A timely detour: an examination of the deviation time associated with shared ride-hailing services in Toronto 

Abstract
To date, most research that has looked at ride-hailing has concentrated on its traditional form and little attention has been given to shared ride-hailing services such as UberPool and Lyft Line. These services promise to substantially reduce the cost of ride-hailing trips by enabling passengers traveling in the same general direction to share rides. Furthermore, they are poised to have vast social and economic implications for cities, as their goal, according to Uber’s former CEO Travis Kalanick, is to take a third of all cars off the road (Smith, 2014). Influential scholars such as Shaheen and Cohen (2019) agree that pooled services behave differently to traditional ride-hailing and that much surrounding their impacts upon cities remains misunderstood. While indeed largely unexplored, many of the alleged benefits of shared ride-hailing are believed to align closely with cities’ sustainable transport objectives (Circella et al., 2018). However, without an adequate analysis of this mode, which necessitates separating shared ride-hailing service from other forms of ride-hailing and distinguishing between successfully matched and unmatched trips, there is a risk that policymakers and researchers will inaccurately assess the potential benefits of pooling and continue to regulate it in accordance to the impacts of traditional ride-hailing.
A particularly unexplored facet of shared ride-hailing is the detour time required to pick up additional passengers. This time penalty, known as deviation time, is important since it affects the attractiveness of shared ride-hailing (Schwieterman and Smith, 2018; Sarriera et al., 2017), and ultimately the extent to which it may offer transportation, environmental, and social benefits. Using a 17-million record dataset of every ride-hailing trip conducted in Toronto between September 2016 and April 2017, we conduct a travel-time difference analysis between successfully matched shared ride-hailing trips and equivalent non-shared trips. Equivalence is assessed using the trip origin and destination start-times and locations, from which we create spatiotemporal buffers. We find pooling to add an average of 3.8 minutes to travel-time, and find this time penalty to vary across time and space in Toronto. Shared trips conducted outside of downtown are more likely to go unmatched, but when matched they will often result in sizeable delays in arrival time. Our preliminary results also indicate that shared ride-hailing demand significantly influences matching propensity, and that pooling reduces vehicle-kilometers traveled, relative to equivalent non-shared ride-hailing trips. By examining the time penalty associated with sharing, this study contributes novel empirical evidence of the effects of shared ride-hailing services on urban mobility systems and contributes to shaping the future of mobility by exposing the intrinsic spatial-related inequalities of this mode.