We have already discussed how roll yield can negatively affect the overall return of a commodity holding The impact of contango or backwardation can be relatively large compared to the overall return.
Petroleum has unfortunately been in the news lately. Nevertheless, Crude Oil performance last year gave us a good illustration of the impact that contango/backwardation can have.
Crude Oil – 2009
Crude Oil’s had a fantastic year in 2009. The spot price bottomed around 35 and topped 80 to finish on a near +100% performance
Many would assume that quick and easy way to double their money was to invest in Crude Oil in 2009 (assuming you could time the top and bottom perfectly). This is without counting the strong effect of contango that would have eaten into the return.
This can be illustrated by the fact that the USO ETF – supposed to reflect the performance, less expenses, of the spot price of West Texas Intermediate (WTI) light, sweet crude oil – did not manage to emulate the levels of performance seen in the Crude Oil spot price in 2009:
This is, of course, because the ETF managers invest in Crude Oil futures and are subject to the same contango, which eats into their returns.
Below is a chart of several prices for Crude Oil in 2009:
- The spot price is the headline price.
- The rolled contract price represents a back-adjusted contract, trading in the front-month contract and rolling over to the next contract before expiry.
- Two other single contracts, with different maturities are also plotted.
All prices are rebased to start 2009 at the same level as the spot price.
Whereas you cannot trade the spot price, you can trade using the 3 other contracts. Note how they all underperform the spot price performance: this is contango in action.
Most people call this a negative roll yield (long positions in a market in contango or short positions in a market in backwardation) because the drift is more apparent at time of rolling over to the new contract (which is priced dearer than the current contract), however the decay induced by the contango is gradual and erodes the price regularly – as can be seen in the further-dated contracts (the premium priced in the future-dated contract deflates gradually to zero at time of expiry).
The commodity yield curve is clearly an incidental impact on an overall trading strategy results, but it can also be used to form the basis of the strategy itself
Term Structure Trading Strategy
One such example of a strategy using term structure (aka. yield curve) as a trading signal is described in the paper
Tactical Allocation in Commodity Futures Markets: Combining Momentum and Term Structure Signals.
ABSTRACT: This paper examines the combined role of momentum and term structure signals for the design of profitable trading strategies in commodity futures markets. With significant annualized alphas of 10.14% and 12.66% respectively, the momentum and term structure strategies appear profitable when implemented individually. With an abnormal return of 21.02%, a novel double-sort strategy that exploits both momentum and term structure signals clearly outperforms the single-sort strategies.
The authors calculate the roll return for various instruments and select the most backwardated and contangoed markets. Only backwardated markets are allowed to be long and contangoed markets to be short. The increase in performance compared to a standard momentum strategy appears interesting.
Term Structure as a Strategy Filter
Another idea to explore is using term structure as a filter to a Trend Following strategy. Similarly to the concept explained above, one could prevent going short strongly backwardated contracts or long strongly contangoed contracts – basically avoid the markets where the odds are stacked against them.
I have not come across any such published test but found this paper:
It gives interesting fundamental insights as to why markets might be in contango or backwardation and studies the impact or prediction power of the roll yield on a passive long investment strategy. One of their conclusions is that yield return rates start having predictive power when considered on a long-term basis (multi-year) as opposed to monthly measurements.