THE IMPACT OF ETHANOL PRODUCTION ON LOCAL CORN BASIS Kathleen Behnke A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Science at the University of Wisconsin June 2010 ABSTRACT OF THESIS The Impact of Ethanol Plants on Local Corn Basis by Kathleen Behnke As the United States searches for a sustainable source of fuel, corn-based ethanol has emerged as an early leader. Over the past decade, ethanol production has risen from 1.5 million gallons in 1999 to 10.6 million gallons in 2009. This growth was primarily fueled by the growth and expansion of starch-based ethanol plants. Accordingly, this resulted in increased demand for corn and today almost 35 percent of U.S. corn production is used for ethanol. The aim of this thesis is to examine the impact local ethanol plants have on corn basis. The basis is the difference between the local cash price and the nearby futures contract price, and it accounts for variation in the supply and demand in the local market relative to the national market. It is predicted the entrance of an ethanol plant into a local cash market will increase corn demand, resulting in an increased cash price. The data set contains cash corn prices from 153 grain buyers in eight different Midwestern states. The data ranges from Fall 1999 through Summer 2009. In addition to being affected by ethanol production, it is predicted basis is influenced to by the ratio of local to national corn production, transportation costs, storage opportunity costs, and seasonal factors. To estimate the impact these variables have on corn basis a spatial error component model is used, which accounts for both the spatial dependencies and panel data structure. The empirical results were plausible and consistent with theoretical expectations. Results show that ethanol production in a 50-mile region of a county centroid has a small yet positive impact on local corn prices. The estimated impact of a 50 million gallon per year plant is a 0.425 ii cent per bushel increase in basis. These findings are smaller than the impacts found in previous work so the impacts were further investigated and shown to be consistent when directly compared to others’ findings. This study concludes local ethanol plants do have a positive price impact; however the research also suggests the price impacts of ethanol production may be felt well beyond the county borders. Additionally, there is evidence the long-term price impacts are much less than the initial short-term price response. iii ACKNOWLEDGEMENTS I would like to express my gratitude to my advisor, Professor Randy Fortenbery, for his support and guidance during this study as well as my academic career. I would also like to extend thanks to Professor Steve Deller and Professor Brent Hueth for serving on my thesis committee. Additionally, I am thankful for the programming support I received from Professor Brian Gould. I owe a special thanks to Kevin McNew and CashGrainBids.com for generously providing me with the data necessary for this project. Finally, I would like to thank the professors, staff, and graduate students in the Department of Agricultural & Applied Economics for their support and assistance throughout this process. iv TABLE OF CONTENTS ABSTRACT OF THESIS ............................................................................................................. ii ACKNOWLEDGEMENTS…………………………………………………………….………iv TABLE OF CONTENTS ............................................................................................................. v LIST OF FIGURES .................................................................................................................... vii LIST OF TABLES ..................................................................................................................... viii CHAPTER 1: INTRODUCTION ................................................................................................ 1 1.1 Rationale................................................................................................................................ 1 1.2 Objectives .............................................................................................................................. 3 1.3 Scope of the Study................................................................................................................. 4 1.4 Organization of the Study ..................................................................................................... 4 CHAPTER 2: BACKGROUND AND LITERATURE REVIEW ........................................... 6 2.1 Background ........................................................................................................................... 6 2.1.1 Corn Background ............................................................................................................ 6 2.1.2 Ethanol History and Policy ............................................................................................. 8 2.1.3 Ethanol Production Process .......................................................................................... 11 2.1.4 Ethanol’s Future ........................................................................................................... 12 2.2 Literature Review ................................................................................................................ 13 CHAPTER 3: ANALYTICAL FRAMEWORK ...................................................................... 17 3.1 Conceptual Model of Estimating Basis ............................................................................... 17 3.2 Empirical Model .................................................................................................................. 23 3.2.1 Panel Data ..................................................................................................................... 24 v 3.2.2 Spatial Methods ............................................................................................................ 24 3.2.3 Spatial Panel Model ...................................................................................................... 26 3.3 Data Sources ........................................................................................................................ 28 CHAPTER 4: EMPIRICAL RESULTS ................................................................................... 32 4.1 Model Validation................................................................................................................. 32 4.1.1 Hausman Test ............................................................................................................... 32 4.1.2 Lagrange Multiplier Tests to Select Model .................................................................. 32 4.1.3 Tests for Spatial Error Correlation and Random Region Effects ................................. 34 4.3 Alternative Models............................................................................................................................ 35 4.3 Parameter Estimates ............................................................................................................ 35 4.3.1 Weights Matrix ............................................................................................................. 40 4.4 Time Comparison ................................................................................................................ 42 4.5 Comparison to Literature .................................................................................................... 44 CHAPTER 5: CONCLUSION................................................................................................... 54 5.1 Summary ............................................................................................................................. 54 5.2 Conclusions ......................................................................................................................... 55 5.3 Suggestions for Further Research ....................................................................................... 56 APPENDIX A ............................................................................................................................... 57 APPENDIX B ............................................................................................................................... 59 APPENDIX C ............................................................................................................................... 63 APPENDIX D ............................................................................................................................... 67 SOURCES..................................................................................................................................... 73 vi LIST OF FIGURES Figure 1.1 Historical Ethanol Production …………………………………………………….…. 2 Figure 2.1 U.S. Corn Production………………………………………………………...…….….7 Figure 2.2 Percentage of U.S. Corn Production Used for Ethanol and Exports…………………..7 Figure 3.1 Corn Basis Map……………………………………………………………………....19 Figure 3.2 Average Annual Corn Basis………………………………………………………….20 Figure 3.3 Counties with Corn Price Observations…………………………………………...…29 Figure 3.4 Counties with Ethanol Plants…………………………………………………………29 vii LIST OF TABLES Table 3.1 Summary Statistics ……………………………………………………………….…..31 Table 4.1 Alternative Model Specifications……………………………………………………36 Table 4.2 Model Estimates ……………………………………………………………………38 Table 4.3 Spatial Weight Variations……………………………………………………………..41 Table 4.4 Compare Monthly and Quarterly Data………………………………………………..43 Table 4.5 Compare Full Sample and Sub-sample Estimates…………………………………….47 Table 4.6 Ethanol Impacts …………………………………………………………………...….49 Table 4.7 Compare Full Sample and Sub-sample Estimates with Interest…………………...….51 Table 4.8 Ethanol Impacts with Interest………………………………………………………....53 Table A.1 Renewable Fuel Standard Program Mandates………………………………………..58 Table B.1 Ethanol Plants Included in the Sample…………..……………………………………59 Table C.1 Grain Elevators………………………………………………………………………..63 Table D.1 Summary Statistics by State…………………………………………………………..67 Table D.2 Summary Statistics by Year…………………………………………………………..69 viii CHAPTER 1 INTRODUCTION 1.1 RATIONALE More than decade ago the United States began an aggressive search to find a practical source of renewable fuel to meet our insatiable energy demands. Alternative fuels such as starch-based ethanol, cellulosic ethanol, and biodiesel are all considered to be potential solutions in a national effort to reduce gasoline usage by 20 percent over the next ten years (Bush, 2007). Corn-based ethanol emerged as an early leader due to the abundance of corn and the popularity of ethanol-gasoline mixes. The national ethanol industry has expanded dramatically over the past 10 years. According to the Renewable Fuels Association (RFA), today there are more than 200 production plants with the capacity to produce almost 13.5 billion gallons. This is up from just 54 biorefineries with a production capacity of 1.7 billion gallons in 2000. The historical increase in production can be seen in Figure 1.1. RFA also reports the ethanol industry supported 400,000 jobs in 2009 and contributed $53.3 billion to the nation’s Gross Domestic Product (GDP). Furthermore, they calculate that despite the tax credit to ethanol producers the industry still contributed a tax surplus of $3.4 billion to the federal treasury. The rise of ethanol in the US has been largely driven by government mandates, tax incentives, and the push to lessen America’s dependence of foreign oil. The government has supported the use of ethanol as a policy to reduce dependence on foreign oil since the 1970’s. In 1 Figure 1.1: Historical Ethanol Production Historical Ethanol Production 12000 Millions of Gallons 10000 8000 6000 4000 2000 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 0 Source: Renewable Fuels Association the 1990’s it became popular to blend ethanol as an oxygenate in conventional gasoline to reduce smog. Ethanol production standards were set in place by the Energy Policy Act of 2005, and then updated as part of the Energy Independence and Security Act of 2007. Currently, ethanol production is scheduled to reach 36 billion gallons by 2022 and in the short-term there are plans to increase production by another 1.4 billion gallons in 2010. Furthermore, according to the 2010 Ethanol Industry Outlook, the 2009 production of 10.6 billion gallons of ethanol reduced the demand for oil by 364 million barrels. These government mandates, coupled with the high crude oil prices, have pushed the biofuels sector to center stage in the discussion of future U.S. energy policy. However, this conversation must consider the implications of energy policy on the agricultural sector. 2 Diverting corn and soy to produce ethanol and biodiesel has an impact on these commodity prices, which in turn affects many other factors in the traditional agriculture markets. Several economic issues are important to stakeholders in the both the ethanol and corn industries. Questions about the how much corn will be needed to continue the growth of the ethanol sector and how the increased demand for corn will affect prices on both a local and national level have increased in importance as the industry continues to expand. 1.2 OBJECTIVES The ever strengthening relationship between the food and fuel markets clearly raises the question of how the biofuels industry affects the price producers receive for corn. The overall purpose of this study is to examine the magnitude of this impact at the local level and measure the extent to which the effect is maintained over time. To this end the specific objectives of are to: 1. develop and estimate a spatial panel model of corn basis; 2. assess the impacts of ethanol plants on the local corn basis; and 3. determine if these impacts are consistent with the short-run impacts found in previous studies. Objective 1 involves the construction and validation of a spatial error components model to analyze the impact ethanol production and other independent variables have on corn basis. Objective 2 requires the implementation of the model to estimate the impact local ethanol production has on local corn price. This study specifically examines the price impact of ethanol production within 50 miles of a county centroid. Finally, objective 3 compares the findings of 3 this study to other work to determine if the impacts previously found are maintained in this more long-run setting. While the topic has been studied before, this particular work is important because it increases both the scope and the depth of the data used. Additionally, by accounting for spatial dependencies and the panel nature of the data more validity can be given to the results. Most importantly, this study will provide greater understanding of the impact of ethanol production on local corn prices. 1.3 SCOPE OF THE STUDY In order to estimate the impact of ethanol plants on local corn basis the study includes data from Illinois, Indiana, Iowa, Kansas, Minnesota, Nebraska, South Dakota, and Wisconsin. Together these Midwestern states account for more than 75 percent of the nation’s corn production (NASS) and are home to more than 70 percent of ethanol production plants (RFA) making them the ideal sample space. To estimate basis changes over time the sample period ranges from October 1999 through September 2009. This allows for estimation from the beginning of the period of rapid ethanol plant expansion. The data is aggregated by season to account for the variation throughout the year. Overall, there are observations from 153 different locations and 40 time periods included in the sample. 1.4 ORGANIZATION OF THE STUDY This chapter has presented the question this thesis addresses as well as objectives, and the general approach to the study. The next chapter provides a more detailed background of the 4 problem and a literature review of the topic. Chapter 3 presents the analytical framework designed to measure the changes in basis, as well as the spatial-panel model needed to properly frame the question. Chapter 4 presents the empirical results and discussion and Chapter 5 concludes with the summary, conclusions, and suggestions for further research. 5 CHAPTER 2 BACKGROUND AND LITERATURE REVIEW 2.1 BACKGROUND Before consumers are able to purchase starch-based ethanol at the pump there are many important production steps. The corn must be planted, harvested, and transported to market. Then the processing plant must buy inputs, produce ethanol, and send it to a blending facility before it can be distributed. This chapter begins by providing a background of the U.S. corn industry. It continues with a background of the ethanol industry, including an examination of government policies, production practices, and a look towards the future. This chapter will conclude with a review of the economic literature investigating the impact ethanol production has on corn prices and this study’s economic contribution. 2.1.1 Corn Background Corn has been an important part of agriculture in the U.S. since it was first introduced from Central America. Not only is corn a food source for humans and animals, it can also be converted to sugar, starch, beverage, or fuel. Previously the low cost of substitute fuels limited the conversion of corn to ethanol, but in response to changing prices and regulatory conditions it has found a place as a source of renewable biofuel. 6 Figure 2.1: U.S. Corn Production U.S. Corn Production 14000 Millions of Bushels 12000 10000 8000 6000 4000 2000 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 Source: USDA National Agricultural Statistics Service Figure 2.2: Percentage of U.S. Corn Production Used for Ethanol and Exports Percentage of U.S. Corn Production Used for Ethanol and Exports 40.0% 35.0% 30.0% 25.0% Ethanol 20.0% Exports 15.0% 10.0% 5.0% 0.0% 1990 1992 1994 1996 1998 2000 2002 Source: USDA Economic Research Service 7 2004 2006 2008 According to the Economic Research Service of the United States Department of Agriculture (USDA), national corn production has increased over the past decade due to greater demand. Figure 2.1 shows changes in production from 1990 through 2009. Production levels fluctuate in response to acres planted, weather conditions, and improved plant technology that allows for greater yields. Additionally, the USDA predicts higher net future returns to corn relative to other crops which provides an economic incentive to expand corn acreage in the coming years. In 2009, 34 percent of national corn production went into ethanol production, 15 percent was exported, 41 percent fell into the category of feed and residual use, and the rest was used for food, sugar, seeds, or was carried over as stocks in 2010. Over the past two decades the relative shares of corn use have shifted, as shown in Figure 2.2. From 1990 to 2009 the amount of corn exported as percent of total production decreased by seven percentage points, despite an increase in total export volume. In contrast, corn used for ethanol as a percentage of total production increased from four percent to 34 percent, becoming the second largest use category. The growth rate in this category is expected to stabilize as fewer new corn-based ethanol plants are built, but the demand for corn in ethanol production will continue to be large. 2.1.2 Ethanol History and Policy The use of ethanol has always been linked to the automotive industry. According to the U.S. Energy Information Association (EIA), Henry Ford built his first vehicle, the Quadricycle, to run on pure ethanol. However this automobile was quickly pushed to the side in favor of vehicles powered by gasoline, a less expensive alternative. This, coupled with the impending Prohibition, began America’s love affair with oil. Post World War II, the commercial ethanol 8 market almost disappeared and did not begin to resurge until the oil crisis of the 1970’s (U.S. EIA). As an effort to decrease dependence of foreign oil, the government has created policies to increase the use of ethanol fuel since 1978. The Energy Tax Act of 1978 provided a tax credit of $0.40 for every gallon of ethanol blended into gasoline at the 10 percent level. The number of ethanol plants began to increase and the Tax Reform Act of 1984 increased the blending credit to $0.60 per gallon. Despite these subsidies many of the new ethanol plants went out of business in the late 1980’s. Additional support for ethanol came from the United States Environmental Protection Agency (EPA). As a response to growing air pollution in 1995, the EPA required an oxygenate be added to gasoline in ten major smog producing regions of the country. This mandated that gasoline be mixed with a 10 percent oxygenating agent. Initially methyl tert-butyl ether (MTBE) was the popular choice, but after groundwater contamination scares from MTBE, ethanol has come to dominate the market for oxygenates (EIA). Another important component in the growth of the ethanol industry has been tax credits such as the Volumetric Ethanol Excise Tax Credit (VEECT) and the small ethanol producer credit. The VEECT was signed into law as part of the American Jobs Creation Act of 2004 and provided gasoline blenders a $0.51 excise tax credit per gallon of ethanol blended with gasoline. The 2008 Farm Bill reduced the credit to $0.45 and it is set to expire December 31, 2010. Additionally, the small ethanol producer credit provides an income tax credit of $0.10 per gallon for the first 15 million gallons of ethanol produced for plants with a capacity of less than 60 9 million gallons per year. Both of these credits have played an important role in encouraging growth in ethanol production and use. To promote even greater ethanol utilization, the government passed the Energy Policy Act of 2005. It mandated the production and sale of four billion gallons of ethanol in 2006, with incremental increases resulting in production of 7.5 billion gallons in 2012. This legislation led to a huge increase in ethanol production and plants were quickly producing a much greater volume than the mandates required. As a response the Energy Independence and Security Act of 2007 updated the mandates requiring almost 13 billion gallons in 2010 and setting the target at 36 billion gallons in 2022. However, the directive specifies that only 15 billion gallons can be corn-based ethanol, the rest will be cellulosic and advanced biofuels. For a deeper look into the Renewable Fuel Standard see Appendix A. The future expansion of the ethanol industry may depend on EPA’s approval of increasing the ethanol content in gasoline from 10 percent (E10) to 15 percent (E15). With RFAs current supply and demand predictions, if 100 percent of gasoline is sold as E10, by 2011 the supply of ethanol would exceed volume needed to create the E10 blend. According to Nurenberg of Ethanol Producer Magazine (2009), this blend wall will make it difficult to reach the Renewable Fuels Standard of 36 billion gallons by 2022. The EPA may allow E15 sometime later this year, but without a change the growth of the industry may be hindered by suppressed demand. It should also be noted that ethanol production and use receives support from state governments. All of the states in the sample have some sort of program supporting biofuels and some even have their own renewable fuels mandates. Other programs include production tax 10 incentives, plant loan assistance, blender’s tax credits, state money for ethanol research and promotion, and mandates for state fleets to shift to renewable fuels (International Institute for Sustainable Development, 2006). 2.1.3 Ethanol Production Process Ethanol is a sugar-based bio-fuel that produces energy when burned. Though its energy content is less than that of pure gasoline, ethanol can reduce tailpipe carbon monoxide emissions by as much as 30 percent (RFA). This is a result of ethanol being composed of 35 percent oxygen, which results in more complete fuel combustion (RFA). Ethanol is commonly blended with gasoline at the 10 percent level (E10) to serve as an oxygenate. Furthermore, flex fuel vehicles can use E85, a blend of 85 percent ethanol and 15 percent gasoline. The majority of American-made ethanol is produced from corn, while a small amount is produced from cheese whey, wood waste, or other grains. Corn-based ethanol plants represent 92 percent of all plants, but almost 99 percent of all U.S. ethanol production. Depending on plant technology, an average of 2.8 gallons of ethanol can be produced from one bushel of corn (RFA). As technology improves, plants are continually striving to produce more ethanol with fewer inputs to improve efficiency and reduce costs. In 2000, plants which produced 40 million gallons of ethanol per year were standard, but by 2005 the new plants were being built with capacities of 50 or 100 million gallons per year (Sneller & Durante, 2006). Today the average plant capacity is slightly less than 70 million gallons of production per year (RFA). According to Eidman (2007) the returns to scale in ethanol production increased between 2003 and 2005, leading to an increase in the size of new plants. In late 2006, an average 60 million gallon per year plant had an investment cost of 11 $1.875 per gallon of output, whereas a 120 million gallon per year plant had a cost of $1.50 per gallon of output. Other important components in determining the profitability of a plant include: the price the plant receives for its outputs, efficiency of the firm, cost of capital and labor, the cost of natural gas to power the plant, and of course, the cost of the main input, corn. In addition to ethanol, a traditional dry mill ethanol plant produces a number of coproducts such as dried distillers’ grains with solubles (DDGS), and carbon dioxide (CO2). On average a plant can produce 18 pounds of DDGS per bushel of corn used in ethanol production and most plants are able to sell it to the livestock industry as a high value feed source. If there is a market opportunity, some plants are able to sell the CO2 to the food processing and bottling industries. Also, ethanol wet mills can produce corn gluten meal, corn gluten feed, sweeteners and corn oil which can also be sold to their respective industries. 2.1.4 Ethanol’s Future The future of ethanol production remains uncertain. The initial rapid growth in the industry was challenged as the price of corn began to rise in 2007. High input costs put a strain on many ethanol producers, and in 2008 a hedging strategy gone wrong caused the bankruptcy of the nation’s largest ethanol producer, Verasun (Taylor & Dorsey, 2010). As the corn-based ethanol industry still faces challenges, the race is on to develop new methods of ethanol production. According to the RFA, there are currently 28 new cellulosic ethanol plants under development and construction which will use wood, switch grass, sugarcane, or other biomaterials. Work completed by De La Torre Ugarte, English, and Jensen (2007) examined the future implications of ethanol production and expansion. Their estimates are based upon the standards 12 set by The Energy Policy Act of 2005 and The Biofuels Security Act of 2007, as proposed at the time of writing, which set ethanol production levels at 10 billion gallons in 2010, 30 billion gallons in 2020, and 60 billion gallons in 2030. Mindful of improved technologies for cellulosic ethanol being developed, the authors examine three scenarios for the future of ethanol production. The first scenario projects cellulose-to-ethanol technology to be commercially available by 2012, thus new ethanol plants would adopt this technology and existing corn-toethanol plants would continue to use corn as their main input. The second scenario examines the impacts of corn-based plants adopting cellulose technologies in 2012 and the third scenario envisions the switch happening in 2015. When compared to the USDA 2006 baseline, all three scenarios estimate the price of corn to increase by at least $0.86 per bushel by 2010. By 2030, the corn price impact is estimated to be $0.62 in the first scenario, $0.52 in the second scenario, and $0.59 in the third scenario. They also estimate the 2030 price of soybeans to increase by $0.89 to $1.23 per bushel and the price of wheat to increase by $0.36 to $0.53 per bushel as a result of competing for acres with higher valued corn. From 2007 to 2030 they project a cumulative net farm income increase of $210 billion and an $8.7 billion reduction in government payments as a result of current ethanol policy. 2.2 LITERATURE REVIEW The record growth of the ethanol industry has generated a wide body of economic literature investigating ethanol’s impacts on crop, land, and fuels prices, its ability to create community economic development, the industry’s potential for long-term sustainability, and 13 more. While all of these pieces are important in understanding the broad impact ethanol has on a national level, here the focus is on the impact ethanol plants have on local corn prices. There is rich literature on the link between the food and fuel markets and the impact ethanol has on national corn prices. A study by Luchansky and Monks (2008) found an interesting shift in the relationship between corn price and ethanol production as the market has evolved. Their findings indicate that on the supply side, ethanol production is not significantly related to corn prices. In contrast, they cite a 1998 study by Rask which found that corn prices strongly influence ethanol production levels, but they conclude that due to government mandates and clean air requirements, production is no longer being heavily influenced by input costs. Rather, it now appears ethanol production levels are playing a role in determining the price of corn. Fortenbery and Park (2008) measured the effect of ethanol production on U.S. corn prices at a national level and found that a one percent increase in ethanol production will cause a 0.16 percent increase in the short-run corn price. The results also show that the great increase in corn price in 2007 is not fully explained by the impact of ethanol production. They conclude that some of the price increase can be contributed to ethanol production, but other supply and demand factors also played a large role. Rather than investigating the impact of ethanol on the national corn price, it is the goal of this study to estimate the effect an ethanol production plant has on the local corn price. Early work on measuring the impact of ethanol plants on local grain prices was conducted by McNew and Griffith (2005) in a study which estimated the impact of 12 ethanol plants that opened in 2001 and 2002. The found, on average, corn prices increased by 5.9 cents in the region and 14 positive price impacts could be felt up to 68 miles away. The price impact at the ethanol plant sites ranged from 4.6 to 19.3 cents per bushel, depending on the local corn supply. In theory, areas with high corn demand or low corn supply experience price impacts that are relatively greater than in areas with less demand or excess supply. The findings of an increased local corn price were supported by a Henderson and Gloy (2009) investigation of ethanol plant impacts on cropland values. Their results indicate an annual impact of 2.3 to 6.4 cents per bushel. They conclude the change is a result of the decreased transportation costs. However, the estimation of positive price impacts due to a local ethanol plant is not a universal finding. In Kansas, O’Brien (2009) found corn prices at elevators located within 60 miles of an ethanol plant were significantly lower than elevators further than 60 miles from a plant. In addition, a similar effect was found in Katchova (2009). In his model, farmers located in the same zip code as an ethanol plant actually received a price 10.9 cents lower than other farmers in the sample. This does not mean ethanol plants negatively affect corn prices, but merely that spatial differences play an important role in determining prices and close proximity to an ethanol plant does not secure higher prices. A study of Iowa ethanol plants by Gallagher, Wisner, and Burbacker (2005) found the corn price impacts were dependent on the location of the plant. They observed nine market areas in Iowa and found evidence that an ethanol plants tends to increase the local corn price. However, in locations where there were already many grain buyers, such as the northwest region of the state, the introduction of an ethanol plant had no statistically significant price impact. Thus the results were mixed and dependent upon pre-existing market conditions. They also 15 found evidence that corn prices declined as distance increased from the Mississippi River on Iowa’s eastern border. Olson, Klein, and Taylor (2007) build a strong theoretical model for basis analysis. The variables include futures price, corn production, corn usage in ethanol, a storage measure, and variables dealing with transportation. McNew and Griffith (2005) include state and national corn production, monthly dummy variables, and an ethanol dummy variable, diesel price, and a sophisticated distance term. In other basis literature, not specific to ethanol production, the relationship between local and national prices in modeled by a production ratio, rather than absolute production values (Fortenbery, 2002). Also, Fortenbery (2002), Kahl & Curtis (1986), and Garcia & Good (1983) stress the importance of including a storage cost or opportunity cost measure in basis analysis. Finally, Kahl & Curtis (1986) also include a price measure. The specific variables used in this study will be further discussed in Chapter 3. The analysis in this thesis is consistent with the current literature’s attempt to assess the economic impacts of ethanol production on local corn prices. This study expands upon the current literature by increasing both the time span and the scope of the data set. This investigation is able to better predict the long term price changes as a result of ethanol production due to the use of a 10 year data set. Also, this study expands the data set to include counties with and without ethanol plants to truly measure how much a nearby ethanol plant actually affects basis. Taking a broad look across time and space at an ethanol production plant’s price impact is this study’s contribution. 16 CHAPTER 3 ANALYTICAL FRAMEWORK Shepherd wrote “within the limits of the perfect market, prices should differ among locations by no more than the cost of transportation; among time periods by no more than the cost of storage; and among product forms by no more than the cost of transformation” (as cited in Davis & Hill, 1974, p. 135). Yet it is found that a simple calculation of transport, storage, and processing costs are not enough to explain corn price differentials across space and time. Factors such as supply and demand on the local, national, and international level also play a huge role in determining the corn basis. This chapter examines these factors, describes the empirical model to be implemented and concludes with a description of the data sources. 3.1 CONCEPTUAL MODEL OF ESTIMATING BASIS The rise in US ethanol production has many important economic consequences. Diverting corn away from its traditional feed, food, and export markets leads to fundamental shifts in agriculture production decisions. Corn is a commodity, thus corn producers are price takers. This means they have no direct influence over the prices they receive. Moreover, most producers sell directly to the local grain elevator, thus the most important price to a producer is the price being offered at a specific location and time. The pricing of corn is further complicated by production and use patterns. Corn is produced only once per year, but there is a relatively constant demand throughout the year. Prices vary across time in response to storage costs, and vary across space in response to whether specific regions have a corn surplus or deficit. 17 In addition to the cash price, another important pricing component is the basis. It is defined as: Basis = Local Cash Price – Nearby Futures Price. (3.1) The local cash price is defined as the price a corn producer would receive for corn on a specified day and location. The nearby futures price is defined as the price of the nearby futures corn contract as traded through the CME Group. The contract trades 5,000 bushels of #2 yellow corn in the months of March, May, July, September and December. The nearby contract describes the contract closest to expiration, not including the current month. When the corn price basis increases, or strengthens, in a local grain market, it indicates the local corn price has increased relative to the futures price. It is important to note the basis merely measures the relationship between the local price and futures price; it indicates nothing about the actual price levels. For example, the basis can be strengthening but the corn price may be dropping. There are a variety of factors influencing the basis at any particular location at any given point in time. Figure 3.1 is a map of the corn basis throughout the Midwest on May 28, 2010. The basis varies across space in response to supply and demand conditions, as well as in regard to transportation costs. The map more or less shows how basis weakens as distance increases from south-western Illinois. The basis also varies over time in regards to changing market conditions. Figure 3.2 shows the basis over the past decade for individual states, plus the full sample. 18 Figure 3.1: Corn Basis Map Source: Center for Agricultural and Rural Development, Iowa State University 19 Figure 3.2: Average Annual Corn Basis Corn Basis 0 1999 2000 2001 2002 2003 2004 2005 -10 2006 2007 2008 Illinois Cents per bushel Iowa -20 Indiana Kansas -30 Minnesota Nebraska South Dakota -40 Wisconsin Full Sample -50 -60 Source: CashGrainBids.com, Calculations by Author Fortunately, many of the factors which influence basis fit into three specified categories provided by Garcia and Good (1983). They determined the magnitude of basis is influenced by cost, stock, and flow factors. Cost factors include storage and transportation costs. The stock factor is the supply measure and includes the amount of corn in storage. Finally, flow factors measure demand and include the rate of market consumption. The model employed here will use historical data that accounts for the cost, flow, and stock factors identified by Garcia and Good. Specifically, the variables of interest are: Production Ratio The basis is determined by both the local cash and futures prices, thus the ratio of local to national production is important in determining price relationships. It is expected that if 20 local stocks make up a relatively greater share of the national stocks, the basis will be lower because corn will need to be transported out of the region. Conversely, if local production is making up relatively less of the national production, it is expected the local cash price will be higher relative to the futures price. Local ethanol Theoretically an ethanol plant has the power to strengthen the local basis is two ways. First, an ethanol plant increases the demand for corn in its region. Second, as Davis and Hill (1973) note, due to the spatial nature of the elevator industry, market structure theory indicates some elements of geographical monopsony may exist. If this occurs, a single firm in a particular region may then exhibit market power and have the ability to influence price. Thus, an ethanol plant represents a new entrant to the market and may dilute market power held by the grain elevator. Diesel Transportation costs play a major role in determining basis because all surplus grain from the local market must be moved to the national or international market. Following economic theory and Gallagher, Wisner, and Burbacker (2005) the local market price is a function of the exogenous national market price less the cost to transport the commodity to market. Hence, for excess supply markets the decline in local price occurs with greater distance to the market, matching the increased transport costs. To account for the cost of transport the average Midwest diesel price is used as a proxy for grain trucking costs. 21 Storage There is a substantial demand for corn year-round, but the commodity is only produced one time per year. This means the price will fluctuate throughout the year in relation to storage. Garcia and Good (1983) note storage costs can include warehouse charges, interest, or insurance. In the model, storage is accounted for by including the prime interest rate, which mimics opportunity costs. Specifically, if the opportunity cost of holding grain is high, producers are expected to sell. This would amount to an increased supply on the cash market, thus a high opportunity cost is expected to cause a decrease in basis. Seasonality Another important consideration, and one which is closely linked to storage, is seasonality. A higher cash price is expected as time increases from harvest as a method to compensate producers for storing the grain. In grain surplus areas, it is expected the basis will be the weakest at harvest and will strengthen throughout the year as the local market reduces its overall supply of grain. As the excess grain is moved out of the market the local cash price is expected to converge with the futures price (Kahl & Curtis, 1986). A model which can be used to explain local corn price basis can be built by combining these factors. The model used in this study is specified as: 22 Basis = β0 + β1 * (Production Ratio) + β2*(Ethanol Production) + β3*(Interest Rate) + β4*(Diesel Price) + ∑7𝑖=1 𝛽𝑖6 * (State) + ∑3𝑖=1 𝛽𝑖7 * (Season) (3.2) As noted in Chapter 2, Kahl & Curtis (1986) include price in their basis model and find it to be statistically significant in grain surplus markets. Higher prices may induce producers to sell, thus weakening the basis. The model will also be estimated with this variable. Selling grain on the local cash market is inherently risky because of the high level of price volatility in the market. In order to reduce risk, producers are able to enter into a hedge and trade the price risk for basis risk. Typically, basis risk is less than price risk. As ethanol plants enter a local grain market the increased demand for corn has the potential to increase the local cash price relative to the futures price, thus strengthening the basis. Investigating the impact ethanol has on corn basis is important so producers can adjust their price expectations. Above all, having a proper prediction of the basis is vital to enable producers to make effective risk management decisions. 3.2 EMPIRICAL MODEL The observations in the dataset vary over time and space. This provides the opportunity to observe changes which occur through time, as well as those that occur across locations. Panel data has an advantage over pure cross-section or pure time-series data in detecting and measuring effects, as it is able to look at a broader picture (Gujarati, 2003). Most importantly, it allows for the investigation of more complicated behavior models. However, this expansive dataset also presents many significant econometric considerations. As OLS is not an appropriate method for panel data, a fixed or random effects approach must be employed (Elhorst, 2003). Additionally, potential spatial dependency must be accounted for in the model. The following sections 23 describe methods for addressing panel data, incorporating spatial methods, and finally combining the two. 3.2.1 Panel Data In order to select the best approach for dealing with panel data both a fixed effects and random effects model must be estimated. Based on prior assumptions of the model, the fixed effects approach uses dummy variables to represent time periods, cross-sections, or a combination of both to account for omitted explanatory variables. Alternatively, the random effects approach represents the lack of knowledge of the true model through the disturbance term. To choose between the two approaches the Hausman test may be used (Gujarati, 2003). 3.2.2 Spatial Methods When using spatial data it is important to account for spatial dependencies. Ignoring these relationships can lead to inefficient and biased estimates, invalid inference procedures, and ultimately lead to drawing the wrong conclusions from the data analysis (Cliff & Ord, 1981). Spatial dependency arises in a sample for a variety of reasons. In this work, the spatial dependency is a result of spill-over effects and spatial externalities. Additionally, the arbitrary county boundaries make it necessary to aggregate data over space to get a full view of the supply and demand conditions in a particular location. Others have used time-series methods to deal with the changes in basis, but Anselin (1988) writes that these specifications are not appropriate due to the multidirectional nature of dependence in space as opposed to the one-directional time movement. Thus it is necessary to account for spatial dependency in basis analysis. 24 3.2.2.1 Spatial Models There are two main types of models to deal with spatial dependency. The first is a spatial lag model which is used when the spatial correlation pertains to the dependent variable and is generally specified as: y = ρWy + Xβ + ε. (3.3) The second is the spatial error model, which is used when the spatial correlation effects the error term. It is typically specified as: y = Xβ + ε ε = λWε + u (3.4) In both models, y is a n x 1 vector of observations of the dependent variable, W is an n x n spatial weights matrix, ρ and λ are spatial autoregressive parameters, X in an n x k matrix of observations of the independent variables, β is the k x 1 vector of regression coefficients, and ε and u are error term vectors. 3.2.2.2 Spatial Weights Matrix The spatial weights matrix, W, is an N x N positive matrix that specifies the neighborhood set for each observation. Each location observation appears as both a row and column. If location i and location j are considered to be neighbors wij will have a non-zero value, and if the locations are not neighbors wij = 0. Also, by convention a location is not considered to be its own neighbor thus the diagonal elements wii = 0. Generally, the weights are standardized so that the elements of each row sum to one, or 𝑤𝑖𝑗𝑠 = 𝑤𝑖𝑗 / ∑𝑗 𝑤𝑖𝑗 . This row 25 standardization allows for the interpretation of the weights by constructing a weighted average of the neighboring values through a spatial lag operator, which can then be applied to the error term Wε (Stakhovych and Bijmolt). 3.2.3 Spatial Panel Model To analyze the effects of time and space, following Anselin, Le Gallo & Jayet (2008), we begin with a basic pooled linear regression model: yit=xitβ + uit , (3.5) For the model i is the cross-sectional index, with i = 1….N and t the time index, with t = 1….T. The total number of observations is NxT. The dependent variable is yit, where each unique observation is denoted at both i and t. The observations of the exogenous variables are contained in1xK vector xit, β is a Kx1 vector of the regression coefficients, and uit is the error term. To properly analyze spatial effects the observations are stacked first by the time period t = 1….T and then by the cross-section i = 1….N which leads to y’ = (y11,….y1N,…yT1,…yTN). The error term consists of spatially autocorrelated residuals, as well as random disturbances. Following Baltagi et al. (2003), the error vector for time t is represented as ut = µ + εt (3.6) εt = λWεt + vt. (3.7) with where ut’ = (ut1, …,utN), εt’ = (εt1,…, εtN) and μ’ = (μ1,…, μN) denote the vector of random region effects which are assumed to be IIN(0, σ2μ) . Using the W-matrix we are able to find λ, the 26 spatial autoregressive coefficient, which will have a positive value less than one. In a panel setting the spatial weights matrix and the spatial autoregressive coefficient are assumed to remain constant over time (Anselin, 1988). Finally, vt’ = (vt1, …, vtN) where vti is i.i.d. over i and t and is assumed to be N(0, σ2v). εt can be rewritten as εt = (IN – λW)-1vt = B-1vt, (3.8) where B=IN – λW and IN is an identity matrix of dimension N. Once the data is stacked the pooled regression can be written as: y = Xβ + u, (3.9) where y is a NT x 1 vector, X is a NT x K matrix, and ε is a NT x 1 vector. The error vector takes the form: u = (iT ⊗IN)μ + (IT ⊗ B-1)v, (3.10) where iT is a vector of ones with dimension T and IT is an identity matrix of dimension T. From this the covariance matrix for u can be written: Ωu = σ2μ (JT ⊗ IN) + σ2v (IT ⊗(B’B)-1), (3.11) where JT is a matrix of ones with dimension T. From this we can rewrite the matrix as: Ωu = σ2v [KT ⊗ (T φIN + (B’B)-1) + ET ⊗ (B’B)-1] = σ2v Σu, (3.12) Where φ= σ2μ/ σ2v, KT = JT/T, ET = IT – KT, and Σu = [KT ⊗ (T φIN + (B’B)-1) + ET ⊗ (B’B)-1] thus: Σu-1 = KT ⊗ (T φIN + (B’B)-1)-1 + ET ⊗ (B’B) 27 (3.13) and: |Σu| = |T φIN + (B’B)-1| *|(B’B)-1|T-1. (3.14) Using these results, Anselin derived the log-likelihood function which was for the model: L= − NT 1 T−1 1 ln 2πσ2v − ln [ |TϕIN + (B′ B)−1 |] + ln |B′ B| − 2 u′ Σu−1 u 2 2 2 2σv where u = y- Xβ. (3.15) 3.3 DATA SOURCES The data set contains information on monthly corn prices, futures prices, diesel prices, interest rates, and ethanol production. Additionally, there is information at an annual level on local and national corn production and quarterly information about national and state stock levels. The variable of interest in the model is the corn basis. The local price data is a collection of daily corn prices compiled by CashGrainBids.com from 153 grain elevators over 129 months. The data ranges from 1999 to present, and the sample used ranges from October 1999 to September 2009. The daily data were aggregated to monthly data. Missing observations, as well months with partially missing data, where interpolated from neighboring counties. In the sample, missing observations represent approximately two percent of the data points. In order to calculate the basis, the local corn price is subtracted from the nearby futures price. All counties with an observation are shown in purple in Figure 3.3. 28 Figure 3.3: Counties with Corn Price Observations Figure 3.4: Counties with Ethanol Plants 29 All information about ethanol production, such as plant location and nameplate capacity comes from the Renewable Fuels Association (RFA). To determine the date production began at the plant, data was used from Ethanol Producer Magazine. If the date was unavailable, the plant’s website was used or the plant was called for the information. Figure 3.4 is a map displaying all counties with at least one ethanol plant in green. The data of national and county level corn production, as well as state and national corn stocks, comes from the USDA National Agricultural Statistics Service (NASS). Any missing data points in the level of county corn production where obtained by using the existing data from the county, and the average percentage change in production from around the state. Data on corn production is available on a county level and information about ethanol production is available at a point location, but to truly understand the local supply and demand conditions which determine the price at a grain elevator a more broad measure is needed. In an effort to get more complete information of these factors a 50 mile buffer ring was drawn around the centroid of each county. From this, to determine the local conditions, data from the county was summed with data from any other counties whose centroid fell within the 50 mile buffer. In the model, diesel prices are used as a proxy for transportation costs. The price used is the average monthly Midwest retail price as provided by the U.S. Energy Information Administration. The prime interest rate acts as a proxy for the storage opportunity cost in the model. This information comes from the Federal Reserve Bank Statistical Release. For much of the analysis the data was aggregated to quarterly time periods. O’Brien (2009) also uses this approach in his analysis as a method to reflect the seasonality of grain marketing. He defines October through December as the fall quarter. At harvest time the final 30 size of the new crop is known and producers begin to make decisions about the use of their crop. January through March is the winter quarter. At this time of year producers are continuing to store or selling to elevators or other grain buyers. Spring includes April through June and is the planting season. Prices are driven by acreage decisions and crop planting conditions, as well as grain stocks and demand. Finally, the summer months of July through September are the period of crop development, where weather and yield predictions dominate the market prices. This method is supported in other basis literature. Garcia and Good (1983) aggregated monthly prices into seasonal observations in an effort to increase the variability among independent variables. Davis and Hill (1974) also separated their data into seasons to estimate basis. Even the when using monthly data, Kahl and Curtis (1986) use seasonal dummy variables to reflect the seasonality of grain pricing. Thus, the aggregation of data into quarterly pieces will reduce some variation, but it is a method with precedence in the literature. Table 3.1 provides summary statistics of the data set. Summary statistics are also available by individual state or year in Appendix D. Table 3.1: Summary Statistics Variable Mean Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price (cents) National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Interest Rate -28.14 227.28 136.32 211.46 10.65 2.14 155.03 5.10 3.71 6.12 31 Std. Dev. 14.52 109.11 155.49 80.86 1.26 0.99 92.38 2.79 2.35 1.95 Min -92 7 0 115.40 8.97 0.07 2.51 0.96 0.08 3.25 Max 16 544 1026 434.20 13.04 4.34 543.99 10.28 13.70 9.50 CHAPTER 4 EMPIRICAL RESULTS The goal of this study is to measure the impact ethanol plants have on local corn basis. This chapter presents a set of results which empirically quantify these effects. First, the model’s validity is assessed and then estimates of the impact are reported. Next, results from varying model specifications are presented and compared to the model’s estimates. Finally, the model is compared against other models in the literature. 4.1 MODEL VALIDATION 4.1.1 Hausman Test As noted in section 3.2.1, when using panel data the Hausman test is necessary to pick between a fixed- or random-effects estimator. To conduct this test both models are estimated and the parameters are compared. The null hypothesis is that there are no omitted variables, which would mean the fixed effects and random effects estimators do not differ substantially. The Hausman test returns a value of 5.96. The test statistic is distributed as chi-squared with six degrees of freedom, which leads to a failure to reject the null hypothesis. Hence, the test does not indicate a serious problem with omitted variables, thus a random-effects estimator is most appropriate. 4.1.2 Lagrange Multiplier Tests to Select Model When using spatial data, an important first step is determining the proper model specification. In order to test the spatial lag model against the spatial error model two Lagrange Multiplier tests are used. LMerror is used to evaluate if the spatial error model is necessary and the LMlag test examines if the spatial lag model should be used (Stakhovych & Bijmolt, 2008). 32 The null hypothesis is no spatial modeling and the tests are chi-square distributed with one degree of freedom. The tests are: 𝐿𝑀𝑒𝑟𝑟𝑜𝑟 = 𝐿𝑀𝑙𝑎𝑔 = 𝑒′ 𝑊𝑒 2 )^ ̂2 𝜎 ( (4.1) 𝑇 𝑒′ 𝑊𝑦 2 )^ ̂2 𝜎 ( (4.2) 𝑛𝐽 where 𝑇 = 𝑇𝑟[(𝑊 ′ + 𝑊)𝑊] (4.3) ′ 1 𝐽 = 𝑛∗𝜎̂2 [(𝑊𝑋𝛽̂ ) 𝑀(𝑊𝑋𝛽̂ ) + 𝑇𝜎̂ 2 ] (4.4) 𝑀 = 𝐼 − 𝑋(𝑋 ′ 𝑋)−1 𝑋 ′ (4.5) e is the vector of OLS residuals, 𝜎̂ 2 = e’e/N, I is an identity matrix of dimensions n x n, and 𝐵̂are the OLS parameter estimates. When running the LM tests for the corn basis data it is found that both the LMerror test and LMlag test reject the null hypothesis of no spatial dependency. Stakhovych and Bijmolt recommend that if both Lagrange Multiplier test statistics are found to be significant the specification associated with the more significant test is the correct. The test statistics are LMerror = 100.29 and LMlag = 93.63, thus the spatial error is used from this point forward. It should also be noted the spatial error model coincides with previous literature on the topic. (Note: These tests are for spatial data, but not developed to account the time series nature of the data. The LM tests were conducted from the final time observation in the data set, but were found to also hold in other time periods.) 33 4.1.3 Tests for Spatial Error Correlation and Random Region Effects The spatial distribution of the data suggests there may be some spatial dependency in the sample. As noted in Baltagi, Song, and Koh (2003), it is important to conduct a Lagrange Multiplier (LM) test because ignoring the spatial correlation and heterogeneity due to random region effects will result in inefficient estimates and misleading inference. The LM test developed in Baltagi et al. (2003) is conducted to simultaneously test for the existence of spatial error correlation and random region effects. In order to test the joint hypothesis H0: λ = 𝜎𝜇2 = 0, the test statistic: 𝑁𝑇 𝐿𝑀𝑗 = 2(𝑇−1) 𝐺 2 + 𝑁2 𝑇 𝑏 𝐻2 is used, where using the OLS residuals, u, 𝐺 = (4.6) 𝑢′ (𝐽𝑇 ⊗𝐼𝑁 )𝑢 𝑢′ 𝑢 − 1 , 𝐻 = 𝑢′ (𝐼𝑇 ⊗ 𝑊+𝑊 ′ 2 ) 𝑢/𝑢′𝑢 and 𝑏 = 𝑡𝑟(𝑊 2 + 𝑊 ′ 𝑊) . This test returns a value of 18825.02 which implies we are able to reject the joint null of no spatial effects and no random effects. Additionally, a marginal LM test is conducted to detect the presence of spatial effects assuming there are no random effects (Baltagi et al., 2003). The test statistic for H0: λ=0 (assuming 𝜎𝜇2 = 0) is 𝑁2 𝑇 𝐿𝑀2 = √ 𝑏 𝐻 (4.7) which returns a value of 93.63. The statistic is asymptotically distributed 𝜒12 thus this null hypothesis is also rejected. 34 4.2 ALTERNATIVE MODELS The tests in Section 4.1 indicate that the use of a spatial error components model is necessary to obtain efficient and unbiased estimates. The data is entered into the maximum likelihood function developed by Anselin (1988) and keeping with the literature (McNew & Griffith, 2005) the W-matrix specifies that any observations within 50-miles of one another will have correlated errors. In the literature there are a variety of model specifications used to estimate basis. The theoretical model discussed in Chapter 3 included a production ratio, transportation costs, storage costs, ethanol production, and state and seasonal dummy variables. The results for this model, along with alternative specifications are displayed in Table 4.1. The theoretical model described above is labeled Model A. Model B uses the ratio of local stocks to national stocks, rather than the production ratio. Model C omits the measure of storage opportunity cost, Model D omits the transportation proxy variable, and Model E omits the seasonal dummy variables. Model F includes the nearby futures price as an additional variable. 35 Table 4.1 Alternative Specifications Intercept Prod Ratio Stocks Ratio Diesel Ethanol Interest Futures Illinois Iowa Kansas Minnesota Nebraska South Dakota Wisconsin Fall Spring Summer 𝛔𝟐𝐯 𝛔𝟐𝛍 λ Likelihood Parameters AIC BIC Model A 20.32 ** -1.29 * -0.10 ** 0.0085 ** -2.16 ** Model B 4.18 -2.20 -0.11 0.0084 -2.19 Model C 5.99 -1.11 * ** ** ** Model D 0.93 -1.18 * -0.12 ** 0.0083 ** -6.46 ** 6.20 ** -6.88 -13.12 ** 12.89 ** -11.03 -7.48 * 3.98 -6.88 -19.72 ** -6.42 ** -17.21 -8.36 * 4.53 -6.20 -22.10 ** -8.83 ** -19.55 -24.35 ** -11.00 ** -22.98 -2.39 2.39 3.13 2.25 5.52 6.37 -1.38 2.97 2.17 11.66 ** 11.65 ** 11.64 17.91 ** 17.83 ** 16.76 0.95 ** 0.95 ** 0.96 -18447.8 -18464.7 -18682.2 15 15 14 36926 36959 37392 36898 36932 37367 *Notes significance at the five percent level **Notes significance at the one percent level 36 ** ** ** ** ** ** ** ** 0.0083 ** -2.40 ** -6.60 -13.02 -7.39 -19.55 -8.16 -21.85 -24.24 -3.24 -0.11 -4.61 11.84 17.71 0.95 -18511.2 14 37050 37025 ** ** * ** ** ** ** * ** ** ** Model E 20.78 ** -1.22 * -0.11 ** 0.0084 ** -2.23 ** -6.65 -12.45 -7.36 -18.90 -7.60 -21.27 -24.04 ** ** * ** * ** ** 11.65 ** 17.43 ** 0.95 ** -18533.4 12 37091 37069 Model F 21.64 ** -1.29 * -0.09 0.0085 -2.27 -0.01 -6.46 -13.12 -7.49 -19.73 -8.37 -22.11 -24.35 -2.68 2.18 -1.74 11.66 17.92 0.98 -18446.5 16 36925 36896 ** ** ** ** ** * ** ** ** ** ** ** ** In order to pick the most appropriate model in a panel setting, Frees (2004) recommends Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC). Both statistics allow for comparison between both nested and non-nested models, making them fitting for this application. They are calculated: AIC = -2 * ln(maximized likelihood) + 2 * (number of model parameters) and BIC = -2 * ln(maximized likelihood) + ln(number of model parameters). For models with the same number of parameters, the AIC is equivalent to maximizing the log likelihood. The preferred model is indicated by the smallest AIC or BIC statistic. The difference between the two is that the BIC gives greater weight to the number of parameters, and using both measures gives greater confidence in the results. Model F returns the lowest AIC and BIC, very closely followed by Model A. The difference between the models is the inclusion of the futures price in Model F, which is found to be statistically insignificant. Due to the closeness of the AIC and BIC, for the remainder of the study Model A will be regarded as the base model. Most importantly, all models indicate that ethanol production is statistically significant and estimate a coefficient value which is similar in magnitude. Overall, the models appear to be relatively similar, with the exception being Model B. When the stocks ratio is used rather than the production ratio Illinois, Iowa, Kansas, and Nebraska have positive values. A more detailed explanation of variable interpretation will follow in Section 4.3. 37 4.3 PARAMETER ESTIMATES As noted in Section 4.2, Model A serves as the base model for the remainder of this study. Table 4.2 reports the coefficients and t-values for this model. All parameters, except the seasonal dummy variables, are statistically significant at the five percent level. Table 4.2 Model Estimates Variable Coefficient 20.3183 Intercept -1.2884 Production Ratio -0.1049 Diesel Price 0.0085 Ethanol Production -2.1638 Interest Rate -6.4601 Illinois -13.1158 Iowa -7.4836 Kansas -19.7159 Minnesota -8.3617 Nebraska -22.0950 South Dakota -24.3472 Wisconsin -2.3943 Fall 2.2472 Spring -1.3769 Summer 11.6639 𝛔𝟐𝐯 17.9099 𝛔𝟐𝛍 0.9489 λ *Notes significance at the five percent level **Notes significance at the one percent level 38 ** * ** ** ** ** ** * ** * ** ** ** ** ** T-values 4.6724 -2.30243 -9.80638 7.65322 -4.85065 -2.86512 -5.06022 -2.24027 -7.11868 -2.53333 -6.9716 -5.32627 -1.05198 1.05091 -0.93145 54.58875 7.43203 22.1769 The parameters of the model, β, can be interpreted as partial derivatives, similar to the least-squares interpretation (Lesage & Pace, 2009). Thus, δyi/δxir = βr for all i, r and δyi/δxjr = 0, for all j≠i and for all variables r. The production ratio measures local corn production as a percentage of national corn production. As expected, the coefficient has a negative sign. This implies an increase in local supply, relative to national supply, will cause the basis to weaken. The coefficient implies a one percent increase in local corn production relative to national production results in a decreased basis of -1.29 cents. The coefficient of diesel price is also negative. Transportation costs greatly influence basis and the results indicate that as transport costs rise, the basis weakens. The coefficient of -0.1 implies that if the price of diesel rises by 10 cents it is expected that the basis will widen by one cent. The ethanol coefficient in positive and indicates an ethanol production positively affects basis. An ethanol plant is predicted to strengthen basis by .0085 cents per million gallons of ethanol production. This means a 50 MGY plant will increase basis by 0.425 cents. The interest rate is negative as expected. It is a proxy for the opportunity cost of storage, and as the opportunity cost of holding grain increases it is expected that more will be sold on the cash market, causing the basis to weaken. A one percent increase in the interest rate will cause the basis to widen by 2.12 cents. All of the states had statistically significant coefficients when compared to the base state of Indiana. The basis is Indiana is fairly strong relative to the other states in the sample (see Figures 3.1 and 3.2) so it is not surprising that the dummy coefficients for the other states are 39 negative. It should also be noted, none of the seasonal dummy variables are statistically significant. The model also estimates λ, σ2v and σ2μ to be statistically significant. The spatial autocorrelation coefficient is denoted as λ, σ2μ is the variance of the random-effects vector and σ2v is the variance of the error vector. Their significance implies they are needed in the model, further verifying the model’s validity. 4.3.1 Weights Matrix An important feature of spatial models, and one that is frequently ignored in the literature, is the construction of the NxN spatial weights matrix. There are numerous ways to build a weights matrix and for this particular model McNew and Griffith (2005) suggest that any locations within 50 miles will be correlated. Nevertheless, the field of spatial econometrics provides no set rules for picking a Wmatrix structure so it is best to try a variety of specifications to test the robustness of the results. This study creates and tests five different weights matrices in the model. Using the latitude and longitude of the county centroid for all counties with a dependent variable observation (N=153), the following spatial weight matrices are constructed: - 50-mile Buffer – For county i, the neighborhood set includes all counties with centroids within 50 miles of the centroid of county i. In this specification, counties have between one and ten neighbors. - Contiguous Counties – For county i, the neighborhood set includes all counties contiguous to county i. 40 - Nearest Neighbors (NN) - For county i, the neighborhood set includes the m counties nearest to county i. The specifications examined are m = 5, 7, and 10. The model was run with these different W-matrix specifications and the results are shown in Table 4.3. Table 4.3 Spatial Weight Variations Intercept Prod Ratio Diesel Ethanol Interest Illinois Iowa Kansas Minnesota Nebraska South Dakota Wisconsin Fall Spring Summer 𝛔𝟐𝐯 𝛔𝟐𝛍 λ Likelihood Value 50-Miles Contiguous N=5 N=7 20.32 ** 11.15 ** 20.28 ** 13.74 * -1.29 * -2.20 ** -3.18 ** -3.30 ** -0.10 ** -0.08 ** -0.10 ** -0.08 ** 0.0085 ** 0.0110 ** 0.0135 ** 0.0113 ** -2.16 ** -1.86 ** -2.69 ** -3.17 ** -6.46 ** -3.75 -4.48 * -0.0004 -13.12 ** -11.02 ** -9.44 ** -4.70 ** -7.48 * -5.87 -6.56 ** -0.02 ** -19.72 ** -17.81 ** -17.00 ** -12.17 ** -8.36 * -10.27 ** -10.37 ** -4.15 * -22.10 ** -22.07 ** -21.65 ** -16.76 ** -24.35 ** -18.42 ** -19.27 ** -15.99 ** -2.39 1.58 2.01 3.07 2.25 2.12 6.16 ** 6.85 * -1.38 -0.52 3.95 5.59 11.66 ** 13.51 ** 12.88 ** 12.82 ** 17.91 ** 26.43 ** 26.11 ** 27.36 ** 0.95 ** 0.97 ** 0.96 ** 0.96 ** -18447 -17902 -17808 -17579 *Notes significance at the five percent level **Notes significance at the one percent level T-values have been omitted from this table 41 N=10 24.83 -2.96 ** -0.07 0.0120 ** -5.90 ** -4.07 * -2.51 3.02 -8.59 ** -1.61 -12.61 ** -15.86 ** 0.03 10.45 16.51 13.36 ** 25.92 ** 0.99 ** -17627 Lasage and Pace (2009) note there is not a formal measure to compare models with different spatial weight matrices because they are not nested models. Still, they recommend comparing the log-likelihood function values. This measure indicates the best model is when the W-matrix is specified with seven neighbors. When comparing different W-matrix specifications the main variable of interest, ethanol production, is always statistically significant. Interestingly, the 50-mile neighbor relationship specified by the literature returns the lowest estimate for ethanol production’s impact on basis. The seven-neighbor model estimates the impact of ethanol production to be 0.0113 per million gallons, which equates to a 0.565 cent increase in basis for a 50 MGY plant. The five-neighbor matrix returns the largest ethanol production impact and predicts a .675 cent increase for a 50 MGY plant. 4.4 TIME COMPARISON For computational ease the data in the study was aggregated in to quarterly time periods as discussed in Section 3.3. As this step reduces variation, it is necessary to verify the results do not significantly change as a consequence. To ensure varying time units do not yield different results, separate models were run using both quarterly and monthly time periods for the states of Illinois, Iowa, and Kansas. These states were selected because of their high number of county observations in the data. Table 4.4 compares the observations over the full sample time period of October 1999 through September 2009 using quarterly observations in Model Q and monthly observations in 42 Table 4.4 Compare Monthly and Quarterly Data Illinois January Model Q 7.3887 (1.02) 2.5985 (0.99) -0.118 ** (-6.53) 0.0053 (1.46) -1.6445 * (-2.21) 3.1845 (0.78) 5.3328 (1.30) 1.7289 (0.40) X February X March X April X May X June X July X August X September X October X November X Intercept Prod. Ratio Diesel Ethanol Prod. Interest Rate Winter Spring Summer 𝛔𝟐𝐯 𝛔𝟐𝛍 λ 11.81 (25.74) 18.56 (3.85) 0.94 (10.91) ** ** ** Iowa Model M 5.6286 (1.31) 2.5905 (1.17) -0.1097 ** (-11.81) 0.0059 * (2.07) -1.5913 * (-4.07) X X X 0.0198 (0.36) 5.3393 (1.86) 1.5001 (0.61) 6.5565 (2.27) 2.5564 (0.88) 4.7021 (1.55) 3.2697 (1.06) 6.3509 (2.18) -6.3120 (-2.07) -4.1511 (-1.42) 4.9603 (1.62) 19.19 (44.97) 19.92 (3.87) 0.94 (9.10) Model Q 16.7189 (1.10) -11.121 ** (-4.22) -0.1511 ** (-3.34) 0.0068 ** (4.53) -2.3777 (-1.27) 9.5181 (0.92) 16.7766 (1.64) 5.2033 (0.53) x x x * x x x x * x * x x x ** ** ** 11.94 (28.61) 16.39 (3.99) 0.97 (8.94) *Notes significance at the five percent level **Notes significance at the one percent level T-values reported in parentheses 43 ** ** ** Kansas Model M 6.4620 * (2.14) -9.3507 ** (-4.40) -0.0864 ** (-14.37) 0.0078 ** (7.55) -2.4114 ** (-9.70) X X X 1.7902 (1.00) 6.1760 (3.57) 2.4707 (1.43) 6.8406 (3.68) 2.0747 (1.07) 5.2190 (2.95) 0.0046 (0.31) 4.2921 (2.30) -5.3528 (-2.84) -3.9794 (-2.00) 4.2610 (2.32) 18.07 (49.96) 23.61 (4.12) 0.97 (14.14) Model Q -8.6216 (-1.53) -8.7416 ** (-4.98) -0.1034 ** (-11.21) 0.0290 ** (3.37) -1.0418 ** (-2.87) 1.5210 (0.79) 3.7244 (1.89) 2.1851 (1.10) X X X X X X X X X X X ** ** ** 19.89 (18.22) 32.49 (2.13) 0.98 (11.17) ** * ** Model M -13.584 ** (-2.46) -9.8474 ** (-8.19) -0.1018 ** (-16.47) 0.0269 ** (4.68) -0.9656 ** (-3.94) X X X 4.0990 (2.20) 6.8035 (3.61) 1.8101 (1.04) 6.4703 (3.39) 3.4880 (1.83) 9.0699 (4.79) 5.1208 (2.71) 8.2566 (4.37) 1.0314 (0.66) 1.8171 (1.04) 6.0732 (3.22) 27.35 (31.83) 37.55 (2.26) 0.98 (11.82) * ** ** ** ** ** ** * ** Model M. A visual inspection of that data shows that generally the variables are similar in sign and magnitude across models. A few notable differences include: ethanol production is not statistically significant in the Illinois quarterly model whereas it is in the monthly model; the interest rate is not statistically significant in the Iowa quarterly model whereas it is in the monthly model; some of the monthly dummy variables are statistically significant whereas none of the seasonal dummy variables are significant at the five percent level. 4.5 COMPARISON TO LITERATURE The impact of ethanol on corn basis found in this study is somewhat surprising when compared to results in previous work. Here it is estimated a 50 MGY ethanol plant within 50 miles of a county centroid has a 0.425 cent impact on local basis, whereas others suggest the impact may be greater. This section aims to not only further validate this study’s results, but also gain insight into these inconsistent findings. To achieve this end, the model employed here is closely compared to the McNew and Griffith (2005) study. The McNew and Griffith study examined the regions surrounding 12 different ethanol plants from March 2000 to March 2003. They find that in the 150-mile square region surrounding the plant the average impact is a 5.9 cent increase in basis. Besides the different estimation of the ethanol production impact, there are some important differences to note between the studies: - The time period used in the McNew and Griffith study is a sub-sample of the time period used in this study. It should be noted the Midwestern average annual corn basis was continually increasing between 2000 and 2003, whereas over the full time 44 period of 1999 to 2009 the average annual basis is decreasing. (See Figure 3.2 for a visual representation.) - To estimate their model, McNew and Griffith use state-level corn production, national-level corn production, monthly dummy variables, a dummy variable for ethanol production, and a sophisticated transportation variable. Differences in this study’s analysis are the use of a corn production ratio, a more crude proxy for transportation, and interest rate as a proxy for storage costs. - The McNew and Griffith study uses locations that were within approximately 75 miles of a new ethanol plant. They specifically look at regions with new ethanol plants whereas the data set used in this study contains counties with pre-existing plants, counties which gain plants during the 10-year period, and some counties which never have a plant. To test the model used in this study a sub-sample of the data was extracted from the data set. This sub-sample only includes counties from Illinois, Iowa, South Dakota, and Wisconsin because these are the states used by McNew and Griffith (they also have one observation in Missouri). Also, in keeping with McNew and Griffiths selection criteria, the sub-sample only includes counties where an ethanol plant opened between Spring 2000 and Spring 2003. Near the end of their report, McNew and Griffith state they were unable to identify whether the price impacts would persist over time due to data constraints. Over half of the plants in their sample had been open for less than six months. They predicted over time the price impact of a plant will diminish as market conditions adjust to the new demand center (p. 176). Thus, they conclude their estimates are likely measures of short-term impacts and not indicative of an ethanol plant’s long-term price impact. 45 Table 4.5 contains the estimates of six different models used to compare the results of this study to the results of McNew and Griffith. To mimic the time period used by McNew and Griffith the data is separated into Time 1 (Fall 1999 – Summer 2003) and Time 2 (Fall 2003 – Summer 2009). Thus the models run are: - Sub-sample in Time 1 - This model is the direct comparison to McNew and Griffith. - Sub-sample in Time 2 – This model investigates whether the impact found by McNew and Griffith persists over time. - Sub-sample over the full time period. - Full sample in Time 1 – This model examines whether the impacts found by McNew and Griffith hold when the model includes counties with and without an ethanol plant. - Full sample in Time 2 - Full sample over the full time period. (Note: Table 4.5 expands over two pages.) 46 Table 4.5: Compare Full Sample and Sub-Sample Estimates Intercept Production Ratio Diesel Ethanol Illinois South Dakota Subsample (N=22) Wisconsin Winter Spring Summer σ2v σ2μ λ Fall 1999 Summer 2003 (Time 1) Fall 2003 Summer 2009 (Time 2) -38.1568 (-6.10) 0.21435 (0.20) -0.01474 (-0.38) 0.13347 (11.51) 15.36167 (4.32) -5.22206 (-1.75) 9.37036 (2.00) 4.08061 (3.36) 4.42004 (3.62) -0.99808 (-0.81) 17.24287 (12.85) 16.57584 (3.00) 0.56 (2.30) 23.34869 (2.80) -13.4581 (-5.34) -0.10721 (-17.33) 0.02065 (6.98) 11.90544 (2.73) -22.607 (-4.39) -14.7958 (-1.98) 2.64308 (2.42) 4.31058 (3.95) 1.96208 (1.78) 42.16872 (15.77) 29.13658 (2.18) 0.30 (1.25) ** ** ** * ** ** ** ** * 47 ** ** ** ** ** ** * * ** ** * Fall 1999 Summer 2009 (Full Time) -13.6663 (-2.64) -3.97666 (-2.59) -0.06996 (-10.31) 0.00751 (2.47) 11.66646 (2.38) -7.41843 (-2.10) 1.43494 (0.28) 2.22307 (1.58) 4.46042 (3.16) 3.37015* (2.39) 23.62253 (20.89) 11.30683 (2.90) 0.86 (5.48) ** ** ** * * * ** * ** ** ** Table 4.5 Continued Fall 1999 Summer 2003 (Time 1) Intercept Full Sample (N =153) Fall 2003 Summer 2009 (Time 2) -15.8176 * 13.93429 (-2.02) (4.60) Production -1.59076 ** -2.50443 Ratio (-2.74) (-3.55) Diesel 0.02141 -0.12038 (0.43) (-17.19) Ethanol 0.00779 * 0.01126 (2.38) (7.68) Illinois -5.61763 * -5.07186 (-2.56) (-2.24) Iowa -14.2573 ** -14.4399 (-5.94) (-6.40) Kansas -7.04125 * -8.46234 (-2.31) (-3.01) Minnesota -21.7964 ** -20.6492 (-8.49) (-8.59) Nebraska -8.54345 ** -12.5121 (-2.87) (-4.70) South Dakota -25.5823 ** -23.0244 (-8.50) (-7.93) Wisconsin -21.5224 ** -25.4419 (-4.94) (-5.96) Fall 3.626 * 1.41794 (2.17) (0.92) Spring 5.30165 ** 4.25605 (5.30) (2.95) Summer 0.00749 2.51643 (0.25) (1.72) 2 5.4353 ** 14.22327 σv (33.80) (41.80) 2 18.24325 ** 24.77653 σμ (7.45) (7.33) 0.93 ** 0.87 λ (21.05) (12.02) * Notes significance at the five percent level **Notes significance at the one percent level T-values reported in parentheses 48 ** ** ** ** * ** ** ** ** ** ** ** ** ** ** Fall 1999 Summer 2009 (Full Time) 5.994 (1.08) -1.10714 (-1.92) -0.12284 (-7.95) 0.00833 (7.43) -6.87933 (-2.77) -11.0317 (-3.49) -6.87786 (-1.42) -17.2101 (-5.07) -6.19979 (-1.29) -19.5501 (-5.20) -22.9762 (-4.62) 3.12733 (0.95) 6.36547 (1.78) 2.16525 (0.85) 11.64203 (54.61) 16.75687 (7.51) 0.96 (29.23) ** ** ** ** ** ** ** ** ** ** In all six models the results of ethanol production, the main variable of interest, come back statistically significant. It should also be noted that McNew and Griffith did not use a measure of storage or storage costs so it was not included here. Table 4.6 shows the impacts a 50 MGY ethanol production plant will have on basis, as estimated by each of the models. Table 4.6: Ethanol Impacts Ethanol Impact on Basis Time 1 Time 2 Full Time Sub-Sample 6.67 1.03 0.38 Full Sample 0.39 0.56 0.42 When looking at the sub-sample in Time 1, which is designed to mimic the sample of McNew and Griffith, a basis improvement of 6.67 cents per bushel is found. This is in the range of the 5.9 cent impact found by McNew and Griffith, because they cite improvements ranging between 1.5 cents and 12 cents for individual plants. Additionally, the 50 MGY size used in these calculations is larger than the average plant size used in the McNew and Griffith sample, yet they indicate the plant size is relatively unimportant. It is found that when all locations are examined in Time 1, the price impact of an ethanol plant is considerably less. This could be for a variety of reasons. First, McNew and Griffith may over attribute some of the increase in basis to ethanol production. Over the time period, both ethanol production and basis are continually increasing so the lack of locations without ethanol plants may lead to over estimating the impacts. A second and more probable explanation is that the lack of price impact from local ethanol production in the full sample does not mean there is no impact; rather it may mean the price impacts from ethanol production are being spread well beyond the 50-mile region defined 49 by the model. This means even counties without an ethanol plant within 50 miles are still receiving positive price impacts from ethanol production. Thus, when including counties in the sample without plants, the direct impact of local ethanol production is diluted as impacts are still being felt by counties further away. Another interesting finding in Table 4.6 is that over time the price impact of ethanol plants seems to decrease. Examining the sub-sample, in Time 1 the impact is 6.67 cents, but in Time 2 the impact of ethanol production is 1.03 cents. As suggested by McNew and Griffith (2005), it appears that the price impact diminishes over time as market conditions evolve to meet new centers of demand. The extended sample period allows for the examination of more longterm price impacts, rather than capturing the short-term price adjustments. These long-term impacts are expected to be smaller, possibly contributing to the difference between this study and previous literature. As previously noted, McNew and Griffith do not account for storage opportunity costs whereas this study takes them into account. The six models were re-run, this time including the interest rate and the results are in Table 4.7. (Note: Table 4.7 expands over two pages) 50 Table 4.7 Compare Full Sample and Sub-Sample Estimates with Interest Fall 1999 Summer 2003 Intercept Production Ratio Diesel Ethanol Interest Illinois South Dakota Sub-sample (N=22) Wisconsin Winter Spring Summer σ2v σ2μ λ -37.01 (-9.56) -1.77 (-1.86) 0.21 (8.11) 0.0388 (4.16) -3.48 (-21.04) 15.26 (5.27) -7.43 (-2.91) 5.56 (1.33) 2.95 (4.27) 2.82 (4.05) -3.67 (-5.19) 9.02 (12.85) 13.28 (3.19) 0.38 (1.31) 51 ** ** ** ** ** ** ** ** ** ** ** ** Fall 2003 Summer 2009 20.39 (2.70) -11.39 (-4.85) -0.094 (-14.30) 0.0182 (6.10) -0.99 (-4.30) 11.31 (2.96) -19.44 (-4.14) -11.33 (-1.68) 2.88 (2.87) 3.91 (3.88) 1.34 (1.32) 58.54 (15.58) 22.26 (2.15) 0.06 (0.24) ** ** ** ** ** ** ** ** ** ** * Fall 1999 Summer 2009 (Full) -1.45 (-0.37) -2.95 ** (-2.56) -0.061 ** (-10.99) 0.0064 ** (2.40) -2.64 ** (-13.88) 11.33 ** (3.97) -7.11 ** (-2.69) 3.00 (0.76) 2.72 ** (2.75) 3.53 ** (3.55) 0.68 (0.67) 24.53 ** (20.73) 10.01 ** (3.08) 0.61 ** (2.42) Table 4.7 Continued Fall 1999 Summer 2003 Intercept Fall 2003 Summer 2009 -16.15 * 14.96 (-2.58) (4.71) Production -1.88 ** -2.55 Ratio (-3.26) (-3.59) Diesel 0.240 ** -0.118 (5.43) (-16.45) Ethanol 0.0075 * 0.0113 (2.29) (7.63) Interest -4.19 ** -0.26 (-17.53) (-1.01) Illinois -4.74 * -5.00 (-2.37) (-2.19) Iowa -14.67 ** -14.46 (-7.10) (-6.30) Kansas -6.96 ** -8.46 (-2.64) (-2.96) Minnesota -23.06 ** -20.67 (-10.27) (-8.43) Full Sample (N =153) Nebraska -9.83 ** -12.62 (-3.93) (-4.67) South Dakota -27.33 ** -23.05 (-10.10) (-7.77) Wisconsin -21.34 ** -25.35 (-5.38) (-5.90) Fall 1.95 1.39 (1.36) (1.05) Spring 2.52 4.11 (1.83) (3.04) Summer -3.30 * 2.41 (-2.40) (1.80) 2 5.11 ** 14.27 σv (33.75) (41.80) 20.45 ** 25.09 σ2μ (7.60) (7.34) 0.90 ** 0.86 λ (7.60) (11.80) * Notes significance at the five percent level **Notes significance at the one percent level T-values reported in parentheses 52 ** ** ** ** * ** ** ** ** ** ** ** ** ** ** Fall 1999 Summer 2009 (Full) 20.32 ** (4.67) -1.29 * (-2.30) -0.105 ** (-9.80) 0.0085 ** (7.65) -2.16 ** (-4.85) -6.46 ** (-2.87) -13.12 ** (-5.06) -7.48 * (-2.24) -19.72 ** (-7.12) -8.36 * (-2.53) -22.10 ** (-6.97) -24.35 ** (-5.33) -2.39 (-1.05) 2.25 (1.05) -1.38 (-0.93) 11.66 ** (54.59) 17.91 ** (7.43) 0.95 ** (22.18) Again the price impacts of a 50 MGY ethanol plant are calculated and can be found in Table 4.8. When accounting for storage opportunity costs the impact of ethanol production in the sub-sample during Time 1 decreases from 6.67 cents to 1.94 cents. This may indicate an unaccounted for factor played an important role in predicting basis, and its absence resulted in an over-estimation of the price impacts. The difference between estimations in the sub-sample and the full sample remains and it is believed that this variation is still a result of the factors described above. Table 4.8 Ethanol Impacts with Interest Ethanol Impact (with Interest) Sub-Sample Full Sample Time 1 Time 2 Full Time 1.94 0.38 0.91 0.56 0.32 0.43 53 CHAPTER 5 CONCLUSION 5.1 SUMMARY The ethanol industry has experienced rapid growth over the past decade. This growth primarily occurred in the starch-based ethanol sector, resulting in much greater demand for corn. While it is certain there is an impact on the national grain sector, the magnitude of this impact on local corn basis, especially in the long run, has been sparsely studied. This thesis aimed to measure the impact ethanol plants have on local corn basis. The specific objectives were to: 1. develop and estimate a spatial panel model of corn basis; 2. assess the impacts of ethanol plants on the local corn basis; and 3. determine if these impacts are consistent with the short-run impacts found in previous studies. Given the above objectives, a description of the ethanol and corn markets, along with a discussion of the linkages between the two was provided. A review of economic literature on the topic reveals ethanol plants have a documented impact on market conditions. Specifically, some research suggests ethanol plants have the ability to strengthen basis, though other research fails to reach these conclusions. These inconsistencies, coupled with the fact that no research looks from the beginning of the ethanol growth period to present, rationalizes this study’s purpose of determining how ethanol plants affect corn basis. With a focus on predicting corn basis, a model was built to account for supply and demand factors which influence the relationship between the local and national prices. These 54 include the amount of ethanol produced within 50 miles, a ratio of local and national corn production, diesel price as a proxy for transportation, interest rate as proxy for storage opportunity cost, and state and seasonal dummy variables. To measure the impact of ethanol on corn basis, data from 153 Midwestern counties over 40 quarters was used. Tests were conducted which supported the use of a random effects model and the hypothesis of spatial dependency. The spatial nature of the data can lead to biased or inefficient estimates when using OLS, thus this study built a spatial error components model to estimate the impact ethanol has on corn basis. The econometric results indicated local ethanol production has a statistically significant and positive impact on local corn basis. The results predict on average the entrance of a 50 MGY ethanol production plant in the Midwest will increase local basis between 0.40 and 0.65 cents, dependent upon the specification of the weights matrix. Still, these results are much smaller than those predicted by previous work. 5.2 CONCLUSIONS The following conclusions are drawn in response to the objectives and are based on the empirical results from Chapter 4, as well as arguments developed in previous chapters. In response to Objective 1, a spatial error component model was built to account for the spatial and panel nature of the data. The Hausman Test verified the use an error component model and Lagrange Multiplier tests supported the use of a spatial error model. In regards to Objective 2, the impact of having an ethanol production plant within 50 miles was measured to be a 0.40 cent basis improvement. 55 In response to Objective 3, it was determined that the long-term price impact of ethanol production is considerably less than the impact found in the short-run. The data was able to closely replicate short-term finding of previous studies, but over time the impacts were found to decrease. The findings also suggest the price impacts of ethanol production reach further than the 50 mile radius assigned by this study. 5.3 SUGGESTIONS FOR FURTHER RESEARCH Additional research in this area will be of benefit to allow for a deeper understanding of the long-run impacts of ethanol on corn basis. Three extensions or modifications of the current study that will allow for a great understanding of the topic are to improve measures of transportation costs, investigate the reach of ethanol price impacts, and provide region specific estimates. Transportation costs are one of the driving forces in determining basis and in this study they were modeled using the proxy of Midwest monthly average diesel price to capture the variation in cost over time. However, transportation costs also vary over space so some measure of distance to a terminal grain market may be useful. The depth of the McNew and Griffith (2005) data set allowed them to account for the specific distance grain travels, but this analysis only partially accounts for difference over space by using state-level dummy variables. Including specific transportation distance to terminal market variables may allow for a more complete analysis of specific price impacts. Additionally, the findings of this study suggest ethanol’s price impact extends beyond the 50 mile county buffer as defined in the model. It would be interesting in future work to include 56 an ethanol production and distance interaction term to estimate the reach of ethanol production price impacts. This type of analysis was conducted in the McNew and Griffith (2005) study, but it would be useful to see this analysis updated with more long-term estimates. Another interesting extension of the current research would be to narrow the scope of the research to investigate the impacts of ethanol in particular regions. McNew and Griffith were able to show a wide range of impacts depending on the plant, so it would be useful to see if that is also true on a long-range scale. It is likely that the impacts of ethanol production on corn basis are greater in corn deficit areas. Finally, to truly understand how ethanol production affects basis more work will need to be done as the industry matures. During the past decade the grain market had to first adapt to the increased demand for corn by the ethanol industry. Then, the industry had to adjust to the large upswing, then downswing in corn prices. Now, without changes in EPA blending policies, it is likely that the growth rate of the ethanol industry has nearly plateaued as the Renewable Fuel Standard Program caps corn-based ethanol production at 15 billion gallons. It cannot be denied that the use of corn in ethanol production has drastically altered the grain market. Today over 30 percent of the nation’s corn goes into ethanol production and increased demand has been shown to be partially responsible for the price run-up in 2007-08. However this study finds that at the most local level, between Fall 1999 and Summer 2009, the mere presence of ethanol production within 50 miles is not likely to induce a large long-term shift in corn basis. 57 APPENDIX A The Renewable Fuel Standard Program (RFS2) was updated from the original RFS program developed by the Energy Independence and Security Act of 2007. It increased the previous mandates to specify production of renewable fuels to equal at least 36 billion gallons by 2022. It breaks the volume requirements into four categories: 1) renewable biofuel – ethanol derived from corn-starch, 2) advanced biofuel – essentially anything but corn-starch ethanol, 3) cellulosic biofuel – fuel produced from cellulose, hemicelluloses, or lignin, 4) biomass-based diesel – diesel from fats and oils, not co-processed with petroleum. Table A.1: Renewable Fuel Standard Program Mandates – Billion Gallons Year Renewable Biofuel Advanced Biofuel Cellulosic Biofuel 9 2008 10.5 0.6 2009 12 0.95 2010 12.6 1.35 2011 13.2 2 2012 13.8 2.75 2013 14.4 3.75 2014 15 5.5 2015 15 7.25 2016 15 9 2017 15 11 2018 15 13 2019 15 15 2020 15 18 2021 15 21 2022 Source: Renewable Fuels Association Biomassbased Diesel 0.1 0.25 0.5 1 1.75 3 4.25 5.5 7 8.5 10.5 13.5 16 58 0.5 0.65 0.8 1 Undifferentiated Total RFS Advanced Biofuel 0.1 0.2 0.3 0.5 1.75 2 2.5 3 3.5 4 4.5 4.5 4.5 5 9 11.1 12.95 13.95 15.2 16.55 18.15 20.5 22.25 24 26 28 30 33 36 APPENDIX B Table B.1: Ethanol Plants Included in the Sample Firm Name Abengoa Bioenergy Corp. Abengoa Bioenergy Corp. Abengoa Bioenergy Corp. Absolute Energy, LLC* ACE Ethanol, LLC Adkins Energy, LLC* Advanced Bioenergy, LLC AGP* Agri-Energy, LLC* Al-Corn Clean Fuel* AltraBiofuels Indiana, LLC Amaizing Energy, LLC* Archer Daniels Midland Archer Daniels Midland Archer Daniels Midland Archer Daniels Midland Archer Daniels Midland Archer Daniels Midland Arkalon Energy, LLC Aventine Renewable Energy, LLC Aventine Renewable Energy, LLC Badger State Ethanol, LLC* Big River Resources Galva, LLC Big River Resources, LLC* Big River United Energy BioFuel Energy - Buffalo Lake Energy, LLC BioFuel Energy - Pioneer Trail Energy, LLC Bonanza Energy, LLC Bridgeport Ethanol Bushmills Ethanol, Inc.* Cardinal Ethanol Cargill, Inc. Cargill, Inc. Castle Rock Renewable Fuels, LLC Center Ethanol Company Year Opened 2002 2007 1993 2008 2002 2002 2007 1995 1999 1996 2008 2005 1981 1982 2009 1978 2002 1980 2007 1995 1981 2002 2009 2004 2008 2008 2007 2007 2008 2005 2008 1995 NA 2007 2008 59 County Sedgwick Buffalo York Mitchell Chippewa Stephenson Fillmore Adams Rock Dodge Putnam Crawford Linn Clinton Platte Macon Lyon Peoria Seward Hamilton Tazewell Green Henry Des Moines Dubuque Martin Hall Finney Morrill Kandiyohi Randolph Washington Wapello Juneau St. Clair State Nameplate Capacity (MGY) KS 25 NE 88 NE 55 IA 110 WI 41 IL 40 NE 100 NE 52 MN 21 MN 42 IN 92 IA 55 IA 420 IA 237 NE 300 IL 290 MN 40 IL 100 KS 110 NE 50 IL 157 WI 48 IL 100 IA 100 IA 110 MN 115 NE 115 KS 55 NE 54 MN 50 IN 100 NE 85 IA 35 WI 50 IL 54 Central Indiana Ethanol, LLC Central MN Ethanol Coop* Chief Ethanol Chippewa Valley Ethanol Co.* Corn Plus, LLP* Corn, LP* Cornhusker Energy Lexington, LLC Dakota Ethanol, LLC* DENCO, LLC Didion Ethanol E Energy Adams, LLC E3 Biofuels East Kansas Agri-Energy, LLC* ESE Alcohol Inc. Gateway Ethanol Glacial Lakes Energy, LLC - Mina Glacial Lakes Energy, LLC* Global Ethanol/Midwest Grain Processors Golden Grain Energy, LLC* Grain Processing Corp. Granite Falls Energy, LLC* Green Plains Renewable Energy Green Plains Renewable Energy Green Plains Renewable Energy Green Plains Renewable Energy Green Plains Renewable Energy Guardian Energy Hawkeye Renewables, LLC Hawkeye Renewables, LLC Hawkeye Renewables, LLC Hawkeye Renewables, LLC Heartland Corn Products* Heartland Grain Fuels, LP Heartland Grain Fuels, LP Heron Lake BioEnergy, LLC Highwater Ethanol LLC Homeland Energy Husker Ag, LLC* Illinois River Energy, LLC Iroquois Bio-Energy Company, LLC 2007 1999 1985 1996 1994 2005 2005 2001 1999 2008 2007 2007 2005 1991 2007 2008 2001 2002 2004 NA 2005 2008 2004 2009 2007 2008 2009 2006 2003 2008 2008 1995 2008 1998 2007 2009 2009 2003 2006 2007 60 Grant Morrison Adams Swift Faribault Wright Dawson Lake Stevens Columbia Gage Saunders Anderson Wichita Pratt Edmunds Codington Kossuth Cerro Gordo Muscatine Yellow Medicine Wells Merrick Valley Page Dickinson Waseca Bremer Hardin Guthrie Butler Sibley Brown Beadle Jackson Redwood Chickasaw Pierce Ogle Jasper IN MN NE MN MN IA NE SD MN WI NE NE KS KS KS SD SD IA IA IA MN IN NE NE IA IA MN IA IA IA IA MN SD SD MN MN IA NE IL IN 40 21.5 62 45 44 60 40 50 24 40 50 25 35 1.5 55 107 100 98 115 20 52 110 100 50 55 55 110 110 90 110 110 100 50 32 50 55 100 75 100 40 KAAPA Ethanol, LLC* Kansas Ethanol, LLC Lincolnland Agri-Energy, LLC* Lincolnway Energy, LLC* Little Sioux Corn Processors, LP* Louis Dreyfus Commodities Louis Dreyfus Commodities Marquis Energy, LLC Mid America Agri Products/Horizon Mid America Agri Products/Wheatland Midwest Renewable Energy, LLC Minnesota Energy* NEDAK Ethanol Nesika Energy, LLC New Energy Corp. North Country Ethanol, LLC* NuGen Energy One Earth Energy Otter Tail Ag Enterprises Patriot Renewable Fuels, LLC Penford Products Pine Lake Corn Processors, LLC Platinum Ethanol, LLC* Plymouth Ethanol, LLC* POET Biorefining - Alexandria POET Biorefining - Ashton POET Biorefining - Big Stone POET Biorefining - Bingham Lake POET Biorefining - Chancellor POET Biorefining - Coon Rapids POET Biorefining - Corning POET Biorefining - Emmetsburg POET Biorefining - Glenville POET Biorefining - Gowrie POET Biorefining - Hanlontown POET Biorefining - Hudson POET Biorefining - Jewell POET Biorefining - Lake Crystal POET Biorefining - Mitchell POET Biorefining - North Manchester 2003 2008 2004 2006 2003 2009 2007 2008 2007 2007 1999 1997 2008 2008 1984 1994 2008 2009 2008 2008 2008 2005 2008 2009 2008 2004 2002 1997 2003 2002 2006 2005 1999 2006 2004 2004 2006 2005 2006 2008 61 Kearney Rice Crawford Story Cherokee Greene Madison Putnam Furnas Perkins Lincoln Renville Holt Republic St. Joesph Roberts Turner Ford Otter Tail Henry Linn Hardin Ida Plymouth Madison Osceola Grant Cottonwood Turner Carroll Adams Palo Alto Freeborn Webster Worth Lincoln Hamilton Blue Earth Davison Wabash NE KS IL IA IA IA NE IL NE NE NE MN NE KS IN SD SD IL MN IL IA IA IA IA IN IA SD MN SD IA IA IA MN IA IA SD IA MN SD IN 40 55 48 55 92 100 45 100 44 44 25 18 44 10 102 20 110 100 57.5 100 45 31 110 50 68 56 79 35 110 54 65 55 42 69 56 56 69 56 68 68 POET Biorefining - Portland POET Biorefining - Preston POET Biorefining - Scotland POET Biorefining- Groton Prairie Horizon Agri-Energy, LLC Quad-County Corn Processors* Redfield Energy, LLC * Reeve Agri-Energy Renew Energy Riverland Biofuels Siouxland Energy & Livestock Coop* Siouxland Ethanol, LLC Southwest Iowa Renewable Energy, LLC * The Andersons Clymers Ethanol, LLC Trenton Agri Products, LLC United Ethanol United WI Grain Producers, LLC* Utica Energy, LLC Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Valero Renewable Fuels Western Plains Energy, LLC* Western Wisconsin Renewable Energy, LLC* * Denotes locally owned plant 2007 1999 1988 2003 2006 2002 2006 1982 2007 2007 2003 2007 2009 2007 2004 2007 2005 2003 2006 2007 2003 2007 2005 2008 2007 2008 2004 2006 Source: Renewable Fuels Association 62 Jay Fillmore Bon Homme Brown Phillips Ida Spink Finney Jefferson Fulton Sioux Dakota Pottawattmie Cass Hitchcock Rock Columbia Winnebago Buena Vista Boone Brookings Floyd Webster O'Brien Montgomery Martin Logan Dunn IN MN SD SD KS IA SD KS WI IL IA NE IA IN NE WI WI WI IA NE SD IA IA IA IN MN KS WI 68 46 11 53 40 30 50 12 130 37 60 50 110 110 40 52 49 48 110 110 120 110 110 110 110 110 45 40 APPENDIX C Table C.1 contains the list of grain elevators which contributing cash corn prices to sample. Table C.1: Grain Elevators State Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois Illinois County Adams Bureau Cass Champaign Christian Clark Dewitt Douglas Grundy Hancock Henry Iroquios Jasper Kane Kankakee Knox La Salle Livingston Logan Marshall Mason McDonough McLean Montgomery Morgan Moultrie Ogle Peoria Piatt Sangamon Shelby Tazewell Whiteside Woodford City Quincy Ohio Beardstown Champaign Stonington Casey Weldon Arcola Minooka Nauvoo Galva Watseka Rushville Elburn Manteno Galesburg Streator Strawn Latham Lacon Easton Bushnell Towanda Raymond Jacksonville Bethany Polo Elmwood Monticello Williamsville Cowden Minier Albany Minonk 63 Company ADM Northern Grain Marketing ADM The Andersons Stonington Coop Grain Huisinga Grain Weldon Coop Okaw Coop Consolidated Grain Colusa Elevator Gateway Coop Watseka Interstate Western Grain Marketing Elburn Coop Farmers Elevator GrainStore Missal Farmers Grain Trainor Grain Farmers Grain ADM Farmers Elevator Biggs and Eas West Central FS Towanda Grain Sorrells Elevator Pisgah Coop Bethany Grain Bocker Grain Inc Ag Land FS TopFlight Grain Culver-Fancy Prairie Coop Tate and Lyle Minier Coop Bunge Ruff Brothers Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Indiana Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Bartholomew Carroll Decatur Fayette Jasper Kosciusko Madison Miami Porter St. Joesph Starke Sullivan Tippecanoe Adair Black Hawk Buchanan Buena Vista Cass Cedar Chickasaw Clayton Delaware Emmet Floyd Franklin Grundy Guthrie Hamilton Hancock Hardin Harrison Henry Humbolt Iowa Jasper Linn Lyon Mahaska Mitchell Monona Columbus Delphi Greensburg Glenwood Remington Warsaw Summitville Amboy Portage South Bend Hamlet Sullivan Lafayette Adair Dunkerton Jesup Alta Massena Clarence New Hampton Clayton Ryan Armstrong Rockford Coulter Beaman Guthrie Center Williams Kanawha Union Modale Mount Union Ottosen Conroy Prairie City Cedar Rapids Little Rock Oskaloosa Stacyville Blencoe 64 Premier Ag The Andersons Lowes Pellets Peavey Co-Alliance Zolman Farms Harvest Land Coop Kokomo Grain Cargill New Energy Corp Starke County Coop ADM Tate and Lyle West Central Coop Dunkerton Coop East Central Iowa Coop First Coop 21st Century Coop River Valley Coop Five Star Farmers Coop Consolidated Grain and Barge Ryan Coop Stateline Coop Farmers Coop AgVantage FS Mid-Iowa Coop Rose Acre Farms Feed Mill Prairie Land Coop North Central Coop Prairie Land Coop United Western Coop Prairie Ag Coop Ottosen Elevator Heartland Coop Heartland Coop ADM Growmark Farmers Coop Society Quad County Grain Northern Country Coop Western Iowa Coop Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Iowa Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Kansas Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Montgomery O'Brien Osceola Palo Alto Polk Scott Sioux Tama Union Van Buren Wapello Webster Winnebago Worth Wright Barton Coffey Dickinson Franklin Gray Kearny Marion Marshall McPhereson Meade Norton Pawnee Reno Russell Sedgwick Sheridan Smith Big Stone Blue Earth Brown Chippewa Dodge Faribault Goodhue Jackson Grant Hartley Ashton West Bend Runnells Davenport Alton Dysart Creston Stockport Eddyville Gowrie Thompson Hanlontown Goldfield Great Bend Le Roy Chapman Ottawa Ingalls Lakin Hillsboro Beattie Moundridge Plains Norton Larned Nickerson Gorham Andale Seguin Athol Barry Lake Crystal Sleepy Eye Clara City Dodge Center Delavan Dennison Heron Lake 65 Hoye Feed and Grain Ag Partners United Farmers Coop Max Yield Coop Runnells Grain Cenex Harvest States Midwest Farmers Coop Tama-Benton Coop DeBruce Grain Roquette Stockport Elevator Cargill West Central Coop Farmers Coop Five Star Coop Gold Eagle Coop Great Bend Coop LeRoy Coop Alida Pearl Coop Ottawa Coop Irsik and Doll Cropland Coop Coop Grain Beattie Coop Mid Kansas Coop Plains Equity Ag Valley Coop Pawnee County Coop Farmers Coop United Ag Andale Farmers Coop Frontier Ag Athol Coop Beardsley Farmers Crystal Valley Coop River Region Coop Farmers Elevator Greenway Coop Watonwan Farm Service Central Valley Coop New Vision Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Minnesota Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska Nebraska South Dakota South Dakota South Dakota South Dakota South Dakota South Dakota South Dakota South Dakota Wisconsin Wisconsin Wisconsin Lac qui Parle Lyon McLeod Mower Nicollet Olmsteade Pipestone Redwood Renville Rock Swift Yellow Medicine Butler Cass Clay Dawson Fillmore Frontier Greeley Jefferson Lancaster Nemaha Phelps Pierce Platte Red Willow Thayer Valley Brookings Brown Grant Hanson Hutchinson Lake Spink Union Columbia Dane Pepin Bellingham Marshall Hutchinson Adams Lafayette Stewartville Jasper Redwood Falls Renville Hills Holloway Clarkfield Bellwood Greenwood Ong Gothenburg Shickley Maywood Spalding Diller Firth Brownville Funk Osmond Humphrey McCook Bruning North Loup Brookings Aberdeen Milbank Emery Dimock Madison Northville Elk Point Cambria Cottage Grove Durand 66 Bellingham Farmers ADM Ethanol Hutch Coop Northern Country Coop United Farmers Coop All American Coop Eastern Farmers Coop Meadowland Farmers Coop Coop Country Farmers New Vision Western Consolidated Coop Prairie Grain Frontier Coop Midwest Farmers Coop Aurora Coop Farmland Service Coop Shickley Grain Ag Valley Coop Country Partners Coop Firth Coop Firth Coop Bartlett Grain Cooperative Producers Battle Creek Farmers Coop Central Valley Ag Frenchman Valley Coop Bruning Grain Country Partners Coop AgFirst Farmers Coop South Dakota Wheat Growers Western Consolidated Coop Cargill Central Farmers Cooperative Madison Farmers North Central Farmers Southeast Farmers Coop Landmark Coop Landmark Coop Countryside Coop APPENDIX D Table D.1 – Summary Statistics by State Illinois Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio Indiana Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio Iowa Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio Kansas Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio 67 Mean St. Dev Min Max -21.70 11.50 -68 6 295.95 81.88 114 544 166.26 154.62 0 637 2.77 0.66 1.21 4.34 194.46 89.98 33.37 543.99 4.49 1.93 1.38 12.46 Mean St. Dev Min Max -16.73 11.94 -54 14 208.94 73.52 83 403 66.60 95.40 0 431 1.97 0.66 0.93 3.45 133.72 69.32 26.45 402.80 3.16 1.66 0.99 9.92 Mean St. Dev Min Max -32.20 12.01 -71 8 289.33 78.38 106 472 184.40 201.09 0 1026 2.72 0.68 1.11 4.32 203.50 81.51 40.93 471.72 4.99 2.53 1.28 13.70 Mean St. Dev Min Max -20.06 16.12 -78 16 39.69 25.97 7 111 16.33 28.18 0 177 0.37 0.24 0.07 1.17 27.46 20.29 2.51 110.72 0.68 0.58 0.08 4.15 Minnesota Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio Nebraska Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio South Dakota Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio Wisconsin Basis Local Corn Production (million bushels) Ethanol (MGY) Production Ratio Local Stocks (million bushels) Stocks Ratio 68 Mean St. Dev Min Max -38.96 12.42 -86 1 253.04 62.13 128 397 169.66 98.55 21 621 2.39 0.56 1.34 3.72 176.16 71.87 45.06 396.52 4.16 1.76 1.55 10.84 Mean St. Dev Min Max -28.42 11.06 -58 13 168.69 58.18 34 331 96.48 121.19 0 574 1.58 0.51 0.38 2.60 115.53 53.31 14.63 330.58 2.86 1.64 0.44 8.58 Mean St. Dev Min Max -39.32 12.13 -69 -5 131.11 50.80 51 275 146.22 115.65 0 436 1.23 0.44 0.49 2.27 90.09 45.54 17.74 274.53 2.18 1.23 0.57 7.74 Mean St. Dev Min Max -48.73 15.94 -92 -26 120.10 10.55 100 137 42.07 30.68 0 81 1.14 0.13 0.97 1.40 88.80 24.93 46.33 136.89 2.31 1.38 1.09 7.63 Table D.2 – Summary Statistics by Year Fall 1999 - Summer 2000 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2000 - Summer 2001 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2001 - Summer 2002 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio 69 Mean St. Dev Min Max -34.34 11.35 -64 -8 200.16 88.40 8 349 68.83 100.08 0 402 139.31 8.53 125.57 148.87 9.43 0.00 9.43 9.43 2.12 0.94 0.09 3.71 142.47 76.34 4.01 349.48 4.74 2.35 1.72 8.04 3.58 2.17 0.10 11.00 Mean St. Dev Min Max -29.86 11.53 -63 3 202.19 89.36 10 354 69.29 100.07 0 402 148.28 4.98 143.80 156.53 9.92 0.00 9.92 9.92 2.04 0.90 0.10 3.57 144.70 76.79 4.40 354.34 5.10 2.46 1.90 8.52 3.32 1.92 0.12 9.84 Mean St. Dev Min Max -22.41 10.88 -51 10 196.34 86.39 9 361 72.89 100.01 0 402 125.83 6.54 115.40 133.33 9.50 0.00 9.50 9.50 2.07 0.91 0.10 3.80 136.40 74.06 4.83 361.23 4.81 2.49 1.60 8.26 3.41 1.99 0.12 9.92 Fall 2002 - Summer 2003 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2003 - Summer 2004 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2004 - Summer 2005 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio 70 Mean St. Dev Min Max -13.05 11.26 -40 15 202.36 100.46 8 387 84.67 100.77 0 402 147.89 6.49 143.73 159.10 8.97 0.00 8.97 8.97 2.26 1.12 0.09 4.32 130.81 84.98 4.21 387.10 4.21 2.44 1.09 7.63 3.88 2.59 0.11 13.49 Mean St. Dev Min Max -17.09 9.52 -36 15 217.25 103.64 7 411 102.20 105.33 0 402 161.99 12.14 147.07 179.43 10.09 0.00 10.09 10.09 2.15 1.03 0.07 4.08 137.63 89.13 2.51 411.41 4.29 2.61 0.96 7.95 4.11 2.64 0.09 12.88 Mean St. Dev Min Max -29.99 12.50 -66 2 252.13 113.54 9 459 120.20 115.09 0 411 220.27 19.29 202.20 251.30 11.81 0.00 11.81 11.81 2.14 0.96 0.08 3.89 180.39 96.15 4.79 459.26 5.66 2.74 2.11 9.45 3.76 2.25 0.09 11.16 Fall 2005 - Summer 2006 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2006 - Summer 2007 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio Fall 2007 - Summer 2008 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio 71 Mean St. Dev Min Max -37.28 13.03 -73 0 238.14 109.01 10 436 145.72 137.24 0 652 270.84 16.63 245.23 289.97 11.11 0.00 11.11 11.11 2.14 0.98 0.09 3.93 167.06 93.93 5.26 436.19 5.78 2.93 1.97 9.81 3.44 2.09 0.10 10.51 Mean St. Dev Min Max -29.11 13.35 -70 16 231.79 111.89 7 436 176.93 157.77 0 706 267.88 16.71 250.73 289.70 10.53 0.00 10.53 10.53 2.20 1.06 0.07 4.14 153.25 95.59 3.79 435.86 4.96 2.85 1.30 8.93 3.88 2.60 0.08 13.70 Mean St. Dev Min Max -39.87 13.09 -92 -6 274.42 129.68 11 544 232.06 179.01 0 816 383.93 47.99 324.60 434.20 13.04 0.00 13.04 13.04 2.10 0.99 0.09 4.17 183.88 109.41 4.56 543.99 5.70 3.23 1.62 10.28 4.08 2.78 0.11 13.70 Fall 2008 - Summer 2009 Basis (cents) Local Corn Production (million bushels) Ethanol (MGY) Midwest Diesel Price National Corn Production (billion bushels) Production Ratio Local Stocks (million bushels) National Stocks (billion bushels) Stocks Ratio 72 Mean St. Dev Min Max -28.40 13.70 -69 11 257.97 118.55 12 524 290.36 223.50 0 1026 248.42 30.22 214.97 293.40 12.09 0.00 12.09 12.09 2.13 0.98 0.10 4.34 173.69 102.20 5.62 524.21 5.75 3.13 1.67 10.08 3.67 2.20 0.12 11.48 SOURCES Anselin, L. (1988). Spatial econometrics : Methods and models. Dordrecht Boston: Kluwer Academic Publishers. Anselin, L., Le Gallo, J., & Jayet, H. (2008). Spatial Panel Econometrics. In L. Matyas & P. Sevestre (Eds.), The Econometrics of Panel Data (625-660). Berlin: Springer-Verlag Baltagi, B., Song, S., & Koh, W. (2003) Testing panel data regression models with spatial error correlation. Journal of Econometrics, 117, 123-150. Burridge, P. (1980) Cliff-Ord test for spatial correlation. Journal of Royal Statistical Society Series – Methodological 42: 107-108. Bush, G. (2007, January 23) 2007 State of the Union Address. United States Congress. Washington D.C. Cash Grain Bids Inc. (2010) Historical Grain Prices. http://www.cashgrainbids.com/. Accessed March 2010. Center for Agricultural and Rural Development, Iowa State University. (2010). [Map illustration Daily Corn and Soybean Basis Maps for Iowa and the Midwest June 2010]. Retrieved from www.card.iastate.edu/ag_risk_tools/basis_maps/index.aspx Cliff, A., Ord, J. (1981) Spatial processes: Models and applications. Pion, London Davis, L., & Hill, L. (1974). Spatial Price Differentials for Corn among Illinois Country Elevators. American Journal of Agricultural Economics, 56(1), 135. Retrieved from Business Source Premier database. De La Torre Ugarte, D., English, B., & Jensen, K. (2007). Sixty Billion Gallons by 2030: Economic and Agricultural Impacts of Ethanol and Biodiesel Expansion. American Journal of Agricultural Economics, 89(5), 1290-1295. doi:10.1111/j.1467-8276.2007.01099.x. Eidman, V. (2007). Economic Parameters for Corn Ethanol and Biodiesel Production. Journal of Agricultural and Applied Economics, 39(2), 345-356. Elhorst, P. (2003). Specification and Estimation of Spatial Panel Data Models. International Regional Science Review, 26(3), 244-268. Energy Independence and Security Act of 2007, H.R. 6, 110th Cong. (2007) Energy Policy Act of 2005, Pub. L. No. 109-58, 119 Stat 594 (2005) Energy timelines: Ethanol. (2008). U.S. Energy Information Administration. Retrieved March 2, 2010, from http://tonto.eia.doe.gov/kids/energy.cfm?page=tl_ethanol 73 Ethanol Producer Magazine. (2010) Plant List. Retrieved from http://www.ethanolproducer.com/plant-list.jsp Federal Reserve Statistical Release (2010) Retrieved from http://www.federalreserve.gov/releases/h15/data/Monthly/H15_PRIME_NA.txt Fortenbery, T.R. (2002). Understanding and Forecasting Soybean Basis: An Empirical Investigation. Working paper. Fortenbery, T.R. & Park, H. (2008). The Effect of Ethanol Production on the U.S. National Corn Price. (Staff Paper No. 523). Retrieved from University of Wisconsin, Department of Agricultural & Applied Economics Website: www.aae.wisc.edu/pubs/sps/ Frees, E. (2004). Longitudinal and Panel Data. New York, NY: Cambridge University Press. Gallagher, P., Wisner, R., & Brubacker, H. (2005) Price relationships in processors’ input market areas: Testing theories for corn prices near ethanol plants. Canadian Journal of Agricultural Economics, 53(2-3), 117-139. Garcia, P. & Good, D. (1983). Proceedings from the NCR-134 Conference of Applied Commodity Price Analysis, Forecasting, and Market Risk Management: An Anlysis of the Factors Influencing the Illinois Corn Basis 1971-1981. Des Moines, IA. Gujarati, D. (2003). Basic Econometrics (4th ed.). New York, NY: McGraw-Hill, Inc. Henderson, J. & Gloy, B. (2009). The impact of ethanol plants of cropland values in the great plains. Agricultural Finance Review, 69(1), 36-48. International Institute for Sustainable Development. (2006). Government support for ethanol and biodiesel in the United States. Genvea, Switzerland: Koplow, Doug. Kahl. K., & Curtis, C. (1986). A comparative analysis of the corn basis in feed grain deficit and surplus areas. Review of Research in Futures Markets 5(3), 220-232. Katchova, A. (2009). Proceedings from the Agricultural & Applied Economics Association’s 2009 AAEA & ACCI Joint Annual Meeting: The Spatial Effect of Ethanol Biorefinery Locations on Local Corn Prices. Milwaukee, WI. Lesage. J., Pace, R. (2009). Introduction to Spatial Econometrics. CRC Press/Taylor & Francis Group. Luchansky, M., & Monks, J. (2009). Supply and demand elasticities in the U.S. ethanol fuel market. Energy Economics, 31(3), 403-410. doi:10.1016/j.eneco.2008.12.005. McNew, K., & Griffith, D. (2005). Measuring the impact of ethanol plants on local grain prices. Review of Agricultural Economics, 27(2), 164-180. 74 Nurenberg, T. (January 2009). Breaking Through the Blend Wall. Retrieved from http://www.ethanolproducer.com/article.jsp?article_id=5151 O’Brien, D. (2009). Proceedings from the NCR-134 Conference of Applied Commodity Price Analysis, Forecasting, and Market Risk Management: The Effects of the Micro-Market Structure for Kansas Elevators on Spatial Grain Price Differentials. St. Louis, MO. Olson, A., Klein, N. & Taylor, G. (2007). Proceedings from American Agricultural Economics Association Annual Meeting: The Impact of Increased Ethanol Production of Corn Basis in South Dakota. Portland, OR. Rapier, R. (2010). Washington's foolish fuel policy. Forbes. Retrieved from http://blogs.forbes.com/energysource/2010/02/16/washington-foolish-fuel-policy/ Renewable Fuels Association. (2010) 2010 Ethanol Industry Outlook. Renewable Fuels Association. (2010) Industry Statistics. www.ethanolrfa.org. Accessed January, 2010. Sneller, T., & Durante, D. (2006) Economic Impacts of Ethanol Production. Ethanol Across America. Solomon, B. D., Barnes, J. R., & Halvorsen, K. E. (2007). Grain and cellulosic ethanol: History, economics, and energy policy. Biomass and Bioenergy, 31(6), 416-425. Stakhovych,S. & Bijmolt, H. (2008). Specification of spatial models: A simulation study of weights matricies. Papers in Regional Science. 88(2), 389-408.\ Taylor, T. & Dorsey, J. (June 2010). The Rise and Fall of Verasun. Retrieved from http://www.ethanolproducer.com/article.jsp?article_id=6623 Tokgoz, S., Elobeid, A., Fabiosa, J., Hayes, D., Babcock, B., Yu, T., et al. (2008). Bottlenecks, Drought, and Oil Price Spikes: Impact on U.S. Ethanol and Agricultural Sectors. Review of Agricultural Economics, 30(4), 604-622. doi:10.1111/j.1467-9353.2008.00436.x. United States Department of Agriculture, Economic Research Service. (2010) Corn: Market Outlook. http://www.ers.usda.gov/Briefing/Corn/2009baseline.htm. Accessed May 2010. United States Department of Agriculture, National Agricultural Statistics Service. (2010) Quick Stats. http://www.nass.usda.gov/. Accessed January 27, 2010. United States Department of Agriculture, Office of the Chief Economist. (2008). USDA agricultural projections to 2017 75 U.S. Energy Information Administration (2010). Energy Timelines: Ethanol. http://www.eia.doe.gov/kids/energy.cfm?page=tl_ethanol Accessed January 28, 2010. U.S. Energy Information Administration (2010). Midwest No 2 Diesel Retail Sales by All Sellers. http://tonto.eia.doe.gov/oog/info/gdu/gasdiesel.asp. Accessed January 28, 2010. 76