Monday, March 07, 2011

The Odd Women

A scientist put together a research team to do some Awesome Science, and they wrote a grant proposal together. The team involved three PIs at different universities, their graduate students, postdocs, undergrad research students, and so on: the usual elements of a research proposal. The proposed research was excellent, and all 3 PIs are talented, respected scientists. Even so, some people doubted whether the proposal would be funded.

There are lots of reasons to doubt whether any particular proposal, however excellent, will be funded, but in this case, there was a reason unrelated to the significance of the proposed work, the research qualifications of the research team, the technical merits of the proposal, or even the lack of enough money to fund all excellent grant proposals. So what was the reason for doubt in this case?

The 3 PIs are women.

The doubt had nothing to do with the qualifications of any one of the PIs (all are highly qualified), but rested on the fact that all 3 PIs are female.

Of course, in the entire history of humankind, including today, no one would blink an eye if there were 3 male PIs. In the physical sciences, most scientists are men, and this has always been the case, so an all-male team of PIs is unremarkable, even today.

Oh sure, nowadays male PIs might describe in a proposal how they would involve female students or postdocs in the research, and they might write about how some of them have even advised females in the past, but the proposal would be funded or not funded depending on the scientific merit of the proposed work. I see proposals like this all the time.

But, even considering that this field of science is dominated by men, especially at research universities, is it so strange that a research project might be led by three women?

And, even if it is unusual, is it a problem?

In fact, it was not a problem. The pessimists were wrong, and the proposal was funded. And it was funded because the science was great and the PIs are all leaders in their field, with substantial track records of excellence and productivity.

Now, some would think the proposal was funded because the 3 PIs are women.

Are you following along? It's confusing, I know, but here is a handy summary:

If the proposal is not funded, it might be because the 3 PIs are women.
If the proposal is funded, it might be because the 3 PIs are women.

Fortunately(?), these statements refer to perceptions, not reality. Today, projects led entirely by women, even in STEM fields in which women are vastly underrepresented, are funded by the NSF if the proposed research is excellent.

We are not yet, however, at the stage where an all-female team of PIs is unremarkable. We are still at the stage where some people wonder if women get grants because grant agencies have to fund some females to meet diversity quotas or worry about projects that involve too many women. Maybe we shouldn't worry about this. Maybe we should just be happy (for now) to get grants. Maybe it is asking too much to have grants and respect?

This started me thinking about what an unremarkable proportion of women would be on a project. Is it equal to or less than the proportion of women in a particular field, or can we crank that number up a bit? In a 3-person PI team in a field in which women represent much less than 1/3 of scientists at research universities, we'd have to round a fractional woman down to zero if we used the proportional scheme.

I'm going to propose that, in fields such as this, ~25 +/- 10 % would be a non-threatening, unremarkable % of women involved in a research team. At this level, most people wouldn't worry that the science won't be any good or that talented men are being excluded for unfair reasons that have nothing to do with their research skills or experience.

But actually, my preference would be to assume that my not-entirely-serious, cynico-sarcastic analysis is flawed -- the proposal with the 3 women PIs was funded, after all -- and let some people go ahead and worry that maybe it was funded because NSF has to toss some money to the girls now and then.

Eventually (soon?), anyone who remarks in a negative way on the oddness of an all-woman PI team will be told that they, in fact, are the odd ones to think there is anything remarkable or problematic about an excellent and productive research team that just happens to involve only women.

Note: The title of this post is an oblique reference to the George Gissing novel,
The Odd Women, in which "odd" doesn't mean that the women are bizarre, but refers instead to the fact that, in late 19th century England, there were more women than men of marriageable age; i.e., an "odd" number of women. The title of this post therefore refers to a possibly problematic circumstance in which there are too many women.


James Annan said...

Well, there is some simple combinatorics here. If the proportion of women in the field is p and the formation of teams is entirely gender-blind then the proportions of 3-person teams with 3,2,1,0 men (and therefore 0-3 women) will be q^3, 3q^2p, 3qp^2, p^3 respectively where q=1-p is the proportion of men (and obviously this generalises to other sizes of teams). So the obvious questions are, what is the proportion of females (at a suitable career stage) in this particular field, and how common are the different gender splits?

If the empirical proportions are widely divergent from the above probabilities, then it would suggest that the process of collaboration is not entirely gender-blind. Not that this is necessarily a problem, or that its absence means everything is peachy, either.

Anonymous said...

Great post.

A related question, what proportion of women do you have to hire before you are perceived as giving an unfair advantage to women or, gasp, "only hiring women?"

In my experience, 25-30%.

Anonymous said...

James's analysis seems reasonable to me. For Anonymous at 8:49, 'what proportion of women do you have to hire before you are perceived as giving an unfair advantage to women or, gasp, "only hiring women?" '
If the gender balance of your hiring is very different from the gender balance of your applicant pool, there is reason to suspect that gender bias affects the hiring process. In some fields, hiring more that 25% women does require some affirmative action, and in other fields hiring more men than women would be indicative of gender bias, as the pools are predominantly female. Since you did not indicate what field you were in, we can't tell whether the pools you hire from are 25-30% women.

Anonymous said...

So there shouldn't be more women in a department or unit or project than corresponds to the proportion of women in that field or else men are being discriminated against? Talk about ceilings.. and that one wouldn't be made of glass.

Anonymous said...

While I agree with James that even if research teams are gender blind a three woman PI proposal is not be statistically anonymous. But, many research teams (at least in my field) are not gender blind -- we work with people we like and who we know well. Most of my friends are male. Most of my outside the university collaborations are with men that I was in grad school or a post doc with or met over beer at a conference (never in the bathroom though!).

At the university it is quite different. Here I work with the people I have come to know locally and who are most interested in working together. More than 50% of these people are women. Through them I have come to know (socially as well as scientifically) more women further afield and so do not expect to remain on all male multi-university proposals for much longer.

For me the social interactions are as important as the scientific ones in getting projects started. Given that I see that as normal as a man I see no reason why that would be different for women and am never surprised to see all woman authored papers or proposals.

Alex said...

Wait, were they saying that a proposal with 3 female co-PIs wouldn't get funded because there's something bad about having 3 female co-PIs? Or were they saying that it won't get funded because the panels are sexist?

The first, of course, is garbage. The second, sadly, might be true in many cases, though obviously it turned out to not be an issue in this case.

Anonymous said...

This makes me think a bit about job search gender balance - where 0-1 women out of 3-4 candidates is seen as 'normal' but 2 or more women is seen as unduly female-biased and I'm in the life sciences where we get about half the PhDs!

Materialist said...

What is the scaling law of diversity importance with team size?
Certainly a collaboration of 100 scientists that excluded an identifiable subgroup would fail the smell test, while just any two scientists collaborating would (probably) be not very remarkable.

Revathi said...

The statistical possibility of three women PIs is low in certain fields like physics or electrical engg; Having said that, Funding panels tend to be notoriously biased on many issues-mainly the reputation of the schools concerned plays a big role. There is some gender bias but generally if a woman makes to the point of being a PI, she is already over the worst.

Anonymous said...

In my opinion, as long as the proportion of women in collaboration X is not consistently statistically different from 50% then there is not an over or under representation.

Unfortunately it seems that in the Physical Sciences there still is a consistent statistical underrepresentation of women.

cesium said...

but does that mean that you would have a problem with an all-female collaboration on a project?

Anonymous said...

The point is that no matter what you do, (some) women will always feel like they were treated differently because of their sex (got/did not get money BECAUSE of being female). As long as you have this view there is no way you can move towards a less gender biased treatment since you define the treatment as gender biased no matter the outcome.

Anonymous said...

I am amazed a physical sciences professor would ask this question.

The first commenter had the right idea. If p is the fraction of women and q the fraction of men in the field;

Mean no. of women in a team
of n members = np

Standard Deviation = npq

They teach this in high school.