I found this interesting.
This study by Deary, Strand, Smith, and Fernandes, which followed over 70,000 English schoolchildren from age 11 to age 16, found that a person’s latent intelligence predicts educational outcomes quite well, with a correlation coefficient of 0.81. A person of average latent intelligence at age 11 had a 58% chance of successfully obtaining five GCSEs by the age of 16. Those whose latent intelligence was one full standard deviation above the mean did so 91% of the time, while those one full standard deviation below the mean only did so 16% of the time.
The study also found that girls and boys with the same latent intelligence had different outcomes. Whereas 61% of girls of average latent intelligence achieved at least five grades of between A* and C on their GCSEs, boys of average latent intelligence only did so 50% of the time. Moreover, girls had an advantage in every subject except Physics, which had no gendered advantage.
That last fact is important because girls’ comparative advantage in areas such as reading is the single largest driver behind the under-representation of women in math-related fields. A counter-intuitive implication of this (at least, to those harboring certain intellectual priors) is that making boys better at reading, the arts, and the humanities is probably the best way to get more girls into math-related careers.