The impact of biofuels on food prices

1

The impact of biofuels on food prices

Universiteit van Amsterdam

MSc Thesis in Economics

Track: International Economics

27/08/2014

Author: Lorenzo Poloni Student number: 10604766

lorenzo.poloni@student.uva.nl

Reader: Dr. Veestraeten D. Second reader: Dr. Klaassen F.

2

Chapter 1: Introduction

Between 2007 and 2008 food prices reached their peak in international markets, after some years of surge that had followed a declining trend that went on since the beginning of the 1970s until the first years of the 2000s. There seem to be various reasons for this increase, but the economic literature seems to agree in listing the production of biofuels among the causes, even if with drastically different opinions regarding the size of its contribution. This connection between biofuels and food prices will be the main focus of this thesis.

Figure 1: FAO food price indexes (2002-2004=100)

Source: Food and Agriculture Organization (FAO)

Biofuels are a product of the processing of various crops, such as sugarcane and corn in Brazil and U.S. and rapeseed and palm oil in Germany and Malaysia, respectively. The fact that these crops may be used for an energy purpose establishes a new link between food prices and liquid fuels, introducing new dynamics in the price pattern of the former (De Gorter, Drabik, Just, 2013; De Gorter et al., 2013). These dynamics will be explained throughout this paper. High food prices and their increased volatility are of fundamental relevance in the global economy since food demand seems to be already destined to grow along with world population and increasing prices are likely to lead to social unrest and worsen the situation of poorer countries, in which almost half of household income is spent on food.

Various nations decided to invest in biofuels as an energy source, subsidizing this activity to face the high costs of this technology in order to stimulate production (De Gorter, Drabik, and

Timilsina, 2013; Cheng, Timilsina, 2011). The result was that crops producers in many cases

adapted their cultivation to the purpose of biofuel production, converting their field production in some cases or just selling their products to different buyers. The augmented crop demand had

0,0 50,0 100,0 150,0 200,0 250,0 300,0 350,0 400,0 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14

3

repercussions on global markets, affecting especially those countries that imported crops for feeding purposes, that saw prices increase.

After the peak food prices fell again, but the trend after 2008 still appears to be an upward one. The general price plunge seems to point to the responsibility of the financial crisis for this fall, according to the literature, but underlying reasons for high food prices to reoccur in the future seem to remain. If biofuel production actually affected food prices, it is likely that it may occur again. In fact biofuel production is still strongly subsidized and data from EIA (U.S. Energy Information Administration) seems to indicate it also substantially increased from 2007. On the other side technological improvement should at least reduce if not eliminate the competition with food production in the future, since the new generation (2nd generation) of biofuels should

produce energy from different biomasses that do not directly compete with food production, such as agricultural and forest residuals and non-crop feedstocks in general.

Second generation biofuels appear to be on their way for a widespread diffusion, but so far the high costs of production and plant creations have been a stumbling block (Cheng, Timilsina, 2011). In fact 1st generation biofuels, which are based on crop-feedstocks, are still predominant on the biofuel market at present time. Given this scenario, the main question arising is how and to what extent food prices are going to be affected by biofuel production when the global economy will have recovered from recession? This is clearly a question that is difficult to find an answer for, given the uncertainty about the overall situation of the world economy, but it may at least be addressed with a comprehensive overview of the present state of the interaction between the biofuel and the food market. This will be the purpose of this work. Thus, the research question that this paper aims to answer is : Has biofuel production affected food prices so far? This may also give insights on what could be the possible outcome of this relation in a near future. The starting point would be an analysis on the present literature on the subject, with major attention to the link between food prices and biofuel production. The strand of the literature on the topic seem to have for a great part focused on the period 2003-2008 (roughly) when the rise in food prices and the increase in the production of biofuels brought attention to this issue.

In order to have an idea about the impact of biofuels on food prices, this link will be analyzed with a regression, taking food prices as the dependent variable and biofuel production as one of the independent variables. The model used for this estimation will be a vector error correction model

4

(VECM) to determine the long-term relation between variables. The short-term effects will be estimated by OLS.

Zilberman et al. (2013), reviewing the findings of several analyses, conclude that biofuels may affect food prices, but with a differing impact depending on crop and location. For this reason, the analysis of this paper will be comprised of two separate regressions, analyzing the main feedstocks that agree to the bulk of the biofuel production.

There are two main categories of biofuels: bioethanol and biodiesel, which come from different feedstocks. The former comes from the direct processing of the feedstock, whereas the latter is mainly based on oils that are obtained from the feedstock, which implies an extra passage in the processing. Due to data limitations, this analysis will consider only the impact of bioethanol on the feedstocks that are used in its production.

According to Rajcaniova, Drabik and Ciaian (2011), Brazil and the U.S. are price setter for world bioethanol, whereas the E.U. is price setter for biodiesel. In Brazil bioethanol production comes almost entirely from sugar cane and in the U.S. the feedstock used is corn for 97% of the total production, according to Ajanovic (2009). For these reasons the idea would be regressing the price of each of these feedstocks, using biofuel production in the area of interest as an independent variable: for instance corn price associated to biofuel production in the US. Focusing on the impact of biofuel production on the price of their relative underlying source is expected to yield more precise results than the impact on aggregate food prices. The regressions would include the variables that were indicated as drivers of food prices in the literature to explain the price variation. The outcome of these regressions should indicate the impact of the production of the main biofuels on their relative feedstock prices. The data analyzed should be taken from a 15 years span. Data for biofuel production tend to start around the beginning of the 1990s and they will be used until when they are available.

Chapter 2 will present on overview of the recent patterns of the two main variables of interest, namely food prices and biofuel production. Chapter 3 will try to present the mechanisms behind food price variations providing a theoretical explanation, Chapter 4 will be dedicated to the empirics and Chapter 5 will report the conclusions.

5

Chapter 2: The patterns of food prices and biofuel production

This chapter will allow to take a glance at the recent trends of the main variables of interest, namely food prices and biofuel production, in order to allow further reasoning.

2.1 Recent trends in food prices

The agricultural improvements introduced by the so-called “Green Revolution” reduced food price volatility and introduced a slowly declining trend in the level of prices that went on since the 70s. After having reached historical lows in 2000 and 2001 due to the Asian financial crisis (Mitchell 2008), food price started rising, reaching their peak between 2006 and 2008, depending on the traded commodity. After the peak a plunge occurred, allegedly due to the global recession, but after that, prices started rising again. High prices in food commodities represent a major concern especially for less developed and developing countries, so that in these recent years of historically high levels, food prices caused riots in many countries and led to policy actions in others, such as ban on exports or tariff reductions on imports (Mitchell 2008). Such policy actions should have also contributed to intensifying the upward pressure on food prices, but there are surely various reasons for this increase and they will be analyzed thoroughly in Chapter 3.

The FAO price index is a weighted average of prices of different kinds of food commodities, in the case represented by figure 2, indexed with respect to the values in 2002-2004=100. The index in Figure 2 gives a representation of food price trends in the last decade and a half. From this aggregate food price index it is possible to notice the aforementioned peak occurring around 2008, the fall in the following year, whose lowest point is still remarkably higher than the values the index departs from, and another surge in 2011 that , at first glance, appears to stabilize prices until 2014. Splitting the index into different products seems to reveal that , on the contrary, price variations have occurred in the last few years also.

6

Figure 2: FAO aggregate food price index (2002-2004=100)

Source: Food and Agriculture Organization (FAO)

Figure 3: FAO food price indexes (2002-2004=100)

Source: Food and Agriculture Organization (FAO)

Taking the indices of different categories of commodities (Figure 3), it is possible to notice some particular features of the price trends. In the first place, the indices seem to follow a similar pattern, even if with some discrepancies. The only exception appears to be for sugar, whose price appears to be behaving in a clearly different way.

The fact that, in general, food prices have had common movements may suggest a common cause that directly determine the price of all the commodities, but this is not necessarily the case. In fact there are great interdependencies between food commodities, so that a variation in the price of single commodity may have repercussions on the price of the others. These interdependencies are mainly due to substitution effects, that may occur both in consumption and in production. Due to these kind of interdependencies, it is also possible that a shock in the market of a single product (or more than one) can indirectly influence other products as well. For example, the rise in the

0,0 50,0 100,0 150,0 200,0 250,0 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 0,0 50,0 100,0 150,0 200,0 250,0 300,0 350,0 400,0 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14

7

price of one single cereal, corn, is hypothesized. If the price of this cereal rises, a consumer may want to buy another kind of grain to substitute it with, let us say wheat. This would increase the demand for wheat and, in theory, also its price. Moreover, a farmer may want to adapt his cultivation toward the most profitable solution. Given a rise in the corn price, a barley producer may want to start producing corn because it has become a more profitable choice. This decreases barley supply, in turn increasing its price. In addition to that, it should also be considered that an hypothetical increase in corn would affect not only final consumers and producers, but also other agents that use corn as an intermediate good in their production chain. In fact corn is by far the most diffused feeding in the world, in addition to being used in industrial production with various purposes, so a higher price would increase the costs of production of livestock feeding and therefore also the price of meat and dairy products. It is interesting to notice how the only food category in the graph that seems to be distant from this kind of logic, except theoretically for the substitution in production, is sugar, which is also the one that presents a dynamic pattern that most clearly differs from the others. On the other side, it should also be noticed that the meat price index seems to be the least volatile. A possible explanation in light of the previous reasoning may be the fact that meat producers may hedge against the risk of higher feeding prices buying such intermediate product in advance via derivative purchases, therefore decreasing the volatility of the price of meat but this does not seem to be coherent with the fact that dairy products follow a different trend.

2.2 Recent trends in biofuel production

Environmental concerns and the will to lessen energy import dependence are the factors that spurred most of the world’s main economies toward the use of alternative energy sources. Efforts in this direction were made already since the 1970s and then more after the increase of oil prices that occurred during the Gulf War. The challenge presented by the use of alternative energy gained importance over the years, effectively becoming a central topic in developed countries’ agenda only in the last 10 to 15 years. Among possible solutions to tackle this issue, biofuels turned out to have been a popular option in recent years.

One of the main factors that favored this solution is that biofuels are a substitute for gasoline, even if not perfect due to the different performance (lower amount of energy per liter), which implies that it is possible to use biofuels with the same type of machines that run on oil, with relatively small modifications to the engines. The fact of having a substitute of this kind would

8

imply reducing dependence on oil, which is an import for the majority of countries, and reduce emissions caused by the combustion of fossil fuels.

The first country to invest in this sector was Brazil, which is currently one of the biggest producers of biofuels. Brazil started the alternative energy program in 1975 due to the major increases in oil prices (Rothkopf and Garten, 2007), with the idea of producing ethanol from sugarcanes instead, a resource widely available in the country and in whose production Brazil is the global leader. The ethanol market in the country got effectively liberalized between 1996 and 2000, contributing to create an internal demand for biofuel (Balcombe Rapsomanikis, 2008).

The United States also started producing biofuels from an early stage, in fact ethanol production is ongoing since 1980 (Renewable Fuels Association).

The main challenge from those years until today has been, and still is in certain cases, reducing the production costs that were prohibitive and prevented the full commercialization. This is also the reason why biofuel production still needs certain policy actions in order to be viable, such as subsidies and blending mandates. A blending mandate is a norm that imposes a minimum

blending percentage of the fuel that is sold in one country, for instance the gasoline sold in Brazil must be currently blended with biofuel in a percentage between 18 and 25 points. Investments in research and technological development managed to systematically decrease production costs over the years and this, thanks to the aforementioned policy actions, lead to a steady increase of the total production over time.

These technological advancements have been able to create economies of scale and thus allowing for further diffusion of biofuels , so governments decided to spur the activity further. During the 1990s U.S. agencies such as the EPA (Environmental Protection Agency) promoted laws to increase the demand and properly develop the biofuel market. Europe, because of the resources available, invested in the production of biodiesel, which per se presents higher production costs. This is the main reason why European production started later on and gained importance in the union policies starting from 2004.

Whereas Brazilian biofuel production steadily grew over the years, other producers saw a

spectacular increase during the last decade. The U.S. in the first place but also European and Asian producers have in fact significantly augmented their production level, increasing the world

9

Figure 4: World biofuel production

Source: U.S. Energy Information Administration, International Energy Statistics Figure 5: Total biofuel production of Brazil, Europe and US

Source: U.S. Energy Information Administration, International Energy Statistics

Looking at the graph that represents the production level of the main producers in recent years (Figure 5), it can be noticed that Brazilian production’s growth suffered a downturn around 2008, which in turn should be the reason why world production’s growth appears to be slowing down between 2008 and 2011, given the relative weight of Brazil in the sector. The reason behind the decrease in production for Brazil appears to be given by the increased price of sugar, so that it became more advantageous after 2008 to sell sugar instead of using it for biofuel production (De Gorter et al. 2013)1_.

1

De Gorter, Harry, et al. "An economic model of Brazil's ethanol-sugar markets and impacts of fuel policies." (2013). 0 500 1000 1500 2000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Th ou sa nd b arre ls pe r d ay 0 100 200 300 400 500 600 700 800 900 1000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Th ou ns an ds b arre ls p er d ay

10

Figure 5 represents the biofuel production in the three major world producers, that is European Union, the United States and Brazil. Also a great number of Asian states are active in the sector, with a significant cumulated production. Taken individually though, the quantities produced in the single Asian countries do not have great relevance. Even if currently the overall quantities

produced are substantially higher than in the years before and have become economically

relevant, the biofuel one is still a developing sector, therefore the size of the market is still limited if compared to other energy sources’. In fact, the current level of technological advancement does not allow for large scale production yet, since it would require too many resources, land in

particular, to satisfy a good percentage of the energetic needs of a developed nation like the U.S. (Cheng, Timilsina, 2010 ). This is the reason why attempts have been made to expand the range of feedstocks that can be used in order to produce biofuels. In fact current bioethanol production is mainly coming from sugar and starch portion of plants and biodiesel production from vegetable oils. Recently though, it has been made viable to produce bioethanol from various lignocellulosic materials (such as wood, grasses, agricultural residues) and biodiesel from animal fats, waste oils and organisms such as microalgae. These kind of technologies are already available, but still far from being fully commercialized, therefore will not be considered when analyzing biofuel production insofar. The products of these technologies are denominated advanced biofuels, or second generation biofuels.

Chapter 3: Reasons for the food price increase

Chapter 3 will present the factors that appear to influence the price of food commodities and the way in which this occurs. A separate section will focus on the possible effect of biofuel production.

3.1 What affects food prices

There seem to be various factors that have contributed to the recent price increase, according to the ample literature on the topic, so that different forces are in play and appear to act both on the demand and on the supply side. Moreover, these forces add up to some particular conditions of the market. In fact, interconnections between the prices and market restrictions are plentiful in the food commodity sector, and this contributes to complicate the economic mechanisms that can be expected. This is an important reason why the results aimed at estimating the effect of the single factors tend to widely differ in light of the assumptions made and the model used. For instance, Gilbert (2010), using a Capital Asset Pricing Model (CAPM) estimates the impact of

11

biofuels as almost irrelevant in the 2007-08 price surge, whereas Mitchell (2008) through a

theoretical analysis attributes responsibility of 70-75% of the food price surge to “biofuels and the related consequences of low grain stocks, large land use shifts, speculative activity and export bans” (Mitchell, 2008, page 17). In the following part the elements that appear to have had a contribution to the price variations will be analyzed, with specific attention to the biofuel sector in a separate section, and the interactions that occur between the different factors will be

investigated. The choice concerning how to include these elements in the empirics of this paper will be presented in Chapter 4.1 and 4.4.

3.1.1 Oil prices

The price of oil is unarguably a factor to take into account, as the literature suggests, while explaining variations in food prices. It is the primary energy source in any stage of food

production, it is a major input for agricultural products such as fertilizers and it directly affects transport costs. It is then evident that if the price of oil goes up, so will the costs incurred by primary producers (such as farmers), but also by the processing industry. Then, in theory, food commodities will be more responsive to an increase in oil price the more they are processed and the more they travel. Mitchell (2008) estimates that higher energy and transport costs raised agricultural production costs by 15-20%. Baffes (2007), analyzes the rates of pass-through between the energy and the agricultural sector: his findings confirm the great responsiveness of food prices, especially during spikes, and the synchronous movement of the cost of energy and food prices. Figure 6 represents a comparison between the prices of crude oil and food. It is noticeable how they present a similar pattern and the strong correlation between the two is particularly evident in the spike occurring in 2008, supporting Baffes’ findings reported in the previous line.

Even though the relation between the oil price and the food price has been widely considered as positive by the literature, other authors beg to differ, in light of their findings. Most notably Zhang et al. (2009) run a VEC (Vector Error Correction) model whose outcome rejects that oil price cause variations in the food price, both in the long-run and in the short-run. This paper is one of the few empirically analyzing this relation in a time span which will be wholly included in this analysis, using monthly data from 1989 to 2008. Cooke and Robles (2009) using a similar model in a similar data span, find a negative coefficient for the oil price, even though their final claim is that such an analysis appears to be inconclusive.

12

Figure 6: Comparison between food and oil prices

Source: FRED Economic Data- Average between Crude Oil prices in Brent and West Texas Intermediate, Food and Agriculture Organization , 1990- 2014

3.1.2 Supply shocks

While discussing agricultural price variation, one cannot overlook the shocks that may occur on the supply side, because of which the crop production in a given year may be greatly affected by climatic effects, weather events such as droughts or floods and others. While referring to the worldwide market, these effects may have a reduced relevance, because it is not unlikely that a hampered production in one area may be offset by a good production elsewhere. This is the reason why the problem of supply shocks has been sometimes disregarded in the literature concerning food price analysis, even though it has been included in many empirical research, such as Cook and Robles (2009), De Gorter et al. (2013), Baier et al. (2009) and others. Nonetheless, the fact that the effect of these shocks may be cumulative, for example if bad conditions occur for more than one year, and may be relevant in magnitude, especially if world production of a certain commodity is concentrated in some areas, make it a variable not to overlook.

The commodities in analysis are corn and sugar. Corn is a diffused crop since it is suitable for various climates and is therefore produced in many countries around the world. It seems reasonable that unpredictable supply shocks may be absorbed by extra production across different nations, thus reducing the effect of external factors like weather events. The same cannot be said for sugar. Brazil is by far the main sugar producer: almost 25% of worldwide sugar production is located there, whereas the 10 main producers account for roughly 75% of total

0 20 40 60 80 100 120 0,0 20,0 40,0 60,0 80,0 100,0 120,0 140,0 160,0 180,0 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 Av er age O il p rice ($ pe r b ar re l) FAO p rice in de x

13

production2. It is then clear that this great concentration makes sugar a commodity extremely susceptible to weather events and climatic shocks. As a matter of fact, also the data about biofuel production in Brazil presented in the previous chapter had a downturn because of an increase in the price of sugar. This increase, according to De Gorter et al. (2013), was mainly due to bad climatic events that affected the production between 2010 and 2012. This corroborates the idea that weather events should indeed be influential in determining the sugar price.

3.1.3 Speculation and financial markets

Among the reasons for the food price increase, speculation appears to play a relevant role. As a consequence of the so-called “Great Moderation”, commodity prices in general increased, in particular in the time span ranging from 2002 until 2007. Excess liquidity and low interest rates due to the perceived safety of the market widened the pool of possible investments and spurred investors to seek return opportunities in all sort of commodity markets (among others) and the food sector was no exception. Investments in food commodities became particularly appealing due to the weakening dollar in the last decade, since most of international agricultural

transactions are carried out in this currency (Baffes et al, 2010). This implies that whenever the dollar depreciates it becomes relatively cheaper to buy and to invest in food commodities for investors and countries whose currency has appreciated with respect to the dollar. Therefore, a dollar depreciation is expected to increase the demand for food commodities and increase the price, not only via the speculation effect. Moreover, decreasing interest rates are supposed to stimulate speculative investments and the effect of a decreasing interest rate should be a currency depreciation, in the long run, so the role the exchange rate may take up is twofold. The intense speculative activity of the last decade is supposed to have inflated prices (Gilbert 2008), so that speculation is a factor that goes beyond the simple dynamics of supply and demand of the traded good.

Due to the volatility of the production that characterizes the agricultural sector, market agents that require food commodities in order to pursue their activity tend to secure their purchases by buying quantities in advance via derivatives. For instance, an industry that processes food wants to be sure of the price that will be paid for the prime matter needed for their activity in the following periods by negotiating it in advance via derivative contracts, not to incur in losses due to a price increase. Master (2008), as reported by Cooke and Robles (2009), reports that in recent years

2 _{Data from Sucden}

14

index funds profusely invested in commodity future markets, influencing the price of the underlying goods. The uncertainty, the expectation of a price increase and an inflated price for commodities might also have incited the hedgers to a stronger activity on the derivative market, negotiating at a higher price because of speculators on the futures market and resulting in a contribution to strengthen the upward trend of the food commodity market. In any case, the amount of futures with a certain underlying commodity traded may represent an informative measure for explaining part of the price increase, capturing part of the effect of speculation (only the one on derivative markets) jointly with the hedging mechanism just explained that may have occurred.

Gilbert (2010) distinguishes between commercial and non commercial positions while referring to food commodities. In fact, investments with a speculative motive typically imply a non-commercial position and impact the spot price, concretely affecting the price level of the commodity. On the other side, future contracts that are typical of commercial positions, in theory should have no effect on the spot price, as stated by Krugman (2008) and reported in the article of Cooke and Robles (2009). The fact that there are upward pressures on the spot rate though, may be able to influence the acquisitions made by future contract and affect the spot price in following periods. The literature suggests different measures to represent the effect of speculation on the price of a commodity, varying from macroeconomic fundamentals such as the interest and the exchange rate, to more elaborate indicators such as the amount of futures traded (open interest ) and the amount of non-commercial position suggested by Gilbert (2010).

3.1.4 Demand growth

The demand for food is supposed to be on the rise due to population growth, especially for

developing countries. In fact populous countries such as China and India presented growth rates of their population respectively above 0.5% and 1.5% for the whole decade 2000-20103, which given the number of inhabitants represents a relevant amount. This in turn implies that such numbers should impact the world food demand and thus affect the price. The demand increase from developing countries though, should be mainly due to the shift in the dietary patterns that arises as a consequence of the increase in the available household income. The economic growth of countries such as BRIC (Brazil, Russia, India and China) and others, arguably translated into an

3 _{Data from World Bank DataBank}

15

increased demand for variety and higher-value goods, such as meat and dairy products. These goods in turn impact the demand for feeds, a great bulk of which is grain-based (Henry and Correa, 1992), so an increased demand for livestock-derived products should put a great upward pressure on the price of grains. This kind of shift in the food commodities demand is expected to have a great impact on world quantities since most of the aforementioned developing countries are extremely populous. So, for instance, an increased meat demand in China alone should be able to have major repercussions on world prices, not only for meat, but also grains. Interestingly though, data do not seem to support this theory while explaining the recent food price increase. Baffes and Haniotis (2010) highlight how food demand growth in the last years did not accelerate neither in China nor in India nor in the world as a whole. In their analysis they consider a data span that starts in 1961 and which is divided in four periods. Their results show that food demand growth has been slowing down over time (including also the demand for meat), with the exception of the last decade, where demand increased for maize and soybeans, possibly for grains, but not for meat and dairy products. Therefore, the rise in consumption of these countries has been more influential before the period 2002-2008. Also, as reported by Abbott et al. (2011), the agricultural policies of these countries have led to the self-sufficiency with respect to some food commodities, especially grains, among which corn, wheat and rice for China. India on the other side

promulgated an export ban in 2007 for some grains, most importantly rice, following the concerns due to the food price increase, once again indicating how these populous country aim to rely on their own agricultural policy rather than resorting to international markets.

3.1.5 Food stocks and market characteristics

The agricultural policies introduced by the so-called “Green Revolution” (1961-1972) incentivized production and led to structural surpluses that started being accumulated as reserves. This favored a steady decrease in food prices for the following three decades which became

particularly evident at the turn of the century (visible in Figure 2 and 3 in Chapter 2), leading in turn to a decrease in the amount of cultivated land, since producing was less profitable, and a decrease in the expenses in R&D in the agricultural sector, which theoretically would reduce productivity growth in the following periods (Abbott et al.,2011)

Hamelinck (2013) cites an Oxfam report concerning the Uruguay round repercussions. According to it, agricultural policies and increased market openness may have harmed less developed countries by driving food prices to such a low level that countries that were less efficient in

16

agricultural production were excluded from the market (from the supply side). This point in particular may be relevant in light of the turmoil generated in some nations by the food price increase, because if this was the case, more countries would have become net importers and, as such, more exposed to the risk of price increases. This may also mean a more rigid demand for certain grains by some countries in case of a price increase.

Food stocks act as a buffer, so that in case of imbalances with demand exceeding the production capacities, food reserves will decrease and the price will increase gradually. The mechanisms explained contributed to narrowing the gap between food production and consumption, eroding reserves over time, in particular after the year 2000 (Abbott et al., 2011). The literature seems to agree on the fact that the reduction of reserves has been influential in increasing the

responsiveness of the food commodity market, so that when shocks occurred international prices have been more affected.

While analyzing the agricultural commodity market, there are some points that one should bear in mind: first of all, protectionist measures are usual praxis for many countries, if not all.

Governments tend to levy import taxes on agricultural goods in order to defend national production and their agricultural sector, or to apply export taxes to secure their internal food demand. The main effect of trade restrictions is naturally a limitation of the quantities traded. An import tax in a country may prevent a foreign nation to export in such country and an export tax may prevent food commodities to reach international markets: in both these cases the effect would be an increase in the price of internationally traded commodities because of the reduced quantities that reach the international market. No export or lower exports from one country reduce world supply. Import taxes in one country diminish internal demand in that country, which reduces demand from that country as a whole. Because of this trade restrictions contribute to price increases.

Another implication is that trade restrictions contribute to create an international market with exchange prices that may drastically differ from those that can be found on local markets. On one hand, local markets tend to be more volatile because more restricted, although they would still be affected by variations in the international market, on the other hand shocks that occur at a local level should be less likely to affect international prices. Secondly, trade restrictions reflect in a lower level of efficiency of food markets both at a local and international level, because it

17

becomes more difficult to sell extra production at advantageous conditions, so that wastes and stock accumulation are incentivized.

Between 30 to 50% of food production is lost along the supply chain, as reported by Hamelinck (2013) in his article, due to different causes. In developed countries undervaluation of food commodities is the main factor leading to wastes, whereas for developing countries the problem tend to be due to inadequate facilities concerning production, transport and/or preservation capacity. Due to this reason, the world would actually be overproducing food commodities, which is something that may limit the price impact of a locally increased demand, like the one that could be generated by biofuel production.

3.2 Biofuels

As explained before, the feedstocks that are currently involved in biofuel production are products that are normally used in the food industry. In fact, not only grains are a direct food source, but also the oils, such as palm and rapeseed oil or canola oil are widely used products, especially among food producers.

The fact that these feedstocks can be used either as a food or an energy source represents the main reason why biofuel production may affect food prices. The biofuel industry creates a new demand for certain feedstocks that adds up to the food demand that is already present.

Considering cultivation fixed in the short run, if a fraction of the total production is used in the biofuel sector, the remainder, which is sold for feeding purposes, will be scarcer and therefore sold at a higher price. According to Abbott (2009), biofuel production became substantial enough to significantly impact food prices since around 2006. This statement, whether the year is precisely identified or not, gives an idea of the fact that the data span available for estimations of the

impact of biofuel production is limited. This is perhaps the main issue for empirical analysis and also the main reason why estimations made so far have been yielding radically different outcomes while regarding the size of the impact of biofuels in the recent food price increase.

Abandoning the assumption of fixed crop production leads to another way in which biofuels allegedly affect food prices. Allowing flexible production, in fact, makes feedstock supply

responsive to changes in price, so that supply can be extended when facing an increased demand by hypothesis due to biofuel production. This can have two effects: converting cultivation to a crop that saw a price increase and creating new cultivations. The first is the so called

land-18

switching effect and it is the producers’ response to the incentive of a crop that has a higher value on the market, so that their cultivation may be changed for a more profitable crop. So, for

instance, if a famer cultivates wheat and corn price rises because corn is used in biofuel

production, the farmer may want to start cultivating corn. This in turn increases the price of wheat since supply is scarcer. It is the substitution effect that was explained in the previous chapter and that can also be held responsible, at least theoretically, for the diffusion of price increases to crops that are not directly involved in biofuel production. It should be noticed that this effect in particular should be present even for a good part of the non-crop feedstocks that are to be used for producing second generation biofuels4. In fact, many of these feedstocks require land usage nonetheless.

As for the second effect, the idea is that if a crop yields more for its producer, there will be a tendency to increase the supply by increasing the amount of cultivated land. This effect has been present in recent years, because the amount of cultivated land decreased in the previous decades in some areas of the world, especially in Europe (Hamelinck, 2013), so that it was possible to restart cultivating abandoned areas. Nonetheless, the amount of arable land is limited in many areas of the world, so that it is difficult to increase overall production by a significant amount just increasing cultivated acreage. The two effects described convey the idea that an increase in the price of a commodity will increase its production, in the first case at the expenses of another food commodity, in the second case not.

A note should be added with regards to the effect of oil price, which is by itself an influential factor with respect to food prices and which enters new dynamics due to biofuel production,

strengthening the link between the agricultural and the energy sector. Many authors underline the importance of the creation of a new link between food and energy brought by a more extensive biofuel production (Baffes and Haniotis, 2010; Von Braun, 2007; Abbott et al., 2009; Cooke and Robles, 2009; FAO 2008 and others).The reason for this is that gasoline, which depends on oil, and biofuels are substitutes, so whenever the price of oil increases, it becomes more profitable to use biofuel instead. Since production costs for biofuel production are high, these two products can act as substitutes only when the oil price trespasses a certain threshold, underneath which gasoline is the “preferred” product. Beyond that threshold, which is when they act as substitutes, biofuel demand increases, which in turn should have repercussions on the price of the feedstock used for

19

biofuel production. Policy measures increase the complexity of these mechanisms and their effect emerges from the empirics in the papers by De Gorter, Drabik, Timilsina (2013). For instance, a subsidy or a tax exemption should not alter, although it may strengthen it in some cases, the theoretical mechanism that should be expected, which is an increase in the price of biofuel due to an increase in the oil price. Interestingly, the results show that the dynamic inverts when a binding blending mandate is present. This inverse relationship exists because whenever oil becomes pricier, so does gasoline, with the result that the demand for the latter will fall. Along with it, so will the quantity of biofuel that needs to be blended with. This decreases the price of the biofuel and of the feedstock behind its production. The fact that these policies are normally both present at the same time makes the effect on prices ambiguous.

If it is true that biofuel production competes with food demand, it may also be true that the effect may be limited, if not negligible, due to the limited size of the biofuel sector when compared to total food production. As previously mentioned in 3.1.5, total food production is actually exceeding the needs of the whole population (even if due to wastes this has no implications in terms of allocation) so that it may be that in practice biofuel production, even if affecting demand, could have a marginal impact on prices. Moreover, the fact that the bulk of biofuel production is concentrated in countries with a wide arable surface, such as Brazil or the U.S. may limit the effect on price by expanding production capacities. Also, in a paper of the European Commission dating back to 20095_{, it is reported that one of the effects of biofuel consumption has been “the re-use of}

recently abandoned agricultural land, or a reduced rate of land abandonment”. Clearly, depending on the magnitude of this phenomena, this should have the ability of at least curbing the effect of biofuels on food prices.

Regardless of the effect that biofuel production may have on food prices, food prices affect biofuel production, at least in theory. Naturally not in aggregate, but it should true be at the level of single feedstocks. In fact if the price of a feedstock increases, also the cost and the price of the biofuel that is produced with such feedstock will rise. Beyond a level that is determined by the oil price, biofuel loses competitiveness against its rival good, gasoline, so that gasoline will be preferred to biofuel (De Gorter, Drabik, Timilsina, 2013). Also, it may become more profitable to directly sell the feedstock whose price has risen instead of using it for fuel production: this mechanism depend

5 _{3 European Commission, 2009. Commission staff working document. Accompanying document to the renewable}

20

heavily on the kind of feedstock. Although subsidies should allow a certain production level and mandates impose a certain biofuel usage, biofuel production should not be unaffected by a rise in the feedstock price. Also in the case of mandates, fuel distributors have the possibility of “buying-out”, which means paying a fee and stop blending with biofuel in the event that conditions will no longer be favorable (Hamelinck, 2013). Since biofuel producers are able to stop production, it seems evident that feedstock prices (as well as other factors) will indeed affect the amount of biofuel produced. The main example is the one of Brazil, that due to a price increase in sugar decreased bioethanol production (De Gorter et al., 2013), or the U.S., that idled around 12% of its total production in the last years due to the high price of corn (Hamelinck 2013).

Chapter 4: The data and the model

The aim of this paper is to determine whether biofuel production affected food prices. The

approach using aggregate data for food prices and biofuel production seems to encounter intrinsic limitations due to various factors, mainly price interconnections between food commodities and imperfect markets, so that it would be difficult to correctly determine the overall impact of biofuel production.

In order to approach this issue, some authors, most eminently Baier, Clements at al. (2009), estimated the total effect with regard to specific crops and countries. Still, the values that are used strongly rely on the assumptions that are made, in particular because the model used requires hypothesized price elasticities, although they can be changed. On the contrary, the attempt that will be made in this paper will be focusing on the level of production in the biofuel sector of a specific country (US, Brazil) and trying to discover the related effect on the relative crop that is used as a feedstock (corn, sugar), at the international level. Naturally this kind of analysis will not be able to estimate the whole effect on food prices, but it will instead focus on yielding a more precise estimate of the impact of biofuel production on its relative feedstock only. The countries chosen for the analysis are the most economically relevant for these dynamics, so that even if the whole estimation will not comprehend worldwide data, it will nonetheless be the most relevant possible for the global price of the feedstock considered.

The empirical analysis will consist of two separate regressions, one for US corn-ethanol and one for Brazilian sugar-ethanol, and will be comprised of two main steps. The first step will be determining a long run relation between the variables considered, under the hypothesis that

21

cointegration may be present between two variables at least. Of course cointegration will be tested for. Once the long-run equilibrium relation is determined, this relation will be included in a regression having the change in corn price as the dependent variable, in order to model the short-term effects.

Although the main aim is to identify a relation between biofuel production and the relative feedstock price, it is fundamental to understand the link that each variable has with the latter, in order to correctly estimate the impact of biofuel production.

4.1 The data for the main variables

The following analysis will be comprised of two separate regressions. The first one will be made using the bioethanol production in the U.S. as a independent variable and the price of corn in international market as the dependent one. Bioethanol production in the U.S. is 57% of the whole world bioethanol production and the 97% of it comes from corn6, thus US production should be a good approximation for the global corn bioethanol production. In fact, another 32% of world bioethanol production comes from Brazil and is almost entirely made using sugar cane instead. This points out the fact that Brazil is the leader in sugar cane bioethanol production, therefore the second regression will be based on that. Bioethanol production in Brazil will be the independent variable and the price of sugar will be the dependent one. It should be noted that the price of sugar includes both the price of sugar coming from sugar cane and sugar beets. Since sugar beets account for 20% of total sugar production worldwide and main sugar beets producers have also been net sugar importers in recent years7_{, the price of sugar (not only from sugarcane) should still}

be a valid measure. Data for sugar and corn will be the world price reported in the World Bank database, nominally expressed in dollars, whereas data for bioethanol production in the U.S have been retrieved from RFA (Renewable Fuels Association) and in Brazil from UNICA (União da indústria de cana-de-açúcar). Data will be annual, since biofuel production reports made data available in this frequency only, and ranging from 1986 to 2013. The logarithmic values of biofuel production are graphically reported in Figure 7 and 9. It can be noticed how US bioethanol production in Figure 7 seems to present a clear upward trend throughout the whole time-span considered, the same cannot be said for any other variable (Figure 8 to 15).

6 _{Data from 2010, U.S. Energy Information Administration, International Energy Statistics, FAPRI (Food and agricultural}

policy research Institute)

22

The amount of bioethanol produced and the food commodity price will be the main variables of interest. The regressions, due to the model chosen, will include a set of variables, presented in Chapter 4.4, that will be considered independent, along with biofuel production. Although the aim would be just relating biofuel production to the commodity price, it is not possible to do so

without understanding the contribution of other variables to food price variations. Figure 7: US bioethanol production, logarithmic values

Data: Historic U.S. fuel Ethanol Production in Millions of Gallons, Renewable Fuels Association (RFA)

Figure 8: International corn price, logarithmic value

Data: World nominal Corn price in $/mt, World Databank

Figure 9: Brazil bioethanol production, logarithmic value

Data: Brazil Ethanol production in Thousand m3_{, UNICA data}

6 7 8 9 10 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e 4 4,5 5 5,5 6 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e 9 9,2 9,4 9,6 9,810 10,2 10,4 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e

23

Figure 10: International sugar price, logarithmic value

Data: World Sugar price in dollar cents/kg, International Sugar Agreement (ISA) price, annual average, World Databank Originally the idea was running a third regression with respect to the European biofuel production. For this regression, the parameters considered would have been the European Union production of biodiesel which accounts for 54% of the total, and the price of rapeseed, which is the feedstock used in the 79% of the cases there8. Crushing seeds and producing rapeseed oil from rapeseed has a cost, as thoroughly explained in De Gorter, Drabik, Timilsina (2013), which implies a difference between the price of the oil and the one of the feedstock. Since rapeseed oil is the factor used both in the food industry and in biodiesel production, the independent variable in this case would have been the price of rapeseed oil. Unfortunately European production started only in the years 2000s and rapeseed oil quotation is only available from 1989, so that this third regression has been dropped due to lack of observations.

4.2 The choice of the model

The choice of the model is an important element to consider for empirical analysis. The majority of the papers on the biofuel topic use various kinds of approaches, mainly Computable General Equilibrium models, or models created ad hoc (Taheripour et al., 2010, Baier et al., 2009, Gilbert, 2010, Cororato and Timilsina, 2012). The starting point though in many cases is an error correction model (ECM), which gets adapted in light of the logic of the paper. The reason for this choice is that this kind of model allows to estimate multiple cointegration between the variables. The main advantage is that the ECM allows to separate the long run effect from the short run effect, which in this case would represent an advantage. In fact, as previously stated, there are some elements related to biofuel production that should have a non-immediate effect on food prices. The short

8 _{Data from 2010, U.S. Energy Information Administration, International Energy Statistics, FAPRI (Food and agricultural}

policy research Institute) -2,5 -2 -1,5 -1 -0,5 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e

24

run effect should be able to capture the effect of an increased demand on the supply (since in the short run the production is fixed), so the effect on the price generated by the repartition of

feedstocks between the biofuel sector and the food commodity sector. The long run effect instead takes into account elements that come from the adaptation to previous market conditions, such as the land-switching effect which means an increased supply (or reduced) of a feedstock. Since for biofuel production only annual data are present, having lags of one year may include part of the long-run effects in the short-term effects of the model, possibly overestimating the short-run component. The data span used will be ranging from 1986 to 2012, where biofuel production should be economically relevant at least at the turn of the century (2006, according to Abbott, 2009). This means that the data span is very limited, so using a model that includes a long-run effect may not be optimal, but this is a shortcoming that cannot easily find a solution. For these reasons the ECM seems to be fitting solution and it will be used with a vector component (VECM, Vector error correction model) in order to allow the presence of more than one cointegrating equation. The outcome of the VECM is the long-run equations on the one hand, and K equations, where K is the number of variables, that correct the long-run equation by their short-term adjustments. Each one of these equations will model the short-term adjustments having the variation of each variable as the dependent variable. It is implied in the VECM that each variable depends on the others, which in the case in analysis is something that cannot be accepted for the short-term equations due to the fact that some variables need to be exogenous and therefore cannot be regressed with respect to the others. For this reason, in this paper the VECM will be used to determine the long-run relation, which should not be affected by this limitation. Since the aim of the paper is to determine the impact of biofuel production on food prices, the specification of the equation having the corn/sugar price variation as the dependent variable will be taken into account to determine the short run effects, which will be estimated by OLS

4.3 The model

The VEC (Vector Error Correction) model is a restricted VAR (Vector Auto-regression) model that allows for the estimation of the short-term effects that prevent the full convergence to the long-run co-integrating relationship. A VAR(p) model with two variables, 𝑌𝑌𝑡𝑡 and 𝑋𝑋𝑡𝑡 ,consists of two

equations, for simplicity it will be a VAR(1), even though two lags should be included in the full model:

25

𝑋𝑋𝑡𝑡 = 𝛿𝛿20+ 𝛿𝛿21𝑌𝑌𝑡𝑡−1+ 𝛾𝛾21𝑋𝑋𝑡𝑡−1+ 𝑢𝑢2𝑡𝑡 (a.2)

If 𝑌𝑌𝑡𝑡 and 𝑋𝑋𝑡𝑡 in the analysis are both stochastic and their first difference is stationary, which can be

tested with an ADF statistic (Augmented Dickey-Fuller), they can be co-integrated by a coefficient 𝛽𝛽, so that 𝑌𝑌𝑡𝑡= 𝛽𝛽𝑋𝑋𝑡𝑡 , which is the long-run relation. So, augmenting the VAR model with the first

difference 𝑌𝑌𝑡𝑡−1= 𝛽𝛽𝑋𝑋𝑡𝑡−1 yields the VEC model.

∆𝑌𝑌𝑡𝑡= 𝛿𝛿10+ 𝛿𝛿11∆𝑌𝑌𝑡𝑡−1+ 𝛾𝛾11∆𝑋𝑋𝑡𝑡−1+ 𝛼𝛼1(𝑌𝑌𝑡𝑡−1− 𝛽𝛽𝑋𝑋𝑡𝑡−1) + 𝑢𝑢1𝑡𝑡 (b.1)

∆𝑋𝑋𝑡𝑡 = 𝛿𝛿20+ 𝛿𝛿21∆𝑌𝑌𝑡𝑡−1+ 𝛾𝛾21∆𝑋𝑋𝑡𝑡−1+ 𝛼𝛼2(𝑌𝑌𝑡𝑡−1− 𝛽𝛽𝑋𝑋𝑡𝑡−1) + 𝑢𝑢2𝑡𝑡 (b.2)

The coefficient 𝛽𝛽 is the one determining the long-term relation between the variables, whereas 𝛼𝛼𝑖𝑖

determines the 𝑖𝑖𝑡𝑡ℎ _{variable’s short-term speed of adjustment. So, 𝛼𝛼}

𝑖𝑖(𝑌𝑌𝑡𝑡−1− 𝛽𝛽𝑋𝑋𝑡𝑡−1) models the

short-term effects. The representation of the model in a vectorial form would be:

∆𝑦𝑦

= П𝑦𝑦

+ 𝜀𝜀

𝑡𝑡� in the example. In the regression K variables will be considered, so that the

vector 𝑦𝑦𝑡𝑡 , which is a 𝐾𝐾 × 1 vector, will be comprised of 𝐾𝐾 variables, cointegrated by 𝑟𝑟 coefficients.

The rank, 𝑟𝑟, determines the number of cointegrating equations that link these variables and it will be chosen running a Johansen cointegration test for each regression.

Also, П = �𝛼𝛼_𝛼𝛼1− 𝛽𝛽𝛼𝛼1

2− 𝛽𝛽𝛼𝛼2� which can also be rewritten as П = 𝛼𝛼𝛽𝛽′, where both α and 𝛽𝛽 are 𝑟𝑟 × 𝐾𝐾

matrices of rank K, where r expresses the rank, or the number of number of linearly independent cointegrating vectors. Therefore the formal specification of the model will be

∆𝑦𝑦

= 𝛼𝛼𝛽𝛽

𝑦𝑦

+ ∑

𝛤𝛤

∆𝑦𝑦

+ 𝑣𝑣 + 𝛿𝛿𝛿𝛿 + 𝜀𝜀

Where 𝑣𝑣, 𝛿𝛿 and 𝜀𝜀𝑡𝑡 are 𝐾𝐾 × 1 vectors that respectively indicate the constant term, the trend and

the disturbance. The number of lags to be included, p, should be p=2 lags, also accordingly with fitting lag-selection criteria. To sum up, beta coefficients define the long term cointegration between variables, where the main variable will be the corn price in the first regression and sugar in the second, with 𝛽𝛽 = 1. The alfa coefficients define the speed of adjustment in the short run, which is the error that the long-run cointegration is corrected for, in the short-run.

26

The identification of the coefficients will allow to understand the kind of interaction between the different variables and the variable for which 𝛽𝛽 = 1.

4.4 The other variables

The variables are to be chosen from the previous reasoning in Chapter 3 and, jointly, should be able to explain the movements of the food commodity price, along with ethanol production. The approximation for the oil price will be made using the annual values of crude oil retrieved in the World Bank database, which is an average of the three main market quotations (Dubai crude, Brent, West Texas Intermediate), equally weighted. A trend is identifiable only at the turn of the century, judging by the graph in Figure 11.

Figure 11: Average nominal oil price in logarithmic value

Data: Average crude oil price in $/bbl, World Databank

The speculation topic is a crucial one: the variable that should include its effect will be the

exchange rate . The idea of including a measure for the open interests in corn derivatives will not be implemented in this paper mainly because it cannot fully capture the effect of speculation and, if used along with the exchange rate, the impact should be relatively small and partly at the expenses of the exchange rate. For the non-commercial position, suggested by Gilbert (2010), the theoretical idea seems solid, but the findings of Gilbert (2010) and also Cooke and Robles (2009) suggest that the impact would also be minimal. For sugar, the option of considering the amount of derivatives traded was not considered because sugar derivatives are not traded on a single

market, so that a reliable measure was not available. Thus only the variable exchange rate should be the one including the effect of speculation in this model.

As it was previously said, most of agricultural products’ prices are denominated in dollars, which is the reason why the exchange rate for both U.S. and Brazil should be with respect to the dollar

2 2,5 3 3,5 4 4,5 5 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 lo g v alu e

27

against a basket of different currencies. A different measure has been used to estimate the Exchange rate in the two regressions. Brazil is the main sugar producer and the extra demand for sugar coming from ethanol production also comes from Brazil, so that sugar for this purpose is paid in reais. On the other side, Brazil sells sugar in dollars on the international market, which is a foreign currency, therefore the exchange rate of interest should be nominal. Thus, annual values of the Exchange rate of the dollar against a weight-traded basket of currencies retrieved from FRED Economic data (Federal Reserve Bank of St. Louis) has been chosen. The same but opposite reasoning applies for the US corn-ethanol regression. The US is the main corn exporter and corn demand for ethanol production comes from the US. The international price of corn is

denominated in dollars, which for the US is the home currency. Consequently, a real measure for the exchange rate should be the one affecting the international price of corn. For the corn-US ethanol regression the value included is an index of the Real effective exchange rate (2005=100) taken from the World Bank database.

Figure 12 : Nominal exchange rate in logarithmic values

Data: Trade Weighted U.S. Dollar Index: Major Currencies, Index March 1973=100 Source: Board of Governors of the Federal Reserve System

Figure 13: Real effective Exchange rate, logarithmic values

Real effective exchange rate index for U.S dollar (2005=100), World Databank 4,2 4,3 4,4 4,5 4,6 4,7 4,8 19 85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 lo g v alu e 4,5 4,55 4,6 4,65 4,7 4,75 4,8 4,85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 lo g v alu e

28

As for the stocks, the dimension that should be interesting for the international price should be a global one, so world reserves of feedstocks have been considered. To do so, a stock-to-use ratio has been used, built as:

�𝐼𝐼𝐼𝐼𝑖𝑖𝛿𝛿𝑖𝑖𝐼𝐼𝐼𝐼 𝑠𝑠𝛿𝛿𝑠𝑠𝑠𝑠𝑠𝑠 + 𝑝𝑝𝑟𝑟𝑠𝑠𝑝𝑝𝑢𝑢𝑠𝑠𝛿𝛿𝑖𝑖𝑠𝑠𝐼𝐼 − 𝛿𝛿𝑠𝑠𝛿𝛿𝐼𝐼𝐼𝐼 𝑢𝑢𝑠𝑠𝑢𝑢 _{𝛿𝛿𝑠𝑠𝛿𝛿𝐼𝐼𝐼𝐼 𝑢𝑢𝑠𝑠𝑢𝑢} � × 100

The stock-to-use ratio may be a rough measure, but it seems to be able to be explicative while discussing the price level. In fact in this way it is possible to include data for world production and consumption in a single indicator. In fact not only the level of stocks per se is considered but also the demand-side factors in play since total use is included, and the supply shocks that may affect total production. This measure has been calculated by using marketing based estimation available at the USDA Foreign Agricultural Service. The total use takes into account consumption, industrial use and total disappearances, so that the stock-to-use ratio should take into account quantities effectively available, which is what should be influential while determining the price. It can noticed how the corn ratio in Figure 14 sees a drop around the year 2000, in line with the statement of Abbott (2011) reported in Chapter 3.1.5, although the author referred to food commodities in general. The ratio for sugar instead appears to be alternating minor rises and plunges (Figure 15).

Figure 14: Corn Stock-To-Use ratio, logarithmic values

Data: STU ratios obtained from data from PSD Online, USDA Foreign Agricultural Service 2,5 2,7 2,9 3,1 3,3 3,5 3,7 3,9 4,1 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e

29

Figure 15: Sugar Stock-To-Use ratio, logarithmic values

Data: STU ratios obtained from data from PSD Online, USDA Foreign Agricultural Service

Finally an additional correction has been introduced for the Brazil-Sugar regression that would take into account the effect of weather events. As emerged from the previous reasoning in

chapter 3.1.2, production concentration has been the decisive factor for including such a measure in this regression and not in the corn-U.S. regression. In a first attempt the idea was adding a measure for the total production of sugar in Brazil, so that shocks due to exogenous events, which are supposed to be very influential, are taken into account. Since sugar production and bioethanol production are competing though, this measure seemed to lead to collinearity, as it was reported by the error correlation test. Therefore the measure adopted will be the amount of sugarcane produced in Brazil. Data for sugar, sugarcane and ethanol in Brazil were retrieved from UNICAdata. In the regressions all the variables are measured by their logarithmic value.

4.5 The Results

In order to run the regressions there are some preliminary steps to take. The first would be checking that the variables considered are stochastic and their first differences stationary. To do so, an Augmented Dickey-Fuller statistic has been used, the results can be checked in Appendix 2. Therefore, a test was run in order to determine the number of lags to include in the regression. Theoretically, the Akaike Information criteria (AIC) should be the one to prefer when operating with a VEC model, since it does not consider the total number of observations and should yield the most fitting number of lags for the coefficient estimation. The reduced number of observation though, which is the main limitation of this model, seems to indicate that the use of the Bayesian information Criteria (BIC) would be a better choice, since it seems to better apply to limited

datasets. Also, since there is a total of 27 observation for the datasets of both regressions, which is already a limited number to determine a long-run coefficient, a maximum of two lags has been set

2,5 2,7 2,9 3,1 3,3 3,5 3,7 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 Lo g v alu e

30

for the model. Since there was no trend that could have been considered for all the variables and for the whole data span, the trend, which is needed for the model, has been considered constant.

4.5.1 The Corn-Ethanol regression

For this first regression, 𝐾𝐾 = 5 variables were included, a Johansen test for cointegration was run (reported in Appendix 2) and yielded 2 cointegrating coefficients (𝑟𝑟 = 2), which leads to the same number of cointegrating equations. In line with this result, the first cointegrating equation was set so that the 𝛽𝛽 coefficient for corn price was normalized to one, whereas for the second the amount of bioethanol produced was set with 𝛽𝛽 = 1. The reason for this is that it seems plausible that these two variables can be made to depend on the others. The beta coefficients indicate the long-run relation between the variables and the values are reported in Table 1.

Table 1: Beta coefficients of cointegrating equation 1, corn-ethanol regression

_ce1 Coefficient Standard Error P-value

Corn price 1 - - US Ethanol 0 - - Stock-To-Use -.3476337 .2517488 0.167 Oil price -.6175837 .114194 0.000 Exchange rate -.6593572 .6530728 0.313 Constant -.3639872 - -

Table 2: Beta coefficients of cointegrating equation 2, corn-ethanol regression

_ce2 Coefficient Standard Error P-value

Corn price 2.78e-16 - -

US Ethanol 1 - -

Stock-To-Use -.2611479 .4592609 0.570

Oil price -1.684028 .2083221 0.000

Exchange rate -2.563672 1.191389 0.031

Constant 9.453534 - -

The beta coefficients reported in Table 1 determine the long run relation between the variables. Let us consider the cointegrating equations before proceeding. Starting from the first