From Heather Sarsons and Guo Xu, in the new AEA Papers and Proceedings:

Women are 7.3 percentage points less likely to provide extreme judgments (column 1). In terms of magnitude, this gap is economically large. Compared to the mean of the dependent vari-able (19.8 percent), this corresponds to a gap of 36 percent. The gap is somewhat smaller for the self-reported confidence level but still nontrivial (column 3). On average, women tend to report a confidence score that is 0.221 points lower than men. This corresponds to a gap of 4 percent when evaluated against the mean...

The results show that both men and women are less confident when asked questions outside of their field but that a confidence gap persists. For example, while men are 5 percentage points less likely to give an extreme answer when speaking on a topic out-side of their primary field, women are 9.2 per-centage points less likely to do so (column 1). For the measure of confidence, men are on aver-age 0.585 points less confident when speaking on a topic outside their primary field (column 2). Once again, that confidence gap is significantly larger for women. We would expect women to be less confident than men when answering questions outside of their field if women actually have a narrower range of expertise than men do. The results, however, do not change when we control for the respondents' breadth of expertise using RePEc data and allow breadth to vary by gender. More importantly, accounting for the differential confidence when moving beyond one's own field "explains away" the level effect of gender.

...it appears that the confidence gap emerges when women are speaking on topics on which they might be less informed.

I would be curious to see if similar results hold for bloggers and Op-Ed writers, or whether it is "the Margaret Thatcher syndrome" at work.