Preferences for Proxy Attributes

• Published in: Management science Vol. 33; no. 2; pp. 198 - 214
• Main Authors: Fischer, Gregory W, Damodaran, Nirmala, Laskey, Kathryn B, Lincoln, David
• Format: Journal Article
• Language: English
• Published: Linthicum, MD INFORMS 01.02.1987; Institute of Management Sciences; Institute for Operations Research and the Management Sciences
• Series: Management Science
• Subjects: Applied sciences; best guess heuristic; Bias; Decision analysis; Decision theory. Utility theory; Entscheidung; Exact sciences and technology; Expected utility; Heuristic; Heuristics; Modeling; multiattribute utility; Operational research and scientific management; Operational research. Management science; Pollutant emissions; Pollution control costs; preference model; proxy attribute; Proxy reporting; Proxy statements; relative importance heuristic; Tradeoffs; Utilities costs; Utility functions; worst case heuristic; Preference theory; Utility theory; Decision analysis; Multicriteria analysis; Decision theory
• ISSN: 0025-1909; 1526-5501
• DOI: 10.1287/mnsc.33.2.198

A proxy attribute is an indirect measure of an ultimate decision objective. Keeney and Raiffa (Keeney, R. L., H. Raiffa. 1976. Decisions with Multiple Objectives: Preferences and Value Tradeoffs . John Wiley, New York.) argued that assessing utility functions over proxy attributes requires complex inferences that may exceed the human capacity for consistent judgment, thus biasing utility assessments. This hypothesis was tested in an experimental study of preferences for pollution control alternatives. Each decision maker made two sets of utility assessments: the first regarding outcomes described by the fundamental attributes "pollution control cost" and "pollution related illness"; the second regarding outcomes described by the fundamental attribute "pollution control cost" and the proxy attribute "pollution emissions level" (which served as an indirect measure of illness). The subjects displayed a near universal bias to overweight the proxy attribute relative to the prescriptions of expected utility theory. The bias was large, resulting in a substantial loss of expected utility in a simulated policy decision making scenario. To account for this bias, we developed three heuristic models of preferences for proxy attributes: the best guess, worst case, and relative importance models. The results strongly favored the relative importance model, according to which decision makers assess scaling constants by relying on general attitudes regarding the relative importance of different decision objectives rather than on well-articulated preferences for rates of substitution between pairs of attributes.