Volatility-of-Volatility: Dynamic Models and Hedging Strategies

Abstract

In this thesis, volatility-of-volatility, as measured by the VVIX Index, is studied as an independent risk-factor for investors. A statistical study of the VVIX time series indicates that, like the well-studied VIX, it exhibits properties of stationarity, mean-reversion, and strong autocorrelation. Further, several stochastic volatility models are tested using Generalized Method of Moments (GMM) in their abilities to explain the behavior of the VVIX. The 3/2-model is found to best fit the VVIX data, a result that is further validated by a simulation-based robustness test. Further, motivated by the recent literature on volatility-of-volatility, the predictability of the VVIX on tail risk hedging returns is explored. After constructing a tail risk hedging portfolio using delta-hedged VIX calls, a vol-of-vol hedging strategy is constructed by exploiting the negative risk-premia predicted by the VVIX. It is found that applying the vol-of-vol hedge significantly improves risk-adjusted returns of a market portfolio, as measured by Sharpe and Sortino ratios, in the period of study. Moreover, it is shown that vol-of-vol-based tail risk hedging is more effective than volatility-driven tail risk hedging when the instruments of choice are VIX calls.