An Estimate of the Degree of Interconnectedness between European Regions: A Bayesian Model Averaging Approach

This paper provides a methodology based on General Variance Decomposition and Bayesian Model Averaging to estimate the degree of economic interconnectedness across different regions. In a sample of 199 European NUTS2 regions in the period 1980-2008 the connectedness appears very heterogeneous and not symmetric,the idiosyncratic component not very relevant (at most 14% of the variance is due to shocks arising and remaining in the same region), as well as the common component. While a clear pattern of core-periphery emerges, this appears not determined by the geographical proximity. The country component is not very significant, heterogeneous across countries, and proportional to countries' size. Finally, the comparison of the estimated connectedness matrix with two spatial matrices generally used in spatial econometrics (a first-order contiguity and a distance-based matrix) reveals that both are very far from representing the actual interconnectedness between European regions.