Using High Frequency Data to Calculate, Model and Forecast Realized Volatility

The objective of this paper is to calculate, model, and forecast realized volatility using high-frequency stock-market index data. The approach differs from existing ones in several ways. First, it is shown that the decay of the serial dependence of high-frequency returns on the sampling frequency is consistent with an ARMA process under temporal aggregation. This is important in modelling high-frequency returns and chosing the optimal sampling frequency when calculating realized volatility. Second, as a result of several test statistics for long memory in realized volatility, it is found that the realized volatility series can be modelled as an ARFIMA process. The ARFIMA's forecasting performance is assessed in a simulation study, and, although it outperforms representative GARCH models, it does so with greater complexity and data intensiveness that may not be worthwhile relative to GARCH's simplicity and flexibility.