Composite indicators are computational models used to measure and compare the performance of countries, organizations, companies, people, etc. They are reliable tools that have recently gained popularity for their effectiveness in solving multidimensional measurement problems, by providing an aggregation of several dimensions into a single index. One of the issues still open in the literature is the identification of appropriate weights to assign to the different dimensions.
The study titled “Sensitivity-based weighting method for composite indicators” conducted by Chiara Gigliarano and Viet Duong Nguyen (Full Professor and Research Fellow at the LIUC School of Economics and Management, respectively) proposes a new weighting method for composite indices. The proposed solution is called “sensitivity-based weights” and gives each dimension a weight proportional to the variance of the output explained by the corresponding input.
The proposed weighting procedure offers several advantages, such as (i) the weights obtained can be easily understood by policy-makers; (ii) the method demonstrates great compatibility with any correlation structure, including the case of poorly correlated inputs (unlike weighting methods based on correlation analysis, such as PCA and FA).
