Skip to main content
European Commission logo print header

Farm Accountancy Cost Estimation and Policy Analysis of European Agriculture

Final Report Summary - FACEPA (Farm Accountancy Cost Estimation and Policy Analysis of European Agriculture)


Executive Summary:

4.1.1 Executive summary:

FACEPA, which is the acronym for Farm Accountancy Cost Estimation and Policy Analysis of European Agriculture, is a research project funded by the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement no 212292. The aim of FACEPA has been to develop tools and methods to analyse production costs in European agriculture using FADN (Farm Accountancy Data Network) data. The project provides information, expertise, and economic models on cost of production for various agricultural products. The research work is performed by 10 partners in nine EU countries. FADN is the European Commission’s system to collect farm accountancy and production data for individual farms in EU member states. The statistical models developed in the project have been validated, tested, and subsequently used for policy analysis on national FADN data for each of the nine partner countries. The FACEPA project is targeted to provide technical support and relevant quantitative information on costs of production in agriculture to policy makers and experts within the Commission. In addition, it is also expected that the various tools (models) that have been developed in this research project will: i) strengthen the Commission’s capacity to estimate costs of production for a wide range of agricultural products; and ii) improve the quality and accuracy of impact assessment of CAP measures. The project have included several activities (tasks) aimed at both specification and implementation of new tools (models), and the development of important (cost) data from a policy point of view.

The contributions of this project to the existing knowledge of cost studies in agriculture intends:

• To give a review of the EU FADN system including a special emphasis on its underlying accountancy framework and cost concepts, and an examination of its usefulness from a statistical point of view.
• The development of a “general” (econometric) model used to estimate cost of production for major agricultural products.
• The implementation of this former model in the form of an operational computer tool with user-friendly interface that can be used by relevant services of the Commission to estimate costs of production for agricultural commodities, the FACEPA model software.
• Extensions and further applications of estimating cost of production in EU agriculture, with respect to farm efficiency and the competitiveness of crop and dairy sectors in the participant countries.
• Applications studying the relationships between the costs of producing commodities across the EU and its impact which agriculture exerts upon the landscape and natural environment (i.e. multifunctionality of agriculture), the development and implementation of farmed-based models that use production cost estimates to evaluate various agricultural measures on agricultural, environmental, financial and socio-economic indicators.

The numerous empirical applications that have been conducted in this project will provide policy makers with a wide range of tools and comparable quantitative information (cost) that can be fed into agricultural policy impact assessment. The “common” cost of production model can be applied to many EU countries, and it will provide homogeneous and comparable cost estimates and thus allow for the estimation of cost of production and cost analysis from a broad European perspective.

Project Context and Objectives:

4.1.2 A summary description of project context and objectives

The FADN data systems are a formidable source of information that makes the study of many facets of EU agriculture possible. Originally, the FADN accounting framework and its underlying data systems were used as a mean to provide information on cost of production and to “predict” annual changes of the CAP support prices (i.e. known as the “objective method”). Academics and policy makers have also increasingly used it for other purposes such as conducting farm performance studies. However, with the successive reforms of the CAP since the early nineties it can be questioned whether farm accounting data generated by FADN are still the most appropriate or of the “highest quality” to conduct economic analysis on EU agriculture. This is a crucial problem which needs to be addressed and this is what this research project intend to do by: i) “reviving” cost of production studies applied to several aspects of EU agriculture; and ii) showing how it is possible to conduct fruitful quantitative policy assessment analysis using FADN data. This effort will essentially be pursued using alternative farm-based models. These two broad dimensions of generating cost of production for agricultural commodities in the EU – i.e. developing appropriate cost of production model and relying on representative farm accounting data for EU agriculture – are addressed by this research project. FACEPA thus intends:

• To address the usefulness and appropriateness of the present FADN data systems to measure cost of production for agricultural commodities.
• To study the feasibility of developing a “general” cost of production model for EU agriculture that is easy to use by practitioners and reliable in terms of generating relevant analysis for agricultural production and policy analysis.
• To test and implement this cost model in an EU context with the idea in mind that it can be applied on a large scale (i.e. several agricultural commodities and large number of member countries).
• To assess the relationship between cost structure and farm performance, farm technology, environmental quality and farm heterogeneity with FADN databases.
• To provide methodological improvements to the above “general” cost of production model
• To undertake the evaluation of agricultural policy measures using FADN data indicators.

The work plan of FACEPA distinguishes ten (10) different work packages. All of them include several sub-tasks in order to optimize the organization of the research and ensure that the different partners work together effectively. Nine (9) work packages are dealing with research activities and one (1) work package refer to management activities. These first nine work packages can be further divided into three groups. The first group, which refers to the first four work packages, deals with the development, implementation, validation, and dissemination of an “economic model for estimating the cost of production of various types of agricultural products using the FADN data”. In the second group, are included the next four work packages which focus on applications and extensions of the former cost model that are relevant when to study performance, policy and farm structure considerations in EU agriculture. Within this group, alternative farm-based models are developed and used to generate estimated cost of production. Finally, the third group includes the last of these nine work packages. Its objective is devoted to the evaluation of (agricultural) public policies using cost estimates obtained in the previous work packages. During the lifetime of the project, 30 deliverables will be submitted.

The main task of the third and last period has been to provide and finalize the theoretical and empirical foundation needed to develop and finalize the “general” cost of production model for the estimation of the cost of production for various types of agricultural products using FADN data. In terms of deliverables the objectives for the reporting period have been to finalize maintaining working papers. Given the considerations and the relevant background materials developed during the FACEPA project, the objectives assigned to the work undertaken during the period consisted of:

• To implement the cost of production models for different products, countries and production systems, and to estimate the model for three pre-established agricultural products including crop products, pigs and dairy.
• To develop, test, and complete the cost of production model software, the FACEPA model, and to finalize the software User Guide. The FACEPA model software aims at calculate coefficients of production and costs of production for the major products from FADN data.
• To analyse the relationship between productions cost structure and farm performance, and the efficiency (with special attention to the effect of economies of scale) and competitiveness of the crop and dairy sectors in the participant countries. Furthermore, to assess the differences between the economic performance of individual and corporate farms in the New Member States, and to provide quantitative insight into the impact of CAP reform on farm economic performance. Also, to provide a country level analysis of EU agriculture’s competitiveness considering various support schemes.
• To develop, verify, and describe the use of PMP model to simulate different scenarios based on the market price variations and on the effects of CAP reform implementation.
• To determine the relationship between the costs of commodity production and provisioning of environmental values, and to contrast the costs of commodity production and associated environmental impacts in marginal and productive agricultural regions.
• To contrast organic and conventional forms of commodity production in terms of costs and environmental performance.
• To develop a methodology aimed at specifying disaggregated input-output (IO) tables of EU agriculture at the regional level, and to implement this former methodology using FADN data and apply it for a EU region.
• Using FADN data, to generate estimated indicators (quantile indicators) on the statistical distribution of cost of production estimates, and in so doing provide information on the distribution of costs according to size and/or scale.
• To specify and to estimate a varying-parameter, cost of production model to address the heterogeneity and diversity of EU farms, a methodological application to estimate cost of production using a semi-parametric random coefficient model.
• To develop key economic models that use production cost estimates to evaluate agricultural policy measures: i) to provide and test farm economic models for evaluating ex-post policy measures, ii) to provide and test farm economic models for evaluating ex-ante policy reforms, and iii) to undertake preliminary comparative analyses on these ex-post and ex-ante evaluations across member states and regions. This research component designs and develops key economic models that use the production cost estimates from WP3 but also from WP5, WP6 and WP7 to evaluate various agricultural measures on agricultural, environmental, financial and socio-economic indicators using FADN data and additional EUROSTAT data.

The expected outputs during the project period associated with the above objectives included:

• All partners presented the results from the implementation and testing of the cost of production model software for their own countries, and delivered their final written contributions.
• To complete the cost of production model software, the FACEPA model, and to finalize the software User Guide. The FACEPA model software aims at calculate coefficients of production and costs of production for the major products from FADN data. The software was established from the works conducted work package 3. The FACEPA model is programmed in SAS language and runs on the Enterprise Guide SAS module.
• To obtain and finalize empirical results/findings on the relationship between cost structure and farm performance, farm technology, environmental quality and farm heterogeneity with FADN databases.
• To implement and complete the use of PMP model to simulate different scenarios based on the market price variations and on the effects of CAP reform implementation.
• To characterise and quantify the relationship between the cost of producing commodities across the EU and the impact which agricultural production exerts upon the landscape a natural environment, and to contrast organic and conventional forms of commodity production in terms of costs and environmental performance.
• To provide applications and/or improvements of the cost of production model/estimates based on FADN databases, and to implement a methodology to produce, disaggregated input-output (IO) tables of EU agriculture at a regional.
• To develop and implement a new methodology to estimate cost of production using a semi-parametric random coefficient model.
• To specify economic models that can be used to evaluate agricultural policy measures. This research component in work package9 designs and develops key economic models that use the production cost estimates from WP3 but also from WP5, WP6 and WP7 to evaluate various agricultural measures on agricultural, environmental, financial and socio-economic indicators using FADN data and additional EUROSTAT data.
• The main outcome of work package 9 consists in providing functioning farm-level economic models based on FADN cost estimates that are appropriate for ex-post as well as ex-ante policy analysis.

Project Results:

4.1.3 A description of the main S&T results/foregrounds

4.1.3.1 Concepts

FADN is a highly harmonised EU-wide survey. Much time and effort is spent in the Member States and at the EC in order to produce comparable and reliable information for agricultural activities. Despite the high level of harmonisation there are several differences concerning the collection of FADN data and the interpretation of FADN regulation which are very important to keep in mind while constructing the cost of production model. In the Member States, the data are collected according to EU-wide standardised guidelines. However, there are national distinctions concerning data availability. Further the accounting system of the FADN is quite different from the one used in financial accounting. In financial accounting the key characteristic is reliability. The aim of the FADN is to serve the decision making process in the EU, so the relevance, the comparability and the usefulness are important requirements. The information served by the FADN is less reliable than the data from financial accounting, because many times estimations have to be made and information has to be updated in order to meet the requirements of the FADN. Additionally, the comparability is strongly influenced by national distinctions. The cost accounting principles of IFRS and FADN are basically the same. Differences can be found at the recording of farm produced feeding stuffs, due to the diverse valuation methods (on farm cost versus farm gate price) of costs. Compared to IFRS, investment subsidies increase income in FADN and profit tax is not deducted from the income in FADN. Concerning assets, according to FADN, only operational resources are considered whereas according to IFRS the total assets of an enterprise (for example including financial assets) are to be indicated in the financial statement.

The requirements of FADN regulations are realised in different ways in the Member States. Data delivery in the Member States for FADN purposes is not totally harmonised. For example: FADN requires a valuation of agricultural land at net selling prices. Some countries follow this requirement while other countries value agricultural land different from these FADN specifications. These countries often distinguish between purchased land and other land. Another example is the difference between Member States in the valuation of quotas and delivery rights due to the very heterogeneous national rules. The valuation of biological assets is also very diverse. The problem of contract rearing also should be mentioned. The model should separate farms with receipts from contract rearing and also consider national distinctions regarding the recording of contract rearing.

Despite the high level of harmonisation there are several differences concerning FADN data collection and the interpretation of FADN regulations which are very important to keep in mind while constructing the cost of production model. It can be stated that FADN data collection has many differences in the Member States, thus special care must be taken to define harmonised European data for the cost model.

• The lower thresholds to get into the sample vary from 1 to 16 European size units (ESU). In some Member States upper thresholds are determined, as well (e.g. Netherlands).
• The FADN's representativeness is considerably different. In seven Member States of EU25 FADN farms represent less than 90 per cent in terms of Standard Gross Margin.
• Member States use different additional criteria for farms to get into the field of survey and the methods for defining the sample are also different.
• The year of FSS varies from Member State to Member State between 1999 and 2006. It has great importance, as there was a large structural change in CEE countries in recent years.

The differences between the Member States also vary regarding activities other than agricultural activities on the holding. This means that harmonisation in the area of “Other income on the holding” is difficult. The inclusion of ”Other gainful activities” varies from Member State to Member State. Member States decide in different ways whether a holding which has other gainful activity should be included in FADN or not. Regarding the cost of production model the most important problem is that the cost items cover various activities and there are no data showing the division of the costs. And due to the different interpretation of other gainful activities it is difficult to determine whether a cost connects to an agricultural or non-agricultural activity.

4.1.3.2 Comparing the characteristics of EU and national FADN

Most work packages in the FACEPA project are making use of the FADN data because it is the only source of micro-economic data for agriculture in the EU which is harmonised. In spite of the existence of common rules adopted by the Member States to inform the European FADN, the majority of countries use special methods for their national needs. No national sample is exactly the same as the EU-FADN over the period 1995-2005. Below two points are addressed related to methodology: the statistical point (sample and weighting) and the accountancy point (depreciation).

The field of survey of the EU FADN is a well-defined group of farms out of the total number of farms and includes commercial farms that exceed a minimum threshold of agricultural production measured in European size units. This threshold varies among Member States, and some countries apply additional criteria for excluding farms from the survey. Given the great variety within the FADN field of observation, stratified sampling is applied to ensure that the sample of farms adequately reflects this heterogeneity. Stratification as well as procedures and methodology to select sample farms vary among Member States. For example, non-random sampling and voluntary participation may introduce sampling bias. The differences in selecting, stratifying and sampling between Member States can also result in differences in national and EU FADN weights and affect representativeness. Using different SGMs in the national FADN has impacts on the selection of farms (by effect on economic size) but also on the specialisation of the farm holding. Germany, for national needs, keeps in the national FADN the farms with a SGM between 8 and 16 ESU which are under the European threshold. Italy excludes from EU-FADN almost empty cells (economic size x type of farming). This represents 0.3% of the national SGM and 489 farms. Until 2003, France used an additional criterion to split the field of observation for selection of the national sample. The sub-sample refers to a mode of data collection. The Netherlands use random selection with a specific stratification. The size classes are different within types of farming; the Dutch FADN uses strata for organic farming and subtypes of farming. Few farms are missing from the field of observation of the European FADN. Germany computes SGMs from the 5-year average instead of the 3-year average in the other countries. The sub-stratification used by the Netherlands in selection is also used for weighting. For the stratification in the weighting system, Hungary uses the legal form of the enterprise to separate private farms from economic organizations.

Because depreciation plays an important role in the Gross Farm Income and because the regulation does not impose any method of depreciation, it is necessary to analyse the applied methods. Depreciation depends on three parameters: estimation of fixed assets (replacement value/acquisition cost), method for computing depreciation (linear/digressive), rate of depreciation. Concerning the estimation of fixed assets, the National FADN is different from EU-FADN for Germany and to a lesser extent for Hungary. These countries adopt the acquisition cost. But in respect to regulations, Germany corrects depreciation to inform the EU-FADN. In general, depreciation based on replacement values is about 1/3rd higher than those based on acquisition costs. For computing depreciation, the FACEPA partners use the same method in the national FADN as the European one, even if each of them prefers the linear method or the digressive method.

Regarding outputs, additional variables are available in all National FADN of the FACEPA partners, except for Bulgaria. The young Bulgarian FADN, imposed by the Accession Treaty, uses the EU-FADN classification. To comply with the EU-FADN, member states often aggregate national output variables. Knowing the details of the products grouped together is useful for two purposes: to specify the content of data but also to point out differences between countries. For organic products the available information is limited. Four countries (Belgium, France, Germany and Italy) limit the data to the European regulation which is a separation variable to mark the farm. For the mixed organic farms it is not possible to distinguish organic products, except for Italy where organic product is an option among several possibilities. The Netherlands uses a sub-sample of organic farms and Germany collects optional information about sales.

4.1.3.3 Statistical problems associated with the EU FADN database

The strengths and weaknesses of the EU FADN database for estimating cost of production from a statistical point of view have been discussed and analysed. To this end, the approach and implementation of sampling and weighting in EU FADN is summarized and size and distribution of used weights are examined. Further the coverage and representativeness of EU FADN data have been analysed, and also the question of the representativeness of EU FADN with respect to organic farming was addressed. The overview of the sampling and weighting used in the EU FADN data system highlights the extent to which Member States use different methodologies for selecting, sampling and stratifying farms, and the possible impacts this may have on representativeness, coverage and weighting of the data. Factors of concern include:

• the country-specific differences between the field of observation and the total population,
• the use of additional criteria and sub-samples as well as different SGMs for sampling plans,
• resulting differences in national and EU FADN weights and the representativeness,
• the potential sampling bias introduced by non-random sampling and voluntary participation in some Member States.

Weighting factors are used to extrapolate the EU FADN sample. These weighting factors also have to be taken into account when specifying a cost of production model which aims to reflect the input-output allocation on the Member State level to prevent distorted results. The larger the variation in the weighting coefficients is, the greater the need for their incorporation in a cost production model that aims to produce information at the level of the Member States. To investigate the size and variation of weighting coefficients across Member States and farm types, a descriptive analysis is carried out for the year 2005. The analysis of the weighting coefficients shows that the variation of weights is high, especially in the nMS. The results also reveal that some Member States have very high weights which might lead to lower reliability of the cost estimates. Some of the causes for the high weights could be very low heterogeneity, sampling procedure and differences in national and EU FADN sampling and weighting methodology. In this case use of national FADN weights can be an option to improve the reliability of the estimates. Further research is needed to analyse the impact of differences in EU and national weights on production costs estimates.

The representativeness of EU FADN has been examined by comparing a set of various structural variables between EU FADN data and the FSS for all 25 Member States. Further, it has assessed the change in the coverage and representativeness through time by comparing the representativeness of the EU FADN in 1995 and 2005. The low coverage for the variable “number of holdings” is suggested to be due to the fact that the number of farms with a SGM below the country-specific ESU threshold is significant. At the same time the much higher coverage for the other variables show that those farms below the ESU threshold have a minor share in UAA or number of dairy cows, for example. The coverage for the “number of holdings” versus the other variables are not surprising, as the fundamental aim of the EU FADN is not to gain a high degree of coverage with respect to the number of holdings in the Member States, but with respect to the agricultural activity measured in total SGM. FADN data covers a larger part of the whole population in the oMS than in the nMS. This is particularly true for the number of farms where the average values for the EU-15 are more than twice as high as for the nMS. The results show that there is considerable potential for focusing production cost estimation on samples of specialized farms, as these often have a very high share in the total production of selected products. Nevertheless sample sizes need to be checked in each case to ensure robust estimates. The findings indicate that, on an EU average, the coverage and representativeness is relatively large for the variables under study. However, considering the single Member States reveals that in some cases significant differences exist cross-sectional. In view of the changes over time, it is shown that the coverage and representativeness increased from 1995 to 2005. It must however be noted that the empirical analysis and the conclusions drawn are based on the comparison of structural variables such as hectares of major crops and numbers of specific livestock between the sample and the population. It gives no final answer whether estimations of costs of production using EU FADN will reflect the true population value. Validating the estimation results using cost calculations from other sources are therefore necessary topics for future research.

Another analyse that was carried out was the coverage of organic farming. An identifier variable for organic farms was introduced in the EU FADN in 2001. In 2006, the sample includes accounts from more than 3000 fully organic farms; however, sample sizes vary strongly between countries. Currently, the number of organic farms is small and hence the sample will only allow an econometric estimation for few countries. Generally, the potential for estimating cost of production based on specialized organic farms is low, due to small sample sizes and the often higher diversity of production structure in organic farms. As organic farming is not a stratification criteria employed when calculating the EU FADN weights, the reliability of these weights might be low, especially in countries where organic holdings represent only a small proportion of farms. Another problem is that some countries do have strata for organic farms on national level which are not taken into account in the EU weighting. More robust and representative estimates may be achieved using national FADNs which in some countries include a higher number of organic farms, and/or allow a weighting of these farms.

4.1.3.4 Implementation and validation of the ‘general’ cost of production model

Within the FACEPA project, considerable resources have been allocated to implement and validate the ‘general’ cost of production model (GECOM model). The outcome of the related work has been published in three connected reports. The first two reports describe the implementation, the validation and the results from the GECOM on the basis of national farm accountancy data networks (FADN) and the EU FADN, respectively. The third report provides an overall synthesis and conclusions.

First, general aspects of model specification and estimation, e.g. the list of outputs and inputs, are shortly summarized. In general, the more detailed the list of outputs the more accurate is the specification of the model. On the other hand, a longer list of outputs increases the probability that estimated cost coefficients of some outputs will be less robust and precise. This trade-off needs to be considered in view of the specific data for which cost estimates are sought, and is thus a decision to be taken for each application. The decision on the list of inputs, specifically the treatment of marketable farm-grown intermediates and the inclusion of subsidies, and the resulting income indicator needs to be taken in view of the research question to be analysed.

The implementation of a seemingly unrelated regression algorithm in the software package SAS allows for a very fast and stable estimation of the GECOM. However, one issue which has been observed by all partners was the frequent occurrence of negative cost coefficients. While often not statistically significant, these negative coefficients proved to be a major concern in the validation of the model. The assumption of a common Leontief technology underlying the production function of all farms clearly constitutes a strong constraint on the applicability. Some of these limitations can be offset by selecting farm samples with homogenous technologies, and therefore as far as possible, farm samples should be stratified according to applied technologies.

During the testing of the GECOM model, it was repeatedly highlighted that even small and/or infrequent data errors can have a significant impact on estimated cost share coefficients. Their identification has proved to be a major challenge, despite extensive plausibility checks by national and EU authorities. For the applications within the FACEPA project, a multivariate methodology based on the Mahalanobi distance measure to eliminate the data errors was tested, which generally improved stability of results over time, however often excluded an undesirably large share of observations.

The validation process showed that to ensure a correct interpretation of results from other sources in relation to GECOM cost estimations, a very careful examination of these sources is needed with respect to the approach used, the definitions of costs and cost categories, the definition and calculation of imputed costs and the scope of the costs allocated. In general, the validation of the GECOM by comparing results to those of other studies as well as by a review of estimates by national experts highlights that the quality of estimates differs by country. Overall, level and trend of total costs of the main products wheat, milk and pigs were judged to be plausible. Generally, estimated cost for crop products were less robust and in several cases implausibly variable over time. In many countries, also the estimates of individual cost components were assessed to be realistic, especially of direct costs, while the values for overheads and depreciation were less reliable.

An emerging key issue for the dissemination of results from the GECOM was that existing national “conventions” for the definition and presentation of production costs differ from the ones used in the FACEPA project. Thus, when presenting results in a national context, appropriate care needs to be taken to thoroughly explain approach and definitions, and to reprocess cost estimates to match national conventions as far as possible. In the subsequent section, the outcome of the comparison of cost estimates based on EU and national FADN is presented and discussed. Differences in production costs estimates based on EU vs. national FADNs may in principle be caused by differences in samples, weighting factors and data. The analysis showed that while in many cases production costs estimates based on EU and national FADN are very similar, in numerous instances results differ significantly. In these latter cases, results based on the national databases were generally judged to be more plausible. This outcome is in line with prior expectations, as data in national FADNs are by nature more differentiated and closer to ‘original’ farm data, and national weights better reflect the actual sampling procedure than this can be the case for the weights derived by the static EU FADN weighting system.

Based on the experiences made with the GECOM application above, this report emphasizes that the generic software tool for implementing the GECOM should offer user-friendly options for carrying out basic data pre-checks, provide flexibility with respect to selection of samples and aggregation of outputs and inputs, and produce a clearly represented overview of the statistical significance of estimated coefficients.

As a general conclusion, the implementation, testing and validation of the GECOM showed that the model can provide plausible estimates of production costs for main products in most countries, reflecting developments over time as well as cost composition, while results for products with smaller output shares were often not convincing and highly variable over time. However, the experiences also showed the indispensable necessity of pre-checking the data in each case, dealing with outliers and taking into account details and changes in the data definition and collection. A general conclusions from the experiences gained is that no “simple” application of one general model is possible for all samples and products. An analysis and validation of results by experts (i.e. of both FADN data and agricultural production systems in the analysed samples) will always be needed.

Key issues for future research concerning the GECOM are the estimation of production costs for meat products, and the search for improved, robust and transparent outlier detection methods. Other estimation approaches (e.g. panel and entropy estimators) and/or model specifications (flexible functional forms) have an obvious potential for dealing with several of the limitations of the current methodology identified in this report, however, more research is required to facilitate a broader and more robust application of these approaches to estimate production costs for all EU Member States.

4.1.3.5 Dissemination and valorisation of the production cost model software

In the European FADN, the costs, detailed by category (seeds, fuel etc), are available for agricultural holdings. The developed computer software is designed to allocate these costs to different productions. The concept and start point was a model built by INRA in the 2000’ years. The software was established from the works conducted by vTI in work package 3 (the cost of production (GECOM) model). This computer software allows generating cost of production for all sorts of agricultural commodities using i) FADN data and ii) the SYSLIN procedure of the SAS software. It is menu-driven and includes a detailed interface so that users of this computer software could operate it using an EXCEL worksheet environment. The interface relies on the JAVA system, and the computer software integrates the recent specifications and developments made by vTI on the generic cost of production model. INRA presented a first and preliminary version of this computer software in May 2010 at the 5th general FACEPA meeting in Uppsala. At the 7th General Meeting in Den Haag on March 31 and April 1, 2011, it was agreed to change the present name of the production cost software and call it from then and onwards for the FACEPA model.

The FACEPA model is programmed in SAS language and runs on the Enterprise Guide SAS module. The model estimates input – output coefficients from EU FADN data. The technical coefficients of production are estimated by using the SAS ‘PROC SYSLIN’ procedure. From these coefficients, it is possible to deduce costs per quintal, cost per hectare or cost per animal. The model runs for a single country over a period of several years or for a single year for one or more countries. It is possible to add a classification variable to the list of variables given, in order to obtain results per sub-group (per region, or per type of farming). It is also possible to obtain the results on a specific field.

The model takes into account automatically the total production. The user defines a list of crops outputs and livestock outputs and the balances are automatically calculated. For inputs, variable costs and fixed costs are distinguished, but the total input is not given. The choice of input determines the income indicator, which is integrated in the model. For example, if the user takes only the intermediate consumptions, the income indicator is the added value. The files include imputed costs for family-owned factors. It is possible to include these costs in the model, but this procedure was not tested by the vTI-team. More generally, it is necessary to be careful in interpreting the results on the fixed costs.

In the FADN data, the values of output are given at the production prices. In the model, subsidies are considered as negative costs. It is possible to select coupled or total subsidies. In the final output, subsidies are included in the "basic prices". This solution is in theory only coherent for coupled subsidies but it is possible to include Single Farm Payments (SFP) and the second pillar payments. Taxes on products are also automatically deducted in the calculation of basic prices.

It is possible to obtain information to judge the significance of the coefficients (Standard error, T value, P value). Different options are possible:

1) It is possible to delete outliers (see vTI reports).
2) The results could be weighted (or not) with the SYS02 variable, which corresponds to the weight of the holding in the sample used.
3) The production includes or not the production used on farm. If the user takes the production used on farms, it is necessary to put in the inputs seed and feed produced on farm.
4) The "Allocation of residuals" option allows you to edit the breakdown of individual costs. The residual difference between the estimated values and the real values calculated for each holding is distributed over the different products proportionally to the gross output (with home-grown consumption) or the gross product (without home-grown consumption). This option also creates three groups of intensity according to the holding coefficients calculated for each production.

4.1.3.6 Extensions and further applications of estimating cost of production

Various efficiency indicators for European Union (EU) countries included in the FACEPA project, Belgium, Estonia, France, Germany, Hungary, Italy, the Netherlands and Sweden have been analysed. The availability of long period datasets between 1990 and 2006, allowed concentrating on the long time trends in technical efficiency especially in Old Member States.

Two main approaches developed over time for analyzing technical efficiency in agriculture have been used: (1) The nonparametric data envelopment analysis (DEA); (2) the parametric stochastic frontier analysis (SFA). While the vast majority of empirical studies on technical efficiency in the agricultural sector mostly have utilized only one method to estimate their efficiencies, both methodological approaches to measure efficiency have been applied. In addition, most studies focus on a single country’s agricultural sector, thus the comparative analysis of the technical efficiency is rather scarce. The relative importance of specific subsectors and rationale of compilation more homogeneous sample have been taken into account, thus the analyse focuses on the field crops and dairy sectors separately. The analyse highlights the importance of easier availability to the farm level data, namely Farm Accountancy Data Network (FADN) data in the EU, which may provide interesting insights for policy makers on farm level technical efficiency and develop more targeted policy to improve efficiency in European agriculture.

Generally, all countries have relatively high levels of mean efficiency ranging from 0.72 to 0.92 for both field crops and dairy farms. Interestingly the majority of countries have better performance in dairy sectors in terms of higher levels of mean efficiency than in field crop production. A slightly decreasing trend however may be observed for all countries. Technical Efficiency estimates are largely in line with those obtained by previous studies.

Further work has been done to the issue of how relative performance of farms fluctuates in terms of technical efficiency over time. One hypothesis may be that many poorly performed farms remaining inefficient and some farmers are performing always very efficiently. Farms which are usually at the bottom or top of the efficiency ranking can be identified. However, the FADN data has an inherent problem for long time period analysis arising from its rotated panel nature, namely that not all the farms are observed for the whole period. In this respect, there is a need to calculate transition matrices in each consecutive year. Surprisingly stability analysis revealed that in average 60% of farms maintain their efficiency ranking in two consecutive years, whilst 20% improve and 20% worsen their positions for all countries. However, these ratios slightly fluctuate around these values for one year to next year. Mobility analysis ranks countries according to the mobility of SFA scores within the distribution. Farms in New Member States are more mobile than those in EU15.

The DEA estimation shows a similar declining trend on the development of technical efficiency over time except for the Swedish dairy sector showing an increasing efficiency trend. The total productivity changes have been investigated in two steps. First, a definite trend in total factor productivity changes was not found. Second, the question whether total factor productivity changes converge or diverge over time was addressed. By using panel unit root tests for estimations reveal a convergence of productivity across old EU member countries during analysed period. Finally, the total factor productivity changes were decomposed into its main elements. Field crop farm indicators generally present a significantly higher volatility than dairy farms. Random effect panel regression of Total Factor Productivity Change on its components shows Technological Change as being the significant positive driver for crop farms, whilst Technical Efficiency Change followed by Technological Change are the most important for dairy farms. In addition, no significant impacts of CAP reforms in 1992 and 2000 on total productivity changes could be found.

Technical efficiency scores have been used to obtain three distinct methods, Stochastic Frontier Analysis (SFA), Data Envelopment Analysis (DEA), Operational Competitiveness Ranking Analysis (OCRA), based on national Farm Accountancy Data Network (FADN) data. This have been done in order to analyse the impact of European Union (EU) accession and the influence of farm classification, more precisely farm type, upon the performance on field crop and dairy farms in three New Member States (NMS), Bulgaria, Estonia and Hungary.

Theoretical and empirical evidence have been provided that farm classification is subject for empirical analysis, because using FADN and conceptual (Hill type) typology may result in considerably different farm structures. The main outcome of this research is that individual farms are not equivalent to family farms as usually assumed in previous research. The findings stress that average size of individual farms is considerably higher than of family farms. Not surprisingly, an ambiguous pattern of farm performance emerged from different approaches irrespective to product groups and country. However, the majority of results confirm that the average performance of individual and family farms is weaker than that of the corporate farms: including companies, cooperatives, intermediate and non-family farms irrespective of the methods, product group and country.

The main conclusion point at for the second stage regressions that, employing efficiency estimates obtained with the three distinct methods (SFA, DEA and OCRA), yield rather diverging results. From a methodological point of view, one would expect that commonly used methods, i.e. SFA and DEA would result in dependent variables with higher explanatory power, and consecutively better specified second stage regressions. This was not the case. Determination coefficients were by far the highest in OCRA regressions, and these also produced the highest number of significant coefficients. Considering SFA and DEA methods, the efficiency scores obtained with the latter seem to be more appropriate for second stage regressions.

First, an assessment of the impact of farm types on farm performance was done. The simple mean comparison estimation showed there are significant differences in farm performance among farms in terms of legal form or farm organization. However, panel regression just partly confirms these results. The main reason is that a considerable number of farm type coefficients are not significant. Reference will be made only to those results, where estimations provide significant results. The impact of family and individual farms on farm performance is rather negative except for Estonian dairy farms, where the opposite effect was observed. The most striking result is that farm size is positively related to performance confirming that scale efficiencies do matter in these countries.

The final interest is the possible impact of the EU accession on the farm performance. With the exception of some regressions having OCRA scores as dependent variable, the EU accession proved to have negative effects upon farm performance, regardless of the country, sector or farm typology considered. Although this might not seem a plausible result at first, it has some logic behind, and it is not unprecedented. Through EU accession farmers got access to higher subsidies, but the public support received by farmers in the frame of the Common Agricultural Policy (CAP) may have a negative influence on their technical efficiency. As it has often been shown in agriculture, public support reduces farmers’ effort, implying greater waste of resources and thus further located from the efficient frontier.

The competitiveness of EU agriculture with respect to farm efficiency was the last analysed subject within work package 5. Most often competitiveness of a country’s given agricultural sector is analysed using trade and price data, the linkage between efficiency and competitiveness is quite ambiguous. In this research however, the analyses was based solely on FADN data, using some ideas – not directly supported by theoretical models - and policy directives, in order to link sector (field crop and dairy) and country specific (Belgium, France, Germany, Italy, the Netherlands, Sweden, Estonia and Hungary) farm performance through CAP payments to competitiveness. The unbalanced panel nature of the dataset allowed the estimation of three distinct models for both field crop and dairy farms: pooled OLS, random effect panel regression and fixed effects panel regression. Where relevant, the emphasised differences between EU-15 countries and New Member States (NMS) were represented in the sample. The decoupling process is captured in the analysis by the inclusion of a dummy variable. Results obtained are quite ambiguous. There are only a few common, easily interpretable results; rather they vary according to models applied and country discussed. The conclusion suggests that although the linkage between the evolution of technical efficiency scores and subsidies received, to have implication upon farms’ competitiveness, the classical gravity and trade based analysis offers better representation of individual counties competitiveness of a given sector.

4.1.3.7 Modelling farm technologies

It is well known that European FADN does not collect the information about the variable costs associated to the different farm activities but only the total variable cost at a farm level. This lack of information makes it difficult to evaluate the production allocation decisions without the use of other external sources (engineering information, literature, etc.). This method has the risk to not be able to differentiate the costs according to the specific farm specialization and size (economic and physic). Moreover, all the policy and market evaluations based on FADN database are based on estimations and not on explicit profitability value. The evaluation of the farm enterprises costs is the main issue of FACEPA project and also the methodology applied in work package 6 takes these concerns into account.

To solve the problem linked to the lack of FADN data on specific production costs per process or enterprise, the classical approach of PMP (by which the farm behavior is estimated in conditions of maximization of the gross margin) has been modified. One important phase of PMP is the calibration process, proposed for estimating the farm production decision component. The lack of information about the analytical costs poses a problem during the calibration phase of the model when the estimation of the cost function requires a non negative marginal cost of all production processes activated by the holding. The standard PMP calibration method has been modified to solve this problem, and in order to generate the observed production plan using the dual structure of the problem proposed by Paris and Howitt. In particular, the dual optimality conditions have been used directly in the estimation phase of the non linear function. This approach (qualified as an extension of the Heckelei’s PMP methodology) avoids the first phase of the classical PMP method by imposing first order conditions directly in the second cost function estimation phase. So, the model considers the information relative to the total farm variable costs available in the European FADN archive. This is an innovation, important because permits to perform analyses using FADN dataset without having to resort to parameters that are exogenous to the model. The application of this innovation in well illustrated in the deliverable “Methodology to assess the farm production costs using PMP farm models” in which a group of farms (35) selected from the Italian FADN dataset, has been used to estimate the costs of production. Information about hectares, yields, prices, subsidies and farm total variable costs have been used for estimating the specific costs associated to the farm production plan. Considering that, differently from European FADN database, in the Italian FADN the costs are allocated among the different production processes from the surveyors, these results have been compared with the estimated costs.

The results have been very interesting and for this reason, the Italian dataset has been used to validate the estimates obtained by the application of the PMP model to the Italian case. In fact, while for Belgium and Hungary case studies the lack of observed specific accounting costs has prevented the possibility to make a sort of validation, for the Italian regions (Veneto, Lombardy and Piedmont) the allocation made by the surveyors present in the national FADN dataset has permitted a comparison between the observed and estimated costs. With this respect, the methodology is not so different from that developed in Italy within the work package 3. Here the results of econometric estimations (GECOM model) have been compared with the observed costs following the same scheme. Inside INEA, responsible for the implementation of Italian FADN, the discussions about the results of both methodologies have been very interesting. The results obtained for Hungary and Belgium have been compared with the output of work package 5: for Belgium the results have been very close while for Hungary the PMP estimates have been much higher than the outcome in work package 5.

The different applications of the PMP model in work package 6 are illustrated in the deliverable “Methodology for the definition of case study farms and model structure for each case study”. Here the PMP model has been applied considering the Farm Type “arable crops” in Italy, Belgium and Hungary. All the information about farms (acreage, prices, yields, other earnings and total variable costs have been used to estimate two type of costs: the specific marginal accounting cost and the hidden marginal costs. While the sum of the first ones is equal to the total variable costs provided by the European FADN, the estimate of hidden costs is related to the part of the cost that eludes the farm accounting system but that is considered an important element inside the farmer’s decision process because it influences the production choices. The hidden cost is an opportunity cost influenced by different factors like the farmer’s experience, his risk attitude, and the market expectations and so on. All these variables are not evident in the accounting results but are taken into account in the observed production plan. So, the hidden cost became a very important component in the assessment of the economic situation of the farm, difficult to estimate.

In addition to this, the analysis has been carried out using a multivariate analysis based on principal component detection and the cluster analysis method which has contributed to reduce the variability of the information used in the estimation phase and to control the outliers. In particular, the estimation procedure seems to be very sensitive to the presence of outliers, so a preventive check is important to minimize the interference of out-of-range value. The cluster analysis permits to reduce this risk. Moreover, the presence of a high variability of the yield for some crops produces unreliable estimates in some cases, including those with a high number of observations. So, when the internal sample homogeneity is not so high it is important to stratify the territory or the sector in order to have an improvement in the statistical significance. An important result of this application is that the PMP model seems to have a good capacity to reproduce the observed accounting costs for cereals and, in general, for crops with a high level of homogeneity in prices and yields. It is important to reduce the variability as much as possible and to find an adequate method to group the farm like sector and territorial stratification or multivariate methods. Also the estimation of the hidden costs can be considered as an important result of this application because of it is an element very difficult to assess but as fundamental in the definition of the production plan.

The deliverable “The effect of the single farm payment on cost function and production function” in work package 6, adds another element to the evaluation. The first is the possibility to change technologies in the production plan and the second concerns the income variation and new land allocation according to policy scenarios including CAP reform and increasing market prices. The possibility to change the production plan, differentiating between crops already grown into the farm and new crops, it is of great interest and the structure of the PMP model implemented in FACEPA permits to have a production function (the average technology) and the related variable cost. An additional “delta” represents the position of each farmer with respect to the average technology. The PMP model estimates the “latent information” related to production function and the associated variable crops that is used when economic conditions become favorable for the activation of new production function. The interesting thing is that the latent information can be used relatively to a given crop that is not grown in the farm. The analysis has been applied to Veneto region, considering arable crops and introducing the condition imposed by the Health Check reform in three scenarios, with different yield levels. It is very important that the result obtained from the new crop represented by sorghum is used for energy production. The PMP model can be used to evaluate under which economic conditions the new crop can be inserted in the production plan of the farms and what can be the impact on the environment.

4.1.3.8 Cost of production and the environment

In work package 7 the deliverable “The influence of landscape services on farm costs: The case of Swedish milk farmers” uses information on the economic performance and biodiversity provision of 304 Swedish milk farmers to report research on the relationship between biodiversity and the cost of farming. The biodiversity indicators are based on biological field studies mapping the existence of valuable species in the agricultural landscape while farms’ marginal costs are estimated using a flexible cost function.

A positive correlation, unconditional as well as conditional, is found between marginal costs and biodiversity. This relationship is valid when it comes to the binary choice of managing permanent pastures or not, and to providing more biodiversity for those already with valuable pastures. For the main indicator - number of species of vascular plants – a one percentage increase in biodiversity is related to a 0.03 per cent higher marginal cost. If instead biodiversity is incorporated into the farm cost function and assumed to be a variable output, then also an increase of biodiversity boosts marginal costs could be found. The results therefore support a competitive relationship between the provision of biodiversity and farms’ cost structure. This also holds for different types of outputs, but it is more pronounced between biodiversity and beef production.

Weighting the number of species with rarity gives a similar but somewhat weaker relationship, which suggests that the landscape characteristics of pastures with rare species are not more unfavourable for modern agricultural production than other pastures. However, from a biological perspective the landscape properties are of course important for biodiversity, not least the possible networks of habitats. If a pasture is not grazed for a number of years many species will be lost, but with surrounding pastures containing the species a re-colonization is possible. A further examination of the biodiversity at neighbouring farms shows that farms located in areas where neighbours have high biodiversity will on average have higher costs than other farms. The elasticity is similar to that of biodiversity at the own farm, a possible explanation for this being that landscape characteristics influence both biodiversity and the farmers’ cost structure.

The deliverable “The disadvantage of farming in marginal agricultural regions and the potential loss of environmental values” discusses the long term development of biodiversity and its dependence on suitable habitats for plant and animal species. In this context, agriculture has an important role as a provider of biodiversity. This has been highlighted in the Swedish national environmental objectives, which explicitly point out the importance of agricultural landscapes. Still, a farm’s provision of biodiversity is a joint production with commercial commodities, which implies that the supply of biodiversity may be sensitive to farm performance. If farms in a biodiversity-rich region are less efficient than others, there is a risk of a decline in agricultural production and hence a risk of a loss of biodiversity in these regions. This deliverable focus on the relationship between Swedish farms’ efficiency and their provision of biodiversity, which is done by combining economic data (from the European FADN database) with information on biodiversity indicators (from the Swedish TUVA database) such as the number of plant species growing in a particular pasture. In total 266 farms with animal production are identified in 2003 and, for these farms, efficiency is estimated using the Data Envelopment Analysis (DEA) method. The efficiency score is then used in a second stage in order to assess the relationship between efficiency and biodiversity.

The first set of results stems from a regional analysis using a meta-frontier framework to determine regional (defined by NUTS1 regions) production frontiers. One finding is a distinct technological pattern showing Northern Sweden as a region with a technological disadvantage when it comes to agricultural production, which is as expected due to the cold climate and short growing season of the northern latitudes. However, only two (out of 60) plant species are primarily dependent (have more than 90 % of their locations) on the northern agricultural landscapes.

The second set of results is based on an analysis with the number of different species of vascular plants as the principal indicator of biodiversity. The main result is a negative correlation between farm efficiency and biodiversity in proximity to the farm. On the other hand, the provision of biodiversity at farm level is not correlated with efficiency. The role of biodiversity in surrounding areas stresses the importance of properties common to all farms in biodiversity-rich areas, which is strengthened by the lack of any significant relationship between biodiversity and efficiency when the landscape properties are removed from the biodiversity indicator.

The overall aim of deliverable “Organic farming: implications for costs for production and provisioning of environmental services” have been to contrast organic and conventional forms of commodity production in terms of costs and environmental performance. Specific objectives have been applied to the ‘general’ cost of production model (GECOM model) developed in the FACEPA project to organic farms, to compare GECOM results for organic farming to data from other national studies as part of a (quasi-)validation, to discuss production costs in organic farming in the light of the structure of the organic farming sector and the respective policy environment in selected EU Member States, and to explore the potential of FADN systems for deriving environmental impacts at farm level, calculating and comparing selected indicators for organic farms.

In many countries, private organic standards play an important role, and these may affect costs if they differ to the EU organic regulation. In several countries (Denmark, France, the Netherlands, the United Kingdom, and Sweden) these private standards with regard to livestock feed and housing are likely to increase respective costs. Other examples identified include social standards in the Italian Organic Standards which may have an impact on labour costs, and additional environmental requirements for organic farms in Poland. On the other hand, in some cases derogations from the EU regulation may reduce costs (e.g. derogation for conventional seed in Poland in 2006). In some countries, certification is subsidised or covered by the state (e.g. Denmark), which reduces costs for farmers accordingly. The availability of data from existing studies on production costs in organic farming for validation purposes is very limited. Information was therefore collected for selected study countries from national experts. Still, the challenges experienced during the collection and processing of cost data from other sources for conventional farming were amplified for organic farming not only due to even fewer sources being available, but also due to the greater importance of methodological issues concerning the treatment of farm-produced production factors and stronger interlinkages between all farm processes. These limitations need to be taken into account when interpreting and using the collected cost data.

The GECOM model has been applied to the EU as well as the German national FADN. In the EU-FADN, a variable identifying organic farms is included since 2000, however only a few countries (Austria, Germany, Denmark, the United Kingdom, France, Italy, Poland, and Sweden) have a data set for organic farms which is big enough for analysis. Only for Austria and Germany is the organic sample big enough for all of the years from 2000-2007, while in most other countries samples are often small in the period 2000-2003. However, the data availability for these countries increases from 2004 onwards. FADN data for the new member state Poland have been included since the country joined the EU in 2004. To increase robustness of results and facilitate interpretation, GECOM estimates have been averaged over the time period where samples were big enough for econometric estimations.

Generally, the production cost estimates for organic milk match the reference data very well, with respect to absolute values as well as with respect to cost structures. Estimated production costs I (excluding cost of labour, land and capital) range from 200 to 300 €/t of organic milk in most of the countries analysed, with Germany having the highest costs (340 €/t) and Poland showing the lowest costs (110 €/t). In Austria and France, cost of milk production is only slightly higher in organic compared to conventional farming, and costs structures of the two farming systems are very similar. In Denmark, Germany, Italy and Sweden, production costs for organic milk are significantly higher than for conventional milk. This is due to higher feed costs (especially for Italy) and, in the case of Germany, higher miscellaneous costs. In Poland, estimated costs of milk production is lower in organic farming than in conventional farming which might be caused by very extensive organic production systems, and the rather high feed costs in conventional farms. The general relations between organic and conventional production costs remain the same when including the costs for labour, land and capital (production costs II, full costs), however the gap to conventional farming increases in Italy (due to higher labour costs), Poland and Sweden (due to higher capital costs) and especially in the case of Germany (due to higher costs for all three factors). Estimated full costs in the old member states range from 350 €/t (France) to 490 €/t (Germany). With the exception of Poland, the market price for organic milk is higher than for conventional milk in all of the countries. The estimates indicate highest subsidies per tonne of milk in Austria and lowest in France and Denmark. Total returns and subsidies cover total costs only in France, Poland and Italy. For wheat, the level and structure of estimated costs and the cost information from other sources match well only for Denmark. The differences for the other countries are partly due to remaining intractable differences in cost aggregation and methodological approaches, however may also be due to the fact that GECOM results for crop products are often less robust.

For a more detailed analysis of production costs for organic wheat and milk, the GECOM model is applied to German national FADN data from 2000 to 2009. To increase the robustness of results, a statistical method for outlier detection was used. An above average rate of outliers was detected for field crop farms, large farms and legal farms (corporate farms). The improvements from the removal of outliers were most obvious for milk, as estimated production costs were much less volatile over years. Production costs as well as returns of wheat are much higher for organic farms than for conventional farms. Conventional farms show much higher costs for fertilizer and crop protection, whereas organic farms have very high costs for contract work and depreciation, and a higher net value added. Production costs as well as returns for organic milk are about 50 €/t higher than those of conventional farms. Organic farms have much higher costs for home-grown feed and slightly higher costs for purchased feed and depreciation, and a slightly higher net value added than conventional farms. The results also indicate a cost advantage of farms which are specialised in organic milk production compared to more mixed farm types. In this deliverable the possibility have been investigated of using farm economic data to provide environmental indicators on which farms can be assessed. A selection of environmental indicators was made based upon previous research. These assess the level of inputs (fertiliser, crop protection, purchased feed), intensity of the agriculture (intensification indicator, LUs per forage area), participation in agri-environmental activities (monetary receipts from agri-environmental schemes), diversity of cropping (Shannon index), and availability of wildlife habitats (proportion of land that is permanent grassland, woodland, or fallow). These indicators were investigated using Farm Business Survey data for England and Wales from 2008-09 and 2009-10. A selection of indicators has been used to compare organic and conventional farms across robust farm types using FBS data. Each indicator was assessed across all farms within the survey and across all organic and all conventional farms. The indicators were then calculated for each farm type and the split of these into organic and conventional.

The results showed that there are statistically significant differences between organic and conventional farms in terms of fertiliser cost, crop protection cost, intensification, and agri-environment scheme payments. These results suggest that organic farms are less intensive with lower fertiliser and crop protection use and tend to be involved in more agri-environment schemes than conventional farms. In contrast there is no significant difference between organic and conventional farms with regards to crop diversity except for mixed and lowland grazing livestock farms where organic farms have a statistically significantly lower diversity. There is also no significant difference between organic and conventional farms in terms of the proportion of land that is woodland, permanent grass or fallow except for general cropping farms where organic farms generally have a higher proportion. With regards to purchased feed costs and livestock stocking densities, whether there is a significant difference between organic and conventional farms depends on the robust farm type. Purchased feed and purchased concentrate costs for dairy farms only show differences of low statistical significance with organic farms having slightly higher costs per livestock unit (possibly due to higher organic feed prices rather than higher usage). For lowland grazing livestock there is a more strongly significant difference with organic farms having lower purchased feed costs. This is also reflected in LFA grazing livestock farms although with a slightly lower significance. In general purchased feed or concentrate costs are not significantly different between conventional and organic mixed farms. Dairy and lowland grazing livestock farms show significant differences in stocking density between organic and conventional management with organic farms tending to have lower stocking densities. The difference for LFA grazing livestock farms is only significant at the 5% level, perhaps reflecting the fact that such farms tend to be unable to support larger stocking densities regardless of management system.

In general it appears from the analysis that organic farms are less intensive than conventional farms, however organic farms appear to have less cropping (and potentially less habitat) variety as reflected by some of the Shannon index results. It would also appear that grazing livestock farms in general may be beneficial to the environment as assessed using this particular set of indicators. It is shown from the analysis presented here that it is possible to use economic data such as the FBS to provide some information on the environmental performance of farms and to compare this across different types if farms and farming systems. In particular it would be of great interest to combine some of the indicators into an overall score that took account of intensity, crop variation, variation in habitat and stocking rates, as well as agri-environment payments. Although an indirect measure of environmental performance may never achieve a perfect assessment a combined score could be weighted to reflect the relative importance of the various factors.

4.1.3.9 Evaluation of public policies

In work package 9 cost functions have been estimated for several EU member countries using the statistical software STATA. The objectives have been as follows: i) to provide and test farm economic models for evaluating ex-post policy measures, ii) to provide and test farm economic models for evaluating ex-ante policy reforms, and iii) to undertake preliminary comparative analyses on these ex-post and ex-ante evaluations across member states and regions. This research component have designed and developed key economic models that use the production cost estimates from work package 3 but also from work packages 5, 6 and 7 to evaluate various agricultural measures on agricultural, environmental, financial and socio-economic indicators using FADN data and additional EUROSTAT data. In particular, this work package have aimed: (a) at providing and testing farm economic models suitable for evaluating ex-post measures that have been implemented as components of: (i) the reform of the Common Agricultural Policy (CAP) for the incumbent Members States since 1992, and (ii) the accession of the new Member States since their adhesion to the EU; (b) at providing and testing farm economic models suitable for evaluating ex-ante measures that would be most likely implemented as components of the continuation of the reform of the CAP; (c) at providing a preliminary comparative analysis on these ex-post and ex-ante evaluations across Member States and regions.

The main outcome of this work package consisted in providing functioning farm-level economic models based on FADN cost estimates that are appropriate for ex-post as well as ex-ante policy analysis. Based on the model developed by UCL, costs functions have been estimated for dairy, cattle and crop farms in Bavaria and Lower Saxony, and ex-post evaluation of the impact of policy changes have been analysed.

4.1.3 A description of the main S&T results/foregrounds

4.1.3.1 Concepts

FADN is a highly harmonised EU-wide survey. Much time and effort is spent in the Member States and at the EC in order to produce comparable and reliable information for agricultural activities. Despite the high level of harmonisation there are several differences concerning the collection of FADN data and the interpretation of FADN regulation which are very important to keep in mind while constructing the cost of production model. In the Member States, the data are collected according to EU-wide standardised guidelines. However, there are national distinctions concerning data availability. Further the accounting system of the FADN is quite different from the one used in financial accounting. In financial accounting the key characteristic is reliability. The aim of the FADN is to serve the decision making process in the EU, so the relevance, the comparability and the usefulness are important requirements. The information served by the FADN is less reliable than the data from financial accounting, because many times estimations have to be made and information has to be updated in order to meet the requirements of the FADN. Additionally, the comparability is strongly influenced by national distinctions. The cost accounting principles of IFRS and FADN are basically the same. Differences can be found at the recording of farm produced feeding stuffs, due to the diverse valuation methods (on farm cost versus farm gate price) of costs. Compared to IFRS, investment subsidies increase income in FADN and profit tax is not deducted from the income in FADN. Concerning assets, according to FADN, only operational resources are considered whereas according to IFRS the total assets of an enterprise (for example including financial assets) are to be indicated in the financial statement.

The requirements of FADN regulations are realised in different ways in the Member States. Data delivery in the Member States for FADN purposes is not totally harmonised. For example: FADN requires a valuation of agricultural land at net selling prices. Some countries follow this requirement while other countries value agricultural land different from these FADN specifications. These countries often distinguish between purchased land and other land. Another example is the difference between Member States in the valuation of quotas and delivery rights due to the very heterogeneous national rules. The valuation of biological assets is also very diverse. The problem of contract rearing also should be mentioned. The model should separate farms with receipts from contract rearing and also consider national distinctions regarding the recording of contract rearing.

Despite the high level of harmonisation there are several differences concerning FADN data collection and the interpretation of FADN regulations which are very important to keep in mind while constructing the cost of production model. It can be stated that FADN data collection has many differences in the Member States, thus special care must be taken to define harmonised European data for the cost model.

• The lower thresholds to get into the sample vary from 1 to 16 European size units (ESU). In some Member States upper thresholds are determined, as well (e.g. Netherlands).
• The FADN's representativeness is considerably different. In seven Member States of EU25 FADN farms represent less than 90 per cent in terms of Standard Gross Margin.
• Member States use different additional criteria for farms to get into the field of survey and the methods for defining the sample are also different.
• The year of FSS varies from Member State to Member State between 1999 and 2006. It has great importance, as there was a large structural change in CEE countries in recent years.

The differences between the Member States also vary regarding activities other than agricultural activities on the holding. This means that harmonisation in the area of “Other income on the holding” is difficult. The inclusion of ”Other gainful activities” varies from Member State to Member State. Member States decide in different ways whether a holding which has other gainful activity should be included in FADN or not. Regarding the cost of production model the most important problem is that the cost items cover various activities and there are no data showing the division of the costs. And due to the different interpretation of other gainful activities it is difficult to determine whether a cost connects to an agricultural or non-agricultural activity.

4.1.3.2 Comparing the characteristics of EU and national FADN

Most work packages in the FACEPA project are making use of the FADN data because it is the only source of micro-economic data for agriculture in the EU which is harmonised. In spite of the existence of common rules adopted by the Member States to inform the European FADN, the majority of countries use special methods for their national needs. No national sample is exactly the same as the EU-FADN over the period 1995-2005. Below two points are addressed related to methodology: the statistical point (sample and weighting) and the accountancy point (depreciation).

The field of survey of the EU FADN is a well-defined group of farms out of the total number of farms and includes commercial farms that exceed a minimum threshold of agricultural production measured in European size units. This threshold varies among Member States, and some countries apply additional criteria for excluding farms from the survey. Given the great variety within the FADN field of observation, stratified sampling is applied to ensure that the sample of farms adequately reflects this heterogeneity. Stratification as well as procedures and methodology to select sample farms vary among Member States. For example, non-random sampling and voluntary participation may introduce sampling bias. The differences in selecting, stratifying and sampling between Member States can also result in differences in national and EU FADN weights and affect representativeness. Using different SGMs in the national FADN has impacts on the selection of farms (by effect on economic size) but also on the specialisation of the farm holding. Germany, for national needs, keeps in the national FADN the farms with a SGM between 8 and 16 ESU which are under the European threshold. Italy excludes from EU-FADN almost empty cells (economic size x type of farming). This represents 0.3% of the national SGM and 489 farms. Until 2003, France used an additional criterion to split the field of observation for selection of the national sample. The sub-sample refers to a mode of data collection. The Netherlands use random selection with a specific stratification. The size classes are different within types of farming; the Dutch FADN uses strata for organic farming and subtypes of farming. Few farms are missing from the field of observation of the European FADN. Germany computes SGMs from the 5-year average instead of the 3-year average in the other countries. The sub-stratification used by the Netherlands in selection is also used for weighting. For the stratification in the weighting system, Hungary uses the legal form of the enterprise to separate private farms from economic organizations.

Because depreciation plays an important role in the Gross Farm Income and because the regulation does not impose any method of depreciation, it is necessary to analyse the applied methods. Depreciation depends on three parameters: estimation of fixed assets (replacement value/acquisition cost), method for computing depreciation (linear/digressive), rate of depreciation. Concerning the estimation of fixed assets, the National FADN is different from EU-FADN for Germany and to a lesser extent for Hungary. These countries adopt the acquisition cost. But in respect to regulations, Germany corrects depreciation to inform the EU-FADN. In general, depreciation based on replacement values is about 1/3rd higher than those based on acquisition costs. For computing depreciation, the FACEPA partners use the same method in the national FADN as the European one, even if each of them prefers the linear method or the digressive method.

Regarding outputs, additional variables are available in all National FADN of the FACEPA partners, except for Bulgaria. The young Bulgarian FADN, imposed by the Accession Treaty, uses the EU-FADN classification. To comply with the EU-FADN, member states often aggregate national output variables. Knowing the details of the products grouped together is useful for two purposes: to specify the content of data but also to point out differences between countries. For organic products the available information is limited. Four countries (Belgium, France, Germany and Italy) limit the data to the European regulation which is a separation variable to mark the farm. For the mixed organic farms it is not possible to distinguish organic products, except for Italy where organic product is an option among several possibilities. The Netherlands uses a sub-sample of organic farms and Germany collects optional information about sales.

4.1.3.3 Statistical problems associated with the EU FADN database

The strengths and weaknesses of the EU FADN database for estimating cost of production from a statistical point of view have been discussed and analysed. To this end, the approach and implementation of sampling and weighting in EU FADN is summarized and size and distribution of used weights are examined. Further the coverage and representativeness of EU FADN data have been analysed, and also the question of the representativeness of EU FADN with respect to organic farming was addressed. The overview of the sampling and weighting used in the EU FADN data system highlights the extent to which Member States use different methodologies for selecting, sampling and stratifying farms, and the possible impacts this may have on representativeness, coverage and weighting of the data. Factors of concern include:

• the country-specific differences between the field of observation and the total population,
• the use of additional criteria and sub-samples as well as different SGMs for sampling plans,
• resulting differences in national and EU FADN weights and the representativeness,
• the potential sampling bias introduced by non-random sampling and voluntary participation in some Member States.

Weighting factors are used to extrapolate the EU FADN sample. These weighting factors also have to be taken into account when specifying a cost of production model which aims to reflect the input-output allocation on the Member State level to prevent distorted results. The larger the variation in the weighting coefficients is, the greater the need for their incorporation in a cost production model that aims to produce information at the level of the Member States. To investigate the size and variation of weighting coefficients across Member States and farm types, a descriptive analysis is carried out for the year 2005. The analysis of the weighting coefficients shows that the variation of weights is high, especially in the nMS. The results also reveal that some Member States have very high weights which might lead to lower reliability of the cost estimates. Some of the causes for the high weights could be very low heterogeneity, sampling procedure and differences in national and EU FADN sampling and weighting methodology. In this case use of national FADN weights can be an option to improve the reliability of the estimates. Further research is needed to analyse the impact of differences in EU and national weights on production costs estimates.

The representativeness of EU FADN has been examined by comparing a set of various structural variables between EU FADN data and the FSS for all 25 Member States. Further, it has assessed the change in the coverage and representativeness through time by comparing the representativeness of the EU FADN in 1995 and 2005. The low coverage for the variable “number of holdings” is suggested to be due to the fact that the number of farms with a SGM below the country-specific ESU threshold is significant. At the same time the much higher coverage for the other variables show that those farms below the ESU threshold have a minor share in UAA or number of dairy cows, for example. The coverage for the “number of holdings” versus the other variables are not surprising, as the fundamental aim of the EU FADN is not to gain a high degree of coverage with respect to the number of holdings in the Member States, but with respect to the agricultural activity measured in total SGM. FADN data covers a larger part of the whole population in the oMS than in the nMS. This is particularly true for the number of farms where the average values for the EU-15 are more than twice as high as for the nMS. The results show that there is considerable potential for focusing production cost estimation on samples of specialized farms, as these often have a very high share in the total production of selected products. Nevertheless sample sizes need to be checked in each case to ensure robust estimates. The findings indicate that, on an EU average, the coverage and representativeness is relatively large for the variables under study. However, considering the single Member States reveals that in some cases significant differences exist cross-sectional. In view of the changes over time, it is shown that the coverage and representativeness increased from 1995 to 2005. It must however be noted that the empirical analysis and the conclusions drawn are based on the comparison of structural variables such as hectares of major crops and numbers of specific livestock between the sample and the population. It gives no final answer whether estimations of costs of production using EU FADN will reflect the true population value. Validating the estimation results using cost calculations from other sources are therefore necessary topics for future research.

Another analyse that was carried out was the coverage of organic farming. An identifier variable for organic farms was introduced in the EU FADN in 2001. In 2006, the sample includes accounts from more than 3000 fully organic farms; however, sample sizes vary strongly between countries. Currently, the number of organic farms is small and hence the sample will only allow an econometric estimation for few countries. Generally, the potential for estimating cost of production based on specialized organic farms is low, due to small sample sizes and the often higher diversity of production structure in organic farms. As organic farming is not a stratification criteria employed when calculating the EU FADN weights, the reliability of these weights might be low, especially in countries where organic holdings represent only a small proportion of farms. Another problem is that some countries do have strata for organic farms on national level which are not taken into account in the EU weighting. More robust and representative estimates may be achieved using national FADNs which in some countries include a higher number of organic farms, and/or allow a weighting of these farms.

4.1.3.4 Implementation and validation of the ‘general’ cost of production model

Within the FACEPA project, considerable resources have been allocated to implement and validate the ‘general’ cost of production model (GECOM model). The outcome of the related work has been published in three connected reports. The first two reports describe the implementation, the validation and the results from the GECOM on the basis of national farm accountancy data networks (FADN) and the EU FADN, respectively. The third report provides an overall synthesis and conclusions.

First, general aspects of model specification and estimation, e.g. the list of outputs and inputs, are shortly summarized. In general, the more detailed the list of outputs the more accurate is the specification of the model. On the other hand, a longer list of outputs increases the probability that estimated cost coefficients of some outputs will be less robust and precise. This trade-off needs to be considered in view of the specific data for which cost estimates are sought, and is thus a decision to be taken for each application. The decision on the list of inputs, specifically the treatment of marketable farm-grown intermediates and the inclusion of subsidies, and the resulting income indicator needs to be taken in view of the research question to be analysed.

The implementation of a seemingly unrelated regression algorithm in the software package SAS allows for a very fast and stable estimation of the GECOM. However, one issue which has been observed by all partners was the frequent occurrence of negative cost coefficients. While often not statistically significant, these negative coefficients proved to be a major concern in the validation of the model. The assumption of a common Leontief technology underlying the production function of all farms clearly constitutes a strong constraint on the applicability. Some of these limitations can be offset by selecting farm samples with homogenous technologies, and therefore as far as possible, farm samples should be stratified according to applied technologies.

During the testing of the GECOM model, it was repeatedly highlighted that even small and/or infrequent data errors can have a significant impact on estimated cost share coefficients. Their identification has proved to be a major challenge, despite extensive plausibility checks by national and EU authorities. For the applications within the FACEPA project, a multivariate methodology based on the Mahalanobi distance measure to eliminate the data errors was tested, which generally improved stability of results over time, however often excluded an undesirably large share of observations.

The validation process showed that to ensure a correct interpretation of results from other sources in relation to GECOM cost estimations, a very careful examination of these sources is needed with respect to the approach used, the definitions of costs and cost categories, the definition and calculation of imputed costs and the scope of the costs allocated. In general, the validation of the GECOM by comparing results to those of other studies as well as by a review of estimates by national experts highlights that the quality of estimates differs by country. Overall, level and trend of total costs of the main products wheat, milk and pigs were judged to be plausible. Generally, estimated cost for crop products were less robust and in several cases implausibly variable over time. In many countries, also the estimates of individual cost components were assessed to be realistic, especially of direct costs, while the values for overheads and depreciation were less reliable.

An emerging key issue for the dissemination of results from the GECOM was that existing national “conventions” for the definition and presentation of production costs differ from the ones used in the FACEPA project. Thus, when presenting results in a national context, appropriate care needs to be taken to thoroughly explain approach and definitions, and to reprocess cost estimates to match national conventions as far as possible. In the subsequent section, the outcome of the comparison of cost estimates based on EU and national FADN is presented and discussed. Differences in production costs estimates based on EU vs. national FADNs may in principle be caused by differences in samples, weighting factors and data. The analysis showed that while in many cases production costs estimates based on EU and national FADN are very similar, in numerous instances results differ significantly. In these latter cases, results based on the national databases were generally judged to be more plausible. This outcome is in line with prior expectations, as data in national FADNs are by nature more differentiated and closer to ‘original’ farm data, and national weights better reflect the actual sampling procedure than this can be the case for the weights derived by the static EU FADN weighting system.

Based on the experiences made with the GECOM application above, this report emphasizes that the generic software tool for implementing the GECOM should offer user-friendly options for carrying out basic data pre-checks, provide flexibility with respect to selection of samples and aggregation of outputs and inputs, and produce a clearly represented overview of the statistical significance of estimated coefficients.

As a general conclusion, the implementation, testing and validation of the GECOM showed that the model can provide plausible estimates of production costs for main products in most countries, reflecting developments over time as well as cost composition, while results for products with smaller output shares were often not convincing and highly variable over time. However, the experiences also showed the indispensable necessity of pre-checking the data in each case, dealing with outliers and taking into account details and changes in the data definition and collection. A general conclusions from the experiences gained is that no “simple” application of one general model is possible for all samples and products. An analysis and validation of results by experts (i.e. of both FADN data and agricultural production systems in the analysed samples) will always be needed.

Key issues for future research concerning the GECOM are the estimation of production costs for meat products, and the search for improved, robust and transparent outlier detection methods. Other estimation approaches (e.g. panel and entropy estimators) and/or model specifications (flexible functional forms) have an obvious potential for dealing with several of the limitations of the current methodology identified in this report, however, more research is required to facilitate a broader and more robust application of these approaches to estimate production costs for all EU Member States.

4.1.3.5 Dissemination and valorisation of the production cost model software

In the European FADN, the costs, detailed by category (seeds, fuel etc), are available for agricultural holdings. The developed computer software is designed to allocate these costs to different productions. The concept and start point was a model built by INRA in the 2000’ years. The software was established from the works conducted by vTI in work package 3 (the cost of production (GECOM) model). This computer software allows generating cost of production for all sorts of agricultural commodities using i) FADN data and ii) the SYSLIN procedure of the SAS software. It is menu-driven and includes a detailed interface so that users of this computer software could operate it using an EXCEL worksheet environment. The interface relies on the JAVA system, and the computer software integrates the recent specifications and developments made by vTI on the generic cost of production model. INRA presented a first and preliminary version of this computer software in May 2010 at the 5th general FACEPA meeting in Uppsala. At the 7th General Meeting in Den Haag on March 31 and April 1, 2011, it was agreed to change the present name of the production cost software and call it from then and onwards for the FACEPA model.

The FACEPA model is programmed in SAS language and runs on the Enterprise Guide SAS module. The model estimates input – output coefficients from EU FADN data. The technical coefficients of production are estimated by using the SAS ‘PROC SYSLIN’ procedure. From these coefficients, it is possible to deduce costs per quintal, cost per hectare or cost per animal. The model runs for a single country over a period of several years or for a single year for one or more countries. It is possible to add a classification variable to the list of variables given, in order to obtain results per sub-group (per region, or per type of farming). It is also possible to obtain the results on a specific field.

The model takes into account automatically the total production. The user defines a list of crops outputs and livestock outputs and the balances are automatically calculated. For inputs, variable costs and fixed costs are distinguished, but the total input is not given. The choice of input determines the income indicator, which is integrated in the model. For example, if the user takes only the intermediate consumptions, the income indicator is the added value. The files include imputed costs for family-owned factors. It is possible to include these costs in the model, but this procedure was not tested by the vTI-team. More generally, it is necessary to be careful in interpreting the results on the fixed costs.

In the FADN data, the values of output are given at the production prices. In the model, subsidies are considered as negative costs. It is possible to select coupled or total subsidies. In the final output, subsidies are included in the "basic prices". This solution is in theory only coherent for coupled subsidies but it is possible to include Single Farm Payments (SFP) and the second pillar payments. Taxes on products are also automatically deducted in the calculation of basic prices.

It is possible to obtain information to judge the significance of the coefficients (Standard error, T value, P value). Different options are possible:

1) It is possible to delete outliers (see vTI reports).
2) The results could be weighted (or not) with the SYS02 variable, which corresponds to the weight of the holding in the sample used.
3) The production includes or not the production used on farm. If the user takes the production used on farms, it is necessary to put in the inputs seed and feed produced on farm.
4) The "Allocation of residuals" option allows you to edit the breakdown of individual costs. The residual difference between the estimated values and the real values calculated for each holding is distributed over the different products proportionally to the gross output (with home-grown consumption) or the gross product (without home-grown consumption). This option also creates three groups of intensity according to the holding coefficients calculated for each production.

4.1.3.6 Extensions and further applications of estimating cost of production

Various efficiency indicators for European Union (EU) countries included in the FACEPA project, Belgium, Estonia, France, Germany, Hungary, Italy, the Netherlands and Sweden have been analysed. The availability of long period datasets between 1990 and 2006, allowed concentrating on the long time trends in technical efficiency especially in Old Member States.

Two main approaches developed over time for analyzing technical efficiency in agriculture have been used: (1) The nonparametric data envelopment analysis (DEA); (2) the parametric stochastic frontier analysis (SFA). While the vast majority of empirical studies on technical efficiency in the agricultural sector mostly have utilized only one method to estimate their efficiencies, both methodological approaches to measure efficiency have been applied. In addition, most studies focus on a single country’s agricultural sector, thus the comparative analysis of the technical efficiency is rather scarce. The relative importance of specific subsectors and rationale of compilation more homogeneous sample have been taken into account, thus the analyse focuses on the field crops and dairy sectors separately. The analyse highlights the importance of easier availability to the farm level data, namely Farm Accountancy Data Network (FADN) data in the EU, which may provide interesting insights for policy makers on farm level technical efficiency and develop more targeted policy to improve efficiency in European agriculture.

Generally, all countries have relatively high levels of mean efficiency ranging from 0.72 to 0.92 for both field crops and dairy farms. Interestingly the majority of countries have better performance in dairy sectors in terms of higher levels of mean efficiency than in field crop production. A slightly decreasing trend however may be observed for all countries. Technical Efficiency estimates are largely in line with those obtained by previous studies.

Further work has been done to the issue of how relative performance of farms fluctuates in terms of technical efficiency over time. One hypothesis may be that many poorly performed farms remaining inefficient and some farmers are performing always very efficiently. Farms which are usually at the bottom or top of the efficiency ranking can be identified. However, the FADN data has an inherent problem for long time period analysis arising from its rotated panel nature, namely that not all the farms are observed for the whole period. In this respect, there is a need to calculate transition matrices in each consecutive year. Surprisingly stability analysis revealed that in average 60% of farms maintain their efficiency ranking in two consecutive years, whilst 20% improve and 20% worsen their positions for all countries. However, these ratios slightly fluctuate around these values for one year to next year. Mobility analysis ranks countries according to the mobility of SFA scores within the distribution. Farms in New Member States are more mobile than those in EU15.

The DEA estimation shows a similar declining trend on the development of technical efficiency over time except for the Swedish dairy sector showing an increasing efficiency trend. The total productivity changes have been investigated in two steps. First, a definite trend in total factor productivity changes was not found. Second, the question whether total factor productivity changes converge or diverge over time was addressed. By using panel unit root tests for estimations reveal a convergence of productivity across old EU member countries during analysed period. Finally, the total factor productivity changes were decomposed into its main elements. Field crop farm indicators generally present a significantly higher volatility than dairy farms. Random effect panel regression of Total Factor Productivity Change on its components shows Technological Change as being the significant positive driver for crop farms, whilst Technical Efficiency Change followed by Technological Change are the most important for dairy farms. In addition, no significant impacts of CAP reforms in 1992 and 2000 on total productivity changes could be found.

Technical efficiency scores have been used to obtain three distinct methods, Stochastic Frontier Analysis (SFA), Data Envelopment Analysis (DEA), Operational Competitiveness Ranking Analysis (OCRA), based on national Farm Accountancy Data Network (FADN) data. This have been done in order to analyse the impact of European Union (EU) accession and the influence of farm classification, more precisely farm type, upon the performance on field crop and dairy farms in three New Member States (NMS), Bulgaria, Estonia and Hungary.

Theoretical and empirical evidence have been provided that farm classification is subject for empirical analysis, because using FADN and conceptual (Hill type) typology may result in considerably different farm structures. The main outcome of this research is that individual farms are not equivalent to family farms as usually assumed in previous research. The findings stress that average size of individual farms is considerably higher than of family farms. Not surprisingly, an ambiguous pattern of farm performance emerged from different approaches irrespective to product groups and country. However, the majority of results confirm that the average performance of individual and family farms is weaker than that of the corporate farms: including companies, cooperatives, intermediate and non-family farms irrespective of the methods, product group and country.

The main conclusion point at for the second stage regressions that, employing efficiency estimates obtained with the three distinct methods (SFA, DEA and OCRA), yield rather diverging results. From a methodological point of view, one would expect that commonly used methods, i.e. SFA and DEA would result in dependent variables with higher explanatory power, and consecutively better specified second stage regressions. This was not the case. Determination coefficients were by far the highest in OCRA regressions, and these also produced the highest number of significant coefficients. Considering SFA and DEA methods, the efficiency scores obtained with the latter seem to be more appropriate for second stage regressions.

First, an assessment of the impact of farm types on farm performance was done. The simple mean comparison estimation showed there are significant differences in farm performance among farms in terms of legal form or farm organization. However, panel regression just partly confirms these results. The main reason is that a considerable number of farm type coefficients are not significant. Reference will be made only to those results, where estimations provide significant results. The impact of family and individual farms on farm performance is rather negative except for Estonian dairy farms, where the opposite effect was observed. The most striking result is that farm size is positively related to performance confirming that scale efficiencies do matter in these countries.

The final interest is the possible impact of the EU accession on the farm performance. With the exception of some regressions having OCRA scores as dependent variable, the EU accession proved to have negative effects upon farm performance, regardless of the country, sector or farm typology considered. Although this might not seem a plausible result at first, it has some logic behind, and it is not unprecedented. Through EU accession farmers got access to higher subsidies, but the public support received by farmers in the frame of the Common Agricultural Policy (CAP) may have a negative influence on their technical efficiency. As it has often been shown in agriculture, public support reduces farmers’ effort, implying greater waste of resources and thus further located from the efficient frontier.

The competitiveness of EU agriculture with respect to farm efficiency was the last analysed subject within work package 5. Most often competitiveness of a country’s given agricultural sector is analysed using trade and price data, the linkage between efficiency and competitiveness is quite ambiguous. In this research however, the analyses was based solely on FADN data, using some ideas – not directly supported by theoretical models - and policy directives, in order to link sector (field crop and dairy) and country specific (Belgium, France, Germany, Italy, the Netherlands, Sweden, Estonia and Hungary) farm performance through CAP payments to competitiveness. The unbalanced panel nature of the dataset allowed the estimation of three distinct models for both field crop and dairy farms: pooled OLS, random effect panel regression and fixed effects panel regression. Where relevant, the emphasised differences between EU-15 countries and New Member States (NMS) were represented in the sample. The decoupling process is captured in the analysis by the inclusion of a dummy variable. Results obtained are quite ambiguous. There are only a few common, easily interpretable results; rather they vary according to models applied and country discussed. The conclusion suggests that although the linkage between the evolution of technical efficiency scores and subsidies received, to have implication upon farms’ competitiveness, the classical gravity and trade based analysis offers better representation of individual counties competitiveness of a given sector.

4.1.3.7 Modelling farm technologies

It is well known that European FADN does not collect the information about the variable costs associated to the different farm activities but only the total variable cost at a farm level. This lack of information makes it difficult to evaluate the production allocation decisions without the use of other external sources (engineering information, literature, etc.). This method has the risk to not be able to differentiate the costs according to the specific farm specialization and size (economic and physic). Moreover, all the policy and market evaluations based on FADN database are based on estimations and not on explicit profitability value. The evaluation of the farm enterprises costs is the main issue of FACEPA project and also the methodology applied in work package 6 takes these concerns into account.

To solve the problem linked to the lack of FADN data on specific production costs per process or enterprise, the classical approach of PMP (by which the farm behavior is estimated in conditions of maximization of the gross margin) has been modified. One important phase of PMP is the calibration process, proposed for estimating the farm production decision component. The lack of information about the analytical costs poses a problem during the calibration phase of the model when the estimation of the cost function requires a non negative marginal cost of all production processes activated by the holding. The standard PMP calibration method has been modified to solve this problem, and in order to generate the observed production plan using the dual structure of the problem proposed by Paris and Howitt. In particular, the dual optimality conditions have been used directly in the estimation phase of the non linear function. This approach (qualified as an extension of the Heckelei’s PMP methodology) avoids the first phase of the classical PMP method by imposing first order conditions directly in the second cost function estimation phase. So, the model considers the information relative to the total farm variable costs available in the European FADN archive. This is an innovation, important because permits to perform analyses using FADN dataset without having to resort to parameters that are exogenous to the model. The application of this innovation in well illustrated in the deliverable “Methodology to assess the farm production costs using PMP farm models” in which a group of farms (35) selected from the Italian FADN dataset, has been used to estimate the costs of production. Information about hectares, yields, prices, subsidies and farm total variable costs have been used for estimating the specific costs associated to the farm production plan. Considering that, differently from European FADN database, in the Italian FADN the costs are allocated among the different production processes from the surveyors, these results have been compared with the estimated costs.

The results have been very interesting and for this reason, the Italian dataset has been used to validate the estimates obtained by the application of the PMP model to the Italian case. In fact, while for Belgium and Hungary case studies the lack of observed specific accounting costs has prevented the possibility to make a sort of validation, for the Italian regions (Veneto, Lombardy and Piedmont) the allocation made by the surveyors present in the national FADN dataset has permitted a comparison between the observed and estimated costs. With this respect, the methodology is not so different from that developed in Italy within the work package 3. Here the results of econometric estimations (GECOM model) have been compared with the observed costs following the same scheme. Inside INEA, responsible for the implementation of Italian FADN, the discussions about the results of both methodologies have been very interesting. The results obtained for Hungary and Belgium have been compared with the output of work package 5: for Belgium the results have been very close while for Hungary the PMP estimates have been much higher than the outcome in work package 5.

The different applications of the PMP model in work package 6 are illustrated in the deliverable “Methodology for the definition of case study farms and model structure for each case study”. Here the PMP model has been applied considering the Farm Type “arable crops” in Italy, Belgium and Hungary. All the information about farms (acreage, prices, yields, other earnings and total variable costs have been used to estimate two type of costs: the specific marginal accounting cost and the hidden marginal costs. While the sum of the first ones is equal to the total variable costs provided by the European FADN, the estimate of hidden costs is related to the part of the cost that eludes the farm accounting system but that is considered an important element inside the farmer’s decision process because it influences the production choices. The hidden cost is an opportunity cost influenced by different factors like the farmer’s experience, his risk attitude, and the market expectations and so on. All these variables are not evident in the accounting results but are taken into account in the observed production plan. So, the hidden cost became a very important component in the assessment of the economic situation of the farm, difficult to estimate.

In addition to this, the analysis has been carried out using a multivariate analysis based on principal component detection and the cluster analysis method which has contributed to reduce the variability of the information used in the estimation phase and to control the outliers. In particular, the estimation procedure seems to be very sensitive to the presence of outliers, so a preventive check is important to minimize the interference of out-of-range value. The cluster analysis permits to reduce this risk. Moreover, the presence of a high variability of the yield for some crops produces unreliable estimates in some cases, including those with a high number of observations. So, when the internal sample homogeneity is not so high it is important to stratify the territory or the sector in order to have an improvement in the statistical significance. An important result of this application is that the PMP model seems to have a good capacity to reproduce the observed accounting costs for cereals and, in general, for crops with a high level of homogeneity in prices and yields. It is important to reduce the variability as much as possible and to find an adequate method to group the farm like sector and territorial stratification or multivariate methods. Also the estimation of the hidden costs can be considered as an important result of this application because of it is an element very difficult to assess but as fundamental in the definition of the production plan.

The deliverable “The effect of the single farm payment on cost function and production function” in work package 6, adds another element to the evaluation. The first is the possibility to change technologies in the production plan and the second concerns the income variation and new land allocation according to policy scenarios including CAP reform and increasing market prices. The possibility to change the production plan, differentiating between crops already grown into the farm and new crops, it is of great interest and the structure of the PMP model implemented in FACEPA permits to have a production function (the average technology) and the related variable cost. An additional “delta” represents the position of each farmer with respect to the average technology. The PMP model estimates the “latent information” related to production function and the associated variable crops that is used when economic conditions become favorable for the activation of new production function. The interesting thing is that the latent information can be used relatively to a given crop that is not grown in the farm. The analysis has been applied to Veneto region, considering arable crops and introducing the condition imposed by the Health Check reform in three scenarios, with different yield levels. It is very important that the result obtained from the new crop represented by sorghum is used for energy production. The PMP model can be used to evaluate under which economic conditions the new crop can be inserted in the production plan of the farms and what can be the impact on the environment.

4.1.3.8 Cost of production and the environment

In work package 7 the deliverable “The influence of landscape services on farm costs: The case of Swedish milk farmers” uses information on the economic performance and biodiversity provision of 304 Swedish milk farmers to report research on the relationship between biodiversity and the cost of farming. The biodiversity indicators are based on biological field studies mapping the existence of valuable species in the agricultural landscape while farms’ marginal costs are estimated using a flexible cost function.

A positive correlation, unconditional as well as conditional, is found between marginal costs and biodiversity. This relationship is valid when it comes to the binary choice of managing permanent pastures or not, and to providing more biodiversity for those already with valuable pastures. For the main indicator - number of species of vascular plants – a one percentage increase in biodiversity is related to a 0.03 per cent higher marginal cost. If instead biodiversity is incorporated into the farm cost function and assumed to be a variable output, then also an increase of biodiversity boosts marginal costs could be found. The results therefore support a competitive relationship between the provision of biodiversity and farms’ cost structure. This also holds for different types of outputs, but it is more pronounced between biodiversity and beef production.

Weighting the number of species with rarity gives a similar but somewhat weaker relationship, which suggests that the landscape characteristics of pastures with rare species are not more unfavourable for modern agricultural production than other pastures. However, from a biological perspective the landscape properties are of course important for biodiversity, not least the possible networks of habitats. If a pasture is not grazed for a number of years many species will be lost, but with surrounding pastures containing the species a re-colonization is possible. A further examination of the biodiversity at neighbouring farms shows that farms located in areas where neighbours have high biodiversity will on average have higher costs than other farms. The elasticity is similar to that of biodiversity at the own farm, a possible explanation for this being that landscape characteristics influence both biodiversity and the farmers’ cost structure.

The deliverable “The disadvantage of farming in marginal agricultural regions and the potential loss of environmental values” discusses the long term development of biodiversity and its dependence on suitable habitats for plant and animal species. In this context, agriculture has an important role as a provider of biodiversity. This has been highlighted in the Swedish national environmental objectives, which explicitly point out the importance of agricultural landscapes. Still, a farm’s provision of biodiversity is a joint production with commercial commodities, which implies that the supply of biodiversity may be sensitive to farm performance. If farms in a biodiversity-rich region are less efficient than others, there is a risk of a decline in agricultural production and hence a risk of a loss of biodiversity in these regions. This deliverable focus on the relationship between Swedish farms’ efficiency and their provision of biodiversity, which is done by combining economic data (from the European FADN database) with information on biodiversity indicators (from the Swedish TUVA database) such as the number of plant species growing in a particular pasture. In total 266 farms with animal production are identified in 2003 and, for these farms, efficiency is estimated using the Data Envelopment Analysis (DEA) method. The efficiency score is then used in a second stage in order to assess the relationship between efficiency and biodiversity.

The first set of results stems from a regional analysis using a meta-frontier framework to determine regional (defined by NUTS1 regions) production frontiers. One finding is a distinct technological pattern showing Northern Sweden as a region with a technological disadvantage when it comes to agricultural production, which is as expected due to the cold climate and short growing season of the northern latitudes. However, only two (out of 60) plant species are primarily dependent (have more than 90 % of their locations) on the northern agricultural landscapes.

The second set of results is based on an analysis with the number of different species of vascular plants as the principal indicator of biodiversity. The main result is a negative correlation between farm efficiency and biodiversity in proximity to the farm. On the other hand, the provision of biodiversity at farm level is not correlated with efficiency. The role of biodiversity in surrounding areas stresses the importance of properties common to all farms in biodiversity-rich areas, which is strengthened by the lack of any significant relationship between biodiversity and efficiency when the landscape properties are removed from the biodiversity indicator.

The overall aim of deliverable “Organic farming: implications for costs for production and provisioning of environmental services” have been to contrast organic and conventional forms of commodity production in terms of costs and environmental performance. Specific objectives have been applied to the ‘general’ cost of production model (GECOM model) developed in the FACEPA project to organic farms, to compare GECOM results for organic farming to data from other national studies as part of a (quasi-)validation, to discuss production costs in organic farming in the light of the structure of the organic farming sector and the respective policy environment in selected EU Member States, and to explore the potential of FADN systems for deriving environmental impacts at farm level, calculating and comparing selected indicators for organic farms.

In many countries, private organic standards play an important role, and these may affect costs if they differ to the EU organic regulation. In several countries (Denmark, France, the Netherlands, the United Kingdom, and Sweden) these private standards with regard to livestock feed and housing are likely to increase respective costs. Other examples identified include social standards in the Italian Organic Standards which may have an impact on labour costs, and additional environmental requirements for organic farms in Poland. On the other hand, in some cases derogations from the EU regulation may reduce costs (e.g. derogation for conventional seed in Poland in 2006). In some countries, certification is subsidised or covered by the state (e.g. Denmark), which reduces costs for farmers accordingly. The availability of data from existing studies on production costs in organic farming for validation purposes is very limited. Information was therefore collected for selected study countries from national experts. Still, the challenges experienced during the collection and processing of cost data from other sources for conventional farming were amplified for organic farming not only due to even fewer sources being available, but also due to the greater importance of methodological issues concerning the treatment of farm-produced production factors and stronger interlinkages between all farm processes. These limitations need to be taken into account when interpreting and using the collected cost data.

The GECOM model has been applied to the EU as well as the German national FADN. In the EU-FADN, a variable identifying organic farms is included since 2000, however only a few countries (Austria, Germany, Denmark, the United Kingdom, France, Italy, Poland, and Sweden) have a data set for organic farms which is big enough for analysis. Only for Austria and Germany is the organic sample big enough for all of the years from 2000-2007, while in most other countries samples are often small in the period 2000-2003. However, the data availability for these countries increases from 2004 onwards. FADN data for the new member state Poland have been included since the country joined the EU in 2004. To increase robustness of results and facilitate interpretation, GECOM estimates have been averaged over the time period where samples were big enough for econometric estimations.

Generally, the production cost estimates for organic milk match the reference data very well, with respect to absolute values as well as with respect to cost structures. Estimated production costs I (excluding cost of labour, land and capital) range from 200 to 300 €/t of organic milk in most of the countries analysed, with Germany having the highest costs (340 €/t) and Poland showing the lowest costs (110 €/t). In Austria and France, cost of milk production is only slightly higher in organic compared to conventional farming, and costs structures of the two farming systems are very similar. In Denmark, Germany, Italy and Sweden, production costs for organic milk are significantly higher than for conventional milk. This is due to higher feed costs (especially for Italy) and, in the case of Germany, higher miscellaneous costs. In Poland, estimated costs of milk production is lower in organic farming than in conventional farming which might be caused by very extensive organic production systems, and the rather high feed costs in conventional farms. The general relations between organic and conventional production costs remain the same when including the costs for labour, land and capital (production costs II, full costs), however the gap to conventional farming increases in Italy (due to higher labour costs), Poland and Sweden (due to higher capital costs) and especially in the case of Germany (due to higher costs for all three factors). Estimated full costs in the old member states range from 350 €/t (France) to 490 €/t (Germany). With the exception of Poland, the market price for organic milk is higher than for conventional milk in all of the countries. The estimates indicate highest subsidies per tonne of milk in Austria and lowest in France and Denmark. Total returns and subsidies cover total costs only in France, Poland and Italy. For wheat, the level and structure of estimated costs and the cost information from other sources match well only for Denmark. The differences for the other countries are partly due to remaining intractable differences in cost aggregation and methodological approaches, however may also be due to the fact that GECOM results for crop products are often less robust.

For a more detailed analysis of production costs for organic wheat and milk, the GECOM model is applied to German national FADN data from 2000 to 2009. To increase the robustness of results, a statistical method for outlier detection was used. An above average rate of outliers was detected for field crop farms, large farms and legal farms (corporate farms). The improvements from the removal of outliers were most obvious for milk, as estimated production costs were much less volatile over years. Production costs as well as returns of wheat are much higher for organic farms than for conventional farms. Conventional farms show much higher costs for fertilizer and crop protection, whereas organic farms have very high costs for contract work and depreciation, and a higher net value added. Production costs as well as returns for organic milk are about 50 €/t higher than those of conventional farms. Organic farms have much higher costs for home-grown feed and slightly higher costs for purchased feed and depreciation, and a slightly higher net value added than conventional farms. The results also indicate a cost advantage of farms which are specialised in organic milk production compared to more mixed farm types. In this deliverable the possibility have been investigated of using farm economic data to provide environmental indicators on which farms can be assessed. A selection of environmental indicators was made based upon previous research. These assess the level of inputs (fertiliser, crop protection, purchased feed), intensity of the agriculture (intensification indicator, LUs per forage area), participation in agri-environmental activities (monetary receipts from agri-environmental schemes), diversity of cropping (Shannon index), and availability of wildlife habitats (proportion of land that is permanent grassland, woodland, or fallow). These indicators were investigated using Farm Business Survey data for England and Wales from 2008-09 and 2009-10. A selection of indicators has been used to compare organic and conventional farms across robust farm types using FBS data. Each indicator was assessed across all farms within the survey and across all organic and all conventional farms. The indicators were then calculated for each farm type and the split of these into organic and conventional.

The results showed that there are statistically significant differences between organic and conventional farms in terms of fertiliser cost, crop protection cost, intensification, and agri-environment scheme payments. These results suggest that organic farms are less intensive with lower fertiliser and crop protection use and tend to be involved in more agri-environment schemes than conventional farms. In contrast there is no significant difference between organic and conventional farms with regards to crop diversity except for mixed and lowland grazing livestock farms where organic farms have a statistically significantly lower diversity. There is also no significant difference between organic and conventional farms in terms of the proportion of land that is woodland, permanent grass or fallow except for general cropping farms where organic farms generally have a higher proportion. With regards to purchased feed costs and livestock stocking densities, whether there is a significant difference between organic and conventional farms depends on the robust farm type. Purchased feed and purchased concentrate costs for dairy farms only show differences of low statistical significance with organic farms having slightly higher costs per livestock unit (possibly due to higher organic feed prices rather than higher usage). For lowland grazing livestock there is a more strongly significant difference with organic farms having lower purchased feed costs. This is also reflected in LFA grazing livestock farms although with a slightly lower significance. In general purchased feed or concentrate costs are not significantly different between conventional and organic mixed farms. Dairy and lowland grazing livestock farms show significant differences in stocking density between organic and conventional management with organic farms tending to have lower stocking densities. The difference for LFA grazing livestock farms is only significant at the 5% level, perhaps reflecting the fact that such farms tend to be unable to support larger stocking densities regardless of management system.

In general it appears from the analysis that organic farms are less intensive than conventional farms, however organic farms appear to have less cropping (and potentially less habitat) variety as reflected by some of the Shannon index results. It would also appear that grazing livestock farms in general may be beneficial to the environment as assessed using this particular set of indicators. It is shown from the analysis presented here that it is possible to use economic data such as the FBS to provide some information on the environmental performance of farms and to compare this across different types if farms and farming systems. In particular it would be of great interest to combine some of the indicators into an overall score that took account of intensity, crop variation, variation in habitat and stocking rates, as well as agri-environment payments. Although an indirect measure of environmental performance may never achieve a perfect assessment a combined score could be weighted to reflect the relative importance of the various factors.

4.1.3.9 Evaluation of public policies

In work package 9 cost functions have been estimated for several EU member countries using the statistical software STATA. The objectives have been as follows: i) to provide and test farm economic models for evaluating ex-post policy measures, ii) to provide and test farm economic models for evaluating ex-ante policy reforms, and iii) to undertake preliminary comparative analyses on these ex-post and ex-ante evaluations across member states and regions. This research component have designed and developed key economic models that use the production cost estimates from work package 3 but also from work packages 5, 6 and 7 to evaluate various agricultural measures on agricultural, environmental, financial and socio-economic indicators using FADN data and additional EUROSTAT data. In particular, this work package have aimed: (a) at providing and testing farm economic models suitable for evaluating ex-post measures that have been implemented as components of: (i) the reform of the Common Agricultural Policy (CAP) for the incumbent Members States since 1992, and (ii) the accession of the new Member States since their adhesion to the EU; (b) at providing and testing farm economic models suitable for evaluating ex-ante measures that would be most likely implemented as components of the continuation of the reform of the CAP; (c) at providing a preliminary comparative analysis on these ex-post and ex-ante evaluations across Member States and regions.

The main outcome of this work package consisted in providing functioning farm-level economic models based on FADN cost estimates that are appropriate for ex-post as well as ex-ante policy analysis. Based on the model developed by UCL, costs functions have been estimated for dairy, cattle and crop farms in Bavaria and Lower Saxony, and ex-post evaluation of the impact of policy changes have been analysed.

Potential Impact:

4.1.4 The potential impact

4.1.4.1 Strategi impact and assistance to policy makers

Analysing the impact assessment of CAP measures is an obligatory, but complicated task for policy makers. Policy makers are therefore often making use of researchers to provide them with information to be able to carry out this task. Researchers on their turn are using agricultural- and econometric models for the impact assessment and are in need of high quality data. These data should originate from a well accepted data provider, preferably the European Commission itself. The Farm Accountancy Data, that is collected by DG-AGRI and stored in the FADN database, contains valuable information on the returns and the costs of a large number of representative farms. These farms for 2004 now cover the most important farm types in all the European Union member states. The agricultural and econometric models mentioned above in general don’t make use of these FADN data directly, but are instead activity based. This means that these models are in need of returns and costs per enterprise (activity). Currently this kind of data is often taken from different sources for which accuracy might be a problem. Because per enterprise (activity) data are not provided directly by the FADN database there is a need to generate this information through an econometric tool in an accurate and well thought way provided by this project.

The calculation of cost of production per enterprise (activity) thus serves as a basis for improving the agricultural- and econometric modelling for measuring the impact assessment of the CAP. However these data not only have their use for modelling, they can also be used themselves for analysing the impact assessment of CAP. The cost of production calculation can for example be used to analyse the relationship between the cost structure and the farm performance and to quantify the relationship between the costs of producing commodities across the EU and the impact on the landscape and natural environment.

The (cost) information that is important for policy makers are at the heart of the results and findings of several project tasks. For instance, some of the deliverables associated with WP5 and WP6 and WP9 (see D5.3 D5.4 D6.3 D9.3 and D9.4) would be highly valuable to policy makers. It is important to stress that that the deliverables D9.1 D9.2 D9.3 and D9.4 (associated with WP9) would show how the cost estimates but other FADN indicators are or can be systematically used in policy impact assessment exercises and how their accuracies could be tested. In a similar vein, all the empirical cost estimates that will be generated under WP3 for crop products, pigs and dairy would likely be valuable to policy makers. Finally, the implementation of this user-friendly computer tool (model) to estimate cost of production would be quite useful to enhance the use of cost estimates in policy units of the Commission.

4.1.4.2 Generation of knowledge

This project and our approach for generating a database with cost of production per activity are not unique. Other research groups all over the world have probably used similar models to determine cost of production data. Besides using models they might also have experience in using other sources. Our aim is to learn from these other groups and to improve our approach with this knowledge. For this reason the project started with an extensive literature review. Different partners in the project team have experience in using econometric modelling for determining costs of production based on FADN data. The strength of this project is that modellers can combine their strength to establish an improved modelling tool for cost of production. This aspect of the project would enhance the empirical knowledge of cost of production in all agricultural sectors of EU member countries (se in particular the tasks of WP3 and the corresponding deliverables D3.1 D3.2 and D3.3). In the same vein, it is worth pointing the role of tasks associated with WP7 (see deliverables D7.2 and D7.3 D7.4) which provides first hand information on the cost of production associated with more environmental-friendly farm systems. Another type of knowledge that has been generated by the project has to do with the model developments performed in this project. This especially concerns the tasks associated with WP3 and WP8.

4.1.4.3 Added value and performing of cost analysis

The project has a functioning public website on [www.ekon.slu.se/facepa]. It includes a popular description of the project, partner information, the public deliverables produced during the project, and more in depth information on the project. It also contains an internal part with a reference library, a discussion forum, and a database with all deliverables that is open only to the various partners and EU officials.

The main user groups of the project’s results are the European Commission, national authorities responsible for the FADN, researchers in agricultural economics, and policy makers and policy advisors in the EU. Project results are also relevant to farmers’ organisations, NGOs working with environmental and rural development issues, international organisations and the international academic community. The results obtained during the lifetime of the project have been shared with the European Commission and have been (where applicable) made publicly available on the project website. Most deliverables are working papers that have been published continuously on the project website after due approval of the Commission. When appropriate, the intent is to publish the working papers in international scientific journals.

4.1.4.4 Estimation of cost of production and potential impact

In the work carried out in the project one potential impact point at and made clear that there are important differences in concepts and definitions used in the different Member State FADNs and the EU FADN. The different studies carried out within work packages also show that the EU FADN is a useful database for estimating the cost of production for various agricultural products at the Member State or even the more regionalized level. The statistical findings, however, also show that for every empirical application using EU FADN, care should be given to the selection of holdings, and weights in relation to the variables of interest.

A general conclusion with focus on the implementation, testing and validation of the GECOM showed that the model can provide plausible estimates of production costs for main products in most countries, reflecting developments over time as well as cost composition, while results for products with smaller output shares were often not convincing and highly variable over time. However, the experiences also showed the indispensable necessity of pre-checking the data in each case, dealing with outliers and taking into account details and changes in the data definition and collection. A general conclusions from the experiences gained is that no “simple” application of one general model is possible for all samples and products. An analysis and validation of results by experts (i.e. of both FADN data and agricultural production systems in the analysed samples) will always be needed.

The production cost software developed in the project, the FACEPA model, has been installed at the FADN services of the EU Commission. The results of the model can be very useful in the process analysing the impact assessment of CAP measures, and allow for the estimation of cost of production and cost analysis from a broad European perspective. In addition to the main user groups of the project’s results, the model can also be a useful tool in the sense of a wider frame of end users.

The structure of the developed PMP model seems to be very interesting due to different reasons. First, it is an alternative to the traditional PMP model and follows an approach that permits to overcome the problem of poor data about farmer behavior. In this respect it is suitable to be used in the framework of FADN dataset that does not collect data about farmer’s risk attitude, experience, etc. Second, the outcome represented by the hidden costs is another important element: this cost component does not appear in the accounting scheme and resumes all the behavioral elements previously mentioned. In the framework of an economic analysis of the farm situation, it is very important to take into account also of the hidden costs. The results show that the PMP model can predict the impact of market price changes or CAP measure on existing crops for which a different technological level is adopted or on new crops that are introduced in a given territory. The possibility to introduce cluster analysis or multivariate techniques makes the PMP model very flexible when it is necessary to analyse a specific territory or farm type or when the sample is not homogeneous. FADN scheme permits to select the sample with respect to these variables and this fact confirm the potential PMP model utility for sectorial and territorial analysis. Moreover, the measurement of the impact of price variation on the farm size and to the level of specialisation allows analysing the positive or negative impact on the environment consequent to a change in the production plan or to crop intensification.

The potential impact linked to policy implications shows that the positive relationship between biodiversity and marginal costs of market goods (such as milk, beef and crop) implies a competitive-economic relationship, i.e. biodiversity and other commodities compete for farm resources. Hence biodiversity does not necessarily have to be reduced by a lower production of other agricultural goods. This relationship is, however, conditional on that no structural brake appears so that land disappears from agricultural production. Since the cost of providing biodiversity increases with the level of supply (of biodiversity), the efficiency of using a flat rate per hectare as a policy tool in order to support biodiversity is questioned. If the support does not cover the additional costs for the most valuable pastures, the results stress that these will not be economically competitive in the long run.

With respect to the agricultural landscape surrounding the farm is very important when it comes to providing biodiversity, zooning may be an efficient policy tool. Plant and animal species will spread among adjacent pastures creating a buffet for changes in the micro-environment of a specific pasture (e.g. some plant species are dependent on annual grazing). From an economic perspective farms located close to each other tend to face the same environmental conditions having impact on the possibilities to efficient farming. The importance of targeting environmental support tools is also underscored by the fact that the costs structure between organic and conventional farmers are very different across EU-Member States.

4.1.4 The potential impact

4.1.4.1 Strategi impact and assistance to policy makers

Analysing the impact assessment of CAP measures is an obligatory, but complicated task for policy makers. Policy makers are therefore often making use of researchers to provide them with information to be able to carry out this task. Researchers on their turn are using agricultural- and econometric models for the impact assessment and are in need of high quality data. These data should originate from a well accepted data provider, preferably the European Commission itself. The Farm Accountancy Data, that is collected by DG-AGRI and stored in the FADN database, contains valuable information on the returns and the costs of a large number of representative farms. These farms for 2004 now cover the most important farm types in all the European Union member states. The agricultural and econometric models mentioned above in general don’t make use of these FADN data directly, but are instead activity based. This means that these models are in need of returns and costs per enterprise (activity). Currently this kind of data is often taken from different sources for which accuracy might be a problem. Because per enterprise (activity) data are not provided directly by the FADN database there is a need to generate this information through an econometric tool in an accurate and well thought way provided by this project.

The calculation of cost of production per enterprise (activity) thus serves as a basis for improving the agricultural- and econometric modelling for measuring the impact assessment of the CAP. However these data not only have their use for modelling, they can also be used themselves for analysing the impact assessment of CAP. The cost of production calculation can for example be used to analyse the relationship between the cost structure and the farm performance and to quantify the relationship between the costs of producing commodities across the EU and the impact on the landscape and natural environment.

The (cost) information that is important for policy makers are at the heart of the results and findings of several project tasks. For instance, some of the deliverables associated with WP5 and WP6 and WP9 (see D5.3 D5.4 D6.3 D9.3 and D9.4) would be highly valuable to policy makers. It is important to stress that that the deliverables D9.1 D9.2 D9.3 and D9.4 (associated with WP9) would show how the cost estimates but other FADN indicators are or can be systematically used in policy impact assessment exercises and how their accuracies could be tested. In a similar vein, all the empirical cost estimates that will be generated under WP3 for crop products, pigs and dairy would likely be valuable to policy makers. Finally, the implementation of this user-friendly computer tool (model) to estimate cost of production would be quite useful to enhance the use of cost estimates in policy units of the Commission.

4.1.4.2 Generation of knowledge

This project and our approach for generating a database with cost of production per activity are not unique. Other research groups all over the world have probably used similar models to determine cost of production data. Besides using models they might also have experience in using other sources. Our aim is to learn from these other groups and to improve our approach with this knowledge. For this reason the project started with an extensive literature review. Different partners in the project team have experience in using econometric modelling for determining costs of production based on FADN data. The strength of this project is that modellers can combine their strength to establish an improved modelling tool for cost of production. This aspect of the project would enhance the empirical knowledge of cost of production in all agricultural sectors of EU member countries (se in particular the tasks of WP3 and the corresponding deliverables D3.1 D3.2 and D3.3). In the same vein, it is worth pointing the role of tasks associated with WP7 (see deliverables D7.2 and D7.3 D7.4) which provides first hand information on the cost of production associated with more environmental-friendly farm systems. Another type of knowledge that has been generated by the project has to do with the model developments performed in this project. This especially concerns the tasks associated with WP3 and WP8.

4.1.4.3 Added value and performing of cost analysis

The project has a functioning public website on [www.ekon.slu.se/facepa]. It includes a popular description of the project, partner information, the public deliverables produced during the project, and more in depth information on the project. It also contains an internal part with a reference library, a discussion forum, and a database with all deliverables that is open only to the various partners and EU officials.

The main user groups of the project’s results are the European Commission, national authorities responsible for the FADN, researchers in agricultural economics, and policy makers and policy advisors in the EU. Project results are also relevant to farmers’ organisations, NGOs working with environmental and rural development issues, international organisations and the international academic community. The results obtained during the lifetime of the project have been shared with the European Commission and have been (where applicable) made publicly available on the project website. Most deliverables are working papers that have been published continuously on the project website after due approval of the Commission. When appropriate, the intent is to publish the working papers in international scientific journals.

4.1.4.4 Estimation of cost of production and potential impact

In the work carried out in the project one potential impact point at and made clear that there are important differences in concepts and definitions used in the different Member State FADNs and the EU FADN. The different studies carried out within work packages also show that the EU FADN is a useful database for estimating the cost of production for various agricultural products at the Member State or even the more regionalized level. The statistical findings, however, also show that for every empirical application using EU FADN, care should be given to the selection of holdings, and weights in relation to the variables of interest.

A general conclusion with focus on the implementation, testing and validation of the GECOM showed that the model can provide plausible estimates of production costs for main products in most countries, reflecting developments over time as well as cost composition, while results for products with smaller output shares were often not convincing and highly variable over time. However, the experiences also showed the indispensable necessity of pre-checking the data in each case, dealing with outliers and taking into account details and changes in the data definition and collection. A general conclusions from the experiences gained is that no “simple” application of one general model is possible for all samples and products. An analysis and validation of results by experts (i.e. of both FADN data and agricultural production systems in the analysed samples) will always be needed.

The production cost software developed in the project, the FACEPA model, has been installed at the FADN services of the EU Commission. The results of the model can be very useful in the process analysing the impact assessment of CAP measures, and allow for the estimation of cost of production and cost analysis from a broad European perspective. In addition to the main user groups of the project’s results, the model can also be a useful tool in the sense of a wider frame of end users.

The structure of the developed PMP model seems to be very interesting due to different reasons. First, it is an alternative to the traditional PMP model and follows an approach that permits to overcome the problem of poor data about farmer behavior. In this respect it is suitable to be used in the framework of FADN dataset that does not collect data about farmer’s risk attitude, experience, etc. Second, the outcome represented by the hidden costs is another important element: this cost component does not appear in the accounting scheme and resumes all the behavioral elements previously mentioned. In the framework of an economic analysis of the farm situation, it is very important to take into account also of the hidden costs. The results show that the PMP model can predict the impact of market price changes or CAP measure on existing crops for which a different technological level is adopted or on new crops that are introduced in a given territory. The possibility to introduce cluster analysis or multivariate techniques makes the PMP model very flexible when it is necessary to analyse a specific territory or farm type or when the sample is not homogeneous. FADN scheme permits to select the sample with respect to these variables and this fact confirm the potential PMP model utility for sectorial and territorial analysis. Moreover, the measurement of the impact of price variation on the farm size and to the level of specialisation allows analysing the positive or negative impact on the environment consequent to a change in the production plan or to crop intensification.

The potential impact linked to policy implications shows that the positive relationship between biodiversity and marginal costs of market goods (such as milk, beef and crop) implies a competitive-economic relationship, i.e. biodiversity and other commodities compete for farm resources. Hence biodiversity does not necessarily have to be reduced by a lower production of other agricultural goods. This relationship is, however, conditional on that no structural brake appears so that land disappears from agricultural production. Since the cost of providing biodiversity increases with the level of supply (of biodiversity), the efficiency of using a flat rate per hectare as a policy tool in order to support biodiversity is questioned. If the support does not cover the additional costs for the most valuable pastures, the results stress that these will not be economically competitive in the long run.

With respect to the agricultural landscape surrounding the farm is very important when it comes to providing biodiversity, zooning may be an efficient policy tool. Plant and animal species will spread among adjacent pastures creating a buffet for changes in the micro-environment of a specific pasture (e.g. some plant species are dependent on annual grazing). From an economic perspective farms located close to each other tend to face the same environmental conditions having impact on the possibilities to efficient farming. The importance of targeting environmental support tools is also underscored by the fact that the costs structure between organic and conventional farmers are very different across EU-Member States.

List of Websites:

http://facepa.slu.se

Professor Yves Surry, yves.surry@slu.se tel +46 18 671795