Similarity judgments in choice under uncertainty: A reinterpretation of the predictions of regret theory

The Regret theory of Loomes and Sugden (1982) predicts choice anomalies implied by other alternatives to expected utility (e.g., violations of the independence axiom). It also makes unique and controversial predictions regarding the rational violation of stochastic dominance and invariance. All these predictions depend critically on assumptions regarding the statistical independence or dependence of the available alternatives. None of the predictions depend on the framing or representation of the alternatives. Leland (1994) shows that a model of choice based on similarity judgments predicts choices implied by Regret theory. In contrast to Regret theory, however, these predictions depend critically on the way the choices are framed and not on the dependence or independence of the alternatives. This paper presents experimental results indicating that the frequencies with which violations of independence, dominance, and invariance occur are, in most cases, insensitive to the statistical dependence or independence of the alternatives but sensitive to the way the choices are presented. These findings support the hypothesis that such behaviors arise as a consequence of reliance upon similarity judgments.