Woke Mind Virus Bank
Woke Mind Virus Bank
Evidence for a page that does not exist yet.
This bank produces evidence. It does not draft prose. No sample sentences, no candidate openings, no page voice.
It answers three questions and nothing else. Commentary, opinion pieces and culture-war argument are out. Every entry carries a number, a population, a year, a source name, and a link a stranger can open.
Q1. Who has measured young men by political ideology on a mental-health outcome, and what did they get? Q2. Which way does the causation run? Q3. What happened to natal males who medically transitioned young?
Verdicts, with the pipeline codes from 02 - System/Research Pipeline for AI Agents.md §2 in brackets:
| Verdict | Meaning | Code |
|---|---|---|
| shown | A citation a stranger can open carries the claim as stated | S |
| partly shown | Part of the claim is carried; part is not, or it holds only in a narrower form | P/C |
| asserted | No evidence offered and none found | P |
| contradicted | The evidence goes the other way | C |
| unverifiable | Not the kind of claim evidence can reach | — |
Two entries (E4, E5) are tabulations run for this bank from public survey files rather than published findings. Both are marked, both name the file and the variable, and both were validated against published aggregates from the same data before being used. They are descriptions of what the data says. They are not published results and a page must not cite them as if they were.
X rows from 01 - Workbench/woke-mind-virus-2026-08-24/grok-x-rows.md appear only where the row carries one of the three answers with a number. The citation is the underlying survey or paper; the X link is the pointer.
Q1. Who has measured young men by ideology?
The owner’s missing cell. It exists in five datasets and is published as a plain table in one of them.
E1. Gimbrone, Bates, Prins & Keyes (2022), “The politics of depression”, SSM–Mental Health 2:100043. The anchor, and the only published table with liberal boys as its own row. Verdict: shown. Monitoring the Future, nationally representative annual cross-sections of US 12th-graders, 2005–2018, N = 86,138. Depressive affect is a four-item self-report scored 1 to 5, alpha 0.74. Among boys, 21.1% called themselves conservative and 17.5% liberal.
Pooled means (SD), Appendix Table A.3:
| Cell | Depressive affect | Self-esteem | Self-derogation | Loneliness |
|---|---|---|---|---|
| Male conservative | 1.89 (0.82) | 4.15 (0.87) | 1.93 (0.98) | 2.53 (1.08) |
| Male moderate | 2.00 (0.86) | 4.11 (0.81) | 2.03 (0.98) | 2.65 (1.07) |
| Male liberal | 2.06 (0.88) | 4.06 (0.87) | 2.12 (1.01) | 2.70 (1.08) |
| Male radical | 2.50 (1.11) | 3.88 (1.05) | 2.40 (1.17) | 2.75 (1.07) |
| Female conservative | 1.81 (0.83) | 4.14 (0.86) | 1.93 (0.99) | 2.67 (1.12) |
| Female liberal | 2.19 (0.92) | 3.93 (0.87) | 2.25 (1.05) | 2.98 (1.10) |
Liberal boys score worse than conservative boys by 0.17 points and worse than conservative girls by 0.25. Liberal girls score worse than conservative girls by 0.38, more than twice the boys’ gap. Adjusted for region, urbanicity, race and GPA, with male and conservative as the reference categories, the liberal-versus-conservative coefficient among boys is b = 0.22 (95% CI 0.12 to 0.32), β = 0.10, E-value 1.85. The much-quoted three-way interaction is borderline: female × liberal × 2014–2018 b = 0.17 (95% CI 0.01 to 0.32), β = 0.04. The plain female × liberal term is b = 0.05 and not significant, and the female main effect is −0.01.
Boys who called themselves radical are the worst male cell at 2.50, worse than radical girls at 2.22 and worse than any liberal cell of either sex. N = 1,040 male radicals, 2.5% of boys. https://doi.org/10.1016/j.ssmmh.2021.100043 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8713953/
E2. What Gimbrone does not print, and what gets quoted from it. Verdict: partly shown. The circulating sentence — “liberal boys were significantly more likely to report depression than conservatives of either gender” — is Musa al-Gharbi’s, from his 2023 essay, not the paper’s (grok row R08, https://x.com/McCormickProf/status/2041826291574251926). The essay’s wording was confirmed verbatim at https://americanaffairsjournal.org/2023/03/how-to-understand-the-well-being-gap-between-liberals-and-conservatives/.
The paper’s own means carry the ordering — 2.06 for liberal boys against 1.89 and 1.81 for conservative boys and girls. The paper never tests that contrast for significance and never prints a trend figure for liberal boys. The only trend numbers in the text are liberal girls, 1.92 (SD 0.89) in 2010 to 2.65 (SD 0.88) in 2018, a rise of 0.73; and conservative boys, 1.75 (SD 0.72) to 2.17 (SD 0.87), a rise of 0.42, the smallest of the four groups. The liberal-boy time series exists only as a line on Figure 1.
One sub-finding does isolate the boys and a page can use it. Among White students, liberal boys without a college-educated parent were the one group whose scores converged with liberal girls’.
E3. Pew’s American Trends Panel Wave 64, as cut by Zach Goldberg and re-graphed by Jonathan Haidt. Diagnosis, not symptoms. Verdict: shown, and the write-up is journalism rather than a paper. Pew ATP Wave 64, fielded 19–24 March 2020, N = 11,537 US adults. The item asks whether a doctor or other healthcare provider has ever told the respondent they have a mental health condition.
Whites aged 18–29, as published in the Washington Free Beacon, 19 April 2021: liberal men 34%, moderate men 22%, conservative men 16%; liberal women 56%, moderate women 28%, conservative women 27%. All young White liberals 46%, moderates 26%, conservatives 21%. Goldberg’s decimals of 33.6% and 56.3% exist only in his X thread (grok row R03, https://x.com/ZachG932/status/1249764370458062850); he published this in no journal, no think-tank paper and no newsletter.
Haidt re-graphed the same wave for all races. Read off his chart, ages 18–29: liberal men about 30%, moderate men about 18%, conservative men about 13%; liberal women about 51%, conservative women about 21%.
Two things a page must carry. This is a reported diagnosis, so it moves with help-seeking as well as with illness, and the same analyst posted in the same thread that he could not rule out liberals simply seeking evaluations more often (row R04, https://x.com/ZachG932/status/1248826478432858112). https://freebeacon.com/politics/white-libs-mental-health/ · https://www.afterbabel.com/p/mental-health-liberal-girls · dataset https://www.pewresearch.org/social-trends/dataset/covid-19-late-march-2020/
E4. Cooperative Election Study 2022. The young-male cell, computed for this bank because nobody published it.
Verdict: shown — a tabulation run here, validated against published aggregates from the same file.
CES 2022 Common Content, N = 60,000, fielded 29 September to 8 November 2022; 59,710 answered item CC22_309f, “Would you say that in general your mental health is…”, with Excellent, Very good, Good, Fair, Poor. Weighted with commonweight, scored 0 to 100 the way Nate Silver scored it (Excellent 100, Poor 0). Public file, no login: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/PR4L8P
Validation first. Recomputing Silver’s published aggregates from the raw file returns liberals 52.5, moderates 58.0, conservatives 67.5 against his published 53, 58, 68; and all-age men 55.7 / 60.6 / 69.1 against his published 56 / 61 / 69. The pipeline reproduces his numbers, so the cells below come from the same method he used.
The cross he never published. Ages 18–29, mean on 0–100, then percent Excellent, percent Poor, percent Fair-or-Poor, unweighted n:
| Cell | Mean | Excellent | Poor | Fair+Poor | n |
|---|---|---|---|---|---|
| Men liberal | 46.9 | 10.9% | 14.4% | 40.8% | 1,348 |
| Men moderate | 54.4 | 17.7% | 10.8% | 31.8% | 1,333 |
| Men conservative | 59.3 | 24.6% | 8.4% | 28.5% | 855 |
| Women liberal | 39.5 | 6.5% | 17.3% | 51.7% | 1,950 |
| Women moderate | 42.2 | 6.8% | 14.3% | 47.7% | 1,518 |
| Women conservative | 54.5 | 16.6% | 7.6% | 29.5% | 811 |
The gap among young men is 12.4 points, about two thirds the size of the 15.0-point gap among young women. On the five-point scale, men 18–29 run very liberal 45.3, liberal 48.5, middle of the road 54.4, conservative 55.9, very conservative 65.1 — monotonic across the whole range. Among young men who seldom or never attend church, the gap is 39.6 against 53.6, so it is not only religion.
Silver’s own post is https://www.natesilver.net/p/what-explains-the-liberal-conservative (18 June 2025). He published sex and generation separately and never crossed them with ideology. The figures circulating from him as “conservative 51% excellent, liberal 20%” are composition shares, not rates: of people reporting excellent mental health, 51% are conservative and 20% liberal.
E5. General Social Survey. Three measures on young men, computed for this bank, and one of them is a null.
Verdict: partly shown — small cells, and the measures disagree.
GSS cumulative file 1972–2024, public download at https://gss.norc.org/us/en/gss/get-the-data.html. Weighted with wtssps. Liberal is polviews 1–3, conservative is 5–7. Every cell below runs n = 74 to 176, so all of it is noisy.
| Measure | Years | Liberal men 18–29 | Conservative men 18–29 |
|---|---|---|---|
| Ever told by a doctor you had depression | 2014, 2018, 2022 | 25.7% (n=134) | 16.9% (n=97) |
| Self-rated mental health, 1 best to 5 worst | 2018, 2021 | 2.92, 33.9% fair or poor (n=129) | 2.50, 15.8% fair or poor (n=74) |
| Days of poor mental health in the past 30 | 2016–2022 | 4.07 days (n=135) | 4.13 days (n=97) |
| Not too happy | 2022–2024 | 29.2% (n=145) | 21.9% (n=125) |
The third row is the strongest counter-cell in this bank. On the one measure that asks for a count of days rather than a rating or a label, young men show no ideology gap at all — 4.07 against 4.13. The same measure shows a large gap among young women, 8.31 days against 5.56. The gap among young men appears on ratings and diagnoses and disappears on the day count.
The happiness gap among young men is also recent. It was −0.1 points in 2010–2016 and 1.9 points in 1972–1989.
E6. Twenge (2024), “Who’s more pessimistic, young liberals or young conservatives?”, Generation Tech. Gimbrone’s data carried to 2022. Verdict: partly shown — the numbers are read off a chart. Monitoring the Future 12th-graders, 1989–2022, six-item depressive-symptoms scale, “high” meaning a mean of 3 or more on five-point items. Percent high in depression in 2021–22, read from Figure 1: liberal young women about 46%, liberal young men about 36%, conservative young men about 21%, conservative young women about 19%. The late-2000s baseline was about 14, 15, 11 and 8. Her own text says liberal young men are “almost twice as likely to be depressed as conservative young men.”
Two things worth noting. This is the only extension of the Gimbrone analysis past 2018 anywhere, and it is a blog post rather than a paper. And at the 1989–94 start of the series liberal boys were the highest of the four groups, so the ordering is not a fixed fact about ideology. https://www.generationtechblog.com/p/whos-more-pessimistic-young-liberals
E7. Lahtinen (2024), “Construction and validation of a scale for assessing critical social justice attitudes”, Scandinavian Journal of Psychology. The only source that measures the woke route itself and has a male cell. Verdict: shown. Two Finnish studies, combined n = 5,878. Study 2 is a nationwide sample of 5,030 people aged 15–84 recruited in October 2022 through a link in Helsingin Sanomat — self-selected, not a probability sample. 2,634 men (52.4%), 2,112 women (42.0%), 120 who answered “other” (2.4%).
On the seven-item Critical Social Justice Attitude Scale, scored 0 to 4: everyone 1.55 [1.52, 1.58]; women 2.13 [2.09, 2.17]; men 1.03 [1.00, 1.06]; Cohen’s d = 1.20. Self-reported wokeness: women 2.11, men 1.11, d = 0.76. Men rejected every item on the scale.
Scale items correlate with anxiety at r = .17 to .28 and with depression at r = .13 to .24. Splitting at the scale midpoint, high scorers (n = 1,785) against low scorers (n = 3,239): anxiety t(3038.9) = −17.00, depression t(3127.0) = −13.72, happiness t(3506.1) = 6.44. The author states the mental-health results held for both sexes separately. In Study 1 (n = 848) the student correlations were r = .39 anxiety, .27 depression, −.17 happiness. https://doi.org/10.1111/sjop.13018 · open preprint with the full tables https://doi.org/10.31234/osf.io/qd4c6
E8. Kaufmann (2024), Centre for Heterodox Social Science Report No. 1, on the FIRE student survey. The magnitude comparison. Verdict: partly shown — the male cell exists only inside models. FIRE’s 2024 College Free Speech Rankings microdata, about 55,100 US undergraduates aged 18–22, fielded January to June 2023. The measure is frequency of feeling anxious or depressed, not a validated instrument. 68% report anxiety at least half the time and 42% depression.
Standardized predictors of a five-item mental-illness factor, N = 52,007: female .19, non-heterosexual .13, higher social class .12, liberal ideology .06. In his worked profile, moving a student from very liberal to moderate or conservative drops chronic anxiety about 9 points; moving the same student from female to male drops a further 16.
Being female predicts poor mental health about three times as strongly as being liberal does. That is the size comparison the page needs, and it comes from the largest student sample in this bank. No plain four-cell table by sex and ideology is published anywhere; the microdata is available by request to data@thefire.org. The same report gives the only non-US measurements found: a YouGov UK sample of 1,552 people aged 18–20, ideology effect .08 standardized, and a US Deltapoll sample of 1,502, effect .13. https://erickaufmann.substack.com/p/the-mental-health-crisis-does-not
Q2. Which way does the causation run?
Three candidate directions. (A) the ideology makes people sick. (B) sick people, or people with the traits that lead to sickness, adopt the ideology. (C) liberals only report more. Only sources that test or measure direction are listed. Correlations without a directional test are not here, however large.
E9. Schaffner, Hershewe, Kava & Strell (2025), “Do conservatives really have better mental well-being than liberals?”, PLoS ONE 20(4):e0321573. The strongest evidence for (C), and it is an experiment. Verdict: shown for (C), on the self-rated-mental-health measure. Study 1: CES 2022, 60,000 US adults, 59,710 answering the item. Controlling for more than two dozen demographic, socioeconomic and recent-life-experience variables cuts the conservative-liberal gap by about 40% and does not remove it. The full-scale gap is 19 points raw and 11 points adjusted.
Study 2: a randomised experiment on a 1,000-person module of the 2023 CES, 500 per arm. One arm rated “your mental health”, the other “your overall mood”. Same scale, same everything else.
- Asked about mental health: conservatives 11 points more likely to answer excellent or very good (p = 0.044) and 12 points less likely to answer fair or poor (p = 0.004).
- Asked about overall mood: the gap disappears. Conservatives slightly less likely to give a top rating (p = 0.225), fair-or-poor gap near zero (p = 0.888).
- The movement is almost all on the conservative side. 64% of conservatives rated their mental health highly against 49% for their mood. Liberals barely moved at the top and fell from 26% to 17% at the negative end (p = 0.049).
A word swap erases the gap. The authors offer two readings and pick neither: conservatives inflate ratings of a stigmatised term, or mental health and mood are different things that ideology predicts differently. Either way, E4 and E5’s rating-based cells inherit the problem, and E1’s symptom scale and E3’s diagnosis item do not. https://doi.org/10.1371/journal.pone.0321573 · https://pmc.ncbi.nlm.nih.gov/articles/PMC12043138/
E10. Goldberg’s stigma-index adjustment, the argument E9 later overturned. Verdict: contradicted by E9. Ipsos Understanding Society Wave 22, fielded 1–4 October 2021 on the KnowledgePanel, N = 1,025 US adults; the stigma models run on N = 960. Goldberg adjusted the mental-health self-rating for a four-item mental-health stigmatisation index. The Democrat-Republican gap goes 0.38 raw, 0.32 with demographics, 0.29 with the stigma index added — the index moves it by 0.03. Republicans score 0.54 SD higher than Democrats on the index itself. He concluded the reporting-style objection was “a dead-end” (grok row R06, https://x.com/ZachG932/status/1679977654022475783, follow-up https://x.com/ZachG932/status/1797734915985027125). None of it is published outside X.
The result stands and the conclusion does not. Controlling for a measured stigma attitude is a weaker test than randomising the word, and the randomised test in E9 wiped the gap out. A page that uses R06 has to note that a stronger design ran two years later and went the other way.
E11. Gimbrone’s own religiosity check, which is evidence for (B), inside the anchor paper. Verdict: shown. Same paper as E1. Adding religiosity to the model halves the liberal-versus-conservative coefficient among boys: b = 0.22 (95% CI 0.12, 0.32) becomes b = 0.11 (95% CI 0.00, 0.22), and β falls from 0.10 to 0.05 — no longer distinguishable from zero. The authors write that after this adjustment the trends held for every group “with the exception of male liberals whose predicted symptom scores moved closer to those of conservatives.”
Half of the boys’ ideology gap is religion, and the girls’ pattern survives the same adjustment. That is one measured confound doing half the work in the paper the whole claim rests on. The E-value of 1.85 says an unmeasured confound would need a risk ratio of 1.85 with both ideology and depression to erase the rest. https://pmc.ncbi.nlm.nih.gov/articles/PMC8713953/
E12. Lahtinen (2024) on whether the woke route adds anything beyond generic leftism. Verdict: contradicted, for the specific route the owner suspects. Same study as E7. The critical social justice scale correlates with anxiety and depression. So does simply placing yourself on the political left, and the author reports it does so at the same level or slightly higher: “being on the political left had similar or slightly higher negative mental health correlations than CSJAS and the global CSJA item.” Being liberal on the liberal-conservative axis correlated less than either.
Measuring woke attitudes directly buys nothing over asking someone whether they are left-wing. If woke ideology were the mechanism, the specific measure should beat the crude one. It does not. https://doi.org/10.1111/sjop.13018 · https://doi.org/10.31234/osf.io/qd4c6
E13. Bernardi, Bridger, Board, Yamamori, Gross & Roiser (2025), “Stressful Politics?” — the only two-way longitudinal test found, and both arrows fire. Verdict: partly shown, both directions. Preprint, 24 October 2025, not peer reviewed. Three panels: British adults via YouGov (N = 1,700 then 1,366, June and August 2023); British youth aged 18–27 via Prolific (N = 1,462 July 2024, N = 978 February/March 2025); American youth aged 18–27 (N = 249 before the 2024 election, N = 141 March 2025). The exposure is finding politics stressful, not ideology.
Cross-sectionally, a one-SD rise in political stress goes with higher depressive symptoms: 0.27 SD in British adults, 0.20 in British youth, 0.25 in American youth, all p < 0.001.
Cross-lagged, in the young British panel:
- depression at wave 1 predicts political stress at wave 2, η² ≈ 0.01;
- political stress at wave 1 predicts depression at wave 2, η² = 0.005, p = 0.033. In the adult panel the second arrow is absent (η² < 0.001, p = 0.907). A separate model finds general stress at wave 1 predicting political stress at wave 2 but not the reverse.
Both arrows run in young people, and the sick-people-turn-to-politics arrow is the larger one. Two waves, non-probability samples, and the American youth follow-up is 141 people. https://doi.org/10.31234/osf.io/hqe3g_v1 · https://osf.io/hqe3g/
E14. Rigoli (2025) and De Neve (2013). Temporal precedence for (B), and a direct test of whether it is a male story. Verdict: partly shown. Rigoli, “Neuroticism Is Linked With Liberal Ideology in Young, but not Old, People in the United States”, International Social Science Journal, 28 October 2025. Three studies: GSS 2022 (n = 1,644), a Prolific sample (n = 600), and a non-US sample. Neuroticism predicts liberal ideology at age 29 with an effect of −0.269, and the effect is gone by age 57. Study 3 found no such effect outside the United States, which the author reads as generational rather than developmental. The three-way test found the neuroticism-ideology link does not differ by gender. Title, journal, date and the three-study design confirmed at Crossref; the coefficients were read from the paper by the research lane and the article itself is paywalled. https://doi.org/10.1111/issj.70025
De Neve (2013), “Personality, Childhood Experience, and Political Ideology”, Political Psychology 34(6). N = 14,672, representative US sample with family clusters allowing sibling fixed effects. Openness predicts liberal ideology and conscientiousness predicts conservative ideology, robust to sibling fixed effects. Childhood trauma interacts with openness in predicting adult ideology. This is the study behind al-Gharbi’s line that people who experienced abuse and trauma as children were more likely to identify as liberal as adults. Closed access; the coefficients could not be read. https://doi.org/10.1111/pops.12075
Al-Gharbi’s own verdict on the whole question, in the essay the grok rows point at: which direction predominates is “empirically unclear” and “hotly contested.”
Q3. Natal males who transitioned young
Six entries, the cap, one per outcome plus the Cass Review and the harm-side paper. Natal male means registered male at birth. Where a study reports natal males separately the natal-male figure is given; where it does not, that is said.
E15. The Cass Review final report, April 2024. What it concluded about the male-typical cases. Verdict: shown. The treatment was designed for boys and is now given to someone else. §14.34: “Because an intervention intended for one group of young people (predominantly pre-pubertal birth-registered males) has been given to a different group, it is hard to know what percentage of these young people might have resolved their gender-related distress in a variety of other ways.” The GIDS audit inside the report puts the current cohort at 73% natal female and 27% natal male at referral; 34.6% of natal males were referred to endocrinology against 24.2% of natal females, so natal males were 34% of the endocrine caseload. Referrals flipped inside seven years: adolescent males outnumbered females 24 to 15 in 2009, and by 2016 adolescent females outnumbered males 1,071 to 426.
Most of the boys desisted. §2.6: early studies found persistence in “approximately 15%” of pre-pubertal children and “the majority of these children became same-sex attracted, cisgender adults”; later stricter studies found 10–33%. Table 8, from Steensma et al. 2013b: of 127 children followed, among birth-registered males 23 persisted and 56 desisted, a 71% desistance rate; among birth-registered females it was 24 and 24. Childhood social transition predicted persistence for boys and not girls: 96% of the boys who desisted had not socially transitioned at referral. The largest boys-only follow-up is Singh, Bradley & Zucker 2021 (Frontiers in Psychiatry 12:632784): 139 clinic-referred boys, 17 persisted (12.2%), 122 desisted (87.8%); persistence did not differ between boys who met full childhood criteria (13.6%) and subthreshold boys (9.8%); 63.6% were androphilic or bisexual in fantasy at follow-up. Pooled across all prior boys’ follow-up studies, 17.4% of 235 boys persisted. https://pmc.ncbi.nlm.nih.gov/articles/PMC8039393/
The boys’ route is entangled with homosexuality and autism. §8.28: in the original Dutch cohort “89% of the 70 patients were same-sex attracted to their birth-registered sex… Only one patient was heterosexual.” §8.30, GIDS natal males with orientation recorded: 42% attracted to males, 39% bisexual, 19% attracted to females. Pooled ASD prevalence in referred youth is 9%, and trans and gender-diverse people are “three to six times more likely to be autistic.”
The blocker indication for boys is narrow and the rest is unproven. §14.58: “there seems to be a very narrow indication for the use of puberty blockers in birth-registered males as the start of a medical transition pathway in order to stop irreversible pubertal changes. Other indications remain unproven at this time.” §14.41: blocking too early in birth-registered males can leave inadequate penile growth and force intestinal vaginoplasty, which carries higher risk.
Two summary conclusions. Para 86: “It has been suggested that hormone treatment reduces the elevated risk of death by suicide in this population, but the evidence found did not support this conclusion.” Para 87: “The percentage of people treated with hormones who subsequently detransition remains unknown due to the lack of long-term follow-up studies, although there is suggestion that numbers are increasing.” §14.25 adds that 98% of those who start blockers proceed to hormones, so blockers “are not buying time to think.”
Evidence quality, from the York systematic reviews commissioned for the Review. Puberty suppression: 50 studies, 1 high quality, 25 moderate, 24 low (https://adc.bmj.com/content/109/Suppl_2/s33). Masculinising and feminising hormones: 53 studies, 1 high quality, 33 moderate, 19 low (https://adc.bmj.com/content/109/Suppl_2/s48). Two of 103 endocrine studies rated high quality. Care pathways, 23 studies covering 6,133 children: 36% received puberty suppression (95% CI 27–45), 51% hormones (40–62), 68% either (57–77), 16% surgery (10–24) (https://doi.org/10.1136/archdischild-2023-326760). Report (official site is behind a bot check; this mirror is byte-identical): https://web.archive.org/web/20250312023258/https://cass.independent-review.uk/wp-content/uploads/2024/04/CassReview_Final.pdf
E16. Suicide. Both sides, and the study that settles which reading survives. Verdict: partly shown for the elevated-risk claim, contradicted for the causal reading.
The natal-male-specific figure. Erlangsen et al. (2023), “Transgender Identity and Suicide Attempts and Mortality in Denmark”, JAMA 329(24):2145–2153. All of Denmark 1980–2021; 3,759 transgender people, 1,975 natal male. All 12 suicide deaths in the cohort were natal males, adjusted IRR 4.5 (95% CI 2.6 to 8.0) against non-transgender men. Zero suicide deaths among natal females. Suicide attempts: natal male aIRR 8.5, natal female 6.8. No treated-versus-untreated comparison and no psychiatric adjustment in the main analysis. https://pmc.ncbi.nlm.nih.gov/articles/PMC10300682/
The figure that circulates, and its author’s objection. Dhejne et al. (2011), PLoS ONE 6(2):e16885. 324 people with surgery and legal sex change 1973–2003, 191 natal male; controls matched 10 to 1 from the general population. Suicide-death adjusted HR 19.1 (95% CI 5.8 to 62.9) — built on 10 suicide deaths against 5 in controls, which is why the interval spans an order of magnitude. Split by cohort, the significance holds for 1973–1988 and disappears for 1989–2003. The paper’s own words: “no inferences can be drawn as to the effectiveness of sex reassignment as a treatment for transsexualism… the results should not be interpreted such as sex reassignment per se increases morbidity and mortality.” Dhejne in a 2015 interview: “It’s very frustrating! I’ve even seen professors use my work to support ridiculous claims.” https://doi.org/10.1371/journal.pone.0016885 · https://www.transadvocate.com/fact-check-study-shows-transition-makes-trans-people-suicidal_n_15483.htm
The study that changes the reading. Ruuska, Tuisku, Holttinen & Kaltiala (2024), BMJ Mental Health 27:e300940. Every Finnish gender-referred person under 23 between 1996 and 2019, n = 2,083 (41.3% natal male), against 16,643 matched controls, mean follow-up 6.5 years. 55 deaths, 20 of them suicides. Suicide 0.3% against 0.1%; rate 0.51 per 1,000 person-years against 0.12; unadjusted suicide HR 4.3 (1.7 to 10.7). Adjusted for specialist psychiatric treatment history the difference vanishes: HR 1.8 (0.6 to 4.8) for suicide, 1.0 (0.5 to 2.0) for all-cause. The authors: “Clinical gender dysphoria does not appear to be predictive of all-cause nor suicide mortality when psychiatric treatment history is accounted for.” The Finnish register authority barred birth-sex stratification, so there is no natal-male figure here. https://pmc.ncbi.nlm.nih.gov/articles/PMC10875569/
The UK claim that a blocker restriction caused a suicide spike. Appleby review for the Department of Health and Social Care, 19 July 2024: 5 suicides in the three years before the Bell ruling and 7 after, “essentially no difference… would not reach statistical significance.” https://www.gov.uk/government/publications/review-of-suicides-and-gender-dysphoria-at-the-tavistock-and-portman-nhs-foundation-trust/review-of-suicides-and-gender-dysphoria-at-the-tavistock-and-portman-nhs-foundation-trust-independent-report
E17. Chen et al. (2023), “Psychosocial Functioning in Transgender Youth after 2 Years of Hormones”, NEJM 388:240–250. The strongest benefit study, and its natal-male result is the exception. Verdict: contradicted, for natal males specifically. Prospective four-site US cohort, no control group. 315 youths, mean age 16, 111 (35.2%) designated male at birth, 83.2% starting hormones at Tanner 5. Retention 81% of possible observations.
Overall annual change: appearance congruence +0.48 on a five-point scale, depression −1.27 on the 63-point BDI, anxiety −1.46 T-points, life satisfaction +2.32 T-points.
The paper’s own sentence: “Depression and anxiety symptoms decreased significantly, and life satisfaction increased significantly, among youth designated female at birth but not among those designated male at birth.” The interaction terms for natal males relative to natal females: depression +1.91 per year (0.33 to 3.50), anxiety +1.56 per year (0.01 to 3.10), life satisfaction −1.86 per year (−3.49 to −0.24). Appearance congruence improved in both.
On the flagship prospective study, the mental-health benefit is a natal-female result. Natal males got the appearance change without the mood change. Two participants died by suicide during the study, recorded in Table 2 along with 11 cases of suicidal ideation. The authors state the study “lacked a comparison group.” https://pmc.ncbi.nlm.nih.gov/articles/PMC10081536/
E18. Regret. The low figure, its natal-male denominator, and the arithmetic error in the meta-analysis. Verdict: partly shown — the rate is low and the measurement cannot see late regret. Wiepjes et al. (2018), “The Amsterdam Cohort of Gender Dysphoria Study (1972–2015)”, Journal of Sexual Medicine 15(4):582–590. One clinic treating over 95% of the Dutch transgender population. 6,793 people, 4,432 natal male. Regret in 0.6% of trans women who had gonadectomy: 11 of 1,742. Trans men 0.3%, 3 of 885. The definition matters as much as the number. A trans woman counted as regretful only if she started testosterone after vaginoplasty and expressed regret. Of the 11, five had social regret, five true regret, one turned out non-binary. 36% of hormone starters never returned to the clinic. Median follow-up was 6.4 years while mean time to regret among the regret cases was 10.8 years, so the window is shorter than the thing being measured. The authors say the figure “could be an underestimation.” https://doi.org/10.1016/j.jsxm.2018.01.016 · free copy https://pure.amsterdamumc.nl/ws/files/158086270/The-amsterdam-cohort-of-gender-dysphoria-study-1972-2015-trends-in-prevalence-treatment-and-regrets.pdf
Bustos et al. (2021), PRS Global Open 9(3):e3477, the source of the famous 1%: pooled regret 1% (95% CI <1% to 2%), transfeminine surgeries 1%, vaginoplasty specifically 2%. https://pmc.ncbi.nlm.nih.gov/articles/PMC8099405/ Its published erratum matters and the abstract still carries the error. A letter (Expósito-Campos & D’Angelo, PRS Global Open 2021;9(11):e3951) showed Bustos had inflated the Wiepjes surgical sample from 2,627 to 4,863. The erratum (2022;10(4):e4340) corrected the table, which brings the real total to about 5,672, not the 7,928 the abstract still states. 23 of the 27 included studies were moderate-to-high risk of bias. https://pmc.ncbi.nlm.nih.gov/articles/PMC8751779/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC9049036/
A natal-male asymmetry runs through both. The per-person rate is low and the case count skews male: 57 of Bustos’s 77 regret cases were transfeminine, and 11 of Wiepjes’s 14. The mechanism by which clinic counts miss cases: in Littman’s 2021 survey of 100 detransitioners, only 24% ever told their clinician they had detransitioned. https://pmc.ncbi.nlm.nih.gov/articles/PMC8604821/
E19. Detransition and discontinuation. Three national datasets, and they disagree about natal males. Verdict: partly shown — every figure is real and they do not reconcile.
| Source | Population | Natal-male discontinuation | Natal-female |
|---|---|---|---|
| van der Loos et al. 2022, Lancet Child Adolesc Health 6:869 | 720 Dutch people who started blockers in adolescence then hormones, 220 natal male, national prescription registry | 4% (9 of 220) | 1% |
| Roberts et al. 2022, J Clin Endocrinol Metab 107:e3937 | 952 US Military Health System patients, 325 transfeminine, mean age 19.2 | 19% at 4 years | 35.6% |
| Kaltiala et al. 2024, BMC Psychiatry 24:566 | All of Finland, 1,359 people, mean follow-up 8.5 years | 10.5% (49 of 467) | 6.5% (58 of 892) |
Roberts and Kaltiala point in opposite directions on sex. In the US military data natal females discontinued nearly twice as often as natal males; in the Finnish register natal males discontinued more, p = 0.004. The Dutch figure of 98% overall continuation is the lowest discontinuation number anywhere and comes from the strictest screening gate. Kaltiala also reports the discontinuation hazard 2.7 times higher in the 2013–2019 intake than the 1996–2005 intake. Definitions differ and that is most of the disagreement: van der Loos counts an absent prescription, Roberts a 90-day gap in one insurance system, Kaltiala a 12-month reimbursement gap. None can say why anyone stopped. https://doi.org/10.1016/S2352-4642(22)00254-1 · https://pubmed.ncbi.nlm.nih.gov/35452119/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC11334601/
The systematic review of the whole literature, Feigerlova (2025), Journal of Sexual Medicine, 15 studies covering 3,804 children and adolescents and 3,270 adults: change of request before starting blockers 0.8% to 7.4%, blocker discontinuation 1% to 7.6%, hormone discontinuation 1.6% to 9.8%, with no natal-male breakout. The author’s own assessment of all 15: heterogeneous definitions, “their numbers were too small to be statistically relevant, their time frame was insufficient, they did not use patient-level data, or they did not consider confounding factors.” https://doi.org/10.1093/jsxmed/qdae186
Who detransitioners are. Littman, O’Malley, Kerschner & Bailey (2024), Archives of Sexual Behavior 53: 78 US adults aged 18–33 who had stopped identifying as transgender at least six months earlier. 71 of 78 (91%) were natal female; seven were natal male. First identified as trans at a mean age of 17.1 and held it 5.4 years. Fewer than 17% met DSM-5 criteria for childhood gender dysphoria by retrospective report; 53% said rapid-onset gender dysphoria applied. 43% of the natal females were exclusively homosexual against 0% of the natal males. Recruited through detransition networks, so the sex ratio describes who volunteers, not a population rate. https://pmc.ncbi.nlm.nih.gov/articles/PMC10794437/
E20. Lewis et al. (2025), “Examining gender-specific mental health risks after gender-affirming surgery”, Journal of Sexual Medicine 22(4):645–651. The paper behind the circulating “107,000 patients” claim. Verdict: partly shown — the numbers are real and the design cannot carry the causal reading put on it. This is the study Genspect posted without a citation (grok row R16, https://x.com/genspect/status/2091244174247137542). TriNetX US electronic health records, adults with an ICD-10 F64 gender dysphoria code, June 2014 to June 2024. 107,583 patients, propensity-matched cohorts with and without surgery, outcomes over two years after surgery.
- Males with surgery against males without: depression 25.4% vs 11.5%, RR 2.203, p < 0.0001; anxiety 12.8% vs 2.6%, RR 4.882, p < 0.0001.
- Females: depression 22.9% vs 14.6%, RR 1.563; anxiety 10.5% vs 7.1%, RR 1.478.
- Feminising individuals — natal males — depression RR 1.783 (p = 0.0298), substance use disorder RR 1.284 (p < 0.0001).
Not retracted and not corrected. No erratum or expression of concern exists on PubMed or Crossref. What exists is three letters and an author reply (J Sex Med 2025;22(9):1708–1715). One of them, from Jorgensen, states that the study period begins before US ICD-10 coding started and that cohorts E and F “could be comprised of cisgender patients who underwent surgery for other indications (eg, a cisgender female who received a mastectomy for breast cancer).”
Three things a page cannot skip. The matching controlled for age, race and ethnicity, not for mental health before surgery, so the people who reached surgery may have been sicker on the way in. ICD codes measure diagnosis-seeking, and post-surgical patients see doctors more. And the database’s sex field is recorded sex, which is often updated after transition, so “males” and “females” in this paper are not reliably natal sex — only the feminising and masculinising cohorts are, and those are the ones the Jorgensen letter says may be contaminated. https://doi.org/10.1093/jsxmed/qdaf026 · https://pubmed.ncbi.nlm.nih.gov/39996623/
The tally
| entries | |
|---|---|
| Q1, young men by ideology | 8 (E1–E8) |
| Q2, direction of causation | 6 (E9–E14) |
| Q3, natal males transitioned young | 6 (E15–E20) |
| Total | 20 |
Figures in circulation that do not survive checking
Guards, so the page does not repeat a number that has already been withdrawn or misread.
- The “56% of young liberals” figure is the women’s cell. In Goldberg’s own breakdown it is 56% for White liberal women aged 18–29 and 34% for the men (E3). Posts giving the 56% without the sex split are quoting the female number as the group number.
- “Liberal boys were significantly more likely to report depression than conservatives of either gender” is al-Gharbi’s sentence, not Gimbrone’s. The paper’s means carry the ordering; the paper never tests that contrast (E2).
- Silver’s “conservative 51% excellent, liberal 20%” are composition shares, not rates. They say that of people reporting excellent mental health, 51% are conservative. His actual means are liberal 53, moderate 58, conservative 68 on a 0–100 scale (E4).
- The stigma objection is not a dead end. Goldberg declared it one in 2023 after a covariate adjustment (E10). A randomised wording experiment in 2025 erased the whole gap (E9).
- The American Family Survey “18 percentage points” figure did not verify. The 2021 through 2024 AFS report PDFs were read directly. The 2022 report, 3,000 adults fielded 8–15 August 2022, contains no mental-health-by-ideology figure at all; its mental-health content is care access and parental concern. Treat the attribution as unsourced (grok row R11, https://x.com/BradWilcoxIFS/status/1579436857355800576).
- Dhejne’s “19× suicide” compares post-operative trans people to the general population, not to untransitioned trans people, rests on 10 deaths, loses significance in the 1989–2003 half of the cohort, and the paper itself says no inference about treatment can be drawn (E16).
- Bustos’s regret meta-analysis says 7,928 patients in its abstract and the corrected total is about 5,672. The erratum fixed the table and the abstract was never changed (E18).
- The uncited “2025 study of 107,000+ patients” is Lewis et al. 2025 and is banked at E20 with its design limits attached.
- Healthy Minds cannot produce the cell at all. Its questionnaire has no political ideology or party item, verified in the 2023–24 and 2024–25 instruments. Any claim sourced to Healthy Minds about respondents’ own politics is wrong at the instrument level.
What was left out and why
The grok rows carry 52 items and seven produced entries. The rest fall into four groups carrying no number on the three questions: the origin and use of the phrase “woke mind virus” (R35–R42), clinician and researcher opinion about aetiology (R18–R29), first-person accounts on both sides (R30–R34, R43–R45), and general young-male crisis figures with no ideology variable in them (R10, R14, R48, R51). They may serve a page in other ways. They are not evidence on the three questions asked here.
Gaps
G1. The cell exists in five datasets and is published as a plain table in one. Gimbrone’s Appendix Table A.3 is the only peer-reviewed table anywhere with liberal boys as its own row (E1). The other four sources are a newspaper write-up of a tweet (E3), two blog posts reading values off charts (E6, E8), and two files nobody had tabulated until this bank did (E4, E5). Walking all 41 works citing Gimbrone found no journal-published replication with a sex split between 2022 and 2026. What it changes: the page can say the boys’ gap is real in every dataset that has the cell. It cannot say the finding has been replicated in the literature, because it has not been.
G2. This was never a data gap. It was a publication gap, and it is now partly closed. The Cooperative Election Study asks 60,000 Americans a year and records ideology, sex and age. Schaffner et al. used that exact file and published no ideology × sex × age table. Silver used it and published sex and age separately. Both cells were computed here in an afternoon (E4). The same is true of the GSS (E5). What it changes: the answer to “has anyone measured this” is now yes, including us. Two cautions attach. The CES dropped the mental-health item in 2024, so that series ends at 2022. And a tabulation run for a bank is not a published finding — a page that leans on E4 or E5 should say who ran it and on what file.
G3. The boys’ gap is largest on the measures that ask for a label and smallest on the one that asks for a count. Set the four measures side by side. Diagnosis reported: 34% against 16% (E3), 25.7% against 16.9% (E5). Self-rating: 46.9 against 59.3 (E4), 2.92 against 2.50 (E5). Symptom scale: 2.06 against 1.89 (E1). Days of poor mental health in the past 30: 4.07 against 4.13 — nothing at all (E5). And a randomised wording experiment erases the whole gap when “mental health” becomes “overall mood” (E9). What it changes: the page cannot treat the boys’ gap as a measured difference in illness. It is largest where the respondent is asked to apply a category to himself and it vanishes on the one item that asks him to count days. That pattern is what a reporting effect looks like, and it is what E9 predicts.
And the test that would settle it has not been run. No measurement-invariance analysis of a standard depression instrument — CES-D, PHQ-9, K6 — across political groups was found. Until someone shows those scales mean the same thing to a conservative and a liberal answering them, every gap in Q1 is a gap in scores whose comparability is assumed rather than demonstrated.
G4. No study measures ideology at one time and mental health later in young men. Gimbrone is repeated cross-sections and its authors say so: “data were cross-sectional. Future longitudinal analyses of the effect of political beliefs on adolescent mental health could help establish causality.” The one two-wave cross-lagged study found (E13) measures finding politics stressful, not ideology, reports no sex split, and finds both arrows significant in young people with the reverse arrow larger. The one study that tested whether the personality-ideology link is a male story found it is not (E14). What it changes: the direction question is open and the page has to say so. Nothing here would let anyone say liberalism makes young men ill, and nothing here would let anyone rule it out.
G5. The woke route has been measured once, in Finland, and it came back empty. Lahtinen’s critical social justice scale is the only instrument found that measures woke attitudes and mental health in the same people (E7, E12). Its male cell exists, the mental-health correlations hold in both sexes, and the woke measure buys nothing over simply asking whether someone is left-wing. Nobody has run an equivalent on young men in the US or the UK, and the Finnish sample was self-selected through a newspaper link. What it changes: the specific mechanism the page is built to test has one measurement and that measurement is against it. Asserting woke ideology as the route asserts something that has been looked at once and not found.
G6. For natal males, the benefit result is the null one, and the discontinuation figures contradict each other. The strongest prospective study reports depression, anxiety and life satisfaction improving in natal females and not in natal males, in its own words, with interaction terms that exclude zero (E17). Discontinuation by sex points one way in the US military data (natal males 19%, natal females 35.6%) and the other way in the Finnish register (natal males 10.5%, natal females 6.5%) (E19). The detransition systematic review covering 7,074 people has no natal-male breakout at all, and neither does the Finnish mortality register (E16, E19). What it changes: a page can say the mental-health case for medical transition rests on results that did not appear in natal males in the one prospective study that split them. It cannot say how often natal males stop, because the two national registers disagree by a factor of two in opposite directions.
G7. Nobody has joined the two halves of the question. The suspected route runs liberalism → transgenderism → harm in young men. This bank found no source measuring ideology and transition in the same population: no study of whether liberal young men are more likely to identify as trans or seek medical transition, and no study of whether outcomes after transition differ by the patient’s politics. The Q1 literature never mentions gender identity as a variable; the Q3 literature never mentions political ideology. What it changes: the middle link is not weakly supported. It is unmeasured. The page can argue for it and it cannot cite anyone.
G8. Nothing anywhere compares treated and untreated adolescents with random or quasi-random assignment. The two designs that come closest both weaken the treatment case. Ruuska’s suicide gap disappears once psychiatric treatment history is controlled for (E16). The 2020 Bränström and Pachankis correction states that comparing surgically treated to untreated dysphoric people “demonstrated no advantage of surgery” and calls the original conclusion “too strong.” What it changes: every number in Q3 is observational, and the page should present them as what happened to people who chose or were selected into treatment, never as what treatment did to them.