Abstract
The need for research on commodity volatility has grown considerably due to the important role and financialization of commodities in global asset markets. This paper examines the volatility forecasting performance of a wide variety of GARCH-based models in the context of biofuel feedstock markets in the presence of structural breaks. Our sample is also extended to several non-renewable energy commodities to evaluate comparatively the volatility forecasting performance across various commodity markets. The model specifications allow for different conditional distribution functions in the rolling window estimations. A break detection algorithm finds significant evidence of structural breaks in the unconditional variance of all commodity returns under study. The out-of-sample analysis, which is based on an up-to-date model comparison testing procedure, reveals that volatility models accommodating structural breaks in the data provide the best volatility forecasts for most cases. Regarding the relevance of distribution functions, the skewed normal distribution dominates in the model confidence sets. Nevertheless, the complex distribution functions do not always outperform simpler ones, although true return distribution is asymmetric and heavy-tailed.
•This paper investigates the volatility forecasting in the context of biofuel feedstock markets in the presence of structural breaks.•The model specifications allow for different conditional distribution functions in the rolling window estimations.•The out-of-sample analysis is based on an up-to-date model comparison testing procedure.•The results support that volatility models accommodating structural breaks in the data provide the best volatility forecasts.