This paper clarifies three mathematical aspects of random utility models (RUM) of discrete choice which have not generally been appreciated in the literature. First it is shown that the early researchers of this subject all used essentially the same assumptions, although they formulated these in different ways; RUM may be defined by the utility distributions, by the probability functions or by the consumer surplus measure and these are equivalent, in a more general context than has previously been shown. Second, the Invariant RUM, as used by these researchers, is shown to exist in two distinct forms, the more general of which is entirely equivalent to McFadden's Generalised Extreme Value (GEV) model family, which therefore includes all Invariant RUM models. Third, the GEV family can be defined in an alternative and more intuitive form, which confirms that this approach is more general than has been thought and includes many models that are not of closed form. The attractive properties of the GEV family, relating utility distributions, probabilities and surplus measures by simple differentiation and integration, can therefore be applied to all models of the Invariant RUM type.Reprinted with permission from Elsevier. For the covering abstract see ITRD E134766
Abstract