> Therefore, we conclude that both variance and [variance ratio] in mathematics performance vary greatly among countries. Confirming our earlier finding ([21]), we also conclude that [variance ratio] is reproducibly essentially unity for some [predominantly muslim or low gender equality] countries. These findings are inconsistent with the greater male variability hypothesis.
If we're trying to draw conclusions about innate differences operationalized as distribution test scores, any kind of biased schooling undermines the ability to draw conclusions from that population.
If culture affects the outcome, it waters down the premise of the variability difference theory.
And what is unusual about Czech Republic, where VR and gender gap were near unity and zero?
Couple of points here. A culture biased along gender lines may bias the outcome thus obscure the effect of genetics on gender differences. Also, no one in their right mind claims that environment has no effect, and so measuring some environmental influence doesn't undermine the greater male variability hypothesis.
>And what is unusual about Czech Republic
I wouldn't hazard a guess, but the "paradoxical effect" of gender equality negatively correlating with female interest in STEM might have some hard-to-tease-out influence here.
Something about the hypothesis itself:
The variability hypothesis seems plausible, but there seems to be little evidence linking it to biological differences.
Variability differences undoubtedly exist, but not universally and not with same degree across cultures. As the measured variance is very culture dependent and even disappears, existing data does not support the hypothesis linking it to biological differences.
There is also strong negative correlation between variance ratio (VR) and mathematical performance. (r = -0.640, p < 0.001). VR = (variance in male scores) / (variance in female scores). The variance is higher where math scores are lower.
source:
Debunking Myths about Gender and Mathematics Performance http://www.ams.org/notices/201201/rtx120100010p.pdf