Assessing the utility of multiattribute utility assessments

Abstract The usefulness of multiattribute utility (MAU) analysis was assessed by comparing the sensitivity to both relevant and irrelevant information of analytically derived judgments and of holistic judgments. Subjects evaluated four hypothetical apartments, described to them in a series of letters, by indicating how much rent they would be willing to pay for each apartment. In addition, they completed a survey that requested judgments about the six attributes that characterized the apartments. From the survey, evaluations of the apartments were derived from an additive MAU model. The holistic rental judgments exhibited little sensitivity to differences among the apartments. However, they did show a “linearity effect,” an ordering of the apartments that apparently reflected an oversimplified interpretation of the information. Derived judgments showed no linearity effect, and did show a sensitivity to apartment differences. There was some evidence that derived judgments were less sensitive if the MAU judgments were made after making the holistic judgments. Holistic and derived judgments were generally uncorrelated, presumably because of the heavy information processing demands imposed by the holistic judgment task.