The gender equality paradox that wasn't

Do women seek STEM jobs mostly because they "have" to? A viral study and popular talking point rests on shaky methodology.

Spend enough time in online spaces, and you may come across a compelling — and counterintuitive — narrative: Sexism and repressive gender norms aren’t to blame for women being underrepresented in STEM fields relative to their population share. The real reason, or so this narrative goes, is that women are finally free not to enter STEM fields at all.

This account stems from a 2018 study in which the authors measured indicators of women’s science participation against the World Economic Forum’s Global Gender Gap Index. The headline finding is the so-called gender equality paradox, or the claim that women in more gender-equal countries are less likely to graduate with STEM degrees or pursue more traditionally masculine careers — and more likely to have typically feminine interests, personalities, and hobbies.

The authors propose an explanation: In more gender-equal countries, higher security and life satisfaction mean women are free to choose the careers they really want over high-paying technical ones. In this reading, efforts to increase women’s participation in STEM run counter to their revealed preferences for feminine careers, such as in education, carework, or administration. This view rapidly garnered traction in the popular discourse, becoming a talking point of both conservatives like Jordan Peterson and liberal writers like Noah Smith and receiving generally favourable coverage in mainstream outlets such as The Atlantic.

For individuals weary of feminist admonishments, the gender equality paradox was an effective counterpoint. But this popular success has outrun serious scrutiny of the methodology of the original paper.

The problem of measurement

The gender-equality paradox appears to be highly sensitive to measurement choices. One group of researchers re-analyzed the paradoxical result using a different measure of gender equality. Instead of the GGGI, which captures relative gender gaps and female political representation, they used a simplified measure: the Basic Index of Gender Inequality, itself first proposed by the authors of the original 2018 gender equality paradox study.

This is a common way to test the robustness of a scientific finding, since credible results should not be dependent on slight differences in measurement. When researchers analyzed the 2018 dataset using BIGI, they found no relationship between gender equality and the percentage of post-secondary STEM graduates who were women. A modest relationship did reappear when BIGI was measured against the ‘propensity’ (as defined by the 2018 paper) of women to graduate with STEM degrees, but it was substantially weaker than the original estimate and existed only in a newer sample.

Even in the original study, using the more favourable ‘propensity’ metric, gender equality as measured by GGGI accounts for just over one-fifth of the variation in the likelihood of graduating with a STEM degree across countries — an insufficient correlation to suggest a causal relationship without further evidence.

Did gender equality “free” women to become nurses? Source

The wider evidence shrinks too

There have been attempts to replicate the gender equality paradox in other domains, such as personality. In these cases, the scientists are testing whether more gender-equal countries have larger male-female gaps in common psychological indicators, like the Big Five personality traits. These studies typically find larger psychological sex differences in more equal countries.

But the methods used to arrive at this result expose another problem with this line of research. These studies rely on difference scores, created by subtracting the male mean from the female mean within each country and comparing that with measures of gender equality. This makes simple causal explanations suspect because the same difference score can arise from radically different conditions. Male math scores could be improving at twice the rate of female ones, or female math scores could be declining while male scores hold steady. The difference score would be the same in both scenarios. If the scores are the result of altogether different trends, they’re unlikely to reflect the same phenomenon.

A 2024 study from the University of Helsinki applied this critique systemically to the gender equality paradox. The authors re-analyzed multiple major studies exploring the paradox and looked at the underlying relationship between each variable – such as science attitudes and personality traits – and gender equality. Across all variables reanalyzed, there was no consistent pattern between them. Critically for the paradox, none showed men and women moving in opposite directions as equality improves.

Another problem: Men and women tend to move together statistically. If education, height, scientific literacy, or personality traits are higher for the men in country A than the men in country B, the women of country A will usually be higher in those traits compared to the women of country B as well — even if the gap between men and women in country A is also larger.

This can create major distortions if not handled carefully. When the averages for men and women are so strongly correlated, calculating a simple difference will remove all of their shared variance. This leaves only the small slice where the sexes diverge. That difference is then standardized by dividing it by what variation remains — and because so much of it was stripped out, the denominator used to calculate the final result collapses. This artificially inflates the apparent relationship and makes it look much bigger than it actually is.

The Helsinki researchers also noted that the studies they re-analyzed had strongly (in some cases, nearly perfectly) correlated means that were not sufficiently accounted for. They reevaluated the original results using more advanced statistical methods that standardized the scale of the variables and broke the difference scores into their core components. They found that the paradox’s magnitude had been overestimated by about 60% on average across studies. For the original 2018 study, the corrected effect for science-related attitudes was nearly half of what was originally reported.

The paradox that wasn’t

The results of the Helsinki study strongly imply, as the authors note, that the gender-equality paradox depends heavily on countries’ baseline differences distorting how the numbers are put together. Intuitively, this is because differences between nations are much greater than the variation between the men and women within a country. Thus, as the standard deviation is compressed and magnifies small sex differences, the resulting pattern will tend to mirror these country-based trends as it is inflated by their larger variation.

Evidence for this implication is provided by more recent research. This 2025 paper makes the key observation that gender roles and behaviours are not a single phenomenon across the globe. Rather, they are highly dependent on cultural context. Instead of treating all countries as one uniform group, the authors divided the world into seven cultural clusters — the Germanic / Protestant west, the Catholic west, Latin America, Orthodox Europe, the Muslim Middle East, Sub-Saharan Africa, and Asia — to control for differences in cultural context

The researchers analyzed multiple major studies published on the gender equality paradox using the same data and variables, but this time they controlled for culture. In practical terms, this means that differences in baseline characteristics — such as laws, cultural norms, or history — were held constant.

This matters mathematically. Clusters like this exist when groups of data points share unmeasured factors that affect the outcome. Ignoring them can be misleading, as the apparent differences may reflect the different cluster baselines, instead of the variable of interest.

One consequence of ignoring these patterns is a well-known statistical phenomenon known as Simpson’s paradox: a statistical quirk where a trend can appear to hold for an entire group, only to vanish or reverse once you break that group into the relevant sub-groups.

Imagine a study alleging class-based discrimination at a prestigious university. In this study, the poorer a student is, the less likely they are to be accepted into a program at the school. In effect, there is a correlation between family income and acceptance rates. But careful readers will notice a problem. Poorer students often have lower baseline test scores and GPAs. If you break the data down by GPA and SAT percentile, the correlation between income and acceptance could disappear or even reverse. In other words, the overall trend is likely a statistical artifact, created by combining students with very different academic profiles.

Researchers can address this problem by restricting analysis to countries that instead share a common baseline, or “intercept,” so that we don’t erroneously compare countries with very different cultural profiles without controlling for confounders. When the 2025 study accounted for this — so that Sweden was compared to Norway rather than to Iran — the gender equality paradox became nonsignificant in six of the datasets and reversed direction in one.

There was still finer analysis: The Germanic/Protestant West was subdivided into five language regions and the analysis was re-performed within that cluster. For the three studies on personality, all had large negative correlations between gender differences and equality, meaning that gender equality led to fewer gender-based personality differences. For the remaining four, the results were again nonsignificant.

The correlation between gender equality and science jobs on a global level seems to have been driven by Western countries, especially Germanic-speaking Protestant ones. Western countries, it seems, have unusually large gender gaps in domains such as STEM representation compared to the rest of the world. Across the board, whether a country was Western had a better correlation with gender gaps than gender equality did. Tellingly, invented predictors that correlated with Western-ness, such as colder average temperatures, also predicted gender gaps about as well as gender equality did. In a sense, the Western countries represented a group of outliers that the “paradox” could not live without.

This makes sense given we know that WEIRD —– that is, Western, Educated, Industrialized, Rich, and Democratic —– cultures are frequently outliers when it comes to factors like personality, cognition, decision-making, and, critically, gender dynamics. For precisely this reason, in psychology, there has been a growing push against extrapolating findings from such WEIRD countries to the rest of the world.

Various theories could explain why the Western nations have larger gender gaps to begin with. One is simply a matter of data quality. Psychological data obtained from Western populations is typically more internally consistent than data obtained from non-Western populations, meaning that the items on a measurement scale correlate more reliably with one another — possibly a result of these metrics generally being designed and fine-tuned in the West. As a result of greater reliability in one cultural context, this data will generally be better at detecting differences between people in that context, both between men and women and within each sex. In the 2025 study, the authors also found that data quality indicators predicted national gender differences as well as or better than measures of gender equality. When data quality alone was included as a control, the relationship between gender gaps and equality was reduced by 80% on average.

Ultimately, if the paradox reflected a real, causal relationship between gender equality and gender differences in STEM participation, vocation choice, and personality, controlling for country baselines should not cause the relationship to consistently collapse. What remains is a much narrower observation — that a specific set of Western, Protestant, industrialized countries have unusually large gender gaps in STEM, personality, interests, and other metrics — which is not really a gender equality paradox. It is a fact about a dozen or so countries in need of its own explanation.

What actually happens as gender equality rises

Careful readers may still need more convincing. After all, absence of evidence is not evidence of absence. But there is a powerful reason to doubt that gender equality causes gender gaps in STEM professions and elsewhere. Countries around the world have grown substantially more gender-equal over time, and single-country trends over time are not as exposed to the same statistical artifacts discussed above. If one really caused the other, we would expect to see at least some evidence that gender gaps grow in a given country as equality improves, and we wouldn’t need to torture cross-sectional OECD data to see a relationship.

But we see the opposite of what the paradox predicts: Where gender equality has any effect at all, it tends to make men and women more alike, not less.

A 2020 study attempted to replicate the gender equality paradox for psychological values over 14 years of data, rather than simply using cross-country correlations for a single year. This is a stronger design: Cross-sectional correlations across countries conflate gender equality with every other way countries differ, while longitudinal analysis like this tracks changes within each country over time, letting each nation serve as its own control.

The researchers studied 32 European nations from 2002 to 2016, and found that rising gender equality did not produce the widening differences the paradox predicts — the relationship was either null or ran the opposite direction. Another study published in 2023 did the same, this time focusing on the career aspirations of boys and girls. The researchers looked at a sample of European countries from 2006 to 2018, and found that as national equality and women’s empowerment improved, the career aspirations of male and female students converged. This effect can also be seen around the world in a number of domains. In the United States, the percentage of women in the STEM workforce has grown dramatically over the last half-century of feminist advancements.

The available evidence suggests that gender equality does not cause larger gender gaps. When gender equality improves in a given country, these gaps typically shrink. Rather, a handful of Western countries are both very egalitarian and report larger baseline sex differences across STEM aspirations, occupational choice, and a number of psychological traits.

It is only when researchers rely on highly questionable cross-country correlations, which are often biased by underlying cultural and socio-economic differences, that we can convince ourselves of any other story. But this is not good science. Correlation is not causation, and post-hoc, just-so stories are not good theories.

---

Gabrielle Sorensen is an economic researcher and freelance writer, with previous bylines in publications such as The Hub and Policy Options. She holds her M.A. in Economics from the University of British Columbia.