Using Constrained Optimization for the Identification of Convergence Clubs

In the last years a central issue in regional economic growth debate is represented by the convergence problem. Many empirical economists have noticed that per-worker GDP of poor regions tend to converge to those of richer regions. However more recently it has been observed that the economic convergence might not be achieved if, in the empirical analysis, we consider the entire data set as one sample. The phenomenon should be analyzed considering regions as belonging to different sub-samples with quite similar economy. Many authors refer to this hypothesis as economic convergence clubs. The definition of these homogeneous groups represents a crucial issue in many regional economic growth studies. The aim of this paper is to propose a method for the identification of convergence clubs for the European regions at NUTS 2 level. The econometric specification used is based on the classical, and spatial augmented version of the conditional β-convergence model. Two different optimization algorithms for the identification of convergence clubs are proposed and compared: Simulated Annealing and Iterated Conditional Modes. Copyright Springer Science+Business Media New York 2013