Intraday realised volatility forecasting and announcements

This paper examines the importance of macroeconomic announcements, nonlinearity and combining, to realised volatility forecasting in equity-, energy- and commodities-mini-futures markets, by using intraday frequency data. We use three evaluation criteria to detect whether the predictions are more accurate on the out-of-sample announcement days or on the all out-of-sample days. The forecasting evaluation dataset starts from 14 October 2009 to 14 October 2011. The findings indicate that there are some announcements on which nonlinear and combined models forecast realised volatility more accurately in announcement days.