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

Source: a written parliamentary answer, 14 January 2022, using data collected 31 March 2021. It states three figures. 146 trans women were in prison in England and Wales. 87, in the men's estate, had a conviction for at least one sexual offence. In the women's estate, the number with such a conviction was reported as "fewer than five."

"Fewer than five" is a disclosure limit, not a count. It is used when the true figure is small enough that publishing it could identify a specific individual. The value it represents could be 0, 1, 2, 3, or 4.

The graphic assigns this figure a value of 5 and adds it to 87, producing 92. Five is not one of the values "fewer than five" can take. The largest permitted is 4, and 87 plus 4 is 91. No government document states 92, and no reading of the source produces it. The true total is somewhere between 87 and 91.

Two further discrepancies in the same source. The offence figures cover only the conviction currently resulting in imprisonment; the government states it could not provide prior-conviction data within the cost limit for a written question. And the 87 includes people with a Gender Recognition Certificate, while the 146 total excludes them. The two halves of the graphic's own source count different populations.

Source: the 2021 census. Respondents were asked whether the gender they identify with matches the sex registered at their birth. 262,000 answered no. Of those, 48,000 wrote in "trans woman."

On 12 September 2024, the Office for Statistics Regulation withdrew this data's accredited status. Its stated finding: the question did not function as intended, and patterns in the responses indicate many respondents, particularly those with lower English proficiency, misread it.

The ONS's subsequent guidance states the figure "should not be used to provide estimates of the sizes of the population who identified as each of trans man, trans woman, non-binary." That is the exact use the graphic makes of it.

Supporting data: of respondents coded as trans man or trans woman, 63% gave a write-in response that did not clearly indicate a trans identity and was consistent with a non-trans reading of their sex response. Separately, 117,775 respondents indicated their gender differs from their birth registration and gave no further identification. Both facts move the true figure in opposite directions by an unknown amount. Neither can be quantified from this dataset.

Holding 92 constant, the result changes as follows:

A sixfold range on the single input the graphic cannot verify.

The prison figures are a count on a single date: 31 March 2021. This measures who is in custody that day, not how many people were convicted that year. The two numbers diverge whenever offences carry different average sentences, because the mix of offences among people in custody right now is not the same as the mix of offences among people convicted in any single year.

Consider two offence types, each convicted 100 times a year, spread evenly across the year. One carries an average sentence of a few months: people convicted of it pass through custody quickly and are replaced by next year's convictions soon after. The other carries an average sentence of several years: people convicted of it are still inside when the next several years of convictions arrive and join them. Walk into a prison on any single day and count how many people are there for each offence, and the longer-sentence offence will make up a far larger share of that day's headcount than it does of a year's convictions, even though both are committed equally often. A single-day count measures a stock: who is in prison right now. It does not measure a flow: how often something happens.

England and Wales publishes real figures that show this gap. In the year ending December 2024, the average custodial sentence for theft was 8.1 months, a series low, while the average for indictable offences overall was 22.0 months. In the year ending June 2024, the average sentence for sexual offences was 69.5 months, the longest of any offence category, against 45.2 months for robbery and 41.4 months for drug offences. Sexual offences and robbery also carried the highest custody rate of any offence group in 2024, 56% and 55%: a conviction for either was more likely to result in any prison time at all, on top of producing more of it once it did.

Both offence types are convicted at the same rate: 100 per year. Theft's 8.1-month average sentence means about 68 people convicted of it are in custody on any given day. Sexual offences' 69.5-month average sentence means about 579 are, more than eight times as many, from the identical number of annual convictions. The only variable in this comparison is sentence length. Every row in the graphic counts people in prison for sexual offences, so every row reflects this same accumulation. The table does not show that trans women's row was inflated more than the other rows. It shows that a same-day headcount cannot produce an offending rate for any row.

To measure an offending rate rather than a custody count requires convictions per year, by transgender status. No such data has been published in England and Wales or in any other jurisdiction. Prison services publish counts taken on a single date. Sentencing data does not record gender identity. HMPPS began publishing the reported biological sex of transgender prisoners in its 2024–25 report, but that is still a snapshot of who is in custody, not a count of who was convicted.

What is not known, for trans women specifically, is what the remaining convictions are once sexual offences are set aside. No official publication breaks the non-sexual convictions of transgender prisoners down by category. A 2023 Freedom of Information request that asked the Ministry of Justice for exactly this detail was answered only with a further breakdown of sexual offence sub-types, not with any category outside it. In the prison population generally, theft and violence against the person are the most common offence categories by volume, and theft carries one of the shortest average sentences of any category, as the table above shows. If the non-sexual convictions of trans women in prison follow a broadly similar mix, that would be consistent with the accumulation effect described above: a larger number of short-sentence convictions passing through the count quickly, next to a smaller number of long-sentence sexual-offence convictions that persist in it. That is a plausible contributor. It is not a demonstrated one. Confirming it would require the same missing data as everything else in this section, trans women's convictions broken down by offence type, which has not been published.

This limitation is not specific to the graphic's authors. It applies to any calculation, made by anyone, using currently available data.

The only study to examine when offences occurred relative to a person living in their identified gender. No equivalent exists for England and Wales.

Correctional Service Canada, 2022. 99 gender diverse federal offenders identified over a 27-month window. 33 with a history of sexual offending. 82% of the 33 were trans women.

What the study recorded is how each person was living and presenting at the time of the offence. That classification rests on self-identification. There is no record, in this study or any corrections dataset in Canada or elsewhere, of a person's gender identity before they entered the system, so the study cannot separate two different groups it may contain: people who adopted a transgender identity after conviction, and trans women who already identified privately as trans but offended before disclosing it. The 94% figure counts everyone in both groups who offended while still presenting as their birth sex.

Read at face value, the finding says that in this group almost all of the sexual offending happened before the person transitioned. Only 2 of the 33, about 6%, offended after they were living as their identified gender. That is the whole of the post-transition evidence the study contains.

The limits matter as much as the finding. This is 33 people in Canadian federal custody, which covers sentences of two years or more, identified over a 27-month window rather than on a single day, so the same buildup of long-sentence cases described above affects who appears in it. Because only two people offended after transition, the study cannot show whether transition, or any specific step within it such as hormones or surgery, raises risk, lowers it, or does nothing. The report records nothing about those steps. What the data support is narrow: in this small group, sexual offending occurred overwhelmingly before transition, not after.

The infographic's method has three steps. 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 "1,916 per million."

Apply those same three steps to the Canadian numbers. Of the 33 gender diverse prisoners with sexual offence histories, 31 offended before transition and 2 offended after. Take that share, present it as a rate, set aside that it describes 33 people, and set aside that the study recorded presentation at the time of the offence rather than any transition step. The headline this produces is "transition reduces sexual offending by 94%."

That headline should not convince anyone, and the reasons are the same ones that apply to the original graphic. It rests on 2 people in the post-transition group. It comes from one country. The study measured nothing about hormones or surgery, so it cannot describe an effect of transition. A share within a small group is not a rate for a population.

The method returns a confident number whichever way it is pointed, including a direction the graphic's own audience would reject. A procedure that yields opposite conclusions from the same kind of data is not measuring the thing it claims to measure.

The graphic's specific numbers fail. A narrower, separately sourced pattern exists in the same data, and it carries its own caveats rather than the graphic's.

Written parliamentary answers show that among transgender prisoners who reported their legal gender as male, the share with a sexual offence conviction has run above the male prison population's roughly 17% in every year published: about 49% in earlier years, 56% in 2023, and 62% (151 of 245) as of 31 March 2024. These figures do not depend on the disputed census estimate, but they are shares within a few hundred people, not an offending rate, and the stock-versus-flow limit from Part 3 applies to them just as it does to the graphic's headline number. A share this size, from a group this small, can move by double digits with a handful of cases either way, and it says nothing about the far larger population of trans women who have no criminal record.

Two factors could inflate the share on their own, apart from how often anyone offends. Sexual offences carry the longest average sentences of any category, so convicted people accumulate in a single-day count more than others do. Recording someone as transgender in custody also requires a case-board review, which a short sentence may end before it happens, while a long sentence, correlated with sexual offences, has time to be recorded. Neither factor is quantifiable from published data, and neither accounts for the whole gap. What remains is a composition figure inside one small custodial subgroup. It is not evidence about the risk posed by trans women as a population, and the same limits would apply to any other small group's prison statistics read the same way.

Answering whether trans women commit sexual offences more often than anyone else requires a count of convictions per year, matched to the number of trans women in the population, held over enough time to be stable. No country publishes this. The prison figures that exist count who is in custody on a single day, which is not the same as how often a group offends. Until a jurisdiction publishes annual convictions by transgender status against a sound population estimate, the question has no numerical answer, and neither this graphic nor any other can supply one.

The graphic's 1,916 per million cannot be used to conclude anything about how often trans women commit sexual offences. The numerator exceeds what its source permits, the denominator has been withdrawn from official use, and the method measures who is in prison on one day rather than who was convicted in a year. Those are three independent failures.

No other jurisdiction's data resolves this. US figures are prison headcounts with the same limitation. The Swedish cohort study followed 324 people who had surgery between 1973 and 2003 and did not report sexual offences separately from other violent crime. No jurisdiction publishes convictions per year by transgender status, which is the figure the question requires.

One finding can be stated. Correctional Service Canada found that 94% of gender diverse prisoners with sexual offence histories offended before they publicly claimed to be trans. That is 31 of 33. Two post-transition cases cannot establish whether transition raises or lowers risk.