Random quasi-linear utility (Q6105385)

In this paper the authors present an extension of the random utility model (RUM); they call it random quasi-linear utility model (RQUM). Section 1 contains an elaborate description of the results. After preliminaries Section 3 contains the definition of RQUM. Preferences with a quasilinear utility representation are considered. In RUM a stochastic choice rule is described by a Borel probability measure defined by Equations (7) or (8); in RQUM the Borel measure is applied to a set of preferences defined by `low maxima' (9). This Borel probability measure is called a low representation of the stochastic choice rule. RQUM accommodates any Borel probability measure; `utility ties are broken down by a convenient lexicographic tie-breaking rule'. Section 4 contains the main results. Theorem 1 shows that a stochastic choice rule has a low representation if and only if 5 axioms are fulfilled, two of which are problem orientated (no complementarity, cross-price neutrality). Theorem 3 (surprisingly there is no Theorem 2) contains results for revealed stochastic quasi-linearity or revealed stochastic preferences. In theorem 4 RQUM is applied for finite data sets. Section 5 has miscellaneous results; especially it is shown, that the axioms in Theorem 1 are logically independent. Section 6 contains the proofs. The authors emphasize that the low representation has `a unique and explicit identification of the observable stochastic choice rule'.