not sure if that's true. do you mean fixed delta skew (e.g. ATM - 25d put) or fixed moneyness (e.g. 100%-90%)?
what I have observed and what I am trying to say is that the negative relationship b/w VIX (in points or in %) and SPX (in % or points) is not linear. it is convex instead. The VIX definition effectively has both vol level and vol skew components. My thought was the skew got to get larger to cause the convexity, especially if the skew is not normalized by ATM vol. However I have no mathematical proof for this.
Was looking through the TVIX prospectus today because of the whole stop issuing new shares thing. "The long term expected value of TVIX is 0." Wow, they are honest!
They should be as they likely need to be. Insane how many shares of the ETNs have been created. That recent XIV pullback was an awesome entry point. Even start of day today was a good entry point with march 19% above spot and april 13% above march. Ridiculously steep curve with a decent margin of safety for spot vix to spike by march settlement.
I hope you were not short TVIX I think this insanity still has ways to go but once it's over, it will be an awesome short.