Global Disruptions in the Transportation Sector: The Effect of Ridehailing Services and the COVID-19 Pandemic

Enabled by information and communication technologies (ICT) and based on the principles of the shared/gig economy, ridehailing services (e.g., Uber, DiDi, Ola) are transforming the travel patterns and the lifestyle patterns of people around the world. Other new technologies based on the same principles, including smartphone app connectivity and access to mobility services for multiple users on an on-demand basis (e.g., food delivery services, micromobility), can potentially have similar transformative impacts. It is thus important to study these services and their impacts carefully to leverage these technologies in creating a more inclusive, sustainable, and resilient transportation system. \tIn the last two years, the COVID-19 pandemic has created an even bigger disruption to the transportation sector (including life in general). While there is a general consensus among the research community that the pandemic is fundamentally transforming the transportation sector, the post-pandemic future of the transportation system remains to be seen. What social and transportation inequities occurred during the disruption of the pandemic? And how can the post-pandemic transportation sector be shaped to become more inclusive, resilient and sustainable? \nThe goal of my dissertation is to create a deeper understanding of these two major disruptions in the transportation sector. I do this by analyzing survey data collected in cities of the United States, Canada, Germany, Chile, Mexico, Brazil, India and China, in combination with socio-demographic and geospatial datasets available in these countries. \nIn the first two studies (Chapter 2-3) I focus on California. First, I explore the factors that affect the use of ridehailing services (Uber, Lyft) as well as adoption of shared (pooled) ridehailing (UberPOOL, Lyft Share) by estimating a semi-ordered bivariate probit model. The model reveals the similarities and differences in the markets for each of the two services. Among the main findings from the study, I find that being white and living in a higher-income household is associated with a higher likelihood of being a frequent user of non-shared ridehailing but does not have statistically significant effects on the likelihood of adopting shared ridehailing. While the likelihood of using both non-shared ridehailing and shared ridehailing is higher in urban areas, residents of neighborhoods with higher intersection density are found to be more likely to only adopt shared ridehailing.\nNext, I estimate an integrated choice and latent variable (ICLV) model to develop an in-depth understanding about the effect of the built-environment on ridehailing use for non-work purposes while accounting for confounding effects such as the preference to own a vehicle and to live in urban locations. My analysis confirms that failure to consider the latent preferences for residential location can lead to biased results. This analysis results in two major findings: 1. individuals living in vibrant and walkable neighborhoods are more likely to replace other modes (possibly active modes) with ridehailing, 2. previous studies may have misestimated the relationships between public transit and ridehailing by ignoring confounding effects. \nIn the following two studies, I move to an international perspective on the impacts of these services and other disruptions in the transportation sector. First, I focus on the adoption of ridehailing in developing countries. To do this, I compile survey datasets from Mexico City, Sao Paulo, Beijing, and Mumbai and estimate a binary logit model of the adoption of ridehailing with discrete segmentation for each country. My analysis shows that younger respondents are more likely to adopt these services in all locations. A number of other factors are found to have significant effects only in selected regions. Among other findings, in Mumbai, respondents who live in zero-vehicle households are more likely to use ridehailing, probably as an effect of social status, while this is not true in the other regions. Women are more likely to use ridehailing than men in Sao Paulo and Beijing, and this effect is significantly stronger in Mumbai. However, in Mexico City, an opposite effect was observed, i.e., men are more likely to use these services than women.\nIn the final study, I focus on the disruption that the COVID-19 pandemic brought to society starting in early 2020. To do this, I focus on one of the major components of disruption the pandemic has caused, the heavy shift to telecommuting. I aim to understand the influence of household and individual socio-demographic characteristics on two related dependent variables: the decision to exclusively telecommute and the frequency of physical commute to work (if not exclusively telecommuting) during the first wave of the pandemic in Canada, Chile, Germany and the U.S. I jointly model the two decisions while accounting for confounding effects, including those associated with different recruiting and sampling methods for each country and unobserved country-specific attributes (e.g., COVID-19 response). In all countries, affluent workers (i.e., high-income, high-educated, or non-essential-workers) are found to have a higher propensity to exclusively telecommute and to report to work at a lower frequency if commuting physically. I also uncover that the effects of a few selected sociodemographic characteristics differ greatly by country, including household size, full/part-time worker status, gender, and vehicle ownership. This study contributes to the academic literature by comparing how the response to the global COVID-19 pandemic (in terms of telecommuting behavior) depended on the local context. Finally, the last two studies converge on one finding from my dissertation – context matters while studying individual behaviors, and it is not always easy to generalize findings and transfer the lessons learned from one location to others. \n

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.