Social Shock Sharing and Stochastic Dominance

Since the seminal paper of Atkinson and Bourguignon (1982), little decisive progress has been achieved in developing empirically efficient stochastic dominance criteria for multidimensional social welfare analysis. By proposing new axioms of 'Social Shock Sharing', this paper provides new intuitive justifications to imposing sign restrictions on partial derivatives of individual von Neumann-Morgenstern utility functions. These new breakthrough findings are exploited to derive necessary and sufficient stochastic dominance criteria for multidimensional social welfare comparisons, up to the sixth order, at least. Equivalent results are derived in terms of multidimensional poverty conditions. Empirically powerful discriminatory criteria are obtained by combining all social shock sharing axioms up to some high order and by deriving a dimension reduction property. An application to Egypt at the beginning of the XXIst century demonstrates the practical substantial gain in discriminating power of the approach by revealing a unambiguous continual improvement in bivariate income-education social welfare over the studied period.