The Empirical Performance of Alternative Extreme Value Volatility Estimators ∗