Thomas Epper (University of St. Gallen, Switzerland)

Abstract:

We study risk taking in the money and the effort domain using an experimental setup with real incentives. In both domains subjects were exposed to triplets of lotteries with equivalent terminal outcome distribution but varying description frame (gain, loss and mixed). We present our results in three parts: In a first part, we propose a series of nonparametric tests that shed light on the description variance and domain-dependence of risk preferences. We examine a number of critical predictions on reference dependence proposed by the literature, such as diminishing sensitivity, reflection and prospect-theory loss aversion. In a second part, we introduce a model-free method to identify rationality types. Our method yields a parsimonious characterization of description (in-)variance in the data. We pose the questions whether individuals' degree of rationality is consistent across outcome domains and whether the documented departures from neoclassic model predictions are quantiatively relevant. In a third part, we calibrate a structural behavioral model and derive an individual-specific index of asset integration. Our comprehensive description of bounded rationality provides novel insights on the descriptive accuracy and external validity of popular theories of choice under risk.