A Generalized Continuous-Multinomial Response Model with a t-distributed Error Kernel

摘要: In multinomial response models, idiosyncratic variations in the indirect utility are generally modeled using Gumbel or normal distributions. This study makes a strong case to substitute these thin-tailed distributions with t-distribution. First, we demonstrate that model t-distributed error kernel better estimates and predicts preferences, especially class-imbalanced datasets. Our proposed specification also implicitly accounts for decision-uncertainty behavior, i.e. degree of certainty decision-makers hold their choices relative variation any alternative. Second, after applying first time, extend this generalized continuous-multinomial (GCM) derive its full-information maximum likelihood estimator. The involves an open-form expression cumulative density function multivariate t-distribution, which propose compute combination composite marginal method separation-of-variables approach. Third, establish finite sample properties GCM (GCM-t) highlight superiority over normally-distributed (GCM-N) Monte Carlo study. Finally, compare GCM-t GCM-N empirical setting related preferences electric vehicles (EVs). We observe accounting behavior results lower elasticity higher willingness pay improving EV attributes than those model. These differences relevant making policies expedite adoption EVs.