An experimental study of stochastic decision making

Deterministic decision theories cannot explain the stochastic component of individuals‟ decisions. So specifications of stochastic choices are incorporated into decision theories and utility functions to describe this kind of stochastic decision making. This paper combines three utility functions as well as three specifications of stochastic choice and applies them to experimental data. It is found that exponential utility function with loss aversion parameter embedded in heteroscedastic Fechner errors has the best fit to the data. At the same time by setting two subgroups and allowing for totally different parameters between the two groups, it is argued that unobserved heterogeneity plays a role in determining the estimation results. Furthermore, the author proposed a general form of Fechner error which combines homoscedastic Fechner errors and heteroscedastic Fechner errors in literature. Further studies are still needed to study the new form of Fechner error.