CAPRI Modelling System Documentation

Transcription

1 CAPRI Modelling Sysem Documenaion COMMON AGRICULTURAL POLICY REGIONAL IMPACT ANALYSIS Edio: Wolfgang Biz Developmen of a egionalised EU-wide opeaional model o assess he impac of cuen Common Agiculual Policy on faming susainabiliy, J05/30/2004 Deliveable 1 Bonn, Augus 2005

2 CAPRI Modelling Sysem Documenaion COMMON AGRICULTURAL POLICY REGIONAL IMPACT ANALYSIS Edio: Wolfgang Biz Wih conibuions of: Macel Adenäue Bonn Jose-Maia Alvaez-Coque, Valencia Wolfgang Biz, Bonn Kamel Elouhichi Guillemo Flichman Eoghan Gavey, Galway Thomas Heckele Bonn Buno Heny de Fahan Tobjön Jansson, Lund & Bonn Guiseppe Palladino, Reggio Emilia Ignacio Peez, Bonn Maco Se Reggio Emilia Chisine Wieck, Bonn Page 2 of 133

3 Acknowledgemens Many people and insiuions conibued in many diffeen ways ove he yeas o he developmen of he CAPRI modelling sysem. Pof. Henichsmeye should be named fis, as he pushed he eam in Bonn o look fo suiable panes, sa pe-ess and woe a poposal using he idea of a combinaion of a EU make model fo agiculual poducs wih egional aggegae pogamming models, which led o he fis CAPRI pojec The wok would have no been possible wihou he expeience and ools developed in he conex of ealie pojecs by eam membes in Bonn such as Hans-Josef Geuel and Andea Zinl. Equally impoan wee elemens of he SPEL/MFSS and RAUMIS pojecs. In he lae case, some of he RAUMIS eam membes conibued diecly o he fis CAPRI pojec. Nex comes he eam in Reggio Emilia who invened he aconym CAPRI a good ademak is almos as impoan as he poduc iself. I is ceainly fai o menion he Euopean Commission nex, which povided he necessay addiional funds beyond he ones invesed by he newok iself. Bu he ole of he Euopean Commission wen beyond ha in many especs. The success of CAPRI would have been impossible wihou he inees and ciical feedback of ou panes a DG-AGRI in G1 and G2. Addiionally, bigge pas of he daa pa wee povided by EUROSTAT and conacs a EUROSTAT bough ciical and helpful commens o he developmen of he new mehodology and algoihm o build up a consisen and complee daa base. DG-ENV financed anohe pojec aound CAPRI and he Euopean Envionmenal Agency oped fo CAPRI as he souce fo he hed size pojecion in he conex of he Clean Ai fo Euope pogam. Remain he many people who conibued wih bis and pieces ove he yeas. I is no necessay hee o menion hem individually, many can be found as auhos of pas of his documenaion anywhee. Bu i seems impoan o sess he fac ha hei abiliy o look a he bigh side of life, o see he poenial of ideas moe han he had wok equied o ge hem woking, and o emain fiendly and helpful even unde sess made he CAPRI newok a unique expeience fo all involved. The nex ound of CAPRI pojecs again funded by DG-RSRCH has aleady saed. The edio hopes ha he success of he pojec will coninue as he fun o wok ogehe wih a devoed newok of eseaches, in many cases, now fiends. Page 3 of 133

4 Table of Conens Table of Conens 4 1 Inoducion Sucue of he documenaion Hisoy of CAPRI Oveview on CAPRI 8 2 The CAPRI Daa Base Poducion Aciviies as he coe Linking poducion aciviies and he make The Complee and Consisen Daa Base (COCO) fo he naional scale Oveview and daa equiemens fo he naional scale Esimaion pocedue Defining uppe and lowe bounds fo he esimaed value Concluding emak on he esimaion pocess Daa and esimaion goups Consisen esimaion of hecaes, yields and goss poducion Consisency of he fam and make balances fo cop poducs Consisency of hed sizes, animal poducion and balance shees The Regionalised Daa Base (CAPREG) Daa equiemens a egional level Daa souces a egional level Daa availabiliy a egional level Reading and soing he oiginal REGIO daa Mehodological poceeding Pices fo oupus and inpus Filling gaps in REGIO Mapping cop aeas and hed sizes fom REGIO o COCO definiions Pefec aggegaion beween egional and naional daa fo aciviy levels Esimaing expeced yields wih a Hodick-Pesco file The wold Daa Base 36 3 Inpu Allocaion Inpu allocaion excluding young animals, feilise and feed Backgound Economeic Esimaion Reconciliaion of Inpus, using Highes Poseio Densiy Esimaos Highes Poseio Densiy esimaion famewok How ae he esuls used in CAPRI? Inpu allocaion fo young animals and he hed flow model Inpu allocaion fo feed Esimaion of fodde pices Feed inpu coefficiens Inpu allocaion fo feilises and nuien balances Nuien balances fo NPK NPK oupu a ail 59 Page 4 of 133

5 3.4.3 The ammonia module Inpu allocaion of oganic and inoganic NPK and he nuien balance Geenhouse Gases 68 4 Baseline Geneaion Model (CAPTRD) Tend cuve Consisency consains in he end pojecion ool Consains elaing o make balances and yields Consains elaing o agiculual poducion Consains elaing o pices, poducion values and evenues Consains elaing o consume behaviou Consains elaing o pocessed poducs Consains elaing o policy Consains elaing o gowh aes Thee-sage pocedue fo ends Sep 1: Unesiced ends Sep 2: Consained ends a Membe Sae level Sep 3: Adding suppos based on exenal esuls and beaking down o egional level Beaking down esuls fom Membe Sae o egional level Calibaing he model o he pojecion Calibaing he egional supply models Calibaing he global ade model 81 5 Simulaion Scenaio Model (CAPMOD) Oveview of he sysem Module fo agiculual supply a egional level Basic ineacions beween aciviies in he supply model Deailed discussion of he equaions in he supply model Calibaion of he egional pogamming models Esimaing he supply esponse of he egional pogamming models Make module fo young animals Make module fo agiculual oupus Oveview on he make model The appoach of he CAPRI make module Behavioual equaions fo supply and feed demand Behavioual equaions fo final demand Behavioual equaions fo he pocessing indusy Tade flows and he Amingon assumpion Make cleaing condiions Pice linkages Endogenous policy insumens in he make model Endogenous aiffs unde Taiff Rae Quoas Oveview on a egional module inside he make model Basic ineacion inside he make module duing simulaions Paamee calibaion and souces fo he behavioual equaions Calibaion of he sysem of supply funcions Calibaion of he final demand sysems Linking he diffeen modules he pice mechanism Sensiiviy of he CAPRI model o he Amingon subsiuion elasiciies Fam Type Pogamming Model: a FADN-based appoach 113 Page 5 of 133

6 6.1 The CAPRI fam ype appoach CAPRI Exploiaion ools 120 Refeences 123 Annex: Code liss 124 Page 6 of 133

7 1 Inoducion 1.1 Sucue of he documenaion The documenaion is sucued as follows. The sho inoducion in chape 1 fis gives an oveview of he CAPRI aciviies followed by a sho descipion of he sysem. The es of he documen follows he pojec wokflow: he diffeen seps of building up he naional and egional daa base (chape 2), he allocaion of diffeen inpus (chape 3) and he pojecion ools needed o esablish a baseline (chape 4) ae discussed. Chape5 deals wih he scenaio impac analysis: descipion of he diffeen modules of he economic model and hei elaionships. In he las wo chapes (chapes 6 and 7) he fam ype appoach and he exploiaion ools used in CAPRI ae biefly pesened. 1.2 Hisoy of CAPRI CAPRI sands fo Common Agiculual Policy Regionalised Impac analysis and is boh he aconym fo an EU-wide quaniaive agiculual seco modelling sysem and of he fis pojec cened aound i 1. The name hins a he main objecive of he sysem: assessing he effec of CAP policy insumens no only a he EU o Membe Sae level bu a sub-naional level as well. The scope of he pojec has widened ove ime: he fis phase (FAIR3-CT : CAPRI ) povided he concep of he daa base and he egional supply models, bu linked hese o a simple make model disinguishing he EU and es-of-he-wold. In paallel, a eam a he FAL in Baunschweig applied CAPRI o asses he consequences of an inceased shae of biological faming sysem (FAIR3-CT : Effecs of he CAP-efom and possible fuhe developmens on oganic faming in he EU). A fuhe elaively small pojec (ENV.B.2/ETU/2000/073: Developmen of models and ools fo assessing he envionmenal impac of agiculual policies, ) added a dis-aggegaion below adminisaive egions in fom of fam ype models, efined he exising envionmenal indicaos and added new ones. A new pojec wih he oiginal newok (QLTR : CAP-STRAT ) efined many of he appoaches of he fis phase, and linked a complex spaial global muli-commodiy model ino he sysem. The applicaion of CAPRI fo suga make efom opions in he conex of anohe pojec impoved he way he complex ABC suga quoa sysem is handled in he model. In 2004, again a lage pojec (FP VI, N : CAPRI-Dynaspa) saed unde he co-odinaion of he eam in Bonn o ende he sysem ecusive-dynamic, dis-aggegae esuls in space, include he new Membe Saes and add a labou module and an indicao fo enegy use. A he same ime, a pojec began o apply CAPRI o analyse he effecs of bi-laeal ade libealisaion wih Medieanean counies (FP VI, N : EU-MedAgPol). In 2005, a pojec fo IPTS/JRC saed o updae and impove he fam ype model laye and o include Bulgaia and Romania. A he same ime, he SEAMLESS pojec (FP VI: ) saed, wih CAPRI used o link esuls wih a complex laye of fam ype models and fom hee o naional, EU and global makes. In SEAMLESS he fam ype laye of CAPRI will be efined and updaed, and a module fo endogenous sucual change 1 Web Sie: hp:// Page 7 of 133

8 is foeseen. In paallel, he eam in LEI, The Hague, The Nehelands, will apply CAPRI in he inegaed pojec SENSOR ( ). Duing he yeas, he sysem was applied o a wide ange of diffeen scenaios. The vey fis applicaion in 1999 analysed he so-called Agenda 2000 efom package of he CAP. Sholy afewads, a eam a SLI, Lund, Sweden applied CAPRI o analyse CAP efom opion fo milk and daiy. FAL, Baunschweig looked ino he effecs of an incease of biological poducion sysems. WTO scenaios wee un by he eam in Bonn in 2002 and Moeove CAPRI was applied o analyse suga make efom opions a egional level, linked o esuls of he WATSIM and CAPSIM models. In 2003, scenaios dealing wih he CAP efom package iled Mid Tem Review wee pefomed by he eam in Bonn (Biz e al. 2003) and adable pemis fo geenhouse gas emission fom agiculue analysed (Péez 2005). The eam in Louvain-La-Neuve, ogehe wih he goup in Bonn, analysed suga make efom opions, applying he make module linked o he egional supply models (Adenaeue e al. 2004). In 2004 followed an analysis of a compulsoy insuance paid by fam agains Food and Mouh disease by SLI and uns dealing wih mehane emission by he eam in Galway, Ieland. In he same yea CAPRI was insalled by DG-AGRI in Bussels and a baseline geneaed in ode o mach DG-AGRI s oulook pojecions. Thee eams should be menioned, as hey povided hei own funds o shae he newok and conibue o he sysem: he eams a FAT, Tänikon in Swizeland, he eam a NILF, Oslo in Noway, and he eam a SLI, Lund in Sweden. If no explicily menioned in he following, he documened feaues had been co-financed by DG-RSRCH. The documenaion as i sands now capues he sae of he sysem in sping 2004 a he end of he CAP-STRAT pojec. I is planned o updae he documenaion on a egula basis if he need aises. 1.3 Oveview on CAPRI The CAPRI modelling sysem iself consiss of specific daa bases, a mehodology, is sofwae implemenaion and he eseaches involved in hei developmen, mainenance and applicaions. The daa bases exploi wheeve possible well-documened, official and hamonised daa souces, especially daa fom EUROSTAT, FAOSTAT, OECD and exacions fom he Fam Accouning Daa Newok (FADN) 2. Specific modules ensue ha he daa used in CAPRI ae muually compaible and complee in ime and space. They cove abou 50 agiculual pimay and pocessed poducs fo he EU (see able 26 in he Annex), fom fam ype o global scale including inpu and oupu coefficiens. The economic model builds on a philosophy of model emplaes which ae sucually idenical so ha insances fo poducs and egions ae geneaed by populaing he emplae wih specific paamee ses. This appoach ensues compaabiliy of esuls acoss poducs, aciviies and egions, allows fo low cos sysem mainenance and enables is inegaion wihin a lage modelling newok such as SEAMLESS. A he same ime, he appoach opens up he chance fo complemenay appoaches a diffeen levels, which may shed ligh on diffeen aspecs no coveed by CAPRI o help o lean abou possibiliy aggegaion eos in CAPRI. 2 FADN daa ae used in he conex of so-called sudy conacs wih DG-AGRI, which define explicily he scope fo which he daa can be used, who has access o he daa and ensue he daa ae desoyed afe he lifeime of he conac. Page 8 of 133

9 The economic model is spli ino wo majo modules. The supply module consiss of independen aggegae non-linea pogamming models epesening aciviies of all fames a egional o fam ype level capued by he Economic Accouns fo Agiculue (EAA). The pogamming models ae a kind of hybid appoach, as hey combine a Leoniefechnology fo vaiable coss coveing a low and high yield vaian fo he diffeen poducion aciviies wih a non-linea cos funcion which capues he effecs of labou and capial on fames decisions. The non-linea cos funcion allows fo pefec calibaion of he models and a smooh simulaion esponse ooed in obseved behaviou. The models capue in high deail he pemiums paid unde CAP, include NPK balances and a module wih feeding aciviies coveing nuien equiemens of animals. Main consains ouside he feed block ae aable and gassland, se-aside obligaions and milk quoas. The complex suga quoa egime is capued by a componen maximising expeced uiliy fom sochasic evenues. Pices ae exogenous in he supply module and povided by he make module. Gass, silage and manue ae assumed o be non-adable and eceive inenal pices based on hei subsiuion value and oppouniy coss. The make module consiss of wo sub-modules. The sub-module fo makeable agiculual oupus is a spaial, non-sochasic global muli-commodiy model fo abou 40 pimay and pocessed agiculual poducs, coveing abou 40 counies o couny blocks in 18 ading blocks (able 19 on page 94). Bi-laeal ade flows and aached pices ae modelled based on he Amingon assumpions (Amingon 1969). The behavioual funcions fo supply, feed, pocessing and human consumpion apply flexible funcional foms whee calibaion algoihms ensue full compliance wih mico-economic heoy including cuvaue. The paamees ae synheic, i.e. o a lage exen aken fom he lieaue and ohe modelling sysems. Policy insumens cove Poduc Suppo Equivalens and Consume Suppo Equivalens (PSE/CSE) fom he OECD, (bi-laeal) aiffs, he Taiff Rae Quoa (TRQ) mechanism and, fo he EU, inevenion socks and subsidized expos. This sub-module delives pices used in he supply module and allows fo make analysis a global, EU and naional scale, including a welfae analysis. A second sub-module deals wih pices fo young animals. As he supply models ae solved independenly a fixed pices, he link beween he supply and make modules is based on an ieaive pocedue. Afe each ieaion, duing which he supply module woks wih fixed pices, he consan ems of he behavioual funcions fo supply and feed demand ae calibaed o he esuls of he egional aggegae pogamming models aggegaed o Membe Sae level. Solving he make modules hen delives new pices. A weighed aveage of he pices fom pas ieaions hen defines he pices used in he nex ieaion of he supply module. Equally, in beween ieaions, CAP pemiums ae e-calculaed o ensue compliance wih naional ceilings. CAPRI allows fo modula applicaions as e.g. egional supply models fo a specific Membe Sae may be un a fixed exogenous pices wihou any make module. The fam ype model laye may be swiched ON o OFF. Equally, he model may be used in a compaaive-saic o ecusive-dynamic fashion. Pos-model analysis includes he calculaion of diffeen income indicaos as vaiable coss, evenues, goss magins, ec., boh fo individual poducion aciviies as fo egions, accoding o he mehodology of he EAA. A welfae analysis a Membe Sae level, o globally, a couny o couny block level, coves agiculual pofis, aiff evenues, oulays fo domesic suppos and he money meic measue o capue welfae effecs on consumes. Oulays unde he fis pilla of he CAP ae modelled in vey high deail. Envionmenal indicaos cove NPK balances and oupu of climae elevan gases accoding he guidelines of he Inegovenmenal Panel on Climae Change (IPCC). Model esuls ae pesened as Page 9 of 133

10 ineacive maps and as hemaic ineacive dill-down ables. These exploiaion ools ae fuhe explained in he las chape. The echnical soluion of CAPRI is cened on he modelling language GAMS which is applied fo mos of he daa base wok and CONOPT applied as solve fo he diffeen consained (opimisaion) poblems. The diffeen modules ae seeed by a Gaphical Use Ineface cuenly ealised in C, which ineacs wih FORTRAN code and libaies which ae ine-alias dealing wih daa base managemen. Typically, hese applicaions geneae unspecific pas of he GAMS code. Exploiaion ools apply addiionally Java apples fo ineacive maps and XLM/XSLT o geneae ineacive HTML ables. Mehodological developmen, updaing, mainenance and applicaion of CAPRI ae based on a newok appoach wih is cuenly cened in Bonn. The eam in Bonn acs as a cleaing house : any changes inoduced in CAPRI ae eviewed by i and, when acceped, become pa of he mase vesion. The mase vesion, coveing daa bases, sofwae and documenaion is disibued o all paicipans of he newok usually in he conex of aining sessions which bing he newok ogehe a leas once pe yea. The CAPRI modelling sysem may be defined as a club good : hee ae no fees aached o is use bu he eny in he newok is conolled by he cuen club membes. The membes conibue by acquiing new pojecs, by qualiy conol of daa, new mehodological appoaches, model esuls and echnical soluions, and by oganising evens such as pojec meeings o aining sessions. So fa he newok appoach woked quie successfully bu i migh need evision if he club exceeds a ceain size. Page 10 of 133

11 2 The CAPRI Daa Base Models and daa ae almos no sepaable. Mehodological conceps can only be pu o wok if he necessay daa ae available. Equally, esuls obained wih a model mio he qualiy of he undelying daa. The CAPRI modelling eam consequenly invesed consideable esouces o build up a daa base suiable fo he puposes of he pojec. Fom he beginning, he idea was o ceae wheeve possible susainable links o well-esablished saisical daa and o develop algoihms which can be applied acoss egions and ime, so ha an auomaed updae of he diffeen pieces of he CAPRI daa base could be pefomed as fa as possible. The main guidelines fo he diffeen pieces of he daa base ae: Wheeve possible link o hamonised, well documened, official and geneally available daa souces o ensue wide-spead accepance of he daa and hei susainabiliy. Compleeness ove ime and space. As fa as official daa souces compise gaps, suiable algoihm wee developed and applied o fill hese. Consisency beween he diffeen daa (closed make balances, pefec aggegaion fom lowe o highe egional level ec.) Consisen link beween economic daa as pices and evenues and physical daa as fam and make balances, cop oaions, hed sizes, yields and inpu demand. Accoding o he diffeen egional layes inelinked in he modelling sysem, daa a Membe Sae level -cuenly EU27 plus Noway- need o fi o daa a egional level -adminisaive unis a he so-called NUTS 2 level, abou 300 egions fo EU25- and daa a global level, cuenly 16 non-eu egions boken down o 27 counies o couny blocks. As i would be impossible o ensue consisency acoss all egional layes simulaneously, he pocess of building up he daa base is spli in hee main pas: Building up he daa base a naional o Membe Sae level. I inegaes he EAA (valued oupu and inpu use) wih make and fam daa, wih cop oaions and hed sizes and a hed flow model fo young animals (secion 0). Building up he daa base a egional o NUTS 2 level, which akes he naional daa as given (fo puposes of daa consisency), and includes he allocaion of inpus acoss aciviies and egions as well as consisen aceages, hed sizes and yields a egional level. The inpu allocaion sep allows he calculaion of egional and aciviy specific economic indicaos such as evenues, coss and goss magins pe hecae o head. The egionalisaion sep inoduces supply oiened CAP insumens like pemiums and quoas (secion 2.4). Building up he global daa base, which includes supply uilisaion accouns fo he ohe egions in he make model, bilaeal ade flows, as well as daa on ade policies (Mos Favouie Naion Taiffs, Pefeenial Ageemens, Taiff Rae quoas, expo subsidies) plus daa domesic make suppo insumens (make inevenions, subsidies o consumpion) (secion 2.5). The basic pinciple of he CAPRI daa base is ha of he Aciviy Based Table of Accouns which oos in he combinaion of a physical and valued inpu/oupu able including make balances, aciviy levels (aceages and hed sizes) and he EAA. The concep was developed Page 11 of 133

12 end of sevenies building on simila appoaches a he fam level a he Insiue fo Agiculual Policy in Bonn and fis applied in he so-called SPEL/EU daa base. 2.1 Poducion Aciviies as he coe The economic aciviies in he agiculual seco ae boken down concepually ino poducion aciviies (e.g. copping a hecae of whea o faening a pig). These aciviies ae chaaceised by physical oupu and inpu coefficiens. Fo mos aciviies, oal poducion quaniies can be found in saisics and oupu coefficiens deived by division of aciviy levels (e.g. sof whea would poduce sof whea and saw, wheeas pigs fo faening would poduce pig mea and NPK compised in manue). Howeve fo some aciviies ohe souces of infomaion ae necessay (e.g. cacass weighs of sows is necessay o deive he oupu coefficien fo he pig faening pocess). Fo manue oupu engineeing funcions ae used o define he oupu coefficiens. The way he diffeen oupu coefficiens ae calculaed is descibed in moe deail below. The second pa chaaceising he poducion aciviies ae he inpu coefficiens. Sof whea, o pick up ou example again, would be linked o a ceain use of NPK feilise o he use of plan poecion inpus, epai and enegy coss. All hese inpus ae used by many aciviies, and official daa egading he disibuion of inpus o aciviies ae no available. The pocess of aibuing oal inpu in a egion o individual aciviies is called inpu allocaion. I is mehodologically moe demanding han consucing oupu coefficiens. Specific esimaos ae developed fo young animals, feilises, feed and he emaining inpus, which ae discussed below. Muliplied wih aveage fam gae pices fo oupus and inpus especively, oupu coefficiens define fam gae evenues, and inpu coefficiens vaiable poducion coss. The aveage fam pices used in he CAPRI daa base ae deived fom he EEA and hence link physical and valued saisics. Howeve in some cases as young animals and manue which ae no valued in he EEA, own esimaes ae inoduced. In ode o finalise he chaaceisaion of he income siuaion in he diffeen poducion aciviies, subsidies paid o poducion mus be aken ino accoun. The CAPRI daa base feaues a ahe complex descipion of he diffeen CAP pemiums allocaed o he individual aciviies. Howeve he poblem of subsidies ouside of CAP fo he EU Membe Saes emains so fa unsolved, bu is on he agenda fo fuue amelioaions. The following able gives an example fo seleced aciviy elaed infomaion fom he CAPRI daa base. Page 12 of 133

13 Table 1 Example of seleced daa base elemens fo a poducion aciviy SW HE [Sof w hea poducion aciviy] Descipion Uni Oupus SWHE Sof whea yield kg/ha STRA Saw yield kg/ha Inpus NITF Oganic and anoganic N applied kg/ha PHOF Oganic and anoganic P applied kg/ha POTF Oganic and anoganic K applied kg/ha SEED Seed inpu cons Euo 1995/ha PLAP Plan poecion poducs cons Euo 1995/ha REPA Repai coss cons Euo 1995/ha ENER Enegy coss cons Euo 1995/ha INPO Ohe inpus cons Euo 1995/ha Income indicaos TOOU Value of oal oupus Euo/ha TOIN Value of oal inpus Euo/ha GVAP Goss value added a poduce pices Euo/ha PRME CAP pemiums Euo/ha MGVA Goss value added a poduce pices plus pemiums Euo/ha Aciviy level and daa elaing o CAP LEVL Hecaes copped 1000 ha HSTY 5.22 Hisoic yield used o define CAP pemiums /ha SETR 8.63 Se aside ae % Souce: CAPRI daa base, Denmak, hee yea aveage Linking poducion aciviies and he make The connecion beween he individual aciviies and he makes ae he aciviy levels. Toal sof whea poduced is he sum of copped sof whea hecaes muliplied wih he aveage sof whea oupu coefficien. In cases like pig mea, as menioned befoe, seveal aciviies ae involved o deive poducion. The poduced quaniies ene he fam and make balances. Poducion plus impos as he esouces ae equal o he diffeen use posiions as expos, sock changes, feed use, human consumpion and pocessing. These balances ae only available a Membe Sae, no a egional level. Poducion esablishes he link o he EAA as well, as aveage fam gae pices ae uni values deived by dividing he values fom he EAA by poducion quaniies. The hee basic ideniies linking he diffeen elemens of he daa base ae expessed in mahemaical ems as following. The fis equaion implies ha oal poducion o oal inpu use (code in he daa base: GROF o goss poducion/goss inpu use a fam level) can be deived fom he inpu and oupu coefficiens and he aciviy levels (LEVL): Equaion 1 GROF io = j LEVL The second ype of ideniies efes o he fam and make balances: j IO j Page 13 of 133

14 Equaion 2 GROFio SEDFio LOSFio INTFio = NETFio NETF + IMPT = EXPT + STCM + FEDM + LOSM + SEDM + HCOM + INDM + PRCM io io io io io io io io io io The fam balance posiions ae seed use (SEDF) and losses (LOSF) on fam (only epoed fo ceeals) and inenal use on fam (INTF, only epoed fo manue and young animals). NETF o ne ade on fam is hence equal o valued poducion/inpu use and esablishes he link beween he make and he agiculual poducion aciviy. Adding impos (IMPT) o NETF defines oal esouces, which mus be equal o expos (EXPT), sock changes (STCM), feed use on make (FEDM), losses on make (LOSM), seed use on make (SEDM), human consumpion (HCOM), indusial use (INDM) and pocessing (PRCM). The hid ideniy defines he value of he EAA in poduce pices (EAAP) as sold poducion o puchased inpu use (NETF) in physical ems muliplied wih he uni valued pice (UVAP): Equaion 3 EAAP = UVAP io io NETF io The following able shows he elemens of he CAPRI daa base as hey have been aanged in he ables of he daa base. Table 2 Main elemens of he CAPRI daa base Aciviies Fam- and make balances Oupus Oupu coefficiens Poducion, seed and feed use, ohe inenal use, losses, sock changes, expos and impos, human consumpion, pocessing Inpus Inpu coefficiens Puchases, inenal deliveies Income indicaos Aciviy levels Seconday poducs Revenues, coss, Goss Value Added, pemiums Hecaes, slaugheed heads o hed sizes Makeable poducion, losses, sock changes, expos and impos, human consumpion, pocessing Pices Uni value pices fom he EAA wih and wihou subsidies and axes Uni value pices fom he EAA wih and wihou subsidies and axes Consume pices Posiions fom he EAA Value of oupus wih o wihou subsidies and axes linked o poducion Value of inpus wih o wihou subsidies and axes link o inpu use Toal evenues, coss, goss value added, subsidies, axes Page 14 of 133

15 2.3 The Complee and Consisen Daa Base (COCO) fo he naional scale Oveview and daa equiemens fo he naional scale The CAPRI modelling sysem is, as fa as possible, fed by saisical souces available a Euopean level which ae mosly cenalised and egulaly updaed. Fam and make balances, economic indicaos, aceages, hed sizes and naional inpu oupu coefficiens ae almos eniely aken fom EUROSTAT. In ode o use his infomaion diecly in he model, he CAPRI and CAPSIM 3 eams developed ou of EUROSTAT daa a complee and consisen daa base (COCO) a Membe Sae level (Biz e al. 2002). The main souces used o build up he naional daa base ae shown in he following able and diagam. Table 3 Daa iems and hei main souces Daa iems Aciviy levels Poducion Fam and make balance posiions Souce Land use saisics, hed size saisics, slaugheing saisics, saisics on impo and expo of live animals Fam and make balance saisics, cop poducion saisics, slaugheing saisics, saisics on impo and expo of live animals Fam and make balance saisics Secoal evenues and coss Economic Accouns fo Agiculue (EAA) Pices Oupu coefficiens Inpu coefficiens Aciviy specific income indicaos Policy daa Deived fom poducion and EAA Deived fom poducion and aciviy levels, engineeing knowledge Diffeen ype of esimaos, engineeing funcions Deived fom inpu and oupu coefficiens and pices Vaious souces (Official Jounal of he EU) Souce: Euosa (hp://epp.euosa.cec.eu.in), seveal bio-physical economeic sudies and Euopean Commission (hp://publicaions.eu.in/geneal/oj_en.hml) Esimaion pocedue COCO was pimaily designed o fill gaps o o coec inconsisencies found in saisical daa and, addiionally, o easily inegae daa fom non-eurostat souces in he model. Howeve given he ask of having o consuc consisen ime seies on yields, make balances, EAA posiions and pices fo all EU Membe Saes, a heavy weigh was pu on a anspaen and unifom economeic soluion so ha manual coecions wee avoided. 3 The Common Agiculual Policy Simulaion Model (CAPSIM) was developed by D. Heinz-Pee Wizke, EuoCae, Bonn (hp:// Page 15 of 133

16 COCO included daa anging fom 1985 o 2002 fo he 14 membe saes of he EU 4 a ha ime, fom he naional daa found in NEWCRONOS 5. Regading he consucion of he daa base, hee pincipal poblems had o be solved: (1) Gaps had o be filled in ime seies, eihe befoe he fis available poin, inside he ange whee obsevaions ae given, o beyond i. (2) Some ime seies wee missing alogehe and had o be esimaed, e.g. when hee ae daa on animal poducion bu none on mea oupu pe head. (3) Minimal coecions of given saisical daa, if no in line wih he accouning ideniies, had o be made. In ode o ake ino accoun logical elaion beween he ime seies o fill, and evenually o make minimal coecions in he ligh of consisency definiions, simulaneous esimaion echniques ae used in his execise. In ode o use o he geaes exen he infomaion conained in he exising daa, he following pinciples ae applied: (1) Accouning ideniies. -posiions of he make balance summing up o zeo, he diffeence beween socks as he sock change and simila esicions- consain he esimaion oucome. (2) Relaions beween aggegaed ime seies (e.g. oal ceeal aea) and single ime seies ae used as addiional esicions in he esimaion pocess. (3) Bounds fo he esimaed values based on engineeing knowledge o deived fom fis and second momens of imes seies ensue plausible esimaes and/o bind esimaes o oiginal daa. Addiionally, bounds ae consuced fom moe disaggegaed ime seies, if he aggegae is missing. (4) As many ime seies as echnically possible ae esimaed simulaneously o use he full exen of he infomaional conen of he daa consains (1) and (2). The fis hee poins can be inepeed as a kind of Bayesian appoach: addiional a pioi infomaion supplemens he esimaion. Howeve in classical Bayesian analysis, he infomaion is expessed as a disibuion of he paamees o esimae. Fo ou pupose, such a concep would be complex and inanspaen, as he fied value and no he esimaed paamee is of majo inees. Fuhe on, he saisical popeies of he esimaos ae in ou case of mino impoance -we do no need good esimaes of he paamees bu consisen, plausible and good fied values- leaving oom fo fuhe expe knowledge infomaion. The eade may noice ha he poblem is quie simila o sysem esimaion in economics. Conside a sysem of supply cuves. Given ex-pos daa, we naually wan he esimaes o fi he given daa as close as possible, bu simulaneously equie he esimaes o be in line wih economic heoy. The lae poin is ypically ensued by wo appoaches: (1) he esimaion equaions ae in line wih some opimisaion poblem in he backgound (fo example pofi maximisaion, i.e. he supplied oupus ae egessed on a funcion of pices whose funcional fom is deived fom fis ode condiions of a pofi maximisaion poblem) and (2) appopiae esicions on he paamees ensue ha he esuling sysem is in line wih 4 In CAPRI Luxemboug is aggegaed o Belgium as a NUTS 2 egion. The 10 new Membe Saes wee included in Daa fo Noway ae pocessed by COCO as well, bu naually, sem fom diffeen souces. Page 16 of 133

17 fis and second ode condiions of a pofi maximisaion poblem. The ulimae aim is he combinaion of a funcional fom and paamee esicions which allows fo boh a good fi and confomiy wih mico-economic heoy. Ou appoach is quie simila as ou goal asks fo consisen esimaes as well. Bu, hee ae wo impoan diffeences, (1) we need o coec he oiginal daa as well -ha would be he esimaed values ex-pos- and (2) i is vey complex o simply impossible o define he full se of consisency condiions ove esicions on he esimaed paamees in ou case. Insead, we inoduce explici daa consains involving he fied values fo each poin and ake he fied values lae as he conen of he daa base. The concep woks in he following seps: 1. Esimae independen end lines fo he ime seies. 2. Esimae a Hodick-Pesco file using given daa whee available and ohewise he end esimae as inpu. 3. Define suppos whee ae (a) given daa, (b) he esuls fom he Hodick-Pesco file imes R² plus he las (1-R²) imes he las known poin. The concep is pu o wok by he minimisaion of nomalised leas squaes unde consains: (1.1) min a, b, c i i i * ( y y ) y ) if y + w i, ( e sdeves ) 2 2 Equaion 4 s.. (1.2) (1.3) a + b T + c T y i l i y * i y 2 u + e * y if no y = y if y (1.4) e l e e u (1.5) Accouning ideniies defined on y * whee: a,b,c ae he paamees o esimae and descibe a polynomial end fi, y ae given and y * fied values, e he eo ems of he esimaion, sdeves is he sandad deviaion of he eos of an unconsained end line and T end. i epesens he index of he elemens o esimae (cop poducion aciviies o goups, hed sizes ec.), sands fo he yea and he subscips l and u ae he indices fo uppe and lowe bounds of he esimaes and eos. The objecive funcion minimises he sum of wo elaive squaed eos: (1) beween coeced and given daa, and (2) diffeences beween end foecass and given es. fied daa. The nomalisaion fo he eos (second em) is based on he sandad deviaion of he eo of an unconsained linea em line. The nomalisaion was necessay and helpful o eflec he fac ha he means of he ime seies eneing he esimaion deviae consideably. The nomalisaion hence leads o minimisaion of elaive eos insead of absolue ones. Page 17 of 133

18 The fied values y* a known poins will only deviae fom he given daa if he accouning ideniies canno be solved wihou coecions. In ha case, nomalised squaed coecions dive he pocess, e.g. o deemine which elemens of he make balance o coec. I should be noed ha fied value y* and eos e ae defined a unknown poins in consain (1.2) ove a polynomial end fi up o degee wo. The equaion guaanees hence compleeness in imes. The degee of he esuling polynomial fom may be less han indicaed depending on he numbe of available obsevaions. The eo ems a unknown poins ae inoduced o allow confomiy beween end esimaes and accouning ideniies. A known poins, he equaions define he eo ems, as usual in egessions. Uppe and lowe bounds esic he esimaion oucome as indicaed in (1.3) and (1.4). Fo ceain seies and obsevaions hey eflec logical bounds -as non-negaiveness o bounds aken fom engineeing knowledge. Fo he emaining cases hey ae consuced fom mean and vaiance of he known poins o avoid cuious foecass. Equaion (1.5) indicaes ha consisency esicions ae added o he fiing pocess. These esicions ae discussed in deails below. Reades familia wih he wok of he CAPRI eam in he las yeas may wonde why he auhos ae using a modified leas squaes esimao and no a Coss (CE) o Maximum Enopy Esimao (ME). The easons ae simila o he poins menioned above egading he applicaion of a Bayesian appoach: he ex-ane knowledge can be expessed mainly elaing o he esimaed value and no in elaion o esimaed paamees. Accodingly, suppos would need o be defined a leas fo he eo ems and he consisency slacks. The auhos ae convinced ha he cuen famewok can be mapped wihou geae poblems in an enopy esimao bu expecs a highe compuaional buden due o he moe complex objecive funcion Defining uppe and lowe bounds fo he esimaed value The iniial appoach fixed obsevaions a given daa and did no include bounds fo he end esimaes. Aleady fis ess showed ha he end oucomes could look ahe awkwad, especially when seveal obsevaions wee missing a he ends of ime seies, and he necessiy of bounds became obvious. If seveal elemens of a make balance ae missing, fo insance, he consisency condiion ceainly influence he oucome of he pocess, bu if o he bee is no clea befoehand. In ode o keep esimaes in a plausible ange, we defined an esimaion coido fo missing obsevaions based on a moving aveage and he vaiance of each ime seies. Naually, i became obvious immediaely ha no all given daa could be fixed. Assume, fo insance, ha all elemens fo a balance ae given, bu he balance is no closed. Such a daa consellaion would yield infeasibiliies. In ode o allow fo he necessay coecion, a igh coido aound all given daa values was inoduced. As he appoach was esed on a gowing numbe of daa ses, hese igh bounds iniially inoduced aound given daa wee moe and moe elaxed o accommodae fo inconsisencies in he oiginal daa, and ules wee inoduced o widen hem depending on daa consellaions. The code was gowing lage and lage wih ahe complex if-else ules depending on possible inconsisencies in he given daa o avoid infeasibiliies. The envisaged anspaency was in dange o be los. Afe a ciical evaluaion, he pocedue was evised, based on he following agumens. Fisly, if coecions on oiginal daa ae allowed even if hese ae aleady consisen, hee may be a sizeable ade-off beween a bee end fi fo he missing daa and coecions of he exising ones. The declaed aim was howeve o coec oiginal daa only when necessay. Secondly, any updae of he oiginal daa may povoke new inconsisencies, hus Page 18 of 133

19 asking fo lage coecion bounds, o he inoducion o evision of ules o define hese bounds. Accodingly, he soluion should be able o deec if and whee oiginal daa povoke infeasibiliies, and inoduce solely coecions a hese poins. The dual soluion fom minimising he sum of infeasibiliies fom he esimaion poblem shown above is used o define opimal coecion coidos. If he poblem is infeasible wih given bounds, he shadow values on (1.2) o (1.5) show which bounds o consains povoke he infeasibiliies. As consisency consains canno be dopped, a feasible soluion can hence only be found if bounds ae elaxed. Founaely, he dual soluion indicaes exacly which bounds o coec. Exacly hese bounds ae sepwise elaxed unil all infeasibiliies ae emoved and he opimisaion can sa. The pocess fis elaxes bounds fo he esimaes a missing poins befoe bounds aound given daa ae elaxed. Founaely, he pocess can be implemened quie easily. The gadien based solve CONOPT3 fis seaches fo a feasible soluion befoe woking on he objecive. If infeasibiliies ae found, shadow values on consains and equaions ae epoed based on esimaed gadiens fom he minimisaion of infeasibiliies. I is hence no necessay o explicily define he Lagangian funcion of poblem (1) in ode o calculae he shadow values. The saing deviaion allowed fo given daa is jus 0.01% imes he coefficien of vaiance of a linea end on he ime seies, in ode o avoid numeical poblems wih fixed vaiables. The pocess has he advanage of self coecion. If an updae inoduces a se of inenally consisen daa, he bounds fom a fome soluion ae no longe elevan, and he esimaed values ae fixed o he given ones. If an updae povokes infeasibiliies no found befoe, he pocess will auomaically look up he minimal coecion necessay o fulfil he consisency famewok. Hence, chances ae gea ha conol coss fo updaes ae small. Naually, he pocedue may yield quie cuious esimaes fo missing daa if oulies ae pesen and povoke pessue on esimaes of missing daa ove he daa consains. The manifold checks on he esuls le us howeve conclude ha such oulies ae ypically subjec o ahe lage coecions hemselves and do no have a sizeable impac on ohe seies. The ypical check is o plo he given daa agains he consisen ones fo he key ime seies, and obvious oulies usually sick o he eye due o hei high deviaion agains he oiginal daa. Discussions if and how an explici saisical oulie es could and should be inoduced in he famewok ae no ye finalised Bounds fo end esimaes The pocess of defining and elaxing bounds is discussed based on an example. The oiginal ime seies shown in ed - "EL HCOM TOMA (Given)"- in he diagam below is ahe ypical fo gaps in he aw daa. Missing values can be found boh in-beween given poins, and a he ails. The dak blue and uquoise seies show he uppe and lowe bounds fo he esimaion coido. Page 19 of 133

20 Figue 1. Example fo bounds on end esimaes in CoCo EL HCOM TOMA Given EL HCOM TOMA Esimy.Up EL HCOM TOMA Esimy.Lo Souce: Own calculaions These lowe and uppe bounds ae geneaed in he following way: In beween he ange of known poins hee he poins 1 o 14 - he esimaion channel coesponds o +/ sandad deviaion of he seies fom he nex obseved daa poin. Fo all ohe poins hee he foecased ones, 15 o 17 - he cene of he esimaion channel is defined by he neaes known moving aveage plus 0.1 imes he sandad deviaion of he ime seies The ieaive pocedue a wok The lowe and uppe limis fo he given poins ae eplaced by vey igh bounds fo he given poins befoe he solve is pu o wok, as seen in he nex diagam, wheeas he esimaion coido fo he unknown poins is unaffeced. Howeve he combinaion of hese bounds wih he consisency consains and all ohe bounds pesen simulaneously fo ohe ime seies yielded infeasibiliies in ou example. Non-zeo shadow values wee found fo hee poins (obsevaions 13, 16 and 17). Fo obsevaion 5 and 15, esimaes ae a lowe bounds, bu no shadow value was aached, so a coecion of he bounds was no necessay. Page 20 of 133

21 Figue 2. Example fo bounds on end esimaes in CoCo, coninued EL HCOM TOMA Esimy.Lo EL HCOM TOMA Esimy.L EL HCOM TOMA Esimy.Up Souce: Own calculaions The hid figue shows he esimaion bounds and esimaed values a he end of he opimisaion sage when all infeasibiliies had been emoved and he objecive funcion is a is opimum. I can be seen ha he bounds had been elaxed fo obsevaions 13, 16 and 17. All he oiginal values ae (almos) fixed. The new poins inoduced ceainly do no look like a end fi, as hey eflec he elaion beween consisency condiions, esimaed gaps and given daa in ohe ime seies. I should be noed ha he lowe bound on poin 5 is acive, pobably pulling he esimaed value fom he end line owads he neighbouing obsevaions. Figue 3. Example fo bounds on end esimaes in CoCo, coninued EL HCOM TOMA Esimy.Lo EL HCOM TOMA Esimy.L EL HCOM TOMA Esimy.Up Souce: Own calculaions Page 21 of 133

22 EL HCOM TOMA Esimy.Lo EL HCOM TOMA Esimy.L EL HCOM TOMA Esimy.Up Souce: Own calculaions Concluding emak on he esimaion pocess We may conclude ha he pocess is a ahe pagmaic one. Fisly, i ceainly ensues ha infeasibiliies ae avoided in mos insances, hus educing he conol cos fo a new daa updae. Secondly, he isk ha oiginal daa ae coeced wihou an inconsisency pesen is close o zeo. Thidly, he sweeping ail poblem of end esimaes is o a ceain exen educed by inoducing bounds. The fis ade-off is he use of an esimao wih unknown lage and small sample popeies. Secondly, he pocess equies eihe a gadien based solve using a wo-sage pocess seaching fis a feasible poin o he Lagangian of poblem (1). Thidly, he poblem as defined above can only be solved by geneal NLP solves, and is hence no easily poable o ohe sofwae plafoms as fo insance a saisical package. And finally, he esuling soluion can ceainly no be easily explained as he chosen esimaes and coecions ae he oucome of a simulaneous opimisaion poblem. Neveheless, he eams involved ae convinced ha given hei cuen esouce endowmen, he soluion is as close o opimal as possible Daa and esimaion goups The daa eneing he esimaion pocess sem all fom EUROSTAT collecions. Physical poducion saisics and balance shees ae fom he ZPA1 domain, pices fom he PRAG domain and he EAA accouns sem fom he COSA domain. Daa ae diecly conveed fom he EUROSTAT fomaed inpu files o GAMS ables wihou inemediae files via a home-wien FORTRAN ouine called DFTCON. The oiginal EUROSTAT codes ae conveed o wo dimensional iem-poduc ype codes, as fa as possible aleady in CAPRI convenions. The esimaion is caied ou independenly fo each membe sae. In ode o educe he compuaional buden and conol coss, he pocess is subdivided addiionally in he following pas: (1) Esimaion of hecaes, yields and goss poducion fo all cop poducs simulaneously. Page 22 of 133

23 (2) Esimaion of fam and make balances fo cop poducs, boken down in he following goups: (2.1) Ceeals (2.2) Indusial cops including oilseeds (2.3) Fuis (2.4) Vegeables (2.5) Wine (2.6) Fodde fom aable land (2.7) A las goup including all hose ime seies which ae no assigned o one of he above menioned (e.g. suga bees) (3) Esimaion of hed sizes, goss poducion oupu and fam and make balances fo animal aciviies and poducs, boken down in he following goups: (3.1) All aciviies and poducs elaed o he poducion and use of milk and sheep and goa mea (3.2) Cale goup (faening and aising aciviies, mea) wihou daiy cows (compised in 3.1) (3.3) Pigs (3.4) Pouly In he following secions he specific daa consains fo he diffeen esimaion poblems will be discussed in fuhe deail Consisen esimaion of hecaes, yields and goss poducion Code: coco\coco_esimc.gms Consisen esimaion of hecaes, yield and cop oupu goss poducion is he fis of hee sepaaely defined esimaion poblems. The main ouline of each of he esimaion poblems is defined above in poblem (1). We will hence concenae on he deailed descipion of he accouning ideniies esicing he esimaion. The simulaneous esimaion of cop aciviy levels, yields and goss poducion is consained by he following equaions 6 : Poducion of oupu equals aciviy level (hecaes) muliplied wih O-coefficiens (Yields) (GofD_) 7 Equaion 5 whee: j GROF = LEVL OUTP * i j j * j, i denoes poducion aciviy 6 As fa as possible we will use he codes and he unis as documened in he daa base. If no, hey will be specified unde he equaion. Fuhemoe, we neglec he ime and Membe sae index in he equaion fo bee eadabiliy. Howeve i should be clea ha each consisency condiion mus hold in evey yea and fo each Membe sae. 7 Name of equaion in he code. Page 23 of 133

24 i denoes poduc GROF physical goss poducion (ypically measued in 1000 ) LEVL aciviy level (measued in 1000 ha) OUTP Oupu coefficien (yield, ypically in kg/ha) Consisency beween hecaes of he aggegae (e.g. CERE, OILS) and sum of hecaes ove componens of ha aggegae (e.g. SWHE, RYEM...) (ConsisL_) Equaion 6 LEVL k = LEVL j j k whee: k denoes he aggegae One of he aggegaes is oal uilisable aeas, so ha adding up of all cop levels o he available land is guaaneed. Consisency beween poducion of aggegae and sum of poducion ove componens of he aggegae (ConsisG_) Equaion 7 GROF k = GROF i i k The esuling daa copy yields, cop aeas and cop poducion ae fixed in any following esimaion and no longe subjec o coecions. In many cases, he daa eneing he esimaion pocess need o be added up fom single ime seies. In all hese cases, he value is only calculaed, if all elemens on he igh hand side ae non-zeo (o if a ceain elemen is zeo eveywhee). Consequenly, aggegae ime seies show ypically a high numbe of gaps. In ode o bind he esimae fo aggegae close o given daa, a sum of all non-zeo daa of componens even if daa fo some componens is missing is used as lowe bound if he consuced bound is no highe. As an addiional safeguad, he uppe limis fo cop yields ae educed o engineeing maxima Consisency of he fam and make balances fo cop poducs Code: coco\coco_esimb.gms This secion descibes he make and fam balances and how consisency of hei elemens is achieved. Fuhemoe, hese balance posiions ae linked o he EAA by uni value pices. The following able, aken fom he EAA handbook, shows o he lef physical esouces, in he middle column physical uses/sinks and indicaes in he igh column if he physical iems in he middle column ae valued in he EAA. The diffeence o he old EAA concep and hence he old SPEL/EU and CAPRI daa base ( ) should be noed. The old concep valued solely sales beween agiculual unis and non-agiculual ones, plus change in socks on fam plus own final consumpion. Any ineacion beween agiculual unis and inbeween aciviies on he same uni wee no aken ino accoun in he old EAA. The new definiion excludes only losses on fam and ina-aciviy use (seeds, milk fo livesock feed, wine gapes, olives fo olive oil, haching eggs, animal by-poducs used in cop poducion as sluy, manue) fom being valued in he EAA. As discussed lae on, ade in young animal beween fams of a Membe saes is excluded as well, in conadicion o he oveall concep. Page 24 of 133

25 Table 4 EAA definiion accoding o EUROSTAT Resou ces Goss poducion - Losses = Usable oupu Uses Sales (oal, excluding ade in animals beween agiculual holdings) Change in socks (wih poduces) Own-accoun poduced fixed capial goods (planaions, yielding epea poducs, poducive animals) Own final consumpion ( of agiculual poducs) Pocessing by poduces (of agiculual poducs, sepaable) Ina-uni consumpion: fo he same aciviy:(seeds, milk fo livesock feed, wine gapes, olives fo olive oil, haching eggs) fo a sepaae aciviy: Agiculual oupu of he agiculual indusy X X X X X Souce: EUROSTAT 2000, p. 42 Cop poducs used in animal feed (ceeals, oilseeds, fodde cops, makeable o no, ec.) Animal by poducs used in cop poducion (sluy, manue) The change inoduced by he new definiion of he EAA allowed fo a simplificaion of he fam and make balances compaed o he old CAPRI and SPEL/EU daa base. The valued posiions ae now poducion minus losses, seed and inenal use (manue, animal flows inside he seco), and a spli up of sock changes, human consumpion and feed beween fam and make balance is no longe necessay. X Pimay, non-pocessed poducs Consisency of fam balance posiions (ConsF_) Fam sands fo he absac naional fam as he aggegae of all individual fams. The fam balance is buil o mimic he valuaion scheme of he EAA. In ode o find he physical equivalen o he EAA NETF, losses on fam (LOSF), seed use on fam (SEDF) and inenal use of animals and manue (INTF) deduced fom goss poducion (GROF). All posiions ae in physical ems. Daa fo he fam posiions ae available in EUROSTAT fo ceeals, only. The INTF posiions ae zeo fo cops by definiion. Equaion 8 GROF = SEDF + LOSF + INTF + NETF i i i i i Page 25 of 133

26 Consisency of make balance posiions (ConsMkb_) The make balance is an accouning sysem which summaises ansacions of all agiculual oupus on makes. Resouces ae ansacion of he agiculual seco (NETF) o makeable poducion in case of seconday poducs (MAAR) plus oal impos (IMPT) plus impos as live animals (IMPL) in case of mea. Uses ae expos (EXPT), seed use on make (SEDM), losses on make (LOSM), feed use (FEDM), indusial use (INDM), pocessing o seconday poducs (PRCM) and sock changes (STCM). Any saisical adjusmens epoed by EUROSTAT ae se o zeo. The eade is eminded ha a disincion beween seed and losses on fam and make is available fo ceeals, only. Equaion 9 NETF i pimaies + MAPRi sec ondaiesimpti + IMPLi Mea = EXPT + SEDM i + PRCM + HCOM i i + LOSM + FEDM + INDM i + STCM i i + SADM i i i Consisency o Economic Accouns of Agiculue (ConsisEAA) The connecion beween he EAA valued posiion (EAAP: EAA value a poduce pices) and he fam balance posiion NETF ae uni values a poduce pices (UVAP): Equaion 10 EAAP = NETF UVAP / 1000 i i * i I should be noed ha he NETF posiion is deived fom daa in EUROSTAT s ZPA1 domain. Accoding o he new EAA handbook, membe saes ae equied o epo boh physical and valued daa along wih uni values pices in he EAA. One may hence quesion ou decision o use ZPA1 daa insead of he physical daa fom he EAA. Fis of all, no all membe saes epo quaniies and uni value pices. Secondly, diffeences beween physical posiion NETF, deived fom he fam and make balances, and he EAA physical values ae sizeable in many cases. Using he EAA daa would hence lead o inconsisencies beween he fam and make balance posiions. Neveheless, we ae lef wih he poblem ha he diffeences exis and ae had o inepe, and can lead o asonishing uni values, boh egading hei level as hei developmen ove ime, especially in a coss-couny compaison. We hope ha some of he diffeences can be claified in fuue by conacs o EUROSTAT. Consisency beween fam and make balances posiions fo seed use and losses and oal losses and seed use (ConsP_) The spli up of posiions in fam and make iems exiss solely fo seeds and losses, he only posiions whee a spli-up is necessay o accommodae he new EAA. As indicaed above, seed and losses on fam ae epoed fo ceeals, only. Equaion 11 LOST SEDT i i = LOSF + LOSM i = SEDF + SEDM i Consisency beween iems of fam and make balance fo aggegaes and componens (ConsisG_) The condiions ensue as above fo he cop poducion esimaion goup consisency beween aggegae and membe of he aggegae, fo example ha ceeals impos ae equal o he impos fo sof whea, baley... i i Page 26 of 133

27 Equaion 12 whee: RESPOS RESPOS k = RESPOS i i k compises all posiions elevan fo fam and make balance (excep pices and SADM) The case fo seconday poducs: Thee ae a few seconday poducs no valued by he EAA compised in he daa base, namely oils and cakes fom oilseeds, sach and ice. Fo hese poducs, an explici connecion beween pocessing of pimaies and makeable poducion is esablished. Pocessing elaion: consisency beween pocessing of pimay poducs (PRCM) and makeable poducion (MAPR) of seconday ones (Pocess_) Equaion 13 j * j PRCM PRCY = MAPR ji i whee: PRCY ji Pocessing yield, e.g. kg of soya oil exaced fom one kg soya Seed use is by definiion no possible fo seconday Consisency condiions fo sock changes and socks Modelling of socks and sock changes is impoan fo boh, pimay and seconday poducs. Sock flow beween he yeas (SocksLM_) Equaion 14 STKM 1 + STCM = STKM whee: STKM Sock level Limi sum of sum of sock changes ove ime o 10% of poducion (SocksAML_. SocksAMH_) Mean("GROF ") + Mean("IMPT ") STCM < *0.1 Mean("MAPR ") Equaion 15 + Mean("GROF ") + Mean("IMPT ") STCM > *0.1 + Mean("MAPR ") The condiions ae inoduced o keep he esimao fom piling up socks ove ime. Page 27 of 133

28 2.3.8 Consisency of hed sizes, animal poducion and balance shees 8 As in he secions befoe, he aim of his pa of he model is o consuc a eliable daa base egading livesock aciviy levels and hei especing I/O coefficiens in line wih naional saisics. In geneal, he layou of his esimaion poblem follows he same seps and uses paly he same equaions as he wo poblems descibed befoehand. Animal aciviy levels elae (a) o an aveage of he counings in he cuen yea (milk poducion, laying hens and sows) especively an aveage of las yea s Decembe couning, cuen yea s July/Augus couning and cuen yea s Decembe couning (suckle and daiy cows) o (b) o slaugheed plus expoed minus impoed heads (faening aciviies: beef, heifes, male and female calves, pok, sheep and goa, pouly) o (c) o young animals aised (male and female calves aising, heifes aising), measued in 1000 es. 1 million (laying hens and pouly faening) heads. The esimaion poblem is paly defined by means of he following aleady known equaions fom chape 3: (1) Poducion of oupu equals = aciviy level muliplied wih O-coefficiens (GofD_) (3) Consisency of poducion of aggegae o sum ove poducion of he componens (ConsisG_) (4) Consisency of make balance posiions (ConsMkb_) (5) Consisency o Ecomomic Accouns of Agiculue (ConsisEAA_) (7) Consisency of aggegae iems of fam and make balance o sum ove iems of componens (ConsisG_) (8) Pocessing elaion: consisency beween esouces of aw poduc and makeable poducion of pocessed ones (Pocess_) (9) Consisency of make balance posiions (ConsMkb_) (10) Sock flow beween he yeas (SocksLM_) (11) Limi sum of sock changes o 10% of poducion (SocksAML_. SocksAMH_) Besides hese, he addiional daa consisency condiions fomulaed below consain he esimaion of hed sized, animal oupus and hei balance shees: Definiion of young animal inpu (IaniH_) The need of each ype of young animal is defined as follows: Equaion 16 GROFiy = SLGH iy + EXPLiy IMPLiy + HEIR. LEVL iy = IHEI whee: SLGH slaugheed heads EXPL expoed heads of live animals IMPL impoed heads of live animals HEIR.LEVL numbe of heifes aised, only added if iy=ihei 8 This secion is mainly based on a daf wien by Tobjön Jansson and Andes Bäcksand (SLI, Sweden). Page 28 of 133

29 Definiion of mea oupu (IaniT_) The mea oupu fo each ype of animal is defined as follows: Equaion 17 aac iy LEVL aac OUTP acc, mea IYANI acc, iy = SLGT iy + EXPT iy IMPT whee: aac iy animal aciviies using young animal caegoy iy as inpu OUTP Mea oupu coefficien YANI Young animal inpu coefficien SLGT slaugheed heads EXPT expoed ons of live animals IMPL impoed ons of live animals The eade should noe he diffeence o he daa base. Hee, he mea oupu is defined pe slaugheed animal, whee in he daa base i is elaed o he aciviy level. Take daiy cows as an example: duing esimaion, he mea oupu coefficien elaes o one cow slaugheed, and oal mea oupu is he cow hed imes he eplacemen ae imes he cacass weigh. In he daa base, he mea oupu coefficien is pe cow (= cacass weigh imes eplacemen ae). Uses equal esouces fo young animals (YaniB_) The balance equals esouces of young animals (own poducion and impo of live animals) wih hei use in faening and aising aciviies. Sock changes ae defined by he EAA as (des)invesmens and have o be booked by definiion on he oupu side. Accodingly, hee ae neve sock changes of young animals used as inpus. Fuhemoe, he EAA does no ake ino accoun sales of young animals fom ohe fams o ades inside he couny, no ineacions beween animal aciviies of he same fam. Solely impos of live animals ae valued by he EAA as coss in animal poducion. Accodingly, flows of all ohe young animals ae by definiion no valued by he EAA seco. Each animal caegoy feaues is own equaion. Inpu and oupu of young animals ae linked ove a coss-se: Equaion 18 GROF iy = GROF oy STCM whee: iy young animals as inpus (booked as coss) oy young animals as oupus (booked as evenues) The equaion saes ha goss need of young animals as inpus - ypically equal numbe of slaugheed plus expoed plus aised heads minus he impoed heads is equal o he oupu of young animals minus sock changes. The disincion beween young animals on he inpu and oupu side allows o calculae he ina-secoal, ina-aciviy and ina-egional income effecs of exchanges of young animals, which ae consolidaed by he EAA. Consisency of numbe of impoed calves (CalvesT_) Impoed calves ae spli up in male and female ones: oy iy Equaion 19 GROF. ICAL = CAMF. LEVL + CAFF. LEVL whee: GROF.ICAL slaugheed plus expoed minus impoed calves Page 29 of 133

30 Fix elaion beween male and female young calves fom daiy and suckle cows (MalFem1_), (MalFem2_) The popoion of male calves in oal calves bon pe daiy and pe suckle cow is kep beween 50 and 52%. Equaion 20 YCAM YCAM Cows Cows ( YCAM ( YCAM cows cows + YCAF + YCAF cows cows ) *0.52 ) *0.50 Definiion of sock changes fo young animals (SocksA_) a) Fo he animal caegoies whee young animals ae used in he same yea (chicken, lambs, pigles) Equaion 21 LEVL LEVL = STCM +1 b) fo heifes and bulls, i is he change in he aising aciviies of he las yea (old animals oupued in he cuen yea sem fom young ones poduced in he yea befoe -1) Equaion 22 LEVL = STCM LEVL 1 c) fo young cows, i is he change in he heifes aising aciviies plus sock change of daiy and suckle cows Equaion HEIR. LEVL HEIR. LEVL DCOW. LEVL SCOW. LEVL DCOW. LEVL SCOW. LEVL = STCM Definiion of poein and fa balance fo milk pocessing on pocessing indusy level (MLKCNT_) Equaion 24 PRCM. MILK * MLKCNT = MAPR i * MLKCNTi i MLKSECO whee: PRCM.MILK milk colleced by daiies (cow and sheep/goa) MLKSECO se of elemens which conain all seconday commodiies poduced fom milk (bue cheese...) MLKCNT able wih poein and fa conen of he diffeen poducs Consisency of EAA value of milk (EAAMLK_) In he EAA, milk coves boh cow and sheep/goa milk. A disincion is made beween cow milk and milk fom sheep and goa in he daa base: Equaion 25 EAAP. MILK = EAAP. COMI + EAAP. SGMI Page 30 of 133

31 2.4 The Regionalised Daa Base (CAPREG) Daa equiemens a egional level CAPRI aims a building up a Policy Infomaion Sysem of he EU s agiculual seco egionalised a NUTS 2 level wih an emphasis on he impac of he CAP. The coe of he sysem consiss of a egionalized agiculual seco model using an aciviy based non-linea pogamming appoach. One feaue of such a highly disaggegaed, aciviy based agiculual seco model is he deailed infomaion esuling fom ex-ane simulaions of policy scenaios concening he oupu and inpu of specific agiculual poducion aciviies and hei elaionships. This infomaion is also a pe-condiion o judge possible impacs of agiculual poducion on he envionmen. Howeve hese sysems equie as well his kind of infomaion (daa) ex-pos, a leas paially. I is especially necessay o define fo each egion in he model, a leas fo he basis yea he maix of I/O-coefficiens fo he diffeen poducion aciviies ogehe wih pices fo hese oupus and inpus. Moeove fo calibaion and validaion puposes infomaion concening land use and livesock numbes is necessay. Given he impoance of he EU as an inenaional playe on agiculual wold makes, neihe wold no EU make pices can be eaed as exogenous o he model. Theefoe, a make module links he supply side of he model wih naional and inenaional makes fo agiculual poducs. Fo he ime being, he smalles make egion in he CAPRI is he Membe Sae level, hough o be a spo make fo all egional unis inen of he Membe Sae. This simplificaion allows o use naional daa o cove he model s make side Daa souces a egional level Aleady duing he fis CAPRI meeing, he REGIO domain of EUROSTAT was judged as he only hamonized daa souce available on egionalized agiculual daa in he EU. REGIO is one of seveal pas of NEWCRONOS and is iself boken down in domains, one of which coves agiculual and foesy saisics. In he agiculual and foesy domain [AGRI] he following ables ae available: Land use [A2LAND] Cop poducion - havesed aeas, poducion and yields [A2CROPS] Animal poducion - livesock numbes [A2ANIMAL] Cows s milk collecion - deliveies o daiies, % fa conen [A2MILK] Agiculual accouns on egional level [A2ACCT] Sucue of agiculual holdings [A2STRUC, A3STRUC] Labou foce of agiculual holdings [A2WORK] Daa availabiliy a egional level The following able shows he official availabiliy of he diffeen ables of REGIO. Howeve he cuen coveage concening ime and sub-egions diffes damaically beween he ables and wihin he ables beween he Membe Saes. Page 31 of 133

32 A second poblem consiss in he elaively high aggegaion level especially in he field of cop poducion. Hence, addiional souces, assumpions and economeic pocedues mus be applied o close daa gaps and o beak down aggegaed daa. Table 5 Official daa availabiliy in REGIO Table Land use Cop poducion (havesed aeas, poducion and yields) Animal poducion (livesock numbes) Cows s milk collecion (deliveies o daiies, % fa conen) Agiculual accouns on egional level Official availabiliy fom 1974 yealy fom 1975 yealy fom 1977 yealy fom 1977 yealy fom 1980 yealy Sucue of agiculual holdings 1983, 1985, 1987, 1989/91, 1993 Labou foce of agiculual holdings Souce: Euosa (hp://epp.euosa.cec.eu.in) fom 1983 yealy Reading and soing he oiginal REGIO daa The oiginal REGIO daa ae soed in an ASCII-foma designed by EUROSTAT fo NEWCRONOS and used in connecion wih he CUB-X, EUROSTAT s daa bowse. The daa can be bowsed and exaced o seveal fomas diecly wih CUB-X (one able each ime). Howeve in he case of he CAPRI-pojec, daa fom seveal ables mus be meged ogehe adding up o some million numbes. CUB-X was neve designed fo such quaniies. Theefoe, he goup in Bonn designed a ool called DFTCON which conves hese files ino a ahe simple foma: In a fis sep, hese files ae soed by egion, yea and oiginal code, so ha hey can be easily accessed by ohe sofwae o pefom exacion fom he oiginal NEWCRONOS daa base. In a second sep hese files ae ead, he oiginal codes ae assigned o eigh chaace sings and he esuling able pe egion and yea is soed in binay compessed fom in he daa base. The esul of hese wo seps ae ables, one fo each egional uni available a NUTS 0 o NUTS 2 level in REGIO and each yea which compise all daa fom he REGIO ables: land use, cop poducion, animal populaions, cow s milk collecion and agiculual accouns. These ables ae soed in a daa managemen sysem designed fo use in agiculual seco modelling Mehodological poceeding The saing poin of he mehodological appoach is he decision o use he consisen and complee naional daa base (COCO) as a fame o efeence poin fo any egionalizaion. In ohe wods, any aggegaion of he main daa iems (aeas, hed sizes, goss poducion and inemediae use, uni value pices and EAA-posiions) of he egionalized daa ove egions mus mach he naional values. Page 32 of 133

33 Given ha saing posiion, he following appoaches ae geneally applied: Daa ene he consisency checks as found in REGIO. This is mainly ue fo animal hed sizes whee REGIO offes daa a he same o even moe disaggegaed level as found in COCO. Gaps in REGIO ae filled ou and daa found in REGIO a a highe aggegaion level as equied in CAPRI ae boken down by using exising naional infomaion. Funcions used ae sucually and (ofen) numeically idenical fo all egional unis and goups of aciviies and inpus/oupus. Economeic analysis o addiional daa souces ae used o close gaps. All he appoaches descibed in he following sub-secions ae only hough as a fis cude esimae. Wheeve addiional daa souces ae available, hei conen should be checked and made available o ovecome he lis of hese easy-o-use esimaes pesened in hee. The pocedues descibed in hee can be hough as a safey ne o ensue ha egionalized daa ae echnically available bu no as an adequae subsiue fo collecing hese daa fom addiional souces Pices fo oupus and inpus Code: capeg\pice_yani.gms The agiculual domain of REGIO does no cove egionalized pices. Fo simpliciy, he egional pices ae heefoe assumed o be idenical o secoal ones 9 : Equaion 26 UVAG = UVAG s Young animal pices ae a special case since hey ae no included in he COCO daa base (he cuen mehodology of he EAA does no value inemediae use of animals) bu ae necessay o calculae income indicaos fo inemediae aciviies (e.g. aising calves). Only expoed o impoed live animals ae implicily accouned fo by valuing he conneced mea impos and expos. Young animals ae valued based on he mea value and assumed elaionships beween live and cacass weighs. Male calves (ICAM, YCAM) ae assumed o have a final weigh of 55 kg, of which 60 % ae valued a veal pices. Female calves (ICAF, YCAF) ae assumed o have a final weigh of 60 kg, of which 60 % ae valued a veal pices. Young heifes (IHEI, YHEI) ae assumed o have a final weigh of 300 kg, of which 54 % ae valued a beef. Young bulls (IBUL, YBUL) ae assumed o have a final weigh of 335 kg, of which 54 % ae valued a beef. Young cows (ICOW, YCOW) ae assumed o have a final weigh of 575 kg, of which 54 % ae valued a beef. Fo pigles (IPIG, YPIG), pice noaions wee egessed on pig mea pices and ae assumed o have a final weigh of 20 kg of which 78 % ae valued a pig mea pices. Lambs (ILAM, YLAM) ae assumed o weigh 4 kg and ae valued a 80 % of sheep and goa mea pices. Chicken (ICHI, YCHI) ae assumed o weigh 0.1 kg and ae valued a 80 % of pouly pices Filling gaps in REGIO Code: capeg\end.gms 9 Thee is no easy way o elax his assumpion if no fuhe daa souces ae available. Page 33 of 133

34 In cases whee daa in REGIO on egional level ae missing, a linea end line is esimaed fo he Membe Sae ime seies in REGIO definiion. The aveage elaion beween egional and naional daa whee boh ae available is hen muliplied wih he naional end value o deive a end esimae a egional level Mapping cop aeas and hed sizes fom REGIO o COCO definiions Code: capeg\map_fom_egio.gms Only some few cop aciviies ae available in REGIO (ceeals wih whea, baley, gain maize, ice; poaoes, suga bee, oil seeds wih ape and sunflowe; obacco, fodde maize; gassland, pemanen cops wih vineyads and olive planaions). The COCO daa base, howeve coves some 30 diffeen cop aciviies. In ode o beak hese aggegaes down o COCO definiions, he naional shaes of he aggegae ae used. As an example, his appoach is explained fo ceeals. Daa on he poducion aciviies BARL (baley), MAIZ (gain maize) and PARI (paddy ice) as found in COCO mach diecly he level of disaggegaion in REGIO. Theefoe, he egionalized daa ae diecly se o he values in REGIO. The diffeence beween he sum of hese 3 aciviies and he aggegae daa on ceeals in REGIO mus be equal o he sum of he emaining aciviies in ceeals as shown in COCO, namely RYE (ye and meslin), OATS (oas) and OCER (ohe ceeals). As long as no ohe egional infomaion is available, he diffeence fom REGIO is boken down applying naional shaes. The appoach is shown fo OATS in he following equaions, whee he suffix sands fo egional daa: Equaion 27 LEVL OATS, CEREAL WHEAT BARLEY LEVL OATS,COCO MAIZEGR RICE = LEVL + LEVL + LEVL OATS,COCO RYE,COCO OCER,COCO Simila equaions ae used o beak down ohe aggegaes and esidual aeas in REGIO 10. One impoan advanage of he appoach is he fac ha he esuling aeas ae auomaically consisen o he naional daa if he ingoing infomaion fom REGIO was consisen o naional level. Founaely, he egional infomaion on hed sizes coves mos of he daa needed o give nice poxies fo all animal aciviies in COCO definiion. REGIOs beak down fo hed sizes is moe deailed han COCO -a leas fo he impoan secos. Regional esimaes fo he aciviy levels ae heefoe he esul of an aggegaion appoach, in opposie o cop poducion Pefec aggegaion beween egional and naional daa fo aciviy levels Besides echnological plausibiliy and a good mach wih exising egional saisics, he egionalized daa fo he CAPRI model mus be also consisen o he naional level. The minimum equiemen fo his consisency includes aciviy levels and goss poducion. Consisency fo aciviy levels is momenaily achieved by fis using he appoaches descibed above o poduce fis esimaes (Levl ) fo he elevan daa iems and hen by calculaing and applying in a second sep coecion facos. 10 If no daa a all ae found, he shae on he uilisable agiculual aea is used. Page 34 of 133

35 Equaion 28 CORR Levl * j, j, LEVL = Levl = Levl j, j, * CORR Levl j, LEVL j, n Code: capeg\cons_levls.gms A specific poblem is he fac ha land use saisics do no epo a beak down of idling land ino obligaoy se-aside, volunay se-aside and fallow land 11. Equally, he shae of oilseeds gown as enegy cops on se-aside needs o be deemined. An enopy esimao is used o disibue he naional infomaion on he diffeen ypes of idling land o egional level, wih he following esicions Obligaoy se-aside aeas mus be equal o he se-aside obligaions deived fom aeas and se-aside aes fo Gandes Culues (which may diffe a egional level accoding o he shae of small poduces). Fo hese cops, aciviy levels ae paially endogenous in he esimaion in ode o allow a spli up of oilseeds ino hose gown unde he se-aside obligaions and hose gown as non-food cops on se-aside. Obligaoy and volunay se-aside canno exceed ceain shaes of cops subjecs o se-aside (a leas befoe Agenda 2000 policy) Fallow land mus equalise he sum of obligaoy se-aside, volunay se-aside and ohe idling land. Toal uilisable aea mus say consan. In some cases, aeas epoed as fallow land ae smalle han se-aside obligaions. In hese cases, pas of gassland aeas and ohe cops ae allowed o be educed. The poceeding fo goss oupu (GROF) is simila o he one fo aciviy levels, as coecion facos ae applied o line up egional yields wih given naional poducion: Equaion 29 CORR O * j, GROF, o = O j, = Levl j * CORR j, GROF. o O j, GROF o, n In case of missing saisical infomaion fo egional yields, naional yields ae used. A special ule is used fo fodde maize yields, whee egional yields ae deived fom naional fodde maize yields, and he elaion beween egional and naional aveage ceeal yields. Fo gassland and fodde fom aable land, missing yields ae deived fom naional ones using he elaion beween egional and naional socking densiies of uminans. The socking densiies ae calculaed by muliplying hed sizes wih live sock unis and dividing he esuling sum of livesock unis by he gassland, fodde maize and ohe fodde on aable land aeas. Code: capeg\cons_yields.gms 11 The necessay addiional infomaion on non-food poducion on se-aside, obligaoy and volunay se-aside aeas can be found on he DG-AGRI web seve. Page 35 of 133

36 Esimaing expeced yields wih a Hodick-Pesco file The inpu allocaion in any given yea should no be linked o ealised, bu o expeced yields. Expeced yields ae consuced using he following modified Hodick-Pesco file: * * 2 * Equaion 30 min hp = 1000 ( y+ 1 y 1 ) + ( y y ) 1< < T 1 whee y coves all oupu coefficiens in he daa base. The Hodick-Pesco file is applied boh a he naional and egional level afe any gaps in he ime seies had been closed. 2.5 The wold Daa Base 2 The global daa base of CAPRI compehends maco-economic daa fo diffeen wold egions, policy daa and global agiculual poducion daa. Seveal daa souces can be menioned: Daa on bilaeal ade beween seleced wold egional aggegaes (main ading playes) ae boowed fom he Wold Agiculual Tade Simulaion Model (WATSIM). Daa on policy vaiables such as applied and scheduled aiffs, aiff ae quoas o bilaeal ade ageemens ae obained fom he AGLINK Model (OECD) and he Agiculual Make Access Daabase (AMAD). Pefeences. Changes in demand behaviou no linked o income o pices changes ae ended using ex pos ime seies on pe capia consumpion, in mos cases in line wih daa found in he EU Pospecs fo Agiculual Makes (Euopean Commission). The pice famewok conained in he make module is based on epesenaive long em ime seies fo wold make pices of majo aw and pocessed agiculual poduc, which ae end foecased. These daa ae necessay fo he consucion of a wold ade model (compehending ceain wold egional aggegaes), which should delive some pice feedback o he Euopean supply sysem. Page 36 of 133

37 3 Inpu Allocaion The em inpu allocaion descibes how aggegae inpu demand (e.g. oal anoganic N feilise use in Denmak) is disibued o poducion aciviies. The esuling aciviy specific daa ae called inpu coefficiens. They may eihe be measued in value ( /ha) o physical ems (kg/ha). The CAPRI daa base uses physical ems and, whee no available, inpu coefficien measued in consan pices. Mico-economic heoy of a pofi maximising poduce equies evenue exhausion, i.e. maginal evenues mus be equal o maginal coss simulaneously fo all ealised aciviies. The maginal physical inpu demand muliplied wih he inpu pice exhauss maginal evenues, leading o zeo maginal pofis. Maginal inpu demands pe aciviy can only be used o define aggegae inpu demand if hey ae equal o aveage inpu demands. The lae is he case fo he Leonief poducion funcion. The advanage of assuming a Leonief echnology in agiculual poducion analysis is he fac ha an explici link beween poducion aciviies and oal physical inpu use is inoduced (e.g. envionmenal indicaos can be linked diecly o individual aciviies o aciviy specific income indicaos, since goss magins can be calculaed). The disadvanage is he ahe igid echnology assumpion. We would fo example expec ha inceasing a cop shae in a egion will change he aveage soil qualiy he cop uses, which in un should change yields and nuien equiemens. I should hence be undesood ha he Leonief assumpion is an absacion and simplificaion of he eal agiculual echnology in a egion. The assumpion is somewha elaxed in CAPRI as wo poducion inensiies ae inoduced. Inpu coefficiens fo diffeen inpus ae consuced in diffeen ways which will be discussed in moe deail in he following secions: Fo niae, phosphae and poash, nuien balances ae consuced so o ake ino accoun cop and manue nuien conen and obseved feilise use, combined wih a simple fixed coefficien appoach fo ammonia losses. These balances ex-pos deemine he effecive inpu coefficiens based on a coss-enopy esimaion famewok. Fo feed, he inpu calculaion is ooed in a mix of engineeing knowledge (equiemen funcions fo animal aciviies, nuien conen of feeding suff), obseved daa ex-pos (oal naional feed use, naional feed coss) and esimaed feed coss fom a FADN sample, combined wihin a Highes Poseio Densiy (HPD) esimaion famewok. Fo he emaining inpus, esimaion esuls fom a FADN sample ae combined wih aggegae naional inpu demand epoed in he EAA and sandad goss magin esimaions, again using a HPD esimaion famewok. 3.1 Inpu allocaion excluding young animals, feilise and feed Backgound Thee is a long hisoy of allocaing inpus o poducion aciviies in agiculual seco analysis, daing back o he days whee I/O models and aggegae fam LPs whee he only quaniaive insumens available. In hese models, he inpu coefficiens epesened a Page 37 of 133

38 Leonief echnology, which was pu o wok in he quaniaive ools as well. Howeve inpu coefficiens pe aciviy do no necessay imply a Leonief echnology. The allocaed inpu demands can be seen as maginal ones (which ae idenical o aveage ones in he Leonief case) and ae hen compaible wih flexible echnologies as well. Inpu coefficiens can be pu o wok in a numbe of ineesing fields. Fis of all, aciviy specific income indicaos may be deived, which may faciliae analyzing esuls and may be used in un o define secoal income. Similaly, impoan envionmenal indicaos ae linked o inpu use and can hence be linked o aciviies as well wih he help of inpu coefficiens. Given he impoance o he inpu allocaion, he CAP-STRAT pojec ( ) compised an own wok package o esimae inpu coefficiens. On a fis sep, inpu coefficiens wee esimaed using sandad economeics fom single fam ecod as found in FADN. Addiionally, ess fo a moe complex esimaion famewok building upon enopy echniques and inegaing esicions deived fom cos minimizaion wee un in paallel. The need o accommodae he esimaion esuls wih daa fom he EAA in ode o ensue muual compaibiliy beween income indicaos and inpu demand pe aciviy and egion on he one hand, and secoal income indicaos as well as secoal inpu use on he ohe equies deviaing fom he esimaed mean of he coefficiens esimaed fom single fam ecods. Fuhe on, in some cases esimaes evealed zeo o negaive inpu coefficiens, which canno be aken ove. Accodingly, i was decided o se up a second sage esimaion famewok building upon he unesiced esimaes fom FADN. The famewok can be applied o yeas whee no FADN daa ae available, and hus ensues ha he esuls will be coninuously used fo he yeas ahead, befoe an updae of he labo-inensive esimaions is again necessay and feasible Economeic Esimaion Sandad economeic mehods wee employed o calculae inpu coefficiens fom single fam ecods found in FADN (wihin a consisen aggegaion famewok, as explained in chape 6). Raw daa wee ansfomed ino CAPRI compaible caegoies. Fixed-Effecs, Random Effecs, Weighed Fixed-Effecs, and Weighed Random-Effecs as well as OLS and WLS models wee esed wih vaying degees of success. Afe finding heeoskedasiciy poblems, deciding o neglec fom using an inecep (in ode o confom o he Leonief echnology assumed by he model) and afe compaing esuls fo plausibiliy, i was decided ha a saighfowad WLS model was he mos suiable fom if a consisen esimaion echnique was o be used fo all esimaions. The main eason fo choosing such a simple WLS esimao ove a weighed andom effecs model wih no fixed effec inecep was he quesion of plausibiliy of esuls. Specificaion ess suggesed, in fac, ha fixed effecs esimaos migh have been used in evey egession, bu apa fom he poblem of disibuing fam specific fixed effec ineceps acoss cop and animal aciviies, hee wee wo (elaed) easons no o use hese esuls. Fisly, he esuls of he fixed effecs specificaions on he whole- wee implausible, wih a lage numbe of negaive coefficiens. Secondly, i was fel ha any possible endogeneiy in he esimaions would pobably have a geae popoionae effec in he fixed effecs esuls. The weigh acually used in he final WLS vesions was oal oupu. Iniial expeimens also evealed a high degee of mulicollineaiy if aciviy levels and oupus wee boh used on he igh hand side. Accodingly, i was decided o use oupu on he igh hand side if possible (so ha egional vaiaions could be incopoaed ino he model). Whee sufficien oupu values wee no available, aciviy levels wee used, using he cieion descibed below. Fuhemoe, because of a clealy deleeious effec on esuls, Page 38 of 133

39 he equivalens of he CAPRI esidual aciviy caegoies OCRO (ohe cops), OFRU (ohe fuis), OCER (ohe ceeals), OVEG (ohe vegeables), ec. wee all dopped fom he esimaions. All egessions wee un using STATA 7.0. Pice indices wee aken fom he COCO daabase in ode o calculae inpu coss in eal ems. The saing sample sizes wee, as follows, all muliplied by 10 (fo he yeas ) unless ohewise saed: AT - Ausia fams BL - Belgium, 2601 fams DE - Gemany, fams --> pice daa fom DK - Denmak, 6625 fams EL - Geece, fams --> pice daa fom FI - Finland, 1324 fams IR - Ieland, 3409 fams --> no pice daa pio o 1995 IT - Ialy, fams PT - Pougal, 6379 fams SE - Sweden, 1191 fams UK - Unied Kingdom, 6668 fams ES - Spain, fams NL Nehelands, 3565 fams FR Fance, fams The following daa cleaning pocedues wee used: The egessos wih less han o equal o 100 obsevaions fo boh aciviy levels and oupu wee excluded. The daa wee uncaed a zeo in ode o eliminae epoed negaive level and oupu values and also epoed negaive eal inpu coss. All non-zeo values wee couned and a choice made beween eihe aciviy level o oupu, as he appopiae igh-hand side vaiable (only one could be use o avoid mulicollineaiy). An aciviy s oupu value was used if he numbe of non-zeo oupu values associaed wih ha aciviy was geae han he numbe of he aciviy s non-zeo levels minus 500. Thus, oupu was always he pefeed opion unless levels wee epoed fo a leas 500 moe obsevaions han oupus. This pocedue was necessay because of a numbe of cases in he daa when only oupu o aciviy level values bu no boh. Seveal egessions wee un o yield esimaes fo coefficiens in each of 11 inpu caegoies: Toal Inpus, Cop Only Inpus, Animal Only Inpus, Seeds, Plan Poecion, Feilize Ohe Cop Inpus, Puchased and Non-Puchased Feeds and Ohe Animal Only Inpus. Page 39 of 133

40 3.1.3 Reconciliaion of Inpus, using Highes Poseio Densiy Esimaos Code: inpus\dis_inpu.gms As a esul of he unesiced esimaion based on FADN, a maix of inpu coefficiens and hei esimaed sandad eos is available. Some of hose coefficiens ae elaed o he oupu of a ceain aciviy (e.g. how much money is spend on a ceain inpu o poduce one uni of a poduc), some of hem ae elaed o he aceage of on aciviy (inpu coss pe aciviy level). The able below pesens a sample of he esuls fom he economeic egessions. These ae he oupu (GROF) coefficiens of 2 aciviies, sof whea and baley, fo 4 inpu caegoies; oal inpus (TOIN), oal ohe inpus (TOIX), cop only inpus (COSC), and feilise (FERT). All coefficiens is saisically significan excep hose in ed. Table 6 Sample of sof whea and baley poducion coefficiens fo 4 inpus (1995 pices) GROF AT BL DE DK EL ES FI FR IR IT NL PT SE UK CS. Whea f TOIN TOIX COSC FERT Baley TOIN TOIX COSC FERT Souce: inpu esimaion, CAPRI modelling sysem Fo example, he TOIN coefficien fo sof whea in Ausia eveals ha on aveage i coss an Ausian fame o poduce an exa onne of whea. These coefficiens should eveal a easonable sense of coss-couny compaaive advanage among aciviies. In able 6, he coefficiens of vaiaion fo sof whea fo TOIN, TOIX, COSC, and FERT wee 34 %, 41 %, 29 % and 44 % especively. Those fo baley wee 21 %, 29 %, 19 %, and 27 % especively. Thus, a high degee of vaiaion fo TOIX and FERT is clea in his sample. This gives an indicaion of he geneal vaiabiliy undelying he esimaed coefficiens. All of he economeic coefficiens wee equied o be ansfomed ino an aciviy level fom, due o he fac ha his is he definiion used in he CAPRI model. Befoe his could be done, i seemed necessay o fill up he maix of esimaed coefficiens because some esimaes wee missing and ohes wee negaive. In ode o his we consuced a numbe of coefficiens ha wee weighed aveages among ceain goups. These mean coefficiens wee he following. 1. Mean coefficiens of aciviy goups. Each aciviy was allocaed o a ceain goup (e.g. sof whea belonged o ceeals). Fo each goup we buil weighed aveages among he posiive esimaes wihin a goup using he esimaed -saisics as weighs. This coefficien only exised if hee was a leas one posiive esimae inside Page 40 of 133

41 ha goup and was hen used o eplace he gaps inside he coefficien maix. If ha mean coefficien was no available, due o no posiive esimae inside a goup a all, he nex ype of mean coefficiens became elevan: 2. Mean coefficiens fo an aciviy among Euopean egions. This second ype of mean coefficiens calculaes weighed aveages among hee ypes of egional cluses. These cluses ae Nohen Euopean Saes, Souhen Euopean saes and all Euopean egions. Again, he esimaed -saisics wee used as aggegaion weighs. Unfounaely, his ype of aveages did no fill all gaps in he coefficien maix as hee wee some aciviies ha had no posiive esimae ove he enie EU. Fo hose he hid ype of mean coefficiens was calculaed. 3. Mean coefficiens fo aciviy goups among egional cluses. Hee we calculaed fo he hee egional cluses he aveages of he fis ype of mean coefficiens. As even he lae ae synheic, we gave each mean of hem he same weigh. Founaely hee was only a small pobabiliy ha his coefficien did no exis fo one of he goups as his was only he case if no coefficien inside a goup ove he enie EU had a posiive esimae, which was no he case. Following hese ules we finally go a maix of esimaed and synheic calculaed inpu coefficiens fo boh, he pe aciviy level and he pe poducion uni definiion. 12 Fo he synheic one hee was no esimaed sandad eo available bu we waned o use hose lae on. So we assumed hem o eflec ha hese coefficiens have only weak foundaion o have a -saisic of 0.5. The pe level definiion was only aken ove if he coefficien was eally esimaed o if no pe poducion uni definiion did exis. To ansfe he lae ino pe aciviy level definiion, we muliplied hem wih he aveage yield ( ) of he especive aciviy. The esuling coefficiens and hei sandad eos wee hen used in he coss enopy appoach descibed below. 13 Missing economeic esimaes and compaibiliy wih EAA figues wee no he only easons ha made a econciliaion of esimaed inpus coefficiens necessay. Moeove he economic sense of he esimaes could no be guaaneed and he definiion of inpus in he esimaion diffeed fom he one used in CAPRI. Theefoe we decided o include fuhe pio infomaion on inpu coefficiens in agiculue. The second se of pios in he inpu econciliaion was heefoe based on daa fom he EAA. Toal coss of a ceain inpu wihin an aciviy in a Euopean Membe Sae was calculaed by muliplying he oal expendiues on ha inpu wih he popoion of he oal expeced evenue of ha aciviy o ha of all aciviies using he inpu. Toal expeced evenue in his case was he poducion value (including make value and pemiums) of he especive aciviy. If his esuled in a ceain coefficien being calculaed as zeo due o missing daa, hen his coefficien would be eplaced by one fom a simila aciviy e.g. a zeo coefficien fo MAIF would be eplaced by he coefficien fo GRAS 12 In addiion, a simila pocedue (using slighly diffeen goups) was applied o consucing coefficiens fo he Ohe aciviies (e.g. OCER, OFRU, OVEG), which had been omied fom he economeic esimaions. They ae given he aveage goup coefficien, unless hee is none; hen hey ae given he aveage nohen o souhen Euopean coefficien as appopiae. 13 Adjusmens wee made fo scaling issues wih egad o eggs fo ceain counies, and gass fo Finland. In addiion, when CAFR, CAFF and HEIR did no have economeic daa, hey assumed he coefficiens and sandad eos of CAMR, CAMF and HEIF especively (CAPRI aciviy code definiions in able 29 o he appendix). Page 41 of 133

42 This kind of pio infomaion ies o give he esuls a kind of economic sense. Fo he same eason he hid ype of pios was ceaed based on sandad goss magins fo agiculual aciviies eceived fom EUROSTAT. Those exised fo nealy all aciviies. The se fom 1994 was used, since his was he mos complee available. Relaive ahe han absolue diffeences wee impoan, given he equiemen o confom o EAA values Highes Poseio Densiy esimaion famewok Given he hee ypes of pio infomaion explained above esimaed inpu coefficiens, daa fom EAA and sandad goss magins-, he choice of a HPD Esimao o econcile esimaed inpu coefficiens seemed o be convenien. 15 The esimaion was caied ou fo all CAPRI aciviies (z) -excluding aciviies ha whee spli up like DCOW ino DCOL and DCOH-, and a numbe of inpus in CAPRI (denoed by X CI,z ) and FADN (X FI,z ) definiion. The lis of inpu definiions can be found in he annex (able 31). Fo each pio we defined 4 suppo poins (k) cened on he value of he pios defined as above. The suppo ange was defined as follows: Fo he economeic esimaes: S XFz,k P XFz + [-100; -1; 1; 100] σ XFz, whee S XF,I,z,k gives he suppo poins fo he FADN inpu X FI,z ha has a sandad eo of σ XFi. Fo he EAA pios: pio *(1+ [-10; -0.1; 0.1; 10]). S XCz,k P XCz (1+ [-10; -0.1; 0.1; 10]), whee S XCz,k gives he suppo poins fo he CAPRI inpu X CI,z. A special eamen was chosen fo he oal inpu coefficien. Hee he suppo ange was half ha fom above. Fo he sandad goss magins: S GM,z,k P GM,z (1+ [-10; -0.1; 0.1; 10]), whee S XCk gives he suppo poins fo he sandad goss magin of aciviy z. We define he a pioi pobabiliy fo each suppo poin o be: AP k = [0.002; 0.49; 0.49; 0.002], in ode o give he ouemos suppo poins less weigh. Poseio pobabiliies ae denoed by PP. The model seup is hen given by: 14 Conay o he economeic esimaed pios, he wo ohe ypes wee diffeen in diffeen yeas, since he econciliaion had o be done fo each yea in he daabase. The second pio ype is yea specific by naue, as he EAA values diffe beween yeas. In case of sandad goss magins, unfounaely, we had hem only fo one yea (1994). So we decided o dive hem ove ime using he popoion of expeced evenue of an aciviy in a ceain yea o ha in he yea The advanage of coss enopy is ha one can define he suppo space ahe wide and give he edges a vey low pio pobabiliy. Page 42 of 133

43 Equaion 31 max H(PP, PP, PP ) = s.. CI,FI,Z,K CI,Z,K FI,Z,K GM,Z,K PP PP PP CI,Z,K FI,Z,K PPCI,Z,K ln + PPFI,Z,K ln APK APK FI,Z,K + PPGM,Z,K ln APK PP = 1, PP = 1, PP = 1 CI,Z,K FI,Z,K GM,Z,K k k k X CI,Z CI,Z,K CI,Z,K FI,Z FI,Z,K FI,Z,K k k Z GM,Z,K GM,Z,K k Z Z CI,Z exo,z CI G1(CI,Z) CI G2(CI,FI) FI G3(CI,FI) CI CI,Z Z Z G1(CI,Z) FI1 G4(FI,FI1) = PP S, X = PP S GM = PP S GM = EREV X X EAA = X LEVL X X CI,Z FI,Z X FI1,Z = X = X FI,Z CI,Z = X FI, Z The fis wo ows of he equaion shown above ae subjec o maximize coss enopy, while he hid ow guaanies ha all pobabiliies sum up o uniy. In he fouh ow, he esimaes fo inpu coefficiens and goss magins ae e-paameeized fom he poseio pobabiliies and he suppo poins. The fifh ow defines goss magins fo an aciviy z as he diffeence beween expeced evenue pe aciviy level (EREV) of ha aciviy and he sum ove all inpus used in ha aciviy. The Se G1(CI,Z) allocaes he inpus used o each aciviy and X exo,z ae inpus, ha ae no esimaed hee, bu canno be negleced in defining goss magins (like young animal inpus). In he sixh ow, we find a saemen which guaanees ha he sum ove all aciviies of hei aciviy levels muliplied wih an inpu gives he oal expendiues on ha Inpu given by he EAA. The sevenh and eighh ows link he inpus in he CAPRI definiion o hose in FADN definiion. The fis of hose wo ae used when he FADN inpus ae an aggegae of CAPRI inpus (defined in he se G2(CI,FI)) o hey have he same definiion and he second one when CAPRI inpus ae an aggegae of FADN inpus. Since esimaed inpus in he FADN definiion exis fo aggegaes and componens of hem, we ensue in he las line ha he sum ove FADN inpus ha belong o an aggegaed FADN inpu (defined in he se G4(FI,FI1)) sum up o he lae. The esimaion is caied ou in GAMS wihin and un fo each yea in he daabase. Some bounds ae fuhe se o avoid esimaes unning ino implausible anges How ae he esuls used in CAPRI? The coss enopy esimaion yields moneay inpu coefficiens fo he feilise ypes (Niae, Phosphae, Poassium), seeds, plan poecion, feeds, phamaceuical inpus, epais, Page 43 of 133

44 agiculual sevice inpu, enegy and ohe inpus. While he lae fou ypes can diecly be used in he CAPRI model, we need special eamens fo he ohe ypes e.g. feilises, because hey ae used in physical unis inside he model, and feeds, since hey ae much moe disaggegaed. Theefoe, he esimaed esuls will go o ohe pas in he egionalisaion. The coss fo feeds go ino he feed imming, whee animal equiemens ae bough ino equilibium wih he conens of he feeding suff as suppos. A simila hing could be done wih he feilise coss in he feilise imming. 3.2 Inpu allocaion fo young animals and he hed flow model Code: capeg\spli_acs.gms Figue 4 shows he diffeen cale aciviies and he elaed young animal poducs used in he model. Milk cows (DCOL, DCOH) and suckle cows (SCOW) poduce male and female calves (YCAM, YCAF). The elaion beween male and female calves is esimaed ex-pos in he COCO famewok. These calves ae assumed o weigh 50 kg (female) and 55 kg (male) a bih and o be bon on he 1s of Januay. They ene immediaely he aising pocesses fo male and female calves (CAMR, CAFR) which poduce young heifes (YHEI, 300 kg live weigh) and young bulls (YBUL, 335 kg). The aising pocessing ae assumed o ake one yea so ha calves bon in ene he pocesses fo male adul faening (BULL, BULH), heifes faening (HEIL, HEIH) o heifes aising (HEIR) on he 1s Januay of he nex yea +1. The heifes aising pocess poduces hen he young cows which can be used fo eplacemen o hed size inceasing on he fis of Januay of +2. The able below he diagam shows a numeical example fo he elaionships. Figue 4. The cale chain Beef Milk Cows Suckle Cows Young cow Beeding Heifes Faening female Calves Female Calf Raising female Calves Young heife Faening Heifes High/Low Male Calf Raising male Calves Young bull Veal Faening male Calves Male adul cale High/Low Beef Souce: CAPRI Modelling Sysem Accodingly, each aising and faening pocess akes exacly one young animal on he inpu side. The aising pocesses poduce exacly one animal on he oupu side which is one yea olde. The oupu of calves pe cow, pigles pe sow, lambs pe mohe sheep o mohe goa Page 44 of 133

45 is deived ex pos, e.g. simulaneously fom he numbe of cows in -1, he numbe of slaugheed bulls and heifes and eplaced in +1 which deemine he level of he aising pocesses in and numbe of slaugheed calves in. The hed flow models fo pig, sheep and goa and pouly ae simila bu less complex, as all ineacions happen in he same yea and no specific aising pocesses ae inoduced. Table 7 Example fo he elaion inside he cale chain (Denmak, ) Male calves used in and bon in DCOWLEVL Numbe of daiy cows DCOWYCAM Numbe of male calves bon pe 1000 daiy cows Numbe of males calves bon fom daiy cows SCOWLEVL Numbe of suckle cows SCOWYCAM Numbe of male calves bon pe 1000 suckle cows Numbe of male calves bon fom suckle cows Numbe of all male calves bon GROFYCAM Numbe of male calves poduced CAMFLEVL Numbe of male calves faened CAMRLEVL Aciviy level of he male calves aising pocess Sum of pocesses using male calves GROFYCAM Numbe of male calves used Female calves used in and bon in DCOWLEVL Numbe of daiy cows DCOWYCAF Numbe of female calves bon pe 1000 daiy cows Numbe of female calves bon fom daiy cows SCOWLEVL Numbe of suckle cows SCOWYCAF Numbe of male calves bon pe 1000 suckle cows Numbe of female calves bon fom suckle cows Numbe of all female calves bon GROFYCAF Numbe of female calves poduced CAFFLEVL Numbe of female calves faened CAFRLEVL Aciviy level of he female calves aising pocess Female calves used in and bon in GROFYCAF Numbe of female calves used Young bulls used in and young bulls poduced in BULFLEVL Aciviy level of he bull faening pocess GROFIBUL Numbe of young bulls used GROFYBUL Numbe of young bulls aised fom calvs CAMRLEVL Aciviy level of he male calves aising pocess Heifes used in and heifes poduced in HEIFLEVL Aciviy level of he heifes faening pocess HEIRLEVL Aciviy level of he heifes aising pocess Sum of heife pocesses GROFIHEI Numbe of heifes used GROFYHEI Numbe of heifes aised fom calves CAFRLEVL Aciviy level of he female calves aising pocess Cows used in and heifes poduced in DCOWLEVL Numbe of daiy cows DCOWICOW Numbe of young cows needed pe 1000 daiy cows Sum of young cows needed fo he daiy cow hed DCOWSLGH Slaugheed daiy cows SCOWLEVL Numbe of suckle cows SCOWICOW Numbe of young cows needed pe 1000 suckle cows Sum of young cows needed fo he suckle cow hed SCOWSLGH Slaugheed suckle cows Sum of slaugheed cows GROFICOW Numbe of young cows used Sock change in daiy cows (DCOWLEVL(+1)-DCOWLEVL() Sock change in suckle cows (SCOWLEVL(+1)-SCOWLEVL() Sum of sock changes in cows Sum of slaugheed cows and sock change GROFYCOW Nume of heifes aised o young cows HEIRLEVL Aciviy level of he heifes aising pocess The able above is aken fom he COCO daa base. In some cases, egional saisical daa o esimaes fo numbe of young animals pe adul ae available, bu in mos cases, all inpu and oupu coefficiens elaing o young animals ae idenical a egional and naional level. Neveheless, expeiences wih simulaions duing he fis CAPRI pojec phase evealed ha a fixed elaionship beween mea oupu and young animal need as expessed wih on bull Page 45 of 133

46 faening pocess oveesimaes he igidiy of he echnology in he cale chain, whee poduces may eac wih changes in final weighs o elaive changes in oupu pices (mea) in elaion o inpu pices (feed, young animals). A highe pice fo young animals will end o incease final weighs, as feed has become compaaively cheape and vice-vesa. In ode o inoduce moe flexibiliy in he sysem, he daiy cow, heife and bull faening pocesses ae spli up each in wo pocessed as shown in he following able. Table 8 Spli up of cale chain pocesses in diffeen inensiies Daiy cows (DCOW) Bull faening (BULF) Heifes faening (HEIF) Low inensiy/final weigh DCOL: 60% milk yield of aveage, vaiable inpus besides feed an young animals a 60% of aveage BULL: 20% lowe mea oupu, vaiable inpus besides feed an young animals a 80% of aveage HEIL: 20% lowe mea oupu, vaiable inpus besides feed an young animals a 80% of aveage High inensiy/final weigh DCOH: 140% milk yield of aveage, vaiable inpus besides feed an young animals a 140% of aveage BULH: 20% highe mea oupu, vaiable inpus besides feed an young animals a 120% of aveage HEIH: 20% highe mea oupu, vaiable inpus besides feed an young animals a 120% of aveage 3.3 Inpu allocaion fo feed The inpu allocaion fo feed descibes how much kg of ceain feed caegoies (ceeals, ich poein, ich enegy, feed based on daiy poducs, ohe feed) o single feeding suff (fodde maize, gass, fodde fom aable land, saw, milk fo feeding) ae used pe animal aciviy level 16. The inpu allocaion fo feed akes ino accoun nuien equiemens of animals, building upon equiemen funcions. The inpu coefficiens fo feeding suff shall hence ensue ha enegy, poein equiemens, ec. cove he nuien needs of he animals. Fuhe on, ex-pos, hey should be in line wih egional fodde poducion and oal feed demand saisics a naional level, he lae semming fom make balances. And las bu no leas, he inpu coefficiens ogehe wih feed pices should lead o easonable feed cos fo he aciviies Esimaion of fodde pices Since he las evision of he EAA, own poduced fodde (gass, silage ec.) is valued in he EAA. Individual esimaes ae given fo fodde maize and fodde oo cops, bu no beak down is given fo fodde on aable land and fodde poduced as gassland as pesened in he CAPRI daa base. The diffeence beween gass and aable land is inoduced, as convesion of gass o aable land is fobidden unde coss-compliance condiions so ha maginal values of gassland and aable land may be diffeen. 16 The eade should noice again ha he aciviy definiion fo faening pocesses ae slaugheed plus expoed minus impoed animals and no sable places. Page 46 of 133

47 The pice aached o fodde should eflec boh is nuiional conen and he poducion coss a egional level. The enopy based esimaion pocess ies o inegae boh aspecs. The following equaions ae inegaed in he esimao. Fisly, he egional pices fo gass, fodde on aable land and saw (fin) muliplied wih he fed quaniies a egional level mus exhaus he vale epoed in he economic accouns, so ha he EAA evenues aached o fodde ae kep unchanged: Equaion 32 FEDUSE fin PFOD = EAAP OFAR, MS + EAAP fin fin GRAS, MS Secondly, he Goss Value Added of he fodde aciviies is defined as he diffeence beween evenues and oal inpu coss based on he inpu allocaion fo cops descibed above Equaion 33 GVAM = YIELD PFOD TOIN fin fin fin fin Nex, he sandad ingediens of a coss enopy esimao ae added: definiion of he esimaed values fom suppos and he poseio pobabiliies, summing up of he poseio pobabiliies o uniy and he definiion of he coss enopy iself Equaion 34 H ( PROB) = k k k k sup sup p p fin, gvam, k fin, pice, k fin, gvam, k fin, pice, k p = 1 = 1 fin, pice, k fin, pice, k p p p fin, gvam, k fin, pice, k fin, gvam, k fin, pice, k log( p log( p = GVAM = PFOD fin, pice, k fin, gvam, k fin fin pq pq k k ) ) The a pioi mean fo he pices of gass and ohe fodde on aable land ae he EAAP values divided by oal poducion volume which is by definiion equal o feed use. The pice of saw fo feed use is expeced o be a 1 % of he gass pice. The oue suppos ae se so ha he highe suppo is a fou imes he a pioi mean. Suppos fo Goss Value Added pe aciviy ae cened aound 150 % of he value of oal inpus as allocaed by he ules and algoihm descibed above, wih ahe wide bounds. The a pioi pobabiliies fo he hee suppos ae se a 1 %, 98 % and 1 %. The wide suppos fo he Goss Value Added of he fodde aciviies mio he poblem of finding good inenal pices bu also he dubious daa qualiy boh of fodde oupu as epoed in saisics and he value aached o i in he EAA. The wide suppos allow fo negaive Goss Value Added, which may ceainly occu in ceain yeas depending on ealised yields. In ode o exclude such esimaion oucomes as fa as possible an addiional consain is inoduced: Equaion 35 YIELD fin PFOD TOIN fin fin gvafac The paamee gvafac is iniialised wih uniy so ha fis a soluion is ied whee all aciviies have evenues exceed coss. If infeasibiliies aise, he faco is sepwise educed Page 47 of 133

48 unil feasibiliy is achieved, o ensue ha he minimal numbe of aciviies wih negaive Goss Value Addeds is esimaed Feed inpu coefficiens Feed use deemines o a lage exen coss of animal poducion as well as he use of ceain cops as fo fodde o ceeals, and of seconday poducs as oilseeds as by-poducs of he milling indusy. Hence, fom he view poin of policy modelling, a plausible descipion of he elaions beween animal poducion, feed use and make pices fo feeds is necessay. The issue has been addessed by a boad ange of mehodological soluions in seco models. Some impoan soluions ae chaaceised as follows: (1) Poducs used fo feed ae modelled as ne-pus, only, depending on cop and animal pices and fuhe facos, e.g. echnological pogess and availabiliy of pimay facos. Hence, he ne-pu appoach models (ina-secoal) feed use implicily. Typically, he elevan paamees ae esimaed based on dualiy in he conex of a pofi funcion whee he poducion possibiliy se is hidden. Plausibiliy checks of hese hidden echnological elaions ae had o do and sill hade o incopoae in he esimaion appoach. (2) Feed use is modelled explicily, bu he undelying echnology is hidden, e.g. when cos funcions fo he feed compound indusy ae esimaed as in BRITZ & SIEBER (1998). Modelling feed use as a funcion of pices is quie common, fo example in mulicommodiy models such as WATSIM of he IAP o he Wold Food Model of he FAO. The echnological elaions ae hidden as in (1), bu somewha easie o check by compaing he change in animal poducion povoked by a pice change wih is effecs on feed use. The nex soluions efe o pogamming models: (3) Feed use is modelled as fix cos pe uni of animal poducion aciviies. Animal poducion will no be affeced diecly by changes in cop poducion and vice vesa, bu only indiecly by an updae of he fix cos coefficiens. (4) Feed use is modelled via feeding aciviies. In ha case, he model ypically simulaneously deemines he opimal levels of animal and cop poducion and he amoun of each oupu fed o he animals. These soluions diffe by: a) The aggegaion level of he feeding suff. Some models, fo example he Geman seco model RAUMIS, have feeding aciviies fo each oupu o each animal aciviy, excluding echnological impossible combinaions such as feeding saw o pigles. The soluion ypically leads o a ahe high numbe of feeding aciviies. Ohes, as he SPEL-MFSS, use aggegaes of aw poducs in he feeding aciviies. Las bu no leas, single aw poducs can be mixed o diffeen pedefined menus whose mix is hen he endogenous vaiable, as in TASM. b) The definiion of he consains. In RAUMIS and SPEL-MFSS (WEBER 1995, pp. 39), equiemens such as enegy o poein ae modelled explicily. If mixes fo ceain animals ae defined befoehand, explici equiemen consains in he model may be lef ou if each individual mix guaanees aleady ha equiemens ae coveed. The highe he numbe of feed use aciviies pe animal aciviy, and he lowe he numbe of equiemen consains, he highe he chance of songly ovespecialised soluions. In ode o ensue a "plausible" mix of he feeding aciviies, bounds on he feeding aciviies ae someimes used, as in SPEL-MFSS Page 48 of 133

49 o in ealie vesions of RAUMIS. The subsiuion possibiliies may be influenced by using PMP on feeding aciviies, as in he new vesion of RAUMIS (CYPRIS 1999). Soluion (3) and (4) may be ieaively coupled o modules which deemine he coss fo soluion (3) o pices used in soluion (4) Soluion in CAPRI Such an ieaive coupling is he case in CAPRI which addesses he quesion of modelling he elaions beween animal and cop poducion, pice and makes as follows: Feed use of non-adable fodde such as gazing is modelled by individual feeding aciviies in he egional supply models. Wheeas feeding aciviies of adable poducs such as whea o soya beans ae aggegaed o five caegoies (ceeals, ich poein, ich enegy, milk based and ohes). Requiemen consains (enegy, poein, lysine ec.) ae inoduced o ensue echnological plausible subsiuion beween feeds. Calibaion o an esimaed feed inpu allocaion ex-pos is guaaneed by PMP calibaion ems. The esimaed feed inpu allocaion guaanees ha he feed posiions of he naional make balances ae me, and ha esuling feed coss ae plausible. Tadable feeding suffs can be sold and bough in unlimied quaniies. No diffeence is made beween adable feeding suffs poduced in he egion and such bough (ne ade appoach). Gas, fodde maize, ohe fodde fom aable land and saw ae assumed o be adable only inside a egional uni. All quaniies poduced in he egions mus be fed o egions o can be los, he lae howeve only in he case of saw. The egional models ae solved independenly fom each ohe wih pices fo oupus and inpus - including feeding suff - fixed. Regional ne ade fom he egional supply model is aggegaed pe Membe Sae and enes he behavioual funcions of he make module including he quaniies of he five feed aggegaes. The link beween he egional supply modules and he make modules is descibed in moe deail below. Make module and egional supply models ae ieaively linked whee pices fom he make module ae used in he supply module geneaing quaniies ha in un ae used in he make module. Fo he five feed aggegaes, he make module deemines - wih each ieaion a new pice and new nuien conens fo he feed aggegaes (fo deails see: HECKELEI, BRITZ & LOEHE 1998). The subsequen secions ae mosly devoed o he quesion how he egional models embedded in he specific CAPRI layou descibed above can be specified in a way ha calibaion of obseved feeding quaniies is achieved, and a plausible simulaion behaviou fo feed use is obained. An impoan feaue of he CAPRI model is a ahe explici, pimal modelling of feeding aciviies in he supply pa as a cos minimising poblem. Subsiuion possibiliies in feeding Page 49 of 133

50 ae modelled by equiemen consains fo each animal caegoy which can be saisfied by an appopiae feed mix. The appoach is common in pogamming models because: o I is known ha he feed compound indusy, exension specialiss and fames use pogamming models o minimise hei feed coss. Hence, hee is hope ha a simplified and aggegaed vesion of such a ype of model will wok in a moe aggegaed conex as well. o Infomaion on nuien equiemens of animals ae published in he lieaue as ae nuien conens of feeding suffs, so ha he consain maix can be specified based on engineeing knowledge. The undelying paamees and funcions ae discussed in he chapes above. Pices fo oupus including feeding suffs ae anyway needed in he conex of a seco model. The dawback of he appoach is a long sanding expeience of modelles wih ovespecialised soluions fom such feed cos minimising models and a long lis of pagmaic icks o ge id of hem. Some ypical poblems of aggegae models apply hee as well: some daa such as availabiliy and qualiy of fodde ae much hade o ge on he aggegae han a single fam level. Published nuien equiemens of animals elae ypically o conolled expeimens, and no o he acual fam pacise. Specific consains included in fam models canno be incopoaed, and he specificaion of models used by he feed compound indusy is no published. When unning a simulaion wih a ypical seco model, he deeminaion of he cop oaion, animal hed sizes and feeding pacise is a simulaneous poblem. 17 How aciviy levels in cop and animal poducion ae calibaed in CAPRI is discussed lae on and we will assume fo simpliciy ha hed sizes (LEVL) of animal caegoies (a) ae given and all feeding suffs (f) ae bough a known pices (PRICE). In ode o shed ligh on he elaion beween ovespecialisaion and calibaion, we will sa wih he assumpion ha physical equiemens (AREQ) fo enegy, poein ec. ae known as well as he nuien conens (NUTR) of he feeding suff. This infomaion seves o define he consains (="c") of ou feed cos minimising poblem which is fomulaed fo one of he egions (="") as follows: Equaion 36 min s.. feed feed aac, feed FEDUSE FEDNG FEDNG AREQ feed, aaac, feed, aaac, c, aac, feed, PRICE NUTR LEVL DAYS aac, feed feed, c aaac, c, aac, c = FEDUSE feed feed whee FEDNG aac,feed, is he quaniy of he feeding suff feed fed o an animal of caegoy aac in egion o be deemined and "bas" ove vaiables denoe known values. FEDUSE ae oal quaniies fed of a ceain ype of feed. The consains sae ha he equiemens of he animals mus be coveed by an appopiae feed mix. The objecive minimises oal feed cos a given animal hed sizes and pices. We should noice ha as long as all feeding suff can be bough in unlimied quaniies, he animal caegoies can be solved independenly fom each ohe wihou affecing he soluion of Equaion 36. A minimum equiemen of 17 Compae e.g. HAZELL & NORTON (1986), 263 ff., KASNAKOGLU & BAUER (1989) Page 50 of 133

51 Equaion 36 fom he view poin of model calibaion is feasibiliy in he base yea. Tha can be achieved by an appopiae adjusmen of he coefficiens NUTR and AREQ. I will be shown below how Coss Enopy economeics can be pu o wok o solve ha poblem. Howeve no only pices and hed sizes ae known in he base yea bu he quaniies fed a naional level as well, a leas fo adable feeding suffs. Thee is lile hope ha solving Equaion 36 fo all egions and adding up he esuling feed use will epoduce he obseved feed quaniies a naional level wihou pio calibaion. Aleady a "sligh" deviaion of 5 % in sensible makes such as ceeals would suely iiae ineesed policy makes. Table 1 shows he deviaion in Gemany obained fom Equaion 36 some yeas back in ess if "hypoheical" equiemens ae used 18. The model is no allowed o use moe of any feed caegoy naionally han obseved. Gas, silage and oos ae assumed o be no adable, and consequenly hei feeding quaniies ae fixed a egional availabiliy. Since milk and suga bee quoas fix sales, he feed quaniies of milk and aw milk ae fixed as he esidual beween sales a quoa level and poducion (hee also exogenous). Table 9 Feed use fom non-calibaed model, Gemany Feed Quaniy used % of base yea DHAY hay STRA saw FCER- ceeals FPRO - poein ich FENE - enegy ich FMIL - milk based FOTH - ohe Souce: CAPRI esuls When inepeing he esuls, one should keep in mind ha he appealing looking 100 % values fo all non-adables (aw milk, oo cops, silage, gaze & gazing) ae due o fixed values and esuls fo ich enegy and milk based feed ae based on (obviously binding) uppe bounds. Noneheless, he pogam "squeezes" 9 Mio. of ceeals, poein ich, milk based and ohe fodde ou. The esuls shall be sufficien o pove ha a calibaion is needed. Inoducing an adding up consain fo oal use of he diffeen feeding suffs a he naional level (FEDUSEN) and fixing is value o base yea levels leads o: 18 An exac definiion of "hypoheical" will be given lae. A sligh incease in max. fibe deegen was necessay fo cows o achieve feasibiliy. The feed aggegaes ae defined accoding o he SPEL-EU daa base (WOLF 1995). Page 51 of 133

52 Equaion 37 min s.. feed aac, feed feed, feed FEDUSE FEDNG FEDNG FEDUSE AREQ aaac, feed, aac, feed, feed, feed, aaac, c, PRICE NUTR LEVL DAYS aac, feed feed, c = FEDUSEN aaac, c, aac, c = FEDUSE feed feed feed feed The las line ensues ha he known quaniies fed a naional level ae me. Unfounaely, Equaion 36 is no longe a cos minimisaion poblem, because quaniies o be fed a naional level and hei pices ae known! Consequenly, he value of he objecive funcion in Equaion 37, Equaion 38 FEDUSE feed, PRICE feed = feed, feed FEDUSEN feed PRICE is a given consan. If we plug he poblem shown above in a solve i canno squeeze ou any quaniies since oal use is fixed a naional level. I will simply disibue he feed ove egions and animals so ha he equiemens of animals ae me. When he fis feasible soluion fo he equiemen consains is found, i will sop. The disibuion will be abiay, wih poenially devasaing consequences fo he feed coss of he egional aciviies. Naually, some of he consains in Equaion 36 will be binding, bu which ones will be abiay as well (as will be he dual values aibued o hem). Reades familia wih he on-going discussion on Posiive Mahemaical Pogamming (PMP) will now end o lay back and elax because hey know aleady a nice soluion o he calibaion poblem. Thei obvious idea will be o use he dual values on he vaiables FEDUSEN o define addiional non-linea ems o be added o he objecive funcion 19. Naually, pefec calibaion will be guaaneed by he PMP mehodology. Bu wha abou he simulaion behaviou? If he model wihou calibaion bounds poduces a soluion damaically diffeen fom he obseved one (see Table 9), he influence of he PMP ems fo feed on he simulaion behaviou will be emendous. Consequenly, he effec of he equiemen consains will be small, had o judge and depend on he quie abiay soluion fom he calibaion sep. The goal o descibe he subsiuion possibiliies in feeding by an appopiae se of consains would suely be missed. Pehaps he bes advice wih such a soluion would be o leave hem ou compleely. Insead, appopiae own and coss-cos ems beween feeding aciviies would need o be inoduced o descibe he cos minimising behaviou in feeding. The echnological subsiuion possibiliies beween feeding suffs which we aimed a descibing by he consains in Equaion 37 would be mosly hidden as dual infomaion in hese cos ems. Bu hee ae fuhe poblems elaed o a PMP appoach. Since he mos expensive feeding suff will always be squeezed ou fis, dual values of he coesponding calibaion consains will be equal o exogenous pices. Coss-egional o ime seies analysis of he duals fom a sample of egional models will hence no poduce any addiional infomaion. feed 19 PMP fo feed ae used, fo example, in he RAUMIS model (CYPRIS 1999). Page 52 of 133

53 Wihou obseved vaiance, a eliable esimaion is no possible. The - a fis sigh - appealing PMP soluion will consequenly no wok well, if no fuhe infomaion is available. The essence of he agumenaion above is ha he use of an aggegaed cos minimisaion model in simulaion uns is only sensible, if he consains based on he coefficiens (NUTR and REQU) can sufficienly explain obseved feed use in he base yea. Theefoe, he "hypoheical" equiemens mus fis be calibaed wih espec o obseved feeding quaniies in he base yea. Then, as a las eso, PMP can be pu o wok. The geneal poblem is hen o define a se of equiemens which calibaes he feed cos LP as close as possible o saisically obseved quaniies. I is solved by he following seps. (STEP 1.) Definiion of nuiional conen (NUTR) of feeding suffs. Wheeas fo ceeals and ohe aded feeding suffs, he infomaion abou hei conen of enegy, poein ec. is quie accuae, doub may be aised concening fodde (gas, gazings, hay, silage ec.), boh elaed o yield esimaes and nuiional conen. These conens ae hence eaed as endogenous vaiable in he esimaion poblem. (STEP 2.) Definiion of nuiional equiemens (AREQ) fo each animal caegoy in each egion based on so-called equiemen funcions as descibed above. Fo ceain animal caegoies, addiional equiemens o consains ae inoduced (lysine, max. dy mae inake, neual deegen fibe ec.). Fo some of he animal caegoies hey eflec diffeences in egional yields pe head, fo example he milk yield pe cow. The undelying funcions ae based on a lieaue seach and can be undesood as he echnological fonie unde a conolled expeimenal envionmen. (STEP 3.) Calibaion of hese hypoheical equiemens and consains so ha hey ae as fa as possible binding fo he obseved feed quaniies. The necessiy of calibaing he heoeical equiemens can be easily undesood when aking ino accoun he conol cos on fam level elaed o wok exacly on he echnological fonie: he nuien conen of feeding suffs mus be caefully checked and he inake of each feed pe animal exacly weighed. Ohewise, one isks o save he animal, o damage hei healh and o educe yields, wih high coss involved. To exclude ha isk, fames will feed secuely moe han heoeically equied. The calibaion pocess is based on he infomaion compised in he se of "hypoheical" equiemens and feeding consains (enegy, cude poein, dy mae; max. dy mae inake) and he known quaniies fed in he base yea. As he lae ones ae "had" daa o mee, hey seve as consains fo ou calibaion model. The Coss Enopy cieion will minimise deviaion fom appopiae saing poins based on he "hypoheical" equiemens and will ake he effec on he coss of he individual poducion aciviies ino accoun. In conas o mos ohe economeic echniques, he Enopy appoach allows he esimaion of paamees in he case of ill-posed poblems, i.e. if he numbe of obsevaions is less han he numbe of paamees (GOLAN, JUDGE & MILLER 1996). The paamee esimaes ae pobabiliy weighed linea combinaions of given suppo poins (SUP). The objecive will seach a poseioi pobabiliies which show he minimal deviaion of he a pioi pobabiliies aached o he suppo poins. Page 53 of 133

54 In ou case, he expeced value fo each paamee E[p] is a linea combinaion of k = 3 suppo poins SUP k 20 weighed wih he a poseioi pobabiliies PROB k : Equaion 39 E[p] = PROB SUP k s.. PROB = 1 k k k k Thee diffeen ypes of paamees ae endogenously esimaed, which equies he definiion of suppos: 1. Animal equiemen. These ae cened aound he elaion beween oal hypoheical need (e.g. he enegy need of all naional heds as deived fom he equiemen funcion) divided by he oal uncoeced enegy need of all fed quaniies. Moe deails ae found below. 2. Feed cos pe animal. These ae based on he FADN esimaes afe consolidaion wih he feed coss epoed in he Economic Accouns as descibed above. 3. Nuien conen fo fodde (gass, fodde maize, ohe fodde fom aable land), based on he able shown in he following. The acual esimaion poceeds sep-wise ove egional aggegaes, since a simulaneous esimaion acoss all egions pe Membe Sae would ceae a ahe lage and somewha inanspaen famewok. Fis, he poblem is solved a Membe Sae level. The esuling esimaed animal equiemens ae hen used o define suppos a NUTS I and lae a NUTS II level. The nuien conen of he fodde fom he Membe Sae level is aken ove o NUTS I and NUTS II level and hence idenical acoss egions o ease he compaison of he feed inpu allocaion. Suppo fo Requiemens Geneally, suppo poins fo enopy poblems ae based on a pioi infomaion. The highe he spead of he suppos, he weake hei influence on he final soluion. As he flaes disibuion is eached when all pobabiliies ae equal o a pioi ones, suppos should be cened on a plausible expeced value fo he paamee o be esimaed. In ou case, he "hypoheical" equiemens ae unfounaely no such plausible expecaions. As explained above, hey epesen a echnological fonie no applicable o he secoal aveage. If he cos minimizing poblem shown above based on he "hypoheical" equiemens can squeeze ou lage quaniies and hence undeesimaes he feeding coss, i is a clea hin ha hese equiemens ae eihe oo low, no complee o he nuien conen of he feeding suff is oo high. The quesion is no if specific equiemens in ou poblem ae "coec" in he sense ha a fame o he feed compound indusy would use i in deemining he opimal mix, bu if hey ae suiable in descibing he aggegaed subsiuion possibiliies fo feeding suff. 20 The vaiance of he ME-esimaes appoaches a lowe limi if he numbe of suppo poins goes o infiniy (GOLAN, JUDGE, MILLER 1986, p.139). In ess wih a simple "wo paamee" - "one consain" ME poblem, he effec of inceasing he numbe of suppo poins was judged significan only up o fou suppo poins. Beyond fou suppo poins, paamee esimaes sill kep changing sysemaically, bu he change was close o he compuaional accuacy of he compue used. Since he compuaional buden of he solve gows damaically wih inceasing numbe of suppo poins, fou suppos wee chosen fo he analysis. Page 54 of 133

55 Given he lage numbe of feeding suff, hei diffeences in conen in nuiens and pices, we would expec he seco woking moe o less exacly on he aggegaed fonie of he "eal" subsiuion se, i.e. ha mos of he consains of poblem (1) would be binding. We mus hence look ou fo a suiable cene poin fo ou suppos fo he individual equiemens fo which we would expec ha he consains ae binding. Neveheless, he suppos should sill elae o ou "hypoheical" equiemens. As he fis sep, we calculae base yea elaions beween oal "hypoheical" equiemens AREQ h and oal deliveies fo each consain c in (1) a he naional level: Equaion 40 REL c = aac, feed LEVL aac, FEDUSE feed REQU NUTR h aac, c, feed Wih jus one secoal consain, say enegy equiemens, he coecion faco expessed by REL could be applied o all equiemen coefficiens diecly and would ensue ha he enegy consain is "jus" binding. No enopy esimaion would be needed. Howeve fo a complex layou wih up o 6 equiemens, fixed availabiliy of ceain feeding suff in ceain egions, applying he facos o all equiemens would lead o infeasibiliies. The elaions fom Equaion 40 ogehe wih hypoheical equiemens a naional level (ms, based on naional aveage yield) ae used o define he suppo poins fo he calibaion sep a naional level: h Equaion 41 SUP = {.4,1.0,1.6 } REL REQU acc c aác, c, ms 0 c aac, c, ms, In he second sep, egional specific suppo poins ae based on he poin esimaes obained a naional level E[REQU ms ]: Equaion 42 SUP = { 0.1,0.7,1.3,1.9 } [ REQU ] E aac, c, ms aac, c, REQU aac, c, aac, c h REQU aac, c, ms Feed coss As feeding accouns fo a lage pa of coss in animal poducion, coecion of equiemens should ake he effec on coss ino accoun. Theefoe, suppo poins fo feeding coss (COST) ae defined based on he feeding coss and evenues epoed in he SPEL-EU daa base. Expecaions ae cened on he maximum of feeding coss, 80% of oal coss and 30% of oal evenue epoed in he daa base: SUP aac,cos, = Equaion 43 {.01,0.7,1.3,1.9 } max ( 0.3 REVENUES,0.8 TOIN, FEDCST ) aac, c 0 a, ms a. ms Expeced values fo equiemens and feeding coss ae defined by he endogenously deemined pobabiliies and he suppo poins defined above: Page 55 of 133

56 Equaion 44 E[REQU E[COST E[NUTR aac,c, aac, feed,c ] = ] = ] = k k k PROB PROB PROB aac,c,k aac,cos,k feed,c,k Pobabiliies obseve he adding up condiions: Equaion 45 k k k PROB PROB PROB aac,c,k aac,cos,k feed,c = 1 = 1 = 1 aac, c, aac, feed, c SUP SUP SUP aac,c,k aacc,cos,k feed,,c,k aac, c, aac, feed, c Following he noaion esablished in pa 1, he equiemens defined pe day and head ae coveed by: Equaion 46 E[ REQU ] DAYS aac, FEDNG E[ NUTR ] aac c aac, c, aac, feed, feed, c feed,, Adding up he fed quaniies ove he hed sizes, he oal feed use pe egion is defined as Equaion 47 FEDUSE feed, = FEDNGaac, feed, LEVL aac, = f, aac Since some of he feeding suffs (silage, gas, ohe oo cops) ae assumed o be no adable beween egions, he above equaion is eaed as an equaliy consain whee FEDUSE is fixed o he obseved poducion quaniies. Fo all ohe feeding suffs, he following adding up consains a he naional level define binding equaliy consains fo he calibaion pocess whee FEDUSEN defines he exogenously fixed, obseved quaniies fed a Membe Sae level: Equaion 48 FEDUSE feed = FEDUSEN Finally, he objecive funcion wih he coss enopy cieion is defined as Equaion 49 H ( PROB) = aac, c, k aac, k feed, c PROB PROB PROB aac, c, k aac,cos, k feed, c feed log( PROB f log( PROB log( PROB feed, c aac, c, k aac,cos, k The equiemens based on he soluion of he enopy poblem descibed in pa 2 as well as he ade pice deeminaion descibed in pa 3 ae hen inegaed in a famewok whee all he egional supply models fo a Membe Sae ae linked ogehe. This simulaneous soluion is used only once fo he calibaion sep wheeas egional models will be solved independenly in simulaion uns. pq k ) pq k ) pq k ) Page 56 of 133

57 Calibaion consains ensue ha he aciviy levels of he base peiod ae me ha obseved naional feed quaniies ae used up ha obseved egional poducion quaniies of gaze and silage ae fed. The "nomal" consains of he famewok which ae idenical o he one applied in simulaion uns consis of he equiemen consains, aea esicions and poliical esicions such as selling quoas fo milk and suga and he CAP se-aside egime. In he fis calibaion sep, opimal egional ade wih hay and saw is deived. Regional pices diffe fom he unifom naional pice. In he nex sep, he egional models ae solved independenly wih egional feed use fixed a he esuls obseved in he fis sep. On he one hand, he dual values of he calibaion consains ae used in he Maximum Enopy esimaion of he quadaic cos funcion fo he aciviies (HECKELEI & BRITZ 1999). On he ohe hand, dual values fo he fixed feed quaniies ae obained. As in he case of poducion aciviies, he maginal values fo feed ae mapped ino nonlinea ems of he objecive 21. In ode o ge a simple and easy o inepe definiion of hese non-linea ems fo feed use, he quadaic ems BF ae based on an own-pice poin elasiciy of -0.5 fo egional feed use. Linea ems AF ensue ha he dual values obained ae me and ha he model calibaes pefecly o he base yea: Equaion 50 BF AF f, f, PRIC = 0.5 FEDUSE = λ feduse(f,b,) f,b f,b, BF f, FEDUSE f,b, Noe ha "PRIC" in (20) is he unifom naional pice as found in he SPEL-EU daa base and diffe hence fom he pice fo he egionally aded feeding suffs hay and saw obained by he mechanism descibed in pa Discussion When judging he soluions discussed above, he oveall model's objecives and sucue as well as he daa availabiliy mus be kep in mind. Lile is known saisically abou feeding pacises in diffeen egions acoss Euope. Besides, he main analyical objecive of he CAPRI modelling sysem is dieced owads policy impacs on egional and aggegae aciviy subsiuion, i.e. on cop levels and hed sizes, and on he esuling impacs on and feed-backs fom he makes. The disibuion of individual feeding suff o individual animal caegoies is of mino impoance in he oveall conex and maes mosly fom he view poin of influencing he oveall allocaion behaviou. The supply module deals wih bulks of aded feeding suff only: ceeals, ich poein, ich enegy and ohe feed. Silage, gaze, oo cops and aw milk ae modelled on a single 21 Posiive Mahemaical Pogamming in he conex of feed disibuion is also used in he Geman seco model RAUMIS, since 1997 (CYPRIS 1999). Howeve non-linea cos ems ae inoduced fo each feeding aciviy (fo example "whea fed o pigs") and no fo oal egional feed use as done hee. Page 57 of 133

58 poduc basis. Wihou fuhe emedies, he few linea equiemen consains would esul in a quie jumpy behaviou of he egional model. A successful inegaion wih he make module would be impossible. Theefoe, pices fo egionally aded feed depend on he ne ade and PMP-ems ae inoduced fo oal egional feed use fo each feeding suff in he supply module. The quaniies fed in he supply module ae exogenous fixed vaiables in he make module which deemines he mix of he bulks, fo example he shae of whea, baley and maize in he ceeals aggegae simulaneously wih he pices fo he componens (BRITZ 1998). The elasiciies used in double log funcions deemining he shaes ae paly based on esimaed cos funcions fo he Geman feed compound indusy (BRITZ & SIEBER 1998). The esuling new pice fo he bulks as well as hei new nuien conen ae hen handed back in he nex oveall model ieaion o he supply module. The oveall sucue is hence a mixue of pimal and dual appoaches. Hopefully, he soluion mainly inegaes he advanages of he diffeen soluions insead of combining all possible daw-backs. 3.4 Inpu allocaion fo feilises and nuien balances In he following secion, he exising envionmenal indicaos in CAPRI, planned and aleady achieved impovemens, and possible fuhe exensions ae biefly discussed. I should be noed ha CAPRI is basically a egionalised agiculual seco model, hus concenaing on he modelling of aggegaed eacions of agiculual poduces and consumes o changes in long em shifes as echnical pogess, income changes and CAP pogams. Mos indicaos ae ahe obus pessue indicaos and can be calculaed easily based on fixed paamees appoaches fom he endogenous vaiables of he egional aggegae supply models. Accodingly, economic (dis)-incenives can be linked o he pessue indicaos o fuhe passive indicaos can be inoduced o he cuen ones changed easily. So fa he link beween insumens of agi-envionmenal insumens and pessue indicaos had been exploed fo he case of geenhouse gas emissions (Péez 2005). Duing he fis phase of CAPRI ( ), NPK balances and oupu of geenhouse gases had been inoduced, and an enegy use indicao was exploed fo Swizeland. The pojec fo DG-ENV ( ) hen led o (1) he impovemen of he cuen sae indicaos -especially ammonia oupu and niae leaching, (2) he inoducion of new ones as a wae balances and chemical indicaos, (3) feasibiliy sudies fo he applicaion of he Nuien Flow Model fo he Nehelands and he bio-physical model CopSys fo egions in Fance, and (4) impoving he inepeaion of envionmenal indicaos by conasing hem wih soil and land-use maps. The following able shows an oveview of he indicaos embedded in he CAPRI sysem afe he finalisaion of he DG-ENV pojec. Page 58 of 133

59 Table 10 Indicaos in he CAPRI sysem Indicao Linked o Fixed a Souce/Commen NPK oupu a ail Ammonia emissions NPK losses by leaching and soil soage Oupu of geenhouse gases (nious oxide, mehane) Wae balances Niae concenaions in gound wae Regional animal populaion and yields (final weighs, milk yield, lengh of poducion peiod) Animal populaion, housing & soage ype, cop level & yields NPK oupu a ail and ammonia emission, N- cop need Animal heds, mineal feilise Meeoology, managemen, iigaion, soil soil ype, gound wae level, niogen supluses Animal ype Membe sae level EU level Unifom coefficiens pe animal ype and pue mineal nuien fo EU Regional coefficiens pe cop aciviy Region, cops and fam ypes Chemical emissions cop poducion Regional coefficiens pe cop aciviy Souce: CAPRI modelling sysem Fam managemen lieaue, opeaionally embedded in sysem IASSA, pooype embedded; Nuien Flow Model (LEI, Nehelands) Opeaional, cuenly wih old emission facos Coune-check wih Euopean Envionmenal Agency, IPCC ules CopWa model, paial coune-check wih CopSys model Case sudies fo he Nehelands and Fance Case sudies fo he Nehelands and Fance Nuien balances fo NPK Nuien balances in CAPRI ae buil aound he following elemens: Expo of nuien by havesed maeial pe cop depending on egional cop paens and yields. Oupu of manue a ail depending on animal ype, egional animal populaion and animal yields, as final weighs o milk yields. Inpu of mineal feilise as given fom naional saisics a secoal level. The Ammonia emission model (see sub-secion 3.4.3) NPK oupu a ail The oupu of P and K a ail is esimaed based on ypical nuien conens of manue: Page 59 of 133

60 Table 11 Nuien conen in manue in kg pue nuien/m³ P K Cale Swine Pouly Souce: Lufa von Wese-Ems, Sand Apil 1990, Naehsoffanfall. These daa ae conveed ino ypical pue nuien emission a ail pe day and kg live weigh in ode o apply hem fo he diffeen ype of animals. Fo cale, i is assumed ha one live sock uni (=500 kg) poduces 18 m³ manue pe yea so ha he numbes in he able above ae muliplied wih 18 m³ and divided by (500 kg *365 days). Fo he diffeen ypes of cale aciviies, i is hence necessay o deemine he aveage live weigh and he lengh of he poducion pocess. Fo calves faening (CAMF, CAFF), he cacass weigh is divided by 60 % in ode o aive a final weigh and a sa weigh of 50 kg is assumed. Daily weigh inceases ae beween 0.8 kg/day and 1.2 kg/day and depend popoionally on aveage socking densiies of cale in elaion o he aveage EU socking densiy fo which a daily weigh incease of 1 kg/day is assumed. Toal emissions pe animal hence incease wih final weighs bu decease pe kg of mea poduced fo inensive poducion sysems wih high daily weigh inceases. The same elaionship holds fo all ohe animal caegoies discussed in he following paagaphs. Fo calves aising (CAMR, CAFR), wo peiods ae disinguished. Fom 50 o 150 kg, a daily incease of 0.8 kg/day is assumed. The emaining peiod capues he gowh fom 151 o 335 kg fo male and 330 kg fo female calves, whee he daily incease is beween 1 kg/day and 1.4 kg/day, again depending on socking densiies. The bull faening pocess capues he peiod fom 335 kg live weigh o final weigh. Daily inceases ae beween 0.8 kg/day up o 1.4 kg/day, depending on final weighs and socking densiies. Cacass weighs as epoed in he daa base ae e-conveed ino live weigh assuming a faco of 54% fo low and 57% fo highe final weighs. The heifes faening pocess capues he peiod fom 300 kg live weigh o final weigh, assuming a daily incease of 0.8 kg/day. Cacass weighs, as epoed in he daa base, ae e-conveed ino live weigh assuming a faco of 54 % fo low and 57 % fo highe final weighs. Suckle cows ae assumed o be whole yea long in poducion and weigh 550 kg, wheeas milk cows ae assumed o have a weigh of 600 kg and ae again fo 365 days in poducion. Addiional daa elae o he addiional NPK oupu pe kg milk poduced by cows and ae aken fom he RAUMIS model: Table 12 Addiional emission of NPK pe kg of milk poduced N P K Souce: RAUMIS Model (hp:// Page 60 of 133

61 The facos shown above fo pigs ae conveed ino a pe day and live weigh faco fo sows by assuming a poducion of 5 m³ of manue pe sow (200 kg sow) and 15 pigles a 10 kg ove a peiod of 42 days. Consequenly, he manue oupu of sows vaies in he model wih he numbe of pigles poduced. Fo pig faening pocesses, i is assumed ha 1.9 m³ ae poduced pe sandad pig wih a final cacass weigh of 90 kg a 78 % mea conen, a saing weigh of he faening peiod of 20 kg (weigh of he pigle), a poducion peiod of 143 days and 2.3 ounds pe yea. The acual facos used depend on ables elaing he final weigh o ypical daily weigh inceases. Fo pouly, i is assumed ha 8 m³ of manue ae poduced by 100 laying hens, which ae assumed o weigh 1.9 kg and say fo 365 days in poducion. Fo pouly faening pocesses, a faening peiod of 49 days o each 1.9 kg is assumed. Fo sheep and goa used fo milk poducion o as mohe animals, he cale facos ae applied by assuming a live weigh of 57.5 kg and 365 days in poducion. Fo faening pocesses, a daily incease of 200 kg and a mea conen of 60 % of he cacass weigh ae assumed. The niogen emission facos fom animal aciviies ae coupled o cude poein inake (IPCC 1997). Accoding o he lieaue (Udesande e al. 1993), hee is a elaion of 1 o 6 beween cude poein and N in feeding. By combining his infomaion wih N eenion aes pe animal aciviy (IPCC 2000, able 4.15), manue poducion aes can be esimaed (N inake minus N eenion). Table 13 Cude poein inake, manue poducion and niogen eenion pe head (EU 15, yea 2001) Cude poein Niogen in manue Niogen eenion BULH BULL CAFF CAFR CAMF CAMR DCOH DCOL HEIH HEIL HEIR HENS (1000 unis) PIGF POUF (1000 unis) SHGM SHGF SOWS SCOW Souce: CAPRI Modelling Sysem Page 61 of 133

62 3.4.3 The ammonia module Code: ammo/ammoda.gms; ammo/ammo.gms The ammonia (NH 3 ) oupu module akes he niogen oupu pe animal fom he exising CAPRI module and eplaces he cuen fixed coefficien appoach wih unifom Euopean facos pe animal ype by Membe Sae specific ones, aking ino accoun diffeences in applicaion, soage and housing sysems beween he Membe Saes. The geneal appoach follows he wok a IASSA. The following diagam shows he NH 3 sinks aken ino accoun by coefficiens. Figue 5. Ammonia sinks in he Ammonia emission module Niogen fom animals Gazing NH 3 NH 3 sable NH 3 soage NH 3 Oganic N applicaion Cop N Need Mineal N NH 3 Souce: CAPRI modelling sysem In Figue 5, whie aows epesen ammonia losses and ae based on unifom o Membe Sae specific coefficiens. A fis Membe Sae specific coefficien chaaceises fo each animal ype he shae of ime spen on gassland and spen in he sable. Fo daiy cows, fo example, he facos ae beween 41 % spen in he sable in Ieland and 93 % in Swizeland. Duing gazing 8% of he exceed N is assumed los as ammonia. The ime spen in he sable is hen spli up in liquid and solid housing sysems. To give an example, 95 % of he Duch cows ae assumed o use liquid manue sysems, wheeas in Finland 66 % of he cows ae in solid sysems. Ammonia losses in boh sysems ae assumed o be idenical pe animal ypes bu diffe beween animals. 10 % ammonia losses ae assumed fo sheep and goa, 12 % fo cale, 17 % fo pigs and 20 % fo pouly. The emaining niae is hen eihe pu ino soage o diecly applied o he gound. No soage is assumed fo sheep and goas and in all emaining cases no-coveed sysems ae assumed wih a loss faco of 6 % of he N bough iniially ino soage. Page 62 of 133

63 Afe soage, he emaining N is applied o he soil, eihe spead o he suface losses a 20% o using applicaion echniques wih lowe (15%) o high (10%) emission educions. Cuenly, i is assumed ha all fames wok wih he sandad echniques. Technically, he undelying calculaions ae embedded as GAMS code in an own module boh called duing updaes of he daa base and model uns. The elevan coefficiens ae soed as a sepaae able, a anspaen soluion allowing o quickly pefom sensiiviy analysis o updaing hem. Table 14 Niogen balance (EU 15, yea 2001) INPUT OUTPUT Impo of niogen by anoganic feilise a 68.2 Expo of niogen wih havesed maeial f Impo of niogen by oganic feilise (in manue) b Niogen in ammonia losses fom manue fallen on gazings g 2.08 Niogen fom biological fixaion* c 2.89 Niogen in ammonia losses fom manue in sable h 7.13 Niogen fom amospheic deposiion d Niogen in ammonia losses fom manue soage i 2.53 Niogen in ammonia losses fom manue applicaion on he field j 8.34 Niogen in ammonia losses fom oganic feilise k=g+h+i+j Niogen in ammonia losses fom mineal feilise l 2.89 TOTAL INPUT e=a+b+c+d TOTAL OUTPUT n=f+k+l+m Souce: CAPRI modelling sysem Nuien losses a soil level (SURPLUS) m=e-f-k-l Inpu allocaion of oganic and inoganic NPK and he nuien balance The inpu allocaion of oganic and inoganic feilize deemines how much NPK oganic and inoganic feilise is applied pe ha of a cop, simulaneously esimaing he NPK availabiliy in manue. Fisly, nuien expo by he havesed maeial is deemined, based on he following facos: Table 15 Expos of nuiens in kg pe on of yield o consan Euo evenues N P K Sof whea Duum whea Rye Baley Page 63 of 133

64 Oas Gain maize Ohe ceeals Paddy ice Saw Poaoes Suga bee Fodde oo cops Pulses Rape seed Sunflowe seed Soya Ohe oil seeds Texile cops Gas Fodde maize Ohe fodde fom aable land Tomaoes Ohe vegeables Apples, pea and peaches Cius fui Ohe fuis Nuseies, flowes, ohe cops, ohe indusial cops Olive oil Table olives Table gapes Table wine, ohe wine 1.9/ / /0.65 Tobacco Souce: CAPRI modelling sysem The facos above ae applied o he expeced yields fo he diffeen cops consuced wih he Hodick-Pesco file explained above. Muliplied wih cop aeas, hey povide an esimae of oal nuien expo a naional and egional level (igh hand side of he figue below). The maximum expos pe ha allowed ae 200 kg of N, 160 kg of P and 140 kg of K pe ha. Ex-pos, he amoun of nuiens found as inpu in he naional nuien balance is hence known as he sum of he esimaed nuien conen in manue plus he amoun of inoganic feilise applied, which is based on daa of he Euopean Feilise Manufacue s Associaion as published by FAOSTAT. In ode o educe he effec of yealy changes in feilize socks, hee yea aveages ae defined fo he NPK quaniies demanded by agiculue. Page 64 of 133

65 Fo Niogen, ammonia losses ae explicily handled as aken fom he Nuien Flow Model developed by LEI and conibued by he pojec financed by DG-ENV. Remaining losses could eihe be un-off (leaching o accumulaion in soil) and as non-ammonia gas losses (nious oxide). Addiional souces of N ae aken ino accoun as well. Table 16 Amospheic deposiion of N pe kg and yea Ausia 20 Belgium 32 Denmak 18 Finland 5 Fance 16 Gemany 29 Unied Kingdom 15 Geece 7 Iland 10 Ialy 12 Nehelands 36 Noway 5 Pougal 3 Spain 6 Sweden 5 Swizeland 18 Souce: CAPRI modelling sysem Figue 6 offes a gaphical epesenaion of hese elaionships. Page 65 of 133

66 Figue 6. Ex-pos calibaion of NPK balances and he ammonia module Hed sizes Final weighs ec. Naional mineal feilise use Amospheic deposiion Cop aeas yields Manue oupu funcions Nuien conen of haves Oupu a ail NH3-Losses Anoganic Biological fixaion Nuien expo NH3-Losses Oganic % Oganic availabiliy Ne of NH3 losses Deliveies = Use % feilisaion Ove expo Poseioi densiy esimao Souce: CAPRI modelling sysem The following equaions compise ogehe he coss-enopy esimao fo he NPK (Fnu=N, P o K) balancing poblem. Fisly, he puchases (NETTRD) of anoganic feilise fo he egions mus add up o he given inoganic feilise puchases a Membe Sae level: Equaion 51 Ned Fnu MS = Ned Fnu The cop need minus biological fixaion fo pulses muliplied wih a faco descibing feilisaion beyond expos mus be coveed by: (1) inoganic feilise coeced by ammonia losses duing applicaion in case of N, (2) amospheic deposiion, aking ino accoun a cop specific loss faco in fom of ammonia, and (3) nuien conen in manue, coeced by ammonia losses in case of N, and a specific availabiliy faco. Equaion 52 cac = NETTRD + NBal + Levl aac cac NuFac fnu AmDep Levl Fnu aac Fnu biofix cac ( 1 NFacFnu, cac ) ( 1+ NuFacG cac ofa gae, gai) (1 NH3Loss NFac Fnu AmDep Cac aac fnu Anog Fnu, ) (1 NH3Loss Manue Fnu, ) ( 1 NavFac ) fnu Page 66 of 133

67 The faco fo biological fixaion (NFac biofix ) is defined elaive o nuien expo, assuming deliveies of 75 % fo pulses (PULS), 10 % fo ohe fodde fom aable land (OFAR) and 5 % fo gassland (GRAE, GRAI). The faco descibing luxuy consumpion of feilise (NuFac) and he availabiliy facos fo nuien in manue (NavFac) ae esimaed based on he HPD Esimao: min HDP = fnu NuFac σ fnu NuFac fnu µ NuFac fnu 2 Equaion 53 fnu fnu NavFac σ NuFacG, σ Nim, σ fnu NavFac fnu µ fnu NuFac fnu NavFac fnu µ 2 Nim ngp µ ngp LEVL UAAR NavFac ngp ngp LEVL ngp 2 NuFacG fnu The expeced means γ fo he availabiliy fo P and K in manue (Navfac) ae cened aound 50 %, fo N a 50 %*40 %+25 %*86%, since 50 % ae assumed o be eleased immediaely, of which 60 % ae los as ammonia and 25 % ae eleased slowly, wih a cop availabiliy of 86 %. These expeced means a naional level ae muliplied wih he egional oupu of he nuien pe hecae divided by he naional oupu of nuien pe hecae so ha he a pioi expecaion ae highe losses wih highe socking densiies. The lowe limis ae almos a zeo and he uppe limis consequenly a he uniy. The sandad deviaion σ is calculaed assuming a pobabiliy of 1% fo a zeo availabiliy and 1% fo an availabiliy of 100%. The expeced mean γ fo he faco descibing ove-feilisaion pacices (Nufac) is cened aound 120 %, wih a 1% pobabiliy fo 160 % and a 1 % pobabiliy fo 80 % (suppo poins) wih define he sandad deviaion σ. Uppe and lowe limis ae a 500% and 5%, especively. A second faco (Nufacg) is only applied fo gassland and ohe fodde fom aable land and cened aound zeo, wih expeced mean of +10% and a -10% wih pobabiliies of 1%. Bounds fo he faco Nufacg ae a -0.5 and 2.5. The las em elaes o he disibuion of oganic N o he diffeen goup of cops. The disibuion is needed fo simulaion uns wih he biophysical model DNDC (Join Reseach Cene Ispa, Ialy) linked o CAPRI esuls in he conex of he CAPRI-Dynaspa pojec. I is impoan o noe ha he CAPRI appoach leads o nuien oupu coefficien a ail aking ino accoun egional specifics of he poducion sysems as final weigh and even daily weigh incease as well as socking densiies. Fuhe on, an impoan diffeence compaed o many deailed fam models is he fac ha he nuien inpu coefficiens of he cops ae a naional level consisen wih obseved mineal feilise use. The nuien balances ae consains in he egional opimisaion models, whee all he manue mus be spead, bu mineal feilise can be bough a fixed pices in unlimied quaniies. Losses can exceed he magniude of he base yea bu ae no allowed o fall below he base yea value. The lae assumpion could be eplaced by a posiive coelaion beween coss and nuien availabiliy of he manue spead. Thee is hence an endogenous coss-effec beween cops and animals via he nuien balances. 2 Page 67 of 133

68 The facos above ogehe wih he egional disibuion of he naional given inoganic feilise use ae esimaed ove a ime seies. Tend lines ae egessed hough he esuling ime seies of manue availabiliy facos of NPK and cop nuien facos fo NPK, and he esuling yealy aes of change ae used in simulaion o capue echnical pogess in feilise applicaion. The following able shows a summay by highlighing which elemens of he NPK ae endogenous and exogenous duing he allocaion mechanism and duing model simulaions: Table 17 Table: Elemens eneing he of NPK balance ex-pos and ex-ane Given: Ex-pos Hed sizes => Manue oupu Cop aeas and yields => Expo wih haves Naional anoganic applicaion Esimaed: Regional anoganic applicaion Faco fo Feilizaion beyond N expo Manue availabiliy Ex_ane Model esul: Hed sizes => manue oupu Cop aeas and yields => Expo wih haves Naional and Regional anoganic applicaion Given: Faco fo Feilizaion beyond expo (ended) Manue availabiliy (ended) Souce: CAPRI modelling sysem Geenhouse Gases Fo he pupose of modelling GHG emissions fom agiculue, a muli-saegy appoach is followed. I is impoan o ake ino accoun ha agiculue is an impoan emie of seveal climae elevan gases ohe han cabon dioxide. Theefoe, wo ypes of polluans ae modelled: mehane (CH 4 ) and nious oxide (N 2 O). The souces consideed ae: CH 4 emissions fom animal poducion, manue managemen and ice culivaion and N 2 O fom agiculual soils and manue managemen 22. In CAPRI consisen GHG emission invenoies fo he Euopean agiculual seco ae consuced. As aleady menioned, land use and niogen flows ae esimaed a a egional level. This is he main infomaion needed o calculae he paamees included in he IPCC Good Pacice Guidance (IPCC, 2000). The following able liss he emission souces modelled: 22 Cabon sinks ae no included since he measuemen of cabon dioxide absopion hough agiculual biomass is highly complex (high unceainy involved, especially in agiculual soils) and has song linkages wih ohe economic aciviies no consideed in his analysis, such as bio-diesel poducion and foesy managemen. Page 68 of 133

69 Table (1) Agiculual geenhouse gas emission souces included in he model Geenhouse Gas Emission souce Code Eneic femenaion CH4En Mehane Manue managemen CH4Man Rice poducion CH4Ric Manue managemen N2OMan Manue exceion on gazings N2OGa Emissions fom synheic feilise N2OSyn Emissions fom oganic animal wase N2OWas Nious Oxide Emissions fom feilise applicaion N2OApp Emissions fom cop esidues N2OCo Souce: CAPRI Modelling Sysem Emissions fom niogen-fixing cops Indiec emissions fom ammonia losses Emissions fom amospheic deposiion N2OFix N2OAmm N2ODep Fo a deailed analysis of hese single emission souces efe o Péez Page 69 of 133

70

71 4 Baseline Geneaion Model (CAPTRD) Code: capd.gms The aim of he CAPRI pojecion ool is o povide a baseline used as compaison poins o compaison ime seies fo counefacual analysis. The baseline may be inepeed as a pojecion in ime coveing he mos pobable fuue developmen o he Euopean agiculual seco unde he saus-quo policy and including all fuue changes aleady foeseen in he cuen legislaion. Concepually, he baseline should capue he complex ineelaions beween echnological, sucual and pefeence changes fo agiculual poducs wold-wide in combinaion wih changes in policies, populaion and non-agiculual makes. Given he complexiy of hese highly ineelaed developmens, baselines ae in mos cases no a saigh oucome fom a model bu developed in conjuncion of end analysis, model uns and expe consulaions. In his pocess, model paamees such as e.g. elasiciies and exogenous assumpions such as e.g. echnological pogess capued in yield gowh ae adjused in ode o achieve plausible esuls (as egaded by expes, e.g. Euopean Commission pojecions). I is almos unavoidable ha he pocess is somewha inanspaen. Two ypical examples ae discussed hee. In he case of he AgLink modelling sysem of he OECD, quesionnaies ae sen ou o he OECD Membe Saes coveing all endogenous and exogenous vaiables of AgLink. The Membe Saes fill in ime seies egading he fuue developmens fo hei especive counies. The values inpued may sem hemselves fom couny specific model baselines, expe consulaions, end analyses o ohe souces in many cases, hei povenience is no known in deail. The OECD hen ses he consan ems in all behavioual equaions of AgLink so ha he couny modules would exacly ecove he values fo he endogenous vaiables fo ha couny found in he quesionnaies a he values inpued fo he exogenous vaiables. Clealy, as he counies will fill in hei quesionnaie wihou knowing abou he fuue expecaions of ohe OECD Membes, he expecaions of he diffeen eams e.g. egading impos/expos o wold make pices may diffe and lead o values a couny level which ae muually no compaible when linked globally ogehe in he modelling famewok. To eliminae such diffeences, he OECD will epeaedly sa AgLink o geneae echnically compaible esuls and eceive commens on hese uns which will lead o updaed daa in he quesionnaies and hus new shif ems in he behavioual equaions. The second example is ha of FAPRI whee a so-called meling down meeing is oganised whee he modelles esponsible fo specific pas of he sysem come ogehe wih make expes. Resuls ae discussed, paamees and assumpions changed unil hee is consensus. Lile is known abou how he pocess woks exacly, bu boh examples undeline he ineacion beween model mechanism and ex-ane expecaions of make expes. This secion explains in deail he mehodology used in CAPRI o consuc a baseline. Befoe eneing ino hese deails i should be saed ha ulimaely almos any pojecion may be educed o a paicula ype of end pojecions, a leas if he exogenous inpus, such as populaion, pices o household expendiue ae also pojeced (usually by ohe eseach eams) as funcions of ime. In his sense end pojecion may povide a fim gound on which o build pojecions and his is exacly hei pupose in ou wok. These ends ae supplemened in he CAPRI baseline ools wih esuls fom ohe baselines, especially fom DG-AGRI. Page 71 of 133

72 The pojecion ool is fed boh by foecass fom diffeen expes o modelling ools, as well by end foecass using daa fom he COCO daabase 23 as ex-pos infomaion. The pupose of hese end esimaes is, on he one hand, o compae expe foecass wih a puely echnical polongaion of ime seies and, on he ohe hand, o povide a safey ne posiion in case no values fom exenal pojecion ae available. Theefoe, end vaiables fo baseline geneaion in he model ae mainly consuced ou of expe daa on pojecions (e.g. FAO, Euopean Commission o Wold Bank) and linea ends of daa conained in he CAPRI daa base. These end vaiables ae simulaneously subjec o he consisency esicions imposed by he mahemaical pogamming model and no made as independen foecass fo each ime seies (e.g. closed aea and make balances). The esuling esimao is hence a sysem esimao unde consains whose popeies ae discussed in he following secion. Noneheless i is o be acknowledged hee ha he end emain mechanical in ha hey y o espec echnological elaionships bu emain ignoan abou behavioual funcions o policy developmens Tend cuve The fis ingedien in he esimao is he end cuve iself which is defined as: Equaion 54 X j, Tend = a j + b j c, j whee he paamees a, b and c ae o be esimaed so ha he squaed deviaion beween given and esimaed daa ae minimized. The X sands fo he daa and epesens a five dimensional aay, spanning up poducs i and iems j (as feed use o poducion), egions poins in ime and diffeen daa saus as Tend o Obseved. The end cuve iself is a kind of Box-Cox ansfomaion, as paamee c is used as he exponen of he end. Fo c equal uniy, he esuling cuve is a saigh line, fo c beween 0 and 1, he cuve is concave fom below, i.e. inceasing bu wih deceasing aes, wheeas fo c > 1, he cuve is convex fom below, i.e. inceasing wih inceasing aes. In ode o peven diffeences beween ime poins o incease shaply ove he pojecion peiod, he paamees c ae esiced o be below 1.2. In a fis pooype of he module, a polynomial end cuve of degee wo was evaluaed. Howeve a quadaic funcion is no necessaily monoone on he foecas ineval so ha a end cuve may fo example show inceasing yields fo he fis pa of he pojecion peiod and afewads a decease. As such oucomes ae puely echnical and no moivaed by a pioi knowledge, i was deemed moe plausible o swich o he fomulaion shown above wih he same numbe of fee paamees as a quadaic end cuve, bu wih monoony guaaneed. The ex-pos peiod coves he peiod fom 1985 owads In ode o cu down he size of he esuling poblem, he ex-ane peiod is defined in en yeas seps (2003, 2010, 2020, 2030), as inemediae yeas can be simply calculaed once he esimaed paamees ae known. 23 Biz, W., Wieck, C., Jansson, T. (2002): Naional famewok of he CAPRI-daa base - he COCO Module, CAPRI Woking Pape 02-04, Insiue of Agiculual Policy, Bonn. 24 The only excepion is he quoa egime on he milk make which has been ecognised in he end pojecions in ha he milk poducion has been deived fom he quoa endowmens (whee cuen quoas ae assumed o pesis unil 2025). Page 72 of 133

73 4.2 Consisency consains in he end pojecion ool The consains in he end pojecion enfoce muual compaibiliy beween baseline foecass fo individual seies in he ligh of elaions beween hese seies, eihe based on definiions as poducion equals yield imes aea o on echnical elaions beween seies as he balance beween enegy deliveies fom feed use and enegy equiemens fom he animal heds. The se of consains is deemed o be exhausive in he sense as any fuhe esicion would eihe no add infomaion o equie daa beyond hose available. The undelying daa se akes ino accoun all agiculual aciviies and poducs accoding o he definiion of he Economic Accouns fo Agiculue. The consains discussed in he following can be seen as a minimum se of consisency condiions necessay fo a pojecion of agiculual vaiables. As discussed above in deail, he full pojecion ool feaues fuhe consains especially elaing o pice feedbacks on supply and demand Consains elaing o make balances and yields Closed make balances define he fis se of consains and sae ha he sum of impos (IMPT) and poducion (GROF) mus be equal o he sum of feed (FEDM) and seed (SEDM) use, human consumpion (HCOM), pocessing (INDM,PRCM), losses (LOSM) and expos (EXPT): Equaion 55 X,IMPT,Tend GROF,Tend + X = FEDM,Tend SEDM,Tend PRCM,Tend INDM,Tend X X X X LOSM,Tend HCOM,Tend,EXPT,Tend + + X X X Whee ae he Membe Saes of he EU, i ae he poducs, he diffeen foecasing yeas. All elemens of he make balances ae expessed as pimay poduc equivalens accoding o he concep of supply uilizaion accouns. Human consumpion of whea does hence include floo bead, pasa ec. ecalculaed ino wha equivalen based on convesion facos. The only expecaions ae oilseeds, whee pocessing o cakes and oils is explicily coveed, and aw milk, whee again, pocessing o he diffeen daiy poducs is included explicily. Secondly, poducion (GROF) is equal o yield imes aea/hed size (LEVL) whee acs ae all poducion aciviies: Equaion 56 X GROF, Tend = acs X acs, Tend X acs, Tend LEVL, A se of equaions elaes o he hecaes fo goups of cop aciviies (ceeals, oilseeds, indusial cops, vegeables, fesh fuis, oal vineyads, fodde poducion on aable land). I defines e.g. ha he oal hecaes of ceeals is equal o he sum of hecaes fo he individual ceeals as sof whea, duum whea, baley and so foh. Equaion 57 cop _ gp, Tend X = X, LEVL, j, Tend LEVL, j cop _ gp Equally, he make balance posiions fo ceain poducs ene adding up equaions fo goups of poducs (ceeals, oilseeds, indusial cops, vegeables, fesh fuis, oal vineyads, fodde poducion, mea). As an example, oal ceeal poducion is equal o he sum ove he poduced quaniies of he individual ceeals. Page 73 of 133

74 Equaion 58 MkBal, Tend X = X, po _ gp, MkBal, Tend i po _ gp Consains elaing o agiculual poducion Adding up ove he individual cop aeas defines he oal uilizable agiculual aea (UAAR,LEVL): Equaion 59 UAAR, Tend X = X, LEVL, cops cops, Tend LEVL, Fuhe consains link he diffeen animal aciviies ove young animal makes: Equaion 60 GROF, Tend X = X, oyan GROF, Tend iyan iyani oyani Whee oyani sands fo he diffeen young animals defined as oupus (young cows, young bulls, young heifes, male/female calves, pigles, lambs and chicken). These oupus ae poduced by aising pocesses, and used as inpus in he ohe animal pocesses (faening, aising o milk poducing). Finally, balances fo enegy and poein equiemens fo each animal ype maac ae inoduced as: Equaion 61 maac, Tend Con, Tend Con Con Tend ( maac, X feed, X feed, = amaac + amaac X yield, ) feed whee Con ae he conens in ems of enegy and cude poein. The lef hand side of he equaion defines oal delivey of enegy o poein fom he cuen feeding pacise pe animal aciviy in egion wheeas he igh hand side he need pe animal deived fom equiemen funcions depending on he main oupu (mea, milk, eggs, pigles bon). The paamees a and b of he equiemen funcions ae esimaed fom engineeing funcions as implemened in he CAPRI modelling sysem, and scaled so ha he balance holds fo he basis peiod. The faco in fon of he equiemens inoduces some inpu saving echnical pogess of -0.4% pe annum. The feeding coefficiens muliplied wih he hed sizes define oal feed use fo he diffeen feeding suffs bulks (ceeals, poein ich, enegy ich, daiy based, ohe) and single nonadable feed (gass, maize silage, fodde oo cops, saw, milk fo feeding, ohe fodde fom aable land): Equaion 62 X FEDM, Tend feed, = maac X maac, Tend feed, X maac, Tend " levl", Finally, he feed use of individual poducs mus add up o he feed use of he bulks menioned above: Equaion 63 FEDM, Tend X = X, feed, o fed FEDM, Tend o, Consains elaing o pices, poducion values and evenues The check of exenal foecass evealed ha fo some poducs, pice pojecions ae no available. I was decided o include pices, value and evenues pe aciviy in he consained Page 74 of 133

75 esimaion pocess. The fis equaion defines he value (EAAG, posiion fom he Economic Accouns fo Agiculue) of each poduc and poduc goup as he poduc of poducion (GROF) imes he uni value pices (UVAG): Equaion 64 X EAAG, Tend = X GROF, Tend X UVAG, Tend The evenues of he aciviies (TOOU, oal oupu) fo each aciviy and goup of aciviies acs ae defined as: Equaion 65 X acs, Tend TOOU, = o X acs, Tend o, X UVAG, Tend o, As fo he make balances, he values fo ceain aggegae poduc goups mus add up: Equaion 66 EAAG, Tend X = X, po _ gp, EAAG, Tend i po _ gp Consume pices (UVAD) ae equal o poduce pices (UVAG) plus a magin (CMRG): Equaion 67 X UVAD, Tend UVAG, Tend = X + X CMRG, Tend Consains elaing o consume behaviou Human consumpion (HCOM) is defined as pe head consumpion muliplied wih populaion: Equaion 68 X HCOM, Tend = X INHA, Tend X INHA, Tend LEVL, Consume expendiues pe capu (EXPE) ae equal o human consumpion pe capu (INHA) imes consume pices (UVAD): Equaion 69 X EXPE, Tend = X INHA, Tend X UVAD, Tend LEVL, As fo he make balances, he pe capu expendiue (EXPE) fo ceain aggegae poduc goups including an aggegaion ove all poducs - mus add up: Equaion 70 EXPE, Tend X = X, po _ gp, EXPE, Tend i po _ gp Consains elaing o pocessed poducs Makeable poducion (MAPR) of seconday poducs (sec) - cakes and oils fom oilseeds, molasses and suga ice and sach - is linked o pocessing of pimay poducs (PRCM) by pocessing yields (PRCY): Equaion 71 X MAPR, Tend sec, = X i sec i PRCM, Tend X PRCY, Tend sec, In case of poducs fom deived milk (mlkseco) bue skimmed milk powde cheese, fesh milk poducs, ceam, concenaed milk and whole milk powde -, fa and poein conen (MLKCNT) of he pocessed milk (COMI cow milk, SHGM sheep & goa milk) mus be equal o he conen of he deived poducs: Page 75 of 133

76 Equaion 72 X = PRCM, Tend COMI, X mlk sec o X MAPR, Tend mlk sec o, MLKCNT, Tend COMI, X + X MLKCNT, Tend mlkseco, PRCM, Tend SHGM, X MLKCNT, Tend SHGM, Consains elaing o policy Thee ae wo consains: fisly, he aceage unde compulsaoy se-aside mus be equal o he se-aside obligaions of he individual cops: Equaion X Tend = cac, X " levl", cac, Tend " OSET ", Tend " se", X " levl",, cac " se", Tend ( X cac ) Secondly, milk poducion is fixed o he milk quoa, modified by evenual unde- o ovedeliveies in he base yea Consains elaing o gowh aes Duing esimaion, some safeguads egading he size of he implici gowh aes had been inoduced: Toal agiculual aea is no allowed o decline a a ae exceeding -0.5 % pe annum. Changes in human consumpion pe capu fo each of he poducs canno exceed a gowh ae of +/- 2% pe annum. Due o some song and ahe implausible ends fo oal mea and ceeals consumpion, he gowh ae hee was esiced o +/- 0.8 % pe annum fo mea and +/- 0.4% pe annum fo ceeals assuming ha end shifs beween single iems ae moe likely han song ends in aggegae food goups. Changes in pices ae no allowed o exceed a gowh ae of +/- 2% pe annum. The numbe of calves bon pe cow is no allowed o exceed a ange of +/- 10 % aound he base peiod value unil he las pojecion yea. Final faening weighs mus fall ino a coido of +/- 20% aound he base peiod value. Song inceases in pok poducion in he pas ae esiced by envionmenal legislaion in foce, noably he niae diecive. Accodingly, inceases wee esiced o +1% fo EU15 Membe Saes (+0.5% fo Denmak and The Nehelands) pe annum. Milk yields pe daiy cows wee esiced by an uppe bound of lies pe cow and yea. Shaes of aable cop on oal aable aea ae bounded by he fomula which allows small shaes o expand o shink moe compaed o cops wih a high shae. A cop wih a base yea shae of 0.1% is allowed o expand o 2.5%, one of 10% only o 25%, and one of 50% o only 70%: Page 76 of 133

77 aab,tend "level", X.up/lo= X aab,tend "level",bas Equaion 74 1 aab,tend 4 1 "level",bas "aab",tend 4 "aab",tend "level",bas X"level",bas X bas ± X max 0.2, las bas 4.3 Thee-sage pocedue fo ends The esimaion pocess is a wo-sage pocedue, whee esuls fom pevious seps feed ino he cuen on Sep 1: Unesiced ends The fis sage esimaes unesiced end cuves. The opimal values of he esimaed end paamees a, b and c ae defined by minimizing squaed eos nomalized wih he mean of he ime seies (fo echnical easons, solely), using he end as weighs: j," Daa" c, j Equaion 75 X j, expos a j + b jexpos SSQ = j," Daa" j, expos X mean The weighing wih he end was inoduced afe a caeful analysis of he esuls of he fis sep. Fis of all, i eflecs he fac ha saisics fom he ealy yeas (mid eighies) ae ofen less eliable hen hose fom lae yeas. Secondly, is moves he cene of gaviy of he esimaion in diecion of he base peiod which is used as a kind of fallback posiion he wose he fi of he above equaion. The esuling paamees povide fisly a saing poin fo he consained esimaions. Secondly, he vaiance of he esuling eo ems defines he weighs fo he nex wo seps. And hidly, he end esimae ogehe wih R² fom ha fis sep is used o define he suppo poin fo he nex seps: 2 expos " Suppo" 2 c, i, j 2 " Daa" Equaion 76 ( ) ( ) X j, exane = R a j + b jexane + 1 R X j, bas The suppo poin is hence he weighed aveage of he end foecas and he base yea values, defined as a five yea aveage aound Sep 2: Consained ends a Membe Sae level The second sep adds he consisency condiions discussed above. In almos all cases, he unesiced end esimaes fom he fis sep would violae one o seveal of he consisency condiions. We need hence now o find esimaes which boh fi ino he consisency consains and exploi in a echnical feasible way he infomaion compised in he ex-pos developmen. Take he second ype of consisency consains as an example, which defines poducion as hecaes/hed sizes imes yield. Clealy, we would like ou ex-ane end esimaes o fulfil ha condiion. Howeve unning independen end esimaes fo baley aea, baley yield and baley poducion will almos ceainly poduce esimaes whee poducion is no equal o yield imes aea. One soluion would be o dop one of he hee esimaes, say yield, and eplace i insead by he division of foecased poducion by foecased aceage. Howeve by doing so, we delibeaely how away he infomaion compised in he developmen of baley yield ove ime. Adding he kind of definiional Page 77 of 133

78 elaions beween he ime seies does hence help us o exploi moe infomaion han is compised in single seies, and efains fom howing away ex-ane pas of he infomaion available. Howeve when esimaing simulaneously he diffeen ends, we need o eflec if he sum of squaes (SSQ) as a penaly funcion sill woks easonable. A nice popey is he fac ha song ends i.e. such wih a high explanaoy powe will dominae weak ones. Howeve as ou las foecased poin is fa away fom he mean, changing slighly he paamees could lead o dasic diffeences in he esimaes wihou a sizeable effec especially on he SSQ when i is aleady small. Especially shaky ends will show values a he ails which can be fa away fom hose obseved ex-pos. We need hence a safeguad which daws ou esimaes o a easonable value in such cases. The confidence ineval fom he end esimae will no help, as i will be cened aound he ail value and simply be quie lage fo bad R². Howeve we may use he agumenaion undelying he usual es saisics fo he paamees elaed o he end (a,b,c). These saisics es he pobabiliy of (a,b,c) being significanly diffeen fom zeo. I can be shown ha hese ess ae songly elaed o R² of he egession. If he zeo hypoheses would be ue, i.e. if he esimaed paamees would have a high pobabiliy of being zeo, we would no use he end line, bu he mean of he seies insead. The easoning behind he es saisics is he basis fo he suppos defined above. We modified i howeve o mach he poblem a hand. Fis of all, we used a hee-yea aveage based on he las known values as he fallback posiion and no he mean of he seies. Secondly, in ypical economeic analysis, es saisics would only be epoed fo he final esimaion layou, some vaiables would have been dopped fom he egession befoehand if ceain pobabiliy hesholds ae undecu. Fo ou applicaions, we oped fo a coninuous ule as i would simply be impossible o analyze manually each and evey end line and decide upon an alenaive esimaion. The coninuous ule daws he esimaes songe in diecion of ou H0 he value is equal o he hee yea aveage aound he las known poins he shakie he esimaed paamees ae. The esuling penaly funcion is defined as minimizaion of he squaed deviaions fom he suppos defined above, weighed wih he vaiance of he eo ems fom he fis sep: Equaion 77 Penaly = j, exane X j," Tend " j, exane X X j," Sep1" " vae" j," Suppo" exane The value used by ha penaly funcion fo each ime poin consiss hence of wo elemens: (1) he diffeence beween he end esimae fiing ino he consisency condiions and he suppos deived fom he unesiced ends, and (2) he vaiance of he eo ems fom he end esimaes. Fo all unesiced end lines, he mean eo will be zeo so ha i canno be used as a cieion. Insead, he vaiance of he eo em is used as a measuemen fo he magniude of he eo ems. I is deceasing wih he mean of he explanaoy vaiable and wih a bee fi of he end cuve. Nomalizing wih he vaiance of he eo ems will hence ensue ha elaive ahe han absolue deviaions ae penalized, and ha deviaions fom he suppo ae penalized songe whee he end had a high explanaoy powe. How is he fis elemen of he em moivaed, i.e. he squaed diffeence beween he esiced end esimaes and he suppos? If R² fo a ceain ime seies is 100%, he penaly is defined as he squaed diffeence beween he esiced end esimae and he unesiced 2 Page 78 of 133

79 one (see definiion of he suppo above). In ohe wods: fo a pefec fi, he esiced end esimae is dawn owads he unesiced end esimae. If R² is zeo, and he end cuve does no explain any of he vaiance and he pobabiliy fo (a,b,c) being equal o zeo becomes maximal. Consequenly, we le he solve find he minimal squaed diffeence beween he base daa poins and he esiced end esimae as he suppo becomes equal o he base daa. The base daa epesen a hee-yea aveage aound he las hee known yeas. Fo all cases in beween, we minimize squaed diffeence fom he weighed aveage of he unesiced end esimae weighed wih R² and he hee-yea aveage weighed wih (1-R²). The weighs ensue ha deviaions fo lines wih a secue unesiced fi ae smalle han fo ime seies wih moe shaky ends. Geneally, all end esimaes ae esiced o he nonnegaive domain. Fo seleced vaiables, insead of using solely he mechanisic coidos shown above, addiional esimaions coidos had been inoduced as discussed above. Oiginally, i was foeseen o add a hid sep whee aggegaion o EU level should be added as an addiional laye of infomaion, wih some elemens as ne ade and impos/expos no planned o be included in he esimaion sep a Membe Sae level. Howeve duing he developmen of he ool, he numbe of simulaneously esimaed iems and hei elaions capued by he consains inceased so ha an inegaion of he individual Membe Sae modules ino one famewok wih addiional adding up consains o EU level became echnically no longe feasible. Insead, he elemens planned o be solely included in he EU aggegaion sep, namely he posiions elaing o ne ade, whee added o he individual Membe Sae modules Sep 3: Adding suppos based on exenal esuls and beaking down o egional level In he final esimaion sep, esuls fom exenal pojecions on make balance posiions (poducion, consumpion, ne ade ec.) and on aciviy levels ae added. Cuenly, hese pojecions ae povided by DG-AGRI. As DG-AGRI is he main clien, i is deemed sensible o foce he pojecions o comply wih he DG-AGRI baseline wheeve he consains of he esimaion poblem allow fo i. Tha is achieved by wo changes o he objecive funcion: 1. Suppos ae eplaced by he esuls of DG-AGRI baseline, he lae popoionally scaled so ha esuls fom he DG-AGRI baseline and he CAPRI daa base ae idenical. 2. Deviaions agains DG-AGRI esuls ae weighed 100 imes highe as end based suppos. Accodingly, he Sep 3 objecive funcion is defined as: Equaion 78 Penaly = j," DG AGRI " j, exane X, i, exane + j," DG AGRI " j, exane X, i, exane X X j," Tend " j, exane j," Tend " j, exane X X X j," Sep1" " vae" X j," Sep1" " vae" j," Suppo" exane 2 j," DG AGRI" exane *100 Page 79 of 133

80 The esuls a Membe Sae level ae hen boken down o egional level, ensuing adding up of aeas and poducion: Equaion 79 Equaion 80 GROF, Tend X MS = X, MS j, Tend X MS = X," levl", MS GROF, Tend j, Tend " levl", Beaking down esuls fom Membe Sae o egional level Even if i would be pefeable o add he egional dimension aleady duing he esimaion of he vaiables discussed above, he dimensionaliy of he poblem endes such an appoach unfeasible. Insead, he sep 3 pojecion esuls egading aciviy levels and poducion quaniies ae aken as fixed and given, and ae disibued o he egions minimizing deviaion fom egional suppos. Thee ae only fou esicions acive: The se-aside obligaions a egional levels Adding up of egional aeas o Membe Sae aeas Adding up of egional poducion o Membe Sae poducion Adding up cop aciviies o uilisable agiculual aea. In ode o keep developmens a egional and naional level compaable, elaive changes in aciviy levels ae no allowed o deviae moe hen 50% fom he naional developmen, in case of yields, developmen is bounded o a +/-20% ange elaive o he naional one. These bounds ae sofened in cases of infeasibiliies. 4.4 Calibaing he model o he pojecion Calibaing he egional supply models The supply side models of he CAPRI simulaion ool ae pogamming models wih an objecive funcion. A calibaion o he esuls of he pojecion ools hus equies ha fis ode opimaliy condiions (maginal evenues equal o maginal coss, all consains feasible) hold in he calibaion poin fo each of he NUTS 2 models. The consequences egading he calibaion ae wofold: (1) elemens no pojeced so fa bu eneing he consains of he supply models mus be defined in such way ha consains ae feasible, and (2) he cos funcions of he models mus be shifed such ha maginal coss and maginal evenues ae equal in he calibaion poin. As explained above, he equiemen funcions used in he pojecion ools ae a linea appoximaion fo he ones used in he simulaion ool; addiional consains esic on op he feed mix in he supply modules. Fuhe on, he feed mix was only pojeced a Membe Sae, no a NUTS 2 level. I is hence necessay o find a feed mix in he pojeced poin which exhauss he pojeced poducion of non-adable feed and he pojeced feed mix of he bulks as ceeals, fis in he equiemen consains and leads o plausible feed cos. In ode o do so, he feed allocaion famewok is e-used. The esuling facos ae soed in exenal files and eloaded by counefacual uns. Page 80 of 133

81 Secondly, mehods boowed fom Posiive Mahemaical Pogamming ae applied o define he diffeence beween maginal evenues and maginal coss in he calibaion poin, and hese diffeences ae added o he aciviy specific consan ems of he non-linea cos funcion. The esuling paamees ae as well soed in exenal files o be eloaded in case of counefacual uns Calibaing he global ade model The pojecion esuls a EU25 level plus Noway, Bulgaia and Romania ae aken as given when he global ade model is calibaed. Tha calibaion sep on he one hand defines bilaeal impo and expo flows fom hese counies o ohe ade blocks, as well as developmen in poducion, feed use, pocessing and human consumpion fo he diffeen egions of he wold no coveed by he pojecion ool. These developmens ae cuenly almos exclusively based on pojecions by he FAO. Page 81 of 133

82

83 5 Simulaion Scenaio Model (CAPMOD) Code: capmod.gms 5.1 Oveview of he sysem The CAPRI simulaion ool is composed of a supply and make modules, inelinked wih each ohe. In he supply module, egional agiculual supply of annual cops and animal oupus is modelled by an aggegaed pofi funcion appoach unde a limied numbe of consains: land, policy esicions such as sales quoas and se aside obligaions and feeding esicions based on equiemen funcions. The undelying mehodology assumes a wo sage decision pocess. In he fis sage, poduces deemine opimal vaiable inpu coefficiens pe hecae o head (nuien needs fo cops and animals, seed, plan poecion, enegy, phamaceuical inpus, ec.) fo given yields, which ae deemined exogenously by end analysis (daa fom EUROSTAT). Nuien equiemens ene he supply models as consains and all ohe vaiable inpus, ogehe wih hei pices, define he accouning cos maix. In he second sage, he pofi maximising mix of cop and animal aciviies is deemined simulaneously wih cos minimising feed and feilise in he supply models. Availabiliy of gass and aable land and he pesence of quoas impose a esicion on aceage o poducion possibiliies. Moeove cop poducion is influenced by se aside obligaions and animal equiemens (e.g. goss enegy and cude poein) ae coveed by a cos minimised feeding combinaion. Feilise needs of cops have o be me by eihe oganic nuiens found in manue (oupu fom animals) o in puchased feilise (aded good). A cos funcion coveing he effec of all facos no explicily handled by esicions o he accouning coss as addiional binding esouces o isk- ensues calibaion of aciviy levels and feeding habis in he base yea and plausible eacions of he sysem. These cos funcion ems ae esimaed fom ex-pos daa o calibaed o exogenous elasiciies. Fodde (gass, saw, fodde maize, oo cops, silage, milk fom suckle cows o mohe goa and sheep) 25 is assumed o be non-adable, and hence links animal pocesses o he cops and egional land availabiliy. All ohe oupus and inpus can be sold and puchased a fixed pices. Selling of milk canno exceed he elaed quoa, he suga bee quoa egime is modelled by a specific isk componen. The use of a mahemaical pogamming appoach has he advanage o diecly embed compensaion paymens, se-aside obligaions, volunay se-aside and sales quoas, as well as o capue impoan elaions beween agiculual poducion aciviies. No a leas, envionmenal indicaos as NPK balances and oupu of gases linked o global waming ae diecly inpued in he sysem. The make module beaks down he wold ino 12 couny aggegaes o ading panes, each one feauing sysems of supply, human consumpion, feed and pocessing funcions. The paamees of hese funcions ae deived fom elasiciies boowed fom ohe sudies and modelling sysems and calibaed o pojeced quaniies and pices in he simulaion yea. Regulaiy is ensued hough he choice of he funcional fom (a nomalised quadaic funcion fo feed and supply and a genealised Leonief expendiue funcion fo human consumpion) and some fuhe esicions (homogeneiy of degee zeo in pices, symmey 25 A deailed descipion can be found in: Wolfgang Biz & Thomas Heckelei (1999): Calibaion of Feed Requiemens and Pice deeminaion of feed in CAPRI, CAPRI woking pape 99-06, available on he pojec web sie. (hp:// Page 83 of 133

84 and coec cuvaue). Accodingly, he demand sysem allows fo he calculaion of welfae changes fo consumes, pocessing indusy and public seco. Policy insumens in he make module include bilaeal aiffs and poduce o consume subsidy equivalen pice wedges (PSE/CSE). Taiff ae quoas (TRQs), inevenion sales and subsidised expos unde he Wold Tade Oganisaion (WTO) commimen esicions ae explicily modelled fo he EU 15. In he make module, special aenion is given o he pocessing of daiy poducs in he EU. Fis, balancing equaions fo fa and poein ensue ha hese make use of he exac amoun of fa and poein conained in he aw milk. The poducion of pocessed daiy poducs is based on a nomalised quadaic funcion diven by he egional diffeences beween he make pice and he value of is fa and poein conen. Then, fo consisency, pices of aw milk ae decomposed ino hei fa and poein conen valued wih fa and poein pices. The make module compises of a bilaeal wold ade model based on he Amingon assumpion (Amingon, 1969). Accoding o Amingon s heoy, he composiion of demand fom domesic sales and diffeen impo oigins depends on pice elaionships accoding o bilaeal ade seams. This allows he model o eflec ade pefeences fo ceain egions (e.g. Pama o Manchego cheese) ha canno be obseved in a ne ade model. The equilibium in CAPRI is obained by leing he supply and make modules ieae wih each ohe. In he fis ieaion, he egional aggegae pogamming models (one fo each Nus 2 egion) ae solved wih exogenous pices. Regional agiculual income is heefoe maximised subjec o seveal esicions (land, feilise need, se-aside, ec). Afe being solved, he egional esuls of hese models (cop aeas, hed sizes, inpu/oupu coefficiens, ec.) ae aggegaed o Membe Sae level models, which ae hen calibaed using Posiive Mahemaical Pogamming (PMP) esimaion echniques. Young animal pices ae deemined by linking hese calibaed Membe Sae models ino a non-spaial EU ade model wih make balances fo young animals, as shown in figue 7. In he second ieaion, supply and feed demand funcions of he make module ae fis calibaed o he esuls fom he supply module on feed use and poducion obained in he pevious ieaion. The make module is hen solved a his sage (consained equaion sysem) and he esuling poduce pices a Membe Sae level ansmied o he supply models fo he following ieaion. A he same ime, in beween ieaions, pemiums fo aciviies ae adjused if ceilings defined in he Common Make Oganisaions (CMOs) ae ovesho. Page 84 of 133

85 Figue 7. Link of modules in CAPRI Pemium Calculao Calculaion of agiculual pemiums depending on CMOs (ceilings, base aeas, ) Levels Supply Module 200 egional opimisaion models (Max. ag. income s.. esicions) Peennial sub-module (economeic esimaion) Aggegaion o MS level Supply Feed Demand Make Module Muli-commodiy spaial make model 11 egional wold aggegaes and EU-15 Membe Saes Pices Young Animal Makes Linked opimisaion models a Membe Sae level Souce: CAPRI Modelling Sysem 5.2 Module fo agiculual supply a egional level Basic ineacions beween aciviies in he supply model Thee ae wo souces fo ineacions beween aciviies in simulaion expeimens: he objecive funcion and consains. In he cuen vesion of CAPRI, he objecive funcion does no compise ine-aciviy ems, i.e. no maginal coss-cos effecs, so ha he majo ineplay is due o consains. The ineacion is bes undesood by looking a he fis ode condiions of a pogamming model including PMP ems: Equaion 81 Rev j = Cos j + ac j + bc Levl The lef hand side (Rev) shows he maginal evenues, which ae ypically equal o he fixed pices imes he fixed yields plus pemiums. The igh hand side shows he diffeen elemens of he maginal coss. Fisly, he vaiable o accouning coss (Cos) which ae fix as hey ae based on he Leonief assumpion. The em ( ac + bc Levl ) shows he maginal non-linea coss, hese maginal coss ae inceasing in he aciviy levels. The emaining em λ a i ij i= 1 capues he maginal coss linked o he use of exhaused esouces and he equal o he sum of he shadow pices λ muliplied he pe uni demand of ha aciviy j fo esouce he maix A being again based on Leonief echnology. The shadow values of binding esouces hence ae he dives linking he aciviies. A cenal ole in he CAPRI supply model plays he land-balance. Is shadow pice appeas as a cos in all cop aciviies including fodde poducing ones, so ha animals ae indiecly j j j + j m i= 1 λ a j i ij m Page 85 of 133

86 affeced as well. The second majo link is he availabiliy of no-makeable feeding suff, and finally, less impoan oganic feilise. The basic effecs ae bes discussed wih a simple example. Assume an incease of a pe ha pemium fo sof whea, all ohe hings unchanged. Wha will happen in he model? The inceased pemium will lead o an imbalance beween maginal evenues (= yield imes pices plus pemium) and maginal coss (=accouning coss, esouce use cos, non-linea coss). In ode o close he gap, as maginal evenues ae fixed, he aea unde sof whea will be inceased unil maginal coss of poducing sof whea have inceased o a poin whee hey ae again equal o maginal evenues. As he maginal coss linked o he non-linea cos funcion ( ac j + bc jlevl j ) ae inceasing in aciviy levels, inceasing he aea unde sof whea will hence educe ha gap. A he same ime, as he land balance mus be kep closed, ohe cop aciviies mus be educed. The non-linea cos funcion will fo hese cops now povoke a counevailing effec: educing he aciviy levels of compeing cops will lead o lowe coss fo hese cops. Wih maginal evenues (Rev) and accouning coss (Cos) fixed, ha will equie he shadow pice λ of he land balance o incease. Wha will be he impac on animal aciviies? Again, he shadow pice of he land balance will be cucial. Fo aciviies poducing non-makeable feed, maginal evenues ae no defined as pices imes yields, bu as inenal feed value imes pices. The inenal feed value is deemined as he subsiuion value of non-makeable fodde agains ohe feeding suff, and depends on hei nuien conen and fuhe feed esicions. Inceasing he shadow pice of land will hence eihe equie o decease ohe coss in poducing fodde o o incease he inenal maginal evenues. Saing i he ohe way aound a high shadow pice of land endes non-makeable fodde less compeiive compaed o ohe feeding suff. As feed coss ae howeve vey slighly inceasing in quaniies fed pe head, feed coss fo animals will incease. Bu as hei seveal equiemen consains involved, some feeding suff may incease and ohe decease. Clealy, he highe he shae of non-makeable fodde in he mix fo a ceain animal ype, he highe he effec. As maginal feed coss will incease, and maginal evenues fo he animal pocess ae no changing, ohe maginal coss in animal poducion need o be educed, and again he non-linea cos funcion will be he cucial pa, as he maginal cos elaed o i will decease if hed sizes dop. To summaize he supply esponse, inceasing pemiums fo a cop will hence incease he copping shae of ha cop, educe he shae of ohe cops, incease he shadow pice of land, lead o less fodde poducion, highe fodde coss and hus educed hed size of animals. Wha will be he impacs coveed by he make? The changes in hecaes will lead o inceased supply of he cop wih he highe pemium and less supply of all ohe cops a given pices, i.e. one upwad and many downwad shifs of he supply cuves. Equally, supply cuves fo animal poducs will shif downwads. On he ohe hand, some feed demand cuve will shif as well, some upwad, ohe downwad. These shifs will move he make module away fom he fome fixed poins whee make balances wee closed. Fo he cop poduc wih he inceased pemiums, inceased supply plus some changes in feed will mos pobably lead o lowe pices, wheeas pices of ohe cops will mos pobably incease. Tha will equie new adjusmens duing he nex ieaion whee he supply models ae solved, wih o a ceain exen counevailing effecs. Page 86 of 133

87 Table 18 Oveview on a egional aggegae pogamming model Cop Aciviies Animal Aciviies Feed Use Ne Tade Consains Objecive funcion + Pemium Acc.Coss vaiable cos funcion ems + Pemium Acc.Coss vaiable cos funcion ems - vaiable cos funcion ems fo feeding + Pice Oupu = 0 Aea - <= UAAR Se aside +/- = 0 Quoas - - <= Ref. Quaniy Feilize needs = 0 Feed equiemens Souce: CAPRI modelling sysem Deailed discussion of he equaions in he supply model Feed block The feed block ensues ha he equiemens of he animal pocesses ae me, and links hese o he makes and cop poducion decisions. The fis ype of equaion ensues ha equiemens (enegy, poein, lysine, minimum and maximum dy mae diffeen fibe equiemens fo uminans) ae me: = 0 Equaion 82 AREQ acc, eq DAYS aac FEDNG acc, feed REQCNT feed, feed eq The lef hand side capues he daily animal equiemens (AREQ) fo each egion animal aciviy acc and equiemen AREQ muliplied wih he days (DAYS) he animal is in he poducion pocess. Boh ae paamees fixed duing he soluion of he modelling sysem. The igh hand side ensues ha he equiemen conen of he acual feed mix epesened by he feeding (FEDNG) of ceain ype of feed o he animals muliplied wih he equiemen conen (REQCNT) in he egions coves hese nuiional demands. Fo enegy and poein, he less han is eplaced by an equal sign o ensue a moe plausible subsiuion inside he feed mix. Two addiional esicions ensue ha he conen of a ceain ype of feed in he mix measued in dy mae is in beween pe-defined uppe and lowe limis (MAXSHR, MINSHR): Page 87 of 133

88 Equaion 83 AREQ DAYS MAXSHR acc,"drma" FEDNG REQCNT acc,feed aac feed,"drma" acc,feed Equaion 84 AREQ DAYS MINSHR acc,"drma" FEDNG REQCNT acc,feed aac feed,"drma" acc,feed Toal feed use (FEDUSE) in a egion is defined as he feeding pe head muliplied wih he aciviy level (LEVL) fo he animal aciviies: Equaion 85 FEDUSE, = LEVL, FEDNG,, feed aac aac aac feed Land balances and se-aside esicions The model disinguishes aable and gassland and compises hus wo land balances: Equaion 86 LEVL " aab" = LEVL, aab aab Equaion 87 LEVL, " gas" = LEVL" gae" + LEVL" gai " Boh land balances mus be exhaused. Fo aable land, idling land no in se-aside (aciviy FALL) is an explici aciviy which closes he balance. Fo he gassland, he model disinguishes wo ypes wih diffeen yields (GRAE: gassland exensive, GRAI: gassland inensive) so ha idling gassland can be expessed of an aveage lowe poducion inensiy of gassland by changing he mix beween he wo inensiies. The obligaoy se-aside esicions inoduced by he McShay efom 1992 and valid unil he implemenaion of he Luxemboug compomise of June 2003 is an explici esicion in he model: Equaion 88 LEVL" ose" = LEVL aab 1 aab SETR SETR aab aab The somewha asonishing way he se-aside ae is inoduced mios he legislaion. A seaside ae of 10% does no imply ha fo one ha of he cop wih he se-aside obligaion 0.1 ha of land mus be pu ino se-aside, bu ha 0.9 ha of he cop mus be combined wih 0.1 ha of idling land. The equaion above implies ha non-food poducion on se-aside akes by assumpion place on volunay se-aside, endeing he analysis of model esuls easie wih no pacical consequences fo simulaion esuls. The equaion above is eplaced fo yeas whee he Luxemboug compomise of June 2003 is implemened by a Membe Sae, whee he level of obligaoy se-aside is fixed insead o he hisoical obligaions. Fo ceain yeas of he McShay efom, he oal shae of se-aside be i obligaoy o volunay on a lis of ceain cops was no allowed o exceed a ceain ceiling. Tha esicion is capued by he following equaion: Page 88 of 133

89 Equaion 89 LEVL " ose" + LEVL " vse" LEVL " nonf " LEVL aab SETR, aab aab MXSETA Feilising block The equaion below is discussed in he inpu allocaion chape in moe deail. Sufficien o say hee ha he fis line coves nuien cop needs minus biological fixaion of leguminosae, and mus be equal o puchases of inoganic feilise educed by ammonia losses in he case of N, he plan available pa of amospheic deposiion in he case of N, and he available nuiens in manue and losses. Equaion 90 cac Levl + NBal aac + Losses cac = NETTRD + AmDep Levl fnu biofix ( Fnu )( 1 NFac ) aac Fnu Fnu cac NFac (1 NH3Loss AmDep Cac aac Fnu, cac Anog Fnu, ) (1 NH3Loss Manue Fnu, NuFac ) fnu ( 1 NavFac ) fnu A second equaion ensues ha a ceain minimum shae of he cop need is coveed by inoganic feilise: cac cac fnu cac, fnu cac Fnu Equaion 91 Levl ( Fnu ) NuFac MINAN NETTRD Balancing equaions fo oupus Oupus poduced mus be sold if hey ae adable acoss egions o used inenally, as in he case of young animals o feed. Equaion 92 ac Levl ac OUTP ac,o o fodde o oyani o fodde = NETTRD + YANUSE + FEDUSE As descibed in he daa base chape he concep of he EAA equies a disincion beween young animals as inpus and oupus, whee only he ne ade is valued in he EAA on he oupu side. Consequenly, he emone expessed as demand fo young animals on he inpu side mus be mapped ino equivalen ne impo of young animals on he oupu side: Equaion 93 Levl ac I aac, yani = aac YANUSE oyani iyani In combinaion wih he sandad balancing equaion shown above, he NETTRD vaiable fo young animals on he oupu side becomes negaive if he YANUSE vaiable fo a ceain ype of young animals exceeds he poducion inside he egion. The objecive funcion The objecive funcion is spli up ino he linea pa, he one elaing o he quadaic cos funcion fo aciviies and he quadaic cos funcion elaing o he feed mix coss: Page 89 of 133

90 Equaion 94 OBJE = LINEAR + QUADRA + QUADRF The linea pa compises he evenues fom sales and he coss of puchases, minus he coss of allocaed inpus no explicily coveed by consains (i.e. all inpus wih he exempions of feilises, feed and young animals) plus pemiums: Equaion 95 LINEAR io io,io ac ac ( ac ac ) = NETTRD PRICE + LEVL PRME COST The quadaic cos funcion elaing o feed is defined as follows: Equaion 96 QUADRF = LEVL aac aac,feed ( a 1 acc,feed + baac,feedfedngaac,feed ) aac,feed 2 FEDNG The maginal feed coss pe animal incease hence linea wih he amoun of feed. Suga bee The cuen Common Make Oganisaion (CMO) fo suga egulaes Euopean suga bee supply wih a sysem of poducion quoas. Two diffeen quoas ae esablished subjec o diffeen pice guaanee (A and B quoas, qa and qb). Suga bees poduced beyond hose quoas (so called C bees) ae sold as suga on he wold make a pevailing pices. The CAPRI sysem feaues an expeced pofi maximisaion famewok ha caes fo yield unceainy as developed by Adenäue (2005). The idea behind his is ha obseved C suga poducions in he pas ae unlikely o be an oucome of compeiiveness a C bee pices ahe han being dependan on he fames incenive o fulfil hei quoa ighs even in case of a bad haves. Regional suga bee quoas ae defined based on a FADN analysis. Expeced pofi of suga bee poducion is hen epesened by: Equaion 97 SugbREV A = p NETTRD A B ( p p ) B C ( p p ) SUGB A A ( 1 CDFSugb( q ))( NETTRDSUGB q ) 2 S A + ( σ ) PDFSugb( q ) A+ B A+ B ( 1 CDFSugb( q ))( NETTRDSUGB q ) S 2 A+ B + ( σ ) PDFSugb( q ) Whee PDFSugb and CDFSugb ae he pobabiliy es. cumulaed densiy funcions of he NETTRD vaiable wih he sandad deviaion σ S. σ S is defined as NETTRD SUGB * VCOF, whee he lae is he egional coefficien of yield vaiaion esimaed fom FADN. p ABC ae Page 90 of 133

91 he pices fo he hee diffeen ypes of suga bee which ae exogenous and linked o he EU and wold make pices fo suga. The vaiable SugbREV subsiues fo he expession NETTRD io PRICE io (if io=sugb) in Equaion Calibaion of he egional pogamming models Since he vey fis CAPRI vesion, ideas based on Posiive Mahemaical Pogamming wee used o achieve pefec calibaion o obseved behaviou namely egional saisics on copping paen, heds and yield and daa base esuls as he inpu o feed disibuion. The basic idea is o inepe he obseved siuaion as a pofi maximising choice of he agen, assuming ha all consains and coefficiens ae coecly specified wih he exempion of coss o evenues no included in he model. Any diffeence beween he maginal evenues and he maginal coss found a he base yea siuaion is hen mapped ino a non-linea cos funcion, so ha maginal evenues and coss ae equal fo all aciviies. In ode o find he diffeence beween maginal coss and evenues in he model wihou he non-linea cos funcion, calibaion bounds aound he choice vaiables ae inoduced. The eade is now eminded ha maginal coss in a pogamming model wihou non-linea ems compise he accouning cos found in he objecive and oppouniy coss linked o binding esouces. The oppouniy coss in un ae a funcion of he accouning coss found in he objecive. I is heefoe no asonishing ha a model whee maginal evenues ae no equal o maginal evenues a obseved aciviy levels will mos pobably no poduce eliable esimaes of oppouniy coss. The CAPRI eam esponded o ha poblem by defining exogenously he oppouniy coss of wo majo esicions: fo he land balance and fo milk quoas. The emaining shadow pices mosly elae o he feed block, and ae less ciical as hey have a clea connecion o pices of makeable feed as ceeals which ae no subjec o he poblems discussed above Esimaing he supply esponse of he egional pogamming models The developmen, es and validaion of economeic appoaches o esimae supply esponses a he egional level in he conex of egional pogamming models fom an impoan ask fo he CAPRI eam. Up o now, hee is sill no fully saisfacoy soluion of he poblem, bu some of he appoaches ae discussed in hee. The wo possible compeios ae sandad dualiy based appoaches wih a following calibaion sep o esimaes based diecly on he Kuhn-Tucke condiions of he pogamming models. Boh may o may no equie a pioi infomaion o ovecome missing degees of feedom o educe second o highe momens of esimaed paamees. The dualiy based sysem esimaion appoach has he advanage o be well esablished. Less daa ae equied fo he esimaion, ypically pices and pemiums and poducion quaniies. Tha may be seen as advanage o educe he amoun of moe o less consuced infomaion eneing he esimaion, as inpu coefficiens. Howeve he calibaion pocess is cumbesome, and he esuling elasiciies in simulaion expeimens will diffe fom he esuls of he economeic analysis. The second appoach esimaing paamees using he Kuhn-Tucke-condiions of he model leads clealy o consisency beween he esimaion and simulaion famewok. Howeve fo a model wih as many choice vaiables as CAPRI ha saighfowad appoach may equie modificaions as well, e.g. by defining he oppouniy coss fom he feed equiemens exogenously. Page 91 of 133

92 5.3 Make module fo young animals The make module fo young animals ensues closed balances fo pigles, calves ec. a Euopean level. The individual egional models may sell o buy young animals in unlimied quaniies a fixed pices duing each ieaion. The make module mus hence geneae pices which lead o an equilibaion of egions wih excess demand and such wih excess supply of young animals. The fis ials wee based on a simple algoihm which was changing pices as a funcion of excess demand o supply a Euopean level. Howeve especially due o he high inedependencies inside he cale chain, hee ae impoan cosspice effecs, which could no be soed ou wih a simple appoach. Tha lef he eam wih wo possible compeios: a kind of muli-commodiy model fo young animals, whee he paamees would need o be esimaed fom simulaion expeimens wih he egional supply models, o a famewok building diecly on he egional pogamming models. The lae seemed moe pomising, despie he fac i is compuaionally infeasible o link all egional models simulaneously. Insead, he Inpu/Oupu coefficiens and all ohe coefficiens appeaing in he consains of he egional pogamming models ae aggegaed o Membe Sae level using aciviy levels as weighs. The esuling models ae hence sucually idenical o he egional models and compise a echnology equal o he weighed aveage ove all egions in ha Membe Saes. Due o he ypical aggegaion bias, hese Membe Sae models will howeve pefom diffeenly in a simulaion fom solving all egional models and hen aggegaing he esuls. Moe specifically, hey will even no epoduce he soluion obained fom he egional models a cuen pices. In ode o ovecome he aggegaion poblem, he Membe Sae models ae calibaed using ideas boowed fom Posiive Mahemaical Pogamming o he cuen esuls fom he egional models in any ieaion. In ode o do so, calibaion bounds ae inoduced aound he aggegaed esuls fo he aciviy levels and he feeding aciviies. Equally, a egionally weighed aveage fo shadow pices of gassland, aable land and he milk quoas is calculaed and added o he coss of he elaed poducion aciviies. Land balances and milk quoas ae hen emoved fom he model. The model is hen solved Afewads, hey ae sacked ogehe wih a se of new equaions epesening make cleaing condiions fo young animals. The shadow pices of hese consains a he opimal soluion hen define he pices fo young animals. 5.4 Make module fo agiculual oupus Oveview on he make model Wheeas he oulay of he supply module has no changed a lo since he CAPRI pojec ended in 1999, he make module was compleely evised. Even if seveal independen simulaion sysems fo agiculual wold makes ae available as OECD s AgLink, he FAPRI sysem a he Univesiy of Missoui o he WATSIM 26 sysem a Bonn Univesiy, i was sill consideed necessay o have an independen make module fo CAPRI. 26 In he beginning, he CAPRI make pa daw on he daa base fom he WATSIM modelling sysem. As he lae is no longe acive, he CAPRI make pa has become an independen wold ade model fo agiculual poducs. Page 92 of 133

93 The CAPRI make module can be chaaceised as a ecusive-dynamic, saic, paial, spaial, global equilibium model fo mos agiculual pimay and some seconday poducs, in oal abou 40 commodiies. The ecusive-dynamic aspec is cuenly only capued in a paial adjusmen appoach on he supply side. I is saic as sochasic effecs ae no coveed and paial as i excludes faco makes, non-agiculual poducs and some agiculual poducs as flowes. I is spaial as i includes bi-laeal ade flows and he elaed ade policy insumens beween he ade blocks in he model. The em paial equilibium model o muli-commodiy model sands fo a class of models wien in physical and valued ems. Demand and supply quaniies ae endogenous in ha model ype and diven by behavioual funcions depending on endogenous pices. Pices in diffeen egions ae linked via a pice ansmission funcion, which capues e.g. he effec of impo aiffs o expo subsidies. Pices in diffeen makes (beef mea and pok mea) in any one egion ae linked via coss-pice ems in he behavioual funcions. These models do no equie an objecive funcion; insead hei soluion is a fix poin o a squae sysem of equaions which compises he same numbe of endogenous vaiables as equaions. The CAPRI make module beaks down he wold ino 40 counies o couny aggegaes, each feauing sysems of supply, human consumpion, feed and pocessing funcions. The paamees of hese funcions ae deived fom elasiciies boowed fom ohe sudies and modelling sysems, and calibaed o pojeced quaniies and pices in he simulaion yea. The choice of flexible funcional foms (nomalised quadaic fo feed and pocessing demand as well as fo supply, Genealised Leonief Expendiue funcion fo human consumpion) and imposiion of esicions (homogeneiy of degee zeo in pices, symmey, coec cuvaue, addiiviy) ensue egulaiy as discussed below. Accodingly, he sysem allows fo he calculaion of welfae changes fo he diffeen agens epesened in he make model. Some of he 40 counies ae blocked o couny aggegaes wih a unifom bode poecion, and bilaeal ade flows ae modelled solely beween hese blocks. Such blocks ae he EU15, EU10, Medieanean and Mecosu counies and an aggegae of Bulgaia and Romania. All ohe counies o couny aggegaes ae idenical o a ade block in he model. Policy insumens in he make module include (bi)laeal aiffs and Poduce/Consume Subsidy Equivalen pice wedges (PSE/CSE). Taiff Rae Quoas (TRQs) ae inegaed in he modelling sysem, as ae inevenion sock changes and subsidised expos unde WTO commimen esicions fo he EU. Subsidies o agiculual poduces in he EU ae no coveed in he make model, bu inegaed in a vey deailed manne in he supply model. The EU ineacs via ade flows wih he emaining 17 egions in he model, bu each of he EU Membe Saes feaues an own sysem of behavioual funcions. The pices linkage beween he EU Membe Saes and he EU pool is howeve simply one of equal elaive changes, no a leas o ende he analysis of esuls moe easy. The make model in is cuen layou compises abou endogenous vaiables and he idenical numbe of equaions. Page 93 of 133

94 Table 19 Regional disaggegaion of he make module Couny/Couny aggegae Euopean Union 15, boken down ino Membe Saes (Luxemboug aggegaed wih Belgium) Euopean Union 10, boken down ino Membe Saes Code EU EU Componens wih own behavioual funcions AT BL DK DE EL ES FI FR IR IT NL PT SE UK CY CZ EE HU LT LV MT SI SK PL Ausia Belgium/Lux Denmak Gemany Geece Spain Finland Fance Iland Ialy Nehelands Pougal Sweden Unied Kingdom Cypus Czech Republic Esonia Hungay Lihuania Lavia Mala Slovenia Slovakia Poland In supply module? Noway NO Yes Bulgaia & Romania Medieanean counies Unied Saes of Ameica Canada Mexico Mecosu counies Res of Souh Ameica Ausalia & New Zealand BUR MED USA CAN MEX MER RSA ANZ China CHN 12. India IND 13. Japan JAP Leas developed counies ACP counies which ae no leas LDC ACP BG RO Yes No Bulgaia Romania Yes Yes Yes 27 A deailed descipion can be found in: C. Tien, B. Heny de Fahan, W.Biz (2001): Regionalisaion of he Res of he Wold Aggegae, CAPRI woking pape 01-01, available on he pojec web sie: hp:// Page 94 of 133

95 Couny/Couny aggegae developed Code Res of he wold ROW 17. Souce: CAPRI modelling sysem Componens wih own behavioual funcions In supply module? The appoach of he CAPRI make module Muli-commodiy models ae as aleady menioned above a wide-spead ype of agiculual seco models. Thee ae wo ypes of such models, wih a somewha diffeen hisoy. The fis ype could be labelled emplae models, and is fis example is Swopsim. Templae models use sucually idenical equaions fo each poduc and egion, so ha diffeences beween makes ae expessed in paamees. Typically, hese paamees ae eihe based on lieaue eseach, boowed fom ohe models o simply se by he eseache and ae fiendly emed as being synheic. Domesic policies in emplae models ae ypically expessed by pice wedges beween make and poduce especively consume pices, ofen using he PSE/CSE concep of he OECD. Wheeas emplae models applied in he beginning ahe simple funcional foms as consan elasiciy double-logs in Swopsim o WATSIM -, since some yeas flexible funcional foms ae in vogue, ofen combined wih a calibaion algoihm which ensues ha he paamee ses ae in line wih micoeconomic heoy. The flexible funcional foms combined wih he calibaion algoihm allow fo a se of paamees wih idenical poin elasiciies o any obseved heoy consisen behaviou which a he same ime ecoves quaniies a one poin of obseved pices and income. Ensuing ha paamees ae in line wih pofi especively uiliy maximisaion allows fo a welfae analysis of esuls. Even if using a diffeen mehodology (explici echnology, inclusion of faco makes ec.), i should be menioned ha Compuable Geneal Equilibium models ae emplae models as well in he sense ha hey use an idenical equaion sucue fo all poducs and egions. Equally, hey ae in line wih micoeconomic heoy. The second ype of model is olde and did emege fom economeically esimaed singlemake models linked ogehe he mos pominen example being he FAPRI modelling sysem. The obvious advanages of ha appoach ae fisly he flexibiliy o use any funcional elaion allowing fo a good fi ex-pos, secondly ha he economeically esimaed paamees ae ooed in obseved behaviou and hidly, ha he funcional fom used in simulaions is idenically o he one used in paamee esimaion. The downside is he fac ha paamees ae ypically no esimaed subjec o egulaiy condiions and will likely violae some condiions fom mico-heoy. Consequenly, hese models ae ypically no used fo welfae analysis. Besides FAPRI, ohe examples of such models ae AgLink a he OECD o he se of models emeging fom AgMemod. The CAPRI make module is a emplae model using flexible funcional foms. The eason is obvious: i is simply impossible o esimae he behavioual equaions fo abou 40 poducs and 40 counies o couny blocks wold wide wih he esouce available o he CAPRI eam. Insead, he emplae appoach ensues ha he same easoning is applied acoss he boad, and he flexible funcional foms allow fo capuing o a lage degee egion and poduc specificiies. As such, he esuls fom economeic analysis o even complee paamees ses fom ohe models could be mapped ino he CAPRI make model. Page 95 of 133

96 5.4.3 Behavioual equaions fo supply and feed demand Supply fo each agiculual oupu i and egion (EU Membe Saes o egional aggegae) is modelled by a supply funcion deived fom a nomalised quadaic pofi funcion via he envelope heoem. Supply depends on poduce pices ppi nomalised wih a pice index. The pice index elaes o all hose goods eihe inpus o oupus which ae no explicily modelled in he sysem: Equaion 98 supply = as + j bs j, ppi p j, index. Supply cuves fo he EU Membe Saes, Noway, Bulgaia and Romania ae calibaed in each ieaion o he las oupu pice veco used in he supply model and he aggegaed supply esuls a Membe Sae level, by shifing he consan ems as. The slope ems bs which capue own and coss-pice effecs ae se in line wih pofi maximisaion, as discussed below. The calibaion of he pice dependen paamees bs is discussed below. The sysem fo feed demand is sucued idenically. Howeve no poduce pices, bu aw poduc pices am1p deemined by he Amingon op level aggegao dive feed demand feed, combined wih changes in he supply of animal poducs weighed wih feed use facos w: Equaion 99 feed = af + j bf j, am1 p p index. j, i w i supply supply Feed use does hence popoionally incease if he supply of mea o milk is inceased, and pice changes dive subsiuion inside of he feed mix. I is planned o eplace ha sysem in he nea fuue by explici enegy and poein equiemen balances linked o enegy and poein shadow pices which will define hen feed incenive pices, as i is aleady ealised fo he fa and poein balances fo daiy poducs. As fo supply, feed demand cuves fo he EU Membe Saes, Noway, Bulgaia and Romania ae calibaed in each ieaion o he las oupu pice veco used in he supply model and he aggegaed feed demand a Membe Sae level, by shifing he consan ems af Behavioual equaions fo final demand The final demand funcions ae based on he following family of indiec uiliy funcions depending on consume pices cpi and pe capia income y 28 whee G and F ae funcions of degee zeo in pices (RYAN & WALES 1996) which will be defined below: Equaion 100 G U ( cpi y) = ( y F) Using Roy s ideniy, he following pe capia Mashallian demands PeCap ae deived: i cal i 28 Pe capia income and oal expendiue ae used as synonyms in he following as he demand is cove all goods and hus exhaus available income. Page 96 of 133

97 Gii Equaion 101 PeCap = F + ( y F ) i i G whee he F i and G i ae he fis deivaive of F and G vesus own pices. The funcion F is defined as follows: Equaion 102 F = d i i cpi i whee he d i epesen he consan ems of he Mashallian demands and can be inepeed as minimum commimen levels o consumpion quaniies independen of pices and income. The em in backes in he pe capia demands in Equaion 101 above hence capues he expendiue emaining afe he value F of pice and income independen commimens d had been subaced fom available income y. The funcion G, based on he Genealised Leonief fomulaion: Equaion 103 G = i j bd ij cpi i cpi wih he deivaive of G vesus he own pice is labelled Gi and defined as: Equaion 104 Gi j = i bd ij cpi i cpi Symmey is guaaneed by a symmeic bd maix descibing he pice dependen ems, coec cuvaue by non-negaive he off-diagonal elemens of bd, adding up is auomaically a( x) given, as Eule s Law fo a homogenous funcion of degee one a( x) = xi, leads i xi o: Equaion 105 i PeCap i cpi i j j Gicpii i = ( y F ) + dicpii G i, G = ( y F) + F = y G and homogeneiy is guaaneed by he funcional foms as well. The expendiue funcion can be deived fom he indiec uiliy funcions and gives: Equaion 106 G y = e( U, y) = F U ( cp y) The funcion is only semi-flexible as he non-negaiviy imposed on he off-diagonal elemens ensuing coec cuvaue will exclude Hicksian complemenaiy, a esicion no deemed impoan in he ligh of he poduc lis coveed. Human consumpion hcom is simply he sum of populaion pop muliplied wih he pe capia demands: Equaion 107 hcom = pop pecap Page 97 of 133

98 5.4.5 Behavioual equaions fo he pocessing indusy Pocessing demand fo oilseeds is modelled by using behavioual funcions deived fom a nomalised quadaic pofi funcion unde he assumpion of a fixed I/O elaion beween seeds, cakes and oils. Consequenly, he pocessing demand poc depends on pocessing magins pocmag which ae diffeences beween he value of he oupus (oil and cake) pe uni of oilseed pocessed and he value of he oilseed inpued: Equaion 108 poc = ac + j bc j, pocmag p index. whee he pocessing magin is defined fom he poduce pices ppi and cushing coefficiens deived fom obseved supply quaniies as: Equaion 109 pocmag seed, = ppi + ppi + ppi seed, seed cak, seed oil, j, supply supply supply supply bas seed cak, bas seed, bas seed oil, bas seed, Finally, oupu of oils and cakes supply depends on he pocessed quaniies poc of he oilseeds and he cushing coefficiens: Equaion 110 supply supply cake, oil, = = poc poc seed, seed, supply supply supply supply bas seed cak, bas seed, bas oil cak, bas seed, Special aenion is given o he pocessing sage of daiy poducs fo he EU Membe saes. Fis of all, balancing equaions fo fa and poein ensue ha he pocessed poducs use up exacly he amoun of fa and poein compised in he aw milk. The fa and poein conen con of aw milk and milk poducs mlk is based on saisical and engineeing infomaion, and kep consan a calibaed base yea levels. Equaion 111 supply " milk", con" milk", fp = supplymlk, conmlk, Poducion of pocessed daiy poducs is based on a nomalised quadaic funcion diven by he diffeence beween he daiy poduc s make pice and he value of is fa and poein conen. mlk fp Equaion 112 supplky j mlk, = am mlk, ( ) + bm ppi con ppi con ppi p mlk, j, j j,"fa" fa, j,"po" po, index, And lasly, pices of aw milk ae equal o is fa and poein conen valued wih fa and poein pices. Page 98 of 133

99 5.4.6 Tade flows and he Amingon assumpion The Amingon 29 assumpion dives he composiion of demand fom domesic sales and he diffeen impo oigins depending on pice elaions and hus deemines bilaeal ade flows. The Amingon assumpion is fequenly used in ha conex, and e.g. applied in mos Compuable Geneal Equilibium models o descibe he subsiuion beween domesic sales and impos. The undelying easoning is ha of a wo-sage demand sysem. A he uppe level, demand fo poducs as whea, pok ec. is deemined as a funcion of pices and income see above. These pices ae a weighed aveage of poducs fom diffeen egional oigins. A he lowe level, he composiion of demand pe poduc i in egion semming fom diffeen oigins 1 is deemined based on a CES uiliy funcion: Equaion 113 U i, = α δ 1M 1 ρ 1 whee U denoes uiliy in egion and fo poduc i due o consumpion of he impo quaniies M semming fom he diffeen oigins 1. If is equal 1, M denoes domesic sales. δ ae he so-called shae paamees, α is called he shif-paamee and ρ is a paamee deived fom he subsiuion elasiciy. Deiving he fis ode condiions fo uiliy maximisaion unde budge consains leads afe some e-aangemens o he following elaion beween impoed quaniies M: Equaion 114 M M 1 2 δ = δ 1 2 P 2 P 1 1, i ( 1+ ρ ) whee he em 1 ( 1+ρ ) denoes he subsiuion elasiciy. As seen fom he equaion, impos fom egion 1 will incease if is compeiiveness inceases eihe because of a lowe pice in 1 o a highe pice 2. The esuling changes in he composiions of impos incease wih he size of he elaed shae paamee δ 1 and wih he size of he subsiuion elasiciy. The CES uiliy funcion is ahe esicive as i has solely one paamee δ pe impo flow. The subsiuion elasiciy 1 ( 1+ρ ) is se exogenously. The δ paamees ae deemined when calibaing he model o known impo flows, wheeas α is used o mee he known quaniies in he calibaion poin. The model compises a wo sage Amingon sysem (see below): on he op level, he composiion of oal demand fom impos and domesic sales is deemined, as a funcion of he elaion beween he inenal make pice and he aveage impo pice. The lowe sage deemines he impo shaes fom diffeen oigins. The subsiuion elasiciy on he op level sage is smalle han fo he second one, i.e. we assume ha consumes will be less esponsive egading subsiuion beween domesic and impoed goods compaed o changes in beween impoed goods. The following able shows he subsiuion elasiciies used fo he diffeen poduc goups. Compaed o mos ohe sudies, we oped fo a ahe elasic subsiuion beween poducs fom diffeen oigins, as agiculual poducs ae geneally moe unifom hen aggegaed poduc goups, as hey can be found e.g. in CGE models., i 1 ρ, i 29 Amingon, Paul S. (1969), A Theoy of Demand fo Poducs Disinguished by Place of Poducion, IMF Saff Papes 16, pp Page 99 of 133

100 Table 20 Subsiuion elasiciies fo he Amingon CES uiliy aggegaos 30 Poduc (goup) Subsiuion elasiciy beween domesic sales and impos Subsiuion elasiciy beween impo seams Bue & Ceam, Mea 4 8 Cheese, fesh milk poducs 2 4 All ohe poducs Souce: Own calculaions Figue 8. Two-sage Amingon Sysem Am Demand (Am1) = Human consumpion + Feed Use + Pocessing ρ [ dp Am2 + dp ] ρ 1 ρ i, = sp 1 i, i,, w i, i,, DSales i, Domesic Sales ( DSales ) Impos (Am2) Am2 = sp dp 1 i, 2 i, i,, 1 Flows 1 ρ ρ i,, 1 Flows(R,R, XX)... Flows(R, R, XX) 1 n The Amingon appoach suffes fom wo impoan shocomings. Fis of all, a calibaion o a zeo seam is impossible so ha only obseved impo flows eac o policy changes while all ohes ae fixed a zeo level. Fo mos simulaion uns, ha shocoming should no be seious. I is planned o ovecome ha poblem by inoducing consan ems in he CES uiliy funcion, and consequenly he shae equaions. Secondly, he Amingon aggegao defines an uiliy a ggegae and no a physical quaniy. Tha second poblem is healed by e-coecing in he esul lising o physical quaniies. Lile empiical wok can be found egading he esimaion of he funcional paamees of Amingon sysems. Hence, subsiuion elasiciies wee chosen as o eflec poduc popeies as shown above. 30 A sensiiviy analysis on hose elasiciies is given in secion 5.7 Page 100 of 133

101 5.4.7 Make cleaing condiions All quaniies in he model ae measued in 1000 meic ons. The quaniy balances fis sae ha poducion mus be equal o domesic sales plus expo flows plus changes in inevenion socks: Equaion 115 supply = isch, dsales + flows 1, + 1 Fuhe on, impos and expos ae defined fom bilaeal ade flows as: Equaion 116 Equaion 117 impos i = flows, 1 expos i = flows, 1 1 1, Finally, he Amingon fis sage aggegae am1, shown in he diagam above, is equal o he domesic consumpion elemens feed, human consumpion and pocessing: i Equaion 118 am 1 = feed + hcon + poc, i Pice linkages All pices in model ae expessed as pe meic on. Impo pices impp 1 fom egion 1 ino egion of poduc i ae deemined fom make pices pmk aking ino accoun bilaeal ad valoem (aiffa) and specific (aiffs) aiffs minus expo subsidies expsub: Equaion 119 impp,, 1 = pmk, 1( + aiffa,, 1 100) + aiffs,, 1 expsub, 1 i i 1 i i i Bilaeal aiffs may be endogenous vaiables if hey ae deemined by a aiff ae quoa (TRQ), see below. Equally, expo subsidies ae endogenous vaiables. Poduce pices ae deived fom make pices using diec and indiec PSEs pice wedges, excep fo EU15, EU10 and Bulgaia and Romania. The eade is eminded ha fo he EU27, he supply model includes a ahe deailed descipion of he diffeen pemium schemes of he CAP, so ha he EU pemiums need no o be modelled as pice wedges in he make pa. Equaion 120 ppi PSEi, = pmk + PSEd + i The aveage pices of impos deived fom he Amingon second sage aggegae ae labelled am2p and defined as oal impo value divided by he Amingon second sage uiliy aggegae am2: Equaion 121 am2 p = 1 flows 1 am2 impp Similaly, he aveage pices fo goods consumed domesically am1p ae a weighed aveage of he domesic make pice pmk weighed wih domesic sales dsales and he Amingon second sage uiliiy aggegae am2 weighed wih he aveage impo pice am2p: 1 Page 101 of 133

102 Equaion 122 am1 p am2 = am2 p am1 + dsales pmk Consume pices cpi ae deived fom he composie good pice index am1p aken ino accoun policy inoduced pice wedges as diec and indiec consume subsidy equivalens plus a fix magin coveing anspo, pocessing and all ohe makeing coss: Equaion 123 cpi, = am1 p CSEd CSEi + cmg,. i Uni value expos ne of bode poecion ae defined as aveage make pices in he expo desinaion minus aiffs as: Equaion 124 uvae = i ( pmk 1 aiffs1, i ) ( 1 aiffa 1 ) 1 expos i flows The uni values expos ae used o define he pe uni expo subsidies expsub as shown in he equaion below. The paamee cexps is used o line up he make equaion wih he subsidies obseved ex-pos. Pe uni expo subsidies hence incease, if make pices pmk incease o expo uni values uvae dop, o if he shae of subsidised expos exps on oal expos incease. How he amoun of subsidised expos is deemined is discussed below. exps expsub i i exp, expos Equaion 125 = ( pmk uvae + c s ) The Amingon aggegao funcions ae aleady shown in he diagam above. The composiions inside of he Amingon composie goods can be deived fom fis ode condiions of uiliy maximisaion unde budge consains and lead o he following condiions: Equaion 126 am dsales dp + ϕ1. w, pmk i = dp am2 p Similaly, elaions beween impo shaes ae deemined by: i 1, i Equaion 127 flows flows i. 1 2 dp = dp 1 2 impp impp ϕ Endogenous policy insumens in he make model On he make side, he amoun of subsidised expos (exps) ae modelled by a sigmoid funcion, diven by he diffeence beween EU make (pmk) and adminisaive pice (padm), see equaion below. The sigmoid funcion used looks like: Equaion 128 Sigmoid( x) = exp ( min( x,0) ( 1+ exp( abs( x) ) ) ) whee x is eplaced by he expession shown below in he equaions. Page 102 of 133

103 The esponse was chosen as seep as echnically possible by seing a high value fo α, i.e. inevenion pices ae undecu solely if WTO commimen (QUTE) and he maximum quaniy of sock changes ae eached. α exps1 1 E β β PADM i Equaion 129 E QuE sigmoid ( pmk PADM ) = i The paamees β ae deemined based on obseved pice and quaniies of subsidised expos. In ode o ensue ha subsidised expos do no exceed acual expos, he following smooh appoximaion is used: Equaion exps + ( ) exps expos exps expos The elaion is shown in he figue below. = 2 1i, Figue 9. Modelling of subsidised expos by a logisic funcion PADM β PADM PMk 50% QUTE QUTE EXPS Puchases o inevenion socks inp depend on he pobabiliy of he cuen make pice pmk o undecu he adminisaive pice padm assuming a nomally disibued make pice wih sandad deviaion sddev and maximal amouns of puchases INTM: Equaion 131 = InM ef ( padm pmk ) sddev ) inp, A decease of he adminisaive pice o an incease of he make pice will hence decease puchases o inevenion socks. Releases fom inevenion socks ind depend on he pobabiliy of make pices pmk o undecu uni value expos uvae,,muliplied wih he cuen inevenion sock size being equal o saing size ink plus inevenion puchases inp: Equaion 132 = ( ink + inp ) ef (( uvae pmk + ) sddev ) ind, γ i Releases will hence incease if wold make pice inceases o he EU make pice dops, and if he size of he inevenion sock inceases. The paamees γ ae deemined fom expos daa on pices and inevenion sock levels. The change in inevenion socks ins eneing he make balance is hence he diffeence beween inevenion puchases inp and inevenion sock eleases ind: i i Page 103 of 133

104 Equaion 133 ins ind, = inp i Endogenous aiffs unde Taiff Rae Quoas Taiff Rae Quoas (TRQs) esablish a wo-ie aiff egime: as long as impo quaniies do no exceed he impo quoa, he low in-quoa aiff is applied. Quaniies above he quoa ae chaged wih he highe Mos-Favoued-Naion (MFN) aiff. CAPRI disinguishes wo ypes of TRQs: such open o all ading panes, and bi-laeally allocaed TRQs. Equally, as fo all aiffs, TRQs may define ad valoem and/o specific aiffs. A make unde a TRQ mechanism may be in one of he following egimes: Quoa undefill: he in-quoa aiff is applied. The willingness o pay of he consumes is equal o he bode pice plus he in-quoa aiff. Quoa exacly filled: he in-quoa aiff is applied. The willingness o pay of consumes and hus he pice paid is somewhee beween he bode plus he in-quoa aiff and he bode pice plus he MFN aiff. The diffeence beween he pice in he make and he bode pice plus he in-quoa aiff esablishes a quoa en. Depending on popey ighs on he quoa and he allocaion mechanism, he quoa en is shaed in diffeen poions by he poduces, impoing agencies, he domesic makeing chain o he adminisaion. Typically, he quoa en can neihe be obseved no is hei knowledge abou disibuion of he en. Quoa ovefill: he highe MFN-aiff is applied. The quoa en is equal o he diffeence beween he MFN and he in-quoa aiff. Again, how he quoa en is disibued o agens is ypically no known. Thee ae a couple of fuhe complicaions, linked o spaial and commodiy aggegaion poblems. In many cases, TRQs ae defined fo vey specific daa qualiies, which ae moe dis-aggegaed as he poduc definiion of he model. TRQs fo beef may efe e.g. o specific cus, aces o even feeding pacises. Tha ypically leads o a siuaion whee boh impos coveed and no coveed by a TRQ mechanism ae aggegaed in he daa base of he model. Consequenly, i is no clea which egime govens he make. Fuhe on, TRQs may be defined fo individual counies whee he model woks on a couny block. Besides he poblem of defining he egime ex-pos, he elaion beween he impo quaniy and he aiff is no diffeeniable bu kinked. Theefoe, again a sigmoid funcion (Figue 9) is applied in he CAPRI make pa Oveview on a egional module inside he make model The esuling layou of a make fo a couny (aggegae) in he make module is shown in he following diagam. Due o he Amingon assumpion, poduc makes fo diffeen egions ae linked by impo seams and impo pices if obseved in he base yea. Accodingly, no unifom wold make pice is found in he sysem. Page 104 of 133

105 Figue 10. Gaphical pesenaion fo one egion of a spaial make sysem Regional Pices Supply Expo subsidies? Inevenion sales Domesic Sales Expo seams Amingon Sage 2 Aggegae Impo Seams and pices Amingon Sage 1 aggegae Pocessing Human consumpion Feed Use Souce: CAPRI modelling sysem Basic ineacion inside he make module duing simulaions As wih he supply module, he main difficuly in undesanding model eacions is based on he simulaneiy of changes occuing afe a shock o he model. Coss-pice effecs and ade elaions inelink basically all poduc makes fo all egions. Wheeas in he supply model, ineacions beween poducs ae mosly based on explici epesenaion of echnology (land balances, feed esicions), such ineacions ae capued in muli-commodiy models in he paamees of he behavioual funcions. Even if he following naaive is simplifying and descibing eacions as if hey would appea in a kind of naual sequence whee hey ae appea simulaneously in he model, we will neveheless analyse he effec of an inceased supply a given pices fo one poduc and one egion. Such a shif could e.g. esul fom he inoducion of a subsidy fo oupuing ha poduc. The inceased supply will lead o imbalances in he make cleaing equaion fo ha poduc and ha egion. These imbalances can only be equilibaed again if supply and demand adjus, which equies pice changes. In ou example, he pice in ha egion will have o dop o educe supply. Tha dop will simulae feed demand, and o a lesse exen, human consumpion. The smalle effec on human consumpion has wo easons: fisly, pice elasiciies fo feed demand ae ypically highe and secondly, consume pices ae linked wih ahe high magins o fam gae pices. The esuling lowe pice a fam gae inceases inenaional compeiiveness. Due o he Amingon mechanism, consumes aound he wold will now incease he shae of ha egion in hei consumpion of ha poduc, and lowe hei demand fom ohe oigins. Tha will pu pice pessue in all ohe egional makes. The pessue will be he highe he highe he impo shae of he egion wih he exogenous incease of supply on he demand of ha poduc. The esuling pice pessue will in un educe supply and simulae demand Page 105 of 133

106 and feed eveywhee, and, wih educed pices, offse paially he inceased compeiiveness of he egion whee he shock was inoduced. Simulaneously, impacs on make fo ohes poducs will occu. Depending on he size of he coss pice elasiciies, demand fo ohe poducs will dop wih falling pices fo a subsiue. A he same ime, educed pices will simulae supply of ohe poducs. The esuling imbalances will hence foce downwads pice adjusmens in ohe makes as well. 5.5 Paamee calibaion and souces fo he behavioual equaions Calibaion of he sysem of supply funcions The supply equaion was aleady inoduced in Equaion 98. The maix bs is equal o he Hessian maix of second deivaives of he nomalised pofi funcion o nomalised pices and mus hence be symmeic by definiion. As bs is equal o he fis deivaive of he supply funcion agains nomalised pices, he supply elasiciies a he calibaion poin ae defined as: Equaion 134 ε j= bs j ppi p j, index. supply i Homogeneiy of supply funcions of degee zeo is given due o he nomalisaion wih a pice index: if all pices and he pice index ae aised by he same pecenage, he supply quaniy does no change. Remains he quesion of cuvaue, which is guaaneed if bs is posiive definie, ensued by a Cholesky decomposiion duing he calibaion pocess. The cuvaue ensues ha maginal pofis ae inceased if one o seveal of he pices ae inceased, and is one of popeies of a pofi funcion deived fom mico-heoy. The calibaion seaches fo minimal squaed deviaions beween he consisen elasiciies and given ones. The uncalibaed elasiciies fo he non-eu egions ae aken fom he Wold Food Model of he FAO, saus Missing own-supply elasiciies ae se o 0.5. I is assumed ha he elasiciy o all emaining poducs including he inpus is -0.25, if no given. Thee ae some fuhe esicions inoduced: Absolue elasiciies ae no allowed o be lage han 10. Reacions in beween ceeals and beween ceeals and meas mus be subsiuive Calibaion of he final demand sysems Accoding o he concep of he Supply Uilizaion Accouns, all pocessing demand by he food indusy is couned as human consumpion. Equally, impos of food poducs ae econveed in pimay poduc equivalens. Human consumpion of a pimay poduc in he make model does hence include all pocessed food poducs deived fom i as pasa, muesl bead ec. ooing in bead. As discussed above, he demand sysem discussed above is homogenous of degee zeo in pices and income, and symmeic if bd is symmeic. The somewha moe cumbesome poof ha uiliy is deceasing in pices and inceasing in income as long as he maix bd has only Page 106 of 133

107 posiive off-diagonal elemens is lef ou in hee. The down-side of he esicion on he sign of he elemens of Pbd is ha fac ha he funcion hen allows fo Hicksian subsiues, only. The funcion is hen clealy no longe flexible which may be seen as a disadvanage in economeic applicaions. Given he poduc lis of he CAPRI make model, he limiaion was even judged as a safeguad agains cuious pice effecs 31 as complemenaiies fo he compensaed demands ae no easy o ague fo. The symmey and non-negaiviy condiions ae imposed duing he calibaion of he paamees o he pice and income elasiciies boowed fom he WFM. The calibaion necessiaes deivaives of Mashallian demands vesus pices and income fom he expendiue sysem above which ae deemined as follows: PeCap y i Gii = G Equaion 135 PeCap cpi whee : Gij j j i Gii = cpi Giji = G j = 1 2, j bd GiiGi 2 G j bd j j ( y F ) cpi cpi i i j j i j The ems fo he own pice effecs ae somewha moe complicaed, and heefoe deemined indiecly via he homogeneiy condiion fo elasiciies duing calibaion. The objecive funcion minimizes squaed diffeences beween given and consisen elasiciies, simulaneously fo he base yea and he las yea of he pojecion peiod. The paamees di ae chosen so ha he funcions calibae o quaniies and pices in he calibaion poin. 5.6 Linking he diffeen modules he pice mechanism As hined a above seveal imes, he make modules and he egional pogamming models ineac wih each ohe in an ieaive way. Basically, he make modules delive pices o he supply module, and he supply module infomaion o updae he supply and feed demand esponse fom he make models. Fo he make module fo agiculual oupus, he updae of he supply and feed demand esponse is pu o wok by changing he consan ems in he behavioual equaions such ha supply and demand quaniies simulaed a pices used duing he las ieaion in he supply module would be idenical o he quaniies obained fom he make module a ha pices. Howeve he poin elasiciies of he aggegaed esponse fom he supply module diffe fom he ones in he make modules which necessiae an ieaive updae. In ode o speed up convegence, he supply side uses a weighed aveage of pices of he las ieaions. 31 As an alenaive, a nomalized quadaic expendiue sysem was esed. Accoding o he family of indiec uiliy funcions discussed above, he funcion G is hen eplaced by a fom quadaic in nomalized pices. Howeve a Cholesky decomposiion is hen necessay o ensue coec cuvaue duing he calibaion pocess, which endes he soluion moe cumbesome. An advanage of he NQ sysem is he fac ha i allows fomally fo complemenaiy in he Hicksian effecs. In pacice, ha would mean ha he Mashallian elasiciies ceaed by he calibaion of he NQ sysem have o be caefully checked fo such complemenaiies o ensue a plausible behaviou of he demand sysem in simulaions. Page 107 of 133

108 The fis vesion of CAPRI fixed supply of EU Membe Saes in he make module duing ieaions. I uned ou howeve ha convegence is achieved fase if supply is pice esponsive even wih diffeing poin elasiciies. One of he opions discussed is o geneae a se of pice elasiciies fom he egional pogamming models and o calibae he paamees of he make module o i. Howeve given he lage amoun of commodiies and egional o even fam ype models, hese sensiiviy analysis would ake quie some ime. The ineacion beween he egional pogamming models and he young animal module was aleady explained above. Basically, i is again an ieaive updae of paamees in a moe aggegae model; howeve he young animal module compises models a Membe Sae level which ae sucually idenical o he egional models. The updae hus equies boh he definiion of a weighed aveage of he I/O coefficiens as well as he applicaion of ideas boowed fom Posiive Mahemaical Pogamming o achieve a poin calibaion. As fo makeable oupus, pices fo young animals used in any ieaion ae a weighed aveage of pevious ieaions. 5.7 Sensiiviy of he CAPRI model o he Amingon subsiuion elasiciies A convenional sensiiviy analysis consiss o un he model using iniial Amingon elasiciies o obain he baseline, hen o eun i unde vaious elasiciy values, all ohe hings held consan, and finally o compae he efeence and simulaion esuls. In ou sensiiviy sudy, he implemenaion of his ype of analysis shows vey small numeical vaiaions on evey vaiable level a less han pecen. This is he eason why we chose o associae exogenous shocks o he sensiiviy analysis. To pefom he sensiiviy analysis, we fis inoduce diffeen ses of Amingon subsiuion elasiciies in he model. Then, we inoduce an exogenous shock by changing, fo example, he policy paamees o he shif facos in he supply equaions. Finally, we compae he eacions of endogenous vaiables (pice, poducion, domesic sales, impos, expos) fo diffeen ses of elasiciies as show in Figue 11. Thee exogenous shocks, associaed o he sensiiviy analysis, ae hus implemened: (i) a 20% decease in supply, (ii) a 10% decease in subsidized expos, (iii) an incease in aiff ae quoas. Fo each shock, he simulaion elaed o he iniial Amingon elasiciies, i.e., scenaio 3, is used as he baseline. Is esuls ae compaed o hose of he sensiiviy uns. The ses of subsiuion elasiciies ae obained by shifing he iniial value of hese elasiciies o moe o less 70 pecen. The use of he same pecenage change - beween he baseline and he ohe sensiiviy uns - allows o evaluae he degee of symmey in he sensiiviy. Lowe values fo elasiciies imply a decease of pefeence and hus a geae difficuly in subsiuing beween demand oigins, wheeas highe values fo elasiciies imply an incease of pefeence and, hus, a geae ease in subsiuing beween demand oigins. Page 108 of 133

109 Figue 11. Illusaion of Sensiiviy Analysis on he CAPRI Make Module Subsiuion elasiciies as pecen of iniial value: Scenaio σ1 a σ2 b 1-70% -70% 2-35% -35% 3 0% 0% 4 +35% +35% 5 +70% +70% σ: decease Pefeence: decease σ: incease Pefeence: incease Exogenous Shocks: 20% decease in whea supply in EU a σ1: Subsiuion elasiciies beween domesic sales and impos b σ2: Subsiuion elasiciies beween impo seams Repos: The simulaion esuls unde vaious subsiuion elasiciies (pecen deviaion o he baseline) Scenaio Pices Poducion Domesic sales Expos Impos Ec. Souce: Own calculaions To keep he discussion eadable, we only pesen he esuls associaed o he lage vaiaion of he elasiciies (i.e. scenaios 1 and 5: ± 70%). High and low values ae specified o epesen 70 pecen moe o less han he iniial values used in he baseline. We sic ouselves also o esuls fo he Euopean Union (EU) and o some key commodiies which pesen a vaiaion highe han 0.1 pe housand. Howeve we poin ou impoan findings fo ohe makes whee necessay. As one would expec, he esuls of sensiiviy depend songly on he exogenous shock associae o he sensiiviy analysis. When pefoming a 20% decease in supply (Table 21), changes in poducion levels ae insensiive o he Amingon elasiciies, excep fo ohe mea and suga poducions which show a change exceeding 2%. The same obsevaion applies o changes in poduce and consume pices. All he pice changes show lile eacions wih less han o aound 2% in eihe diecion, excep fo change in he poduce pices of ohe mea and suga which incease o 3% and 10% especively. Like changes in poducion and pices, changes in domesic sales ae pacically invaian wih espec o changes in he Amingon elasiciies, excep fo change in he ice domesic sales which shows a eacion exceeding 11%. Page 109 of 133

110 Table 21 Deviaion of he simulaion esuls o he baseline unde high and low subsiuion elasiciies wih a 20% decease in supply Elasiciy of subsiuion a WHEAT BARLY SUGA RICE MEAO Poduce pice Low 1,5% 0,3% 10,5% 1,6% 3,2% High -0,6% -0,1% -4,1% -0,9% -1,7% Consume pice Low 0,2% 0,0% 1,9% 0,2% 1,4% High -0,1% 0,0% -0,8% -0,1% -0,8% Poducion Low 1,1% 0,1% 2,7% 0,8% 2,3% High -0,4% 0,0% -1,1% -0,4% -1,2% Domesic sales Low 0,4% -0,2% 0,0% 11,4% 2,1% High -1,1% 0,0% 0,0% -5,5% -1,2% Expos Low 7,0% 2,8% 12,7% -11,8% 0,3% High 4,8% -0,4% -5,1% 4,3% -0,1% Impos Low -24,5% -7,9% 0,6% -10,1% -12,2% Souce: CAPRI esuls High 25,1% 0,2% -0,3% 4,8% 6,7% a Low elasiciy of subsiuion: -70% of he iniial value High elasiciy of subsiuion: +70% of he iniial value The same sensiiviy esuls, peaining o changes in pices, poducion and domesic sales ae obained wih he wo ohe exogenous shocks which consis in a 10% decease in subsidized expos and an incease in aiff ae quoas. As shown in Table 22 and Table 23, change in all hese vaiables do no exceed 2% excep fo changes in he domesic sales of skim milk powde and ice which vay by 5 o 7% unde a 10% decease in subsidized expos, and changes in poduce and consume pices of cheese unde an incease in aiff ae quoas. Page 110 of 133

111 Table 22 Deviaion of he simulaion esuls o he baseline unde high and low subsiuion elasiciies wih a 10% decease in subsidized expos Elasiciy of subsiuio n WHEAT BARLY MILS CHES RICE BEFM Poduce pice Low 0,9% 0,6% 0,2% 0,3% 0,8% 1,6% High -0,7% -0,5% 0,1% -2,4% -0,3% -1,4% Consume pice Low 0,1% 0,1% -0,2% 0,2% 0,1% 0,8% High -0,1% 0,0% 0,1% -1,5% -0,1% -0,7% Poducion Low 0,5% 0,4% 0,1% 0,2% 0,3% 0,9% High -0,4% -0,3% -0,1% -0,7% -0,1% -0,8% Domesic sales Low -1,2% -1,0% -7,1% -0,2% -5,4% -0,6% High 1,3% 0,8% 6,7% -0,5% 6,3% 1,0% Expos Low 14,4% 18,5% 20,1% 7,2% 22,4% 14,6% High -14,4% -16,4% -18,7% -4,1% -18,7% -16,5% Impos Low 21,4% 16,1% 14,0% 13,8% 8,1% 10,2% Souce: CAPRI esuls High -29,4% -24,4% -13,1% -0,9% -9,3% -15,4% a Low elasiciy of subsiuion: -70% of he iniial value High elasiciy of subsiuion: +70% of he iniial value As expeced, he main changes in vaiables ha ae affeced by he Amingon elasiciies ae hose of ade flows. Independenly of he shock and make ypes, he lages changes concen impo and expo quaniies and, hence, ae he moe sensiive o elasiciies. Expo changes ae sensiive o changes in he Amingon elasiciies. Of couse, impo changes ae even moe affeced. The lages effecs on ade changes ae obseved fo mos commodiies whose ade is lage and chaaceised by high iniial Amingon elasiciies such as in he case of ceeals. As shown in Table 22, he lages effecs on ade changes ae obseved when pefoming a 10% decease in subsidized expos. Fo some makes, such as he whea make, he effec on impo changes can each 30%. Mos of he lage effecs on expo changes ae found in makes chaaceized by lile ade such as he ice make. Unde his shock, makes wih highe elasiciies show lowe effecs on expo and impo changes and lage effecs on domesic sales changes, and convesely fo makes wih lowe elasiciies, lage effecs on expo and impo changes and lowe effecs on domesic sales changes. This means ha, unde a shock of 10% decease in subsidized expos, highe values of Amingon elasiciies imply an incease of pefeence in domesic sales agains impos, which esuls in a decease in expos. Page 111 of 133

112 Table 23 Deviaion of he simulaion esuls o he baseline unde high and low subsiuion elasiciies wih an incease in aiff ae quoas a Elasiciy of subsiuion b BARLY MILS CHES BTCR SUGA Poduce pice Low 0,0% 0,1% -5,3% 2,8% 0,0% High 0,0% 0,0% 0,8% -1,5% 0,0% Consume pice Low 0,0% -0,6% -3,2% 1,6% 0,0% High 0,0% 0,1% 0,5% -0,9% 0,0% Poducion Low 0,0% 0,1% -0,6% 0,5% 0,0% High 0,0% -0,1% 0,1% -0,2% 0,0% Domesic sales Low -0,1% 0,5% -1,1% 0,6% 0,0% High 0,0% -0,1% 0,2% -0,2% 0,0% Expos Low 0,1% -0,7% 6,7% 19,9% 0,0% High 0,0% 0,0% -0,5% -23,7% 0,0% Impos Low 0,8% -0,4% -1,2% -15,7% 0,0% Souce: CAPRI esuls High -0,3% 0,1% 0,3% 5,0% 0,0% a The pecenage of he incease in he TRQ applied fo each commodiy depends on he impos and he aiff ae quoas in he base yeas b Low elasiciy of subsiuion: -70% of he iniial value High elasiciy of subsiuion: +70% of he iniial value As shown in Table 23, when pefoming an incease on aiff ae quoas (TRQ), effecs on he changes in mos of he vaiables ae no sensiive o he Amingon elasiciies. I means ha effecs of he TRQ on model oucomes unde diffeen ses of Amingon elasiciies ae maginal. Wih espec o symmey in he opposie change in Amingon elasiciies, we obseve ha he pecenages of change in vaiable levels vesus hei iniial values do no show much symmey. Fo mos of he vaiables and commodiies, changes ae lage in he lowe subsiuion elasiciies (-70%) han in he highe subsiuion elasiciies (+70%), as expeced since he elaive change in paamees is lage in he fome han in he lae. Excepions appea on changes in impos unde he assumpion of a decease in subsidized expos, which eac convesely, i.e. changes ae less in he lowe subsiuion elasiciies han in he highe subsiuion elasiciies (Table 22). In sum, all he effecs on he changes in vaiable levels emain low compaed o he changes applied on he Amingon elasiciies (± 70%). The model oucomes ae hus compaaively insensiive o he acual magniude of he Amingon elasiciies. Page 112 of 133

113 6 Fam Type Pogamming Model: a FADN-based appoach 6.1 The CAPRI fam ype appoach The main aims linked o he inoducion of fam ypes in he sysem is o amelioae he analysis of agiculual policies linked o sucual vaiables as fam size o socking densiy, impove he eliabiliy of envionmenal indicaos and allow fo income analysis a fam ype level. In ohe wods, he inoducion of daa fo single fams fom he FADN daa base educes he aggegaion bias of he model a egional level. The fam goup models could be classified by a numbe of indicaos like he economic impoance (fams wih high agiculual income agains hose wih lowe ones), envionmenal impac (classic agains ecological faming) and many ohes. The sandad gouping in FADN is based on specialisaion (e.g. specialised in pig poducion), which migh be suppoed on he following agumens: Fis of all, he esuling goups ae aleady clealy defined accoding o official Euopean documens (Commission Decision 2003/369/EC) and esuls obained can be easily compaed o ohe sudies, secondly, he gouping is based on sandad goss magins, educing he sochasic impac of weahe o pice changes on he gouping fo single yeas, and as a hid poin, i can be agued ha envionmenal impacs ae ofen linked o fam specialisaion. Bu even wih he fam ypology accoding o Euopean sandads applied, a numbe of issues need o be addessed fo is applicaion in CAPRI: (1) Numbe of fam goups defined fo each egion. Clealy, he amoun of deail inceases wih he numbe of fam ypes, in line wih compuing ime and managemen coss o handle he addiional infomaion. Due o such esouce and echnical esicions, in CAPRI i was decided o choose no moe han five fam ypes (he mos epesenaive) plus a mixed emaining goup epesening all ohe fams fo he modelling sysem (and allowing consisen aggegaion of egional daa). (2) Level of ypology: Fo simpliciy and a bee compaison o FADN, we use he same hee digi ypology as defined in FADN. Consequenly, 50 diffeen ypes of specialisaion can be found in CAPRI (see able 24). The following diagam shows he elaion beween he FADN daa base and he elemens of he CAPRI daa pocesso. Page 113 of 133

114 Figue 12. Inegaion of fam ypes in he CAPRI daa base CoCo REGIO CAPREG 5 Types pe egion Consisency (Levels, Oupu) Assignmen o NUTS FADN ecods CAPRI Souce: Own calculaions In a fis inegaion sep, ex-pos daa on NUTS 2 level fom he CAPRI daa on aciviy levels and oupu wee seleced fo abou 50 poducion aciviies. Fuhe on, an exacion pogam povided he necessay daa fom he FADN daa base. The second inegaion sep consised in a non-linea opimisaion pogam which ensued maching aciviy levels (hecaes, hed sizes) and poducion quaniies beween CAPRI and FADN. Pa of he poblem a his sage elaed o he diffeen egional beakdown of CAPRI and FADN: wheeas he CAPRI daa base efes o adminisaive NUTS egions, he FADN daa base has is own se of FADN-egions. In ode o incease he numbe of fams available pe ype and egion and, a he same ime, pevening poblems wih confidenialiy limis, he algoihm used in CAPRI disibued he aggegaion weighs fo each fam ove seveal FADN-egions. A specific fam in he newok may easily epesen fams no only in he FADN-egion whee he fam is siuaed bu in ohe egions as well (wihin he boundaies of a NUTS 2 egion). In ode o mach he CAPRI daa base which is in majo elemens deived fom he REGIO daa base a EUROSTAT i was necessay o change he aggegaion weighs and aciviy daa of single FADN ecods. Minimising squaed diffeences ensued ha he changes wee no bigge hen necessay. Afe ha sep, he single fam ecods wee aggegaed o specialised fams pe egion (see able 24) and he five mos fequen fam ypes wee seleced, wih he fequency elaing o he aggegaion weighs. This sep is necessay only once fo a given base yea. Afewads, an addiional algoihm ensues ha inpu use aggegaed ove he fam ypes maches he inpu use a NUTS 2 level. These algoihms ae inegaed in he so-called egionalisaion sep in CAPRI, which combines he COCO daa base (wih is ime seies a naional level) wih infomaion fom REGIO and ohe souces a egional level. Page 114 of 133

115 Table 24 Fam ypes found in he sysem 131 Specialis COP (ohe han ice) 132 Specialis ice 133 COP and ice combined 141 Specialis oo cops 142 Ceeals and oo cops combined 143 Specialis field vegeables 144 Vaious field cops 201 Specialis make gaden vegeables 202 Specialis flowes and onamenals 203 Geneal make gaden copping 311 Qualiy wine 312 Wine ohe han qualiy 313 Qualiy & ohe wine combined 314 Vineyads fo vaious ypes of poducion 321 Specialis fui (ohe han cius) 322 Cius fuis 323 Fuis & cius fuis combined 330 Olives 340 Vaious pemanen cops combined 411 Milk 412 Milk & cale eaing 421 Cale eaing 422 Cale faening 431 Daiying wih eaing & faening 432 Reaing & faening wih daiying 441 Sheep 442 Sheep & cale combined 443 Goas 444 Vaious gazing livesock 501 Specialis pigs 502 Specialis pouly 503 Vaious ganiues combined 601 Make gadening & pemanen cops 602 Field cops & make gadening 603 Field cops & vineyads 604 Field cops & pemanen cops 605 Mixed copping-mainly field cops 606 Mixed copping-mainly make gadening o pemanen cops 711 Mixed livesock-mainly daiying 712 Mixed livesock-mainly non-daiy gazing 721 Mixed livesock-ganivoes & daiying 722 Mixed livesock-ganivoes & non-daiy gazing 723 Mixed livesock-ganivoes wih vaious livesock 811 Field cops & daiying 812 Daiying & field cops 813 Field cops & non-daiy gazing 814 Non-daiy gazing & field cops 821 Field cops & ganivoes 822 Pemanen cops & gazing livesock 823 Vaious mixed cops and livesock 999 Res Souce: FADN (hp://euopa.eu.in/comm/agiculue/ica/index_en.cfm). Page 115 of 133

116 In he CAPRI modelling sysem, fam ypes ae eaed echnically as a fuhe beakdown inside NUTS 2 egions (pseudo-egions): he aciviy levels in each fam ype feaue own inpu and oupu coefficiens and ae independenly opimised fo maximal pofis (emplae appoach of he CAPRI supply module). Afe a model un, he fam ype esuls ae aggegaed o NUTS 2, Membe Sae and EU level. I should be noed ha he elaion beween NUTS 2 and Membe Saes is geogaphical; he disaggegaion hus povides localised effecs in space. Fam ype daa howeve canno be linked o specific locaions in he NUTS 2 egions, even if hey beak down consisenly oupu, in physical and valued ems, aciviy levels, and economic and envionmenal indicaos. An impovemen in ha espec would equie a complee link wih a Geogaphical Infomaion Sysem plus inensive economic analysis o ceae mapping algoihms beween spaial specifics (soil ypes, local climae, slope, aliude..), poducion pogam and fam specialisaion. Some wok in his diecion is being undeaken in CAPRI-Dynaspa and, possibly, in SEAMLESS. Figue 13 shows he coding scheme. Membe Saes ae labelled wih wo chaace codes accoding o EUROSTAT sandads (AT fo Ausia, BL fo Belgium and Luxemboug, DK fo Denmak, DE fo Gemany,...). Regions inside a Membe Sae eceive a 3-digi code (fis posiion: NUTS 1 level, second: NUTS 2 level, hid: NUS III level) following he EUROSTAT NUTS classificaion scheme. The fam ypes ae labelled wih alphanumeical hee-digis code as well, whee he 000 efes o he egional level. Figue 13. Aggegaion fom fam ypes o NUTS 2 and Membe Sae Membe Sae MS MS MS MS MS MS MS MS MS MS MS Souce: CAPRI Modelling Sysem Moeove he sysem aggegaes acoss egions all fams of he same specialisaion, allowing fo he analysis of effecs fo fams of a ceain specialisaion acoss Euope. In ode o add Page 116 of 133

117 addiional layes of infomaion, he specialised fam ypes can be also aggegaed, as shown in able 25. Table 25 Aggegaed fam ypes used fo impac assessmen Code Descipion Fam ype included A10 Specialis COP (ohe han ice) o vaious field cops 133,144 A13 Specialis Rice o Rice & COP 132,133 A14 Roo cops 141,142 A23 Pemanen cops & vegeables 143,201,202,203,311,312,313,31 4,321,322,323,330,340 A41 Daiy 411,412,431 A42 Cale faening & aiing 421,422,432 A44 Sheep & goas 441,442,443, Specialis pigs 501 A52 Specialis pouly 502,503 A60 Field cops divesified 601,602,603,604,605,606 A70 Livesock divesified 711,712,721,722,723 A80 Livesock & cops divesified 811,812,813,814,821,822, Vaious Souce: CAPRI modelling sysem Figue 14 shows he elaion beween he fam ypes and ohe elemens of he modelling sysem. Inside he sysem, fam ypes ae aggegaed o NUTS 2 and Membe Saes, o allow a link o he policy and make module. These aggegaions allow exploiing he esuls fom fam ypes in maps and ables elaing o geogaphical unis. All esuls ae soed in he daa base managemen sysem as well and can be easily accessed. Page 117 of 133

118 Figue 14. Inegaion of fam ypes in he CAPRI modelling sysem CAPRI Regionalised Daa Base Mapping ool HTML ables EXCEL ec. CAPRI Modelling Sysem Fam ypes Aggegaion NUTS II Aggegaion MS,EU CAPRI Resul Daa Base DAOUT Souce: Own calculaions Figue 15 shows he dominan fam ypes pe couny. Fo easons of suvey eseach he fam ypes menioned in able 25 ae fuhe combined. I clealy shows ha daiy is a dominan fam ype in noh of Euope. An excepion is Denmak whee specialised COP, livesock and cops divesified and specialised pigs and pouly ae he dominan fam ypes. Cale faening, eaing, sheep and goas ae he dominan fam ypes in Ieland and Unied Kingdom. In he souh of Euope, Pougal, Spain, Ialy and Geece, pemanen cops and vegeables is he dominan fam ype. Also in Fance and o a lesse exen in Belgium/Luxemboug and he Nehelands, his fam ype is elaively impoan. The heeogeneiy of fam ypes seems o be quie big in Fance (diffeen fam ypes have abou he same weigh) and small in Ieland. Page 118 of 133

119 Figue 15. Fam ypes in EU15 counies Souce: Own calculaions Page 119 of 133

120 7 CAPRI Exploiaion ools The usefulness of a lage and complex modelling sysems such as CAPRI is closely linked o quesion how esuls can be accessed. Round abou non-zeo numbes pe NUTS II egion ae he oucome of a single scenaio un, ogehe wih he esuls of he make model, moe hen half a million non-zeo numbes ae poduced. Wihou nicely developed ools o view ino he daa heap, uses would simply be los when ying o analyse he esuls. The exploiaion ools fo he CAPRI modelling sysem have developed ove ime: Fom he vey beginning, all esuls ae soed backed in a binay daa base fom which all o seleced esuls fom one o seveal uns can be loaded in a muli-dimensional viewe called DAOUT. Expos of seleced daa fom he viewe via clip-boad o ohe applicaion ae vey simple. On op, bulk expos ae possible o exenal file foma. DAOUT was no specifically developed fo CAPRI, bu is pa of a Daa Base Managemen Sysem developed a Bonn Univesiy since he 70ies, and used in ohe modelling sysem as RAUMIS; CAPSIM o WATSIM as well. Figue 16. DAOUT Geneal oveview Souce: CAPRI Modelling Sysem Since 2001, esuls ae also pesened in ine-linked HTML ables which allow fo dilling down ino he esuls. The laes vesion of ha ool builds upon XLM/XSLT, and he esuling ables can be copied via he clipboad in ohe applicaions. Especially he laes vesion of Micosof Office peseves he fomaing of he able. The descipion below does no only descibe he ool, bu explain how uses may manipulae Page 120 of 133

121 exising ables fo inoduce new ones. The XML files ae geneaed by GAMS duing he un, so ha no fuhe echnical seps ae necessay o view he daa. Figue 17. The XML Tool Geneal oveview Souce: CAPRI Modelling Sysem; Copyigh of he XSLT/XML able ool Vesion 2.0: Wolfgang Biz, Insiue fo Agiculual Policy, Bonn, Gemany, 2005 The egional dimension of he modelling sysem asks fo addiional exploiaion ool which was developed in 1999 as he CAPRI mapping ool. A small Java Apple loads files geneaed diecly by GAMS, and geneaes ineacive maps. Page 121 of 133

122 Figue 18. The mapping ool geneal design Souce: CAPRI Modelling Sysem Some of hese exploiaion ools will be expoed in is cuen fom o wih sligh modificaions. Fuhe echnical deails ae povided in he CAPRI web sie. Page 122 of 133