For the reasons you stated, we will never know for sure to what extent sex differences in STEM achievement and racial differences in academic achievement are genetic. In that case, we should not jump to the conclusion that any difference is either a flaw in the measuring instrument ("the SAT is biased") or the result of unequal opportunities ("rich kids score better because of test prep").
We can know how cultural bias influences STEM achievement gender gaps by
cross-referencing data against other countries. Any differences due to genetics should be consistent.
Boys outperform girls in mathematics in 35 of the 65 countries and economies that participated in PISA 2009. In five countries, girls outperform boys, and in 25 countries there is no significant difference between the genders.
We can find out more by doing controlled studies into such things as teacher bias:
http://www.livescience.com/19552-girls-math-teachers-bias.html
