Spatial disparities in the European agriculture: a regional analysis

Abstract

This article examines the territorial imbalances in European agriculture during the period 1980 to 2001, by means of the information provided by various methodological instruments which allow us to overcome the drawbacks of conventional convergence analysis. The results obtained reveal that the regional distribution of productivity in the agricultural sector is characterized by the presence of positive spatial dependence. This fact implies that the European regions in close spatial proximity register similar levels of the variable under study, which highlights the relevance of geographical location in this context. The empirical evidence presented also shows that regional disparities have remained almost constant during the time interval considered. However, the increase in density around the European average explains the observed reduction in the degree of bipolarization, while intra-distribution mobility is relatively limited. Finally, the analysis carried out allows us to assess the role of variables such as country of origin, investment per worker in the agricultural sector, regional per capita income or the size of the agrifood industry, in explaining the dynamics of the distribution under analysis.

Acknowledgements

The authors thank an anonymous referee for their helpful comments and suggestions. Financial support from Spanish MEC (Coordinated Project SEJ2005-08738-C02-01 and 02) is gratefully acknowledged.

Notes

¹ A review of the main conclusions reported in this literature can be found in Armstrong (2002) or Terrasi (2002).

² Article 2 of the Treaty of the European Union specifically states that: ‘the Community mission will be to promote (…) the harmonious, balanced and sustainable development of economic activities, sustained growth (…), a high degree of convergence of economic performance (…)’.

³ In relation to this, see Paci (1997), Paci and Pigliaru (1999) or Gil et al. (2002).

⁴ The only exception, to our knowledge, is the work by Colino and Noguera (2002). There are, however, several examples of research efforts using national-level data to investigate this issue within the European setting, among which we could mention Schimmelpfenning and Thirtle (1999), Gutierrez (2000), Ball et al. (2001) or Aldaz and Millán (2003).

⁵ This policy, the ultimate aim of which was to increase productivity in European agriculture, in line with the provisions of the Treaty of Rome, used 46.5% of the Community budget in 2002.

⁶ NUTS is the French acronym for ‘Nomenclature of Territorial Units for Statistics’, a hierarchical classification of subnational spatial units established by Eurostat. In this classification, NUTS-0 corresponds to country level and increasing numbers indicate increasing levels of subnational disaggregation.

⁷ In any event, lack of information for the whole of the time interval considered, has obliged us to omit from our analysis the countries newly incorporated into the European Union in May 2004, as well as the former East German Länder and the French Overseas departments. We also decided not to include data for Brussels or Inner London, two regions with practically negligible levels of agricultural sector employment, but which, nevertheless, over time, register major fluctuations in the variable of interest, thus affecting the interpretation of our results.

⁸ This is in fact the option taken, e.g. by López-Bazo et al. (1999) or Rey and Montouri (1999).

⁹ In order to check the robustness of the conclusions obtained, we considered various spatial weight matrices. In particular, we constructed two more matrices W based on the 15- and 20-nearest neighbours. Nevertheless, the results are in all cases very similar to those reported in the text.

¹⁰ In fact, this result would be of particular importance if, following the approach adopted by Paci (1997), Colino and Noguera (2002) or Gil et al. (2002), the aim of the analysis were to examine regional disparities in European agriculture through the estimation of convergence equations. Indeed, the information supplied by Table 1 raises serious doubts as to the consistency, unbiasedness and/or efficiency of estimations obtained without taking into account the existence of spatial dependence. A more detailed analysis of this issue can be found in Anselin (1988).

¹¹ To further confirm this finding, we also constructed the Moran's scatterplots for the distribution under analysis. This is a graph on which the standardized values of the variable to be analysed are plotted on the horizontal axis and the spatial lag of the same variable on the vertical axis. Thus each of the quadrants represents a different type of spatial association. According to Figs 16 and 17, which appear in the Appendix, there is a noticeable concentration of regions in Quadrants I and III, both at the beginning and at the end of the study period, thus confirming a predominating tendency in European agriculture toward spatial clustering involving regions with similar values of farm productivity, while there are relatively few cases of major disparity between the agricultural productivity of any given region and the average of its neighbours.

¹² Note that , given that .

¹³ Dalgaard and Vastrup (2001) have demonstrated that the joint use of the SD of the logs and the coefficient of variation does not prove redundant in this setting, since these two dispersion statistics could yield different conclusions.

¹⁴ A similar result is found by Paci (1997) in a more limited geographical and temporal setting than that covered in the present study. Indeed, similar conclusions have been reached in national analyses. For the Spanish case, see, e.g. Mas et al. (1994), Raymond and García-Greciano (1994) or Salinas-Jiménez (2003).

¹⁵ All estimates were based on Gaussian kernel functions, while the smoothing parameter was determined in each case following Silverman (1986, p. 47). It is worth mentioning that the results obtained are robust to the kernel function used.

¹⁶ Note that in this case, there is no overlapping between the various groups, since the decomposition of the Gini index into between-group and within-group inequality is exact (Pyatt, 1976).

¹⁷ The optimal partitioning for the two-group case is characterized by the fact that the productivity level that separates the two groups coincides with the sample average. By adopting this criterion for the classification of the various regions considered, it is possible to explain an average of 71% of total inequality, measured in terms of the Gini index. Thus, the within-group inequality left unexplained by this partition is about 29%.

¹⁸ For example, for the European case, Lopez-Bazo et al. (1999) and Le Gallo (2004) estimate various transition matrices to analyse regional mobility in terms of per capita income and aggregate productivity.

¹⁹ Gaussian kernel functions were used in all cases, while the smoothing parameter was selected following Silverman (1986, p. 86).

²⁰ Though estimations were repeated for different transition periods, the results in all cases proved very similar to those just discussed. They are not reported here for lack of space, but are available from the authors upon request.

²¹ Having reached this point, however, it should be noted that any comparison of Figs 7 and 11 must be based exclusively on the shape of the distributions, since there is no point in comparing the density levels that appear on the vertical axes.

²² This last variable was proxied by the contribution to regional gross value added made by the food, beverages and tobacco industry (sectoral classification NACE-CLIO R17). It should be noted in this respect that, despite our best efforts, we were unable to obtain data at a higher level of sectoral disaggregation to cover the entire geographical and temporal scope of the present study.

²³ This type of methodology has been used, e.g. by Ezcurra et al. (2005) to investigate the causes of regional disparities in welfare in the European Union.