I have a feeling that the variability is at least partially modulated by factors such as (lack of) encouragement and gender bias. Imagine a spinning flywheel (this is a terrible analogy but I think it will illustrate my point) with particles on it that jump off the edge at random velocities. The added velocity of the flywheel eventually makes them end up at a certain radial distance on the ground. Regardless of the speed at which the flywheel spins, the average position of the particles on the ground will be at the center of the flywheel. But if the flywheel is spinning fast, the variability will be greater than if it is spinning slowly.

I think in a similar way, encouragement and bias -- like the speed of the flywheel -- can account for this variability, rather than intrinsic differences in the capacity for mathematical reasoning.

That said, it is interesting why the gap seemed to bottom out at 4 to 1 despite supposed programs to encourage young women to go into mathematics and science.