What kinds of commodity futures portfolio allocation strategies work? In her December 2015 paper entitled “Long-Short Commodity Investing: A Review of the Literature”, Joelle Miffre summarizes recent academic studies that analyze the performance of long-short commodity futures strategies. She focuses on strategies exploiting roll yields, inventory levels, hedging pressure or momentum. She also surveys alternative strategies based on risk, value, liquidity, sensitivity to inflation or skewness, plus some combination strategies. She relies mostly on Sharpe ratio to compare strategies. Based on results from about 50 studies, she concludes that:
- Regarding theory of storage (physical and financing costs) approaches:
- The main conclusion of roll yield research is that equally weighted commodity futures portfolios that are long high roll yields and short low roll yields generate much higher gross annual Sharpe ratios (0.39 to 0.76 over different samples) than long-only portfolios.
- Alternatively sorting commodities based on standardized inventories generates gross annual Sharpe ratio 0.46 during 1971-2010 (but is complex compared to using roll yields).
- Regarding the theory of hedging pressure, portfolios that are long (short) commodity futures for which hedgers are net short (long) generate average gross annual Sharpe ratio 0.51 during 1992-2011. There is, however, some contradictory research.
- With respect to trend following:
- Trend following signals overlap with those generated by the theories of storage and hedging pressure, but there are differences that suggest incremental information.
- Relative momentum portfolios that are long (short) equally weighted commodity futures with the highest (lowest) returns over the past 1 to 12 months generate average gross annual Sharpe ratio 0.50.
- Time-series (absolute or intrinsic) momentum portfolios that are long (short) commodity futures with positive (negative) past returns generate average gross annual Sharpe ratio 0.52. Regression indicate that relative and time-series momentum signaling differs.
- Trend indicators based on moving averages, channels and proximity to 52-week high also
generate attractive gross annual Sharpe ratios.
- Regarding other indicators:
- Betting against beta (measured with respect to a commodities basket) does not work. However, portfolios that are long (short) commodity futures with low (high) total or idiosyncratic volatilities produce attractive results.
- Portfolios that are long (short) commodities that appear cheap (expensive) relative to their respective long-run prices work.
- Portfolios that are long (short) the least (most) liquid commodity futures work.
- Portfolios that are long (short) commodity futures that are most (least) sensitive to changes in inflation rate work.
- Portfolios that are long (short) commodity futures with the most negative (positive) past return skewness work well (gross annual Sharpe ratio 0.78).
- Regarding strategy enhancements:
- Selecting specific contracts with the highest roll yield (rather than nearest and next nearest) enhances gross annual Sharpe ratio (0.68).
- Strategies that allocate based on near-term roll yields but to distant rather than near-term contracts substantially boost gross annual Sharpe ratio.
- Combining momentum and roll yield by buying (selling) winner (losers) with contracts along the term structure that are most backwardated (contangoed) boosts return and reduces volatility.
- Multi-sorts of different indicators enhance performance.
- Using change in roll yield and change in momentum rather than static roll yield and momentum appear to enhance long-short strategies (roughly doubling gross annual Sharpe ratio).
In summary, the body of research on commodity futures investment strategies shows that: (1) portfolio performance depends mostly on inventory levels, roll yields, hedging pressures and past performance; and, (2) long-short strategies generally outperform long-only.
Cautions regarding conclusions include:
- Studies generally present gross, not net, performance. Implementation costs would reduce performance. Studies that do tackle costs may not be realistic for individual investors.
- The review does not address either standalone or inherited/aggregate data snooping bias, which is especially of concern for strategy enhancements.
- Some strategies/enhancements may have capacity issues.
