Female convenors invite more female speakers (or do they?)

Women are underrepresented in most academic fields, including my own (economics), and especially at higher levels. People are always trying to find simple interventions to improve the situation. According to a new study in the microbiology journal mbio (reported somewhat breathlessly in the Atlantic) one such intervention is to ensure that women are amongst the organisers. This is argued to increase the probability of inviting female speakers (a worthy intermediate goal) which, they argue, will advance women in academia generally (the end goal):

“Put at least one woman on the team that organizes a scientific symposium, and that team will be much more likely to invite female speakers,” said study co-author Arturo Casadevall, M.D., Ph.D., professor and chair of microbiology & immunology at Einstein, director of the College of Medicine’s Center for Immunological Sciences, and attending physician at Montefiore Medical Center, the University Hospital for Einstein.

[from the press release]

As far as I can see, however, they haven’t shown this. The study shows a simple bivariate correlation between the presence of a woman on an organising panel and the proportion of invited speakers who are women. In the paper, the authors are quite careful to note that their finding is a correlation only, but in a classic example of the “science news cycle” at work, the press release slips quietly into the language of causation.

The issue with this is that there is potentially a very important unexamined confounding factor. Scientific displines are not internally homogenous: some subfields attract a disproportionate number of men, while some subfields attract a disproportionate number of women (relative to the field as a whole). If you compare a seminar on development economics with one on financial economics you’ll see what I mean. There are bound to be similar divides in microbiology (though they may not be as pronounced as in economics): perhaps immunology leans towards men, while epidemiology leans towards women, or the reverse, or something else entirely. Some differences will exist.

But convenors and speakers for a session likely come from the same subpopulation (people interested in that subfield). Then a subfield that has a higher proportion of women will be straightforwardly more likely to have both a woman on the organising panel AND female invited speakers - even if all these positions are filled completely at random. It need not be, as the authors tentatively suggest, that “women as conveners more often consider gender and make conscious efforts to find female speakers”. Indeed there need be no causative mechanism from ‘women as convenors’ to ‘women as speakers’ at all, since both could be effects of a common third cause.

The reality is probably not so stark. Commonsense and anecodotal experience suggests the effect the authors claim does exist. But the quantifiable causative effect of a female organiser is unlikely to be as large as the simple correlation would imply, and is possibly much smaller. Of course that doesn’t mean it’s not a good idea anyway - but this paper doesn’t make the case.

(c) Jorge Cham, www.phdcomics.com