Someone hasn't done their literature survey very well. It's the leaky pipeline that's most obvious, but we have quite a few convincing (not "anecdotal" nor unduly "small sample") papers using unimpeachable methodology* that demonstrate bias against women from very early in academia, e.g. enquiring about doctoral study, or applying for the kind of position students customarily use to support themselves. Here are two from this year.
http://www.pnas.org/content/109/41/16474.fullhttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=2063742* You base your study on paperwork, and make copies of it that differ only in a name, which can then be male or female (you have to take care that your chosen names don't give other differing impressions that might cause bias, of e.g. ethnicity or class, but are comparable apart from gender; but that's not too hard to do in prestudy). If your (appropriate and sufficiently large) sample evaluates the paperwork with male names higher than the identical paperwork with female names, game over: there's no explanation other than bias.
