CoChat
Sign Up
PreprintCited by 34Cognitive Abilities and TestingPsychometric Methodologies and TestingRace, Genetics, and Society

‘National IQ’ datasets do not provide accurate, unbiased or comparable measures of cognitive ability worldwide

Open access
Sear, Rebecca·2022
DOI 10.31234/osf.io/26vfb
Open this paper in CoChat — free accountOpen PDF

Abstract

In 2002, Lynn and Vanhanen produced a dataset which claimed to provide average ‘national IQs’ for nation-states worldwide. Despite extensive critique, this dataset, and subsequent updated versions, have been used in many empirical publications. Here I evaluate the latest version, produced in 2019 by Lynn and Becker, and show that this dataset is not fit for purpose. The primary data sources are inadequate for calculating ‘national IQs’. Most data originates from samples which are wholly unrepresentative of their national populations. Many are convenience samples with small sample sizes, collected over a wide date range (the oldest from the 1940s), often including only children and often including individuals selected because they had particular characteristics (i.e., were deliberately chosen to be unrepresentative). The transformation of cognitive test data from these sources into ‘national IQs’ involved a series of questionable assumptions and arbitrary adjustments. Some of these assumptions were based on Lynn and Becker’s prior beliefs about intelligence, including that regional and racial differences in IQ exist. These methods will result not just in error in the dataset but systematic bias. Even if representative data and appropriate methods were available for constructing ‘national IQs’, it is impossible to generate comparable measures of cognitive ability from culturally diverse populations across the globe. ‘National IQ’ datasets therefore do not provide comparable, accurate and unbiased measures of cognitive ability worldwide, and should not be used to draw inferences about global variation in intelligence.

AI snapshot

Written from the paper’s full text.

TL;DR

The 2019 Lynn and Becker “national IQ” dataset is not fit for purpose because its primary sources are mostly small, unrepresentative samples and its construction lacks a transparent, unbiased methodology. The dataset therefore cannot provide comparable, accurate, or unbiased measures of cognitive ability worldwide and should not be used for inference.

Key findings

  • The 2019 dataset provides “measured IQ” for 131 countries and extrapolated IQs for 72 more, but 37% of samples have fewer than 1,000 individuals, 74% fewer than 5,000, 63% include only individuals under 18 years, and 30% of countries are estimated from a single sample.
  • Only 32.7% of samples are classified by Lynn and Becker as “national”; 12 samples were “foreign,” and regional bias is severe: sub-Saharan Africa has an average IQ of 70, 40% single-sample countries, 48%/88% of samples under 1,000/5,000, and 72% of samples under 18 years.
  • Six countries have IQs estimated from fewer than 100 individuals (Angola 19, Dominican Republic 34, Greenland 40, Uzbekistan 51, Republic of Congo 88, “Netherlands Antilles” 96), and 14 more from fewer than 200.
  • No clear search strategy or consistent inclusion/exclusion criteria are stated; of 11 studies Wicherts et al. selected as reasonable quality, only 2 appear in the 2019 dataset, and transformations of raw test data may introduce further bias, as in the Gambia example.

Methods

  • The paper evaluates the 2019 Lynn and Becker “national IQ” dataset (version 1.3.3, June 2019), including the secondary data and methodology published in *The Intelligence of Nations* and on the associated website.
  • It examines the dataset itself, summarizes data by World Bank world region, and assesses sample sizes, ages, sample types, and transformations into “IQ” scores.
  • It consults primary sources for samples from sub-Saharan Africa, chosen because of the region’s implausibly low scores.

Limitations

  • The paper states no formal limitations section.
  • The text makes evident that primary-source inspection was limited to sub-Saharan Africa; other regions were assessed mainly through Lynn and Becker’s own classifications and the dataset summaries.
  • The paper is a critical evaluation of an existing dataset rather than a new empirical data collection or independent cross-national cognitive testing study.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.31234/osf.io/26vfb

Cites

40 of 40 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

3 of 31 cited sources checked · 1 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Greater exposure to formal education is associated with higher scores on cognitive tests.

    “for example, greater exposure to formal education is associated with higher scores (Brinch & Galloway, 2012;Matarazzo & Herman, 1984).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Brinch & Galloway, 2012 source
    • cites Matarazzo & Herman, 1984 source
  • Interpretationcitation

    The calculated 'national IQ' of the Gambia of 49.78/50.57 is on the borderline of 'moderate intellectual disability' (IQ range ~35-49).

    “which is on the borderline of 'moderate intellectual disability' (IQ range ~35-49).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Patel et al., 2020 source

      The Clinical features > Moderate intellectual disability passage gives an IQ range beginning at 35–49 and extending to 50–55, supporting the numerical comparison. But the paper describes assessment of individuals, not a calculated national IQ for the Gambia; its Definition and Conclusions also caution that IQ alone cannot establish intellectual disability or its severity. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
  • Findingreasoning

    Many samples in the 2019 'national IQ' dataset include individuals chosen because they had particular characteristics, i.e., deliberately chosen to be unrepresentative.

    “often including individuals chosen because they had particular characteristics (i.e., samples which were deliberately chosen to be unrepresentative).”
    • Also stated in: § Summarising the dataset · plain
      “Samples were therefore often selected by study authors because they had particular characteristics; in other words, they were deliberately chosen to be unrepresentative of the national population.”
    Quote in the paperSubstantial claim
  • Findingown data

    The majority of samples in the 2019 dataset are biased towards younger individuals.

    “the majority of samples are biased towards younger individuals”
    Quote in the paperSubstantial claim
  • Findingreasoning

    The extent of biases in the 2019 'national IQ' dataset differs by world region.

    “The extent of these biases differs by world region”
    Quote in the paperSubstantial claim
  • May combine two claims. 24
  • Classified as probably not a substantial claim. 10

  • May combine two claims.

    Premisecitation

    Psychometric tests are culturally specific, affected by education, motivation, and a wide range of other environmental factors, and improve with coaching and on re-testing.

    “Psychometric tests are culturallyspecific, affected by education, by motivation, as well as a wide range of other environmental factors, and improve with coaching and on re-testing”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Anum, 2014
    • cites Dramé & Ferguson, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Duckworth et al., 2011 source

      The Abstract reports that incentives increased IQ scores, supporting the claim that test performance is affected by motivation. The Discussion also identifies attentiveness, anxiety, emotional ability, and physical health as relevant to performance. Our copy does not establish that the tests are culturally specific or affected by education, nor does it report improvement with coaching or re-testing; its discussion of cross-cultural differences concerns test motivation, not the cultural specificity of the tests. (LLM, conf=0.90)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Findingown data

    Sub-Saharan Africa has a particularly high proportion of single-sample countries, of small sample sizes and samples including only children in the 2019 dataset.

    “Sub-Saharan Africa, for example, has a particularly high proportion of singlesample countries, of small sample sizes and samples including only children (Table 1).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The 2019 'national IQ' dataset is biased, with sample sizes and ages of participants varying by world region.

    “and that the dataset is biased, with sample sizes and ages of participants varying by world region.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    Botswana's 'national IQ' in the 2019 dataset is estimated from one sample of 140 adolescents sampled in South Africa, used because they were from an ethnic group common in Botswana.

    “Botswana's 'national IQ' is estimated from one sample of 140 adolescents sampled in South Africa, and were apparently used to calculate the 'national IQ' of Botswana, not because there was evidence that there were from Botswana, but because they were from an ethnic group which is common in Botswana.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    The authors of the 2019 'national IQ' dataset include no clear methodology or search strategy for locating studies in the literature, nor are consistent inclusion or exclusion criteria stated.

    “The authors include no clear methodology or search strategy for locating studies in the literature, nor are consistent inclusion or exclusion criteria stated.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The majority of samples in the 2019 'national IQ' dataset are small and unrepresentative of national populations.

    “the majority of samples are small and unrepresentative of national populations.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Lynn and Becker conclude that it is probably impossible to introduce policies to increase the fertility of highly educated high IQ career women who remain childless.

    “The principal problem is the large number of highly educated high IQ career women who remain childless.It is probably impossible to introduce policies to increase the fertility of these women who have been educated out of their reproductive function.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    The authors of the dataset themselves draw explicit policy implications from these claims; Chapter 4 of The Intelligence of Nations is devoted to these implications, and the authors make five suggestions for 'strategies to increase national IQs'.

    “The authors of the dataset themselves draw explicit policy implications from these claims: Chapter 4 of The Intelligence of Nations is devoted to these implications, and the authors make five suggestions for”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingreasoning

    Some studies included in the dataset were conducted with the explicit aim of exploring how cultural factors affect test scores, and so provide evidence of the cultural specificity of these tests.

    “Some studies were conducted with the explicit aim of exploring how cultural factors affect test scores, for example, and so provide evidence of the cultural specificity of these tests.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    The 'national IQ' dataset has been used to draw conclusions about innate differences in intelligence between groups, particularly between races, and how these supposedly innate differences lead to a range of other economic, health and social outcomes.

    “Yet, this dataset has been used to draw conclusions about innate differences in intelligence between groups, particularly between races, and how these supposedly innate differences lead to a range of other economic, health and social outcomes”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Figueredo et al., 2020 source
    • cites Kanazawa, 2006

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Lynn, 2006

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    The fourth and fifth suggestions for increasing national IQs are 'positive eugenics; and negative eugenics'.

    “The fourth and fifth suggestions are”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Findingnone given

    Category labels were only provided for codes 1-5 in the online manual, though several samples were coded '6' in the spreadsheet.

    “Category labels were only provided for codes 1-5 in the online manual, though several samples were coded '6' in the spreadsheet”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    The 'national IQ' of Mali is based on a single study in which the authors state that use of the Ravens may substantially underestimate the intelligence of children in Mali and that this can be particularly problematic when comparisons are made across cultures using the same test and norms.

    “Results indicate that use of the Ravens may substantially underestimate the intelligence of children in Mali.This can be particularly problematic when comparisons are made across cultures using the same test and norms”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Dramé & Ferguson, 2019 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingreasoning

    Lynn and Becker discuss a range of eugenic policies which have previously been used by governments, without any discussion of either the wholly flawed science or of the human rights violations which underly eugenic policies.

    “Lynn and Becker discuss a range of eugenic policies which have previously been used by governments, without any discussion of either the wholly flawed science or of the human rights violations which underly eugenic policies.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Many of the studies included in the 'national IQ' dataset provide clear evidence that the cognitive tests used are unsuitable for comparative work, and some include very clear statements that their samples should not be used for comparative purposes.

    “Many of the studies included in the dataset provide clear evidence that the cognitive tests used are unsuitable for comparative work, and some include very clear statements that their samples should not be used for comparative purposes.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    TIMSS and PIRLS are surveys of student achievement conducted in many countries but not all, excluding almost all countries in sub-Saharan Africa.

    “These are surveys of student achievement conducted in many countries but not all, and excluding almost all countries in sub-Saharan Africa.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Table 2 may overestimate the proportion of 'national' samples; for example, a sample of 74 orphans from a single orphanage in Eritrea is coded in the dataset as 'national'.

    “An inspection of the primary sources also suggests Table 2 may overestimate the proportion of 'national' samples.For example, a sample of 74 orphans from a single orphanage in Eritrea is coded in the dataset as 'national'.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Lynn and Becker conclude that eugenic immigration policies designed to reduce immigration from supposedly 'low IQ' to 'high IQ' countries are likely to be ineffective.

    “They also conclude that eugenic immigration policies -designed to reduce immigration from supposedly 'low IQ' to 'high IQ' countries -are likely to be ineffective (p334).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    In version 1.3.3 of the 2019 dataset, IQ scores which were previously estimated to be <60 have been recoded to 60 in the UW column.

    “Note that in version 1.3.3 of the 2019 dataset, IQs scores which were previously estimated to be <60 have been recoded to 60 in the UW column”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Lynn and Becker conclude that it is probably impossible to introduce policies to reduce the fertility of those with low intelligence.

    “It is also probably impossible to introduce policies to reduce the fertility of those with low intelligence”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    It is impossible to replicate the authors' strategy or to understand why they included some studies and not others in the 2019 'national IQ' dataset.

    “it is impossible to replicate the authors' strategy or to understand why they included some studies and not others.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Data from 12 different types of cognitive tests are used to construct the 2019 'national IQ' dataset, with several tests having multiple sub-scales, only some of which may have been used in any study.

    “Lynn and Becker (2019) report that data from 12 different types of cognitive tests are used to construct the dataset, with several tests having multiple sub-scales, only some of which may have been used in any study (p17, Table 4).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The 2019 'national IQ' dataset provides 'measured IQ' for 131 countries and extrapolated IQs for a further 72.

    “This 2019 version of the dataset provides 'measured IQ' for 131 countries, and extrapolated IQs for a further 72”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The 2019 'national IQ' dataset does not provide comparable, accurate and unbiased measures of cognitive ability worldwide.

    “This 'national IQ' dataset therefore does not provide comparable, accurate and unbiased measures of cognitive ability worldwide”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The 2019 'national IQ' dataset is intended to provide accurate, unbiased and comparable measures of cognitive ability worldwide.

  • May combine two claims.

    Premisecitation

    Samples in the 2019 dataset had lower IQs if they had younger ages, were tested earlier in time, used older test norms and came from rural samples.

    “samples had lower IQs if they had younger ages, were tested earlier in time, used older test norms and came from rural samples.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Lynn & Becker, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The primary sources are not necessarily poor quality pieces of work themselves, but they cannot be used to produce nationally-representative, globally comparative estimates of cognitive ability.

    “It is not that the primary sources are necessarily poor quality pieces of work themselves, but that they cannot be used to produce nationally-representative, globally comparative estimates of cognitive ability.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    TIMSS and PIRLS are not suitable for calculating 'national IQ' as they represent educational attainment of children in school and are not intended to be used as measures of national intelligence.

    “These are not suitable for calculating 'national IQ' as they represent educational attainment, of children in school (so again not representative samples), and are not intended to be used as measures of national intelligence.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    'National IQ' datasets do not provide comparable, accurate and unbiased data on cognitive abilities worldwide.

    “'National IQ' datasets do not provide comparable, accurate and unbiased data on cognitive abilities worldwide.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    The nature of the biases in this particular dataset mean its use may lead to racist conclusions and legitimise eugenic arguments, even where that is not the intention of the authors.

    “The nature of the biases in this particular dataset mean its use may lead to racist conclusions and legitimise eugenic arguments, even where that is not the intention of the authors.”
    Quote in the paperNot a substantial claim
29 of 29

  • The paper relies on this without support.

    The 2019 'national IQ' dataset is intended to provide accurate, unbiased and comparable measures of cognitive ability worldwide.

    “I evaluate the dataset by considering its potential to provide comparative, accurate and unbiased measures of intelligence worldwide”
1 of 1

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Overall single-sample percentage conflicts with regional figures

    Minor

    The regional percentages and country counts imply that about 21% of the 131 countries have IQs estimated from a single sample, not the 30% reported in the total row. Rounding the regional percentages cannot account for the gap.

  • Some regional single-sample percentages are impossible for their country counts

    Minor

    The table defines this percentage over countries, but no whole-number count out of 17 countries rounds to 10%: one country is about 6% and two are about 12%. The East Asia & Pacific row therefore cannot have both values as printed.

2 of 2

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

  • Cites a retracted work: 10.1073/pnas.1018601108

    Info

    The cited work was retracted on 2025-05-01 (notice 10.1073/pnas.2508040122), not before this paper was published (2022-05-17).

    Cited work 10.1073/pnas.1018601108 · retracted 2025-05-01 · notice 10.1073/pnas.2508040122 · this paper published 2022-05-17

1 of 1

Cites

1 retracted

Flagged first, then by citations

Research works printed but not linked: 2

  1. Dendir, S. (2013). WIDER Working Paper No. 2013/086 Children's endowment, schooling, and work in Ethiopia. Helsinki: The United Nations University World Institute for Development Economics Research (UNU-WIDER). Retrieved from http://hdl.handle.net/10419/93694www.econstor.eu
  2. Ebbesen, C. L. (2020). Flawed estimates of cognitive ability in Clark et al. Psychological Science, 2020. Retrieved June 21, 2020, from https://psyarxiv.com/tzr8c

Cited by

Sorted by citations

Related

Authors
Sear, R.
Affiliations
London School of Hygiene & Tropical Medicine (GB)
Corresponding author
Sear, Rebecca
ORCID
Sear, Rebecca
Publisher
Center for Open Science
Published
2022-05-17
DOI
10.31234/osf.io/26vfb
OpenAlex
W4280624454
Cited by
34
References
40
Citations by year
2026: 3 · 2025: 3 · 2024: 22 · 2023: 5 · 2022: 1
Type
posted content
Language
en
Open access
Licence
Locations
  • PsyArXiv (OSF Preprints) (submittedVersion)
  • OSF Preprints (OSF Preprints) (submittedVersion)
  • PsyArXiv (OSF Preprints) (submittedVersion, cc-by)
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Cognitive Abilities and Testing
  • Social Sciences › Decision Sciences › Management Science and Operations Research › Psychometric Methodologies and Testing
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Race, Genetics, and Society
Keywords
national IQ, cognitive ability, intelligence testing, sample representativeness, convenience samples, measurement bias, cross-cultural comparability

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-29.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-29.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review