Combining Response times and Choice Data Using a Neuroeconomic Model of the Decision Process Improves Out-of-sample Predictions

A basic problem in empirical economics involves using data from one domain to make out-of-sample predictions for a different, but related environment. When the choice data are binary, a canonical method for making these types of predictions is the logistic choice model. We investigate if it is possible to improve out-of-sample predictions by changing two aspects of the canonical approach: 1) Using response times in addition to the choice data, and 2) Combining them using a model from the neuroeconomics literature, called the Drift-Diffusion Model (DDM). We compare the out-of-sample prediction accuracies of both methods using both simulations and real experimental data. In both cases we find that the DDM method outperforms the Logit prediction method. Furthermore, although the improvement in prediction accuracy is small for the case in which items have very similar or very different values, the DDM method improves prediction accuracy substantially for intermediate cases.