What independent random utility representations are equivalent to the IIA assumption?
John K. Dagsvik
Photo: Theory and Decision
Theory and Decision pp. 1-5
This paper discusses random utility representations of the Luce model (Luce, Individual choice behavior: a theoretical analysis, 1959). Earlier works, such as McFadden (Frontier in econometrics, 1973), Yellott (J Math Psychol 15:109–144, 1977), and Strauss (J Math Psychol 20:35–52, 1979) have discussed random utility representations under the assumption that utilities are additively (or multiplicatively) separable in a deterministic and a random part. Under various conditions, they have established that a separable and independent random utility representation exists if and only if the random terms are type III (type I) extreme value distributed. This paper analyzes independent random utility representations without the separability condition and with an infinite universal set of alternatives. Under these assumptions, it turns out that the most general random utility representation of the Luce model is a utility function that is an arbitrary strictly increasing transformation of a separable utility function (additive or multiplicative) with extreme value distributed random terms.