A trichotomic segmentation approach for estimating composite indicators

Abstract

Composite indicators (CIs) are commonly used for benchmarking of countries over the years, summarizing in a single measurement, complex social, economic, environmental etc. concepts by involving several thematically related sub-indicators. When estimating CIs and for a few specific countries, it is possible to have a strong indication and belief about their performance, prior to obtaining their scores. Based on that, a number of countries are likely to occupy the top-ranking positions; some will remain at the bottom list, while others may range in intermediate places. This initial preference information imposes a trichotomic segmentation that divides the countries under assessment into categories of superior, inferior and of ambiguous future performance. In this paper, we introduce the trichotomic segmentation as initial preference information to estimate the values of the CIs. We build on the popular Benefit of the Doubt (BoD) method with common weights and develop a two-goal linear programming model, that next to the evaluation of the common weights for the sub-indicators, estimates cut-off points for the CI values that distinguish the superior and inferior countries. The proposed model maintains the advantages of the common variable weighting and produces scores that induce better discrimination of the countries, having also a significant correlation with the original CI scores. The proposed methodology is applied to reassess the Digital Economy and the Society Index.