Principal investigator: Bruno Benevit
Original title: Exploring travelers' responses to a prepeak discount fare policy and optimizing the pricing strategy to ease peak congestion: The case of Beijing subway
Authors: Xiangming Yao, Linshan Chen, Peng Zhao, Qingru Zou and Zijia Wang
Location of the Intervention: China
Sample Size: 11 million active passengers
Sector: Transportation Economics
Primary Variable of Interest: Congestion
Type of Intervention: Price discrimination
Methodology: Nonlinear integral programming
Summary
Public transportation services are susceptible to congestion, which typically occurs at the beginning and end of business hours. This situation is particularly frequent in metro systems in metropolitan areas. Therefore, measures to redirect demand for these services, such as adopting time-based pricing differentiation, can mitigate this problem. This study analyzed the impacts of adopting a discount policy for travel before the morning peak hours on the Beijing Metro, China. The results demonstrated that the pre-peak discount fare policy on the Beijing Metro was not effective, indicating that adopting a discount time closer to peak hours would help reduce morning congestion.
- Policy Problem
Peak hours on public transport often lead to congestion at the beginning and end of business hours. This is a recurring problem for most metro systems. Given the limitations of expanding the supply of this type of service in certain types of transport, especially those most dependent on infrastructure expansion such as metro systems, traffic demand management measures can mitigate this situation (YAO et al., 2025).
Individuals react differently when the price of a ticket changes. In general, when the price increases, the number of passengers decreases, but this response varies depending on the context. Some research shows that, in the short term, the reduction in the number of passengers is smaller, while in the long term more people may change their transportation habits. Thus, adopting price differentiation based on time of day allows for a less concentrated distribution of demand throughout the day, adapting to the supply capacity of the transportation system.
Unlike road traffic, where transport demand management (TDM) measures are well established, applying surcharges to public transport presents specific challenges. Surcharges can lead passengers to migrate to other modes of transport, exacerbating congestion in other modes. Therefore, many TDM strategies in public transport systems adopt promotions or incentives instead of fare increases to address this problem.
Beijing's metro, one of China's largest subway systems, also faces this situation, experiencing recurring congestion during peak hours. In this context, transit agencies in the region initiated a discount policy in 2016 to alleviate congestion during the morning rush hour. However, the impact of the policy did not meet the expected results compared to similar policies in other locations (GE et al., 2015). Therefore, understanding traveler behavior at the individual level is fundamental to improving policies of this nature.
- Policy Implementation Context
Beijing's subway system has three lines: BT, CP, and 6. A pre-peak fare discount policy was implemented on the Beijing subway lines BT and CP to try to reduce overcrowding in the mornings, especially between 7 am and 9 am. In 2016, a 30% discount was offered to passengers boarding before 7 am at 16 stations on two lines, but the effects on demand were limited. The following year, the discount was increased to 50%, and eight more stations on line 6 were included in the program.
It is worth noting that the BT Line of the Beijing subway system, the subject of a study in this work, has 13 stations, including two connecting stations. The line experiences significant congestion in the morning heading towards the city center, especially between 7:00 AM and 8:30 AM. Its capacity is limited due to its aging infrastructure, with six-car trains that can accommodate up to 238 passengers per car. The total travel time from end to end of this line is approximately 30 minutes and 35 seconds.
Advances in the use of smart card data have enabled new methodologies for developing optimized fare strategies. This information has helped to consider passenger heterogeneity and the direct and indirect impacts of fare changes on demand for the metro system. The choice of discounts, rather than surcharges during peak hours, was motivated by the fear that higher fares would lead passengers to opt for buses or cars, potentially worsening road congestion.
Despite repeated efforts, demand patterns changed little, and the reduction in overcrowding fell short of expectations. Because the policy did not achieve the desired results, it was not expanded to the entire network, prompting further analysis of its effectiveness.
- Evaluation Details
The study used data from smart cards The data covers the six-month periods before and after the policy's implementation, spanning from June 2016 to June 2017. Records from employee and temporary cards, representing less than 5% of the total, were excluded as they could not be linked to regular users. Additionally, only weekday data was considered, as the policy did not apply on weekends and holidays. Records that did not allow for tracking of movements during the two analyzed periods were also removed. In total, there were over 12,49 million active cards in June 2016 and 13,51 million in June 2017, of which approximately 81% could be tracked.
To better understand passenger travel patterns and their responses to fare changes, subway users were classified into groups based on three main characteristics: (i) intensity of use, (ii) temporal travel behavior, and (iii) spatial patterns. Intensity of use measures how frequently each passenger uses the subway, including the average number of trips per day, the average number of days of use per week, and the stability of that use over time. Passengers who use the subway more frequently and regularly tend to have greater loyalty to the system, while those with more irregular patterns may be less dependent on this mode of transport.
Temporal characteristics were analyzed to identify travel patterns. The time of the first trip of the day was considered an indicator, helping to identify passengers who begin their journeys during peak hours, commonly workers or students. In addition, variations in departure times over time were identified, helping to distinguish users with more predictable routines from those with more flexible schedules. Finally, spatial characteristics were considered based on the OD coverage rate, which indicates the variety of origins and destinations used by a passenger.
The data revealed distinct patterns among Beijing subway passengers. Regular users made frequent and stable trips throughout the week, generally workers or students, while occasional users showed lower frequency and more variable schedules. Regarding temporal patterns, passengers who always start their journeys during peak hours are likely workers, while those with greater schedule flexibility show variation in their commutes. Spatial patterns indicated that passengers with fixed routes, such as home-work or home-school, show little variation in their origins and destinations. Those with more varied routes may be service providers or professionals with multiple workplaces.
- Method
The developed model adopted a non-linear integer quadratic programming to optimize the distribution of demand in the subway, considering the pre-peak fare discount policy. The formulation included variables related to the stations offering discounts, the discount cutoff time, and the percentage applied. The objective function sought to minimize the difference between the observed occupancy rate on the trains and the desired rate, in order to reduce overcrowding. Constraints were incorporated to ensure the consistency of the model, such as the equality of some parameters in certain scenarios and the limitation of the shift in passenger departure times.
The study segmented the analyzed period into 10-minute intervals, allowing for tracking the variation in demand over time. To calibrate the model, smart card data from August 2015, prior to the implementation of the discount policy, were used. This data enabled the identification of origin-destination data and travel patterns. Furthermore, the influence of the discount policy on passenger decisions was incorporated into the model through fare elasticity at departure time, considering the limitation on the maximum time users are willing to anticipate their trip.
Train capacity was modeled taking into account existing infrastructure, including travel time between stations, stopping time, and passenger capacity per carriage. Additionally, the model was tested for five fare discount scenarios before peak hours. Each scenario varied in terms of the station with the discount, the cutoff time, and the discount percentage, ranging from a scenario with no discount to one with complete flexibility in discounts and cutoff time.
- Main results
The model results indicated that the current discount scenario does not have a significant effect on reducing crowding, a result consistent with that observed in the Beijing subway. A comparison between the scenarios revealed that changing the discount rate has a limited effect, while changes to the discount cutoff time generate more significant impacts.
In the scenario where the cut-off time was adjusted, the reduction in overcrowding was significantly greater than in the scenarios that only increased the discount percentage. The most flexible scenario showed the best performance in reducing overcrowding, although the authors highlight the possibility of compromising passenger convenience due to different schedules and rates between stations.
The fare elasticity analysis showed that the discount can shift some passengers from peak hours to earlier times. However, this effect is limited, as most passengers do not anticipate their trip by more than 30 minutes. Given that the actual peak hour occurs around 8 am, the impact of the implemented policy is insignificant.
As a consequence of these results, the authors highlighted that systems like the Beijing subway, where demand is high and concentrated in a short period of time, make it difficult to modify passenger behavior through fare incentives. The Beijing subway's fare policy already offers a significant discount compared to other systems, such as those in Hong Kong, New York, and Washington, which apply smaller discounts outside of peak hours. Furthermore, the Beijing government subsidizes approximately 50% of the subway's operating costs, allowing for greater flexibility in fare setting.
To improve the efficiency of the discount policy, it would be necessary to adjust the cutoff time for the benefit so that it coincides with the actual distribution of demand. The experience of the Melbourne Metro, which implemented a free ticket for travel before peak hours, showed more satisfactory results in redistributing demand. However, this strategy involves financial challenges, as the loss of revenue can be substantial.
- Lessons in Public Policy
In this article, the authors analyzed the impacts of adopting a fare discount policy for trips taken before the morning peak hours on the Beijing subway using a nonlinear programming model. The model results indicate that the policy had a limited impact on reducing demand during peak hours, consistent with the behavior observed during the policy's implementation. The effectiveness of the measure was restricted by the discount's time limit, which did not cover most of the peak demand period. The results also suggest that the magnitude of the discount has a limited influence on passengers' decisions.
The evidence presented in this article provided fundamental insights into the relationship between fare incentives and the redistribution of demand in public transportation, aiding in the development of strategies to mitigate overcrowding. The authors highlighted that adjustments to the discount's cutoff time and expanding the time frame covered by the policy can increase its effectiveness.
References
GE, Y.-E. et al. SOLVING TRAFFIC CONGESTION FROM THE DEMAND SIDE. PROMET – Traffic&Transportation, v. 27, no. 6, p. 529–538, 21 Dec. 2015.
YAO, X. et al. Exploring travelers' responses to a prepeak discount fare policy and optimizing the pricing strategy to ease peak congestion: The case of Beijing subway. Transportation Research Part A: Policy and Practice, vol. 191, p. 104335, Jan. 2025.
