Standard cost-benefit analysis is inadequate for large scale social risks (natural disasters, technological accidents, terrorism) because it rests on the unrealistic assumption of perfect risk sharing. As a consequence, it focuses only on the aggregate effects of the risk, while it neglects both the distributional consequences (who suffers more?) and the interdependencies of individual risks (how many people will suffer?).

This paper proposes a new approach to address those limitations, the equity-enhanced mean variance rule. The rule is derived by extending the classic Arrow-Pratt approximation of the certainty equivalent to a framework that accounts for heterogeneity and distributive justice. Two approaches to equity are compared and implemented: the first expresses concerns for the heterogeneity of individual risks (ex-ante equity); the second represents concerns for the collective realization of individual risks and resulting inequalities or correlated effects (ex-post equity). The equity-enhanced mean variance rule will depend on the type of equity concerns, and on few, easily computable parameters: risk and inequality indices, Society's inequality aversion and individuals' average risk aversion. As an application, I discuss the equity-efficiency tradeoff associated to the National Flood Insurance Program in the United States.