Semi-parametric Spatial Auto-covariance Models of Regional Growth Behavior in Europe

We propose a semi-parametric spatial auto-covariance specification of the growth model to examine the growth behaviour of European regions in the period 1988-2000. This specification simultaneously takes account of the problems of non-linearities and spatial dependence. We obtain two main results. First, there is a trade-off between the identification of non-linearities and the estimation of the spatial parameters. Second, even when controlling for spatial dependence, there is evidence of strong non-linearity between regional growth and its predictors. In particular, the relation between growth and unemployment is negative but not globally linear: regions with an unemployment rate over a certain threshold appear to be 'locked' within an 'underdevelopment trap' with similar lower growth rates.