Relationship between corn prices and ethanol production the case of the United States.

(1)

Relationship Between Corn Prices and Ethanol Production

The Case of the United States

Julija Simionenko 10391762

MSc in Economics Thesis

Supervisor: Prof. Dr. Menno P. Pradhan Secondary reviewer: Dhr. Dr. D.J.M. Veestraeten

May 2014

Abstract: The Food vs. Fuel dilemma has been one of the most debated subjects over the last decade. A lot of research has been conducted but there is still no unanimous consensus about biofuels production effect on food prices. This thesis, therefore, analyzes the dilemma by examining relationship between United States corn and ethanol markets. The study uses different econometric techniques to estimate the effect that ethanol production has on corn prices, and checks their validly given the available data. Study shows improvement over previous econometric approaches as it combines demand and supply specific effects together with macroeconomic factors. Research results suggest that there is a significant but rather small effect, which does not explain most of the variation in corn prices.

(2)

List of Figures

1. World trade (exports) in total merchandise, billion US dollars.

2. World soybeans, maize, rice and wheat prices (nominal dollars per metric ton) 3. Crude oil, Natural gas, and Fertilizers prices

4. World net per capita Production Index Number (2004-‐2006 = 100)

5. World Ethanol and Biodiesel production (million gallons), together with Crude oil prices (U.S. dollars/bbl.).

6. World meat and dairy price indices

7. U.S. ethanol and gasoline prices, U.S. dollars per gallon.

8. U.S. Ethanol production, capacity (million gallons), and the number of plants. 9. U.S. Corn production and use for feed grain, fuel ethanol, and exports.

10. U.S. Corn and soybeans harvested area (million hectares). 11. U.S. Corn ending stocks, million bushels.

(4)

Introduction

The recent world food crisis raised a lot of debates in the past years. Major food commodities prices increased dramatically during the years 2007-2008. While some prices doubled on a yearly basis, others more than doubled in the period of a couple of months! This rapid price increase of the major food groups puts a lot of pressure on the poorest members of societies, who spend the majority of their income on food. This resulted in economic and social instability not only in poor nations but also in developed countries. According to D. Headey and Sh. Fan (2010) research, even though food prices are now lower than their 2008 peak, real prices have remained significantly higher in 2009 and 2010 than they were prior to the crisis, and various simulation models predict that real food prices will remain high until at least the end of the next decade.

The discussions usually concern the reasons that could have caused such price hikes. The most popular causes recognized by studies are unfavorable weather conditions and increase in oil prices. Some countries experienced lower yields than their trend growth rates would have predicted. For example, Australia faced around 50% lower yields in the years 2005 – 2006 due to the droughts. Poorer harvests in Ukraine and India also put pressure on the markets. In addition, increases in crude oil prices raised the costs of grain production. This affected production via higher fertilizers and transportation costs, since both require fuel as their input.

Other assessed reasons include growing population and households’ income in developing countries, like China and India. It is believed that the increase in income leads to higher demand for resource intensive foods, like meat and dairy. This not only raises the price of these goods directly but also the price of grain, since it is used as the feed for livestock. Even thought the world population is growing rapidly, the pace of food growth exceeds it, and together with globalizing trade countries are choosing to keep lower food stockpiles. This in turn contributes to the higher food prices. Further findings include depreciation of the United States dollar, financial speculations, trade distortions, and finally, one of the most controversial theories – creation of biofuels.

(5)

The latter cause received a lot of attention from media as well as from global organizations around the world. A lot of research has been conducted but there is still no unanimous consensus about its effect on food prices. This raises the well-known “food vs. fuel” dilemma.

The basic idea behind it is that biofuels compete with food production and this in response raises food prices. First-generation biofuels are made from edible crops; therefore they reduce the supply of crops left for food. The competition also occurs for natural resources like arable land and water. The land that could originally be used for growing various food and feed crops is now being transformed into growing inputs for biofuels. One of the biggest examples of this effect can be observed in the United States maize market. The ethanol made of corn production has begun to rise sharply starting the year 2000 and by the year 2012 it was using around 40% of that year’s U.S. corn yield (Carter, Miller, 2012). The world corn prices seem to experience the same growth trend over this period. Since United States is the biggest corn producer, exporting 20% of its annual yield, changes in U.S. corn market have a significant effect on world’s corn and other staple food prices (USDA, 2013).

Despite the above-mentioned facts and examples, some economists still doubt the significance of the effect that biofuels have on food prices and tend to lay more weight on other causes. These findings suggest that biofuels creation accounted only for vague increase in global commodities prices and should not be considered as a serious threat. Other literature, on contrary, supports the evidence and estimates that biofuels have been the major cause of the 2007-2008 world food crisis. Some literature examples include works of A. Babcock and F. Fabiosa (2011), who found that expansion of corn ethanol from subsidies and market-based expansion of the corn ethanol industry accounted for 36% of the average increase that was seen in corn prices from 2006 to 2009; S. Baier, M. Clements, Ch. Griffiths, J. Ihrig (2009) estimated that over the years 2007-2008 the increase in worldwide biofuels production pushed up corn, soybean and sugar prices by 27%, 21% and 12% respectively; Lipsky (2008) estimated that the increased demand for biofuels accounted for 70% of the increase in maize prices and 40% of the increase in soybean prices and etc. There is an equal amount of research conducted to support both views. Moreover, it is hard to compare results due to the differences in approaches, periods considered, methods and techniques of estimation.

(6)

This thesis will try to analyze the relationship between biofuels production and food prices. Due to the limitations of the paper and availability of some data the analysis will concentrate on the United States ethanol and corn markets. Research question of the paper can be formulated as: “Is there a causal relationship between ethanol production and corn prices in the United States?”. Analysis of this particular market is essential for determining biofuels production effects around the world. According to Allianz, maize (corn) is ranked as the primary staple crop across the globe. Since the U.S. is the biggest global corn producer, changes in U.S. corn market have significant effects to the world’s corn supply and prices. Corn is also known as the major feed grain for livestock. Therefore changes in its prices indirectly affect the price level of meat and dairy products.

This analysis is important, because most other studies have not been able to provide accurate estimation of the effect that biofuels have on food prices, holding everything else constant. The study will be conducted in three complementary stages, using different econometric techniques, and different assumptions in each stage. The analysis is going to cover quarterly data from 1980s, when ethanol production in United States started, until the year 2012. Dependent variable will be the log of U.S. quarterly corn price (measured in $/bushel) and independent variable will be the log of U.S. quarterly ethanol production (measured in millions of gallons). Other exogenous variables will be added to solve omitted variable bias problem, simultaneous causality is addressed by 3SLS estimation of the system of simultaneous equations. To account for non-stationarity of some variables, the relationship is going to be additionally tested via VECM.

The structure of the paper will consist of three main parts: literature analysis, empirical analysis and conclusions. Literature analysis will explain the “food vs. fuel” dilemma in more detail, review other studies addressing the issue, and analyze main statistics. Furthermore, situation in the United States markets will be examined as well. Empirical analysis part will focus on the review of the data, main variables, and models. This will include clarification of the assumptions behind each model, interpretation of the research results. At the end of the paper conclusions and suggestions will be drawn.

(7)

1. Literature analysis

1.1 World Food Crisis Overview

The global food price crisis of the years 2007-2008 can easily be called one of the major crises that hit the entire world over the last few decades. It created not only economical, but also social and political instability in developing as well as developed countries. A total of 842 million people in 2011-2013, or around one in eight people in the world, were estimated to be suffering from chronic hunger, regularly not getting enough food to conduct an active life (FAO, 2013). Sudden increase in major staple prices puts the most pressure on the poorest members of societies as they spend majority of their income on food products.

Poverty and hunger eradication is the primary focus of the eight Millennium Development Goals set by United Nations. This objective is especially relevant among developing countries, since world’s largest poverty levels are concentrated in these areas. One of the most recognized ways to tackle the issue is to increase income received by farmers. The World Bank annual report (2007) analysis found that more than three quarters, that is 78%, of those living in extreme poverty lived in rural areas, with nearly two thirds of the extreme poor deriving their livelihoods from agriculture. Governments should therefore concentrate on their countries’ agricultural sector development, making it less dependable on international food markets.

However, this matter was neglected as most of the world was enjoying trade liberalization and low international food commodity prices. For more than 30 years international prices of staple foods were at their lowest levels and most developing countries stopped encouraging investment and growth of their own agricultural sectors. On contrary, at the expense of the agriculture governments were raising revenues by investing in manufacturing and other areas.

By simply relying on imports, some countries like India and China even started reducing their food stocks. Unfortunately, the recent crisis, followed by increase in international and therefore import prices, forced countries to reconsider their policies and raised serious concerns about their food security.

(8)

Figure 1. World trade (exports) in total merchandise, billion US dollars.

Source: World Trade Organization

A crisis this big was last seen back in the years 1972-1974, during the “Soviet wheat deal”. Afterwards prices were declining for the next two decades and have reached their lowest rate during the period 2000-2002. Unfortunately, this low-price period was about to be over. The prices started growing rapidly after the year 2003, and in 2007-2008 reached record high levels, followed by another upward surge in 2011. The nature of this crisis, however, is not how expensive prices are relative to their historical trend, but how quickly they have risen, together with the related problem of behavioral adjustments by consumers and producers (D. Headey and Sh. Fan, 2010).

These price spikes are best observed among the major crops. Figure 2 shows dynamics of the world prices of the following primary grains: maize, wheat, rice and soybeans. As it is seen from a graph, the biggest and the most rapid surge appeared in rice prices. The rice prices started rising in late 2007. Between January and April of 2008, in a period of three months, prices grew by almost 160%. The total increase compared to pre-crisis price level was more than 230%. However, the prices started declining afterwards even though they still have not reached the previous levels. Maize prices started growing at the end of the year 2006. During the two-year period they increased by more than 150%. Even though in June 2008 there was a slight decline, prices did not return to the pre-crisis levels. After exactly two years, in June 2010, another surge began. This time, in less than a year prices doubled, reaching their all time high rate of more than 300% of the pre-crisis price level that was recorded in the year 2006.

1,000 4,000 7,000 10,000 13,000 16,000 19,000 1980 1984 1988 1992 1996 2000 2004 2008 2012

(9)

Soybeans prices began rising in the last quarter of 2006 and experienced changes similar to those in the maize market. In less than two years prices changed by more than 170%. After the crisis prices did not stabilize and in August 2012 reached the highest level, which accounted for around 310% of the previous price.

Prices of wheat began rising later than rice and soybean prices, around mid 2007, and have reached their peak in March 2008. In 10 months period there was recorded 120% upsurge in wheat price level. Afterwards prices did not remain constant and continued to fluctuate. During the years 2010-2011 they rose again, fortunately, did not return to the levels observed during crisis years.

Figure 2.World soybeans, maize, rice and wheat prices (dollars per metric ton)

Source: The World Bank

Most of the major field grains are used not only for food, but also as a feed for livestock. Therefore, following the increase in coarse grain prices, meat and dairy prices rose as well. Their changes, however, were not as extreme as changes in grain prices.

The causes of the crisis were widely discussed in the local governments, international institutions and media. The main drivers were indicated, but the extent to which they affected price level is still debatable. Most studies recognize increasing crude oil prices, droughts and depreciation of the United States dollar as major causes. Understanding these possible triggers is essential for the further empirical analysis, which is going to be conducted in the second part of the thesis.

55 155 255 355 455 555 655 1980 1984 1988 1992 1996 2000 2004 2008 2012 Soybeans Maize Rice Wheat

(10)

Petroleum price changes are thought to have direct and indirect effects on food prices. First of all, petroleum is one of the major inputs in food production. Second, oil, together with natural gas, is used in manufacturing of the fertilizers. With the increase in oil prices, fertilizers became more expensive as well. Higher oil prices stimulated the demand for natural gas, which consequently led to higher natural gas prices. Therefore, production of food had to deal with two of its main inputs becoming more expensive.

Figure 3. Crude oil, Natural gas, and Fertilizers prices

Source: The World Bank

Extreme weather changes in some major grain producing countries have contributed to the recent crisis. Droughts and lack of precipitation reduced harvests of major crops producing countries and thus world crop supply. For instance, Australia faced around 50% lower wheat yields in the years 2005 -2006 due to the droughts. D. Headey and Sh. Fan (2010) state, that as a counter-seasonal southern hemisphere exporter, it is quite possible that the Australian drought had a particularly sharp effect on prices, especially given that other countries also experienced lower yields than their trend growth rates would have predicted. Prices were pushed up even more when some major international crop producers decided to isolate and protect their domestic markets from high international prices and insufficient supply by imposing

trade restrictions. Countries like Ukraine, Argentina and India introduced quotas or

completely restricted some of their grain exports to other countries in order to make sure that their own domestic demand was satisfied. Moreover, in response to decreasing grain yields, precautionary imports from major global consumers also rose. 0 20 40 60 80 100 120 140 160 180 200 1980 1984 1988 1992 1996 2000 2004 2008 2012 Crude oil Natural Gas Fertilizer

(11)

U.S. dollar depreciation, according to most studies, played an important role in

price increase. First, the dynamics of U.S. dollar and food commodities prices are similar. When the dollar was weaker the price level was higher and when dollar appreciated the prices seemed to decline. Second, as U.S. dollar depreciated, international goods became even more expensive to United States consumers. At the same time, prices of United States production have become cheaper for the rest of the world. Hence, as D. Headey and Sh. Fan (2010) state, conversion to, for example, euros would cut off 20-30% of the nominal increase in U.S. dollar– denominated food prices. Most obvious causes of U.S. dollar decline are low real interest rates and increasing United States trade deficit.

Low interest rates are also distinguished as a possible driver of commodities prices, including food. The reasoning behind is that when interest rates are low, money flows out of interest-bearing instruments and into foreign currencies, emerging market stocks, other securities, and commodities, including food commodities (D. Headey and Sh. Fan, 2010). This shift drives these assets’ prices upwards. However this reason is still questionable and is not considered as the major cause of the crisis. Another crisis trigger indicated by some studies is speculation in the financial markets. It is a common knowledge that food markets mainly operate by using forward contracts, which secure producers from possible price changes in the future. Over time, this forward contract market evolved into futures contract market. It allows trading forward agreements as separate financial products. The main concern about it is that participants who are not involved in agriculture-related work are permitted to take part in trade. This in turn allows them to speculate on food commodities price trends. However, the studies conducted on this topic are not very accurate and often doubtful. There is also a bulk of evidence including findings of D. Mitchel (2008), which suggest that these speculative activities and export bans would probably not have occurred as they usually appear as a symptom of an existing turmoil and were largely responses to rising prices.

Some critics claim that the source of the crisis lies in world population growth rates. It is not a secret that increasing population requires additional supply of food products. However, such reasoning is not supported by most data, since world

(12)

population growth rate has been decreasing since 1980s and food production per capita index seemed to be increasing from 1960s.

Figure 4: World net per capita Production Index Number (2004-2006 = 100)

Source: Food and Agriculture Organization of the United Nations

J. von Braun (2008) and other economists share the view that the dominant force in recent food price crisis was rapid economic growth in many developing countries that has pushed up middle-class consumers’ purchasing power, generated rising demand for food, and shifted food demand away from traditional staples towards higher-value foods like meat and milk. Even though consumers switched to more resource intensive food, demand for grains still followed the increase, since it is used as a feed for livestock.

Debates on biofuels production have escalated among policy makers and media outlets immediately after the crisis. Biofuels are recognized by a lot of studies to have one of the biggest effects on food prices. Increasing creation of this renewable fuel is blamed for diverting grain used for food towards fuel. This discussion is better known as “food vs. fuel” dilemma, which is going to be a further interest of this paper.

In many ways, including triggers and drivers, crises of the years 2007-2008 and 1972-1974 were similar. Both included higher oil prices, depreciation of the U.S. dollar, lower interest rates, unfavorable weather conditions, and decreasing food stocks. What is more, both crises experienced demand shocks in U.S. grain markets: from biofuels creation in 2007-2008 and from Soviet Union in 1972-1974. Even though crises shared some resemblances, the world could not prevent the former crisis from happening. This raises concern that economy stabilization will once again lead to the apathy towards agricultural investments and development in advanced and developing countries.

70 80 90 100 110 120 1975 1980 1985 1990 1995 2000 2005 2010

(13)

1.2 Food vs. Fuel Dilemma

In the recent years, bioenergy did not only draw attention as a sustainable energy source, but also as the potentially significant contributor to the recent world food price crisis. Biofuels are blamed for diverting farmland and crops that were originally used for food and feed, to fuel production. Biofuels and food price debate is a long-standing, controversial one in the literature, with wide ranging views. This is due to the number of impacts and feedback loops involved that positively or negatively affected the price system (HLPE, 2013). Although many vide-range organizations and observers, including Word Bank and other civil society organizations, identify biofuels as an important factor of the crisis, the real debate exists on the extent to which biofuels affect agricultural prices and volatility.

Biofuel is a liquid, renewable fuel that can be made from living organisms or metabolic byproducts – organic or food waste products. There can be distinguished two types of biofuels: first-generation and second-generation. The literature regarding former world food price crisis has been mainly focusing on first generation biofuel and its effects, as it uses grains and oilseeds for its creation. Thus, it competes for land and resources with other agricultural activities, including production of other forms of bioenergy, other economic activities, urbanization and, increasingly, with land protection for environmental objectives, especially biodiversity and carbon sequestration (HLPE, 2013). Second generation biofuels creation, on the contrary, is widely supported, since it uses non-food crops, crop residues and waste, therefore, not putting much pressure on food commodities prices.

Biofuels were first started to produce as a response to increasing oil prices, desire to establish oil independency and decrease green-house-gas emissions. As oil and correspondingly gasoline prices started rising, world began to look for cheaper substitutes that could compete with oil production. This in turn increased demand for biofuels. Just in one decade, during the years 2001-2010, world biofuel production has increased about five and a half times. The sharpest rise occurred in 2007-2008, when ethanol production increased by more than 30% and biodiesel production rose by almost 50%.

(14)

Figure 5. World Ethanol and Biodiesel production (million gallons), together with

Crude oil prices (U.S. dollars per bbl.)

Source: The World Bank and Earth Policy Institute

In addition, increasing international oil prices raised concerns about countries’ national energy security. Having access to cheap energy is essential for well functioning modern economies. One way to increase energy security is to reduce dependence on a single energy source, which, for example, could be reached by developing native renewable fuel resources. According to G. Quaiattini (2008), a former president of Canadian Renewable Fuels Association, a healthy supply of alternative energy sources gives the power to combat gasoline price spikes.

To support the domestic biofuels market several countries even imposed different subsidies, tariffs, tax reliefs and mandates, which stimulated biofuels demand and supply. Now there are more than 60 countries that developed biofuels policies (HLPE, 2013). These policies encouraged biofuels production even with record high and steadily increasing food prices. This implies that countries, which were importing food, especially those in Sub Saharan Africa, as they import greater share of their food, were at mercy of domestic policies that governed biofuels production in food exporting countries.

Imposed biofuels policies caused no less debate than the biofuels production itself. Some critics believed that support for biofuels should be eliminated or at least reconsidered in light of certain social and economic concerns associated with rising

0.0 20.0 40.0 60.0 80.0 100.0 120.0 0 5,000 10,000 15,000 20,000 25,000 30,000 1995 1997 1999 2001 2003 2005 2007 2009 2011 Biodiesel Ethanol Crude oil

(15)

food prices and starving population in developing countries. In October 2007, professor of social science and economics Jean Ziegel, who has been the UN's independent expert on the right to food, even called for a five-year moratorium,

aiming to ban the conversion of land for the production of biofuels, calling it "crime against humanity" and the "catastrophe" for poor people. The same year, G. Monbiot, a journalist for The Guardian, also suggested a five-year freeze on biofuels targets and incentives at least until the second-generation fuel market evolves. But even then targets should be set low and be increased only continuously.

Besides possible negative aspects, bioenergy is also considered to be a potentially significant contributor towards economic development of rural areas. It can serve as a mean of poverty reduction through wider economic growth, higher employment, increase in income received by farmers and stabilization of oil prices. With regards to the potential for poverty reduction or exacerbation, biofuels rely on many of the same policy, regulatory or investment shortcomings that impede agriculture as a route to poverty reduction (ODI, 2009). Since most of these limitations need policy adjustments and improvements at a country level rather than global, each country should be analyzed individually in order to determine biofuels production impact on poverty level.

There are direct and indirect ways, in which biofuels can affect food commodities prices. Rapid increase in demand for and production of biofuels, particularly bioethanol from maize and sugarcane, has had a number of effects on grain supply and demand systems (Rosegrant, 2008). The first direct effect is that demand for biofuels reduces crops availability for food and feed purposes. Together with higher biofuels demand, increases the demand for grains and oilseeds, as they are the primary production inputs. This in turn raises the aggregate demand for crops regardless of their usage.

Second, limited resources like arable land, restrain the supply and therefore availability of crops. Consequently, greater demand and relatively slower increase in supply result in higher overall crops price level. Higher price level implies that fewer people can afford these staples, and therefore, all excess supply is absorbed by demand for crops for biofuels production. The result is that people end up consuming less for a higher cost, which as discussed earlier, mostly affects the poorest

(16)

members of the society, and thus raises concerns over poverty level.

Previously mentioned supply and demand-side effects have led towards numerous consequences that were indirectly triggered by the growing production of first generation biofuels. One of these indirect effects is substitution, which explains why price growth spread to other crops that were not even used in biofuels production. On the consumption level, substitution appears when consumers switch from commodities with higher prices towards cheaper substitutes. For example, higher maize prices have caused food consumers to shift from maize, which is still a significant staple food crop in much of the developing world, to rice and wheat (Rosegrant, 2008). On the production level, rising prices created an opportunity for farmers to obtain higher profits. For instance, higher maize prices made maize more profitable to grow, causing some farmers to shift from rice and wheat (and other crop) cultivation to maize cultivation (Rosegrant, 2008). While maize displaced soybeans in the U.S. other oilseeds displaced wheat in EU and other wheat exporting countries (Mitchell, 2008).

Changes in acreage of the main cereals and oilseeds resulted in higher prices of all

available crops. However, multiple studies, including those made by FAO and IIASA (using GAEZ classification), estimated that there is still a gross balance of about 3.2 billion hectares of good land not used for growing crops (N. Alexandratos and J. Bruinsma, 2012). Other authors consider these results to be rather optimistic. For example, Young (1999) argued that the results estimated large areas of potential cropland even in regions where low land availability had already resulted in the conversion of areas highly unsuitable, such as areas of steep slopes. The analysis also failed to account properly for competing land uses and that the use of global datasets leads to broad areas being considered as suitable for cropping even though only portions of each “cell” would be suitable (HLPE, 2013).

Another knock-on effect of significant concern is that biofuels have added substantially to the depletion of grain stocks (Mitchell 2008). According to FIPRI researchers, because stocks are a residual, stock declines in some countries primarily reflect causes, such as rising demand or insufficient supply (D. Headay, Sh. Fan 2010). Increasing biofuels production obviously contributed to the higher demand for grains, therefore causing it to grow faster than supply. Grain stocks

(17)

started declining in the years 2005-2006 as demand rose faster than production of grains. Land shifts from soybeans towards maize in U.S. and from wheat to oilseeds in EU also slowed the production of these crops, which otherwise could have kept their stock levels higher. However, it is not certain that in the absence of biofuels farmers would have chosen to produce the same levels of crops.

Grains not only are inputs for biofuels production, but also serve as feed to livestock. As grain prices increase, so do the prices of products produced from the livestock. Therefore changes in crops prices have a direct effect on costs of meat and dairy products. Figure below shows changes in meat and dairy products prices, which happened to coincide with the dynamics of grain prices and rising production of biofuels.

Figure 6. World meat and dairy price indices

Source: Food and Agriculture Organization 0.0 50.0 100.0 150.0 200.0 250.0 300.0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Meat Price Index

(18)

1.3 U.S. Ethanol and Corn Markets Overview

To understand the global effect of biofuels on food prices it is essential to analyze the main markets where production of different types of biofuel takes place. This and remaining parts of the paper, therefore, are going to further concentrate on the United States ethanol production and its impact on corn prices, since in the U.S., two commodities, corn and soybean oil, account for over 90 percent of biofuels production (USDA testimony), with corn being the main feed stock for U.S. ethanol production (S. Baier et al., 2009). This particular U.S. markets analysis is important for several reasons:

• Dominance of the U.S. grain markets. The United States accounts for about one third of global maize production (D. Mitchell, 2008), and heavily dominates global exports of maize (60 percent) and wheat (25 percent), and although those of Argentina and Brazil have overtaken U.S. soybean exports in recent decades, the United States is still the world’s third largest soybean exporter (D. Headey, Sh. Fan, 2010). Therefore, U.S. prices are often quoted as international prices for most grain, except rice, where Thai prices are being used. For this reason, international grain prices are very sensitive to events happening in U.S. grain markets.

• The U.S. being the world’s largest producer of ethanol from maize. Together with Brazil and European Union, it is a leader in the development and use of biofuels. However, European biofuels production is concentrated on biodiesels and it uses oilseeds for its production. Biofuel production in other parts of the world is either relatively small or uses different crops (for example, sugarcane for ethanol in Brazil), which have not experienced price surges (D. Headey, Sh. Fan, 2010). Hence, changes in U.S. ethanol from maize production have a significant effect on international corn prices.

• Importance of U.S.-specific factors. Events happening in U.S. economy or grain markets are thought to be possible suspects of the recent world food price crisis. Major instances include the sharp U.S. dollar depreciation, accumulation of dollar reserves by foreign countries, dynamics in commodity futures markets and even unfavorable weather conditions that contribute to poor grain yields. Changes in trade and demand shocks are also of the

(19)

primary importance, especially, since the 1972 Soviet wheat deal. Most market observers now consider exports to be the great uncertainty underlying commodity supply, demand, and price forecasts (Schnepf, 2005).

For the previously stated reasons, most of the economists and researchers (including works of Abbott, Hurt, Tyner, 2008; Mitchell 2008; Schnepf 2008; von Braun 2008a) conclude that the diversion of the U.S. corn from food and feed to ethanol creation constitutes the largest source of international biofuel demand and the largest source of demand-induced price pressure (D. Headey, Sh. Fan, 2010).

The U.S. ethanol production has been rising slowly since the 1980s, when it first started. But later, the rapid growth occurred from the year 2004 to 2010, which led to the dramatic increase in U.S. corn based ethanol production. During this period the production rose almost 3 times (by 289%, that is from 3400 millions of gallons in 2004 to 13230 millions of gallons in 2010).

The main reason behind it was the high profit margins from producing ethanol. Once oil prices exceeded 60 US dollars, biofuels have become more competitive and farmers started diverting grain to biofuel production, especially since high oil prices were expected to persist (Schmidhuber, 2006). Greater demand for ethanol has led to higher ethanol prices, which meant greater profit opportunities for not only ethanol producers but also maize growers, as corn is the main feedstock in ethanol production. The following figure shows U.S. ethanol and gasoline prices for the period 2000-2012. High profit margins coincide with the years of ethanol production surge implying that high profits could have been responsible for the increase in ethanol production.

(20)

Figure 7. U.S. ethanol and gasoline prices, U.S. dollars per gallon.

Source: USDA

The increasing number of ethanol plants and ethanol production capacity in the U.S support the fact that higher profits encouraged ethanol producers to expand their business. The chart bellow shows the number of ethanol plants in U.S. and their capacity from 1999 through 2012. Since then the number of ethanol plants has grown almost 3,8 times, reaching 211 in the year 2012. The capacity of these plants has been increasing at a faster rate than has the number of plants, meaning the average plant size was increasing. Production was remarkably close to plant capacity, especially considering that both were tracked at the end of each year; so plants built late in the year contribute largely towards capacity but only partially toward production (RFA, 2013).

Figure 8. U.S. Ethanol production, capacity (MM gallons), and the number of plants.

Source: Renewable Fuels Association 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 Ja n -0 0 Ju l-0 0 Ja n -0 1 Ju l-0 1 Ja n -0 2 Ju l-0 2 Ja n -0 3 Ju l-0 3 Ja n -0 4 Ju l-0 4 Ja n -0 5 Ju l-0 5 Ja n -0 6 Ju l-0 6 Ja n -0 7 Ju l-0 7 Ja n -0 8 Ju l-0 8 Ja n -0 9 Ju l-0 9 Ja n -1 0 Ju l-1 0 Ja n -1 1 Ju l-1 1 Ja n -1 2 Ju l-1 2 Ja n -1 3 Ethanol Gasoline 0 50 100 150 200 250 - 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 1999 2001 2003 2005 2007 2009 2011 Ethanol Plants Capacity Production

(21)

Since corn is a major feedstock in U.S. ethanol creation, increase in ethanol supply is directly responsible for the larger corn demand. Each year the greater share of produced corn has been diverted to the production of fuel ethanol. In early the 2000s this part accounted for only 6% of total corn production, while in 2009 and 2010 it reached 35% and 40% respectively. The data also suggests that over the last decade all the increase in corn production was mostly absorbed by exports and fuel ethanol, while the share of corn for feed has been declining. As a matter of fact, in 2004 feed accounted for 52% of total corn production, while in 2011 it dropped only to 37%, implying that greater part of corn is now used for fuel.

Figure 9. U.S. Corn production and use for feed grain, fuel ethanol, and exports.

Source: Earth Policy Institute and USDA

The higher demand for corn consequently caused an increase in corn prices. Higher prices, in turn, implied higher profits and therefore, encouraged farmers to switch from other grains to growing maize. In the U.S. most of the farmland for maize however, was diverted from soybean production. Changes in acreage imply that there is a constant competition for arable land between the two staples in U.S. This can be best observed over the last decade when biofuels boom occurred. During this period sharp rises in corn area ware followed by the declines in land used for soybean production. D. Mitchel (2008) estimated, that a rapid expansion of maize area by 23% in 2007 resulted in a 16% decline in soybean area, which reduced

0 50 100 150 200 250 300 350 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

(22)

soybean production and contributed to the 75% rise in soybean prices from April 2007 to April 2008. Even though corn area declined the following year it continued to grow the year after and is still expanding on the expense of soybeans. As discussed earlier, these price increases also spread to the other markets, causing price hikes among other food commodities.

Figure 10. U.S. Corn and soybeans harvested area (million hectares).

Source: Earth Policy Institute and USDA.

Since demand of most staple grains is typically considered to be inelastic, an increase in their price level rarely affects the quantities demanded. Thus, without any big changes in demand, higher prices tend to persist. Similarly, grain supply in most literature is also often treaded as inelastic, even if there are possibilities for land diversion. Moreover, because most grains are limited to a single annual harvest, new supply flows to market in response to a postharvest price change must come from either domestic stocks or international sources (D. Headey, Sh. Fan, 2010). Hence, supply elasticities tend to be highly inelastic in grain markets, making them very vulnerable to relatively small shocks, especially when stocks are low (D. Headey, Sh. Fan, 2010). In the U.S. rising ethanol production has put pressure on corn market that resulted in lower corn stocks and higher prices. Over the last decade the biggest stock decline appeared in 2006, when in one-year period stock levels fell by almost 34%. Another major decline appeared in 2011-2012, when stocks dropped by 28% compared to the levels in 2010. It is worth noting, that some of the declines are attributed to the droughts in the U.S.

22 24 26 28 30 32 34 36 38 1995 1997 1999 2001 2003 2005 2007 2009 2011 Corn Soybeans

(23)

Figure 11. U.S. Corn ending stocks, million bushels.

Source: USDA, Economic Research Service

Besides earlier mentioned demand shocks and market effects, U.S. corn yields and therefore price levels were also affected by a number of droughts that hindered the country throughout the years. Some of biggest droughts that especially affected U.S. corn production over the last decade were recorded in the years 2002, 2005, and 2012. The graph below shows U.S. corn yield for 2000-2012, from which can be seen that lower yields coincide with those years of droughts.

Figure 12. U.S. Corn yield, tons per hectare.

Source: Earth Policy Institute and USDA 0 400 800 1200 1600 2000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

(24)

1.4 Review of Other Studies on Biofuels

Food commodity price inflation has been a popular subject over the years. Although there is now a large amount of literature written on a topic, the extent to which different factors, including biofuels, affect the price volatility is still not clearly understood.

Several studies have tried to simulate the effect of various biofuels scenarios on food price level, but such research is difficult to compare. The studies can vary substantially in terms of time periods considered, prices used (export, import, wholesale, and retail), coverage of food products, the currency in which prices are expressed, and whether prices are real or nominal (Schepf 2008). Some of the analysis missed main points, were not conducted carefully or failed to include crucial factors.

To address biofuels effect on food prices, the scientific community has used numerous different tools and approaches. According to HLPE researchers (2013), all these have been mobilized to explore the causes of the recent global food price crisis and can be separated into four main groups:

1. First group of papers varies in approach, but was mainly designed to analyze the role of biofuels creation in food prices hikes. Studies include elasticity calculations and simplistic economic models (e.g. Baier, Clements, Griffiths and Ihrig, 2009; Hochman, Rajagopal and Zilberman, 2011).

2. Second category usually uses methodology based on changes in demand and supply factors. Analysis concentrates on the period since 2005 to determine what had most plausibly led to the increase in food prices (e.g. Headey and Fan, 2010; Abbott, Hurt and Tyner, 2008).

3. Third group contains papers that use various world agricultural models to assess economic consequences of biofuels (e.g. Rosegrant, 2008; Babcock and Fabiosa, 2011; Birur, Hertel, and Tyner, 2008; Baffes and Haniotis, 2010). 4. The last set of work uses statistical models and methods to analyze

relationships between crop prices and other factors such as oil prices, biofuels prices and etc. (e.g. Mallory, Hayes, and Irwin, 2010; Hochman, Kaplan, Zilberman, 2013; Fortenbery and Park, 2008; Kristoufek, Janda and

(25)

Zilberman, 2012).

The literature on biofuels is rather broad and since it is based on different assumptions and methods, studies suggest different results and outcomes. Since this study will be using econometric approach to analyze the relationship between U.S. ethanol and corn markets, it is worth to study similar works of group 4 accurately and in more detail.

Ch. L. Gilbert’s research focused on agricultural price booms and factors that might

have possibly caused them. The analysis was carried out in two complementary stages. In the first stage author performed explanatory univariate analysis with the help of Granger causality tests. The goal of this analysis was to isolate the variables that were responsible for the price changes and were important to further analysis. The data for this part were chosen quarterly, covering the period 1971-2008, and included three dependent variables presenting three agricultural price indexes: IMF agricultural Food Price Index, Grains Price index (average of corn, rice and wheat indexes), Vegetable Oil Index (average of palm oil, soybean oil and sunflower oil indexes).

These potential explanatory variables were selected from demand-side factors as well as other factors that are common for agricultural products, and which could have been responsible for the price spikes. This list includes: world GDP volume, oil prices, real U.S. dollar exchange value, money supply, and open interest over there main Chicago futures markets (corn, soybeans, wheat). The author also notes that factors affecting aggregate food price level and the price level of separate commodities might differ (Ch.L. Gilbert, 2010).

Before the analysis, ADF unit-root test was performed, which explains why some of the variables were used as their first difference. The Granger causality test was based on ADL (2,2) equations; the period was split in half and tested separately implying that same variable might have different effect at different timing (Ch.L. Gilbert, 2010).

The second stage covers the analysis of the period 2006-2009, during which the major price boom occurred. To counter the small number of observations, the author decided to switch from quarterly to monthly data. As previously, food prices were

(26)

measured using IMF food price indexes. Three main explanatory variables were chosen: exchange rate, oil price, and index of futures positions on 12 major U.S. agricultural futures markets. To deal with possible simultaneous causality, the author constructed three equations for endogenous variable, where only exchange rate was considered to be exogenous. All the equations were estimated using OLS, 2SLS, and 3SLS.

The effect of biofuels production in this analysis remains in the residuals of the equations, which suggested that demand for biofuels was responsible for at most ¼- 1

3 of the rise in food prices. Other factors reflected in the residuals might have also contributed to this increase. To sum up, the results of the analysis do not support imposing restrictions on biofuels production (Ch.L. Gilbert, 2010).

T. Randall Fortenberry and H. Park conducted their research with the help of a

system of equations representing corn supply and demand. The work focuses on the estimation of a short-run corn price elasticity associated with production of ethanol. Authors used quarterly data for their analysis that covered 11-year period, from 2nd quarter December 1995 to 1st quarter September 2006. The quarters were created to reflect the marketing year rather than calendar year.

The model itself consists of five log-log type equations that represent five endogenous variables used in the model: price, supply, export, feed and industrial demand. The demand as noted before was constructed from three endogenous variables: corn for feed, export and industrial use demands. Further specification of five equations is as follows: price (supply, feed, export and industrial demands, previous period price, dummy variables representing different quarters); supply (previous period corn price, previous period interest rate, previous period supply, dummy variables); feed (price of corn, price of soybean meals, numbers of broilers, cattle and hogs, dummy variables); industrial demand (corn price, ethanol production, population, trend, dummy variables); exports (corn price, rest of the world wheat production, previous period exports, dollar exchange rate, GDP per capita, dummy variables). This system of simultaneous equations was decided to estimate with the help of 3SLS model. Before the estimation, identification was verified by calculating order and rank conditions (T.R. Fortenberry, H. Park, 2008).

(27)

The results of the analysis suggest that the effect of each demand factor varies significantly. The biggest effect appeared to be related with corn used for industrial purposes, where export had a second largest impact and feed effect was not even statistically significant.

Later, by substituting all the equations into the corn price equation, authors managed to separately calculate ethanol production effect on corn price. The final result showed that 1% increase in ethanol production is associated with 0.16% increase in corn price in short term. This however, did not explain the entire changes corn price, which means that demand/supply side effects are not sufficient in explaining the corn price changes (T.R. Fortenberry, H. Park, 2008).

The main difference of the study by M.L. Mallory, D.J. Hayes and S.H. Irwin is that the link between corn and ethanol prices was analyzed in futures prices of at least one-year maturity. The authors performed their research in two stages. The first stage used market efficiency theory, which implies that long-run equilibrium relationship between ethanol and corn markets is subject to zero-profit condition for a competitive industry. Using this break-even condition they constructed corn and ethanol price equations. Furthermore, the cost of carry arbitrage conditions that are specific for ethanol, corn and natural gas futures markets, were used to forecast corn spot prices from long-term relationship. These conditions suggest that the price of futures contracts should account for equilibrium carry, which must include the compensation for physical costs of storage, interest and convenience yield. Therefore, futures prices can be expressed as a sum of spot prices and carry (M.L. Mallory, D.J. Hayes, S.H. Irwin, 2010).

The second stage includes econometric analysis used to test the validity of the results obtained by equations constructed in the previous stage. The data set covers daily time series with 932 observations from January 3, 2007 to September 8, 2010. The variables were chosen nearby and one year to maturity ethanol, natural gas and corn prices. Before the analysis data were tested for cointegration using ADF model and Johansen test for cointegration. Afterwards, VECM was used to test for long-term and short-long-term relationships between three variables.

(28)

The results of the analysis support the idea that ethanol, natural gas and corn prices are governed by a break-even relationship in forward markets and concluded that there is a clear link between prices of corn and ethanol, which started in the mid-2006 and continued to the fall of 2010 (M.L. Mallory, D.J. Hayes, S.H. Irwin, 2010).

G. Hochman, S. Kaplan, D. Zilberman analyzed factors impacting food commodity

prices via statistical model that also accounts for the previous works on a subject. Already existing research provides prior information that is considered as a base for more refined statistical analysis of the relationship between food and biofuels markets.

Authors first established the multi-market food commodity model consisting of supply and demand (food/feed, biofuels, inventory) equations that were later used to derive the empirical equation of world price of different crops. This main log-linear type equation includes such variables like exchange rate, trade policy index, energy price, gasoline price, maximum temperature, GDP per capita and previous period world price of crops. Authors chose panel data set for each country in each year.

An important feature of this analysis is explicitly taking into account and incorporating the adjustments in inventories of different food commodities. For this reason, previous period food-commodity prices were included into estimation, such that the higher the price in the previous period, the lower the inventories at the end of the current period. Therefore, expectations for needed inventory increase and prices spike (G. Hochman, S. Kaplan, D. Zilberman, 2013).

Finally, using the prior data, authors employed Bayesian estimation techniques to estimate the posterior distribution of the parameters, and quantify and assess the importance of different factors in explaining crop prices.

One of the main focuses was on the importance of the inventories, which was analyzed in two key commodities: corn and soybeans. In the corn market this effect turned out to be substantial and also significant at a 5% level. Introducing corn ethanol into analysis directly showed very little effect however, when biofuels were introduced indirectly via gasoline prices, the effect grew significantly and accounted for almost 20%. Other factors included GDP per capita, which did not have a

(29)

significant effect of 9%; exchange rate 13%; energy price 5%; and fertilizers accounted for 4% (G. Hochman, S. Kaplan, D. Zilberman, 2013).

The following table summarizes the main points of earlier discussed studies. The limitations of these studies will be addressed in the further research of this paper.

Literature analysis summary

Coverage Approach and estimation techniques Results Limitations Source: Ch. L. Gilbert

“How to Understand High Food Prices”

Quarterly time series: 1971 – 2008 Monthly time series: 2006 – 2009 Granger causality test. OLS, 2SLS, 3SLS. Global biofuels production accounted for at most 1/4 – 1/3 of the increase in global food prices.

The effect of biofuels remains in the residual, therefore cannot be estimated precisely

Biofuels effect on aggregate and separate commodities price level can differ.

Biofuels effects in different countries differ.

Small amount of explanatory variables might result in omitted variable bias. Source: T. Randall Fortenberry and H. Park

“The Effect of Ethanol Production on the U.S. National Corn Price” Quarterly time series: 1995 – 2006 Demand – Supply factors analyzed via system of simultaneous equations (3SLS) 1% increase in U.S. ethanol production is associated with 0.16% increase in corn price in the short term.

Short time frame of the analysis. Lack of macroeconomic variables. Non- stationarity of the data.

Source: M.L. Mallory, D.J. Hayes, S.H. Irwin

“How market efficiency and theory of storage link corn and ethanol markets” Daily time series: January 3, 2007 – September 8, 2010 Structural price model (based on market efficiency and supply of storage theories) and time-series VECM.

Corn and Ethanol prices are governed by breakeven conditions. This relationship is maintained in the forward markets.

Short time frame of the analysis. Relationship between the prices does not explain to what extent ethanol production effects corn prices.

Source: G. Hochman, S. Kaplan, D. Zilberman “The causes of recent food commodity crises”

Yearly panel data: 1991 - 2010 Multi-market commodity model of supply and demand equations. Bayesian estimation techniques. Ethanol production alone did not show significant effects. Significant effect was noted only through gasoline prices.

Possibility of simultaneous causality between the variables.

Biofuels effect in different countries might differ due to different policies.

(30)

2. Empirical Analysis

This part of the thesis uses empirical analysis to test the relationship between corn price and ethanol production in the United States. The research is carried out in three stages, using econometric techniques. Each stage contains a different estimation approach, which was chosen according to the different assumptions about the proposed link between the two commodities. Previous research results and methods are also taken into account and serve as a base in choosing various variables for the analysis. In the beginning of this section, the methodology of research is presented in more detail, followed by the description of data and variables. At the end, the results and validity of each method is compared given the conditions of the available data.

2.1 Methodology

As mentioned before, the main research question about the relationship between corn prices and ethanol production is analyzed in three different stages. This approach was chosen after careful analysis of the existing literature, which suggested different assumptions regarding this link. The effect is first estimated using a simple Ordinary Least Squares (OLS) regression analysis; in the second stage the model is augmented by introducing system of simultaneous equations and Three Stage Least Squares (3SLS) estimation; and finally the long run and short run relationship between both commodities is examined via Johansen Test of Cointegration and Vector Error Correction Model (VECM).

OLS estimation

Time series OLS regression analysis is the first and the most basic strategy to start with. The main assumption behind it is that the causality between dependent and independent variables runs only one way. Meaning, current ethanol production has a direct effect on contemporaneous corn prices, but contemporaneous corn prices do not effect current ethanol production. This assumption might be valid for a couple of reasons. First of all, most corn sales are made in futures contracts, which imply that ethanol producers would be getting their corn supply at the prices they agreed on a few periods ago, and so spot prices of corn would not affect their current production. Secondly, if ethanol producers are holding sufficient levels of corn stock, then corn spot prices will not have an effect on their current production levels. However, this

(31)

only remains an assumption since there is no available data regarding the corn stocks held by ethanol producers or prices at which they make their purchases. If this assumption holds, then independent variable together with control variables will provide consistent unbiased estimate, given that all the other OLS assumptions hold. Specification of the estimated model is as follows:

𝑌! = 𝛽!+ 𝛽!𝑋!+ 𝛾!𝑊!!+ ⋯ + 𝛾!𝑊!"+ 𝑢! ,

where, 𝑌_! is a dependent variable at time t; 𝑋_! is an independent variable at time t; 𝑢!is error term; and 𝑊! is k number of different control variables at time t, so that

𝐸[𝑢!|𝑋!, 𝑊!!, … , 𝑊!"] = 0 . This allows for 𝑋! coefficient to have unbiased causal

interpretation, even though coefficients on control variables might not provide correct, unbiased interpretation.

3SLS estimation

In the previous case OLS estimation was conducted assuming that independent variable 𝑋_! was exogenous and conditional mean assumption holds. However, majority of the existing literature agrees on the opposite case, which implies that simultaneous causality among corn price and ethanol production exists, and therefore OLS estimates are inconsistent and subject to omitted variable bias.

One of the options to address this endogeneity problem is instrumenting the endogenous variable 𝑋_! with one or a set of exogenous variables. There are a few different methods that account for simultaneous causality. Probably the most popular and the most widely used in the literature is Two Stage Least Squares (2SLS) model. This model provides consistent and efficient results for single equation estimation, but ignores information related with endogenous variables. Another approach used by researchers is Seemingly Unrelated Regressions (SUR). This estimation differs from 2SLS, because it suggests correlation among error terms across different equations in the system, but does not account for endogeneity within each equation. The most superior method, which is a combination of the previously described estimation techniques, is 3SLS model. It is used to estimate the systems of equations that include endogenous variables on both sides of the equations and also accounts

(32)

for cross-equation error correlations. Therefore, it was this particular model that was chosen for the further analysis.

Specification of the model is as follows:

𝑌_!! = 𝛽_! + 𝛽_!𝑌_!!+ ⋯ + 𝛽_!"𝑌_!"+ 𝛾_!𝑋_!!+ ⋯ + 𝛾_!𝑋_!"+ 𝑢_!! 𝑌_!! = 𝛽_!+ 𝛽_!𝑌_!!+ 𝛾_!𝑊_!! + ⋯ + 𝛾_!𝑊_!"+ 𝑢_!!

…………

𝑌!" = 𝛽!+ 𝛽!𝑌!!+ 𝛾!𝑍!!+ ⋯ + 𝛾!𝑍!"+ 𝑢!",

where 𝑦! = (𝑌!!, 𝑌!!, … , 𝑌!") is a vector of endogenous variables in the system,

𝑧_! = (𝑋_!!, … , 𝑋_!", 𝑊_!!, … , 𝑊_!", … , 𝑍_!!… , 𝑍_!") is a vector of all the endogenous variables in the system, 𝑢! = (𝑢!!, 𝑢!!, … , 𝑢!") is a vector of error terms across all the

equations. The model is estimated using structural equation specification form, with the entire set of exogenous variables representing instruments. As always, it is assumed that exogenous variables are uncorrelated with error term 𝐸 [𝑢_!, 𝑧_!] = 0, while endogenous variables are 𝐸 [𝑢!, 𝑦!] ≠0. Before the estimation, each equation in

the system was tested for identification by checking rank and order conditions. VECM

The goal of this estimation is to test for long run and short run relationships between the corn prices and ethanol production. Careful analysis of the existing research and data suggest that neither of two variables are stationary time series. This is also confirmed by performing ADF test for unit root (Annex 1). The results of this test for multiple variables are summarized in the next part of this thesis.

Second step is to check the variables that are both integrated of order 1 for cointegration. This was performed using Johansen Test for Cointegration, which supported the hypothesis and concludes that variables are cointegrated with rank 1. Final step of the analysis is made using VECM and is summarized as follows:

∆𝑌!= 𝛽!"+ 𝛽!!∆𝑌!!!+ ⋯ + 𝛽!!Δ𝑌!!!+ 𝛾!!∆𝑋!!!+ ⋯ + 𝛾!!ΔX!!!+ 𝛼! 𝑌!!!− 𝜃𝑋!!!

(33)

∆𝑋_! = 𝛽_!"+ 𝛽_!"∆𝑌_!!!+ ⋯ + 𝛽_!!Δ𝑌_!!!+ 𝛾_!"∆𝑋_!!!+ ⋯ + 𝛾_!!ΔX_!!!+ 𝛼_! 𝑌_!!!− 𝜃𝑋_!!! + 𝑢_!!

where (𝑌!− 𝜃𝑋!) is called error correction term and helps to predict future values of

∆𝑌_! and/or ∆𝑋_! (J.H. Stock and M.M. Watson, 2012). This model is crucial in the analysis, as non-stationary data might result in spurious correlations among variables, which might lead to unreliable estimates of the coefficients in the regressions.

2.2 Data

This study uses quarterly time series data on different variables across the United States. The sample covers the period of 33 years and ranges from 1980q2 to 2012q4. The data set is adjusted to coincide with U.S. corn marketing year, so that the first quarter starts in September, second in December, third in March and fourth in June. This type of quarter specification also matched seasonality.

The data on most of the U.S. agricultural commodities and livestock is obtained from a database of the Unites States Department of Agriculture. The global prices of other commodities, like fuel, fertilizers and natural gas, were gathered from World Bank commodity price data report, so called “Pink Sheet”. Climatic Research Unit at University of East Anglia provided historical U.S. rainfall statistics. Data on U.S. population and exchange rates were gathered from Economic Research Division at Federal Reserve Bank of St. Louis, federal funds effective rate from Board of Governors at Federal Reserve System, real GDP per capita from Bureau of Economic Analysis at U.S. Department of Commerce and author’s calculations. Data on historical corn futures market open interest is from Chicago Board of Trade and finally production of ethanol in U.S. from USDOE, Energy Information Administration. Most of the variables in the models are used in a form of logs so that regression coefficients could be interpreted as direct estimates of elasticities, rather than changes in units across time. This adjustment allows reading and comparing the results more conveniently. Furthermore, in VECM some variables are also converted to their first differences as dataset of these variables turned out to be not stationary. The results of ADF unit root test are summarized in the Annex 1 at the end of the paper.

Relationship between corn prices and ethanol production the case of the United States.