Working Papers
Expected Utility for the Lab (with Horst Zank), [Link to paper]
Abstract: A large literature has documented choices that are inconsistent with the standard assumption that agents maximize their expected utility (EU). Central to this issue is the independence axiom, which EU maximisers must satisfy but is often violated in practice. We characterise EU for lotteries with three outcomes, a setting that applies to many empirical tests of the independence axiom. For a weak order, our preference requirements are implications of the independence axiom, namely, monotonicity with respect to first-order stochastic dominance, best-outcome mixture independence, and best-worst replacement invariance. Our laboratory experiment shows that when the independence axiom is decomposed into these weaker properties, violations are less frequent: 70% of participants adhered to best-outcome mixture independence, and 65% to best-worst replacement invariance. This suggests that the commonly observed violations of the EU assumption could be due to axioms other than independence, such as continuity or transitivity. When participants were given the opportunity to revise their choices, they corrected an average of 30% of initial violations—significantly higher than the 19% of initial adherences that were subsequently reversed. This suggests that violations may stem more from misunderstandings than from genuine preferences against these properties. Additionally, we find that the Common Ratio Paradox—a well-known example of violations of the independence axiom—is less pronounced under a different experimental condition, highlighting the importance of careful experimental design in understanding decision-making behaviour under risk.
Loss Aversion: A Possibility-Impossibility Result (with Horst Zank), draft available upon request.
Abstract: A foundation for loss aversion is proposed. We adopt a framework that combines advances from reference-dependence as in prospect theory (Kahneman and Tversky 1979) with the treatment of prospects as in disappointment aversion (Gul 1991). In the theory of loss aversion (TLA), prospects are seen primarily as leading to a pure gains component, a pure losses component or the reference point. The index of loss aversion transforms the expected utility of the pure components and aggregates these separate measures into an overall value for a prospect. For the domain of simple prospects that offer at most one gain or one loss, where many empirical regularities have been observed, the model agrees with the cumulative version of prospect theory. The general version of TLA invokes weighting functions for the overall probability of gaining, respectively, the overall probability of losing, and identifies the loss aversion function as a transformation of a basic utility for risky outcomes. This provides an extension of original prospect theory for preferences that satisfy first-order stochastic dominance and continuity in probabilities. In contrast to disappointment aversion or cumulative prospect theory, continuity in outcomes imposes significant constraints on the TLA-model: both loss aversion and probability weighting vanish, such that preferences must agree with expected utility.
Work in Progress
Loss Aversion with Shifting Reference Points.