Data aggregation in constructing composite indicators: A perspective of information loss

Composite indicators (CIs) have been widely accepted as a useful tool for performance comparisons, public communication and decision support in a wide spectrum of fields, e.g. economy, environment and knowledge/information/innovation. The quality and reliability of a CI depend heavily on the underlying construction scheme where data aggregation is a major step. This paper analyzes the data aggregation problem in CI construction from the point of view of information loss. Based on the “minimum information loss” principle, the distance-based and entropy-based aggregation models for constructing CIs are presented. The entropy-based aggregation model has also been extended to deal with qualitative data. It is shown that the proposed aggregation models have close relationships with several popular MCDA aggregation methods in CI construction, although our proposed models seem to be more flexible while more complex in application. Two case studies are presented to illustrate the use of the proposed aggregation models.