High-frequency data, frequency domain inference, and volatility forecasting

Although it is clear that the volatility of asset returns is serially corre lated, there is no general agreement as to the most appropriate parametric model for characterizing this temporal dependence, In this paper. we propos e a simple way of modeling financial market volatility using high-frequency data. The method avoids using a tight parametric model by instead simply f itting a long autoregression to log-squared, squared, or absolute high-freq uency returns. This can either be estimated by the usual time domain method , or alternatively the autoregressive coefficients can be backed out from t he smoothed periodograrn estimate of the spectrum of log-squared, squared, or absolute returns. We show how this approach can be used to construct vol atility forecasts, which compare favorably with some leading alternatives i n an out-of-sample forecasting exercise.