Roll yield, popularized by the Goldman Sachs Commodity Index (GSCI), has become a source of debate within the commodity trading community. Even the definition of roll yield, a seemingly simple concept, has come into question by some who want to redefine it in unique ways. There are those who send alerts when roll yields go negative and those who make calls for traders to do their “contango homework.” Conventional wisdom holds that negative roll yield is an indicator of lower prices. Academics even claim that roll yield is a source of commodity market return.
This article will attempt to demonstrate, using GSCI (now the S&P GSCI) roll methodology, that roll yield, or more generally, futures curve shape, is not indicative of market returns. Conventionally defined, roll yield is just another measure of the complex character of commodity futures markets.
Additional questions remain as to the GSCI computation methodology and its implications for commodity asset class returns. Those are for another day.
What is roll yield? Simply put, roll yield is the price difference between delivery months in a roll. It is the shape of the futures price curve for a commodity futures contract (see “Contango vs. backwardation,” below).
Roll yield equals the difference in price between the new contract you are buying and the old contract being sold. For example, suppose on Aug. 31, 2012 you were to buy December corn at $7.99 ¾ and sell your existing September corn position at $8.02 ¾. What is your roll yield? It is simply the percentage difference between the two prices (new price/old price -1) or (799.75/802.75) -1 = -0.374%.
This is an example of negative roll yield. The 2012 curve is largely backwardated. The same trade this year has positive roll yield (364.75/359-1 = +1.6%) and this year’s curve is in contango. Over the last five years, corn was backwardated twice, in contango twice and largely mixed in 2010. In that time, corn rose two years and fell two years. Factors driving corn markets are many, varied and complex but, prima facie, curve shape does not look like one of them. Corn is not a special case.
The West Texas Intermediate crude oil futures price curves show a truly varied mix: 2010 clearly in contango, 2013 clearly backwardated and the rest are a mix (see “Crude facts,” below). Prices rose four years and fell one. These charts show the lack of effect of curve shape on long term prices.
What about the short term?
What about the GSCI roll which is monthly in energies and less frequent in other commodities? Does short-term roll yield really have an effect on return? Let’s look at the GSCI roll. We will use the corn and crude oil rolls, which actually are representative of the whole index which includes, of course, many commodities.
GSCI (and most commodity indexes) roll on the fifth to ninth business day of the month. Twenty percent is rolled each day, so, by the last day of the roll period, the entire position is rolled over from the old month to the new month. The GSCI is calculated on the basis of exchange settlement positions.
Corn rolls five times a year, as follows: In February the March contact position is rolled into the May contract; in April, May into July; in June, July into September; in August, September into December and in November, December into March.
From May 2009 to August 2014, the GSCI conducted 27 rolls of its corn position. To measure the effect of roll yield, we first measure the average roll yield for the five-day roll period (using closing prices). Next we measure the actual return of the new position (see “What roll yield,” below). During the June 2014 roll period, the GSCI sold July corn at a five-day average price of $4.481 per bushel. GSCI bought September corn at $4.444. This equals a roll yield of -.81%. On the August roll GSCI sold its September position at $3.568 for a loss of 19.72%.
In general, during the 27 periods shown, corn had negative roll yield eight periods and negative return 16 periods. The correlation coefficient for the GSCI corn roll yield to corn return is negative -0.22, indicating independence or no correlation (see, “What roll yield?” above).
Crude oil rolls monthly in the GSCI. For the May 2009 to August 2014 period, WTI crude oil was rolled more than 64 times. For the 64 roll periods shown, crude oil had negative roll yield only 12 times, yet it had negative return for 27 periods (see “Rolling the dice,” below). Again, there is little apparent causal relationship between negative roll yield and negative return. The correlation coefficient between roll yield and return is a very low 0.126, nearly reflecting independence.
While limited to the two important commodities shown here, results do not significantly differ for most commodities in the GSCI or any long-only commodity index for that matter. In theory, the lack of causality between roll yield and returns is fundamental.
It would take a weird kind of serial dependence for curve shape to matter. In fact, curve shape, roll yield, backwardation or whatever you want to call it, is just one characteristic of the complex nature of futures markets.
George D. Rahal is a longtime commodity trader and founder and principal of Commodity Trading Advisor, Vista CTA LLC. Reach him at email@example.com.