Final Report Summary - SPARD (Spatial Analysis of Rural Development Measures)
Executive Summary:
Summary 244944 - SPARD (Spatial Analysis of Rural Development Measures)
Stronger accountability requirements and EU budget constraints increase the pressure towards policies targeted on specific objectives, e.g. provision of public benefits. The intention of the call KBBE-2009-1-4-02 was to enhance the capacity for effective targeting of Rural Development Prpgrammes (RDPs) within the EU through spatial analysis. It was explicitly aimed for applying a specific method for data analysis with focus to spatial determinants, which up to now had not yet been used in a systematic manner for the evaluation of RDPs: Spatial econometric modeling. With the introduction of the Common Monitoring and Evaluation Framework (CMEF) the data monitoring situation for the RPP 2007-2014 together with improved availability of additional spatial data was expected to having reached a state feasible to meeting the requirements for applying spatial econometric analysis at different scales. The FP7 collaborative project SPARD (Spatial Analysis of Rural Development Measures) was carried out from 04/2010 – 06/2013 and was organized in 6 work packages under collaboration of 9 Partner institutions from 8 countries. The main objectives and results were as follows:
1. To provide a framework for organizing the collection and the use of regional key baseline data and evaluation results of Rural Development Programmes (RDPs) and other statistical and economic information in a systematic, clear and concise way. SPARD collected, quality checked and harmonized the available CMEF data until 2011, and related relevant EUROSTAT and spatial data at NUTS3 and developed the SPARD Data Viewer for data retrieval and download (access via www.spard.eu or www.spard-is.eu).
2. To explain the causal relationships between regional characteristics and needs, on the one hand, and the RDPs implementation and success in their spatial dimension, on the other. SPARD developed an analytical framework combining RDP structure and intervention logic with requirements for spatial econometric modeling. Reports and papers on targeting strategies and performance at EU and case study scales were developed and a modeling study on cost effectiveness of AEM participation levels was conducted.
3. To develop and apply a spatial econometric modeling approach at different scales. . The vast majority of project activities was dedicated to the development, elaboration and validation of spatial econometric models for the three measures 121 (modernization), 214 (agri-environmental measures) and 311 (diversification). At EU- scale expenditure and impact models were tested, at programming scale in six case study regions (Brandenburg (DE), North Holland (NL), Emilia Romagna (IT), Midi Pyrénées (FR), Scotland (UK) and Eastern Slovenia (SI)) participation, expenditures and impact models. Results indicate spatial dependencies and neighborhood effects, which advice the use of spatial econometrics for evaluation purposes. Results also hint to time lags, and strong differences between regions. One serious obstacle that that sets clear limitations to a broad application of the method in evaluation so far is the insufficient data availability particularly for impact indicators. The preferred scale for spatial econometric analysis for RDP evaluation is NUTS4-5/ LAU2. SPARD research, particularly at case study scale brought various new insights into structural and spatial (geophysical, natural, socio-economic) determinants for participation in RDPs.
4. To build a tool that will help policymakers, both at EU and Member States/ regional level, to design better targeted RDPs. A key physical output of the SPARD project is the information system SPARD-IS (www.spard-is) that comprises research results and the SPARD Data Viewer. The SPARD website www.spard.eu in addition offers access to policy briefs, animated ppts and publications. A Special Issue with the journal of Regional Studies is in preparation.
Project Context and Objectives:
Project context and overall objectives
Stronger accountability requirements and EU budget constraints will increase the pressure towards CAP policies targeted on specific objectives such as the provision of public benefits (environmental, rural, social). The intention of the call KBBE-2009-1-4-02 was to enhance the capacity for effective targeting of Rural Development (RD) policies within the European Union through spatial analysis of Rural Development Measures (RDMs). It was explicitly aimed for applying a specific method for data analysis with focus to spatial determinants, which up to now had not yet been used in a systematic manner for the evaluation of Rural Development Programmes (RDPs): Spatial econometric modelling. Spatial econometrics can contribute to the analysis of RDPs because it facilitates an ex-post evaluation of RDMs, using the potential spatial dependence and spatial heterogeneity in these measures and their impacts. Econometric analyses of RDPs are relatively scarce, mainly due to a lack of reliable data at the regional level (e.g. Ederveen et al. 2002).
With the RDP 2007-2014 the conditions of the EU RDP monitoring had changed insofar, that the data situation was expected to having reached a state feasible to meeting the requirements for applying spatial econometric analysis at different scales. The Member States (MS) first elaborate National Strategic Plans and then prepare RDPs, which must take into account the overarching European goals outlined in the EU Strategic Objectives for RD, as well as ensure a balance between the four axes of RD. In line with this strategic approach, a system of monitoring and evaluation of rural development policy has been developed: the Common Monitoring and Evaluation Framework (CMEF), according to which MS are requested to collect indicators on characteristics, needs, expenditures and results. This system will guide MS towards a more effective system for assessing progress towards European and national objectives, ensuring the accountability of public spending through RDPs, and improving RDP performance. Additionally, managing authorities in some MS have created GIS-based databases with a huge amount of data related to area-based measures. What was lacking was the analysis in which these data sets are combined with other indicators at a high level of regional disaggregation.
The main objectives of the FP7 collaborative project SPARD were
• to provide a framework for organising the collection and the use of regional key baseline data and evaluation results of Rural Development Programmes (RDPs) and other statistical and economic information in a systematic, clear and concise way
• to explain the causal relationships between regional characteristics and needs, on the one hand, and the RDPs implementation and success in their spatial dimension, on the other
• to develop and apply a spatial econometric modelling approach at different scales
• to build a tool that will help policymakers, both at EU and Member States/ regional level, to design better targeted RDPs.
Six work packages contributed to fulfilling these objectives. The following text specifies objectives of the work packages.
WP2 Knowledge - and Data Base
The overall objective of WP2 was to provide an information infrastructure for RDP performance evaluation containing CMEF indicators and additional ones from national RD management authorities. The main tasks refer to
• the development and maintenance of an information infrastructure providing user friendly remote data access for RDP performance evaluation indicators
• data delivery: support, harmonisation, and storage for RDP performance evaluation.
Task 2.1: Design and development of a data warehouse for the project teams und end users from the European Commission
Task 2.1 had already been completed in period 1.
Task 2.2: Design and development of an indicator data base for the project teams und end users from the European Commission
Task 2.2 had already been to a large extent completed in period 1, and had submitted a first version of a living document (D2.2). This living document D2.2 has been updated and submitted in its final form now.
Task 2.3: Data warehouse maintenance and data delivery
and Task 2.4: RDP indicator base maintenance
Task 2.3 and Task 2.4 are to be considered as the continuation of Task 2.1 and Task 2.2. CMEF data until 2011 have been integrated into the data warehouse “SPARD Data Viewer”. The functionalities have been developed and established in line with the requirements and recommendations of the end-users, based on the second end user meeting in Bled in September 2011. On the Final SPARD workshop in Brussels in June 2013 the SPARD-Data Viewer was presented to a larger group of end users from DG Agri and ENRD.
WP3 Analytical framework for causal and spatial relationships
The general objective of WP3 was to develop the conceptual model framework, to what four
specific objectives were related:
• to establish logical links between political objectives, rural development measures, and impacts
• to formulate hypotheses about the causal relationships between expenditures, needs, characteristics and policy results
• to identify target areas and target groups of rural development measures
• to improve the understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
Task 3.1 Development of the conceptual model (by ZALF and UEDIN with partners with key roles in WP2, WP4 and WP5
Task 3.1 had already been completed in period 1.
Two tasks of WP3 have been carried out and completed by due time within the second reporting period.
Task 3.2 Deduction of political target areas and target groups of rural development measures (by ZALF supported by LEI, UNIBO, UEDIN, INRA, UL, AIT)
Objective of this task was to identify possible data sources (monitoring data, geo-referenced data) that enable an empirical analysis of target areas and target groups of rural development measures. It aimed at generalizing the outcomes of the empirical analysis into common procedures and protocols to identify target areas and target groups in different regional contexts, e.g. based on disparities in baseline indicators.
This task has been successfully carried out and completed by due time under lead of the WP3 coordinator ZALF, supported by all SPARD partners who contributed with case study research to the WP5.
The deliverable report D3.2 analyzed and categorized EU targeting approaches for rural development measures. In particular, different types of targeting strategies proposed for rural development measures in the literature, were compiled to an overview of targeting strategies relevant for rural development. For all SPARD case study regions a an assessment of the targeting performance was carried out.
Task 3.3 Improved Understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF (by ZALF)
In this task, shortcomings of the CMEF in considering the spatial component of participation are addressed. Task 3.3. has the objective to demonstrate how the result indicator ‘area under successful land management’ can be calculated through spatially explicit bio-economic farm modeling in order to facilitate ex post (evaluation of observed implementation) and ex ante (scenarios of different implementation levels) analyses of EU rural development measures, whereby the focus of the deliverable report is specifically on agri-environmental measures (measure code 214). This task has been successfully carried out and completed by due time by WP3 leader ZALF. Focus of the D3.3. report is specifically on agri-environmental measures (measure code 214). Taking the district of Ostprignitz-Ruppin, Germany as example (NUTS3 administrative level), the report analyzes how different implementation levels (no implementation, observed implementation, full implementation) of the two agri-environmental measures “grassland extensification” and “organic farming” influence the performance of selected economic and environmental indicators. A bio-economic farm model has been applied.
WP4 Development of a Spatial Econometric Model for RDP Analysis
Three objectives had been set for WP4:
• Building a theoretical model for the functional relations for relevant RDP indicators (and the linkage
between these indicators)
• Estimating these relations using spatial econometrics
• Testing the effect of Rural Development Measures
Three tasks of WP4 have been carried out and completed by due time within the second reporting period. They were mainly related to the following objectives
• to develop a quantitative model for the assessment of the various RD measures and their impacts, using spatial econometric modelling and
• to carry out a number of evaluations/ assessments at EU-27 scale.
In terms of deliverables D4.3 D4.4 and D4.5 have been completed.
Task 4.1. Define the econometric test to assess the impact of RDPs
Task 4.1 had already been to a large extent completed in period 1, and had submitted a first version of a living document (D4.1). This living document D4.1 has been updated and submitted in its final form now.
Task 4.2. Analyse the database for spatial patterns (ESDA)
Task 4.2 had already been completed in period 1.
Task 4.3. Specification and estimation of the model at NUTS0 level - EU wide
Objective of task 4.3 is the specification and estimation of the spatial econometric model (based on Task 4.1 and 4.2) focusing on the variation between the member states. The difference in impact of RD Measures is aimed to be explained at member state level. This task also includes analysis of the development of indicators over time.
During the first reporting period WP4 had developed a theoretical model for the functional relations relevant for RDP indicators, including the linkage between these indicators. Building on this theoretical model we formulated a quantitative model to assess the impact of RD measures using spatial econometrics. The model was used to assess the impact of three different measures at EU-27 level. In WP4 we analysed which relations between Rural Development Indicators (and other data available) are affected by spatial interactions and thus have to be tested using spatial econometrics. The collection of data on impact indicators and baseline indicators was not always straightforward. There was no database available that included all relevant data. The search for and collection of relevant baseline indicators took more time than expected for Task 4.3.
Task 4.4. Specification of the model for the case studies - EU-depth (NUTS2 and NUTS3 level)
In this task the objective was to specify the spatial econometric model in a generic form for WP5 (based on Task 4.1 and 4.2) that focuses on the relation between environmental variables and the Rural Development Indicators at a lower aggregation level (NUTS3). The necessary information provision was assigned to the case studies, where estimation of the model was foreseen to take place in WP5.
In Task 4.4 the econometric work of Task 4.3 was linked to the spatial econometric analyses of the case studies of Task 5.2. Task 5.2 provides a summary. In Task 4.4 we analysed the similarities and differences of the tasks 4.3 and 5.2 in more detail. The specifications modelled in Task 4.3 are in principle also applicable at case study level, but it turned out during SPARD that there are hardly any impact indicators available at low aggregation levels to be able to explore spatial econometric analyses at case study level. A suitable level would be NUTS4-5.
For 4.4 we trained the WP5 researchers in spatial econometric modelling and compared the EU wide analysis with the case study work. For the indicators analysed both in the EU-wide and the case studies study (e.g. agricultural labour productivity and tourism)we analysed the level of the system and the level of the expected spill overs. In Task 4.4 the emphasis was also on the explanation of uptake and participation in RDP measures. This urged for different kind of models (discrete econometric models etc.) Moreover, participation models taught us about participation but not on the actual impacts according to CMEF. The main issue however was the lack of availability of impact indicators at a sufficiently low aggregation level.
Task 4.5. Describe general methodology for the use of spatial econometrics in Rural Development Programmes (by LEI and VUA)
Objective of D4.5 was, to describe, based on experiences from the previous tasks (Task 4.1-Task 4.4) a general methodology how spatial econometrics can be used successfully for a better evaluation of Rural Development Plans.
In Task 4.5 we focussed on the recommendations for use within the EU. We identified the current challenges in RDP evaluation and analysed how the SPARD project could be a solution for these challenges.
Task 4.5 discussed the use of spatial econometrics for the impact assessment of RDP measures. First, we discussed the advantages of using spatial econometrics, and the limitations and pitfalls of using spatial econometrics. Then, the use of econometrics was discussed based on the methodological challenges formulated by Lukesch and Schuh (2010).
WP5 Validation in Case Studies Areas
The objectives of this work package were to prove that
• the methodology is feasible at different scales of application/ levels of aggregation;
• the modelling results are reliable for further specification by using and processing of data of higher or different quality (more disaggregated, higher spatial resolution, specific properties).
The validation was carried out in 6 case study areas at the main programming level (i.e. the level in which Rural development plans are designed). The selected case study areas are Brandenburg (2x NUTS 2, Germany), North Holland (NUTS 2, The Netherlands), Emilia Romagna (NUTS 2, Italy), Midi Pyrennees (NUTS 2, France) Eastern Slovenia (NUTS 2, Slovenia) and Scotland (NUTS2, UK).
The work package was organised in 4 tasks:
• Task 5.1 - Data screening and qualitative identification of causal relationships;
• Task 5.2 - Calibration of model and estimation;
• Task 5.3 – Feasibility of ex-ante analysis;
• Task 5.4 - Results/tool evaluation.
Of these, task 5.1 was carried out during the first reporting period, while the remaining tasks were carried out during the second reporting period.
All objectives of WP5 have been achieved. All activities have been carried out and expected deliverables submitted, though with some delay compared to the initial planning.
Task 5.1 - Data screening and qualitative identification of causal relationships
Task 5.1 had already been accomplished during the first period.
Task 5.2 - Calibration of model and estimation
In this task, the spatial econometric model were estimated in the 6 case study areas. Models were estimated based on disaggregated information (mostly at NUTS5 level where possible). The process of data collection was rather lengthy and this caused serious delays in some case study areas. However this allowed to have a much more complete and verified data set. Though preliminary estimation were already available in early 2012, several updates of the data bases were available at later stages, and new estimates were produced. The final estimations and a draft aggregate report were finalised before the meeting in Den Haag (October 2012) and the final deliverable (D5.2) submitted in March 2013. In order to improve quality and harmonisation of the results, UNIBO developed a common living document collecting methodological choices of each case study and commonly agreed features of the model. UNIBO also organised a WP5 dedicated meeting in Bologna (4-6 July 2012) to discuss preliminary results and improve harmonisation of further model development.
Task 5.3 – Feasibility of ex-ante analysis
This task was devoted to understanding the feasibility of using the tool for ex-ante analysis and to support task 5.4. In particular, in one selected case study (Emilia Romagna, Italy) a mathematical programming model for optimal spatial targeting of Agri-environmental measures was developed, focusing in particular on integration of information generated through spatial econometrics. The results are presented in deliverable D5.3 submitted in March 2013.
Task 5.4 – Results/tool evaluation
In this task, results of tasks 5.2 and 5.3 have been discussed in stakeholder/ end-user/ expert meetings (one in each case study area). This has allowed to gather policy/stakeholders feedback about the tool and an evaluation of the tool and indications for further developments. The results are presented in deliverable D5.4 submitted in march 2013.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
The objective of WP6 was to develop the stand-alone modeling tool SPARD-DSS incl. a Graphical User Interface (GUI) that allows end users to conduct ex-post evaluations and ex-ante assessments to demonstrate CMEF indicators at different spatial scales, causal relationships at horizontal cross-country and vertical in-depth level.
Four specific objectives were set:
• Process design and requirement analysis of the interactive SPARD-DSS using software-prototyping and methods of participative end user involvements
• Developing a conceptual approach of the SPARD-DSS Tool based on requirement analysis on (a) analytical objectives, (b) functionality, (c) graphical design (incl. 'look and feel'), compatibility (e.g. interfaces) that result in a tailored domain structure of the software architecture
• Programming of the SPARD-DSS based on the process-oriented outcome of the requirement analysis
• Compatibility testing to technical setting of data management system (work package 2).
Four tasks of WP6 have been carried out and completed by due time within the second reporting period.
Task 6.1: Software Prototyping and EC Stakeholder (end user) process design
Task 6.1 had already been accomplished during the first period.
Task 6.2 and 6.4 Development of the Conceptual Approach of the SPARD DSS and Programming the SPARD-IS
Objective of this task was to develop the SPARD-IS conceptually. Based on the requirement analysis of task 6.1 SPARD-IS was defined with the following functionalities:
• The SPARD-IS lists policy briefs, explains and visualises the CMEF framework. Data on expenditures are shown in tables, maps on selectable single indicators.
• The SPARD-IS maps were generated in a dynamic setting and they can be geographically localised in a visual map layer.
• The SPARD-IS demonstrates in a short introduction the applied spatial econometric modelling approach using an ESDA example.
• The spatial results are carefully summarized in policy messages, which explain essential key findings of different thematic areas. They are retrievable in downloadable PDF factsheets to be visualised in an overlay as additional information level.
• The factsheets of the case studies are presented in overlays and they will be downloadable to the local (client) computer.
• The case study areas are visualised in an interactive Google Map-application. The description of case study areas are presented as a factsheet in an overlay. The factsheets are also downloadable as PDF.
• The SPARD-IS integrates the SPARD Dataviewer as one major component for data analysis, which can be applied and steered by end user-demands.
Task 6.3: Internal Interface Definition of WP 2 and External Interfaces
Objective of this task was to provide technical linkages with the data management system to ensure a compatibility on jointly used software. Direct use of gathered data of the data management system can be therefore provided through individually defined interfaces and action protocols that allow direct data use, data retrieval and easy up-date functionalities. External interfaces to other Impact Assessment Tools will be considered on potential system compatibility and / or data compatibility. This task has been successfully carried out and completed in due time under lead of the WP6, supported by AIT who contributed with the know-how on data management systems.
The deliverable report D6.3 provides a Documentation of the interfaces to the data management system. The objective of task 6.5 was to test the SPARD-IS at various levels. The test have been conducted at the level of (1) textual summaries of SPARD results, (2) internal technical functionalities of the SPARD-IS and (3) external demonstrations of the SPARD-IS to potential end users for continuous improvements within the project lifetime.
Project Results:
WP1 project management
Three deliverable reports list details on coordination and general administration performed and outputs, communication and dissemination strategy and outputs, and the final project workshop. WP1 project management performed professional scientific and administrative coordination and dissemination, in which the website www.spard.eu , six thematic policy briefs and a special issue publication (>20 papers, journal of Regional Studies, submission Sept. 2013) are key products.
WP2 Knowledge - and Data Base
Four deliverables were the output of WP2. The central deliverable is the SPARD data Viewer.
With the establishment of a well functioning data storage, filtering and thematic retrieval function, an infrastructure was created providing user friendly remote data access for CMEF indicators at NUTS3 (and higher aggregation levels). The “SPARD Data Viewer” is located at http://sf5.arcs.ac.at/spard_site/dataviewer/ and accessible via the SPARD website www.spard.eu and the SPARD-IS at www.spard-is.eu.
WP3 Analytical framework for causal and spatial relationships
Three deliverable reports were produced in WP3, out of which D3.2 is the central deliverable.
Deduction of political target areas and target groups of rural development measures
The first part of the D3.2 report (chapters 1, 2 and 3) analyzes what types of targeting strategies are proposed in the literature, provides an overview of important observations regarding the targeting performance in the EU and presents examples of methods that may facilitate the decision-making on target groups and areas. The second part (chapters 4, 5) provides an assessment of the targeting performance in the SPARD case study regions and performs an attempt to assess expenditure targeting for rural development at EU-level.
The results show that the targeting mechanisms employed by the SPARD case study regions for rural development reach from relatively simple approaches based on eligibility criteria only, to more complex and selective targeting mechanisms based on zoning policies or scoring systems. The choice for a particular approach is discussed against various influencing factors, such as administration cost, distributional motives, or spatial variability in terms of benefits and cost. The expenditure analysis identified some slight correlations between spending for a particular axis and available CMEF indicators, which were mostly in line with expectations as they support typical phenomena reported in the literature (cf. D3.1 report), while some appeared to be accidental or indirect effects.
Overall, correlation is not necessarily an indication for a causal link between axis-spending and indicator performance, and there is always some uncertainty associated to whether indicator performance is the result of an expenditure pattern, or whether the expenditure pattern is made according to the indicator performance. The experience from the targeting assessment in the SPARD case studies and the analysis of the correlation of RD expenditure and the indicators provided by the CMEF in the 2007-2013 period are used to derive recommendations for the further development of the CMEF to better support RD evaluation and analysis in the future.
Improved understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
Taking the district of Ostprignitz-Ruppin, Germany as example (NUTS3 administrative level), the report analyzes how different implementation levels (no implementation, observed implementation, full implementation) of the two agri-environmental measures “grassland extensification” and “organic farming” influence the performance of selected economic and environmental indicators. Background is that the CMEF is aimed at collecting standardized information of the interrelations between successful measure uptake ( analyzed by the selected CMEF result indicator) and indicators assessing or measuring the sustainability impact of the measures (analyzed by prescribed CMEF impact indicators), but still the basic understanding of kind of interrelations and scope of potential improvement is insufficient. Cost-effectiveness and cost- efficiency plays a major role in this context.
The results show that grassland extensification at its observed implementation level attracts primarily participants which fulfill the measure prescriptions already before participation, therefore only low compliance costs occur and the additional benefit for the environment is small. An assumed extensification of the entire regional grassland area (full implementation) leads to more significant environmental results, as then also previously very intensively managed grasslands are affected. However, this scenario would at least double current budget costs for this measure and is therefore not desirable. The second measure, organic farming, is currently applied on less than 10% of the regional area, too little to make a significant impact on the regional scale. A full implementation would lead to significant environmental benefits but would also increase current expenditure at least tenfold. For each scenario, the total area with environmental improvement was reported (= “area under successful land management”). A general limitation in our analysis, however, is that it has a focus on improvement effects while possible maintenance effects, which are frequently stated objectives of agri-environmental measures (e.g. preventing land abandonment), could not be quantified. The CMEF at its current state is not designed to consider maintenance effects, requiring a continuative discussion in the future of what is actually meant by “successful” land management.
WP4 Development of a Spatial Econometric Model for RDP Analysis
Five deliverable reports present the results from WP4. Central deliverable is D4.3.
In Task 4.3 spatial econometric models for the measures 121, 214 and 311 were elaborated and specified.
For labour productivity, the Cambridge econometric database was used. Data were available at NUTS2 level. For nitrogen surplus, only data at NUTS0 level are available for an EU-wide analyses. Fortunately, we were able to collect data for a period of time to ensure a sufficient number of observations to explore spatial econometric techniques. The High Nature Value index was not available at NUTS2 level but was constructed based on available indicators according to the definition in the literature. Data were collected from different sources (Eurostat and Cambridge Econometrics, Farm Structure Survey). The Tourism indicators were available at Eurostat. Land use variables, for instance, were derived from Corine databases. Additional variables on attractiveness were collected separately. Finally, the RDP expenditures turned out not to be readily available at NUTS2 level. In this sense, we took advantage of the work in WP3 of SPARD.
Thereafter the models for the impact indicators were specified based on the relevant theoretical models and estimated. For each analysis, we set up different specification due to differences across data availability, the level of data available, the time period of available data, missing data etc. The basic approach was to first estimate a ‘static’ model, and then a ‘dynamic’ model in which time-lagged dependent variables were included. For both types of temporal analysis, the necessity of applying spatial econometric techniques was explored.
For nitrogen surplus, the number of countries to be included was limited, so that we constructed a panel data set in order to be able to explore spatial econometric techniques. Spatial panel data techniques are more advanced econometric techniques. The tourism analyses was applied to four indicators of tourism which could be derived from the Tourism indicators available at Eurostat: inbound tourism in hotels, inbound tourism in rural accommodations such as holiday houses and camping sites, domestic tourism in hotels and domestic tourism in rural accommodations. In addition, we also explored spatial heterogeneity in our specifications. Especially in the labour productivity analyses, different regimes in Europe were extensively explored. In the tourism analyses, we included country specific dummy variables and other spatial variables. Also, for agricultural labour productivity, nitrogen surplus and HNV farmland, the impact of cross axis expenditures were tested. There was no evidence for cross axis impacts.
Four spatial econometric impact models have been developed based upon the CMEF assessment framework of the RDP (D4.3).
The first model is on labour productivity, relating the spendings on measure 121 (modernization of agricultural holdings) to changes in labour productivity in agriculture over de RDP period 2000-2010. We reviewed three labour productivity models (Solow/Swan; Mankiw/Romer/weil; Islam) and developed a model for our analysis in which the labour productivity in 2010 is explained by the labour productivity in the starting year (2000), GDP (Gross Domestic Product) per capita, investments, RDP spendings, the population density, the farm size, the type of crops. To capture the spatial dependence in our data (both in labour productivity data as in the RDP expenditure data) we added spatial variables into our model: e.g. motorway density and added spatial regimes.
The impact of measure 214 (agri environmental measures) on water quality is estimated relating the expenditures to a reduction of nitrogen surplus per hectare. The estimation is performed at NUTS0 level, because EU wide data on nitrogen surplus over the period 2000-2009 are only available at member state level. In our model the change in nitrogen surplus is related to GDP, the investments, the RDP expenditures on: measure 214, other measures within axis 2 and measures within axis 1. A spatial panel data estimation is done for the period 2001-2008.
The impact of measure 214 (agri environmental measures) on biodiversity is estimated relating the expenditures to the change of high nature value farmland (HNV). Information on HNV farmland is not available for the entire EU over the programme period, we constructed a HNV-index. This index is based on the stocking density and crop diversity at NUTS2 level. HNV in 2010 is explained by HNV in 2000, GDP/ha, investments, the RDP expenditures on: measure 214, other measures within axis 2 and measures within axis 1.
In the tourism model we relate the expenditure on measure 311 (diversification into non-agricultural activities) and 313 (encouragement of tourism activities) to the change in nights spent in the region. Our model is based on the Nissan et al., 2011 model. Tourism indicators were: inbound tourism in hotels, inbound tourism in rural accommodations such as holiday houses and camping sites, domestic tourism in hotels and domestic tourism in rural accommodations. The Night spent in 2010 are in our model a function of the capacity (bed places) in the starting year, capacity growth, population, GDP/capita, unemployment and spatial information. The model is estimated at NUTS2 level.
To summarize findings from WP4 in general: the indicators related to agriculture show relatively strong spatial clustering effects. Looking at the share of agriculture in total value added, we found a moderate amount of spatial autocorrelation at the conventional NUTS2 level (a Moran’s I of 0.48) but all spatial clusters disappeared when we looked at NUTS1 and NUTS0 regions. Unfortunately, analyses at lower levels are not feasible at a pan-European scale, but within SPARD these will be undertaken in the case studies.
From all this, there are the following lessons we can learn:
1. It is important to carefully select the way an indicator is taken into account. GVA in the primary sector measured in absolute terms shows different levels of clustering than when measured in relative terms (0.17 against 0.48). This also holds for the number and share of self-employed persons (0.14 against 0.62).
2. Likewise, the spatial scale matters enormously. Looking at the share of agriculture in total value added, we found a moderate amount of spatial autocorrelation at the conventional NUTS2 level (a Moran’s I of 0.48).
3. In general, the indicators related to agriculture show relatively strong spatial clustering effects. In particular when looking at the economic and physical size of farms, high values for Moran’s I are found. Outside agriculture, strong spatial clustering effects are found for life-long learning and the level of higher educated adults (0.73 and 0.77 respectively).
4. The queen and rook contiguity matrices are not preferrable weight indexes for EU wide estimations, because they cannot handle islands well. The Gabriel weight matrix is preferred, because islands can be linked to the nearest observations.
5. At NUTS2 level the correlation of spending and the dependent variables is apparently negligible or weak. The spending on RDP measures is low, compared to other investments.
6. Our spatial econometric analysis rely on the presence of good quality data. It was not possible to include all relevant characteristics in our analysis due to data availability.
7. Spatial dependence is relevant for all impact indicators analysed at EU level, as well as for the RDP expenditures of all axis.
8. In addition, for agricultural labour productivity, nitrogen surplus and HNV farmland, the impact of cross axis expenditures were tested. There was no evidence for cross axis impacts.
9. Spatial econometrics is very useful in RDP-evaluation because it provides insights in the counterfactuals (e.g. if no RDP expenditures were available), and it provides disentangles the insights in the effects of several determinants simultaneously.
WP5 Validation in Case Studies Areas
The results of WP5 are reported in deliverables D5.1 D5.2 D5.3 D5.4. Central deliverable is D5.2 comprised of a cross cutting report and with six extended reports from case study modelling as annexes.
The study carried out in task 5.2 provided estimation of econometric models for several RDP measures (121, 214, 311) in regions of 6 different countries, approximating results with participation, payment and impact indicators. All case study results are described in separate deliverable report annexes of D5.2 and an animated presentation is available on www.spard.eu and http://www.youtube.com/watch?v=5ic8N-Z3Omg. The results highlighted some relevance of spatial issues and some potential of spatial econometrics in contributing to explain participation to RDPs. It also showed several limitations of application, due mainly to lack of data availability, many of which however not specific of spatial analysis, but rather common with any exercise aimed at explaining in detail the drivers of RDP effects.
The analysis also allowed to better identify (several) data and evaluation gaps, which could be the basis for further better oriented research and policy support activity. Some of these issues, particularly those related to RDP-tailored model specification, matching with priority perception by decision-makers and use of models’ results for ex-ante analysis, will be further developed already in within the remaining activities of the SPARD project.
Task 5.3 highlighted the importance of spatial differentiation to explain the determinants of farmers’ participation to AEMs schemes and the relevance of considering this differentiation in optimisation tools searching for optimal incentive-compatible targeting. It also showed the weaknesses of this approach, and the need for further improvement; the main limitation (and connected line of improvement), concerns the difficulty in accounting for policy design variables in spatial econometric and hence in feeding them into optimisation models.
The reactions collected with the stakeholders/local end-users in task 5.4 was rather positive. The outcome mainly points at the idea that more refined techniques for analysis can in fact support a better understanding and design of RDP; spatial econometrics can contribute in this direction. However the issue has to be cast in a consistent and improved process of data collection and management, monitoring and evaluation of RDPs.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
WP6 delivered five deliverable reports, the central output is the SPARD-IS.
The SPARD-IS start page guides the user through the entire menu. By clicking on the entrance button the user will gain a complete overview to all categories which are integrated in the SPARD-IS. Some exemplary screenshots are presented below in order demonstrate the structure and technical implementation of the SPARD-IS (http://sf5.ait.ac.at/spard_site/results/spardisstart.html). In order to illustrate the technical interfaces of the SPARD-IS the design will be briefly summarized in this section. Based on this section, which explains the software environment and the functionality of the SPARD-IS, the relations of the system components can be described via the necessary interface protocols.
Software-Environment:
• The SPARD-IS is a Client-Server-Model as the major concept, which divides programmes into the client-component and the server-component. In the case of SPARD-IS the server hosts data in a database and provides services to the client.
• The conceptual approach of the SPARD-IS is based on the CMS Drupal 6. The SPARD Dataviewer which will be part of results, findings and data presentation uses the Java Webstart technology and the Client-Server-Model.
• SPARD-IS will be developed using web technologies CMS Drupal, PHP, HTML, CSS and Javascript to avoid the disadvantage of likely run time errors. The SPARD-IS will run in browser applications only and does not require any additional runtime environment.
• The SPARD Dataviewer is a client-software which is executed on a local machine. The data viewer retrieves data requests on a remote database server. The end user selects required sets of information within the SPARD Dataviewer and the client sends a data request to the database. The database collects the requested data and sends them back to SPARD Dataviewer. The viewer uses the received data for its listing in the graphical user interface (GUI).
• The SPARD-IS, the SPARD Dataviewer and the database server will be hosted at AIT. The maintenance of the server and the database including database management system will be provided by AIT. An apache server software will be used in order to host and execute SPARD-IS. PostgreSQL (http://www.postgresql.org/) will be used as database management system. Apache (http://httpd.apache.org/) as web server, and PostgreSQL were chosen because both are open-source software. They are free of charge, transparent and well documented.
Given in the settings of the above described system environment, altogether six interfaces are defined as major bottlenecks for the SPARD-IS. Hereby a high performance in terms of system response time and system robustness can be guaranteed.
As direct response to the demonstration of the SPARD-IS at the final meeting as well as during bilateral discussions the following essential findings have been recorded:
- At a glance the SPARD tool looks good in terms what was possible within project time given the current circumstances that end users were difficult to identify due to low in-house demand at EU Commission level for a tailored SPARD Decision Support System.
- Technically the SPARD tool looks easy to handle and access. Few compatibility problems may happen from time to time, which should be resolved.
- SPARD-IS gives a good overview, sometimes the menu could have been slightly improved in terms of comprehensiveness and bar positions.
- It was wise that the SPARD did not build a tool as stand-alone version, but an internet based tool.
- Data quality is most important part. The data quality is in general not sufficient at RDP evaluation level to work representatively at region-explicit level.
- More region-explicit indicators, and especially output indicators, are needed.
- The emphasis should be given more on output, not on input indicators (this should be emphasized, because contradictor effects e.g. on emission GHG could be levelled out). The impacts of measures should be accumulated and the overall effect should be taken into account.
- Cluster of representative farms would be better than the actual system is established (the system has to be changed with the same budget efforts). Problem-oriented regions should be build (e.g. according to high risk of erosion, leaching etc.) in order to act also problem-oriented in terms of funding of RDP.
- Region explicitness is essential
- EU-Commission mechanisms are not ideal in terms of data quality (huge data gaps) and the collaboration of institutions within the regions should be improved.
- The institutional collaboration is the problem, often one-dimensional thinking from certain institutions and not multi-dimensional thinking with collaboration among institutions.
Potential Impact:
WP2 Knowledge - and Data Base
Resulting from the termination of the project before the finalization of the RDP 2007-2013 period, the data base is incomplete for future evaluation purposes that could benefit from the harmonized EU-27 data base at NUTS3, and its functionalities. On the Final SPARD workshop, the consortium pointed to this fact in the discussions with end users. Also the fact that plans for the further development of the CMEF towards 2014-2020 and related scaling issues were not yet definite at the end of the project lifetime, set limits to the exploitation potential, which are independent from the successful completion of all WP2 tasks according to DG Agri demands.
WP3 Analytical framework for causal and spatial relationships
Deduction of political target areas and target groups of rural development measures
Particularly with regard to rural development measures in the EU, the D3.2 report has identified various fields that need clarification or improvement if rural development measures are to be monitored and evaluated appropriately in the future.
As general findings regarding improved targeting we want to point to the following:
Clarification of objectives
Targeting strategies depend on clear objectives. Like it cannot be evaluated whether an RD measure achieves the results it was designed for if objectives are not sufficiently clear (ECA 2011), effective targeting cannot be implemented either. Target areas and target groups need to adequately consider all objectives of an RD measure, if necessary according to their priority. If a horizontal measure equally serves multiple objectives, it is counterproductive to base the target area on one objective alone. Recommendations for improvement are:
• the number of (eligible) objectives per measure has to be reduced
• the optimal solution would be to design separate instruments for the individual objectives
• To assist decision-making, however, it may help to rank objectives according to their priority
• Prime candidates for targeting are those RD measures with few and relatively precisely formulated objectives
Competitive funding mechanism
Particularly the lack of inducing competition among participants is likely to reduce the overall quality of applications and is unlikely to optimize value for money. A competitive process is more likely to produce a cost-effective result compared to a situation when budgets are sufficient.
• If all applications are funded, except they miss out for formal reasons, there is essentially no competition between participants and the likelihood that less suitable (with regard to objectives) projects/parcels are funded increases
• A competitive process may be associated with more administration effort, however, some member states have started establishing a competitive process which just lacks proper enforcement.
• The decision whether or not to establish a competive mechanisms depends on the variability of benefits and costs and their correlation. The higher the geo-phyiscal heterogeneity, the more e.g. a zoning approach makes sense. If there is much variability in benefits and/or costs, efforts towards more competition via application of additional selection criteria should definitely be explored. If less variability is given, eligibility criteria alone may be sufficient.
Equity versus cost-effectiveness
The low targeting performance in three of the SPARD case study regions (chapter 4 of this report) may also be an indication for hidden distributional motives. Ensuring equity among farmers or administrative units or completeness out of fairness reasons, typical distributional motives, will therefore often speak against rigorous targeting strategies.
• There is a clear conflict between cohesion objectives and the need for cost-effectiveness. E.g. investment needs for smaller farms often do not coincide with investment sustainability (= cost-effectiveness), which is often the decisive factor for allocating funding
• This conflict with also contributes to explaining why horizontal measures are dominating over specifically targeted measures in the case of measure 214.
Additional observations of use for the further development of the CMEF made by the SPARD project include
Data issues
• The conditions for collecting environmental indicators are different across member states, e.g. bird monitoring data not equally available in the EU member states, often depends on private initiatives, otherwise too costly
• Data gaps are not always due to methodological problems, instead certain indicators such as labor productivity in agriculture are not reported by some member states by referring to the agreement, which states that data are delivered voluntarily
• The update cycles of the indicators are different and the indicators published in one year often refer to different reporting years, which is indicated but makes the comparability often difficult (unless indicators do not change much over the years)
Synergies between measures should be indicated
• If single RDP measures are implemented in conjunction with other measures, synergies and thus high benefits can occur (ECA 2006). However, if a combined application of measures occurs, this needs to be indicated by the member states and should be made explicit in the CMEF as otherwise it cannot considered in evaluation studies.
Separation of maintenance and improvement objectives
• If the CMEF is to be used appropriately it also needs to be clarified (by the implementing authorities) whether different objectives exist for the same measure depending on the sub-region/zone where a measure is applied. It also needs to be made clear whether the objectives aim at improvement (reduction in pollution, increase in income) or maintenance (stabilize farm numbers or incomes), as this requires different interpretation and processing strategies for the CMEF indicators.
• If the objective is improvement, a positive change in relevant results and impact indicators values is interpreted as a positive effect. If the objective is maintenance, no change in relevant indicators is already a positive effect.
Assignment of indicators to administrative regions
• SPARD CSA analyses have shown that we need spatially disaggregated indicators if we want to analyze and understand spatial impacts. Data analyses and recommendations usually have to be provided for administrative or planning regions. Results and impacts, however, may not be limited to the boundaries of such areas. For example, a farmstead may be located in one district, but the farm has most of its farm area in the neighboring district. Should a change in a farm-level indicator then be attributed to the district where the farmstead is located or to the district where the farm has most of its land? If the real area shares are known, relative percentages can be calculated. However, usually such spatially explicit information is not available. Instead aggregated figures e.g. on the total number of farms implementing a RD measure or the total number of supported hectares are reported, without knowing where the farmstead and thus the decision-maker is located.
Continuous vs. accumulated measures
• RDP measures may be continuous or permanent in nature or they may represent a single event. Examples for relatively continuous measures are AEM. An investment, in contrast, is a single time event. If expenditures for investment aids in a region decline in successive years, this may not necessarily be an indication for less interest of the farmers or a wrongly targeted policy, but may also reflect that the potential for investment aids is currently met.
• Investment effects (on such indicators as gross value added, gross fixed capital formation, labor productivity; all agriculture-related) will not always become immediately visible but sometimes several years after the actual investment. Such effects could not be considered in the analysis in SPARD, as we could not obtain a coherent dataset over more than one programming period. Instead, expenditure in the current period (years 2007-2010) was related to CMEF baseline indicator performance, while possible effects of expenditure in earlier period could not be considered, which may lead to an overestimation of effects. Ideally, accumulated figures should be considered here.
Improved Understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
As for potential impacts for the further development of the CMEF and its application, as well as for policy recommendations for the upcoming RD period 2014-2020, D3.3. concludes as follows:
After the first practical test of the CMEF in the programming period 2007-2013, it is likely that several adjustments will be brought forward by the EU member states, resulting from their experience with the application of this framework. However, the general tasks for future versions of the CMEF will remain the same, namely to provide EU-wide evidence for the effectiveness of rural development measures in order to demonstrate the European taxpayers that adequate value for money is provided through these measures. The analysis in this report demonstrated how bio-economic modelling applied in a spatially explicit approach can assist in providing the result indicators needed for the CMEF (“’Area under successful land management”). It supports both ex post (observed implementation of a measure) and ex ante analysis (different implementation scenarios, e.g. no implementation, full implementation etc.). The analysis was done for only one NUTS3 region; therefore it is very far away from providing results at the EU scale. However, as agri-environmental measures deal with regionally specific environmental objectives, high spatial resolution and regionally relevant management prescriptions it is impossible to simulate meaningful results with only one model for the whole EU. The overall goal should therefore not be to develop such a tool but to assist the authorities responsible for agri-environmental measures in calculating the indicators to be delivered to the EU in order to ensure that the indicator values reflect the specific regional circumstances in an adequate way.
However, the analysis also shows that even if this route is taken, it still remains a challenge to provide conclusions for both improvement and maintenance effects of agri-environmental measures. Improvement is usually easier to model once the general model assumptions are clear. Maintenance of a state -or prevention of deterioration of a state, however, is much more difficult to model, as the assumptions for this type of effect are much more complex. As agri-environmental measures often have maintenance objectives, conclusions regarding their effectiveness based on improvement effects alone, may therefore be misleading. Any analysis should therefore formulate its results and recommendations with caution and under consideration of these points.
At EU-level, if the analysis at the regional level in the EU member states was performed with such a high level of accuracy and caution, and expressed in relative and not absolute terms, e.g. % of potential achieved instead of number of hectares enrolled in a measure, the CMEF would finally get closer to its objective of supporting EU-wide monitoring and evaluation (Uthes 2013).
WP4 Development of a Spatial Econometric Model for RDP Analysis
Outcomes of the WP4 research hold the potential to facilitate decision making on the application of the method for evaluation purposes. With the contribution to the SPARD policy briefs, WP4 points concisely to potential impacts and application related lessons learnt. Such are e.g. the advantages and limitations of using spatial econometrics in RDP evaluation, relevant findings, indicating the existence of spatially determined conditions of RDP measures implementation and impacts, and data related requirements that have to be met by the CMEF if spatial econometrics is aimed to be applied.
WP5 Validation in Case Studies Areas
The main potential impact and societal implications of the work performed goes in the direction of improvements in RDP policy evaluation and design. The main implications of this are in the field of better efficiency intended as value for money from public expenditure and higher effectiveness in tackling economic, social and environmental objectives of rural development. The pathway through which project results can contribute to this are well established thanks to the high stakeholders involvement and exchange about project results, involving in particular EC and regional staff in charge of RDP design and evaluation.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
As for potential impacts with regard to the dissemination of SPARD results via the SPARD-DSS it will be important to consider the following development and perceptions:
In all the SPARD-IS was well perceived at the final workshop in Brussels as a tool to present the comprehensive results of the SPARD project. The SPARD-IS was originally planned to be established at the EU Commission level, but the major problem was a lacking stable demand-driven end user group, which would have been able to continuously accompany the development of the SPARD-IS. Based on the conducted requirement analysis it was evident that DG AGRI has in-house services and that external tools for decision making need more time of being implemented at Commission level. This decision on the development of SPARD-IS instead of a tailored SPARD Decision Support System for selected end users was taken commonly within the SPARD consortium. The most-adequate and best fitting approach was elaborated to illustrate the results of SPARD (1) in a comprehensive way, (2) to maximise the outreach of SPARD results to the wider public and last but not least (3) to raise awareness of potential end user at the level of the EU Commission. With regard to the technical development of the SPARD-IS this task was important base to implement the system as web-based tool. The protocols supported the technical linking-up between different software components and assured an efficient functionality.
In all the SPARD-IS was well perceived at the final workshop in Brussels as a tool to present the comprehensive results of the SPARD project. The most-adequate and best fitting approach was elaborated to illustrate the results of SPARD (1) in a comprehensive way, (2) to maximise the outreach of SPARD results to the wider public and last but not least (3) to raise awareness of potential end user at the level of the EU Commission. The wide dissemination of the SPARD results can raise awareness of the project and feedback from different institutions, who work in the similar research area, can link-up with the SPARD consortium in order to exchange knowledge at as open research platform.
List of Websites:
The project website: www.spard.eu
The Information System SPARD-IS: www.spard-is.eu
Project co-ordinator: Dr. Annette Piorr, Leibniz-Centre for Agricultural Landscape Research (ZALF e.V.) Eberswalder Str. 84, 15374 Muencheberg, Germany. Email: apiorr@zalf.de
Project Officer: Dr. Hans-Jörg Lutzeyer, DG RTD
Summary 244944 - SPARD (Spatial Analysis of Rural Development Measures)
Stronger accountability requirements and EU budget constraints increase the pressure towards policies targeted on specific objectives, e.g. provision of public benefits. The intention of the call KBBE-2009-1-4-02 was to enhance the capacity for effective targeting of Rural Development Prpgrammes (RDPs) within the EU through spatial analysis. It was explicitly aimed for applying a specific method for data analysis with focus to spatial determinants, which up to now had not yet been used in a systematic manner for the evaluation of RDPs: Spatial econometric modeling. With the introduction of the Common Monitoring and Evaluation Framework (CMEF) the data monitoring situation for the RPP 2007-2014 together with improved availability of additional spatial data was expected to having reached a state feasible to meeting the requirements for applying spatial econometric analysis at different scales. The FP7 collaborative project SPARD (Spatial Analysis of Rural Development Measures) was carried out from 04/2010 – 06/2013 and was organized in 6 work packages under collaboration of 9 Partner institutions from 8 countries. The main objectives and results were as follows:
1. To provide a framework for organizing the collection and the use of regional key baseline data and evaluation results of Rural Development Programmes (RDPs) and other statistical and economic information in a systematic, clear and concise way. SPARD collected, quality checked and harmonized the available CMEF data until 2011, and related relevant EUROSTAT and spatial data at NUTS3 and developed the SPARD Data Viewer for data retrieval and download (access via www.spard.eu or www.spard-is.eu).
2. To explain the causal relationships between regional characteristics and needs, on the one hand, and the RDPs implementation and success in their spatial dimension, on the other. SPARD developed an analytical framework combining RDP structure and intervention logic with requirements for spatial econometric modeling. Reports and papers on targeting strategies and performance at EU and case study scales were developed and a modeling study on cost effectiveness of AEM participation levels was conducted.
3. To develop and apply a spatial econometric modeling approach at different scales. . The vast majority of project activities was dedicated to the development, elaboration and validation of spatial econometric models for the three measures 121 (modernization), 214 (agri-environmental measures) and 311 (diversification). At EU- scale expenditure and impact models were tested, at programming scale in six case study regions (Brandenburg (DE), North Holland (NL), Emilia Romagna (IT), Midi Pyrénées (FR), Scotland (UK) and Eastern Slovenia (SI)) participation, expenditures and impact models. Results indicate spatial dependencies and neighborhood effects, which advice the use of spatial econometrics for evaluation purposes. Results also hint to time lags, and strong differences between regions. One serious obstacle that that sets clear limitations to a broad application of the method in evaluation so far is the insufficient data availability particularly for impact indicators. The preferred scale for spatial econometric analysis for RDP evaluation is NUTS4-5/ LAU2. SPARD research, particularly at case study scale brought various new insights into structural and spatial (geophysical, natural, socio-economic) determinants for participation in RDPs.
4. To build a tool that will help policymakers, both at EU and Member States/ regional level, to design better targeted RDPs. A key physical output of the SPARD project is the information system SPARD-IS (www.spard-is) that comprises research results and the SPARD Data Viewer. The SPARD website www.spard.eu in addition offers access to policy briefs, animated ppts and publications. A Special Issue with the journal of Regional Studies is in preparation.
Project Context and Objectives:
Project context and overall objectives
Stronger accountability requirements and EU budget constraints will increase the pressure towards CAP policies targeted on specific objectives such as the provision of public benefits (environmental, rural, social). The intention of the call KBBE-2009-1-4-02 was to enhance the capacity for effective targeting of Rural Development (RD) policies within the European Union through spatial analysis of Rural Development Measures (RDMs). It was explicitly aimed for applying a specific method for data analysis with focus to spatial determinants, which up to now had not yet been used in a systematic manner for the evaluation of Rural Development Programmes (RDPs): Spatial econometric modelling. Spatial econometrics can contribute to the analysis of RDPs because it facilitates an ex-post evaluation of RDMs, using the potential spatial dependence and spatial heterogeneity in these measures and their impacts. Econometric analyses of RDPs are relatively scarce, mainly due to a lack of reliable data at the regional level (e.g. Ederveen et al. 2002).
With the RDP 2007-2014 the conditions of the EU RDP monitoring had changed insofar, that the data situation was expected to having reached a state feasible to meeting the requirements for applying spatial econometric analysis at different scales. The Member States (MS) first elaborate National Strategic Plans and then prepare RDPs, which must take into account the overarching European goals outlined in the EU Strategic Objectives for RD, as well as ensure a balance between the four axes of RD. In line with this strategic approach, a system of monitoring and evaluation of rural development policy has been developed: the Common Monitoring and Evaluation Framework (CMEF), according to which MS are requested to collect indicators on characteristics, needs, expenditures and results. This system will guide MS towards a more effective system for assessing progress towards European and national objectives, ensuring the accountability of public spending through RDPs, and improving RDP performance. Additionally, managing authorities in some MS have created GIS-based databases with a huge amount of data related to area-based measures. What was lacking was the analysis in which these data sets are combined with other indicators at a high level of regional disaggregation.
The main objectives of the FP7 collaborative project SPARD were
• to provide a framework for organising the collection and the use of regional key baseline data and evaluation results of Rural Development Programmes (RDPs) and other statistical and economic information in a systematic, clear and concise way
• to explain the causal relationships between regional characteristics and needs, on the one hand, and the RDPs implementation and success in their spatial dimension, on the other
• to develop and apply a spatial econometric modelling approach at different scales
• to build a tool that will help policymakers, both at EU and Member States/ regional level, to design better targeted RDPs.
Six work packages contributed to fulfilling these objectives. The following text specifies objectives of the work packages.
WP2 Knowledge - and Data Base
The overall objective of WP2 was to provide an information infrastructure for RDP performance evaluation containing CMEF indicators and additional ones from national RD management authorities. The main tasks refer to
• the development and maintenance of an information infrastructure providing user friendly remote data access for RDP performance evaluation indicators
• data delivery: support, harmonisation, and storage for RDP performance evaluation.
Task 2.1: Design and development of a data warehouse for the project teams und end users from the European Commission
Task 2.1 had already been completed in period 1.
Task 2.2: Design and development of an indicator data base for the project teams und end users from the European Commission
Task 2.2 had already been to a large extent completed in period 1, and had submitted a first version of a living document (D2.2). This living document D2.2 has been updated and submitted in its final form now.
Task 2.3: Data warehouse maintenance and data delivery
and Task 2.4: RDP indicator base maintenance
Task 2.3 and Task 2.4 are to be considered as the continuation of Task 2.1 and Task 2.2. CMEF data until 2011 have been integrated into the data warehouse “SPARD Data Viewer”. The functionalities have been developed and established in line with the requirements and recommendations of the end-users, based on the second end user meeting in Bled in September 2011. On the Final SPARD workshop in Brussels in June 2013 the SPARD-Data Viewer was presented to a larger group of end users from DG Agri and ENRD.
WP3 Analytical framework for causal and spatial relationships
The general objective of WP3 was to develop the conceptual model framework, to what four
specific objectives were related:
• to establish logical links between political objectives, rural development measures, and impacts
• to formulate hypotheses about the causal relationships between expenditures, needs, characteristics and policy results
• to identify target areas and target groups of rural development measures
• to improve the understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
Task 3.1 Development of the conceptual model (by ZALF and UEDIN with partners with key roles in WP2, WP4 and WP5
Task 3.1 had already been completed in period 1.
Two tasks of WP3 have been carried out and completed by due time within the second reporting period.
Task 3.2 Deduction of political target areas and target groups of rural development measures (by ZALF supported by LEI, UNIBO, UEDIN, INRA, UL, AIT)
Objective of this task was to identify possible data sources (monitoring data, geo-referenced data) that enable an empirical analysis of target areas and target groups of rural development measures. It aimed at generalizing the outcomes of the empirical analysis into common procedures and protocols to identify target areas and target groups in different regional contexts, e.g. based on disparities in baseline indicators.
This task has been successfully carried out and completed by due time under lead of the WP3 coordinator ZALF, supported by all SPARD partners who contributed with case study research to the WP5.
The deliverable report D3.2 analyzed and categorized EU targeting approaches for rural development measures. In particular, different types of targeting strategies proposed for rural development measures in the literature, were compiled to an overview of targeting strategies relevant for rural development. For all SPARD case study regions a an assessment of the targeting performance was carried out.
Task 3.3 Improved Understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF (by ZALF)
In this task, shortcomings of the CMEF in considering the spatial component of participation are addressed. Task 3.3. has the objective to demonstrate how the result indicator ‘area under successful land management’ can be calculated through spatially explicit bio-economic farm modeling in order to facilitate ex post (evaluation of observed implementation) and ex ante (scenarios of different implementation levels) analyses of EU rural development measures, whereby the focus of the deliverable report is specifically on agri-environmental measures (measure code 214). This task has been successfully carried out and completed by due time by WP3 leader ZALF. Focus of the D3.3. report is specifically on agri-environmental measures (measure code 214). Taking the district of Ostprignitz-Ruppin, Germany as example (NUTS3 administrative level), the report analyzes how different implementation levels (no implementation, observed implementation, full implementation) of the two agri-environmental measures “grassland extensification” and “organic farming” influence the performance of selected economic and environmental indicators. A bio-economic farm model has been applied.
WP4 Development of a Spatial Econometric Model for RDP Analysis
Three objectives had been set for WP4:
• Building a theoretical model for the functional relations for relevant RDP indicators (and the linkage
between these indicators)
• Estimating these relations using spatial econometrics
• Testing the effect of Rural Development Measures
Three tasks of WP4 have been carried out and completed by due time within the second reporting period. They were mainly related to the following objectives
• to develop a quantitative model for the assessment of the various RD measures and their impacts, using spatial econometric modelling and
• to carry out a number of evaluations/ assessments at EU-27 scale.
In terms of deliverables D4.3 D4.4 and D4.5 have been completed.
Task 4.1. Define the econometric test to assess the impact of RDPs
Task 4.1 had already been to a large extent completed in period 1, and had submitted a first version of a living document (D4.1). This living document D4.1 has been updated and submitted in its final form now.
Task 4.2. Analyse the database for spatial patterns (ESDA)
Task 4.2 had already been completed in period 1.
Task 4.3. Specification and estimation of the model at NUTS0 level - EU wide
Objective of task 4.3 is the specification and estimation of the spatial econometric model (based on Task 4.1 and 4.2) focusing on the variation between the member states. The difference in impact of RD Measures is aimed to be explained at member state level. This task also includes analysis of the development of indicators over time.
During the first reporting period WP4 had developed a theoretical model for the functional relations relevant for RDP indicators, including the linkage between these indicators. Building on this theoretical model we formulated a quantitative model to assess the impact of RD measures using spatial econometrics. The model was used to assess the impact of three different measures at EU-27 level. In WP4 we analysed which relations between Rural Development Indicators (and other data available) are affected by spatial interactions and thus have to be tested using spatial econometrics. The collection of data on impact indicators and baseline indicators was not always straightforward. There was no database available that included all relevant data. The search for and collection of relevant baseline indicators took more time than expected for Task 4.3.
Task 4.4. Specification of the model for the case studies - EU-depth (NUTS2 and NUTS3 level)
In this task the objective was to specify the spatial econometric model in a generic form for WP5 (based on Task 4.1 and 4.2) that focuses on the relation between environmental variables and the Rural Development Indicators at a lower aggregation level (NUTS3). The necessary information provision was assigned to the case studies, where estimation of the model was foreseen to take place in WP5.
In Task 4.4 the econometric work of Task 4.3 was linked to the spatial econometric analyses of the case studies of Task 5.2. Task 5.2 provides a summary. In Task 4.4 we analysed the similarities and differences of the tasks 4.3 and 5.2 in more detail. The specifications modelled in Task 4.3 are in principle also applicable at case study level, but it turned out during SPARD that there are hardly any impact indicators available at low aggregation levels to be able to explore spatial econometric analyses at case study level. A suitable level would be NUTS4-5.
For 4.4 we trained the WP5 researchers in spatial econometric modelling and compared the EU wide analysis with the case study work. For the indicators analysed both in the EU-wide and the case studies study (e.g. agricultural labour productivity and tourism)we analysed the level of the system and the level of the expected spill overs. In Task 4.4 the emphasis was also on the explanation of uptake and participation in RDP measures. This urged for different kind of models (discrete econometric models etc.) Moreover, participation models taught us about participation but not on the actual impacts according to CMEF. The main issue however was the lack of availability of impact indicators at a sufficiently low aggregation level.
Task 4.5. Describe general methodology for the use of spatial econometrics in Rural Development Programmes (by LEI and VUA)
Objective of D4.5 was, to describe, based on experiences from the previous tasks (Task 4.1-Task 4.4) a general methodology how spatial econometrics can be used successfully for a better evaluation of Rural Development Plans.
In Task 4.5 we focussed on the recommendations for use within the EU. We identified the current challenges in RDP evaluation and analysed how the SPARD project could be a solution for these challenges.
Task 4.5 discussed the use of spatial econometrics for the impact assessment of RDP measures. First, we discussed the advantages of using spatial econometrics, and the limitations and pitfalls of using spatial econometrics. Then, the use of econometrics was discussed based on the methodological challenges formulated by Lukesch and Schuh (2010).
WP5 Validation in Case Studies Areas
The objectives of this work package were to prove that
• the methodology is feasible at different scales of application/ levels of aggregation;
• the modelling results are reliable for further specification by using and processing of data of higher or different quality (more disaggregated, higher spatial resolution, specific properties).
The validation was carried out in 6 case study areas at the main programming level (i.e. the level in which Rural development plans are designed). The selected case study areas are Brandenburg (2x NUTS 2, Germany), North Holland (NUTS 2, The Netherlands), Emilia Romagna (NUTS 2, Italy), Midi Pyrennees (NUTS 2, France) Eastern Slovenia (NUTS 2, Slovenia) and Scotland (NUTS2, UK).
The work package was organised in 4 tasks:
• Task 5.1 - Data screening and qualitative identification of causal relationships;
• Task 5.2 - Calibration of model and estimation;
• Task 5.3 – Feasibility of ex-ante analysis;
• Task 5.4 - Results/tool evaluation.
Of these, task 5.1 was carried out during the first reporting period, while the remaining tasks were carried out during the second reporting period.
All objectives of WP5 have been achieved. All activities have been carried out and expected deliverables submitted, though with some delay compared to the initial planning.
Task 5.1 - Data screening and qualitative identification of causal relationships
Task 5.1 had already been accomplished during the first period.
Task 5.2 - Calibration of model and estimation
In this task, the spatial econometric model were estimated in the 6 case study areas. Models were estimated based on disaggregated information (mostly at NUTS5 level where possible). The process of data collection was rather lengthy and this caused serious delays in some case study areas. However this allowed to have a much more complete and verified data set. Though preliminary estimation were already available in early 2012, several updates of the data bases were available at later stages, and new estimates were produced. The final estimations and a draft aggregate report were finalised before the meeting in Den Haag (October 2012) and the final deliverable (D5.2) submitted in March 2013. In order to improve quality and harmonisation of the results, UNIBO developed a common living document collecting methodological choices of each case study and commonly agreed features of the model. UNIBO also organised a WP5 dedicated meeting in Bologna (4-6 July 2012) to discuss preliminary results and improve harmonisation of further model development.
Task 5.3 – Feasibility of ex-ante analysis
This task was devoted to understanding the feasibility of using the tool for ex-ante analysis and to support task 5.4. In particular, in one selected case study (Emilia Romagna, Italy) a mathematical programming model for optimal spatial targeting of Agri-environmental measures was developed, focusing in particular on integration of information generated through spatial econometrics. The results are presented in deliverable D5.3 submitted in March 2013.
Task 5.4 – Results/tool evaluation
In this task, results of tasks 5.2 and 5.3 have been discussed in stakeholder/ end-user/ expert meetings (one in each case study area). This has allowed to gather policy/stakeholders feedback about the tool and an evaluation of the tool and indications for further developments. The results are presented in deliverable D5.4 submitted in march 2013.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
The objective of WP6 was to develop the stand-alone modeling tool SPARD-DSS incl. a Graphical User Interface (GUI) that allows end users to conduct ex-post evaluations and ex-ante assessments to demonstrate CMEF indicators at different spatial scales, causal relationships at horizontal cross-country and vertical in-depth level.
Four specific objectives were set:
• Process design and requirement analysis of the interactive SPARD-DSS using software-prototyping and methods of participative end user involvements
• Developing a conceptual approach of the SPARD-DSS Tool based on requirement analysis on (a) analytical objectives, (b) functionality, (c) graphical design (incl. 'look and feel'), compatibility (e.g. interfaces) that result in a tailored domain structure of the software architecture
• Programming of the SPARD-DSS based on the process-oriented outcome of the requirement analysis
• Compatibility testing to technical setting of data management system (work package 2).
Four tasks of WP6 have been carried out and completed by due time within the second reporting period.
Task 6.1: Software Prototyping and EC Stakeholder (end user) process design
Task 6.1 had already been accomplished during the first period.
Task 6.2 and 6.4 Development of the Conceptual Approach of the SPARD DSS and Programming the SPARD-IS
Objective of this task was to develop the SPARD-IS conceptually. Based on the requirement analysis of task 6.1 SPARD-IS was defined with the following functionalities:
• The SPARD-IS lists policy briefs, explains and visualises the CMEF framework. Data on expenditures are shown in tables, maps on selectable single indicators.
• The SPARD-IS maps were generated in a dynamic setting and they can be geographically localised in a visual map layer.
• The SPARD-IS demonstrates in a short introduction the applied spatial econometric modelling approach using an ESDA example.
• The spatial results are carefully summarized in policy messages, which explain essential key findings of different thematic areas. They are retrievable in downloadable PDF factsheets to be visualised in an overlay as additional information level.
• The factsheets of the case studies are presented in overlays and they will be downloadable to the local (client) computer.
• The case study areas are visualised in an interactive Google Map-application. The description of case study areas are presented as a factsheet in an overlay. The factsheets are also downloadable as PDF.
• The SPARD-IS integrates the SPARD Dataviewer as one major component for data analysis, which can be applied and steered by end user-demands.
Task 6.3: Internal Interface Definition of WP 2 and External Interfaces
Objective of this task was to provide technical linkages with the data management system to ensure a compatibility on jointly used software. Direct use of gathered data of the data management system can be therefore provided through individually defined interfaces and action protocols that allow direct data use, data retrieval and easy up-date functionalities. External interfaces to other Impact Assessment Tools will be considered on potential system compatibility and / or data compatibility. This task has been successfully carried out and completed in due time under lead of the WP6, supported by AIT who contributed with the know-how on data management systems.
The deliverable report D6.3 provides a Documentation of the interfaces to the data management system. The objective of task 6.5 was to test the SPARD-IS at various levels. The test have been conducted at the level of (1) textual summaries of SPARD results, (2) internal technical functionalities of the SPARD-IS and (3) external demonstrations of the SPARD-IS to potential end users for continuous improvements within the project lifetime.
Project Results:
WP1 project management
Three deliverable reports list details on coordination and general administration performed and outputs, communication and dissemination strategy and outputs, and the final project workshop. WP1 project management performed professional scientific and administrative coordination and dissemination, in which the website www.spard.eu , six thematic policy briefs and a special issue publication (>20 papers, journal of Regional Studies, submission Sept. 2013) are key products.
WP2 Knowledge - and Data Base
Four deliverables were the output of WP2. The central deliverable is the SPARD data Viewer.
With the establishment of a well functioning data storage, filtering and thematic retrieval function, an infrastructure was created providing user friendly remote data access for CMEF indicators at NUTS3 (and higher aggregation levels). The “SPARD Data Viewer” is located at http://sf5.arcs.ac.at/spard_site/dataviewer/ and accessible via the SPARD website www.spard.eu and the SPARD-IS at www.spard-is.eu.
WP3 Analytical framework for causal and spatial relationships
Three deliverable reports were produced in WP3, out of which D3.2 is the central deliverable.
Deduction of political target areas and target groups of rural development measures
The first part of the D3.2 report (chapters 1, 2 and 3) analyzes what types of targeting strategies are proposed in the literature, provides an overview of important observations regarding the targeting performance in the EU and presents examples of methods that may facilitate the decision-making on target groups and areas. The second part (chapters 4, 5) provides an assessment of the targeting performance in the SPARD case study regions and performs an attempt to assess expenditure targeting for rural development at EU-level.
The results show that the targeting mechanisms employed by the SPARD case study regions for rural development reach from relatively simple approaches based on eligibility criteria only, to more complex and selective targeting mechanisms based on zoning policies or scoring systems. The choice for a particular approach is discussed against various influencing factors, such as administration cost, distributional motives, or spatial variability in terms of benefits and cost. The expenditure analysis identified some slight correlations between spending for a particular axis and available CMEF indicators, which were mostly in line with expectations as they support typical phenomena reported in the literature (cf. D3.1 report), while some appeared to be accidental or indirect effects.
Overall, correlation is not necessarily an indication for a causal link between axis-spending and indicator performance, and there is always some uncertainty associated to whether indicator performance is the result of an expenditure pattern, or whether the expenditure pattern is made according to the indicator performance. The experience from the targeting assessment in the SPARD case studies and the analysis of the correlation of RD expenditure and the indicators provided by the CMEF in the 2007-2013 period are used to derive recommendations for the further development of the CMEF to better support RD evaluation and analysis in the future.
Improved understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
Taking the district of Ostprignitz-Ruppin, Germany as example (NUTS3 administrative level), the report analyzes how different implementation levels (no implementation, observed implementation, full implementation) of the two agri-environmental measures “grassland extensification” and “organic farming” influence the performance of selected economic and environmental indicators. Background is that the CMEF is aimed at collecting standardized information of the interrelations between successful measure uptake ( analyzed by the selected CMEF result indicator) and indicators assessing or measuring the sustainability impact of the measures (analyzed by prescribed CMEF impact indicators), but still the basic understanding of kind of interrelations and scope of potential improvement is insufficient. Cost-effectiveness and cost- efficiency plays a major role in this context.
The results show that grassland extensification at its observed implementation level attracts primarily participants which fulfill the measure prescriptions already before participation, therefore only low compliance costs occur and the additional benefit for the environment is small. An assumed extensification of the entire regional grassland area (full implementation) leads to more significant environmental results, as then also previously very intensively managed grasslands are affected. However, this scenario would at least double current budget costs for this measure and is therefore not desirable. The second measure, organic farming, is currently applied on less than 10% of the regional area, too little to make a significant impact on the regional scale. A full implementation would lead to significant environmental benefits but would also increase current expenditure at least tenfold. For each scenario, the total area with environmental improvement was reported (= “area under successful land management”). A general limitation in our analysis, however, is that it has a focus on improvement effects while possible maintenance effects, which are frequently stated objectives of agri-environmental measures (e.g. preventing land abandonment), could not be quantified. The CMEF at its current state is not designed to consider maintenance effects, requiring a continuative discussion in the future of what is actually meant by “successful” land management.
WP4 Development of a Spatial Econometric Model for RDP Analysis
Five deliverable reports present the results from WP4. Central deliverable is D4.3.
In Task 4.3 spatial econometric models for the measures 121, 214 and 311 were elaborated and specified.
For labour productivity, the Cambridge econometric database was used. Data were available at NUTS2 level. For nitrogen surplus, only data at NUTS0 level are available for an EU-wide analyses. Fortunately, we were able to collect data for a period of time to ensure a sufficient number of observations to explore spatial econometric techniques. The High Nature Value index was not available at NUTS2 level but was constructed based on available indicators according to the definition in the literature. Data were collected from different sources (Eurostat and Cambridge Econometrics, Farm Structure Survey). The Tourism indicators were available at Eurostat. Land use variables, for instance, were derived from Corine databases. Additional variables on attractiveness were collected separately. Finally, the RDP expenditures turned out not to be readily available at NUTS2 level. In this sense, we took advantage of the work in WP3 of SPARD.
Thereafter the models for the impact indicators were specified based on the relevant theoretical models and estimated. For each analysis, we set up different specification due to differences across data availability, the level of data available, the time period of available data, missing data etc. The basic approach was to first estimate a ‘static’ model, and then a ‘dynamic’ model in which time-lagged dependent variables were included. For both types of temporal analysis, the necessity of applying spatial econometric techniques was explored.
For nitrogen surplus, the number of countries to be included was limited, so that we constructed a panel data set in order to be able to explore spatial econometric techniques. Spatial panel data techniques are more advanced econometric techniques. The tourism analyses was applied to four indicators of tourism which could be derived from the Tourism indicators available at Eurostat: inbound tourism in hotels, inbound tourism in rural accommodations such as holiday houses and camping sites, domestic tourism in hotels and domestic tourism in rural accommodations. In addition, we also explored spatial heterogeneity in our specifications. Especially in the labour productivity analyses, different regimes in Europe were extensively explored. In the tourism analyses, we included country specific dummy variables and other spatial variables. Also, for agricultural labour productivity, nitrogen surplus and HNV farmland, the impact of cross axis expenditures were tested. There was no evidence for cross axis impacts.
Four spatial econometric impact models have been developed based upon the CMEF assessment framework of the RDP (D4.3).
The first model is on labour productivity, relating the spendings on measure 121 (modernization of agricultural holdings) to changes in labour productivity in agriculture over de RDP period 2000-2010. We reviewed three labour productivity models (Solow/Swan; Mankiw/Romer/weil; Islam) and developed a model for our analysis in which the labour productivity in 2010 is explained by the labour productivity in the starting year (2000), GDP (Gross Domestic Product) per capita, investments, RDP spendings, the population density, the farm size, the type of crops. To capture the spatial dependence in our data (both in labour productivity data as in the RDP expenditure data) we added spatial variables into our model: e.g. motorway density and added spatial regimes.
The impact of measure 214 (agri environmental measures) on water quality is estimated relating the expenditures to a reduction of nitrogen surplus per hectare. The estimation is performed at NUTS0 level, because EU wide data on nitrogen surplus over the period 2000-2009 are only available at member state level. In our model the change in nitrogen surplus is related to GDP, the investments, the RDP expenditures on: measure 214, other measures within axis 2 and measures within axis 1. A spatial panel data estimation is done for the period 2001-2008.
The impact of measure 214 (agri environmental measures) on biodiversity is estimated relating the expenditures to the change of high nature value farmland (HNV). Information on HNV farmland is not available for the entire EU over the programme period, we constructed a HNV-index. This index is based on the stocking density and crop diversity at NUTS2 level. HNV in 2010 is explained by HNV in 2000, GDP/ha, investments, the RDP expenditures on: measure 214, other measures within axis 2 and measures within axis 1.
In the tourism model we relate the expenditure on measure 311 (diversification into non-agricultural activities) and 313 (encouragement of tourism activities) to the change in nights spent in the region. Our model is based on the Nissan et al., 2011 model. Tourism indicators were: inbound tourism in hotels, inbound tourism in rural accommodations such as holiday houses and camping sites, domestic tourism in hotels and domestic tourism in rural accommodations. The Night spent in 2010 are in our model a function of the capacity (bed places) in the starting year, capacity growth, population, GDP/capita, unemployment and spatial information. The model is estimated at NUTS2 level.
To summarize findings from WP4 in general: the indicators related to agriculture show relatively strong spatial clustering effects. Looking at the share of agriculture in total value added, we found a moderate amount of spatial autocorrelation at the conventional NUTS2 level (a Moran’s I of 0.48) but all spatial clusters disappeared when we looked at NUTS1 and NUTS0 regions. Unfortunately, analyses at lower levels are not feasible at a pan-European scale, but within SPARD these will be undertaken in the case studies.
From all this, there are the following lessons we can learn:
1. It is important to carefully select the way an indicator is taken into account. GVA in the primary sector measured in absolute terms shows different levels of clustering than when measured in relative terms (0.17 against 0.48). This also holds for the number and share of self-employed persons (0.14 against 0.62).
2. Likewise, the spatial scale matters enormously. Looking at the share of agriculture in total value added, we found a moderate amount of spatial autocorrelation at the conventional NUTS2 level (a Moran’s I of 0.48).
3. In general, the indicators related to agriculture show relatively strong spatial clustering effects. In particular when looking at the economic and physical size of farms, high values for Moran’s I are found. Outside agriculture, strong spatial clustering effects are found for life-long learning and the level of higher educated adults (0.73 and 0.77 respectively).
4. The queen and rook contiguity matrices are not preferrable weight indexes for EU wide estimations, because they cannot handle islands well. The Gabriel weight matrix is preferred, because islands can be linked to the nearest observations.
5. At NUTS2 level the correlation of spending and the dependent variables is apparently negligible or weak. The spending on RDP measures is low, compared to other investments.
6. Our spatial econometric analysis rely on the presence of good quality data. It was not possible to include all relevant characteristics in our analysis due to data availability.
7. Spatial dependence is relevant for all impact indicators analysed at EU level, as well as for the RDP expenditures of all axis.
8. In addition, for agricultural labour productivity, nitrogen surplus and HNV farmland, the impact of cross axis expenditures were tested. There was no evidence for cross axis impacts.
9. Spatial econometrics is very useful in RDP-evaluation because it provides insights in the counterfactuals (e.g. if no RDP expenditures were available), and it provides disentangles the insights in the effects of several determinants simultaneously.
WP5 Validation in Case Studies Areas
The results of WP5 are reported in deliverables D5.1 D5.2 D5.3 D5.4. Central deliverable is D5.2 comprised of a cross cutting report and with six extended reports from case study modelling as annexes.
The study carried out in task 5.2 provided estimation of econometric models for several RDP measures (121, 214, 311) in regions of 6 different countries, approximating results with participation, payment and impact indicators. All case study results are described in separate deliverable report annexes of D5.2 and an animated presentation is available on www.spard.eu and http://www.youtube.com/watch?v=5ic8N-Z3Omg. The results highlighted some relevance of spatial issues and some potential of spatial econometrics in contributing to explain participation to RDPs. It also showed several limitations of application, due mainly to lack of data availability, many of which however not specific of spatial analysis, but rather common with any exercise aimed at explaining in detail the drivers of RDP effects.
The analysis also allowed to better identify (several) data and evaluation gaps, which could be the basis for further better oriented research and policy support activity. Some of these issues, particularly those related to RDP-tailored model specification, matching with priority perception by decision-makers and use of models’ results for ex-ante analysis, will be further developed already in within the remaining activities of the SPARD project.
Task 5.3 highlighted the importance of spatial differentiation to explain the determinants of farmers’ participation to AEMs schemes and the relevance of considering this differentiation in optimisation tools searching for optimal incentive-compatible targeting. It also showed the weaknesses of this approach, and the need for further improvement; the main limitation (and connected line of improvement), concerns the difficulty in accounting for policy design variables in spatial econometric and hence in feeding them into optimisation models.
The reactions collected with the stakeholders/local end-users in task 5.4 was rather positive. The outcome mainly points at the idea that more refined techniques for analysis can in fact support a better understanding and design of RDP; spatial econometrics can contribute in this direction. However the issue has to be cast in a consistent and improved process of data collection and management, monitoring and evaluation of RDPs.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
WP6 delivered five deliverable reports, the central output is the SPARD-IS.
The SPARD-IS start page guides the user through the entire menu. By clicking on the entrance button the user will gain a complete overview to all categories which are integrated in the SPARD-IS. Some exemplary screenshots are presented below in order demonstrate the structure and technical implementation of the SPARD-IS (http://sf5.ait.ac.at/spard_site/results/spardisstart.html). In order to illustrate the technical interfaces of the SPARD-IS the design will be briefly summarized in this section. Based on this section, which explains the software environment and the functionality of the SPARD-IS, the relations of the system components can be described via the necessary interface protocols.
Software-Environment:
• The SPARD-IS is a Client-Server-Model as the major concept, which divides programmes into the client-component and the server-component. In the case of SPARD-IS the server hosts data in a database and provides services to the client.
• The conceptual approach of the SPARD-IS is based on the CMS Drupal 6. The SPARD Dataviewer which will be part of results, findings and data presentation uses the Java Webstart technology and the Client-Server-Model.
• SPARD-IS will be developed using web technologies CMS Drupal, PHP, HTML, CSS and Javascript to avoid the disadvantage of likely run time errors. The SPARD-IS will run in browser applications only and does not require any additional runtime environment.
• The SPARD Dataviewer is a client-software which is executed on a local machine. The data viewer retrieves data requests on a remote database server. The end user selects required sets of information within the SPARD Dataviewer and the client sends a data request to the database. The database collects the requested data and sends them back to SPARD Dataviewer. The viewer uses the received data for its listing in the graphical user interface (GUI).
• The SPARD-IS, the SPARD Dataviewer and the database server will be hosted at AIT. The maintenance of the server and the database including database management system will be provided by AIT. An apache server software will be used in order to host and execute SPARD-IS. PostgreSQL (http://www.postgresql.org/) will be used as database management system. Apache (http://httpd.apache.org/) as web server, and PostgreSQL were chosen because both are open-source software. They are free of charge, transparent and well documented.
Given in the settings of the above described system environment, altogether six interfaces are defined as major bottlenecks for the SPARD-IS. Hereby a high performance in terms of system response time and system robustness can be guaranteed.
As direct response to the demonstration of the SPARD-IS at the final meeting as well as during bilateral discussions the following essential findings have been recorded:
- At a glance the SPARD tool looks good in terms what was possible within project time given the current circumstances that end users were difficult to identify due to low in-house demand at EU Commission level for a tailored SPARD Decision Support System.
- Technically the SPARD tool looks easy to handle and access. Few compatibility problems may happen from time to time, which should be resolved.
- SPARD-IS gives a good overview, sometimes the menu could have been slightly improved in terms of comprehensiveness and bar positions.
- It was wise that the SPARD did not build a tool as stand-alone version, but an internet based tool.
- Data quality is most important part. The data quality is in general not sufficient at RDP evaluation level to work representatively at region-explicit level.
- More region-explicit indicators, and especially output indicators, are needed.
- The emphasis should be given more on output, not on input indicators (this should be emphasized, because contradictor effects e.g. on emission GHG could be levelled out). The impacts of measures should be accumulated and the overall effect should be taken into account.
- Cluster of representative farms would be better than the actual system is established (the system has to be changed with the same budget efforts). Problem-oriented regions should be build (e.g. according to high risk of erosion, leaching etc.) in order to act also problem-oriented in terms of funding of RDP.
- Region explicitness is essential
- EU-Commission mechanisms are not ideal in terms of data quality (huge data gaps) and the collaboration of institutions within the regions should be improved.
- The institutional collaboration is the problem, often one-dimensional thinking from certain institutions and not multi-dimensional thinking with collaboration among institutions.
Potential Impact:
WP2 Knowledge - and Data Base
Resulting from the termination of the project before the finalization of the RDP 2007-2013 period, the data base is incomplete for future evaluation purposes that could benefit from the harmonized EU-27 data base at NUTS3, and its functionalities. On the Final SPARD workshop, the consortium pointed to this fact in the discussions with end users. Also the fact that plans for the further development of the CMEF towards 2014-2020 and related scaling issues were not yet definite at the end of the project lifetime, set limits to the exploitation potential, which are independent from the successful completion of all WP2 tasks according to DG Agri demands.
WP3 Analytical framework for causal and spatial relationships
Deduction of political target areas and target groups of rural development measures
Particularly with regard to rural development measures in the EU, the D3.2 report has identified various fields that need clarification or improvement if rural development measures are to be monitored and evaluated appropriately in the future.
As general findings regarding improved targeting we want to point to the following:
Clarification of objectives
Targeting strategies depend on clear objectives. Like it cannot be evaluated whether an RD measure achieves the results it was designed for if objectives are not sufficiently clear (ECA 2011), effective targeting cannot be implemented either. Target areas and target groups need to adequately consider all objectives of an RD measure, if necessary according to their priority. If a horizontal measure equally serves multiple objectives, it is counterproductive to base the target area on one objective alone. Recommendations for improvement are:
• the number of (eligible) objectives per measure has to be reduced
• the optimal solution would be to design separate instruments for the individual objectives
• To assist decision-making, however, it may help to rank objectives according to their priority
• Prime candidates for targeting are those RD measures with few and relatively precisely formulated objectives
Competitive funding mechanism
Particularly the lack of inducing competition among participants is likely to reduce the overall quality of applications and is unlikely to optimize value for money. A competitive process is more likely to produce a cost-effective result compared to a situation when budgets are sufficient.
• If all applications are funded, except they miss out for formal reasons, there is essentially no competition between participants and the likelihood that less suitable (with regard to objectives) projects/parcels are funded increases
• A competitive process may be associated with more administration effort, however, some member states have started establishing a competitive process which just lacks proper enforcement.
• The decision whether or not to establish a competive mechanisms depends on the variability of benefits and costs and their correlation. The higher the geo-phyiscal heterogeneity, the more e.g. a zoning approach makes sense. If there is much variability in benefits and/or costs, efforts towards more competition via application of additional selection criteria should definitely be explored. If less variability is given, eligibility criteria alone may be sufficient.
Equity versus cost-effectiveness
The low targeting performance in three of the SPARD case study regions (chapter 4 of this report) may also be an indication for hidden distributional motives. Ensuring equity among farmers or administrative units or completeness out of fairness reasons, typical distributional motives, will therefore often speak against rigorous targeting strategies.
• There is a clear conflict between cohesion objectives and the need for cost-effectiveness. E.g. investment needs for smaller farms often do not coincide with investment sustainability (= cost-effectiveness), which is often the decisive factor for allocating funding
• This conflict with also contributes to explaining why horizontal measures are dominating over specifically targeted measures in the case of measure 214.
Additional observations of use for the further development of the CMEF made by the SPARD project include
Data issues
• The conditions for collecting environmental indicators are different across member states, e.g. bird monitoring data not equally available in the EU member states, often depends on private initiatives, otherwise too costly
• Data gaps are not always due to methodological problems, instead certain indicators such as labor productivity in agriculture are not reported by some member states by referring to the agreement, which states that data are delivered voluntarily
• The update cycles of the indicators are different and the indicators published in one year often refer to different reporting years, which is indicated but makes the comparability often difficult (unless indicators do not change much over the years)
Synergies between measures should be indicated
• If single RDP measures are implemented in conjunction with other measures, synergies and thus high benefits can occur (ECA 2006). However, if a combined application of measures occurs, this needs to be indicated by the member states and should be made explicit in the CMEF as otherwise it cannot considered in evaluation studies.
Separation of maintenance and improvement objectives
• If the CMEF is to be used appropriately it also needs to be clarified (by the implementing authorities) whether different objectives exist for the same measure depending on the sub-region/zone where a measure is applied. It also needs to be made clear whether the objectives aim at improvement (reduction in pollution, increase in income) or maintenance (stabilize farm numbers or incomes), as this requires different interpretation and processing strategies for the CMEF indicators.
• If the objective is improvement, a positive change in relevant results and impact indicators values is interpreted as a positive effect. If the objective is maintenance, no change in relevant indicators is already a positive effect.
Assignment of indicators to administrative regions
• SPARD CSA analyses have shown that we need spatially disaggregated indicators if we want to analyze and understand spatial impacts. Data analyses and recommendations usually have to be provided for administrative or planning regions. Results and impacts, however, may not be limited to the boundaries of such areas. For example, a farmstead may be located in one district, but the farm has most of its farm area in the neighboring district. Should a change in a farm-level indicator then be attributed to the district where the farmstead is located or to the district where the farm has most of its land? If the real area shares are known, relative percentages can be calculated. However, usually such spatially explicit information is not available. Instead aggregated figures e.g. on the total number of farms implementing a RD measure or the total number of supported hectares are reported, without knowing where the farmstead and thus the decision-maker is located.
Continuous vs. accumulated measures
• RDP measures may be continuous or permanent in nature or they may represent a single event. Examples for relatively continuous measures are AEM. An investment, in contrast, is a single time event. If expenditures for investment aids in a region decline in successive years, this may not necessarily be an indication for less interest of the farmers or a wrongly targeted policy, but may also reflect that the potential for investment aids is currently met.
• Investment effects (on such indicators as gross value added, gross fixed capital formation, labor productivity; all agriculture-related) will not always become immediately visible but sometimes several years after the actual investment. Such effects could not be considered in the analysis in SPARD, as we could not obtain a coherent dataset over more than one programming period. Instead, expenditure in the current period (years 2007-2010) was related to CMEF baseline indicator performance, while possible effects of expenditure in earlier period could not be considered, which may lead to an overestimation of effects. Ideally, accumulated figures should be considered here.
Improved Understanding of the relationships between EU RD expenditures and resulting effects and impacts applying the CMEF
As for potential impacts for the further development of the CMEF and its application, as well as for policy recommendations for the upcoming RD period 2014-2020, D3.3. concludes as follows:
After the first practical test of the CMEF in the programming period 2007-2013, it is likely that several adjustments will be brought forward by the EU member states, resulting from their experience with the application of this framework. However, the general tasks for future versions of the CMEF will remain the same, namely to provide EU-wide evidence for the effectiveness of rural development measures in order to demonstrate the European taxpayers that adequate value for money is provided through these measures. The analysis in this report demonstrated how bio-economic modelling applied in a spatially explicit approach can assist in providing the result indicators needed for the CMEF (“’Area under successful land management”). It supports both ex post (observed implementation of a measure) and ex ante analysis (different implementation scenarios, e.g. no implementation, full implementation etc.). The analysis was done for only one NUTS3 region; therefore it is very far away from providing results at the EU scale. However, as agri-environmental measures deal with regionally specific environmental objectives, high spatial resolution and regionally relevant management prescriptions it is impossible to simulate meaningful results with only one model for the whole EU. The overall goal should therefore not be to develop such a tool but to assist the authorities responsible for agri-environmental measures in calculating the indicators to be delivered to the EU in order to ensure that the indicator values reflect the specific regional circumstances in an adequate way.
However, the analysis also shows that even if this route is taken, it still remains a challenge to provide conclusions for both improvement and maintenance effects of agri-environmental measures. Improvement is usually easier to model once the general model assumptions are clear. Maintenance of a state -or prevention of deterioration of a state, however, is much more difficult to model, as the assumptions for this type of effect are much more complex. As agri-environmental measures often have maintenance objectives, conclusions regarding their effectiveness based on improvement effects alone, may therefore be misleading. Any analysis should therefore formulate its results and recommendations with caution and under consideration of these points.
At EU-level, if the analysis at the regional level in the EU member states was performed with such a high level of accuracy and caution, and expressed in relative and not absolute terms, e.g. % of potential achieved instead of number of hectares enrolled in a measure, the CMEF would finally get closer to its objective of supporting EU-wide monitoring and evaluation (Uthes 2013).
WP4 Development of a Spatial Econometric Model for RDP Analysis
Outcomes of the WP4 research hold the potential to facilitate decision making on the application of the method for evaluation purposes. With the contribution to the SPARD policy briefs, WP4 points concisely to potential impacts and application related lessons learnt. Such are e.g. the advantages and limitations of using spatial econometrics in RDP evaluation, relevant findings, indicating the existence of spatially determined conditions of RDP measures implementation and impacts, and data related requirements that have to be met by the CMEF if spatial econometrics is aimed to be applied.
WP5 Validation in Case Studies Areas
The main potential impact and societal implications of the work performed goes in the direction of improvements in RDP policy evaluation and design. The main implications of this are in the field of better efficiency intended as value for money from public expenditure and higher effectiveness in tackling economic, social and environmental objectives of rural development. The pathway through which project results can contribute to this are well established thanks to the high stakeholders involvement and exchange about project results, involving in particular EC and regional staff in charge of RDP design and evaluation.
WP6 End-User Involvement and SPARD-Information System (SPARD-IS)
As for potential impacts with regard to the dissemination of SPARD results via the SPARD-DSS it will be important to consider the following development and perceptions:
In all the SPARD-IS was well perceived at the final workshop in Brussels as a tool to present the comprehensive results of the SPARD project. The SPARD-IS was originally planned to be established at the EU Commission level, but the major problem was a lacking stable demand-driven end user group, which would have been able to continuously accompany the development of the SPARD-IS. Based on the conducted requirement analysis it was evident that DG AGRI has in-house services and that external tools for decision making need more time of being implemented at Commission level. This decision on the development of SPARD-IS instead of a tailored SPARD Decision Support System for selected end users was taken commonly within the SPARD consortium. The most-adequate and best fitting approach was elaborated to illustrate the results of SPARD (1) in a comprehensive way, (2) to maximise the outreach of SPARD results to the wider public and last but not least (3) to raise awareness of potential end user at the level of the EU Commission. With regard to the technical development of the SPARD-IS this task was important base to implement the system as web-based tool. The protocols supported the technical linking-up between different software components and assured an efficient functionality.
In all the SPARD-IS was well perceived at the final workshop in Brussels as a tool to present the comprehensive results of the SPARD project. The most-adequate and best fitting approach was elaborated to illustrate the results of SPARD (1) in a comprehensive way, (2) to maximise the outreach of SPARD results to the wider public and last but not least (3) to raise awareness of potential end user at the level of the EU Commission. The wide dissemination of the SPARD results can raise awareness of the project and feedback from different institutions, who work in the similar research area, can link-up with the SPARD consortium in order to exchange knowledge at as open research platform.
List of Websites:
The project website: www.spard.eu
The Information System SPARD-IS: www.spard-is.eu
Project co-ordinator: Dr. Annette Piorr, Leibniz-Centre for Agricultural Landscape Research (ZALF e.V.) Eberswalder Str. 84, 15374 Muencheberg, Germany. Email: apiorr@zalf.de
Project Officer: Dr. Hans-Jörg Lutzeyer, DG RTD
