Your first two sources are interesting, though somewhat tangential to the question I was asking.
Of the key findings of the first, most provide little insight on the question I'm asking: why women are less willing than men to work gratis?
Claims 1.2.3 and 1.2.4 simply suggest women are less skilled than men, which should also preclude them from working for pay. 1.2.5 is no explanation at all - many corporate environments with plenty of women are considerably more hostile than open source communities. 1.2.6 is an incorrect stereotype - see my link to the ATUS.
1.2.1 is interesting - it suggests women strongly prefer a collectivist environment to an individualist one. As a person who's done little more than fix typos, it's worth asking: is wikipedia also a more individualist environment than Demand Media [1]? Is OSS more individualist than the typical corporate environment?
1.2.2 seems to contradict other parts of the report - women both get a "high level of attention" and are unable to get attention (further in the report.) Not sure what to make of this one. [2]
The second report strongly suggests women can't cut it in software - they lack the dedication (claim 1), are unwilling to work with mentors who are different from them (claim 2), they can't handle a text interface (claim 4), they can't handle competition (claim 6), and other women won't help them (claim 7). But the problem is, all of these claims apply equally well to for-pay software [3]. (Claim 5 is actually not a cause, but a result.) Only claim 3 suggests an actual difference between gratis software production and for-pay production - the use of gendered pronouns. (I'll hypothesize that HR does search and replace to eliminate these in big companies.)
So near as I can tell, the only relevant claims made by these reports are that either gendered pronouns or an individualist environment lead women to be unwilling to contribute. It wouldn't surprise me at all if there is an individualism/collectivism gap between OSS and for-pay software. That's an interesting hypothesis.
Another testable implication of that theory: women would be less likely than men to work as individual (i.e., non-accenture) contractors or work for startups, preferring instead the collectivist environments of big companies. Is this true?
[1] I recall reading that most Demand Media workers are women. Would love it if someone with more knowledge could confirm/deny this.
[2] Also finding the methodology of this report weak. To determine attention received, why take a poll rather than just running an experiment? I.e., post some newbie questions on mailing lists with male and female names, and measure the results.
[3] Maybe there is a gap caused by corporate use of Visual Studio as opposed to OSS use of emacs/vi?