How gender ideology corrupted government data

Blurring the line between sex and gender has serious consequences.

When, in September 2024, the Office for National Statistics (ONS) downgraded data from the census for the first time in its 120-plus-year history, it shocked many. Having spent years ignoring advice and warnings about how to gain accurate data on both biological sex and the transgender population, the ONS asked such a confusing question on gender identity that it rendered the findings meaningless. It transpired that a sizeable number of people picked up as trans weren’t: their level of English simply meant they hadn’t understood the question.

While this part of the 2021 census, which as a whole cost taxpayers almost £1bn, was undoubtedly an expensive failure of data collection, it was far from being a one-off. In February 2024 the government commissioned the UCL professor Alice Sullivan to review how data on sex and gender identity was collected by public bodies. Her report, published on 19 March, revealed that scores of official statistics and data sets have been corrupted over the past decade. “The term ‘sex’ has lost its ordinary meaning in data collection,” she wrote, with implications not just for public policy, but for safety and safeguarding. Some of the greatest risks have been to children.

In the 1990s, “gender” crept into official data collection as a synonym for sex. Figures shared with the New Statesman show that before 1990 just 2 per cent (or less) of questions used the word gender instead. That increased to 16 per cent during the 1990s, and 37 per cent between 2000 and 2009. More fundamental changes have taken place over the past ten years. First, “gender” began being used not interchangeably with sex, but as a proxy for gender identity. Then, questions on sex were replaced by ones explicitly about gender identity. From 2020, the majority of questions did not ask about sex at all. A quarter asked about respondents’ “gender identity” – a contested and ill-defined concept – and 41 per cent asked about “gender”. Only 17 per cent asked explicitly about sex. How did the UK sleepwalk into collecting years’ worth of dodgy data?

Some institutions were influenced by lobbying from trans-supportive groups and by ideologically driven employees. Others simply took their lead from the institution trusted to have the highest standards of data collection: the ONS. Many, Sullivan wrote, cited definitions used by the ONS – and its 2021 census question in particular – “as their touchstone for devising questions on sex and gender”. They were confident the ONS “would have carried out extensive and rigorous testing of the questions”. The ONS did test its proposed wording – on 300,000 households, in 2019. “Some respondents found this question difficult to answer” and it was given a “medium” risk rating for its “potential for impact on data quality”. The ONS asked it anyway.

Some of the instances of poor data practice in the report are jaw-dropping. Department for Education statistics on looked-after children allowed gender to be “recorded according to the wishes of the child” in some circumstances. (This was changed for the 2024-25 reporting year when binary sex was recorded, but it’s extraordinary it was ever deemed acceptable not to know the sex of society’s most vulnerable young people.) The ability of individuals to change the gender listed on their NHS records has not just led to the “loss of vital healthcare data” – undermining data used to improve drugs safety, risks and lifetime outcomes – but has also put transgender patients at risk. Lab results can be misinterpreted, and trans people whose health records do not record their birth sex are not called for sex-specific cancer screenings (such as cervical smears). Wes Streeting has banned under-18s from being able to change their gender marker on safeguarding grounds following publication of Sullivan’s report.

Peter Kyle, the Science Secretary, has shared the report with all government departments. There is plenty in it they could and should act upon. Take criminal justice: we do not know the sex of 1 per cent of offenders deemed most risky to the public, and we cannot compare police data because of inconsistent collection. Official figures on sex in the Criminal Justice Statistics bulletin are “a mix of self-reported and officer identified”. Of 21 police forces that responded to a freedom of information request in 2023, 13 answered that self-declared sex would be recorded for the suspects of crimes. Only one force recorded the sex of rape suspects “irrespective of either a [Gender Recognition Certificate] or self-declared gender identity”.

Sullivan’s recommendations are not rocket science. Some boil down to: “Ask a straightforward question.” Data on sex should be collected – by default – in all research commissioned by government, she says. And attempts to gain important information on gender self-identification should be asked in a separate, clearly worded question. As Sullivan told the BBC, there’s no reason to see her suggestions as “a trade-off” between sex and gender identity. They are, “two distinct variables… [and] we should be recording both”.

A once-in-a-decade opportunity to gain an accurate picture of the trans community, and better address its needs, was squandered. Ideology has been allowed to damage data. It is impossible to know how much public money has been wasted in the process. We cannot correct the past, but we can ensure sex is accurately recorded in the future. Sullivan is surely right: “Everybody, regardless of their politics, needs accurate data for good policymaking.”

[See also: The Palantir problem]