Sample Bias

(Post in question #1)

(Post in question #2)

The super scary “autogynephilia-confirming” Hostdefe Individualism Index research. Apparently, there is a strong correlation between individualism in a society and what percent of transgender individuals are straight vs non-straight. Unfortunately, it seems that Anne Lawrence (the only author to publish on this topic) has a case of “sample bias” (here being used to describe the sample of studies she used).


Cuypere et. al (which happens to be included in Lawrence’s original paper) found that 10 of the transsexual subjects had a ‘homosexual’ orientation, while 4 were bisexual, 6 were ‘heterosexual’ and 2 were asexual. Consequently, the proportion of non

(A large number of these are going to be older studies, but I think it’s interesting that 1988 was chosen as the cutoff date, especially considering that the oldest study is exactly from 1988, almost as if the range was chosen after the studies were collected. It also does not match up with DSM publication dates: the DSM-III was published in 1980 while the DSM-IV was published in 1994, and the DSM-V not until 2013 (but of course Lawrence couldn’t have known that in 2008. Coherent dates would be 1980-1994, 1994-2008 or 1980-2008 [the most dubious of the three])

Sorenson et. al 1982 found that 2/3 of Danish male-to-female transsexuals had homosexual experiences and 1/3 have not, while Denmark has a fairly high Hofstede score at 71. If we look at Brown’s chart, we would expect the percent of transsexuals without homosexual experiences to be 50%, but the true number is fairly below that.

Sorenson 1981 found that:

At the time of follow-up 17 (12) of the patients lived alone. Six
(2) cohabited with men and of these five (1) were married. None of the married
transsexuals had had difficulties with the authorities in getting married. Ten (4)
reported having cohabited with men for more than 6 months postoperatively


Seventeen (10) of the operated 23 (14) males had male partners. Two (0) of
these had had sexual relationships with females, though these were not steady

Given that the country the study took place in has a Hofstede score of 71 (as reported above), we can see the disparity between the 73% “homosexual” (straight) trans women percent, and thus 27% non-“homosexual”. The graph predicts a value of about 50% for a HII of 71, so a value of 27% is far below that.

Nieder et. al 2011 creates a problem for Blanchardianists. They are forced to choose between the Hofstede Individualism Index hypothesis and their repeated assertion that ‘homosexuality’ and early onset dysphoria are two nearly concentric circles, while ‘non-homosexuality’ and late onset dysphoria are two nearly concentric circles. Table 7 shows that 42.3% of transsexuals in Belgium were ‘late-onset’, which if were to follow typical Blanchardianist dogma means that they are ‘non-homosexual’. Similarly, the value for Germany is 30%. Belgium has a HII of 75, which predicts a “NHS” value of around 60%, an almost 20% difference. Germany’s HII value is 67, predicting a value of around 45%, yielding a difference of about 15%. This considerably drags down the trend line.

O’Gorman 1982 has several relevant statistics here. 47% of the patients were “late-onset”, and again, we will use onset as a proxy for sexual orientation to test Blanchardianist assumptions and expose a contradiction. So by this logic, 47% of patients were non-heterosexual. Hofstede did not publish values for his analysis of Northern Ireland, but some studies have used the average of the Republic of Ireland and the United Kingdom as an approximation, yielding a value of 79.5. Using the chart, the predicted percent of NHS transsexuals would be 65%, giving us a different of over 15%.

Hoenig et. al 1970 (an old study, I know!) found that 78% of the trans participants (in England) were a Kinsey 5 or 6 (in relation to ‘birth sex’) and therefore 22% NHS. The HII for the UK is 89, which again plugging this value into the graph yields the prediction of about 90% NHS, almost the complete opposite of what we found!

Hoenig et. al 1974 had 74% as ‘homosexual’ (as well as 83.3% crossdressing [high rates of crossdressing among straight trans women are replicated in places like Thailand and Singapore]), thus 26% NHS. This is similarly contrary to the predicted value of 90% from the HII score.

[Some other interesting tidbits in this study show that personality disorders are somewhat correlated with transsexuality (mostly OCD, BPD, HPD, and the bizarre “sensitive” personality disorder that I can only find in reference to BPD). The data is pretty inapplicable today because of how vastly criterion for personality disorders has changed]

Dixen et. al 1984 found that 75% of trans women currently in a relationship were in a relationship with a man (i.e. “homosexual” in the eyes of Blanchardianists), and 25% were in a relationship with a woman. The percent of “NHS” should be about 25%, contrary to the 90% prediction given the United States’ HII score of 91.

Kuyper et. al 2013 has some confusing results that take a bit of interpretation to get a proportion of “NHS” to “HS” trans individuals. Taking a few liberties with the definitions as most of the data is unavailable (which of course will make this analysis somewhat speculative), we note that individuals with an ambivalent gender identity are twice as likely to be homosexual or bisexual (which is not replicated in Joel et. al 2018 – they find a non-significant correlation). Table 4 shows that the power of this association remains constant, and strong when considering ‘homosexual’ alone, so we will drop the bisexual portion. People with ambivalent gender identities are more than twice as likely to have a ‘homosexual’ sexual orientation, indicating at least a 66.6% – 33.3% split in ‘HS’ – ‘NHS’. Netherlands has a HII of 80, yielding a predicted ‘NHS’ percent of ~65-66%, contrary to what the study estimates. A similar analysis with the given “more likely”, shows that those with incongruent identities are at minimum 51 ‘HS’ – 49% ‘NHS’, also contradicting the estimated value from the graph. If someone ends up getting access to the data set in the study (which you can often do by emailing the author[s]), then they could confirm by using the smaller subsample of individuals with ambivalent/incongruent gender identities that have discontent with their sexed attributes, that want to use hormones and get surgeries, and both have discontent and want to undergo transition.

Cuypere et. al 1995 indicates that 45.5% of trans women are male-attracted, implying that 54.5% are ‘NHS’, contrary to the 66% predicted. This is a disparity of 11.5%.

King et. al 2018 found that 70.5% of trans women have had sexual experiences only with men. This yields a “NHS” value of 29.5%. Now while Hofstede himself never estimated cultural dimensions  for Uganda, there are some estimations and more modern calculations. Rarick 2013 found a HII score of 30, while an average of other Eastern African done by Hofstede yields 27. Due to the time difference between Hofstede’s research and the Rarick study, and the cultural differences between countries, I’ll be using the average between the two: 28.5. An HII of 28.5 predicts a proportion of ‘NHS’ individuals of about 4%, which is a fairly large different from the actual number.

Silva-Santisteban et. al 2011 was primarily focused on HIV, but had some clues that we can use to estimate the percentage of ‘HS’ vs ‘NHS’ in Peru. Table 3 indicates that 10.1+10.8=20.9% of trans women engaged in insertive anal sex the last time they had sex. According to most Blanchardianists and some research, almost all ‘homosexual’ (straight) trans women are ‘bottoms’, do not use their genitalia / have strong aversions to using their genitalia, and thus would not engage in insertive anal sex. If we take that assumption, then the percentage of ‘NHS’ trans women is 20.9%. Peru has a HII of 19, one of the lowest scores in the world. This low score predicts a ‘NHS’ percent of 0-1%, leading to a disparity of 19.9-20.9%.

Zhang et. al 2016 found that 67.2% of Chinese transgender individuals classified themselves as homosexual, meaning that 32.8% are ‘NHS’. We should note that inclusion criterion for the study was having sex with a man, which will very likely exclude a large number of transgender individuals who have not had sex with men. Furthermore, we note that 45.9% of transgender individuals have had vaginal sex in their lifetime, indicating that there might even be some bisexual individuals calling themselves ‘homosexual’ (but this is just a hypothesis, not any definite reasoning). Regardless, China’s HII value of 20, which predicts a ‘NHS’ value of ~2%. This is a 30.8% disparity at the very minimum (remember inclusion criterion).

Prabawanti et. al 2014 found that ‘88% currently only had sex with men’ implying that 12% of waria in Indonesia don’t exclusively have sex with men. This indicates a ‘NHS’ value of 12%, which contrasts with the predicted value of 0% from the country’s HII of 14. This is a 12% difference.

Landen et. al 1998 found that 78.5% of MtFs are sexually attracted to the ‘same’ biological sex, which implies a ‘NHS’ rate of 21.5%. Given that Sweden’s HII is 71 which predicts a rate of 45%, this indicates a disparity of 23.5%.

Three sets of studies by similar sets of authors (all leaders authors are Colizzi) [1, 2, 3] all find that Italy has a proportion of “homosexual” transsexuals greater than 93%, indicating a “NHS” value of 7% or less. This is in contrast with the predicted 55% from the country’s 75 Hofstede Individualism metric, indicating a disparity of 48% or greater.

Another study from Italy, Fisher et. al 2013, found similar results: 82.6% male-attracted trans women. Consequently, the percent of “NHS” trans women was 17.4%, which indicates a disparity of 37.6%. The differences between the studies is likely a result of sampling differences and location.

Cussino et. al 2017 replicates these previous results, finding that 91.2% of trans women in the country are male-attracted, indicating that only 8.8% of the sample was ‘NHS’. Again, Italy’s predicted score is 55%, indicating a disparity of 46.2%.

Bonierbale et. al 2013 indicates that trans women in France are predominantly ‘homosexual’; 64%, thus indicating that ‘nonhomosexuals’ are 36%. This is in contrast to the predicted value of 45% from the country’s HII of 71. This yields a disparity of about 9%.

Wierckx et. al 2011, using a sample of trans women from Sweden, found that 44% of the trans women in the sample were heterosexual (that is, attracted to men), meaning 56% were ‘NHS’. This is in contrast to a predicted NHS% value of 45% (from 71), indicating a disparity of 11%.

Turan et. al 2015 found that 98% of trans women were ‘heterosexual’, but did not clarify which meaning they were using. Contextualization with studies from the same author can elicit some information: This study uses the pronoun ‘her’ in reference to a female-to-male patient, indicating that biological sex (in their view) prevails for the authors. Consequently, the implication is that ‘heterosexual’ refers to lesbian trans women, trans women who are attracted to women. Given the Hofstede score of 37, predicting a NHS% value of 6%, this elicits a disparity of 92%.

TODO: Re-add UK & US studies, look for stuff from Australia & New Zealand again. Maybe South Africa,

General Problems with the Study:

Luckily, some of these are noted in the “limitations” section of the paper, but the flaws seem understated

  1. Lawrence uses more recent figures for sexual orientation among trans people (ranging from 1988 to 2008 as noted above), while using the original figures for Hofstede’s Index (calculated over the period 1967-1973, giving significant credibility to my use of older studies), leaving the unstated assumption that sexual orientation and culture remained the same over the time period(s).
  2. She fails to consider a hypothesis she brings up in other papers: that ‘heterosexual’ trans women ‘lie’ to get treatment, and that this might vary by country and have an influence on the proportions.
  3. She uses various metrics to measure sexual orientation: stated orientation, clinician judgement, sexual experience, sexual partnership, marriage to a woman, preference. Stated orientation (and all other metrics), as outlined above, is often questioned by Blanchardianists, including Lawrence. Clinician judgement is susceptible to inaccuracy due to warped representations and bias. Marriage to a woman as an indication of non’homosexuality’ is misleading because it ignores how often gay men marry women because of societal conditions.
  4. The study indicating that all Brazilian trans women are straight has the possible confound of the fact that gender clinics in Brazil often deny lesbian trans women (see Lobato et. al 2006), which can reperpetuate a lower amount of female-attracted trans women feeling able to go to clinics.

Here’s a table of all of the relevant excluded studies and the calculations in a simpler format.


Author(s) Year Sexuality Reported/Inferred Method of Determination Country IDV Score Predicted % NHS% ΔNHS
King et. al 2018 Reported* Sexual experiences Uganda 28.5* 4% 29.5% 25.5%
Cussino et. al 2017 Reported Stated Italy 76 55% 8.8% 46.2%
Zhang et. al 2016 Reported Stated China 20 2% 32.8% 30.8%
Turan et. al 2015 Reported* Unknown Turkey 37 6% 98% 92%
Colizzi et. al 2015 Reported Stated Italy 76 55% 6% 49%
Colizzi et. al 2015 Reported Stated Italy 76 55% 7% 48%
Colizzi et. al 2014 Reported Stated Italy 76 55% 4% 51%
Prabawanti et. al 2014 Reported* Sexual experiences Indonesia 14 0% 12% 12%
Fisher et. al 2013 Reported Mixture Italy 76 55% 17.4% 37.6%
Bonierbale et. al 2013 Reported Stated France 71 45% 36% 9%
Wierckx et. al 2011 Reported Stated Sweden 71 45% 56% 11%
Landen et. al 1998 Reported Patient files from experts Sweden 71 45% 21.5% 23.5%
Cuypere et. al 1995 Reported Stated Belgium 75 66% 45.5% 11.5%
Dixen et. al 1984 Reported* Relationship United States 91 90% 25% 65%
Sorensen et. al 1982 Reported* Sexual experiences Netherlands 71 50% 33% 17%
Sorensen 1981 Reported* Relationship Netherlands 71 50% 27% 23%
Hoenig et. al 1974 Reported Stated United States 91 90% 26% 64%
Hoenig et. al 1970 Reported Kinsey Scale United Kingdom 89 89% 22% 67%
Silva-Santisteban et. al 2011 Inferred Type of anal sex Peru 19 0% 20.9% 20.9%
Nieder et. al 2011 Inferred Age of Onset Belgium 75 60% 42.3% 20%
Nieder et. al 2011 Inferred Age of Onset Germany 67 45% 30% 15%
O’Gorman 1982 Inferred Age of Onset Northern Ireland 79.5 65% 47% 18%
Overall Average 34.4%
Overall Average Excluding Turan et. al 31.7%
Overall Average Excluding Inferred 37.9%
Overall Average Excluding Inferred and Turan et. al 34.8%
Overall Average Only Using Direct Determination 35.6%

Unfortunately for me, there is another graph to analyze, which I will do at another time. I have some more planned analyses, but I’ve been on long enough of a hiatus (real life gets in the way!)