Impacts of stockholdings behaviours on agricultural market volatility: a dynamic Computable General Equilibrium approach

Transcription

1 Impacs of sockholdings behaviours on agriculural marke volailiy: a dynamic Compuable General Equilibrium approach Fabienne FEMENIA INRA and Agrocampus Oues, UMR 302 SMART The successive CAP reforms quesion is price sabilizing aspec and more and more aenion is being paid o privae risk managing insrumens such as sorage. The effecs of sorage have already been widely sudied in economic lieraure. Bu almos none of hese sudies do accoun of he links beween producers, households and sockholders decisions as can a CGE model. Furhermore, he iner emporal decisions of hese agens are generally no really aken ino accoun when sudying he effecs of an insrumen like sorage allowing for iner emporal arbages. Finally, almos all of hese sudies focus on he effec of sockholdings on exogenous price volailiy and assume raional expecaions, which do no allow for he represenaion of he endogenous par, induced by marke funcioning, of risk; ye his endogenous risk has ofen been used o jusify public inervenion in agriculural markes. In his paper we build a model addressing hese issues and conduc some illusraive simulaions. Some of our resuls sand in conras wih he conclusions of previous economic sudies concerning he effecs of sorage on marke volailiies and reveal he imporance of he role played by expecaions of economic agens and by he links beween he iner emporal decisions when sudying such an insrumen. Key Words: Sockholding behaviours, General equilibrium, Endogenous risk, Iner emporal decisions

2 Inroducion One of he objecives of he European Common Agriculural Policy (CAP) when i was inroduced was o ensure he sabiliy of agriculural incomes in he European Union (EU). A sysem of public insrumens was used o conrol he quaniies of agriculural producs supplied on European markes and hus o guaranee sable agriculural prices and incomes. Ye in 992 he EU sared o replace his price suppor scheme by a sysem of paymens more decoupled from producion and prices. This decoupling of farm paymens was reinforced by he 2003 CAP reform and, as shown by he 2008 Healh Check, is likely o be coninued in he fuure. Ye, as shown noably by CHAVAS and KIM (2006), suppressing a price suppor program leads o increase price volailiy. The successive reforms of he CAP hus quesion is price sabilizing aspec. Some privae insrumens could however be used by European agriculural producers o manage he price risk bu have no been exensively used unil oday, noably because of he exisence of public insrumens (ANDERSON, 992). Privae sorage is one of hese insrumens. Indeed sockholding behaviours allow for iner emporal arbages (ANDERSON, 992): when prices are low he demand for socks is high and when prices rise he quaniies sored are pu back ono he marke which comes o miigae he rise. This mechanism was formerly used by he European Union o sabilize marke prices hrough public sockholdings. In fac, as shown by LENCE and HAYES (2002), his public sorage subsiued speculaive sockholdings. One can hus presume ha privae sorage will be more and more used on European markes. The effec of privae sorage on marke dynamics is an imporan poin and hus needs o be aken ino accoun in economic evaluaions of agriculural policies. Ye mos of hese economic evaluaions are based on Compuable General Equilibrium (CGE) models which rarely incorporae sockholding behaviours. However, in a marke economy, prices resul from several decisions of economic agens acing on several markes poenially linked beween hem; moreover, as poined ou by MAKKI e al. (996), sorage and rade can ac complemenarily on price volailiy. Accouning for all hese relaionships, as can a CGE model do, hus seems crucial when one ges ineresed in he effecs of sorage on price risk. One of he reasons why CGE models do no include sockholdings behaviours is ha mos of hem were no originally aimed a simulaing shor-run policy effecs (Herel e al., 2005), as a maer of fac, hese models are generally no fully dynamic (FEMENIA and GOHIN, 2009). The purpose of his aricle is o ackle his issue by inroducing sockholding behaviours in a dynamic CGE model able o accoun for agriculural price volailiy. Furhermore, he economic lieraure idenifies wo kinds of phenomenon explaining agriculural marke prices volailiy: his volailiy can be due o exogenous random shocks like climaic hazards. Bu price flucuaions can also be endogenous, ha is o say linked o marke funcioning and o expecaion errors from economic agens. Indeed, in agriculural secors here is a ime lag beween producion decisions and harvess, his ime lag implies ha producers have o base he decisions on expeced raher han on observed marke prices, and he possible expecaion errors can induce price flucuaions. This phenomenon was formalised by Ezekiel (938) in his Cobweb heorem. Boh of hese wo sources of volailiy are linked in he sense ha economic agens can someime make misakes because exogenous shocks occur beween producion decisions and harvess which generaes price flucuaions and if hey are no raional hese flucuaions will spread over ime. Whereas he impacs of sockholdings behaviours on exogenous price volailiy have been quie largely sudied in he economic lieraure (WILLIAMS and WRIGHT, 99, DEATON and LAROQUE, 992), he sudies dealing wih he effecs on endogenous volailiy are much rarer. Though, non raional speculaive behaviours of privae sockholders can be said o desabilize markes (see RAVALLION, 987 for insance). In fac we have no been able o find more han one paper addressing he issue of he modelling of sockholding behaviours in an imperfec expecaion 2

3 framework (MITRA and BOUSSARD, 2009). Furhermore none of he aforemenioned works are conduced in a general equilibrium framework. The model we build in his aricle has several characerisics allowing he inegraion of hese differen elemens. Fs, we depar from a widely used general equilibrium framework: he GTAP model and daabase (HERTEL, 997). Then, we rely on he work of FEMENIA and GOHIN (2009) o build a dynamic model aking iner emporal decisions of economic agens ino accoun. These iner emporal decisions are based on imperfec expecaions which enables a represenaion of he endogenous aspec of marke volailiy, he exogenous par being inroduced via exogenous shocks. Finally, privae sockholding behaviours are inroduced ino he model. In addiion o be conduced in a general equilibrium, our work essenially differeniaes from he previous work dealing wih sorage and endogenous volailiy by he specificaion of he iner emporal behaviours of all economic agens and by he iming of sockholders decisions; hose are aken once harvess are pu on markes and no conemporarily o producion decisions as in MITRA and BOUSSARD (2008). Once we build his model, we run some simulaions o sudy he effecs of European whea sorage when exogenous produciviy shocks occur in he Res of he World. These simulaions reveal ha considering imperfec expecaions and accouning for he general equilibrium links beween secors and for he iner emporal dimension of he decisions of economic agens can lead o resuls differen from wha is commonly found in he economic lieraure, noably hose concerning he ransmission of marke flucuaions beween secors/ regions or from prices o producion quaniies. Furhermore, our resuls end o confm ha he high auocorrelaions observed beween agriculural prices can be aribued o sockholdings behaviours, which previous sudies (DEATON and LAROQUE, 992) had suspeced bu had no been able o reproduce. The remaining of he aricle is organised as follows: in he nex par we describe he characerisics of he model, namely is dynamic characerisics, he way marke price volailiy is inroduced, how privae sockholdings behaviours are modelled and finally he execuion of he model; a second par of he aricle is devoed o he resuls of he simulaions we have conduced and o some sensiiviy analysis of hese resuls; finally we conclude.. The model Mos of he CGE models used oday o assess he effecs of agriculural policies are no adaped o deal wih price volailiy and, a foriori, wih he effecs of sockholdings on his volailiy, because hey are mosly saic, do no open he possibiliy o represen he endogenous aspec of price volailiy and do no inroduce sockholding behaviours. To ackle his issue we build a model able o accoun for he dynamic evoluion of markes and for he iner emporal decisions of economic agens. This model is also suied o he represenaion of exogenous and endogenous price volailiy and includes sockholdings behaviours. Our saring poin is a version of he widely used GTAP framework (HERTEL, 997) adaped o he sudy of agriculural markes: he GTAP AGR framework. The main differences beween his model and ours are described in wha follows.. Characerisics of dynamic behaviours The fs concern when dealing wih price flucuaions is o model marke evoluion period by period, ha is why we rely on he work of FEMENIA and GOHIN (2009) who build a dynamic CGE model based on he GTAP framework. In his model secoral capial socks accrue from one period o anoher in each region: 3

4 K K I, wih K he capial sock, I he new invesmen and he depreciaion rae of capial, he subses i, r and denoing respecively he secor, he region and he ime period concerned. New invesmens a one period will hus increase capial socks for he nex period. Capial socks increase as ime goes by as long as secoral invesmens are higher han he depreciaion of capial sock. When invesmens equal capial depreciaion, capial socks sop increasing and hen remain consan; his fs seady sae period will be referred o as period T. Using capial accumulaion as a link beween periods is quie a usual way o inroduce dynamics in CGE modelling (see for insance he Linkage model from he World Bank or he Mage model from he CEPII). However mos of he exising dynamic CGE models do no accoun for iner emporal decision processes of economic agens and are hus no able o accoun for he formaion of he expecaions. As in FEMENIA and GOHIN (2009), his drawback is overcome in our model: invesmen decisions of producers and saving decisions of households are based on iner emporal arbages. Indeed, o ake his invesmen decision he producer seeks o maximize he presen value of his fm (DEVARAJAN and GO, 996), which corresponds o he discouned value of all his expeced fuure profis (capial income) minus his expeced fuure invesmen coss: I ² max wk K PI PI I r 2 K s K K K I Wih r he ineres rae, wk he capial income, PI he price of invesmen and an adjusmen parameer: he erm I ² PI 2 K represens he adjusmen cos of capial (MCKIBBIN and WILCOXEN, 998). Solving his opimisaion problem leads o a condiion deermining opimal invesmen in our CGE model: I I I ² wk PI r PI PIr K K 2 K ² As in seady sae invesmen equals capial depreciaion and prices are sable, from period T his equaion becomes: wkrt PIrT r PIrT r 2 Households also base he saving decisions on an iner emporal rade-off. Indeed hey spend a par of he income hey earn a one period o consume goods, which brings hem some uiliy, and save he remaining par. The par of he income saved a one period will be used laer o consume and represens a fuure uiliy. So, he represenaive household in each region seeks o maximize he value of is iner emporal uiliy, subjec o an iner emporal budge consrain: max Ur u Q r Er s P Q S r r r r r Wih a ime preference parameer (households have a preference for immediae uiliy), Q he quaniy consumed, P he composie consumer price, E he oal income (including ineres earned from foreign asses, facor reurns, disribued profis and ax receips) and S 4

5 savings. The fs order condiion of his program allows o deermine he evoluion of savings: E S E r r S. The aforemenioned equaion, combined wih he fac ha in r r r a seady sae savings equal invesmen, allows o derive he levels of savings for all periods. These, combined wih a foreign deb accumulaion period by period, are he main characerisics which enable our model o reproduce he dynamic evoluion of markes..2 Modelling of he marke prices volailiy Two sources of price volailiy on agriculural markes are idenified in economic lieraure (BUTAULT and LE MOUËL, 2004): price flucuaions can be due o exogenous sochasic shocks and can also be generaed by non raional marke behaviours. These wo aspecs are inroduced in our model. The fs par of his secion is devoed o he inroducion of exogenous disurbances in he model and he second one o he modelling of non raional behaviours..2. Inroducion of exogenous sochasic disurbances in he model Many economiss have argued ha flucuaions on agriculural markes were essenially due o demand and supply shocks (MOSCHINI and HENNESSY, 200). Indeed, he ime lag beween producion decisions of farmers and he harvess induces a shor-erm rigidiy of he agriculural supply, which can hardly adjus o marke price changes. Furhermore mos agriculural producs are saples and demand for hese goods is quie inelasic. Because of hese wo characerisics agriculural markes are very sensiive o marke shocks: a supply decrease due o a climaic hazard for insance will resul in a large price increase. This phenomenon is formalized by he King s law. Ye agriculural producion is exposed o several epidemic and climaic risks and hese exogenous shocks occur quie frequenly, hus generaing price flucuaions. Our purpose here is o inroduce random supply shocks in he dynamic model o incorporae exogenous price flucuaions. In our model he agriculural echnology is represened by a nesed CES producion funcion. The fs nes combines producion facors o creae value added; he second one combines he aggregae facors wih inermediae consumpions o produce oupu: VA a K b L c T Y VA IC Wih VA he value added, L he labour facor, T he land facor, Y he quaniy produced and IC he aggregae inermediae consumpion. a, b, c and are share parameers, and deermine respecively he degree of subsiuabiliy beween capial, labour and land and beween value added and inermediae consumpion, finally and are produciviy parameers. Supply shocks are inroduced in our model hrough he produciviy parameer. Indeed we assume ha hese shocks can be linked o produciviy shocks. We hus inroduce random disurbances such ha shock wih schock he shocked produciviy parameer and assume ha N 0, ² which implies ha 5

6 shock flucuaes around wih a variance equal o ² shock ² ², ha is o say ha he values calibraed from he GTAP daabase correspond o average expeced values over many years..2.2 Inroducion of imperfec expecaions The iner emporal dimension of decision processes in our model imply ha agens have o form expecaions abou he fuure pah of economy a he ime decisions are made. Many sudies dealing wih uncerainy assume raional expecaions (WRIGHT, 200, WILLIAMS and WRIGHT,99, PRATT and BLAKE, 2007), which means ha ha economic agens have he same knowledge as economiss abou markes funcioning and ha expeced prices are hose corresponding o he economic model (MUTH, 96). However, according o some auhors, expecaions of economic agens are no raional in he Muh sense (BOUSSARD, 996, ROSSER and KRAMER, 200), due o he coss generaed by he acquisiion of informaion. As formalized by Ezekiel (938) in his famous Cobweb heorem he non raionaliy of farmers can cause expecaion errors o spread over ime and induce endogenous flucuaions of marke prices. This endogenous price volailiy has besides ofen been used o jusify public inervenions in agriculural markes (BOUSSARD, 2006). Assuming ha farmers have he righ informaion concerning he own produciviy (ha hey know he disribuion of he exogenous shocks affecing he producion) seems quie obvious. On he oher hand, we consider ha he expecaions abou marke prices are non raional and hence incorporae endogenous volailiy ino our model. As poined ou by NEWBERY and STIGLITZ (98), if some farmers have imperfec expecaions, he exisence of privae sockholdings behaviours can induce serial correlaion and make pas prices informaive. So, even if exogenous produciviy shocks are independen over ime he use of pas informaion o form expecaions abou he fuure is jusified in our case. For ha purpose we rely on NERLOVE s work (958) who proposed a formalisaion for adapive expecaions based on pas informaion. These Nerlovian expecaions are such ha agens ake he pas errors ino accoun o form he new expecaions: P ˆ P ˆ P P ˆ P P ˆ, ˆP denoing expeced prices and P observed marke prices, 0 can be seen as a measure of he adjusmen speed of expecaions. In fac he lower, he slower expecaions adjus o marke changes. An exreme case of Nerlovian expecaion arises when equals : he economic agen only considers he previous period o form his expecaion. These are called naïve expecaions..3 Inroducion of sockholding behaviours We focus now on he inroducion of sockholdings behaviours in our dynamic CGE model. We disinguish privae speculaive sockholdings which are held by privae sockholders seeking o make profi from price changes and public sockholdings only aimed a sabilizing marke prices. To accoun for sockholdings i is fs necessary o represen he behaviours of privae sockholders, he fs par of his secion is devoed o his issue. Then in a hd par we explain how a new sorage secor is inroduced ino he model. Finally we discuss he oher elemens ha have o be inroduced ino he model o accoun for sockholdings. 6

7 .3. Deerminaion of sockholdings behaviours A new agen, he sockholder, is inroduced ino he model. There is one represenaive sockholder in each region. This agen holds socks, can sell a par of hese socks or buy oher socks a curren marke price a each period. Le ST be he quaniy sored and k he uniary sorage cos in he region. Le A be he quaniy bough and V he quaniy sold by he sockholder. These bough and sold quaniies affec he socks: ST ST A V or ST ST wih A V The sockholder seeks o maximize his iner emporal profi which corresponds o he discouned sum of his sales minus his purchases and he sorage coss. His program can hus be expressed as: max V A k ST r i s / ST ST A V r or max s / ST ST r i k ST r The Lagrangian associaed o his program is: L kr ST ST ST r i i Solving he opimisaion program hus leads o he following fs order condiions: L ˆ P 0, 0, i, r, r r L ST P 0, k r r k r r We find here he sandard relaionship explaining sockholding behaviours (Williams and Wrigh, 99): if he cos of buying goods a ime and soring hem during one period is less han he (discouned) price a which hese goods can be sold a ime + ( P kr r ), hen sockholders will buy goods hus increasing he curren prices unil P kr r. On he conrary, if P kr hen sockholders will sell he socks hus lowering r curren marke prices unil P kr or unil he socks are null in which case he r marke is in equilibrium even if P kr. These consideraions allow us o explain r why sockholding behaviours come o miigae marke price volailiy. We hus have he complemenariy problem: 7

8 ST P k 0 r ST P k r r r 0 The sockholding behaviour will hus be in he model as: P kr ST, as we r assume non raional expecaions P ˆ is such as ˆ P P, because we assume ha, conrary o producers and like households, sockholders face marke prices a ime hey ake he decisions. Here our work differeniaes from MITRA and BOUSSARD (2008), who also assume ha sockholders form imperfec expecaions, indeed in he work, sorage is assumed o occur a he ime producion decisions are aken and no once harvess are pu on he marke..3.2 Creaion of a sorage service secor Soring a commodiy generaes coss paid by privae or public sockholders and made up, for insance, of he ren of grain silos and of he wages of workers who carry ou sock handling. In order o deermine hese facor incomes we inroduce a sorage service secor in our model. This secor uses labour and capial facors which are combined hrough a Consan Elasiciy of Subsiuion (CES) funcion o produce he service good. The opimisaion problem of producers in his secor can hus be wrien as: min wl L wk K STr STr STr STr r s Y d K d L s Y s K K I STr ist STr STr r r r r STr r r STr r STr STr i ST r wih Y STr he supply of sorage service, L STr and K STr he quaniies of labour and capial, wl STr and wk STr he uniary income, d a share parameer, a produciviy parameer and he elasiciy of subsiuion beween labour and capial. Solving his program leads o he zero profi condiion: PSTr ST wlstr LSTr wkstr K STr, i which s equaion will allow us o deermine he uniary sorage coss kr P STr. Furhermore, as he capial sock in his secor, as in oher secor, is subjec o adjusmen coss, sorage capaciy a one period is limied even if no sorage bound is explicily imposed. The specificaion of his sorage service secor differeniaes our work from Herel a al. (2003) who also incorporae sockholdings in a CGE model bu consider ha sorage is cosless and fix and exogenous sorage capaciy limi., 8

9 .3.3 Equilibrium condiions To ake sockholding ino accoun in our CGE model some condiions insuring markes equilibrium have o be modified. Fs, supply and demand of goods for sorage modify he equilibrium marke prices. So, he marke equilibrium condiions deermining marke prices now includes beginning-of-period socks on he supply side and end-of-period socks on he demand side. Then, as in HERTEL e al. (2005), in our model privae sockholdings are considered as a form of invesmen and are hus financed by savings. This comes o modify he equaion ensuring he equaliy beween invesmens and savings a world level and deermining he world ineres rae. So, when sockholdings are inroduced in he model, equaion Y M Q P becomes S PI I r becomes Y M ST Q ST P, and equaion r r i S PI I P ST kr ST r, wih Y he quaniy of good i r r i r i r i produced in region r a ime, M ne impors and Q he quaniy consumed in region r..4 Execuion of he model Our dynamic model is solved period by period and in wo seps for each period. This sequencing of he model resoluion deserves some explanaion. As we already menioned, consumers and producers base he decisions on expeced fuure marke prices. If a produciviy shock occurs a, say, he fs period, afer agriculural producers have decided how much o produce base on he expecaions abou he fuure, hen he effecive realized oupu quaniies won be he same as he ones farmers had expec. On he oher hand he oher economic agens observe marke condiions and hus know he shocked marke prices a ime hey ake he decisions. Thereby he aforemenioned dynamic model has o be solved in wo seps: in a fs sep he model is solved wih a produciviy value equal o E shock, he oucome corresponds o wha agriculural producers plan for he fuure period and hus provides he level of facors hey will use o produce; hen in a second sep he model is solved wih he shocked produciviy bu his ime he levels of facors used are se equal o hose deermined in he fs sep and are hus exogenous; he agriculural supply is hen deermined by he producion funcion, he oucome of he model corresponds o wha effecively happens on markes, a leas for he fs period. In he second period, he fs sep is re-execued aking ino accoun he new levels of sockholdings and capial socks which resul from he fs period and he new expecaions of agriculural producers, and he second sep is re-execued aking ino accoun he new value of he random produciviy parameer. And so on. This sequencing of he model resoluion allows us o accoun for he ime lag exising beween producion decisions and harvess in agriculural secors. 2. Simulaions and resuls The main purpose of his aricle is o build a fully dynamic general equilibrium model aimed a assessing he effecs of sockholding behaviours on agriculural prices volailiy and able o See Femenia and Gohin, 2009 for more informaion on closure rules and on he necessiy of inroducing his endogenous ineres rae in he case of imperfec expecaions 9

10 accoun for he endogenous dimension of his volailiy. Having described he srucure of his model in he fs par, his second par is devoed o he resuls of some simulaions which are conduced as an illusraive purpose, in order o have a fs insigh of he impacs of he model specificaions on he simulaed effecs of sockholdings behaviours on marke volailiy. 2. Definiion of simulaions In hese simulaions we focus on he European whea secor, which is assumed o be he only secor producing a sorable commodiy, and sudy he impacs on his secor of sochasic supply shocks arising in oher regions of he world during 25 periods. Assuming ha sockholdings only concern one secor and one region in he world is obviously unrealisic, bu we recall ha hese simulaions are conduced as an illusraive purpose; his assumpion is made in order o beer idenify he differen simulaed effecs and so o ease he inerpreaion of he resuls. 2.. Daa We use he 6 h version of he GTAP daabase which conains daa corresponding o he year 200 as a benchmark. Daa are aggregaed in 27 secors, among which 2 are agriculural secors, and in 3 regions: he European Union (EU), he Unied Saes (US) and he Res of he World (RoW). As menioned in par, we add a new secor producing sorage services. As he GTAP daabase was iniially aimed a being used in a saic framework we need o make some assumpion o calibrae he daa for our dynamic model: we assume ha he iniial ineres rae r, he ime preference parameer and he uni capial insallaion cos are all equal o 5%. Then, as in Femenia and Gohin (2009), we assume ha he iniial poin is a seady sae. This assumpion, which faciliaes he calibraion of he oher dynamic parameers, implies ha prices are sable and, in fac, ha privae sockholdings are null. However, he CES form of he producion funcion in he sorage service secor does no allow for null producion. To overcome his issue we also assume ha some precauionary whea socks, represening 0% of whea producion, are held by he public secor in he European Union. These precauionary socks are consan over ime and hus have no effec on price volailiy. Finally, in he sandard case, he expecaion adjusmen parameer is se o /0, and he elasiciy of subsiuion beween producion facors in he sorage service secor is se o 0.8. We conduc some sensiiviy analysis o hese parameers, he resuls of which are presened in he las par of his secion Characerisics of markes volailiy The price volailiy in our models resuls from producion shocks occurring in he Res of World s whea secor. These shocks can lead agriculural producers o make misakes when hey anicipae forhcoming prices. A fs sep in our simulaions is hus o generae he shocks affecing he produciviy parameer. The value of whea, RoW whea, RoW calibraed from he GTAP daabase, and 0

11 corresponding as we have seen o he mean value of he random parameer shock whea, RoW, is.95. We recall ha shock,,, wih N 0, ². whea RoW whea RoW Calibraing he value of ² is no a rivial ask. Indeed, he daa available, like hose from he Food and Agriculure Organizaion (FAO) ha have been used by Herel e al. (2003) o characerize he exogenous producion volailiy in here model, concern he producion quaniies or yield, bu hese daa resul in fac from producers decisions, among ohers, and no only from exogenous shocks. So, as we will see laer, he volailiy of quaniies produced can be much higher han he volailiy of produciviy shocks, especially when marke agens are assumed o have imperfec expecaions. For hese reasons, in our sandard case we se he value of ² o 0.9% an hen conduc some sensiiviy analysis of he resuls o his value. The 25 sochasic exogenous shocks are hus generaed according o a normal disribuion N 0, 0.9%. They are ploed on Figure and Table represens he main characerisics of he disribuion. 6% 4% 2% 0% % -4% -6% -8% Figure - Produciviy shocks Table below repors he main characerisics of he 25 produciviy shocks generaed according o he above menioned disribuion. Mean S. D. a Fs-order a-c b -0.4 % 2.7 % -0.8 Table -Disribuion characerisics of produciviy shocks a Sandard Deviaion, b Auocorrelaion 2..3 Benchmark resuls Before focusing on he impacs of sockholding behaviours, some aenion mus be paid o he oucome of our dynamic CGE model before he inroducion of sorage. These resuls will be used as a benchmark o assess he effecs of privae sorage. The evoluions of whea oupu in he hree regions of he world are represened on figure 2 and he evoluions of whea price are represened on figure 3. The fs hing o noe is ha, even if exogenous produciviy shocks occur in he RoW only, he whea producion flucuaes in he same way in he EU and, o a lesser exen, in he US. The same phenomenon arises for prices: price flucuaions are synchronous. Besides, as illusraed in able 2, whea prices in all regions are highly correlaed. This synchronism is of course parly due o rade exchanges beween regions, bu, while if agens were raional his rade could allow a dampening in marke flucuaions a world level via a risk sharing mechanism, in case of imperfec expecaion marke flucuaions synchronise and are

12 amplified a world level. This illusraes one imporan criicism agains agriculural rade liberalisaion (Boussard e al., 2005). 30% 80% 20% 40% 0% 0% -0% % -40% % -80% EU25 RoW US Figure 2- Whea oupu (% change compared o he baseline) EU25 RoW US Figure 3- Whea price (% change compared o he baseline) 2 EU RoW US EU RoW US Table 2 - Correlaion beween whea prices Then, produciviy shocks are comprised beween -6% and +6% (see figure ) bu oupu flucuaions are much more imporan: a producion increase of more han 20% is observed in he RoW in period 20 and a nearly 5% decrease arises in he EU in period 23, as shown on figure 2. The mechanisms explaining hese marke evoluions are as follows. If a one period a negaive produciviy shock a has occurred in he RoW, leading o whea price and capial income in whea secor o increase, a he nex period whea agriculural producers re-adjus he price expecaion according o he previously observed price increase: hey expec marke price higher han he iniial price, and so plan o produce more. Then, if a posiive produciviy occurs leading he harves o be even larger han ha which producers had expeced, he whea marke price decreases. A he same ime he increase of capial income observed during he previous period leads producers o expec an increase of capial income for he forhcoming periods and so o make new invesmens which will lead o an increase in he capial for he hd period. This increase of capial sock can lead producers o no decrease he producion as much as hey should if hey expec a price decrease for he fuure. So he marke volailiy originaing from exogenous produciviy shocks is amplified by he linked imperfec price and facor reurn expecaions, and his endogenous aspec can even generae sudden price peaks as in periods 5,, 8 and 24 (see figure 3). Indeed, hese periods follow periods where posiive shock occur (see figure ), so whea producers expec a marke price decrease, his price expecaion, combined wih he fac ha invesmens in previous period were small and ha he capial sock has jus decreased, leads hem o plan o produce less han iniially. Because of a negaive produciviy shock, he realized harves is acually much lower han iniially see figure 2). As whea demand is quie price inelasic, his large decrease of producion induces a very large price increase. Following hese peaks, producers re adjus he expecaions and marke prices recover lower levels for he nex periods. In he EU and he US realized producers oupu are equal o wha hey plan bu, as marke prices are affeced by hose of he RoW, his does no preven hem from making expecaion errors leading o endogenous price flucuaions. These resuls illusrae he relaive imporance of endogenous compared o exogenous marke flucuaions. Indeed, as shown by in able 3, exogenous produciviy shocks in he RoW whea secor characerized by a 2.7% sandard deviaion can generae oupu flucuaions

13 characerized by a sandard deviaion 3 imes higher. This comes o illusrae he difficuly o calibrae he disribuion of produciviy shocks based on producion daa. The sandard deviaions around 26% for whea are in accordance wih he flucuaions observed on figure 3. Oupu changes Price changes Mean S. D. Mean S. D. Fs-order a-c EU.9% 7.74% 2.05% 26.77% 0.03 Whea RoW.3% 8.82% 2.07% 26.3% 0.03 US.53% 3.27%.77% 25.99% 0.03 EU -0.05% 0.97% 0.64% 4.59% 0 Oilseeds RoW -0.3% 0.85% 0.6% 4.37% 0 US -0.07% 0.48% 0.63% 4.48% 0 EU -0.05% 0.82% 0.39% 3.4% 0 Oher Cereals RoW -0.8%.32% 0.62% 4.59% 0 US -0.07% 0.45% 0.45% 3.33% 0 Table 3 - Disribuion characerisics of oupu and price changes compared o he iniial values Table 3 also shows ha oher secors relaed o whea are also impaced in all regions: as he whea producion is a he mean higher han is iniial value, he oilseeds and oher cereals producions are lower a he mean which leads o some mean prices increases. The sandard deviaions of oupu and prices in hese secors are no negligible, even if no as high as in he whea secor. So exogenous produciviy shocks arising in he RoW whea secor spread o all regions and o several secors, generaing marke flucuaions amplified by he non raionaliy of marke paricipans. As a maer of fac regional incomes are also impaced by hese shocks even if hey occur in only one region, one secor: he sandard deviaion of regional income change is equal o 0.08% in he EU, 0.% in he RoW and 0.2% in he US (see Table 6). 2.2 Impac of sockholdings Having described he oucome of he model wihou sorage, we now rely on he impacs of sockholding behaviours concerning he European whea secor on he resuls Sandard case In wha we call our sandard case, we se he hisorical weighing parameer o /0 for all agens and he subsiuion elasiciy beween labour and capial in he sorage service secor is se o 0.8. Some sensiiviy analysis of he resuls o hese parameers, as well as o he volailiy of producion shocks, will be presened in he nex pars. Figures 4 and 5 below represen he flucuaions of whea oupu and price when sockholdings behaviours are inroduced in he model. We can noice on Figure 4 ha oupu flucuaions in he whea secor seem o be slighly smoohed compared o he case wihou socks for all regions, bu he mos ineresing resul appears on Figure 5: he European whea price is no synchronized wih RoW s and US s ones now. This is also refleced in he correlaion beween prices repored in Table 4: whereas whea prices in he RoW and in he US are sill highly correlaed, he correlaions beween whea price in he EU and prices in oher regions are reduced by abou 25%. Sockholdings behaviours end in fac o disconnec he European whea marke from world markes. 3

14 40% 80% 20% 40% 0% % -40% % -80% EU25 RoW US EU25 RoW US Figure 4 - Whea oupu (% change compared o he baseline) Figure 5 - Whea price (% change compared o he baseline) 4 EU RoW US EU RoW US Table 4 - Correlaion beween whea prices We can also see on Figure 5 ha price decreases in he European whea secors are very limied bu do no oally disappear. They are in fac bounded by he expecaions of sockholders concerning he fuure whea price and he sorage coss. Indeed, le us recall ha if sockholders expec a price rise, hey buy 0,25 0,20 whea unil: P kr r, which 0,5 prevens he whea price o decrease below 0,0 0,05 kr. On he oher hand, we can sill r 0, observe some price peaks in he EU as in oher regions, hese peaks however lower w hea socks han before he inroducion of sorage, which can his ime be aribued o sock ous as illusraed by Figure 6. Privae sockholdings behaviours hus seem o limi he occurrence of price peaks due o he effec of large producion shocks on he non raional behaviours of producers, while inducing oher peaks due o sock ous. The disribuion characerisics of oupu and price changes in he whea, oilseeds and oher cereals secors are repored in able 5. Figure 6 - Whea socks in he EU (in millions of onnes) Oupu changes Price changes Mean S. D. Mean S. D. Fs-order a-c EU 6.80% 7.27% 0.44% 2.28% Whea RoW 2.8% 2.46% -0.73% 2.89% US.0% 3.59% -.08% 9.98% 0.47 EU -0.48%.00% 0.5% 3.29% 0.07 Oilseeds RoW -0.07% 0.79% 0.24% 2.95% 0.03 US -0.02% 0.44% 0.26% 3.02% 0.03 EU -0.25% 0.69% 0.40% 2.04% Oher Cereals RoW -0.09%.9% 0.8% 3.08% 0.04 US -0.05% 0.40% 0.4% 2.% Table 5 - Disribuion characerisics of oupu and price changes

15 A fs hing o noe is ha, compared o he disribuions observed before he inroducion of sorage in he model (Table 3), prices are now lower a he mean in all regions and all secors. This resul sands in conras wih he idea, found in several sudies, ha price sabilizaion would no change mean price. In fac, as poined ou by WILLIAMS and WRIGHT (99), as long as he mean price is endogenous and he responses and feedbacks of economic agens are aken ino accoun, a sabilizaion mechanism canno keep his price unchanged. Our resuls come o illusrae he poin. In our simulaion framework, he mean price decrease is furhermore accenuaed by he limiaion of he huge sudden price increases due o non raional expecaions when sorage is no allowed. Then, he effecs of sockholding behaviours commonly found in he economic lieraure are ha sorage ends o sabilize price and desabilize producion (WILLIAMS and WRIGHT, 99).While our resuls sugges ha sockholdings behaviours effecively limi price flucuaions, paricularly in he European whea secor, oupu flucuaions are no sysemaically increased: hey increase in he US and RoW whea secors and in he EU oilseeds secors bu decrease in all oher secors presened in he able. One explanaion for he decrease of oupu flucuaions is ha he sabilizaion of price also allows producers o sabilize he expecaions and so o sabilize he producion. So, even if sockholding behaviours can play a desabilizing role as advocaed in he economic lieraure, hey also induce an improvemen of oher agens expecaions and he wo phenomenons inerac. Moreover, as poined ou by Newbery and Sigliz (98), agriculural producers are more concerned wih he sabiliy of he income han wih price or producion sabiliy. Here our resuls sugges, as presened in Table 6, ha he inroducion of sorage allows o sabilize farmers incomes in all cases, even when he producion is desabilized. Wihou sockholdings Wih sockholdings Whea Oilseeds Oher Cereals Whea Oilseeds Oher Cereals EU 6.97% 2.0% 2.00% 3.02%.40%.28% RoW 4.30% 2.47% 6.79% 32.04%.64% 4.42% US 7.%.90% 3.78% 2.29%.26% 2.34% Table 6 - Sandard deviaion of farm income changes A las poin ha differs in our resuls from wha is found in he economic lieraure is ha decrease of price volailiy induced by sockholdings behaviours is shared by all regions, even if more imporan in he EU; his differs from he view expressed by TYERS and ANDERSON (992) for insance of a risk sharing beween regions. Here sockholders allow all marke agens o improve he expecaion and hus decrease he endogenous par of price volailiy. Finally, according o DEATON and LAROQUE (992) he high fs order auocorrelaion observed on agriculural prices could be due o sockholdings behaviours. However hese auhors have no been able o reproduce his auocorrelaion wih he sorage model assuming raional expecaions. MITRA and BOUSSARD (2009) who assume imperfec expecaions have neiher been able o reproduce his feaure of agriculural prices disribuion. None of hese sudies rely on a general equilibrium model. Ye, as shown by Tables 3 and 5, in our dynamic CGE framework he inroducion sockholdings leads he fs order auocorrelaion of whea prices o be muliplied by a leas 20 in he whea secors and o also increase in he oher secors. The iner emporal dynamic dimension of our model hus allows o ake ino accoun he link beween period induced by sockholders behaviours. 5

16 2.2.2 Sensiiviy Analysis The simulaions presened above sugges ha modelling he effec of sockholdings behaviours in a dynamic iner emporal CGE framework assuming non raional expecaions can lead o resuls on marke risks quie differen from hose commonly found in he economic lieraure. Indeed, hese resuls sugges ha sorage effecively allows o sabilize prices and induces a desabilizaion of income, bu i does no necessarily desabilize oupu. Furhermore, in our resuls here is no evidence of a ransmission of price volailiy from he secor concerned wih sockholdings o oher secors or regions. However o run hese simulaions some parameers deermining he variabiliy of exogenous shocks, he form of sockholders expecaions and he elasiciy of sorage service supply, have been se o arbary values. Some sensiiviy analyses are now conduced o es he sensiiviy of our resuls o hese values. Sensiiviy o he variabiliy of supply shocks In our sandard case he supply shocks implemened in he RoW whea secor are generaed according o a normal disribuion wih a sandard deviaion equal o 3%. We now run oher simulaions for differen values of, namely %, 2%, 4% and 5%. We have repored in Table 7 below he changes in sandard deviaion of oupus, prices and incomes induced by he inroducion of sorage for he differen volailiies of exogenous shocks. Regarding hese resuls, i appears ha he volailiy of produciviy shocks has a huge influence on he effecs of sorage on price volailiy. Indeed, when he shocks are highly volaile ( 4% and 5% ), sockholdings behaviours end o sabilize whea, oilseeds and oher cereals prices in all regions, as in our sandard case, whereas when hey are less volaile ( % and 2% ), price volailiies increase wih he inroducion of sorage. On he oher hand, in all cases price volailiies increase or decrease simulaneously in all secors: here is no kind of price risk ransfer. We can also see from hese resuls ha a desabilizaion of prices induces a desabilizaion of oupu, which is due o he desabilizaion of agens expecaions, bu he conrary is no necessarily rue: even when sockholdings behaviours reduce price risk, hey do no necessarily sabilize oupu. However, farm income is sabilized in all cases. 6

17 % 2% 4% 5% EU 2.6% 5.7% -45.2% -65.9% Whea RoW 3.9% 5.% -4.3% -40.9% US 3.8% 3.2% -20.2% -45.9% EU 0.0% 7.8% -36.7% -52.9% Prices Oilseeds RoW.9% 2.6% -32.8% -56.0% S.D. US 0.0%.6% -36.8% -55.7% EU 0.9%.0% -29.3% -50.4% Oher RoW 0.9% 0.3% -32.7% -52.9% Cereals US 0.9% 0.6% -33.3% -52.4% EU 8.4% 8.4% -32.2% 6.4% Whea RoW 6.0% 0.3% 6.5% -34.7% US 9.8% 4.2% -20.0% -37.% EU 4.8% 28.6%.6% -45.9% Oupu Oilseeds RoW 3.0% 2.2% -.8% -52.2% S.D. US 0.0% 25.0% -0.8% -48.0% EU 3.7% 23.% 8.8% -3.5% Oher RoW 4.5% 7.4% -9.7% -48.6% Cereals US 0.0% 4.3% -0.0% -4.7% EU -.43% -.3% -.6% -.0% Whea RoW -.82% -.90% -.98% -.99% US -.43% -.6% -2.0% -2.07% EU -2.04% -2.02% -2.08% -.84% Income Oilseeds RoW -2.05% -2.05% -2.06% -2.04% S.D. US -2.05% -2.04% -2.06% -2.08% EU -.92% -.70% -2.09% -2.08% Oher RoW -2.05% -2.06% -2.04% -2.02% Cereals US -2.02% -2.00% -2.06% -2.07% Table 7 Changes in Sandard Deviaions induced by he inroducion of sorage in he model 0,4 0,4 0,3 0,3 0,2 0,2 0, 0, 0, % % 4% 5% Figure 7 - Whea socks in he EU (in millions of onnes) Sensiiviy o he expecaions of sockholders In our sandard case we consider ha sockholders form he expecaions in he same way han he oher agens in he model: he hisorical weighing parameer is se o /0 for all of hem. However, as shown by CHAVAS (999), expecaions of economic agens are heerogeneous, and one can presume ha speculaive sockholders may have expecaions differen from oher agens. We invesigae here he impacs of differen sockholders expecaion schemes on he volailiy of markes. We hus run some simulaions considering differen values for he hisorical weighing parameer, namely /3, /5,/30 and /50. The 7

18 changes in sandard deviaions of prices, oupus and incomes induced by sockholdings behaviours are repored in Table 8. A fs hing o noe is ha in all cases here he inroducion of sorage allows o sabilize prices in all regions, and ha he lowes is, ha is o say he more pas informaion is aken ino accoun by sockholders, he more he price volailiy is reduced by he inroducion of sorage in he model. Oupu flucuaions are also dampened when decreases, even if we can sill no conclude wheher sockholdings behaviours end o increase oupu variabiliy or no. As for /0, he volailiy of farm income decreases for /30 and /50 in all regions wih he inroducion of sorage in he EU. However, for /3 and /5, ha is o say when sockholders reac more quickly han oher agens o price news, or said in anoher way, when hey ake less accoun of pas informaion han oher agens, sockholding behaviours in he EU whea secor end o increase he volailiy of farm income in ha secor, bu no in oher secors. These resuls come o confm wha FEMENIA and GOHIN (2009) poined ou in he paper on he role of expecaion schemes: he dynamic of markes is smoohed when agens slowly reac o price news. /3 /5 /30 /50 EU -34.6% -48.0% -54.% -55.% Whea RoW -4.5% -9.4% -6.2% -4.% US -2.0% -26.% -23.% -20.9% EU -24.2% -29.9% -35.0% -36.9% Prices Oilseeds RoW -24.2% -27.2% -32.9% -33.% S.D. US -26.4% -30.3% -36.6% -38.% EU -20.5% -23.7% -28.3% -29.4% Oher RoW -23.6% -27.0% -32.5% -33.6% Cereals US -23.4% -27.0% -32.6% -33.7% EU 27.4% 60.7% -24.4% -25.3% Whea RoW 4.2% 28.8% 44.2% 46.4% US 40.% 7.% 8.6% 9.5% EU -.0% -5.9% -5.9% -5.9% Oupu Oilseeds RoW -6.% -6.8% -9.% -8.3% S.D. US -2.2% -4.4% -.% -.% EU 3.% 2.%.0%.0% Oher RoW -2.4% -3.5% -5.9% -5.9% Cereals US -2.% -4.2% -8.3% -8.3% EU 46.93% 24.89% % % Whea RoW -29.5% % -24.5% -2.08% US % -36.0% % % EU % % -5.36% -5.09% Income Oilseeds RoW % % % -54.0% S.D. US -3.88% % % % EU -3.80% % -6.88% -6.35% Oher RoW % % % % Cereals US % % % % Table 8 - Changes in Sandard Deviaions induced by he inroducion of sorage in he model Sensiiviy o he elasiciy of sorage supply In his las analysis we focus on he sensiiviy of our resuls o he elasiciy of he sorage supply. Indeed, he sorage service secor, labour and capial are used o allow he sorage of physical quaniies. These facors are combined hrough a CES producion funcion characerized by a subsiuion elasiciy beween facors. A change in hus induces a change in he supply elasiciy of he sorage service. In our sandard framework, is se o 0.8, which is quie high. One could argue ha sorage resources, like silos for insance, do no 8

19 adjus so easily. Tha is why we run some simulaions for lower values of, namely 0.4 and 0.7 in order o es he sensiiviy of our resul o he elasiciy of sorage supply. Table 9 repors he changes in sandard deviaions of prices, oupus and income for hese differen simulaion seings. As one could expec he highes is, ha is o say he more easily sorage supply can adjus o sockholders demand, he more sockholdings behaviours allow o sabilize prices and oupus. However, as previously we find here ha in all cases he inroducion of sorage in he model leads o sabilize prices and farm incomes in all secors and all regions, and ha i does no necessarily desabilize oupu. The main conclusions we can draw from our model abou he effecs of sockholdings behaviours on marke risk hus seem robus o he level of sorage supply elasiciy EU -48.0% -54.% Whea RoW -9.4% -6.2% US -26.% -23.% EU -29.9% -35.0% Prices Oilseeds RoW -27.2% -32.9% S.D. US -30.3% -36.6% EU -23.7% -28.3% Oher RoW -27.0% -32.5% Cereals US -27.0% -32.6% EU 69.4% 64.7% Whea RoW 6.9% 27.2% US 0.3% 0.7% EU 8.7% 5.2% Oupu Oilseeds RoW -3.8% 2.3% S.D. US -5.6% -3.3% EU 76.3% 269.% Oher RoW -4.7% 0.0% Cereals US -8.8% -6.7% EU -.38% -.36% Whea RoW -2.% -2.% US -.93% -.95% EU -.92% -.76% Income Oilseeds RoW -2.07% -2.09% S.D. US -2.0% -2.02% EU -2.06% -.92% Oher RoW -2.09% -2.0% Cereals US -2.03% -2.04% Table 9 - Changes in Sandard Deviaions induced by he inroducion of sorage in he model Conclusion The successive reforms of he CAP quesion is price sabilizing aspec and more and more aenion is paid o privae risk managing insrumen such as sorage. The effecs of privae sorage on marke volailiy have already been widely sudied in economic lieraure. Bu almos none of hese previous sudies do accoun for he links beween producers, households and sockholders decisions as can a CGE model do. Furhermore, he iner emporal decisions of hese agens are generally no really aken ino accoun which is a drawback when sudying he effecs of an insrumen like sorage allowing for iner emporal arbages. Finally, almos all of hese sudies focus on he effec of sockholdings on exogenous price volailiy and assume raional expecaions, which does no allow for he represenaion of he endogenous, induced by marke funcioning, par of risk; ye speculaive 9

20 sockholders have someimes been blamed o increase he volailiy on agriculural markes because of he non raional behaviours. To address hese issues we build a dynamic CGE model, aking he iner emporal decisions of economic agens ino accoun, including imperfec expecaions and privae sockholding behaviours and hen conduc some simulaions. These simulaions reveal some ineresing resuls, which, even if hey do no compleely conradic he oucome of previous sudies, come o miigae hem. Whereas several sudies conclude ha sabilizaion schemes preserve mean prices, in our framework his is no he case: he expecaion improving role of sockholdings behaviours end o limi he occurrence and magniude of sudden price peaks due accumulaion of expecaion errors from oher agens; price peaks sill exis under sorage bu hey are mosly due o sock ous phenomenons, and his end o decrease marke prices a he mean. Concerning he effecs of privae sockholding behaviours on marke volailiy, which is a cenral issue in he curren debaes on agriculural marke risks, we find, like MITRA and BOUSSARD (2009) ha sockholding behaviours can in fac have a desabilizing effec on prices when exogenous shocks are small bu he effec becomes sabilizing for bigger shocks; and in ha case he price sabilizing effec is all he more imporan sockholders ake much pas informaion ino accoun o form he expecaions. Then, some sudies like WILLIAMS AND WRIGHT (99), for insance, argue ha if sockholding behaviours sabilize prices, hey also desabilize producion; his is no necessarily he case in our framework: when sabilizing prices, sockholdings behaviours also sabilize producers expecaions and producion. In he same way, he improvemen of expecaions allows o sabilize marke prices in oher secors han he secor of he sorable commodiy (whea in our simulaions), here is no ransfer of volailiy beween regions or beween secors. Furhermore, we find ha, sockholdings behaviours in he EU end o sabilize income in each region, excep if sockholders ake less accoun of pas informaion han oher agens o form he expecaions. In addiion o hese refinemens bring o he sudy of he effecs of sockholdings behaviours on marke volailiy, our framework allows o overcome some shorcomings of exising sudies. Indeed, because of he iner emporal dynamic dimension of our CGE model, we find here evidence of he serial correlaion of prices induced by sockholdings behaviours, advocaed by NEWBERY and STIGLITZ (98), ha previous sudies like hose of DEATON and LAROQUE (992), assuming raional expecaions, or MITRA and BOUSSARD (2009), assuming imperfec expecaions, could no reproduce. Furhermore, our resuls reveal he difficuly o esimae a disribuion of exogenous (for insance climaic) producion shocks based on producion daa, especially if marke agens are no fully raional, because hese daa already resul from many decisions based on agens expecaions. Finally, we mus acknowledge ha he simulaions presened here are jus conduced as an illusraive purpose: he sorage akes place in he EU whea secor only and he produciviy shocks concern whea in he res of he world only, furhermore we have no aken ino accoun he risk aversion of economic agens which can have an impac on he decisions. Furher work should be done o overcome hese limiaions and build a CGE model aimed a sudying he effec of commodiy price sabilisaion program, which as poined ou by several auhors (NEWBERY and STIGLITZ, 983, WILLIAMS and WRIGHT, 99, WEAVER and HELMBERGER, 977) should no be no wihou accouning for sorage aciviies. 20