Travel and energy implications of ridesourcing service in Austin, Texas

https://doi.org/10.1016/j.trd.2019.03.005Get rights and content

Highlights

  • 1.5 million individual ridesourcing rides used to estimate net effect on energy use.

  • Commuting and between-ride “deadheading” account for 19% and 26% of ridesourcing VMT.

  • Ridesource vehicles are 2 MPG more efficient and twice as likely to be hybrid-electric.

  • Ridesourcing increases energy use by an estimated 41–90% compared to prior mode.

Abstract

This paper identifies major aspects of ridesourcing services provided by Transportation Network Companies (TNCs) which influence vehicles miles traveled (VMT) and energy use. Using detailed data on approximately 1.5 million individual rides provided by RideAustin in Austin Texas, we quantify the additional miles TNC drivers travel: before beginning and after ending their shifts, to reach a passenger once a ride has been requested, and between consecutive rides (all of which is referred to as deadheading); and the relative fuel efficiency of the vehicles that RideAustin drivers use compared to the average vehicle registered in Austin. We conservatively estimate that TNC drivers commute to and from their service areas accounts for 19% of the total ridesourcing VMT. In addition, we estimate that TNC drivers drove 55% more miles between ride requests within 60 min of each other, accounting for 26% of total ridesourcing VMT. Vehicles used for ridesourcing are on average two miles per gallon more fuel efficient than comparable light-duty vehicles registered in Austin, with twice as many are hybrid-electric vehicles. New generation battery electric vehicles with 200 miles of range would be able to fulfill 90% of full-time drivers’ shifts on a single charge. We estimate that the net effect of ridesourcing on energy use is a 41–90% increase compared to baseline, pre-TNC, personal travel.

Introduction

In the late 2000s pervasive use of smartphones, GPS systems, and digital mapping/routing applications gave rise to ridesourcing services. In a ridesourcing service, a smartphone application connects an individual seeking a ride with a nearby driver willing to provide transportation. The application gives real-time information about wait and travel times, enables passengers to pay via credit card, and allows passengers and drivers to evaluate each other once travel has been completed. The providers of such services classified as “transportation network companies” or TNCs (Rayle et al., 2016). In some cities, the application allows travelers to share the ride (and cost) with a stranger. In this case, the ridesourcing trip can be referred to as “ridesharing”.

Ridesourcing services have grown dramatically over the past decade. As of early 2018, Uber—the largest U.S. provider—reported service in 633 cities worldwide, with exponential growth in the number of rides provided. Uber accumulated its first 1 billion rides in about 6 years, the next billion in 6 months, and 3 billion more in about another year (Uber, 2017). In early 2018, Lyft—the second-largest U.S. TNC—reported among its “community” 23 million passengers, 1.4 million drivers, and service in all U.S. states and more than 300 airports (Lyft, 2018), and in September 2018 announced their first billion rides. China’s Didi Chuxing, Uber’s main competitor in global growth, reported having more than 450 million users and 21 million drivers, with its 25 million daily rides eclipsing the combined rides given by the rest of the world’s ridesourcing services combined (Sreeharsha and Isaac, 2018). Continued evolution of this concept consists of a range of similar services including conventional car-sharing, bike-sharing, scooter-sharing, and microtransit services (e.g., on-demand shuttle buses) (Burgstaller et al., 2017, Clewlow and Mishra, 2017). Taken together, these types of emerging travel services are referred to as “mobility as a service” (MaaS); many envision travelers eventually purchasing access to a portfolio of travel services through a monthly subscription (Goodall et al., 2017). The TNC service providers have expressed their intention to eventually replace their contract drivers with automated vehicles, thereby greatly reducing the cost of providing a ride (Uber, 2019, Lyft, 2019). Using mesoscopic simulation experiments, Loeb et al., 2018 estimated that shared autonomous electric vehicles could serve 40% of all trips under 50 miles in Austin Texas, with a nearly 20% increase in VMT. Such services could lead to significant reductions in taxi ridership (Contreras and Paz, 2018, Nie, 2017), one of the major competitors of TNCs.

TNC companies typically do not release their data to the public or researchers, but a number of independent studies show that these services greatly impact urban transportation patterns and trends. For example, recent studies in San Francisco suggest that around 6000 ridesourcing vehicles operated in the city at peak times—outnumbering taxis by more than a factor of 15. In addition, ridesourcing vehicles accounted for around 10% of all person-trips and vehicle miles traveled (VMT), as well as about half of the change in traffic congestion in San Francisco between 2010 and 2016 (San Francisco County Transportation Authority, 2017, San Francisco County Transportation Authority, 2018). Another study of seven major U.S. cities found that 21% of adults used ridesourcing services and 9% more used them with friends but had not installed the application themselves; of ridesourcing adopters in metropolitan areas, 24% used the service on a weekly or daily basis (Clewlow and Mishra, 2017). A 2016 Pew Research Center study estimated that 21% of urban Americans and 15% of suburban Americans had used ridesourcing services (Smith, 2016).

Ridesourcing can decrease energy use in the following ways. In the cities where ridesharing (or pooling) is offered, sharing a ride with a stranger can as much as halve vehicle miles of travel compared to two travelers driving their own personal vehicles along a similar route. Second, in the medium term, by concentrating VMT in fewer vehicles, the owners of those vehicles have an incentive to purchase more efficient automobiles; high-mileage use means that the initial increase in the purchase price of a more efficient vehicle will be offset sooner by lower fuel costs. Finally, in the long term, riders may retire an existing vehicle, and, no longer facing a fixed cost for their transportation needs, may eliminate trips they made previously with their own vehicle (Cervero et al., 2007, Xue et al., 2018). An analysis of U.S. state-level data found that per-capita vehicle registrations decreased 3% after entry of ridesourcing services in metropolitan areas, with no effect on VMT (Ward et al., 2018).

On the other hand, ridesourcing services can increase energy use by a number of different ways. Miles driven without a paying passenger onboard are referred to as “deadheading”; deadhead miles can offset any energy benefits of ridesourcing services, and can lead to greater traffic congestion. There are two major components of deadheading: commuting since some drivers commute relatively long distances into urban areas to begin and end their shift of driving; and between-ride deadheading, which includes drivers circling areas waiting for riders to summon them, potentially driving longer distances as directed by their mobile app to take advantage of surge pricing, and the additional miles driven after accepting a ride request to pick up the passenger.

The few studies analyzing deadheading from ridesourcing services vary considerably in terms of data sources, methods, and areas of study. Among recent studies, estimates of the fraction of total ridesourcing VMT that are deadhead miles range from 36% to 45%, yet none of these studies capture all empty miles accumulated by ridesourcing vehicles. Henao and Marshall (2018) calculated a conservative (lower end) empty mile rate of 41% for a single Uber/Lyft driver in Denver due to deadheading and accounting for commuting only at the end of shifts. Komanduri et al., 2018 used the RideAustin data set to estimate that 37% of all ridesourcing VMT are deadhead miles, including commuting at the beginning and end of shifts. Their estimate assumed a uniform commute distance of two miles, which also applied when picking up passengers after more than 30 min of idle time. They also accounted for straight-line distances between the rides without adjusting for the real street network distance. Cramer and Krueger, 2016 obtained from Uber the city-wide average percentage of drivers’ miles driven with a passenger in five cities.2 They report empty miles rates of 45% in Seattle and 36% in Los Angeles, excluding miles driven with the Uber app turned off, which might exclude empty commuting miles at the beginning and end of shifts. Drawing primarily on operational data collected by the New York City Taxi and Limousine Commission, a non-peer-reviewed report estimates an empty VMT rate of 40% for weekday ridesourcing rides that started and/or ended in Manhattan’s Central Business District, again not accounting for commuting miles (Schaller, 2017). In San Francisco, a method based on the public-facing APIs of Uber and Lyft was used to estimate an empty VMT rate of 20% for rides starting and ending within the city’s core area (San Francisco County Transportation Authority, 2017). However, their method excludes commuting, and understates the share of deadheading as the authors attributed the distance from ride request to passenger pickup to passenger miles instead of to empty miles.

Whether ridesourcing increases VMT and energy use depends on the mode of travel replaced by the ridesourcing trip. Taxis likely generate comparable, or perhaps even more deadheading miles than TNCs: taxis are more likely to circle areas seeking riders, while ridesource vehicles in theory can wait parked until a ride is requested through the smart phone app. Private auto trips often add additional VMT from searching for parking at the end of the trip, depending on the destination, whereas ridesourcing drivers never have to search for parking (although they may have to find a safe location to drop off their passengers). In principle, deadheading and ridesourcing inefficiency can be reduced over time by TNC providers better matching driver supply with passenger trip demand, thereby reducing the distance driven to reach riders (Xu et al., 2019).

A third factor of ridesourcing that may increase energy use is that such trips may substitute for a trip in a more energy-efficient mode or even induce new travel; for example, a trip shifted from public transit to ridesourcing service would, on average, increase the amount of energy consumed per passenger during that trip.3 On the other hand, ridesourcing service may supplement conventional public transit service, by providing first-/last-mile travel to or from transit stops. An analysis using transit agency data found that ridership increased 5% two years after Uber entry, but ranged from a 7% increase to an 8% decrease based on the size of the metro area and its transit system (Hall et al., 2018). In addition, by encouraging more to drive part-time (as opposed to fewer full-time taxi drivers), ridesourcing services reduce the average number of rides provided per driver or vehicle, which may increase traffic congestion. The stop-and-go driving from increased traffic congestion leads to a slight increase in energy use (Gately et al., 2017).

Table 1 summarizes six aspects of ridesourcing services that influence their VMT and energy use. The first three likely result in reduced VMT and energy use, while the last three likely result in increased VMT and energy use, relative to travel before the introduction of ridesourcing services.

Our analysis uses data of individual rides conducted by a TNC, RideAustin, in Austin Texas, to quantify VMT and energy use effects of ridesourcing. TNCs Uber and Lyft pulled their services out of the Austin, Texas market on May 8, 2016 (Solomon, 2017). One month later a local non-profit company, RideAustin, was formed to provide such transportation services in the city. In addition to RideAustin, several other providers began operation in Austin in the same timeframe. An online survey of nearly 1840 riders, conducted in November and December 2016, found that 42% used one of these new ridesourcing services after Uber and Lyft left the market, and of those that used one of the new services, 47% used the RideAustin service (Hampshire et al., 2017). Lavieri et al., 2018 also used RideAustin data to model the spatiotemporal distribution of ridesourcing rides, the demographics of passengers, and mode substitution in relation to public transit.

This study estimates the fraction of RideAustin drivers’ VMT without passengers, as well as the VMT attributed to drivers’ travel into the city at the beginning and the end of their shift, using both measured and estimated distances. We categorize drivers based on their weekly working hours and compare travel patterns of occasional, frequent, and full-time drivers. In addition, the distribution of rides to and from certain major locations and land uses in Austin are analyzed. Vehicle fuel economy of the RideAustin vehicles is reviewed and compared with that of the personally owned fleet in the region, to determine if vehicles used through the service are more efficient than comparable vehicles registered in Austin. Finally, the impacts of the RideAustin ridesourcing service on net energy use, based on estimates of five of the six factors shown in Table 1, are estimated. This study demonstrates the value of detailed data on individual rides provided by TNCs, as well as other necessary data, to fully assess the net impact of this emerging type of transportation service on mobility and energy use of urban transportation systems.

Section snippets

Data and methods

We obtained publicly-available data from RideAustin on nearly 1.5 million individual rides made over an 11-month period from June 2016 to April 2017 (RideAustin, 2017). The RideAustin data on nearly 1.5 million rides, conducted by almost 5000 drivers, includes the datetime stamp at five points along each ride: when the ride request was dispatched to a driver; when the driver accepted the dispatched request; when the driver reached the rider; when the ride started; and when the ride was

Results and discussion

In this section we describe the detailed methods used to estimate the effect of ridesourcing service on between-ride and commute deadheading, more efficient vehicles, and modal shift, as well as an analysis of ride distributions by location and land use.

Conclusions

Data on 1.5 million individual rides were used to estimate the additional miles RideAustin drivers travel to reach a rider, as well as the additional miles driven between consecutive rides (referred to as deadheading). Relative to the distance from passenger pick up to drop off, RideAustin drivers traveled 21% more miles just to reach their passenger(s) after accepting the ride request; and drivers traveled an estimated 55% more miles between the end of a ride and the start of the next ride

Acknowledgements

This work was authored by the Lawrence Berkeley National Laboratory under Contract No. DE-AC02-05CH11231 and the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, under Contract No. DE-AC36-08GO28308, for the U.S. Department of Energy (DOE). This report and the work described were sponsored by the DOE Vehicle Technologies Office (VTO) under the Systems and Modeling for Accelerated Research in Transportation (SMART) Mobility Laboratory Consortium, an

Declarations of interest

None.

References (47)

  • Alternative Fuels Data Center, 2018. Average Per-Passenger Fuel Economy of Various Travel Modes. URL...
  • G.S. Bauer et al.

    Cost, energy, and environmental impact of automated electric taxi fleets in Manhattan

    Environ. Sci. Technol.

    (2018)
  • S. Burgstaller et al.

    Rethinking mobility: the “pay as you go” car: ride hailing just the start

    Goldman Sachs Groups Equity Res.

    (2017)
  • R. Cervero et al.

    City CarShare: longer-term travel demand and car ownership impacts

    Transp. Res. Rec. J. Transp. Res. Board

    (2007)
  • City of Austin, 2010. Austin Open Data Portal. URL...
  • Clewlow, R.R., Mishra, G.S., 2017. Disruptive Transportation: The Adoption, Utilization, and Impacts of Ride-Hailing in...
  • J. Cramer et al.

    Disruptive change in the taxi business: the case of Uber

    Am. Econ. Rev.

    (2016)
  • Feigon, S., Murphy, C., 2016. Shared Mobility and the Transformation of Public Transit, TCRP J-11/TASK...
  • George, S.R., Zafar, M., 2018. Electrifying the Ride-Sourcing Sector in...
  • Goodall, W., Fishman, T.D., Bornstein, J., Bonthon, B., 2017. The rise of mobility as a service. Deloitte Review, Issue...
  • Hampshire, R., Simek, C., Fabusuyi, T., Di, X., Chen, X., 2017. Measuring the Impact of an Unanticipated Disruption of...
  • Henao, A., 2017. Impacts of Ridesourcing – Lyft and Uber – on Transportation including VMT, Mode Replacement, Parking,...
  • A. Henao et al.

    The impact of ride-hailing on vehicle miles traveled

    Transportation (Amst)

    (2018)
  • Cited by (97)

    • On ride-sourcing services of electric vehicles considering cruising for charging and parking

      2023, Transportation Research Part D: Transport and Environment
    • Patterns of electric vehicle charging on transportation network companies in the US

      2023, Transportation Research Part D: Transport and Environment
    View all citing articles on Scopus
    1

    Current address: Department of City and Regional Planning, University of North Carolina at Chapel Hill, New East Campus Box 3140, Chapel Hill, NC 27599, United States.

    View full text