L'incertitude au-delà du risque : essais expérimentaux sur la prise de décision dans l'ambiguïté

Through a series of controlled experiments, this thesis explores how individuals react to ambiguity in different contexts and how their beliefs and attitudes influence decision-making under uncertainty. The first chapter, (In)Consistency of Beliefs and Attitudes under Uncertainty: Every Cloud Has a Silver Lining, is a joint work with Emmanuel Kemel. We investigate whether individuals hold consistent beliefs and attitudes under ambiguity. We develop a unified theoretical and experimental framework that allows for deviations from Expected Utility theory by combining source theory and prospect theory. Using a natural ambiguous source—the voter turnout in the 2019 European Parliament elections in the UK—we find that while typical inconsistencies like the fourfold pattern of risk attitudes and loss aversion persist under ambiguity, the elicited subjective probabilities are well-calibrated when these deviations are properly accounted for. The second chapter, Attitudes Towards Natural Sources of Uncertainty for Gains and Losses, is a joint work with Mohamed El Guide, Yassine Kaouane and Hayat Zouiten. We explore how ambiguity attitudes change when outcomes shift from gains to losses, focusing on natural sources of uncertainty. Employing a novel method that jointly measures ambiguity attitudes and beliefs using matching probabilities, we find that ambiguity attitudes are consistent across outcome domains. Specifically, the duality condition holds, indicating that ambiguity aversion in gains corresponds to ambiguity aversion in losses, while the reflection effect is rejected. These results suggest that ambiguity attitudes are stable across different contexts and robust to the sign of outcomes. The third chapter, Hedging Against Ambiguity? Exploring Randomization through Ambiguity Attitudes and Perception, is a joint work with Elias Bouacida. We investigate whether individuals use randomization as a strategy to hedge against ambiguity and whether this behavior is influenced by ambiguity attitudes or ambiguity perception. We distinguish between subjective and objective ambiguity perception and elicit beliefs within a multiple priors model. Using a within-subject experimental design, we find that participants often choose to randomize across all conditions, even when randomization theoretically should not provide a hedge against ambiguity. Neither ambiguity attitudes nor ambiguity perception significantly influence the decision to randomize; instead, compound risk attitudes may play a role. Given our observations of randomization behavior, an avenue for future research is to further explore this phenomenon using simpler experimental designs to better understand the underlying factors.