Field Notes

question

Is the gender-equality paradox a measurement artifact?

StatusWorkingReviewdue UpdatedStands on5 sourcesSupportsnotes

The gender-equality paradox (GEP) is the finding that countries with higher levels of gender equality, as measured by indices such as the World Economic Forum’s Global Gender Gap Index, often show larger sex differences in STEM participation, not smaller ones (Stoet and Geary 2018).

Richardson, Reiches, Bruch, and colleagues (2020) re-analysed the same data and argued that the paradox is partly an artifact of measurement choices: which countries are included, how STEM participation is operationalised, and whether the GGGI captures cultural attitudes toward women in STEM or merely aggregate outcomes such as parliamentary representation.

Herlitz, Honig, Hedebrant, and Asperholm (2024) conducted a systematic review of 54 articles and new analyses of 27 meta-analyses and large-scale studies. They found that more psychological sex differences are larger, rather than smaller, in countries with better living conditions, with economic indicators such as GDP being the most sensitive predictors. This finding is broader than the specific STEM-paradox claim and harder to attribute to measurement artifact in a single index.

Why it matters

The GEP debate is the strongest empirical test of whether sex differences are socially constructed. If the paradox is real, the pure social-construction account is harder to sustain: differences that are entirely cultural should shrink as cultures become more equal, not grow. If the paradox is a measurement artifact, the social-construction account survives unscathed.

This matters for the vault’s synthesis in Male variability, education, and life outcomes, which treats the GEP as evidence that biology and culture interact in complex ways that defy simple accounts in either direction. If the GEP collapses on closer inspection, that limb of the argument weakens.

What is already known

  • Stoet and Geary (2018) documented the paradox using UNESCO and WEF data.
  • Richardson et al. (2020) challenged it on measurement grounds; Stoet and Geary published a reply in the same issue.
  • Herlitz et al. (2024) found that the broader pattern — more sex differences are larger in richer countries — holds across multiple psychological domains and is not limited to the STEM-participation operationalisation that Richardson et al. criticised.
  • The Yalcinkaya and Adams (2020) cultural-psychological model offers an alternative explanation (self-expression values in WEIRD settings lead to identity-based career choice) that predicts larger gaps without requiring the biological account the paradox is often used to support.

Substantially yes, and the artifact has now been named

The question asked whether the paradox is a measurement artifact. The answer is largely yes, on two independent grounds that were not both visible when this note was written.

The STEM version rests on a metric its authors had to correct. The GenderSci Lab could not reproduce Stoet and Geary’s women-in-STEM figures from the UNESCO data cited, and reported this to Psychological Science, which required a corrigendum.1 The correction clarified that the headline quantity was never the percentage of women among STEM graduates, as the paper had said, but a constructed ratio the authors call a propensity: a/(a + b), where a is the share of women graduates taking STEM and b the share of men graduates taking STEM. Richardson and colleagues then showed the correlation has small effect sizes and loses significance under alternative measures of both gender equality and women’s STEM participation,2 and Marsh and colleagues found the same sensitivity in mathematics self-concept.

The general method cannot support the general claim. This is the deeper result and it postdates the note’s framing. Ilmarinen and Lönnqvist decomposed the difference-score correlation that nearly the whole literature relies on, across fifteen variables in three domains, and found the strong headline predictions were driven by one component: men’s and women’s country means correlate between .93 and .97.3

What survives

Not nothing, and the note should not overcorrect. Ilmarinen and Lönnqvist make a methodological claim, not the claim that no sex difference varies by country; their proposal is decomposition, not abandonment. The open part of this question is now narrower and sharper: whether any cross-country pattern survives when the constituent slopes for men and women are modelled separately, rather than whether the difference-score correlations are trustworthy. They are not.

The vault’s own use of this literature needs adjusting on one point. Male variability, education, and life outcomes cites Herlitz and colleagues (2024) as the review that makes the pure social-construction account harder to sustain. Ilmarinen and Lönnqvist cite that same review among the works raising alarms about difference scores, which suggests the vault’s summary of it is one-sided. Reading Herlitz directly is now the highest-value next step.

  • Read Herlitz et al. (2024) in full and establish whether it endorses the paradox, qualifies it on difference-score grounds, or both. The vault currently represents it as endorsement.
  • Find any study that models men’s and women’s country slopes separately, as Ilmarinen and Lönnqvist propose, and report what remains of the pattern.
  • Check whether any study has replicated the GEP finding using an alternative gender-equality index that Richardson et al. would accept as valid.
  • Test the Yalcinkaya and Adams (2020) self-expression model against the biological-variability account: both predict larger gaps in richer countries, but for different reasons, and a study that discriminates between them would advance the question beyond “is the paradox real or not.”
  • Find whether the most recent wave of PISA data (post-2022) changes the cross-national picture.

  1. Corrigendum: The Gender-Equality Paradox in Science, Technology, Engineering, and Mathematics Education, Psychological Science, 2020, setting out the propensity formula a/(a + b). ↩

  2. Richardson, Reiches, Bruch, Boulicault, Noll, and Shattuck-Heidorn, “Is There a Gender-Equality Paradox in Science, Technology, Engineering, and Math (STEM)? Commentary on the Study by Stoet and Geary (2018)”, Psychological Science 31, no. 3 (2020): 338–341. Stoet and Geary replied in the same issue. ↩

  3. Deconstructing the gender-equality paradox (Ville-Juhani Ilmarinen and Jan-Erik Lönnqvist, “Deconstructing the gender-equality paradox,” Journal of Personality and Social Psychology 127, no. 1 (2024): 217–237, doi:10.1037/pspp0000508); saved record. ↩

Built on 5 sources (5 external).

Working out connections…