The presentation is based on Riege and Brekke (2019), “Ambiguous additivity neglect: Contrasting two ways of measuring probability assessments”. The presentation will emphasize the results on ambiguity.
Many natural processes are not characterized primarily by risk where we know the outcome is subject to a known probability distribution. We often simply do not know. This distinction between risk and uncertainty dates back to Knight and Keynes. That subject have preference over risk versus uncertainty (in this context usually called ambiguity) and was first demonstrated empirically by Ellsberg using urns with bead of unknown colour. While the theoretical literature on ambiguity has increased vastly, the empirical studies mostly follow Ellsberg, finding rather modest amount of ambiguity aversion. In this paper we rather use natural events, rather than Ellsberg urns. And while the experiment was not designed to quantify ambiguity aversion, our data indicate an ambiguity aversion far beyond what has been observed with Ellsberg urn.
The paper contrasts two different traditions on probability assessments: asking participants for probabilities (psychology) versus deriving probabilities from participant’s preferences (economics), through investigating additivity neglect and ambiguity aversion. In five experiments, participants estimated five mutually exclusive and exhaustive outcomes of a real future event, followed by a choice between betting on either one of the uncertain outcomes just estimated (ambiguous lottery) or on a risky lottery. Participant’s probability estimates displayed additivity neglect adding to 160%-223% across experiments, but participants did not prefer betting on the ambiguous lottery when making choices, even when objective winning chances were much lower than their own estimates. Rather participants made choices as if their assigned probabilities added to only 28%, indicating strong ambiguity aversion. People behave differently in judgments and choices, indicating a “judgment-decision discrepancy” for the two traditions of probability assessments. Participants also showed a tendency to avoid risky lotteries with 50% winning probability.