Read on Twitter
Gender inclusivity in survey design! A thread!
(This is aimed at people doing statistical analysis and unsure how to account for trans people. If you're not doing stats it will look different. This could still be useful though.)
(This is aimed at people doing statistical analysis and unsure how to account for trans people. If you're not doing stats it will look different. This could still be useful though.)
First step: ask why you're collecting information about gender at all. Will it be theoretically useful to you? Are you expecting to use it in your analysis? If not, don't ask.
But we do see gendered effects expressed a lot, so you might want to at least control for it, right?
But we do see gendered effects expressed a lot, so you might want to at least control for it, right?
Second step: create the most parsimonious number of categories you can to make sure everyone who takes your survey can answer this question.
This means something along the lines of "woman", "man," "non-binary person," and an open-ended "other."
I can hear you objecting...
This means something along the lines of "woman", "man," "non-binary person," and an open-ended "other."
I can hear you objecting...
...that there won't be enough people in NB or any particular category of "other" to be statistically relevant and you'll have to drop those responses from your analysis anyway.
You can't know if you don't ask, but more importantly: if you don't do this, you're introducing bias.
You can't know if you don't ask, but more importantly: if you don't do this, you're introducing bias.
Either people will answer in a way that isn't accurate, which is a validity problem, or they'll drop out, which is a selection bias problem. If more NB folks are, say, young, your sample's about to skew older. Etc. Either way, you want them in your dataset.
Even if you can't draw statistically significant conclusions with the other categories - and who knows, you might be able to! - you can use them as controls. (Analytically, you'll probably be recoding the "other" responses.)
But now I hear you saying, wait, what about trans people who aren't non-binary?
Well, you already asked for gender. Next question: why do you need to know if someone is transgender?
If it's not analytically useful to you, you don't need to know. "Women" includes trans women.
Well, you already asked for gender. Next question: why do you need to know if someone is transgender?
If it's not analytically useful to you, you don't need to know. "Women" includes trans women.
Unless you're trying to parse gender effects pretty finely and specifically looking at trans people vs cis people, you don't need to know trans status.
If you do need to, ask it as another question. "Are you transgender?" works pretty well.
If you do need to, ask it as another question. "Are you transgender?" works pretty well.
So if you do need to know both of these things, we have a pair of questions:
Which of these best describes your gender? (4 answers)
Are you transgender? (yes, no, other)
That's it. Now people can participate in your survey.
Which of these best describes your gender? (4 answers)
Are you transgender? (yes, no, other)
That's it. Now people can participate in your survey.
You're going to have to code a bunch of "other" responses by hand to run your analysis, but these questions are pretty loaded for a lot of people, so having that open-ended category allows people to participate who couldn't otherwise.
Also, trans people are very used to people doing this extremely badly, which means you'll get credit for not being entirely hamhanded and terrible.
Ask only what you need to know + ask it in a way that will get you valid answers = people respond who would otherwise drop out.
Ask only what you need to know + ask it in a way that will get you valid answers = people respond who would otherwise drop out.
