Abstract: In the standard literature for modeling heterogeneity, it is commonly assumed that the distribution of types in a population is invariant across menus. This assumption does not accommodate behavioral patterns such as context dependence. In this paper, I run an experiment to test whether the distribution of risk attitudes changes depending on the construction of the menu. Specifically, when an unattractive certain option is added to a menu, does the average decision maker exhibit higher levels of risk aversion? In other words, can the certain option be used as a decoy to switch decision makers away from a high-risk option? My pilot study shows a statistically significant relation between the presence of a certain option and a decrease in the likelihood of a high-risk option being chosen. What's more, I find violations to core axioms that characterize random utility in standard literature, which evidences the need for menu dependence to model heterogeneity.

Abstract: Constant threshold representations (CTRs) describe agents with limited perception who cannot distinguish small differences in utility. Unlike pure utility maximization, this model can be difficult to fit into empirical or simulated data sets as limited perception makes it impossible to observe any differences between a truly-best element in a menu and slightly worse alternatives. I implemented a novel algorithmic approach that attempts to fit CTR into general data sets. While I do not provide any theoretical complexity guarantees, my initial findings suggest that my algorithm works reasonably fast in practical settings. The algorithm is twofold, with the first part that efficiently suggests a utility ranking to complete binary relations and the second part that constructs a utility function based on binary relations built from the first step. I plan to extend this research for other structures, like true multi-utility models where choices are path independent.