That's an interesting paper. I am not a statistician or professional scientist, however, I was surprised at the amount of story telling involved in their analyses, and how often they would exclude certain data based on those stories. Is doing this common in studies like this? It seems one could make the same data say anything by using this approach.

I am particularly perplexed by their discussion of what they term "the variability hypothesis", which states that because:

1- a higher statistical variance in the distribution of some attribute in one population vs another population leads to a larger difference in relative representation of the two populations the further from the mean you go (this is true a priori), and

2- men have a higher variance in "math ability" compared to women,

then we will observe a significant difference in representation between the sexes at very high levels of ability. The authors mention that there is in fact a negative correlation between gender inequality and variance in ability (i.e. the more equal a society is, by their metrics, the higher the difference in variance in performance is between the sexes), which seems to elegantly support the variability hypothesis, to my understanding. I wonder if there has been any follow-up to this paper?