Evaluation of Travel Time Reliability using “Revealed Preference” Data & Bayesian Posterior Analysis

EasyChair Preprint 979

Abstract

In Swedish context, the value of delay is deemed equalling to the value of travel time reliability (VoR), which mostly derived from Stated Preference (SP) studies. According to our knowledge, there are several issues with the SP method for obtaining VoR, for example, its deficiency in harmonizing the stated choices with the actual choices. On the other hand, Revealed Preference (RP) data from ticket sales has its limit in socioeconomic information of travellers and scenario variation. This project aimed to use a RP method to estimate VoR from several selected railway corridors, with Bayesian posterior analysis to infer socioeconomic differences between passengers given on their actual choices.

The data in the study are from two sources, ticket sales data from a major passenger operator SJ, and data on train movements from Trafikverket’s (Swedish Transportation administration) TFÖR database. Both data sources are for the whole year 2009. The data includes 60 545 individual observations on traveler’s route choice for two specific trip relations. The chosen trip relations are long-distance non-commuting trips with travel distances between 200 and 250 kilometers.

The project is a “proof-of-concept” for possible use of ticket sales data for the estimation of VoR. We can conclude that the estimated VoR – 1,13 times value of travel time, is in compliance with results from previous international studies using SP and/or RP data. The simulated distribution of VoR from posterior analysis also clearly indicates a bimodal pattern of valuing travel reliability, probably due to socio-economic characteristics or trip purposes.

Keyphrases: Bayesian model, Mixed Logit, Value of reliability (VoR), posterior analysis, revealed preference (RP)

@booklet{EasyChair:979,
  author    = {Sida Jiang and Christer Persson and Karin Brundell-Freij},
  title     = {Evaluation of Travel Time Reliability using “Revealed Preference” Data & Bayesian Posterior Analysis},
  howpublished = {EasyChair Preprint 979},
  year      = {EasyChair, 2019}}