A multicommodity model of futures prices: Using futures prices of one commmodity to estimate the stochastic process of another

This article proposes a multicommodity model of futures prices of more than one commodity that allows the use of long-maturity futures prices available for one commodity to estimate futures prices for the other. The model considers that commodity prices have common and commodity-specific factors. A procedure for choosing the number of both types of unobservable-Gaussian factors is presented. Also, it is shown how commodities with and without seasonality may be jointly modeled and how to estimate the model using Kalman filter. Results for the West Texas Intermediate-Brent and for the West Texas Intermediate-unleaded gasoline models presented show strong improvements over the traditional individual commodity models, with much lower out-of-sample errors and better volatility estimates, even when using fewer factors. (c) 2008 Wiley Periodicals, Inc.