Identifying commuting behavior segments for TDM program design : university case study

Abstract

Cities and larger employers provide transportation services to diverse users with widely different commuting behavior patterns. Although it may introduce complexities in policy design and implementation to treat different users in various ways, the knowledge of the heterogeneity among them offers us new potentials in optimizing service design and improving user experience. In this research, the case of the Massachusetts Institute of Technology (MIT) has been utilized as an example to explore the potentials of identifying commuting behavior segments and offering actionable policy recommendations. In order to understand the conditions of MIT transportation services, the mid-term impacts of the AccessMIT program are evaluated using the MIT Commuting Surveys conducted in 2014, 2016, and 2018. Then, this research investigates the discrepancy between self-reported commuting diaries and actual commuting behavior utilizing both active and passive mobility data.

Finally, this thesis applies emerging methodologies to segment commuting behavior clusters using a longitudinal representation of multi-year passive mobility data and applies the proposed methodology to a sample of MIT employees. This research reveals three key findings. First, the impact of the AccessMIT program launched by MIT in 2016 has sustained itself and had a positive mid-term impact on changing employees' commuting mode choices and improving their satisfaction rates. Yet this impact varied across different employee groups. For example, the decrease in the single-occupancy vehicle (SOV) mode choices of administration, service, and medical staff happened immediately after the launch of AccessMIT in 2016, but that of faculty happened much slower. Second, the discrepancy between self-reported and actual commuting behavior is not substantial when examining all MIT employees in the aggregate.

However, it varies largely among different groups of employees (e.g., different employee types). Third, the application of the up-to-date clustering methodologies identifies 9 commuting behavior clusters. These 9 clusters carry distinct temporal commuting patterns. For example, aspiring meanderers saw an apparent decrease in the parking frequency while determined riders had a high transit frequency and a very low parking frequency, which have been both steady. Moreover, to offer actionable policy recommendations for next-stage transportation demand management (TDM) at MIT, this thesis supplements the empirical analysis with a comprehensive profiling process using both active and passive mobility data and socio-demographic characteristics..