Sunday 9 April 2023

The Errors Will Ramify For Years

A lot of people have been in touch to say that of course Margaret Thatcher was most recently portrayed by a drag queen, since gender self-identification was the inexorable logic of the self-made man or the self-made woman, and since a comparable figure, emerging in the Britain of the 2020s, would be assumed to be a transwoman, just as Thatcher herself emerged in the Britain of everything from Danny La Rue and Dick Emery to David Bowie and The Rocky Horror Show. Quite so. As to the Britain of the 2020s, a stalwart of old school Trotskyism at Counterfire, which is as sound on this and on Brexit as the Morning Star is, has sent me this, by Michael Biggs:

Did you realise that one in every 67 Muslims is transgender? That adults with no educational qualifications are almost twice as likely to identify as transgender as university graduates? That the London boroughs of Brent and Newham are home to higher proportions of transgender people than Brighton and Oxford? These are some of the astonishing results from the 2021 census of England and Wales, which was the first in the world to ask about gender identity.

The Office for National Statistics (ONS) released detailed census data for England and Wales on Tuesday. These data deepen the problems raised by Alice Sullivan, professor of sociology at University College London, and myself after initial results were released in January. As the census is the gold standard of data collection, to query its numbers seems sacrilegious. The results for gender identity, however, are scarcely credible.

The problem began with the question itself. The ONS did not ask a plain question like: ‘Are you transgender?’ Instead, it chose a convoluted formulation: ‘Is the gender you identify with the same as your sex registered at birth?’

Those who answered ‘no’ could then write in their gender identity. This question assumes that everyone has a gender identity. It also assumes that everyone was registered at birth. As the human rights campaigner Maya Forstater has emphasised, some immigrants were not registered at birth. The question could have puzzled many respondents outside the professional and managerial classes. How many of them mistakenly answered in the negative

Anomalies appear when we look at the distribution of the transgender population by local authority. Newham and Brent top the list, with 1.5 per cent and 1.3 per cent respectively. Brighton and Hove, with 1 per cent, ranks only twentieth. Yet Brighton is the LGBTQ capital of Britain, home to the country’s longest-running Trans Pride celebration. It is also the site of two universities, including the University of Sussex where some transgender activists and their allies effectively ousted philosopher Kathleen Stock for writing about sex. Could Brighton really have a less salubrious climate for trans people than Newham? It seems unlikely.

This data is also hard to tally with that gathered from a petition to reform the Gender Recognition Act, launched by trans activists in 2021, which attracted 118,000 signatories from England and Wales. By comparing the distribution of signatures to the distribution of the transgender population according to the census, we see that the correlation between the two distributions across all 331 local authorities is close to zero. Brighton and Hove, for example, had more signatories, relative to adult population, than anywhere else except the City of London. Newham and Brent, by contrast, had relatively few signatories, ranking 303rd and 304th respectively.

If places like Newham and Brent had so few inhabitants willing to sign a pro-transgender petition, what accounts for their prominence in the census transgender figures? What these boroughs have are many immigrants for whom English is a second language, and who are therefore liable to be confused by a convoluted question on gender identity. The strongest predictor of the transgender population across 331 local authorities, as measured by the census, is the proportion of people whose main language is not English.

Not all the numbers are contaminated. The census disaggregates the transgender population into five categories. Two of the categories are helpful for determining the number of trans people in a particular area: those who wrote in ‘non-binary’ or some other novel identity like ‘gender queer’, comprising 48,000 respondents. They are distributed geographically much as we would expect, with Brighton and Hove having the largest concentration. It is the remaining three categories which are likely to include some confused non-transgender respondents: trans man, trans woman, and those who specified no identity, totalling 214,000 respondents.

The census data released this week added to the list of anomalies. Muslims are almost three times more likely than non-religious people to identify as transgender. Black people are four times more likely than white people to identify as transgender. In every case, the census results contradict what we know from other data. For example, referrals to the Tavistock’s Gender Identity Development Service include far more white and non-religious youth than the overall population of the same age. By contrast, those census results are what would be expected if a number of people with poor English were confused by the question and inadvertently classified themselves as transgender.

The ONS has also finally this week released customised data showing the tabulation of gender identity by proficiency in English. As predicted, those who speak English ‘not well’ or ‘not well at all’ were most likely to be counted as transgender: 2.2 per cent of them, compared to 0.4 per cent of those whose main language is English (or Welsh in Wales). Adults whose main language is not English made up only 10 per cent of the overall population, but according to the census they contributed 29 per cent of the transgender numbers.

How did the ONS manage to produce such implausible data on gender identity? In ‘a case study of policy capture’, the statisticians were guided by lobby groups like Stonewall. It is surely no coincidence that the gender question replicated, with minor variation, Stonewall’s definition of ‘cisgender’: ‘Someone whose gender identity is the same as the sex they were assigned at birth’.

The question, according to the ONS, was ‘evaluated via community testing at LGBT History Month events’. But did the ONS consider how immigrants whose first language is not English – and who may be blissfully ignorant of esoteric concepts like ‘cisgender’ – might understand the question, or rather could misunderstand it?

The decennial census is the bedrock of British statistics. Errors in data collection not only impede our understanding of society, but also misallocate government funding. Newham Borough Council, for example, will come under pressure to devote more of its budget to the substantial number of trans people recorded by the census, when – if my argument is correct – those numbers have been significantly inflated. The problem does not end with the census, because its question on gender identity has become the default for taxpayer-funded surveys in England and Wales. The errors will ramify for years.


  1. Did you answer the gender identity question?

    1. Yes. The question was voluntary, because they wanted most of us to ignore it.

      If one tenth of the people who had answered at all had ticked No, then all the talk forever thereafter would have been of "one person in every 10". Or every five. Or every three.

      It was vitally important that we ticked the Yes box. Yes, each of us identified with the sex registered at birth. And the State needed to know that.