Spatial HAC estimator: analysis of convergence of European regions

This paper applies a nonparametric heteroskedasticity and autocorrelation consistent (HAC) estimator of error terms in the context of the spatial autoregressive model of GDP per capita convergence of European regions at NUTS 2 level. By introducing the spatial dimension, it looks at how the equilibrium distribution of GDP per capita of EU regions evolves both in time and space dimensions. Results demonstrate that the global spatial spillovers of growth rates make an important contribution to the process of convergence by reinforcing the economic growth of neighboring regions. Results are even more pronounced when the convergence in wage per worker is considered. The choice of kernel functions does not significantly affect the estimation of the variance-covariance matrix, while the choice of the bandwidth parameter is quite important. Finally, results are sensitive to the weighting matrix specification, and further research is needed to give a more rigorous justification for the selection of the weighting matrix.