A Multicriteria Approach to the Human Development Index Classification

Abstract

This paper proposes a new Human Development Index (HDI) classification method using as a basis the same dimensions and weights used by United Nations Development Programme (UNDP), however, avoiding the compensatory effect and misclassification that can occur when UNDP approach is applied. The proposed approach also allows working with any set of dimensions and weights. The classification was performed by the Elimination Et Choix Traidusaint la Realite (ELECTRE TRI), a multicriteria method, in which classes were defined with the visual aid of a non-parametrically tool, the Kernel Density Estimation and the class profiles were calculated by the Jenks Natural Breaks algorithm. The proposed classification method achieved better results than the one adopted by the UNDP, because the countries could be classified accurately in their natural classes without occurring the compensation effect. However, when comparing with the HDI index we have obtained an accuracy of 76.60% for the first proposed model and 78.72% for the second one. Our proposal purely used the same dimensions and weights adopted by UNDP. The proposed method is more robust, closer to reality and can eliminate the compensatory effect between dimensions, thus it can provide more credibility to the HDI. The combination of the ELECTRE TRI method with statistical tools to define classes and class profiles for the HDI is unprecedented in the literature.