Modification to the human development index

Abstract

The Human Development Index (HDI) constructed by the United Nations Development Programme (UNDP) is the most popular measure of human development. The HDI has undergone a substantial number of revisions over time, but in its current form, it is a measure made up of four variables: gross national income in constant purchasing parity dollars, life expectancy at birth, mean years of schooling and expectcd years of schooling. The first indicator measures the income facet of human development, the second indicator measures the health dimension and the third and fourth indicators measure the education dimension. Each variable is scaled to take a value between 0 and 1. The geometric mean of the scaled values of the two education indicators is defined as the education index. The geometric mean of the education index and the two other variables is taken as the HDI value for a given country. In 2011, based on the values of the above HDI, 187 countries were ranked and grouped into four categories. In spite of its simplicity, popularity and wide acceptability, a number of authors have noted some major shortcomings of the UNDP's HDI such as: the narrowness of scope, the arbitary nature in which weights are attached to different dimensions, the unreliability of data used in computations and the inability to compare HDI values over time due to frequent revisions to the composition and construction methodology. Upon looking these shortcomings the present study identified 17 new variables in addition to the four variables currently used to develop an alternative HDI to broaden the dimensions of human development measured. Upon carrying out Principal Components Analysis (PCA), these 21 variables were reduced to 13 variables which were deemed to be adequate for the purpose of measuring human development effectively. Thereafter, two alternative indices were constructed by applying PCA and Factor Analysis (FA) to this system of 13 variables. Under PCA, two PCs, the first accounting for 71% and the second accounting for 12% of the variance of the 13 variables were identified. Since the first PC constituted of all the variables with similar weights and the same sign, it was considered as an appropriate index for