Threshold Autoregressive Models of the Commodities Futures Basis

The focus in this paper is on the time series dynamics of the basis for commodity futures. These have special interest since regulation of commodity markets is much laxer than is typical for stock markets. However, although such futures contracts have been traded for several decades, they have been subject to far less scrutiny that stock index futures. We redress the balance by assessing the controversial hypothesis of Keynes and Hicks that such markets will typically exist in a state of backwardation, occurring when the contemporaneous futures price is below the expected future spot price. A two-regime threshold autoregressive (TAR) model of the basis is applied to five heavily traded agricultural commodities: cocoa, coffee, corn, oats and wheat. Moreover, a novel step methodology is used to estimate the expected future spot rate and account for the unobserved convenience yield. Over the period 1990 to 2005 statistically significant evidence is found supporting Keynes' theory. Strikingly, the great majority of series in contango regime are mean reverting, whilst those in the backwardation regime behave like a random walk. We conclude that significant differences in basis dynamics are present among differing regimes