Passenger Vehicle Sharing: Nodes, Links, Opportunities, and Relevance

Passenger Vehicle Sharing, also known as carpooling, consistently garners significant attention due to its association with social benefits. Increasing vehicle sharing is, to some degree, a cost-effective way to enhance transportation capacity. Understanding the factors that influence private passenger vehicle sharing can help inform local traffic patterns and planning decisions. This dissertation’s first contribution is to demonstrate the regional factors that influence one specific type of vehicle sharing: demand for park and ride facilities. A key finding of the first paper presented herein is that while one can successfully model such demand based on a variety of traffic, demographic, and land use variables, the findings were somewhat hampered by a lack of detailed data.

To overcome this lack of data, a new metric is considered. Traditionally, vehicle sharing has been measured by the percentage of vehicles with multiple occupants, but an alternative metric—link-based vehicle occupancy—can also be used to quantify this sharing. Because passenger throughput can be estimated as the number of vehicles moving on a link per hour multiplied by vehicle occupancy, the metric of vehicle occupancy aligns with the paradigm of moving people in addition to moving vehicles. However, occupancy data are not widespread. Traditionally, occupancy has been obtained through self-reported survey data or manual observations, and as of 2023, affordable, technology-based solutions, while offering promise, are not yet widely implementable on a cost-effective and reliable basis. In response, this dissertation explores the use of an appealing, but imperfect, source of vehicle occupancy: crash data.

This data source can be used for generating occupancy at various geographical scales, from as small as a Census block group to statewide. Because of the twin challenges of small sample sizes (for small geographical units) and potential bias (owing to differences in occupancy between crash-involved vehicles and all vehicles), this dissertation then seeks to make three additional contributions. First, research has been conducted to develop methods for systematically extracting average vehicle occupancy from crashes and controlling potential bias to evaluate the corridor use of vehicle sharing. This involves assessing the occupancy of vehicles on specific links within transportation networks, which has been challenging due to limited data availability. Second, the dissertation examines how this occupancy can influence public decision making by quantifying how better vehicle occupancy data can influence project prioritization, given that such projects rely on estimated improvements in person throughput and person delay reduction. Finally, this research examines how land use affects vehicle occupancy, finding that the influence of key variables (e.g., travel time to work or percent of low-income households) differ by region. Returning to the premise that transportation supply is based in part on vehicle occupancy, this finding leads to region-specific considerations that localities may wish to use to inform their land development review process.