A multiplicative components model for oil price forecasting

This thesis proposes a new model for forecasting nominal oil prices inspired by the financial volatility literature. We propose a multiplicative components model where the conditional expectation of the oil price is decomposed into two components: a long term component that is derived by market fundamentals and a short term component which takes into account the information in futures prices and how they have departed from actual spot prices in the immediate past. In an extensive out-of-sample exercise we compare the performance of our model to an array of models: random walk, AR(1), ARMA(1,1), VAR(1), VECM(1) and Brent Futures. To assess its out-of-sample predictive ability, we use the unconditional predictive ability test of Diebold and Mariano (1995), the conditional predictive ability test of Giacomini and White (2006) and the model confidence set (MCS) of Hansen et al. (2011). The model outperforms all benchmarks in accuracy using the mean absolute error loss function and has the least bias according to the Mincer and Zarnowitz (1969) test. The results suggest that the multiplicative components model could potentially be a leading forecasting model for oil prices