I am currently reading the paper of Derman and al for my master thesis on Variance Swap. At one point one says that "The variance vega is largest when the option is ATM", considering here a call option on a stock.
I must say, I am having some difficulties to understand that. The only reason I came up with is that vanilla option are almost linear in volatility at ATM.
In order to clarify my questions which is not clearly explicit.