Preference Reversals

Choice between Gambles

Expected utility theories of decision making assume that if individuals have sufficient information and are given sufficient time to process that information, they will make choices that maximize their interests. Since the choices people make reflect their best interests, those choices should be stable irrespective of how the preferences are elicited. For example, if given a choice between Options A and B, the agent prefers A, the agent should, when asked how much he or she would be willing to pay for the two options, pay more for Option A than for Option B. However, a great many studies have shown that the method of eliciting a preference can affect the choices people make. In the classic form, experimental participants are asked to choose between pairs of gambles. One choice has a high probability (P Bet) of winning a modest sum of money; the other has a low probability of winning a large sum of money. In the example below, the expected value (the probability of the outcome multiplied by the value of the outcome) of each choice is the same:

When asked to choose between the two gambles most people prefer the P Bet. However, when asked to state the lowest amount of money they would sell the gambles for, or how much the gambles are worth to them, people tend to assign the $ Bet a higher monetary value. Thus, preferences elicited as a choice are reversed when those preferences are elicited in another form and violate expected utility theory's axiom of invariance.

Examples

The effect is not restricted to monetary decisions and occurs when the same individuals make the same choice twice and when the choices made by different individuals are elicited in different ways. For example, in a study of personnel selection, participants were asked to imagine that they were company executives choosing between two candidates, who had been scored by a selection committee along two dimensions (technical knowledge and human relations), for a position as a production engineer. One group of participants made their choices by choosing directly between the two candidates or by a matching technique. A second group received the same information, except that one of the four personnel scores was missing and had to be filled in to make the two candidates equivalent. The choices made directly tended to be reversed in the group whose choices were inferred by their evaluation of the missing attribute.

Preference reversals are not limited to situations in which preferences are elicited using direct choices and evaluations. They occur also when evaluations are made along different dimensions. For example, college students were asked to evaluate various hypothetical consumer- and health-related scenarios in terms of monetary value and life expectancy. The participants were asked to state how much they would be willing to sacrifice financially and in terms of life expectancy in order to gain an AIDS vaccine, immunity to tooth decay, a treatment that gives 20/20 vision, and one that provides immunity to cancer. Consumer items included airline tickets, theater tickets, vacations, and a date with a favorite celebrity. Although it might seem odd to compare different kinds of decisions along these two dimensions, the preferences elicited by them should be rank ordered in the same way on both scales, since any valuation scale reflects subjective utility. That is, the rank order of preferences on the monetary scale should be the same as the rank order of items along the life expectancy scale. The results, however, showed that health items were ranked higher than consumer items in terms of life expectancy value but [Page 899]the consumer items were ranked higher on the monetary scale than the health items. This result indicates that the features of a decision (e.g., health benefits) are given more weight when they are meaningfully related to the scale on which they are evaluated (e.g., life expectancy).

...