Laissez les bons temps rouler!
One of the great myths of commodity futures trading is the “roll return,” which Bloomberg writes about here (bonus SWP quote–but he left out the good stuff!, so I’ll have to fill that in here). Consider the oil market, which is currently in a contango, with the November WTI future ($45.99/bbl) trading below the December ($46.52/bbl). This is supposedly bad for those with a long ETF position, or an index position that includes many commodities in contango, because when the position is “rolled” forward in a couple weeks as the November contracts moves towards expiration, the investor will sell the November contract at a lower price than he buys the December, thereby allegedly causing a loss. The investor could avoid this, supposedly, by holding inventory of the spot commodity.
The flip side of this allegedly occurs when the market is in backwardation: the expiring contract is sold at a higher price than the next deferred contract is bought, thereby supposedly allowing the investor to capture the backwardation, whereas since spot prices tend to be trending down in these conditions, holders of the spot lose.
This is wrong, for two reasons. First, it gets the accounting wrong, by starting in the middle of the investment. The profit or loss on the November crude futures position depends on the difference between the price at which the November was sold in early October and bought in early September, not the difference between the price at which the November is sold and the December is bought in early October. Similarly, in November, the P/L on the December position will be the difference between sell and buy prices for the December futures, not the difference between the prices of the December and January futures in early December. You need to compare apples to apples: the “roll return” compares apples and oranges.
On average and over time, the investor engaged in this rolling strategy earns the risk premium on oil (or the portfolio of commodities in the index). This is because the futures price is the expected spot price at expiration plus a risk premium. The rolling position receives the spot price at expiration and pays the expected spot price at the time the position is initiated, plus a risk adjustment. On average the spot price parts cancel out, leaving the risk premium.
Second, the expected change in the price of the spot commodity compensates the holder for the costs of carrying inventory, which include financing costs (very small, at present), and warehousing costs, insurance, etc. Net of these costs, the P/L on the position includes a risk premium for exposure to spot price risk, and in a well-functioning market, this will be the same as the risk premium in the corresponding future.
Moving away from commodities illustrates how the alleged difference between a rolled futures position and a spot position is largely chimerical. Consider a position in S&P 500 index futures when the interest rate is above the dividend yield. (Yes, children, that was true once upon a time!) Under this condition, the S&P futures would be in contango, and there would be an apparent roll loss when one sells the expiring contract and buying the first deferred. Similarly, comparing the futures price to the spot index at the time the future is bought, the future will be above the spot, and since at expiration the future and the spot index converge to the same value, the future will apparently underperform the investment in the underlying. But this underperformance is illusory, because it neglects to take into account the cost of carrying the cash index position (which is driven by the difference between the funding rate and the dividend yield). When buys and sells are matched appropriately, and all costs and benefits are accounted for properly, the performance of the two positions is the same.
Conversely, in the current situation using the roll return illogic, the rolled position in S&P futures will apparently outperform an investment in the cash index, because the futures market is in backwardation. But this backwardation exists because the dividend yield exceeds the rate of financing an investment in the cash index. The apparent difference in performance is explained by the fact that the futures position doesn’t capture the dividend yield. Once the cost of carrying the cash index position (which is negative, in this case) is taken into consideration, the performance of the positions is identical.
Back in 1992, Metallgesellschaft blew up precisely because the trader in charge of their oil trading convinced management that a stack-and-roll “hedging” strategy would make money in a backwardated market, because he would be consistently selling the future near expiration for a price that exceeded the next-deferred that he was buying. This “logic” was again comparing apples to oranges. By implementing that “logic” to the tune of millions of barrels, Metallgesellschaft became the charter member of the billion dollar club–it was the first firm to have lost $1 billion trading derivatives.
So don’t obsess about roll returns or try to figure out ways to invest in cash commodities when the market is in a contango/carry. Futures are far more liquid and cheaper to trade, so if you want exposure to commodity prices do it through futures directly or indirectly (e.g., through ETFs or index funds). Decide on the allocation to commodities based on the risk it adds to your portfolio and the risk premium you can earn. Don’t worry about the roll. If you decide that commodities fit in your portfolio, laissez les bon temps roullez!
When you compare ETFs that invest in oil futures differently, the performance is wildly different. For instance, USO is down -27.5% on a 1 year basis, but USL is only down -11.4%. The difference is that USL is much smaller and takes equal weighted positions in the first twelve months of futures contracts.
It certainly makes it seem like how you handle contango has significant implications for performance.
Comment by John Hall — September 26, 2016 @ 3:12 pm
@John-USL should be less volatile because it is an average across the curve.
Can you do a post on whether or not the commodity ETF’s are safe to speculate in. I always thought there was not enough physical gold backing up GLD- for example.
Comment by TomHend — September 26, 2016 @ 5:55 pm
“the futures price is the expected spot price at expiration plus a risk premium”
Hmm. I beg to differ. I would say that the futures price is determined mainly by the current spot price and storage costs.
Comment by 1970jaw — September 27, 2016 @ 1:13 am
Wait, you are arguing that Mettalsgesselshaft blew up because they thought they had a money maker while expected profit was actually zero, since it was all equivalent? No, they blew up since they were massively short contango and once the market flipped from backwardation to contango, the margin calls killed them. In other words, it was the risk of the spread trade, and not the mild expected underperformance. With enough size, you can blow up on any direction, too, whether you are collecting or paying the risk premium.
Comment by Krzys — September 27, 2016 @ 8:55 am
accounting profit versus true economic profit. pesky isn’t it.
Comment by Jeffrey Carter — September 27, 2016 @ 9:19 am
@Krzys-I was retelling the bullshit justification that the head trader (Benson) gave to the rather clueless MG management, not giving my view. Believe me, I know all about what happened to MG, and wrote about it 20+ years ago.
I got in a battle with Merton Miller over this. He had defended the company’s strategy. I said it was a huge spread trade that increased their risk. I won, in the court of informed opinion.
Mert was very sore about this. When I came into his office to push back on what he was saying (this was before I wrote the article) he threw me out: part of the reason I wrote the article was to prove him wrong. (I was a visiting prof at Chicago at the time.) I made a point of sitting next to him at lunch at the Quadrangle Club for the next few months, just to piss him off. Robert Lucas asked me “what’s Mert’s problem with you?” LOL.
@1970jaw. 1. That’s the pure cash-carry arb story. When stockouts are a possibility, it doesn’t work. If it did, commodity pricing would be boring. 2. The statements are not inconsistent, at all. The drift in the spot price relative to the risk free rate plus storage costs is a risk premium. The drift in the futures price is a risk premium. The risk premiums have to be the same to prevent arbitrage.
Very interesting paper. Incredible that Merton Miller got this wrong. What were his arguments?
However, I think that you are a little harsh with the “roll return” guys.
The reason is that most “investors” have a view on the spot price, i.e. they ignore the shape of the futures curve.
The “roll return” argument just states that your true profit over a certain time horizon will differ from the change in the spot price by the “roll return”, i.e. the comulative convenience yield (positive or negative).
And this is correct. If you expect a doubling of the spot price over three months and the futures trades at twice spot, you will not make any money buying an ETF, hence the negative roll return. As you know, investing successfully in a commodity requires not only correct forecasting of the spot price, but an opinion on all the other factors affecting the market price (storage etc.)
But the products are sold based on spot price forecasts.
Also do not forget that products with fixed roll dates such as ETFs or index products are likely to be front run by traders (that’s what happened to all the Commodity index investors a few years back, once they got large enough to matter)thus increasing your transaction costs.
I have a question regarding your “risk premium” argument: Is it clear which sign the risk premium has in commodity markets?
I mean, a positivie risk premium compensates you for being long an asset.
It is not clear to me whether in commodity markets this is always the case. Theoretically at least there could be a risk premium for being short a certain futures contract, in which case the risk premium would be negative. Or am I missing something?
Comment by Viennacapitalist — September 28, 2016 @ 2:32 am
PS
I just read had a look at the paper. It seems that Miller ignored the effect of basis risk on Firm Value, merely focusing on spot price risk (which, of course, was 0 by definiton).
Comment by Viennacapitalist — September 28, 2016 @ 2:39 am
“the futures price is the expected spot price at expiration plus a risk premium”
Is it? I thought it was the price today for delivery in the future.
Comment by Green As Grass — September 28, 2016 @ 5:07 am
@TheProfessor You’re right that USL has lower standard deviation. I calculated it with monthly data since January 2008 and it is 8.5% (not annualized) compared to 9.8% for USO and WTI. USO also has a higher correlation with WTI at 97.5% compared to 96% for USL.
Nevertheless, there is a huge disconnect between spot prices and ETF prices. If you want to profit from rising oil prices, ETFs are not the best way. From the beginning of the year through August, WTI was up 20% but USO was down -5.8% and USL was up 4%. Big difference.
Comment by John Hall — September 28, 2016 @ 2:21 pm
This all seems reasonable: The futures term structure contains a lot of information. Back out storage costs, interest rates, dividends, etc., and eventually you are only left with a risk premium.
But the key question is whether that risk premium is reasonable or not. It seems odd to ignore this cost— “Hey, you should get homeowner’s insurance. Don’t worry about that price—that’s just a risk premium that the insurance company collects!”
Consider VIX futures. Since the end of 2009, your investment in continuously rolling VIX futures would have earned you a return of about -50% per year. That’s a total loss of 99.6%. The VIX index is down about 8% per year over the same time period. The difference is the risk premium– After all, few people want to be short the VIX, since it will spike if the market falls. But I wouldn’t argue that this risk premium is reasonable, given that I’m being charged over 40% per year.
It’s a similar story with natural gas, as the Bloomberg article mentions. UNG, the natural gas ETN that rolls natural gas futures, is down about 98% since inception. Is that a reasonable risk premium? Just because whatever’s left in the roll yield is a risk premium doesn’t mean the premium is one an investor should pay without thinking about it.
Comment by Eric Newman — September 29, 2016 @ 3:35 pm