Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 64 of the 64 references it cites were found in a registry or an index; 33 of the 106 sources its claims cite were checked for support.

Simpson’s gender-equality paradox

Open access
Berggren, Mathias, Bergh, Robin · 2025 · Proceedings of the National Academy of Sciences

Abstract

Several cross-country examinations have found larger gender differences in Western countries. More recently, it has been argued, from an evolutionary standpoint, that equality may paradoxically increase gender differences, because it provides more freedom for men and women to pursue innate preferences. However, this paradox has primarily been examined with this cross-country methodology, opening up for other cultural differences to drive the results. For instance, measures developed in Protestant and Germanic-speaking countries, may not work the same in other cultural clusters of countries, and may not have the same statistical qualities there (e.g., in terms of reliability). Here, we reanalyze the results from multiple studies on the gender-equality paradox with country-level data available. We find that gender differences covary more strongly with cultural regions and data quality than gender equality, and that any variable higher in the West appears to achieve similar correlations as gender equality. Also, controlling for cultural regions consistently and strongly attenuates the association with gender equality, including to become statistically nonsignificant, or to switch direction. In other words, the baseline associations differ between and within cultural clusters (a Simpson's paradox), suggesting there is no simple causal relation between gender equality and expressed gender differences. Similarly, controlling for data quality indicators strongly attenuates the paradox. Finally, we show that, with and without controls, there is no consistent paradoxical association across many of the largest cross-cultural studies on gender differences, including newly analyzed data. The same is true for other country development variables considered in the gender-equality paradox literature.

Overview

CoChat Summary

## TL;DR

The paper argues that the gender-equality paradox is not supported once cultural clusters and data quality are accounted for, and that the apparent global pattern is a Simpson’s paradox. Across reanalyzed cross-country studies, cultural regions and data quality often predicted gender differences as well as or better than gender equality, while the main personality replication did not replicate in the original positive direction.

## Key findings

  • Across the reanalyzed studies, cultural regions were “always more strongly associated” with gender differences than gender equality, and controlling for cultural regions reduced the GEP correlations by 93 to 99% in the reported examples.
  • In the Germanic/Protestant West, there was no correlation between gender equality and gender differences at the regional level (ps > 0.07), but after controlling for language regions, all personality studies showed negative correlations of rs = −0.79 to −0.88 (ps < 0.002).
  • Data quality had similar or stronger associations with gender differences than gender equality; when both were entered together, the unique explained variance from gender equality dropped by 51 to 98% (average 80%).
  • In the new large personality replication using the Gosling-Potter Internet Personality Project, the baseline association was nonsignificantly negative, cultural regions were weak, and data quality had a stronger positive correlation; the original personality pattern “did not replicate.”

## Methods

  • The authors reanalyzed multiple prior cross-country GEP studies (1–6), plus ESS values data from refs. 32 and 33, reconstructing baseline associations and then adding controls for cultural regions and data quality.
  • Countries were grouped into seven broader cultural regions: Asia, Catholic West, Germanic/Protestant West, Latin America, Muslim Middle-East, Orthodox Europe, and Sub-Saharan Africa; some uncertain cases were tested in robustness checks.
  • For additional outcomes and predictors, they used the same general approach with linear regression, tested associations with gender equality and other developmental variables, and controlled for false discovery rate; they also analyzed new Big Five data from a dataset with over 9,000,000 participants and k = 131 countries.

## Limitations

  • The paper states that causality cannot be established from these correlational cross-country comparisons, and that stronger evidence would require longitudinal, natural-experiment, or experimental designs.
  • The cultural-region grouping is described as having a subjective aspect, though the authors report a robustness check for uncertain cases.
  • Evident from the text: data quality was only directly available or estimable in limited ways for some studies, and the paper notes that its indicators likely do not capture all relevant variation in data quality.

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

Integrity signals

Registry

Crossref · 10.1073/pnas.2422247122

Cites

64 of 64 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1. (Top row) Scatterplots between simulated gender equality (x-axes), and gender differences (y-axes) with three cultural regions (red circles–green triangles–blue squares) when gender equality and gender differences have a positive- (Left), no- (Middle), and negative (Right) association, but where there exists some confound that causes regions’ mean-level gender equality and gender differences to be positively correlated (for example because the gender difference dimension assessed is more culturally meaningful in regions higher in gender equality), leading to an apparent positive correlation in the aggregated datasets. Solid lines show the aggregated association, dashed lines show the association within regions. The Left figure shows a pattern allowed if gender equality causes gender differences. The Right- and Middle figures are examples of Simpson’s paradox, where the aggregated dataset provides an incorrect picture of the direction of the relationship between the variables when this confound is controlled for, and thus challenges an underlying causal effect in the positive direction. (Bottom Left) Share of male STEM college students across countries in 2015 against the Global Gender Gap Index used in (5) with a regression line (the positive relationship is significant). (Bottom Right) Share of U.S. male STEM bachelor students from 1966 to 2021, a period of increasing gender equality for one of the countries high in gender equality in the Bottom Left figure (see SI Appendix, section 1.2 for additional figures over time and further discussion). In relation to the Top figure: consider the blue-squares to be repeated measures of the United States and that different color-shapes correspond to different countries.
Fig. 1. (Top row) Scatterplots between simulated gender equality (x-axes), and gender differences (y-axes) with three cultural regions (red circles–green triangles–blue squares) when gender equality and gender differences have a positive- (Left), no- (Middle), and negative (Right) association, but where there exists some confound that causes regions’ mean-level gender equality and gender differences to be positively correlated (for example because the gender difference dimension assessed is more culturally meaningful in regions higher in gender equality), leading to an apparent positive correlation in the aggregated datasets. Solid lines show the aggregated association, dashed lines show the association within regions. The Left figure shows a pattern allowed if gender equality causes gender differences. The Right- and Middle figures are examples of Simpson’s paradox, where the aggregated dataset provides an incorrect picture of the direction of the relationship between the variables when this confound is controlled for, and thus challenges an underlying causal effect in the positive direction. (Bottom Left) Share of male STEM college students across countries in 2015 against the Global Gender Gap Index used in (5) with a regression line (the positive relationship is significant). (Bottom Right) Share of U.S. male STEM bachelor students from 1966 to 2021, a period of increasing gender equality for one of the countries high in gender equality in the Bottom Left figure (see SI Appendix, section 1.2 for additional figures over time and further discussion). In relation to the Top figure: consider the blue-squares to be repeated measures of the United States and that different color-shapes correspond to different countries.CC BY-NC-ND · PubMed Central
Fig. 2. CI for partial correlations between gender equality and gender differences controlled for cultural regions in the main studies (1-6, and European Social Survey, ESS). For (6), the standardized slope-coefficient in a weighted linear regression is used instead of the correlation (Materials and Methods).
Fig. 2. CI for partial correlations between gender equality and gender differences controlled for cultural regions in the main studies (1-6, and European Social Survey, ESS). For (6), the standardized slope-coefficient in a weighted linear regression is used instead of the correlation (Materials and Methods).CC BY-NC-ND · PubMed Central
Fig. 3. Relationships between some data quality indicators and measured gender differences in countries: Cronbach’s alpha, or (in 4) the proportion of floor-responders to a patience staircase-item. Barplot of the frequency of complete responders on that item is also shown to the Bottom Right (over 40,000 participants chose the floor option).
Fig. 3. Relationships between some data quality indicators and measured gender differences in countries: Cronbach’s alpha, or (in 4) the proportion of floor-responders to a patience staircase-item. Barplot of the frequency of complete responders on that item is also shown to the Bottom Right (over 40,000 participants chose the floor option).CC BY-NC-ND · PubMed Central
Fig. 4. (Top) The evolutionary explanation for the GEP states that gender equality and related developmental predictors increase gender differences in preferences (and sometimes on other dimensions). Other associations are unspecified. (Bottom) These results imply that cultural clusters (and data quality) drive the association, while other developmental variables are merely correlated with cultural clusters. No other systematic direct relationships are found (except potentially some negative ones between gender equality and gender differences).
Fig. 4. (Top) The evolutionary explanation for the GEP states that gender equality and related developmental predictors increase gender differences in preferences (and sometimes on other dimensions). Other associations are unspecified. (Bottom) These results imply that cultural clusters (and data quality) drive the association, while other developmental variables are merely correlated with cultural clusters. No other systematic direct relationships are found (except potentially some negative ones between gender equality and gender differences).CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

16 Valid · 5 Not valid · 26 Not checked · 106 Own results and interpretations

63 more claims · 4 with their cited sources checked

Create a free account to see all 153 claims and to verify the 6 references still open.

See all claims — free account
  • It may state two separate claims as one. 55
  • Probably not a substantial claim. 41

Create a free account to see these claims.

Citations

Cites

Show all 63 references

Cited by

See all 17

Metadata

Authors
Berggren, M., & Bergh, R.
Affiliations
Uppsala University (SE)
Corresponding author
Berggren, Mathias
Venue
Proceedings of the National Academy of Sciences
Published
2025-06-05
Volume
122
Issue
23
Pages
e2422247122-e2422247122
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
OpenAlex
W4411054864
Cited by
17
References
63
Citations by year
2026: 11 · 2025: 5
FWCI
16.64
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • Diva portal (Dalarna University Library) (submittedVersion, other-oa)
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Evolutionary Psychology and Human Behavior
  • Social Sciences › Psychology › Social Psychology › Cultural Differences and Values
  • Social Sciences › Gender Studies › Gender and Feminist Studies
Keywords
gender-equality paradox, gender differences, gender equality, Simpson's paradox, cultural regions, data quality, cross-cultural comparison, evolutionary psychology, country-level data, measurement reliability
MeSH terms
Gender Equity, Cross-Cultural Comparison, Female, Humans, Male, Sexism

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Simpson’s gender-equality paradox. https://app.cochat.ai/papers/10.1073/pnas.2422247122
The paper
Berggren, M., & Bergh, R. (2025). Simpson’s gender-equality paradox. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2422247122

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.