Distribution dynamics: a spatial perspective

It is quite common in cross-sectional convergence analyses that data exhibit spatial dependence. Within the literature adopting the distribution dynamics approach, authors typically opt for spatial prefiltering. We follow an alternative route and propose a procedure based on an estimate of the mean function of a conditional density for which we develop a two-stage non-parametric estimator that allows for spatial dependence estimated via a spline estimator of the spatial correlation function. The finite sample performance of this estimator is assessed via Monte Carlo simulations. We apply the procedure that incorporates the proposed spatial non-parametric estimator to data on per capita personal income in US states and metropolitan statistical areas.