I studied engineering and there was very few women (<20%) in my class. You can't blame companies for having a similar hiring ratio. And the recruitment of the college I went through ("grandes ecoles" in France) is purely based on a math+physics competitive exam, so there is very little selection bias in the first place. So I am not convinced discrimination is the key to explain gender imbalance in engineering jobs.
What if the exams and job interviews are designed by people who are bad at team work? During my studies and also in my various jobs I observed that there is little team work, that many people even openly don't want to do it. (Managed to do it for some things, e.g. Functional Analysis) Maybe there are more ways to approach problems and at least for "real work" it seems that working in a team leads to better results. (Think all this agile thing with pair programming, retros etc etc...)
Don't forget Correlation isn't equivalent with Causation, actually there is neither a <= nor a =>.
I don't see how the exam's and job interview's focus (or lack thereof) on teamwork is relevant. How to make a better exam isn't really on trial here, because no matter how much data suggests that teamwork is as important as technical ability in an engineering role, no company or school will have a moral responsibility to ensure their exams reward good teamwork. The exam is still equally fair to any gender so long as it is totally anonymous.
