From the study:
"For outsiders, we see evidence for gender bias: women’s acceptance rates are 71.8% when they use gender neutral profiles, but drop to 62.5% when their gender is identifiable. There is a similar drop for men, but the effect is not as strong."
Yours is a good point, however, that seems to have been missed in the reporting on this study. The men's acceptance rates did also drop (significantly even) when their gender was revealed. So it appears that men might also benefit from having a gender-neutral identity.
I don't find this especially convincing. Some of the issues here are reasonable things to look at; some aren't.
>1) I am surprised that there are only 20x more PRs by men. I would say it’s more like 100-200x more men on GitHub then women.
So they're comparing data (which may be erroneous, but is actual data), with a feeling: "I would say". And the data types are even different! The study compared the ratio of pull requests between genders; this person quibbles over the ratio of users between genders.
>2) But the women that are there are usually always employed by some small to medium tech company. So this study is possibly comparing mostly professional women with mostly hobbyist men.
Perhaps. It would be interesting to see data here. Again, saying "are usually" isn't based on data.
>4) In my experience a lot of people on GH with feminine sounding names are actually men from Eastern Europe (Sasha, Nicola, etc) or India (Sandy, Sunny, etc).
This is a complaint against a methodology that the study didn't use. From the study (https://peerj.com/preprints/1733v1.pdf):
>...we took a different approach – linking GitHub accounts with social media profiles where the user has self-reported gender.
You can say this wasn't a big enough ratio (they identified 35% of users' genders this way), or that people will report incorrect data, but if you're talking about "feminine sounding names", you're not addressing this article.
Point five has nothing to do with gender. Perhaps there's an unstated point there.
Point seven is easily testable; points eight and nine are testable, but not as easily.