Improving Parameter Estimates and Model Prediction by Aggregate Customization in Choice Experiments

Abstract

We propose aggregate customization as an approach to improve individual estimates using a hierarchical Bayes choice model. Our approach involves the use of prior estimates to build a common design customized for the average respondent. We conduct two simulation studies to investigate conditions that are most conducive to aggregate customization. The simulations are validated by a field study showing that aggregate customization results in better estimates of individual parameters and more accurate predictions of individuals' choices. The proposed approach is easy to use, and a simulation study can assess the expected benefit from aggregate customization prior to its implementation.