[Symmetry]

The mental model I'm using here is that IQ is the result of something like (Genetic g) * diminishing_returns(nutrition). And that nutrition is something like K1*(social_factor) + K2 + noise(). So noise is going to result in smaller deviations when the social_factor is higher.

Before reading that website I would have expected that variations in the first term for nutrition would be small compared to the second term, so I wouldn't have necessarily expected this to show up between affluent and poor kids in the US. But between affluent kids in the US and kids from sub-Saharan Africa I certainly would have.

EDIT: In fact, I would have guessed the working-class/middle-class IQ differences were an order of magnitude lower.