Base period and reference scenario

Because of the comparative-static modelling approach of the CAPRI system, two basic settings are analysed. On the one hand, energy use and its driving forces as well as tech-nical efficiency and its link with profitability are examined in the base period of the model settings. This enables us to provide information which for many parameters is consistent with statistical data. A three-year average around the year 2002 is selected as the base period . This allows us to avoid further exogenous information influencing the energy-use results of the model (such as technological progress or shifts in consumption patterns for energy sources, etc.) and is an ex-post analysis.

Nevertheless, the different scenarios modelled are compared to a «reference» scenario, and are all ex-ante impact-assessment simulations for the year 2013. Consequently, to allow for consistent comparison results that do not include changes resulting from trend projec-tions, the reference scenario is also scaled to the year 2013. Ideally, the most likely devel-opment of exogenous parameters should be considered in this reference scenario, and the endogenous response of the model should most probably match future developments. In the CAPRI model, two approaches are chosen to create a feasible reference solution, with both expert assessment (consultation on the most likely developments for a specific varia-ble) and trend analysis being performed. Expert data of this sort can be retrieved from inter-national organisations dealing with the issue in question, such as the EU Commission, the FAO or the World Bank. The three major sources of information contained within the frame-work are ex-post developments found in the time series generated by the model, the exoge-nous forecast provided by the EU Commission (DG Agri), and impacts of the current policy in place, including future changes already decided upon. Essentially, the approach is as transparent as possible, and represents a way of automatically generating a CAPRI base-line, with manual interventions in the process being reduced to the minimum. These dif-ferent elements are combined in a sequential process (based on Britz et al., 2007a):

1) Firstly, future changes in policy that have already been decided on are implemented for the current base year in an ex-post simulation with the sub-model «CAPMOD».

This step will therefore answer the question of what the consequences would be in the current base year of changes already decided upon for the future. The correspon-ding changes for the different endogenous results, e.g. for activity levels, production, yields, demand or trade, are stored, and internally termed «policy shifts». For the cur-rent situation, this means that the CAP for the year 2013 after the reform 2003/2004 would be implemented in 2001.

2) Secondly, a projection module known as CAPTRD is used. The results from step 1, the

«policy shifts», along with trend estimates and base-year values, define support for

«Highest Posterior Density Estimators» to produce a mutually consistent set of projec-tions for activity levels, production, yields and market-balance coefficients at the diffe-rent regional and farm-type levels.

3) In a third and final step, the projection is expanded to an overall level, including exo-genous projections for non-EU countries. Finally, the modelling system is calibrated exante to the results of the different projection steps.

The interrelationships are shown in Figure 7.

In technical terms, baseline generation aims to capture either structural or technologi-cal inter relationships, or changes in preference concerning agricultural products. Because of its market-model structure, worldwide demand shifts linked to population growth as well as policy-scheme adaptations must be covered. In keeping with the complex nature of such interrelationships, baseline generation is not purely a model-calculation-process outcome, but depends largely on expert knowledge of trend analysis and model runs in which external parameters such as elasticities and technological-progress data are adjusted and verified. Furthermore, COCO database trend-estimation results are applied for the ad-justment process (see Britz et al., 2002). The overall aim of such a process is to increase trend-estimate capacity from the pure prolongation of time series towards a comprehen-sive basis for policy-change assessment, including a safety net for cases in which no values from external projection are available. Consequently, the resultant estimator is a technical-system estimator considering a number of restrictions. Except for the milk market, where the strict limit is dictated by the current quota system, the trends can be viewed indepen-dently of the current policy framework. Consequently, a major component of the trend-estimate process is the trend curve itself, as shown in Equation 13. The curve can be seen as a Box-Cox transformation (Britz et al., 2007), as its parameter c is used as the exponent

Time series

Source: Britz et al.

(2007a).

of the trend. Consequently, with c being equal to unity, the result curve represents a straight line where for 0 < c s 1, a concave shape from below with decreasing rates results. Finally, c>1 results in a convex curve from below with increasing rates. Never theless, c is restricted to 1.2 in order to prevent sharp increases between two different time points.

Equation 13 CAPRI baseline-trend curve

X

Represents the data subjected to trend estimation (five-dimensional array spanning i, j, r, t, data status

Source: Britz et al. (2007).

A wide range of constraints ensure the projection of agricultural variables, some of the most important of which are listed below (see Britz et al., 2007 for details):

• Constraints related to closed market balances: these constraints ensure that the sum of imports and production are equal to the sum of feed and seed use, as well as human consumption, processing, losses and exports. This assumes that production quantity equals yield times production area or number of animals. Furthermore, pro-duct-group data for e.g. oilseeds must be consistent with the sum of their individual production-activity parameters.

• Agricultural-production constraints: these ensure on the one hand that the utilisable agricultural area equals the sum of the scope of all production activities, and on the other hand that the scope of the individual animal-production activities is consistent with the provision of young animals. Furthermore, the sum of animal-specific feed

j

requirements in terms of energy and crude protein times the number of animals in each category, including a range of animal-specific feeding patterns, must be covered by the bulk of feedstuff-component (e.g. cereals, protein- and energy-rich fodder, etc.) delivery. Here, a number of feedstuff components (e.g. grass, maize silage, fodder root crops, etc.) are considered to be non-tradable.

• Constraints related to prices, production values and revenues: these cover consisten-cies with external forecast information, mainly with positions within the EAA.

• Further constraints: these cover consumer behaviour, processed products, policy and growth rates, in order to ensure consistency and plausibility and maintain safeguards for the simulation process (see Britz et al., 2007).

Based on this fairly general observation, the projection tool applies a top-down ap-proach which first produces projections at the rather aggregated level of Member States and even the EU, then breaks these down into the regional, and finally farm-type, levels.

One reason for this approach is that it is far less demanding and high-quality forecasts can be achieved when projecting larger aggregates, e.g. the development of overall cereal pro-duction, yield and acreage at EU level. A second reason for this approach is rooted in tech-nical constraints. The projection is based on highest posterior density (HPD) estimators (Heckelei, et al. 2005), which are solved as constrained non-linear optimisation problems.

The function of the HPD is further expressed in Chapter 3.5. As for the exogenous infor-mation processed in the model, a number of parameters are considered, including general economic ones as well as trade-related data or domestic-policy-scheme changes of the EU CAP. For the CAP, the Luxembourg Proposal basically contains the current CAP adjustments displayed in CAPRI. For this analysis, the decoupling scheme is highly relevant, since the full decoupling of suckler cows in the EU-25 is simulated in Chapter 5.3. Details on the exo-genous parameters and the implementation of the 2003 CAP reform are given in Appen-dix 5 and AppenAppen-dix 7. Reference is made to the so-called hybrid model, which has been adopted instead of a single farm premium in a number of countries. With their dynamic hybrid system, Germany, Finland and the UK will end up with uniform regional payments, so that the previous farm-specific payments for beef fattening will be spread out among all forms of eligible land, with accordingly minimised incentives to continue beef farming.

Sweden keeps 75 per cent of the beef premiums coupled and 40 per cent of the slaugh-ter premiums farm-specific. Denmark also keeps 75 per cent of the beef premium coupled, and 75 per cent of the decoupled part is allocated to farm-specific premiums. In addition, 90 per cent of slaughter premiums for adult cattle are allocated to grassland, whilst 90 per cent of those for calves are kept farm-specific. Consequently, there is less distribution of the previous coupled payments to other types of farming in these countries. All other mem-ber states use farm-premium schemes, so that the historical direct payments will remain in the farming system. A different approach is chosen for the energy-related coefficients, all of which are based on the base-period values and adapted according to trend projections as described in Chapter 3.6.

3.3 Estimating direct energy use