2022: 90.0% men
2023: 77.1% men
2024: 83.7% men
edit: literally why would someone flag this lol
23 Likes
2022 (as per the announcement, it says 2156 participated, so that means 2156-1649-95-88/2156 = 15% haven’t answered it)
2023 (assuming everyone answered since there is 14% as no answer)
2024 (all of 2290 respondants answered)
I think we have definitely improved this question’s choices. (0.7% not answered, is great to see)
so @cafkafk the real values are
2022: 76.5% men ~15% no answer (90% among answered)
2023: 77.1% men ~14% no answer (89.6% among answered)
2024: 83.7% men ~0.7% no answer (84.3% among answered)
10 Likes
Ohh I didn’t realize! It does seem that 2022 had all bins optional (at least that’s what it’s announcement says).
Also, I do think it’s cool how the survey has generally improved on these questions, from binning everything other than male/female into “other”, to including a not answered portion in 2023, and then making several improvements in 2024, including male/female being replaced with man/woman, differentiating between not wanting to respond and not answering, and making non-binary/non-conforming a distinct category.
Survey improvements aside, that does seem to indicate it’s even more bleak, since it seems that having more people answer didn’t decrease the relative share of men. Also that suggests that the demographic change isn’t transient but a trend, going in the wrong direction 
8 Likes
Side note: according to DebianWomen/Projects/Statistics - Debian Wiki and Debian Project -- debian.org Developers LDAP Search, female participants currently accounts for 2.9% of Debian Developers. (Not to mention that they at least have this project)
2 Likes
I’ve observed several leaving to participate in other communities, like Lix or Aux, so it may have been indicative of a bigger trend, or there were so few in the first place that that migration significantly reduced the % on its own. I don’t know how many of such users still participate here, anecdotally it seems not many. In any case until the decisionmaking apparatus changes, I would not expect this trend to reverse.
3 Likes
In any case until the decisionmaking apparatus changes, I would not expect this trend to reverse.
I think this as relatively little to do with any decision-making apparatus. I believe it depends much more on the underlying distribution you are sampling from.
Just to illustrate what I am thinking of. JetBrains (like many others) is doing similar surveys. One such example is here, specifically the section Gender distribution in tech over the last 3 years: The State of Developer Ecosystem in 2023 Infographic | JetBrains: Developer Tools for Professionals and Teams
It does look remarkably similar to the results that we have here.
4 Likes
I’m not sure we’re interpreting the same data here. What @phanirithvij showed is that the actual relative share between those who did answer did in fact improve significantly towards equal by ~5 points and that that’s greatly accelerated compared to the change between 2022 and 2023 where it only improved by 0.4 points (which likely might as well have been noise).
2 Likes
I was not commenting on percent that didn’t answer, but on trends related to gender participation, but ultimately I’ve realized it’s impossible to say anything truly interesting without the underlying data, and even then I don’ think we’re enough people not to be easily affected by noise.
Regardless, the approximate number of men compared to other genders in the project isn’t ideal (despite how much worse debian may be!), I think there’s no debating men seem to dominate in numbers, regardless of the exact percentage.
5 Likes
Outreach, lots more outreach and making the community and the place more inclusive in general would be a great step. We need to make sure women have a seat at the table in the community and projects. The same applies for minorities as well.
9 Likes
We already have Summer of Nix. Sounds like a good thing we could use to reach out to people who are less likely to bump into Nix and get them excited.
1 Like
Also there is https://www.outreachy.org/ which is great at targeting people underrepresented in general, and if we could get a partnership with them it would be huge for diversity. IIRC guix does this, in a style like summer of Nix.
11 Likes
I find this renewed discussion on that question a bit strange.
As I mentioned above: if you are sampling from an underlying distribution that has 90% male (however defined) individuals, and you are sampling in a non-biased fashion, your expected value in your sample is going to be 90% male. In other words: we should rationally expect to get roughly 90% male participants in an industry that is 90% male overall if we are unbiased in our sampling.
What’s more: if you introduce a sampling bias against men, for whatever reason, and you assume that men don’t just disappear from the software industry altogether… well these men need to end up somewhere else, right? Logically. In other words, if we “succeed” in getting more women into the project, for example due to the introduction of some sampling bias, assuming the underlying distribution stays the same, the men will end up in other places. And these places will then “fail” to attract women. Even though their supposed “failure” is actually caused by our introduction of biased sampling.
As for trends, a linear regression with three datapoints is a bit of a wild guess. And if you do a polynomial regression, the result with three datapoints is arbitrary for practical purposes. So the data that we have says nothing about trends, really.
In addition to that the questionnaire has changed as we aren’t asking the same questions year after year. So we don’t even know if we are getting data for the same variables.
If we really wanted to take this entire thing more seriously we’d have to ask: what is really causing the sampling bias in the underlying distribution in software with respect to the general population? And can we, as a project, even intervene on that? Because if we can’t then we shouldn’t worry too much, but people should rather campaign on national / societal politics or something. But if we can, then we would actually know what to do and we could start doing that.
15 Likes
I think we can do stuff that is a net positive like outreachy without having to fight the entire system, what’s so bad about helping underrepresented people, it really doesn’t hurt anyone else, it’s not like outreachy is taking spots from men or something?
Also stop saying there are no concrete proposals. I’ve proposed outreachy, it doesn’t get more concrete than that!
12 Likes
Part of my point was actually that it indeed does “hurt” someone in the sense that if everyone is competing for the same 10% of women in the overall industry, one improving their percentage necessarily means someone else is decreasing theirs.
That’s a false premise, if there’s more welcoming spaces then women wouldn’t quit the industry at the current rates. This community can and should be one of those spaces.
Also last time I checked, people can use more than one technology at the same time.
EDIT: I’ll also leave this as my last response on this thread since it’s going to be another one of those discussions where we waste energy expanding on a quite self-evident premise.
14 Likes
I don’t think it’s the case that we’re competing for 10%… As in I neither agree that it’s a finite pool nor 10%… Just providing opportunities for more women fresh to the industry that might otherwise go in other directions hurts no one, I know plenty of women who’ve simply moved to different engineering things because tech just was too unpleasant for them.
7 Likes
That is an interesting data point. Could you share the source?
EDIT: I’ll also leave this as my last response on this thread since it’s going to be another one of those discussions where we waste energy expanding on a quite self-evident premise.
What is the self-evident premise? Most of these things are not self-evident. People just assume stuff. But if you really want to do something that is effective and doesn’t have a whole host of negative unintended consequences, you need to think this through.
I don’t think it’s the case that we’re competing for 10%… As in I neither agree that it’s a finite pool nor 10%
The pool is most certainly finite, even though you might assume some rate of annual replenishing here, plus people leaving. But if that annual replenishing and the people leaving don’t change the underlying distribution much, then it does not matter really to my argument.
That being said, the 10% is something I just assumed based on incomplete evidence from looking at this one survey from JetBrains. I don’t know that it is 10% for sure. But I don’t think it’s totally implausible either.
(That being said: I’ll probably log-off for today as well. But think it was important to make my point, because I am really not convinced there is a massive problem on our end. Nor am I convinced we, as a project, could do a whole lot to influence the overall gender distribution in software.)
3 Likes
This claim would do well with a citation
Generally speaking, for instance, you’ll see women’s representation be much larger in industry than in open source projects, and it will vary widely by different projects, countries, and many other factors, so yea, one jetbrains survey really can’t be generalized to draw large conclusions.
4 Likes
This claim would do well with a citation
Not really. The pool is always finite. There is always going to be a finite number of people in every way you might categorize them.
Generally speaking, for instance, you’ll see women’s representation be much larger in industry than in open source projects, and it will vary widely by different projects, countries, and many other factors, so yea, one jetbrains survey really can’t be generalized to draw large conclusions.
I agree. I made an assumption for the sake of an argument. That was a crass generalization. I don’t think it was totally implausible though. And the argument stands with 5% or 15% as well.
We could however attempt to arrive at a more realistic estimate. That might better be an entirely different thread though. For that we would need to review a lot more data. And ask a few more important questions. Like where do we actually get new people from?
We could attempt some kind of a meta-review about known distributions of all sorts of characteristics of people, including gender.
(that’s a serious long-term idea btw. … anyways, I am off now)
…like yes there is a global maximum (x amount of people exist in the universe) that if you measure instantaneously is finite in that instant, but the distribution of that really is not finite in the way you’re saying it is… and it’s a very long stretch to imply that makes it harmful to adopt outreachy…
3 Likes