Assessment of user benefits in presence of random taste heterogeneity: Comparison of parametric and nonparametric models

The ultimate objective of most practical transport modeling is the evaluation of the economic impacts of alternative policy measures. During the past decade, significant advances have been made in the ability of operational models to accommodate flexible patterns of taste heterogeneity. The majority of applications use parametric distributions in which the parameters of the distribution are specified a priori. However, in reality individual tastes rarely follow a standard distribution, especially in the extremes of the distribution in which behavior is likely to differ substantially from that of the rest of the population. The major problem in these cases is the suitability of estimated models for determining economic welfare because they may yield erroneous or counterintuitive results. Recent research proposes using nonparametric models to capture randomness in individuals' tastes, in which the shape of the unknown distribution is defined as part of the estimation process. But there is still no evidence regarding the implication of using these more advanced model structures for determining economic welfare. The aim of this paper is to investigate to what extent imposing a well-shaped distribution of individual tastes affects the appraisal of transport policy measures. In particular the focus is on random taste heterogeneity. With the use of simulated data, an investigation is done on how parametric and non-parametric models perform under different assumptions for true taste distributions. Finally, implications of the accuracy of alternative methods of computing benefit measures are critically analyzed.