Modeling Jump and Continuous Components in the Volatility of Oil Futures

Forecasting the volatility of crude oil futures using intraday data

“…For the sake of forecasting, we are only aware of two papers considering intraday data. Martens and Zein (2004) rely on ARFIMA models and, closest to our study, Tseng et al (2009) follow a strategy similar to ABD. The authors fit the HAR-CJ model using the realized-range as an alternative to the so-called realized variance.…”

“…26) where continuous and squared jumps components are separated at different horizons. This is also the model used in Tseng et al (2009) using oil data. Recall that in our work we use the convention in Patton and Sheppard (2011) so that there is no overlap between lags for the different component of the HAR model.…”

“…Second, our results have implications for the management of oil contingent claims as well as for risk management activities (see Giot and Laurent (2003) and Cabedo and Moya (2003)) and portfolio selection where commodities play an increasingly important role in recent years. Third, we investigate the importance of disentangling jumps, that are present in oil futures returns, from the continuous component in line with Andersen et al (2007) (ABD hereafter) for traditional assets and Tseng et al (2009) for oil, as jumps are indeed present in oil price data. This is an important issue as jumps are perceived very differently ('bad volatility') from the Brownian part ('good volatility') and their forecasting power is still under intense debate (see Corsi et al (2010)).…”

Modelling the volatility of TOCOM energy futures: A regime switching realised volatility approach

“…They report that incorporating implied volatility can significantly improve short term (daily and weekly) volatility forecasts, while including other market variables improves long term (monthly) volatility forecasts. Two other studies by Tseng et al (2009) and Ma et al (2017) employ the HAR Realised Range-Based Volatility (HAR-RRV) and its variations to model and forecast crude oil futures prices. Both studies report that HAR-RRV with inclusion of jump and sign components perform better than simple HAR-RV in predicting volatility of crude futures.…”

“…Anderson et al 2007propose a version of HAR-RV model that conditions the RV on "continues RV" and "jump components", known as HAR-RV-CJ, to model volatility of S&P 500, currencies and T-Bonds. Tseng et al (2009) and Sévi (2014) also apply the HAR-RV-CJ to model oil price volatility. To estimate a HAR-RV-CJ, first the realised volatility must be decomposed into continuous and jump components.…”