The Gender Equality Paradox in STEM: The Original Study, The Correction, The Rebuttal and The Blog

  • <<
  • >>

560873.jpg

 

The controversial 2018 “Gender Equality Paradox” study that made it into mainstream news has now received a 1,113-word correction after Harvard University researchers could not replicate the findings.

In the initial paper published in APS, Gijsbert Stoet (University of Essex, UK) and David Geary (University of Missouri) used a publicly available international database to examine male and female representation in STEM fields. Their results indicated that countries with more gender equality had fewer women earning STEM degrees, while countries with less gender equality actually had more women studying science and technology—something the authors dubbed the “Gender Equality Paradox.”

They further expanded on the paradox as such: countries with more gender equality, like the U.S., Sweden and Norway tend to be richer, so women have the financial freedom to pursue their true interests, which lie in the humanities. Conversely, women in countries like Turkey and Algeria with less gender equality, a lower GDP and less economic opportunities enter STEM fields for the high pay, regardless of true interest.

Sarah Richardson and colleagues at the GenderSci Lab at Harvard University have spent over a year conducting a critical analysis of the Gender Equality Paradox hypothesis, including trying and failing to replicate the data and results, publishing a commentary on the correction of the original study and, most recently, authoring a series of posts on the GenderSci Lab blog regarding the paradox.

Based on their collective expertise and research, Richardson and her team are firm in their conclusion: The Gender Equality Paradox is only a paradox if you start with particular assumptions.

One key problem (and four other issues)

In their blog, Richardson and graduate student Joe Bruch identify five specific problems within Stoet and Geary’s study, including spurious correlation; lack of longitudinal design; inadequate robustness of the finding to changes in data, measures and assumptions; and ecological fallacy. However, it is the number one problem that has gained the most mainstream attention—issues with replicating the findings.

Stoet and Geary used country-level tertiary degree statistics to compare women’s achievement in STEM around the world using UNESCO statistics. The database is common and used often for this type of analysis, but when the GenderSci Lab tried to match the numbers to the UNESCO data, they could not.

Not mentioned in the original article but revealed in the correction, Stoet and Geary actually produced their own novel metric, termed a “propensity measure.” This metric adds together the percentage distribution of all female graduates who received a degree in STEM to the percentage of all male graduates who received a degree in STEM, making this number the denominator, and then takes the female percentage from this.

Take Algeria, for example. According to UNESCO, in 2015, Algeria had 287,914 tertiary graduates, 63% of whom were female. Of 180,554 female graduates, 27% obtained a degree in STEM. Of the male graduates, 39% earned STEM degrees. Thus, the percentage of women among STEM graduates was 54%. The propensity measure, however, calculates a female STEM degree rate of 41%.

Even using the propensity measure the second time around, the GenderSci Lab was still not able to replicate the data, alleging the numbers used in the original paper were lower than the percentage of women among STEM graduates across all nations by an average of nearly 9%. In fact, 19 countries showed significant variation, with 15 of them moving ten spots or more in the gender equality rankings, according to the GenderSci Lab team’s own analysis using the propensity measure metric.

“These changes in country rankings are important because they invite alternative hypotheses regarding the correlation between [gender equality score] and women’s achievement in STEM to that proposed by Stoet and Geary,” Richardson and Bruch write. “What we observe is that there is huge variation in the gender gap in women’s and men’s tertiary degrees, no matter how it is measured, across countries.  These variations do not conform to simple patterns, suggesting that the question of women’s and men’s inequalities in STEM fields represents a distinct and rich domain of gender inequality that is not easily represented along a single dimension and with a single measure.”

Seeking to address Richardson’s commentary on their correction, Stoet and Geary authored a reply that defends the propensity metric as well as their original paper.

“It is important to control for differences in the overall number of women and men who attend college, which varies from nation to nation. Accordingly, in our original article, we chose a calculation method that adjusts for this potential confound,” wrote Stoet and Geary. “That said, our specific approach has no bearing on the conclusions.”