Transgender population overestimated by ½

update: sequel article here


RuPaul, and Caitlyn Jenner are without doubt the most famous of transgender people.  Though there are undeniably many hundreds of thousands of others in the U.S.  It is though a small portion of Americans (~½%), but they are decent friends living within our communities and positively touch upon our lives.  Now with the election of President Trump, there is a mild anxiety on reversals of the policy advantages this group gained under President Obama.  The MSM news focuses on two topics, the first is the equal access transgender people should have at both work and school, and in public bathrooms.  And the second is on the estimate of just how many transgender people are there.  This article focuses on the latter math topic, with the most popular guess -based on work from untrained regional health employees, and UCLA lawyers and policy flunkies- now stands at >1.5 million adults (or >0.6% of Americans).  We assert here that this statistical measure is highly dubious and likely overestimates the number by ½!  While championing the rights of those disenfranchised (particularly for equal love and safety), it is also critical to note that resulting overestimations is not victimless.  It equally implies that hundreds of thousands of the majority-straight Americans are now be assumed to be transgender, and that unfairly snips at the resource distribution that would otherwise go to government's enumerated straight people.

This website has efficaciously refuted or corrected other poor statistics estimates from a lengthy list of otherwise meaningful sources [e.g., Presidential election forecasts, Oxford mortality study, ESPN Superbowl prediction, NASA’s climate estimates, Warren Buffett’s accounting measure, Wall Street strategists, facebook’s many flawed metrics, solicited articles in nature magazine (here, here), National Review’s “upgraded middle-class”, etc.]  And the 1.5 million estimate from a UCLA think tank here will be no different.  Let’s discuss this through topical segments below.

The older study total

The same UCLA institute published a 2011 study estimating that the adult LGB community was 3.5% of the population, and the some-what related survey of transgender population was an additional 0.3%.  But both of these statistics were deceptive even then.  Comparable studies on the LGB community (in the U.S., or internationally) had no overarching consensus that could validate any of the findings.  Other official estimates from more reliable sources were as low as in the 1% range, for example.  This is important because there is a large difference and hence reliability, among these small portions (1% range, and 3.5%).  In absolute terms, it means that our estimate of the number of LGB Americans can be anything from 4 million, to 8 million, or in some cases even outside of that!

Additionally, the national 0.3% estimate for transgender people was based on a survey in just two bi-coastal and liberal states: California and Massachusetts.  Note that the 0.3% adult transgender estimate comes to just under 700k people.

The newer study total

This is the problem with not calling out flawed studies from the get-go: the idiocy gets worse when left unchecked.  Now a new 2017 study, applying more recent data, takes some of these complications and doubles-down on them (as we’ll see in the next segment).

The newer study, a few years after the older study above, explores adult data from 19 states.  And the serially-linked estimates have nearly tripled!  This is ridiculous, even relative to the wide confidence interval purported on the older study.  Blending in many more states now beyond California and Massachusetts, the less populated conservative states now bring their new estimate to 1.4 million transgender adults (at 0.6% it is more than double the 0.3% older study estimate.) 

Was there a tripling in California’s transgender adults in a few years’ time?  While we can expect some higher comfort level (lower societal stigma) in self-responding as transgender, we know that triple is an impossible multiple when attempting to substantiate this with the age-based Census or CDC study.  This leaves us with the reality that the survey design was tainted to begin with.  And in only a policy sense we should consider if the enhancement of the number of transgender people comes from previously categorized LGB people, even though each person’s unique condition is of course something we should be careful not to discount.  The resulting UCLA estimate could likely be overestimating the transgender population by something in-between, or by ½, or more.

Confidence from fictitious polynomials

It is important to note that the UCA surveyors never actually surveyed teenagers in the 50 states (plus District of Columbia), before releasing a perplexing checkered-chart showing where in the U.S. transgender youths live (because apparently where one is born determines to what extent one considers themselves to be transgender?)  They used a multi-degree polynomial -across subjective- time ranges to extrapolate via uninhibitedly swinging around the number of transgender youth associated with a state.  In other words, they somehow determined that through life people go through predictable waves of being transgender or not, with the peak being earliest in life at nearly twice the rate of those middle-age (before turning up and then down again, later in life…) 

This is again all imprudent conclusions that novice researchers with an Excel have arranged.  Natural demographic curves, which the media and policy makers should appreciate, are never so inexplicably complex.  The UCLA researchers should have stopped their work only in the states they collected adult samples.  Now they have estimates -literally- all over the map, with confidence intervals that make no sense.

In fact, they decided to go the route of using just a 5,000 simulation to spread an adult survey in 19 states, down to teenagers in all 50 states.  This is again foolhardiness, protracted.  We should note again that not only are the age confidence internals too narrow given the total confidence interval, but of the 50 states where the teenage youth were extrapolated, 6% are above the 95% CI and 16% are below it.  And all of the estimates among the national adult age groups (18-24, 25-64, 65+) have nearly a 0.6%, 2-tail confidence spread, when the entire adult population is also at 0.6%.  So that does fit, and neither does the fact that California -the engine powering this probability survey- has a blown out 0.9% confidence spread (they use a similar decorative, non-parametric fit term of credibility spread).  See map below. 

How can a 22% (6%+16%) “error” fall outside of the 5%-allowed error (100% - 95% confidence)?  It shouldn’t be!  And also why are 6% and 16% so far off from one another?  Again, shouldn’t be!

Look at the wackiest result from Iowa at 0.4% transgender youth rate, while their neighbor Minnesota is at 0.9% rate.  Is there probabilistically that much of a societal difference between these individuals on one side of the state line versus the other?  Of course not, and it should be noted that the value for Iowa is hopelessly below the entire national 95% confidence interval, while the entire Iowa 95% confidence interval is below the interval for Minnesota!

These misfitting components of the disordered survey results reflects a meager survey design, undependable and empty results, and a likely overestimation of the number of transgenders in order to boost the think tank’s policy goals.  Except here they use increasingly corrupt math to go overboard on measuring an important topic, and we’ll later learn as we have in the many examples this website has rebutted (listed above) that the consequences for disloyal “policy research” is that this often enduringly backfires on one’s agenda as long as famous independent websites -such as this- to fact-check.