On the upsurge of U.S. food prices revisited

Economic Modelling 42 (2014) 272–276

Contents lists available at ScienceDirect

Economic Modelling journal homepage: www.elsevier.com/locate/ecmod

On the upsurge of U.S. food prices revisited Jungho Baek a,1, Won W. Koo b,2 a b

Department of Economics, School of Management, University of Alaska Fairbanks, United States Center for Agricultural Policy and Trade Studies, Department of Agribusiness and Applied Economics, North Dakota State University, United States

a r t i c l e

i n f o

Article history: Accepted 27 June 2014 Available online xxxx Keywords: Commodity price Exchange rate Energy price Food price Johansen cointegration analysis Vector error-correction model

a b s t r a c t The objective of this study is to empirically examine the effects of changes in exchange rate, commodity price and energy price on five U.S. food prices — cereal/bakery, meats, dairy, fruits/vegetables and beverages. The Johansen cointegration analysis and a vector error-correction (VEC) model are applied to monthly data for the 2001–2010 period. Results show the existence of stable long-run relationships among the selected variables. We also find that energy and commodity prices have influenced U.S. food prices mainly through changes in prices of cereal/ bakery, meats and dairy. Finally, exchange rate is found to have been a significant factor influencing U.S. food prices. The Energy Independence and Security Act of 2007 is one of main driving forces for the recent food price inflation, which has affected negatively consumers, especially low income households, in the United States. © 2014 Elsevier B.V. All rights reserved.

1. Introduction During the period 2006–2008, U.S. consumers had experienced the rapid spike in food prices not witnessed for almost two decades. The Consumer Price Index for all food (food CPI), for example, increased by 5.0% between 2007 and 2008, the highest increase since 1990. Prior to this period, the food CPI in the U.S. had never increased above the average annual rate of 4.0% over the past 15 years. As a result, the food CPI increased much faster than the Consumer Price Index for all items (overall CPI) (3.3%). This sharp upward trend in U.S. food prices had mainly been led by increases in prices for cereal/bakery (7.3%), meats/poultry (5.0%), eggs (21.5%) and dairy products (7.7%). A large body of literature has analyzed the so-called 2006–08 price surge in the U.S. food markets (for example, Abbott et al., 2008; Hanrahan, 2008; Headey and Fan, 2008; Herndon, 2008; Lipsky, 2008; Mitchell, 2008; Rosegrant, 2008; Schnepf, 2008; Trostle, 2008). These studies have typically attempted to identify the causations of the sharp hikes in U.S. food prices. The results from these studies generally suggest that, among other things, rising energy prices (i.e., crude oil prices), increased farm commodity prices (due mainly to significant growth in ethanol production) and the weak U.S. dollar have been the important factors driving up the rise in food prices. Trostle (2008), for example, shows that the U.S. dollar's global weakness has helped U.S. major commodities become more competitive on export markets, thereby enhancing foreign demand for U.S. commodities and hence

1 2

E-mail addresses: [email protected] (J. Baek), [email protected] (W.W. Koo). Tel.: +1 907 474 2754; fax: +1 907 474 5219. Tel.: +1 701 231 7448; fax: +1 701 231 7400.

http://dx.doi.org/10.1016/j.econmod.2014.06.018 0264-9993/© 2014 Elsevier B.V. All rights reserved.

prices. In addition, Mitchell (2008) finds that the rapid increase in crude oil prices has significantly increased the competitiveness of ethanol, which in turn has pushed up demand for farm commodities (i.e., corn) and thus prices. One common feature of the studies reviewed above is that they have mostly used descriptive statistics, graphical methods and simulation methods at best in tackling the issue. In other words, relatively little attention has been paid to conduct econometric work on this topic. To our knowledge, Baek and Koo (2010), and Lambert and Miljkovic (2010) are perhaps the only two published articles that have introduced an econometric technique to empirically quantify the impacts of market factors on U.S. food prices. Using a cointegrated vector autoregression (CVAR) model, for example, Baek and Koo (2010) find that commodity price, energy price and exchange rate have significant impacts on U.S. food prices. Using the same econometric method, Lambert and Miljkovic (2010) show that commodity prices and food industry wages have played key roles in influencing U.S. food prices during the 1970–2009 period, but energy prices have not. One deficiency of these studies, however, is that they have commonly used aggregate food prices (i.e., food CPI) in linking food price inflation and market factors.3 Given that the impacts of market factors (i.e., prices of energy and commodities and exchange rate) on food prices vary depending on different food products, their empirical findings may suffer from aggregation bias 3 Unlike Lambert and Miljkovic (2010), we have excluded industry wages from the final model. It is because agriculture is known as one of the most capital intensive industries in the U.S. economy; hence, changes in industry wages may have little impacts on U.S. agricultural prices. In addition, the USDA recently reports that the increase in the 2006–2008 food prices is primarily the result of high energy prices. Thus, it seems sufficient enough to include energy prices as a key factor in the model.

J. Baek, W.W. Koo / Economic Modelling 42 (2014) 272–276

of data, thereby resulting in a misinterpretation of the dynamics of food prices in response to changes in the market variables.4 In this study, we take one step further and attempt to quantify the effects of various market factors on U.S. food prices within the context of disaggregating price series for food in the United States. A special attention has been given to the assessment of dynamic linkages between changes in prices of energy and agricultural commodities and exchange rate, and changes in prices of five major food groups such as cereal/bakery, meat, dairy, fruits/vegetables and beverages. To achieve this objective, we use Johansen's maximum likelihood cointegration analysis and vector error-correction (VEC) models as Baek and Koo (2010) did. The cointegration approach takes into account data nonstationarity and allows us to explore the dynamic relationships among a group of variables without imposing a priori structural restrictions on the model (Sims, 1980). It is hoped that this careful study should lead to a better understanding of the 2006–08 food price surge in the United States. The remaining sections present theoretical framework, empirical methodology, data, empirical findings and concluding remarks. 2. Theoretical framework To illustrate theoretical relationships among five major food groups, prices of energy and agricultural commodities, and exchange rate, the individual consumer's demand for food group j, which maximizes consumer utility subject to its budget constraint, is first defined as: i Dj


 ¼ f P j; Y

ð1Þ

where Dij is individual i's demand for food group j — in this study, for example, j = cereal/bakery, meat, dairy, fruit/vegetable and beverage; Pj is the price of food group j; and Yis the individual's disposable income. An aggregate demand for food group j in a region is a product of the individual demand for food group j and the population in the region: r

i

D j ¼ D j  POP

r

ð2Þ

where Drj is the aggregate demand for food group j in region r; and POPrdenotes the region's population. We then define the total supply of food group j in a region, which is largely determined by agricultural goods and fisheries available for food production in the region.
 r S j ¼ f P j ; AF

ð3Þ

where Srj is the aggregate supply of food group j in region r; and AF is agricultural and fishery products produced for food products in the region. It is important to emphasize that, since total agricultural and fishery products produced are generally used for food production, biofuel production and exports, agricultural products used for food production can be defined as the total agricultural production (AP) minus the sum of agricultural and fishery products used for production of biofuel and exports as follows: AF ¼ AP ðCP Þ−AT ðCP; ERÞ−GðCP; ENP Þ

depend on the price of agricultural commodity (CP) and the price of energy (ENP). Accordingly, AFcan be specified as a function ofCP, ENP and ER. Finally, Eqs. (2)–(4) and the market equilibrium conditions for the demand for and supply of each food group yield the following relationship: P j ¼ f ðCP; ENP; ER; Y Þ:

4 It is worth mentioning that in the early 2011, Journal of Policy Modeling published a special issue containing several papers that analyzed the agri–energy–food price nexus (see Sieber and Dominguez, 2011). Schade and Wiesenthal (2011), for example, investigate the interrelationship between commodity prices (e.g., oil and feedstock prices) and biofuel using the Monte Carlo method. However, these papers focus on the European agricultural markets.

ð5Þ

Since the U.S. government has mandated the renewable fuel standard at 10% of gasoline used under the Energy Independent and Security Act (EISA) of 2007, biofuel production is sensitive to the price of energy (ENP). As the price of energy increases, biofuel production will increase and also increase prices of food since agricultural products are used for biofuel production rather than food production. Exchange rate (ER) is also sensitive to food prices, because U.S. exports of agricultural products are strongly related to the value of the U.S. dollar against the importer's currency. Similarly, agricultural commodity prices (CP) are highly correlated with food prices. The relationship between the prices of the food groups (Pj) depends upon whether they are complementary or substitute. It should be pointed out here that, since a small portion of individual household income is used for food in the U.S., we drop disposable income from our empirical modeling. 3. Empirical methods The Johansen cointegration approach (Johansen, 1995) starts with an unrestricted vector autoregression (VAR) of zt involving up to k lags: zt ¼

k X

Φi zt−i þ Ψwt þ α þ ut

ð6Þ

i¼1

where z t is a (1 × n) vector of jointly determined (endogenous) variables – in this study, zt = [CELt, MET t, DRYt, FRUt, BEVt]; wt is a vector of exogenous variables – in this study, wt = [ERt, ENPt, CPt]; α is a vector of constant term; and ut is a vector of normally and independently distributed error term. CELt, METt, DRYt, FRUt and BEVt represent cereal/bakery, meat, dairy, fruit/vegetable and beverage, respectively. ERt, ENPt and CPt represent exchange rate, energy price and commodity price, respectively. Because the right-hand side of each equation in Eq. (1) consists of a common set of regressors including lagged and predetermined variables, ordinary least squares (OLS) is efficient to estimate each equation (Harris and Sollis, 2003). It is should be pointed out, however, that, if variables in zt are nonstationary, then OLS regression among the series results in a spurious regression problem (Wooldridge, 2006). To avoid this problem, Engle and Granger (1987) show that, even in the case that all the variables in a model are nonstationary, it is possible for a linear combination of integrated variables to be stationary; in this case, the variables are said to be cointegrated and hence the problem of spurious regression does not arise. Accordingly, the first requirement for the use of the Johansen approach is that the variables must be nonstationary. Eq. (6) can be reparameterized using the lag operator (L;Δ = 1 − L) as follows:

ð4Þ

where AT represents exports of agricultural and fishery products, which is defined as a function of the price of agricultural commodity (CP) and relevant exchange rate (ER); and Grepresents agricultural and fishery products used for non-food purpose, mainly biofuel production and

273

Δzt ¼

k−1 X

Γi Δzt−i þ Πzt−k þ Ψwt þ αþut

ð7Þ

i¼1

where Γi = − (I − Φ1 − … − Φi), (i = 1, …, k ‐ 1) and Π = − (I − Φ1 − … − Φk). The system specified this way contains information on both short-run and long-run adjustments to changes in zt, via the estimates of Γi and Π, respectively. That is, Π = αβ ', where α represents the speed of adjustment to equilibrium and β ' is a matrix of long-run coefficients such that the term β ' zt − k represents up to (n − 1) cointegration relationships in the system (Johansen, 1995). Eq. (7) is said to be cointegrated of rank r, if Π has a rank r. If Π has a rank 0

274

J. Baek, W.W. Koo / Economic Modelling 42 (2014) 272–276

b r b n, for example, r ≤ (n − 1) cointegration vectors exist in β; in other words, r columns of β consist of r linearly independent combinations of the variables in zt along with (n − r) nonstationary vectors. Hence, testing for cointegration is identical to a consideration of the rank of Π, which is determined by the likelihood ratio test (Johansen, 1995). If all variables in zt are I(1) and cointegrated, following Engle and Granger (1987), Eq. (7) can be reformulated into a vector errorcorrection (VEC) representation as follows: Δzt ¼

k −1 X

 0  Γi Δzt−i þ α β zt−1 þ Ψwt þ αþut

ð8Þ

i¼1

where β ' zt − 1 is a measure of the error or deviation from the equilibrium, known as an error-correction term. The term is stationary since the series are cointegrated. Eq. (8) explicitly takes into account the dynamics of short-run adjustments toward a long-run equilibrium. If at any time the equilibrium achieves, for example, β ' z t − 1 is equal to zero. During periods of disequilibrium, on the other hand, β ' zt − 1 is nonzero and measures the distance the system is away from equilibrium during time t; in this case, an estimate of α provides information on the speed of adjustment, suggesting how the variable zt changes in response to disequilibrium. 4. Data The U.S. Consumer Price Indices (CPIs, 2005 = 100) are used as proxies for prices of cereal/bakery, meat, dairy, fruits/vegetables and nonalcoholic beverages, and are collected from the Bureau of Labor Statistics (BLS) in the U.S. Department of Labor. Following Baek and Koo (2010), the index of prices received from all farm products (2005 = 100) is used as a proxy for U.S. commodity prices and is taken from the National Agricultural Statistics Service (NASS) of the U.S. Department of Agriculture. The U.S. index for energy (2005 = 100) is used as a proxy for U.S. energy price and is obtained from the BLS. Finally, the exchange rate is the real effective exchange rate (2005 = 100) and is collected from the International Financial Statistics (IFS) published by the International Monetary Fund (IMF). The dataset contains 110 monthly observations for the period November 2001 to December 2010 (2001: M11–2010:M12). All variables are in natural logarithms. 5. Empirical results As noted above, the first step in cointegration analysis is to identify if the selected variables are nonstationary. The existence of a unit root is determined using the Dickey–Fuller generalized least squares (DF–GLS) test (Elliot et al., 1996). The MAIC criterion (Ng and Peron, 2001) is used to determine the appropriate lag-length truncation in each model that includes a constant and a linear time trend. The results show that the null hypothesis of a unit root cannot be rejected for all level series Table 1 Results of Dickey–Fuller generalized least squares (DF–GLS) test. Variable

Level

First difference

Lag

ln ln ln ln ln ln ln ln

−1.83 −2.11 −2.06 −0.89 −1.21 −2.50 −1.07 −2.12

−3.26a −4.10a −5.42a −2.89b −6.00a −5.27a −4.17a −5.94a

3 1 1 10 1 2 8 1

CELt METt DRYt FRUt BEVt ERt ENPt CPt

Note: CELt, METt, DRYt, FRUt, BEVt, ERt, ENPt and CPt represent cereal/bakery price, meat price, dairy price, fruits/vegetables price, beverages price, exchange rate, energy price, and commodity price. The 5% and 10% critical values for the DF–GLS, including a constant and trend, are −3.02 and −2.73, respectively. The lag order is chosen by the MAIC criterion. a Denotes rejection of the null hypothesis at the 5% significance level. b Denotes rejection of the null hypothesis at the 10% significance level.

at the 5% significance level, but can be rejected after first-differencing (Table 1). From these findings, we conclude that all the eight variables are nonstationary and integrated of order one, or I(1); hence, cointegration analysis can be pursued on them. It is worth mentioning that from a policy perspective, it is important to know whether an economic time series has a unit root process or not. For example, if the selected variables have unit roots, then the levels of those variables in the coming year can be highly correlated with what those variables were many years ago. This means that a policy that causes a discrete change in the selected variables can have long-lasting effects. Prior to implementation of the Johansen test, the specification issue to be addressed is the determination of endogeneity (exogeneity) of the selected variables for the VAR model. In our case, based on economic reasoning, exchange rate, energy price and commodity price are generally thought to behave exogenously in U.S. food markets; in other words, these three variables significantly influence the movements of U.S. food prices, but may not be affected by U.S. food prices. This reasoning is also consistent with the empirical findings of Baek and Koo (2010). Therefore, the VAR model is first defined as five endogenous variables(CELt, METt, DRYt, FRUt, BEVt), with other three variables (ERt, ENPt, CPt) being exogenous.5 By eliminating insignificant exogenous variables based on an F-test in the system, the final VAR model is then specified and used for cointegration analysis. Johansen's maximum likelihood cointegration test is applied to determine the number of independent cointegration relationships in the five-dimensional system. For this, we select the VAR lag lengths of four (k = 4) as the optimal lag, based on both the AIC and diagnostic tests on the residuals; this model does not reveal any significant departures from serial correlation and homoskedasticity (Table 2). The results indicate that there are three cointegrating vectors (r = 3) at the 10% significance level (Table 2).6 This represents three stable equilibrium relationships to which the endogenous variables (i.e., five food prices) have a tendency to return in the long-run.7 From policy perspective, the existence of long-run relationships among the disaggregating food prices implies that any market shock on specific food price has impacts on the overall food price. In this respect, government policies implemented by overlooking this relationship could lead to undesirable outcomes such as inefficient resource allocations and welfare losses. For example, the U.S. energy policy under the Energy Independence and Security Act of 2007 (EISA) was designed to reduce the dependency on foreign oil by increasing production of renewable energy, including corn-based ethanol in the United States. However, the policy was one of main forces driving the 2006–2008 food price inflation, which affected negatively consumers, especially low income households in the U.S. Before turning our attention to the VEC model, for completeness, we conduct the test for the long-run weak exogeneity in order to check whether or not the five variables can be treated as endogenous in the cointegrating vectors. This test can be done by restricting a parameter in speed-of-adjustment to zero (αi = 0). The results show that all the variables can be rejected at the 5% significance level (Table 3). This indicates that the five variables are not weakly exogenous to the system; in other words, all the variables should be treated as endogenous variables in the system as we have done in this study. We also apply the test for the long-run exclusion to examine whether any of the endogenous variables can be excluded from a cointegrating vector. The null hypothesis is formulated by restricting the matrix of long-run coefficients to 5 To some, inclusion of both commodity price and energy price in the model may cause multicollinearity problem. But the correlation between these two variables over the sample period is found to be 0.42, which seems to be a bit low. Combined with the fact that multicollinearity does not lead to biased estimation in regression modeling (Wooldridge, 2006), this should somehow mitigate our concern with multicollinearity. 6 Unlike the trace test, the maximum eigenvalue test does not lead to a consistent test procedure (Doornik and Hendry, 2001). Hence, we only report the trace statistics to test the null hypothesis. 7 Non-normality of residuals does not bias the results of the Johansen estimation (Gonzalo, 1994).

J. Baek, W.W. Koo / Economic Modelling 42 (2014) 272–276 Table 2 Results of Johansen cointegration test. Null hypothesis

Eigenvalue

Trace statistics

r=0 r≤1 r≤2 r≤3 r≤4 Serial correlation Heteroskedasticity Normality

0.370 0.251 0.198 0.103 0.059 1.14 [0.17] 0.54 [0.99] 90.17 [0.00]a

121.17 [0.00]a 72.17 [0.01]a 41.52 [0.07]b 18.07 [0.35] 6.51 [0.41]

Parentheses are p-values. Serial correlation of the residuals of a whole system are examined using the F-form of the Lagrange multiplier (LM) test, which is valid for systems with lagged independent variables. Heteroskedasticity is tested using the F-form of the LM test. Normality of the residuals is tested with the Doornik–Hansen method (Doornik and Hendry, 1994). a Denotes rejection of the null hypothesis at the 5% significance level. b Denotes rejection of the null hypothesis at the 10% significance level.

zero (βi = 0). The results show that the null hypothesis can be rejected for all five endogenous variables at the 5% significance level (Table 3), indicating that all the five endogenous variables are relevant to the cointegrating space and cannot be excluded from the long-run relationship. With the three cointegration relationships obtained from the cointegration rank test, the VEC model in Eq. (3) is estimated using a general-to-specific procedure (Hendry, 1995). More specifically, since exchange rate, energy price and commodity price are treated as exogenous variables in the model, the VEC model is first estimated conditional on these three variables. By eliminating all the insignificant variables based on an F-test, the parsimonious VEC (PVEC) model is then estimated using full-information maximum likelihood (FIML). The number of lags included in the PVEC model is the same as that used in the cointegration test (4 lags). The multivariate diagnostic tests on the estimated model as a system indicate no serious problems with serial correlation (FAR(175, 287) = 0.87, p-value = 0.83) and heteroskedasticity (FARCH(1170, 123) = 0.46, p-value = 0.99). The results of the PVEC model show that cereal/bakery price is negatively correlated with lagged prices of dairy and nonalcoholic beverages, but positively correlated with its own lagged price (Table 4). Energy price and commodity price influence cereal/bakery price significantly. Meat price is negatively correlated with lagged beverage price, but positively correlated with lagged dairy price. It is also significantly affected by exchange rate and commodity price. Dairy price is negatively correlated with lagged beverage price, but positively correlated with its own lagged price. Exchange rate and prices of energy and commodity have significant effects on dairy price. Fruit/vegetable price is negatively correlated with lagged dairy and beverage prices, but positively correlated with lagged prices of cereal/bakery, fruit/vegetable and beverage. Finally, beverage price is negatively correlated with lagged cereal/bakery, fruit/vegetable and beverage prices. Our findings show that the relationships among the five food groups are unclear in general; that is, fruits/vegetables and beverages have

Table 3 Results of weak exogeneity and exclusion tests. Variable

Weak exogeneity H 0 : αi = 0

Exclusion H0 : βi = 0

ln ln ln ln ln

9.63 [0.02]a 16.69 [0.00]a 10.08 [0.02]a 14.47 [0.00]a 12.70 [0.01]a

20.36 [0.00]a 13.37 [0.00]a 15.17 [0.00]a 36.76 [0.00]a 21.30 [0.00]a

CELt METt DRYt FRUt BEVt

Note: αi and βi represent the speed of adjustment to equilibrium and a matrix of long-run coefficients. Values are the likelihood ratio (LR) test statistic based on the χ2 distribution. p-Values are given in brackets. a Denotes the rejection of weak exogeneity at the 5% level.

275

complementary relationships with cereal/bakery, meat and dairy (negative sign) in consumption, but the relationships among cereal, meat and dairy are not clear. In addition, meat and dairy may be substitute, but cereal/bakery may have a complementary relationship with dairy (negative sign), but substitute relationship with meat. In addition, the results show that exchange rate has a significant effect on all food prices except cereal/bakery; that is, depreciation of the U.S. dollar leads to make U.S. food products more competitive in the world market, thereby increasing demand for U.S. food products and hence prices. Given the significant exchange rate impact, growing global economic uncertainty is expected to affect commodity and food prices through changes in the value of the U.S. dollar. As most European economies are recently moving further into recession, for example, appreciation of the U.S. dollar is likely to lessen pressure on U.S. food prices through increases in U.S. agricultural imports. Finally, energy and commodity prices are found to play important roles in determining prices of cereal/bakery, meat and dairy; that is, an increase in prices of energy and agricultural commodity leads to a rise in those food prices through the increased costs of producing those products. For example, increased price of corn results in a rise in production cost of meat, including beef, pork, and poultry, thereby leading to increased prices of the meat products. Similarly, a rise in corn price leads to increases in prices of dairy products since corn is a main feed for dairy cattle. In addition, U.S. energy policy under the Energy Independence and Security Act (EISA) of 2007 encouraged the U.S. energy industry to increase ethanol production from corn. In 2011, approximately 5 billion bushels of corn, a 40% of total corn produced in 2011 (Economic Research Service, 2011), was used to produce about 13.95 billion gal of ethanol in the U.S. (Energy Information Administration, 2011). This additional demand for corn raised the price of corn which also increases prices of other crops, including wheat and soybeans, since these crops compete with one another for crop land in the United States. Further, increased petroleum prices are likely to expand grains-based biofuel production, particularly under the EISA, and increase further pressures on commodity prices and thus food prices. In this light, we further infer that the recent strong linkage between energy policy and commodity/ food prices caused the food price inflation through changes in these three food prices. This finding thus explains why it is important to employ disaggregating price data for food in examining the 2006–08 food price inflation appropriately. Notice that the error correction terms for the four endogenous variables are negatively significant at the 5% significance level (except meat price). This implies that, with a shock to the U.S. food market, food prices tend to recover to their long-run equilibrium positions. This finding also confirms the cointegration relationships found earlier and the validity of the error-correction presentation. 6. Concluding remarks Although the 2006–08 price surge in the U.S. has been widely investigated, relatively little attention has been paid to empirical analysis of this issue. The main contribution of this study is, therefore, to examine the effects of such market factors as exchange rate, commodity price and energy price on U.S. prices of five food products in the framework of dynamic time series. For this purpose, we use the Johansen cointegration analysis and a vector error-correction (VEC) model with monthly data for 2001–2010. Cointegration analysis shows the presence of stable longrun relationships among the selected variables. The negatively significant coefficients of error-correction terms in the VEC model further validate the existence of equilibrium relationships among the variables. We also find that energy and commodity prices have significant impacts on U.S. food prices mainly through changes in prices of cereal/bakery, meats and diary; hence these three products have played key roles in creating the recent connection between energy and commodity/food markets in the U.S. Finally, exchange rate is generally found to be a significant factor influencing U.S. food prices.

276

J. Baek, W.W. Koo / Economic Modelling 42 (2014) 272–276

Table 4 Results of parsimonious VEC (PVEC) models. Δ ln CELt Δ ln CELt − 2 Δ ln CELt − 3 Δ ln CELt − 4 Δ ln DRYt − 1 Δ ln DRYt − 3 Δ ln DRYt − 4 Δ ln FRUt − 1 Δ ln FRUt − 2 Δ ln BEVt − 1 Δ ln BEVt − 2 Δ ln BEVt − 3 Δ ln BEVt − 4 Δ ln ERt Δ ln ERt − 2 Δ ln ERt − 4 Δ ln ENPt Δ ln ENPt − 4 Δ ln CPt Δ ln CPt − 1 Δ ln CPt − 2 Δ ln CPt − 3 Trend Constant ec1 ec2 ec3 a b

Δ ln METt

Δ ln DRYt

Δ ln FRUt

Δ ln BEVt

0.140 (1.71)a 0.553 (2.29)b −0.341 (−2.71)b −0.123 (3.15)b

0.247 (2.70)b

−0.256 (−2.31)b −0.266 (−2.25)b

0.105 (2.43)b 0.499 (5.76)b −0.380 (−2.68)b −0.244 (−4.29)b −0.109 (−1.91)a

−0.547 (−2.99)b

−0.115 (−2.45)b −0.234 (−2.44)b

0.396 (2.17)b

−0.146 (−1.58)

−0.228 (−3.38)b 0.098 (1.82)a 0.103 (2.56)b 0.135 (3.11)b

−0.206 (−2.52)b

−0.229 (−2.11)b −0.233 (−2.10)b

0.030 (2.83)b 0.045 (1.81)a 0.027 (1.85)a 0.032 (2.14) 0.047 (3.28)b

0.073 (4.11)b 0.063 (3.88)b 0.063 (3.52)b 0.058 (3.52)b

−0.207 (−7.00)b

0.377 (3.18)b 0.031 (2.79)b

−0.644 (−2.50)b

0.038 (2.38)b

−0.085 (−2.50)b

0.101 (2.84)b

−0.038 (−7.05)b

−0001 (−3.12)b −0.599 (−6.19)b −0.216 (−6.22)b

−0.001 (−0.01) 0.053 (3.18)b −0.028 (−3.22)b

Indicates significance at the 10% level. Indicates significance at the 5% level.

References Abbott, P.C., Hurt, C., Tyner, W.E., 2008. What's driving food prices? Farm Foundation Issues Report. Baek, J., Koo, W.W., 2010. Analyzing factors affecting U.S. food price inflation. Can. J. Agric. Econ. 58 (3), 303–320. Doornik, J., Hendry, D., 1994. PcFiml Version 8: Interactive Econometric Modeling of Dynamic System. Institute of Economics and Statistics, Oxford. Doornik, J., Hendry, D., 2001. PcGive 10: Empirical Econometric Modeling. Timberlake Consultants, London. EIA, 2011. Production and Consumption of Biofuel. U.S. Department of Energy. Elliot, G., Rothenberg, T.J., Stock, J.H., 1996. Efficient tests for an autoregressive unit root. Econometrica 64 (4), 813–836. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation, estimation, and testing. Econometrica 55, 251–276. ERS, 2011. Feed Data Base. U.S. Department of Agriculture. Gonzalo, J., 1994. Five alternative methods of estimating long-run equilibrium relationships. J. Econ. 60, 203–233. Hanrahan, C.E., 2008. Rising food prices and global food needs: the U.S. response. Congressional Research Service. Harris, R., Sollis, R., 2003. Applied Time Series Modeling and Forecasting. John Wiley and Sons, Chichester. Headey, D., Fan, S., 2008. Anatomy of a crisis: the causes and consequences of surging food prices. Agric. Econ. 39, 375–391. Hendry, D.F., 1995. Dynamic Econometrics. Oxford University Press, Oxford, UK. Herndon, C.W., 2008. The ethanolization of agriculture and the roles of agricultural economists. J. Agric. Appl. Econ. 40 (2), 403–414.

Johansen, S., 1995. Likelihood-Based Inference in Cointegrated Vector Autoregressive Models. Oxford University Press, Oxford, UK. Lambert, D.K., Miljkovic, D., 2010. The source of variability in U.S. food prices. J. Policy Model 32, 210–222. Lipsky, J., 2008. Commodity prices and global inflation. Remarks by the First Deputy Managing Director of the IMF at the Council on Foreign Relations. International Monetary Fund (IMF), New York City. Mitchell, D., 2008. A note on rising food prices. Policy Research Working Paper No. 4682. The World Bank. Ng, S., Peron, P., 2001. Lag length selection and the construction of unit root test with good size and power. Econometrica 69, 1519–1554. Rosegrant, M., 2008. Biofuels and grain prices: impacts and policy responses. Testimony for the U.S. Senate Committee on Homeland Security and Governmental Affairs. International Food Policy Research Institute. Schade, B., Wiesenthal, T., 2011. Biofuel: a model based assessment under uncertainty applying the Monte Carlo method. J. Policy Model 33, 92–126. Schnepf, R., 2008. High agricultural commodity prices: what are the issues? CRS Report for Congress. Congressional Research Service. Sieber, S., Dominguez, I.P., 2011. Impact assessment of agri-food policies in Europe: methods, tools and applications. J. Policy Model 33, 1–6. Sims, C., 1980. Macroeconomics and reality. Econometrica 48, 1–48. Trostle, R., 2008. Global agricultural supply and demand: factors contributing to the recent increase in food commodity prices. Economic Research Service Report WRS-0801. USDA. Wooldridge, J.M., 2006. Introductory Econometrics: A Modern Approach. South-Western College Publishing, Mason, OH.