Time Series Modeling of Cash and Futures Commodity Prices

Commodity prices exhibit differing levels of mean reversion and unit root tests are a standard part of the analysis of commodity price series. Changing underlying means are inherent in commodity prices and can create biased estimates if not correctly specified when performing unit root tests. Prominent financial models include terms for both mean reversion and unit roots but assume that mean reversion occurs gradually over time. Other models such as the popular error correction models require the researcher to determine if prices are either mean-reverting or follow a unit root process. We discuss the models commonly used for commodity prices and how their assumptions align with how commodity spot and futures prices actually behave. We argue for using panel unit root tests for futures prices as they allow for differing underlying means across futures contracts. Cash prices are difficult as none of the currently available models captures their likely stochastic process. Current models, however, can still be useful as close approximations.