[qiqing]

I would like to note that the STEM gender gap is much lower among new immigrants students (particularly from Asia and Eastern Europe) than their US-born counterparts, at least anecdotally. And I think this biases me towards the non-negativity approach argued for in the article. Much of the data about lower STEM performance from females in the U.S. comes from 1970's and 1980's data, when females also took fewer courses advanced courses. By 2000, female high school students were taking calculus at the same rate as males, though still lagged behind in physics. [1]

Perceptions are generally not nuanced and often lag 10-15 years behind. If you take the interaction of gender and ethnicity into account, you actually see an interesting reversal at the 99th percentile for Asian Americans where there are where there are 1.37 F for every 1.25 M. There is another study that shows Asian females score higher on STEM subject exams when they had to write a short essay about being Asian prior to the exam, but scored lower when they had to write about being female. (I'll try to add the citation later when I find it.)

Overall, I think the macro strategy of making female role models more visible is an effective one, and may well pay off in the next decade or so. Another reasonable strategy may well involve immigration reform with a special eye towards people who have focused on STEM subjects. Perhaps extra H1B slots for techie females? (I haven't thought this one through yet, but it may be worth exploring.)

Side note: It is also interesting to note that there are a substantially higher percentage of female Iranian mathematicians, and some of those mathematicians have noted that some of this is due to the cultural perception that highly theoretical pursuits (including theoretical computer science and theoretical physics) are not "too masculine" and are less likely to be dissuaded from pursuing the field.

1. http://dericbownds.net/uploaded_images/hyde.pdf