Is preference reversal just stochastic variation?

This paper investigates whether the preference reversal phenomenon can be accommodated by a stochastic model of expected utility. The model is based on Loomes and Sugden's (European Economic Review, 1995) theory of random preference. Its central assumption is that each individual has a set of preference orderings and a probability distribution over that set. Each decision is made according to a preference ordering drawn at random from that set. There are probability distributions over sets of preference orderings, call that preference distributions, that predict the observed asymmetric reversal patterns. These preference distributions are however hard to justify. Moreover they cannot explain the symmetric patterns of reversal that have been observed after repetition and feedback in some experiments (which different, more easily justifiable preference distributions can explain). The model casts doubts on a widely used measure of reversals, and on some conclusions based on that measure, such as the famous observation by Grether and Plott (American Economic Review, 1979) that incentives made preference reversal stronger.