We measure past performance as the average return over different ranking periods (1, 3, 6 or 12 months) and hold the long-short portfolios over different holding periods (1, 3, 6 or 12 months). These permutations result in 16 momentum strategies.
We implement our momentum strategies on 31 commodity futures. We include in our analysis 5 oil and gas commodities (heating oil, light crude oil, natural gas, regular gas, unleaded gas), 6 metal commodities (aluminum, copper, gold 100 oz, palladium, platinum, silver 1000 oz), 13 agricultural commodities (cocoa, coffee C, corn, cotton #2, milk, oats, orange juice, soybean meal, soybean oil, soybeans, sugar #11, wheat, white wheat), 4 livestock commodities (feeder cattle, frozen pork bellies, lean hogs, live cattle), diammonium phosphate, lumber and western plywood. The sample starts in 31st January 1979 and ends in 30th September 2004.
Fig.1 presents the average returns that our momentum strategies generate for different ranking and holding periods. With 13 out of 16 strategies having positive and significant average returns, it is clear from the graph that momentum strategies are profitable in commodity futures markets. With an average return at 15%, the strategy that ranks commodity futures based on their performance over the past 12 months and holds the momentum portfolio for 1 month is the most profitable.
Vice versa, the strategy that ranks commodity futures based on their past 12 months average returns and hold the portfolios for 12 months, does not seem to be working (average return of -3% a year). The 13 strategies that show significant performance at the 10% level offer an average return of 9.4%. This is quite outstanding given that over the same period an equally-weighted portfolio of commodity futures lost 2.5%! This shows that tactically allocating wealth towards the best performing commodities and away from the worst performing ones generates a much higher return than a simple buy-and-hold strategy.
The possibility remains that the momentum profits are a compensation for systematic risks. We address this concern by looking at the risk-adjusted returns of the strategies; namely, the return that the strategies generate once we have factored out the risks of the equity, bond and commodity markets. Fig.2 indicates that the risk-adjusted returns are of the same magnitude as the mean returns of Fig.1.
We can therefore conclude that the momentum profits are not a compensation for systematic risk. The strategy with a 12 month ranking period and a 1 month holding period is the one that performs the best on a risk-adjusted basis. It earns a staggering 16% average return net of risks (or 16% alpha).
Which contracts do the momentum strategies recommend trading? We attempt to answer this question by relating the buy and sell recommendations of the momentum strategies to the propensity of the contracts to be in backwardation or in contango.
With this in mind, we create for each commodity a position dummy that assigns a positive value to the commodity when it is bought, a negative value to the commodity when it is sold and a value of 0 if the strategy ignores the commodity altogether. We then relate the position dummies to the roll-returns2 in an attempt to find out if there is a relation between (1) the commodities that are in the long-short portfolios and (2) backwardation and contango. Fig.3 presents the correlations between the position dummies and the roll-returns. A positive correlation indicates that when the roll-return is positive (namely, the market is in backwardation), the momentum strategy buys the commodity and when the roll-return is negative (namely, the market is in contango), the momentum strategy sells the commodity. Clearly, the correlations are for the most part positive (they are equal to 36% on average).
This indicates that the momentum strategy buys backwardated contracts and sells contangoed contracts. This makes a lot of sense as a backwardated commodity gives a positive return tomorrow if bought today, while a contangoed commodity gives a positive return tomorrow if sold today. Likewise, the momentum profits identified in Figs.1 and 2 are generated from long position in backwardated markets and short positions in contangoed markets. This suggests that a strategy that buys the most backwardated contract and shorts the most contangoed one is likely to be profitable. We see this as an interesting avenue for future research.