Hedging with Two Futures Contracts: Simplicity Pays


  • We gratefully acknowledge financial support from FWO and ICM, and help with the data from the USDA. Comments from Van's committee (Michael Brennan, Geert Dhaene, and Gunther Wuyts) have substantially improved earlier drafts. We also received useful comments from Klyuev Vladimir, Eric de Bodt, Michel Levasseur and other participants at the All China Economics International Conference (Hongkong), workshops at K.U. Leuven and ESC/Université de Lille, and the 2010 EFMA conference in Aarhus, and from an anonymous referee for this Journal. All remaining errors are the authors' responsibility. Correspondence: Van Thi Tuong Nguyen.


We propose to use two futures contracts in hedging an agricultural commodity commitment to solve either the standard delta hedge or the roll-over issue. Most current literature on dual-hedge strategies is based on a structured model to reduce roll-over risk and is somehow difficult to apply for agricultural futures contracts. Instead, we propose to apply a regression based model and a naive rules of thumb for dual-hedges which are applicable for agricultural commodities.

The naive dual strategy stems from the fact that in a large sample of agricultural commodities, De Ville, Dhaene and Sercu (2008) find that GARCH-based hedges do not perform as well as OLS-based ones and that we can avoid estimation error with such a simple rule. Our semi-naive hedge ratios are driven from two conditions: omitting exposure to spot price and minimising the variance of the unexpected basis effects on the portfolio values. We find that, generally, (i) rebalancing helps; (ii) the two-contract hedging rules do better than the one-contract counterparts, even for standard delta hedges without rolling-over; (iii) simplicity pays: the naive rules are the best one–for corn and wheat within the two-contract group, the semi-naive rule systematically beats the others and GARCH performs worse than OLS for either one-contract or two-contract hedges and for soybeans the traditional naive rule performs nearly as well as OLS. These conclusions are based on the tests on unconditional variance (Diebold and Mariano, 1995) and those on conditional risk (Giacomini and White, 2006).