This study examines two important issues underlying realized volatility and correlation estimators. First, an empirical inquiry is conducted to assess whether Bax and Eurodollar futures tick-by-tick data can be characterized as marked-point processes. Second, ARMA, neural network, GARCH-BEKK, and naive volatility and correlation forecasts are compared in an out-of-sample context when a trader prices an interest rate spread option based on those forecasts and simultaneously delta-hedges her position. Other loss functions are also considered. Competing volatility forecasts are also compared to implied volatilities.

Lawrence Kryzanowski, Simon Lalancette, Minh Chau To, Performance Attribution using an APT with Prespecified Macrofactors and Time-Varying Risk Premia and Betas, The Journal of Financial and Quantitative Analysis, Vol. 32, No. 2 (Jun., 1997), pp. 205-224