A state space forecasting approach to commodity futures trading.

Abstract

State space forecasting originated in the mid-1970's from engineering models based upon the Kalman filter. To date, the application of state space forecasting to commodity and financial markets has been limited. This study examines a system for trading futures contracts using state space forecasts of commodity futures prices. The system is evaluated for a speculative asset (gold) and a nonspeculative commodity (copper). Price forecasts are developed from multivariate state space models, and the variables considered in the models are those which, according to economic theory, may influence price movements. It is demonstrated that the relative importance of the different economic variables changes over time. Through simulated trading of copper and gold futures contracts, it is also shown that significant profits can be generated from the state space forecasting approach, especially when prices are trending upward or downward fairly continuously. Relaxing the position limit of one contract substantially increases profits. The results suggest that the copper and gold futures markets may be inefficient in the semistrong sense.