Essays on Decision-Making Under Risk

Abstract

This thesis investigates choice behaviour under risk within both expected and nonexpected utility (EU) frameworks. Empirical research has repeatedly uncovered systematic deviations, such as loss aversion and probability distortion, from the predictions of the standard EU model. To address these discrepancies, the thesis employs both theoretical and experimental methods. Chapter 2 introduces a preference foundation to identify a measure for loss aversion by combining elements of reference-dependence, central to prospect theory (Kahneman and Tversky, 1979), with the disappointment aversion approach (Gul, 1991). Prospects are viewed as two-stage objects of choice leading to a pure gains component, a pure losses component, or the reference point. The loss aversion index transforms the basic EU of these components and aggregates these separate measures into an overall value. This framework can incorporate probability weighting without relying on rank-dependence and, nevertheless, satisfies first-order stochastic dominance. Notably, continuity at the reference point implies both loss aversion and probability weighting functions to be linear, and the model collapses to standard EU with a kink. Hence, there is a trade-off between empirical realism and maintaining the technical property of continuous utility at the reference point. Chapter 3 concerns the independence axiom—a cornerstone of EU theory that is frequently violated. It identifies weaker, testable implications of the axiom, such as best-outcome mixture independence and best-worst replacement invariance, and empirically examines them through a laboratory experiment. Results show that 70% of participants conformed to best-outcome mixture independence, while 65% satisfied best-worst replacement invariance; however, only 53% adhered to both simultaneously. The study also finds that the well-known Common Ratio Paradox vanishes under different experimental conditions. Moreover, allowing participants to revise their choices—after being given explanations—reduced violations by 30%, indicating that misunderstanding of the axiom may also explain observed violations.

Date of Award	3 Jul 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	David Delacretaz (Co Supervisor) & Horst Zank (Main Supervisor)