This is a good meta-analysis of 100 different gendered studies on math performance involving more than 3 million participants, so it has a really nice sample; done by U of Wisconsin:
Hyde, J. S., Fennema, E., & Lamon, S. J. (1990). Gender differences in mathematics performance: a meta-analysis.
Psychological bulletin,
107(2), 139.
Since it's from 1990, it's scanned rather than typed -- meaning I can't just copy/paste. I can email it if you'd like, or I could paste the abstract:
The "lower performance... in high school" is examined deeper in the paper. The authors go on to attribute the issue to social and cultural issues, not biological differences between sexes.
Most recently the same authors made another meta-study of cognition (46 studies in this one) examining the notion that men are cognitively different from women against the hypothesis that they're similar instead:
Hyde, J. S. (2005). The gender similarities hypothesis.
American psychologist,
60(6), 581.
This paper's available to freely read without a subscription:
http://humanbehaviors.free.fr/References - Articles/The gender similarities hypothesis.pdf
The abstract reads:
Which is a fairly conservative way to put the actual findings: that cognitive differences between the sexes is virtually nonexistent; and complicated by (again) social, rather than biological, factors.
This was the conclusion: