Aware of the server issues that had plagued the 2016 census, I filled out my online 2021 census form a day early. I was diligent, and I filled it out as though it were today, August 10th. It wasn’t hard: census day is also Day 45 of the second Sydney lockdown, so none of us are going anywhere today. No-one is going to work, no-one is going to school, and certainly no-one is coming to visit the three of us stuck in my little two-bedroom apartment.
I entered everyone’s name in the correct order on the online form. I stared at my section, the sense of anxiety rising as I knew it would. I’ve been dreading this census for months, and now here it is. With a little bitterness I note that the form doesn’t police the order in which I fill out the sections, so I can leave the part about me until the very end.
I filled out the information for the children first. With the schools closed I took the opportunity to talk with my eleven year old son and my eight year old daughter about the Australian Bureau of Statistics, its role, and its function. We talked about why we have a national census, and why it’s important for planning and service provision. We talked about data privacy. I asked the kids their opinions about the pros and cons of releasing identifying information to the public after 99 years. Should I, as their parent, refuse permission on their behalf now – while they are still young – so that they can make a better informed choice when they get older? Should they make a choice now? If I were to let them make the choice themselves, what choice would they want to make? I want my children to grow up making informed choices. Both of their parents are statisticians: in our family we take these things seriously.
It was a nice family moment. I could focus on the kids for a while, and not think about what would happen next. But inevitably, the kids returned to their ipads and I went back to the census. There it was. The question I’d been dreading.
The kids are absorbed in their activities, so they don’t notice I’m crying.
There is a public policy need to measure biological sex in the census. Planning for obstetric services, for example, is fundamentally tied to a persons reproductive biology. Of course, the truth of the matter is that while humans are a sexually dimorphic species, the actual biology is quite complicated and doesn’t reduce neatly to a binary “male or female” dichotomy. For various reasons, there’s a small proportion of the species who are naturally intersex, and their sexual characteristics are such that they can’t easily be assigned to either category. For this reason, the census question on sex has an explanatory note:
If these options do not describe the person, they can select something other than male or female.
This note contains a link that takes you to an expanded version of the question that includes a third response option:
☐ Non-binary sex
The language used here is peculiar. “Non-binary sex” isn’t really a term used anywhere in the LGBTIQ community or in the medical literature. It’s a strange hybrid of two quite unrelated terms: “intersex” is the term that refers to people whose anatomical characteristics are not easily classified as male or female, whereas “non-binary” is the term that refers to people who do not describe their gender as male or female in their everyday life. Conflating these two things is a bad idea. Intersex people and non-binary people are different groups, and they have different needs. The question as framed is quite misleading. On twitter, I saw a lot of non-binary people (who I’d guess are probably not intersex) indicating that they chose the “non-binary sex” option, so the 2021 census is almost certain to grossly overestimate the size of the intersex population. Statistically, this terminology is extremely poorly chosen: the data from anyone who chooses this option can’t be used for planning, and it can’t be cross-correlated with other survey instruments that measure sex and gender correctly. I hate to say it, but the “non-binary sex” data will necessarily have to be excluded from all statistical, scientific, and demographic work that uses the census. If you ask the wrong question, it doesn’t matter how honestly and diligently people answer it, your data will be garbage.
Even more peculiar is the fact that the public messaging by the ABS is… confused, at best. On twitter, when asked by a person whose gender is non-binary why no questions were asked about number of children they had given birth to, the ABS responded in a way that made clear that the question is fundamentally about biological sex:
That\'s correct. In testing, this question caused distress for those whose non-binary sex characteristics means they're unable to have children. In close consultation with the LGBTIQ+ community, the form was designed to not present this question.--- Australian Bureau of Statistics (@ABSStats) August 8, 2021
Elsewhere on twitter, the ABS encourages a transgender woman whose passport and birth certificate have different gender markers to choose the option that best represents her:
This question relates to sex, there's no question on gender identity, although this is collected in other ABS surveys. See https://t.co/m7fZZVKVK3. We encourage you to answer with what you think best represents you.--- Australian Bureau of Statistics (@ABSStats) August 8, 2021
This is very peculiar, particularly given that the ABS has published standards for assessing sex and gender that are extremely clear about the intended referent of the question. I’ll quote the opening section in full because it is very precise:
The Standard for Sex, Gender, Variations of Sex Characteristics and Sexual Orientation Variables, 2020 (“2020 Standard”) has been developed by the Australian Bureau of Statistics (ABS) to standardise the collection and dissemination of data relating to sex, gender, variations of sex characteristics and sexual orientation.
This 2020 Standard replaces the Standard for Sex and Gender Variables, 2016, with updated sex and gender variables, as well as the introduction of variables for variations of sex characteristics and sexual orientation.
This product presents statistical standards for four variables:
- Variations of sex characteristics
- Sexual orientation.
The 2020 Standard describes the four variables and their associated conceptual issues and definitions. The standard for each variable includes the concept(s), definition(s), questionnaire modules, classification, coding structure, and output categories to be used in ABS interviewer-based and self-enumerated collections. The 2020 Standard also provides guidance on deriving cisgender (cis) and trans and gender diverse (trans) counts using the sex and gender variables.
The four variables included in the 2020 Standard, when cross-classified with other variables, can provide comprehensive data on a particular topic, issue or population group. This information can be used for a range of purposes, including:
- informed decision making and planning
- policy formulation and monitoring
- social, population and economic research and analysis
- program provision and evaluation (e.g. health services).
The 2020 Standard can be used by other government, academic and private sector organisations in their own statistical collections to improve the comparability and quality of data.
If you read through the standard properly, you’ll see that it’s very good. If you measure all four of these variables you are able to obtain an accurate picture of the LGBTIQ population. But you do need to measure all four variables. If you don’t measure all four, you end up with a statistical mess, because it is now impossible to extract the thing you actually care about. If a legally and medically transitioned transgender woman (e.g., me) follows the advice provided by the ABS on twitter, she would probably choose “female”. Except if you read further into the ABS standard, this will be encoded as “female sex”. Here’s the data dictionary the ABS provides for the variant of the question that was asked in the census:
Preferred code Alternative code Label Definition 1 M Male Persons who reported their sex as male. 2 F Female Persons who reported their sex as female. 3 X Non-binary sex Persons who reported their sex as non-binary sex.
The data dictionary is unambiguous: this is a question solely about biological sex. It is not a question about gender. The ABS is very precise in differentiating between sex and gender, and their standard makes the distinction as follows:
The terms sex and gender are interrelated and often used interchangeably, however they are two distinct concepts:
- Sex is understood in relation to sex characteristics. Sex recorded at birth refers to what was determined by sex characteristics observed at birth or infancy
- Gender is about social and cultural differences in identity, expression and experience. This is discussed further in the Gender variable.
While they are two related concepts, caution should be exercised when comparing counts for sex with those for gender. Caution should also be exercised when comparing counts for sex recorded at birth and the sex of a person at the time of completing a survey, as a person’s sex may change over the course of their lifetime.
As a transgender woman, my sex recorded at birth was “male”. However, my passport and birth certificate both read “female”. My sex chromosomes are (almost certainly) XY, but my hormonal system runs on estrogen. I have breasts, but I don’t have a uterus. I was not intersex at birth, but because of the medical treatment I’ve received my body now possesses some quite substantial “variations of sex characteristics”. On the surface, I could make a case for any of the three responses, but the context – in which the ABS classifies this as a question about sex, not gender – is informative. In context, the question is not intended to represent transgender identities in any sense: it’s asking you about your body. That’s it. If there had been any intention of taking transgender inclusion seriously, the ABS standard clearly states that the census version of the question is the wrong one to use:
For collections requiring cis and trans outputs, sex recorded at birth is the required sex question. This is discussed further in the Cisgender and Trans and Gender Diverse Classification section.
The census is written in a way that does not allow cis and trans outputs to be extracted, so the question used doesn’t specify “sex recorded at birth”. Nevertheless, the variant of the question that the census used does not appear alongside the gender questions. Rather, it is classified as a sex question. I’m not a statistically naive person, and I know how to read an ABS standard. Irrespective of what the ABS account says on twitter, the actual standard makes clear that I’m expected to call myself a male. But even if I do, the question is still pointless for transgender or intersex people. Very few people read the ABS standards, so in practice they’ll act in accordance with the “commonsense” recommendations that the ABS twitter account is giving. As a consequence transgender people will mostly (but not exclusively) respond with their self-identified gender, non-binary people and intersex people will both choose non-binary sex, and everyone else will just choose what they always choose.
I hate to say it, but from the perspective of the LGBTIQ population, the census data are junk.
Lyle Shelton must be very proud.
How did the Australian Bureau of Statistics mess this up so badly? They aren’t an amateur organisation. On the contrary, the ABS is filled with people who take these matters very seriously and who care deeply about ensuring data quality. That’s hardly a surprise since the ABS is the primary body for collecting, curating, and disseminating statistical information about the Australian population. The people who work there are statistically sophisticated, and as their own standards indicate, they are fully aware of the importance of separating sex, gender, variations of sex characteristics, and sexual orientation. In fact, precisely because it is well understood that these things need to be disambiguated in administrative data, it was proposed that the census should measure all four variables. The proposal was rejected.
You can click through and read the official reasons, but I honestly don’t believe that these are the real reasons. On twitter, Casey Briggs noted that the Deputy Australian Statistician Teresa Dickinson recently gave an interview to the ACEMS podcast in which this topic was discussed.
For those interested in more on the sex question in the Census, this interview with the Deputy Australian Statistician on the @ACEMathStats podcast goes into a bit of detail. The main discussion of this issue is from approx 29:40 https://t.co/Xp2tTU3bdR--- casey briggs (@CaseyBriggs) August 9, 2021
This is what she said:
“We won’t be collecting information on gender or sexual orientation… It’s the government that makes these decisions and after our very long process of assessing and testing questions that was where we ended up.”
She went on to say that other ABS surveys do include more inclusive measures of gender and sexual orientation… just not the census. She refers to the UK, NZ, Canada which all do include these measures in their census. We don’t. For Australia, she goes on to point out that there’s a long history of the census not including important questions when they are proposed, and gives the example of the “income” question, which was proposed for the 1971 census, which was testing very well with the general population, but was rejected because a government Minister had been part of the test group and he didn’t like being asked the question. It was his view that “held sway”, according to the Deputy Australian Statistician, and so it did not appear until the 1976 census. Her perspective on this is that “sometimes these things take a while to percolate”.
What should we make of all this?
The ABS standard is awfully clear on the subject, and the ABS does ask the questions properly in other contexts. Upon being asked about the topic, the Deputy Australian Statistician was at pains to point out that the ABS does not have autonomy in choosing the census questions, and that it is the government who decides. She goes on to give an example from the 1971 census that sounds a lot like political interference in the census. I cannot prove it, and nobody has said so explicitly, but my best guess is that the ABS knows perfectly well that the census question is wrong. I think the statisticians were overruled by the politicians.
If that’s what happened, it is a serious matter. People need to trust that the ABS makes choices solely on statistical grounds, and that the census is not subject to partisan political meddling. Ministers should not be overruling statistical advice any more than they should be overruling health advice. I sincerely hope that’s not what has happened here, but I rather suspect that it is.
Staring at the screen, I agonise about what answer to provide. All three options are misleading. It feels viscerally painful. It hurts. I don’t want to lie on the census, but the question isn’t asking me anything that I can answer properly. There isn’t any way I can respond that properly describes what I am. Someone in the government made the deliberate decision to make that impossible for me to do. I did my best to answer honestly, and I still don’t know if the answer I gave was the right one. For someone who really does care about the integrity of the census, that hurts.
Later that night, my kids ask me why I’m crying. I try to explain, as best I can. As a statistician, how do I tell my kids that the census is pointlessly cruel for reasons that appear to be nothing short of political interference? As a trans woman, how do I tell them that their government despises one of their parents? When my eleven year old son asks me why they didn’t just let people be who they are and respond honestly about our identities, what am I supposed to say? I’m usually good with words, but this time I don’t have any. My eight year old daughter does, though.
“That’s fucking stupid,” she says.
I can’t argue with her there.