Georgia Institute of Technology
The teleworking treatment effect on travel behavior: An in-depth exploration of endogenous switching regression models
Abstract
dc:description.abstractThe idea of teleworking (TWing) was proposed in the 1960s as a way of reducing travel, and the number of teleworkers has steadily climbed over the years. In the past three years, the COVID-19 pandemic has boosted teleworking to unprecedented levels. As the pandemic appears to have been brought under control, many “pandemic teleworkers” have chosen to keep teleworking even though this work arrangement is no longer required by their employers and/or the risk of serious illness from the coronavirus has been dramatically reduced. The reshaped teleworker composition (e.g., in terms of sociodemographic traits, occupation, TWing frequency, and TWing motives) will unavoidably lead to travel demand changes (e.g., in net volume, peak-hour distribution, mode distribution). The objectives of this dissertation fall into two categories: policy-oriented and methodology-oriented. The policy-oriented objectives aim to understand TWing-related motives, factors influencing teleworking frequency, and the behavioral outcome of teleworking. The methodology-oriented objectives aim to improve and compare endogenous switching regression models, which we adopt to quantify the impact of teleworking on vehicle-miles driven (VMD). Specifically, to understand “why do people telework?”, we apply a latent class choice model and identify five heterogeneous teleworker segments based on TWing-related motives, namely travel-dominant, flexibility-dominant, career-dominant, workplace-discouraged, and family-dominant. With the same model, we also identify factors that influence the teleworking frequency for each motive segment, which explains “what influences teleworking frequency?” In particular, we find that factors such as gender, education, and job characteristics have heterogeneous impacts on teleworkers with different motives. Next, to deepen the understanding of teleworking-induced behavioral changes on VMD (“what are the teleworking outcomes?”), we adopt three endogenous switching regression models (ESRMs), namely, binary probit switching regression, ordered probit switching regression, and multinomial logit switching regression. Compared to non-teleworkers, we consider the behavioral difference between “non-usual” and “usual” teleworkers. Also, we separate teleworkers based on teleworking-related motives (specifically, travel stressed or not) to compare results when the outcome variable (VMD) likely conforms to the teleworking motivation versus when it does not. Moreover, we expand the application of endogenous switching regression models, such as by allowing more flexible model specifications; developing analytical representations of, and simulation procedures for estimating, the back-log-transformed treatment effect; and improving estimator efficiency by implementing simultaneous maximum likelihood estimation instead of using the prevalent two-stage estimation approach. The results of the dissertation provide a unique perspective on understanding the post-pandemic teleworker segmentation, factors that have heterogeneous impacts on teleworking frequency, and teleworking treatment effects on travel behavior accounting for the influence of TWing-related motivations and frequency. The application, improvement, and comparison of ESRMs will also be relevant to numerous instances of self-selection beyond the present context.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Civil and Environmental Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Wang, Xinyi
- Advisor dc:contributor.advisor
-
- Mokhtarian, Patricia L.
- Committee members dc:contributor.committeemember
-
- Mei, Yajun
- Watkins, Kari
- Pendyala, Ram
- Guhathakurta, Subhrajit
Subjects
dc:subject × 12Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/72741