Official figures estimating the number of transgender people in England and Wales have been stripped of their accredited status after an investigation found that too many people misunderstood the 2021 census question.

In an embarrassing setback, the Office for National Statistics has requested a reclassification of the finding that there were 262,000 trans people in England and Wales on census day, equivalent to 0.55% of the population.

After an investigation by the statistics watchdog, this will no longer be treated as an official statistic, but will be deemed “experimental”. It is the first time this has happened to census data since the Office for Statistics Regulation was formed before the 2011 census. The ONS had claimed the data would be useful to “aid allocation for resources and policy development”.

Concern about the figures emerged when the census appeared to show that the east London borough of Tower Hamlets had the highest proportion of trans people, at one in 100 – more than in Brighton, long recognised as a popular place for LGBTQ+ people to live.

The figures are used by policymakers to better understand their populations’ needs for services. For example, Tower Hamlets council issued a “state of the borough” report based on the now questionable data telling residents that 2,643 people in the borough aged over 16 identified as trans men, trans women, non-binary and other gender identities.

After an investigation by the OSR, the ONS has now conceded that “some respondents may not have interpreted the question as intended, notably people with lower English language skills in some communities”.

Dr Michael Biggs, a professor of sociology at Oxford University, said the admission of error was “long overdue” and called for the question to be rewritten where it is used in other official data gathering.

“The same question is used in the GP patient survey, and has been used to claim that trans people are more likely to suffer from dementia. This result could easily be an artefact of the bad question– older immigrants with poorer English are more likely to misunderstand the question and more likely to have dementia,” he said. “So there are very real consequences of this flawed question.”

For the first time the 2021 census for England and Wales set a voluntary question about gender identity that asked: “Is the gender you identify with the same as your sex registered at birth?” Of those who answered yes, about 2% did not speak English well. But of those who answered no, around 13% did not speak English well.

The gender-critical feminist Kathleen Stock criticised the question as being “couched in obscure genderese” that would “stump non-initiates”. Stonewall, the LGBTQ+ charity, said the figures “finally paint an accurate picture of the diverse ‘Rainbow Britain’ that we now live in”.

The overall proportion of trans people in England and Wales gleaned from the census (0.55%) is in the same ballpark as the 0.44% for Scotland. Statisticians north of the border posed a clearer question in their 2022 census: “Do you consider yourself to be trans, or have a trans history?”

The problem comes when the figures are used on a local basis.

The Office for Statistics Regulation concluded: “The statistics do not comply with important quality aspects of the code of practice for statistics.” It said the question “did not work as intended”.

“The evidence indicates that people may have found the question confusing and therefore gave a response that did not reflect their gender identity. This appears to be more likely for people who do not speak English as their first language.”

The OSR said the ONS had “not shown sufficient willingness to engage constructively with the emergent evidence … We consider it regrettable that ONS’s defensiveness has created an impression of bias to some external observers.”

Emma Rourke, deputy national statistician at the ONS, said: “The available evidence does show there is potential for bias in how the gender identity question was answered by those in England and Wales who responded that they did not speak English well. Our request to reclassify better reflects the findings coming from the first census question of its kind and our developing understanding of measuring this complex and important topic.”