[origin_path]

The census is a set of forms. It isn't like there's someone in front of you trying to guess your age, gender etc. What "self-identify" means here is that you're allowed to pick whatever you want regardless of biological truth, whereas normally lying on a census can result in a fine.

[ZeroGravitas]

This literally is the history of race on census forms.

A census taker would decide what race you are, from categories thought up by other people.

https://www.census.gov/data-tools/demo/race/MREAD_1790_2010....

> While people nowadays are enumerated by race based on self-identification, until 1950 their race on the census was mainly determined by their census enumerator.[170] During this time multiracial people who were White and of another race were usually marked down as belonging to the other race due to the One drop rule.[170] The instructions provided to enumerators endorsed this practice.

[Quarrelsome]

I think "biological truth" is loaded here. In terms of a census there is an ambiguity about what it is that you're counting, are you counting people biologically born a given gender, people who are going to pair with the opposite sex, people who prefer a particular prefix or people that can give birth to children?

It appears in this case the census is content in counting people who prefer a particular prefix and only counting that (which can be self-identified) achieves its aims.

Tbh I don't massively see how this changes how one might treat census data (especially since the numbers remain small enough). I would suggest that if your research requires data points for people born as women then maybe you should be looking for other datasets such as births.