Are Gay Men More Likely to Sexually Abuse Children?

One fact decides the shape of everything else here. No national offender dataset in the United States or the United Kingdom records the sexual orientation of people convicted of sexual offences.

In the US, the FBI's Uniform Crime Reports and the National Incident-Based Reporting System record the age, sex and race of known offenders, and ethnicity in NIBRS. There is no orientation field. NIBRS does carry a victim-offender relationship code for a homosexual relationship, and hate crime reporting carries anti-gay and anti-lesbian bias codes. Neither is an offender attribute. The first describes how two people in an incident were connected, the second describes the motive behind an offence, and both get quoted as though they answered this question.

The Bureau of Justice Statistics publishes arrest, conviction, prison and recidivism series, including a nine-year follow-up on sex offenders. Those break offenders down by sex, race, age and offence type. BJS does ask about sexual orientation in the National Inmate Survey and the National Crime Victimization Survey, and it asks in order to measure who is being victimised. Sex offender registries under SORNA require a long list of identifiers. Orientation is not among them.

Britain is the same. Police recorded crime and Ministry of Justice criminal justice statistics classify defendants by sex, age and ethnicity. The ONS has answered freedom of information requests by saying its crime figures are about victims rather than perpetrators and directing enquirers to the Ministry of Justice, which does not publish the breakdown either.

One source comes close, and it is worth looking at precisely because of how close it gets.

Those are the figures for everyone in prison. HMPPS does not cross-tabulate them by offence, so they cannot tell you whether men convicted of sexual offences against children look different from the rest of the prison population. HMPPS also states that it assumes the quality of this variable is low, since orientation is self-declared in custody and the declarations run toward heterosexual.

So the rate that both sides of this argument reach for is not held by any official system. Everything that follows is an attempt to work around that absence.

Three pieces of work carry almost the entire public argument. All three are from between 1978 and 1994, all three are small, and all three are quoted for things they do not say.

A retrospective chart review at a Denver children's hospital child abuse clinic, covering one year from July 1991. Reviewers went through the records of 352 children and noted whether the alleged adult offender was described as gay or lesbian.

The authors concluded that the children in this group were unlikely to have been molested by identifiably gay or lesbian people, which is an accurate description of what they found. Advocacy summaries round it to a population claim, and the roundings have included "98% of abuse is heterosexual" and "under 1% of molesters are gay". Neither is what a one-year caseload at one hospital can produce.

The design has two specific weaknesses, and they are not the ones usually raised. Orientation came from what caregivers had said and what staff had written down, after the allegation and probably because of it, and nobody asked the alleged offenders anything. And the caseload is dominated by cases inside the family, which is the setting where a heterosexual partner of a relative is the modal answer by construction. Two events in 269 cases will not support a precise figure whichever way they fall.

One hundred and seventy-five men convicted of sexual assault against a child, referred by Massachusetts courts to a forensic facility for people judged a substantial risk. Each was assessed for adult sexual orientation and for the sex of his victims. This is one of very few studies in either direction that asked about the offender's adult orientation instead of inferring it from the victim.

In none of the 22 bisexual men did attraction to men exceed attraction to women. The authors raised the possibility that homosexuality and homosexual pedophilia are mutually exclusive, and that the adult heterosexual male is the greater risk to a child. Paul Cameron later cited this paper as evidence of a 3:2 ratio and claimed that 54% of the molestations in it were committed by homosexual or bisexual practitioners. Gregory Herek, checking the paper, reported that the figure appears nowhere in it. The study most often marshalled for the elevated claim concluded the reverse of it.

It is still a clinical sample of men whom courts had already flagged as dangerous, orientation was assessed by interview rather than by any independent measure, and a cell containing zero people cannot be turned into a risk ratio.

Kurt Freund's laboratory at the Clarke Institute in Toronto used phallometry, which measures penile volume change while a man views standardised images of people of different ages and sexes. The 1989 study compared men who preferred adult men with men who preferred adult women, and measured how each group responded to children of their non-preferred age.

Gay men who preferred physically mature partners responded no more to male children than straight men who preferred physically mature partners responded to female children. That is a direct test of the claim, on a physiological measure rather than self-report, and it came out negative. The samples were small, partly recruited through a gay bookstore, and measured arousal in a laboratory rather than behaviour in the world. It remains the closest thing in the literature to an experiment on the question.

The argument turns on a word doing two jobs.

Sexologists treat erotic interest as having two axes that vary independently. One is the age of the preferred partner: pedophilia for prepubescent children, hebephilia for pubescent ones, teleiophilia for physically mature adults. The other is the sex of the preferred partner: androphilia for men, gynephilia for women.

A gay man, in ordinary speech, is an androphilic teleiophile. A homosexual pedophile, in the clinical literature, is a man whose interest in children happens to be directed at boys. The second term says something about the sex of the children and nothing whatever about whether the man has an adult orientation at all. Collapsing the two is the single move that generates the popular claim.

A National Research Council expert panel put this directly in 1993. The premise that male molesters of male victims are homosexual in orientation is not supported, because most men who molest boys do not report sexual interest in adult men. Specialists in the field increasingly write "male-male molestation" instead, for exactly this reason.

There is a further finding here that does real work later in the argument. Pedophilic men differentiate between boys and girls less than adult-oriented men differentiate between men and women. A man whose interest is in children is comparatively likely to offend against either sex; a man whose interest is in adults is comparatively unlikely to be attracted to both. So the proportion of offences with male victims cannot be read as a proportion of offenders who are gay, because the sorting that holds for adults does not hold for children.

The terminology has caused trouble inside psychiatry too. DSM-5 text described pedophilia as a sexual orientation; the American Psychiatric Association corrected it in November 2013, saying the wording was an error and should have read sexual interest.

The serious version of this argument is not Paul Cameron's, and it is worth clearing him out of the way first, because he is the reason most of the numbers are in circulation and he is not a source of any of them.

Cameron's 1985 review in Psychological Reports concluded that men who practised homosexual acts were at least 12 times more apt to molest a child, and around 16 times with a correction for bisexual men. He got there by treating every male offender with a male victim as gay or bisexual, then dividing by an assumed gay population share of somewhere between 1% and 4%. The first step is the inference the National Research Council rejected. Separately, he was expelled from the American Psychological Association in 1983 after failing to cooperate with an ethics investigation, the American Sociological Association passed a resolution in 1986 about misrepresentation of research in his work, and a federal judge in Baker v. Wade described his sworn statistical claims as distortions. That is relevant to weighing his numbers and it is not an argument against the underlying question.

The real case is made by researchers who accept everything in Part 3 and then point at a genuine puzzle.

Across forensic series, men who offend against male children are roughly 20 to 30% of all offenders against children. James Cantor, who is one of the clearest exponents of the pedophilia-is-not-homosexuality position, states that share plainly and accepts it. Large probability surveys put the gay share of adult men at about 2 to 4%. Those two numbers are not close to each other, and anyone who wants to dismiss this claim has to account for the gap rather than talk past it.

The victim-sex ratios that produce it are consistent across independent samples. The Kinsey Institute's 1965 institutional study found 199 offenders against female children against 96 against male children. Freund's 1984 seven-year referral series found 292 and 165. A Canadian federal prison census of men who had offended against under-12s found 66.3% with female victims only, 14.7% with male victims only, and 19.0% with both.

This is the one paper in the literature that turns the puzzle into a number. Working from roughly 465 non-psychotic sex offenders referred to the Clarke Institute, Freund and Watson started from an observed offender ratio of 1.93:1 female-victim to male-victim, adjusted it using the measured sensitivity of their phallometric test in each group to estimate the share who were true pedophiles, which moved the ratio to about 1.35:1 or 1.44:1, then applied a correction for apprehension risk. Offenders against boys report far more victims, so on this reasoning each one is caught more readily and the group is over-represented among the convicted.

The result was a ratio of roughly 11 heterosexual to 1 homosexual among true pedophiles, against a gynephile-to-androphile ratio in the general male population that they took as about 20:1. Pedophilia, on this calculation, is more common among men whose erotic development ran toward males.

Then the authors wrote the caveat into their own abstract: this "would not indicate that androphilic males have a greater propensity to offend against children". That sentence is the one that goes missing when the paper is quoted. Elsewhere in the same discussion they refer to the idea that androphiles offend more as a myth refuted by earlier work.

The calculation also has a soft joint that is worth naming, because it does most of the heavy lifting. The apprehension-risk correction rests on Abel's victim counts, which are means: about 150 victims per offender against boys against about 20 against girls. Means in this literature are dragged upward by a handful of extremely prolific men. The Australian Institute of Criminology has reported the corresponding medians as roughly 4.4 and 1.3. Using means rather than medians is what takes the ratio from about 1.4:1 to 11:1. The authors described the study as exploratory, built in an imaginary group of about 98 incest-only offenders because such cases had stopped being referred, and said the results might be temporary. Nobody has replicated or updated it in the 34 years since.

Erickson, Walbek and Seely reviewed 568 case files from a Minnesota state hospital treatment programme in 1988 and reported, in passing, that 86% of offenders against males described themselves as homosexual or bisexual. This is the only study in the literature that asked offenders to self-identify, which is why it is the strongest thing a critic has that is not Cameron's own work. It also does not say how the self-description was elicited, does not separate gay from bisexual, concerns only the male-victim subgroup of about a quarter of the sample, and was incidental to the paper's actual purpose. Men under forensic evaluation have a documented reason to prefer any label to pedophile, and "gay" is available.

The second is administrative and current. In 2024 to 2025, gay and lesbian offenders were 5% of starts on HMPPS accredited sexual offending programmes, 44 people, and bisexual offenders 4%, 42 people. Across all programme types both sat at around 1%, and among prisoners who declared an orientation the combined gay, lesbian and bisexual share was about 2.7%. So inside that treatment caseload the share is several times the prison figure. The counts are small, orientation is self-declared, a programme caseload is not a conviction count, and the same under-declaration HMPPS warns about applies to the denominator. It still points the way it points, and leaving it out would be dishonest.

Every number in this article arrived through a door, and each door admits a different population.

None of these is a sample of people who sexually abuse children. They are samples of people who were caught, referred, treated or surveyed, and most child sexual abuse never reaches the authorities at all. Nothing is known about the orientation of the people in that larger group, and nothing can be, on the data anyone currently holds.

Self-report distorts this in both directions at once, which is the part usually reported in only one. Men in evaluation or treatment are motivated to claim any self-descriptor other than pedophile, so the incentive there runs toward over-claiming a gay or bisexual identity. HMPPS says that prisoners under-report gay identity and that its figures skew heterosexual, so the incentive there runs the other way. Both effects are documented, neither is quantified, and they do not cancel to zero in any known ratio.

Third-party report has its own version. In the Denver study, the people describing an alleged offender's orientation were doing so after the allegation surfaced, and what they believed may have been reshaped by the discovery in either direction.

To make a population claim you would need a defined population, a defined outcome, the same orientation measure applied to offenders and to everyone else, and an adjustment for the fact that different offenders are caught at different rates. No study here has more than two of those four. The reason is not laziness. Pedophilia has a low base rate in the general population, which makes an adequately powered study outside forensic settings very hard to run, and the researchers in this field say so themselves rather than being accused of it.

There is a source of bias here that runs in one direction and is almost never adjusted for. For most of the twentieth century, consensual sex between men was itself a sexual offence.

In the United States, sodomy statutes were used against consensual adult conduct until Lawrence v. Texas in 2003. Lawrence made the statutes unenforceable. It did not vacate the convictions already entered or clear anyone from a registry. Idaho, Mississippi and South Carolina have continued to require people convicted before 2003 under those statutes to register as sex offenders, and the unwinding has happened case by case: a federal court ruled in 2021 on a registration requirement flowing from a 1994 Idaho crimes-against-nature conviction, and South Carolina settled with the ACLU in April 2022 to remove people listed solely for sodomy, in a suit claiming at least 18 others were affected. California's SB 145 dealt with a live version of the same thing, where registration was mandatory for oral and anal sex with mid-teen minors and discretionary for vaginal intercourse at the same ages.

In England and Wales, sex between men was partially decriminalised in 1967 with an age of consent of 21. It fell to 18 in 1994 and reached 16, the heterosexual age, only on 8 January 2001. For 34 years, conduct that was legal for a mixed-sex couple produced convictions for a same-sex one. Convictions for buggery and gross indecency continued throughout, and the published tables do not separate adult-adult from adult-minor cases.

Lord Sharkey estimated around 65,000 men convicted under the repealed laws, of whom about 15,000 were alive in 2016. The disregard scheme of 2012 and the pardons under the Policing and Crime Act 2017, extended in 2023, allow those convictions to be set aside. The pardon is automatic for the dead and opt-in for the living, so a man who has not applied still has the conviction, and no scheme rewrites a table published in 1990.

The effect on any figure drawn from a long conviction series or a registry is predictable in direction and unknown in size. It inflates the same-sex side. Nobody has published a national estimate of how much. The only relevant number is from the Williams Institute's 2022 survey of 964 registrants, self-selected and online, in which 9% of LGBTQ respondents reported a sodomy charge against 3% of everyone else. That is a survey, not an audit of a registry, and it should not be used as one.

Some of what is missing here is missing in a way that defeats the claim. Some of it is missing in a way that defeats the answer to the claim, which is the part that tends to go unsaid.

Nobody knows the orientation distribution of people who abuse children and are never caught, and since detection is not uniform, nobody can rule out that it differs from the convicted population. Nobody has audited a registry to establish how many entries rest on conduct that is legal today. And the question has never been asked about women at all. Groth noted in the 1970s that there was practically no reportage of sexual molestation of girls by lesbian adults, and that is still roughly where it stands: a 2019 review by Tozdan, Briken and Dekker describes female-perpetrated child sexual abuse as understudied and surrounded by a culture of denial. Half of the question has not been put.

Then the part that cuts against the answer people want. The American Psychological Association says there is no scientific evidence that gay men or lesbians are more likely than heterosexual people to molest children, and the American Academy of Pediatrics said in a 2011 clinical report that homosexual adults are no more likely to sexually abuse children than heterosexual adults, citing the Denver study. Both statements are accurate about the state of the evidence. They are statements about an absence, and they are routinely quoted as though a measured equality had been found. No such measurement exists. The honest position is that the claim is unsupported, not that the opposite has been established.

What would settle it is easy to specify and nobody has done it. Add an orientation field to conviction records, ask the same question the same way in a population survey, and correct for the difference in how readily different offenders are caught. Two of those three are administrative decisions.

Until then the argument runs on a 1994 chart review with two cases in it, a 1978 sample of 175 men with a zero in one cell, and a 1992 calculation whose own authors said it might be temporary and added a sentence that everyone quoting it leaves out.