A Comparison of Three Probabilistic Models of Binary Discrete Choice Under Risk

This paper compares the “out-of-context” predictive success of three probabilistic models of binary discrete choice under risk. One of the models is the conventional homoscedastic latent index or “strong utility” model that is widespread in applied econometrics: This model is “context-free” in the sense that its error part is homoscedastic with respect to decision sets. The other two models are also latent index models, but their error part is heteroscedastic with respect to decision sets, and in that sense are “context-dependent” models. Context-dependent models of choice under risk arise from several different theoretical perspectives. Here I consider my own “contextual utility” model (Wilcox 2009) and the “decision field theory” model (Busemeyer and Townsend 1993). A new experiment is performed on 80 subjects. Two-thirds of the data is used to estimate models at the individual level, and these estimates are used to predict the remaining third of choices. The data is divided up so that the decision sets in the estimating data and the prediction data have interestingly different contexts. The context-dependent error models consistently outperform the context-free error model in prediction. JEL Classification Codes: C25, C91, D81