Hedging Spot Corn: An Examination of the Minneapolis Grain Exchange s Cash Settled Corn Contract

Transcription

1 Journal of Agribusiness 21,1(Spring 2003):65S Agricultural Economics Association of Georgia Hedging Spot Corn: An Examination of the Minneapolis Grain Exchange s Cash Settled Corn Contract Dwight R. Sanders, Mark R. Manfredo, and Tracy D. Greer This research examines the potential basis behavior and hedging effectiveness for the Minneapolis Grain Exchange s (MGE s) cash settled corn contract. MGE futures cash settle to the National Corn Index (NCI) calculated by the Data Transmission Network (DTN). Focusing on seven regions in Illinois, the data suggest that NCI futures offer potential advantages over the existing Chicago Board of Trade (CBOT) corn futures. In particular, nearby basis variability could be reduced by 4 per bushel from 8.6 to 4.6 per bushel, and unconditional hedging effectiveness may increase from an average of 79% for the CBOT to 93% for the NCI. These results are statistically significant, and likely to be economically important given that agribusiness firms such as grain merchandisers and country elevators traditionally have very low margins. Key Words: basis behavior, cash settlement, corn futures, new contracts Historically, nearly all hedging of feed grain price risk has occurred in the Chicago Board of Trade s (CBOT s) corn futures market. This market reflects prices at the terminal level (i.e., Illinois River delivery). Therefore, the CBOT corn futures market implicitly contains a transportation component versus country elevator or first-handler prices. Differences arise in the price of corn at the first-handler level vis-à-vis the terminal due to transportation costs from the location of the first handler (e.g., country elevator) to the river terminal, variable quality factors, and local supply and demand conditions (Leuthold, Junkus, and Cordier, 1989). If the embedded transportation cost is volatile, then this dimension of basis risk (i.e., spatial basis risk) is added for those firms hedging prices at alternative points in the supply chain. Furthermore, the contract specifications for the CBOT corn contract, as well as the other grain contracts traded at the CBOT, implicitly give sellers of futures contracts choices regarding delivery timing within the contract month, location of delivery, Dwight R. Sanders is assistant professor, Department of Agribusiness Economics, Southern Illinois University, Carbondale; Mark R. Manfredo is assistant professor, Morrison School of Agribusiness and Resource Management, Arizona State University, Mesa; and Tracy D. Greer is graduate assistant, Department of Agribusiness Economics, Southern Illinois University, Carbondale.

2 66 Spring 2003 Journal of Agribusiness and quality differentials on the grain to be delivered (Hranaiova and Tomek, 2002). In fact, Hranaiova and Tomek find that these delivery options embedded in the CBOT contract may impact basis convergence and thus basis risk. In response to like concerns, the Minneapolis Grain Exchange (MGE) recently introduced a cash settled corn futures contract based on the National Corn Index (NCI). The NCI is calculated by the Data Transmission Network (DTN), reflecting country elevator bids for U.S. No. 2 yellow corn. The NCI futures contract differs from the CBOT corn futures contract in two primary ways. First, the contract is cash settled to the NCI index as opposed to the delivery settlement that underlies the CBOT contract. Cash settlement is appealing because it eliminates embedded delivery options granted to the seller (Hranaiova and Tomek, 2002) and allows for simultaneous expiration of futures and options. Cash settlement also alleviates the need to monitor costly physical delivery processes, which facilitates a monthly expiration cycle. Second, the NCI futures reflect pricing at a different transfer point in the supply chain namely, the first-handler or country elevator level. Therefore, the NCI may provide a more well-behaved basis for producers and local elevators than the existing terminal-level corn futures contract offered by the CBOT. Traditional commodity markets are increasingly focused on quality characteristics and identity preservation (e.g., genetically modified organisms or GMOs). This, in combination with recent changes in hedge accounting practices (i.e., Federal Accounting Standard 133), increases the importance of using hedging instruments that result in highly effective hedges. 1 As these markets evolve, it is crucial that the risk management industry innovate to meet the hedging demands of agribusinesses (see Parcell, 2002). The NCI contract offers an alternative hedging tool for producers and local elevators marketing and merchandising cash corn. It is important that the performance of this and other new risk management tools be evaluated and results communicated to the agribusiness community. Toward this end, the primary objective of this research is to provide such a performance assessment of the Minneapolis Grain Exchange s cash settled corn contract. In doing this, we first examine the NCI and the MGE futures in terms of construction, calculation, and settlement procedures. In order for cash settlement to be effective, the settlement procedure must be free of exploitation and accurately depict cash market prices (Peterson, 1995; Rich and Leuthold, 1993). Moreover, the index used for settlement must be a reliable indicator of the commercial value of the commodity (Garbade and Silber, 2000). Second, we examine the potential basis behavior and hedging effectiveness of the NCI futures contract relative to the existing CBOT corn futures contract. Following traditional regression approaches to hedging performance, such as those used by Ditsch and Leuthold (1996), we focus on hedging spot corn transactions over a one- 1 A full discussion of Federal Accounting Standard 133 is beyond the scope of this paper. However, the standard states that the ineffective portion of a hedge (i.e., basis risk) does not receive traditional hedge accounting treatment; rather, those dollars are marked-to-market in earnings (International Treasurer, 1998). Therefore, it is increasingly important that hedge effectiveness, and the methodology by which it is determined, be well defined and documented by agribusiness firms.

3 Sanders, Manfredo, and Greer Hedging Spot Corn 67 month horizon. Because historical NCI futures price data are not yet available, the underlying NCI cash index is used as a proxy for the futures a common practice in the examination of new cash settled futures contracts (Schroeder and Mintert, 1988; Kimle and Hayenga, 1994; Ditsch and Leuthold, 1996). Furthermore, the cash markets examined are available for seven regions in Illinois, which likely have different basis patterns due to their unique geographical locations. Finally, the results are evaluated and interpreted from the economic perspective of a potential user the country elevator. The simulated basis levels and the examination of hedge ratios together provide a guide to the potential hedging effectiveness of the NCI contract for spot corn transactions in Illinois. Additionally, the research offers a framework for practitioners to evaluate hedge effectiveness in other locations. The information provided by this study should prove valuable to the MGE and the agribusiness industry in terms of expected basis levels, basis variability, minimum variance hedge ratios, and ultimately the potential hedging effectiveness of the NCI futures contract. Furthermore, while most academic research has looked favorably upon cash settled futures contracts in livestock markets (Schroeder and Mintert, 1988; Kimle and Hayenga, 1994; Ditsch and Leuthold, 1996), and the Chicago Mercantile Exchange has two successful cash settled livestock futures in its lean hog and feeder cattle contracts, research concerning cash settled grain futures is scarce (Chaherli and Hauser, 1995). The examination of the NCI provides an interesting research opportunity because it is the first cash settled futures contract introduced for a major U.S. feed grain. It also represents a different risk transfer point in the grain marketing system than existing futures contracts. The results of this research are important as an informational guide to the MGE, potential users of the NCI for hedging spot corn transactions, and to risk management educators, extension agents, and market advisors. Before assessing the potential for the NCI futures, however, it is first helpful to understand the construction of the NCI index and the specifications of the MGE s NCI futures contract. The National Corn Index and MGE Futures Contract 2 The Data Transmission Network (DTN) calculates the National Corn Index, which has been reported daily since The NCI is the simple average price for U.S. No. 2 yellow corn from all bids collected by DTN. These bids are collected primarily through direct telephone calls to elevators, although some bids are received via , fax, and the internet. On average, DTN collects bids from 1,630 elevators nearly 90% of all U.S. elevators. Elevators in seven states Iowa, Illinois, Nebraska, Kansas, Minnesota, Indiana, and Ohio represent 75% to 80% of the bids collected. 2 The information in this section was drawn from the Minneapolis Grain Exchange s website ( on March 21, The specific numbers reflect the MGE s audit of the DTN data collection process on April 23S25, April 30SMay 2, and July 2S5, 2001.

4 68 Spring 2003 Journal of Agribusiness The single largest owner of the corn bids (i.e., elevator ownership) comprises only 3.3% of those collected, and thus no single firm dominates the index. The integrity of the index is maintained through controls put in place by DTN, as well as a form of self-auditing inherent in the bid collection process. For instance, bids falling outside of a designated range are flagged for confirmation. If the bid cannot be confirmed, it is not included in the day s data and subsequent calculation of the NCI. Furthermore, the MGE notes the bid process is largely self-auditing in the sense that DTN provides the individual elevator bids to DTN subscribers. Most DTN subscribers are producers and other agribusiness firms. Consequently, to retain credibility with their customers, elevators are unlikely to report a bid to DTN at which they are unwilling to transact. Indeed, it could prove very costly to an elevator s core business to provide DTN with anything other than its actual posted prices. Given the data collection process and index construction, by all reasonable standards, DTN s NCI appears to meet the requirements for a good cash price i.e., it reflects commercial value and is not prone to manipulation (Peterson, 1995; Rich and Leuthold, 1993). Therefore, the NCI is a valid candidate to underlie a cash settled futures contract. The Minneapolis Grain Exchange s National Corn Index futures contract (NCI futures) cash settles to a simple average of the last three daily NCI prices published during the contract month. The settlement price is rounded to the nearest quarter cent using standard rounding techniques. Cash settlement occurs on the business day following the last trading day of the month. As an example, the March 2002 contract cash settles to a simple average of the daily NCI prices on the 26th, 27th, and 28th of March. Those prices were reported, respectively, as , , and per bushel. Thus, the NCI futures cash settled on April 1st at a price of per bushel. After the last day of trading, all open futures contracts are marked to the underlying cash index value. So if the purchaser (or seller) of a March futures contract held the contract to expiration, the contract would automatically be evened-up at on April 1st. There is no exchange of physical product cash flows are simply exchanged through traders futures accounts. In contrast to the CBOT s corn futures, a contract is listed for every calendar month for NCI futures. The NCI futures trade exclusively on the MGE s new electronic platform, MGExpress, whereas the CBOT contract predominantly trades in a physical trading pit via open-outcry. Data To evaluate the potential usefulness of the NCI futures, cash transaction quotes are collected from an independent third party: the Illinois Department of Agriculture Market News. These data are from a daily survey of roughly 100 country elevators in Illinois over seven geographical regions (see figure 1). The disparity between the geographical regions should provide uniquely different basis levels and potentially different basis behavior. For instance, the Northern and Little Egypt regions are approximately 250 miles apart and can experience very different local supply and demand conditions.

5 Sanders, Manfredo, and Greer Hedging Spot Corn 69 \ / SOUTH CENTRAL SOUTHWEST Source: University of Illinois Marketing and Outlook website. A more detailed, colorized map of these Illinois regions is availabie online at Figure 1. Illinois cash price regions The Illinois spot data are available from June 1997 through May The analysis focuses on a monthly hedging horizon, resulting in 60 observations. Specifically, prices are drawn from the third to the last business day of each month. This corresponds to the first day of the three-day averaging period for cash settlement of the NCI futures. This is the day when the NCI futures should most closely converge with the underlying index before being influenced by the averaging settlement process (Kimle and Hayenga, 1944). CBOT corn futures prices are also collected on this day. The price levels reflect the nearest to maturity futures contract (without entering the delivery month), and price changes are calculated to reflect changes in the price of the nearby contract. Care is taken such that price changes are not impacted by contract rollover.

6 70 Spring 2003 Journal of Agribusiness Price (Cents/Bushel) NCI CBOT Western Illinois Jun-97 Dec-97 Jun-98 Dec-98 Jun-99 Dec-99 Jun-00 Dec-01 Jun-01 Dec-00 Figure 2. Corn price levels, June 1997SMay 2002 Since the NCI futures do not have a history, the underlying NCI must be used as a proxy for the cash settled futures contract. Clearly, the underlying NCI is not a futures price and does not reflect possible carrying charges, premia, or biases which may exist in actual futures prices. Nonetheless, using the underlying index as a proxy for the futures is common in examining the potential performance of a new cash settled contract (Schroeder and Mintert, 1988; Ditsch and Leuthold, 1996). Furthermore, the monthly delivery cycle and cash settlement feature of the futures should result in a predictable convergence of the NCI futures and the underlying index (Kahl, Hudson, and Ward, 1989). Therefore, any bias this creates should be minimal. Figure 2 illustrates three of the time series described above: cash prices from the Western Illinois region, nearby CBOT corn futures, and the NCI. As observed from this figure, it is clear the CBOT terminal-level prices can trade at a significant premium to elevator prices, reflecting the embedded transportation costs. Otherwise, the prices tend to move together as a result of shifting fundamental supply and demand factors in the overall corn market. Methodology and Results Basis Variability The potential performance of the NCI futures contract is determined by examining both basis variability and hedging effectiveness for the various regions in Illinois presented in figure 1. In examining basis variability, the basis is calculated as the

7 Sanders, Manfredo, and Greer Hedging Spot Corn 71 Table 1. Basis Summary Statistics, Seven Illinois Regions ( /bushel), June 1997SMay 2002 Illinois Regions Description Northern North Central South Central Western West/ Southwest Little Egypt Wabash Average A. National Corn Index (NCI) Mean Std. Dev Maximum Minimum!14.27! !8.06!4.82! !4.17 Range B. Chicago Board of Trade (CBOT) Mean!24.58!16.63!15.34!21.49!17.62!10.76!12.41!16.98 Std. Dev Maximum!12.00!2.00!0.50!8.25! !0.68 Minimum!44.50!36.50!34.00!42.00!39.00!38.50!35.00!38.50 Range F-test a a F-test for equality of variance between the NCI basis and the CBOT basis. The NCI basis is statistically less volatile than that of the CBOT for all regions at the 5% level. cash price from the respective Illinois location, CP t, minus the futures price, FP t. The basis summary statistics are presented in table 1. In panel A of table 1, the NCI is the assumed futures price. In panel B, the futures price is the nearby CBOT corn futures. For example, the Western Illinois basis using the CBOT futures (panel B, table 1) has a mean of!21.49 per bushel with a standard deviation of 8.54 per bushel. This is in contrast to the NCI basis for Western Illinois (panel A, table 1), which has a mean of 3.46 and standard deviation of 4.61 per bushel. For the Western Region, the NCI basis is more stable than the CBOT basis, with a standard deviation nearly one-half the size (4.61 versus 8.54 per bushel) and a much smaller range of values (22.72 versus per bushel), where the range is defined as the absolute difference between the maximum and minimum basis values. The above result is relatively consistent across regions. Over the seven regions, the average standard deviation of the NCI basis is 4.59 per bushel, while the average standard deviation of the CBOT basis is 8.56 per bushel, resulting in an average reduction in basis volatility of nearly 4 per bushel. 3 Reduced basis variability is illustrated with time-series plots in the seven panels of figure 3. 3 The basis is also examined in terms of log-relative percentages, where the basis, ln(cp t /FP t ), represents a percentage of the futures price (see Garcia and Sanders, 1996). The average standard deviation of the NCI basis across the seven regions is 2.3%, and the average CBOT basis variability is 5.1%. Using percentages does not change the presented results.

8 72 Spring 2003 Journal of Agribusiness 20 NORTHERN REGION Basis (Cents/Bushel) Jun-97 NCI CBOT Dec-98 Jun-98 Dec-97 Jun NORTH CENTRAL REGION Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Basis (Cents/Bushel) NCI Jun-98 Dec-97 Jun-97 CBOT Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Jun-99 Dec WESTERN REGION Basis (Cents/Bushel) NCI Jun-98 Dec-97 Jun SOUTH CENTRAL REGION CBOT Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Jun-99 Dec-98 Basis (Cents/Bushel) Jun-98 Dec-97 Jun-97 NCI CBOT Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Jun-99 Dec WEST/SOUTHWEST REGION Basis (Cents/Bushel) Jun-99 Dec-98 Jun-98 Dec-97 Jun WABASH REGION NCI CBOT Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Basis (Cents/Bushel) Jun-99 Dec-98 Jun-98 Dec-97 Jun-97 NCI CBOT Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Basis (Cents/Bushel) Jun-99 Dec-98 Jun-98 Dec-97 Jun-97 NCI CBOT LITTLE EGYPT REGION Dec-01 Jun-01 Dec-00 Jun-00 Dec-99 Figure 3. Basis levels, seven Illinois regions, June 1997SMay 2002

9 Sanders, Manfredo, and Greer Hedging Spot Corn 73 Table 2. Delivery Month Basis Summary Statistics, Seven Illinois Regions ( /bushel), June 1997SMay 2002 Illinois Regions Description Northern North Central South Central Western West/ Southwest Little Egypt Wabash Average A. National Corn Index (NCI) Mean! Std. Dev Maximum Minimum!14.27! !8.06! !2.47 Range B. Chicago Board of Trade (CBOT) Mean!22.53!14.25!12.87!19.18!15.53!7.87!9.82!14.58 Std. Dev Maximum!13.00!2.00!0.50!8.25! !3.32 Minimum!40.75!30.75!25.25!34.75!28.75!25.75!22.75!29.82 Range F-test a Note: Delivery basis is for month-end February, April, June, August, and November for the entire sample, plus October and December beginning in October of a F-test for equality of variance between the NCI basis and the CBOT basis. The NCI basis is statistically less volatile than that of the CBOT at the 10% level for all regions and at the 5% level for all regions except the South Central and Wabash. The figure 3 graphs suggest the CBOT and NCI bases share some similar patterns, but the NCI basis is generally more stable. 4 The relative stability of the NCI basis could be partially due to temporal differences. That is, hedges placed in the NCI are always a few days from expiration when lifted, whereas the CBOT hedges may be as much as eight weeks from expiration. For instance, at the end of May, the nearby CBOT future is the July contract, and it is four weeks until first notice day of delivery. At the end of September, the nearby CBOT future is the December contract, and it is eight weeks until delivery. This temporal difference may bias the results toward the NCI contract. One could argue this is simply an advantage of the NCI s contract design monthly expiration cycle and cash settlement. However, it is important to consider the temporal differences to make a fair comparison. To adjust for the temporal factor, we compare relative basis variability only at the end of calendar months where the nearby CBOT contracts are about to enter 4 The CBOT altered the delivery terms of its corn futures contract in calendar year 2000 from negotiable warehouse receipts to shipping certificates at Illinois River points. There is not yet enough data available to accurately determine how this change in delivery specifications may impact basis behavior and the performance of the CBOT futures contract.

10 74 Spring 2003 Journal of Agribusiness delivery. 5 For instance, at the end of February, the CBOT nearby basis is calculated versus the expiring March contract and delivery is only a few days away. Therefore, basis convergence should be nearly complete. Limiting the analysis to this subsample reduces the number of monthly observations from 60 to 29. The results are presented in table 2. The average standard deviation of the NCI basis shows a modest reduction from 4.59 in the full sample to 4.45 per bushel in the subsample. In contrast, the average CBOT basis shows a more meaningful reduction from 8.56 in the full sample to 6.77 per bushel in the subsample. Thus, in months facing immediate delivery, the NCI reduces basis risk by only 2.32 per bushel versus 3.97 in the full sample. So, temporal considerations account for nearly one-half of the NCI s reduction in basis variability. Nonetheless, in both the full sample (table 1) and subsample (table 2), the reduction in variance between the NCI basis and CBOT basis is statistically significant at the 10% level for all regions (based on an F-test). Therefore, although the magnitude of the reduction is smaller in months immediately prior to delivery, the general conclusions are not altered. Clearly, however, it is important to examine and account for temporal considerations when comparing alternative risk management tools. Hedging Effectiveness The second step in evaluating the NCI futures is to look at potential hedging effectiveness. Hedging effectiveness has traditionally been examined with the following first-difference regression model (Leuthold, Junkus, and Cordier, 1989): 6 (1) CP t ' α % β FP t % e t, where CP t is the change in the cash price being hedged over interval t, FP t is the change in the futures price in which the hedge is placed over interval t, and e t is a random error term or residual basis risk; β is the minimum variance hedge ratio, and α measures systematic trends in the cash price. The R 2 from estimating equation (1) is a measure of ex post hedging effectiveness or risk reduction associated with applying the minimum variance hedge ratio. In this research, the measurement interval is monthly, and FP t is represented by changes in the NCI and nearby CBOT corn futures; CP t is the change in the Illinois cash price for a given region. The estimated β is the ex post minimum variance hedge ratio, and the estimated α captures any systematic trends in the basis over the 5 From 1997 through 1999, the CBOT listed delivery months for March, May, July, September, and December. Then, starting in calendar year 2000, the CBOT also listed futures contracts for November and January delivery. These data are included in the data set. Thus, beginning in 2000, there is also a pending delivery settlement at month-end October and December. 6 Augmented Dickey-Fuller tests suggest all of the price series (in levels) contain a unit root and are nonstationary. The price change series and bases series were found to be stationary. Therefore, equation (1) is correctly specified in first differences (Meyers and Thompson, 1989).

11 Sanders, Manfredo, and Greer Hedging Spot Corn 75 Table 3. Unconditional Hedging Effectiveness Regressions: Results from Estimating Equation (1), CP t = α + β FP t + e t, June 1997SMay 2002 Illinois Regions Description Northern North Central South Central Western West/ Southwest Little Egypt Wabash Average A. National Corn Index (NCI) Hedge Ratio, β (Std. Error) (0.046) (0.034) a (0.038) (0.044) a (0.039) (0.045) (0.034) Adjusted R B. Chicago Board of Trade (CBOT) Hedge Ratio, β (Std. Error) (0.070) (0.059) (0.053) (0.064) (0.064) (0.071) (0.067) Adjusted R Note: The symbol ( ) denotes statistically different from 1.00 at the 5% level using a two-tailed t-test. a Estimated with White s heteroskedastic consistent estimator. sample period. The R 2 is a measure of the in-sample hedging effectiveness, and it is analogous to using the simple correlation coefficient as suggested by Ederington (1979). A higher R 2 implies a stronger correlation between cash and futures, and thus a lower residual basis risk. In this study we use price changes (cents per bushel) to facilitate practitioners intuitive interpretation (Shafer, 1993), to reflect the cashflow nature of the hedging process (Brorsen, Buck, and Koontz, 1998), and to avoid econometric issues associated with the nonstationarity of price levels. 7 The results of estimating equation (1) are presented in table 3, where panel A reports the NCI results and panel B reports the CBOT results. Looking at the CBOT results (panel B), the estimated hedge ratios are generally less than 1.00, but none of them are statistically different from unity (two-tailed t-test, 5% level). The adjusted R 2, or hedging effectiveness, ranges from 74.3% in Northern Illinois to a high of 85.4% in South Central Illinois. Turning to the NCI results (panel A), the estimated hedge ratios tend to be larger than unity, with those in the South Central, West/ Southwest, and Wabash regions statistically greater than 1.00 (two-tailed t-test, 5% level). Hedging effectiveness ranges from a low of 90.2% for Northern Illinois to a high of 94.7% for the North Central region. In comparing the NCI and CBOT results, two observations are made. First, the hedge ratios for the NCI are consistently greater than those of the CBOT. 8 However, 7 Each regression is tested for heteroskedasticity using White s test and serial correlation using a Lagrange multiplier test. If the null of no heteroskedasticity is rejected at the 5% level, then the model is reestimated using White s heteroskedastic consistent estimator. If the null of no serial correlation is rejected, then the model is reestimated using the Newey-West estimator, which is consistent in the presence of serial correlation. These procedures allow for efficient estimation of standard errors under a variety of forms of heteroskedasticity and autocorrelation (Hamilton, 1994, p. 283). 8 The mean FP t for the NCI and CBOT corn futures are!0.73 and!4.18 per bushel, respectively. The standard deviations of the FP t for the NCI and CBOT corn futures are and per bushel, respectively. Neither the means (pairwise t-test, 5% level) nor the standard deviations (F-test, 5% level) are statistically different. Therefore, the presented regression results are not likely caused by a difference in variance of the independent variables.

12 76 Spring 2003 Journal of Agribusiness in only three cases are any of the estimated ratios different from Hence, the naive equal and opposite hedging strategy, or bushel-for-bushel, may be just as effective ex ante as using the estimated hedge ratios (Collins, 2000). Second, hedging effectiveness as measured by the R 2 is notably greater for the NCI than for the CBOT across all seven regions. For instance, in the Northern and Wabash regions, the difference in the adjusted R 2 s is roughly 15%. The largest difference in hedging effectiveness occurs in the Little Egypt region (difference of 16.1%) and the smallest difference in hedging effectiveness is in the South Central region (difference of 8%). The simple OLS regression results from estimating equation (1) are based on unconditional variances and covariances of the price change series, CP t and FP t. Meyers and Thompson (1989) suggest a generalized or conditional approach to estimating hedge ratios where other independent variables are added to equation (1). They propose economic theory, hypothesis testing, and common sense (p. 864) as guides to specify the other variables that should enter the generalized model. In this case, it makes sense to include a set of monthly dummy variables to adjust for seasonal tendencies in cash prices as well as lagged levels of the basis. 9 Incorporating these variables, the conditional hedging model is specified as follows: (2) CP t ' α % β FP t % φ Basis t&1 % ' δ i Month i,t % e t, where CP t and FP t are defined as in equation (1). Month i,t is a set of monthly dummy variables representing February through December (i = 2, 3,..., 11, 12). Basis t!1 is the basis level at the end of the prior month. 10 For the CBOT data series, care is taken to calculate the basis with the contract in which hedges held during month t would be placed. For instance, at the end of April, the basis variable in equation (2) is calculated with the July contract because that is the contract in which hedges held during May would be placed. The estimated coefficients on Basis t!1 are expected to be negative: a weak basis results in a relative strengthening of cash prices the following month. The monthly dummy variables should be negative during harvest, when cash prices are suppressed versus futures prices. Likewise, post-harvest months (i.e., November and December) should show an increase in cash prices relative to futures prices. Monthly dummy variables should also compensate for the above-mentioned temporal differences between the NCI and CBOT due to the CBOT s contract expiration cycles. 9 The goal here is not to do an exhaustive search on all possible conditioning variables. Theory dictates many economic variables, such as stock levels and interest rates, can impact the basis level. However, we want to utilize those conditioning variables that make sense to a hedger and are easily observed by practitioners. Therefore, the conditioning variables used (monthly dummies and lagged basis) are appealing independent variables to add to the specification. As suggested by Meyers and Thompson (1989), we also tried lagged values of FP t and CP t ; however, the estimated coefficients were not statistically different from zero. 10 The basis series used in equation (2) are stationary (augmented Dickey-Fuller test). The inclusion of the lagged basis makes equation (2) similar to an error correction model, where the CP and FP are cointegrated and the lagged value of the basis is the error correction term.

13 Sanders, Manfredo, and Greer Hedging Spot Corn 77 Table 4. NCI Conditional Hedging Effectiveness Regressions: Results from Estimating Equation (2), CP t = α + β FP t + φ Basis t!1 + Σδ i Month i,t + e t, June 1997SMay 2002 Independent Variable, Coefficient Northern North Central South Central Illinois Regions Western West/ Southwest Little Egypt Wabash Average FP t, β (Std. Error) (0.043) a (0.026) (0.028) (0.035) a (0.029) a (0.037) (0.026) Basis t!1, φ (Std. Error)!0.397 ** (0.114)!0.287 ** (0.100)!0.482 ** (0.115)!0.284 ** (0.092)!0.254 ** (0.089)!0.186 * (0.093)!0.099 (0.071)!0.284 Constant, α! ! ! February, δ ** ** * * * March, δ ! !0.062! April, δ ! May, δ * ** June, δ 6!2.332!1.132!0.021! ** July, δ * August, δ * ** ** ** * September, δ !1.468!5.864 **!2.250!0.063 October, δ ** ** ** ** ** November, δ ** ** ** ** ** * ** December, δ 12!0.002! ! Adjusted R Notes: The symbol ( ) denotes statistically different from 1.00 at the 5% level using a two-tailed t-test. Single and double asterisks ( * ) denote statistically different from zero at the 10% and 5% levels, respectively, using a two-tailed t-test. a Estimated with White s heteroskedastic consistent estimator. The estimated conditional hedging regressions [equation (2)] are presented in table 4 for the NCI and in table 5 for the CBOT. The conditional hedge ratios for the NCI (table 4) are not materially different from the unconditional hedge ratios (table 3). This finding is consistent with Myers and Thompson (1989), who suggest that unconditional hedge ratios estimated with price changes are close approximations to conditional hedge ratios. For the NCI, the estimated parameter on the lagged basis is negative and statistically different from zero at the 10% level in six of the seven regions. This suggests a weak basis is followed by a strengthening of the cash price given the change in the futures price (i.e., the basis strengthens). The results also suggest a fairly consistent seasonal component. In particular, cash prices tend to strengthen in November, immediately following harvest. The inclusion of the conditioning variables also improves the level of hedging effectiveness as measured by the adjusted R 2. With the unconditional model (table 3), the average R 2 across the seven regions is 92.9% compared to an average of 95.3% with the conditional model (table 4). The CBOT results (table 5) are comparable to those for the NCI. That is, the conditional hedge ratios are not meaningfully different from the unconditional hedge ratios

14 78 Spring 2003 Journal of Agribusiness Table 5. CBOT Conditional Hedging Effectiveness Regressions: Results from Estimating Equation (2), CP t = α + β FP t + φ Basis t!1 + Σδ i Month i,t + e t, June 1997SMay 2002 Independent Variable, Coefficient Northern North Central South Central Illinois Regions Western West/ Southwest Little Egypt Wabash Average FP t, β (Std. Error) (0.068) a (0.050) (0.047) b (0.035) (0.049) b (0.065) a (0.049) b Basis t!1, φ (Std. Error)!0.019 (0.013)!0.018 ** (0.008)!0.019 ** (0.007)!0.017 ** (0.009)!0.016 ** (0.007)!0.019 * (0.011)!0.018 ** (0.008)!0.018 Constant, α!0.612!0.839!1.346!0.432! !0.553!0.587 February, δ March, δ April, δ * May, δ June, δ 6!4.915!3.448!1.559!3.850! !1.826 July, δ 7!2.614!2.543!1.301!3.260!2.884!4.201!3.313!2.874 August, δ !0.573!3.971! September, δ ! !2.591!7.461 *!4.185!1.637 October, δ * ** * November, δ ** ** ** ** ** ** ** December, δ ** * Adjusted R Notes: The symbol ( ) denotes statistically different from 1.00 at the 5% level using a two-tailed t-test. Single and double asterisks ( * ) denote statistically different from zero at the 10% and 5% levels, respectively, using a two-tailed t-test. a Estimated with White s heteroskedastic consistent estimator. b Estimated with the Newey-West estimator. (panel B, table 3). There is a statistically significant negative relationship between basis levels and subsequent changes in cash prices. Also, all else equal, cash prices tend to increase during November. Whereas the inclusion of conditioning variables resulted in only a modest improvement in the NCI s hedging effectiveness (92.9% to 95.3%), it results in a rather dramatic shift in the CBOT s hedging effectiveness. The average hedging effectiveness across the seven regions improves from 79.1% for the unconditional CBOT hedges to 87.5% for the conditional hedges. Clearly, when hedging spot corn transactions, it is important to consider seasonality and the basis level that exists when hedges are placed. The comparison of conditional hedging effectiveness between the NCI and CBOT futures shows an advantage, albeit reduced, to the NCI futures. The average conditional hedging effectiveness for the NCI futures is 95.3% (table 4) versus 87.5% for the CBOT futures (table 5). This is still a relatively important advantage to the NCI futures. Consider that with the CBOT futures, the remaining residual risk is 12.5% of the underlying price risk ( ). In contrast, with the NCI, the remaining residual risk is only 4.7% of the underlying cash price risk ( ). Therefore,

15 Sanders, Manfredo, and Greer Hedging Spot Corn 79 the reduction in residual basis risk by using NCI futures is more than one-half. This finding implies the NCI futures may provide improved price protection for hedgers of spot corn in Illinois. Importantly, all of the conditioning variables used in equation (2) can be easily observed and obtained by practitioners. For example, assume it is October 31, and a grain merchant in Northern Illinois is placing a one-month storage hedge by selling December CBOT corn futures. Further, assume the current cash price is trading at a basis of 20 per bushel versus the December futures (or 20 under the December futures). Then, the expected change in the cash price is 0.902( FP per bushel. Thus, given the change in the futures price, the basis is expected to strengthen by 13.2 per bushel. This is useful information for merchants planning their business strategies. Economic Significance of the Results The results suggest that the use of NCI futures can yield greater hedging effectiveness and lower basis variability than the CBOT corn futures for hedging spot corn in Illinois. These results are statistically significant, but are they economically significant? Ultimately, this is the question that will determine if the NCI futures succeed or fail. The economic importance of alternative risk management tools is not easy to evaluate. The average basis standard deviation across the seven Illinois locations is 4.59 per bushel with the NCI and 8.56 per bushel with nearby CBOT futures. Grain merchandising is a notoriously low-margin business, with return on sales estimates for country elevators ranging from 0.75% (Ginder, 2002) to 1.6% (Robert Morris Associates, 2000). If we assume an average corn price of $2.25, then this would imply a profit margin ranging from 1.7 to 3.6 per bushel. With this small a margin, it would seem that reducing basis variability by 4 per bushel, or nearly one-half, would be economically important. Similarly, one would expect that a 10% increase in hedging effectiveness, as measured by R 2, is of economic importance. However, there are many economic issues involved in a hedging program other than basis variability which may prevent firms from adopting alternative hedging tools such as the NCI futures contract. First and foremost, any economic benefit must be carefully weighed against the true costs of using a new futures contract, such as liquidity and other transaction costs (Pennings and Meulenberg, 1997). Second, the internal cost of a firm altering its hedging program may be quite high. For instance, it is costly to retrain traders in terms of understanding expected basis levels and basis behavior. Collectively, the indirect costs associated with using new contracts may outweigh the benefits from a decrease in basis risk and overall increased hedging effectiveness.

16 80 Spring 2003 Journal of Agribusiness Summary and Conclusions This study provides an assessment of the Minneapolis Grain Exchange s National Corn Index futures contract. This is accomplished through an examination of the underlying cash index, the proposed NCI futures, as well as an examination of basis behavior and hedging effectiveness for seven diverse geographical regions in Illinois. Overall, our findings show the NCI accurately reflects commercial corn prices (U.S. No. 2 yellow) at the country elevator or first-handler level. Furthermore, the construction of the index, using over 1,600 daily elevator bids and the self-auditing process, makes manipulation of the NCI improbable. Therefore, it is a good candidate for a cash settled futures contract. Moreover, using past values of the NCI as a proxy for futures prices, it is found that the NCI futures may provide a better hedging tool than current CBOT corn futures in terms of lower basis variability and increased hedging effectiveness. The average basis variability across the seven regions is 4.59 per bushel for the NCI versus 8.56 per bushel for the CBOT futures. Similarly, the conditional hedging effectiveness (adjusted R 2 ) across the seven regions averages 87.5% for nearby CBOT futures versus 95.3% for the NCI. Thus, the residual basis variability is reduced by more than one-half when hedges are placed in the NCI futures as opposed to the existing CBOT contract. Although these findings suggest that a successful NCI futures contract may offer substantial benefits to country-level grain merchants, the probability of success for this contract still remains low. Most notably, the NCI futures face strong tradition and liquidity hurdles in opposing the incumbent CBOT corn futures. Grain merchandisers have been using CBOT futures for years and have developed a confidence in the contract. Furthermore, large bid-ask spreads and difficulty filling orders resulting from low liquidity as often found with new futures contracts could seriously hamper the success of the Minneapolis Grain Exchange s NCI futures contract. However, if cultural obstacles can be mitigated and critical liquidity achieved, then the NCI futures may prove to be a valuable hedging tool. References Brorsen, B. W., D. W. Buck, and S. R. Koontz. (1998). Hedging hard red winter wheat: Kansas City versus Chicago. Journal of Futures Markets 18, 449S466. Chaherli, N. M., and R. J. Hauser. (1995). Delivery systems versus cash settlement in corn and soybean futures contracts. OFOR Paper No , Office for Futures and Options Research, University of Illinois. Collins, R. A. (2000). The risk management effectiveness of multivariate hedging models in the U.S. soy complex. Journal of Futures Markets 20, 189S204. Ditsch, M. W., and R. M. Leuthold. (1996). Evaluating the hedging potential of the lean hog futures contract. OFOR Paper No , Office for Futures and Options Research, University of Illinois.

17 Sanders, Manfredo, and Greer Hedging Spot Corn 81 Ederington, L. H. (1979). The hedging performance of the new futures markets. Journal of Finance 34, 157S170. Garbade, K. D., and W. L. Silber. (2000). Cash settlement of futures contracts: An economic analysis. Journal of Futures Markets 20, 19S40. Garcia, P., and D. R. Sanders. (1996). Ex ante basis risk in the live hog futures contract: Has hedgers risk increased? Journal of Futures Markets 16, 421S440. Ginder, R. (2002). TH-BH elevator finance survey and grain market trends. Iowa State University Extension, Ames, IA. Hamilton, J. D. (1994). Time Series Analysis. Princeton, NJ: Princeton University Press. Hranaiova, J., and W. G. Tomek. (2002). Role of delivery options in basis convergence. Journal of Futures Markets 22, 783S809. International Treasurer. (1998). Federal Accounting Standard 133: The impact on risk management strategy and technique. A Best Practice Guide pamphlet through the Journal of Global Treasury and Financial Risk Management, published by International Treasurer, New York. Kahl, K. H., M. A. Hudson, and C. E. Ward. (1989). Cash settlement issues for live cattle futures contracts. Journal of Futures Markets 9, 237S248. Kimle, K. L., and M. L. Hayenga. (1994). Cash settlement as an alternative settlement mechanism for the live hog futures contract. Journal of Futures Markets 3, 347S361. Leuthold, R. M., J. C. Junkus, and J. E. Cordier. (1989). The Theory and Practice of Futures Markets. Lexington, MA: Lexington Books. Meyers, R. J., and S. R. Thompson. (1989). Generalized optimal hedge ratio estimation. American Journal of Agricultural Economics 71, 858S868. Parcell, J. (2002). Emerging IP markets: The Tokyo Grain Exchange s non-gmo soybean contract. Proceedings of the NCR-134 Conference on Applied Commodity Price Analysis, Forecasting, and Market Risk Management. Online. Available at agecon.lib.umn.edu/cgi-bin/pdf_view.pl?paperid=4800. Pennings, J. M., and M. T. G. Meulenberg. (1997). The hedging performance in new agricultural futures markets: A note. Agribusiness: An International Journal 13, 295S300. Peterson, P. (1995). Observations on cash settlement. In B. W. Brorsen (ed.), Proceedings of the NCR-134 Conference on Applied Commodity Price Analysis, Forecasting, and Market Risk Management (pp. 1S2). Stillwater, OK: Oklahoma State University. Rich, D. R., and R. M. Leuthold. (1993). Feeder cattle cash settlement: Hedging risk reduction or illusion? Journal of Futures Markets 13, 497S514. Robert Morris Associates. (2000). Robert Morris Associates Annual Statement Studies. One Liberty Place, Philadelphia, PA. Schroeder, T. C., and J. Mintert. (1988). Hedging feeder steers and heifers in the cashsettled feeder cattle futures market. Western Journal of Agricultural Economics 13, 316S326. Shafer, C. E. (1993). Hedge ratios and basis behavior: An intuitive insight? Journal of Futures Markets 13, 837S847.