The aim of the paper is twofold. First, it estimates the process of agricultural and economic convergence and of inequalities between the two across a sample of 259 EU-25 regions from 1991-2007. Than, it investigates the main sources of convergence between agricultural and economic labour productivity at the regional level. The empirical study is developed taking into account the effects of spatial heterogeneity and spillover effects.

Data

The regional per worker gross domestic product and the agricultural value added per worker represent the variables of the analysis. Their value in 1991 and 2007 has been drown from the Cambridge Regional Economics data.

Methods

The empirical analysis starts testing the neoclassical hypothesis of absolute b-convergence by an Ordinary Least Squares (OLS) approach for the agricultural sector and the overall economy and inequality between the two. The econometric specification makes reference to a version of the model developed by Barro and Sala-i-Martin and Sala-i-Martin. Spatial nonstationarity and heterogeneity across the regions of the sample is detected by the Geographically Weighted Regression (GWR) approach that replaces global OLS regression coefficients by local parameters. This technique, addressing the issue of estimating locally different parameters, is of specific interest for detecting territorial approach to convergence. Mapping the local parameters estimated by GWR, it is investigated whether spatial heterogeneity is linked to the concept of convergence clubs, characterised by possible multiple, locally stable, steady state equilibria and their spatial autocorrelation.

The main sources of convergence between agricultural and economic labour productivity at the regional level are investigated examining the extent to which employment structure has become more similar across regions and the harmonization process of sectoral employment across territorial units.

Main results

Various tests aimed at investigating the presence of spatial effects lead to a GWR model as more appropriate model specification over the OLS. Neglecting these effects, results are unreliable. More precisely, OLS estimates reveal quite distinct geographic pattern of residuals, affected by a spatial variation of the relationships in the model.

The analysis has pointed out a prevailing negative correlation between agricultural and economic speed of convergence at the regional level mainly due to convergence in productive structure across regions with a transfer of resources from agriculture to sector characterised by a higher productivity, process that has been more marked in the poor than in the rich regions.