This dissertation investigates substantive questions developed from Kahneman and Tversky's behavioral choice theory. Behavioral choice theory postulates systematic departures from economically rational behavior when consumers face choices described incompletely or probabilistically. Previous research relies nearly exclusively on monetary options, which are intrinsically unidimensional and exhibit monotone utility. These special properties are likely to influence the frequency of preference reversals and other so-called non-rational behaviors in human decision-making.
Four contributions emerge from this research. First, I extend the idea of risky choices from monetary to non-monetary options and build a theoretical framework with a foundation in prospect theory and reason-based choice. Second, I test the effect of multidimensional vs. unidimensional non-monetary options on choice focusing on both within- and between-dimensional risk. Third, I examine loss aversion across segments and relate an aggregation fallacy to contradictory results in the literature. Fourth, I suggest an extension of Kahneman and Tversky's behavioral choice theory by incorporating options with missing information.
I use three discrete choice experiments to generate decision schema by segments of individuals sharing similar utility functions. Latent class discrete-choice models isolate the direction and magnitude of value for each attribute (level) of a set of multi-attribute options. They do so in choice domains involving both monetary and non-monetary attributes and operate effectively at both the aggregate and segment levels. As such, they support the rigorous design of experiments that circumvent the need to rely on monetary gambles.
Study 1 investigates the influence of monetary (vs. non-monetary) goals on multidimensional risky choice when full information on reference points is available to an individual. Findings support goal-driven behavior, but reveal only limited evidence to support predictions from prospect theory. Findings also suggest that loss aversion is goal-dependent; i.e., loss aversion is higher for attributes that match consumer goals than for those that do not. Furthermore, the effect is more pronounced for high (vs. low) need for cognitive closure individuals.
Study 2 tests the effect of present (vs. absent) reference points on risky multidimensional choice. Exogenously supplied reference points may help a consumer assess alternatives by offering an explicit demarcation between gains and losses. This baseline may also reduce perceived risk. Results indicate higher partworth utility for levels of an attribute when a reference point is provided for that attribute than when it is not. The study also finds that individuals exhibit varying levels of loss aversion for monetary options but contrary to prospect theory no significant loss aversion for non-monetary options. Similar to study 1, study 2 only partially corroborates expectations from prospect theory.
Study 3 extends the application of prospect theory to choice alternatives characterized by missing information. Missing information means that neither outcome values nor outcome probabilities are known. Results suggest that the information context in which choices are made induces a meta-order on preference judgments: risk ≻ uncertain ≻ absent.
Overall, the results reported here suggest that the literature – citing findings based uniquely on choices between monetary options – may not hold for non-monetary options and that new behavioral theory is needed to explain risky multidimensional choice.