Valuing volatility spillovers

We show that volatility spillovers are large enough to matter to investors. We demonstrate that standard deviations of returns to mean-variance portfolios of European equities fall by 1-1.5% at daily, weekly, and monthly rebalancing horizons when volatility spillovers are included in covariance forecasts. We estimate the conditional second moment matrix of (synchronized) daily index returns for the London, Frankfurt and Paris stock markets via two asymmetric dynamic conditional correlation models (A-DCC): the unrestricted model includes volatility spillovers and the restricted model does not. We combine covariance forecasts from the restricted and unrestricted models with a wide range of assumed returns relatives via a polar co-ordinates method, and compute out-of-sample realized portfolio returns and variances for testing. Diebold-Mariano tests confirm that most risk reductions are statistically significant. Stochastic dominance tests indicate that portfolios accounting for volatility spillover would be preferred by risk adverse agents.