Data literacy · Peer-reviewed sources
Are National IQ Rankings by Country Reliable?
No. Essentially every country-by-country IQ table in circulation traces back to one family of datasets compiled by Richard Lynn and David Becker, most recently revised in 2019. The figures average a heterogeneous collection of studies using different tests, different age ranges, and different sampling methods, with values imputed from neighbouring countries where no study existed at all. A peer-reviewed analysis by Sear, Lawson, Kaplan & Shenk (2022) concludes that datasets built this way cannot accurately measure population intelligence.
Where the numbers come from
Richard Lynn (1930–2023) was a British psychologist who spent several decades assembling per-country IQ averages from whatever published studies existed, with David Becker collaborating on the later revisions. The 2019 revision is the version most often republished, and it is the reason the same set of numbers appears on many unrelated pages: there is not a competitive field of national IQ datasets, there is essentially one.
The collection strategy is the root of the problem. Because the aim was full country coverage, the dataset accepts whatever study was available — which means the inputs differ in the test instrument used, the age range tested, how the sample was drawn, how large it was, and what year it was collected. For some countries the underlying sample numbers in the low hundreds. For countries with no study at all, a value was estimated from geographic or demographic neighbours rather than measured.
Averaging results from instruments that were never designed to be compared, then filling the gaps by inference, produces a table that looks precise to three significant figures and rests on evidence that cannot support one.
The peer-reviewed critique
The most substantive published challenge is Sear, Lawson, Kaplan & Shenk (2022), which argues that national IQ datasets of this type cannot accurately measure population intelligence. Its three central objections are that the underlying studies are too heterogeneous to average, that many samples are too small or unrepresentative to describe a country of millions, and that imputing missing countries introduces large biases. The authors recommend researchers stop treating the aggregates as defensible measures.
These are methodological objections, and they stand independently of the separate ethical criticism of Lynn's broader work documented by the Southern Poverty Law Center. Set the ethics aside entirely and the data quality is still contested.
The retraction record
In June 2024, STAT News reported that researchers had called on academic journals to retract multiple papers built on Lynn-derived national IQ data. Scholarly bodies including the European Human Behaviour and Evolution Association have publicly distanced themselves from that body of work. As of 2026, several journals have issued expressions of concern; full retractions vary by journal and by paper.
The signal worth carrying away is not that the numbers are forbidden but that they have been formally contested in the literature that produced them. A figure under active retraction review is not a figure to cite in a decision.
What to use instead
If the underlying question is how educational outcomes or research capacity compare across countries, four primary-source measures are well documented, regularly updated, and designed from the outset to be compared:
- PISA scores (OECD) — Tests 15-year-olds in reading, mathematics, and science on a standardised instrument in 80+ countries, every three years, with full methodological documentation. View the data
- Mean years of schooling (UNESCO and World Bank) — Average years of formal education completed by adults aged 25 and over — a direct measure of attainment rather than a proxy for ability. View the data
- R&D intensity (World Bank) — Research and development expenditure as a share of GDP, which captures national investment in knowledge production. View the data
- Tertiary attainment (OECD) — Share of adults holding a college or university qualification, from Education at a Glance. View the data
None of these is a measure of intelligence, and none should be presented as one. They measure schooling and research investment — which is, in most cases, what a policy or journalistic comparison was actually reaching for when it grabbed an IQ table.
Reading a page that publishes them
A per-country IQ figure presented in the same layout and typography as a Census population count reads as though it carries the same evidentiary weight. It does not. Population counts come from systematic enumeration with published methodology and margins of error; the IQ figure next to it may rest on a few hundred test-takers, or on no measurement at all for that country.
The general habit worth keeping is the one on our guide to verifying a demographic figure: ask which dataset, which vintage, and which table a number came from. Applied to a national IQ ranking, that question does not have a good answer, and that is the finding.
Frequently Asked Questions
No. Essentially every country-by-country IQ table in circulation traces back to one family of datasets compiled by Richard Lynn and David Becker, most recently revised in 2019. The figures are averages drawn from a heterogeneous collection of underlying studies that used different tests, different age ranges, and different sampling methods, with samples for some countries numbering in the low hundreds. Where no study existed at all, values were imputed from neighbouring countries. A peer-reviewed analysis by Sear, Lawson, Kaplan and Shenk (2022) concludes that datasets built this way cannot accurately measure population intelligence.
Richard Lynn (1930–2023), a British psychologist, spent several decades assembling per-country IQ averages from published studies, with David Becker collaborating on later revisions. The 2019 revision is the version most often republished. Because the collection strategy accepted whatever study existed for a given country, the inputs vary enormously in test instrument, sample size, sample representativeness, and year — and for countries with no study, a value was estimated from geographic or demographic neighbours rather than measured.
The most substantive published challenge is Sear, Lawson, Kaplan and Shenk (2022), which argues that national IQ datasets of this type cannot accurately measure population intelligence: the underlying studies are too heterogeneous to average, many samples are too small or unrepresentative to describe a country, and the imputation of missing countries introduces large biases. The authors recommend that researchers stop treating aggregated national IQ figures as defensible measures. Other critiques have raised similar concerns about study quality and imputation method.
STAT News reported in June 2024 that researchers had called for the retraction of multiple papers built on Lynn-derived national IQ data. Scholarly bodies including the European Human Behaviour and Evolution Association have publicly distanced themselves from that body of work. As of 2026 several journals have issued expressions of concern, while full retractions vary by journal and by paper. The practical point for a reader is that the dataset has been formally contested in the peer-reviewed literature, not merely disputed online.
Four primary-source measures are well documented and comparable across countries: PISA scores from the OECD, which test 15-year-olds in reading, mathematics and science on a standardised instrument across more than 80 countries; mean years of schooling, published by UNESCO and the World Bank; research and development spending as a share of GDP, from the World Bank; and tertiary education attainment from OECD Education at a Glance. None of these is a measure of intelligence, but they are well-sourced measures of educational attainment and research investment, which is usually what a policy or journalistic comparison actually needs.
No. We publish figures we can trace to a primary statistical agency — the Census Bureau, the World Bank, the BEA, the CDC — and no such source produces a national IQ series. The same standard rules out crime rankings built from opinion surveys and composite "happiness" indices. Where a subject has a defensible primary source and where it does not is set out on our methodology page.
Related
This page summarises the published academic and journalistic record on country-level IQ datasets and links each claim to its source. It describes the dataset and the critique of it; it makes no claim about any individual or population.
Demographic data you can verify
Get demographics updates by email. No spam, unsubscribe anytime.