On the coherence of composite indexes of well-being: Multiverse analysis for formative models of measurement

Composite indexes are widely used in socio-economic research, yet methodological choices often affect their reliability, leading to variability in results and uncertainty in policy applications. This study applies Multiverse Analysis to systematically assess the robustness of the Italian Equitable and Sustainable Well-being system. 68 alternative specifications for a formative index are fitted. Results show that highly performing provinces exhibit significantly greater variability across the specifications, indicating that index-based assessments may be highly sensitive to methodological assumptions. The choice of aggregation function (e.g., arithmetic vs non-linear means) does not substantially impact uncertainty. However, techniques based on Principal Component Analysis (PCA) inflate the uncertainty in the results, as predicted by authors who question the suitability of PCA for formative measurement. • Composite indexes are alternative to GDP, but there is uncertainty in modelling them. • Multiverse Analysis is expanded for formative models to map model uncertainty. • 68 alternatives are modelled for an index of Italian sustainable well-being framework (BES). • Analysis of variance finds higher variance of performance in highly performing provinces. • BES is found valid for monitoring distress, not as incentive for development.