Uncertainty Analysis Applied to the Representation of Multidimensional Social Phenomena

Abstract

Abstract The representation of social phenomena has been associated with substantial conceptual and operational difficulties. Social phenomena have a complex multidimensional nature that leads to different conceptualizations and measurements. This problem makes it difficult to choose the subindicators to be considered in the composite indicator of the phenomenon. In addition, subindicators can be normalized, weighted, and aggregated in different ways. There is no answer in the literature about which combination of methods is most appropriate to represent a given phenomenon. This research aims to answer which normalization and aggregation methods combination offers the best representation of the social exclusion process considering its multidimensionality. Fifteen subindicators of social exclusion were aggregated and normalized by different methods, generating thirty-one composite indicators. Three criteria measured the performance of the composite indicators: external validity with the average household income indicator, the actual conditions of the environment observed by the urban landscape analysis, and the prediction errors of the composite indicator. The results show significant differences in the capacity of a composite indicator to represent situations of social exclusion. It is possible, however, to represent social exclusion more consistently from subindicators normalized by the min–max technique and aggregated by the geometric mean.