Substitutability vs. complementarity: Rethinking STI composite indicators building with an application of sensitivity analysis

The paper challenges the common view that the arithmetic mean aggregation's rule for building STI indicators is the fairest approach. We argue, on the contrary, that it hides a silent "substitution assumption" among the sub-indicators involved. According to this perspective Casadio and Palazzi (2004) provided a new composite indicator based on a "concave mean". This approach overcomes the "substitution bias" generated by the arithmetic aggregation. As the concave mean is a highly non-linear formula, the paper provides a Sensitivity Analysis (as suggested by Saltelli et al., 2004) to detect sub-indexes importance in this special case. We perform an application to a specific STI indicator comparing results from the concave mean and other aggregation rules.