Ethanol and field crops: Is there a price connection?

Food Policy 63 (2016) 53–61

Contents lists available at ScienceDirect

Food Policy journal homepage: www.elsevier.com/locate/foodpol

Ethanol and field crops: Is there a price connection? Andrea Bastianin a, Marzio Galeotti b,⇑, Matteo Manera c a

University of Milan and Fondazione Eni Enrico Mattei, Italy University of Milan and IEFE-Bocconi, Italy c University of Milan-Bicocca and Fondazione Eni Enrico Mattei, Italy b

a r t i c l e

i n f o

Article history: Received 26 January 2016 Received in revised form 20 May 2016 Accepted 30 June 2016 Available online 15 July 2016 Keywords: Ethanol Field crops Granger causality Forecasting Structural breaks

a b s t r a c t We analyze the relationship between the prices of ethanol, corn, soybeans, wheat and cattle in Nebraska, U.S. We focus on long-run price relations, short-run Granger causality, in-sample and out-of-sample predictability linkages between returns on ethanol, field crops and cattle. Since the ethanol market has been subject to many policy interventions, our analysis takes structural breaks and parameter instabilities into account using modern statistical techniques. We find no evidence that ethanol returns Granger cause food price variations. Conversely, both in-sample and out-of-sample results suggest that ethanol is Granger caused and can be predicted by returns on corn. No linkages between ethanol and cattle are found. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction In recent years agricultural commodities have been a source of concern because of the tendency of prices to increase and become more volatile. The popular press and reports from international and non-government organizations have been voicing the responsibility of the significant expansion of biofuel production in causing increases of food prices, thus putting at serious risk the plight of millions of poor.1 This has sparked the ‘‘Food versus Fuel” debate. According to the underlying view, the introduction of biofuels has strengthened the linkages between fuel and agricultural markets. In particular, because of the very rapid expansion of U.S.-produced ethanol whose main input is corn, the increased conversion of maize into ethanol reduced the supplies of food and increased food prices. The boom of U.S. ethanol production has been largely policydriven. The gasoline used in American cars nowadays contains up to 10% ethanol (E10). At the root of this is the Energy Policy Act (EPA) of 2005 which established a mandate known as the Renewable Fuel Standard (RFS). The RFS program originally required 7.4 billion gallons of renewable fuel to be blended annually into gasoline by 2012, to help to reduce greenhouse-gas emissions and cutting oil imports. In 2007 the U.S. Congress passed the Energy Independence and Security Act (EISA) scaling up the RFS mandate to 13.2 billion gallons of corn-based ethanol annually by 2012 and to an unprecedented 36 billion gallons by 2022.2 ⇑ Corresponding author. 1 2

See e.g. ‘‘As high as an elephant’s eye”, The Economist October 16th, 2010. See Janda et al. (2012) for a more detailed discussion of biofuel policies.

http://dx.doi.org/10.1016/j.foodpol.2016.06.010 0306-9192/Ó 2016 Elsevier Ltd. All rights reserved.

In this paper we study the relationship between the prices of ethanol, corn, soybeans, wheat and cattle in Nebraska from January 1987 through March 2012. Due to the very high correlation with national prices, this state is often used as a representative case study for the U.S. We begin by analysing the stochastic properties of the price series considered. The results of unit root tests indicate that field crops and cattle prices are integrated of order one. As to ethanol, its price can be best described as being stationary around a broken trend. The break date, June 2005, can be associated with major policy changes in the U.S. ethanol market since, as noted above, the EPA was first voted in April 2005 and finally signed into law in August of the same year. Next, we study the long-run price relationship between ethanol and the other commodities building on the bound testing approach recently proposed by Pesaran et al. (2001). This procedure allows the investigation of level relations even when cointegration cannot be established because prices have different orders of integration. The results reveal that in the post-break period there is evidence of a level relationship running from the price of corn to the price ethanol, but not vice versa. Finally, we determine whether ethanol has predictive power for the other price series, or vice versa. To this end we evaluate short-run relationships between returns on ethanol, field crops and cattle both in-sample and out-of-sample via Granger causality testing. As for the out-of-sample analysis, we study the predictive content of different models and compare them against some benchmark specifications. We find no evidence of Granger causality and predictability running from ethanol to the other commodities. On the contrary, returns on ethanol are predictable by using

54

A. Bastianin et al. / Food Policy 63 (2016) 53–61

field crops. We do not find any linkages with cattle. These results are consistent with previous studies that have tested in-sample Granger causality (as mentioned in the next section and in surveys papers by Serra and Zilberman, 2013; Zilberman et al., 2013), as well as with Ubilava and Holt (2010) and Bastianin et al. (2014), which is the only contributions dealing with forecast evaluation. Our study complements and expands the literature along the following lines. Firstly, given that the ethanol market has been subject to several policy interventions, we control for structural breaks and parameter instabilities both when analysing price relations and when carrying out Granger causality testing. We implement the well-known procedure of Zivot and Andrews (1992) to test the null hypothesis of a unit-root against the alternative that prices are broken-trend stationary. In addition, standard tests of noGranger causality are complemented with a recently proposed test by Rossi (2005) that is robust to parameter instabilities. Secondly, we thoroughly analyze out-of-sample relationships between ethanol, field crops and cattle with the goal of confirming previous results with a different empirical strategy. As said, all existing studies have considered causality only within-sample analyses. Nothing is said about the out-of-sample performance of the estimated models, although it can be argued that this perspective is more consistent with the definition of causality originally put forth by Granger (1969). Our approach is thus in line with the one advocated by Ashley et al. (1980) who note that, since the definition of causality introduced by Granger (1969) is a statement about predictive ability, in-sample testing has to be considered merely as a first step, always to be complemented with an out-of-sample analysis. Moreover, out-of-sample tests can be seen as raising the bar relative to in-sample evaluation: if one model is superior to another it must forecast more accurately.3 Indeed, forecast evaluation can identify not only predictive ability, but also changes in predictive ability due to structural changes that insample tests cannot detect by construction (Clark and McCracken, 2005). Our forecast evaluation is based on the Clark and McCracken (2001) test which has been shown to have more power than in-sample tests of Granger causality in the presence of structural breaks and parameter instabilities (Chen, 2005). Thirdly, we include cattle prices in the analysis. We note that Serra and Zilberman (2013) identified the lack of studies about ‘‘meat” as one gap in their survey of the econometric literature dealing with biofuels. They point out that since biofuel feedstocks are used as production factors also for food products, more studies should focus on the transmission of price shocks along the food market chain (e.g. in their survey only three articles include ‘‘meat” in the analysis). The remainder of the paper is organized as follows. Section 2 discusses the causal nexus between ethanol and food prices and reviews the relevant literature concerning the ethanol and food prices relation debate. Section 3 describes the data. Section 4 presents the econometric methodology and the empirical results of the in-sample analysis. Forecasting ability and Granger causality out-of-sample are taken up in Section 5. Policy implications are illustrated in Section 6. Concluding remarks complete the paper. 2. Motivation and related literature On a given portion of arable land farmers grow corn. Corn can be used to produce ethanol, in which case it is supplied to refineries and used as an input, or alternatively can be further processed 3 This point was also stressed by Granger (1980) who showed that in-sample tests of the null hypothesis of no-causality are essentially tests of goodness of fit which, in the case of a rejection of the null hypothesis, inform the analyst that one variable could help to improve the forecasts for the other variable. However the author noted that ‘‘this is quite different from actually producing improved forecasts” (Granger, 1980, p. 348).

to be converted into flour for food purposes, into other ‘‘food” uses such as feedstock for cattle – cows are mainly fed with corn – and into various industrial uses (Anderson et al., 2008). A large portion of growth in corn demand is associated with growth in ethanol production because most U.S. ethanol is made from corn. Policyinduced shifts in the demand for ethanol or higher gasoline prices foster ethanol production, increasing its supply. More ethanol plants and production translates into more demand for corn, which in turn increases corn prices, ceteris paribus. Higher corn prices make corn more profitable to grow, causing some farmers to shift from other crops to corn production. This will also push food, seed, and industrial users to shift from corn to other commodities, increasing their prices. This is the rationale for the ‘‘Food vs. Fuel” claim (Mitchell, 2008; Baffes and Haniotis, 2010). Under this view causality would run from ethanol to corn prices and from corn price to the price of food and other corn-based products.4 Could causality run in the opposite direction? In principle, yes. A food policy ‘‘away” from wheat, a supposedly close substitute of corn, could bring about a demand increase in the corn market and in turn an increase in its price (again assuming a fixed amount of arable land). As a consequence, the price of a fundamental input to the refinery process would increase, leading to higher ethanol prices. After the ethanol price boom many economic studies have econometrically assessed the relationships, if any, between biofuels and agricultural prices. Surveys are provided by Serra and Zilberman (2013) and Zilberman et al. (2013). Zhang et al. (2009) estimate a vector error correction model on U.S. weekly data for corn, oil, gasoline, ethanol, and soybean prices over the period March 1989–December 2007. For the pre-ethanol boom period 1989–1999 the authors find that corn and ethanol prices cointegrate and that the former Granger-causes the latter. No evidence of a long-run relationship between the two variables is found in the ethanol boom period 2000–2007; however, in this period fuel prices (ethanol, oil, and gasoline) Granger cause corn prices. Zhang et al. (2010) analyze short- and long-run impacts of fuels on agricultural commodities in the U.S. They rely on monthly price data for corn, rice, soybeans, sugar, wheat, ethanol, gasoline, and oil from March 1989 through July 2008. Neither long-run, nor short-run Granger causality is found between fuel and agricultural commodity prices. Saghaian (2010) uses monthly time series data on oil, ethanol, corn, soybean, and wheat prices collected from January 1996 through December 2008. The author shows that all variables are integrated of order one and cointegrated. Pairwise Granger-causality tests indicate that while there is a feedback relationship between corn and ethanol prices, there are unidirectional relationships running from soybeans and wheat price series to ethanol. On the contrary, ethanol does not Granger cause soybeans or wheat prices. Using monthly data from 1990 to 2008 Serra et al. (2011) investigate whether ethanol, corn, oil, and gasoline prices in the U.S. are characterized by a long-run equilibrium relationship and whether the adjustment toward this equilibrium relationship is nonlinear. Rising energy prices are found to cause an increase in corn prices; however, corn price increases also boost ethanol prices. Since corn production is relatively inelastic, at least in the short run, an increase in the size of the ethanol market will induce a growth in corn price that in turn will yield higher ethanol prices. Kristoufek et al. (2012a) study the relationships between the monthly prices of agricultural commodities (corn, wheat, sugar cane, soybeans, sugar beets), biodiesel, ethanol and related fuels. Both in the short and medium term the price of corn Grangercauses the price of ethanol, but not vice-versa. The authors also 4 Of course this assumed relationship rests upon a few implicit assumptions, such as the presumption that amount of arable land is fixed over the short-run (Abbott, 2014).

A. Bastianin et al. / Food Policy 63 (2016) 53–61

show that an increase in the price of corn positively affects the price of ethanol and that this effect is relatively short-lived. Weekly data are used by Kristoufek et al. (2012b) who show that, for the period between November 2003 and February 2011, Granger causality runs from corn to ethanol, but not vice-versa. This result is confirmed by Vacha et al. (2013), who study time and frequency dependent correlations between biofuels, agricultural commodities and fossil fuels by means of wavelet coherence analysis. These authors find that the price of production factors (U.S. corn and German diesel) lead the price of biofuels (ethanol and biodiesel), but not vice-versa. Wixson and Katchova (2012) test causality and asymmetric price transmission in the U.S. from January 1995 to December 2010 with monthly price data for soybeans, corn, wheat, oil, and ethanol. Their results indicate that Granger causality runs from returns on corn and soybeans to returns on ethanol, but not vice-versa. Finally, Ubilava and Holt (2010) is the only study that analyzes the relationship between energy and corn prices both in-sample and out-of-sample. Their dataset contains weekly U.S. futures prices for the period October 2006–June 2009. The authors rely on non-linear models and conclude that the inclusion of oil and ethanol prices in the model does not improve corn price forecasts. 3. Data We use publicly available monthly time series of nominal spot prices for ethanol, corn, soybeans, wheat and cattle recorded in Nebraska from January 1987 through March 2012 (December 2010 for cattle). The price of ethanol is expressed in dollars per gallon, the prices of field crops (i.e. corn, soybeans and wheat) are in dollars per bushel, while the cattle price is expressed in dollars per hundredweight. The price of ethanol was sourced from the Nebraska Energy Office database, while prices of crops and cattle are available from the National Agricultural Statistics Service maintained by the U.S. Department of Agriculture. From this source we also took the dollar value of production of field crops and cattle that we used to construct the time-varying weights of two price indices used to summarize the price dynamics of these commodities. The ‘‘crops price index” (price index 1) includes the three field crops, while the ‘‘crops and cattle price index” (price index 2) adds cattle to the first index.5 Our empirical analysis focuses on the state of Nebraska for three main reasons. First, the prices of field crops, cattle and ethanol in Nebraska are highly correlated with national prices and are therefore often used as a representative case study for the U.S.: see Serra et al. (2011), McPhail (2011), and Enders and Holt (2013). Second, the importance of Nebraska in the U.S. ethanol industry: as of February 2011 the nameplate capacity of Iowa, Nebraska and Illinois was equivalent to 26.10%, 12.97% and 9.02%, respectively of the nation’s total (13,596 million gallons per year; source: Nebraska Energy Office). Third, data covering a very long span of time are available.6 Fig. 1 shows that the price history of ethanol can be ideally divided in two periods. The first one runs from 1987 through the early 2000s and is characterized by relatively stable prices and 5 More details about the data and the construction of the price indices are provided in an appendix available from the authors upon request. 6 Although Iowa is the most important ethanol producer in the U.S., Nebraska data are preferable for the aim and the methods of our study. Iowa’s time series start only in 2006 and contain less than 300 weekly observations. There are two problems with these data: first, we cannot appreciate the market developments prior to 2006 when the ethanol marked had already boomed; second, in order to have a reasonable number of observations for both estimation and forecast evaluation, out-of-sample tests should be carried out on samples starting after 2008, namely after the burst of the oil price bubble. Having such a volatile period in the evaluation sample makes the forecast evaluation more interesting and renders benchmark models more difficult to beat (see Section 5).

55

low volatility. During the more volatile second period, instead, the dynamics of the Nebraska ethanol market can be described as a succession of price ups and downs. Over the period January 1987–March 2012 the price of ethanol was on average 1.53 dollars per gallon and displayed a coefficient of variation equal to 0.35. If we compute these statistics for the period running through June 2005, we get an average price of 1.23 dollars per gallon with a coefficient of variation equal to 0.14. For the second sub-period from July 2005 through March 2012, the average price and coefficient of variation increase to 2.15 dollars per gallon and 0.22, respectively. A joint inspection of Fig. 1 and of Panel (a) of Table 1 shows that the second period started with a price increase that culminated at a record high of 3.58 dollars per gallon in June 2006. Cattle price with a coefficient of variation 0.15, is the least volatile series. Price index 1 displays two main peaks; the highest one occurred in June 2011, two months before both corn and soybeans prices reached their maxima (6.93 and 13.30 dollars per bushel, respectively). The second peak was recorded in March 2008, three months before ethanol reached 2.9 dollars per gallon. This peak is the result of field crops reaching record price levels: corn reached 5.4 dollars per bushel in June 2008, while soybeans and wheat prices settled at their record levels (13.3 dollars per bushel in July 2008 and 9.84 dollars per bushel in March 2008, respectively). Descriptive statistics for the percentage change of log-prices (i.e. returns) are shown in Panel (b) of Table 1. Correlation coefficients are reported between returns on price indices, field crops, cattle and contemporaneous, lagged and lead returns on ethanol, as well as the corresponding p-values indicating the probability of rejecting the null hypothesis of zero correlation. The contemporaneous correlation coefficients are all positive, but statistically nil at the 95% confidence level. Also, the correlations between leads of ethanol and returns on price indices, field crops and cattle are all statistically not different from zero. On the contrary, the correlation coefficients between lagged ethanol returns, corn and soybeans are positive and statistically significant. 4. Methodology and empirical results Our empirical strategy first aims at assessing whether or not a long-run cointegrating or, more generally, level relationship exists among the (log of the) price of ethanol, field crops, cattle and the two aggregate price indices. We next study short-run Granger causality and predictability links based on returns (i.e. percentage first differences of log prices).7 In both cases we pay special attention to structural breaks and parameter instabilities by using methods capable of taking them into account. In terms of notation prices are represented by Pj,t where j = E (ethanol), 1 (price index 1), 2 (price index 2), C (corn), S (soybeans), W (wheat), B (cattle). Lower case letters denote the logarithm of prices (i.e. pj,t = ln Pj,t), while the percentage log-return on commodity j is denoted by rj,t = 100
Dpj,t = 100
(pj,t  pj,t1). 4.1. Stationarity To investigate the statistical properties of the log-price series we begin by using standard unit root tests: augmented Dickey and Fuller (ADF), Phillips and Perron (PP), Kwiatkowski, Phillips, Schmidt, and Shin (KPSS). For the majority of price series both the ADF and the PP test do not reject the null hypothesis of a unit root and the KPSS rejects the null hypothesis of trend stationarity.8 7 In the case of no cointegration Granger causality can be studied on the basis of first differences only, ignoring any linkages between price levels, that would otherwise have been captured with an error correction term. 8 For the first difference of log-prices the null of a unit root can always be rejected.

56

A. Bastianin et al. / Food Policy 63 (2016) 53–61

Fig. 1. Ethanol price, price indices, field crops and cattle prices. Note: All prices are in current dollars and have been multiplied by 100 and divided by their value in January 1987 to put them on a common scale.

Table 1 Descriptive statistics.

Panel (a): Prices Average Coef. Var. Min (Date Min) Max (Date Max) No. obs.

Ethanol (1)

Price Index 1 (2)

Price Index 2 (3)

Corn (4)

Soybeans (5)

Wheat (6)

Cattle (7)

1.53 0.35 0.89 (01/1987) 3.58 (06/2006) 303

102.63 0.51 47.46 (01/1987) 287.70 (06/2011) 303

347.25 0.28 215.08 (01/1987) 723.11 (12/2010) 288

2.70 0.40 1.43 (02/1987) 6.93 (08/2011) 303

6.71 0.33 4.00 (10/2001) 13.30 (08/2008) 303

3.84 0.38 1.99 (11/1999) 9.84 (03/2008) 303

76.52 0.15 58.60 (09/1998) 104.00 (12/2010) 288

0.15 38.80 0.06 (0.2799) 0.07 (0.2198) 0.03 (0.6378)

0.06 55.55 0.10 (0.0781) 0.04 (0.4949) 0.08 (0.1679)

0.09 60.25 0.05 (0.4122) 0.16 (0.0060) 0.06 (0.3118)

0.05 93.56 0.06 (0.2738) 0.15 (0.0085) 0.09 (0.1247)

0.08 66.21 0.05 (0.3982) 0.03 (0.5533) 0.06 (0.3126)

0.19 17.67 0.07 (0.2586) 0.05 (0.4122) 0.05 (0.4280)

Panel (b): First difference of log-prices Average 0.09 Coef. Var. 82.05 Corr (Et, Xt) – (p-value) – Corr (Et, Xt+1) – (p-value) – Corr (Et+1, Xt) – (p-value) –

Notes to the table: (i) Price index 1 excludes the price of cattle, while price index 2 includes it; (ii) ‘‘Corr (E, X)” denotes the correlation coefficient between ethanol and the series on columns (2)–(7).

57

A. Bastianin et al. / Food Policy 63 (2016) 53–61

Because of the market and policy changes discussed in the Introduction, structural breaks cannot be ruled out for our series. Zivot and Andrews (1992) (ZA) propose a test for a unit root against broken-trend stationarity when the time of break is unknown. Although for price indices, field crops and cattle the outcomes of this test are in agreement with those of the ADF and PP tests, the null hypothesis of a unit root in the price of ethanol is strongly rejected according to the ZA test.9 This statistic selects June 2005 as the break date. Interestingly, as reported above, this date can be associated with major policy changes that have affected the U.S. ethanol market as noted above: indeed, the Energy Policy Act (EPA) was first voted on April and finally signed into law on August 2005. The EPA has increased the amount of biofuel that must be mixed with gasoline sold in the United States. The EPA and the increasing restrictions on MTBE as a fuel oxygenate might be responsible for the rapid growth in U.S. ethanol production and use over the last decade (Solomon et al., 2007). Given that the EPA might be the cause of the break in the ethanol price series in June 2005, in Table 2 we use such date to split the sample in a pre- and post-break period and present the ADF, PP and KPSS test. The sample split highlights that ethanol is stationary before and after the break. The same can be said for Price Index 1 and the price of field crops before the break. For the post-break sample we can conclude that price indices and field crops have a unit root, while the results for cattle price are mixed. The main implication of these results is that we cannot apply cointegration techniques to study the relationship between these price levels.

4.2. Long-run Rejection of the null hypothesis of unit root implies that series can be analyzed in their first differences and that standard cointegration techniques cannot be applied. This however does not mean that we cannot study level relationships among our series. In fact, the ‘‘bound testing” approach proposed by Pesaran, Smith, and Shin (2001) (PSS) can be used to check whether a (level) relationship exists between the price of ethanol and other variables. The PSS test can be conveniently applied regardless of the order of integration and of cointegration of variables. In order to carry out the test we need to choose which variable is ‘‘long-run forcing”. For instance, the assumption that ethanol is long-run forcing for corn implies that the latter commodity has no long-run impact on the former. In this case a test of the null hypothesis of no long-run relationship running from ethanol to corn can be carried out after estimating the following model:

DpC;t ¼ k0 þ k1 t þ /pC;t1 þ dpE;t1 þ xDpE;t þ

p1 X

cj DpC;tj

j¼1

þ

q1 X

gj DpE;tj þ et

ð1Þ

j¼1

The bound test is obtained by calculating the F-statistic for the joint significance of lagged prices, that is H0: / = d = 0. Pesaran et al. (2001) provide two sets of critical values: a lower critical value bound, which assumes that all regressors are I(0), and an upper critical value bound, for cases when all regressors are I(1). A value of the F-statistic below the lower bound implies that H0 cannot be rejected and hence no long-run relationship exists; on the contrary, when the F-statistic exceeds the upper bound, H0 is rejected 9 The results of the ADF, PP, KPSS, and ZA tests are not reported here to conserve on space. They are nevertheless presented in an appendix that is available from the authors.

Table 2 Unit root and stationarity tests before and after the break. Test

ADF C

PP C&T

C

Panel (a): First sub-sample, January 1987–June 2005 Ethanol 0.013⁄⁄ 0.045⁄⁄ 0.002⁄⁄⁄ Price Index 1 0.010⁄⁄⁄ 0.055⁄ 0.014⁄⁄ Price Index 2 0.026⁄⁄ 0.118 0.035⁄⁄ Corn 0.003⁄⁄⁄ 0.015⁄⁄ 0.040⁄⁄ Soybeans 0.018⁄⁄ 0.068⁄ 0.059⁄ Wheat 0.054⁄ 0.190 0.064⁄ Cattle 0.061⁄ 0.146 0.205 Panel (b): Second sub-sample, Ethanol 0.003⁄⁄⁄ Price Index 1 0.548 Price Index 2 0.771 Corn 0.778 Soybeans 0.843 Wheat 0.284 Cattle 0.048⁄⁄

July 2005–March 2012 0.015⁄⁄ 0.037⁄⁄ 0.310 0.574 0.571 0.804 0.845 0.742 0.763 0.761 0.674 0.263 0.168 0.238

KPSS C&T

C

C&T

0.008⁄⁄⁄ 0.075⁄ 0.166 0.145 0.183 0.217 0.412

S S S S S S S

NS S S NS S S NS

(December 2010) 0.130 S 0.339 NS 0.570 NS 0.745 NS 0.525 NS 0.625 S 0.523 S

S S NS NS NS NS S

Notes to the table: (i) prices in logarithms; (ii) values in the table are p-values of the null hypothesis that a series has a unit root; (iii) for the KPSS test ‘‘S” and ‘‘NS” respectively denote non-rejection and rejection of the null hypothesis of trend stationarity at 95% confidence level; (iv) ‘‘C” and ‘‘T” respectively indicate whether a constant and/or a trend have been included in the test equation.

and it is concluded that there is a long-run relationship between prices. The test is inconclusive when the F-statistic falls within the lower and upper bound. To implement the PSS test we estimate by OLS two sets of six bivariate models like (1), after having determined the number of lagged first differences of both the dependent and the independent variables with the Schwarz Information Criterion. The test is carried out twice: in one case we assume that ethanol is long-run forcing for the other six series, in the other case we assume that it has no long-run impact on the other variables (i.e. the series j – E is long-run forcing for ethanol). In Table 3 the price of ethanol is assumed to be long-run forcing for the other variables. For each sub-sample the bound test is carried out with and without the trend in the test equation.10 In most cases the F-test lies below the lower 5% critical value bound, thus implying that, when ethanol is assumed to drive the other variables in the long-run, no level relationship can be detected in any of the sample periods. In the remaining cases the value of the statistic lies between lower and upper bound and the test is inconclusive.11 Table 4 shows the PSS test outcomes when price indices, the price of field crops or cattle respectively are long-run forcing for ethanol. Some very interesting results emerge: (i) the sample split helps identifying a level relationship from corn to ethanol in the more recent post-break period, while no level relationship is detected over the full sample or before the break; (ii) a postbreak level relationship with ethanol is identified also for wheat; (iii) ethanol and soybeans do not share any level relation; (iv) level relationships between ethanol and the two price indices are found in both sub-samples, but not in the full sample. 4.3. Short-run We now turn to study whether short-run movements of the ethanol price affect the price of field crops and cattle, or vice versa. 10 As the distribution of the bound-F test depends on the exogenous variables included in the test equation, the authors provide a set of critical values that assume that both a trend and a constant are included in the model, and another set that is based on the assumption that only the constant term is included. 11 The PSS test was carried out for the full sample as well. While the results are not shown here (but are available from the authors), the null hypothesis was never rejected.

58

A. Bastianin et al. / Food Policy 63 (2016) 53–61

Table 3 Bound tests for the existence of a level relationship (ethanol long-run forcing).

Panel (a): Constant Price Index 1 Price Index 2 Corn Soybeans Wheat Cattle

Jan. 1987–Jun. 2005

Jul. 2005–Mar. 2012

p, q

p, q

1, 1, 1, 1, 1, 3,

1 1 1 1 1 1

Panel (a): Constant and trend Price Index 1 1, 1 Price Index 2 1, 1 Corn 1, 1 Soybeans 1, 1 Wheat 1, 1 Cattle 3, 1

F-test

Table 4 Bound tests for the existence of a level relationship (commodity i long-run forcing).

F-test

p, q

F-test

p, q

F-test

2, 2, 2, 2, 2, 1,

6.25(c) 6.68(c) 4.75(a) 4.66(a) 5.25(b) 11.70(c)

1, 1, 1, 1, 1, 1,

1 1 2 1 1 1

5.71(b) 7.70(c) 9.30(c) 5.58(b) 7.17(c) 7.92(c)

7.57(c) 7.85(c) 6.34(a) 6.22(a) 6.75(b) 12.59(c)

1, 1, 1, 1, 1, 1,

1 1 2 1 1 1

8.84(c) 19.53(c) 9.53(c) 5.81(a) 7.44(c) 13.32(c)

6.09(c) 5.96(c) 3.16(a) 3.96(a) 3.66(a) 3.27(a)

1, 1, 2, 1, 1, 1,

1 1 1 1 1 1

2.45(a) 0.57(a) 0.77(a) 0.40(a) 2.51(a) 4.85(a)

6.08(a) 6.15(a) 3.04(a) 3.63(a) 3.64(a) 3.65(a)

1, 1, 2, 1, 1, 1,

1 1 1 1 1 1

1.93(a) 1.48(a) 2.39(a) 1.01(a) 1.00(a) 6.26(a)

Panel (a): Constant and trend Price Index 1 2, 1 Price Index 2 2, 1 Corn 2, 1 Soybeans 2, 1 Wheat 2, 1 Cattle 1, 1

We start by testing for Granger causality in the usual in-sample fashion. When evaluating the null hypothesis of no Granger Causality (GC) running from ethanol to commodity i we estimate the following models:

i ¼ 1; 2; C; S; W; B

ð2Þ

Tests of the null hypothesis of no GC running from commodity j to ethanol are based on the estimation of the models:

r E;t ¼ aE þ bj r j;t1 þ cE r E;t1 þ uE;t

Jul. 2005–Mar. 2012

Panel (a): Constant Price Index 1 Price Index 2 Corn Soybeans Wheat Cattle

Notes to the table: (i) the null hypothesis of the bound F-test is of no level relationship; (ii) (p, q) denote the lags of the first differenced regressors; (iii) (a) indicates that the statistics lies below the 0.05 lower bound (i.e. the null is not rejected), (b) that it falls within the 0.05 bounds (i.e. the test is inconclusive), and (c) that it lies above the 0.05 upper bound (i.e. the null is rejected); (iv) the asymptotic critical value bound for the F-statistic is [4.94, 5.73] for the case with constant and [6.56, 7.30] for the case with constant and trend; see Pesaran et al. (2001), Table CI (iii) and CI(v).

r i;t ¼ ai þ bE r E;t1 þ ci r i;t1 þ ui;t

Jan. 1987–Jun. 2005

j ¼ 1; 2; C; S; W; B

ð3Þ

Since Eqs. (2) and (3) represent a bivariate Vector Autoregressive model of order one estimation relies on OLS using the entire sample of observations January 1987–March 2012 (December 2010 when either cattle or price index 2 entered the specification).12 GC testing is carried out on the basis of F-tests of the null hypothesis that the estimates of the b’s in a given model are jointly equal to zero. Table 5 presents the outcome of tests of no GC running from ethanol to returns on price indices, field crops and cattle (Panel (a)) and the evidence as to whether the price of field crops or cattle have in-sample predictive power for ethanol (Panel (b)). If we consider the evidence based on ethanol price models in Panel (a) the null of no GC is rejected for corn and soybeans. The strongest rejection is recorded for soybeans that displays a pvalue lower than 1%, while the null that corn is not Grangercaused by ethanol is rejected only at 5% significance level. The GC test based on commodities price models (Panel (b)) rejects the null hypothesis of no causality only for corn and soybeans. In general the p-values associated with these rejections are higher than those we obtained when looking at GC from ethanol to corn and soybeans. Taken together these results point to the existence of a feedback relationship between ethanol, corn and soybeans. However, we concluded above that ethanol shows predictive ability only for soybeans. 12 All standard errors are heteroskedastic and autocorrelation consistent. In the appendix available from the authors a set of robustness checks is also provided that includes Granger causality testing in a multivariate Vector Autoregressive framework and forecast evaluation for models where the number of lags has been selected via minimization of the Schwartz Information Criterion. In addition, the impact of changing the size of the estimation sample on forecast evaluation has been considered.

1 1 1 1 1 1

Notes to the table: see Table 3.

One problem with the standard approach above is that in the presence of instabilities traditional GC testing may have little or no power, as shown by Rossi (2005). It will be recalled that the price series for ethanol was found to be characterized by structural breaks. To cope with this problem we preliminarily test for the null of parameter stability in Eqs. (2) and (3) using Andrews’ (1993) Quandt Likelihood Ratio (QLR) test.13 Upon this outcome we then carry out a test for the joint null hypothesis of stability and no GC proposed by Rossi (2005).14 Rejection of the null hypothesis can occur either when parameters are not stable or when, even if they are constant, are different from zero. In both cases a rejection is evidence of GC.15 When considering the ethanol price model the QLR test for the null hypothesis of stability of the constant and lagged ethanol returns as reported in Panel (a) of Table 6 does not provide evidence of parameter instability. In addition, Rossi’s test never rejects the joint null hypothesis of parameter stability and no GC running from ethanol to the other variables. Since the statistic is carried out on the constant and the parameter associated with the first lag of ethanol returns, a non-rejection of the null hypothesis means these parameters are stable and that a model including only the AR(1) term cannot be rejected. On the basis of the above tests we can conclude that (i) parameters relating ethanol to price indices, field crops and cattle are stable and that (ii) there is weak evidence in favour of GC running from ethanol to corn and soybeans. We say ‘‘weak” for two reasons: first, the results of Rossi’s tests contradict those from traditional testing procedures. Second, even if one took these results literally, the only conclusion that can be drawn is about the expected predictive power of lagged returns on ethanol. We address below the issue of out-of-sample predictive ability of models with and without lagged ethanol returns, which helps to understand whether the previous weak evidence of GC actually translates into improved forecasts thanks to the information provided by ethanol. 13 This test is the maximum of a series of Chow test statistics testing the hypothesis of no break at date s, where s is such that [0.15T] 6 s 6 [0.85T], where T is the number of observations and [.] denotes the integer part of a number. We use the QLR statistic to test the stability of the a and b parameters of Eqs. (2) and (3). The distribution of the test is non-standard and has been tabulated by Andrews (1993). 14 Like the QLR test, Rossi’s statistic is based on a sequence of tests of the joint null hypothesis of no break at date s, where [0.15T ] 6 s 6 [0.85T ], and of no GC. This amounts to ask whether the a’s and b’s are constant throughout the sample period and the b’s are jointly statistically different from zero. The test, referred to as optimal Exponential Wald test, has a non-standard distribution which has been tabulated by Rossi (2005). 15 If parameters are not constant over time it means that either before or after the break they are different from zero, and hence that GC is present at least in one of the two subsamples.

59

A. Bastianin et al. / Food Policy 63 (2016) 53–61 Table 5 Granger causality tests. Price Index 1

Table 6 Parameter instability and Granger causality tests. Price Index 2

Corn

Soybeans

Wheat

Cattle

Panel (a): Ethanol price models GC test 0.12 1.64

3.23

7.23

1.04

0.25

Panel (b): Commodities price models GC test 1.35 0.05

7.62***

6.49**

0.20

0.16

Notes to the table: Results refer to Eq. (2) for Panel (a) and Eq. (3) for Panel (b); ‘‘GC test” is the F-statistic for Granger Causality testing. * Rejection of the null hypothesis at the 10%. ** Rejection of the null hypothesis at the 5%. *** Rejection of the null hypothesis at the 1%.

Turning to commodities price models in Panel (b) of Table 6 the QLR test provides very limited evidence of parameter instability and the Rossi’s test never rejects the null hypothesis of stability and no GC. 5. Ethanol, field crops and cattle prices: granger causality and forecasting ability In this section we aim to determine whether ethanol has predictive power for the other series, or vice versa. We do so by comparing the Mean Squared Forecast Error (MSFE) of forecasts from models (2) and (3) with three benchmarks: a first order autoregressive model, AR(1), and two random walk models, without and with drift, which we denote as RW and RWD respectively.16 We evaluate the statistical significance of MSFE differentials (i.e. MSFE of each estimated model in (2) and (3) minus the MSFE of the benchmark model b with b = AR(1), RW, RWD) with the Clark and McCracken (2001) encompassing test for nested models. The null hypothesis of the test is that the additional information used by a model does not improve the forecasting performance of the benchmark model. A rejection implies that the additional regressors that enter a model have out-of-sample forecasting power for the dependent variable. The distribution of the test is non-standard; critical values are provided by Clark and McCracken (2000).17 Results of the out-of-sample evaluation of the models in (2) and (3) against AR(1), RW, and RWD benchmarks are displayed in Table 7. We present two sets of results: the Root Mean Squared Forecast Error (RMSFE) ratio and the p-values associated with the Clark and McCracken (2001) test. A RMSFE ratio lower (greater) than one suggests that the model including lagged ethanol returns is better (worse) at forecasting the return of a given commodity than the benchmark. Since Panel (a) of the table shows that majority of RMSFE ratios are greater than one, the benchmarks are rarely out-performed by models that make use of ethanol to forecast price indices, field crops or cattle. A notable exception is soybeans, for which the model that includes ethanol leads to more accurate forecasts than the AR(1) and RWD benchmarks. 16 For Eq. (2) the benchmark AR(1) models can be written as ri,t = ai + qi ri,t1 + ui,t where i = 1, 2, C, S, W, B. For Eq. (3) the benchmark AR(1) model is rE,t = aE + qE rE,E,t = aE + qE rE,t1 + uE,t. The RW forecast for t + 1 is rj,t = 0, while the RWD model is rj,t = aj where j = E, 1, 2, C, S, W, B. Forecasts are obtained using a rolling forecasting scheme: a window of R = b0.5Tc observations is used to estimate models (2) and (3) and to generate one-month ahead forecasts. The iteration of this procedure, that moves the estimation window forward one month at a time, produces a set of forecast vectors, one for each model, of size (T-R). We have also considered windows of size R = b0.7Tc and R = b0.8Tc: since varying the size of the estimation sample does not affect our results, we only show results for the case R = b0.5Tc. We thank a referee for suggesting this robustness check. 17 The distribution of the test, denoted ENC-NEW, is non-standard and depends on the forecasting scheme, on the ratio between the size of the estimation sample (R) and the size of evaluation sample (T-R), and on the number of restrictions that need to be imposed on the bigger model so as to obtain the benchmark. When evaluating the predictive performance of models against the RW and the RWD benchmarks, the autoregressive term is dropped from models in (2) and (3).

Price Index 1 Price Index 2 Corn

Soybeans Wheat Cattle

Panel (a): Ethanol price models QLR test 9.65 9.76 Break date – – Rossi’s test 2.36 4.45

6.59 – 2.68

Panel (b): Commodities prices models QLR test 9.15 4.26 Break date – – Rossi’s test 1.91 1.00

12.45* 10.84 06/1991 – 5.48 4.87

2.78 – 4.17

7.33 – 1.52

3.57 – 0.72

5.55 – 1.41

6.17 – 3.96

Notes to the table: (i) The null hypothesis of the QLR test is that parameters are stable; Rossi’s test is a joint test of stability and no Granger Causality; (ii) For Panel (a) the tests involve both constant and the lagged first difference of (log) ethanol price. The dependent variables are reported as column headers. They refer to Eq. (2) in the paper; (iii) For Panel (b) the tests involve both the constant and the lagged first difference of all explanatory variables excluding the lagged value of ‘‘Ethanol”. The dependent variable ‘‘Ethanol” is the first difference of the log price of ethanol. The columns corresponds to Eq. (3) in the paper; (iv) When the QLR test rejects the null hypothesis we report the time of the break (i.e. mm/yyyy) below the value of the statistic. Rejection of Rossi’s null hypothesis indicates evidence in favour of Granger Causality. ** Rejection of the null hypothesis at the 5%. *** Rejection of the null hypothesis at the 1%. * Rejection of the null hypothesis at the 10%.

However, ethanol does not improve the forecasting performance above that of the simple RW specification. The encompassing test rejects the null hypothesis when forecasts based on ethanol for Price Index 2 and soybeans are evaluated against the AR(1) model as well as the RWD benchmark. Therefore, for both these cases there is evidence that ethanol improves the out-ofsample predictive ability of some of the benchmarks. On the basis of the evidence in this and the previous subsections we can conclude that there is only very limited evidence of Granger causality and out-of-sample predictability running from ethanol to field crops. In addition, we can conclude that ethanol has no predictive power, neither in-sample, nor out of sample for wheat, cattle and Price Index 1. Ethanol forecasts have been obtained in the same way as the forecast of field crops, cattle and price indices just presented; RMSFE ratios and the encompassing tests are shown in Panel (b) of Table 7. When forecasting ethanol, models that use corn or soybeans lead to lower losses than the benchmarks: RMSFE ratios are below unity and the test always rejects the null hypothesis. Therefore those models out-perform the benchmarks. To sum up, we are now able to draw more robust conclusions on Granger causality between the prices of ethanol, field crops and cattle. First, the overall predictive in-sample and out-of-sample ability of ethanol for the other variables is very low. Second, corn and soybeans have both in-sample and out-of-sample predictive power for ethanol. Therefore, we can argue that ethanol is Granger caused by corn and that there is a feedback relation between ethanol and soybeans. Third, no causality is found between ethanol, wheat, cattle and price indices. Lastly, we can state that these relations are quite stable over time.18 6. Conclusions and policy implications In this paper we studied the relationship between the price of ethanol, field crops and cattle in Nebraska from January 1987 through March 2012 both in-sample and out-of-sample. Our inter18 An appendix available from the authors provides additional results based on the Toda and Yamamoto (1995) procedure to test the null hypothesis of no GC within a Vector Autoregression for log-prices. The procedure can be applied to series, whether integrated or cointegrated. The results confirm the findings presented here.

60

A. Bastianin et al. / Food Policy 63 (2016) 53–61

Table 7 Root mean squared forecast error ratios and encompassing test. Price index (excl. cattle)

Price index (incl. cattle)

Corn

Soybeans

Wheat

Cattle

R = [0.5T] Panel (a): Ethanol price models AR(1) 1.0072 Random Walk (RW) 1.0071 RW with drift 1.0065

1.0037* 1.0056 0.9988**

1.0066 1.0101 1.0064

0.9946** 1.0022 0.9976*

1.0047 1.0102 1.0062

1.0055 1.0068 1.0045

Panel (b): Commodities prices models AR(1) 1.0005 Random Walk (RW) 1.0019 RW with drift 1.0002

1.0066 1.0096 1.0064

0.9890** 0.9902** 0.9885**

0.9897** 0.9916** 0.9899**

1.0046 1.0054 1.0037

1.0087 1.0123 1.0090

R = [0.7T] Panel (a): Ethanol price models AR(1) 1.0061 Random Walk (RW) 1.0058 W with drift 1.0041

0.9936 0.9891 0.9873*

1.0091 1.0134 1.0061

0.9849** 0.9982 0.9939

1.0014 1.0044 0.9999

1.0031 1.0001 1.0012

Panel (b): Commodities prices models AR(1) 1.0017 Random Walk (RW) 1.0025 RW with drift 1.0005

1.0014 1.0028 1.0009

0.9919 0.9933 0.9913*

0.9886* 0.9880* 0.9861**

1.0027 1.0040 1.0021

1.0045 1.0059 1.0040

R = [0.8T] Panel (a): Ethanol price models AR(1) 1.0037 Random Walk (RW) 1.0019 RW with drift 1.0034

0.9936 0.9895 0.9884*

1.0014 1.0001 1.0043

0.9815** 0.9979 0.9951

0.9998 0.9958 0.9969

1.0032 0.9993 1.0013

Panel (b): Commodities prices models AR(1) 0.9985 Random Walk (RW) 0.9996 RW with drift 0.9965

1.0012 1.0015 1.0000

0.9949 0.9966 0.9935

0.9898 0.9913 0.9882*

1.0025 1.0043 1.0012

1.0106 1.0113 1.0099

Notes to the table: (i) Figures in the table are Root Mean Squared Forecast Error (RMSFE) ratios. (ii) For Panel (a) a RMSFE ratio grater than one indicates that the benchmark is on average more accurate than the forecast of the model including ‘‘Ethanol”. The models refer to Eq. (2) in the paper; (iii) For Panel (b) a RMSFE ratio grater than one indicates that the benchmark is on average more accurate than the forecast of the model. The models refer to Eq. (3) in the paper; (iv) When the benchmark is either the RW or the RWD, the models do not include the lagged value of the dependent variables; (v) Critical value of the test are from Clark and McCracken (2000). A rejection indicates that the model is better than the benchmark. *** Rejection of the null hypothesis of the encompassing test at the 1% significance level. * Rejection of the null hypothesis of the encompassing test at the 10% significance level. ** Rejection of the null hypothesis of the encompassing test at the 5% significance level.

est was on the causality between biofuel and food prices. To that end a careful econometric analysis was carried out. Unit-root tests in the presence of possible structural breaks and parameter instabilities produced different results: field crops and cattle prices have a unit-root, while ethanol price is stationary around a broken trend. Interestingly, the break date occurs in June 2005, roughly around the same time when Energy Policy Act was first voted. Since prices have different orders of integration we relied on the bound testing approach of Pesaran et al. (2001) to study longrun level relationships between ethanol and the other commodities. In the post-break period we found a level relationship running from the price of corn to the price ethanol. The existing literature has largely analyzed the biofuel-food price relation in-sample. Somewhat surprisingly, almost nothing has been said about the out-of-sample performance of the proposed models, although it is well known that in-sample predictive ability does not necessarily imply out-of-sample predictive accuracy (Chao et al., 2001). We carried out a careful out-of-sample evaluation of price models which indicated that ethanol is predictable by using the price of field crops, but not vice versa. No sensible linkages between ethanol and cattle can be established. Overall these results are consistent with many previous studies (Serra and Zilberman, 2013). Out-of-sample evaluation is relevant for policy making. Policy interventions, whose effects are intended to last for some periods in the future, are inherently forward looking. The behavior of an empirical model at the end of the sample is then crucial for policy design, since the effectiveness of any policy has to be judged in

terms of its ability to act on a target variable, given current market conditions. For instance, extensive analytical and Monte Carlo evidence is available, which suggests that out-of-sample tests have power for identifying the ‘‘congruency” of a model at the end of the sample (Clark and McCracken, 2012). Parameter stability of models used for policy interventions is another issue of interest for policy analysts. Out-of-sample model evaluation prevents the researcher from selecting models which have in-sample predictive ability, while their out-of-sample performance is inferior to simpler, purely time-series benchmark models (such as the random walk for variables that are integrated of order one). Simulation results show that, in presence of parameter instabilities, some out-of-sample tests (among which the Clark and McCracken, 2005, tests employed in our paper) are more powerful than the in-sample test of Granger-causality which has been routinely used so far in the biofuels literature (Chen, 2005).19 One important issue which has been source of controversy and research interest has been the impact of quantitative polices targeting an increase in the production of biofuels. Especially in corn-rich U.S., an increased use of corn-based ethanol was seen 19 Bastianin et al. (2014) study the predictability in the price distribution of ethanol, field crops and cattle by asking whether ethanol returns can be used to forecast different parts of field crops and cattle returns distribution, or vice versa. They construct density forecasts using Conditional Autoregressive Expectile models estimated with Asymmetric Least Squares. Their results show that both the centre and the left tail of the ethanol returns distribution can be predicted by using field crops returns. On the contrary, there is no evidence that ethanol can be used to forecast any region of the field crops or cattle returns distributions.

A. Bastianin et al. / Food Policy 63 (2016) 53–61

as a way to reduce the country energy dependence from imported oil. In 2005 a Renewable Fuel Standard was introduced which, according to its critics, was ultimately responsible for an increase in world food prices. Many papers have considered the effects of a biofuel mandate such as the RFS. An econometric study of causality between ethanol and food prices, like the present one, cannot provide direct evidence to bear on the impact of biofuel mandates – a quantity instrument – on food prices at the root of the ‘‘Food vs Fuel” (FvsF) debate. The reason is that changes in ethanol prices do not identify the direction of ethanol production changes: for instance, a decrease in the price of ethanol is consistent with either a decrease or an increase in ethanol production.20 Our econometric findings, however, are of policy relevance in the above respect as they have some bearing on the FvsF albeit indirectly. In our study we found that the price of ethanol does not cause/has no predictive ability for the price of corn. We also find that the price of corn causes the price of ethanol. What are the implications of these findings? Suppose first that the price of corn increases. Being a fundamental input to bio-refineries this will cause, ceteris paribus, the supply of ethanol to decrease. If demand for ethanol does not decrease (or increases less), then the price of ethanol will go up. Under these circumstances we may observe a correlation between corn and ethanol prices with causality running from the former to the latter. This is our first main finding. What may have caused the price of corn to increase? Many possible factors, but hardly a Renewable Fuel Standard. Suppose now that the production of ethanol increases. There are to possibilities for this to occur: (i) a demand increase, with the consequence that the ethanol price goes up, (ii) a supply increase, causing the ethanol price to go down. In both cases, an expanded production of ethanol leads to an increase in the demand for corn. Assuming a supply of corn unchanged, this pushes the price of corn up. Hence, we will have the price of ethanol to cause the price of corn if supply is unchanged/does not decrease. This is our second main finding. We would not observe this causal effect if corn supply increased to match the demand increase. If the initial increase in the production of ethanol was due to a RFS mandate then we could or could not observe a causal effect from ethanol price to corn price. Our econometric findings leads us conclude that the price of ethanol does not cause the price of corn. In summary, our econometric results seem to be consistent with a RFS mandate to be ex-post unrelated to corn price dynamics. In this sense, this paper has provided evidence that partly and indirectly bears on the ‘‘Fuel versus Food” debate. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.foodpol.2016.06. 010. References Abbott, P., 2014. Biofuels, binding constraints and agricultural price volatility. In: Chavas, J.P., Hummels, D., Wright, B. (Eds.), The Economics of Food Price Volatility. University of Chicago Press, Chicago, pp. 91–131 (forthcoming).

20

A referee, to whom we are grateful, pointed out this aspect.

61

Anderson, D., Anderson, J.D., Sawyer, J., 2008. Impact of the ethanol boom on livestock and dairy industries: what are they going to eat? J. Agric. Appl. Econ. 40, 573–579. Andrews, D.W.K., 1993. Tests for parameter instability and structural change with unknown change point. Econometrica 61, 821–856. Ashley, R., Granger, C.W.J., Schmalensee, R., 1980. Advertising and aggregate consumption: an analysis of causality. Econometrica 48, 1149–1167. Baffes, J., Haniotis, T., 2010. Placing the 2006/08 Commodity Price Boom into Perspective. World Bank Policy Research Working Paper No. 5371. Bastianin, A., Galeotti, M., Manera, M., 2014. Causality and predictability in distribution: the ethanol-food price relation revisited. Energy Econ. 42, 152– 160. Chao, J., Corradi, V., Swanson, N.R., 2001. Out-of-sample tests for Granger causality. Macroecon. Dyn. 5, 598–620. Chen, S.S., 2005. A note on in-sample and out-of-sample tests for Granger causality. J. Forecast. 24, 453–464. Clark, T.E., McCracken, M., 2000. Not-For-Publication Appendix to Tests of Equal Forecast Accuracy and Encompassing for Nested Models (mimeo). Federal Reserve Bank, Kansas City. Clark, T.E., McCracken, M., 2001. Tests of equal forecast accuracy and encompassing for nested models. J. Economet. 105, 85–110. Clark, T.E., McCracken, M.W., 2005. The power of tests of predictive ability in the presence of structural breaks. J. Economet. 124, 1–31. Clark, T.E., McCracken, M.W., 2012. Consistent testing for structural change at the ends of the sample. Adv. Economet. 30, 133–169. Enders, W., Holt, M.T., 2013. The evolving relationship between agricultural and energy commodity prices: a shifting-mean vector autoregressive analysis. In: Chavas, J.P., Hummels, D., Wright, B. (Eds.), The Economics of Food Price Volatility. University of Chicago Press, Chicago, pp. 135–187 (forthcoming). Granger, C.W.J., 1969. Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37, 424–438. Granger, C.W.J., 1980. Testing for causality: a personal viewpoint. J. Econ. Dyn. Control 2, 329–352. Janda, K., Kristoufek, L., Zilberman, D., 2012. Biofuels: policies and impacts. Agric. Econ. 58, 367–371. Kristoufek, L., Janda, K., Zilberman, D., 2012a. Correlations between biofuels and related commodities before and during the food crisis: a taxonomy perspective. Energy Econ. 34, 1380–1391. Kristoufek, L., Janda, K., Zilberman, Z., 2012b. Mutual Responsiveness of Biofuels, Fuels and Food Prices. Working Paper N.38. Centre for Applied Macroeconomic Analysis, Australian National University. McPhail, L.L., 2011. Assessing the impact of US ethanol on fossil fuel markets: a structural VAR approach. Energy Econ. 33, 1177–1185. Mitchell, D., 2008. A Note on Rising Food Prices. World Bank Policy Research Working Paper No. 4682, July. Pesaran, H.M., Smith, R.J., Shin, Y., 2001. Bounds testing approaches to the analysis of level relationships. J. Appl. Economet. 16, 289–326. Rossi, B., 2005. Optimal tests for nested model selections with underlying parameter instabilities. Economet. Theory 21, 962–990. Saghaian, S.H., 2010. The impact of the oil sector on commodity prices: correlation or causation? J. Agric. Appl. Econ. 42, 477–485. Serra, T., Zilberman, D., 2013. Biofuel-related price transmission literature: a review. Energy Econ. 37, 141–151. Serra, T., Zilberman, D., Gil, J.M., Goodwin, B.K., 2011. Nonlinearities in the U.S. cornethanol-oil-gasoline price system. Agric. Econ. 42, 35–45. Solomon, B.D., Barnes, J.R., Halvorsen, K.E., 2007. Grain and cellulosic ethanol: history, economics, and energy policy. Biomass Bioenergy 31, 416–425. Toda, H.Y., Yamamoto, T., 1995. Statistical inference in vector autoregressions with possibly integrated processes. J. Economet. 66, 225–250. Ubilava, D., Holt, M., 2010. Forecasting Corn Prices in the Ethanol Era. Department of Agricultural Economics, Purdue University, unpublished manuscript. Vacha, L., Janda, K., Kristoufek, L., Zilberman, D., 2013. Time-Frequency Dynamics of Biofuels-fuels-food System. Working Paper 27. Centre for Applied Macroeconomic Analysis, Australian National University. Wixson, S.E., Katchova, A.L., 2012. Price Asymmetric Relationships in Commodity and Energy Markets. mimeo. Paper presented at the 123rd European Association of Agricultural Economists, Dublin, Ireland. Zhang, Z., Lohr, L., Escalante, C., Wetzstein, M., 2009. Ethanol, corn, and soybean price relations in a volatile vehicle-fuels market. Energies 2, 320–339. Zhang, Z., Lohr, L., Escalante, C., Wetzstein, M., 2010. Food versus fuel: what do prices tell us? Energy Policy 38, 445–451. Zilberman, D., Hochman, G., Rajagopal, D., Sexton, S., Timilsina, G., 2013. The impact of biofuels on commodity food prices: assessment of findings. Am. J. Agric. Econ. 95, 275–281. Zivot, E., Andrews, D.W.K., 1992. Further evidence on the great crash, the oil price shock, and the unit-root hypothesis. J. Bus. Econ. Stat. 10, 251–270.