EUR 23381 EN - 2008
The Common Agricultural Policy SIMulation (CAPSIM) Model:
Database for Agricultural Sector Modelling
Heinz Peter Witzke, Axel Tonini, Andrea Zintl
The mission of the IPTS is to provide customer-driven support tothe EU policy-making process by researching science-based responsesto policy challenges that have both a socio-economic and ascientific or technological dimension. European Commission JointResearch Centre Institute for Prospective Technological StudiesContact information Address: Edificio Expo. c/ Inca Garcilaso, s/n.E-41092 Seville (Spain) E-mail: [emailprotected]Tel.: +34 954488318 Fax: +34 954488300 http://ipts.jrc.ec.europa.euhttp://www.jrc.ec.europa.eu Legal Notice Neither the EuropeanCommission nor any person acting on behalf of the Commission isresponsible for the use which might be made of thispublication.
Europe Direct is a service to help you find answers to yourquestions about the European Union
Freephone number (*):
00 800 6 7 8 9 10 11
(*) Certain mobile telephone operators do not allow access to 00800 numbers or these calls may be billed.
A great deal of additional information on the European Union isavailable on the Internet. It can be accessed through the Europaserver http://europa.eu/ JRC44285 EUR 23381 EN ISBN978-92-79-09072-1 ISSN 1018-5593 DOI 10.2791/21945 Luxembourg:Office for Official Publications of the European CommunitiesEuropean Communities, 2008 Reproduction is authorised provided thesource is acknowledged Printed in Spain
http://europa.eu.int/citizensrights/signpost/about/index_en.htm#note1#note1
The Common Agricultural Policy SIMulation
(CAPSIM) Model: Database for Agricultural Sector Modelling
Heinz Peter Witzke (EuroCARE and University of Bonn)
Axel Tonini
(European Commission DG JRC-IPTS)
Andrea Zintl (EuroCARE and University of Bonn)
Institute for Prospective Technological Studies
2008
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
ii
Executive Summary Agricultural sector models are used to analysethe impacts of policy changes and are often based on partialequilibrium models. These models are strongly relying on modellingdatabases, compiled from several statistical sources among whichEurostat is one of the most frequently used. Eurostat data areknown for their high reliability given that they have been checkedand revised over many years. However considering the European Union(EU) enlargement it is difficult to guarantee the same qualitystandards for the new Member States, especially in the first yearsbefore and after accession. Data are often incomplete and even ifthey are included they are likely to be revised in the comingyears. As a consequence data processing is necessary in order toupdate country coverage, items and products and constantly checkthe reliability of the data. The aim of the present technicalreport is to describe the underlying techniques and methodsdeveloped for data selection and data preparation for the CommonAgricultural Policy SIMulation (CAPSIM) model (see Witzke and Zintl2007). CAPSIM was developed in the early 1980s by EuroCARE and theUniversity of Bonn on behalf of DG ESTAT. In 2006, the CAPSIM modelwas transferred from DG ESTAT to the European Commission's JointResearch Centre, Institute for Prospective Technological Studies(JRC-IPTS) in order to extend the model to new Candidate Countriesto the European Union accession and to further develop themodelling tools for CAP analysis. In terms of country coverage thedatabase covers: EU-27, Croatia, Former Yugoslav Republic ofMacedonia, as well as other Western Balkan countries, and Turkey.In terms of items and products the database is set up in a moredisaggregated form in order to extend the dairy commoditiescoverage in view of the ongoing Common Agricultural Policy reform.The first steps of data processing are shared between the CommonAgricultural Policy Regional Impact (CAPRI) and CAPSIM modellingsystems and teams. The modelling database is implemented in aroutine called "Complete and Consistent Database" (COCO) toestablish database' completeness and consistency based on thevarious type of official data. The routine consists in two steps:inclusion and combination of input data and calculation of completeand consistent data. The first step being related to datacollection is thoroughly described in this report. The second stepon the calculation of complete and consistent data is not treatedin this report since it is part of a mere modelling step ratherthan an input data selection step. The present technical report isstructured as follows. The first part is an introduction where theaim and objectives are presented and explained. The second part ofthe technical report focuses on providing the characteristics ofthe database, presenting the structure of the database, theextension to dairy commodities as well as a description of dataquality for a selected number of countries. The third focuses ondata selection and data preparation by using several real examplesand explaining how supplementary data are used.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
iii
Table of contents 1 INTRODUCTION 1 2 CHARACTERISTICS OF THEDATABASE 3
2.1 STRUCTURE 3 2.2 EXTENSION TO DAIRY COMMODITIES 4 2.3 DATAQUALITY FOR SELECTED COUNTRIES 6
3 DATA SELECTION AND DATA PREPARATION 10 3.1 DATA SELECTIONROUTINES 10
3.1.1 Eurostat data selection 10 3.1.2 Additional country dataselection 12 3.1.3 Current data selection check 14
3.2 FIRST PART OF COCO DATA COLLECTION 14 3.2.1 Including datafrom New Cronos 14 3.2.2 Data from additional sources 15 3.2.3Completion with additional sources 18 3.2.4 Assigning data todatabase array 22
3.3 SUPPLEMENTARY DATA SELECTION 31 3.3.1 Bulgaria and Romania32 3.3.2 Croatia, the FYROM and Turkey 33
REFERENCES 36 4 ANNEX 1: ITEMS OF THE DATABASE 37 5 ANNEX 2:COUNTRY FILES FOR CROATIA, THE FYROM AND TURKEY 74 6 ANNEX 3:SUPPLEMENTARY DATA FILE FOR ROMANIA AND BULGARIA 74 7 ANNEX 4:ADDITIONAL: COUNTRY FILES FOR ALBANIA, BOSNIA AND
HERZEGOVINA, MONTENEGRO, SERBIA, AND KOSOVO 74 8 ANNEX 5: CAPSIMCODES AND ABBREVIATIONS 74
Data columns currently used in CAPSIM 74 Data rows currentlyused in CAPSIM 78 OTHER ABBREVIATIONS 81
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
1
1 INTRODUCTION The aim of this technical report is to describethe underlying techniques and methods establishing a database to beused for the Common Agricultural Policy Simulation Model (CAPSIM)(see Witzke and Zintl 2007). CAPSIM is one of the in-housemodelling tools at the Joint Research Centre, Institute forProspective Technological Studies (JRC-IPTS) to support analyses onthe Common Agricultural Policy (CAP). In order to carry out policyrelevant modelling analyses, the CAPSIM database had to be updated,checked and extended, in particular for Romania, Bulgaria whor*cently joined the European Union (EU) and for CandidateCountries. Establishing the database for CAPSIM requires thefollowing steps: checking existing datasets, most importantlyEurostat, for completeness and quality problems; supplementingofficial Eurostat data wherever necessary or useful with other datasources (often from National Statistical Offices (NSOs) andAgricultural Ministries (AMs) in the countries considered in thistender); organising the complete modelling database on a readableformat directly from the General Algebraic Modelling System (GAMS)as well as from standard software packages (Excel or Access). Interms of country coverage, the database includes EU-25 MS (MemberStates), Bulgaria and Romania and it is extended to Croatia, theFormer Yugoslav Republic of Macedonia (FYROM) and Turkey and anupdate is provided for other Western Balkans (WBs). The database isset up in a more disaggregated form, in terms of items andproducts, with a particular focus on dairy commodities. Eurostatdata are available for a long ex-post period for all EU-15 MS andare highly reliable, as they have been checked and revised overmany years. However, Eurostat data on the 10 New MS (NMS) accedingin 2004 still cannot reach the same quality standards. In someareas, the coverage is incomplete and Eurostat1 is permanentlychecking and updating data already available. In the meantime,Bulgaria and Romania became new EU MS on 1 January 2007, Croatia,the FYROM and Turkey have the status of Candidate Countries andother WBs are potential candidates over a medium term horizon.Evidently, data problems are even more serious in these countriesthan in the NMS. In Bulgaria, Romania, Croatia and the FYROMsubsistence and semi-subsistence agriculture play major roles interms of agricultural production, land and labour utilization. Theavailability of data, for this part, is very limited and does notcover at all the dairy sector. Some official data have changedremarkably for the Balkan countries according to the latest datareleases, questioning, what a year ago, appeared to be valuableinformation. In contrast to Bulgaria and Romania only a limiteddataset is offered for Croatia, the FYROM and Turkey at NewCronos2. Therefore, the data collected from the NSOs, AMs and othersources for these countries will form the main basis for theestablishment of a sound database for CAPSIM. Techniques andmethods for establishing this database, combining the differentdata sources, are described in detail below. At this point, it maybe noted that national and Eurostat data, different Eurostatdomains and sometimes even the numbers in a single 1 In 2005 aspecial tender (Ref. 2004/S 42-036276/EN) was launched by Eurostatto include the NMS in the modelling database for CAPSIM. 2 NewCronos is the principal Eurostat database, which contains highquality macroeconomic and social statistics data. The data areorganised into 'key indicators on EU policy' and nine statisticalthemes, which are then subdivided into collections of datatables.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
2
market balance are not necessarily consistent with each other.For a number of years already, the first steps of data processingfor model initialization are shared between the Common AgriculturalPolicy Regional Impact (CAPRI) and CAPSIM modelling systems andteams. The modelling database is implemented in a routine called"Complete and Consistent Database" (COCO) to establish databasecompleteness and consistency based on the various types of officialdata (see Britz 2005, section 2.3). This routine allows unitconversion, trend based completions, and mechanical corrections ofpresumed data errors while imposing some minimal technicalconsistency in terms of adding up constraints for areas and soforth. The COCO module is basically divided into two main parts:including and combine input data according to some hierarchicalcriteria, calculating complete and consistent time series whileremaining close to the raw data. The first part, closely related tothe collection of raw data, forms a bridge between raw data anddata consolidation to impose completeness and consistency and isdescribed in this report. The second part is a modelling activityinvolving optimisation problems in a normalised least squaresframework and is not covered in this report.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
3
2 CHARACTERISTICS OF THE DATABASE
2.1 STRUCTURE This section presents how the database isstructured by listing and explaining all items covered in thedatabase. Reference is made particularly to data selection, andconsistency check.
Area of production The area levels for crop products areselected from Eurostat's crop production statistics as far aspossible. Missing aggregates (e.g. "cereals") are computed based onadding up their available components. Residuals (e.g. "other oilseeds") are calculated from the aggregate data minus the componentsselected. Additional auxiliary data are used to define some otherproducts (e.g. apples, pears and peaches for the aggregate "apples,pears & peaches").
Harvested production Data are downloaded for crop productionfrom Eurostat's statistics. However, for consistency with marketbalances (see below), the above data will be overwritten withusable production from the market balance sheets, potentiallycorrected for losses and seeds on farm, if usable production isavailable. Aggregates and residuals are calculated for levels.
Crop yields (t/ha) Crop yields are calculated as a result of allthe area levels and production data available.
Meat production This data area includes animal herd size,slaughtering, imports and exports. Data for slaughtering and tradeare taken from Eurostat's meat production statistics. Nonetheless,for consistency with the market balances these data will beoverwritten with usable production from balance sheets if usableproduction is available.
Animal output per activity The specific output of meat and younganimals per animal, for all animal activities, are calculated bydividing the total output by the corresponding activity levels. Forthe dairy cows activity the output of milk, per cow, is alsoprovided.
Animal input per activity The required input of young animalsfor each animal activity is derived from the slaughteringstatistics including trade in live animals.
Balance positions Balance positions for crops, animals, derivedproducts and milk products include:
Usable production Losses total Total imports Losses farm Totalexports Animal feed Final stocks Industrial use Change in stocksProcessing Total domestic use Agricultural sales Seed total Grosshuman consumption Seed - farm
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
4
Economic Accounts for Agriculture (EAA) positions Valued dataare downloaded from Eurostat's EAA domain. They are linked toproduction data, from crop and meat production statistics toanalyse the income generated. Occasionally, diverging definitionsfor product aggregates (e.g. pulses) in the EAA and productionstatistics require modifications. The database includes data for:EAA at producer prices, EAA subsidies, EAA taxes, EAA at basicprices, EAA at constant year 2000 producer prices, andEAA priceindices (2000=100). Finally, production quantities from the valuedEAA positions, are selected to check if they are aligned to thephysical statistics and sometimes as a fall back position ifproduction statistics and market balances are empty: EAAquantities.
Price data Selling (producer) prices: selling prices areselected from Eurostat's "Agricultural prices annd price indices"domain for all crop and animal products. They are mainly used tocomplete the series on unit values at producer prices from EAA (seebelow). Purchase prices of the agricultural production: purchaseprices for input, such as those for fertilizers or energy, are alsoselected from Eurostat's "Agricultural prices and price indices"domain. Unit values at producer price and at basic price: unitvalues are the quotient of production value and physicalproduction. They are available at producer prices and at basicprices. Unit values at consumer price level: these will becalculated in the COCO routine based on expenditure and consumptiondata.
Household expenditures Data from Eurostat's "Economy andfinance" domain are selected from: breakdown of final consumptionexpenditure of households by consumption; harmonized indices ofconsumer prices; harmonized indices of consumer prices - itemweights and breakdown by food prices. And from the "Agriculture andFisheries, Food: From farm to fork statistics" domain: relativeprice level indices of food products (EU-15=100). The above dataare used to calculate the unit values at consumer prices.
Feed use Data from Eurostat's "Agriculture and Fisheries, Food:From farm to fork statistics, Input into agriculture: feedingstuffs, seed, fertilisers, plant protection products" have beenselected: feed use of disaggregated feeding stuffs.
Exchange rate (1 Euro = ... NC) Where available, the annualexchange rates are from Eurostat's EXINT domain.
Inhabitants Population data are given in Eurostat's AUXINTdomain.
Labour Agricultural labour (total hired) is available fromEurostat's COSA domain as this enters the calculation of Eurostatsagricultural income index.
GDP price index The GDP price index is taken from Eurostat'sAGGS domain.
2.2 EXTENSION TO DAIRY COMMODITIES The CAPSIM model in view ofthe CAP Health Check proposals (see EU Commission 2007) has beenshaped by the JRC-IPTS3 in order to put emphasis on 3Methodological Development of the CAPSIM model and scenarioanalysis, (Tender No J05/34/2006), Institute for ProspectiveTechnological Studies (IPTS), Seville.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
5
dairy policy issues. This involved, among other elements, adisaggregation of the product list. In the scope of this, the dairyproducts coverage is extended to include market balances for wheypowder and casein, relying on a pragmatic approach due to the lackof hard data.
Available information Production data from Eurostat (domain:ZPA1, sub-domain: APRO_MK_POBTA = Milk collection (all milks) anddairy products obtained). External trade data, based on the Foodand Agriculture Organization (FAOSTAT) integrated core database(avoids aggregation from Eurostat COMEXT4 database and problems inobtaining trade data for non-EU countries). EU wide demand for useof whey powder in other compound feed (about 300000t) and forcasein use in cheese production (about 11000t) and feedstuffs(about 20000t) from a special survey from 1989 to 1991 (apparentlyno comparable survey had been carried out subsequently) undertakenby the "Zentrale Markt- und Preisberichtstelle fr Erzeugnisse derLand-, Forst- und Ernhrungswirtschaft GmbH" (ZMP) (see ZMP 1993).From 1990 to 2006 ZMP market balance data identify total demand andmilk component replacing feed for EU-12, EU-15 or EU-25 from 2000to 2006. Default protein contents for casein and whey powder are85% and 10.9% respectively (fromhttp://www.slmb.bag.admin.ch/slmb/archiv/index.html, chapter 4,Milchdauerwaren, table 4.2) fat contents are 0.8% and 1.1%respectively.
Assumption The ZMP information is the best estimate available onthe demand composition for the whole EU. This historicalinformation on certain shares and ratios is first used, to estimatecomplete time series of EU wide market balances (see below). Thetable below shows the raw data situation.
Table 1 Raw data for whey powder and casein market balancesTable ZMP_EU(*, Rows,TT1)
* EU (12) EU (12) EU (12) EU (15) EU (15) EU (15) EU (25) EU(25)1989 1990 1991 1995 1996 1997 2004 2005
USAP . WHEP 910 930 926 1280 1310 1280 1600 1630IMPT . WHEP 8 819 9 3 1.9 2 2EXPT . WHEP 25 38 32 54 76.4 116.1 306 330DOMM . WHEP893 900 913 1235 1236.6 1200.8 1355 1302MilkReplacer . WHEP 570 550600 812 846 832 920 920CompoundFeed . WHEP 200 220 180FEDM . WHEP770 770 780HCOM . WHEP 123 130 133STCM . WHEP 0 0 0 35 0 -35 -600
USAP . CASE 151 108 113 145.7 136.0 134.0 179 175IMPT . CASE 3663 59 62.7 58.0 58.0 51 41EXPT . CASE 75 72 70 82.0 75.0 70.0 9292DOMM . CASE 112 99 102 126.4 119.0 122.0 137 124PRCM . CASE 9 99FEDM . CASE 20 14 17HCOM . CASE 83 76 76 White background:"Zentrale Markt- und Preisberichtstelle fr Erzeugnisse der Land-,Forst- und Ernhrungswirtschaft GmbH" (ZMP). Yellow background: ZMPMarket Balances for 2000 2006. For detailed information on codingsee Annex 1. Data calculation for all missing items is asfollows:
Whey Powder Compounded feed is calculated from the"CompoundFeed" and domestic use ("DOMM") ratio from the years1989-1991 multiplied by the current "DOMM" data.
4 The Eurostat COMEXT database contains detailed foreign tradedata as reported by the EU 25 Member States.
http://www.slmb.bag.admin.ch/slmb/archiv/index.html
See AlsoFactbox: Egypt's economic strategy for 2024-2030Questão 2: As afirmativas abaixo representam relação entre grandezas que envolvem proporcionalidades. As grandezas podem ser: grandezas diretamente... - Desenvolvimento de Recursos HumanosRiding The Infrastructure Train Is Positive Journey For UK AdvisorThe Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
6
Consequently feed on the market ("FEDM") is the sum of thequantity used for milk replacing feed ("MilkReplacer") and"CompoundFeed". Human consumption ("HCOM") is finally the residualusing "DOMM" and "FEDM".
Casein Processing ("PRCM") is calculated from the "DOMM" and"PRCM" ratio from the years 1989-1991 multiplied by the current"DOMM" data. "FEDM" is correspondingly calculated from the "DOMM"and "FEDM" ratio from the years 1989-1991 multiplied by the current"DOMM" data. Finally "HCOM" is the residual using "DOMM" and thecalculated "FEDM" and "PRCM". The data calculated are shown in thefollowing table, highlighted in green:
Table 2 Completed market balances for whey powder and caseinTable ZMP_EU(*, Rows,TT1)
* EU (12) EU (12) EU (12) EU (15) EU (15) EU (15) EU (25) EU(25)1989 1990 1991 1995 1996 1997 2004 2005
USAP . WHEP 910 930 926 1280 1310 1280 1600 1630IMPT . WHEP 8 819 9 3 1.9 2 2EXPT . WHEP 25 38 32 54 76.4 116.1 306 330DOMM . WHEP893 900 913 1235 1236.6 1200.8 1355 1302MilkReplacer . WHEP 570 550600 812 846 832 920 920CompoundFeed . WHEP 200 220 180 273.8 274.2266.3 300.5 288.8FEDM . WHEP 770 770 780 1085.8 1120.2 1098.31220.5 1208.8HCOM . WHEP 123 130 133 149.2 116.4 102.5 134.893.6STCM . WHEP 0 0 0 35 0 -35 -60 0
USAP . CASE 151 108 113 145.7 136.0 134.0 179 175IMPT . CASE 3663 59 62.7 58.0 58.0 51 41EXPT . CASE 75 72 70 82.0 75.0 70.0 9292DOMM . CASE 112 99 102 126.4 119.0 122.0 137 124PRCM . CASE 9 9 910.9 10.3 10.5 11.9 10.7FEDM . CASE 20 14 17 20.6 19.4 19.9 22.420.2HCOM . CASE 83 76 76 94.9 89.3 91.6 103.2 92.9 Having completedthe time series of EU market balances, it was necessary to set upmarket balances on the MS level which were consistent with these EUmarket balances. Domestic use could be calculated from the givendata (on production and trade). But the decomposition of thisdomestic use has only been estimated on the EU level (see above)whereas data on MS specific particularities are unavailable, forexample, an above average use of casein for feed in a particularMS. It was necessary, therefore, to apply the EU shares of a givenyear to all countries. MS particularities are covered, however, interms of the significance of casein and whey powder for nationaldairy markets, because total domestic use may be calculated. Inthis way, it is possible to use the existing data, even though theyare not as complete as for other dairy products where Eurostatoffers complete market balances ready to use.
2.3 DATA QUALITY FOR SELECTED COUNTRIES This section provides ashort qualitative overview assessment for a selected number ofcountries. The process for the establishment of a database hasensured that all available sources have been explored to themaximum extent possible in the time available. Regular contactsbetween the contractor, the NSOs and AMs allow exploring allofficial data available in the respective countries. Recentdevelopments and current status on agricultural statistics may becharacterised as follows:
Romania Gathering of agricultural data in Romania (includingdata for subsistence or semi-subsistence) is still in the processof being transformed from indirect to on farm data collection. Inthe past, data were mainly provided by the local authorities whokept rural household registers. Currently, the NSO is implementingnumerous surveys,
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
7
which will provide more reliable and comparable data in the nearfuture. Anyhow, this is a major undertaking as there are 3.6million people counted as being employed in Romanias farmingsector, representing 32.1% of the total countrys labour force. Thesector, however, only contributes to about 8.5% of the total GDP(2005). The above figures prove that there are significant numbersof small subsistence or semi-subsistence agricultural units (withonly a few cattle or sheep and producing mainly or solely for ownconsumption). Data on meet and milk production need furtherverification. The share of direct sales of milk is, by far, thegreatest of all the EU MS. From the total production of about 6billion litres, less than 25% is delivered to dairies, the restbeing direct sales in informal markets and used forself-consumption. During recent months, the Romanian NationalInstitute for Statistics together with the statistical unit of theMinistry of Agriculture undertook major efforts in upgrading thescope and quality of its databases. Furthermore, the provision ofcorresponding data to the Eurostat New Cronos database has beenintensified. However, it is well known that quality of data on milkand meat production is critical. Data on milk and meat production,collected in national agencies of Romania, were mainly used forverifying New Cronos data and completing data on ewe, goat, sheepmeat and milk production.
Bulgaria The Ministry of Agriculture in Bulgaria is responsiblefor agricultural statistics and the main data collector of data.Statistical methods are still in the process of full harmonisationwith EU requirements. Therefore, some data are still missing or arenot fully comparable with EU data. As there are significant numbersof small subsistence or semi-subsistence agricultural units, (theseare still more than 70% of the total number of farms with only afew cattle or sheep and producing mainly or solely for ownconsumption) data on meet and milk production need furtherverification. During recent months, the Agricultural StatisticsDirectorate of the Ministry of Agriculture in Bulgaria improved thescope and quality of its databases. Furthermore, the provision ofcorresponding data to the Eurostat New Cronos database has beenintensified. As for Romania, particular attention is given onverifying the quality of critical data on milk and meat production.The dairy and meat industry in Bulgaria is still undergoingimportant restructuring. The European Commission (EC) granted 208Bulgarian milk and dairy producers and 378 meat and meat productproducers a transition period to meet EU standards. These milk anddairy producers have until 2009 to meet EU standards. For Bulgaria,data on national production on milk and meat were used to verifyNew Cronos data and completing data on ewe, goat, sheep meat andmilk production.
Croatia Approximately 10% of the working age population earntheir income from the agriculture and food industry. The share ofa*gricultural labour in the total labour force was about 7.3 % in2003 (excluding the food industry). The Croatian Central Bureau ofStatistics (CBS) is currently in the process of adjusting itsmethods to EU standards. Therefore, several data are currentlyunder revision or unavailable. several parts of agricultural andrural statistics are still under construction (e.g. product balancesheets, rural development indicators, etc.) and are, therefore,hardly available. In some cases (for example, the calculation ofa*gricultural area) the CBS and the Croatian Ministry of Agricultureuses different sources for the calculation of indicators and dataare significantly different according to these institutions. Dueto
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
8
the fact that the CBS is currently upgrading and adjusting itsstatistics, data availability for 2006 is only limited. Therefore,a delay in data collection occurred and data collected in theannexed Excel file are as complete as possible, considering thesecirc*mstances. In 2005 the CBS, for the first time, collected dataconcerning private family farms by using the face-to-face methodand a selected sample. This meant they gave up the long lastingmethod of collecting data by using the expert estimation methodcarried out by agricultural experts, on the basis of cadastre data.Consequently, significant differences in data on land areas of somecrops, vineyards and orchards emerged. Therefore, 2005 data shouldrefer to actually used land areas, meaning a break in the timeseries. Unfortunately, the survey exclusively relied on the newmethodology with no attempts at estimating a correction factor,which would be required to make proper backwards calculations. Infact, the same occurred with the livestock surveys. From 2006, thegross indigenous production (GIP) is calculated to include theimport and export of live animals. Fortunately, data for 2002-2005have also been recalculated in order to establish a new timeseries. Further back in time is currently impossible but the twobridge years must give a solid base for appropriate analysis. Inaddition, estimates on slaughtering of livestock, for the yearsmentioned, were expressed in numbers and weight according to thebalance of the number of livestock.
The FYROM Agriculture counted for around 10% of GDP during2000-2004. Adding related marketing and processing activitiesincreased this share to approximately 15%-16% of GDP. Agriculturesshare in employment according to the International LabourOrganisation (ILO) methodology indicates that approximately 17% ofthe total workforces earned their income mainly from agriculture,hunting and fisheries in 2005. The actual share in employment islikely to exceed this but, there are no definitive statistics onthe informal economy or on the subsistence farming sector. Itshould be noted that over 45% of the population live in rural areaswith limited employment opportunities outside the agriculturalsector. Implementation of an agricultural census was postponedseveral years ago. It has now been implemented in the early summerof 2007. Therefore, there are no reliable data on agriculturalstructures. Implementation of sample surveys is restrained by thedelay in the implementation of a census. There are very limiteddata on prices and agricultural incomes. The database on ruraldevelopment is weak and non-existent in parts. Due to the heavyworkload by the implementation of a census, the agriculturaldepartment experience significant delays in providing updates tothe Excel-database.
Turkey In terms of employment, agriculture is the largest sectorin Turkey with more than 30% of the workforce. The overall systemof agricultural statistics in Turkey is undergoing a substantialreorientation and reorganization. In the past, data were rarelycollected on the immediate farm level. Village administrationsprovided most of the agricultural data: an important characteristicto be considered for an assessment of data reliability. The lessdeveloped situation in certain rural areas, as well as thestructure of agriculture limits data collection and agriculturalstatistics in Turkey. Implementation of surveys in these regions isa difficult exercise. However, a farm structure survey wasimplemented at the end of 2006 and data are currently underpreparation. The actual finalisation of the farm structure surveyin Turkey allowed access to actual data by the end of autumn 2007.A preliminary data-table has been elaborated in close cooperationwith the Turkish Statistical Institute (TURKSTAT).
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
9
Turkey is the largest producer of milk and dairy products in theBalkan Middle east region. There are a number of modern processingplants for dairy products but the majority (more than 2/3) of milkproduction is still sold in farms or in informal markets withoutany hygiene control.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
10
3 DATA SELECTION AND DATA PREPARATION The following chapter isexpected to help the user understand the procedures of dataselection and data preparation. It also explains how data input ismanaged and combined in the first part of the COCO-module,particularly the new data collected.
3.1 DATA SELECTION ROUTINES Data for the CAPRI and CAPSIM modelsare selected from Eurostat for all countries and additionally fromnational sources (e.g. NSOs and AMs) for Romania, Bulgaria,Croatia, the FYROM and Turkey. About 90 MB of data from Eurostatare downloaded from the New Cronos database in TSV (Tab Separated)format. This data format is converted by the program TSVCON into animport file (NIM) format. For each Eurostat domain (ZPA1, COSAetc.) the conversion is created combining all selected TSV-filesfor each domain. Selection on data and mapping to model codes isdefined in the coco.ass file. The program IMPSEL is running foreach domain using the unique mapping tables and correspondingNIM-files.
3.1.1 Eurostat data selection About 90 MB of data from Eurostatare downloaded from the New Cronos database.
First step: Data download As the study requires complex sets ofdata, all data are downloaded in TSV-format, as offered by Eurostatfor special users. The TSV-format is a flat file format for timeseries, identified by New Cronos codes. Data can be selected forall the EU-27 MS and for some WBs, whereas the extent ofavailability differs by country. Data for the WBs within New Cronosare quite limited. The following themes and domains of New Cronosare accessed for the database: Agriculture and Fisheries
COSA domain EAA, agricultural labour input statistics, unitvalue statistics for agricultural products
PRAG domain Agricultural prices and price indices ZPA1 domainAgricultural products (ZPA1) FOOD domain Farm to fork statistics(FOOD)
Economy and Finance BRKDOWNS domain Breakdown of finalconsumption expenditure of
households by consumption AUXIND domain Inhabitants AGGS domainPrice index of gross domestic production (GDP) PRICE domainHarmonized indices of consumer prices EXINT domain Exchangerates
Second step: Format conversion The second step of dataprocessing is the conversion of the TSV-format to an intermediateflat file format (NIM-format). This step is necessary because theTSV-format is neither uniformed in coding nor sorted, which wouldcomplicate further selection of data. The TSVCON program works foreach domain (e.g. ZPA1) and converts the multiple TSV-files of thisdomain to a single NIM-file. No selection of data is performed inthis step, but the coding of each data is prepared. Thedifferent
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
11
codes for MS are also unified. Finally the NIM-file entries aresorted for fast selection.
Table 3 Part of the file "apro_mt_pann.tsv" from the ZPA1 domain(TSV-format):
prodmeat,meatitem,unit,geo\time2007a00 2006a00 2005a00 2004a002003a00 2002a00 2001a00 2000a00 1999a00
v10,sl,1000t,bg : : : : : : : 66.000 63.000v10,sl,1000t,cz :79.712 81.03 96.66 109.505 109.495 106.045 108.162127.000v10,sl,1000t,dk : 128.702 135.976 150.082 146.645 153.586153.420 153.910 156.670
The data highlighted in the above table show total cattleslaughtered (1000t in carcass weight) for Bulgaria in the 2000s.The result of the conversion program TSVCON for the years 2000s isshown in the following table.
Table 4 Part of a file in NIM-format from the ZPA1 domain:42.000 BG0002000ZPA1.APRO_MT_PANN.V10.EL.1000HD
7.000 BG0002000ZPA1.APRO_MT_PANN.V10.EL.1000T422.000BG0002000ZPA1.APRO_MT_PANN.V10.SL.1000HD
66.000 BG0002000ZPA1.APRO_MT_PANN.V10.SL.1000T59.000BG0002000ZPA1.APRO_MT_PANN.V10.GP.1000T
The value "66.000" highlighted is coded by the string"BG0002000ZPA1.APRO_MT_PANN.V10.SL.1000T" where: BG000 country code(Bulgaria) 2000 year ZPA1 domain code APRO_MT_PANN sub-domain codeV10 product code (cattle) SL slaughtering 1000T unit code(1000t)
Third step: data selection The third step is to work on dataselection and is performed by the IMPSEL program. A mapping tablelinking New Cronos codes to COCO codes defines the subset of dataseries subsequently used (file "coco.ass").
Table 5 Part of the "coco.ass" file:
ASS TARGET=SLGH.CATT, SOURCE=ZPA1.apro_mt_pann.V10.SL.1000HD;*Slaughtered total cattle, headsASS TARGET=SLGT.CATT,SOURCE=ZPA1.apro_mt_pann.V10.SL.1000t,* Slaughtered total cattle,1000t in carcass weight The keyword "ASS" triggers an assignment tothe "TARGET" COCO code from "SOURCE", i.e. the New Cronos code. Inthe above example, the first assignment should be interpreted asfollows: SLGT.CATT Total cattle slaughtered, 1000t in carcassweight Model column code: SLGT Model row code: CATTZPA1.apro_mt_pann.V10.SL.1000T ZPA1 domain Sub-domain:apro_mt_pann
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
12
Product: V10.SL Unit: 1000T The IMPSEL program is appliedseparately for each domain (COSA, ZPA1, PRAG etc.). Program inputfor each domain is the unique mapping table and respective NIM-file(see above). The result of this procedure is a set of CommaDelimited (CSV) files directly readable by the GAMS program and byExcel and Access. The extraction from the "zpa1.csv" file belowshows the results of the IMPSEL program for the database item"BG000.SLGT.CATT".
Table 6 Part of the "zpa1.csv" file: * Selected from SPEL importfile at 29.09.07 1994 1995 1996 1997 1998 1999 2000BG000.SLGT.IPIG207.00 256.00 252.00 227.00 248.00 267.00 243.00BG000.SLGT.ICAL64.00 57.00 63.00 48.00 46.00 54.00 60.00BG000.SLGT.CATT 85.0065.00 80.00 57.00 56.00 63.00 66.00BG000.SLGT.SHEP 37.00 35.0042.00 40.00 40.00 40.00 39.00BG000.SLGT.GOAT 6.00 10.00 9.00 11.0013.00 18.00 20.00
3.1.2 Additional country data selection Additional data fromnational sources for Romania, Bulgaria, Croatia, the FYROM andTurkey (NSOs, AMs and some expert data) are introduced. Theapproach is reviewing and assessing what is available at Eurostat,collecting missing data from additional sources in these countries,if possible, and estimating important missing data if necessary.Only a limited number of critical data (for milk and dairyproducts) for Romania and Bulgaria were added, based on informationfrom the NSOs of these countries. National data for Croatia, theFYROM and Turkey are the dominating source as Eurostat coverage isquite fragmentary.
Romania and Bulgaria Data problems in Romania and Bulgaria arisefrom significant shares of subsistence and semi-subsistenceagriculture, where a great share of production is consumed by thefarmers households or sold in local informal markets. For milk,meat, fruits and vegetables the share of on farm consumption andsale in informal markets is above 50%. The data available atEurostat were coded by country code &"000", e.g."BG000.SLGT.CATT". Most efforts made to collect supplementarynational data in meetings with the Statistical Offices and the AMshave been focussed on the meat and milk sectors, as these are seenas the most problematic in terms of data quality, due to the highshare of on farm consumption and sale in informal markets in thesecountries. Furthermore, crop sector data are in general easier toobtain. All data were collected in a separate sheet, where the timeseries were coded for default database input by country code&'sup', model column code and model row code,e.g."BGsup.SLGT.CATT" (Bulgaria, total cattle slaughtered in1000t). The collected expert data were compared with the dataavailable at Eurostat and a specific coding for data replacement inthe COCO routine was used in the update file as follows: Thestandard region code (country code &"000") for Eurostat datawas renamed to country code &"ori", e.g. "BG000" to "BGori";The series to replace the original Eurostat series was coded ascountry code &"rev", e.g. "BGrev". In the example below,Eurostat's data for total cattle slaughtered (1000t in carcassweight) are revised by the FAO data series. The decision is madegiven: FAO data fit perfectly to country data (sum of data fromslaughter houses and farms) for 2004; more recent years areavailable from FAO than from Eurostat.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
13
The example is quite typical. Time series in the milk and meatsector are often very short or affected by serious breaks, in sucha way that supplementary information, in particular for recentyears, is urgently needed.
Table 7 Part of the Bulgarian update file 1995 1996 1997 19981999 2000 2001 2002 2003 2004 2005 2006
*meat
BGori.SLGT.CATT 65 80 57 56 63 66* FAO data fits to the expertdata (sl.house + farm) in 2004 and more years are available than inNewCronosBGrev.SLGT.CATT 63.3 77.83 55.52 54.5 61.1 60.4 68.5 23.6528.51 30.77 29.78 22.91* 30.89*BGsup.SLFT.CATT23.26*BGsup.SLMT.CATT 7.624 where: BGori.SLGT.CATT Eurostat data ontotal cattle slaughtered, 1000t in carcass weight BGrev.SLGT.CATTFAO data to revise Eurostat data BGsup.SLFT.CATT Country data(slaughtering on farm) BGsup.SLMT.CATT Country data (slaughteringin slaughter houses) The last two items were summed up (marked inlight brown in Table 7) to compare with the FAO data for the year2004. See section 3.3 for more detailed information onsupplementary data collection.
Croatia, the FYROM and Turkey Comprehensive sets of additionaldata (mainly collected from NSOs and AMs) for Croatia, the FYROMand Turkey were organised in specifically designed Excel sheets(see section 3.3.2). The input for the COCO routine was collectedfor each country in a specific sheet. The time series wereidentified, by default, as supplementary expert data by countrycode &'sup', model column code and model row code, e.g.MKsup.CERE.LEVL. For data replacement update files for each countrywere created similar to the example for Romania.
Table 8 Part of the FYROM update file 1995 1996 1997 1998 19992000
MK000.BAR1.LEVL 36 37MK000.BAR2.LEVL 19 12MKori.BARL.LEVL 51 5351 50* For the years 1995 and 1996 barley is only available as BAR1and BAR2. The values for both yaers are summed up.MKrev.BARL.LEVL55 49 51 53 51 50*MKsup.BARL.LEVL 54.874 48.916 50.936 53.54150.289 49.998MK000.BEAN.LEVL 18 17 19 20 19 18MK000.BWHE.LEVL2MKori.CERE.LEVL 186 170 219 217 216 217* CERE.LEVL is obviouslythe sum of all cereals, but for the years 1995 and 1996 barley wasmissing (see BARL.LEVL above)MKrev.CERE.LEVL 241 219 219 217 216217 where, for example, MKori.CERE.LEVL Eurostat data forproduction area of cereals, 1000ha MKsup.CERE.LEVL Country data,1000ha MKrev.CERE.LEVL Data to revise Eurostat data, 1000ha In theselection from New Cronos, the area for barley (MKori.BARL.LEVL)for the years 1995 and 1996 was missing. These two years werecompleted by the sum of the areas for winter barley(MK000.BAR1.LEVL) and spring barley (MK000.BAR2.LEVL). As the areafor cereals is the sum of all cereals, the area of barley is notadded up for the years 1995 and 1996. The selection from NewCronos
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
14
(MKori.CERE.LEVL) is revised by the corrected data(MKrev.CERE.LEVL). In both cases the data received from thecountries are ignored (the lines are marked as comment) as they donot differ in a relevant way from Eurostat, which is the defaultsource.
3.1.3 Current data selection check Before starting the COCO datacollection routine, it is recommendable to check the currentselection from Eurostat to avoid severe errors when running thecompletion and consistency calculations. A reasonable approach isto compare the current selection with the previous one. The usercan opt to applying various software tools. After an update fromEurostat, frequent problems found are new scaling problems andtyping errors. They should be corrected or deleted. Furthermore,the Eurostat coding system frequently changes. Sometimes only thecodes for sub-domains were updated or re-arranged, which can becorrected by a couple of simple changes in the coco.ass file (seesection 3.1.1, third step). More severe are modified definitionsfor single items, which have to be reflected on a case-by-casebasis in the COCO program. In bilateral contacts with counterpartsfrom the FYROM, Croatia and Turkey, it was frequently stated thatthere is no established or regular data transfer to New Cronos.Some counterparts sent data using official transmission tables butdid not see these updates in their domain.
3.2 FIRST PART OF COCO DATA COLLECTION As mentioned in theintroduction, the focus here is on the first part of COCO includingand combining input data according to some hierarchical criteriafrom various sources. A recurrent characteristic of COCO is tosolve the problem; if the first best source has gaps in aparticular country, or even is entirely empty, it will use thesecond or even third best source if useful. Typical problems andsolutions in the GAMS code are set out as examples of codefragments using the COCO codes in Annex 1.
3.2.1 Including data from New Cronos The program starts byimporting data from Eurostat prepared beforehand, by the dataselection routines and manual checkings. The different domains areprocessed step by step and corrections made on selected data forall MS5. The example below shows how an obvious data error isremoved from Denmark's ZPA1-data. When checking the harvestedproduction from permanent grassland in the New Cronos selectionfrom spring 2007, two evident statistical breaks were identified:in 1992 production decreased from about 8000 to about 4000 (1000t);in 1994 production decreased again from about 4000 to 1000 (1000t).To avoid subsequent errors in the completion and consistencyprocedures, data for the years previous to 1994 are deleted fromthe selection.
5 Eurostat offers data for Belgium and Luxembourg separately,whereas the database combines both countries to the model region"BL000" (Belgium and Luxembourg). The key reason is that Eurostatoffers data mainly for the aggregate Belgium and Luxembourg up tothe year 1999, especially for all market balances. Furthermore,Luxembourg has a rather small agricultural sector (2004 totaloutput was about EUR 250 million) with some similarities toBelgium.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
15
After specific corrections for a particular MS, someaggregations applicable to all MS are performed. The COCO item"APPL" is aggregated from auxiliary data for apples, pears andpeaches and COCO code "FRUI" is aggregated from auxiliary data forfruit trees, plus soft fruits, plus strawberries. The example below(aggregation of "APPL") illustrates a general rule in COCO, notalways followed in the raw data (compare table 6 above). Series areonly aggregated if all components are given or if the componentseries is entirely empty:
Other corrections on ZPA1 data applicable to all MS deal withspecific problems on wine (scaling, changing codes over time),paddy rice (included in total rice balance data) and sheep and goatslaughterings (aggregated in some MS, disaggregated in others). Thesecond Eurostat domain is COSA and only needs a few corrections.The example below shows once more corrections for Denmark, where inthe EAA selection on price indices and data, at constant prices,statistical breaks were identified in the series for crops andfodder from arable land. The data for the years before the breakare deleted from the selection of Eurostat raw data.
The last Eurostat domain is PRAG which also requires somecase-by-case corrections, e.g. price data for eggs are sometimesgiven for 1000 eggs and have to be converted to the price for1000t.
3.2.2 Data from additional sources For the ten NMS additionaldata are included in separate files. In the framework of theEurostat contract "Extension of the agricultural sector model toCandidate Countries and establishment of a dataset for use in theagricultural sector modelling for Candidate Countries, Lot 1:Extension of the agricultural sector model and incorporation ofAcceding Countries " (see Witzke, H.P., Zintl, A. 2006) the COCOdatabase was extended for the NMS. Eurostat's contractor Ariane II,was working at the same time for "Lot 2: Establishment of a datasetfor use in agricultural sector modelling for Acceding Countries"and provided supplementary data from NSOs in NMS. This ARIANE IIinformation is still used if no data are available from Eurostat.Specific update files are included for Romania, Bulgaria, Croatia,the FYROM and Turkey, defining revisions on data from Eurostat (seesection 3.1.2).
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
16
A data check for Bulgaria suggested to revise the data for themeat sector (slaughtered tons and heads) based on FAO, because:country data from the Excel sheets are only available for the year2004; FAO reasonably fits to the country data in 2004 (confirmingreliability at least for 2004); FAO data are available for moreyears than New Cronos. National data are used for Romania both inthe meat and in the milk sectors. National information is availablefor 2004 and 2005 in the meat sector for all cattle components andFAO data are used because: FAO data fit to Eurostat's data from1995 to 1998; FAO data cover more recent years; FAO data reasonablymatch national data for 2004 and 2005. National data and FAO seriesare combined by using a correction factor equal to a ratio fromnational information to FAO data for 2004 and 2005. A similarprocedure is applied to the sheep and goat sectors. Moreimportantly are the corrections in the raw milk balance (RMLK)based on national data, which sometimes appeared to be morecomplete and plausible than Eurostat in this area, even though thenational balance positions are only available for 2004.Essentially, national information is used to disaggregate thoseEurostat series considered plausible. This holds for the Eurostattime series for total human consumption of raw milk ("HCOT.RMLK")which is disaggregated with data for direct sales ("HCOM") and homeconsumption ("CONS") as collected for 2004 in the countrysheets:
A similar procedure is applied to the other elements of the rawmilk balance ("BUTF", "CHSF", "OPRF", "INTF", "PRCM", "LOSM")which, according to New Cronos, are estimated by applying thenational share to the production of raw milk. Additional detaileddata were collected for Croatia, the FYROM and Turkey insupplementary Excel sheets (see section 3.3) from NSOs and the MAs.Data were compared with already available data at Eurostat andupdate files similar to Romania and Bulgaria were included (seesection 3.1.2). Data from Eurostat for the other WBs are currentlyignored. All data are collected in special questionnaires fromNSOs, the MA and expert knowledge in the countries and a separatefile is included for each country.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
17
The module balkan_zpa1prag.gms' prepares the WBs data (from bothcountry groups) to match the COCO definition. These adjustmentsfall into three categories:
1) Similar to EU-27 MS there are many case-by-case adjustmentscorrecting different scaling and definitions:
The production of eggs must be switched from mio eggs to1000t;
Meat prices must be converted from live weight to carcassweight;
Specific corrections for all WBs are necessary for wine andfruits;
Additional corrections are specific to single countries (forexample: implausibly high slaughter weights for sheep and goats(> 60 kg) in Montenegro, fixed with average slaughter weightsfrom sister country aggregates).
2) In many cases, market balances are simply incomplete. As afall back solution, domestic demand is calculated from productionand net trade and disaggregated with shares taken from a sistercountry aggregate (Romania, Bulgaria, Greece, Slovenia,Hungary):
In a first step, the sub-module "aggreg_for_balkan.gms" buildsthis aggregate from COCO results of neighbouring MS (EL000, SI000,HU000, RO000, BG000);
Trade data are frequently missing in the WBs, such that FAO dataare included as a fall back solution to make sure that trade dataare given;
Production of oilcakes and sugar is estimated from raw products,if missing, using the sister country aggregate processingcoefficients;
The production of milk products is estimated from processingcoefficients in Serbia which has a quite complete series;
Finally, missing market balance positions are estimated usingshares from the sister country aggregates.
3) Price information is also completed relying on theneighbourhood country aggregates, in a four step procedure:
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
18
Step 1: For empty time series, prices from the neighbourhoodaggregates are scaled with a price ratio of lead products (cereals,milk) in the same year;
Step 2: For gaps, prices from the neighbourhood aggregates arescaled with the ratio from other years (potentially created throughlead products);
Step 3: For the aggregate "CERE", an average price is firstassigned to other cereals and secondly the group price iscalculated;
Step 4: For all other groups, prices from the neighbourhoodaggregates are scaled by a ratio of known group elements.
As regards trade, FAO data are included as fall back informationfor all MS. Trade is often a crucial point in data selection,because it is very volatile and difficult to estimate with smoothtrends. The advantage of FAO is that trade is usually well coveredin FAO. Trade series have been completed to 2005 based on aFAOSTAT2 project6.
The module "calc_fao.gms" aggregates FAO data to match COCOdefinition.
3.2.3 Completion with additional sources For the ten NMS ARIANEII data are used in case New Cronos data are still missing. Whencompleting a New Cronos time series (ZPA1 and COSA) with ARIANE IIdata (and similarly in other cases) a correction factor is used, aslong as there is at least one overlapping year. The factor issimply the quotient of the sums from ARIANE II data and ZPA1 datafor overlapping years. If there are no overlapping years tocalculate a correction factor, a decision on which series to usehas to be made. Longer series, with more recent data are, ofcourse, preferable to shorter series with older data. To undertakethis decision according to a clear rule, each year has been given aweighting factor representing the presumed reliability of a datapoint which evidently declines for older data. The year 2005, ischosen as a reference point in the calibration of the weightingfunction (i.e. weight = 1 for 2005) because it is the most recentyear of the database. The mechanical decision rule determines theweighted sum of observations according to data availability fortime series from ARIANE II or ZPA1/COSA. From a conceptual point ofview, the weights reflect the contribution of a particularobservation to the utility function used for the selection betweentwo series:
=>=)t(Tt
2)t(Tt
121
)t(u)(U)t(u)(U2Series1Series XXf
where Ti(t) is the set of years with non-zero observations Xitin series i. The weighting function giving the utilitycontributions has been specified as u(t) = 0.85(2005-t) decliningfrom u(2005) = 1 to a value of u(1990) = 0.09, that is to a smallnumber, when the economic (and statistical) transition processbegan. This should reflect the assessment indicating that old dataare of very limited reliability for setting up a modellingdatabase, given that the most recent year 2005 has been selected tobe most relevant for further use. Having specified a form for u(t)this will determine which 6 Consistency and Completeness of SUA andTrade Matrices Based on Entropy Estimators, Division: ESSD, Index:0600163
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
19
data source is to be preferred according to a transparentcriterion linked to data availability. To come back to our specificexample, the rules can be summarized as follows:
1. If there are overlapping years:
use ARIANE II adjusted by a conversion factor before and afterthe last known data;
for years in gap, use ARIANE II only if the correction factor isclose to one (gap filling is easy, even without ARIANE IIdata).
2. If there are no overlapping years:
if ZPA1 is inferior according to the weighted sum ofobservations, delete ZPA1 series;
if ZPA1 is empty (after deletion or right from the beginning)use ARIANE II.
3. If there are overlapping years:
use ARIANE II adjusted by a conversion factor before and afterthe last known data;
for years in gap, use ARIANE II only if the correction factor isclose to one (gap filling is easy, even without ARIANE IIdata).
4. If there are no overlapping years:
if ZPA1 is inferior according to the weighted sum ofobservations, delete ZPA1 series;
if ZPA1 is empty (after deletion or right from the beginning)use ARIANE II.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
20
Table 9 Table of weighting factors
The years define the rows on the table. The first column"weights" includes the weight for each year (after 2005 the weightfor all years is 1). The following columns show the sum of weights,for a specific number of observations. The sum of weights for fiveobservations can be found in column "sum5", e.g. line "1996",column "sum5" is the value (0.86) from1992 to 1996.
Example: Three ARIANE II observations are available from 2001 to2003 and two ZPA1 observations are available from 2005 to 2006. Thesum of the three weighting factors for ARIANE II results in 1.86and for the two ZPA1 data in 2.00. This will cause ARIANE II datato be considered less reliable than the ZPA1 series which onlyincludes two observations, but more recent than the three fromARIANE II. For all MS and the WBs usable production is a veryimportant item. Missing data will cause severe troubles. Therefore,completion of data from the ZPA1 market balance statistics (modelcode "USAP") has also been tried with quantity information givenfrom the agricultural account statistics (model code "EAAQ"). Acorrection factor is
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
21
used, calculated from overlapping years, for this procedure. Inthe New Cronos selection from spring 2007, only the FYROM has anydata in the national account statistics, however, data forquantities are still missing. Therefore, this completion step isirrelevant for WBs in 2007. Nonetheless, the following codefragment illustrates the general COCO principle that differentsources should only be merged if a correction factor can becalculated.
The above extraction from the GAMS file shows the completiondata applied to usable production of crops (collected in setASS_ROWS_T), where a time shift is needed to combine both domains.In a first step, a correction factor is calculated as follows:
Correction factor = sum (t, source1) / sum (t; source2) for allt, where data are available for both sources
Missing data for years in the first source (ZPA1) are filled bydata from the second source (COSA), multiplied by this correctionfactor (or divided by its reciprocal value). Finally, FAO data areused for data still missing in ZPA1. The same procedure as forARIANE II data is applied, including correction factors andweighting table. FAO is the most important source to complete tradedata. If trade data are not available from FAO, they are estimatedby a weighted moving average: weights decrease when distance to thegaps (the years with missing trade data) increases. Domestic usecan be calculated7 from imports, export and usable production. Ifonly domestic use is given for some products, the sub-positions,such as industrial use, processing, human consumption, feed onmarket, total seed and total losses are allocated with theaverage
7 Extraordinary jumps are also deleted and completion of timeseries is left to subsequent trend estimations. To avoidinconsistencies in aggregates, the groups for exports, import anddomestic use are recalculated.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
22
shares in ZPA1 data for other years, from the same country. As afall back solution, the average shares are used for othercountries. As data for oilseeds are critical for all countries, theimplied processing coefficient is checked for plausibility. If thenational coefficient is lower than 60% or above 150% the averagecoefficient for all EU-15 MS, the data for usable production of thecountry are corrected by multiplying the processing data with theaverage EU-15 coefficient. Domestic use and all sub-positions aresubsequently re-calculated.
3.2.4 Assigning data to database array So far data has beenhandled by domains. The next parts of COCO involve moving datacomputed for all domains (ZPA1, COSA and PRAG) to a single GAMSarray "data" for subsequent completion steps and consistencycalculations. First, the prices from the PRAG domain are copied.Only the prices for sub-products are available for cheese andconcentrated milk and have to be aggregated. Prices for apples,pears and peaches are combined for model code "APPL". A credibilitythreshold, in terms of the average price for all MS, is defined forall prices to avoid use of implausible prices. Separate modules forcrops, milk, animals, economic account positions, residualpositions and crop yields are applied to assign input data to the"data" array. Additionally, arrays for lower and upper limits, forsubsequent model calculations in COCO part 2, are also assigned.The approach of this section did not change, in principle, only insome details for the selection of spring 2007.
Sub-module coco_crops Levels of area, production data and marketbalance positions from Eurostat ZPA1 domain, are assigned by thissub-module.
Area levels
The aggregates fodder on arable land ("FARA"), annual greenfodder ("FANG") and perennial green fodder ("FPEG") are deleted if,at least, one of their components is available. Inappropriateaggregation has been frequently found in past experiences withEurostat data such that aggregates are added up, if possible, fromthe sub-components (e.g. aggregate FARA is added up from thesub-components FANG and FPEG). In the model database, the item"FARA" includes additional data for root crop levels, which have tobe added to the data selected from ZPA1 for "fodder on arableland":
For all single products and groups of products, the crop areasfrom Eurostat's production statistics are copied to the "data"array. Data from Eurostat's land use statistics are the second bestchoice for missing area levels. For "ROOF" (fodder
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
23
roots), "OFAR" (other fodder on arable land), "GRAS" (grassing)and "TAGR" (table grapes) levels of the components are added up asfollows: ROOF ROO1, ROO28 OFAR TGRA, FCLV, FLUC, FPGO, fa*gO GRASPMEA, PPAS TAGR VINY TWIN Still missing years, in the time series,for "FARA" are added up by: FARA MAIF, ROOF, OFAR Adding up is onlypermitted for years where all components ever observed areavailable.
Gross production
Then, the gaps in the time series for "LOSF" (losses on farm)and "SEDF" (seed on farm) are closed by estimates from the averageshare of the usable production. The model code "GROF" (grossproduction) can be assigned as follows: Products with marketbalance USAP (t+1) (balance sheets) Remaining products GROF(production statistics) Missing data in the time series for cropsare filled up by multiplying the "GROF" from production statisticsby a correction factor (see above for definition of correctionfactor) from balance sheets and production statistics. This impliesthat a mixture of balance sheets and production statistics isaccepted when a correction factor is calculated. A special case isthe production of "OOIL" (other oils), derived as residual from oilseeds, rape, sunflowers and soya seed. The calculation of productsfrom grass and grazing activities ("PMEA", "PPAS", "TGRA", "FCLV","FLUC", "FPGO", "fa*gO" and "MAIF") for the gross production isdivided into three steps:
calculate yield if "GROF" and "LEVL" are available in ZPA1 ormay be derived (e.g. as residuals) from others;
fill up missing yields with the help of average yields in otherEU countries;
calculate "GROF" as a product from yield and area. Finally,gross production for "ROOF", "OFAR", "FARA" and "GRAS" is added upfrom the components (for explanation see: levels) and "FORA"(fodder from arable land) is established as the sum of "FARA" and"GRAS".
Balance positions
All balance positions for crops and animals, except milkpositions, are assigned to the "data" array. Specific treatmentsare necessary for fruits, textiles and olives. for fruits, best fitto production, data is obtained by adding up the balance sheets forfruits form those for fresh fruits, plus nuts, minus table grapes.Balance sheets for textiles are added up from those for flax andcotton. The processing of olive for oil, is calculated fromproduction (from the production statistics) plus imports, minusexports. For all products (except fodder), where only domestic useis available from the balance sheets, it is assumed that allproduction is used for human consumption. On the contrary, domesticuse where not already given, is added up from its components. Ifstill missing, it is calculated backwards from usable production,imports, exports and stock changes. Finally, in case of stillmissing data for usable production and 8 See Annex 1 for furtherdetailed information on model codes.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
24
gross production, an attempt is made to calculate this fromdomestic use and trade data.
Sub-module coco_milk Data for dairy products and milk areassigned in this sub-module from Eurostat's ZPA1 domain.
Production and market balances of milk and dairy products
Gross production of "COMI" (cow and buffalo milk) is the sum of"CMLK" (cow milk) and "BMLK" (buffalo milk). Gross production of"SGMI" (sheep and goat milk) is the sum of "EMLK" (ewe milk) and"GMLK" (goat milk). Data are only added up for the years where,either components are available or one is irrelevant. First,deliveries to dairies ("PRCM") of "COMI" and "SGMI" are derivedfrom the processing volume of raw milk according to farm balance.The second best solution is to add up the data for processing fromthe components in the dairy collection data (e.g. collection of"CMLK"+"BMLK") and finally, as the third best solution, in case ofstill empty series "PRCM", deliveries are set to production minus1%. The implied low subsistence share of 1% may be an acceptablefall back solution for the EU-15 MS but, it is clear thatparticular efforts would be needed in case of empty delivery seriesin the WBs or NMS. Fortunately, there is no country, not even amongthe WBs, where no delivery data were available at all or estimated.Usually, there are even data to disaggregate the non deliveredparts of raw milk into direct sales (e.g. HCOM.COMI), feed use(INTF.COMI), farm cheese, butter and other processing products(INDM.COMI) and finally losses and home consumption of liquid milk(LOSM.COMI). For WBs, this disaggregation often had to be estimatedwith shares taken from the sister aggregate country mentioned insection 3.2.2 (Romania, Bulgaria, Greece, Slovenia, Hungary). Thebalance sheets for "COMI" and "SGMI" are formed from the balancesheet for "RMLK", using a ratio of cow and buffalo milk, sheep andgoat milk to the whole milk production. The marketable productionfor this aggregate milk, at the dairy level, is set to the sum ofthe processing volumes from cow and buffalo milk, sheep and goatmilk (from the farm balance). Whereas production data anddeliveries to dairies may be distinguished into COMI and SGMI, thedairy statistics on derived products obtained or associated marketbalances do not permit such distinction. As a consequence, thedairy sector is treated as if all raw milk from cows, sheep etc.was collected and merged into single raw milk at dairy ("MILK").The marketable production for this aggregate milk, at the dairylevel, is set to the sum of the processing volumes from cow andbuffalo milk, sheep and goat milk (from the farm balance).Processing is calculated from the use of milk at dairies for theproducts "FRPR" (fresh milk products), "MANU" (manufactured milkproducts), "OTUS" (other uses: skimmed milk, buttermilk and whey)and exports of milk (giving a fourth potential source forprocessing of raw milk). Finally, the balance sheets for thesecondary milk products are taken directly from the data selected.If the usable production is missing, it is calculated from thedomestic use, imports, exports and stock changes.
Fat and protein content of milk products
This section of the COCO code has been revised, in the contextof reviewing all dairy related data. The content of milk productsis initialised using two types of information: raw data on fatcontent of dairy products (and protein content for raw
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
25
milk) from Eurostats statistics obtained on milk collection andproducts; default technical coefficients for the content of milkproducts, in terms of milk fat and protein (this is the onlyinitial information for protein, apart from raw milk, wherestatistical data are available). The initial information on the fatcontent of dairy products is provided complete and reliable inthree steps: taking the simple ratio of production of milk productsmeasured in tons of fat and in product weight; discarding contentsconsidered implausible, compared to the standard technicalinformation as follows:
implausible data are those for dairy products with contents <40% of the standard coefficient or > 250% of the standardcoefficient. These thresholds are a result of the single dairyproducts hom*ogeneity. Stricter thresholds exist for butter withonly 5%, because butter is quite hom*ogenous; completing series witha moving average or the technical coefficient if the series wasentirely empty.
Sub-module coco_anim Data for animal activities and products areassigned in this sub-module. All assigned data are from Eurostat'sZPA1 domain.
Slaughtered animals (heads and tons)
For tons of slaughtered meat of the main animal categories("IPIG", "ILAM", "ICAT" and "ICHI"), the usable production ("USAP")from the balance sheets is assigned as the first best sourcebecause this is likely to be consistent with market balances. Thesecond best source for produced meat is taken from the slaughteringstatistics. In this case, and for cattle categories ("ICOW","IBUL", IHEI", "ICAL") the data are taken from the slaughteringstatistics ("SLGT") and corrected by the relation betweenslaughtered quantities of individual cattle categories (fromslaughtering statistics) to total slaughtered quantities for cattle(from balance sheets). Missing slaughtering data for "IPIG", "ILAM"and "ICHI" are calculated by the same procedure as for the cattlecategories. Export and imports of live animals expressed in carcassweight for all animal categories ("ICOW", "IHEI", "IBUL", "ICAL","IPIG", "ILAM", "ICHI", "ICAT" and "LAMB") are derived from theslaughtering statistics. At the same time, export and import oflive animals for total beef production are taken from the balancesheets for
http://dict.leo.org/ende?lp=ende&p=eL4jU.&search=inhom*ogeneity
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
26
beef. For pork, poultry meat, sheep and goat meat, total importsand exports of live animals from the slaughtering statistics areused. Finally, total slaughtered heads for all animal categories(e.g. "GROF.IPIG") is set to the sum of all slaughtered heads, plusexported heads, minus imported heads. Accordingly, the productionof meat is calculated ("MEAT.IPIG") by the sum of all slaughteredtons, plus exported tons, minus imported tons.
Activity levels and slaughter weights: cattle
The gross production of beef is equal to the production of meatfrom cattle, as defined above. For dairy cows and suckler cows, thelevel of the herd size is the average of two selected countings(May-June and December counting). The input coefficient for dairycows ("DCOW.ICOW") and suckler cows ("SCOW.ICOW") reflects thenumber of slaughtered heads, in relation to the total herd size ofcows. The default value for missing time series is 0.2, assumingthat the live span of a cow is five years. The input coefficient ismultiplied with the herd size to assign the total slaughtered headsof cows. The output coefficient in terms of young animals issubsequently the quotient from meat production and totalslaughtered heads. For heifers and bulls for fattening, the herdsize is equal to the number of slaughtered heads. If theslaughtered heads series is empty, 45% of the total slaughtering(net of cow and calves if available) is used as a default value,based on the assumption that from 100% of slaughtered cattle, 10%9might be calves and the rest (net of cows) is divided into halffemale and half male. Heifers for raising will replace the dairyand suckler cow herd size, therefore the herd size has to reflectthe herd sizes for dairy and suckler cows for the next two years.The extraction below shows the implementation in the GAMS code.
Subsequently, the number of heifers needed as input("GROF.IHEI") for each year is defined by the sum of the herd sizesof heifers for raising and heifers for fattening. The herd size offemale calves for raising, for the current year, is set to thenumber of heifers of the following year. The number of young bullsraised, is equal to the number of adult male cattle in the nextyear. The herd size for male calves for slaughtering is calculatedby taking the maximum of: 10% of all slaughtered heads of allcalves (lower security bound); total calves slaughtered (incl.export, minus imports) plus the herd sizes from male and femalecalves for raising, multiplied with a factor of 0.51 (typicalpercentage of male calves which are known to be slightly morefrequent than females), minus the herd size of male calves forraising. For male calves, the same procedure using a factor of 0.49is applied. The extraction below shows the part for female calves.9 This percentage has been considered a reasonable default value inthe range observed in important MS, e.g. in the 2000 the percentagefor France was 34%, for Germany 9,7% and for Spain 4.6%
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
27
Consequently, the total number of calves needed, as input foreach year, is defined by the sum of the herd sizes of calves forraising and calves for fattening. As calves are slaughtered withinone year, the calves output is equal to the slaughtered heads. Theoutput coefficient of dairy cows in terms of calves is assigned bythe quotient of the output of calves and herd size of cows. Theoutput coefficients of calves for fattening in terms of beef10 aredefined as the corresponding production of beef divided by thenumber of slaughtered calves. For the other animal categories, thebeef outputs are calculated from the meat production divided by theherd sizes. Finally, the total production of beef is added up fromall products of the output coefficient and the herd size. As theherd size for dairy cows is now available, the yield for cow milkcan be calculated as usual from the ratio of production over theactivity level.
Activity levels and slaughter weights: sows and pigs forfattening
The herd size of sows is the average number of sows according tothe four possible annual counting (April, May/June, August andDecember). The herd size of pigs for fattening is assigned to theslaughtering of all pigs minus the slaughtering of sows. The outputcoefficient per sow is the number of slaughtered pigs plus thestock changes on farm divided by the herd size. Thus, the totaloutput of pigs (= pigs born) is the product of the outputcoefficient of sows and their herd size. The input needed is theoutput of the current year, minus the difference in herd size ofthe next year and the current year. The total production of pork isset to the meat production from pigs. The production of pork frompigs for fattening is calculated as a difference from the totalmeat production and the production of pork from sows, assuming thata sow produces 120 kg of meat.
10 As the distinction between veal and beef is not supported inmany statistics, both in COCO and in CAPSIM there is only theoutput beef (including veal).
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
28
Activity levels and slaughter weights: laying hens and poultryfor fattening
The hen herd size is assigned by the herd size of the last yearDecember counting. The output coefficient for eggs from hens is thegross production of eggs, assigned from usable production from thebalance sheets, divided by the herd size. The input coefficientreflects the assumption that 80% of the laying hens are replacedeach year. The herd size of poultry for fattening is the differencefrom all slaughtered poultry and slaughtered hens. The outputcoefficient of hens is calculated from slaughtered poultry andchange in stock to next year. The input coefficient is thedifference between the output coefficient and the change in herdsize. The total production of poultry meat is set to the meatproduction of poultry, derived from slaughtering, plus exports,minus imports.
Activity levels and slaughter weights: sheep and goats
The herd size of sheep and goats for milk is the average numberof animals according to the two possible annual countings (May/Juneand December). The number of slaughtered lambs (sheep and goats) isthe total slaughtering, minus the slaughtering of adults. Theactivity level of sheep and goats for fattening is set to thenumber of slaughtered lambs. The total output of sheep and goatmeat is set to the meat production.
Sub-module coco_eaa In this sub-module, EAA data from EurostatsCOSA domain are assigned. The corresponding unit values arecalculated according to EAA values and the quantities. As a fallback option, selling prices from the Eurostat PRAG domain areadjusted and completed to serve as a fall back solution if needed.EAA positions In a first step, all positions for all products ofthe EAA, except price indices, are assigned to the "data" array.For a number of aggregates, special assignments are needed: CERECERE PARI INDU INDC SUGB PULS + OLIV OILS OILC + OLIV APPL APPS +PEAR + PEAC FRUI FFRU + TROP Kidney beans are a special case. Thisproduct has to be extracted from vegetables and added to pulses tobe in line with the grouping in the market balances.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
29
For all missing data at producer price, the values at basicprice minus subsidies, plus taxes are taken. The rule, vice versa,is applied for missing data at basic price. The COSA domain offers,in addition, some values unrelated to the market balances, such asmaintenance of machinery and buildings. They are copied to the"data" array without any adjustments.
Unit values at producer price
First, the unit value is calculated as a quotient from the valueat producer price and the quantity as selected from the COSAdomain. For the correction of the calculated unit value, the secondlargest and second smallest unit value from all MS is identifiedand an average unit value is calculated from those unit values,which are lower than the second highest and greater than the secondsmallest unit value. Unit values deviating by more than a certainthreshold (= 100%, permitting a fairly high range of variation)from this average are deleted:
The same procedure is applied to unit values at basicprices.
Fall back option: adjusted selling prices
To serve as a fall back option for the EAA unit values, theprices as selected from the PRAG domain are corrected with theaverage ratio between producer prices and selling prices. To takecare of gaps, the time series are filled with the moving averageapproach introduced above in the context of trade information fromFAO. Finally, if price indices are still missing for single items,those from product groups are used.
Energy
If there are data missing on the use of energy several fall backoptions are prepared. If EAA positions for energy are missing,first the average relation between the specific
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
30
position and total energy use of the particular MS is used. Ifstill missing, the average relation between the specific positionand the total energy use for all MS in that year is used. The pricefor gas is used to present the gas consumption in mio l, which isassigned to gross and net production of gas. The gross and netproduction of all other non physical items from EAA, is calculatedby the quotient of EAA value and price index.
Sub-module coco_resid This sub-module calculates residuals fromthe given data for aggregates and sub-positions. First, it isassumed that residuals can be calculated for all product andactivity groups. A corresponding set defining those groups isinitialised as GRC(MS,groups,item) = YES (e.g.GRC(MS,"CERE","LEVL") = YES) initialise the calculation of theresidual levels for the group "CERE" in all MS). Removed fromcalculations are (and GRC(.) is set to no):
All years, where at least one time series of a componentincludes gaps;
Groups, where it is not possible to construct the group positionfrom the components;
Groups, where Eurostat data are smaller than the sum ofsub-positions. Subsequently, the residual level is defined as adifference between the group level and the sum of individual crops,where CROPC_TO_GRPC is a cross set, linking column groups to singleactivities.
The farm and market balance positions for residual outputs(OCER, OOIL etc.) are calculated correspondingly as the differencebetween group positions and sum of individual productpositions.
Sub-module coco_cropyields This sub-module calculates the yieldsfor all crop activities. Yields are evidently calculated for eachcrop activity by dividing the gross production by the productionlevel for this activity. Additionally, a Hodrick-Prescott (HP)filter is applied for the smoothing out of problems with yieldsfrom activities with small production areas. This is a smalloptimisation program with tight bounds around observed productionand area data ( 100t or 100 ha). The HP objective consists inminimising second differences which strongly penalise peaks in thedata, as frequently encountered (partly due to rounding errors)with small areas or quantities. The tight leeway around observedvalues is irrelevant for moderately important crops in the sensethat, the result will be almost identical to the original data. Forunimportant crops, however, the HP filter term will operate to somesmoothing of peaks in the data and thus, in general, to moreplausible yields for these crops11.
Sub-module cor1_gras In most countries grass is the mostimportant crop in terms of area use yet, often the data on grassareas and production are one of the weakest parts of cropstatistics. When relying solely on statistical data, the COCOdatabase frequently showed
11 For example in France, in 2000, 100 ha only represented0.002% of the soft wheat area, but 100 ha of tobacco represented 16% of the total area, as tobacco is irrelevant in France. Thisirrelevant item will be those where unrealistic yields will befrequently found and where deviations from Eurostat data will beacceptable.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
31
unbelievable grass yields in some MS. This sub-module assignsgrass yields, based on expert knowledge, to be used as prioriinformation together with statistical data in part 2 of the COCOroutine. The key information is expert data12 on typical grassyields in dry matter for 2002 in all EU-27 MS and WBs. To convertthis expert information, for a single year, into expert time seriesfor grass yields, the expert data for 2002 are linked to the yieldsof activity aggregate cereals (stored temporarily here on"CERE,GRAS"), assuming that long run yield growth and yearlyfluctuations may thus be approximate.
The yields for pasture, meadows and other fodder on arable landare adjusted accordingly.
3.3 SUPPLEMENTARY DATA SELECTION The section focuses on thecontent and organisation of the supplementary data organised inExcel files. The standard software used may facilitate use of thesedata beyond the immediate purpose as input for COCO. Supplementarydata were collected for: Bulgaria and Romania, Croatia, the FYROM(and as far as possible the other WBs), and Turkey. For Bulgariaand Romania most data are already available from New Cronos andconsidered to be of satisfactory quality. This does not hold formilk and meat production such that supplementary data werecollected for these sectors of agriculture. The method used wasquite different for Croatia, the FYROM and Turkey. The bulk of thedataset, for these countries, was collected from national sources.This was prepared with a set of data collection tables designed inthe initial project phase.
12 These were estimates worked out in September 2006 by OeneOenema and Gerard Velthof from Alterra, Wageningen, in the contextof a service contract for DG-ENV (Integrated measures inagriculture to reduce ammonia emissions , No070501/2005/422822/MAR/C1) with the participation of EuroCARE.
The Common Agricultural Policy SIMulation (CAPSIM) Model:Database for Agricultural Sector Modelling
32
3.3.1 Bulgaria and Romania The data tables collected for Romaniaand Bulgaria are structured as follows:
Content
Romania milk sector
Table 1 Dairy farm structure Table 2 Milk util
![The Common Agricultural Policy SIMulation (CAPSIM) …publications.jrc.ec.europa.eu/repository/bitstream/JRC44285/jrc... · EUR 23381 EN - 2008 The Common Agricultural Policy SIMulation - [PDF Document] (2024)](data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/h8AAvMB+NzmkbcAAAAASUVORK5CYII=)