Analysing Human Development Index from MultiCriteria Decision Making point of view

Fatma Ezzahra Hadj Ammar1 and Meltem Öztürk2
1École supérieure en sciences et technologies de l’informatique et du numérique
2Université Paris-Dauphine, PSL Research University, CNRS, LAMSADE, Place du
Maréchal de Lattre de Tassigny, F-75775 Paris cedex 16, France


Human Development Index (HDI) [1] is a composite indicator (aggregation of three dimensions: health, education and standard of living) developed by United Nations Development Program (UNDP) to assess the level of development of a country. It orders 189 countries by proving for each country an overall score. These scores are then used to classify the countries in 4 categories : very high, high, medium and low human development [2]. UNPD hopes to use this indicator in order to stimulate debate about government policy priorities.

In this work, we analyze this indicator from MultiCriteria Decision Making (MCDM) point of view since the design of a composite indicator is a multicriteria problem. In particular, we will try to answer three questions:

  • do we need a complex aggregation operator and sophisticated normalizations to deduce an order on the countries (our analysis is based on the notion of pareto dominance)?
  • The computation of HDI has undergone several modifications since 1990 (arithmetic mean has been replaced by geometric mean with different types of normalization; for example, for the normalization of gross national income per capita, the Atkinson transformation has been replaced by the logarithm). What are the impacts of such modifications?
  • The classification of countries in 4 categories is done in a rather arbitrary way. Can we obtain the same classification if it was based on an explicable and nonutilitarian classification method like ELECTRE TRI? If so, what would be the parameters of the model (we will show that if we want to use ELECTRE TRI which is a less compensatory but more explainable approach than the one actually used, we will not be able to obtain the current classification of the countries).

We will conclude with a discussion on contributions that multicriteria decision making can have (with it’s modeling methodologies, algorithmic tools, axiomatic analysis, …) for the design of composite indicators.

[2] For instance some data for 2020: Norway (rank1, very high), Australia (rank 8, very high), United States (17, very high), France (26, very high), Greece (32, very high), Turkey (54, very high), Tunisia (95, high), India ( 131, medium), Afghanistan (169, low)