We’re missing our best chance for gender parity on software teams.
I was lucky to be able to visit London twice last year, to do conference talks. It’s a lovely city with brilliant people, and beyond that, their subway system (called the tube) is amazing. No matter where I was in the city, and no matter where I wanted to go, I could always get there on 2 trains and it would take about 40 minutes.
My trips were both busy, so I spent a lot of time in tube stations. I grew very fond of a particular automated announcement: “Mind the gap, please.” This refers to the space between the platform and a stopped train, which at some stations, is large enough that getting on or off feels like jumping a ditch. If you’re not paying attention, you could fall right into it.
Custom Software development has a gap problem too. But we’re not paying attention, and I’m afraid we’re going to fall right in.
An Un-Automated Announcement
I do software development in the startup and open source world. One thing I love about it is that absolutely no formal qualifications are required. It’s open to everyone! In fact, we almost prefer folks without formal cred. Our heroes are college dropouts and self-taught developers. You don’t need a 4-year computer science degree or a résumé in a word doc, we say! Github is your résumé! Contribute useful things to open source, and people will offer you jobs!
It sounds like an ideal place to target if you’re switching careers. Large companies and some specific sub-fields are still likely to require a CS degree, but it’s possible to be successful in this field without one. It’s the tech version of the American Dream. If you contribute to the community, do some open source and get it up on Github, and of course work really really hard, you too could become internet famous!
And perhaps even a race car driver.
Living the Dream
It sounds fantastic, but something about it is off. It has produced a community of developers that is overwhelmingly male. The most widely cited study concludes that contributors to open source projects are 2% female. Startups have depressingly similar demographics. It’s not unusual to find a ten-person dev team with no women at all.
In the industry overall, women make up 26% of developers. That means for every woman, there are about 3 men. That is already pretty unbalanced; studies have shown that we do better creative work in diverse groups, so any time we have an axis that out of balance, we’re missing out.
But my world’s not even at 74/26. We’re at 98/2. For every woman, there are 49 men. That’s so extreme, and we’ve gotten so used to it, that it’s hard to even imagine what real innovation looks like. Imagine if we didn’t make another car sharing app, or another way to sext, and instead worked on actual hard problems! Poverty, disease, hunger, homelessness. Let’s do some fucking good in the world. Diverse teams (and not just along the gender axis) make that possible.
Lots of people are worrying about why, overall in the industry, we’re only at 26%. Personally, I’m more interested in why the gender balance in the communities I care about are so far off from the mainstream programming numbers. Our 2% makes 26% look amazing. How can we make up that 24%?
They don’t have stars on thars
Let’s start with a slightly easier question: what’s different in our culture compared to big company cultures? I guess the headline betrays my bias — I don’t think the startup/OSS world and the mainstream/”big company” world are as far apart as we’d like to think they are.
For example, many folks have said that being successful with OSS or at a startup requires uncompensated work outside of standard hours. This is true. It’s also true that women in general have less time to give, since it’s fucking 2015 but somehow women still do most of the housework and childcare. But those “overtime requirements” are not structurally different from Microsoft’s death marches where developers, QA, and everyone else worked 16-hour days including weekends. I’m not saying either scenario is good, but both sides can require extra time outside of work, so that does not explain the difference between 2% and 26%.
Another factor people often cite to explain this spread is that startups and OSS projects have a higher tolerance for gender-based harassment than larger companies, so women stay away. Large companies, after all, have proper Human Resources departments. Startups generally don’t — more often they’ve got one person who’s the office manager, black-box tester, and also does all the HR tasks in the spaces between. Open source projects typically don’t have anyone resembling HR at all. And many nasty incidents of sexual harassment at startups and in open source projects have been quite public.
But this is fundamentally a visibility problem. Anyone who’s ever reported gender-based harassment to a corporate HR department knows that HR is not there to help you. HR is there to cover the company’s ass. If you don’t know that already, I sincerely hope you never find out. HR is just as likely to “solve” a harassment problem by firing the victim as by firing the perpetrator. The only difference from a startup scandal is that it doesn’t show up on Twitter. (But it does occasionally show up in the LA Times.)
It’s shitty on both sides, and women know how it works where they are. This doesn’t explain a 24-point difference either.
A few other possibilities that come up occasionally:
- Difference in level of abstraction? Nope, there’s low-level and high-level work on both sides.
- Job security? Nope, a big company will lay you off as casually as a small one.
- Static types? ::troll::
There’s one obvious difference left. In fact, we discussed it above! The startup and OSS worlds prefer informal qualifications, while large companies prefer formal ones.
o_0
Could the very thing I love so much about my community be what’s killing it?
Let’s Consult Science!
As the child of scientists, I prefer research-based explorations of this topic to anecdotes. Fortunately, over the last fifty years, there has been some fascinating research on how men and women, in general, are treated differently. What this research says is the following:
Given a man and a woman with equivalent skills and experience, the base level of competence that people assume is lower for the woman than for the man.
Men and women both do this, and it is largely unintentional. It applies whenever we’re evaluating people or their work, such as when we’re meeting them at an event, interviewing them for a job, reviewing their code, or considering their talk proposal.
This surprised me, because I certainly don’t feel like I always judge women more harshly. I wondered if it was one of those pseudo-science studies with a tiny sample size. But the research that supports it has been replicated time and again, and goes back almost 50 years. Using statistical techniques every grad student has to learn their first semester, they isolate all the variables and still end up with statistically significant differences ascribable only to gender. I recommend this survey paper as a starting point for further reading. It refers to many studies that replicate this problem across race, age, and physical ability, in addition to gender.
This difference in perceived competence is called the credibility gap. It affects all women in tech, so by itself, it doesn’t explain the 24% difference between large companies and startups/OSS. But let’s examine some of the ways this gap manifests itself. We’ll start to see the differences in the ways this affects different communities.
What do you think when you see a woman at a meetup?
I first noticed myself make this kind of mistake a few years ago at an SF Ruby developer event. I was line for the bar and I started talking to the woman in front of me. Before she even opened her mouth, though, I knew she was a recruiter, or maybe a product manager, or maaaaybe QA.
Welp, I bet you know how that went. She was a developer with way more experience than me. If the person in front of me in line had been a man, I wouldn’t have gotten into trouble. My default assumption would have been that he was a developer.
Why does it matter? Once you find out, you can adjust, right? And then we’re all good? But the science tells us that this is actually a bug in our wetware. We don’t notice it, but the credibility gap is still there, even after we find out our assumptions were wrong. All the technical women I meet are starting with a deficit of perceived competence. They have to prove themselves more than similarly skilled men do, just to overcome that deficit and get themselves up to the same level of perceived competence that the men enjoyed by default. Men get the benefit of a doubt; women, in general, do not.
This research has been replicated with race. Folks assume a white person is more competent than a person of color. This is why women of color have an especially hard time — they have deficits from two credibility gaps to dig out of before they’re even regarded at the same level as an equivalently-skilled white man.
The problem extends, of course, into hiring. The difference is that when we’re hiring, in addition to our informal impressions of competence, we also consider formal qualifications, such as whether or not they have a CS degree and who they’ve worked for in the past. Stronger formal qualifications can partially compensate for a credibility gap.
And here’s where the OSS/startup world has a peculiar vulnerability. We don’t place as much value on formal qualifications, which means we rely mostly on our subjective impressions of competence. “But we also look at their code!” you might say, and that is true, but code is not value-neutral. This research has been replicated when studying how folks evaluate people’s work products. That means code review has this bias if you know the gender of the developer. And code review (remember, Github is your résumé!) is baked into our hiring processes in a way it’s not at larger companies.
So while folks in the open source world are more likely to be self-taught, and more likely to value that in others — even women — the reality is, our heavier reliance on informal qualifications means we have a harder time hiring women than the rest of the industry. Or to put it another way: women and minority developers face barriers when they’re self-taught that most of us never realized existed.
We all discriminate this way. Women are biased against other women as much as men are. We’re all swimming in the same pool — the prevailing culture, which has historically valued men over women and white people over people of color. None of what we do as individuals is intentional. That’s what’s sneaky about this. It’s a set of subconscious biases that are “in the water.” The only way you can see their effects is by looking at the overall trends.
98/2 is that trend. And it doesn’t look good.
What we might be missing
None of this is news, exactly, since the research goes back to the 60s, but it’s vastly more important for us to think about right now. The graduates of coding “bootcamps” are much more gender-diverse than programmers at large. Many new female developers, lacking formal qualifications, are targeting the open source and startup worlds. Here’s our chance to get some gender diversity onto our teams! But because of these biases we carry, we are not in a good position to evaluate women realistically, even alongside the men who graduated the bootcamp with them.
There is something we can do
I am not advocating that we depend more heavily on formal qualifications. We can work on this problem without losing the spark that makes our community so appealing. These biases are subconscious, but that doesn’t mean we can’t do something about them.
The first thing you can do is notice the gap – recognize and acknowledge these assumptions when you make them.
Just noticing is actually pretty hard. I don’t always see it when I make the less-technical assumption about a woman, particularly if it turns out to be true. It took that mistake at the SF Ruby event to see that my assumptions, even if usually true, were hurting my chances of connecting with other female developers. Then I started seeing it when I evaluated résumés. Between two bootcamp grads, one male and one female, more often than not I found a reason to prefer the man. Because of the science, I now have to believe that it’s not just a matter of their qualifications.
Start watching your own interactions of this type and noticing what you assume about the people involved.
The second thing you can do is step across the gap – make a conscious effort to combat those assumptions.
When I meet a woman at an event these days, I assume she’s technical. No matter what she looks like, I always assume up rather than down. Once I establish more context, of course, I adjust. This means that when she’s technical, she’s starting at the right place instead of with a deficit. The price I pay is that I briefly give more credibility than is warranted to recruiters, but I can live with that, given the incredibly positive upside.
When I’m hiring, I have an HR intern (or the external recruiter) strip anything that could indicate gender or race from the résumés before they get their initial evaluation. For the ones that make the first cut, I have the recruiter print out code from Github, with the username redacted. This has resulted in a tremendous increase in the number of women who make it through to an actual interview. The price I pay here is administrative — someone not on the hiring team has to do all that work. It can be a way for recruiters to add a lot of extra value, but even if I have to pay for it myself, the cost pales in comparison to the economic benefit of a diverse team.
And lastly, when I’m evaluating conference proposals, I use software that conceals demographic information during the first round of reviews. The cost? Sometimes a frequent speaker who writes a shitty proposal doesn’t get selected, and sometimes they’re upset. Again, I’ll take that, given the incredible upside.
I still screw it up
I’ve been thinking about this a lot, and trying to consciously adjust for my bias. That’s what I call minding the gap. But I still mess it up, particularly with race. That’s ok, because it’s still way better than laboring under unconscious bias that I don’t even know about. I try to acknowledge my screw-ups when they happen, and point out what was wrong. Then I’ll move on. Over time, I hope these small acts will adjust my bias more towards reality.
Minding this gap doesn’t mean staring into it all the time. I’m trying to get somewhere, after all. But if I ignore it, I’m looking at the world and seeing a warped reality. And clear eyes are good for business.
Postscript
A few years ago, before I was working with DevMynd(software development company), I started a consulting company with another developer. Since the main advantage of running a company is you can put whatever you want on your business cards, we gave ourselves fun titles that didn’t sound technical.
After a few months of enthusiastically handing out my new cards and repeating my joke-y title to anyone who would listen, I started noticing something weird when I met with people outside the Ruby community. Every single person assumed my male co-founder was the one with more programming experience. Even though nothing in the titles suggested that, and even though it wasn’t true. But somewhere, he got the benefit of a doubt, and I did not.
I am lucky to have an enormous amount of perceived competence within the Ruby-language community. I am, as the kids say, “internet micro-famous.” This gives me enough flexibility to choose who I work with and what I work on. So when I noticed these consistent wrong assumptions about my programming skills, it didn’t feel threatening.
But it did give me pause. I remembered what it felt like at the beginning of my career when everything was so uncertain. When something like this would have felt like a clear sign that I wasn’t in the right place. And I kind of wonder why I stayed.
