Carsharing Usage Patterns and Dynamics in Seoul: Application of Carsharing Bigdata

Abstract

Carsharing services have been introduced as new mobility options in metropolitan areas worldwide. In order for the new mobility option to operate successfully and attract sufficient demand to sustain service, it is imperative to understand where carsharing stations should be located, what factors are important for bolstering demand and mobility patterns of carsharing members. To provide insights for this understanding, this study attempts to identify dynamic usage patterns of carsharing members, factors that influence carsharing operators’ selection of carsharing station locations, and demand based on one-month rental transaction data in Seoul, South Korea. Identification is conducted by three different methodologies: descriptive analysis with linear regression, Heckman Selection model and Multilevel Modelling. By analyzing the data in three different ways, a better understanding of weekly and daily usage patterns and associated influencing factors was created. Comparative analyses are also conducted by segmenting the transaction data into subsets: 1) workday and non-workday rentals and 2) rentals made by age groups of members 3) rentals made during the time of the day. These comparative analyses were necessary for operators to understand usage and mobility patterns of carsharing members along with dynamic nature of carsharing demand in Seoul and thus enabling them to properly design and manage a service. The results revealed the associated factors influencing usage frequency, station location and mobility patterns of carsharing members and untapped various opportunities for carsharing service providers. The estimated models suggest that the supply level of public transit service is positively related to both location selection for carsharing stations and demand. Meanwhile, population density is negatively associated with both selection and demand. A rather weak association between carsharing station location and demand by members in their 40s and older are found, suggesting that carsharing operators mainly target young users when selecting carsharing service locations. Female and old age members are traveling longer on average, specifying that gender and age are a significant determinant of average distance traveled by carsharing members. Further results show that carsharing is used for longer trips in morning time and during non-workdays. Finally, this study also provides evidence that station surrounded by a high percentage of leisure and greater number of subway stations generate rentals which travel longer on average. Based on these findings, estimations can be made about the potential of both opportunities and whether carsharing fits within current mobility patterns or if operators need to change their strategies to incorporate carsharing into the mobility system. The produced outcomes of this study will facilitate the development of a carsharing system in six ways: 1) constructing a theoretical basis when designing two-way carsharing systems in the context of urban configurations and user characteristics, 2) providing insights about how to optimally locate stations or reevaluate the existing sites to maximize user availability, 3) improving system efficiency by understanding the dynamic nature of demand by identifying target users and times, 4) provide a quantitative framework to assess whether carsharing is ensuring sustainable and balance travel habits specifically in terms of travel distance. 5) providing insights about target time of the day and day of the week when carsharing is used for long-distance, and finally 6) helping carsharing operators, local governments, transport planners and environmentalist to develop a better understanding of how to improve the efficiency of the carsharing system and maximize the social, environmental and financial benefits of service. This study may lack generalizability in its findings, but this limitation may be overcome by encompassing temporally and spatially wider data. More detailed information about users such as income and employment status will provide additional knowledge about carsharing behaviors in sample populations.