For the last few years, all the official student surveys coming from my university have offered three options for gender: male, female, and transgender. I appreciate that they’re trying—it’s better than only listing male and female—but arg, no, that’s not how it works.
By itself, “transgender” isn’t a gender; rather, it’s a descriptor meaning that your gender identity doesn’t match up with the gender you were assumed to have at birth. Transgender people are male and female and non-binary, not an extra separate gender.
I’ve seen other surveys that attempt to do better by offering four options: male, female, transgender male, and transgender female. However, that kind of setup implies that trans people aren’t “real” members of their gender. It would be somewhat less problematic if the first two were specifically listed as “cisgender male” and “cisgender female”—but if you really need to know whether your survey respondents are cis or trans, consider breaking that into another question. The Human Rights Campaign has a good example of a survey approach that separates “What is your gender?” from “Do you identify as transgender?”
Of course, as alluded to above, gender is not binary, and your survey also needs an option for people who are non-binary / genderfluid / genderqueer / agender / etc. (Not to imply that these terms are interchangeable, because they are not, just that at bare minimum there needs to be some kind of “outside the gender binary” selection available.) If your survey design allows it, an additional option with an open text field will help you avoid unintentionally excluding anyone. And personally, I’d also like to see a “prefer not to say” choice for gender, as is common on some of the other demographic questions.