For the Korean won US dollar exchange rate log-return, the continuous jump decomposition is applied both to the realized variance (RV) for business time of Korean stock trading and to the RV for non-business time, i.e. overnight, of no Korean stock trading. Different dynamics of the business-time jump and non-business-time jump are analyzed. The decomposition proves to produce substantially better out-of-sample forecasts for the exchange rate RV than the existing continuous jump decomposition applied to the whole day RV. The different jump dynamic is analyzed in terms of volatility spillovers from the RV of the KOSPI to the RV of the exchange rate for stock trading time and to that for overnight.
- Continuous jump decomposition
- High frequency data
- Overnight realized variance
- Volatility forecasting
- Volatility spillover