Image credit: Casey terHorst

By Casey terHorst

Let’s say you go out to your study system and count the abundances of two species. If you found a 50:50 ratio of two species, would you assume that they were equal competitors? Hopefully you wouldn’t rush to that conclusion without more information, like predation levels, relative fecundities, and recruitment success. Maybe one species represents 99% of the incoming recruits, but then gets hammered in competition with the other species. If we want to conclude there’s competition going on, then we need a null model that takes all these factors into account and predicts abundances in the absence of competition. In ecology, we use null models all the time. The FSU Mafia scared everybody away from inferring competition without null models in the 1980s. The 2000s saw the rise of Neutral Theory as an important null model in community ecology. Population geneticists could never conclude anything about genetic differences among populations without a null model of the genetic differences within populations. Null models exist so that we can look at a pattern and make reasonable conclusions about the processes leading to those patterns.

Ok, so any good ecologist uses null models. But what’s our null model for determining whether there’s gender bias in the field of ecology? When I look at my own department, I often say that we have pretty good gender balance. Last year, three of our four hires were women. Great! I just went through and counted up our faculty and we have 24 men and 17 women. I often convince myself that’s close to 50:50. If three men retired and were replaced with women, we’d have 21 men and 20 women. Hey, that’s about 50:50. We’re doing a great job. Right?

Right?

Hold on a minute. Let’s think this through like ecologists.

For one, we’re still at a 24:17 ratio. That’s a low sample size. Although it seems skewed, it’s technically not different from 50:50 (for the record, a chi-square test says it’s not, P=0.27). But even if we did conclude that this is a 50:50 ratio, is that the right null hypothesis? Recently, I’ve seen a lot of cases where a 50:50 gender ratio, or something close to that is taken as evidence of a lack of gender bias in ecology. In a series of recent posts, Dynamic Ecology crunched numbers on recent faculty hires in North America. In one post examining how many new hires have Nature/Science/ PNAS papers:

Before anyone asks: there was no hint of gender imbalance in these data. Women comprised almost exactly the same percentage of new ecology faculty hires with Nature/Science/PNAS papers as they did of all new faculty hires. And the 20 new hires with first-authored Nature/Science/PNAS papers were split 11:9 men:women.

Ok, those numbers are close to 50:50, but they’re still skewed towards men. Those data are statistically indistinguishable from 50:50 (P=0.65), but that’s because of low sample size. If the data set were 1100 men and 900 women, this would be significant evidence of gender imbalance (P<0.001).

I don’t mean to pick on Dynamic Ecology here (ok, I do a little…they’ve had a lot of posts in the pasts with similar ratios/conclusions, and a huge readership and sphere of influence). It’s not just this one blog, but a much more widespread issue. In many journal articles and social media posts, and in my own head, a 50:50 gender ratio is taken as evidence that there’s no gender discrimination.

Setting issues of sample size and polling bias and any other quality-of-data issues aside, what is the null model for examining gender bias? When I look at my Intro Bio classroom, more than 60% of the students are women. In fact, across majors, 60% of college graduates are women. When I look at the seniors in my Ecology class, they’re often about 75% women. Women are highly overrepresented among graduate students in ecology and evolution. Yet men are still overrepresented in faculty positions.

There are many different ways we could construct a null model, and each has merit. Undergraduate gender skew? Graduate students? Post-docs? No matter which model you use though, a 50:50 gender ratio is strong evidence for bias against women.

Regularly, in discussions of gender discrimination in science, I hear ecologists say, “yeah, but we do a much better job in ecology than they do in chemistry or engineering”, to which I say “So what?”. That’s a pretty low bar to clear and it’s clear from these so-called gender balanced ratios that we still have a long way to go. We’re not doing a good enough job.

As ecologists, we know better than most that diverse and balanced groups perform better than others. So we should care about equal gender representation. But equality itself should not be the goal, so much as ending discrimination based on gender. And race. And sexuality. And class. And everything else that intersects to oppress underrepresented groups.

We use patterns to infer process. Null models help us do a better job of that. We use gender ratios to ask whether there has been a history of discrimination or other factors that have affected advancement of women. If we look at the current representation of women on the faculty in academia, even in the best case scenarios, it is clear that something prevents women from moving from the graduate level to the faculty level (spoiler alert: it’s patriarchy). We need to keep asking ourselves about how to address these issues, rather than congratulating ourselves about getting kinda sorta close to a 50:50 ratio.

How do you think we should construct a null model for gender ratios? Or do you disagree and think that 50:50 is the goal? Let us know on Twitter (@rapidecology).

Author Biography: Casey terHorst is an assistant professor at California State University, Northridge. He is a community ecologist and evolutionary biologist focused on what affects diversity, both in nature and in STEM fields. You can find him on Twitter (@ecoevolab).