Forecasting volatility with noisy jumps: an application to the Dow Jones Industrial Average stocks

Empirical high-frequency data can be used to separate the continuous and the jump components of realized volatility. This may improve on the accuracy of out-of-sample realized volatility forecasts. A further improvement may be realized by disentangling the two components using a sampling frequency at which the market microstructure effect is negligible, and this is the objective of the paper. In particular, a significant improvement in the accuracy of volatility forecasts is obtained by deriving the jump information from time intervals at which the noise effect is weak.