Author ORCID Identifier

Date Available


Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation




Civil Engineering

First Advisor

Dr. Gregory Erhardt


US traffic fatal deaths have steadily risen since 2010, with the past few years witnessing an unusual trend increase. To reverse such a dangerous trend, one must understand how and why road crashes occur and which factors are causing them. Emerging transportation technologies have shown the potential to improve mobility and safety. However, such technologies are not inherently beneficial and could worsen road safety if not effectively implemented. One such transportation technology that warrants investigation is the rise of ridesharing services, also called Transportation Network Companies (TNCs). The primary goal of the dissertation is to explore the statistical relationship between road safety outcomes and TNC service components like curbside pick-ups and drop-offs (PUDO) or through the TNC-involved vehicles miles traveled (VMT). It evaluates the relationship between TNC service components like PUDO and Tot TNC VMT with four main types of road crash frequency: the total number of road crashes, fatal and severe injury crashes, crashes involving pedestrians and bicyclists, and crashes involving drink-driving using San Francisco (SF) county data. A fixed-effect Poisson Regression Model with a robust covariance matrix compares San Francisco (SF) county's 2010 safety outcomes when TNCs were negligible to safety outcomes for the exact locations in 2016 for which spatially detailed TNC data is available. Dependent variables like Total Crashes, Fatal and Injury Crashes, Pedestrian and Bicyclist Crashes, Alcohol-involving (DUI) Crashes, and Property Damage Only (PDO) Crashes are evaluated using the model, controlling for vehicle speed, Total VMT, and TNC service components, namely TNC VMT and PUDO. We apply that model to 2010 and 2016 scenarios and counterfactual scenarios that estimate what would have occurred in 2016 without specific aspects of TNC operations. The results show that TNCs indirectly increased total crashes by 4% due to higher exposure and 7% due to changes in vehicle speeds. The direct effect of TNCs on crashes offsets these increases, reducing crashes by 14%, but this effect depends upon the model specification and is insignificant in other specifications tested. The results for other types of crashes are similar in direction but lower in significance. Overall, the results suggest that TNCs are a minor factor in road safety outcomes, at least within the limits of what we can measure with the available data. This finding is broadly consistent with past research on the topic. These results interest engineers, planners, and policymakers seeking to improve road safety. Those aiming to reduce traffic crashes would be well-advised to avoid getting distracted by TNCs in one direction or another and instead focus on known solutions, including road design, vehicle technology, and reducing exposure through reducing vehicle miles traveled.

Digital Object Identifier (DOI)