TY - JOUR

T1 - I can board, but I'd rather wait

T2 - Active boarding delay choice behaviour analysis using smart card data in metro systems

AU - Chen, X.

AU - Jiang, Y.

AU - Bláfoss Ingvardson, J.

AU - Luo, X.

AU - Anker Nielsen, O.

PY - 2023/8/31

Y1 - 2023/8/31

N2 - In a crowded metro network, it is not unusual to observe that passengers actively choose not to board but wait for the next train for a seat, even if there is vacant standing room on the arriving train. We analyse such behaviour using a logit-based choice model based on revealed preference data collected from the smart card records and the operational timetables. The choice model considers waiting time, fluctuating crowding levels, and passengers’ expected seat availability at each station on their trip. The revealed preference data are collected based on an existing time component framework, which can estimate passengers’ itineraries by dividing passengers’ travel time into time components (i.e. access, egress, boarding delay, and transfer-walking times) and analysing their uncertainty. We improve the time component framework by developing methods for estimating distributions corresponding to each time component. Using Chengdu Metro as a case, we find that the extra waiting time resulting from active boarding delay and standing time is valued 50.5% more positively and 25.3% more negatively than the in-vehicle sitting time, respectively. By comparing our findings with studies focused on passive boarding delays caused by fully loaded trains, we suggest that extra waiting time due to active and passive boarding delays should be explicitly distinguished in practice. The estimation results of the distributions for the time components indicate that access and egress walking times follow different distributions at given stations, as opposed to the assumption in most prior studies.

AB - In a crowded metro network, it is not unusual to observe that passengers actively choose not to board but wait for the next train for a seat, even if there is vacant standing room on the arriving train. We analyse such behaviour using a logit-based choice model based on revealed preference data collected from the smart card records and the operational timetables. The choice model considers waiting time, fluctuating crowding levels, and passengers’ expected seat availability at each station on their trip. The revealed preference data are collected based on an existing time component framework, which can estimate passengers’ itineraries by dividing passengers’ travel time into time components (i.e. access, egress, boarding delay, and transfer-walking times) and analysing their uncertainty. We improve the time component framework by developing methods for estimating distributions corresponding to each time component. Using Chengdu Metro as a case, we find that the extra waiting time resulting from active boarding delay and standing time is valued 50.5% more positively and 25.3% more negatively than the in-vehicle sitting time, respectively. By comparing our findings with studies focused on passive boarding delays caused by fully loaded trains, we suggest that extra waiting time due to active and passive boarding delays should be explicitly distinguished in practice. The estimation results of the distributions for the time components indicate that access and egress walking times follow different distributions at given stations, as opposed to the assumption in most prior studies.

KW - Public transport

KW - Active boarding delay

KW - Crowding

KW - Itinerary inference

KW - Smart card records

KW - Revealed preference

U2 - 10.1016/j.tra.2023.103747

DO - 10.1016/j.tra.2023.103747

M3 - Journal article

VL - 174

JO - Transportation Research Part A: Policy and Practice

JF - Transportation Research Part A: Policy and Practice

SN - 0965-8564

M1 - 103747

ER -