Modeling intraday volatility: A new consideration

This paper addresses the limitations of Andersen and Bollerslev's sequential estimation approach for modeling an intraday volatility process. A new approach that utilizes the interaction effect between the periodicity and the heteroskedasticity is proposed. Our method improves the subsequent ARCH structure in the sequential method by integrating the filtration (deseasonalization) process and the ARCH process in a united setting and optimizing the model parameters for the raw series instead of the filtered series. The proposed approach is tested by using 10-min returns of the NASDAQ and S&P 500 indexes. Preferences on using our approach for various forecasting horizons are strongly supported. (C) 2011 Elsevier B.V. All rights reserved.