Exchange rate volatility, macro announcements and the choice of intraday seasonality filtering method

Filtering intraday seasonality in volatility is crucial for using high frequency data in econometric analysis. This paper studies the effects of filtering on statistical inference concerning the impact of news on exchange rate volatility. The properties of different methods are studied using a 5-minute frequency USD/EUR data set and simulated returns. The simulation results suggest that all the methods tend to produce downward-biased estimates of news coefficients, some more than others. The study supports the Flexible Fourier Form method as the best for seasonality filtering.