Regional convergence and agglomeration in Argentina : a spatial panel data approach

This paper examines the per capita income convergence process among 23 Argentinian provinces over the period 1983-2002. The purpose of this work is to apply new estimation methods following two-step procedure as in Badinger et al. (2004). We combine a spatial filtering of variables to remove the spatial correlation (Getis and Griffith, 2002) and suitable estimators for dynamic panels (using first-differenced and system GMM estimators). Our estimations on filtered variables reveal a conditional convergence process between Argentinian provinces and a positive and significant impact of agglomeration variables on growth rate. Hence, our results show that ignoring spatial distortions due to geographic proximity misleads estimations and underestimates the speed of convergence specifically for provinces which are distant from Buenos Aires. Moreover, we improve estimations of agglomeration effects when spatial autocorrelation is removed.