I think a particularly valuable part of this question is the encouragement for everyone on Academia SE to take part in the SE survey, which allows a space for some short comments, as well as an opportunity to volunteer for follow-up research. The more that active users engage with that, the better a picture we'll get of what's working and what opportunities there are for change.
The more that this is qualitative research, rather than just counting people up in categories, the more useful I think this effort will be. That sort of user experience work can surface complex issues and potential solutions. On the other hand, it's hard to tell what benchmark Academia SE demographics should be compared to.
- For instance, it's a worldwide site, so the meaning of certain race/ethnicity categories is hard to interpret.
- People of South Asian and East Asian descent are 16.5% of the SE developers (according to Nat's great answer), but 35% of all people in the world live in China and India.
- Should it be benchmarked against the world's English-speaking population? Maybe not, since most academics around the world are incentivized to work in English.
- More deeply, asking about race/ethnicity is not based on some idea that all people who are white or of European descent are interchangeable or share a deep set of characteristics. Rather, the category provides useful information in the context of a given country; a particular race/ethnicity means they are likely to have been treated in a certain way or that we are more likely to be able to predict other correlated traits.
- So far in the academic literature, I haven't seen treatment of how to ask meaningful world-wide questions about race/ethnicity, other than tailoring the questions to different countries/cultures to capture the distinctions that matter in that society. (I am very aware that many people find the idea of "race" itself offensive, particularly Europeans.)
- SE users are self-selecting in a lot of ways.
- I speculate that Academia SE users are more likely to be early career rather than late career academics, which might make the pool more demographically diverse. (Not sure the extent to which undergraduates and grad students are involved.)
- Because SO and SE are built on programming questions, people who do programming and computational work are more likely to discover and sign up on Academia SE. I believe that even within academic disciplines in the U.S., more quantitative and computational work tends to be done by men. (Definitely my own experience; I believe I've seen documentation of this, and would edit in a reference if anyone has one off-hand.)
- Are there elements of SE culture that are further causing self-selection? In bad ways?
A lot of measurements seem like their meaning should be self-evident, but that's rarely the case. It's also very easy for people to list statistics and for others to infer blame from those statements. Further, when people know stats but aren't used to social science or stats about people, it's easy to come up with calculations that are technically correct but misleading or misinterpreted. It may be most productive if we can agree on some basic descriptive facts and withhold normative judgment from those numbers alone.
When I saw this question, I became really worried it would explode into the conflict we've seen on other parts of the site in response to this issue. My initial thoughts were that the most productive way forward might be to not press the issue but to keep doing useful things, like Atlanta's rebranding itself as "The city too busy to hate."
I don't think avoiding the real problems people experience is the right thing to do, but in this case I think that those problems are better explored by people discussing their experiences, rather than reading into demographic numbers.
Edit: I took a look at the questions in the poll Massimo Ortolano commented about, and those seem like they may be helpful for Academia SE to understand its users.
Edit: Why am I hesitant about this conversation? Because it's being addressed in different ways across SE/SO. This answer on SE Meta covers a lot of it. Then there are various ones (responses to the blog post and others on the topic) that demonstrate how volatile these discussions can be:
SE Meta 1, SE Meta 2, SE Meta 3, Interpersonal Skills (IPS) Meta 1, IPS Meta 2, IPS Meta 3, SO Meta 1, SO Meta 2, SO Meta 3, SO Meta 4, SO Meta 5, SO Meta 6, SO Meta 7