Journal of Risk and Uncertainty, 6:115-143 (1993) 9 1993 Kluwer Academic Publishers Is Choice the Correct Primitive? On Using Certainty Equivalents and Reference Levels to Predict Choices Among Gambles

Choice is viewed as a derived, not a primitive, concept. Individual gambles are assigned subjective certainty equivalents (CE1); the choice set X has an associated reference level [RL(X)] based on the CEls of its members; the outcomes of each gamble are recoded as deviations from the RL(X); and new CEes are con-structed. The gamble having the largest CE2 is chosen. The CEs are described by the rank- and sign-dependent theory of Luce (1992b). The concept of RL is studied axiomatically. The model predicts many behavioral anomalies and is tested with data sets of Mellers, Chang, Birnbaum, and Ord6fiez (1992). Key words: certainty equivalents, reference l vels, rank- and sign-dependent utility Since World War II, many of us trying to understand ecision making in contexts of risk and uncertainty have assumed that the basic primitive should be choice. Choices are said to reveal our preferences; they are viewed as objective; and we have theorized about them at considerable length. This approach has been followed almost without exception among economists and other mathematically inclined theorists working on such prob-lems. It has been less universal among psychologists--in particular, those of a strong experimental disposition. To them it has seemed equally natural and appropriate to as