Dominance Axioms and Multivariate Nonexpected Utility Preferences
研究多元分布上的非期望效用偏好，提出两个等价的支配公理，导出局部效用函数的可加可分结构，并证明风险中性不意味着效用函数在非随机结果上是仿射的。
This paper deals with nonexpected utility preferences over multivariate distributions. We present two equivalent dominance axioms, implying an additively separable structure of the local utility functions. They also imply that nonexpected utility functionals directly depend on the marginals of the multivariate distributions. We define an invariance axiom, show that it is equivalent to the property that all local utility functions are ordinally equivalent, and that it implies an additively separable expected utility functional when the dominance axiom is assumed. An interesting property of multivariate preferences is that risk neutrality does not imply affinity of the utility function over nonstochastic outcomes.