How rural the EU RDP is? An analysis through spatial funds allocation Beatrice Camaioni, Roberto Esposti, Antonello Lobianco, Francesco Pagliacci, Franco Sotte Department of Economics and Social Sciences Università Politecnica delle Marche Ancona (Italy) Paper prepared for presentation at the 2nd AIEAA Conference Between Crisis and Development: which Role for the Bio-Economy 6-7 June, 2013 Parma, Italy WELFARE, WEALTH AND WORK A NEW GROWTH PATH FOR EUROPE A European research consortium is working on the analytical foundations for a new socio ecological growth model OUTLINE 1. Introduction 2. Defining EU rural areas 3. Dataset 4. Methodology: defining Peripherurality 5. Main results: territorial patterns 6. Conclusions 2 Parma, June 6 th, 2013
1. Introduction 1. Defining rural areas: o EU rural areas are going through major transformations: need for specific and clear definitions o In spite of their importance, geographical issues have largely been ignored in defining rural areas (Copus, 1996; Ballas et al., 2003; Bollman et al., 2005; Vidal et al., 2005; Copus et al., 2008). o A geographical analysis could help policymakers in better framing EU policies. The second pillar of the CAP (EAFRD) is supposed to support rural areas in facing new challenges (Sotte, 2009; Esposti, 2011). 2. Analysis of RDP expenditures spatial allocation (ex-post allocation) o Analysis is performed at the local level (i.e., the NUTS 3 level) according to the national/eu decision-making process (top-down approach) o Analysis of the real capacity of local areas in better using RDP funds (bottom-up approach) 3 Parma, June 6 th, 2013 3 2. Defining rurality at the EU level (I) The role of density in assessing rurality No homogeneous definition of rural areas at the EU level (Montresor, 2002). Definitions from OECD (1994; 1996; 2006) and Eurostat (2010) are the most widely used: both are based on density. However, dichotomised urban-rural typologies are largely outdated (Sotte et al., 2012). Multi-dimensional approaches to define rurality The FAO-OECD Report (2007) and The Wye Group (2007) suggest multidimensional approaches in defining rural areas. A collection of quantitative analysis on definitions of rural areas is provided by Copus et al. (2008). 4 Parma, June 6 th, 2013
2. Defining rurality at the EU level (II) Geographical approaches to compute peripherurality The work is aimed at providing new insights on the EU rural areas by linking together socio-economic indicators with geographical ones. Three main thematic areas are considered: 1.Economic indicators and role of agriculture (sector-based approach); 2.Land use and landscape features, e.g., share of agricultural areas or forests on the total surface compared to share of artificial areas (territorial approach); 3.Accessibility / remoteness, according to different territorial scales, e.g. EU level, national level, sub-national level (geographical approach) By linking together these perspectives a comprehensive Peripherurality Index can be computed. 5 Parma, June 6 th, 2013 3. Dataset (I) The dataset: variables and statistical sources Source: personal elaboration 6 Parma, June 6 th, 2013
3. Dataset (II) Geographical coverage Data are collected at the NUTS 3 level (NUTS 2006 Classification, Commission Regulation (EC) No 105/2007) 1,288 observations out of 1,303 NUTS 3 regions in the EU-27. Drawbacks NUTS 3 level allows to properly analyse EU rural areas. However, EU policies are mainly implemented at the NUTS 0 and 2 level. Many missing values: they are replaced with data at the regional / national level (e.g., unemployment rate; per capita GDP; average farm size; average SGM). Some variables may suffer from the existence of a Country effect. It may alter both the PCA and the CA. Time coverage Last available figures refer to years 2007 to 2010. 7 Parma, June 6 th, 2013 4. Methodology: defining peripherurality (I) 1. Summing up rural areas features: Principal Component Analysis PCA is a variable reduction technique that maximizes the amount of variance accounted for in the observed variables by a smaller group of variables (Hotelling, 1933; Pearson; 1901). Principal components (PCs) are linear combinations of the original variables. Each observation can be assigned a score on a standardized scale. 2. Computing the Peripherurality Index From the PCs, a comprehensive Peripherurality Index is computed: euclidean distance from an ideal urban benchmark (the average value for the two EU global MEGAs: Paris and London). 2 PR index i = ( x ) p ip xubp where: -X ip represents the score of the i-th NUTS 3 region in the p-th component; -X ubp represents the score of the urban benchmark in the p-th component. 8 Parma, June 6 th, 2013
4. Methodology: defining peripherurality (II) 3. Identifying rural typologies: cluster analysis Unsupervised learning approach (Tryon, 1939; Johnson, 1967). According to a distance matrix on p measurable characteristics, objects are grouped in such a way that those in the same group are more similar to each others than to those belonging to other clusters. Two approaches (Kaufmann and Rousseeuw, 1990; MacQueen, 1967): o Hierarchical approaches recreate a whole hierarchy of clusters (agglomerative vs. divisive clustering). Graphical output: dendrogram; o Partitioning approaches (K-means and k-medoids clustering) create k nonoverlapping clusters, through iterative algorithms. 9 Parma, June 6 th, 2013 5. Main results: PCA PCs extraction Five PCs are extracted according accounting for 67% of the total variance 10 Parma, June 6 th, 2013
5. Main results: The Peripherurality Index According to the 5 PCs, the PR Index is computed moving from the Urban benchmark (London and Paris). The greater the index, the more rural and/or peripheral the region. Highest values: Mediterranean regions, new MSs, Northern Scandinavia Source: own elaboration (software R) (EuroGeographics for administrative boundaries) 11 Parma, June 6 th, 2013 5. Main results: Cluster analysis (I) Algorithm for cluster analysis A hierarchical cluster analysis is performed according to the 5 PCs o Euclidean distance o AGNES-algorithm (AGGlomerative NESting) + Ward s method (Lance and Williams, 1966; Ward, 1963). o No standardization is applied. Seven clusters are identified. Defining typologies: cluster centres according to the 5 PCs 12 Parma, June 6 th, 2013 Economic / geographical centrality Demogr. shrinking / ageing Manufacturing in rural areas Land Use: forests vs. agric. areas Urban dispersion Peripheries -3.25-0.65-0.68 0.08-0.43 Green lungs -0.10-0.07-0.41 1.43-1.40 Cities 3.42-1.47-1.29-0.15 0.97 Remote regions -6.33-0.89 0.00-0.77 1.89 Agricultural districts 1.10-0.01 0.85-1.06-0.72 Shrinking regions 0.38 4.09-1.70-1.10 0.46 Industrial core 0.54 0.42 1.16 1.10 0.53
5. Main results: Cluster analysis (II) The territorial distribution of the clusters 13 Parma, June 6 th, 2013 Source: own elaboration 5. Main results: the spatial allocation of the RDP expenditures (I) Allocation of the total expenditures from EAFRD for years 2007 to 2009 is considered (Source: European Commission). Data have been collected at the NUTS 3 level (same level of territorial disaggregation). Three specific indexes of expenditures intensity are computed (source: Eurostat Farm Structure Survey / Eurostat National Accounts): RDP expenditure ( ) per unit of utilized agricultural area (UAA in ha); RDP expenditure ( ) per unit of agricultural labour work (annual work unit, AWU); RDP expenditure ( ) per unit of agricultural gross value added (Agricultural GVA, in.000 ). 14 Parma, June 6 th, 2013
5. Main results: the spatial allocation of the RDP expenditures (II) 1. 2. 3. Source: own elaboration (software R package spdep) (EuroGeographics for administrative boundaries) 15 Parma, June 6 th, 2013 5. Main results: the spatial allocation of the RDP expenditures (III) Average RDP expenditures per typology of urban-rural regions (Eurostat, 2010) and per cluster EAFRD Expenditures Exp. per UAA Exp. per AWU Exp. per Agri GVA PR regions 130.76 3,048.21 154.70 IR regions 111.33 2,997.10 117.72 PU regions 101.07 2,625.86 89.82 Peripheries 135.55 1,801.96 137.12 EU green lungs 147.76 4,720.54 180.06 Cities 120.35 3,273.43 96.04 Remote regions 60.44 533.27 124.88 Agricultural districts 69.89 2,067.80 76.20 Shrinking regions 152.09 7,797.13 242.67 Industrial core 141.77 2,714.20 127.04 16 Parma, June 6 th, 2013
5. Main results: the spatial allocation of the RDP expenditures (IV) Pearson correlation coefficients: RDP expenditures and indicators of remoteness / rurality Exp. per UAA Exp. per AWU Exp. per Agri GVA Density 0.033 (0.245) 0.091 (0.001) -0.009 (0.760) PC1 - Economic and geographical centrality -0.028 (0.325) 0.133 (1.85e-06) -0.117 (2.5e-05) PR Index -0.023 (0.032) -0.073 (0.009) 0.090 (0.001) 17 Parma, June 6 th, 2013 6. Conclusions This analysis sheds new light on rural/peripheral areas across Europe Multi-dimensional approach in defining rural areas: through the Peripherurality Index, a more complex geography at the EU scale emerges The RDP seems to be not so green as it was supposed to be: urban/central regions show a greater capacity in spending RDP funds, although results vary according to the different measurements of rurality According to these results, an empirical/econometrical analysis could help in defining the main economic drivers affecting such a spatial allocation Next steps: testing the causal relationships in the spatial allocation of RDP expenditures at the EU level! 18 Parma, June 6 th, 2013
Thanks for your attention Francesco Pagliacci Università Politecnica delle Marche f.pagliacci@univpm.it 19 Parma, June 6 th, 2013