"A meta-analysis by Flore and Wicherts (2015) concluded that the average reported effect of stereotype threat is small, but also that the field may be inflated by publication bias. They argued that, correcting for this, the most likely true effect size is near zero.[19]"
The third reference wasn't available, so can't comment.
These types of "studies" tend to suffer from causal inversion, i.e. "wet roads cause rain", in arguing that the perception of these fields as male-dominated causes women to not enter them.
The simpler and more plausible explanation is stereotyp accuracy: people correctly perceive these fields as male-dominated because women tend to not enter them. That such a perception is an effective deterrent seems implausible given the many other fields that used to be completely male-dominated and are now female-dominated (for example veterinary medicine).
As a feminist, I also can't get on-board with a view of women as such delicate fragile flowers that such perceptions would prevent them from doing something they like, and it doesn't match the women in my life...but maybe I have been blessed with unusually non-delicate women, who knows?
Finally, who says it is female underrepresentation? Maybe it is male overrepresentation? And in fact, there is strong evidence that this is the case. Both men and women who score well in both verbal and math SATs tend to go into non-STEM fields. People who only do well in math tend to go into STEM fields, partly because they don't have a choice. And that group is largely male.
Yes, there is debate about these things, as there often is in the social sciences. This is from the same criticism section on wikipedia:
"In a 2009 meta-analysis, Gregory M. Walton and Steven J. Spencer argued that studies of stereotype threat may in fact systematically under-represent its effects, since such studies measure "only that portion of psychological threat that research has identified and remedied. To the extent that unidentified or unremedied psychological threats further undermine performance, the results underestimate the bias."[23] Despite these limitations, they found that efforts to mitigate stereotype threat significantly reduced group differences on high-stakes tests.[23]"
Please also note that nobody is claiming that stereotype threat is the only factor in STEM performance.
I'm no expert in the field either, and I welcome discussion. But I would prefer a better faith argument than copying the most "gotcha" paragraph from the criticism section of wikipedia.
If you could verbalize your skepticism more clearly I think this discussion would be more fruitful.
EDIT: I see you have edited your comment and added a lot. The above was my response to your first draft.
>As a feminist, I also can't get on-board with a view of women as such delicate fragile flowers that such perceptions would prevent them from doing something they like, and it doesn't match the women in my life...but maybe I have been blessed with unusually non-delicate women, who knows?
This feels especially in bad-faith. These are statistical effects, at society scale. Considering how clear the effectiveness of repeated messages is at a social level (e.g. marketing, religion, fake news, ideology), it should not surprise you that ideas of stereotypes influence people's behaviour. Humans fall for marketing as a society, are we all "delicate fragile flowers"?
https://en.wikipedia.org/wiki/Female_education_in_STEM
And if that isn't enough, a quick search on the web found a number of published papers on the subject:
https://www.frontiersin.org/articles/10.3389/fpsyg.2017.0071...
https://journals.sagepub.com/doi/abs/10.1177/194855061244073...
https://link.springer.com/article/10.1007/s11199-011-0051-0