Over at FiveThirtyEight, they have a number-crunching analysis of the number of papers (co)authored by women in the arxiv preprint server, including a breakdown of first-author and last-author papers by women, which are perhaps better indicators of prestige. The key time series graph is here:
This shows a steady increase (save for a brief drop in the first couple of years, which probably ought to be discounted as the arxiv was just getting started) from a bit over 5% women in the early 90’s to a bit over 15% now. The more detailed discussion in the article is worth reading, and mostly stands on its own.
One thing, though, that I wish they had included was a reference to this graph from the American Institutes of Physics showing basically the same trend:
That’s showing the fraction of physics Ph.D.’s earned by women over the years, and rises from a bit over 10% in the early 90’s to around 20% now. The data on women in faculty positions is less complete, but shows a similar trend.
The FiveThirtyEight piece, by Emma Pierson, covers a lot of issues, but I wish they’d dealt a bit more with this change over time. Because in some ways, that tells you a lot about the underlying dynamics– if the number of papers featuring women as authors simply tracks the number of women in physics in general, that’s one thing. If it rises more slowly than you would expect from the number of women in physics, that would be saying something else, and much less positive. Absent that, it’s hard to know what to really think about the trend Pierson reports.
Of course, it’s a difficult matter to tease this out, and there’s also an issue of subfield distribution– the arxiv started out as exclusively high-energy theory, and has expanded over time to cover a lot more of physics and math, but it’s by no means complete– when I spot an interesting paper in AMO physics, there’s only about a 50% chance that I’ll be able to find an arxiv copy. That’s going to affect the pool from which they’re drawing, which affects what you would expect to see in terms of authorship.
But this kind of basic analysis is a good starting point, and it’s always nice to have more data in the discussion.