Back in December, Alex Bond on wrote a post* about calculating one’s gender gap; his post was inspired by this recent article in Nature, which I found really interesting and encourage you to read. In his post, Alex looks at the gender gap in his own academic record – that is, he looks at the female:male ratio of his supervisors, committee, field crew, and coauthors. This seemed like an interesting exercise, so I figured I’d try it (with slightly different categories). My results:
Supervisors: I have an infinite ratio, as my undergraduate advisor, PhD advisor and coadvisor, and postdoc advisor were all men. Not off to a great start in terms of gender balance, am I?
Committee members: I do better here (though, admittedly, it wouldn’t be possible to do worse!) I had six committee members (yeah, 6 was a little crazy). 2 women, 4 men (0.5 female:male ratio).
Postdocs I’ve had in my lab: 2 women, 1 man (for female:male ratio of 2)
Grad students I’ve had in my lab: 2 women, 1 man (once again giving a ratio of 2)
Undergrads in my lab: 27 women, 9 men (giving a ratio of 3**)
Okay, now on to coauthors, which is what the Nature article focused on. There are two ways to count this: I can tally the people with whom I’ve written manuscripts (just counting each person once, even if I’ve written a bunch of papers with that person). Or, alternatively, I can tally them by coauthorship (that is, weighting the counts of people I’ve collaborated with on multiple papers by the number of times we’ve published together).
Counting each person just once: 30 women, 24 men, for a ratio of 1.25
Counting each person by number of coauthorships: 68 women, 74 men, for a ratio of 0.919***. (The shift when weighting by coauthorships isn’t too surprising, given that the person I’ve published with the most is Spencer Hall.)
For comparison, the female:male ratio for the US, according to the Nature article, was 0.428. Clearly I’m doing better than average, due in part to having collaborated extensively with one woman in particular (Carla Cáceres), and in part because my lab members tend to skew female (see above).
What do I take away from this? I’m not sure, to be honest. It’s interesting to know, I guess, but I don’t think I’ll change anything I do in response. But I guess that’s okay, since my gender ratio (for coauthors) is pretty good.
It does make me wonder what someone would do if they weren’t happy with their gender gap. Early on, mentors are key coauthors. I personally didn’t consider gender when choosing my advisors and, if I were to do it all over again, I would choose to work with the same advisors without any hesitation at all. I do know some women who intentionally chose to work with women, but that never really occurred to me. And, besides, they chose to work with women for reasons other than having a roughly even gender ratio of coauthors. (This interesting post from scitrigrrl at Tenure, She Wrote deals with the topic of whether male and female mentors fill different roles.) And, for collaborators, I haven’t really considered gender either. One of my closest collaborators is male, and the other is female, but that’s an accident. But, I guess if I had a really big gender gap, that might be a sign that I need to examine whether I have biases in terms of how I am choosing potential collaborators.
Do you think there’s value in calculating an individual’s gender gap? If you were unhappy with yours, are there things you think you will change in response?
*Who says there isn’t an ecology blogosphere? (Okay, fine, this wasn’t exactly a prompt reply, suggesting the infrequency of posting might be part of the issue!)
** I have wondered at what point this ratio would be too high. I think it’s normal for me to tend to have more female undergraduates in my lab, both because biology undergrads skew female, and because female students often prefer to work with female faculty.
I’m not entirely sure about the value of calculating this other than to make me go “hmmmmmmmm”. To my surprise, my gender ratio was 0.49, I had thought it would be higher as I’ve co-authored a lot of papers with women. But when I look in detail, they tend to be the same women repeatedly! So a weighted ratio would be closer to 1.
I think there’s value in being self-aware and reflective about your own behavior. So I guess if calculating your own gender gap encourages a bit of self-reflection (particularly about possible implicit biases), there’s some value in that. But the sample sizes for any given individual usually are very small, so that the confidence interval around your own gender gap usually is going to be really wide. Plus, individuals normally have very good reasons for choosing to work with other individuals. Like Meg, I wouldn’t go back and choose any differently than I have when it comes to my choices of supervisors, committee members, collaborators, postdocs, grad students, or technicians. And like Meg, I too am unclear what I could’ve or should’ve done differently. For all these reasons, personally I found that calculating my own gender gap was kind of ineffective as a tool for prompting self-reflection. I didn’t even feel like I needed to go through the exercise of the formal calculation, since a glance back at my records was enough to satisfy me that the formal calculation wouldn’t reveal anything so glaring as to give me even a moment’s pause. Again just speaking personally, I’m more likely to be prompted to personal reflection by data from entire fields rather than personal data, because the sample sizes are bigger.
In his original post, Alex notes that the gender gap in scientific publishing as a whole has to come from *somewhere*. That is, the gender gap for science as a whole is just an aggregate of individual gender gaps. That’s true, of course. But the question is what follows from that fact. For instance, as Alex himself notes, some of those numbers are more or less out of one’s control. Indeed, to the extent that the field as a whole has a gender gap, it might be correspondingly difficult for any individual to eliminate a personal gender gap. Further, the gender gap in science as a whole presumably has other causes besides implicit or explicit sex discrimination by individuals in whom they choose to work with (see Ceci and Williams: http://www.pnas.org/content/108/8/3157.full). But perhaps I’m making a false assumption here, that the sole or main point of calculating one’s own gender gap is to prompt one to reflect on and change one’s own behavior. Perhaps the main point is to prompt one to reflect on any and all causes of gender gaps, including those that one can’t much affect by one’s own choices of who to work with?
There is of course the problem of small sample size, and let’s not talk about independence of the data points. Still, it seems to me this kind of analysis could be a good tool to reflect about our choices as scientists.
What I don’t get entirely is this though: if you look at the whole scientific population, I find it logical to define gender gap as a significant difference from an equal distribution. If we see such a gender-gap, we should reflect about the reasons for it.
However, if we have already established that we have such a gender-gap in the population, it doesn’t demand additional explanation why individual scientists display the same gap – if there is a female:male ratio for the US as a whole of 0.428, and both sexes have equal publication behavior, then gender gaps for individuals MUST scatter around that ratio of 0.428, unless females coauthor more paper than males.
Thus, the more relevant indicator seems to me to test not against 1, but against the population-average, which tells us whether there is a non-random pattern in our choice of “partners”. –> this is what I would call “bias”, on top of the population gap.
An exception may be PhDs and undergraduates, where you could enforce a strict gender-balance of 1 if you wanted and then measure whether you achieved it, but this would assume that you applied a deliberate bias towards a ratio of 1, and then we wouldn’t require a statistical test because choices were not random anyway.
Re: possibly enforcing a strict 1:1 gender balance among one’s UG and graduate students: I couldn’t do that even if I wanted to, not even over some extended period of time. For instance, this summer every single UG student who’s expressed an interest in working in my lab has been female. That’s only slightly unusual; most years UG interest in summer assistantships in my lab skews heavily female. Conversely, the vast majority of prospective graduate students ever to express serious interest in working with me happen to have been male (the sample size there is very small; I don’t exactly have prospective grad students beating a path to my door…).
yes, I also think it’s impractical, it was just to illustrate the point that I would only expect a 1:1 ration if a) the base population has a 1:1 ratio b) I discriminate between male and female in my choice of whom I’m working with, with the aim of having a 1:1 ratio.
Thanks for writing the post, Meg. A few thoughts in relation to the comments so far.
Yes, small sample size is an issue, As with many things in ecology, the interpretation is all about scale. The Nature piece was one extreme (science), and I think the level of individual researcher (or even perhaps each paper/project) would be the smallest unit. At what scale does it matter / should we be concerned? Paper? Individual? Department? Discipline? Journal? All of Science™? A question for which I don’t have the answer.
Similarly, when calculating, is it done at the scale of the collaborator, or the paper? Or the career? Or, as some have done, careers over time? Lots of potentially confounding variables.
And ideally, this would be tested against some population, whether that’s the general population, or the population of ecology undergraduates, or … again, the question of scale (score another for John Wiens).
Well, “scale”, or sample size, is one issue. But isn’t “what are the causes and what should individuals do about them?” other, separate issues? I guess that’s what I’m struggling with a little. I can see how calculating one’s own gender gap could prompt reflection. But it doesn’t seem to give much guidance as to what to reflect *on*, or how to act (which presumably is the ultimate goal of reflection here, right?)
But perhaps my reaction here just says something about me? Different people presumably are prodded into reflection, and action, by different things…
Do you have any further thoughts on what the calculation of personal gender gaps should *do*? At the level of the individual calculating it, I mean. If, say, someone calculates their own gender gap and just kind of mentally shrugs, is that a good thing? A bad thing? Or am I just being dense and asking the wrong question here?
I think addressing causes & possible solutions are also scale-dependent, though. Many at the department/university level, HR departments scrutinize short lists and I know of some that have asked for an explanation of why no women were represented. You mentioned receiving a large proportion of women applicants to your lab, and you consequently take on a larger proportion of women students. In my mind, this is an application of the gender gap at the individual level (assuming, of course, that your data on student applicants isn’t biased by perception) – whether consciously or not, you made the calculation & concluded that your recruitment practices were unbiased.
So where on the scale of individual to science as a whole does this break down? Or is it the slow aggregation of very minor differences at the individual level that, when summed together, give the data presented in the Nature piece?
If your research group had shown something deviating considerably from 1:1, i neither direction (again, small sample sizes aside), could you attempt to recruit more students of the less represented gender? And again, over what time scale would one assess this – the current lab composition? The last 5, 10, 20 years? I don’t know, but perhaps someone with a bit more knowledge (and experience) on this will chime in.
I agree that the time scale question is interesting. There have been some times when my lab skewed heavily towards women. I think there may have even been a semester when it was 100% women, but I don’t remember. But now it has a much more even ratio. With high turnover and small sample sizes, there’s likely to be a bit of fluctuation.
An important thing to think about is seniority. In science in general (and I suspect ecology in particular), older scientists skew more heavily male than younger scientists. So, you’re more likely to have more men in the mix of supervisors than in the mix of supervisees. And the older you are, the more man-heavy your supervisor collection is likely to be.
But it’s tricky. Meg’s example is a good case. You could say that it’s not surprising she has a low female:male ratio for supervisors and committee and higher ratios for the folks she supervises due to generational effects. However, you could also infer that the pattern is due to the “leaky pipeline” — many more women interested and involved in biology early in their careers, but are systematically and disproportionately knocked out of the system as they progress.
I agree with above comments that say that if you really want to be able to do interpretation, you’ve got to be able to compare against some broad statistics, which I suspect are missing (or at least not easy to find) for ecology.
That said, I purposefully chose to put a woman on each of my committees. It would have been easy to fill them with all men. And the women who were included were fully qualified and made sense to have, in terms of my research. Given that I could go either way, I used gender to break ties. I felt it important to have some diversity of career perspective (for lack of a better term) when getting advice/feedback from my committees.
Yes, I agree that the strong difference in gender ratio between my supervisors and committee (that is, people senior to me) and my lab members (that is, younger folks) is interesting. I think it reflects both the loss of women in academia with increasing rank, and a demographic shift. Related to that, I’m planning on submitting this as one of my Friday links:
Great link and really nice to see so many points tied together in one post!
The thing I’ve never seen anyone comment on is the generational thing brought up at the very beginning. If twenty years ago the typical ecologist was a 38-year-old male ecologist and today it’s a 55-year-old male ecologist, what does that say about the future? From these stats it looks like positions in ecology have been fairly stagnant as ecologists as a whole simply got older. An optimist would say that the gender (and other diversity) issues may fall away rapidly as the baby boom generation retires. A pessimist might say that those 55-year-olds are going to cling with all their might to their positions for another 20+ years. Is the average ecologist in 20 years going to be 72 years old? ‘Cause if so, there’s not much hope for gender (and other) parity in the next couple decades.
Hate to confirm your depressing line of thought, but profs don’t retire: https://dynamicecology.wordpress.com/2013/08/16/friday-links-profs-dont-retire-and-more/
I’m surprised that no one’s pointed out that the gender ratio of the regular contributors to Dynamic Ecology is 0.5
Without wanting to give away any big announcements, that number may soon become another illustration of the temporal variability of gender ratios… 🙂
But if you weight by “coolness of avatar”, our ratio is skewed heavily female. 🙂
And if you weight by number of posts…..or average length of posts…..? 🙂
Yeah, I definitely don’t pull my weight on either of those!
But you DO have a cool avatar!
’tis true! I have to contribute *something* here, don’t I? 🙂
We value your contributions way more than that 🙂
Only a fool would weight by those things. 🙂 Plus, if you’re going to weight by those things, your analysis of gender ratios will be confounded by the fact that one of the guys who blogs here is me. So really, you might want to weight by “Are you Jeremy Fox or not?” 🙂
And for the record, I agree that our Jeremy Fox:not Jeremy Fox ratio is too high, and the blog would be better if it came down. 🙂 And in semi-seriousness, I suspect some readers agree. I suspect that there are readers (maybe many readers?) who wait eagerly for the next post from Brian or Meg, but who skip some of my stuff because too much of anything kind of gets old.
Pingback: What we’re reading: Convergent evolution of thrifty yeast, the surprising history of amphibian-killing fungus in Brazil, and novels as biology homework | The Molecular Ecologist