Multiattribute risk linearity

In applying multiattribute utility theory, one typically assumes the condition of mutual utility independence and uses a multiattribute utility function that is additive or multiplicative. This paper proposes a model in which this condition is not satisfied, that is, the risk attitude for one attribute depends on the amounts of the other attributes. It introduces a condition on this dependence called ‘multiattribute risk linearity’ that implies a logarithmic form for the multiattribute utility function, and then shows that a logarithmic utility function can be determined by assessment procedures that require the same number of indifference assessments as those for a multiplicative utility function.