Composite indicators as generalized benefit-of-the-doubt weighted averages

Composite indicators (CIs) are usually computed as arithmetic (weighted) averages of (often normalized) sub-indicators. Several studies criticized this procedure for implying requirements and properties that are often hard to maintain in practical applications. Recent studies explored the multiplicative aggregation and more specifically the geometric (weighted) average as aggregator in CI-building. This paper takes this exploration one step further by considering other members of the family of generalized (weighted) averages as aggregator in the construction of CIs. It is argued that the choice for a particular version of the generalized weighted average enables to reflect decision makers' attitudes in the evaluation. For the Human Development Index, results show that the choice for a specific version of the generalized (weighted) average as aggregator in CI-construction impacts country rankings.