Is It True That Trans Women Are More Likely to Be Sex Offenders?

Answering the question takes two figures, so without them there is no actual answer. No country publishes how many trans women are convicted of sexual offences each year. England and Wales, whose figures both graphics use, does not have a count of how many trans women there are that is accepted as accurate.

If you were sexually abused as a child, you can talk to someone about it. In the United States, the Childhelp National Child Abuse Hotline is open at any hour: call 800-422-4453, or text GO to the same number. In the United Kingdom, the National Association for People Abused in Childhood (NAPAC) runs a free support line for adult survivors on 0808 801 0331. It is open Monday to Thursday, 10am to 9pm, and Friday, 10am to 6pm.

The January graphic appeared in January 2023 and is still being shared. The June graphic was published on 15 June 2023 and carries the same three rows for England and Wales: women, men and trans women.

The June graphic's author writes a Substack under the name Scientific Parent of a Trans Kid. In May 2023 he set out to check whether the January graphic's linked sources supported it, then rebuilt it from Ministry of Justice figures and the census.

The January graphic reaches its 1,916 per million by dividing 92 trans women in prison for a sexual offence by a census count of trans women. It gives no source for the 92. The figure is in neither of the documents the graphic links to, the prison population tables and the census release.

When the June graphic's author went looking for it in May 2023, the only source that turned up was a Ministry of Justice reply. It gave 56 trans women in prison on 31 March 2021 whose principal offence was sexual.

The 92 is not an official figure; no government count gives it. It comes from the blog post that first published the graphic, by Mar Vickers on Wings Over Scotland on 10 January 2023. The blog post added two numbers together. One was 87, the trans women in men's prisons with a sexual offence in a government answer to Parliament. The other was 5, the blog's own figure for women's prisons.

The government answer did not give 5 for women's prisons. It gave fewer than five, so its figures cannot add up to 92.

The 87 also counts more widely than the male row it is set against. It takes in anyone with a sexual offence among the offences they were imprisoned for, while the male row counts each prisoner once, by their main offence.

The June graphic's author later suggested that the 92 matched figures the Ministry of Justice released between April and June 2023. But the graphic had been published in January, before those figures were released.

The January graphic's male and female rows are divided by all-ages population totals, 29.5 million men and 30.4 million women. But its trans row comes from a census question put only to people aged 16 and over. The error understates the male and female rates, so the trans row looks larger by contrast than it should even on the January graphic's own numbers.

Dividing the male row by the 16-and-over population instead of the all-ages one raises the male rate from 395 per million to 496. It also cuts the ratio of the trans row to the male row from 4.8 to 3.9.

The January graphic also takes its male prison figure, 11,660, from June 2021 rather than March 2021. The March 2021 count, the closest to census day, gives 11,558. This error is small and runs against the graphic's own claim, since the larger June count raises the male rate.

The January graphic's denominator, the number it divides by, is 48,000, the rounded figure the census published. It is the same count of respondents classified as trans women that the June graphic gives unrounded as 47,572. That count carries the same problems, examined in the section on the June graphic.

Download this conclusion: the January graphic

This card checks the numbers in the January graphic. The 92 has no official source. The Office for National Statistics (ONS) says the 48,000 is not a reliable or accurate count of trans women. The real 56 over the analyst’s own population estimates gives 464 or 289 per million. Men counted the same way come to 496. And a prison count is a different group from trans women in the community.

Download the card (PNG, 1,080 × 1,350)

The June graphic was built by the author who found the January graphic's problems, and it fixes them. It takes prison figures from March 2021 rather than June, and population totals limited to people aged 16 and over. It also puts 56 in place of the unofficial 92.

The 56 comes from a Ministry of Justice Freedom of Information (FOI) response. It counts the prisoners in England and Wales on 31 March 2021 who were legally male, recorded as identifying as female, and had a sexual offence as their principal offence. A second FOI response in June 2023 confirmed the scope: legally male transgender prisoners identifying as female only, excluding non-binary, male and unrecorded identities.

The 56 counts each prisoner once, under a single principal offence. So it leaves out any prisoner with a sexual conviction from an earlier sentence, or whose sexual offence is not the one recorded as principal. It therefore understates how many have a sexual conviction at all.

Two further limits apply to the 56. HM Prison and Probation Service (HMPPS) counts as transgender only prisoners known to staff who have been through a local case board. A local case board is a meeting of staff, chaired by a prison manager, that plans a transgender prisoner's care and management. HMPPS says its figures are likely to underestimate the true number.

Prisoners with a Gender Recognition Certificate, which legally recognises a person's affirmed gender in the UK, are excluded from the transgender totals entirely. The Ministry of Justice was asked in July 2023 how many of the ten certificate holders were serving a sentence for a sexual offence. It answered "five or fewer" for both a principal and any sexual offence.

The 47,572 is the count of census respondents aged 16 and over who answered that their gender identity differed from their sex registered at birth. Each also gave a write-in answer that the Office for National Statistics (ONS) coded as trans woman.

On 12 September 2024 the Office for Statistics Regulation removed accredited official statistics status from the census gender identity estimates, at the ONS's own request. It did so after a review found the question had not worked as intended. The review also found that some respondents, more often those without English as a first language, may have misunderstood it.

ONS guidance of 26 March 2025 states that the estimates should not be used to size the population identifying as each of trans man, trans woman and non-binary. That is exactly what the June graphic does with the figure.

The ONS does not say the number is too high. Linking census returns to birth registrations confirmed that some respondents had been recorded as trans in error. But the ONS could draw no conclusion about the scale or effect, and published no corrected figure.

Of those the ONS had coded as trans man or trans woman, almost two-thirds, 63%, wrote in something that was not a clearly trans answer, the 2025 guidance found. What they wrote also matched their response to the sex question. For example, a respondent recorded as a trans woman might have ticked female for sex and written "woman" in the identity box. The guidance gives no separate figure for trans women.

A separate ONS cross-tabulation, published 25 January 2023, adds a category mismatch. Of the people classified as trans women, 66.16% answered female to the census sex question and 33.84% answered male. This does not mean the 47,572 were not registered male at birth. The sex question was self-completed, and its guidance advised using the sex on a birth certificate or Gender Recognition Certificate. Nothing shows how many of these respondents did.

Nothing in either dataset establishes that the 56 sit inside the 47,572. The 56 are defined by legal gender, while the census offers a self-reported sex and a free-text identity write-in.

The larger figures in that table are the author’s own. Four weeks before the June graphic, he had published a broader estimate from the same census table: “120,805 trans-identified males”. He reasoned that the census cannot establish the birth sex of the respondents who reported a different gender but were not coded as a trans man or a trans woman. He counted half of that group as male. Divide 56 by his estimate and the rate falls below the male rate printed on the June graphic itself.

He set the broader estimate aside when an FOI confirmed that the 56 counted only prisoners identifying as female. That is the whole of the case for the 1,177 per million now in circulation. He wrote at the time that the precise definitions used and the reliability of the census numbers had a large effect on the calculated rates.

Download this conclusion: the June graphic

This card checks the June graphic’s numbers. The 56 is a real count. The ONS says the 47,572 is not a reliable or accurate count of trans women. The analyst’s own earlier population estimates give 464 or 289 per million. Both are below his 490 for men. And the prisoners are not the same demographic as trans women in the community.

Download the card (PNG, 1,080 × 1,350)

The graphics count who is in prison on one day, and that depends as much on how long people stay as on how many arrive. A one-day count alone cannot show how many people were convicted.

Sentence length is what pulls the one-day count and the number convicted apart. Sexual offences carry some of the longest sentences in the system, so they crowd any one-day headcount.

In England and Wales in the year ending March 2026, the average custodial sentence was 7.8 months for theft and 73.9 months for sexual offences. For indictable offences as a whole, the more serious offences that can be tried at the Crown Court, it was 22.3 months.

The Ministry of Justice reports that 23% of prisoners serving an immediate custodial sentence in England and Wales on 30 June 2026 were there for a sexual offence. It also reports that 99% of those admitted to prison under an immediate custodial sentence for a sexual offence between January and March 2026 were male. Nobody reads those figures and concludes that nearly one in four men is a sex offender.

A prison holds a sample selected by what carries a long sentence, not a sample of the public. From the other end, the June graphic's own male row puts 490 men per million in prison for a sexual offence in March 2021. That is about one in two thousand, or 0.05%.

The same holds for women, a group nobody accuses of mass criminality. The Prison Reform Trust reported on women sent to prison under sentence in the year to June 2018. More went for theft than for violence against the person, robbery, sexual offences, fraud, drugs and motoring offences combined. Nobody concludes from that that most women are shoplifters.

That is the reading the graphics ask you to switch off. They take a share inside the prison population and present it as a fact about people outside it. Done to any group, that produces a number that looks alarming.

Download this conclusion: the prison is not the population

What a share inside a prison does and does not tell you about the people outside it, as one image.

Download the card (PNG, 1,080 × 1,350)

Take the same 100 convictions a year, with every sentence served in full. Sexual offences then leave more than nine times as many people in custody on any given day as theft. Early release shortens both figures and does not change which way the effect runs. The only variable in this comparison is sentence length.

The graphics compare one group of sexual offenders with another, and it is tempting to conclude that the buildup of long-sentence cases cancels out between the rows. It cancels only if the two groups serve the same average time. A headcount is the number of people arriving each year multiplied by how long each one stays. So dividing one headcount by another gives a ratio of offending rates only when the stays are equal.

If the 56 are serving longer sentences than the male row's 11,502 men, or were convicted longer ago, their row is inflated relative to the men's at identical offending rates. If their sentences are shorter, the reverse holds. Nothing published compares time served between the two groups. So a one-day headcount cannot produce an offending rate for any row, and the ratio between the rows is not a ratio of offending rates either.

Measuring an offending rate rather than a headcount requires convictions per year by transgender status. Sentencing data records sex, and nothing that shows whether an offender is transgender.

Download this explanation

This section as one image. It shows why a count of who is in prison on one day cannot be turned into a rate of offending. The England and Wales sentence lengths on it show the size of the effect.

Download the card (PNG, 1,080 × 1,350)

Canada’s prison service has recorded, for its gender diverse prisoners with a sexual offence, whether they offended while living as their biological sex. No equivalent exists for England and Wales.

Correctional Service Canada’s 2022 study covered a very small group of gender diverse federal offenders, of whom a minority had a history of sexual offending. Most of that minority were trans women. Gender diverse, the service’s term, includes offenders who are transgender, non-binary, gender fluid or intersex.

The study recorded whether each person was living as their biological sex at the time of the offence. The group rests on self-identification, and Correctional Service Canada describes these offenders as self-disclosed.

How someone was living is not a record of how they identified privately. So the study cannot separate two groups it may contain. The groups are people who adopted a transgender identity after conviction, and trans women who already identified as trans but offended before disclosing it. The 94% counts everyone in both groups who offended while still living as their birth sex.

The record also cannot isolate anyone whose declaration that they are transgender is false. The England and Wales prison policy on transgender prisoners requires a case board to examine the credibility of a declaration, including counter-evidence, where there is any concern about sincerity. Lord Justice Holroyde ruled on that policy in the High Court in 2021. He wrote that its boards could be expected to be astute to detect a prisoner who, for sinister reasons, was merely pretending.

No published figure exists either way. Nothing counts how often a declaration is insincere, and nothing counts how often the boards get it right. Anyone falsely recorded as transgender would sit inside these counts and inside the graphics built on them. There is no way to say from the data how many that is.

The study's offenders are federal prisoners, which in Canada means sentences of two years or more. They were identified over a 27-month window rather than on a single day. So the same buildup of long-sentence cases described in the section on prison headcounts affects who appears in it.

The 6% is too small to show whether transition, or any step within it such as hormones or surgery, raises risk, lowers it, or does nothing. The study's full report also records nothing about those steps for the people it studied.

Both graphics use the same three steps. They take a group of people in prison, take a share within that group, and present the share as though it were a rate for a whole population. That is how a prison headcount and a disputed population figure become "1,177 per million." Apply the same three steps to the Canadian finding: take the 94%, present it as a rate, and the headline is "transition reduces sexual offending by 94%."

That headline should not convince anyone, and the reasons are the ones that apply to the two graphics. It rests on one small group of prisoners in one country, and the published findings say nothing about hormones or surgery, so they cannot describe an effect of transition. A share within a small group is not a rate for a population.

The same three steps give opposite answers from the same kind of data, so they do not measure offending in either direction.

Download this explanation

The Canadian test as one image: what Canada found, what the graphics’ three steps turn it into, and why the same reasons that sink that headline sink the graphics.

Download the card (PNG, 1,080 × 1,350)

The June graphic's arithmetic is correct. Its author went looking for the January graphic's sources, published what he found, and updated it in public five times over the next ten weeks. Three of those updates rested on new Freedom of Information responses. He marked the January graphic unverified rather than defending it. He fixed the prison figure's month, fixed the all-ages population base, and replaced the unofficial 92 with a figure he could point to in a government document.

The June graphic removes the 56 from the male prison row, leaving 11,502 men. It also removes the 47,572 from the census count of men aged 16 and over, leaving 23.48 million. So the groups being compared do not contain the same people twice.

Both graphics divide by a census count that the Office for National Statistics (ONS) now says should not be used to estimate how many trans women there are.

The June graphic’s author had an alternative of his own, and it turns the graphic the other way. His earlier estimate was “120,805 trans-identified males”. Divide 56 by that estimate and the rate is 464 per million. That is below the 490 his chart prints for other men. The 120,805 is an estimate too, and the ONS guidance leaves no census figure that can size this population with confidence.

Even if nobody disputed the prison count or the population it is divided by, a one-day prison headcount is not a rate of offending.

Run Canada’s 94% through the graphics’ three steps and you could print “transition reduces sexual offending by 94%”. The graphics’ own audience would reject that headline, and every reason for rejecting it applies to the graphics.

A prison records someone as transgender only once they are in its custody and the prison service knows it, whether at reception or later. The record describes how the person is living on the day it is made. It does not say how they were living when they offended.

Correctional Service Canada identified the offenders in its study within its in-custody population. Its accommodation is driven by what each offender asks for, whether or not a prisoner pursues hormones or surgery.

In England and Wales, HM Prison and Probation Service (HMPPS) counts a prisoner as transgender on two conditions. Staff must know them to be living in, or presenting in, a gender different from their legal gender. The prisoner must also have had a local case board, a staff meeting on their care and management. HMPPS records that status as at the date of the count rather than the date of conviction.

Both the request and the record usually come in prison, and always after the offence. In Canada’s adult criminal courts, the median sexual assault case in 2017/2018 took 247 days from the first court appearance to a decision. A federal sentence begins only after that. In England and Wales, Crown Court defendants dealt with between April and June 2026 waited a median of 365 days from the offence to the completion of their case.

The male row they are measured against has no equivalent entrance. Nobody joins the male row by declaring anything in custody.

Each graphic divides its prison count by a count of trans women living in the country, and asks the reader to take those prisoners as a sample of them. The England and Wales counts do not record how the prisoners counted were living when they offended. In Canada’s study, 94% of those with a history of sexual offending had committed their offences while living as their biological sex.

So the significant majority of those counted as trans sex offenders were not trans women already living as women when they offended. A rate only works when the top and bottom numbers count the same group. Even setting aside the other issues, any rate these counts produced would not be about trans women who already live as women in everyday life, because they are a tiny minority among those counted as trans sex offenders.

When the figure is a percentage instead

A second version of the claim drops the rate and quotes a share: some percentage of transgender prisoners are in for a sexual offence. That version has no denominator to attack, and it fails for the reason the section on prison headcounts showed with men and women.

A share inside a prison describes the people the courts sent there. It describes nobody else, in any direction, for any group. And the share being quoted here describes a narrower group still: people a prison service has recorded as transgender while in its custody.