A (critical) look to composite indicator construction for European Regions

The universe of composite indicators is large and variegated, due to the multiple methods used to aggregate the individual components. The choice of the aggregation method is crucial for the construction of the composite indicator and affects the resulting ranking. In this paper we propose a comparison between five different aggregative methods with the purpose of investigating differences and similarities between them from a qualitative and quantitative point of view. To illustrate our proposal, we use five Eurostat variables at regional (NUTS2) level and we compute the correlation and the pairwise Euclidean distance between the composite indicators corresponding to the five aggregation methods considered. Moreover we identify, among the composite indicators, the one that is, respectively, most correlated and closest to the other ones. Results show that the composite indicators identified by the the correlation and distance criteria do not coincide. This suggests that the topological features of the composite indicator is quite affected by the aggregation method. Therefore the choice of the method must be the result of a well-determined objective.