> I feel like this is missing the point that "gender" in a questionnaire is not asking you what you think you are, but what you really are.
No, in this case it's a combination of what you are (i.e. your gender identity), and what whoever decides on how much they'll pay you thinks you are, mostly. But there's all sorts of variables in there still - a single variable probably isn't wonderful. Is a trans woman usually paid more or less than a cis woman? Is a genderqueer person usually paid more or less than X, etc etc? Does their identity filter into it or only their presentation - maybe people pick up on subconscious cues?
> There is a scientific definition of the word gender that is being ignored here.
The scientific community, since they accepted the existence of transgender people, mostly uses "sex" for the biological makeup of a person - and I'll point out that even that's not entirely binary, with the generally entirely ignored group of intersex people.
> I don't think people are interested in the .1% that is transgender as it will probably be too low to make valid statistics.
This sort of assumption has the issue of othering groups of people - asserting that certain groups of people are "other". This has issues well past individual studies - it's a societal issue where we decide that certain groups of people are not worth listening to. There's a set of comics written by a queer author on the subject here: http://www.robot-hugs.com/notice/
When you're designing a form, it's not that difficult to accept options that you haven't thought of straight away - in fact, it's bad data science not to, as otherwise you get a bunch of people who either refuse to respond or box themselves in to whatever you force them to be, even when it's inaccurate, and you can't distinguish them from anyone else. After all - the question "which religion are you?", as might've been used in times past, would ignore atheists, and it'd be even worse if it only accepted the top 3-4 religions in a given country.