Multiple Regimes in Cross-Region Growth Regressions with Spatial Dependence: A Parametric and a Semi-parametric Approach

This paper studies the distribution dynamics of development across European regions over the period 1975-2000. Regional development is measured in terms of both per capita GDP (Y/P) and its components: labour productivity and employment ratio (that in turn can be decomposed in terms of activity and unemployment rate). The Core/Periphery pattern in the European Union is firstly investigated and a comparative analysis in terms of income, productivity, employment and unemployment rates of the two partitions is carried out. Moreover, for each variable as well as for each partition, a nonparametric beta convergence analysis is applied. Synthetically, the results confirm the lack of regional convergence in per capita incomes, the presence of a negative quasi-linear relationship between growth rates and initial levels of labour productivity and a U-shaped relationship between growth rates and initial levels of unemployment rates. As it is well known, however, b-convergence analysis does not allow any test of multiple equilibria, such as “emerging twin peaks”, in the growth process. Equilibrium multiplicity can be properly assessed by using nonparametric techniques of analysis of the cross-regional distribution. In particular, a way to quantify the intra-distribution dynamics is the multivariate kernel, which estimates the joint density of regional income, productivity and (un)employment distribution at time t0 and t0+t. The results of this analysis suggest that over the period considered the regional growth pattern in Europe has followed a polarisation process rather than a convergence path. This appears particularly true in the case of per capita incomes and unemployment rates. Finally, in order to “explain” polarisation, conditional multivariate kernels are estimated. In particular, the role of spatial contiguity and regional sectoral specialisation is investigated.