A capacity approach to territorial resilience: the case of European regions

“The high resilience allows tests of those novel combinations because the system-wide costs of failure are low. The result is the condition needed for creative experimentation” (Holling 2001).

Regional Studies recently dedicated a themed issue right at the “resilience revisited” (Bailey and Turok 2016), where different authors analyze the resilience processes following this approach in different English, Turkish, Canadian, Australian and European regions.

The World Bank developed a composite indicator of people’s preparation for risk at country level which comprises measures of assets across four components of socio-economic systems: human capital, physical and financial assets, social support, State support (World Bank 2014).

These regions belongs to 21 European countries, that do not include Bulgarian, Cypriot, Croatian, Lithuanian, Latvian, Maltese, Norwegian, Rumanian regions and the French islands, for the presence of too many missing values.

The Cliff and Ord model is described by two equations: \(Y={\lambda }WY+X\beta +u\) and \(u={\rho }Wu+\varepsilon \), where \(\varepsilon \) is assumed to be independent and identically distributed or independent but heteroscedastically distributed, and the heteroscedasticity is of unknown form. In the right member of the first equation among the regressors, a variable of spatial lag is inserted. It is defined pre-multiplying the variable to regress (Y) with the spatial-weighting matrix (W). This endogenous variable (WY) is the weighted average of the dependent variable values observed in other statistical units. By means of the second equation, the model takes into account also that the error terms (u) are generated by a spatial autoregressive process. These two equations estimate simultaneously the values of each statistical unit. Consequently, the values of the dependent variable are not independent from each other. The W matrix is constructed as a matrix of contiguity, considering the geographical proximity of the European regions, and weighing the proximity with the geographical coordinates of the centroid of each region. The parameter \({\lambda }\) measures the intensity of spatial interactions and spillovers effects. The parameter measures the spatial autocorrelation of the error term. The parameter \({\beta }\) takes into account that in the dependent variables the values of each region are determined simultaneously from each other (Drukker et al. 2013).

This estimator is consistent and robust when the error term is heteroscedastically distributed.

The environmental resilience cannot be taken under consideration because of missing data. This causes the reduction in the number of European regions considered.