question on paper

Ideally, IP3 of double-balanced mixer is better than that of single-balanced without degeneration (since output of transconductance stage is linearly proportional to input). HOWEVER, in reality, there are hot-carrier related velocity saturation and mobility degradation that introduce sub-square-law behavior that makes the double-balanced mixer have worse IP3 than single-balanced. (this is according to the paper)

My question is, I don't understand why mobility degradation would affect double-balanced more than single-balanced. And is there anyplace I can find more info about this?