The recent fluctuations in commodity prices heavily affected Oil & Gas (O&G) companies’ returns. However, integrated O&G companies are not only exposed to the downturn of the oil prices since a high level of integration allows to obtain non-perfectly positive correlated portfolio. This paper aims to test several different stochastic processes to model the main strategic commodities in integrated O&G companies: brent, natural gas, jet fuel and diesel. The competing univariate models include the log-normal and double exponential jump-diffusion model, the Variance-Gamma approach and the geometric Brownian motion (GBM) with non-linear GARCH volatility. Given the effect of correlation between these assets, we also estimate multivariate models, such as the Dynamic Conditional Correlation (DCC) GARCH, DCC-GJR-GARCH and the DCC-EGARCH. We find that: (i) the asymmetric conditional heteroskedasticity model substantially improves the performance of the univariate jump-diffusion models and (ii) overall, the multivariate approaches are the best models for our strategic energy commodities, mainly the DCC-GJR-GARCH.
Keywords: Commodities, Fat-Tails, Jumps, Maximum Likelihood Estimation.
This paper studies the volatility dynamics of futures contracts on crude oil, natural gas and electricity. To accomplish this purpose, an appropriate Bayesian model comparison exercise between seven stochastic volatility (SV) models and their counterpart GARCH models is performed, with both classes of time-varying volatility processes being estimated through a Markov chain Monte Carlo technique. A comparison exercise for hedging purposes is also considered by computing the extreme risk measures (using the Conditional Value-at-Risk) of simulated returns from the SV model with the best performance—i.e., the SV model with a t- distribution—and the standard GARCH(1,1) model for the hedging of crude oil, natural gas and electricity positions. Overall, we find that: (i) volatility plays an important role in energy futures markets; (ii) SV models generally outperform their GARCH-family counterparts; (iii) a model with t-distributed innovations generally improves the fitting performance of both classes of time-varying volatility models; (iv) the maturity of futures contracts matters; and (v) the correct specification for the stochastic behavior of futures prices impacts the extreme market risk measures of hedged and unhedged positions.
Keywords: Bayesian econometrics, Commodities, Energy markets, Futures contracts, Markov chain Monte Carlo, Stochastic volatility, Utilities
JEL classification: C11, C52, C58, Q40, Q41