The assumption of independence of irrelevant alternatives in a sequential contingent valuation format should be questioned. Statistically, most valuation studies treat nonindependence as a consequence of unobserved individual effects. Another approach is to consider an inferential process in which any particular choice is part of a general choosing strategy of a survey respondent. A stochastic model is suggested, consistent with the reflexivity, transitivity, and continuity axioms of utility analysis. An application of this theoretical model to the valuation of watershed ecosystem restoration demonstrates that an empirical model recognizing reflexivity and transitivity, and also allowing for continuity, shows the highest in-sample predictive ability.
