Deterministic and Stochastic Methods for Estimation of Intra-Day Seasonal Components with High Frequency Data

We introduce a model for the analysis of intraday volatility of exchange rates returns, based on the structural time series methodology. The stochastic seasonal component is useful to model intra-day effects which may be different from one day to the other. The model is estimated with high frequency data for the Deutsche mark-U.S. dollar exchange rates for 1993 and 1996. The structural time series model performs well in terms of coherence with the theoretical aggregation properties of GARCH models, it is effective both in terms of one-period and multi-period forecasting ability and in terms of describing reactions to announcements of US employment reports