Arrow's theorem on the optimality of deductibles: A stochastic dominance approach

Summary We provide a new proof for the optimality of deductible insurance that does not depend on the expected-utility hypothesis. Our model uses only first- and second-degree stochastic dominance arguments.This paper was partially written while Schlesinger was visiting at the University of Toulouse. Financial support for this visit from the Fédération Francaise des Société d' Assurance is gratefully acknowledged. The authors also thank Louis Eeckhoudt, Ed Schlee and an anonymous referee for helpful comments.