WP 2025-03 Mixtures of Risk Preferences with a Bayesian Hierarchical Model

AUTHORS: Xiaoxue Sherry Gao, Glenn W. Harrison and Brian Monroe

ABSTRACT: In the quest to better characterize the heterogeneity of risk preferences at the level of the individual agent there are two econometric methods that can be combined. One uses mixture models to allow for individuals to use some combination of two or more popular models of risk preferences. The other uses hierarchical Bayesian models to generate informative priors for individual estimation from the behavior of similar subjects from the same population. We illustrate the joint implementation of these approaches with several data sets from experimental economics, and show that there is clear evidence that individual subjects use a mixture of two models of risk preferences, rather than each subject making all choices consistently with one or other of the model. The extent to which the two models are combined varies across individuals, with implications for descriptive and normative inferences from estimated risk preferences.