Oil price expectations in explosive phases

Accurate oil price expectations are of great importance for a variety of economic and financial applications. We find that state-of-the-art market-based expectations only weakly outperform a simple no-change benchmark. This gives rise to changing the perspective from an unconditional to a conditional evaluation method. This consideration is relevant when the forecasting methods potentially behave very differently conditional on certain time-varying economic states. Strikingly, it seems that the no-change benchmark outperforms market-based expectations systematically during turbulent market phases. The entertained conditional predictive ability framework allows us to study the role of important state variables for the time-varying performance explicitly. Among these are established variables from the related oil market literature, covering oil price change measures, volatility as well as supply and demand. Additionally, we suggest a novel and complementing indicator for oil price explosiveness. Our results robustly indicate the existence of conditional time-variation. Furthermore, they underline the importance of the new indicator reflecting temporary exuberance and subsequently collapsing oil prices. We find similar results when evaluating expectations obtained from the Energy Information Administration. Besides various robustness checks and extensions, the practical usefulness is further illustrated in an out-of-sample oil price forecasting exercise. Our findings may have consequences for a variety of economic and financial applications e.g. construction of expectational shocks and testing for speculative oil price bubbles.