Abstract

We investigate the relative impact of geographic features on the location of production in the European Union. Specifically, we attempt to quantify how much of the spatial pattern of GDP can be attributed to exogenous first-nature elements alone and how much can be derived from endogenous second-nature factors. In order to disentangle both effects empirically, we control for second-nature factors. A method based on the decomposition of the per capita GDP variance is applied to a panel of 1,171 European NUTS 3 regions for 2006. We demonstrate that variable and model misspecifications may bias results unless proper allowance is made for spatial autocorrelation and spatial heterogeneity, as well as multicollinearity and endogeneity.