Subjective–Objective Method of Maximizing the Average Variance Extracted From Sub-indicators in Composite Indicators

Abstract

This research presents an innovative method for constructing composite indicators: the Subjective–objective method of maximizing extracted variance (Sommev). Sommev’s hybrid weighting approach fills an important gap within a highly controversial area of the composite indicators’ literature, which criticizes the statistical assignment of weights disconnected from theory and the errors and judgmental biases inherent in the expert opinion-based weighting approach. These innovations contribute to a more coherent and consistent operationalization of the theoretical framework of multidimensional phenomena, reconciling the non-compensability between sub-indicators and the maximum retention of original information through statistically defined weights, in which the expert’s opinion is considered, but does not determine the sub-indicator’s weights. Twenty simulations were carried out to analyze the application of the method in representing social exclusion in a Brazilian city. Composite indicators constructed by Sommev retain twice as much information as those constructed with equal weights or weights defined by experts. This increased informational capacity favors a more comprehensive representation of the multidimensional phenomenon, having a high potential for application in solving problems of a multidimensional nature in the social, economic, and environmental areas.