Incorporating spatial information in entropy-weighted composite indicators: An application to education and training in Italy

This study introduces a novel methodology in which a composite indicator derived using the Entropy Weight Method is decomposed into a location-specific component and a spatial component. Although EWM is widely recognized for its robustness in constructing data-driven composite indicators, standard applications typically overlook spatial information. To address this limitation, we combine EWM with spatial filtering techniques to disentangle location-specific effects from spatial spillovers, thereby improving the ability of the composite indicator to capture regional dynamics. This decomposition makes it possible to assess the relative contribution of local characteristics and spillover effects to the overall magnitude of the indicator in each region. In addition, we propose a random permutation test to evaluate the statistical significance of the spatial component. The methodology is applied to the construction of a composite indicator for education and training across 107 Italian provinces, using data from the Benessere Equo e Sostenibile (BES) dataset. The results show that the proposed approach provides policy-relevant insights that support targeted interventions and efficient resource allocation.