This paper examines the three major explanations for the disparity between willingness-to-pay (WTP) and willingness-to-accept (WTA) observed in contingent value surveys and laboratory experiments: a belief that the results must be biased in some fashion, Hanemann's (1991) substitutes hypothesis, and the loss aversion model proposed by Tversky and Kahneman (1991). Starting from the assumption that individuals make utility maximizing choices, we develop structural equations that yield parametric tests of the hypotheses within a single, non-experimental framework. The approach is flexible enough to incorporate a variety of functional form and distributional assumptions and can be applied to either data from either open-ended bids or dichotomous choice questions. The usefulness of the approach is demonstrated using data from a survey that asked both WTP and WTA questions. The results provide weak support for loss aversion.
