Inconsistent Trade-Offs Between Attributes: New Evidence in Preference Assessment Biases

One of the fundamental postulates of rational choice is that preferences manifested by an individual towards alternatives should only depend on the merits of these alternatives and not on extraneous, irrelevant factors. Violations of this basic principle, so-called preference reversals, have puzzled researchers for over twenty years and raised concerns about the use of preference modeling in decision analysis. The present work seeks to further determine the nature of these phenomena, in particular the role played by response mode in certain types of preference reversals. Hershey and Schoemaker (1985) found Probability and Certainty Equivalents to differ systematically and attributed this difference to a framing effect. Here, we generalize their experimental design to control for framing effects and study biases on a larger scope. Our results show that biases do not disappear in the absence of framing, instead they reveal a clear and pervasive bias occurring under more controlled experimental conditions than previously known: direct trade-offs between two attributes X and Y are biased depending on whether X is traded off against Y, or Y traded off against X. From among several hypotheses, the data lend support to the general principle of compatibility (Tversky et al. 1988; Slovic et al. 1990), which implies that an attribute receives more relative weight when it is used as "currency" in trading off.