Title: On the long memory feature through temporal aggregation
Authors: Aleksandr Pereverzin - University of East Anglia (United Kingdom) [presenting]
Abstract: One of the key features of empirical work with economic or financial time series is that the time series under consideration is often aggregated in time. The effect of temporal aggregation on time series which are characterized by a long memory dynamics is studied. The aim is to investigate if a long memory property of time series is invariant to the sampling frequency or aggregation scheme. We combine the time and frequency domain analysis and generalize the up to date theoretical knowledge about the temporal aggregation in discrete-time long memory ARFIMA processes. Monte Carlo simulation is conducted to validate the theoretical implications about the effects of temporal aggregation on long memory processes and estimating the memory parameter of aggregated series in the time and frequency domains. We concentrate our empirical analysis on the high frequency foreign exchange data. Several various tests are used to investigate the long memory dynamics of the foreign exchange rates absolute and squared returns series on the various levels of temporal aggregation. Our results have implications for financial risk management dealing with volatility modeling and forecasting.