This paper proposes a futures-based unobserved components model for the commodity spot price which proves to have superior forecasting ability. The commodity spot price is decomposed into long-term and short-term components, while the futures price is decomposed into expected future spot price and risk premium. Under this model, information from the whole futures curve could be utilized to improve forecasting accuracy of the spot price. Applying this model to oil market data, I find that the model forecasts outperform in multiple dimensions the benchmark of the literature optimal forecast (the random walk) as well as simple futures prices forecasts. The model forecasts overall have smaller error variation over the 20-year sample period, and are also more possible to have smaller absolute error when compared to the benchmark forecasts period by period.
|Name||Discussion Paper in Economics|
|Publisher||University of Aberdeen|
- Spot Price
- Futures Price
- Futures Curve
- Unobserved Components Model
- Stochastic Processes