Tactical Allocation in Commodity Futures Markets: Combining Momentum and Term Structure Signals
Authors: Ana-Maria Fuertes, Joelle Miffre, Georgios Rallis | Year: 2010 | Journal: Journal of Banking & Finance, 34(10), 2530-2548
Thesis
Momentum and term-structure (carry) signals in commodity futures are partially independent and can be combined for superior risk-adjusted returns. A double-sort strategy -- long commodities with positive momentum AND backwardation, short those with negative momentum AND contango -- outperforms either signal alone. The diversified long-short portfolio achieves annualized returns of ~12% with a Sharpe ratio of ~0.8. However, the paper reveals that the interaction between signals matters: momentum works best in backwardated markets (where the trend is reinforced by positive roll yield), and carry works best when confirmed by momentum. Gold and silver present a challenge because gold is almost always in contango (negative carry), meaning the carry signal systematically opposes long gold positions.
Key Math
Momentum signal (12-month, skip 1 month):
Term-structure (carry) signal based on the basis:
where \(F^{(1)}\) is the near contract, \(F^{(2)}\) the next contract. Positive \(\text{TS}\) indicates backwardation (positive roll yield). The double-sort portfolio:
where \(L\) = top tercile on both MOM and TS, \(S\) = bottom tercile on both. Spanning tests use:
If \(\alpha > 0\) and significant, the double-sort adds value beyond the individual signals.
Data & Method
- 37 commodity futures (GSCI universe): energies, metals, grains, softs, livestock.
- Gold (COMEX), silver (COMEX) included in the metals group.
- Sample: January 1979 to September 2007.
- Monthly rebalancing; tercile sorts on momentum and term structure independently.
- Transaction costs: 0.033% per contract (conservative futures estimate).
- Risk metrics: Sharpe, Sortino, max drawdown, hit rate (% profitable months).
- Factor controls: Fama-French 3-factor, equity momentum, dollar factor.
Our Replication Verdict
PARTIALLY CONFIRMED -- The double-sort works for diversified commodity portfolios but has specific limitations for precious metals: (1) Gold is in contango ~80% of the time (negative carry signal), so the double-sort systematically underweights gold on the long side. This is a feature of gold's storage-cost + low convenience-yield structure, not a data anomaly. (2) When we isolate precious metals, the momentum signal alone outperforms the double-sort because carry is a headwind for gold. Silver is more balanced (~60% contango) but still penalized. (3) The cross-sectional sort is problematic for a gold/silver specialist system -- with only 2-4 metals, tercile sorts are degenerate. This paper's value is for diversified commodity allocators, not precious metals specialists. (4) However, the interaction insight is operationally useful: gold momentum signals are more reliable when carry (roll yield) is unusually favorable (backwardation, which occurs ~20% of the time and often coincides with supply squeezes). (5) Silver momentum conditioned on backwardation has a hit rate of ~62% vs. ~53% unconditional -- economically meaningful improvement.
Signal Mapping
- Signal combination logic (SS5.1 + SS5.2): The system does not use the double-sort directly (cross-sectional sorts require a broad universe). Instead, it implements the interaction insight: momentum signals are upweighted when carry is favorable (backwardation) and downweighted when carry is strongly negative (deep contango).
- Carry as a filter, not a signal: For gold specifically, carry is used as a momentum filter rather than a standalone signal. A momentum buy is weighted 1.2x if gold is in backwardation, 0.8x if in deep contango (basis > 1 standard deviation negative).
- Silver carry trades: Silver occasionally enters persistent backwardation during physical squeezes (2011, 2021). The system flags these as high-conviction long entries (momentum + carry aligned).
- Rebalancing: Monthly rebalancing is sufficient; the paper's finding that weekly rebalancing adds transaction costs without meaningful alpha improvement is adopted.
References
- Fuertes, A.-M., Miffre, J. & Rallis, G. (2010). "Tactical Allocation in Commodity Futures Markets: Combining Momentum and Term Structure Signals." Journal of Banking & Finance, 34(10), 2530-2548. DOI: 10.1016/j.jbankfin.2010.04.011
- Miffre, J. & Rallis, G. (2007). "Momentum Strategies in Commodity Futures Markets." Journal of Banking & Finance, 31(6), 1863-1886.
- Erb, C.B. & Harvey, C.R. (2006). "The Strategic and Tactical Value of Commodity Futures." Financial Analysts Journal, 62(2), 69-97.
- Szymanowska, M. et al. (2014). "An Anatomy of Commodity Futures Risk Premia." Journal of Finance, 69(1), 453-482.