Journal of Transport Geography 95 (2021) 103147

Contents lists available at ScienceDirect

Journal of Transport Geography

journal homepage: www.elsevier.com/locate/jtrangeo

Evaluating the equity implications of ridehailing through a multi-modal

accessibility framework
Bilal Abdelwahab a, Matthew Palm b, *, Amer Shalaby a, Steven Farber c
a
Civil and Mineral Engineering, University of Toronto, 35 St George Street, Toronto, ON M5S 1A4, Canada
b
Department of Urban Studies, Worcester State University, 486 Chandler Street, Worcester, MA 01602, United States of America
c
Human Geography, University of Toronto Scarborough, 1265 Military Trail, Toronto, ON M1C 1A4, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: In rapidly-growing metropolitan regions, it is crucial that transportation-related policies and infrastructure are
Accessibility designed to ensure that everyone can participate equitably in economic, social, and civil opportunities. Ride­
Ridehailing hailing services are touted to improve mobility options, but there is scant research that incorporates this mode
Equity analysis
within an accessibility framework. This paper employs a generalized cost measure in a multi-modal accessibility
First/last mile
Transportation equity
framework, namely Access Profile Analysis, to assess the role of ridehailing in providing job access to historically
under-resourced parts of Toronto, Canada, referred to by the city as Neighborhood Improvement Areas (NIAs).
Ridehailing is analyzed both as a mode of commute and as a feeder to the transit network (a first-mile solution).
The results indicate that there are two main determinants of the extent to which ridehailing provides additional
accessibility over transit: the transit level of service at the origin zone and the zone's proximity to employment
opportunities. The ridehailing mode is shown to increase accessibility especially to closer destinations (jobs),
with the highest improvement seen in the city's inner suburbs. On the other hand, integrating ridehailing with
public transit does little to improve access to jobs. Compared to the rest of the city, NIAs experience a higher
accessibility improvement from ridehailing alone, but not from its integration with transit. Nonetheless, job
accessibility remains lower in NIAs than in other areas – even after the introduction of ridehailing.

1. Introduction measure the potential benefits (or lack thereof) of the mode. We
demonstrate how this can be achieved by extending the access profile
Ridehailing services such as Uber and Lyft hold the potential to analysis (Neudorf, 2014) approach to calculating a multi-modal,
improve mobility for under-resourced households (Brown, 2019; King maximum accessibility profile for each traffic analysis zone (TAZ) and
et al., 2019). Transit agencies are beginning to pilot integrations of these planning district (PD) in the City of Toronto. This enables us to measure
services into transit systems, sometimes justifying these pilots from an how accessibility by ridehailing compares to that of public trans­
equity standpoint, such as the Pinellas-Suncoast Transit Authority's Late portation, while accounting for the generalized costs of travel (e.g.
Shift and Direct Connect programs (Curtis et al., 2019, p. 52). Trans­ combining pecuniary with travel time costs). To understand the poten­
portation equity researchers measure the distribution of accessibility tial equity impacts of introducing ridehailing, we investigate changes to
provided by transportation systems to evaluate the social equity of those the accessibility profiles of historically under-resourced neighborhoods
systems (Pereira et al., 2017). Little work measures how the introduction using a neighborhood classification system developed for social policy in
of ridehailing services has altered the landscape of transportation the City of Toronto.
accessibility within cities, especially using an equity lens. This method also allows us to assess the additive benefit of
Despite the recent surge in research activity using accessibility to combining two modes in comparison to either mode alone, enabling
assess equity, as well as recent progress being made in understanding measurement of the accessibility benefits of emerging first-mile part­
inequalities in actual patterns of ridehailing use, few papers have nerships between ridehailing and transit (Curtis et al., 2019). The next
attempted to place ridehailing within an accessibility framework to section overviews equity research on ridehailing, highlighting a gap in

* Corresponding author.
E-mail addresses: [email protected] (B. Abdelwahab), [email protected] (M. Palm), [email protected] (A. Shalaby), steven.farber@
utoronto.ca (S. Farber).

https://doi.org/10.1016/j.jtrangeo.2021.103147
Received 21 April 2021; Received in revised form 9 July 2021; Accepted 11 July 2021
Available online 15 July 2021
0966-6923/© 2021 Elsevier Ltd. All rights reserved.
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

knowledge on the likely impacts of ridehailing on accessibility. We then 2.1. Measuring multi-modal accessibility
introduce the use of generalized costs in accessibility analysis, followed
by a detailed overview of our data inputs. Our methods section details Most equity studies on accessibility deploy gravity-based approaches
how we construct access profiles and then convert them into accessi­ to the calculation of accessibility (Foth, 2013; Hu, 2015; Karner, 2018).
bility scores. We then present results at the TAZ scale, followed by In an effort to better incorporate travel costs in accessibility measures,
planning district aggregations. Our conclusions summarize the value of gravity models give more weight to destinations that are closer or
our new approach and synthesize policy recommendations from our cheaper to reach. Monetary costs have a significant impact on the travel
application to ridehailing. opportunities of under-resourced households (Verbich and El-Geneidy,
2017), yet most gravity-based accessibility studies account for travel
2. Literature review times only, not travel costs. This is important for ridehailing in partic­
ular, as the mode's price structure can be a deterrent for many potential
In the transportation context, equity is traditionally conceptualized riders (Dong, 2020). El-Geneidy et al. demonstrate how to incorporate
as the fairness of the distribution of benefits and burdens of trans­ fares into transit accessibility modelling through a generalized cost (GC)
portation systems among individuals or groups in society (Martens, approach (El-Geneidy et al., 2016). The relatively higher monetary costs
2016). Equity scholars have embraced accessibility – the ease of of ridehailing compared to transit necessitates the use of a generalized
reaching activity destinations – as an important measure of system cost approach to considering the accessibility benefits of both modes
benefits when conducting equity analysis (Karner, 2018; Martens and jointly. Specifically, we question whether the higher fares of ridehailing
Golub, 2018; Pereira et al., 2017). However, most research on equity are worth the potential travel time improvements over transit.
and ridehailing focuses on rider demographics and revealed travel be­ Measuring the accessibility benefits of ridehailing as a complement
haviors. Ridesharing users tend to be younger, wealthier and more- to public transit also requires a multi-modal approach that can measure
highly educated than the general population (Conway et al., 2018; the additive contribution of each mode towards a location's overall
Young and Farber, 2019). However, scholars have identified many po­ accessibility. Prior literature on equity and accessibility is almost
tential equity benefits associated with ridehailing and shared-mobility entirely mode-specific, and public transit focused studies generally as­
more generally (Shaheen et al., 2017), and data provided by rideshar­ sume that travelers walk the first/last mile (Foth, 2013; Hu, 2015;
ing companies has yielded a more complex picture, with riders from Karner, 2018). In scanning this literature, we identified a new,
neighborhoods of color found to be using these services more frequently underutilized approach most appropriate for this study's task, namely
than riders from other neighborhoods (Brown, 2018). Ridehailing the Access Profile Analysis (APA). This approach is proposed recently by
companies also provide greater geographic coverage than traditional Neudorf (Neudorf, 2014) based on earlier work (Black and Conroy,
taxis (Pan et al., 2020), can usually provide greater coverage than 1977). It involves calculating generalized cost measures for each mode
transit, and may be more effective at delivering services given real-time of travel and plotting the cumulative number of destinations reachable
matching of supply and demand.. The extant literature makes it difficult along a generalized cost axis, a theoretical example of which we present
to conclude whether the introduction of ridehailing improves or worsens in Fig. 1. The maximum number of destinations reachable across each
transportation system equity. generalized cost level constitutes a maximum accessibility level for the
This question is further complicated by ridehailing's dual role as a origin. The percent of the plot area under this maximum level can be
complement and substitute to public transit (Gehrke et al., 2019; Hall expressed as a single indicator of accessibility that accounts for time and
et al., 2018; Li et al., 2021). As a complement, ridehailing can provide monetary costs across all modes. Lee applied the APA to Toronto,
mobility to carless residents in transit-poor areas or provide a first/last demonstrating how policies as divergent as transit improvements, land
mile service for transit. In Toronto, unsurprisingly, observed ridehailing use intensification, and parking price reductions can be compared by
trips appear to complement transit in areas of the city where transit their impacts on accessibility under a uniform rubric (Lee, 2019).
services are more limited (Young et al., 2020a), yet most ridehailing In sum, agencies are exploring how to integrate ridehailing with
trips occur within the urban core, competing directly with fixed transit. public transit to advance equity goals, but the literature's existing
In other regions, however, ridehailing does not appear to cover service methods for measuring accessibility are unsuitable for understanding
gaps (Barajas and Brown, 2021; Jiao and Wang, 2021). Some U.S. transit the impact of a new mode on overall accessibility, particularly one that
agencies use equity to justify first/last mile ridehailing partnerships trades reduced travel times for increased monetary costs compared to
(Curtis et al., 2019), yet the potential equity benefits of these programs public transit. This study fills this gap by applying a novel approach to
are unexamined (Palm et al., 2020), as are the accessibility benefits they calculating multi-modal accessibility in one of North America's largest
provide, gaps this paper aims to fill. The additive accessibility benefits of cities, Toronto. Specifically, this study applies the APA method to
a new mode of travel necessitates a multi-modal approach to calculating answer two questions: one, does ridehailing outperform public transit in
accessibility that controls for modal differences in monetary costs. One providing accessibility for under-resourced communities; and two, can
recent approach proposes an adaption of the location quotient (Isser­ the use of ridehailing as an ingress mode improve transit accessibility for
man, 1977) and GINI coefficient (Delbosc and Currie, 2011) to assess the these communities?
impact of ridehailing on accessibility in Rio de Janeiro (de Oliveira
Souza et al., 2021). The authors find that ridehailing “generates a sig­ 3. Data and study area
nificant improvement in the distribution of job accessibility in regions
with low supply and variety of public transportation” (de Oliveira Souza This section begins by laying out all data sources used in a multi-
et al., 2021, p. 9). While a novel and valuable contribution, their modal APA. The following subsection details how source data are
methods offer only an ability to compare the inequities (GINI co­ transformed into APA curves and curve-derived indices.
efficients) of job access by transit versus by ridehailing. They do not
present a clear understanding of the equity impacts of how ridehailing
works additively or in combination with transit. Additionally, their 3.1. Data
analysis did not include ridehailing wait times. As such, we drew upon a
different thread of accessibility methods less frequently deployed in the Data used in this research includes model-generated origin-destina­
equity literature to devise our approach. We turn to this literature next. tion (O–D) data, observed ridehailing trip-level data, and information
on employment and land use. The spatial unit of analysis is the traffic
analysis zone (TAZ) (“Data Management Group – Transportation
Tomorrow Survey,”, n.d.).

2
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Fig. 1. Theoretical Access Profile Analysis (APA) for a given origin.

3.1.1. Transit and auto generalized cost data 8, 2016 and March 31, 2017. The wait times for each trip were provided,
The O–D data for the transit and auto modes was obtained from the rounded to the nearest minute. Wait times of all ridehailing trips,
Travel Modelling Group (TMG) at the University of Toronto. It is esti­ including trips on shared services that have slightly higher wait times
mated from the TTS-demand-based network assignments using EMME (Young et al., 2020b), that started in the city during the morning peak
software for 2016 data. Since the study focuses on job access, travel period were averaged by TAZ to produce wait-times for each TAZ.
conditions pertaining to the weekday morning peak period (6:00–9:00 Ridehailing IVTT is based on the auto mode IVTT modeled by TMG,
AM) was used. All trips in the data are assumed to begin and end at zonal rather than on reported travel times within the City's dataset. This is
centroids except for trips that start or end at transit stations. The station done for two reasons. First, many O–D pairs do not have ridehailing
ends of these trips are represented by their exact locations. trips recorded in the city data, but accessibility analysis requires travel
The TMG provided O–D data for auto trips (in-vehicle travel time, time estimates for all feasible O–D pairs. Second, it allows for a fair and
trip distance, operating cost, and tolls) and transit trips (transit unbiased comparison with transit by ensuring both modes' travel times
perceived travel time) – for all pairs of origins and destinations. The are estimated under the same modelling assumptions.
operating cost of auto travel in the data is C$0.068 per kilometer, and it
includes gas, maintenance, and tire replacement costs. This is likely 3.1.3. Other data
lower than true operating costs (CAA, 2013), but we retain it as it comes Employment and land use data is obtained from the Data Manage­
from the same model as other auto mode data. Additionally, parking cost ment Group (DMG) of the University of Toronto based on estimates
data was provided by the TMG. The parking cost is the destination zone's using TTS and Census data. It includes the population count, number of
average hourly parking rate multiplied by eight hours, assuming eight jobs, and area of each TAZ.
hours of work on a typical weekday. The parking costs do not reflect
employer subsidized parking, or alternative parking options, like 3.2. Study area
monthly rentals. The implication is that they likely over-estimate
parking costs in suburban zones, where employer-provided parking is This study examines job access in the City of Toronto, Canada. Tor­
more common, and they become more accurate towards downtown, onto is Canada's most populated city, with a population of more than 2.7
where employer-provided parking is rarer. million (Statistics Canada, 2017). The city is the core of a larger region
The transit perceived travel time is the weighted sum of the in- known as the Greater Golden Horseshoe Area (GGHA). In this study,
vehicle travel time (IVTT), wait time, transfer time, walk time, and origins are limited to zones within the City of Toronto, while destina­
fare (converted into time units). Multipliers for different trip compo­ tions are jobs in the whole GGHA. There are employment opportunities
nents were estimated using a particle swarm optimization procedure throughout the city and the GGHA, with the downtown area (planning
that runs EMME to minimize the RMSE of the boarding differences by district 1) and Toronto Pearson International Airport (near the city's
line for the AM and PM time periods. Wait time has a weight of 3.25 western borders) as two of the densest employment centers (see Fig. 2).
(each minute of waiting time is perceived as 3.25 min). Walk time is Toronto has a well-established public transit network that operates
weighted based on each walking link's length as defined by the TMG's multiple subway lines, streetcars, buses, and regional rail. Transit's level
transit assignment models. Both IVTT and fare have a weight of 1.0, and of service, measured by the number of arrivals per hour in each zone
the value of time (VOT) used for computing the transit generalized cost during averaged across a one-week period, is shown in Fig. 3. Transit
is C$14/h. This VOT is the average wage for the services/non- level of service is best in the downtown core and generally decreases as
professional sector, the lowest-income sector, in Ontario for 2016. It is distance from downtown increases.
calculated based on employee wages data (Statistics Canada, n.d. as C Ridehailing began operating in Toronto in 2016. Uber arrived in
$13.98 in 2016 dollars. March 2016 and Lyft started service in December 2017, with each of­
fering multiple ride options. Use grew rapidly, with the number of rides
3.1.2. Ridehailing generalized cost data more than doubling to 4.8 million a month between October 2016 and
Ridehailing wait times were estimated from a database of all thirteen August 2018 (Li, 2019).
million ridehailing trips made in the City of Toronto between September

3
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Fig. 2. Spatial distribution of employment in the GGHA with Toronto Planning Districts.

Fig. 3. Transit level of service.

3.2.1. Defining under-resourced areas domains of indicators: Economic Opportunities, Social Development,
Toronto is divided into 16 planning districts (PDs) and 625 TAZs. For Participation in Decision-Making, Physical Surroundings, and Healthy
social planning purposes, the City of Toronto divides Toronto into 140 Lives (City of Toronto, 2014). All indicators were based on publicly
neighborhoods. The City and local stakeholders designated 31 of these available data, aggregated to the neighborhood level. TAZs that fall
neighborhoods as Neighborhood Improvement Areas (NIAs), or loca­ within NIAs are shown in all maps in green. This study makes use of the
tions facing systematic inequalities and needing immediate city invest­ classification of NIA neighborhoods to identify communities that the city
ment (City of Toronto, 2015). Determination of NIA status was based on has historically under-invested in.
an inequality index derived from a principle components analysis of 5

4
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

4. Methods The O–D transit service data included IVTT, wait time, walk time,
and transit fare. This data was converted back into monetary terms to
The analysis included three stages: estimating a generalized cost match the GC units of the rest of the modes.
(GC) of travel, plotting access profiles, and calculating accessibility
GCTransit = (IVTT + Wait*WeightWait + Walk Time*WeightWalk )*VOT + Fare
indices. Four modes of travel are considered: auto (driving), public
transit, ridehailing (all the way), and the integration of ridehailing (for (4)
the first mile) with public transit, henceforth called the integrated mode. In the integrated mode, commuters are assumed to use ridehailing
for the first mile of the trip only if the resulting GC for the combined
ridehailing and transit is lower than the GC by transit alone for the same
4.1. Generalized cost trip. For each O–D pair, calculating the integrated mode's GC is per­
formed through the following steps:
GC consists of both monetary and temporal components, with the
temporal components of different trip stages weighted differently in 1. Calculating the GC required to travel from origin to all rapid transit
order to reflect commuters' relative perceptions. The weighted travel stations (subway, BRT, or commuter rail stations) by ridehailing,
time is converted into its equivalent monetary value using a value of 2. Calculating the GC required to travel from all stations to the desti­
time (VOT), and the final GC values are expressed in Canadian dollars. nation by transit,
The VOT is assumed to be 14C$/h across all modes for two reasons. First, 3. Adding the two parts of the trip together and choosing the station
this is the VOT used by the region's travel demand model when calcu­ that minimizes the GC, and
lating input data for the transit and driving modes. Applying a separate 4. Comparing the resulting GC to the GC of the transit all-the-way op­
VOT to ridehailing risks biasing the analysis. Second, while average tion for that O–D pair and choosing the lower of the two.
VOTs from observed and stated preference surveys on ridehailing can be
higher (Alonso-González et al., 2020; Lazarus et al., 2021; Schwieter­ The first part of such a trip is referred to as the access part or first
man, 2019), our motivation is to understand ridehailing's impact on mile, while the second part is referred to as the transit part.
under-resourced neighborhoods where residents live on lower incomes. {
VOT tends to rise with income (Börjesson et al., 2012), requiring us to GCIntegrated =
x, if x < y
(5)
y, otherwise
select a low VOT that reflects under-resourced households' financial
constraints. C$14/h reflects the average hourly wage for the lowest
where: x is the sum of ridehailing GC for the access part and the transit
wage segment of the Canadian workforce, making it an appropriate
GC for the transit part of the trip, while y is the transit GC of traveling
choice for this task (Statistics Canada, n.d.).
from origin to destination. It is noteworthy that intra-modal transfers
Auto GC consists of IVTT, operating cost, highway toll, and parking
within the transit network in Toronto (which is operated by the Toronto
cost.

GCAuto = IVTT*VOT + Operating Cost + Toll Cost + Hourly Parking Cost*8 hours (1)

Ridehailing GC is calculated using a combination of ridehailing and Transit Commission) are free within a two-hour period.
auto data since the ridehailing data alone was insufficient to calculate Since trips are assumed to start/finish at zonal centroids in the O–D
the full cost of ridehailing. It includes fare, highway toll cost, IVTT, and data, the cost of travel within the same zone is calculated separately.
wait time. Fare is based on the fare structure listed on Uber's website for Simplistic assumptions are made due to the complexity of models that
UberX, the company's lowest-cost service besides UberPool (rides shared were used to produce the O–D data. In order to account for intrazonal
with other customers), and is composed of a fixed fee (base and booking travel, zones are assumed to be circular, and the expected intrazonal
fees) of C$5.25, a ride time charge of C$0.18 per minute, and a distance distance is assumed to be equal to the corresponding radii.
charge of C$0.81 per kilometer. Toll costs are based on the costs of using √̅̅̅̅̅̅̅̅̅̅̅̅̅
Highway 407, the only toll road in the greater Toronto Area, and outside Distanceintrazonal = Area/ (6)
π
of the City of Toronto study area used in this article. This is a suburban
east/west highway that circumvents travel through the non-tolled, more For the auto mode, the speed that is used to find intra-zonal travel
central, and congested arterials and highways. These costs are incor­ time of a given zone is assumed to be the mean speed of auto travel to its
porated into Uber fares, and so are incorporated into the GC for ride­ nearest five zones. Travel speed to 5 nearest neighbours was selected to
hailing as well. Contrary to transit wait time, ridehailing waiting time is smooth out localized affects due to varying zone shapes and sizes.
unweighted (i.e., weight = 1.0), considering that ridehailing is an on- Overall, the number of intrazonal opportunities for zones is small rela­
demand service and that customers typically do not have to wait for tive to extrazonal opportunities, especially in larger sized peripheral
their rides at a bus stop or a station. The auto O–D data used for ride­ zones where densities are quite low, but where our speed estimation
hailing includes in-vehicle travel time, distance, and toll charge to approach is likely least accurate. For transit, intrazonal trips are
produce a full O–D matrix. assumed to be made by foot at a walking speed of 5 km/hr. The average
intrazonal walk time was 11 min. As such, the intra-zonal distances of all
GCRidesharing = Fare + Toll Cost + (IVTT + Wait)*VOT (2) zones are computed then used to compute the other GC components for
each mode.
where:

Fare = 2.50 C$ [base fare] + 2.75 C$ [booking fee] + 0.18 C$/minute + 0.81 C$/kilometer (3)

5
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

4.2. Access profile analysis index in this study. X is the GC value up to which the area under the
curve is to be calculated (e.g., if we are interested in measuring acces­
The main output of APA is an access profile graph for each unit of sibility within C$30 of GC, X would be equal to 30). A(X) is the cumu­
analysis, in this case Traffic Analysis Zones (TAZ), showing one curve for lative proportion of jobs that can be reached within a GC value of X.
each mode. Each mode's curve on the graph represents the cumulative Although the cumulative number of accessible jobs (in absolute terms)
number of reachable jobs using that mode (on the vertical axis) across a was used in the plots as it was deemed a more meaningful representa­
range of GC values (on the horizontal axis) as exemplified in Fig. 1 tion, A(X) is calculated as a proportion of the total number jobs in the
above. A zone (and, therefore, the jobs in it) is deemed ‘reachable’ Greater Golden Horseshoe Area, resulting in smaller, more convenient
within a GC value, x, if the GC from the origin zone to that zone is less accessibility index values. Finally, xX is the weighted average of the GC
than or equal to x. When the APA data is calculated for all TAZs, it is then of traveling to jobs that are within a GC value of X.
aggregated by planning district (PD) and by Neighborhood Improve­ As such, the accessibility index between a GC of zero and X of a
ment Area designation. When aggregated, the zonal APA values are specific origin zone depends on two things: the number of reachable
weighted by their population. activities up to a budget of X and the difference between X and xX . This
way, the index gives more weight to closer jobs than it gives to further
4.3. Accessibility indices ones. As the weighted-average travel time to jobs (xX ) shifts to the left on
the GC axis, the difference between X and xX grows, and the accessibility
While access profiles are useful in comparing the accessibility that is index increases. In other words, as more jobs are reached within lower
provided by multiple modes in a geographic area, they are not useful in GC values, the accessibility index value increases, indicating better
directly comparing different areas. To overcome this limitation, we accessibility, and vice versa. The value of the threshold X, Black and
measure accessibility by calculating the area under the curves of access Conroy (Black and Conroy, 1977) suggest, should depend on whether
profiles (Black and Conroy, 1977; Neudorf, 2014). This area is used as an the scope of the analysis is local or regional accessibility.
evaluation of the overall performance of a particular mode in providing We set X to C$50 for regional accessibility because the APA curves
access to employment opportunities from a specific geographic area. for the auto mode begin leveling off at that point, meaning that there are
Since the curves are plotted using data points rather than mathe­ minimal accessibility benefits to driving beyond a GC of C$50 for most
matical equations, exact integration to obtain the area under the curve origins across Toronto and overall (see Fig. 4). To assess local accessi­
may not be possible. Alternatively, Black and Conroy (Black and Conroy, bility, we set X to half of the regional cut-off, C$26 (rounded up as we
1977) approximate the area under the curve using the following equa­ measure in $2 increments).
tion, which is derived from the integral of access profiles: Accessibility indices are computed for all TAZs, PDs, NIAs, non-NIA
( ) neighborhoods, and the whole city. To analyze the net benefit added by
Ki = A(X) X − xX (7) ridehailing, access profiles are calculated for the maximum of the two
modes (transit and ridehailing), whichever provides access to more jobs
in each GC increment. This represents the uppermost of the two curves
Ki is the measure of accessibility at zone i and is called the accessibility

Fig. 4. Access profiles of PD 1, PD 5, PD 9, NIAs, and non-NIAs.

6
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

for the whole range between zero and C$50. From this new access destinations within the $10 to $30 GC range. In contrast, the integrated
profile, accessibility indices are calculated then compared with those of mode tracks very closely with transit in every part of the city, revealing
transit to find the ridehailing benefit. This represents the benefit of that integration does not improve accessibility over what transit pro­
having the option of ridehailing (but not necessarily using it). vides by itself, mainly because of the dense and high-quality surface
feeder routes and the fare-free transfers to connect with rapid transit.
KMax(transit,rideshailing) − KTransit
Ridehailing Benefit (%) = *100 (8) NIAs experience lower accessibility across all modes, although they
KTransit
appear to gain a greater benefit than non-NIAs from the addition of
The integrated mode, by definition, is the maximum of transit alone ridehailing, at least at low GCs of between $10 and $22. Overall, ride­
and ridehailing-plus-transit, whichever provides better access. Thus, the hailing offers better accessibility than transit within the lower GC range,
benefit of integration can be calculated by directly comparing the and this improvement is greater in inner suburbs and NIAs.
accessibility indices of the integrated mode with those of the transit
mode.
5.1. Ridehailing from origin to destination
Acc.IndexIntegrated − Acc.IndexTransit
Integration Benefit (%) = *100 (9)
Acc.IndexTransit Notwithstanding the utility of the APA graphs, we undertake an
additional procedure to quantify the net impact of ridehailing on
5. Results and discussion accessibility. Accessibility indices are calculated at the TAZ level using
the curves of all modes (in addition to the maximum of the transit and
Ridehailing is analyzed by itself (ridehailing mode) and as a feeder to ridehailing curves). We begin with results for local access, defined as
the transit network (integrated mode). Fig. 4 provides an illustrative accessibility within or below a GC cost of C$26. We then move onto
sample of access profiles for different parts of Toronto, including PD 1 regional accessibility, or access within or below a GC cost of C$50.
(downtown), PD 5 (midtown), PD 9 (inner suburbs), all NIAs, and all
non-NIAs. We observe that the transit and integrated curves track closely 5.1.1. Local access
together in all profiles and for most of the GC range. This results from The disaggregate results for local access are mapped in Fig. 5 below.
public transit offering a lower GC compared to the ridehailing first mile Aggregate results for local access are summarized in Table 1. Fig. 5 and
for 93% of origin-destination pairs. Additionally, the number of reach­ the second from-the-right column in Table 1 refer to the accessibility
able jobs below a GC of C$10 is typically in the order of a few thousand, benefit that ridehailing provides over transit (see eq. 8 for calculation).
making the curves in that range seem close to zero considering the Ridehailing provides the greatest increase in local accessibility in
vertical access extends to 3 million jobs. A noticeable rise in the acces­ TAZs that straddle the edges of the city, while providing the lowest in­
sibility curves for all but the auto mode typically starts between C$8 and creases in areas adjacent to rapid transit. Ridehailing improves non-auto
C$12. The GC value before which no destinations are accessible is job access between zero to 500% across different TAZs. The largest
referred to as the barrier to entry (Neudorf, 2014), and is caused by concentration of neighborhoods benefiting from ridehailing are in the
factors such as fixed fare, parking costs at the destination, and expected city's west end. Three factors contribute to results there: close proximity
wait times. to the second largest employment center in the region (the airport, see
The auto-mode unsurprisingly provides greater access at each GC Fig. 2), the absence of rapid transit in most of these zones (see Fig. 3),
level compared to other modes. In Toronto's core (PD 1), and midtown and the fact that the airport and most of the jobs around it are located
(PD 5), ridehailing offers small improvements over transit at low GCs, or outside of Toronto, requiring transit commuters to these jobs to pay
nearby locations, but otherwise tracks below it. Ridehailing improves multiple fares.
upon transit the most in inner suburbs (PD 9), particularly for These dynamics are also apparent when accessibility benefits are
aggregated to the planning district scale, the results of which are

Fig. 5. Ridehailing's added benefit maps for local accessibility (GC under $26).

7
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Table 1
Accessibility indices for GC threshold of C$26 (local).
Geography Auto Transit Ridehailing alone Integrated Max of transit and ridehailing alone Ridehailing alone benefit (%) Integration benefit (%)

PD 1 6.31 2.30 2.50 2.31 2.65 14.94 0.41

PD 2 6.20 1.42 1.39 1.43 1.64 15.47 0.91
PD 3 6.48 0.78 0.84 0.80 0.99 26.91 2.69
PD 4 6.23 1.46 1.39 1.48 1.67 14.96 1.66
PD 5 5.70 0.45 0.73 0.46 0.77 72.94 3.00
PD 6 5.40 1.11 1.14 1.13 1.35 21.32 1.67
PD 7 5.87 0.23 0.39 0.23 0.42 86.37 1.64
PD 8 6.31 0.32 0.57 0.32 0.67 113.23 1.24
PD 9 5.97 0.18 0.55 0.18 0.55 206.96 0.00
PD 10 6.23 0.39 0.63 0.39 0.70 80.01 0.01
PD 11 6.17 0.89 0.70 0.90 1.09 22.51 0.93
PD 12 5.75 0.53 0.77 0.53 0.84 58.91 0.07
PD 13 4.65 0.52 0.55 0.52 0.70 34.23 0.10
PD 14 4.13 0.21 0.36 0.21 0.38 74.82 0.13
PD 15 3.54 0.11 0.24 0.11 0.24 114.45 0.00
PD 16 4.55 0.28 0.57 0.28 0.57 104.67 0.02
Citywide 5.71 0.86 0.97 0.87 1.12 29.90 1.04
NIA 5.54 0.45 0.63 0.45 0.70 56.97 0.72
Non-NIA 5.76 0.97 1.06 0.98 1.23 26.53 1.09

Fig. 6. Ridehailing's added benefit maps for regional accessibility.

provided in Table 1. For local access, ridehailing improves non-auto job 5.1.2. Regional access
access between 15 and 200% across the city's planning districts. Table 1 For regional access, ridehailing improves the non-auto accessibility
confirms the visual patterns present in the APA curves and TAZ analysis: of TAZs between 0 and 30%, as plotted in Fig. 6. This benefit is signif­
the downtown and surrounding core receive the lowest benefit from the icantly lower than the benefits in local access, reflecting how ridehailing
addition of ridehailing, while the inner suburbs experience the greatest. provides better access than transit at low generalized costs along the
PD 9, which encompasses the northwestern corner of the city, sees a APA curve. As with local access, ridehailing provides the greatest benefit
benefit of 206%, nearly seven times the city's average. This reflects how to TAZs along the western and eastern ends of the city. Unlike with local
ridehailing provides this area with access to the nearby airport at a lower access, ridehailing increases regional access for several central TAZs as
GC than public transit. well.
Citywide, NIA residents experience a greater benefit from ridehailing At the planning district scale, ridehailing's impact on regional access
than non-NIA residents, a result due in part to the over-representation of appears more modest, ranging between 2.39% and 10.66%. These
NIAs in inner suburbs with no rapid transit. This suggests that ride­ values are far less than the local access benefit (15%–207%), reflecting
hailing's accessibility benefits are equitably distributed across the city. the inferiority of ridehailing at higher generalized costs. As with local
Nevertheless, NIAs have lower overall accessibility even with the ride­ access, PD 9 benefits the most in regional access.
hailing option in place. This highlights the need to improve transit Regional access improvements from ridehailing are equitable across
services in the NIAs to close the gap with other city neighborhoods with the city. NIAs show an improvement of 6%, higher than both non-NIAs
respect to job accessibility. and the city overall. As with local access, NIAs have lower transit
accessibility than non-NIAs, and ridehailing does not close this gap.

8
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Fig. 7. Histograms of distance to nearest rapid transit stations among top 20% of TAZs benefiting in local accessibility (top left) and regional accessibility (top right),
and bottom 20% of TAZs benefiting in local accessibility (bottom left) and regional accessibility (bottom right).

Table 2
Accessibility indices for GC threshold of C$50 (regional).
Geography Auto Transit Ridehailing alone Integrated Max of transit and ridehailing alone Ridehailing alone benefit (%) Integration benefit (%)

PD 1 22.25 12.31 11.03 12.41 12.68 3.00 0.79

PD 2 21.71 11.09 10.07 11.21 11.36 2.39 1.09
PD 3 21.79 9.71 10.14 9.85 10.47 7.83 1.36
PD 4 21.72 11.09 10.67 11.22 11.51 3.73 1.12
PD 5 20.77 8.39 8.64 8.49 9.04 7.83 1.29
PD 6 20.90 9.86 8.83 9.99 10.13 2.78 1.33
PD 7 20.65 7.77 7.41 7.90 8.30 6.77 1.68
PD 8 21.07 8.76 8.38 8.85 9.41 7.45 1.00
PD 9 20.19 7.82 6.98 7.82 8.66 10.66 0.03
PD 10 20.69 9.23 8.30 9.23 9.76 5.76 0.03
PD 11 20.85 10.60 9.01 10.66 10.92 3.00 0.62
PD 12 20.26 9.11 7.82 9.13 9.47 3.95 0.18
PD 13 19.23 8.29 6.18 8.30 8.50 2.52 0.16
PD 14 18.77 7.16 5.59 7.18 7.34 2.50 0.17
PD 15 17.01 4.79 3.07 4.79 5.05 5.36 0.06
PD 16 18.32 6.42 5.11 6.44 6.90 7.48 0.27
City 20.63 9.35 8.39 9.43 9.79 4.68 0.80
NIA 20.20 8.46 7.56 8.50 8.97 6.00 0.41
Non-NIA 20.75 9.59 8.62 9.68 10.01 4.37 0.89

TAZ distance to the subway helps explain the patterns in Figs. 5 and and regional). From these distributions, it appears that ridehailing offers
6. For both local (X = C$26) and regional (X = C$50) access, we notice the highest benefit for zones that are between 1 and 7 km for both local
that, generally, there are two groups of zones that experience a low and regional access.
ridehailing benefit: those that are located far away from subway lines
(and from jobs), and those that are located very close to them. In be­
5.2. Integrating ridehailing and transit
tween, there is a range of distances where ridehailing is most beneficial.
This range is determined by looking at the zones with the highest ride­
Since the access profile of the integrated mode is very close to that of
hailing benefit values and plotting a histogram of the distances between
transit in all geographies, it appears that integrating ridehailing with
these zones and their nearest rapid transit stations (see Fig. 7). The top
transit does not considerably improve access. This is reflected in the
and bottom 20% of zones are used for each type of accessibility (local
accessibility indices (see Tables 1 and 2) that show a minimal

9
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Fig. 8. Integration's added benefit maps for local (top) and regional (bottom) accessibility.

integration benefit in comparison with the benefit of ridehailing alone. employment (such as in PDs 1, 9, and 10).
The results also tell us that the integrated mode is generally more
beneficial for origins that are closer to rapid transit lines. However, PD 1 6. Limitations
benefits less from integration than some of the other PDs that are not as
close to rapid transit as PD 1 is, as will be discussed below. Having the Our method contains several limitations that can be addressed
option to integrate ridehailing and transit together does not reduce the through further research. First, our method is sensitive to the value of
disparity in access between NIAs and other zones – neither for local nor time used in the analysis. Value of time generally increases with income
regional destinations. This can be seen in the higher integration benefit (Börjesson et al., 2012), and so our question of equity for under-
that is experienced by non-NIAs compared to NIAs and by comparing resourced communities led us to apply a low value of time relative to
accessibility indices of ‘integrated’ and transit. other studies (Alonso-González et al., 2020, p.; Lazarus et al., 2021;
The maps in Fig. 8 summarize the disaggregate (TAZ-level) added Schwieterman, 2019). This decision penalized ridehailing for its higher
benefit of integration for local and regional access. Similar to the ride­ fares, but reflects observed behaviour in that low-income household are
hailing mode, integration improves access to employment locally more less likely to use this mode (Young and Farber, 2019). Practitioners
than it does regionally. It appears that integration benefit depends more interested in accessibility across all income groups can deploy a heter­
on the location relative to rapid transit and less on the proximity to ogenous VOT to our method by weighting an average VOT by each or­
employment as no benefit is seen near high concentrations of igin's average income.

10
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

Second, our analysis cannot capture other barriers and disincentives Deep subsidies would be required to make ridehailing work as feeder
to ride-hailing. For example, Toronto transit fares include a two-hour mode to rapid rail from the more remote origins across the city, areas
free transfer window that might discourage under-resourced riders that include a significant number of NIAs. From the disaggregate
from switching to ridehailing. Nor does our analysis account for surge regional accessibility indices, it seems that the benefit of integration is
pricing, which impacts ridehailing at certain times of day and in certain higher for origins that are between one and seven kilometers away from
places. In-vehicle travel times are also higher for shared services like rapid rail. Moreover, integration neither closes nor reduces the access
UberPool (Young et al., 2020b), which our current model does not ac­ gap that exists between NIAs and non-NIAs.
count for.
Most of these limitations present opportunities to extend our
approach to policy applications around subsides, surge pricing, and Acknowledgement
pooling incentives. The generalized cost formulas can be readily adapted
to evaluate the benefits of ridehailing subsidies, and the impacts to This work was supported by an XSeed Grant from the University of
accessibility of surge pricing and shared services. Toronto and an Early Researcher Award from the Ontario Ministry of
Research, Innovation and Science.
7. Conclusions
References
This paper offers an advancement in accessibility analysis for policy
by using a multi-modal, generalized cost approach to measure the equity Alonso-González, M.J., van Oort, N., Cats, O., Hoogendoorn-Lanser, S., Hoogendoorn, S.,
implications of a new mode of travel. This technique is rare, if not new, 2020. Value of time and reliability for urban pooled on-demand services. Transport.
Res. Part C Emerg. Technol. 115, 102621. https://doi.org/10.1016/j.
in both the accessibility and equity literatures. Our application dem­ trc.2020.102621.
onstrates how this approach can yield new insights by standardizing Barajas, J.M., Brown, A., 2021. Not minding the gap: does ride-hailing serve transit
accessibility benefits across modes that offer riders clear trade-offs in deserts? J. Transp. Geogr. 90, 102918. https://doi.org/10.1016/j.
jtrangeo.2020.102918.
time versus monetary costs. We also show how our method enables Black, J., Conroy, M., 1977. Accessibility measures and the social evaluation of urban
measurement of the accessibility benefits of a combination of modes structure. Environ. Plan. A Econ. Space 9, 1013–1031. https://doi.org/10.1068/
over and above the benefits provided by either mode alone. The multi- a091013.
Börjesson, M., Fosgerau, M., Algers, S., 2012. On the income elasticity of the value of
modal GC approach will be valuable for researchers seeking to mea­
travel time. Transp. Res. A Policy Pract. 46, 368–377. https://doi.org/10.1016/j.
sure the accessibility impacts of new transportation technologies as they tra.2011.10.007.
become more widespread and integrated into existing transportation Brown, A.E., 2018. Ridehail Revolution: Ridehail Travel and Equity in Los Angeles.
University of California Los Angeles.
systems.
Brown, A., 2019. Redefining car access: ride-hail travel and use in Los Angeles. J. Am.
Overall, this study finds that transit and ridehailing accessibility Plan. Assoc. 85, 83–95. https://doi.org/10.1080/01944363.2019.1603761.
levels are similar when measured by generalized cost. Despite the CAA, 2013. Driving Cost. Canadian Automobile Association, Ottawa, Canada.
overall similarity in accessibility, we find that ridehailing, when used in City of Toronto, 2014. TSNS 2020 Neighborhood Equity Index Methodological
Documentation. Social Policy Analysis and Research, City of Toronto, Toronto,
isolation, improves accessibility in neighborhoods characterized by poor Ontario.
transit service but close proximity to peripheral employment centers. In City of Toronto, 2015. Toronto Strong Neighbourhoods Strategy 2020.
Toronto, ridehailing alone extends transit's local accessibility by 15%– Conway, M., Salon, D., King, D., 2018. Trends in taxi use and the advent of ridehailing,
1995–2017: evidence from the US National Household Travel Survey. Urban Sci. 2,
200% and regional accessibility by 2.4%–10.7%. Toronto's transit-rich 79. https://doi.org/10.3390/urbansci2030079.
downtown experiences the least benefit from the addition of rideshar­ Curtis, T., Merritt, M., Chen, C., Perlmutter, D., Berez, D., Ellis, B., 2019. Partnerships
ing while those in PD 9, an inner suburban area with poor transit con­ Between Transit Agencies and Transportation Network Companies. https://doi.org/
10.17226/25425.
nectivity to a nearby employment hub, benefit the most. Data Management Group – Transportation Tomorrow Survey [WWW Document], n.d.
The accessibility benefits of ridehailing are greater in NIAs, meaning URL http://dmg.utoronto.ca/transportation-tomorrow-survey (accessed 11.11.18).
they are vertically equitable. On average, ridehailing improves access to de Oliveira Souza, C., Vitorino Guimarães, G., da Cruz Saldanha, L.E., Almeida Corrêa do
Nascimento, F., Floriano dos Santos, T., Vieira da Silva, M.A., 2021. Analysis of job
jobs for residents of Toronto's NIAs by 30% and 5% for local and regional accessibility promoted by ride hailing services: a proposed method. J. Transp. Geogr.
jobs, respectively. Zones that are in Neighborhood Improvement Areas 93, 103048. https://doi.org/10.1016/j.jtrangeo.2021.103048.
(NIAs) benefit from ridehailing substantially more than their non-NIA Delbosc, A., Currie, G., 2011. Using Lorenz curves to assess public transport equity.
J. Transp. Geogr. 19, 1252–1259. https://doi.org/10.1016/j.jtrangeo.2011.02.008.
counterparts, suggesting that ridehailing disproportionately improves
Dong, X., 2020. Trade Uber for the bus?: an investigation of individual willingness to use
accessibility for these communities. However, the total level of non ride-hail versus transit. J. Am. Plan. Assoc. 86, 222–235. https://doi.org/10.1080/
private auto accessibility remains lower in NIAs compared to non-NIAs, 01944363.2019.1687318.
highlighting the need for further transit service improvements in NIAs. El-Geneidy, A., Levinson, D., Diab, E., Boisjoly, G., Verbich, D., Loong, C., 2016. The cost
of equity: assessing transit accessibility and social disparity using total travel cost.
Many agencies are exploring options to coordinate or subsidize Transp. Res. A Policy Pract. 91, 302–316. https://doi.org/10.1016/j.
ridehailing as a first-mile service for transit (Curtis et al., 2019). Our tra.2016.07.003.
findings cast doubt on the value of these investments in the City of Foth, N., 2013. Transit Accessibility, Social Equity & Job Type.
Gehrke, S.R., Felix, A., Reardon, T.G., 2019. Substitution of ride-hailing services for more
Toronto, where the TTC's suburban bus routes are among the most sustainable travel options in the greater Boston region. Transp. Res. Rec. 2673,
heavily used in North America, and where the existing subsidy for bus 438–446. https://doi.org/10.1177/0361198118821903.
trips is low by North American standards. While we did not conduct a Hall, J.D., Palsson, C., Price, J., 2018. Is Uber a substitute or complement for public
transit? J. Urban Econ. https://doi.org/10.1016/j.jue.2018.09.003.
cost comparison between subsidies required for transit trips vs. ride­ Hu, L., 2015. Job accessibility of the poor in Los Angeles. J. Am. Plan. Assoc. 81, 30–45.
hailing trips, in the Toronto case, our research suggests that the subsidy https://doi.org/10.1080/01944363.2015.1042014.
required will be far higher than those required for an average bus trip on Isserman, A.M., 1977. The location quotient approach to estimating regional economic
impacts. J. Am. Inst. Plann. 43, 33–41. https://doi.org/10.1080/
the TTC. This finding may not hold for operators with higher subsidies 01944367708977758.
per bus trip, or even for the TTC on some of the higher cost routes and Jiao, J., Wang, F., 2021. Shared mobility and transit-dependent population: a new equity
off-peak operating hours (see for example: Toronto Transit Commission, opportunity or issue? Int. J. Sustain. Transp. 15, 294–305. https://doi.org/10.1080/
15568318.2020.1747578.
2020, p. 19). The finding is also sensitive to the specific way that we
Karner, A., 2018. Assessing public transit service equity using route-level accessibility
modeled the integrated mode, replacing bus access to rapid transit by measures and public data. J. Transp. Geogr. 67, 24–32. https://doi.org/10.1016/j.
lone-rider ridehailing access, as opposed to solutions that allow for ride jtrangeo.2018.01.005.
and cost sharing among passengers. In 93% percent of TAZs, using King, D.A., Smart, M.J., Manville, M., 2019. The poverty of the carless: toward universal
auto access. J. Plan. Educ. Res. https://doi.org/10.1177/0739456X18823252.
transit alone is more beneficial than integrating ridehailing and transit, Lazarus, J.R., Caicedo, J.D., Bayen, A.M., Shaheen, S.A., 2021. To pool or not to pool?
at least for riders whose value of time approximates the minimum wage. Understanding opportunities, challenges, and equity considerations to expanding the

11
B. Abdelwahab et al. Journal of Transport Geography 95 (2021) 103147

market for pooling. Transp. Res. A Policy Pract. 148, 199–222. https://doi.org/ Illinois. Transp. Res. Rec. 2673, 295–304. https://doi.org/10.1177/
10.1016/j.tra.2020.10.007. 0361198119839344.
Lee, J., 2019. Exploring the Accessibility Gap: Quantifying Transport Disadvantage in the Shaheen, S., Bell, C., Cohen, A., Yelchuru, B., 2017. Travel Behavior: Shared Mobility and
City of Toronto. https://doi.org/10.1017/CBO9781107415324.004. Transportation Equity (Technical Report). FHWA.
Li, W., 2019. Empirical Analysis of the Relationship Between Ride- Hailing and Public Statistics Canada, 2017. Census in Brief: Municipalities in Canada with the Largest and
Transit in Toronto. Thesis. University of Toronto, Toronto. Fastest-Growing Populations between 2011 and 2016 [WWW Document]. URL.
Li, W., Shalaby, A., Habib, K.N., 2021. Exploring the ridership impacts of ride-hailing on https://www12.statcan.gc.ca/census-recensement/2016/as-sa/98-200-x/201600
multimodal public transit in Toronto. Transportation 1–25. 1/98-200-x2016001-eng.cfm.
Martens, K., 2016. Transport Justice: Designing Fair Transportation Systems. Routledge. Statistics Canada, n.d. Table 14-10-0306-01 Employee Wages by Occupation, Monthly,
Martens, K., Golub, A., 2018. A fair distribution of accessibility: interpreting civil rights Unadjusted for Seasonality [WWW Document]. URL https://www150.statcan.gc.
regulations for regional transportation plans. J. Plan. Educ. Res. https://doi.org/ ca/t1/tbl1/en/tv.action?pid=1410030601 (accessed 8.14.19).
10.1177/0739456X18791014. Toronto Transit Commission, 2020. 2021 Annual Service Plan. Chief Strategy and
Neudorf, J., 2014. Understanding Accessibility, Analyzing Policy: New Approaches for a Customer Officer. Toronto Transit Commission, Toronto, Ontario.
New Paradigm. University of Waterloo. Verbich, D., El-Geneidy, A., 2017. Public transit fare structure and social vulnerability in
Palm, M., Farber, S., Shalaby, A., Young, M., 2020. Equity analysis and new mobility Montreal, Canada. Transp. Res. A Policy Pract. 96, 43–53. https://doi.org/10.1016/
technologies: toward meaningful interventions. J. Plan. Lit. https://doi.org/ j.tra.2016.12.003.
10.1177/0885412220955197, 088541222095519. Young, M., Farber, S., 2019. The who, why, and when of Uber and other ride-hailing
Pan, R., Yang, H., Xie, K., Wen, Y., 2020. Exploring the equity of traditional and ride- trips: an examination of a large sample household travel survey. Transp. Res. A
hailing taxi services during peak hours. Transp. Res. Rec. J. Transp. Res. Board. Policy Pract. 119, 383–392. https://doi.org/10.1016/j.tra.2018.11.018.
https://doi.org/10.1177/0361198120928338. Young, M., Allen, J., Farber, S., 2020a. Measuring when Uber behaves as a substitute or
Pereira, R.H.M., Schwanen, T., Banister, D., 2017. Distributive justice and equity in supplement to transit: an examination of travel-time differences in Toronto.
transportation. Transp. Rev. https://doi.org/10.1080/01441647.2016.1257660. J. Transp. Geogr. 82 https://doi.org/10.1016/j.jtrangeo.2019.102629.
Schwieterman, J.P., 2019. Uber economics: evaluating the monetary and travel time Young, M., Farber, S., Palm, M., 2020b. The true cost of sharing: a detour penalty
trade-offs of transportation network companies and transit service in Chicago, analysis between UberPool and UberX trips in Toronto. Transp. Res. Part D: Transp.
Environ. 87, 102540. https://doi.org/10.1016/j.trd.2020.102540.

12