Random Scale Heterogeneity in Discrete Choice Models

Methods to model preference heterogeneity in discrete choice models continue to receive a significant amount of attention in the literature. The main interest lies in modeling variations in relative sensitivities across respondents, e.g. leading to different valuations of travel time savings (VTTS). Recently however, the question as to what extent the heterogeneity captured by random parameter models is not exclusively preference heterogeneity, but scale heterogeneity, has also attracted attention. In this paper, the authors present a conceptually straightforward modeling framework that jointly allows for random scale heterogeneity and heterogeneity in relative sensitivities. The authors results on two separate datasets show how this leads not only to significant improvements in model fit, but also helps to disentangle the two types of heterogeneity, with obvious benefits for interpretation. Finally, the two case studies also highlight how not accounting for scale heterogeneity may lead to counter-intuitive findings in terms of inter-coefficient correlation.