On the Measurement and Properties of Ambiguity in Probabilistic Expectations

Abstract

Survey respondents’ probabilistic expectations are now widely used in many fields to study risk perceptions, decision-making processes, and behavior. Researchers have developed several methods to account for the fact that the probability of an event may be more ambiguous for some respondents than others, but few prior studies have empirically compared the approaches. This article contrasts two of the most prominent methods using data from an experiment embedded in a recent Web survey of 926 volunteer panelists. Specifically, we comparatively evaluate the descriptive and relational properties of ambiguity scores obtained by placing follow-up questions after items eliciting expectations that ask either for (1) a range of probabilities that the respondent is confident to contain the true probability or for (2) a verbal response indicating assuredness. Our results show that these two methods produce measures that have more similarities than differences. Both methods yield ambiguity scores that (1) are not strongly associated with exposure to sources of relevant information, (2) are correlated across seemingly unrelated events, and (3) are consistently related to the level of reported risk.