Captivity impact on modelling mode choice behavior

This study investigates the effect of accounting for captive riders (i.e. those who do not have a private vehicle available or cannot drive for any reason), in modeling mode choice (MC) behavior for individuals’ trips in Mansoura city, Egypt. To achieve this objective, MC models are calibrated and validated using data collected from Mansoura. The calibration process is carried out in two stages. In the first stage, MC models are built on the choice riders (i.e. those who have access to a private car) data with the choice set fixed for all travelers. In the second stage, MC models are built whereas the private car mode is excluded from the choice set of captive riders. Besides, the travel time comparability ratio is used to compute the travel attributes (i.e. travel time and travel cost) for non-chosen modes instead of computing them from the average trip speed. The developed MC models, for both captive and choice riders, are compared based on their log-likelihood values and their prediction accuracy. The results show that the MC models that consider captivity perform better in predicting the mode choice of travelers with prediction accuracy of about 90%, for different sample sizes. Furthermore, estimating the attributes of the non-chosen alternatives using the travel time comparability ratio is more effective than estimating them from modes' average speeds. Moreover, the value of time (VOT) is estimated for different mode users, and it is found to be 15.55, 8.32, and 6.24 Egyptian Pounds for private car, taxi, and microbus users, respectively.