Convergence Analysis as Distribution Dynamics When Data are Spatially Dependent

Abstract

Conditional distributions for the analysis of convergence are usually estimated using a standard kernel smoother but this is known to be biased. Hyndman et al. (1996) thus suggest a conditional density estimator with a mean function specified by a local polynomial smoother, i.e. one with better bias properties. However, even in this case, the estimated conditional mean might be incorrect when observations are spatially dependent. Consequently, in this paper we study per capita income inequalities among European Functional Regions and U.S. Metropolitan Statistical Areas through a distribution dynamics approach in which the conditional mean is estimated via a procedure that allows for spatial dependence (Gerolimetto and Magrini, 2009).

Keywords: Regional convergence, Distribution dynamics, Nonparametric smoothing, Spatial dependence

JEL Classification: R10, O40, C14, C21