26 July 2017

Are There Really 15,500 Transgender Individuals In Current Military Service?

A Denver Post story in reaction to President Trump's ban on transgender individuals serving in the military states that:
Transgender people are banned from the military, yet an estimated 15,500 are serving.
This is probably greatly exaggerated and is probably at least an order of magnitude too high. The Denver Post's source, however, is disclosed and can be examined. The basic problem is that neither of its core data sources, Gates (2011) for the overall prevalence of transgender individuals in the general population, and the National Transgender Discrimination Survey (NTDS) from 2008 (whose respondents are all gender non-conforming and are part of a large convenience sample with no demographic information on respondents' ages), are particularly precise sources of data. The NTDS Survey isn't really designed to provide information that fits the needs of the Williams Institute, although isn't a horrible place to look for a rough estimate.

The Denver Post's estimate is based upon this report from the Williams Institute, which clarifies that this is out of activity duty, reserves and national guard. This is broken out as:

7,300 Active Duty (assigned male)
1,300 Active Duty (assigned female)
5,300 Guard/Reserves (assigned male)
1,400 Guard/Reserves (assigned female).

As I noted in a previous post on the topic (relying on data cited here):
The DSM-IV (1994) quotes prevalence of roughly 1 in 30,000 assigned males and 1 in 100,000 assigned females seek sex reassignment surgery in the USA. The most reliable population based estimate of the incidence occurrence is from the Amsterdam Gender Dysphoria Clinic. The data, spanning more than four decades in which the clinic has treated roughly 95% of Dutch transsexuals, gives figures of 1:10,000 assigned males and 1:30,000 assigned females. 
In September 2007 however, Olyslager and Conway presented a paper at the WPATH 20th International Symposium demonstrating that the data from this and similar studies actually implies much higher prevalence rates, with minimum lower bounds of 1:4,500 assigned males and 1:8,000 assigned females across a number of countries worldwide. They also present other evidence suggesting the actual prevalence might be as high as 1:500 births overall.
My source has been updated since my last post to add this point:
The most recent (2016) systematic review of prevalence, leading to a meta-analysis of 27 studies, found estimates per 100,000 population of 9.2 (95% CI = 4.9–13.6) for surgical or hormonal gender affirmation therapy and 6.8 (95% CI = 4.6–9.1) for transgender-related diagnoses. Of studies assessing self-reported transgender identity, prevalence was 871 (95% CI = 519–1,224); however, this result was influenced by a single outlier study. After removal of that study, the figure dropped to 355 (95% CI = 144–566). Significant heterogeneity was observed in most analyses.
This is 1 per 114 people with the outlier study, and one per 288 people without the outlier study, and with the self-reported surveys producing much higher estimates than estimates based upon therapy or third-party diagnostic related estimates.

The Williams Institute relies on "Gates (2011) estimates that approximately 700,000 adults in the US are transgender.", and on the NTDS survey result that "this group is 60% male assigned at birth and 40% female assigned at birth". The adult U.S. population in 2011 was 237,801,767. So, Gates is estimating that 1 in 340 adult Americans is transgender, which is probably significantly too high. Likewise, the assigned male to assigned female ratio of 60-40 in NTDS is probably low relative to other more reliable estimation efforts. 

The Williams Institute footnotes this conclusion from Gates (2011) as follows:
Estimates of the size of the transgender population from national population-based surveys do not exist. This estimate is based on two state-level population-based surveys in which questions regarding transgender status implied a gender transition or at least discordance between sex at birth and current gender presentation.
There are currently 1,219,510 personnel classified as male in active duty U.S. military service and 210,485 personnel classified as female in active duty U.S. military service.

If the Williams Institute figures are correct then 1 in 167 people assigned as male in the U.S. active duty military is transgender and 1 in 162 people assigned as female in the U.S. active duty military is transgender. This is about 3 times the rate at which individuals are transgender in the highest credible estimate for the general population. It is a rate 27 times higher for assigned males and 49 times higher for assigned females than a more conservative minimum lower bound for prevalence of transgender individuals in the general population.

The Williams Institute estimate is also notable for showing no assigned gender disparity, when all of the conservative estimates show that assigned males are two to three times as likely to be transgender as assigned females.

Now, admittedly, the military is not a random sample of the general population. For example, the number of high school graduates enroll in higher education is about ten times as great as the number of people who will ultimately enlist in the military for any given cohort of high school graduates (admittedly the numbers do not fully overlap). This is about 4-5 to one for men and about 30-1 for women.

Since the military is one of the most traditionally masculine pursuits, it makes sense that transgender individuals assigned as female who are not able to medically transition would enlist at higher rates than similarly situated transgender individuals assigned as male, so the lack of the prevalence distinction in the Williams Institute report between assigned males and assigned males that is found in the general population isn't necessarily something that discredits its data.

But, given that the military is also well known to be an institution that has been hostile to LGBT individuals and that the active duty service affords the least opportunity to conceal one's gender identity, it is surprising to see a prevalence of transgender presence in the active duty military that is three times the highest credible estimate of prevalence in the general population. Naively, one would expect a lower prevalence than in the general population with a rate of 1 per 1,000 or so, which would imply that there were about 1,220 assigned males and 210 assigned females in the U.S. military who were transgender, for a total of about 1,430 or less than 10% of the numbers estimated by the Williams Institute.

The methodology suggests that the estimate is overstated because the survey is not a random sample and is probably biased towards people who LGBT gender identities. According to the Williams Institute:
The primary data source for the estimates of transgender military service is the NationalTransgender Discrimination Survey (NTDS), which was conducted by the National Gay and Lesbian Task Force and the National Center for Transgender Equality. This 70-item survey was distributed in cooperation with over 900 organizations across the United States and also was announced through listservs and online communities. It was made available both online and on paper in English and Spanish. The survey was fielded over six months beginning in fall 2008 and resulted in 6,546 valid responses, which is the largest sample of transgender people in the US to date. . . . 
As a purposive sample of transgender adults in the US, estimates derived directly from the NTDS could be biased if the true demographic characteristics of the transgender population differ from the characteristics of transgender respondents to the survey. For example, relative to the US population, NTDS respondents are younger and report higher levels of education. Both factors would be associated with lower levels of lifetime military service. Given the lack of demographic data on the transgender population derived from population-based sources, it is not possible to determine if the age and educational attainment levels of NTDS respondents are different from the general US population because younger and more educated transgender individuals were more likely than others to have completed the survey (known as selection bias) or if transgender individuals are, in fact, younger and more likely to have higher levels of education compared to the general population.  
More than 93% of NTDS respondents provided information using an online web-based survey. 
The primary risk is that survey respondents to a 70 question online survey announced on listservs and online communities in the fall of 2008 overstates has a higher proportion of LGBT respondents than the general public. Indeed, this is almost certainly the case.

The report goes on to state that:
In the US, approximately 5.8% of all adults who have ever served in the armed forces are currently on active duty and 4.4% are now serving in the Guard or Reserve. An estimated 86.8% are veterans who served on active duty in the past and 3.0% are retired from Guard or Reserve service.  
The number of transgender adults in each category is estimated by applying these same proportions to the estimated number of transgender individuals who report any service in the armed forces. 
Transgender military service 
Analyses of the unadjusted NTDS data show that 29.6% of respondents assigned male at birth reported that they have served in the armed forces along with 6.0% of those assigned female at birth. In total, 20% of NTDS respondents reported some type of military service. Assuming NTDS reported rates of military service are true of the transgender population in the US, Figure 2 shows adjusted estimates of military service for the transgender population (separated by sex assigned at birth) and for adult men and women in the US. When figures are adjusted such that the age and educational patterns of the US adult male and female population are applied to the NTDS sample, an estimated 21.4% of transgender individuals have served in the military. The adjusted estimates suggest that 32.0% of those assigned male at birth and 5.5% of those assigned female at birth have served.
By comparison, in the general population "1.4 percent of all female Americans have ever served in the armed services, compared to 13.4 percent of all male Americans." (The Williams Institute study claims that 19.7% of adult men and 1.7% of adult women have served, with a combined 10.4% of all adults, and does not make an age adjusted comparison of military service rates to its sample's demographics despite the fact that military service rates for people under age 45 who are disproportionately represented in their data source are less than half as likely to have served in the military.) It is also surprising giving the educational level of the respondents which significantly exceeds that of the general population since people with college educations are less likely to have served in the military than members of the general public.

Thus, if the Williams Institute data is accurate, transgender individuals assigned as male at birth are about 2.2 times as likely to serve in the military as members of the general population and transgender individuals assigned as female at birth are about 4 times as likely to serve in the military as members of the general population.

Keep in mind that for most of the period from 1940 to 1973, there was a draft in place for men, so many people served in the military involuntarily. For example, about 2.2 million of the men who served in the Vietnam War were drafted, 1.5 million of the men who served in the Korean War were drafted, and 10 million of the men who served in World War II were drafted. There were also significant numbers of interwar conscripts. These numbers compare to 22 million people alive today who are military veterans, although, of course, some people who were conscripted for military service from 1940 to 1973 are no longer living.

Conscription makes it relatively less likely that there would be a huge difference between the likelihood that transgender individuals assigned as male who served in the military than non-transgender individuals during the period from 1940 to 1973.

Also, the NTDS sample is younger than the gender public, yet the percentage of young people who have served in the U.S. military is significantly lower than the percentage of all people who have served in the U.S. military.

Using data from the Census and the Veteran's Administration it is clear that rate of military service for men under the age of 45 is well under 10%. So, the NTDS survey suggests that transgender individuals assigned as male have an age adjusted military service rate about three times as great as members of the general public.

In its defense, the Williams Institute does offer some corroboration of its counterintuitive conclusion that transgender individuals are much more likely to enlist in the military than the general population. It states:
There is other evidence that transgender individuals represent a larger portion of those in the military than their proportion among adults in the US population. In a survey of transgender people assigned male at birth, Shipherd et al. found that 30 percent had served in the military, which is similar to military service among transgender people assigned male at birth in the NTDS. A recent study by Blosnich et al. reviewed all health records of veterans receiving health care through the Veterans Health Administration (VHA) from 2000 through 2011 and found a prevalence of Gender Identity Disorder (GID) five times that of the US general population. Though individuals with GID diagnoses may or may not identify as transgender, the substantially higher prevalence of GID among veterans in the VHA system provides further evidence that transgender people are overrepresented in the US military. 
But, even if the enlistment rate is much higher for transgender individuals than for non-transgender individuals, the estimated prevalence of transgender individuals in the general population from Gates (2011) is almost surely an overestimate.

While this is not impossible, it is very unlikely that there are really 15,500 transgender individuals serving in the active duty military today. It is far more likely that respondents are either atypical of the gender population with regard to military service, perhaps due to the way that survey respondents were recruited, or that they are simply not accurate in their reporting of their military service (as over reporting of all kinds of events is a common problem in survey based studies).

Thus, to get to a 15,500 active duty transgender soldiers in the U.S. military, you need to assume somewhat more than the highest credible estimate of the prevalence of transgender people in the general population and a military enlistment rate for transgender individuals that is three times that of the general population for individuals of a comparable age, as the NTDS sample suggested.

Neither of these assumptions is likely to be true, and while this makes the decision to bar transgender individuals from military service look like it harms more people than it actually does, it also dramatically exaggerates the likely cost of accommodating transgender individuals in military service. 

The estimate also fails to distinguish between individuals who are transgender but "in the closet" and those who are "out". The number of "out" transgender soldiers in the U.S. military appears to be much lower (e.g. there appears to be only one "out" transgender soldier in all of Fort Carson in Colorado). Realistically, there are probably several dozen to a couple hundred "out" transgender individuals serving in the U.S. military. Obviously, military policies on whether transgender individuals are allowed to serve in the military influences this (and they were only allowed to do so late in the Obama Administration).

No comments: