I am what you might call a “heavy user” or, as I prefer to think of it, a “key stakeholder” of the Office of National Statistics’ website and data releases. As a result, I have been following the ONS’ efforts to achieve an accurate picture of the total size of the United Kingdom’s trans population with considerable interest.
So, I’ve been somewhat taken aback by the reporting and commentary over the issue, and around the ONS’ recent court defeat, which has forced it to revert to using the mandatory question asking for sex, rather than gender. (A reference to “use the sex recorded on one of your legal documents such as a birth certificate, Gender Recognition Certificate, or passport” was rewritten to exclude passports on the grounds that this could allow “self-identification through the back door”.)
The central challenge that the ONS faced is that sex and gender are different things, but the differences between the two are not commonly understood by most people, regardless of their gender identity. Indeed, one problem the ONS faced is that the two words in Welsh, rhyw (sex) and rhywedd (gender) are very similar and the difference between the two was even less well-understood.
In practice, of course, that doesn’t matter, because for most people, their sex is the same as their gender, so that they don’t really understand the distinction between the two things: they are going to give you the “right” answer no matter how you phrase the question.
Further compounding the problem is that if people dislike or don’t understand the question, they can and do give unhelpful answers whether through confusion or as a result of protest. For example, the reason why the census now separates the question of national identity from that of ethnicity is to stop black, mixed, and Asian people from writing in “Scottish Asian, Black British, Mixed English” and so forth. This is not an issue of “feeling” but of the ONS’ ability to produce and maintain robust and useful datasets.
To that, the ONS had an added headache: whether to persist with the question as is, which is one of the longest-running continuous data series in the world, or to refine it to get a more accurate answer. One of the longest-running debates in this world is whether you are better off persisting with a flawed data series that you can compare over time, or starting afresh with a better question but losing some of the ability to make comparisons over time.
To take the question of ethnicity: at every census since 1991, when an ethnicity category was first introduced, the ONS has introduced a new “high-level” category. This gives us a more accurate picture of the United Kingdom’s diversity at the time of the census in question, but makes it harder to compare patterns over time.
A similar and fraught debate is playing out in the British Jewish community. In 2001, the Institute Jewish Policy Research (JPR) asked that Jews be counted as an ethnic category, not a religious one, but were rebuffed by the ONS, a decision I believe to have been the wrong one. In 2021, the JPR’s position is that they now have 20 years of comparable data on the British Jewish population, and that it is better to persist with a flawed but comparable dataset rather than start again.
So there is a serious and difficult debate to be had about whether the ONS is better off persisting with the “female/male” question, even though we know that a large number of people answer it as if they are being asked “man/woman”, simply because it can be compared better over time.
We know that people take the current question to be one about their gender, not their sex. I see no reason to believe that people in 1981 were more aware that the question was about sex, not gender, not least because people in 1981 were much less open about sex and gender even than they are today. But is it better to have a question that we know people are not answering “properly” that we have asked over a long period of time, or is it better to have a different question that gives us an accurate picture of people’s sex and gender?
As a heavy user/key stakeholder of the ONS’ services, I am strongly in favour of a flawed, long running data series over a new, but more accurate one. I think the benefits of strong comparisons over time, which aid scrutiny of policy decisions, outweigh the benefits of having a ‘better’ question. But I accept that for many British Jews and for many people, regardless of their sex or gender identity, they would prefer a more accurate question now rather than a longer-running data series.
As such, I think it’s a good thing for data collection purposes that a judicial review forced the ONS to maintain the current question, which asks respondents to answer what their sex is based on either their birth certificate or their gender recognition certificate. This means that the answer in 2021 will more closely resemble that in 2011, 2001, and so on.
At no point has the ONS failed to recognise the importance of sex as a category for policymakers – indeed the importance of sex as an important and separate category for policymakers to be aware of has been a golden thread of everything the ONS has published on the issue. (An exhaustive, if somewhat wonkish, run-through of all its literature can be read here.)
The reality is that, if you are genuinely concerned about a blurring of sex and gender, you should be concerned about the current question because, as the ONS’ question testing has showed, people do not understand the current question, and different people answer it differently, while the legal documents that people are directed to use (a birth certificate or a gender recognition certificate) can have different answers, and in any case, provide answers to different questions.
Sex and gender are already blurred in the census and always have been: “self-identification” already takes place. (And when you think about it, this should be obvious: when you completed the census yesterday, at no point does the census website produce a small robot to judge the pigmentation of your skin, your sexual characteristics or your supposed level of piety.)
The issue has not been that the ONS has been overly sensitive towards the “feelings” of trans people, and the assumption that a more aggressive or robust question will get you a more robust or “accurate” answer is simply wrong: the ONS tested a more aggressive question, asking people what they “consider” their gender identity to be, and respondents did not understand what the question was asking.
Imperfect questions have always been a feature of the census, with people having to refer across multiple categories to draw conclusions. We will, at the end of this census, have an imperfect but long-running gauge of sex and gender, and our first, but still imperfect gauge of the number of trans people in England and Wales. Helpfully for the interests of gauging question effectiveness, the Scottish census is taking a different approach, and we will be able to draw a factual and better-grounded conclusion about how best to ask this question across the UK as a result.
The difficult challenge of how best to reliably collect census data will continue across essentially every issue, but lambasting the ONS, and assuming that only a belligerent question will get the “right” answers as far as sex and gender are concerned wasn’t the right approach to the issue this time and it won’t help to finesse the question in 2031 either.