How policies affect international biofuel price linkages

How Policies Affect International Biofuel Price Linkages1

Miroslava Rajcaniova Slovak Agricultural University, Slovakia [email protected]

Dusan Drabik Charles H. Dyson School of Applied Economics and Management Cornell University, USA [email protected] and

Pavel Ciaian European Commission (DG Joint Research Centre), Spain [email protected]

This version: October 16, 2011. This is an updated and extended version of the paper entitled “Interlinkages among International Biofuel Prices: The Role of Biofuel Policies” that was earlier posted on AgEcon Search as a selected paper for the 2011 AAEA annual meeting.

1

We are grateful for helpful comments received in the 2011 AAEA and EAAE meetings in Pittsburgh, PA and Zurich, Switzerland, respectively. We also acknowledge financial support from the Slovak Ministry of Education under projects APVV-0706-07 and VEGA1/3765/06. The authors are solely responsible for the content of the paper. The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission.

1 Electronic copy available at: http://ssrn.com/abstract=1945022

International Biofuel Price Interlinkages: The Role of Biofuel Policies

Abstract We estimate the role of biofuel policies in determining which country is the biofuel price leader in world markets using a cointegration analysis and the Vector Error Correction (VEC) model. Weekly prices are analyzed for the EU, U.S, and Brazilian ethanol and biodiesel markets for the 2002-2010 and 2005-2010 time periods, respectively. The U.S. blender’s tax credit and consumer tax exemption in Brazil appear to play a role in determining the ethanol prices in other countries. For biodiesel, our results demonstrate that the EU policies – consumer tax exemption and blend target – tend to form the world biodiesel price.

Key words: biofuels, biofuel polices, mandate, tax credit, tax exemption, price leadership, VEC JEL: C32, Q16, Q17, Q47

2 Electronic copy available at: http://ssrn.com/abstract=1945022

1. Introduction Noticeable changes in world markets for biofuels and their feedstocks have occurred since ambitious biofuel consumption targets were established in the United States, the European Union, and elsewhere in the last half decade. Several policies were introduced to incentivize biofuel production in order to meet the targets, including mandates, consumption subsidies, and various production subsidies. As a result, the link between prices of biofuels and feedstocks has strengthened, thereby affecting food prices. Meanwhile, international trade in biofuels has intensified significantly. The link between biofuel and feedstock prices has been analyzed, for example, by de Gorter and Just (2008); Lapan and Moschini (2009); and Mallory et al. (2010), where an analytical formula that theoretically explains the link between ethanol and corn prices in the United States is derived. On the other hand, recent studies by Balcombe and Rapsomanikis (2008); Busse et al. (2010); Ciaian and Kancs (2011); Serra et al. (2011); or Zhang et al. (2010) – using econometric techniques – empirically study the nexus between fuel, biofuel, and feedstock (food) prices. Another strand of literature – for example, Abbott et al. (2009); Wright (2011); Yano et al. (2010); and Hertel and Beckman (2011) – attempts to explain the link between biofuel policies and oil prices on the one hand, and food price levels and volatility on the other. The intensifying trade in biofuels has arisen because of different levels of biofuel blend requirements together with heterogeneous biofuel policies in biofuel producing countries. For example, because of high sugar cane prices, Brazil has recently switched to being a major importer and curtailed its exports substantially; thus, making it possible for the United States to supply the markets that used to import ethanol from Brazil. This turned the U.S. from an importer to an exporter of ethanol, and the U.S. exports are even expected to surpass Brazil in 2011 (Reuters, 2011). Another example of an international interaction of biofuel

3

policies were the “splash and dash” exports of U.S. (and non-U.S.) biodiesel to the European Union before mid-2009 (de Gorter et al., 2011). A key question pertaining to all of the three observed impacts of biofuel policies discussed above is how a biofuel price is determined. In analyzing the U.S. biofuel blend mandate and blender’s tax credit in a closed economy framework, de Gorter and Just (2009) argue that the tax credit is not additive with a biofuel price premium due to the mandate. This implies that only one biofuel policy at a time determines the biofuel price. But what biofuel policy and which country determine the world biofuel price when international trade is considered? Although the answer to this question has important implications for an analysis of biofuel policies, we are only aware of two such studies: Kliauga et al. (2008) and de Gorter et al. (2011). The former finds that the U.S. tax credit for ethanol determined ethanol market prices in the United States and Brazil; the latter concludes that the European Union is the price leader for biodiesel, with the market prices initially determined by an EU tax exemption and later by the biofuel targets. A drawback of the foregoing studies is that they base their empirical conclusions only on simple differences of the observed and predicted market prices, albeit a supporting theory is provided for those differences. In this paper, we use time series econometric techniques (variance decomposition and cointegration analysis)2 to answer two questions: which country is the price leader for the world biofuel (ethanol, biodiesel) price, and which biofuel policy (blender’s tax credit, tax exemption, or biofuel mandate) determines the world market price. Unlike Kliauga et al. (2008) who analyzed the market with only two countries, we consider the United States, Brazil, and the European Union. Owing to data availability, however, we only analyze biodiesel

2

Cointegration analysis and the vector error correction model that we use in our paper are well established in the literature on energy markets. For example, price interdependencies at the global level as well as among different types of energy markets – such as oil, gasoline, or natural gas – are analyzed in Serletis and Herbert (1999); Asche et al. (2001); Paul et al. (2001); Asche et al. (2003); Siliverstovs et al. (2005); or Cuddington and Wang (2006).

4

prices in the European Union (proxied by Germany) and the United States. For both biofuels, we investigate a longer time period, as well as data with a higher frequency (weekly) than has been the case so far. The procedures to identify the price leader and the price-determining biofuel policy differ from the previous studies; this makes it possible to check the robustness of these earlier studies. We find that over the period analyzed in our study, 2002-2010, ethanol price was codetermined by the United States and Brazil, mostly via the U.S. tax credit or Brazilian tax exemption. This result differs from that in Kliauga et al. (2008) to the extent that they hypothesize that only the United States determined the ethanol price. For biodiesel, in the period 2005-2010, we find that the European Union establishes the world price (consistent with de Gorter et al., 2011). Although for biodiesel – relative to ethanol – it is difficult to determine convincingly a policy dictating the biodiesel price, the results suggest the EU biofuel target. The remainder of the paper is organized as follows. In the next section, we briefly describe biofuel policies in the relevant countries under study. In Section 3, we discuss existing theory of biofuel price formation and formulate three hypotheses to empirically test this theory. Section 4 describes the data used, and Section 5 presents the econometric techniques followed. Empirical results are presented in Section 6. Finally, Section 7 provides concluding remarks. 2. Biofuel Policies There is a variety of policies that can directly or indirectly affect either biofuel production or consumption. The first category includes blenders’ tax credits, tax exemptions, mandates, and production subsidies targeted at biofuel production; the second group consists of policies such as import tariffs, production subsidies on biofuel feedstocks, or research and development subsidies. In our study, we focus on the first three policies listed: blenders’ tax credits, tax exemptions, and mandates. Furthermore, we only describe the policies used in the coun-

5

tries covered in the paper: the United States, the European Union, and Brazil.3 A blender’s tax credit is a subsidy on biofuel consumption, thereby benefiting fuel consumers. The tax credit is administered by fuel blenders who lower the price of the final fuel (mixture of a biofuel and a fossil fuel) by an amount of the tax credit proportional to the share of a biofuel in the fuel blend. The federal blender’s tax credit for ethanol in the United States currently amounts to 45 cents per gallon; an additional average state subsidy, which comes mostly in the form of a tax credit, is 7 cents per gallon. Biodiesel blenders enjoy a tax credit of $1 per gallon of biodiesel blended with regular diesel (it was temporarily suspended in 2010, but reenacted retroactively in December of 2010). A tax exemption in the European Union and Brazil represents a reduction in the fuel excise tax collected at the pump level. The economic impacts of a blender’s tax credit and tax exemption are identical in a closed economy – both constitute a biofuel consumption subsidy, but differ substantially in an open economy framework.4 The level of the tax exemption varies across EU countries and between biofuels, but it is declining as governments wish to recoup fiscal revenues from fuel taxes that were foregone because of exemptions. For example, a tax exemption for biodiesel in Germany declined from €0.47 per liter to €0.29 per liter between 2005 and 2009. For Brazil, Kliauga et al. (2008) report a (consumption weighted) average tax exemption of $0.67 per liter, which is approximately 2.7 times the U.S. tax credit. A biofuel mandate, another widely used biofuel policy, is often used in combination with either a blender’s tax credit, as it is the case in the United States, or a tax exemption, as it happens in the European Union and Brazil. Although the U.S. biofuel mandate is set as a 3

According to the data from the U.S. Energy Information Administration, in 2009, the United States, Brazil, and the EU27 combined represented 92 and 93 percent of the world ethanol production and consumption, respectively. In the same year, the share of the EU and U.S. in world’s biodiesel production and consumption amounted to 67 and consumption 75 percent, respectively (http://www.eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=79&pid=80&aid=1&cid=regions,&syid=2005&eyid =2009&unit=TBPD). 4 The reason is that once the world market price of a biofuel is established by one country (A), a tax credit or a tax exemption in the other country (B) cannot affect it, but acts as a production subsidy in the case of a tax credit and a fuel consumption subsidy with a tax exemption (Kliauga et al. (2008); de Gorter et al. (2011).

6

consumption mandate, that is as a quantitative requirement, in practice the U.S. Environmental Protection Agency implements it as a blend mandate by annually specifying a minimum of volumetric percentage of a biofuel in the final fuel mix (Tyner, 2010). The European Union uses the latter form of a biofuel mandate, requiring that a pre-specified share of energy content of the fuel be a biofuel. For instance, the blend equivalent of the U.S. ethanol consumption mandate has been set to 7.95 percent in 2011 (Reuters 2010). For comparison, a mandatory 10 percent minimum target is set in the European Union for the share of biofuels in transport fuel consumption by 2020 (Directive 2009/28/EC). In Brazil, 25 percent of gasoline fuel consumption has to come from ethanol (known as the E25 fuel). 3. The Theory of Biofuel Price Formation and the Hypotheses Tested Autarky Theoretical studies show that biofuel polices are a key factor determining the biofuel price. For example, de Gorter and Just (2009) developed a theoretical model that explains the link between the U.S. biofuel policies – a blender’s tax credit, a biofuel mandate, or their combination – and the biofuel price. In their model, biofuel (ethanol) and fossil fuel (gasoline) are assumed to be perfect substitutes and differ only in miles traveled per gallon of either fuel. They conclude that the price of the biofuel is determined either by a tax credit or a binding (consumption or blend) mandate but never by both at a time. Consider first a case where a blender’s tax credit on ethanol, tc, (or a tax exemption) is the binding biofuel policy. Then the ethanol market price, P E, is directly linked to the gasoline (oil) price, P G, and the fuel consumption tax, t (de Gorter and Just, 2008): PE   PG  (1   )t  tc

(1)

where the coefficient measures miles traveled per gallon of ethanol relative to a gallon of gasoline.5 An implication of equation (1) is that if the fuel tax and the tax credit do not

5

For corn ethanol, ≈ 0.70, while for biodiesel it is somewhat higher, ≈ 0.90.

7

change over time (as it has been the case for the countries included in our study, at least for some time), then the volatility in the world oil price should be transmitted into the ethanol market price. Alternatively, assume there is only a biofuel mandate that dictates a certain amount of biofuel to be blended. Although the economics of a consumption mandate differs somewhat from that of a blend mandate (de Gorter and Just 2009), a common outcome of the two forms of the mandate is that – unlike with a blender’s tax credit – the biofuel price is not directly linked to the gasoline (oil) price: the link is completely severed under a consumption mandate (because the biofuel market price is determined by the intersection of the ethanol supply curve and a fixed mandate level), and it is partially severed under a blend mandate insofar as a change in the gasoline (oil) price affects the fuel demand. The intuition behind this result is that the biofuel price is determined more by the biofuel supply than by the gasoline price for a given mandate quantity. In the case of a blend mandate, only with an inelastic biofuel supply is the biofuel price be tightly linked to a gasoline price (see the appendix and table 1). Finally, suppose (as it is usually the case in reality) that a blender’s tax credit (a tax exemption) is combined with a biofuel mandate. For a mandate to bind, the difference between the ethanol market price under the mandate and the gasoline market price without the mandate has to be greater than the amount of a tax credit – otherwise the relationship (1) holds. International Trade International trade in biofuels makes the forgoing discussion of the biofuel price formation more complex. The question of interest in our paper is to find out empirically how the biofuel price is established when countries trade in biofuels. Kliauga et al. (2008) and de Gorter et al. (2011) argue that only one country’s policy and market situation determines the world biofuel price; hence either of the following situations (but never both) should hold:

8

(i)

The world biofuel market price is linked to the gasoline/diesel (oil) price through a tax credit (or tax exemption), and is determined in the country with a combination of the highest consumer price paid for gasoline (diesel) and the highest net subsidy (the combination of the lowest fuel tax and highest biofuel tax credit/tax exemption).

(ii)

The world biofuel market price is determined by a binding mandate if the induced biofuel price is higher than under the case (i).

The general implication of the relationships (i) and (ii) is that the impact of biofuel policies (tax exemptions, tax credits, or price premia due to biofuel mandates) on biofuel prices are not additive: the market price of a biofuel is not determined by the sum of each country’s tax exemption, tax credit, or a mandate price premium. On the basis of the theoretical predictions (i) – (ii), we formulate three empirically testable hypotheses: Hypothesis 1: A country with the highest biofuel price is the price leader for other countries. The price leader country sets the world biofuel price if its biofuel policy (tax credit/tax exemption or mandate) generates a price that is higher than in other countries. The arbitrage in biofuels will lead to equalization of prices across countries, up to transportation costs and tariffs. Transportation costs and tariffs imposed by the price leader, if any, lower the biofuel price in non-price leading countries. These costs do not, however, affect the direction of causation of the price relationships between countries; they may only weaken these relationships.6 For example, if the United States is the price leader for ethanol, then ethanol prices in other countries are likely to be lower by the sum of transportation costs and tariffs, or may be independent of the U.S. price if the transportation costs or tariffs are prohibitive.7 We can identify the biofuel price leader empirically by testing price interdependencies between countries. Fossil fuel and/or biofuel prices in the price leading country are expected to cause bio-

6

In principle, export subsidies might invert the causality of prices summarized in (i) and (ii). But this type of trade policy was not applied during the period analyzed in our paper. 7 This is because – unlike a traditional textbook analysis – biofuel prices are linked forward to the oil price, and ethanol demands are not added to find the world price.

9

fuel prices in other countries. Hypothesis 2: If the fossil fuel prices determine biofuel prices at least in the price leading country, then the relationship (i) holds; that is, the tax credit (or tax exemption) determines biofuel prices.8 The hypothesis 2 says that if a fossil fuel is an important determinant of biofuel prices, at least in the price leading country, then the tax credit/tax exemption in the price leading country determines the biofuel prices. This is because the net subsidy on a biofuel in the price leading country provides the most favorable biofuel price at the world level; this price is expected to be followed by other countries. The net subsidy creates a gap in biofuel prices between price leading country and other countries, thus creating opportunity for trade. The adjustment in the level of the biofuel price will be driven by changes in the fossil fuel price, however, because the net subsidy does not typically change significantly over a longer period, whereas fossil fuel price adjusts almost daily. Consider the price relationship in equation (1). Totally differentiating equation (1) yields dPE dPG   ; this implies that the relationship between fossil fuel and biofuel prices in the price leading country is linear and is determined only by the conversion coefficient  . Since  is approximately equal to 0.7 for ethanol and 0.9 for biodiesel, the price causation from the fossil fuel to biofuels is expected to be strong if indeed the hypothesis 2 holds. If so, then our econometric estimations are expected to show that the fossil fuel price determines biofuel price in the price leading country, whereas either the fossil fuel or biofuel prices (or both) in the price leading country determine biofuel prices internationally. Hypothesis 3: If a biofuel price of a country determines biofuel prices in other countries,

8

Given that a tax credit/tax exemption (and also a fuel tax) tends to be fixed over a longer period, the varying fossil fuel prices in the price leading country determine the biofuel prices in other countries.

10

then the mandate determines biofuel prices.9 Hypothesis 3 says that the mandate implemented in the price leading country determines the world biofuel prices. With a binding mandate, the biofuel price tends to be isolated from the fossil fuel market; hence, it will not respond to fossil fuel price changes very much. To illustrate this, we use the model in de Gorter and Just (2009) (see the appendix) and simulate a price response of ethanol to changes in the gasoline price in the price setting country (table 1)10. The results show that the price response of ethanol is negative and very small, between -0.1 and -0.01; hence, it is relatively immune from the gasoline price. Internationally, the ethanol price of the price leading country will determine ethanol prices in other countries if indeed mandate is the policy that determines the ethanol prices worldwide. In summary, if the hypothesis 3 holds, we should observe that the relationship between the fossil fuel price and biofuel price is statistically not significant, or negative in the price setting country, whereas the price transmission among countries will be effected through biofuels: the ethanol (biodiesel) price of the price leader will influence ethanol (biodiesel) prices in other countries. The key difference between the hypotheses 2 and 3 is that the former implies significant interdependencies between biofuel and fossil fuel prices as compared to the latter, which states that the fossil fuel price is not driving biofuel prices because the mandate tends to isolate the biofuel and fossil fuel markets from each other. Note that the hypothesis 3 is a subset of hypothesis 2 and, therefore, it cannot be identified if both policies interchangeably determine biofuel prices over the analyzed period.11 Empirically, we can identify the hypothesis 3

9

This follows from the fact that if the mandate determines biofuel price in the price leading country, then the biofuel and fuel markets are isolated from each other. In this case, the fossil fuel price will be independent of the biofuel price. 10 In our simulations, we assume an exogenous gasoline price. 11 If the hypothesis 1 holds, it implies that biofuel and fosil fuel prices are correlated in price leading country; further, they are correlated with biofuel and fosil fuel prices in other countries. Under the hypothesis 3, biofuel and fossil fuel prices are not correlated in the price leading country, nor in other countries. Only biofuel prices among countries are correlated.

11

only if the mandate is a dominant policy. 4. Data Our data consist of weekly price observations for oil (January, 2002 to December, 2010), ethanol and gasoline (January, 2002 to December, 2010), and biodiesel and diesel (June, 2005 to December, 2010) for the European Union, the United States, and Brazil. All EU data are proxied by German prices extracted from the Bloomberg database (ethanol), UFOP12 (biodiesel), and the EU Commission’s Oil bulletin (gasoline and diesel). The U.S. Gulf ethanol and biodiesel prices come from the Bloomberg database, while the U.S. Gulf gasoline and diesel prices are from the U.S. Energy Information Administration. The oil prices are NEMEX futures prices extracted from the U.S. Energy Information Administration. Finally, Brazilian ethanol and gasoline prices are for Sao Paolo (the biggest ethanol producing state in Brazil) and come from the Center for Advanced Studies in Applied Economics and the Brazilian National Agency of Petroleum, Natural Gas, and Biofuels, respectively. 5. Cointegration Because biofuel and fossil fuel prices are interdependent, applying a standard regression approach would violate the exogeneity assumption about regressors. A general approach to analyze interdependences between endogenous variables is the Vector Autoregressive (VAR) model whereby the causality between the current and past values of the variables is examined. The standard requirement for the VAR estimation is stationarity of the time series. But even if the individual time series are not stationary, a combination of two non-stationary time series may be stationary (Engle and Granger, 1987). In this special case, the time series are said to be cointegrated; that is, there exists a long-run equilibrium relationship between them, and a Vector Error Correction (VEC) model – that adds error correction features to the VAR model – can be estimated.

12

The Union for Promotion of Oil and Protein Plants.

12

To test for stationarity of time series, we use two unit root tests: the augmented Dickey-Fuller (ADF) test and the Phillips-Perron (PP) test. The number of lags of a dependent variable is determined by the Akaike Information Criterion (AIC). If the time series are not stationary, we employ the Johansen's cointegration method to examine the long-term relationship between the price series. This method allows us to test for cointegration of several time series and does not require that they be of the same order of integration. The number of cointegrating vectors is determined by the lambda max test and a trace test. Both tests use eigenvalues to compute associated test statistics. The null hypothesis of the test statistics is the existence of at most r cointegrating vectors. We follow the Pantula principle to determine whether the time trend and the constant term should be included in the model. We first perform a bivariate Johansen cointegration test on the pairs of prices. With the patterns obtained from the bivariate case, we test for multivariate cointegration. The bivariate case ignores a possible integration of two markets through a third market; that is, there might exist a long-run cointegration relationship that ties several markets together, whereas such a relationship is not found between two markets alone (Harris, 1995) Then, we estimate a VAR model for the cointegrated variables in which we include a mechanism of the error correction model. We use the AIC and the likelihood ratio (LR) tests to determine the number of lags in the VEC model. In the event that the two tests yield different results, we consider each possibility and first follow the AIC. We test the adequacy of our VEC model by a series of tests: the Lagrange multiplier test for autocorrelation in the residuals; the Jarque-Berra test for normality in residuals; and the stability test of the VEC model estimates. A possible cointegration relationship does not automatically imply a bidirectional causal relationship among the price series. Causality tests show whether or not a country is the price leader and which countries are price followers (or it can well be that none of the

13

countries dominates the others) (Ciaian and Kancs, 2011). If two variables are cointegrated, the Granger causality in at least one direction must exist. This Granger causality can be detected through the VEC model derived from the long run cointegrating vectors. Statistical significance of the differenced explanatory variables provides information about the short run causal relationships between the variables, while the significance of the lagged error correction term explains the long run causal relationships. However, Granger causality detected through F-tests and t-tests of the VEC model may identify the Granger causality only within the sample period. We therefore employ the Variance Decomposition technique in order to measure the effect of shocks to each price on the current and future values of the own and other prices. We perform a decomposition of the variance for each price in the VEC model; the variance is caused by shocks to the other prices after 1 to 48 weeks. By this, we examine the price relationships summarized by the hypothesis 1; that is, whose country prices cause biofuel prices in other countries. 6. Empirical Results The correlation coefficients for various price pairs for ethanol are reported in Table 2. The results confirm a high and positive correlation (0.717) between ethanol prices in Brazil and Europe. There is also a positive correlation between the EU and U.S. ethanol prices (0.649), as well as between the United States and Brazil (0.614). The correlation between biodiesel prices in the EU and U.S. markets is even stronger (0.969) (Table 3). While the correlation analysis is informative, it is not sufficient to draw conclusions about the relationship between the prices. Hence, we first analyze the characteristics of the time series. The use of non-stationary time series could lead to statistically significant results due to a spurious regression. The Dickey-Fuller and Phillips-Perron tests confirm that all our time series are non-stationary; we stationarized them by taking first differences. In order to test for interdependencies among prices, we examine whether there exist a

14

cointegrating vector among fossil fuel and biofuel prices. The results show that gasoline and ethanol prices are cointegrated, except for two price pairs: EU – Brazilian ethanol price and EU ethanol – Brazilian gasoline price (Table 4). Both the trace and the likelihood ratio tests reject the absence of cointegration relationship between the EU and U.S. biodiesel prices, and the EU and U.S. diesel prices at the one percent significance level. The rest of dieselbiodiesel price series is found not to be cointegrated. The results of the multivariate Johansen cointegration test show that the series under consideration are cointegrated of rank 3 in the case of gasoline and ethanol, and of rank 1 for the biodiesel and diesel prices. The cointegration analysis shows that fossil fuel and biofuel prices are interlinked. However, the cointegration analysis cannot predict the direction of causality between the price series. To identify the causality relationships, we estimate the VEC model. Variance Decomposition We estimate two VEC models. In the first model, we include domestic gasoline (diesel) prices and biofuel prices for the analyzed countries. In the second model, we replace domestic gasoline/diesel prices with the world oil price. The second model is used as a robustness check to account for a possible correlation between gasoline (diesel) prices and oil price; this correlation could potentially confound the price relationships. To examine the three hypotheses, we perform the variance decomposition of the price relationships for both models, using the VEC results. The variance decomposition quantifies the effect of exogenous shocks in the own and other prices on the current and future values of biofuel prices. According to the results from the first model reported in Table 5, in all three countries ethanol prices are most responsive to their own lagged values; that is, the own lagged values (own shocks) explain between 50 and 99 percent of the biofuel price variance. The own price effect decreases over time, however. The rest of price series – that is, gasoline prices and

15

non-own ethanol prices – participate individually between 1 and 50 percent in the variance of ethanol prices, depending on the country and time horizon considered. The U.S. ethanol price contributes to the variance of the European ethanol prices after 48 weeks by 11.91 percent. The relative variance in ethanol prices in Europe is then caused by the shocks to the Brazil ethanol prices, 9.5 percent, and partially by the shocks to the gasoline market in the three countries, almost 7 percent in total. On the other hand, the relative variance in the U.S ethanol prices is resistant to the shocks to the EU ethanol prices (0.12 percent after 48 weeks). The Brazilian ethanol price contributes 13.54 percent of the variance of the U.S. ethanol prices after 48 weeks. The relative variance in the U.S. ethanol prices caused by the shocks to Brazilian gasoline prices is even stronger, 22.28 percent. The variance decomposition results further indicate that the U.S. ethanol prices also react relatively strongly to the U.S. gasoline prices (11.01 percent after 48 weeks), but their reaction to the EU gasoline price is weak (less than 1.22 percent). The Brazilian ethanol prices are most responsive (after the own price) to the EU gasoline price (17.01 percent after 48 weeks) and Brazilian gasoline (6.35 percent).13 The rest of prices determine to a lesser extent the Brazilian ethanol prices. Brazilian policies appear to determine ethanol prices in other countries most strongly, followed by the United States and the European Union (hypothesis 1). Brazilian ethanol and gasoline prices contribute to the variance of the EU and U.S. ethanol prices by 10.2 percent and 35.82 percent, respectively. The U.S. ethanol and gasoline prices contribute to the variance of the EU and Brazil ethanol prices by 14.4014 and 3.18 percent, respectively. The European Union plays a less significant role in determining ethanol prices in other countries (less than 2 percent) except for the EU gasoline, which is an important driver of Brazil ethanol (17.01 percent). The results of model 2 reported in table 6 confirm these price relationships. 13

The effect of the EU gasoline price might be weaker because our model does not include Brazilian sugar prices, which are likely to affect ethanol prices as well. 14 14.40% = 11.91% U.S. ethanol + 2.49% U.S. gasoline.

16

The oil price is an important determinant of ethanol prices in Brazil (40.14 percent) and the United States (9.89 percent) which in turn determine EU ethanol price (by 7.09 percent and 25.53 percent, respectively). This confirms that Brazil and the United States are the key price leaders on the ethanol market. This, however, contradicts the hypothesis 1, as the highest ethanol price over the analyzed time period was observed in the EU (proxied by Germany). A possible explanation is that, under then prevailing market conditions, the EU ethanol prices were isolated due to high EU import tariffs on ethanol.15 According to model 1, both the United States and Brazil influence the EU ethanol price through their ethanol prices. The United States and Brazil affect each other primarily through the gasoline prices, whereas Brazil affects the U.S. ethanol price mainly through the gasoline price (table 6). The results in table 5 confirm that for ethanol, hypothesis 2 tends to prevail relative to hypothesis 3. Both gasoline prices and other countries’ ethanol prices co-determine ethanol prices, indicating that the tax credit (or tax exemption) (hypothesis 2) and not mandates (hypothesis 3) determine ethanol prices. Gasoline prices explain 34.51,16 6.66, and 25.27 percent of the ethanol price variation in the United States, European Union, and Brazil, respectively. On the other hand, the other countries’ ethanol price contributes with 13.66,17 21.41, and 2.31 percent to the ethanol price variation in the United States, European Union, and Brazil, respectively. In order for the hypothesis 3 to hold, the impact of gasoline prices on ethanol prices would need to be small. Both the U.S. and Brazilian gasoline prices, however, significantly affect the ethanol price; this supports the rejection of the third hypothesis. Note that because the hypothesis 3 is a subset of hypothesis 2, both policies could interchangeably affect biofuel prices, but we cannot identify their separate effects. Given the fact that gasoline

In general, denatured ethanol imported to the EU is subject to an import tariff of €10.2 per hectoliter, whereas undenatured ethanol is subject to €19.2 per hectoliter import duty. 16 34.51% = 1.22% EU gasoline + 11.01% USA gasoline + 22.28% Brazilian gasoline. 17 13.66% = 0.12% EU ethanol + 13.54% Brazil ethanol.

15

17

price plays an important role in biofuel price formation, the tax policy is likely dominating the mandate. This is confirmed by results from model 2 that indicates that oil price explains a significant share of the ethanol prices variation in price leading countries: it explains 40.14 percent of the price variation in Brazil and 9.89 percent in the United States (table 6). With regards to biodiesel from model 1, akin to the ethanol, most of the variance in all biodiesel prices can be explained by its own innovations (more than 68 percent) (Table 7). The effect of the shocks to the EU biodiesel price on the current and future values of U.S. biodiesel price is much stronger than vice versa. The European biodiesel price contributes to the variance of the U.S. biodiesel prices after 48 weeks by 26.54 percent, while it is only 0.09 percent vice versa. The EU biodiesel is significantly affected by the EU diesel price (12.78 percent). Other prices play only a minor role in determining biodiesel prices in the European Union and the United States. The results from model 2, where diesel prices are replaced by the oil price, confirm the overwhelming role of the EU biodiesel price in determining biodiesel price in the United States (83.5 percent) (Table 8). With the results from model 2, we cannot determine unequivocally what policy plays a role in biodiesel price formation. The first model indicates that the EU tax policy might be effective (hypothesis 2) because of a significant impact of the EU diesel on the EU biodiesel, 12.78 percent (Table 7). Further, the favorable EU biodiesel price sets the price for U.S. biodiesel given by the high EU – U.S. biodiesel causation effect (26.54 percent) reported in Table 7. Note that the hypothesis 3 is a subset of the hypothesis 2, which could imply that both policies may interchangeably affect biodiesel prices. On the other hand, based on model 2, the mandate is found to cause biodiesel prices in the EU (hypothesis 3) because the oil price explains an insignificant share of the EU biodiesel price, 0.06 percent. In turn, the EU biodiesel price is a key driver of the U.S. biodiesel (83.5 percent) (Table 8). Overall, the results are inconclusive: both tax exemption and the mandate may drive biodiesel prices. Yet, the

18

model appears to point to a strong role of the mandate. 7. Concluding Remarks This paper has investigated the price leadership in the world biofuel – that is, ethanol and biodiesel – markets, as well as analyzed which biofuel policy determines the biofuel price. We have used a cointegration analysis and a variance decomposition technique to analyze empirically weekly price series for the Unites States, Brazil, and Canada. Because the empirical approach followed in this paper does not depend on the theory outlined in Kliauga et al. (2008) and de Gorter et al. (2011), our results can be viewed as a check of hypotheses of the earlier papers. Our results imply that the U.S. and Brazilian ethanol polices (mostly the U.S. blender’s tax credit and Brazilian tax exemption) share the price leadership, but the Brazilian appears impact appears to be stronger. In this respect, our findings are only partially consistent with those of Kliauga et al. (2008), as they hypothesized that the United States is the price leader for the ethanol. The difference in results is partly due to a different time period analyzed in their study (2004 -2008). For biodiesel, our results support the hypothesis of de Gorter et al. (2011) about the European Union’s price leadership. However, the results are inconclusive as to what biofuel policy plays the leading role in the biodiesel price formation. Yet, the EU mandate may dominate the EU tax exemption. There are many aspects of real markets that have been abstracted from our paper, such as the effect of biofuel feedstocks on the biofuel prices, or a greater number of countries trading in biofuels than the three we consider. These considerations as well as extension of the period analyzed are left for future research. References

19

Abbott, P.C., Hurt, Ch., Tyner, W.E. (2009). “What's Driving Food Prices?” March 2009 Update, No 48495, Issue Reports, Farm Foundation, http://econpapers.repec.org/RePEc:ags:ffispa:48495. Asche, F., Osmundsen, P., Tveteras, R., (2001). “Market integration for natural gas in Europe”. International Journal of Global Energy Issues 16(4): 300–312 Asche, F., Gjolberg, O., Volker, T., (2003). „Price relationships in the petroleum market: an analysis of crude oil and refined product prices”. Energy Economics 25(3): 289–301 Babcock, B.A. (2008). “Breaking the Link between Food and Biofuels.” Briefing Paper 08BP 53, July 2008, Center for Agricultural and Rural Development, Iowa State University Balcombe, K., Rapsomanikis, G., (2008). “Bayesian estimation of nonlinear vector error correction models: The case of sugar-ethanol-oil nexus in Brazil”. American Journal of Agricultural Economics 90(3): 658–668. Busse, S., Brümmer, B., Ihle, R. (2010). "Interdependencies between fossil fuel and renewable energy markets: the German biodiesel market," DARE Discussion Papers 1010, GeorgAugust University of Göttingen, Department of Agricultural Economics and Rural Development (DARE). Ciaian, P., and Kancs, D. (2011). “Interdependencies in the Energy-Bioenergy-Food Price Systems: A Cointegration Analysis”. Resource and Energy Economics 33(1): 326–348 Cuddington, J.T., Wang, Z. (2006). “Assessing the degree of spot market integration for U.S. natural gas: evidence from daily price data”. Journal of Regulatory Economics 29(2): 195– 210 de Gorter, H., Just, D.R. (2008). 'Water' in the U.S. ethanol tax credit and mandate: implications for rectangular deadweight costs and the corn-oil price relationship, Review of Agricultural Economics 30(3): 397–410. __________. (2009). “The Economics of a Blend Mandate for Biofuels.” American Journal of Agricultural Economics 91(3): 738–750. __________. (2010). “The Social Costs and Benefits of Biofuels: The Intersection of Environmental, Energy and Agricultural Policy.” Applied Economic Perspectives and Policy. 32(1): 4-32. de Gorter, H., Drabik, D. and Just, D.R. (2011). “The Economics of a Blender's Tax Credit versus a Tax Exemption: The Case of U.S. “Splash and Dash” Biodiesel Exports to the European Union”. Applied Economic Perspectives and Policy (2011) ppr024 first published online September 6, 2011 doi:10.1093/aepp/ppr024 Directive 2009/28/EC of the European Parliament and the Council of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and 2003/30/EC. Engle, Robert F., Granger, Clive W. J. (1987). "Co-integration and error correction: 20

Representation, estimation and testing", Econometrica, 55(2): 251-276 Harris, R. (1995). “Using Cointegration Analysis in Econometric Modelling”. London: Prentice Hall/Harvester Wheatsheaf. Hertel, T., and Beckman, J. (2011). “Commodity Price Volatility in the Biofuel Era: An Examination of the Linkage between Energy and Agricultural Markets.” Working paper, NBER. Kliauga, E.M., de Gorter, H., and Just D.R. (2008). “Measuring the ‘Subsidy’ Component of Biofuel Tax Credits and Exemptions”. Paper presented at the International Agricultural Trade Research Consortium meeting, Scottsdale Arizona, 7-9 December 2008 http://www.ilr1.uni-bonn.de/iatrc/IATRC_Scottsdale_degorter_just_kliauga.pdf Lapan, H., Moschini, GC. (2009). "Biofuels Policies and Welfare: Is the Stick of Mandates Better Than the Carrot of Subsidies?," Staff General Research Papers 13076, Iowa State University, Department of Economics. Mallory L.M., Hayes J.D., Irwin, H.S. (2010). "How Market Efficiency and the Theory of Storage Link Corn and Ethanol Markets," Midwest Agribusiness Trade Research and Information Center (MATRIC) Publications 10-wp517, Iowa State University Paul, J.R., Mijkovic, D., Ipe, V. (2001). “Market integration in US gasoline markets”. Applied Economics 33(10): 1335–1340. Reuters (2010). “U.S. to Use More Ethanol in 2011, But Smaller Market Share”. Washington, Monday July 12, 2010. http://www.reuters.com/article/2010/07/12/us-usa-ethanol-targetidUSTRE66B4XR20100712 Reuters.(2011). “U.S. ethanol exports to surpass Brazil this year”. http://www.reuters.com/article/2011/08/25/us-usa-ethanol-exportsidUSTRE77O3EO20110825 Serletis, A., Herbert, J., (1999). “The Message in North American Energy Prices”. Energy Economics 21(5): 471–483 Serra, T., Zilberman, D., Gil, J. M. “Price volatility in ethanol markets”. European Review of Agricultural Economics, 38 (2): 259–280 Siliverstovs, B., L'Hegaret, G., Neumann, A., von Hirschhausen, C., (2005). “International market integration for natural gas. A cointegration analysis of prices in Europe, North America and Japan”. Energy Economics 27(4): 603-615 Tyner, W.E. (2010). “Comparison of the US and EU approaches to stimulating biofuels”. Biofuels 1(1): 19–21 Xiaodong, D., and Hayes, D. (2009). “The impact of ethanol production on us and regional gasoline markets” Energy Policy 37(8): 3227–3234. Wright, B.D. (2011). “The Economics of Grain Price Volatility”. Applied Economic Perspectives and Policy 33(1): 32–58. 21

Yano, Y., D. Blandford, and Surry, Y. (2010). “The Impact of Feedstock Supply and Petroleum Price Variability on Domestic Biofuel and Feedstock Markets – the Case of the United States.” Working Paper 2010:3, Department of Economics, Swedish University of Agricultural Sciences (SLU), Uppsala. Zhang, Z., Lohr, L., Escalante, C., and Wetzstein, M. (2010). “Food versus fuel: What do prices tell us?” Energy Policy, 38(1): 445–451

22

Appendix To illustrate how biofuel price relationships are affected by a blend mandate, we perform a comparative static exercise for the impact of an exogenous fossil fuel (without loss of generality, we assume gasoline) price change on the biofuel price (ethanol) in the price leading country. The derivations are based on the model developed in de Gorter and Just (2009). The equilibrium conditions in the fuel market with an exogenous gasoline price and a binding blend mandate are given by: PF    PE  t c   1    PG S E  PE    D F  PF 

(A1)

where P F , P E , and P G denote price of fuel (combination of ethanol, and gasoline), ethanol, and gasoline, respectively; tc denotes a tax credit;  denotes a blend mandate (e.g., 10 percent), and SE and DF denotes ethanol supply and fuel demand functions, respectively. Totally differentiating the system in (A1) yields:

dPF   dPE  1    dPG S E ' dPE   D F ' dPF

(A2)

where D F ' and S E ' are the first derivatives of the fuel demand and ethanol supply functions, respectively, with respect to their arguments. Solving system (A2) for dPE dPG , and transforming the solution into to the elasticity form yields:

1    DF  0 dPE  dPG  PF   SE DF PE

(A3)

where  DF denotes demand elasticity of fuel, and  SE denotes supply elasticity of ethanol. We extract elasticities from de Gorter and Just (2009) to calibrate the equation (A3). The results are reported in table 1.

23

Table 1. The Magnitude of dPE dPG for the United States with a Binding Blend Mandate and an Exogenous Gasoline Price

Year

dP E /dP G dP E /dP G dP E /dP G Ethanol Share of Fuel Ethanol Supply Ethanol Gasoline Fuel price Consumption ( ) Elasticity ( SE ) price (P E ) price (P G ) (P F ) (for DF = -0.10) (for DF = -0.26) (for DF = -0.40)

2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2008-09 Source: calculated

0.015 0.021 0.025 0.029 0.038 0.048 0.070

13.60 9.30 8.60 8.60 6.90 5.10 3.10

1.59 1.13 1.25 1.60 1.62 2.61 2.40

0.95 0.76 0.96 1.13 1.49 1.99 3.00

0.96 0.77 0.97 1.14 1.49 2.02 2.96

-0.012 -0.015 -0.015 -0.016 -0.015 -0.024 -0.024

-0.031 -0.040 -0.038 -0.041 -0.039 -0.063 -0.063

-0.048 -0.062 -0.059 -0.063 -0.060 -0.096 -0.097

24

Table 2. Correlation between Ethanol and Gasoline Prices EU U.S. Brazilian EU Variable ethanol ethanol ethanol gasoline EU ethanol 1.000 U.S. ethanol 0.649 1.000 Brazilian ethanol 0.717 0.614 1.000 EU gasoline 0.857 0.736 0.757 1.000 U.S. gasoline 0.826 0.762 0.723 0.976 Brazilian gasoline 0.919 0.673 0.860 0.919 Source: calculated

U.S. gasoline

Brazilian gasoline

1.000 0.880

1.000

Table 3. Correlation between Biodiesel and Diesel Prices

Variable EU diesel U.S. diesel EU biodiesel U.S. biodiesel Source: calculated

EU diesel 1.000 0.969 0.860 0.778

U.S. diesel

EU biodiesel

U.S. biodiesel

1.000 0.795 0.735

1.000 0.897

1.000

25

Table 4. Cointegration Results Trace test EU ethanol – Brazilian ethanol EU ethanol – U.S. ethanol EU ethanol – EU gasoline EU ethanol – U.S. gasoline EU ethanol - Brazilian gasoline U.S. ethanol – Brazilian ethanol U.S. ethanol - Brazilian gasoline U.S. ethanol - EU gasoline U.S. ethanol - U.S. gasoline U.S. gasoline - EU gasoline Brazilian ethanol – Brazilian gasoline Brazilian ethanol - U.S. gasoline Brazilian ethanol - EU gasoline Brazilian gasoline – U.S. gasoline Brazilian gasoline - EU gasoline

r=0 16.032 *** 20.947 20.044 21.934 15.767 *** 20.435 24.028 29.183 29.695 47.864 23.071 20.790 20.962 29.856 29.985

max

r=1 3.031 4.778 5.873 5.482 4.331 3.970 4.491 8.420 7.552 8.626 4.127 3.745 5.637 4.430 5.314

** ** ** ** ** *** *** *** ** ** ** *** ***

test

r=0 13.001 *** 16.169 14.171 16.452 11.436 *** 16.465 19.537 20.763 22.143 39.238 18.943 17.045 15.325 25.426 24.672

EU biodiesel – EU diesel 12.2925 *** 3.4793 8.8132 EU biodiesel – U.S. biodiesel 32.533 3.5984 *** 28.9346 EU biodiesel – U.S. diesel 12.812 *** 4.4231 8.3888 EU diesel – U.S. biodiesel 17.4631 *** 3.0728 14.3903 U.S. diesel - EU diesel 27.8459 4.0961 *** 23.7498 U.S. biodiesel – U.S. diesel 14.2895 *** 2.5705 11.719 Source: calculated Note: * significant at a 10% level, ** significant at a 5% level, *** significant at a 1% level

*** *** *** ***

r=1 3.031 4.778 5.873 5.482 4.331 3.970 4.491 8.420 7.552 8.626 4.127 3.745 5.637 4.430 5.314

** * ** ** ** *** *** *** ** ** * *** ***

3.4793 3.5984 *** 4.4231 3.0728 4.0961 *** 2.5705

26

Table 5. Variance Decomposition Results for Ethanol and Gasoline Prices weeks relative variance in percentage of forecast variance explained by innovations in ∆ EU ∆ USA ∆ Brazil ∆ EU ∆ USA ∆ Brazil ethanol ethanol ethanol gasoline gasoline gasoline 1 ∆ EU ethanol 95.09 2.48 2.43 0 0 4 91.59 3.08 4.86 0.13 0.3 12 82.56 6.74 9.08 0.91 0.38 24 76.11 10.06 9.51 2.21 1.55 48 71.94 11.91 9.5 3.47 2.49 1 ∆ USA ethanol 0 98.83 1.17 0 0 4 0.12 98.38 0.39 0.13 0.88 12 0.04 94.84 1.46 0.24 2.38 24 0.06 75.88 7.3 0.86 4.88 48 0.12 51.82 13.54 1.22 11.01 1 ∆ Brazil ethanol 0 0 100 0 0 4 1.22 0.41 96.07 1.31 0.02 12 1.49 3.04 84.76 9.17 0.91 24 1.17 2.43 77.35 13.61 2.25 48 1.05 1.26 72.43 17.01 1.91

0 0.03 0.33 0.56 0.7 0 0.09 1.04 11.01 22.28 0 0.97 0.64 3.2 6.35

Source: calculated

27

Table 6. Variance Decomposition Results for Ethanol and Oil Prices weeks

relative variance in

1 ∆ EU ethanol

percentage of forecast variance explained by innovations in ∆ EU ethanol ∆ USA ethanol ∆ Brazil ethanol ∆ Oil 95.14

2.53

2.33

0

4 12 24 48

92.95 85.7 76.45 67.15

3.28 8.38 16.21 25.53

3.64 5.84 7.13 7.09

0.13 0.08 0.21 0.23

1 ∆ USA ethanol 4 12 24 48

0 0.04 0.26 1.78 6.15

98.38 98.65 96.77 91.77 83.5

1.62 0.86 0.38 0.36 0.45

0 0.44 2.58 6.09 9.89

0 0.79 1.26 1.22 0.85

0 0.21 1.56 3.15 3.55

100 98.97 93.96 79.97 55.46

0 0.04 3.23 15.66 40.14

1 ∆ Brazil ethanol 4 12 24 48 Source: calculated

28

Table 7. Variance Decomposition for Biodiesel and Diesel Prices weeks relative variance in percentage of forecast variance explained by innovations (response) ∆ EU ∆ USA ∆ EU ∆ USA biodiesel biodiesel diesel diesel 1 ∆ EU biodiesel 100 0 0 0 4 98.63 0.11 0.38 0.87 12 93.55 0.08 5.62 0.74 24 87.32 0.09 10.33 2.26 83.8 0.09 12.78 3.33 1 ∆ USA biodiesel 1.19 98.81 0 0 4 9.61 87.93 0.17 2.29 12 21 74.22 0.89 3.89 24 24.79 69.93 2.56 2.73 48 26.54 67.85 3.72 1.89 Source: calculated

29

Table 8. Variance Decomposition Results for Biodiesel and Oil Prices weeks

relative variance in (response)

percentage of forecast variance explained by innovations in (impulse) ∆ EU biodiesel

∆ USA biodiesel

∆ Oil

1 4 12 24 48

∆ EU biodiesel

100 99.74 99.75 99.12 98.11

0 0.18 0.21 0.85 1.83

0 0.09 0.04 0.03 0.06

1

∆ USA biodiesel

1.61

98.39

0

10.47 38.42 67.18 83.5

89.48 60.43 29.7 11.77

0.05 1.15 3.12 4.73

4 12 24 48 Source: calculated

30