from Econbrowser:
Commodity arbitrage
Scott Irwin is the Laurence J. Norton Chair of Agricultural Marketing at the University of Illinois. He has been doing some fascinating research on the relation between spot and futures prices in agricultural markets that may shed some light on the role of speculation in recent commodity price movements. We are delighted that Scott agreed to share some of the results of his research with Econbrowser readers.
Commodity Arbitrage
by Scott Irwin
The performance of commodity futures markets during delivery periods is normally little-noticed and followed only by a few economists with an inordinate interest in these markets. However, that has changed in recent months due to skyrocketing interest in commodity markets in general and the odd behavior of cash and futures prices during delivery periods for grain futures contracts at the CME Group (formerly the CBOT). by Scott Irwin
First, some background. Most storable commodity futures markets, such as those for corn, soybeans, and wheat, are still based on physical delivery. If I buy one December 2008 corn futures contract and hold the contract to the delivery period (normally the first half of the expiration month), then I, the "long," will receive 5,000 bushels of corn at one of the specified delivery locations from the "short" on the other side of the contract. The short (seller) must supply the corn either out of their inventory or purchase the grain in the cash market. It is important to note that the vast majority of futures contracts are offset before the delivery period and only a small percentage of futures contracts typically are settled by actual physical delivery. However, the delivery process is nonetheless essential as it ties the futures price to the cash price at delivery locations. In a perfect market with costless delivery at one location and one date, arbitrage should force the futures price at expiration to equal the cash price. Otherwise there would be a violation of the law of one price.
In reality, delivery on grain futures contracts is not costless (who really wants corn on a barge at an Illinois River shipping station) and complicated by the existence of grade, location, and timing delivery "options" that have a demonstrated value to sellers of contracts. So it is better to think of a zone of convergence between cash and futures prices during delivery periods, with the bounds of convergence determined by the cost of participating in the delivery process.
This is where things get interesting. The charts shown below are drawn from an ongoing research project to examine the delivery performance of CBOT grain futures contracts over the last several years. (My collaborators on the project at the University of Illinois are Darrel Good, Phil Garcia, and Eugene Kunda.)
The charts show the difference between cash and futures prices (the basis) on the first and last day of the delivery period for corn and wheat futures contracts expiring between December (Z) 2001 and March (H) 2008 and soybean futures contracts expiring between November (X) 2001 and March (H) 2008. (For those readers unfamiliar with the letter designations for contract months, the monthly codes can be found here.) As an example, the soybean delivery location basis for the November 2001 contract (in the Illinois River North of Peoria shipping zone) was -25.75 cents per bushel on the first day of delivery and -16.75 cents/bushel on the last day of delivery. Note that a negative basis means the futures price is greater than the cash price and a positive basis means that the futures price is less than the cash price. Also keep in mind that there are two delivery zones for corn, four for soybeans, and three for wheat. Convergence patterns at the presented location for a given commodity are representative of the convergence patterns at other locations.
In each of the three markets, convergence generally is within reasonable bounds through 2005 (ignoring problems created by hurricane Katrina in September 2005). Starting in early 2006, convergence performance deteriorates in all three markets, reaching a nadir in September (U) 2006 when the cash price of wheat ended up 90 cents below futures on the last day of the delivery period. Corn and soybean contracts recovered somewhat by late 2006 and early 2007, but wheat continued to perform very poorly. It is interesting to observe that these patterns reversed during the remainder of 2007 and early 2008. Now it is the soybean market that is performing the worst, with the cash price 85 cents below futures on the last day of the delivery period for the March (H) 2008 contract.
While the recent convergence failures are dramatic, in isolation each episode is not necessarily damaging to the overall economic functioning of the markets. The real economic damage is associated with increased uncertainty in basis behavior as markets bounce unpredictably between converging and not converging. This is damaging because, as first pointed out by Holbrook Working many years ago, basis in storable commodity futures markets should provide a rational storage signal to commodity inventory holders. If the difference between the current cash price and futures for later delivery is wide (cash well below futures) this should be a signal to store and vice versa. However, this depends on the signal being accurate. That is, the basis should narrow over time thereby earning "the carry" for someone holding stocks of the commodity and simultaneously selling the futures.
The following chart dramatically illustrates the deterioration in basis predictability for soybean futures contracts since 2006.
The x-axis measures the level of the delivery location basis on the day after the preceding contract expires (except new crop November contracts, which start on the first trading day of October). The y-axis measures the change in the delivery location basis from the day after the preceding contract expires to the first day of delivery. If delivery location basis is perfectly predictable, then all points will lie on a line with a slope of -1 that runs through the origin. In other words, if basis is -50 cents/bushel two months before expiration, the change in the basis over the subsequent two months should be +50 cents/bushel. The blue regression line indicates the soybean futures market performed reasonably well before 2006 compared to this theoretical benchmark. The red regression line shows the precipitous drop in basis predictability over the last two years. Not only does basis change by far less than the initial basis (slope = -0.36), the wide scatter of points indicates very little precision in predicting the change. More formally, R2 is an indicator of hedging effectiveness and it drops from a respectable 78% pre-2006 to only 19% post-2006. I believe this is an underlying reason for much of the current uproar over convergence problems.
An obvious question is what caused this mess. One line of thinking is that this reflects a temporary imbalance in the markets due to the extraordinary structural changes going on in commodity markets. While it is certainly true that the markets have gone through once-in-a-generation changes in price levels, extraordinary tight supplies, and unprecedented speculative investment, this still does not answer why arbitrage was unable to bring cash and futures prices together during delivery. It is difficult to envision an increase in delivery costs that would explain the magnitude of convergence failure we have seen. Another line of argument is that the influx of investment in commodity futures by so-called long-only index funds has created bubbles in futures prices. Aside from the fact that there is a seller for every buyer of a futures contract, the fund/bubble argument has a difficult time explaining the fact that convergence problems do not have the same pattern over time in corn, soybeans, and wheat. The third line of argument is that contract specifications need to be changed to increase storage premiums paid by takers of delivery or that takers should be compelled to ship grain instead of holding it in storage. These may be useful changes, but it is not yet clear how these factors could explain the observed convergence failures.
Congress is now interested and the CFTC is holding a hearing on April 22nd. Let's hope that cooler heads prevail and we seek a better understanding of the source of convergence problems before reaching conclusions about what changes are needed or mandated.
