The aim of this paper is to generalize the results of Peter and Tijs [10] and [9]. We propose a characterization that holds both for n-agent problems and for a larger class of problems. The axioms used in our characterization are weaker because they are implied by the characterization in the...