Queering the Census: Demographic considerations of adding (and changing) questions on gender and sexuality

In light of many countries debating the addition, expansion, and/or implementation of gender and sexuality census questions, this paper seeks to answer the following question: Through the census, how can we ethically and meaningfully capture “hidden” populations, and, in particular, sexual and gender minorities? What can we learn from the current generative moment, spurred by discourse across and within countries? Using censuses from the US, UK, New Zealand, and Australia—four countries presently at different stages of census development around sexual and gender identity, I evaluate considerations of counting the LGBTQ+ population. I chose these countries based on a comparative quadrant: On one axis, a country could either change (or not change) a census question on sexual orientation and gender identity (SOGI) and on the other, a country could either add (or not add) a census question on SOGI. In this case, the US did not add or change; the UK added but did not change; Australia changed but did not add; and New Zealand has intent to add and change. Using these four countries as case studies, I have been able to analyze the parallel developments of SOGI questions and the challenges/best practices that arise. Some of the initial concerns and challenges at the outset of adding these sociodemographic questions include, from an ethical side: i) privacy concerns and undercounting, ii) the weaponization and politicization of data on marginalized populations, and iii) respondent burden. From a methodological side, there are i) debates on whether queerness can be quantified, ii) survey specific concerns (such as question ordering, response options, question type—like multiple choice or free response, etc...), iii) census specific concerns (such as whether a respondent of a certain age is allowed to answer a sexuality question; whether the head of household responds on behalf of all household members; data protection and transfer between local and national governments, etc...), and iv) comparability/consistency across longitudinal data sets (e.g., whether this meets the needs of data users, particularly from the empirical social sciences, who might be reliant on comparability with past “sex” questions). This paper addresses several of these concerns and aggregates best practices. For example, I address the debates surrounding two-step gender questions vs. gradational gender questions vs. open-ended gender questions. I further explore the potential of using machine learning on open-ended gender responses to address concerns of data concordance, and I recapitulate the existing literature that discusses the miscoding rates associated with conflated sex/gender questions (e.g., on the General Social Survey). I also bring in some preliminary evidence from a survey that I ran using some of the census questions on nationally representative samples in the UK, US, and Australia. In all, this paper describes and aggregates challenges/best practices of SOGI questions through comparative analysis of censuses from the US, UK, Australia, and New Zealand. By understanding the discourse across four countries presently at different stages in the development process of gender and sexuality questions, this paper attempts to follow the recommendations that result from this generative moment.