Trilogy for Troubleshooting Convergence: Manipulation, Structural Imbalance, and Storage Rates

Transcription

1 Trilogy for Troubleshooting Convergence: Manipulation, Structural Imbalance, and Storage Rates Scott H. Irwin 1 Paper prepared for presentation at Protecting America s Agricultural Markets: An Agricultural Commodity Futures Conference, April 5 6, 2018, Overland Park, Kansas. Abstract: Historically unprecedented episodes of non-convergence occurred during in Chicago Board of Trade (CBOT) corn, soybean, and wheat futures contracts and Kansas City Board of Trade (KCBOT) wheat futures contracts. A trilogy of explanations has been offered to troubleshoot these episodes manipulation, structural imbalances, and low storage rates. Theoretical and empirical analysis shows that convergence failures were generated by a disequilibrium between the higher market value of storage in the physical market for grain compared to the storage rate paid to holders of the delivery instrument for grain futures contracts. How to adjust storage rates higher in recognition of this market reality is a highly contentious issue in the grain industry. Key words: convergence, delivery, grain futures, storage, VSR JEL categories: D84, G12, G13, G14, Q13, Q41 1 Scott H. Irwin is the Lawrence J. Norton Chair of Agricultural Marketing, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign. Three people deserve special thanks for their assistance. Aaron Smith helped with updating the price of storage computations. Fred Seamon answered numerous questions about grain futures contract specifications and collected the updated data on shipments and stocks. Hongxia Jiao helped with updating the futures and cash price data. Correspondence can be directed to Scott Irwin. Postal address: 344 Mumford Hall, 1301 W. Gregory Dr. University of Illinois at Urbana-Champaign, Urbana, IL Phone: (217) , sirwin@illinois.edu.

2 1. Introduction Convergence of futures and spot (cash) markets during the delivery period is a bedrock principle of commodity futures markets. In a competitive market, arbitrage should force the futures price at expiration to equal the cash price. Pirrong, Haddock, and Kormendi (1993, p. 11) note that, The emphasis on convergence stems from a belief that the prime function of any forward market is to enable producers, processors, and merchants to hedge to reduce the price risk that they face and to facilitate price discovery. Achievement of both objectives requires the futures price and spot prices to be closely related. In short, the central functions of a futures market are threatened without convergence of futures and spot prices during delivery. As shown in Figures 1 through 4, Chicago Board of Trade (CBOT) corn, soybean, and wheat futures contracts and Kansas City Board of Trade (KCBOT) wheat futures contracts, respectively, experienced several episodes of non-convergence since the mid-1980s. 2,3 The most severe episodes were concentrated in , with non-convergence frequently well outside any reasonable bound based on the cost of delivery. Non-convergence also exceeded 50 cents per bushel in all four markets at least once. The poster child for non-convergence during this period was CBOT wheat. In September 2008, the CBOT wheat futures price expired an astounding $2.50 above the cash price at the cheapest-to-deliver location. The duration and magnitude of these convergence failures was unlike anything seen in the modern record of grain futures markets. Heated public and academic debate ensued as to the possible causes of the convergence failures. Many blamed new financial index traders in grain futures markets. For example, a widely-publicized report by the United States Senate Permanent Subcommittee on Investigations (USS/PSI 2009) claimed commodity index trading caused the non-convergence in wheat markets. The USS/PSI report maintained that index fund capital overpowered the normal functioning of delivery arbitrage. Others argued that the grain futures markets were broken and questioned whether the contracts could remain a useful hedging tool. There were equal concerns that the price discovery function of agricultural futures markets was seriously threatened, which would have ripple effects through the agricultural sector, as the forward contracts generally preferred by grain producers to hedge downside 2 While the both the CBOT and KCBOT are now part of the CME Group, Inc., the CBOT and KCBOT remain self-regulatory organizations that are approved by the Commodity Futures Trading Commission (CFTC) to list corn, soybean, and wheat futures contracts for trading. For this reason, we refer to the original exchanges associated with the respective contracts. 3 See the Data Appendix for details on the construction of the basis series shown in Figures

3 price risks are priced off of futures. Despite these fears, average daily trading volume in the CBOT corn, soybean, and wheat contracts doubled between September 2005, when nonconvergence first appeared, and September 2008 when non-convergence was at its worst. The CBOT and KCBOT made various changes to grain futures contract specifications in an attempt to address the non-convergence problems. The number of warehouse receipts and shipping certificates that a trader could hold was limited, delivery locations were expanded for CBOT wheat, and the KCBOT changed to a seasonal storage rate system for its wheat contract. By far the most fundamental change was the implementation of a variable storage rate (VSR) rule for CBOT wheat beginning with the September 2010 contract (Seamon, 2009). The VSR is keyed to the level of calendar spreads (difference in price across futures contract maturities on a given date) in the period immediately preceding the expiration of the nearby contract. This change was highly contentious in the grain industry when it was proposed, and by all indications, has remained controversial. This discussion makes it clear that non-convergence has been a major issue in grain futures markets for now well over a decade. Three general categories of explanations have been offered to troubleshoot these large non-convergence episodes. The first is manipulation in the form of traditional corners and squeezes. The second is a structural imbalance in contract design or market conditions that favors one side of the market. The third is futures contract storage rates that are set below the market-clearing price of storage in the physical market. The purpose of this paper is to review non-convergence episodes in grain futures markets since 2005 and determine which of the trilogy best explains recent problems. Without this determination, the risk is that changes to grain futures contracts will not address the underlying problem. We begin with a discussion of the delivery system for grain futures markets. 2. Delivery in Grain Futures Markets For grain futures contracts with physical delivery, such as CBOT corn, soybeans, and wheat and KCBOT wheat, the contract terms include the deliverable grades, delivery territory, and period of delivery. This is a consequence of the fact that futures contracts are standardized forward contracts with all terms fixed except price (Peck, 1985, p. 11). The delivery process ties futures and cash prices together. In a perfect market with costless delivery at one particular location and date, arbitrage should force the futures price at expiration to equal the cash price. A well-designed contract will involve few actual deliveries because the terms of the contract balance the commercial interests of long and short futures position holders. As Hieronymus (1977, p. 340) notes, A futures contract is a temporary substitute 3

4 for an eventual cash transaction. In markets that work, delivery is rarely made and taken; futures contracts are entered into for reasons other than exchange of title. The first key aspect of the delivery process for grain futures contracts is that delivery is not satisfied directly by physical grain, but instead by delivering a warehouse receipt in the case of KCBOT wheat or a shipping certificate in the case of CBOT corn, soybeans, and wheat. 4 This is in contrast to other commodity futures markets, such as WTI crude oil, that use a demand certificate system. In this alternative system, futures contracts are essentially transformed into short-term forward contracts during the delivery period and the seller (short) must make physical delivery of the commodity to the buyer (long). This is sometimes referred to as a forced load out delivery system. A warehouse receipt is a legal document that provides proof of ownership (title) of a certain grade and quantity of a commodity at a given storage facility; e.g., 5,000 bushels of number one hard red winter wheat in a warehouse in Hutchinson, Kansas. Crucially, warehouse receipts used in the futures delivery process are negotiable, and thus transferable between parties. A shipping certificate is also a legal document, but rather than representing actual grain in storage, it gives the holder the right but not the obligation to demand load-out of the designated commodity from a particular shipping station; e.g., 5,000 bushels of number two yellow corn loaded on a barge at a shipping station on the Illinois River at LaSalle, Illinois. 5 The advantage of a shipping certificate is the flexibility it offers to makers of delivery because the grain can be sourced over time and space. Like warehouse receipts, shipping certificates are transferable. Neither warehouse receipts nor shipping certificates have expiration dates, and hence, can be held indefinitely. Only commercial firms approved by the CBOT and KCBOT as regular for delivery are allowed to issue warehouse receipts or shipping certificates. Firms must meet certain exchange requirements to be eligible for regularity, such as a minimum net worth of $5 million, and have storage warehouses or shipping stations within the delivery territory of the futures contract. Regular firms are the source of all delivery instruments for their designated warehouses or shipping stations. This means that a regular firm that is short is the only party that has the ability to make an original delivery with a newly issued delivery instrument. The exchange does not allow non-regular firms to issue delivery instruments because there is no guarantee that these other firms have access to sufficient physical commodity and 4 CBOT corn and soybean delivery was based on warehouse receipts prior to the March 2000 contract. CBOT wheat delivery was based on a warehouse receipt prior to the July 2008 contract. 5 In the case of a shipping certificate, title to the grain does not change hands until load out of grain occurs at the shipping station. 4

5 financial resources to complete the delivery process. If firms were to promise delivery and not follow through, the contract would quickly fail. Regular firms are typically large commercial grain firms, such as Cargill, Bunge, and Archer Daniels Midland. For makers of delivery that are not a regular firm, he/she must buy a receipt or certificate from a regular firm, another holder of a receipt or certificate, or have taken delivery on a previous long futures position. Regular firms issuing delivery instruments must either: (i) have an equal quantity of grain in storage at the delivery location, (ii) have an equal amount of grain in storage at a facility near the delivery location, or (iii) be able to source the grain as needed to fulfill their contractual obligation. In the case of (iii), the exchange enforces this requirement by stipulating the maximum number of delivery instruments that a firm may issue. The maximum for a firm issuing warehouse receipts is determined by the amount of warehouse storage space it controls, while the maximum for a firm issuing shipping certificates is based on the loading capabilities at registered shipping stations. These restrictions imply that firms holding short futures positions cannot make unlimited deliveries. Grain futures contracts traded at the CBOT and KCBOT specify a par delivery grade and location for each contract. The par grade is No. 2 yellow corn for CBOT corn, with premiums and discounts for grades above and below No. 2. The par grade is No. 1 yellow soybeans for CBOT soybeans, with discounts for grades below No. 1. The delivery territory for corn futures contracts is the section of the Illinois River from terminals in Chicago and Burns Harbor, Indiana south to Pekin, Illinois and St. Louis. For soybean futures contracts the delivery territory is extended an additional 200 miles along the Illinois and Mississippi Rivers to St. Louis, Missouri. Chicago and Burns Harbor deliveries occur at par for both corn and soybeans, with discounts for deliveries on the Illinois River running between 2 and 6 cents per bushel. St. Loius is at a premium to par for both corn and soybeans. The par grade is No. 2 soft red winter wheat for CBOT wheat, with a 3 cent per bushel premium for No. 1 soft red winter and premiums for other classes of wheat. In July 2009, the delivery territory for the CBOT wheat futures contract was expanded from facilities in Chicago, Toledo, and St. Louis to include facilities in a 12-county area of Northwest Ohio, the Ohio River from Cincinnati to the Mississippi River, and facilities on the Mississippi River from south of St. Louis to Memphis. Par locations for CBOT wheat are Chicago, Toledo, and Ohio River facilities. The par grade is No. 2 hard red winter wheat for KCBOT wheat, with a premium for No. 1 hard red winter. Warehouses in Kansas City are par delivery locations for KCBOT wheat, with warehouses in Wichita, Hutchinson, and Salina/Abilene deliverable at discounts of 6 to 9 cents per bushel. 5

6 The initiation of the actual delivery process of CBOT and KCBOT grain futures contracts is made at the discretion of the short position holder. The long position holder, however, can force delivery by refusing to offset his or her futures position until expiration of the futures contract is imminent. The delivery process consists of a three-day sequence: 1) Intention Day where the short declares their intention for delivery to the clearinghouse, 2) Notice Day where the Clearinghouse notifies the oldest outstanding long position holder with an invoice for delivery, and 3) Delivery Day where the seller and the buyer exchange delivery instruments and payment. The first three-day sequence can be initiated two business days before the first business day of the expiration month and the last three-day sequence can be initiated on the business day prior to the 15th calendar day of the expiration month. This results in a total delivery period of about 10 business days for each contract. The preceding discussion highlights the flexibility built into the delivery process for CBOT and KCBOT grain futures contracts, as delivery can occur on multiple days, with different grades, and at various locations. Standard arbitrage theory predicts that delivery for a grain futures contract will occur at the cheapest-to-deliver location, grade, and date within the delivery period, as this will provide makers of delivery (shorts) the lowest cost alternative for sourcing the grain to satisfy delivery obligations (Stulz 1982; Johnson 1987). 6 If futures are above the cash price, the cash commodity is bought, futures sold, and delivery made. If the cash price exceeds futures, then futures are bought and the buyer stands for delivery. This type of arbitrage should prevent the law of one price from being violated. However, both longs and shorts involved in the delivery process incur costs, which in turn determine arbitrage bounds for the convergence of cash and futures prices at cheapest-todeliver times, grades, and locations. Irwin et al. (2011) estimate the direct cost of delivery for CBOT grain futures contracts to be 6 to 8 cents per bushel; so, a contract can be considered to have converged if cheapest-to-deliver date, grade, and location basis is above or below zero by 6 to 8 cents. There is one final feature of the delivery system for CBOT and KCBOT grain futures contracts that is important to note. Because grain is costly to store, a long taking delivery on a grain futures contract incurs a storage cost for as long as it holds the delivery instrument. This is obviously the case for a warehouse receipt because it represents title to grain in store. It is also true for a shipping certificate because grain must be held by the short or be readily sourced if load out is demanded by the long. The fee is assessed daily and the rate is set by CBOT or KCBOT rule rather than by the market. This daily storage rate is actually the maximum that can be charged by regular firms, but there is little evidence that 6 The value of these delivery options to the short (timing, location, and grade) in grain markets may vary over time (Hranaiova and Tomek 2002; Hranaiova, Jarrow, and Tomek 2005). 6

7 regular firms have ever charged less than the maximum fee to takers of delivery. Historically, the storage rate on grain futures contracts has been fixed for long periods of time. For example, the maximum storage rate for CBOT corn futures contracts was fixed at 15/100 of a cent per bushel per day (about 4.5 cents per month) from the early 1980s until June 2008, with a brief period of a lower rate in The CBOT broke with its long practice of fixed storage rates and implemented a VSR rule for wheat beginning with the September 2010 contract (Seamon, 2009). Under VSR, if the average spread between the expiring and next nearby contract during the specified averaging period is more than 80 percent (less than 50 percent) of full financial carry then the daily storage rate is increased (decreased) by 10/100 of a cent for the next nearby contract. If the average spread between the expiring and next nearby contract during the averaging period is between 50 and 80 percent of full financial carry then the daily storage rate remains the same. There is no upper limit for the maximum allowable storage rate under VSR, but the minimum rate is 16.5/100 of a cent per bushel per day (about 5 cents per month). The KCBOT implemented a similar VSR rule for wheat starting with the March 2018 contract. 3. Troubleshooting Recent Non-Convergence Episodes With this background on CBOT and KCBOT grain futures delivery systems in hand, we can troubleshoot the non-convergence problems in grain futures markets since We are specifically interested in which of the trilogy of explanations can best account for these episodes. The first explanation is manipulation in the form of traditional corners and squeezes. The second explanation is a structural imbalance in contract design or market conditions that favors one side of the market. The third explanation is futures contract storage rates that are set below the market-clearing price of storage in the physical market Manipulation Grain futures markets have a long history of idiosyncratic pricing anomalies that have arisen due to market manipulation in the form of corners and squeezes. (e.g., Paul, 1976; Hieronymus, 1977; Pirrong, 2004), so it is a logical starting place when attempting to explain the non-convergence problems in grain futures markets since In a classic manipulation, a trader or group of traders acquire market power by building up large long positions in futures and the cash market at delivery locations. Once having cornered the market, the trader or group of traders can use their market power to squeeze the shorts in the market and force prices during the delivery period to be much higher than otherwise would be the case. As shown in Figure 5, the classic signature of these episodes is a short-run artificiality in: i) the level of expiring futures prices compared to cash prices in the delivery area, ii) the level of cash prices in the delivery area relative to more distant cash prices, and iii) the level 7

8 of expiring futures prices compared to prices for the next to expire futures contract. The artificiality seldom lasts more than one contract cycle because it is difficult to prevent additional supplies from being moved into deliverable position. It is also why secrecy and surprise is important to a successful corner and squeeze. Reviewing Figures 1 through 4, it is easy to dismiss traditional manipulation as an explanation for non-convergence problems over The magnitude and persistence of non-convergence in all four markets was too large to be explained by a series of corners and squeezes. It simply defies belief that the kind of artificiality shown in Figure 5 could explain the non-convergence episodes. As a result, it is not too surprising that manipulation of the traditional type was never seriously considered as an explanation for the convergence failures over Structural Imbalance The next possible explanation for the non-convergence problems is a structural imbalance in contract design or market conditions that systematically favors one side of the market. Hieronymus again (1977, p. 341) provides an important perspective: Delivery on futures contracts is a sampling of value process. The objective is to get a representative sample. There must be a sufficient amount of the commodity move to and through the delivery points that no one can control and distort the price. The amount must be large enough that the price is representative of the value of the commodity generally so that the relationships with prices at other points of commerce are rational. If these conditions do not hold, Hieronymus (1977, p. 340) warns, When a contract is out of balance the disadvantaged side ceases trading and the contract disappears. Exchanges are certainly aware of this critical dimension of contract success and invest considerable resources in structuring and modifying contracts to reflect commercial market activities and changes in hedging effectiveness. The initial response to the non-convergence problems of by exchange staff, market participants, regulators, legislators, and academic researchers was to focus on potential structural problems with grain futures markets. Much of the discussion centered on the emergence of large-scale participation by a new type of speculator in grain futures markets financial index investors. These investors desired long-only exposure to an index of commodity prices for portfolio diversification, inflation hedging, and return enhancement. Aulerich, Irwin, and Garcia (2013) document the huge growth of the participation by index 8

9 investors in grain futures markets. For example, they report that the net long position of index investors in CBOT wheat increased from an average of 25,702 contracts in 2003 to 134,408 contracts in 2005, over a fivefold increase. The rapid growth in CIT positions is also apparent in CBOT wheat as a percentage of total open interest (long), which increased from 25 to 55 percent over the same period. The growth of financial index positions in other grain futures markets was comparable to that in CBOT wheat. By any measure, the growth of financial index investment was a major structural change in the market participants in agricultural futures markets. 7 It was widely argued at the time that the wave of buying pressure from financial index investors created large and long-lasting bubbles in commodity futures prices ( the Masters Hypothesis ). The U.S. Senate s Permanent Subcommittee on Investigations (USS/PSI 2009, p. 2) concluded that this was the cause of non-convergence in the CBOT wheat futures market: This Report finds that there is significant and persuasive evidence to conclude that these commodity index traders, in the aggregate, were one of the major causes of unwarranted changes here, increases in the price of wheat futures contracts relative to the price of wheat in the cash market. The resulting unusual, persistent and large disparities between wheat futures and cash prices impaired the ability of participants in the grain market to use the futures market to price their crops and hedge their price risks over time, and therefore constituted an undue burden on interstate commerce. Accordingly, the Report finds that the activities of commodity index traders, in the aggregate, constituted excessive speculation in the wheat market under the Commodity Exchange Act. In essence, the USS/PSI report concluded that the structural change in market participation associated with financial index investment was so large that it overpowered the normal functioning of delivery arbitrage. Based on these findings, the Subcommittee recommended: 1) the phase out of existing position limit waivers for index traders in wheat; 2) if necessary, the imposition of additional restrictions on index traders, such as a position limit of 5,000 contracts per trader; 3) the investigation of index trading in other agricultural markets; and 4) the strengthening of data collection on index trading in non-agricultural markets. 7 See Irwin and Sanders (2012) for a detailed discussion of this point. 9

10 Spurred on by charges like that in the USS/PSI (2009) report, a rapidly expanding literature has developed to analyze the influence of financial index positions on commodity futures prices. Consider that no less than six review papers related to this topic have been published since 2011 (Irwin and Sanders, 2011; Fattouh, Kilian, and Mahadeva, 2013; Irwin, 2013; Cheng and Xiong, 2014; Will et al. 2016; and Hasse, Zimmerman, and Zimmerman, 2016). Since the heart of the matter according to the USS/PSI report is the existence of large bubbles in grain futures prices, we focus on direct tests for the existence of bubbles. A recent study by Etienne, Irwin, and Garcia (2015) is instructive. The authors applied a new statistical testing procedure to detect and date-stamp bubbles in corn, soybean, and wheat futures markets during The test detects bubble periods based on departures from a random walk process in daily futures prices. Figure 6 is drawn from their study and it plots futures prices for the five grain futures markets included in the study and all statistically significant bubble periods (shaded bars). While bubbles do occur during some high price periods, such as June 2008 in corn and soybeans, there is also no evidence of bubbles in other periods over when prices reached historical highs and nonconvergence was at its worst. This pattern is especially evident in CBOT wheat, where bubbles did not occur at any point during 2008 when prices reached historical highs and non-convergence reached $2.50 per bushel. Overall, Etienne, Irwin, and Garcia (2015) find that grain futures markets experienced price explosiveness only about two percent of the time and when bubbles did occur, they were generally short-lived and small in magnitude. This indicates grain futures markets were occasionally frothy but were not frequently bubbly as the term is conventionally used. This directly contradicts the conclusion of the USS/PSI report regarding the cause of non-convergence. Other recently published studies that test for bubbles in agricultural futures prices also fail to find evidence of the type of large bubbles alleged in the USS/PSI report (e.g., Areal, Balcombe, and Rapsomanikis, 2016; Li et al., 2017). 8 An alternative argument regarding the role of financial index investors is that large index positions did not cause large bubbles in the level of futures prices, but instead caused a structural change in the term structure ( carry ) of grain futures prices. The term structure refers to the calendar spreads between prices for futures contracts with different maturities. 8 Even though there is very little evidence of large bubbles in grain futures prices during this does not necessarily mean that financial index investment did not have any impact in these markets. For example, index investment could have changed the level of risk premiums in the markets. Despite the theoretical logic of this kind of impact, there is limited empirical evidence supporting it, or any other impact for that matter, in agricultural futures markets. See Sanders and Irwin (2017) for a recent summary of previous empirical findings and a comprehensive set of empirical tests on the impact of financial index investment in agricultural futures markets. 10

11 This argument relies upon well-established patterns of hedger and speculator trading in grain futures markets. Petzel (2009, pp. 8-9) provided a useful synthesis of this argument: Seasoned observers of commodity markets know that as non commercial participants enter a market, the opposite side is usually taken by a short term liquidity provider, but the ultimate counterparty is likely to be a commercial. In the case of commodity index buyers, evidence suggests that the sellers are not typically other investors or leveraged speculators. Instead, they are owners of the physical commodity who are willing to sell into the futures market and either deliver at expiration or roll their hedge forward if the spread allows them to profit from continued storage. This activity is effectively creating synthetic long positions in the commodity for the index investor, matched against real inventories held by the shorts. We have seen high spot prices along with large inventories and strong positive carry relationships as a result of the expanded index activity over the last few years. This implies that the initiation of large positions by index funds in a crowded market space was the source of non-convergence, not bubbles per se. The crowded market space argument can be cast in the one-period supply of storage model of Working (1948, 1949). In Figure 7, the market value of physical storage is plotted on the y-axis and the amount of grain storage (inventory) is plotted on the x-axis. The market value of physical storage consists of two components warehousing cost and convenience yield and this determines the expected change in cash market prices. Warehousing cost is the fixed component of physical storage and includes the rental fee for warehouse space, handling and in- and out-charges, and insurance. Convenience yield is the variable component of physical storage and is typically motivated as an option value generated by transactions costs associated with sourcing the commodity (e.g., Kaldor, 1939; Telser 1958; Routledge, Seppi, and Spatt, 2000). Two types of option values are particularly relevant in grain markets. First, having grain in storage allows firms to take advantage of merchandizing opportunities that require immediate access to grain. Second, filling physical storage with with one grain imposes an opportunity cost because that space cannot be used to store another type of grain (Paul, 1970). The first option increases willingness to hold inventory and is likely to have high value when inventory levels are low, whereas the second reduces willingness to hold inventory and is likely to have high value when inventory levels are high. The net of these two option values determines convenience yield. It is usually assumed that the first type is larger than the second, which implies a positive convenience yield that subtracts from warehousing costs. Positive convenience yield can become large enough that 11

12 the market value of physical storage actually turns negative. For example, it is costly for a soybean processor to shut down their plant and the processor may be willing to pay a large premium for having stocks on hand even though the current price of soybeans is high. Petzel s argument boils down to long futures positions of financial index investors being offset by short futures positions of commercial hedgers, who are hedging their long holdings of physical grain inventories. Hence, the buying pressure of index investors ultimately causes an increase in physical inventories because commercials are not willing to take outright short positions. In terms of the storage model depicted in Figure 7, this shifts out the demand for storage curve. Since the supply of storage curve is assumed to be fixed and upward sloping throughout its range, both the level of storage and the market value of storage increases. This hypothesis was thoroughly tested by Irwin et al. (2011) and Garcia, Irwin, and Smith (2015). Irwin et al. (2011) conducted a battery of Granger causality tests between futures spreads, used to measure the market value of storage, and index positions in CBOT corn, soybeans, and wheat and KCBOT wheat over Not a single case of statistically significant and positive causality from index positions to spreads was found. Garcia, Irwin, and Smith (2015) estimated a reduced-form regression model over of the difference between the market value of physical storage and the futures storage rate and a set of conditioning variables that included financial index positions. Again, no evidence of a significant impact of index positions was found for CBOT corn, soybeans, and wheat and KCBOT wheat. Some studies test for a narrower type of spread impact on the part of financial index investors. Specifically, studies examine spread behavior during the window when index positions are rolled from the expiring to the next nearest contract ( the 5-day Goldman Roll ). These studies generally find that find that spreads in agricultural futures markets are either unaffected or narrow following index rolls (Stoll and Whaley, 2010; Aulerich, Irwin, and Garcia, 2013; Hamilton and Wu, 2015; Sanders and Irwin, 2016). The latter result is attributed to a sunshine trading effect (Admati and Pfleiderer, 1991), whereby large traders that preannounce their intentions attract additional potential counterparties. Two studies (Mou, 2010; Brunetti and Reiffen, 2014) report evidence of expanded spreads after index rolls. The previous discussion indicates there is little evidence that the wave of financial index investment that washed over commodity futures markets starting in the mid-2000s unbalanced grain futures markets and created a series of large bubbles and/or increased the market value of storage. However, this does not preclude other forms of structural imbalance in the grain futures markets from playing a role in the non-convergence episodes. As noted earlier, a key consideration is the adequacy of commercial flows through delivery locations, so that sampling of value in the delivery process can occur efficiently and at relatively low 12

13 cost. Figures 8 through 10 show the annual commercial shipments of grain through facilities regular for delivery of CBOT corn, soybeans, and wheat, respectively, from 1975 through Declining commercial activity in Chicago and Toledo led to a change in delivery locations in 2000 for corn and soybean contracts, and the magnitude of commercial activity at delivery locations increased sharply as a result. Shipments for corn have dropped dramatically in the last decade, to the point where commercial flows at delivery locations are comparable to the level when Chicago and Toledo were the main delivery points. In contrast, soybean shipments have stayed roughly constant since the change in delivery locations in Commercial activity through facilities regular for delivery of CBOT wheat was very small before the addition of Northwest Ohio, Ohio River, and Mississippi River locations in For most of the time over , shipments at CBOT wheat delivery locations were less than 30 million bushels per year. Even with the addition of the new locations in 2009 it has been unusual for annual shipments at delivery locations to exceed 100 million bushels. This is not much bigger than corn and soybean flows before the 2000 change in delivery locations. It is interesting to note that concerns about commercial flows at CBOT wheat delivery locations actually stretch back to the 1920s (Gray and Peck, 1981). The fundamental problem is that changes in wheat production patterns, transportation logistics, and trade flows have left the CBOT wheat contract with an increasingly narrow commercial flow of wheat to draw upon in the delivery process. The CBOT has attempted to address the structural problem of declining commercial flows of wheat several times, first with the addition of Toledo as a delivery location in July 1973, then with the addition of St. Louis in July 1993, and most recently with the addition of delivery locations for wheat in Northwest Ohio and at selected Ohio and Mississippi River locations in July The addition of the new delivery locations for CBOT wheat in 2009 did approximately double total commercial flows, but they were still substantially less than corn and soybean flows. Convergence continued unabated after the addition of the wheat delivery locations in 2009; but this was not surprising since concerns about commercial flows for CBOT wheat predated the non-convergence of Further, it was not obvious how flows could have changed enough over such a short period of time to contribute to the convergence failures. Another structural issue is the continuing role of Chicago as a par delivery point for CBOT corn, soybeans, and wheat contracts. Figures 11 shows that total commercial flows of corn, soybeans, and wheat through Chicago delivery locations have shrunk to less than 20 million bushels in recent years, indicating that Chicago is well outside normal commercial flows of 9 Shipments for Toledo wheat from could not be located. Missing observations for these years were replaced by the average level of shipments over and

14 grain in the U.S. The impact of this decline has long been a concern for the hedging effectiveness of the CBOT grain futures contracts. Writing 25 years ago, Pirrong, Haddock, and Kormendi (1993, p. 31) described the problem this way: Prices in two markets can move idiosyncratically, however, when there are no commodity movements between them, if demand shocks in the market are less than perfectly correlated. These idiosyncrasies create basis risk when one of the locations is a futures delivery point. A change in commodity flows such as the decline of the Great lakes/chicago region as a major transshipping point and the concomitant rise of the Gulf therefore may affect price relations dramatically, and consequently increase or decrease the amount of risk hedgers away from delivery points (i.e., out of position hedgers) must bear. Since other delivery locations have been added to all three CBOT grain futures contracts since this was written, it is not a given that Chicago as a par delivery point is materially harming the performance of the contract. If Chicago functions as a safety valve location and rarely serves as the cheapest-to-deliver location, then any performance problems are likely minimized. However, if deliveries still regularly occur in Chicago this indicates that Chicago cash grain prices could have an outsized influence on futures prices and create additional basis risk for hedgers in non-delivery locations. Since the role of Chicago has been a concern for decades, it likely has little to do with the non-convergence episodes. Nonetheless, it is another example of possible structural imbalances in the current design of CBOT corn, soybean, and wheat futures contracts Storage Rates The key role that storage rates for grain futures contracts play in delivery decisions has long been emphasized in the academic literature (e.g., Peck and Williams 1991, 1992; Pirrong, Haddock, and Kormendi 1993; Hranaiova and Tomek, 2002). The importance of storage rates has also been emphasized by the grain industry in response to episodes of non-convergence that predated For example, the CBOT lowered storage rates for CBOT corn and soybean futures contracts in March 2000 when the delivery instrument was changed from a warehouse receipt to a shipping certificate and delivery locations were re-focused on the Illinois River waterway system. This resulted in an extended period of non-convergence in both markets during (see Figures 1 and 2) that ended only after the grain industry prodded the CBOT to restore storage rates back to their pre-march 2000 levels. While there has always been a general awareness of the relationship between futures storage rates, spreads, and non-convergence, the magnitude of the convergence failures in

15 seemed too large to be explained by storage rates that were too low. In addition, most of the debate about the factors driving non-convergence focused on the role of financial index investors. The first major breakthrough occurred when Irwin et al. (2009, 2011) discovered an empirical relationship between calendar spreads and delivery location basis. 10 Following standard practice in the grain industry, Irwin et al. computed the spread in grain futures prices between the expiring contract and the first deferred contract and expressed the spread as a percentage of the full carry for storing grain between the delivery period for the two contracts. Full carry is computed as the sum of interest opportunity costs for storing grain and the maximum storage rate that can be charged under the contract specifications. Irwin et al. found that non-convergence systematically appeared in CBOT corn, soybean, and wheat futures markets whenever nearby spreads began to exceed about 80 percent of full carry. This result is reproduced in Figure 12 for CBOT corn, soybeans, and wheat over March 1986 through December 2017 and KCBOT wheat over March 1996 through December Delivery location basis averaged between 1.8 and 4.1 times higher when the level of the nearby spread exceeded 80 percent of full carry. Based on the initial findings in Irwin et al. (2009), the CBOT instituted the Variable Rate Storage (VSR) rule for its wheat contract in July As discussed in the earlier section on grain delivery systems, this allowed the contract storage rate for CBOT wheat to adjust up and down based on the percent of full carry for the nearby spread. Specifically, if the average spread between the expiring and next nearby contract during the specified averaging period is more than 80 percent (less than 50 percent) of full financial carry then the daily storage rate is increased (decreased) by 10/100 of a cent for the next nearby contract. If the average spread between the expiring and next nearby contract during the averaging period is between 50 and 80 percent of full financial carry then the daily storage rate remains the same. After its implementation in July 2010, the storage rate increased quickly from about 6 cents per bushel to a peak of 20 cents per bushel in July and September The massive non-convergence failures in CBOT wheat before implementation of the VSR rule subsequently disappeared and generally did not reappear. While it was clear after implementation of the VSR rule for CBOT wheat that adjusting contract storage rates upward was the key to solving the non-convergence problems plaguing grain futures contracts during , the underlying market dynamics that created the 10 Aulerich, Fishe, and Harris (2011) provided another important contribution to understanding non-convergence at this time. 11 We follow the procedure used by the CBOT to compute full carry and add 200 basis points to the LIBOR interest rate when calculating interest opportunity costs. 15

16 problem in the first place were not well understood. In particular, it was still not clear precisely how low storage rates could generate non-convergence of the magnitude experienced during this period. Without a clear economic linkage between the two, the case for raising storage rates to fix non-convergence rested on a purely empirical relationship of uncertain foundation. A significant breakthrough occurred when Garcia, Irwin, and Smith (2015) developed a dynamic rational expectations model of commodity storage and showed that the convergence failures were generated by a disequilibrium between the market value of storage in the physical market for grain and the storage rate paid to holders of the delivery instrument for grain futures contracts. The essential insights from Garcia, Irwin, and Smith s model (GIS hereafter) can be illustrated graphically. Specifically, Figure 13 shows the disequilibrium created by low futures storage rates using the same one-period supply of storage model discussed in the previous section. The only difference from the model in the previous section is that a line has been added to represent the storage rate for the grain futures contract. Panel A shows a case with low storage demand and a market value of physical storage below the storage rate on the futures contract. In this case, the spread in the futures market equals the market value of physical storage and the delivery location basis is zero. Panel B shows a case with high storage demand and a market value of physical storage that is higher than the fixed storage rate allowed on the futures contract. In this case, the spread in the futures market expands to the maximum allowed by the futures contract storage rate ( full carry ), but this is still below the market value of physical storage. The futures spread cannot go any higher than full carry; otherwise, risk-free arbitrage would be possible between futures contracts. The only way for the disequilibrium to be resolved is for the delivery location basis to take on a positive value such that the sum of the futures spread and the delivery location basis equals the market value of physical storage. Otherwise, the futures market will offer inventory holders a lower return for storage than is offered in the physical cash market. Another way of saying the same thing is that holders of the delivery instrument can effectively store grain at below-market storage rates. In equilibrium, this storage windfall opportunity will result in futures prices being bid up above prices in the delivery cash market by exactly the amount of the windfall. A simple numerical example is helpful for understanding the impact of low futures storage rates. To begin, assume that the market value of physical storage is 5 cents per month, the futures contract storage rate is 5 cents per month, and expiring futures and the physical cash price at the delivery location are both $5 per bushel, which means the delivery location basis is zero and convergence is perfect. Also, assume that the one-month ahead futures price is $5.05, reflecting both the maximum futures storage rate and the value of physical storage. Whether one buys physical bushels in the delivery cash market and holds for one 16

17 period or buys cash in the delivery market and sells ahead using the one-month ahead futures contract, the return to storage is expected to be 5 cents per month. Now, assume the market value of storage increases to 10 cents per month. In order to assure that the futures market offers the same storage return as the physical cash market, both the expiring futures price and the one-month ahead futures increase by 5 cents, the difference between the market value of physical storage and the futures storage rate. Therefore, the expiring future price increases to $5.05 and the delivery location basis increases to 5 cents. The onemonth ahead futures price increases to $5.10, consistent with the maximum storage rate in the futures market of 5 cents per month. However, if someone now purchases in the delivery location cash market at the going price of $5 and sells at the one-month ahead price of $5.10, they will earn the same market value of physical storage of 10 cents per month as someone who buys and holds in the delivery location cash market. The model in Figure 13 and the previous numerical example are limited to one period in order to simplify the analysis. The analytics are much more complex for the multi-period case, but the essential insight from the GIS model is that the current delivery location basis widens by the expected value of positive wedges between the price of storage in the physical market and the futures contract storage rate. Consider a highly simplified example where the wedge between the price of physical storage and the maximum storage rate is 5 cents per month and this wedge is expected to last for 12 months. The current delivery location basis does not widen by 5 cents, but instead by 60 cents = 5 cents x 12 months in order to reflect the cumulative value of the expected disequilibrium. This is an important insight because it shows how relatively modest wedges between the physical price of storage and the contract storage rate can generate a surprisingly wide delivery location basis if the wedges are expected to persist for a lengthy period of time. GIS conducted econometric tests for CBOT corn, soybeans, and wheat and KCBOT wheat that supported the predictions of their model, with the expected present value of wedges closely mapping the magnitude of non-convergence. These findings can be illustrated graphically with the aid of Figures 14 through 17 for CBOT corn, soybeans, and wheat and KCBOT wheat, respectively. These figures contain two lines for each of the four markets. The red dashed line is the storage rate for each of the four grain futures markets. This rate for CBOT corn and soybeans was fixed between 4.5 and 5 cents per month for almost the entire 1986 through 2017 period. The same was true for CBOT wheat until VSR was implemented in July Storage rates for KCBOT wheat were fixed until July 2011 when a seasonal storage rate was adopted for the contract. Since that time, storage rates for KCBOT wheat between July and December have been 9 cents and 6 cents the rest of the 17

18 year. The blue line in each of the figures is an estimate of the value, or price, of storage in the physical market. It is estimated based on the following relationship: Market Value of Physical Storage = Current Futures Spread + Current Delivery Basis Next Period Delivery Basis. Note that if non-convergence is only expected to last one period, the delivery basis in the next period is expected to be zero, which is exactly the same relationship shown in Panel B of Figure 13. The change in the basis is relevant when the non-convergence is expected to persist for multiple periods, which in practice should be the norm. The actual computation of the estimate of the market value of physical storage proceeds in three steps. First, the estimate of the market value is computed for each day of the delivery period according to the above relationship. Second, estimates for the first five days of each delivery period are averaged. Third, a centered three-contract moving average is applied to the 5-day averages in order to smooth out the noise associated with each delivery period. Positive wedges between the market value of physical storage and the futures contract storage rate can be easily observed in Figures 14 through 17. The largest wedges occurred in , precisely when non-convergence was at its worst, as shown earlier in Figures 1 through 4. GIS conclude that the market value of physical storage was high during this period because inventories were also high, consistent with a high demand for storage. The plot for CBOT wheat (Figure 16) also convincingly demonstrates that the upward VSR adjustments in storage rates after July 2010 followed increases in the market value of storage in the physical wheat market. Some in the grain trade have vehemently argued that the causality was reversed, when the evidence clearly indicates this is not the case. This analysis also shows why non-convergence in CBOT wheat during could not be solved by adding delivery points, limiting the holding of delivery certificates, or forcing delivery load out, as many advocated, but instead, the solution was most effectively addressed by raising storage rates. Another interesting puzzle is why grain market futures trading volume could increase in the midst of the severe non-convergence problems of The GIS model suggests a solution to the puzzle. In short, traders can do the math and add the difference between market and contract storage rates to the delivery location basis. This requires a certain level of market sophistication regarding the relationship between futures prices, cash prices, and storage rates. Nonetheless, some market participants may have lacked the ability to decode the message from market prices, and as a result, may have been very confused about how to interpret market signals. This could have adversely affected stockholding, price discovery, and risk management strategies. 18