A spatially filtered mixture of β-convergence regressions for EU regions, 1980–2002

Assessing regional growth and convergence across Europe is a matter of primary relevance. Empirical models that do not account for structural heterogeneities and spatial effects may face serious misspecification problems. In this work, a mixture regression approach is applied to the β-convergence model, in order to produce an endogenous selection of regional growth patterns. A priori choices, such as North—South or centre-periphery divisions, are avoided. In addition to this, we deal with the spatial dependence existing in the data, applying a local filter to the data. The results indicate that spatial effects matter, and either absolute, conditional, or club convergence, if extended to the whole sample, might be restrictive assumptions. Excluding a small number of regions that behave as outliers, only a few regions show an appreciable rate of convergence. The majority of data show slow convergence, or no convergence at all. Furthermore, a dualistic phenomenon seems to be present inside some States, reinforcing the “diverging-convergence” paradox.(This abstract was borrowed from another version of this item.)