CAPRI (Common Agricultural Policy Regional Impact) is an agricultural sector model covering the whole of EU27, Norway and Western Balkans at regional level (250 regions) and global agricultural markets at country or country block level.
The main reason was the fact that there was no economic model for the agricultural sector available
The first operational version was released in 1999. Since then, manifold improvements were implemented in CAPRI.
The transparent link between regional or even farm type aggregate programming models covering EU's agriculture in a rather differentiated way with a complex global model for trade in agricultural products. The spatial downscaling part to 1×1 km, which covers crop shares, yield, stocking densities, fertilizer applciation rates and detailed environmental indicators
Simulation results cover areas cropped and herd sizes along with output and inputs coefficients and income indicators for each agricultural activity and each region; prices, supply and demand positions at country level; environmental indicators (gas emissions, N,P,K balances) at regional level; producer and consumer prices, supply and demand positions as well as bilateral trade flows with attached prices, transport costs and tariffs globally for each trade block; the costs of Common Agricultural Policy broken down to individual of policy instruments; a welfare analysis for all countries or country blocks in the system and many other interesting aspects of agriculture.
To simulate and compare ex-ante for a medium term horizon (8-11 year) impacts of different set of agricultural policies.
A full-flegded model documentation can be downloaded here. Any remaining questions may be sent via e-mail to a member of the CAPRI network, if they cannot be answered with the FAQ list.
CAPRI users both work in research institutions and EU Commission services.
The majoritiy of publications based on CAPRI can be found here.
CAPRI features a Grapical User Interface which allows to generate tables, graphs and maps from results. On top, results are stored in a data base and can from there exported e.g. to spreadsheets.
Not necessary. There a good chances that somebody in the network as already answered your or similar questions with simulation runs, and you may only need to download these results to generate from there your own tables, maps and analysis. You may be lucky in scanning the list of publications based on CAPRI and contacting the author(s).
Yes, you may. CAPRI is written in GAMS, and all you need is a GAMS license to change the code. Such changes are typically necssary for any serious analysis when new policies are analysed. The version system hosting CAPRI allows to set up “branches” for such projects.
However, please keep in mind that you should not automatically assume to find support from experienced members of the CAPRI network in case of such changes. Equally, please reflect that your changes are not automatically ported to a new master version. In doubt, contact a member of the CAPRI network before any changes.
That has been done in past, but requires manyfold changes if that region is to be included in the supply part. To do so, availability of data in harmonized definitions is typically the main bottleneck. It is far less demanding to change the regional break-down of the international trade model as here uniform sources at global level are used which are in most cases available at single country level.
Yes and No. In the overall modelling system, the global trade model is responsible for simulating market clearing prices, so that prices are endogenous. However, during solution of the regional supply models, prices are fix and given, and the farmers react as price takers. An iterative link between the supply and market models ensures convergence between the prices used in the supply models and the ones generated by the market models.
There are several reasons. Firstly, the different modules can be maintained and improved indepedently. Secondly, different methodological solutions can be chosen. The supply side is based on explicit profit optimisation under constraints, which has e.g. adventages when capturing the effect of policy instruments as quotas or set-aside or linking it to engineering data or results from bio-physical models. The international trade model is a square system of equations based on behavioural functions, which allows to capture many products, market and interactions simultaneously. Thirdly, it would be most probably technically not feasible to solve the whole system as one supermodule. And not at least, you may use the supply side of CAPRI indepedent from the market models for analysis, even running the system for single countries, only.
CAPRI uses a template on the supply side, i.e. each model has the same structure (equations, variables) but its own set of parameters. For the market model, the same functional forms for the behavioural functions are used for each of the regions, again a template approach.
Yes and No. As the model is spatial, i.e. featuring bi-lateral trade flows with their own import prices, market clearing is not based on central anchor prices - typically called world market prices - as in net trade models. However, average world producer or consumer prices are calculated after simulation runs.
Not in the current master version. There was a version some years back, but the model was never used or able to endogenously simulate the switch between production systems. Further on, the model does not differentiate between irrigated and not irrigated agriculture. Each major arable crop is however present in an intensive and extensive variant.
Yes, each crop production activities in each region is presented by a low and high yield variant which substitute against each other. On top, yield results shown for higher regional aggregates (Member States, Regions) are a weighted average over the NUTS II regions, so that yields above regional level may change even if no yield changes are simulated per NUTS II region. And last not least, yields for some major crops are shifted as a function of elasticities and output prices during iterations between supply and market part.
The typical simulation horizon with CAPRI is mid-term (about 10-15 years ahead), but depending on the specific scenario assumptions the model is also suitable for long-term projections and simulations in the context of climate change.
No, the experimental version of a recursive-dynamic system was abandoned. The model is comparative-static.
The current master version covers NUTS II regions and farm tyes. After a serious effort in 2009/2010, the farm type version is now fully operational. It breaks down the consistently each NUTS II region into up to 9 most important farm type differentiated by farm size and specialization plus a residual type.
A so-called aggregate programming model for each region which explicitly maximizes profit under constraints. These models are solved independently from each other at fixed prices for outputs and inputs. They cover about 50 production activities. The models are non-linear both in constraints and the objective function.
Arable and grass land - linked to the land supply curve -, set-aside, production quotas, requirement constraints for feeding of animals, nutrient constraints (N,P,K) for crops. Emission standards can be added on demand.
The current master comprises a new module where total agricultural land can shrink and expand, and substitution between arable and grass land is possible. Arable land can idle as fallow land, and the intensity of grass land use can change. A standard sensitibvity analysis at the end of each run reports the marginal impact of the land supply curve.
The profit depends on revenues from selling products to the markets, revenues from premiums allocated to production activities, the costs of intermediate inputs based on the fixed and given input coefficients, the costs of the endogenously determined feed quantities (feed input coefficient are not fixed, but vary depending on feed prices and requirement constraints), and finally, the values of cost function depending on the actual mix of activities. For sugar beet, a special sub-module is based on maximizing expected revenues to cover yield uncertainity and the A,B,C sugar beet regimes of the EU.
Production quotas for milk; A and B selling quotas for sugar beet in conjunction with A,B,C, prices; set-aside; the different premiums of the Common Agricultural Policy. Recently, major programs from Pillar II of the CAP (Less Favourate Area support, Natura 2000 payments to agriculture, Agri-Environmental Measures) had been roughly integrated.
Activity levels (cropped areas of a low and high yield variant, animal herds), quantities of feedingstuff and intermediate input purchases, quantities of outputs sold, feed input coefficients, mix of organic and anorganic fertilizer per crop, and implicitly, average regional crop yields.
The premium module allocates the different premium schemes of the CAP to individiual production activities, taking into account at the one hand EU, national and/or regional ceilings (maximum guaranteed areas or herds, budget ceilings), and at the other the definition of the premium payment mode (paid per ha/head/salughtered animal/quantity produced/reference yield) etc.. The model covers in the range of 30 different schemes stemming from many different EU regulations. Entitlement of the so-called “Single Farm Premium” of the CAP are endogenously integrated in the supply module.
The model treats de-coupled payments as a subsidy to land. If the subsidies per ha for all crops would change by the same amount, solely the shadow price of land would change as long as the entitlements are binding. There is currently no consideration of risk or wealth effects of these subsidies.
A so-called spatial multi-commodity model. Supply and demand for a large number of agricultural raw and processed are determined simulaneously with market clearing prices. Spatial means covering bi-lateral trade flows.
A CGE model (Computable General Equilibrium Model) covers firstly all products and actitivities in an economcy, not just agricultural ones, and covers factor markets as well. Secondly, supply is determined in conjunction of a production function and first order condition for profit maximisation, whereas in the CAPRI market model, supply functions are used with prices as driving variables (dual approach). Thirdly, CGE models are typically less differentiated regarding agricultural products, and less explicit in modelling policy instruments. Finally, agricultural MCM such as CAPRI are defined in physical quantities and related prices, whereas CGEs typically only cover volumes and price/quantity indices.
The market model endogenously determines bi-lateral trade flows based on the Armington assumption.
Consumers differentiate products by origin and their willingness to pay depends on the product's origin, e.g. wine from Europe differs in quality and price from US wine. Import shares depend on import prices.
Specific and ad-valorem import tariffs, EU export subsidies, sales to and purchases from EU intervention stocks, certain preferential trade agreements, a greater number of Tariff Rate Quotas, flexible levies for cereals and the minimum entry price scheme for fruits and vegetables of the EU.
Due to the bi-lateral trade presentation, the number of variables and equation will increase quadratically with the number of regions. Therefore, the CAPRI market model has “clustered” the countries in some regions (e.g. the EU15 Member States) to trade blocks. The model only capture trade flows, transport costs, tariffs, export subsidies and import prices at the level of these trade blocks. However, a trade block may be broken down to individual countries with own behavioural equations.
It covers about 40.000 equations and variable which are solved simultaneously. Some of the equations are also highly non-linear. Nevertheless, with good starting values, it can be solved in a few seconds.
The parameters for the CAPRI market model had not been estimated by the CAPRI team, but are borrowed from other models or publications.
The supply, processing, feed and final demand function in the CAPRI market model are flexible in the sense they comprise enough parameter to be calibrated to any regular set of own and cross price elasticities. Regular means that the parameter set is calibrated as to be in line with the assumption of profit maximising producers and utility maximising consumers. For insiders: that captures curvature of the Hessian.
Certain parameters of the supply and feed demand functions for the EU27, Norway, Turkey and Western Balkans are updated in between iterations between the supply and market module so that these functions provide a second order proxy for the supply response from the regional programming models.
The current master covers fixed demand for bio-fuels from first generation feedstock, which generate demand for cereals, sugar and wine for bio-ethanol, and vegetable oils for bio-diesel. Substitution between feedstocks is modelled by a CES-demand function. In autumn 2010, a version with trade in bio-fuels, and endogenous supply and demand for bio-fuels will integrated in the master version.
A software component which based on statistical methods distributes consistently regional results at the NUTS II level (crop areas, stocking densities, fertilizer application rates) to clusters of 1×1 km grid cells to allow for environmental impact assessment and linkage to bio-phyiscal models. (see also )
So-called Homogenous Soil Mapping Units, uniform in soil parameters, slope class, land cover class and adminstrative unit. They may comprise between one single 1×1 km grid cell and several ten thousands ones, dependig on the spatial variance of the delineation feastures
The model covers currently all of EU27 (EU25 plus Bulgaria and Romania), Norway, Turkey and Western Balkans broken down to about 280 adminstrative regions (NUTS II). The international trade model is globally closed according to FAO data, and covers around 60 countries or country aggregates. The farm type version of the model covers around 1900 farm type respectively NUTS II models.
CAPRI covers about 40 agricultural products, covering all of agriculture according to the definition of the Economic Accounts for Agriculture. On top, there is a limited number of processed products included (dairy, oils and cakes, bio-ethnal and bio-diesel and the related by-products).
Yes, inputs as feed, N,P,K fertilizer, diesel or plant protection costs are allocated to individual production activities. However, not all of these are available as time series, so are e.g. feed quantities only allocated for the base period. CAPRI covers and allocates all intermediate inputs according to the definition of the Economic Accounts for Agriculture.
Several sources are combined in an statistical approach which ensures consistency to the Economic Accounts of Agriculture or other statistics on feed and fertilizer use, inter alia: (a) econometric estimates based on single farm data from the European Farm Accouting Data Network, (b) engineering information (e.g. requirement function for animals or nutrient contents of crops), © standard gross margins.
No, prices for all outputs and inputs are identical for all NUTS II regions inside of one Member State as they are derived from the Economic Accounts for Agriculture. The only exemption are fodder prices which reflect production costs.
CoCo, the acronym for “Complete and Consistent”, is the name for the software package which biulds up time series at national level (areas cropped, herd sizes, output coefficients, prices, market balances). The main data source for CoCo are statistics from EUROSTAT.
CAPREG is the name for the software package which builds up time series at regional level (areas cropped, herd sizes, output and input coefficients). The main adata source for CoCo are statistics from EUROSTAT.
Unfortunately, there is no central data base for EU policy instruments. Most data had been manually edited based on EU legislation.
The international part is mostly sources by FAOSAT (market balances, trade flows, estimation of transport costs from bi-lateral import and export unit values). Tariffs are based on AMAD.
The core of CAPRI is written in GAMS. In order to interact easily with the system, a Graphical User Interface realised in Java allows steering the system and explotiation of its results. The system is Windows based. The data base and scenario results are stored in a GAMS specific file format called GDX.
The installation and management of an installation is based on the software versioning system SVN.
You should have around 5-6 GByte of free hard disk to store a version and results from several runs. 2GByte of RAM are necessary if you run the full system including the international market model (otherwise, the model will swap on the hard disk and execution becomes very slow).
A full run (all 280 regions, international trade model used, iterations between supply and market part until convergence) takes about an 5-10 minutes on a fast multi-corPC. That includes all post-model analysis and preparation of the inputs files for the table and mapping tools. Some scenarios might converge slowly, or even lead to infeasibilities in the market part, and run time can then increase then dramatically. Down-scaling the scenario to 1×1 grid cells takes about a working day. The farm type version takes around half an hour for a normal solution - but the same comments about convergence and feasibility apply.
Currently, version 23.2 and later are officially supported.
CAPRI is solved with CONOPT.
CAPRI is handled as open source. Specific project deliverables may however only be accessible to the client, or be blocked for certain periods until publicly available. Access rigths are defined on the Software-Versioning server.
The philosopy of CAPRI is a network approach, where users and developpers share a common responsability to maintain the product, improve it and use it for reseach and policy support. A team in Bonn, headed by Wolfgang Britz, co-ordinates these activities.
A network of European researchers managed by Institute for Food and Resource Economics, University of Bonn. CAPRI will live as long as projects will generate the resources for its survival.
The different projects applying CAPRI devote part of their funds to finance maintenance activities (data base updates, software maintenance and improvements, generation of the reference run, documention etc.).
Yes, it is. CAPRI is handled as open source as its development and maintenance was almost 100% paid by taxpayer's money, typically by project co-financed by the EU and public reseach institutions. Certain components and data may however at least for restricted periods be accessible only for certain clients. The network tries to base CAPRI whereever possible on license free products and data sets, however, you need a GAMS license to run the model. The development of the system is and was funded mainly by EU research programs. But potential users should be aware that is requires quite some time until one may use a complex system as CAPRI for policy impact analysis.
CAPRI is hosted on a software version server from which a local version may be downloaded and then incrementally maintained. Once you have installed the necessary software (Tortoise SVN, JAVA, GAMS) you can obtain a running version of the model (including underlying GAMS code and database) using the following steps:
The folder “ts_docs” contains some further background information as well as information on selected scenarios that are supplied via the aforementioned repository
Generally, yes. However, as there is an upper limit for the number of participants, priority is given for network members involved in current projects, and, clearly clients as Commission services. Training session usually take part once a year in late summer/early autumn.
Last Updated:Tuesday, October 12, 2010