ISSN 2280-6180 (print) © Firenze University Press ISSN 2280-6172 (online) www.fupress.com/bae Bio-based and Applied Economics 3(2): 93-117, 2014 DOI: 10.13128/BAE-12944 Agricultural and oil commodities: price transmission and market integration between US and Italy Franco rosa1,*, Michela VasciaVeo1, robert D. WeaVer2 1 Department of Food Science, Economic Unit, University of Udine, Italy 2 Department of Agricultural Economics, Sociology and Education, Penn State University, USA Abstract. Purpose of this article it to get some evidences of market interaction between United States and Italy using the time series analysis of spot prices spanning from Jan- uary 1999 to May 2012 for crude oil and three ag-commodities: wheat, corn and soy- bean. These crops have been selected for their relevance in ag-commodity exchanges between US and Italy markets. The integration between US and Italy agricultural mar- kets is hypothesized for the consistent volume of crop traded between these two coun- tries while the price transmission is related to the leading price signals of the CBT (Chicago Board of Trade). The integration between oil and ag-commodity markets is suggested both by the large use of energy intensive inputs, (fertilizer, seed, machinery) in production of these ag-commodities, and their use in biofuel production. The results suggest: a) for US market the evidence of market integration between crude oil and US ag-commodities; b) for Italy the integration with US ag-commodity markets and less evidence of integration with the oil market. These results are valuable information both for the agents and policy makers contributing to improve the information accuracy to predict the price movements used by marketing operators for their strategies and poli- cy makers to set up policies to re-establish conditions of market efficiency and allocate these ag-commodities in alternative market channels. Keywords. Agricultural commodity prices, time series analysis, cointegration, price transmission, market integration. JEL Codes. C22, Q11, Q13 1. Introduction Since 2006, the biofuel market in the United States has established a link between the prices of crude oil and grains such as corn characterized by the co-evolution of ag-com- modity prices (Abbott et al., 2008). The massive production of energy, mainly liquid fuels, from agricultural commodities has continued to strengthen these links between agricul- tural and energy markets and defined a dominant feature of current conditions in the agricultural sector. A resulting trend has been noted as a stronger dependency between crude oil, gasoline, and ag-commodity prices (Tyner and Taheripour, 2008). Brazil pro- * Corresponding author: franco.rosa@uniud.it. Assess overall Price Transmis- sion Full Research Article 94 F. Rosa, M. Vasciaveo, R.D. Weaver vides a useful example of a well-integrated, long-term agro-energy market with the oil and cane-bioenergy (ethanol and electricity) market integration to define an energy market in which oil and sugar cane prices exhibit strong comovement. There are other countries where these price links have become increasingly strong: the prices of wood pellets and, to a lesser extent, the wood chips in Austria have been following with a growing degree of correlation the prices for heating fuel in 2006 and 2007 (Schmidhuber, 2007). Prices of oil and agricultural commodities sharply rose in 2007, peaking in the sec- ond half of this year for some products and in the first half of 2008 for others. The causes of price spikes during 2007-2008 include some which were exogenous to the agriculture (macroeconomic), growth of food demand by the BRIC countries, and speculation on the oil prices and other factors (Piot-Lepetit and M’Barek, 2011; OECD, 2008). Others causes are due to physiological changes in market conditions from period to period, natu- ral shocks such as weather, pests or regulatory restrictions in domestic markets (FAO et al., 2011). These events raise new questions concerning ag-commodity price movement. First, have new trends been established for ag-commodity prices? Second, to what extent have recent price shock been temporary or permanent? Third, how have these shocks and possibly persistent trends affected comovement of commodity prices? Fourth, how has the nature of unanticipated shocks changed? Price transmission depends on the market efficiency and it may have limited by a number of causes, at least in the short-run. For example, these conditions may have included supply availability to the final consumer, demand circumscribed by bottlenecks in the distribution, logistic problems in transpor- tation, blending systems (E10), spatial arbitrage, and political constraints like the border measures and subsidies affecting the exchanges. Thus, the purpose of this paper is to con- trinbute to the understanding of the transmission of oil prices to agricultural markets. We expect that our results will contribute to better understand the nexus of agricultural and energy markets and its consequences for trading relations between US and Italy. We choose to study Italy and US as settngs that provide the basis for a test of the hypothesis of market integration and price transmission (Yang and Leatham, 1999). During the 2009- 2010 commercial campaign, Italy imported 60% of soft wheat, 87% of soybeans and 20% of maize from USA (Associazione Nazionale Cerealisti, 2011). The paper is organized as follows: section 2 reviews relevant literature; section 3 presents the time series methodol- ogy used; section 4 describes the price series used and provides a preliminary descriptive analysis of price correlations; section 5 presents results; and final section 6 provides con- clusions and suggestions for future research. 2. Literature review Market efficiency exists if price pass through between markets is complete such that they differ only by the transaction costs (Ardeni, 1989). Rational expectations and com- petitive storage theory supports the hypothesis that commodity stocks, expected prices and hauling costs are keydrives of commodity prices in equilibrium. Importantly, short- ages can induce substantial price shocks, Helmberger and Weaver (1977) and Helm- berger et al.(1982). It is widely acknowledged that the increased use of central commod- ity exchanges affects the extent and speed of transmission to market levels in response to leading price signals (Rapsomanikis et al., 2006). Deaton and Laroque (1991) have used 95Agricultural and oil commodities: price transmission and market integration the storage model to show that prices are not normally distributed, because the stock- holding behaviour by risk-averse agents generates an autoregressive pattern which is much stronger than what can be explained by the storage activity of risk-neutral agents. This market behaviour induces shocks in supply and demand that are correlated over time. On the supply side, this correlation is also induced by correlated shocks while on the demand side persistence in demand for working stocks induces intertemporal corre- lations. However, prices jumps may be induced by speculative demand by producers in response to anticipated stock-outs, see Helmberger and Weaver (1982). The decline of the dollar value (Trostle, 2008) and the speculation stemming from increased futures market volume are further factors contributing to recent agricultural commodity price movement (Robles et al., 2009). Further, supply side factors include relatively lower growth in agri- cultural production and yield, increases in energy prices that have induced increased farm production costs (Tyner and Taheripour, 2008; Sumner, 2009; von Braun et al., 2008), and climatic events (Trostle, 2008). The indirect price transmission through energy feed stock substitutes (e.g. sugar) led to increased demand for land and other limited resources diverting them from other agricultural crops, reducing their supply and driving up their prices (Schmidhuber, 2007). The growth of the biofuels production is an important driver of recent corn and oilseed demand growth (Gilbert, 2010; Zhang et al., 2010; Ciaian and d’Artis, 2011a,b). Biofuel policies, encouraging farmers to produce feed-stocks for biofuel, have increased the dependency between agricultural and energy prices (Yu et al. 2006; Campiche et al., 2007; Zhang and Reed 2008; Gilbert, 2010; Gohin and Chantret, 2010; Nazlioglu, 2011). Policy has also conditioned price transmission as trade restriction, import tariffs, export subsidies or taxes, and macroeconomic exchange rate policies have impacted the efficiency of arbitrage by insulating the domestic markets and hindering the transmission of price signals (Sarris, 2013). Esposti and Listorti (2013) investigate the role of the trade policy by analyzing the agricultural price transmission in presence of bubbles, using Ital- ian and international weekly spot (cash) price over the years 2006–2010. They observe that the bubble has had only a slight impact on the price spreads and that the temporary trade policy measures, when effective, have limited this impact. These interventions are responsible for excess demand or supply schedules of domestic commodity markets pos- sibly generating asymmetric price responses with nonlinear price adjustment (Quiroz and Soto, 1996; Sharma, 2002, Rapsomanikis et al., 2006, 2011; Harri et al., 2009; Gutierrez et al., 2013). The market integration between oil and ag-commodities has been explored using econometrically estimated demand and supply models based on partial or comput- able general equilibrium models (Lapan and Moschini, 2012; de Gorter and Just, 2010; Hertel et al., 2010). These models incorporate calibrated price elasticities and long-run assuptions to simulate dependence of agricultural commodity prices to oil shock prices. An alternative approach used to explore market efficiency is time series analysis to test market integration, price transmission, cointegration, asymmetric response, and causal nexus among markets (2011; Ciaian, 2011b; Gilbert, 2010; Gohin et al., 2010; Goodwin, 1992; Granger and Lee, 1989; Harri et al., 2009; Minot, 2011; Rapsomanikis et al., 2006; Sarris, 2013; Zhang et al., 2008). Here, we use time series analysis to test the hypothesis of market integration between US and Italy and in a broader sense to verify the efficiency of agricultural markets for some ag-commodities selected for their importance in trade 96 F. Rosa, M. Vasciaveo, R.D. Weaver (Tomek and Myers, 1993; Rosa, 1999; Nazlioglu, 2011; Rosa and Vasciaveo, 2012; Esposti and Listorti, 2013; Gutierrez et al., 2013). Whether agricultural and food commodity prices are unjustifiably volatile and unre- lated to the market fundamentals has been extensively considered by Balcombe (2009, 2013) and Gilbert (2010). Persistence of these effects on prices has been considered by Serra and Gil (2012), Algieri (2012), Chatellier (2011), Balcombe (2009), Listorti and Esposti (2012), Rosa and Vasciaveo (2012). In presence of excess volatility beyond that which can be accounted for by changes in market fundamentals, the prices may be driv- en by fad or speculative bubbles, and commodity prices may become inefficient signals for resource allocation (Gilbert and Morgan, 2010; Balcombe and Fraser, 2013). Time- varying volatility of commodity price series leads to autocorrelation patterns in the con- ditional variance of price innovations where the variance is conditional on an informa- tion set available at the time forecasts are being formed. Engle (1982) has termed this conditional heteroscedasticity and developed the autoregressive conditional changes in economic fundamentals. Time-varying volatility in commodity prices has the same gen- eral effect on statistical inference as any other form of heteroscedasticity causing a loss of efficiency and estimated standard errors may be biased (Engle, 1982). Excess kurtosis causes also problems whenever inference requires a particular distributional assumption on the disturbance terms. Although the normal distribution is typically chosen, the actu- al distribution of commodity prices appears to have fatter tails than the normal. This can be a particular problem in maximum likelihood estimation of commodity market mod- els. (Myers, 1992). 3. Methodology Our analysis uses cointegration and vector error correction models to explore spatial market relationships and price transmission (Rapsomanikis et al., 2006). The analysis is performed in the following three steps: 1) we determine whether univariate price series are nonstationary or I(1) (if both price series are not I(1), they cannot be cointegrated); 2) if they are both stationary or I(0), we examine their dynamic interrelationship (Leucci et al., 2013) with the vector autoregressive (VAR) model; 3) if the series are both I(1), the null hypothesis that they are not cointegrated is tested with the Johansen procedure; 4) and if the results suggest evidence of long-run relationship between variables, we estimate vector error correction models (VECM). The scheme of this approach is reported in Figure 1. 3.1. Unit root test The first step of the analysis is to test for stationarity and whether each series is inte- grated with the same order. We employ several unit root tests to consider robustness of our inference including: augmented Dickey-Fuller (1979) [ADF], Phillips-Perron [PP] (1988) and Kwaitkowski-Phillips-Schmidt-Shin [KPSS] (1992). These tests (except for KPSS) examine a null hypothesis of a unit root against the alternative of I(0) stationarity. Stationarity is the null hypothesis for KPSS. If the ADF statistic has a negative sign, as the absolute value increases the level of confidence for rejection of the hypothesis of unit root 97Agricultural and oil commodities: price transmission and market integration is increased. The usefulness of ADF is limited in the presence of an explosive root (Bal- combe and Fraser 2013). The ADF test follows from µ β α ε= + + + +∑− − = y t y c yt t i t i t i k 1 1 (1) where μ is the constant, β is the coefficient on the time trend, k is the lag order of the autoregressive process, Dyt-i is the lagged difference of y whose magnitude is measured by c and ε is the error. The unit root test is carried out under the null hypothesis α = 0 against the alternative hypothesis of α < 0; nonstationarity is rejected when α is signifi- cantly different from 1. A common problem with conventional unit root tests is that they do not allow for any break in the data generation process. If a structural break is hypothesized, the conven- tional ADF test is biased toward the acceptance of the null resulting in a dramatic loss of power. Further, the ADF allows for higher-order autoregressive processes including lags of the order k that have to be pre-determined. Assuming the break time to be exogenous, Perron (1989) suggests that the power to reject a unit root decreases when the stationary alternative is true and the structural break is ignored. Following Perron’s characterization of the form of structural break, Zivot and Andrews (ZA, 1992) formulate three different characterizations of the trend break: i) model A, “the crash model”, allows the break in the intercept; ii) model B, “the changing growth model”, allows for a one-time change in the slope of the trend function with the two segments joined at the break point; and iii) model C, “the mixed model”, combines simultaneously the one-time changes in the level Figure 1. Scheme of the price analysis to test the market integration and price transmission condi- tions.                   Assess overall Price Transmission reject accept If not the same Johansen test H0: no cointegration Specify and estimate VECM; assess dynamic, speed of adjustment If both I(0) If both I(1) Test for the order of integration (ADF, PP, KPSS, ZA) No cointegration Estimate VAR on levels Estimate VAR on differences Source: Own elaboration of the scheme proposed by Rapsomanikis et al. (2006). 98 F. Rosa, M. Vasciaveo, R.D. Weaver with the slope of the trend function of the series2. The aim of this procedure is to sequen- tially examine evidence of breakpoint candidates and select the one that gives most weight to the trend stationary alternative. Hence, to test for a unit root against the alternative of a one-time structural break, Zivot and Andrews propose the following regression equations (derived from equation 1) corresponding to the three cases noted above: µ β α γ ε= + + + + +∑− − = y t y DU c yt t t i t i t i k 1 1 (Model A) µ β α θ ε= + + + + +∑− − = y t y DT c yt t t i t i t i k 1 1 (Model B) µ β α θ γ ε= + + + + + +∑− − = y t y DT DU c yt t t t i t i t i k 1 1 (Model C) where DUt is an indicator dummy variable for a mean shift occurring at each possible break-date while DTt is the corresponding trend shift variable. The ZA unit root test is an endogenous structural break test with unknown timing in the individual series that uses the full sample and different dummy variables for each plausible break date. The break time is selected where the t-statistic from the ADF test of unit root is at a minimum (most negative), then a break date is chosen where the evidence is least favourable for the unit root null. The null hypothesis is that the series is integrated without an exogenous struc- tural break against the alternative that the series can be represented by a trend-stationary process with only one break point occurring at some unknown time. The ZA test is a vari- ation of PP’s original test with the endogenous implementation of structural breaks in the analysis: the date of the break is determined with the t-statistics test of the unit root, with respect to the criteria of minimum values. The ZA test regards every point as a potential break-date and runs a regression for every possible break-date sequentially. 3.2. Cointegration analysis: the Johansen test Cointegration analysis examines whether two series are linked to form an equilibrium relationship. The intuition of cointegration is that two price series cannot evolve in oppo- site directions for very long time if they are cointegrated. This condition is examined by estimation of the static regression between I(1) variables: µ α ε= + +y xt t t (2) 2 For the three models, Zivot and Andrews estimate the testing equation by allowing the break to take place beginning successively in the second, third, fourth, and so on, observation, up to observation T - l, where T stands for the total sample size used in the estimation and l are the lags. The alternative specifications are esti- mated by OLS, and the length of the lag (k) for the difference terms is determined by starting at k = 8, and working backwards until significant values are identified. The estimate of the breakpoint is that particular obser- vation corresponding to the minimum t-value for the one period lagged term, for each model A, B, and C. In order to test the unit root hypothesis, this minimum t-value is compared with a set of asymptotic critical values from the work of Zivot and Andrews (1992). 99Agricultural and oil commodities: price transmission and market integration where xt is a vector of independent variables. The system is cointegrated if the errors εt are I(0). In this case, equation (2) may be interpreted as a long-run equilibrium condi- tion of the process y(t). The Johansen cointegration test uses the vector autoregressive (VAR) model with k lags assuming the variables are I (1) written in error-correction form (Johansen, 1995). To determine the presence of cointegration between variables, the lag length (k) is determined with the Schwarz Bayesian Criterion (SBC or BIC test, Schwarz, 1978) and then the cointegration rank (r) is estimated. 3.3. Gregory Hansen test Gregory and Hansen (1996) propose cointegration tests which are an extension of the Zivot and Andrew (1992) unit root tests to incorporate a single structural break in the underlying cointegrating relationship. The GH test extends the ADF*, Zt* and Zα* type tests designed to test the null of no cointegration against the alternative of cointegration in presence of a single structural break. These authors consider three variations of equation (2) that includes dummies for the structural change: Model C: Level Shift: µ θ α ε= + + +y DU xt t t t (3a) Model C/T: Level Shift with Trend µ θ β α ε= + + + +y DU t xt t t t (3b) Model C/S: Regime Shift (Intercept and Slope coefficients change) µ θ α α ε= + + + +y DU x DU xt t t t t t1 2 (3c) where t is time subscript, ε is an error term and DU is a dummy variable. Model C entails a level shift in the equilibrium relationship, model C/T adds a trend component to the previous model whilst model C/S deals with the regime shift by adding a change in the slope coefficients. The structural change is endogenously determined by the smallest value (the largest negative value) of the cointegration test statistics across all possible break points. 4. Data and descriptive analysis To perform the empirical analysis, we use weekly spot prices3 of three ag-commodi- ties and the oil prices for the period spanning from 1999 to 2012; this frequency has been used to capture more accurately the price movements and linkages (Nazlioglu, 2011). Soft wheat, maize and soybeans are selected for their importance in ag-commodity trade between US and Italy: wheat is highly energy intensive and is a key product for human nutrition while corn and soybean are the most important ag-commodities for animal feeding and biofuel. Table 1 reports the list of variables used in the analysis. 3 Spot prices are used because most of the transactions in Italy are made in these markets. For more details see Rosa and Vasciaveo (2012) 100 F. Rosa, M. Vasciaveo, R.D. Weaver Table 1. Description of ag-commodity price series. Variable Description Source Italian corn price (cit) Weekly average of spot prices in €/ton of national hybrid corn-market at the origin (cit) DATIMA provided by ISMEA1 soybean price (sit) Weekly average of spot prices in €/ton of soybeans with 14% of moisture--market at the origin (sit) DATIMA provided by ISMEA wheat price (wit) Weekly average of spot prices in €/ton of good mercantile wheat--market at the origin (wit) DATIMA provided by ISMEA US corn price (cus) Weekly average of spot prices converted in €/ton of US yellow no. 2 corn at the Gulf of Mexico (cus) FAO International Commodity Price Database soybean price (sus) Weekly average of spot prices converted in €/ ton of US no. 1 yellow soybean at the Gulf of Mexico (sus) FAO International Commodity Price Database wheat price (wus) Weekly average of spot prices converted in €/ ton of US no. 2 soft red winter wheat at the Gulf of Mexico (wus) FAO International Commodity Price Database Oil price oil Weekly spot prices of Brent crude oil converted in €/barrel (oil) US Energy Information Administration (EIA, 2012) DATIMA is a collection of statistical databases including Italian agricultural market data and foreign trade; ISMEA is the Italian agri-food market Institute To be comparable, the US agricultural and oil commodities price series quoted in $ are converted into euro currency, using the official $/€ exchange rate4 and converted to natural logarithms. Visual inspection of the price series reported in Figure 2 suggests a nonlinear trend component exists for each of the series. Figure 2 also suggests a relative- ly steady price period existed during 2005, followed by wider fluctuations to the end of 2008, and wider fluctuations in the final stage for all commodities prices. The wider oil price variability does not seem to affect the fluctuation of the ag-commodity prices. These observations motivate the need to examine the existence of structured breaks that define sub-samples to examine better the effect of volatility. 4.1. Testing for presence of bubbles Figure 2 also suggests the possible presence of bubbles. During the period 2006-10, sub-periods of explosive price are apparent as also noted by Huchet-Bourdon (2011). Bub- bles have been noted as occurring in 2006 when levels of agricultural and food prices increased sharply followed by a collapse, as well as between 2006 and 2008, 2008-2010 and more recently in autumn 2012. (Phillips, P.C.B., Shi S., Yu J., 2012). A number of 4 Available at http://www.statistics.dnb.nl/index.cgi?lang=uk&todo=Koersen 101Agricultural and oil commodities: price transmission and market integration tests have been used to identify the sharp increases and declines in prices also known as explosive bubbles. We followed Phillips, Shi and Yu (2012), PSY hereafter, who developed a method to test for explosive behavior and date the origin and collapse of bubbles. This method is used to check for presence of multiple bubbles of the PCB5 type in a sample data. (Phillips, Shi and Yu, 2012; Gilbert and Morgan, 2010). We apply the more recent generalized sup-augmented Dickey-Fuller (GSADF) test proposed by PSY, for explosive bubbles with variable windows widths in the recursive regression: α β Φ ε= + + ∑ +− = −Y r r r rY rr Y, ,t t i t t1 2 1 2 1 1 1 2 1 eτ ~ N (0,s 2) (4) Here the null hypothesis of nonstationarity (H0: br1,r2 = 0 ) is tested against the alter- native hypothesis H1: br1,r2 > 0 which implies explosive behaviour. Our results reveal evi- dence of bubble behaviour for wheat, rice soybean oil and rapeseed oil price series during the first month of 2008. Beyond fundamentals, the GSDAF test does not provide sufficient evidence to infer whether these bubbles are the result of a trend and may persist in the ag- commodity market. We have tested with the PSY test the series used for this analysis and results are reported in Table 2 for different length of time series and window widths. The analysis provides insights to price behaviour during the examined periods and their consequences for the analysis. The results of Table 2 do not support the hypothesis that bubbles occurred during the sample period with an exception for sit with window width 0.1, however for window width of 0.4 the values are substanitially below the critical threshold at 99 and 95% critical values. The period 2006:1-2008:52 is also examined for bubbles as past literature reports more evidence of price volatility during this period (Gil- 5 PCB is the acronym for price collapsing speculative bubbles that are nonlinear processes (Evans,1991 explosive during the phase of bubble eruption, but they may be stationary over the whole sample period. Figure 2. Index of current prices of some agri commodities and oil prices  Source: own elaborations. cit, wit, sit, cus, wus, sus: €/ton; for oil: €/barrel; Jan 04, 2002= 100. 102 F. Rosa, M. Vasciaveo, R.D. Weaver Table 2. Results of the GSADF recursive test with one lag. Series Lag Window Nr of observations Test statistics GSADF Critical values 99 95 90 cus 1 0.1 168 1.1787 2.9822 2.2381 2.0277 cus 1 0.4 168 1.1788 1.9957 1.4001 1.1322 cit 1 0.1 168 2.0310 2.9822 2.2381 2.0277 cit 1 0.4 168 -0.0909 1.9957 1.4001 1.1322 sit 1 0.1 168 3.5970 2.9822 2.2328 2.0278 sit 1 0.4 168 0.2132 1.9975 1.4001 1.1322 sus 1 0.1 168 1.8955 2.9822 2.2328 2.0278 sus 1 0.4 168 -0.1245 1.9957 1.4002 1.1392 wit 1 0.1 168 2.1124 2.9822 2.2328 2.0278 wit 1 0.4 168 -0.5607 1.9957 1.4001 1.1392 wus 1 0.1 168 0.9183 2.9822 2.2328 2.0277 wus 1 0.4 168 0.0573 1.9957 1.4002 1.1392 oil 1 0.1 168 1.2566 2.9822 2.2328 2.0278 oil 1 0.4 168 0.1011 1.9976 1.4002 1.1392 Time series 1999:1-2012:52. Shaded values are above the critical values. Table 3. Results of GSADF recursive test with one lag. Series Lag Window Nr of observations Test statistics GSADF Critical values 99 95 90 cus 1 0.1 156 2.5989 3.0681 2.2972 1.9947 cus 1 0.4 156 2.5989 2.0938 1.4633 1.1652 cit 1 0.1 156 2.1584 3.0681 2.2972 1.9947 cit 1 0.4 156 3.4991 2.1114 1.4575 1.1524 sit 1 0.1 156 2.3889 3.1413 2.2660 1.9882 sit 1 0.4 156 2.3889 2.0939 1.4633 1.1652 sus 1 0.1 156 1.9957 3.1413 2.2660 1.9882 sus 1 0.4 156 2.1787 3.0682 2.2972 1.9947 wit 1 0.1 156 4.7593 3.1413 2.2660 1.9882 wit 1 0.4 156 4.7593 2.0939 1.4633 1.1652 wus 1 0.1 156 3.1313 3.1413 2.2660 1.9882 wus 1 0.4 156 2.1584 2.0938 1.4633 1.1652 oil 1 0.1 156 3.0871 3.1413 2.2660 1.9882 oil 1 0.4 156 3.0871 3.0682 2.2972 1.9947 Time series 2006:1-2008:52. Shaded values are above the critical values. 103Agricultural and oil commodities: price transmission and market integration bert, 2010, Rosa and Vasciaveo, 2012). Results are reported in Table 3. For the series cus, cit, sit the test values are above the critical values, using the windows width 0.4 but below critical values using the windows width 0.1; a possible explanation is that the smaller win- dow width includes price values less volatile compared to the larger window. For wheat, the results are above the critical values for both window widths. These results are more difficult to explain because in contrast with wus and other ag-commodity in Italy (Areal et al., 2013), the 2007-2008 US wheat market experienced reduced stock levels. Reduced production levels, in conjunction with very low carryover stocks, resulted in an extremely tight global market and is likely to have affected the expectations of market operators in Italy. Another possible explanation is the interaction between spot and future markets. Given an high share of wheat open interest held by noncommercial traders in an already tight market, the demand for long-term wheat future contracts may have affected spot prices and generated bubbles due to strengthening inventory demand. The tests performed by Areal et al. (2013) have revealed weak presence of multiple bubbles in the food prices finding that when present, bubbles have been quite short, continuing between two and fourth months before collapsing. 4.2. Stationary and structural break tests We next consider the order of integration and testing the stationary condition6 with the unit root test for levels and first differences. A number of tests are used with results reported in Table 4. We find all variables are integrated of first order I(1)7 Table 5 reports the results of the Zivot-Andrews test with one break. Minimum ZA statistics for the levels of the variables reject the hypothesis of structured breaks imply- ing the evidence of the unit root tests may be accepted with the exception of oil and sus. Allowing for the identified breaks and a deterministic trend for these products, the null hypothesis of unit root process is rejected. The test is performed in three versions report- ed in previous section 3. A structural break is found in the US soybeans series, the esti- mated date is July 2004 (week 29) fitted with drift (model A) and a change in the trend slope and drift (model C). The oil series is stationary with a break in October 2008 (week 40) and change in trend slope and drift. A possible explanation of the 2004 structural break is the massive growth in biofuel production in the US starting with 2004. The suc- cessive 2008 break in the oil series corresponds to the oil price peak. The other price series are found to be I(1) confirming the results of traditional unit root tests. While sus and oil price series are stationary in model C, this condition is not so evident in model A (only for oil) or in model B; for this reason it is conservatively assumed that all the variables are integrated of order one I(1). 4.3. Preliminary evidence of comovement among Italian and US ag-commodity prices The graphic evidence of comovement in levels for the historical price series (Figure 2) suggests strong comovement with a moderate deviations from cyclical long-run move- 6 This condition implies that the mean, variance and autocorrelation of the series do not change over time; 7 The differences of the alternative tests used in this analysis are not contradictory about stationary condition. 104 F. Rosa, M. Vasciaveo, R.D. Weaver ments, non-linear trend components, and wider range fluctuation in latest periods. Oil price patterns could have affected the agricultural markets during the period (1999-2012) and changed the price dynamics generated by market fundamentals (Headey and Fan, Table 4. Unit root test results. Levels First differences ADF PP KPSS ADF PP KPSS Intercept cus -0.31 -1.28 1.93* -13.84* -33.22* 0.14 sus -1.42 -1.35 2.20* -30.21* -30.99* 0.08 wus -2.23 -2.15 1.79* -28.81* -28.75* 0.03 cit -1.77 -2.08 1.18* -16.35* -16.41* 0.05 sit -1.28 -1.02 2.22* -14.74* -21.10* 0.06 wit -1.48 -1.77 1.03* -18.51* -19.42* 0.06 oil -1.23 -1.17 2.48* -21.42* -21.42* 0.04 Trend & intercept cus -1.59 -2.91 0.44* -13.88* -33.66* 0.05 sus -3.06 -3.08 0.36* -30.20* -31.02* 0.03 wus -3.36 -3.29^ 0.12^ -28.80* -28.73* 0.02 cit -2.47 -2.76 0.12^ -16.35* -16.37* 0.03 sit -2.73 -2.46 0.17° -14.73* -21.09* 0.04 wit -2.07 -2.36 0.16° -18.50* -19.41* 0.04 oil -2.78 -2.71 0.14^ -21.41* -21.41* 0.04 Schwarz Information Criterion to determine the optimal lags for ADF test; the bandwidth for PP and KPSS tests is selected with Newey-West using Bartlett kernel (by default). */°/^ denote statistical sig- nificance at 1, 5 and 10% respectively.   Table 5. Zivot Andrews one break test. Model A Change in drift Model B Change in trend Model C Change in drift and trend Critical value 1% 5% 10% Model A -5.34 -4.80 -4.58 Model B -4.93 -4.42 -4.11 Model C -5.57 -5.08 -4.82 The asymptotic critical value for Zivot and Andrews (1992) test at different levels of significance cus -3.63 -3.10 -3.42 sus -4.98** (2004: w29) -4.15 -5.48*** (2004: w29) wus -3.78 -3.50 -3.87 cit -3.20 -2.83 -3.82 sit -4.03 -3.02 -4.03 wit -3.18 -2.81 -3.26 oil -3.27 -3.30 -5.16** (2008: w40) ***/** denote statistical significance at 1% and 5% respectively; break date in brackets. 105Agricultural and oil commodities: price transmission and market integration 2008; OECD, 2008). The price correlation between two variables, here given by the prices of ag-commodities is measured linearly with the Pearson correlation coefficient (r). Given nonstationarity of the underlying series of price levels, we examine correlation of station- ary first differences.8 The correlation coefficients between the oil price and each ag-com- modity price are computed over the whole sample and for sub- samples with critical 5% values; the results are reported in Table 6. For the entire period (1999-2012), we find the Pearson correlation in price differenc- es are small with values that vary in the range between 0.06 (dwit- dsus) and 0.37 (dcus- dwus) with two correlations below the critical 5% value: dsus-dcit and dsus-dwit. These results are consistent with the hypothesis of innovations in prices commove, though such comovement is small in magnitude. We also find that the innovations in Italian ag-com- modity prices are not influenced by those of the oil prices, though we find evidence of comovement in innovations in oil and US ag-commodity prices. For the subsample period 1999-2004, we find the Pearson correlation values vary in the range between -0.03 (dwus and dcit) and 0.53 (dwus-dcus). However, compared to the full sample period we find more coefficients (9) are below the critical value indicating no comovement. Across coun- tries, we find evidence of comovement between dwit and dcit and the US ag commodity prices, as well as with oil prices. For Italian prices, we find that only for the (dcit, dwit) pair can we reject the hypothesis of correlation. For the period 2004-2008, the estimated Pear- son correlation values vary in the range between -0.02 (dcus-dwit) and 0.40 (dsus-dcus). Critical values indicate there are seven correlation coefficients below the critical value, in each case for pairs of Italian and US prices (dcit and dwit with US ag-commodity prices). For the period 2008-2012, we find the Pearson correlation values vary in the range between 0.11 (dsus and dwit) and 0.40 (dwit-dcit; dwus-dcus; doil-dsit). With respect to evidence of comovement across oil prices and commodity prices for entire period and sub-periods we find evidence of weak comovement of oil and ag-commodity prices that is weakest dur- ing the sub-period 1999-2004 and stronger in the later sub-periods; correlation is smaller with each of the Italian ag commodities compared to US. The analysis suggests that for sub-periods energy and agricultural market price comovement seems to become stronger in more recent periods. The general conclusion is that the US ag commodity price innova- tions appear to be more correlated with those of oil prices while little evidence of such cor- relation is found between Italian ag commodity price innovations and oil price innovations. 4.4 Market integration The cointegration analysis is used to examine the comovements between oil and agri- commodity prices (Johansen and Juselius, 1990). Some of the series checked for unit roots are found to be stationary with a breaking trend, then the Johansen and Gregory- Hansen (GH) tests are used to check for the presence of cointegration for all pairwise price series that accounts for a break in the cointegration relationship. 8 In a bivariate time series characterized by nonstationarity, correlation of in nonstationary levels is meaning- less as by definition the series are not generated by population data generation processes that are invariant with respect to time. In the absence of a population counterpart, correlation would result in spurious inference (Johansen, 1989). 106 F. Rosa, M. Vasciaveo, R.D. Weaver Table 6. Pearson correlation coefficients for first difference series of ag-commodity prices. Series 99/01/08 - 12/05/25; two tail critical value 5% = 0,0742*; n = 699 d_cit d_wit d_sit d_cus d_wus d_sus d_oil 1.0000 0.2597 0.2037 0.1153 0.1420 0.0706 0.1130 d_cit 1.0000 0.1544 0.0827 0.1907 0.0648 0.0751 d_wit 1.0000 0.2329 0.1603 0.2513 0.2096 d_sit 1.0000 0.3655 0.2150 0.2135 d_cus 1.0000 0.2186 0.2149 d_wus 1.0000 0.2151 d_sus 1.0000 d_oil Series 99/01/08 - 04/07/16; two tail critical value 5% = 0,1154*; n = 289 d_cit d_wit d_sit d_cus d_wus d_sus d_oil 1.0000 -0.0451 0.2196 -0.0258 -0.0333 0.0346 -0.0831 d_cit 1.0000 0.1664 0.0463 0.0921 0.0380 0.0411 d_wit 1.0000 0.1564 0.1312 0.1860 0.0508 d_sit 1.0000 0.5333 0.3176 0.1840 d_cus 1.0000 0.2001 0.2180 d_wus 1.0000 0.1676 d_sus 1.0000 d_oil Series 04/07/23 - 08/10/03; two tail Critical value 5% = 0,1323*; n = 220. d_cit d_wit d_sit d_cus d_wus d_sus d_oil 1.0000 0.4456 0.1257 -0.0064 0.1300 0.0769 0.1330 d_cit 1.0000 0.0302 -0.0227 0.1644 0.0577 -0.0529 d_wit 1.0000 0.2964 0.1082 0.2573 0.1080 d_sit 1.0000 0.3518 0.3994 0.2014 d_cus 1.0000 0.3007 0.1356 d_wus 1.0000 0.2280 d_sus 1.0000 d_oil Series 08/10/10 - 12/05/25; two tail Critical value 5% = 0,1424*; n = 190. d_cit d_wit d_sit d_cus d_wus d_sus d_oil 1.0000 0.3850 0.2688 0.2402 0.3013 0.1207 0.2538 d_cit 1.0000 0.2399 0.1363 0.2978 0.1076 0.1949 d_wit 1.0000 0.2550 0.2356 0.3585 0.3905 d_sit 1.0000 0.3943 0.1662 0.2519 d_cus 1.0000 0.1970 0.3013 d_wus 1.0000 0.2893 d_sus 1.0000 d_oil * Shaded values are below the critical values. Source: Author computation. 107Agricultural and oil commodities: price transmission and market integration Table 7. Johansen trace test for cointegration in price level: 1999:w1-2012:w21. cus sus wus cit sit wit cus sus 18.87 wus 27.79** 26.44** cit 26.60** 27.15** 30.74** sit 12.74 25.91** 32.52*** 29.56** wit 23.93* 29.33** 42.17*** 31.24*** 28.69** oil 15.36 17.73 21.16 23.11 16.38 19.50 The critical values are 31.15, 25.87 and 23.34 for 1%, 5% and 10% respectively (MacKinnon-Haug- Michelis, 1999, critical values). ***, ** and *denote statistical significance at 1%, 5% and 10% level of significance, respectively. Null is no cointegration. The results of the trace test reported on Table 7 indicate that all ag-commodity prices are pair wise cointegrated with the exception of sus and cus and sit and cus. These results suggest that Italian and US agricultural markets are integrated while there is not statistical evidence that oil market affects the ag-commodity markets in US or in Italy. An absence of cointegra- tion is found for cus and sus. This is consistent with a dominance of ethanol demand for corn that drives a wedge between the two markets. Table 8 reports the results of the cointegra- tion test for the first sub-period and shows the null hypothesis of no cointegration is rejected for US corn and wheat with Italian commodity prices. US soybeans is cointegrated with the Italian ag-commodities, confirming the results obtained during the entire period of observa- tion and is also cointegrated with the oil prices; US corn is cointegrated with US soybean and wheat. The Italian ag-commodity markets appear to be cointegrated with the US soybean market, and a significant cointegration exists between IT soybeans and oil prices. Table 8. Johansen trace test for cointegration: period 1999:w1-2004:w29. cus sus wus cit sit wit cus sus 26.10** wus 30.80** 25.95** cit 21.47 34.33*** 20.40 sit 17.32 39.44*** 15.48 18.18 wit 17.24 34.24*** 17.60 23.18 17.75 oil 21.37 30.75** 18.89 22.44 24.10* 17.03 ***, ** and *denote statistical significance at 1%, 5% and 10% level of significance, respectively. The results reported in Table 9 for the sub-period (2004-2008) suggest a different market condition: Italian corn, soybeans and wheat prices are cointegrated with their cor- responding Italian prices though not in all cases with US prices and never with the oil 108 F. Rosa, M. Vasciaveo, R.D. Weaver prices. The pairwise cointegrations between soybean and corn in the Italian markets sug- gest also that the price movements of these two major ag-commodity markets are moving together. The first observation inherent in this period is that US and Italian ag-commodity prices move together and US commodity prices are not in general cointegrated with Ital- ian ag-commodity prices. We also find an absence of cointegration between US commod- ity price pairs. Table 9. Johansen trace test for cointegration: period 2004:w30-2008:w40. cus sus wus cit sit wit cus sus 15.60 wus 19.70 12.93 cit 21.70* 23.18 12.69 sit 13.71 24.74* 34.77*** 31.06** wit 23.95* 24.37* 30.09 16.53 30.44** oil 15.32 19.03 12.66 10.85 15.24 17.36 ***, ** and *denote statistical significance at 1%, 5% and 10% level of significance, respectively. For the last sub-period presented in Table 10, the situation is consistently different. The null hypothesis of absence of cointegration with oil is rejected for all products except cit. These findings are consistent with the increasing use of ag-commodities in biofuel production that has generated more interdependence between oil and ag-commodity mar- kets. We also find US prices to be cointegrated except between soy and wheat. US corn is cointegrated with Italian corn and with wheat while Italian soy appears not to be cointe- grated with US corn prices. For each product, US and Italian markets appear cointegrated. No evidence of cointegration across Italian product prices is found. Table 10. Johansen trace test for cointegration: 2008:w41-2012:w21. cus sus wus cit sit wit Cus Sus 27.75** Wus 29.53** 16.70 Cit 24.50* 9.51 14.86 Sit 18.05 47.26*** 11.59 9.20 Wit 42.80*** 19.27 62.59*** 13.35 9.62 Oil 28.80** 36.68*** 36.22*** 20.46 32.39*** 35.91*** ***, ** and *denote statistical significance at 1%, 5% and 10% level of significance, respectively. 109Agricultural and oil commodities: price transmission and market integration These results are supported by the observations of other authors. Campiche et al. (2007) found that while there is no evidence of cointegration among the variables for the period 2003-2005, corn and soybean prices are cointegrated with crude oil prices in the next period 2006-2007. Harri et al. (2009) found robust evidence of cointegration between crude oil and corn, soybeans starting in April 2006. Nazlioglu (2011) examined the coin- tegration between oil and three key ag-commodity prices and found evidence of corn and soybean price cointegration with the oil prices during the period 2008-2010. Structural break timing has been determined a priori in the previous papers. Here, we examine evi- dence of breaks within the context of our cointegration models, using Gregory-Hansen tests based on equations (3a – 3c) that allow for identification of structural break for the entire period 1999-2012. Results are reported in Table 11. Table 11. G-H cointegration test with one structural break9 for US ag-commodities and oil: 1999:w1-2012:w21. cus-oil sus-oil wus-oil ADF* C -3.45 -4.21 -4.23 C/T -3.84 -5.38** (2004: w34) -4.26 C/S -4.06 -4.94* (2008: w10) -4.65 Zt* C -4.69** (2010: w19) -4.44* (2007: w39) -3.81 C/T -5.52*** (2004: w22) -5.72*** (2004: w33) -3.85 C/S -5.72*** (2004: w37) -5.20** (2007: w39) -4.03 Zα* C -42.20** (2010: w19) -40.26** (2007: w39) -28.61 C/T -56.96** (2004: w22) -61.15*** (2004: w33) -28.91 C/S -62.39*** (2004: w37) -52.52** (2007: w39) -31.49 ***/**/* statistical significance at 1%, 5% and 10% level of significance, respectively; break dates in brackets. For oil and cus price, the ADF* test did not reject the null hypothesis of no cointegra- tion with model in the versions C, C/T and C/S whereas Zt* and Zα* type test results sug- gest the rejection of the null for each of the three models. Significance of structural breaks were found for May and September 2004 (week 22 and 37) and May 2010 (week 19). For soybeans and oil, the three tests do not support the rejection of the null hypothesis of cointegration and structural break evidence is found for August 2004 (week 33); besides, Zt* and Zα* fail to reject the null in the regime shift model with a break in July 2007 (week 39). For the long-run relationship between wheat and brent prices, no evidence was found 9 Model C: Level shift, Model C/T: level shift with trend, Model C/S: Regime shift. Null hypothesis: no cointegra- tion. For ADF* and Zt* tests, critical values in Model C are: -5.13 at 1%, -4.61 at 5% and -4.34 at 10%; in Model C/T:-5.45 at 1%, -4.99 at 5% and -4.72 at 10%; in Model C/S: -5.47 at 1%, -4.95 at 5% and -4.68 at 10%. Critical values for Zα* test are -50.07, 40.48, -36.19 respectively at 1, 5 and 10% in Model C; -57.28, -47.96 and –43.22 at 1, 5 and 10% in Model C/T; -57.17, -47.04 and -41.85 at 1, 5 and 10% in Model C/S. The optimal lag length for ADF* test was selected by Akaike information criterion. 110 F. Rosa, M. Vasciaveo, R.D. Weaver for cointegration; a possible explanation is the wheat prices are less dependent on energy prices because only a limited quantity is used in the ethanol production. Table 12. G-H cointegration test with one structural break for IT ag-commodities and oil. cit-oil sit-oil wit-oil ADF* C -3.77 -4.11 -3.75 C/T -4.01 -4.13 -3.84 C/S -4.27 -4.20 -4.27 Zt* C -3.58 -3.69 -3.36 C/T -3.58 -3.71 -3.33 C/S -3.73 -3.84 -3.59 Zα* C -24.72 -27.95 -21.93 C/T -24.63 -28.21 -22.15 C/S -28.14 -29.42 -25.67 ***/**/* statistical significance at 1%, 5% and 10% level of significance, respectively; break dates in brackets. The results of Gregory Hansen tests reported in Table 12 do not in any case support the inference of cointegration between the crude oil and the Italian ag-commodity prices. The results of Table 13 suggest the cointegration between the Italy and US ag-commodity markets. These findings are consistent with those obtained by running the cointegration test without structural breaks. Results appear to be more robust for wheat and soybean commodities (confirmed by all the three tests). For corn, evidence of cointegration follows only from the Zt* and Zα* tests. Table 13. Cointegration test with one structural break between Italian and US ag-commodities. cit-cus sit-sus wit-wus ADF* C -3.89 -4.83** (2010: w19) -5.29** (2001: w34) C/T -4.21 -4.96* (2010: w19) -5.26** (2001: w14) C/S -4.43 -6.11*** (2008: w29) -5.56*** (2004: w28) Zt* C -4.81** (2008: w31) -7.11*** (2010: w19) -5.70*** (2001: w19) C/T -5.10** (2003: w27) -7.04*** (2010: w19) -5.71*** (2001: w19) C/S -4.86** (2008: w26) -7.63*** (2010: w19) -5.98*** (2004: w29) Zα* C -45.21** (2008: w31) -90.15*** (2010: w19) -61.46*** (2001: w19) C/T -50.52** (2003: w27) -88.94*** (2010: w19) -61.49*** (2001: w19) C/S -46.00* (2008: w26) -103.52*** (2010: w19) -67.48*** (2004: w29) ***/**/* denote statistical significance at 1%, 5% and 10% level of significance, respectively. Break dates in brackets. 111Agricultural and oil commodities: price transmission and market integration 5. Price transmission Market imperfections may interfere with the price adjustment process in many ways: asymmetric response, speed adjustment, biased information, decisions of storage and inventory holding, policy intervention and others (Granger et al., 1989). Market condi- tions determine price transmission. If the condition of market efficiency holds, the price change in one market is instantaneously and completely transmitted to the related market and the price difference will reflect only the transfer cost (Fama, 1970; Goodwin, 1992). The cointegration condition of long-run equilibrium requires that the integrated pair-wise series comove together. Price transmission is tested with a cointegration error correc- tion model (Rapsomanikis et al., 2006; Minot, 2011). The hypothesis of price transmis- sion from US (here assumed to be the leading market) versus Italy (domestic market) is empirically justified by the large volume of unidirectional commodity flow that has cre- ated a strong dependency of Italy on US exports of these ag-commodities: more than the 90% of the entire volume of Italy’s ag-commodity trade is with US. The price transmission analysis is performed by using the vector error correction model VECM in the following general form: α Π Γ ε= + + ∑ +− = −p p pt t kk q t k t1 1 (5) where pt is a n x 1 vector of n price variables; Δ is the difference operator, Δpt = pt – pt-1; εt is a n x 1 vector of error terms; α is a n x 1 vector of estimated parameters that describe the trend component; Π is a n x n matrix of estimated parameters for the long-term rela- tionship and the error correction adjustment; and Γk is a set of n x n matrices of esti- mated parameters for the short-run relationship between prices, one for each of q lags of the model. The VECM provides a basis for evaluation of relationships across cointegrated series given that cointegration implies that the two prices move closely in the long-run, though in the short-term the two series could drift apart. This approach is appropriate if the following two conditions are held: i) all variables are nonstationary and integrated of order one I(1), following a random walk; ii) the variables are cointegrated in a linear combination that satisfies the stationary con- dition. The cointegration equation is: Pd = α + β Pw + ε10 (6) Pd and Pw are the prices representative of two spatially separated markets integrated of the same order and the error term ε is stationary, β, the cointegrating vector is the price response of dominated market to price changes of the leading market in the long-run. Since prices are expressed in logarithms, β is the long-run elasticity of the domestic price with respect to the US price or the long-run elasticity of price transmission. The expected value for imported commodities is a β value ranging between 0 < β < 1; for β = 0 the 10 this equation is comparable to eq (2) of the previous section (Goodwin, 1992). 112 F. Rosa, M. Vasciaveo, R.D. Weaver US market has influence on the Italian markets, for β=1 the price change in US market is entirely passed to the Italian market (Ravallion, 1986); considering the lagged effect i.e. β = 0.5 the 50% of the change in US price will be transmitted to the Italian price in the long-run (Minot, 2011). The regression equation form for VECM model is: α θ β δ ρ ε( )= + − + + +− − − −p p p p ptd td tw tw td t1 1 1 1 (7) where ptd is the natural logarithm of the Italian (domestic) price of corn, soybeans and wheat respectively, ptw is the natural logarithm of the US (world) price of the same Italian commodities, α, θ, β, δ, and ρ are parameters to be estimated and εt is the error term, the expression in parenthesis (pdt-1- βpwt-1) is the deviation from the long-run equilibrium. The following two terms measure the short term impact of the lagged increments (Δ) of the natural logarithm of international and domestic prices (Conforti, 2004). The error correc- tion coefficient (θ) measures the speed of adjustment, expected to fluctuate in the range between -1<θ<0. If the lagged error correction term (the term in parentheses) is positive, the domestic price is too high given the long-term relationship, then the negative value of θ “corrects” the error by making it more likely that the Δptd is negative. The larger θ is in absolute value (closer to 1), the more quickly the domestic price (pd) will adjust to the value consistent with its long-run relationship to the world price (pw). The coefficient of change in the world price (δ) is the short-run elasticity of the Italian price relative to the US price and represents the percentage adjustment of domestic price one period after a one percent shock in international price. The expected value is 0<δ<β (Minot, 2011). The coefficient of the lagged change in the domestic price (ρ) is the autoregressive term, indicating the change of the Italian price caused by the change of the corresponding price in the next period, the expected value ranges between -1<ρ<1. Table 14 reports the results for the transmission of US prices to the corresponding Italian prices. The unit root tests reported above suggest that each domestic price is nonstationary, and the Johansen cointegration test is used to test for a long-run relationship between the Italian and the US prices. The results suggest that all the domestic prices have a long-run relationship with the US prices for the corresponding commodity. The long-run elasticity of price transmis- sion is statistically significant for all the commodities and very high for soybeans (0.96) and wheat (0.74) meaning that a high percent change of the US price is passed through to the Italian price in the long-run. The speed of adjustment coefficient (θ) is negative as expected for sit and wit and statistically significant at 1% level while for corn there is a slightly positive θ. The value of short run adjustment coefficient (δ) is in the expected range but is not significant for all pairs of commodities. The auto-regressive term is statis- tically significant for all the variables and is higher for corn. Summarizing the results obtained from the transmission model, for each commodity we find the long-run relationship values (β coefficient) are larger than those that indicate short-run transmission (δ). An important role is performed by the autoregressive term meaning that for the corn market in Italy, the 42% of the change of corn price in period t, is transmitted to period t+1; for the soybean market the value of LR adjustment increases to 0.96 and the short run autoregressive value declines to 0.013; for the wheat market, the LR adjustment value is 0.74 and the autoregressive value is 0.026. 113Agricultural and oil commodities: price transmission and market integration 6. Conclusion A number of studies have presented results that support the hypothesis of integration among the ag-commodity markets; more recently many researchers have demonstrated the growing interaction between oil and the ag-commodity price, and more difficulties in predicting the price changes of ag-commodities. Since the price comovements are becom- ing increasingly complex, this research has been dedicated to test the hypothesis of mar- ket integration and price transmission between US and Italy for oil and some relevant ag-commodities. Our results offer traders and policy makers more reliable information to improve their decisions in market trade and policy formation. Time series analysis has been used for testing the initial hypothesis that the oil price is an exogenous signal driv- ing the ag-commodity prices. This is intuitively justified by the large amount of energy inputs used for the ag-commodity production (e.g. fuel, fertilizer, seed, machinery) and the growing quantity of ag-commodities used for biofuel production. The link between US and Italy ag-commodity markets is grounded on the large volume of ag-commodities flowing from US to Italy and the recognized leadership of US prices settled at the CBT. However, price transmission is a more complex phenomenon embedding the comove- ment, completeness, speed of adjustment asymmetries, in a contest of rapid market change (Engle and Granger, 1987; Johansen and Juselius, 1990). In short our results highlight the importance of identifying sample breaks in time series. As clearly illustrated in Table 9 and 10, results and inferences are not robust across sample periods. Intuitively, this highlights the need to empirically determine the time location of structural breaks. By definition, the presence of such a break implies a change in the underlying data generating mechanism and, therefore, a change in parameter values and perhaps functional form of any relationships. Based on identified structural breaks, we find cointegration to vary across sample periods for both cointegration between US and Italian markets within US or Italian domestic settings. These results also highlight the sample dependence of structural estimates, an intuitive statement that is often overlooked when results are compared across papers using different sample periods. Table 14. Transmission of the US food prices (world) versus the Italian food prices (domestic). Commodity Unit root in Italian prices ADF PP KPSS ZA Long-run relationship Johansen test Error correction model Long-run adjustment β Speed of adjustment θ Short-run adjustment δ Auto- regressive term ρ cit yes yes 0.569* 0.002 0.015 0.416* sit yes yes 0.963* -0.038* 0.013 0.202* wit yes yes 0.738* -0.017* 0.026 0.296* * statistically significant at 1% level. 114 F. Rosa, M. Vasciaveo, R.D. Weaver References Abbott, P., Hurt, C., and Tyner, W. (2008). What’s driving food prices? Farm Foundation Issue Report. Farm Foundation, Oak Brook, IL. Algieri B. (2012), Price volatility, speculation and excessive speculation in commod- ity markets: sheep or shepherd behaviour, Zentrum fur Entwircklungforshung Reserearch, University Bonn, DE. Ardeni P.G. (1989), Does the Law of One Price really hold for commodity prices?. Ameri- can Journal of Agricultural Economics, 71: 303-28. Areal F.J., Balcombe K., and Rapsomanikis G. (2013), Testing for bubbles in Agriculture Commodity Markets, MPRA paper n° 48015 on line http://mpra.ub.uni-muenchen. de/48015. Associazione Nazionale Cerealisti. (2011). La mappa dell’import in Italia dei cereali, semi oleosi e farine proteiche. Source: http://www.slideserve.com/magdalena/la-mappa- dell-import-in-italia-dei-cereali-semi-oleosi-e-farine-proteiche. Balcombe K.(2009), The nature and determinants of volatility in agricultural prices: and empirical study from 1962 to 2008. In: Sarris, A. and Morrison, J. (eds.), The evolv- ing structure of the world agricultural trade. Rome, Italy FAO, 09-136. Balcombe, K., and Fraser, I. (2013). Explosive Root Regimes and Volatility Shifts in Eco- nomic Time Series ,(Draft), School of Economics University of Kent Canterbury. Campiche, J.L., Bryant, H., L., Richardson, J. W., and Outlaw, J. L. (2007). Examining the evolving correspondence between petroleum prices and agricultural commodity price. The American Agricultural Economics Association Annual Meeting, Portland, OR, July 29-Aug. 1. Chatellier V., (2011). Price volatility, market regulation and risk management: challenges for the future of CAP. International agricultural policy 1: 33-50. Ciaian, P. and d’Artis, K. (2011a). Interdependencies in the energy-bioenergy-food price systems: A cointegration analysis. Resource and Energy Economics, 33(1): 326-348. Ciaian, P. and d’Artis, K. (2011b). Food, energy and environment: Is bioenergy the missing link? Food Policy, 36(5): 571-580. de Gorter, H. and Just, D. R. (2010). The Social Costs and Benefits of Biofuel: the Intersec- tion of Environmental, Energy and Agricultural Policy. Applied Economic Perspec- tives and Policy, 32(1): 4-32. Conforti, P. (2004). Price transmission in selected agricultural markets. Commodity and Trade Policy Research Working Paper No 7. Rome: Food and Agriculture Organisa- tion. Deaton, A. and Laroque, G. (1995). Estimating a Non Linear Rotational Expectations Commodity price model with Unobservable State Variables. Journal of Applied Econometrics, 10: 9-40. Dickey, D. A., and Fuller, W. A. (1979). Distribution of estimators for time series regres- sions with a unit root. Journal of American Statistical Association, 74: 427-431. EIA. (2012). Biofuels Issues and Trends Independent Statistics. Source: http://www.eia.gov/ biofuels/issuestrends/pdf/bit.pdf. Engle R.F.,and Granger C.W.J., (1987). Cointegration and error correction: representation, estimation and testing. Econometrica, 55: 251-276. 115Agricultural and oil commodities: price transmission and market integration Esposti, R. and Listorti, G. (2013). Agricultural price transmission across space and com- modities during price bubbles. Agricultural Economics, 44: 125–139. Fama, Eugene F. (1970). Efficient capital markets: A review of theory and empirical work, Journal of Finance, 25 (2), 383-417. FAO, IFAD, OECD IMF, WFP UNCTAD. (2011). “Price Volatility in Food and Agricul- tural Markets: Policy Responses” FAO, Rome. Gilbert, C.L. (2010). How to understand high food prices. Journal of Agricultural Econom- ics, 61: 398–425. Gilbert C.L. and Morgan C.W. (2010), Has food price volatility risen? revised version 8 April 2010. Workshop on Methods to Analyze price volatility. Seville. Spain. January 2010. Goodwin B.K. (1992). Multivariate Cointegration Tests and the Law of One Price in Inter- national Wheat Markets. Review of Agricultural Economics, 14: 117-124. Gohin, A. and Chantret, F. (2010). The long-run impact of energy prices on world food markets: The role of macro-economic linkages. Energy Policy, 38(1): 333-339. Granger, C.W. J. and Lee, T.,H.,(1989). Investigation on production, sales inventory rela- tionships using multi-cointegration and non symmetric error correction models. Journal of applied econometrics, 4, 145-159. Granger, C.W.J. and Newbold P., (1974), Spurious regression in econometrics. Journal of Econometrics, 2, 111-120. Gregory, A. W. and Hansen, B.E. (1996). Residual-Based Tests for Cointegration in Models with Regime Shifts. Journal of Econometrics, 70: 99-126. Gutierrez L. and Piras F. (2013). A global weath market model (GLOWMM) for the analysis of Wheat export prices, paper presented at the Annual Conference AIEAA”, Parma. Harri, A., Nalley, L. and L. Hudson, D. (2009). The Relationship between Oil, Exchange Rates, and Commodity Prices. Journal of Agricultural and Applied Economics, 41(2): 501–510. Headey, D. and Fan, S. (2008). Anatomy of a Crisis: the Causes and Consequences of Surging Food Prices. Agricultural Economics, 39 (suppl): 375-391 Helmberger, P.G., Weaver R.D., and Haygood K.T. (1982). Rational Expectations and Com- petitive Pricing and Storage. American Journal of Agricultural Economics, 64: 266-270. Helmberger P. and Weaver R. (1977). Welfare Implications of Commodity Storage under Uncertainty. American Journal of Agricultural Economics, 59: 639-651. Hertel, T., Tyner, W. and Beirut, D. (2010). The Global Impact of Biofuel Mandate. The Energy Journal, 31: 75-100. Huchet-Bourdon, M. (2011). Agricultural Commodity Price Volatility: An Overview. OECD Food, Agriculture and Fisheries Working Papers, No. 52, OECD Publishing Johansen S., (1989). Correlation, regression and cointegration of non stationary economic time series. Unpublished paper, Univ. Copenhagen, Denmark. Johansen, S., Juselius, K. (1990). Maximum Likelihood Estimation and Inference on Coin- tegration with application on Demand for Money. Oxford Bulletin of Economics and Statistics, 52: 169-209. Kwaitkowski, D., Phillips, P.C.B., Schmidt, P. and Shin, Y. (1992). Testing the null hypoth- esis of stationarity against the alternative of a unit root. Journal of Econometrics, 54:159-178. 116 F. Rosa, M. Vasciaveo, R.D. Weaver Lapan, H. and Moschini, G. (2012). Second Best Biofuel Policy and the Welfare Effect on Quantity Mandates and Subsidies. Journal of Environmental Economics and Manage- ment, 63: 224-241. Leucci, A.C., Ghinoi S., Sgargi D. and Junior V.J.W. (2013). Variation and links among food and energy international process. An analysis through VAR models from 2000 to 2012, paper presented at the 2.nd Annual Conference AIEAA, Parma. McNew, K. and Fackler, P.L. (1997). Testing Market Equilibrium: is Cointegration Inform- ative?. Journal of Agricultural and Resource Economics, 22: 191-207. Listorti, G. and Esposti, R., (2012), Horizontal price transmission in agricultural markets: fundamental concepts and open empirical issues. Bio-base and applied economics, 1, 81-108. Minot, N. (2011). Transmission of World Food Price Changes to Markets in Sub-Saharan Africa. IFPRI Discussion Paper 01059 Paper Series, No: 4682. Source: http://ssrn. com/abstract=1233058. Minot, N. (2012). Food price volatility in sub-Saharan Africa: Has it really increased? International Association of Agricultural Economists (IAAE) Triennial Conference, Foz do Iguaçu, Brazil, August 18-24. Myers, R.J. (1992). Time Series Econometrics And Commodity Price Analysis. Depart- ment of Economics The University of Queensland. Nazlioglu, S. (2011). World Oil and Agricultural Prices: Evidence From Nonlinear Causal- ity. Energy Policy 39: 2935-2943. OECD (2008). Rising Food Prices: Causes, Consequences and Responses. OECD Policy Brief. Perron, P. (1989). The great crash, the oil price shock and the unit root hypothesis. Econo- metrica, 57:1361-1401. Phillips, P.C.B. and Perron, P. (1988). Testing for a unit root in time series regressions. Biometrica, 75: 335-346. Phillips, P.C.B., Shi S., and Yu J. 2012. Testing for multiple bubbles. Cowles Foundation Discussion Paper No. 1843. Yale University. Piot-Lepetit, I.M., and Barek, R. (eds.). (2011). Methods to Analyse Agricultural Com- modity Price Volatility. Springer Science+Business Media, LLC 20. Quiroz, J.A. and Soto, R. (1996). International Price Signals in Agricultural Markets: Do Governments Care? Unpublished mimeo, GERENS and ILADES/Georgetown Uni- versity. Rapsomanikis, G., Hallam, D. and Conforti, P. (2006). Market Integration and Price Transmission in Selected Food and Cash Crop Markets of Developing Countries: Review and Applications. In: FAO and Edward Elgar (eds.), Agricultural Commod- ity Market and Trade: New Approaches to Analyzing Market Structure and Instabil- ity. Cheltenham, UK. Rapsomanikis, G. (2011). Price transmission and volatility spillovers in food markets. In: A. Prakash (edit.). Safeguarding food security in volatile global markets. FAO. Rome. Ravallion, M. (1986). Testing marketing integration. American Journal of Agricultural Eco- nomics 68 (2): 292–307. Robles, M., Torero M. and von Braun, J. (2009). When speculation matters. Issue Brief 57, 117Agricultural and oil commodities: price transmission and market integration Washington, DC: International Food Policy Research Institute. Source: http://www. ifpri.org/sites/default/files/publications/ib57.pdf. Rosa, F. (1999). Testing the Quality-Price Relations in Parmigiano and Padano Cheese Markets, Journal of International Food & Agribusiness Marketing 10(3): 19-43. Rosa, F. and Vasciaveo, M. (2012). Volatilità dei prezzi agricoli: un confronto fra prodotti e paesi dell’UE, Agriregionieuropa, 31. . Sarris, A. (2013). Food commodity price instability and food security. 2nd Annual Con- ference AIEAA, Parma. Serra, T. and Gil, J.M. (2012) Price volatility in food markets: can stock building mitigate the price fluctuation?. Paper presented to the International Association of Agricul- tural Economists, Foz do Iguaçu, Brazil, August 18-24. Sharma, R. (2002). The transmission of World Price Signals: Concept Issues and Some Evidences from Asian Cereal Markets. OCDE Global Forum on Agriculture. OCDE, CCNM/GF/AGR 10. Schmidhuber, J. (2007). Biofuels: An Emerging Threat to Europe’s Food Security? Impact of an Increased Biomass Use on Agricultural Markets, Prices and Food Security: A Longer-Term Perspective. Notre Europe Sumner, D.A. (2009). Recent commodity price movements in historical perspective. Amer- ican Journal of Agricultural Economics 91 (5): 1250-1256. Tomek, W.G. and Myers, R.J. (1993). Empirical Analysis of Agricultural Commodity Pric- es: A Viewpoint. Review of Agricultural Economics 15 (1): 181-202. Trostle, R. (2008). Global Agricultural Supply and Demand: Factors Contributing to the Recent Increase in Food Commodity Prices. USDA, Economic Research Service, Report WRS-0801. Tyner, W.E. and Taheripour, F. (2008). Policy options for integrated energy and agricul- tural markets. Review of Agricultural Economics 30: 387-396. von Braun, J. et al. (2008). High Food Prices: The What, Who, and How of Proposed Poli- cy Actions. IFPRI. Source: http://www.ifpri.org/publication/high-food-prices. Yang, J. and Leatham, D.J. (1999). Price Discovery In Wheat Futures Markets, Journal of Agricultural and Applied Economics, Southern Agricultural Economics Association, 31(02): 359-370. Yu, T.H., Bessler, D. A. and Fuller, S. (2006). Cointegration and Causality Analysis of World Vegetable Oil and Crude Oil Prices. Selected Paper prepared for presentation at the American Agricultural Economics Association Annual Meeting, Long Beach, California July 23-26 Zhang, Q. and Reed, M. (2008). Examining the Impact of the World Crude Oil Price on China’s Agricultural Commodity Prices: The Case of Corn, Soybean, and Pork. Pro- ceedings of the Southern Agricultural Economics Association Annual Meetings, Dallas, TX, February 2–5 Zhang, Z., Lohr, L., Escalante, C. and Wetzstein, M. (2010). Food versus fuel: what do prices tell us? Energy Policy 38: 445–451. Zivot, E. and Andrews, D.W.K. (1992). Further Evidence on the Great Crash, the Oil Price Shock and the Unit Root Hypothesis. Journal of Business and Economic Statistics 10: 251-270.