Sectoral productivity convergence between European regions : does space matter?

The aim of our analysis is the evaluation of the total and sectoral convergence of labour productivity between 182 regions of EU12 in the period 1991-2006. The selected sectors are agriculture, manufactory, market and non-market services. We adopt a β- and sígma-convergence approach along with a methodology based on Getis' spatial filters that allows decomposing variables into their spatial and a-spatial components ensuring their spatial independence. This last point is fundamental to avoid i) omitted variables and/or problems of bias and/or inconsistency of coefficients in growth regressions, ii) bias in the computation of variance. The cited econometric approach also permits to identify spatial regimes of regions with high and low productivity, respectively 'core' and 'periphery'. Our results show significant sígma-convergence in aggregate labour productivity, market and non-market services in the first years and a slightly divergence in the second period. Sigma-divergence is present only in manufactory sector when spatial factors are not considered, while in agriculture a strong convergence is perceivable. The analysis highlights that if we do not take into account spatial effects, sígma-convergence is overestimated. In the second part of the paper, we evaluate β-convergence for total labour productivity and for each sector. We considered five cases: the whole sample (EU12) without dummies, with country dummies and with spatial regime dummies, the 'core' and the 'periphery'; finally we take into account singularly each spatial regime. The same estimation performed with and without spatially filtered variables leads to different results. While β-convergence process takes place in all cases and with all techniques, countries dummies are statistically significant only when spatial effects are not considered. In case of spatially filtered variables, the dummies, whose aim is to take into account the specificity of an agglomeration of regions, lose their significance, both theoretical and statistical, because the spatial factors embedded in each variable, and strictly connected with the country or convergence club to which they belong, are removed. The interesting point, however, is the lack of significance of dummies for spatial regimes. This finding, common both to filtered and unfiltered variables and to all sectors, is a little surprising because the identified spatial regimes are interpretable (and often interpreted) like convergence clubs. In conclusion, this paper shows that economic structure has to be considered together with spatial structure. These two factors affect themselves reciprocally and, for a full and reliable explanation of regional economic dynamics both must be formally included in the analysis.