Abstract

Stated choice experiments (SCEs) increasingly use probabilistic choices—on a scale from 0 to 100—instead of simple yes/no decisions. The motivation is intuitive: in many hypothetical scenarios, respondents face incomplete information, and expressing choice probabilities allows them to reflect their uncertainty about what they would choose.

In this work, we provide the first nonparametric identification results for what is called the resolvable uncertainty—the part of respondents’ uncertainty that they expect to be resolved when they take the actual decision. We use this result in two ways. First, we examine existing applications, measure the extent of resolvable uncertainty, and test common parametric assumptions—which we reject in most cases. Second, using an application on job preferences among high-skilled students, we show that ignoring respondents’ uncertainty can distort counterfactual policy predictions.

Finally, we revisit a key assumption behind SCEs: that the experiment is ceteris paribus—that is, the distribution of resolvable uncertainty does not change across hypothetical scenarios. Our findings suggest that this assumption may not hold in practice, with important implications for how we interpret stated choice data.

For further information please contact: giovanna.tramontano@unibocconi.it