A Mode Choice Model Separating Taste Variation and Stated Preference Reporting Error

Abstract

Because of survey efficiency and ease of setting travel level-of-service, SP data has been expected to be suitable for predicting the furture demand of new travel modes. However, it contains more biases than RP data caused by setting ypothetical situations in the survey. Especially, it has been pointed out that the predictingmodels based on SP data overestimate the furture demand for new travel modes. Therefore, this paper proposed to separate alternative-specific constant term into individual-specific taste term and unobserved SP reporting error term, which has not been considered clearly in conventional travel prediction models. Furthermore, it was also tried to consider individual taste variation by Mass Point approach and to remove Panel Conditioning bias by introducing previous RP choice results into SP model. Throughout the empirical analysis based on SP panel survey data on the use of New Transit System in Hiroshima city, it was shown that the proposed model provided a higher goodness-of-fit index and temporal transferability than the conventional ones.