Today, $VVIX closed at its lowest level in over three years. More interesting, it is in its lowest quartile using all historical data (dating to 2006 when options on $VIX were first listed and their implied vol could be calculated).

As has been discussed recently, $VIX has been range-bound in its upper quartile. Buyers have been unwilling to pay more than the current elevated levels for optionality absent a catalyst, and sellers have been risk-averse since January.

Realized vol-of-$VIX confirms this. Its current level (67) is low historically (roughly the 10th percentile). Furthermore, it has been falling steadily since April. ($VVIX has fallen steadily since May.)

However, while realized volatility-of-$SPX is predictive of $VIX (a Granger causality test is significant), realized volatility-of-$VIX is not predictive of $VVIX. If low realized vol-of-$VIX isn't causing the local minimum in $VVIX, what is?

One possibility is put/call ratio. Unlike $SPX options, where the ratio routinely exceeds 1, $VIX calls are bought more than puts, by as much as 2:1 or 3:1. The calculation of $VVIX is based on the par variance (of $VIX) swap rate, which weights puts more heavily than calls.

For a regular grid of strikes and two options with the same price, the lower strike option (which is the put since only out-of-the-money options are used in the calculation) will receive a larger weight owing to the red term above.

Therefore, if there are more call buyers (or maybe more realistically fewer put buyers) than usual (I don't have enough historical put/call ratio data for $VIX options to test), $VVIX may be falling solely due to the composition of the options on which it is based.

If this is indeed happening, a second-order effect may be the error introduced by approximating an integral by a sum over a discrete grid of strikes. If fewer puts are traded, strikes below a certain level become ineligible and approximation worsens and is biased downward.

This is just a conjecture. If someone has historical put/call ratio data for $VIX options, please let me know if there's anything to it or if I can add it to the list of plausible hypotheses that don't hold when tested.