum skills), against the share who achieve advanced proficiency (i.e. the proportion who pass a global benchmark for advanced skills).

Here we see that those countries where a larger share of students attain minimum proficiency, tend to also be countries where a larger share of students attain advanced proficiency. Better education lifts all boats.

Low-income, low-performing countries are clustered at the bottom of the global scale: the distribution of test scores within these countries is shifted down, relative to high-performing countries. The challenges are therefore much larger in these countries. Less than half of students in Sub-Saharan Africa reach the minimum global threshold of proficiency; and very, very few students achieve advanced skills.

Rich countries, on the other hand, tend to be less clustered. For example, Belgium and Canada have roughly similar average outcomes; but Canada has a higher share of students that achieve minimum proficiency, while Belgium has a larger share of students who achieve advanced proficiency. This shows that there is significant information that average scores fail to capture. The implication is that it's not enough to focus on average outcomes to assess challenges in education quality.

You can compare achievement above minimum, intermediate, and advanced benchmarks, country by country and over time, in these three line charts:

Chart 4: Should we really care about test scores?

The three charts above give an overview of the distribution of student test scores across countries and time. Why is this distribution important? What do we really learn from test scores anyway?

We have lots of good evidence on the private benefits from education. Education helps people be more productive and get better-paid jobs. And education also has positive spill-over effects on society. For example, there is evidence of positive social returns in terms of child mortality and corruption.

But the data here shows that, if we want to increase education, it's not enough to send kids to school. It's important that attending school also translates into acquiring specific skills.

The evidence shows that it is actually education in the form of skills, rather than mere school attainment, what really matters for predicting individual earnings and economic growth – and standardized tests are an imperfect but informative way of measuring skills.2

In the fourth and final chart we show the correlation between learning outcomes and GDP per capita over time. Each line in this plot tracks one country. As we can see most lines point north-east – this means GDP and learning outcomes often move together. Conditioning on baseline levels of GDP and schooling, the correlation is even stronger (see, for example, this chart from Hanushek and Woessmann using a similar dataset).3

Wrapping up

For me, the main message from these four charts is that the momentous education expansion that poor countries have achieved in terms of access to schooling has not been fully matched by an expansion in terms of quality, and this has important consequences for development opportunities in these countries.

In a way, this is an huge opportunity: Since school enrolment is growing and baseline learning outcomes are low, policies that increase education quality in poor countries can have a large positive impact, both for people in these countries, and for people in other parts of the world, via positive externalities.

About the data

Measuring learning outcomes in a way that enables us to make comparisons across countries and time is difficult. There are several international standardised tests that try to measure learning outcomes in a systematic way across countries; but these tests are relatively new, and they tend to cover only specific geographical areas and skills.

One possible approach to learn from all these overlapping but disparate international and regional tests, is to put them on a consistent scale, and then pool them together across skills to maximize coverage across years and countries. This is exactly what Nadir Altinok, Noam Angrist and Harry Patrinos did in a new working paper: Global Data Set on Education Quality (1965–2015). They collected data from a large set of psychometrically-robust international and regional student achievement tests available since 1965, and they linked them together in a common measurement system. The four charts above rely on their data.