Long Memory and Structural Breaks in Commodity Futures Basis and Market

Commodity futures market efficiency has commonly been investigated in the standard I(1)-I(0) cointegration framework and it has provided inconclusive and conflicting results. However, recent empirical studies have found that the spot-futures basis is a fractionally integrated or long memory process across asset classes. Given a stationary spot return, this implies market inefficiency (Maynard and Phillips, 2001). The first contribution of this paper is that it examines the impact of long memory basis behaviour on commodity futures market efficiency. The second contribution is that we evaluate the evidence for long memory in the basis using a stochastic multiple break model suggested recently by Bai and Perron (1998, 2003, 2004) since it is argued that long memory behaviour of the basis may be generated by the presence of structural breaks. Daily spot and futures prices are obtained for a range of futures contracts over the 1/1/1996-31/12/2005 sample period: Soybeans, Cocoa, Heating oil, Gold, Live cattle, Feeder cattle, Hog, S&P 500, US Dollars/Pound and US Dollars/Yen. The conventional cointegration approaches to test the unbiasedness hypotheses are first applied including the error correction model and the Brenner and Kroner (1995) framework. The evidence shows mixed results. Parametric and semi-parametric tests and a multiple mean break model are utilized to evaluate evidence for long memory and structural breaks in the basis. Our findings reveal that the statistical properties of the basis can be characterised by both long memory and structural breaks. This is a new finding in the futures market and is not consistent with market efficiency. Finally, our findings also illustrate that nonstorable commodities tend to perform better than storable commodities in terms of price discovery