Volatility prediction based on scheduled macroeconomic announcements

We investigate the impact of scheduled macroeconomic announcements to the volatility of exchange rates by introducing a flexible model with the following characteristics. For each macroeconomic index we estimate cutoff points in the surprise component of the announcement that specify the degree the volatility process is affected. This degree is quantified by a jump of unknown size that occurs before the announcement and a jump of unknown size that occurs at the time of the announcement and then dies out exponentially with unknown rate. We make inferences by using a population Markov chain Monte Carlo reversible jump algorithm and we illustrate our methodology by predicting the volatility of exchange rates using fifteen United States macroeconomic announcements. The empirical study includes extensions to multiple time series with many exchange rates and comparisons with competing existing models.