Thread by @segm_ebm

Recent claims of a “free fall” in youth trans identities may have been greatly exaggerated. A sample of 45K+ students shows trans & nonbinary identities at an all-time high. The claimed drop may stem from flawed weighting and poorly designed survey questions. SEGM's analysis⬇/1

Per the NCHA data, in 2025, 8% of women & nearly 5% of men attending U.S colleges had a non-"cisgender" identity. While there is no evidence of a drop in transgender/nonbinary identities, the data suggest that we may be approaching a plateau. /2

segm.org/transgender-id…

Youth with "nonbinary"- type identities far outnumber those who identify as "trans men" or "trans women." Since 2022, the nonbinary numbers have leveled off, leading to an overall "trans identification" plateau— just as the cross-sex identity, esp. FtM, is still increasing./3

The data, which come from the National College Health Assessment in the U.S., provide three different ways to estimate "transgender" ("non-cisgender") identification. It yields a range of possible estimates and suggests that at least 1 in 20 do not identify with their sex. /4

Our conclusions using the American College Health Association's National College Health Assessment data contradict two recent analyses that concluded that trans identity in youth is in a "free fall." We examine the reasons for our discordant conclusions. /5

First, we look at the analysis by Kaufmann. We conclude that, at best, what Kaufmann observed is a decline in nonbinary trans identification, but even that finding needs to be rigorously examined. /6

Next, we look at the analysis by Twenge. We observe that her conclusion of a "free fall" in trans and nonbinary identities hinges entirely on the application of statistical weights. Raw data has either only marginally significant findings or suggest an increase, not a fall. /7

It is impossible to know whether the weights improved or distorted the estimates. However, it is concerning when the entire finding is dependent on a weight designed for a population of 60K, but applied to a very small group (trans-identified youth are <0.2% of the sample). In 2024, the weights, when applied to this subgroup, had a particularly unusual effect, sharply deflating raw numbers./8

More generally, applying weights is a double-edged sword. It can help rebalance non-representative samples, but it can also potentially distorting findings, especially when weights are applied to very small sub-samples. The weights in Twenge’s analysis range from 0.0001 (discounting a response to near zero) to 15 (i.e., 1 response counts for 15). With fewer than 200 trans-identified individuals in the sample, even a few responses with extreme weights can render the entire estimate unstable or invalid. /9

While we believe the data source we used (ACHA-NCHA) is more robust than those relied on by Kaufmann and Twenge, our analysis has numerous limitations. Further, it only looks at U.S. college students. Trends among non-college youth may differ. These questions merit careful analysis using reliable data sources and sound methodologies. /10

Social media platforms have largely displaced peer-reviewed journals as de facto forums for debate in gender medicine. We thank @epkaufm and @jean_twenge for initiating this important discussion and debate.

Although we conducted extensive internal review, we are not immune to error and welcome questions, discussions, and rigorous debate! /end

segm.org/transgender-id…