Corn Price Behavior Volatility transmission during the boom on futures Markets

Transcription

1 Corn Price Beavior Volatility transmission during te boom on futures Markets Oliver von Ledebur 1 and Jocen Scmitz 1 vti-ma Institut für Marktanalyse und Agrarandelspolitik, Joann-Heinric von Tünen Institut, Braunscweig, Germany, oliver.ledebur@vti.bund.de Juristisce und Wirtscaftswissenscaftlice Fakultät, Martin Luter Universität Halle-Wittenberg, Halle/Saale, Germany. jocen_scmitz@gmx.de Paper prepared for presentation at te 113 t EAAE Seminar A resilient European food industry and food cain in a callenging world, Cania, Crete, Greece, date as in: September 3-6, 9 Copyrigt 9 by [Oliver von Ledebur and Jocen Scmitz ]. All rigts reserved. Readers may make verbatim copies of tis document for non-commercial purposes by any means, provided tat tis copyrigt notice appears on all suc copies.

2 Corn Price Beavior Volatility transmission during te boom on futures Markets Oliver von Ledebur 1 and Jocen Scmitz 1 vti-ma Institut für Marktanalyse und Agrarandelspolitik, Joann-Heinric von Tünen Institut, Braunscweig, Germany, oliver.ledebur@vti.bund.de Juristisce und Wirtscaftswissenscaftlice Fakultät, Martin Luter Universität Halle-Wittenberg, Halle/Saale, Germany. jocen_scmitz@gmx.de Abstract. Since a number of factors impacted agricultural markets drastically. Among tese are structural canges in global demand and repeated supply constraints tat supported te observed positive development of agricultural prices. Given te increasingly interdependent global markets, te question arises of in ow far an isolated view of a single market, wen analysing price volatility, is sufficient? Te paper is a contribution to te debate on te recent commodity price bubble and te relationsip among commodity futures markets for agricultural raw materials. More particularly, te transmission of price volatility between commodity future markets is analysed. Te background question is weter and to wat extent te volatility of agricultural commodity prices at different market places ave been transferred during te drastic price canges of 8. In tis analysis te volatility of te maize future price at tree different commodity futures excange is modelled as a multivariate GARCH - process. By doing so, interactions between stock markets in different venues are incorporated. Te results of te econometric analysis are discussed against te background of te developments in agricultural and biofuel policy. Keywords: Commodity Futures, Corn, Time Series, Price volatility transmission, multivariate GARCH.. 1. Introduction In te past few years major canges could be observed on te agricultural markets. Witin a sort period of time, te price level of agricultural raw materials rose wit serious consequences for te entire agricultural sector. Te FAO, te EU Commission, IFPRI, te World Bank among oter international organisations point in teir analyses to te driving reasons for canges in te price dynamic on te agricultural markets to an increase of bot te price level and te price volatility, (EU-Commission, 8a; EU-Commission, 8b, p.6ff; FAO, 8, p.55-57; Robles et al., 9; OXFAM, 8; Rabobank, 8, p.8ff). Te price canges on te agricultural markets took place in te course of a general increase in raw materials prices (see Figure 1 and Figure ). 5 = Index Commodity Food Price Index Crude Oil Price index Fertilizers Index M1 M8 1M3 1M1 M5 M1 3M7 4M 4M9 5M4 5M11 6M6 7M1 7M8 8M3 8M1 Source: Own calculations based on IMF (9) and World Bank (9). Figure 1: Montly price index for fertilizer, crude oil and food (Index 5 = 1, -8) 1

3 5 = Index Commodity Food Price Index Weat Corn Pork Beef Soy beans M1 M8 1M3 1M1 M5 M1 3M7 4M 4M9 5M4 5M11 6M6 7M1 7M8 8M3 8M1 Source: Own calculations based on IMF (9) and World Bank (9). Figure : Montly price index for food agricultural products (Index 5 = 1, -8) A series of factors ad an impact on te agricultural markets during te period of price increase. Among tese fundamentals, like te structural canges in global demand and repeated supply bottlenecks supported te observed positive price development in an understandable manner. Oter factors wic influenced te agricultural market volatility in te past years were alsomentioned. Among tese are canges in te market and trade policy in many countries. Te elimination of surpluses in te EU occurred essentially at te same time as te world s most limited market supply situation set in on te cereals market. Several traditional grain exporting countries prevented a relaxing of te world market situation troug te implementation of export taxes, at tis time increasing te existing scarcity, te supply insecurity and ultimately te market volatility. Also "new" market interventions, suc as te additional supply of agricultural raw materials for fuel use as a consequence of te promotion of bio energy, aled to an increase on insecurity. In addition te increase of speculative investors on te futures markets must also be mentioned. Many market observers believe tey influenced te price level and volatility on te futures markets as well as te pysical markets. In te lively debate on tis topic, neiter a consensus on te level of te influence of policy instruments in te areas of renewable energy on te price level of agricultural markets, nor a clear causal interaction between te increased activities of speculative investors on te futures markets were found (FAO, 7, p.48; Rabobank, 8, p.9; USDA, 8, p.). In tis paper te question of te network of futures markets of agricultural raw materials will be considered in te framework of a discussion on te most recent price developments. In particular te transmission of price volatility between markets will be considered. Te corn market was cosen as te key market for tis analysis. Tis product plays a central role worldwide for te breeding industry but also in te area of substitution of fossil fuels wit biogenous fuels from renewable resources. Te empasis of te approac of substituting fossil fuels wit renewable (agrarian) raw materials forced in te United States is based on te use of etanol stem from corn starc. Te Cicago futures market takes an exposed position in world trade wit agricultural raw materials. On te Cicago Board of Trade (CBOT) a large part of worldwide futures trade wit corn is realized since te USA is by far te largest corn producer and exporter. Te contract prices tat are generated on te CBOT in te course of te futures trading tus ave a broad reacing signal function for te corn market. In addition to tis excange, tere are two oter excanges of interest. Tese are te Futures Excange in São Paulo (Bolsa Mercantil e de Futuros, BMF, ere denominated BRAZ) in Brazil and te Parisian

4 Excange MATIF (Marce a Terme d Instruments Financiers). Te Brazilian agriculture markets distinguis temselves wit te worldwide largest export surplus. Te Frenc excange takes te role of leading market wit regard to cereals markets witin te European Union, also a major player on agricultural markets. Te question emerges of weter and to wat extent te price volatility between international trading centres during te drastic price canges of 8 were carried over. Tis issue as only been dealt wit rougly in te scientific literature (Baffes, 7; EU-Commission, 8a; EU-Commission, 8b; Garcia and Leutold, 4). Studies on te volatility of agricultural commodity prices until now ave mainly concentrated metodically on an univariate approac. At te eart of te analysis is a modelling of volatility as a GARCH process 1 wic was supplemented wit exogenous factors. Some examples of tis are Crain and Lee (1996), Goodwin and Scnepf () and Boudouk et al. (3). Te influence of exogenous factors on te volatility of futures prices stood at te centre of tese studies. As factors of influence, for example, trade volume, stock inventories, and te government programs in te agricultural sector were ere identified. Against tis metodical background, interactions between excanges at different trading centres were not contemplated. In ligt of te worldwide increase in independent markets te question emerges of ow adequate an isolated consideration of single market is. For tis reason we take a different approac in tis study. It appears to be urgently necessary to illustrate te relevant markets simultaneously and to document teir interdependencies. In order to acieve tis for te corn market, we illustrate te markets in a multivariate eteroscedasticity model. Te modelling approac of a BEKK model is useful for tis purpose. A positive definite covariance matrix H t witout all too restrictive parameters and a foregoing of an excess of parameters lets tis model appear to be suited for tis question. Tis paper is divided into te following areas. Te next section introduces te metods of te BEKK model used in detail. Section 3 presents te data used and presents te estimated results. A discussion on te results closes tis section. In te final section, some future researc pats are outlined.. Model GARCH models serve as te backbone of volatility modelling. Toug te approac by (Engle, 198), it is possible to model te (unobserved) second moment. Te resulting variance is dependent on te amount of currently available information. Tis type of model can be caracterized in two formulas. Te first equation is described as a mean equation and illustrates te first moment of te process (Equation 1). In tis specification only a long term trend component μ is assumed. Te second equation is described as te variance equation. It serves as te second moment of te process (Equation ). In an ARCH(p) process tis is te total p delayed information. Te known information set is generated from te returns up to te time point t-1. Te returns are calculated as r t = log(f t /F t 1 ). Te futures price at time point t is called Ft and r t describes te illustrated returns at time point t: rt = μ + ε t Equation 1 ε = t z t t ε t I t 1 ~ N(, ) 1 GARCH stands for Generalized Autoregressive Conditional Heteroscedasticity BEKK models are named after Baba, Engle, Kraft und Kroner 3

5 Te resulting variance of r t yields te generalization of te model by Bollerslev (1986) permits te inclusion of past variances in addition to te consideration of past innovations. Tis leads to te general univariate GARCH (p,q) model. =α +αε + K+α ε +β + K+β t 1 t 1 p t p 1 t 1 q t q Equation Te transfer into a multivariate GARCH model (Engle and Kroner, 1995) takes place wit a generalization of te resulting variance matrix H t. H t = Equation 3 Eac element of H t depends on p delayed values of te squared ε t, te cross product of ε t and on q delayed values of te elements from H t. We did not make use of te possibility to draw in exogenous factors into te resulting variance equation. In general, a multivariate GARCH (1,1) model witout exogenous factors can be presented as follows as a BEKK model. For reasons of clarity time indicators are not included in te presentation. A model wit te time delay of only a lag (t-1) was modelled. H t ' = C C a11 + b11 a b ' ε1 ε ε1 a 33 ε 3ε1 ' 1 1 b ε ε 1 ε ε ε ε 1ε 3 a ε ε 3 ε b b a b + a Equation 4 Troug te model construction via te quadratic form it is acieved tat te resulting variancecovariance matrix H t is positive definite. Tis ensures tat all variances and covariances are always positive. In compact form, te above equation can also be written in tis manner: H ' ' ' = C C + Aε t 1ε t 1 A + ' B H t t 1 B Equation 5 Te matrices A, C and B possess te dimension (n n). C is a (lower) triangular matrix. In te model assumed ere, we are dealing wit te matrices A and B as diagonal matrices. A generalization of te model is possible. Furter interactions could be implemented, but ten te matrices A and B are not diagonal anymore and results a muc more complex matrix H t. Apart from te acievement of a positive definite matrix H t, tere is anoter advantage of te BEKK specification. Due to te diagonal BEKK model assumed ere, a cecking of te stationary nature of te process is determined solely troug te diagonal elements of matrices A and B. Te diagonal BEKK n model is stationary if a bii k i k = ( ii k +, ) < 1 1, (Engle and Kroner, 1995, p.133). In accordance wit te questions stated in te introduction, tree trading centres (excanges) were studied (n=3). Furtermore as 4

6 described above only one time lag was included. Te resulting variance and covariance equations are as follows: + 11 = c1 + a11ε 1 b = c + a11aε ε1 b11b = c3 + a11a33ε 3ε1 b11b = c4 + aε b + 3 = c5 + aa33ε 3ε bb = c6 + a33ε 3 b333 Equation 6 Equation 7 Equation 8 Equation 9 Equation 1 Equation 11 Te indexes used ere recede to te notation used in equation 3. Te following presentations are equivalent in te framework of te model as proposed ere: vec(h t ) = vec(h t ) = H t Te Matrix H t contains redundant expressions. Tus no distinction is made between 1 / 1, 31 / 13 or 3 / 3. For tis reason te vec expression, corrected for te redundant expressions, is te same as te vec expression. Te empirically estimated BEKK GARCH model is tus based on a multivariate version of equation 1 and equations 6 to 11 (Derived from Equation 4). 3. Data and Results Te interactions of volatilities between te traded price quotations on te excanges in te USA (CBOT), Brazil (BRAZ) and Europe/France (MATIF) were studied. Te topic of te study was te price quotations wic ran out in Marc 8 and was dealt on all tree excanges wit te same running time. Excange quotations from Marc 7, 7 to Marc 5, 8 were available. As a courtesy of ZMP, te Central Market and Price Reporting Centre (Zentrale Markt- und Preisberictsstelle) provided te data. Price quotations are given in US-Dollars. Eac Futures contract is based on a different amount of corn quantity. One contract in Europe stands for 5 tons of corn. In Brazil 45 units of 6 kilogram bags/coffers will be traded by one contract, tis is equivalent to 7 tones. In te United States te unit per contract is 5 busels. Tis is equivalent to 17 tones (1 busel = 5,4 kilogram). Tese different units of measurement explain te observed price levels per unit of weigt on tese markets. Due to olidays etc., on some days price notations were not available for all tree markets. Tus, all excange quotations for tis day were deleted. Overall at te start, 45 excange quotations were available. After te validation, 6 observations remained for te estimates. In te framework of a GARCH estimate, tis is just a small sample. At te MATIF and BRAZ trading centres, te appropriate futures were not placed and traded earlier. Tus 45 daily notations were available. Tis is due to te different saping of te contracts in te future excanges. Te time development of prices per unit of weigt for corn between Marc 7, 7 and Marc 5, 8 is presented in te following Figure 3. Te value of one future contract consists of te price per unit of weigt multiplied wit te corresponding unit of trading. One can recognize a continual price increase on te Cicago excange. Tis increase began in October 7. A comparably strong and permanent price increase cannot be observed on te oter excanges. Tis price development also affects te returns, te logaritmic difference of te price level of te process. 5

7 /7 6/7 8/7 11/7 1/8 MATIF /7 6/7 8/7 11/7 1/8 BRAZ /7 6/7 8/7 11/7 1/8 CBOT Source: ZMP (9) Figure 3: Corn prices per tons in France (MATIF), 6 Kg in Brasil (BRAZ) and per busel in USA (CBOT) According to Figure 3, te processes sow a very clear non-stationary beaviour. Stationarity of time series data can be tested by Unit-Root tests. Te validation of stationarity was conducted by te Augmented-Dickey-Fuller-Test (ADF-test). For tis test te null ypotesis is non-stationarity. Results are given in Table 1. Te results clearly endorse te non-stationary beaviour of te series. Since te estimation of te model using price levels is not indicated, stationary variables are to be used. Te correspondent returns are significant stationary at te 1 percent level and are suited for modelling. Table contains a summary of te data for te returns of te corn prices. Most clearly evident are te results of te Jarque-Bera statistics in Table. According to tem, te assumptions of normal distribution on te basis of 5% level of significance can not be rejected for te CBOT data. Tis finding is contradictory to te stylized facts of finance market data. Especially regarding te returns, te absence of normal distribution is assumed. Tis is manifested in a iger Kurtosis (Kurtosis >3) meaning more probability mass in te flanks of te distribution of returns. Higer Kurtosis is given for te remaining returns distributions (see Table ). Table 1: Unit-root tests for price level and returns MATIF BRAZ CBOT Future Return Future Return Future Return Observations ADF - Test 1, -1,935 1,5-13,36 1,4-14,3167 (,9411) (,) (,9464) (,) (,9594) (,) Prob. Value in parentesis (Prob.value) Source: own calculations. Table : Summary of returns for te selected excanges ( ) Mean,,,14 Median,6,,1 Max,645,71,457 Min -,657 -,788 -,537 Std. Deviation,167,18,185 Skewness -,844 -,94 -,595 Kurtosis 5,355 5,4 3,843 Jarque-Bera 41,69 53,8667 3,68 prob. Value,,,195 Source: own calculations. MATIF BRAZ CBOT 6

8 Te reasons for suc surprising results could ave teir origin in te political-institutional framework conditions. Here canged framework conditions applied for te corn market, particularly in te USA. Te massively extended production of etanol on te basis of corn starc requires correspondent amounts of raw materials. In te USA corn and soy are competing for land in seeding. Already in te period October 6 to May 7 one could observe a simultaneous price increase for bot products on te futures excanges and also on te pysical markets, wic ave been interpreted as an expression of demand pressure evident troug te competition for land during te sowing period (Teis, 7, p.48-49). Te time period considered ere comes wit a furter increase in te price of crude oil wic drove te etanol boom furter. Against te background of tis price development, te price competitiveness of te biogenic fuel as a substitute for fossil fuel increased (see Figure 4). In ligt of te expected furter demand, increasingly iger prices for corn contracts were offered on te Cicago excange in order to secure te supply of corn for processing and feeding. Barrel US etanol production Crude oil (Barrel) US$/Barrel Jan 7 Feb 7 Mrz 7 Apr 7 Mai 7 Jun 7 Jul 7 Aug 7 Sep 7 Okt 7 Nov 7 Dez 7 Jan 8 Feb 8 Mrz 8 Apr 8 Mai 8 Jun 8 Source: Renewable Fuel Association (9) and IMF (9) Aug 8 Sep 8 Okt 8 Nov 8 Dez 8 Jul Figure 4: US - Etanol produktion and crude oil price development, 1/7 1/8. Te increase of corn price in CBOT at te time period October 7-Marc 8 is coeval wit a devaluation of te US Dollar, also against te Brazilian Real (Figure 5). Tis excange rate development leads to a relative price advantage of te US priced product wic finally may ave led to an increase in foreign demand for US corn. Tis applies particularly against te background of te tense supply situation of feeding stuff in 7/8. In te European Union, weat as been substituted by corn and corn as been increasingly imported. Due to gene modified corn cultivation, Argentinean corn did not get access to te European markets at tis time. Tis led to an increased demand for Brazilian corn. According to extremely ig world market prices Brazilian corn could be transported to EU market witout any tariff burden 3. Tese factors togeter led to a significant price impact as can be seen in Figure 3. 3 Te calculation of tariff rates for corn, by te EU commission, is based on FOB prices of te US goods in te Gulf of Mexico, transport costs to Rotterdam and te administrated intervention price (CAP Monitor, 9). Tis link to te US corn market as been liftet under te prevalent market conditions 7

9 ,6,4 BRR per US$,, 1,8 1,6 1,4 1, 1, Jan 7 Feb 7 Mrz 7 Apr 7 Mai 7 Jun 7 Jul 7 Aug 7 Sep 7 Okt 7 Nov 7 Dez 7 Jan 8 Feb 8 Mrz 8 Apr 8 Mai 8 Jun 8 Jul 8 Aug 8 Sep 8 Okt 8 Nov 8 Dez 8 Source: MAPA - Ministério da Agricultura; Pecuaria e Abastecimento, 9. Figure 5: Excange rate development Brazilian Real to US Dollar, 1/7 bis 1/8. A direct European demand pressure on te US corn market is excluded since te import of corn from te USA does not take place due to existing rules in te area of consumer protection since te end of te 199s. Te EU Framework regulations on genetically modified organisms (GMO) is continually being expanded and updated. A series of legal regulations exist wit te goal of protecting public ealt and te environment. An important branc of te EU laws on GMOs deals wit te release of genetically modified organisms into te environment. In, an autorization procedure was introduced dealing wit te release of GMOs or any type of product made of GMOs. Te proibition, or rater te nonautorization of some so-called Bt Corn varieties led to an actual end of te US exports into te EU (Wirtscaftswoce, 5) as can be seen in Figure 6 and Figure 7. Tis politically induced market development in te USA led to te solidification of expectations for increasing prices, wic ended up among oter tings in less price deviations. Tis may ave led to te fact tat for te futures price on te Cicago Excange te assumption of normal distribution of returns could not be rejected. Mio. ton Oter extra-eu Imports USA Canada So. African Cust. Union Serbia Ukraine Brazil Argentina EU-1 EU-1 EU-15 EU-5 Source: own calculations based on COMTRADE (9) Figure 6: Development of Corn Imports from Non EC/EU Member States,

10 1% 9% 8% 7% 6% 5% 4% 3% % 1% % EU-1 EU-1 EU-15 EU-5 Oter extra-eu Imports USA Canada So. African Cust. Union Serbia Ukraine Brazil Argentina Source: own calculations based on COMTRADE (9) Figure 7: Development of Corn Imports from Non EC/EU Member States, Regarding te variance and covariance equations described in Section, te interdependence of te markets can now be cecked. Particularly te covariance equations provide insigt into te interactions of te markets. From te estimated parameters given in Table 3 te following can be seen: In order to better compreend te results we briefly address te equations 6, 9 and 11. Equation 6 is te variance Equation for CBOT, Equation 9 for MATIF and Equation 11 for BRAZ. Parameters a 11, and c 1 for te CBOT Market are all not statistically significant at te five percent level. Tis means first, tat te according variance equation is partly void (Equation 6). Te returns on te Cicago Futures Market were not marked by conditional eteroscedasticity in te time period considered. Te conditional variance of prices in CBOT is caracterised only by its own lagged variance. As parameter a 11 is insignificant, information socks are not accounted for. Tis finding igligts again te peculiarity of tis excange at tis time. Tis as a broad reacing meaning since tese equations illustrate te spill-over effects of te Cicago Market on MATIF and Brazil. Tus in te time period considered, no spill-over of price or information socks from Cicago (e.g. updated arvest forecast in te USA) took place on te development of prices in te market places MATIF and Brazil. Neverteless CBOT is suc important tat te oter markets considered are influenced via te covariance. Te results sow tat only te lagged conditional variance of CBOT influences te covariance. Te politically induced market development in te USA combined wit te tigt supply situation caused a partial decoupling of te US market from te oter markets analysed ere. Due to te significance of te Futures excange in Cicago (it is te global key market) it came s to a noticeable influence of te oter market places only via te covariance. 9

11 Table 3: Estimated parameters of te BEKK model Coefficient Prob. Value µ 1,1,85 µ,4,475 µ 3,6,1 c 1,6,6539 c,18,576 c 3,37,593 c 4,47,93 c 5 -,5,91 c 6, 1, a 11 -,7,4179 a,33, a 33,479, b 11,9855, b,916, b 33,8745, Source: own calculations. In contrast to te trading centre in Cicago, clear interactions between MATIF and Brazil exist. Tis can be seen due to te significant parameter values a, a 33, b and b 33, wic appear in Equation 1. Tus at least an indirect (via a covariance) influence could be seen for bot markets. Here bot components of te covariance Equation are relevant. Information socks tat occur on one of bot market places impact te volatility of te oter one. Te lagged development of te variance of te oter market alone is not determining it s future development. Eac variance equation sows also te special situation during te time period studied. For te trading centres in Europe and Brazil a GARCH (1,1) Process could be identified for eac. In bot time series a certain amount of eteroscedasticity is apparent. Tis is normal in financial-econometric studies. Tis result must be seen in te context of te above described political framework conditions. From te perspective of te autors, te import ban on genetic corn into Europe as well as te simultaneous etanol boom led to strongly canged price development of futures on te Cicago Futures Excange. Tis resulted in a decoupling of te price development on te excanges in Europe and Brazil. Tis decoupling was ultimately measured and confirmed by te model. Even te single consideration of te Cicago prices via a (univariate) GARCH (1,1) model sowed no resulting eteroscedasicity (te results of estimates are not provided ere). 4 Perspectives Te multivariate analysis framework used ere permits te illustration of te price volatility on te futures excanges as an interplay of many mutually influencing trading centres. Te results sow tat on te futures markets for te considered time period very unusual impact interactions existed and tese allow te assumption tat te institutional framework conditions influenced te markets, or rater isolate tem from eac oter. It could successfully be sown tat te volatility of future prices at different market places do impact eac oter. Hence, an additional building block of te explanation of te volatility could 1

12 be identified. Tis work extended te existing researc of Crain and Lee, 1996; Goodwin and Scnepf, ; Boudouk et al., 3 troug a multivariate analysis. Neverteless not all facets of te determinants of price volatility are clear. In te future, it must also be cecked weter te interactions identified ere are time independent or not and if not, wic factors of influence play a role. Tis olds in particular true for te analysis of te markets during and after te turns in te crude products markets as a consequence of global finance market crises. Also te expansion of te framework of te analysis in order to identify interactions between futures of different (agricultural) raw products is a topic for furter researc. Acknowledgment We would like to tank Wienke von Scenk, Martin Scraa, Josef Efken and Alexander Gruber for valuable comments on former drafts of tis paper and Gerlinde Kubitza for te pacient data preprocessing. Te views expressed erein do not necessarily represent tose of te vti or University of Halle. References Baffes J. (7) Oil Spills on Oter Commodities. World Bank Policy Researc Workking Papers. World Bank, Wasington. p. Bollerslev T. (1986) Generalized Autoregressive Conditional Heteroscedasticity. Journal of Econometrics, vol. 31, pp Boudouk J., Ricardson M., Sen Y., Witelaw R.F. (3) Do Asset Prices Reflect Fundamentals? Fresly Squeezed Evidence from te OJ Market. NBER Working Paper Series 9515, pp. 1-5 CAP Monitor (9): CEREALS Agra Infoma Ltd, Tunbridge Wells, Kent, England. Crain S.J., Lee J.H. (1996) Volatility in Weat Spot and Futures Markets, : Government Farm Programs, Seasonality, and Causality. Te journal of finance vol. 51, pp Garcia P, Leutold, R.M. (4) A Selected review of agricultural commodity futures and options markets. Europ. Rev. of Agr. Economics vol. 31 (3), pp Goodwin B.K., Scnepf R. () Determinants of endogenous price risk in corn and weat futures markets. Journal of Futures Marktes vol., pp Engle R. (198) Autoregressive Conditional Heteroscedasticity wit Estimates of te Variance of UK Inflation. Econometrica vol. 5, pp Engle R, Kroner K. (1995) Multivariate simultaneous generalized ARCH. Econometric Teory vol. 11, pp EUROPEAN COMMISSION (8a) Hig prices on agricultural commodity markets: situation and prospects - A review of causes of ig prices and outlook for world agricultural markets. Brussels, Directorate-General For Agriculture And Rural Development, Directorate L. Economic analysis, perspectives and evaluations, L.5. Agricultural trade policy analysis, July 8 ttp://ec.europa.eu/agriculture/analysis/markets/index_en.tm EUROPEAN COMMISSION (8b) November update on recent agricultural commodity and food price developments in te EU. Brussels, Directorate-General For Agriculture And Rural Development, Directorate L. Economic analysis, perspectives and evaluations, L.. Economic analysis of EU agriculture, November 8 ttp://ec.europa.eu/agriculture/analysis/markets/index_en.tm FAO Food and Agriculture Organisation (8) Food Outlook Global Market Analysis. Rome, Economic and Social Development Department, May issue ttp:// 11

13 FAO Food and Agriculture Organisation (8b) Crop Prospects and Food Situation. Rome, Economic and Social Development Department, July issue ttp:// FAO Food and Agriculture Organisation (7) Food Outlook Global Market Analysis. Rome, Economic and Social Development Department, November issue ttp:// Greene, William H. (8) Econometric Analysis. 6t ed. Pearson Prentice-Hall, Inc. IMF International Monetary Found (9) Primary Commodity Prices. ttp://imf.org/external/np/res/commod/index.asp MAPA Ministério da Agricultura, Pecuaria e Abastecimento (9) ttp:// OXFAM (8) Double-Edged Prices, Oxfam Briefing Paper, October 8 RABOBANK (8) Outlook 9 Down to Eart. Coöperatieve Centrale Raiffeisen-Boerenleenbank B.A (Rabobank) RFA Renewable Fuels Association (9) Montly U.S. Fuel Etanol Production/ Demand. ttp:// Robles M., Torero M., von Braun J. (9) Wen speculation matters. IFPRI Issue Brief 57, February 9 Teis R. (7) Stet eine neue Kursrallye bevor? DLG-Mitteilungen, 11/7, pp USDA (8) Global Agricultural Supply and Demand: Factors Contributing to te Recent Increase in Food Commodity Prices. USDA-ERS, WRS-81, May 8/July 8 (revised) (ttp:// WB WORLD BANK (9) Commodity Markets Review. ttp://go.worldbank.org/komguhxsg Wirtscaftwoce (5) EU-Auflagen stoppen Import von US-Futtermitteln. ttp:// downloaded at