Impact of Changes in Energy Input Prices on Ethanol Importation and Prices

Transcription

1 Journal of Agribusiness 29, 2 (Fall 2011) Agricultural Economics Association of Georgia Impact of Changes in Energy Input Prices on Ethanol Importation and Prices Elizabeth A. Yeager and Allen M. Featherstone This study examines the link between ethanol markets in the United States and Brazil. Vector autoregression is used to explore the relationship between ethanol prices in the United States and Brazil and the primary feed stocks used in production for January 2005 through February Vector autoregression is also used to explore the relationship between the two ethanol markets and the Brazilian real/united States dollar exchange rate. Causality tests, impulse responses, and forecast error decomposition are used to determine the economic implications. Results indicate the markets are not as closely linked as many have hypothesized. Key words: ethanol prices, vector autoregression Ethanol as a fuel source has been considered since the early 1900s. However, it was not until the last decade that rapid growth in ethanol use in the United States occurred because ethanol production had not been economically profitable. The United States and Brazil are the largest producers of ethanol in the world. Ethanol accounts for at least 40% of Brazilian automobile fuel, and gasoline sold in Brazil has at least 20% ethanol added to it (Clean Fuels Development Coalition, 2007). Additionally, Brazil is the world s largest ethanol exporter, exporting about one billion gallons of ethanol annually. Historically, the largest importer of ethanol from Brazil has been the United States importing 453 million gallons in 2006 and 185 million gallons in 2007 (Hofstrand, 2009). Imports from Brazil have decreased in recent years to approximately 95 million gallons for the 2009/2010 market year (United States Department of Agriculture, 2011b). The Renewable Fuels Standard (RFS) was amended in 2007 through the Energy Independence and Security Act and requires the RFS to increase to 36 billion gallons in 2022 (Renewable Fuels Association n.d.). U.S. lawmakers rationale for this standard is a desire to decrease dependence on foreign oil, lower greenhouse gas emissions, and produce more jobs for Americans. In 2007, U.S. ethanol imports were 7% as a percent of U.S. production. The percent of U.S. imports from Brazil was about 2.5 times higher than from any other country at 42% (Hofstrand, 2009). U.S. ethanol imports were less than 2% Elizabeth A. Yeager is an assistant professor in the Department of Agricultural Economics, Purdue University. Allen M. Featherstone is a professor in the Department of Agricultural Economics, Kansas State University.

2 182 Fall 2011 Journal of Agribusiness as a percent of U.S. production in 2010, but the percent of U.S. imports from Brazil remained high at 54% (United States Department of Agriculture, 2011b). In 2011, blenders received a tax credit of $0.45 per gallon of ethanol used regardless of the ethanol origin. An ad valorem tax and secondary tariff of $0.59 per gallon was imposed on ethanol imported from Brazil to limit exporters from obtaining the tax credit though Brazilian ethanol gets around that requirement by moving through countries in the Caribbean and Central America (Hofstrand, 2009; and Renewable Fuels Association, 2010). While the tariff is a point of contention, arguments have been made that even if the tariff was removed additional ethanol would not be imported. The objective of this study is to determine how closely ethanol prices follow changes in the price of the feedstock used in production: corn prices in the United States and sugar prices in Brazil. Additionally, the Brazilian real (BRL)/U.S. dollar (USD) exchange rate and ethanol prices are examined. The exchange rate may be a contributing factor to changes in imports due to currency changes from 2003 through the present. At the beginning of 2011, the USD purchased slightly over 1.6 BRL after a steady drop from the 3.5 BRL rate in February Studies have shown that oil price fluctuations are correlated with exchange rate changes (Amano and van Norden, 1998; Chen and Chen, 2007; Huang and Tseng, 2010; and Lizardo and Mollick, 2010). However, an area yet to be fully explored is whether the exchange rate has a role in determining ethanol prices. Methods and Data A vector autoregression (VAR) model was used to explore the relationship between U.S. and Brazilian ethanol prices and the Brazilian real (BRL)/U.S. dollar (USD) exchange rate. Additionally, the VAR model was used to capture the interdependencies between the input prices (corn and sugar cane) and the price of ethanol in the United States and Brazil. VAR allows for the simultaneous estimation of the variables in the system. VAR estimation treats all variables as endogenous and allows the lags of every variable to influence every other variable in the system (Featherstone and Baker, 1987). It is a data intensive approach that does not impose any a priori knowledge (Sims, 1980). Typically, the only initial restriction placed on the model is a maximum lag length. The estimated coefficients themselves cannot be interpreted individually. The coefficients tend to alternate in sign on successive lags and are further complicated by cross-equation relationships (Sims, 1980). Therefore, impulse response functions and forecast error decompositions are used to identify the response to an exogenous shock to one of the variables and the size and order of influence among the variables, respectively. The equations below illustrate the three-equation system for exchange rate (Xrate), ethanol prices in Brazil (BReth), and ethanol prices in the United States (USeth):

3 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 183 (1) (2) (3) where t is time in months; n is number of lags; k, a, b, c and d are estimated parameters; e 1t, e 2t and e 3t are the error terms for each equation. This model was used because the exchange rate between the BRL and USD, the ethanol price in Brazil, and the ethanol price in the United States affect the decision to import Brazilian ethanol into the United States. This model was estimated using R (Pfaff, 2008; and R Development Core Team, 2011). The number of lags was estimated using the Hannan-Quinn information criteria that trades-off fit for parsimony in parameters (Hannan and Quinn, 1979). The estimated lag length for the system was two though up to 12 lags were tested. The data used for equations (1) (3) were in levels and the data were stationary, in other words the eigenvalues were all less than one. The following equations illustrate the VAR estimation for the four-equation system for corn prices in the United States (UScorn), ethanol prices in the United States (USeth), ethanol prices in Brazil (BReth), and sugar cane prices in Brazil (BRsug): (4) (5) (6) (7) where t is time in months; n is number of lags; k, a, b, c, d and f are estimated parameters; e 4t, e 5t, e 6t and e 7t are the error terms for each equation. This model was chosen to examine the relationships between the prices of the feedstock used in production of

4 184 Fall 2011 Journal of Agribusiness ethanol and the ethanol prices. The data used for equations (4) (7) were in levels and the data were stationary. The estimated number of lags using the Hannan-Quinn criteria was two, with 12 lags being tested. Monthly exchange rates between the BRL and USD were obtained from the United States Department of Agriculture Economic Research Service (2011a). Monthly data on ethanol and corn prices in the United States were obtained through the Agricultural Marketing Resource Center for January 2005 through February 2011 (Hofstrand and Johanns, 2011). Corresponding monthly data on anhydrous ethanol and sugar cane prices in Sao Paulo, Brazil were obtained through the Center for Advanced Studies on Applied Economics (2011) and the Institute of Agricultural Economics (2011), respectively. The U.S. ethanol prices are in dollars per gallon and the corn prices are in dollars per bushel. The Agricultural Marketing Resource Center obtains the monthly ethanol prices from the USDA Iowa Ethanol Plant Report and the corn prices from the National Agricultural Statistics Service. The ethanol prices from Brazil are in Brazilian real per liter and the sugar cane prices are in Brazilian real per ton. Summary statistics for these variables are shown in Table 1. The average exchange rate during the 74 month sample period was 2.02 R$/$. The average corn price was $3.38/bushel, sugar cane price was R$32.50/ton, U.S. ethanol price was $1.91/gallon, and Brazilian ethanol price was R$0.91/liter. Table 1. Variable Definitions and Summary Statistics, January February 2011 Standard Variable Definition Mean Minimum Maximum Deviation Xrate Exchange rate (R$/$) UScorn U.S. corn $/bushel BRsug Brazil sugar cane R$/ton USeth U.S. ethanol $/gallon BReth Brazil ethanol R$/liter Figure 1 shows the relationship between corn prices and U.S. ethanol prices. The prices appear to move closer together beginning in August The correlation coefficient increases from 0.21 before August 2007 to 0.77 after August Figure 2 shows the relationship between sugar cane prices and ethanol prices in Brazil. The prices appear to move closely together during the entire time period, and the correlation between the prices is This may be due to the fact that many sugar cane processing facilities in Brazil can switch back and forth from sugar to ethanol production based on the current prices of each. Figure 3 presents the relationship between the ethanol prices in each country. The ethanol prices appear to generally move together with a correlation

5 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 185 coefficient of Figure 4 illustrates the relationship between the ethanol price in the United States and the BRL/USD exchange rate, and Figure 5 illustrates the relationship between the ethanol price in Brazil and the BRL/USD exchange rate. Both the U.S. and Brazilian ethanol prices move inversely of the exchange rate. Economic theory would suggest the U.S. ethanol price would move inversely of the exchange rate, but the Brazilian ethanol price should move in the same direction as the exchange rate. The correlation coefficient for the U.S. ethanol price and the exchange rate is and the correlation coefficient for the Brazilian ethanol price and the exchange rate is Figure 1: Corn and ethanol prices in the United States, January 2005 through February 2011

6 186 Fall 2011 Journal of Agribusiness Figure 2: Sugar cane and ethanol prices in Brazil, January 2005 through February 2011 Figure 3: Ethanol prices in the United States and Brazil, January 2005 through February 2011

7 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 187 Figure 4: Ethanol prices in the United States and the BRL/USD exchange rate, January 2005 through February 2011 Figure 5: Ethanol price in Brazil and the BRL/USD exchange rate, January 2005 through February 2011

8 188 Fall 2011 Journal of Agribusiness Results Exchange Rate and Ethanol Prices The coefficients for the estimated VAR system of equations (1) (3) are shown in Table 2. The individual coefficient estimates are difficult to interpret in a VAR system, therefore, causality tests, impulse responses, and forecast error decomposition were examined to understand the economic implications. Table 2. Estimated VAR Coefficients, Test Statistics, and Matrices of Residuals for Exchange Rate, Ethanol Price in Brazil, and Ethanol Price in the United States Statistic Exchange Rate Equation Brazil Ethanol Equation U.S. Ethanol Equation Adjusted R-squared Granger causality for exchange rate a Granger causality for Brazil ethanol b Granger causality for U.S. ethanol c Independent Variable Regression Coefficients Intercept ** Time trend * Xrate t *** Xrate t ** BReth t *** BReth t * USeth t * 1.114*** USeth t ** Covariance Matrix of Residuals Xrate BReth USeth Xrate BReth USeth Correlation Matrix of Residuals Xrate BReth USeth Xrate BReth USeth ***, **, and * indicate significance at the less than 0.1% level, 1% level, and 5% level, respectively. a F- value for testing H 0: Xrate do not Granger-cause USeth Breth. b F-value for testing H 0: BReth do not Granger-cause Xrate Useth. c F-value for testing H 0: USeth do not Granger-cause Xrate BReth.

9 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 189 Granger causality tests were conducted to identify if any of the variables could be used in predicting other variables, and the results indicated the variables did not cause any of the other variables besides themselves. The lack of causality among the variables may be a result of the nature of the ethanol markets in the two countries. The United States places taxes on a small amount of ethanol imported from Brazil to protect the domestic market. This may damper what otherwise may be a stronger relationship between the markets. The impulse response identifies the responses over time in all the variables to a onestandard-deviation increase in one of the variables (Featherstone and Baker, 1987). The errors are correlated across equations, so the system needs to be transformed to an orthogonal form for the impulse response and forecast error decomposition (Sims, 1980). The system was triangularized with the variables ordered as exchange rate, Brazilian ethanol price, and U.S. ethanol price. The ordering for the variables was based on prior knowledge of the variables exogeneity and comparisons of alternative orderings (Featherstone and Baker, 1987; Sims, 1980). Statistically, the results were not sensitive to the imposed ordering. 1 Figures 6, 7, and 8 illustrate the impact of a shock in one variable on the other variables. Figure 6 shows that a shock in the exchange rate results in almost no response in the Brazilian ethanol price and a decrease in the U.S. ethanol price. The decrease in the U.S. ethanol price is expected because an increase in the BRL/USD exchange rate makes U.S. ethanol relatively more expensive and Brazilian ethanol less expensive. Figures 7 and 8 indicate that a positive shock in either ethanol price results in a small positive response by the other ethanol price and a very minimal response by the exchange rate. Thus, the markets are not closely linked and are not correlated with the exchange rate between Brazil and the United States. 1 Tables presenting the results from the impulse response functions under alternative orderings are available from the authors upon request.

10 190 Fall 2011 Journal of Agribusiness Figure 6: Response of exchange rate, Brazil ethanol price, and United States ethanol price to a shock in exchange rate Figure 7: Response of exchange rate, Brazil ethanol price, and United States ethanol price to a shock in the Brazil ethanol price

11 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 191 Figure 8: Response of exchange rate, Brazil ethanol price, and United States ethanol price to a shock in the United States ethanol price Forecast error variance decomposition identifies the importance and order of influence among the variables (Sims, 1980). The forecast error decompositions for a 12- month period are presented in Table 3. Up to 24 months were examined; however, the changes were minimal after the 12 month period and are not presented in the table. The order of the variables is important in the forecast error decomposition as in the impulse response function and should be ordered from the most to least exogenous to the system (Featherstone and Baker, 1987). A variable that is strictly exogenous would have a value of 1.00 in its column of the table and its own innovations would explain all of the variance (Sims, 1980). The variables were ordered exchange rate, Brazilian ethanol price, and U.S. ethanol price based on a priori beliefs and comparisons of alternative orderings. 2 It is evident in the first two sections of Table 3 that the exchange rate and Brazilian ethanol price are exogenous to the system because after 12 months almost 94% of the variation in the exchange rate is explained by its own forecast error and 91% of the variation in the Brazilian ethanol price is explained by its own forecast error. The own forecast error explains approximately 60% of the variation in the U.S. ethanol price and the exchange rate explains about 33%. These results confirm the findings of the Granger causality tests and the impulse response functions that the prices are very exogenous and the markets are not closely linked, though the exchange rate does affect the U.S. ethanol market. 2 Results were robust to alternative orderings. Tables presenting the results for the alternative orderings are available from the authors upon request.

12 192 Fall 2011 Journal of Agribusiness Table 3. Proportions of k-months-ahead Forecast Error Attributed to Innovations in Respective Series Months Ahead (k) Proportion of Error Explained by: Xrate BReth USeth Xrate BReth USeth

13 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 193 Ethanol, Corn, and Sugar Cane Prices The coefficients for equations (4) (7) are presented in Table 4. Granger causality tests were conducted and the results indicate U.S. corn prices Granger-cause U.S. ethanol, Brazilian sugar cane, and Brazilian ethanol prices. Results also indicate Brazilian ethanol prices Granger-cause U.S. ethanol, Brazilian sugar cane, and Brazilian ethanol prices. Table 4. Estimated VAR Coefficients, Test Statistics, and Matrices of Residuals for Corn Price in the United States Ethanol Price in the United States, Ethanol Price in Brazil, and Sugar Cane Price in Brazil Statistic U.S. Corn Equation U.S. Ethanol Equation Brazil Ethanol Equation Brazil Cane Equation Adjusted R-squared Granger causality for U.S. corn a 4.110*** Granger causality for U.S. ethanol b Granger causality for Brazil ethanol c 2.681* Granger causality for Brazil sugar cane d Independent Variable Regression Coefficients Intercept * 0.278** Time trend * 0.055* UScorn t *** 0.391*** UScorn t *** USeth t *** USeth t * BReth t *** BReth t * 5.330* BRsug t *** BRsug t * Covariance Matrix of Residuals UScorn USeth BReth BRsug UScorn USeth BReth BRsug Correlation Matrix of Residuals UScorn USeth BReth BRsug UScorn USeth BReth BRsug ***, **, and * indicate significance at the less than 0.1% level, 1% level, and 5% level, respectively. a F- value for testing H 0: UScorn do not Granger-cause USeth BReth Brsug. b F-value for testing H 0: USeth do not Granger-cause UScorn BReth Brsug. c F-value for testing H 0: BReth do not Granger-cause UScorn USeth BRsug. d F-value for testing H 0: BRsug do not Granger-cause UScorn USeth BReth.

14 194 Fall 2011 Journal of Agribusiness The impulse response functions are illustrated through Figures 9, 10, 11, and 12. The variables were ordered U.S. corn, U.S. ethanol, Brazilian ethanol, and Brazilian sugar cane based on prior beliefs of the exogeneity of the variables. 3 The sugar cane price was the most responsive to a shock in the variables, but the responses were statistically insignificant. Minimal responses were observed in most cases implying the two ethanol markets are not closely linked. Figure 9: Response of United States corn price, United States ethanol price, Brazil ethanol price, and Brazil sugar cane price to a shock in the United States corn price 3 Tables presenting the results from the impulse response functions under alternative orderings are available from the authors upon request.

15 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 195 Figure 10: Response of United States corn price, United States ethanol price, Brazil ethanol price, and Brazil sugar cane price to a shock in the United States ethanol price Figure 11: Response of United States corn price, United States ethanol price, Brazil ethanol price, and Brazil sugar cane price to a shock in the Brazil ethanol price

16 196 Fall 2011 Journal of Agribusiness Figure 12: Response of United States corn price, United States ethanol price, Brazil ethanol price, and Brazil sugar cane price to a shock in the Brazil sugar cane price The forecast error decompositions for a 12-month period are presented in Table 5. The corn price is almost entirely exogenous to the system and after 12 months over 94% of the variation in the change in corn price is explained by its own forecast error. The own forecast error for the U.S. ethanol price explained about two-thirds of its price and the corn price explained about 32%. The own forecast error explains over half of the variation in the ethanol price in Brazil while the corn price explains approximately 29% and the ethanol price in the United States explains about 12%. The own forecast error for sugar cane explains about 24% of the variation after 12 months, the corn price explains about 41%, and the U.S. and Brazilian ethanol prices explain about 17% and 18%, respectively. The relatively large effect of corn prices on sugar cane prices may partially be attributed to the use of high-fructose corn syrup, derived from corn, as a substitute to cane sugar in food and beverage manufacturing.

17 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 197 Table 5. Proportions of k-months-ahead Forecast Error Attributed to Innovations in Respective Series Months Proportion of Error Explained by: Ahead (k) UScorn USeth BReth BRsug UScorn USeth BReth

18 198 Fall 2011 Journal of Agribusiness BRsug Conclusion This research examined the link between U.S. and Brazilian ethanol prices. The results indicate that the price of ethanol in Brazil is not significantly impacted by a shock in the price of U.S. ethanol; therefore, it is unlikely United States ethanol demand is impacting the Brazilian ethanol market. Similarly, the price of ethanol in the U.S. was not significantly impacted by a shock in the price of ethanol in Brazil. Recent business ventures have been undertaken by two major U.S. agribusinesses. Monsanto purchased a Brazilian company that will allow them to focus more on sugar cane breeding and applied genomics, and ADM has formed a joint venture to build sugar cane plantations, mills, and ethanol distilleries in Brazil (Hofstrand, 2009). The business ventures may allow Monsanto and ADM to diversify some of their risk away from the U.S. market since the Brazilian market is not closely linked. The correlation coefficient between the U.S. and Brazilian ethanol prices was 0.32; however, the relationship did not appear to be very strong based on the impulse response functions. Graphically, the prices were responding to each other, but the changes were not statistically significant. This may partially be the result of other factors not included within the model such as the impacts of the Renewable Fuels Standard, oil or unleaded fuel prices, and the sugar market. Ethanol production has been targeted as a primary contributor to the increased corn prices in recent years. This study sheds light on the fact

19 Yeager and Featherstone Changes in Input Prices on Ethanol Prices 199 that an increase in the price of ethanol has a small effect on the change in the price of corn. Overall, the ethanol markets in the United States and Brazil are relatively independent of each other. References Amano, R.A., and S. Van Norden. (1998). "Exchange Rates and Oil Prices." Review of International Economics 6(4), Center for Advanced Studies on Applied Economics. (2011). CEPEA - Centro de Estudos Avancados em Economia Aplicada. Online. Available at [Retrieved April 22, 2011]. Chen, S.-S., and H.-C. Chen. (2007). "Oil prices and real exchange rates." Energy Economics 29(3), Clean Fuels Development Coalition. (2007). "The Ethanol Fact Book A Compilation of Information About Fuel Ethanol." American Coalition for Ethanol. Online. Available at [Retrieved January 4, 2011]. Featherstone, A.M., and T.G. Baker. (1987, August). "An Examination of Farm Sector Real Asset Dynamics: " American Journal of Agricultural Economics 69(3), Hannan, E.J., and B.G. Quinn. (1979). "The Determination of the Order of an Autoregression." Journal of the Royal Statistical Society. Series B (Methodological) 41(2), Hofstrand, D. (2009, May). "Brazil's ethanol exports." Ag Decision Maker Iowa State University. Online. Available at [Retrieved January 4, 2011]. Hofstrand, D., and A. Johanns. (2011, April). "Fuel and Grain Price Historic Comparisons." Agricultural Marketing Resource Center. Online. Available at prices_trends_and_markets.cfm. [Retrieved April 22, 2011]. Huang, A.Y., and Y.-H. Tseng. (2010). "Is Crude Oil Price Affected by the US Dollar Exchange Rate?" International Research Journal of Finance and Economics 58, Institute of Agricultural Economics. (2011, July 7). Online. Available at [Retrieved July 15, 2011]. Lizardo, R.A., and A.V. Mollick. (2010, March). "Oil price fluctuations and U.S. dollar exchange rates." Energy Economics 32(2), Pfaff, B. (2008, July). "VAR, SVAR and SVEC Models: Implementation Within R Package vars." Journal of Statistical Software 27(4), R Development Core Team. (2011). "R: A language and environment for statistical computing." R Foundation for Statistical Computing. Online. Available at [Retrieved April 1, 2011]. Renewable Fuels Association. (n.d.). "Renewable Fuels Standard." Renewable Fuels Association. Online. Available at [Retrieved January 4, 2011].

20 200 Fall 2011 Journal of Agribusiness Renewable Fuels Association. (2010, May 24). "Using More Brazilian Ethanol Would RAISE Gasoline Prices for D.C. Drivers - With or Without the Tariff." Renewable Fuels Association. Online. Available at [Retrieved January 3, 2011]. Sims, C.A. (1980, January). "Macroeconomics and Reality." Econometrica 48(1), U.S. Department of Agriculture. (2011a, February). "Agricultural Exchange Rate Data Set." United States Department of Agriculture, Economic Research Service, The Economics of Food, Farming, Natural Resources, and Rural America. Online. Available at [Retrieved April 22, 2011]. U.S. Department of Agriculture. (2011b, July 25). Feed Grains Data: Yearbook Tables. Online. Available at [Retrieved August 14, 2011].