Eliciting Preferences of Ride-Hailing Users and Drivers 2
Eliciting Preferences of Ride-Hailing Users and Drivers
About KAPSARC
Legal Notice
The King Abdullah Petroleum Studies and Research Center (KAPSARC) is a
non-prot global institution dedicated to independent research into energy economics,
policy, technology and the environment across all types of energy. KAPSARC’s
mandate is to advance the understanding of energy challenges and opportunities
facing the world today and tomorrow, through unbiased, independent, and high-caliber
research for the benet of society. KAPSARC is located in Riyadh, Saudi Arabia.
This publication is also available in Arabic.
© Copyright 2020 King Abdullah Petroleum Studies and Research Center (“KAPSARC”).
This Document (and any information, data or materials contained therein) (the
“Document”) shall not be used without the proper attribution to KAPSARC. The
Document shall not be reproduced, in whole or in part, without the written permission
of KAPSARC. KAPSARC makes no warranty, representation or undertaking whether
expressed or implied, nor does it assume any legal liability, whether direct or indirect,
or responsibility for the accuracy, completeness, or usefulness of any information that
is contained in the Document. Nothing in the Document constitutes or shall be implied to
constitute advice, recommendation or option. The views and opinions expressed in this
publication are those of the authors and do not necessarily reect the ofcial views or
position of KAPSARC.
3
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Systemic transformations are taking place in passenger auto travel. The rise of transportation
network companies (TNCs) such as Uber and Lyft has begun to fundamentally alter key aspects
of individuals’ behavior including vehicle ownership, distance traveled, adoption of alternative fuel
vehicles and use of public transit. Such shifts will have major impacts on light-duty passenger vehicles,
which account for one-quarter of global oil demand, more than any other sector. To assess the system-
level impacts of TNCs, the ‘micro-decisions’ of users and drivers need to be better understood. This study
contributes to the nascent eld by using a large, unique sample to estimate a) individuals’ preferences for
being a rider, driver, or non-user of TNC services; b) the propensity of TNC users for ‘ride-pooling;’ c) TNC
drivers’ likelihood of switching to vehicles with better fuel economy; and d) the drivers’ decisions to buy, rent
or lease new vehicles with driving for TNCs as a major consideration. We analyze a sample (N=11,902) of
survey respondents in TNC-served areas in the United States, providing a case study for other countries,
including Saudi Arabia, where personal auto travel is the dominant form of transportation.
The key ndings of this study:
A population-weighted statistical analysis indicates that ride-hailing services are mainly attracting
personal vehicle users as riders, without substantially affecting the demand for public transit. Among
ride-hailing users, around 10% reported postponing the purchase of a new car due to the availability of
TNC services.
Key Points
70
60
50
40
30
20
10
0
Ride-hailing Drive personal
vehicle
Public
transit
Walk or bike Car-sharing or
car-pooling
Would not
make the trip
Ride-hailing Drive personal vehicle Public transit Walk or bike Car-sharing or car-pooling
%
%
%
%
%
%
%
%
Mobility option used if the primary option is not available
Source: KAPSARC analysis.
Eliciting Preferences of Ride-Hailing Users and Drivers 4
Eliciting Preferences of Ride-Hailing Users and Drivers
Key Points
Fitting a multinomial logistic regression model on the propensity of being a rider, driver or non-user
of TNC services, we nd that the likelihood of being a TNC user rises according to age for someone
younger than 44 years, but the pattern reverses for those 44 years and older.
In terms of ‘ride-pooling,’ older TNC users with higher household vehicle ownership who live in
suburban areas are less likely to choose to ride pool.
The majority (65%) of TNC drivers who drive daily for TNCs indicated that their occupation was a major
consideration in vehicle purchase decisions.
Households with drivers who hold postgraduate degrees, drive daily and live in metropolitan areas are
more likely to switch to fuel-efcient vehicles.
5
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Summary
The ‘ride-hailing’ services offered by
transportation network companies (TNCs)
such as Uber and Lyft have rapidly disrupted
personal transportation, particularly in cities.
Schaller (2018) reports that TNCs provided 2.6
billion rides in 2017 in the United States (U.S.), a
37% increase from 2016. The rapid increase in the
adoption of TNC services can be attributed to the
ease of access offered by smartphone applications
and the higher availability of cars and drivers
compared to regulated, traditional taxi services.
Proponents of TNCs emphasize that ride-hailing
services offer consumers alternative transport
options, especially in high-demand areas with
unreliable and inefcient transit services (Alemi et
al. 2018). They also assert that TNCs allow users
to give up vehicle ownership and/or reduce vehicle
usage—not only by replacing personal vehicle use
but also by facilitating multi-modal travel (e.g., ride-
hailing as a means to reach public transportation).
Furthermore, TNCs offer options for ‘ride-pooling,’
in which multiple customers share a single vehicle;
this has immense potential to reduce vehicle miles
traveled (VMT) and thus greenhouse gas emissions
and trafc congestion (Martin and Shaheen 2011).
However, a few studies have highlighted the adverse
effects of TNC services, including induced demand
for travel and a reduction in public transit ridership in
certain areas.
Transportation planners and policymakers are
uncertain about the impacts of TNC services
on energy use, the environment and trafc
problems (Conway et al. 2018). Declines in VMT
and vehicle ownership can reduce greenhouse
gas emissions (as established in ‘car-sharing’
literature), but induced demand can negate
such gains. These system-level impacts are
manifestations of individual-level decisions, both
by customers—namely how frequently to use ride-
hailing and whether the rides replace driving or
public transit—and by TNC drivers, such as how
fuel-efcient a vehicle to drive. However, in the
absence of sufcient data on individual preferences,
transportation authorities and other stakeholders are
unable to quantify the systemic effects of TNCs.
This study takes an important step toward bridging
this gap by analyzing a survey of TNC riders and
drivers conducted by Strategic Vision Incorporated,
consisting of a large sample (N=11,902) of the
U.S. population residing in TNC-served areas. The
survey included information on sociodemographic
characteristics, personal attitudes toward adopting
TNC services as a user or driver, and changes in
travel mode and vehicle ownership preferences after
using these services. Using this revealed preference
data, we investigate the associations between
sociodemographic characteristics and the following:
a) preferences for being a rider, driver, or non-user
of TNC services; b) ride-hailing users’ propensity
for ride-pooling; c) drivers’ choices to switch to
vehicles with better fuel economy; and d) drivers’
decisions to buy, rent or lease a new vehicle, with
driving for TNCs as a major contributing factor. We
t multinomial logistic regressions to answer the rst
question and binary logistic regressions to answer
the remaining three. We also observe non-linear
relationships by accounting for interaction effects of
continuous covariates (e.g., age and income) with
binary covariates (e.g., gender) while estimating
preferences.
Results indicate that younger individuals who have
higher levels of education, live in metropolitan
areas, and belong to more afuent families are
more likely to use ride-hailing services. However,
the relationship between the individuals’ probability
of being a rider, driver or non-user of TNC services
and their age is downward parabolic: it increases
until the age of 44 and then decreases. Households
Eliciting Preferences of Ride-Hailing Users and Drivers 6
Eliciting Preferences of Ride-Hailing Users and Drivers
that own more vehicles are less likely to be either
riders or drivers. Exploring the inclination of the ride-
hailing users to ride-pool, we nd that older travelers
with higher household vehicle ownership who live in
suburban areas are less likely to pool rides. In terms
of interaction effects, females with an education
level below postgraduation are more likely to ride-
pool than their male counterparts if they are younger
than 34 years, but this pattern reverses if they are
over 54.
Younger and married TNC drivers who drive daily
and own a higher number of vehicles are more likely
to switch to fuel-efcient vehicles, other things
being equal. These results are consistent with
previous studies eliciting preferences for electric
vehicles. Further, interaction effect estimates
reveal that postgraduate drivers who live in
metropolitan areas view fuel efcient vehicles
more favorably if they are aged below 48 years.
Finally, younger and lower income drivers have a
higher probability of considering driving for TNCs
whilst purchasing a car.
Summary
7
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Introduction
This section summarizes literature relevant
to individuals’ preferences surrounding the
use of transportation network company
(TNC) services and the subsequent impacts
on their mobility decisions. We rst discuss the
evolution of TNCs and proceed to describe the
sociodemographic and geographic characteristics of
individuals who have a higher tendency to use TNC
services. We then review the literature on how these
services are changing the landscape of urban travel
patterns by affecting vehicle ownership preferences
and demand for other travel modes. We conclude
with a review of studies focusing on ride-hailing
drivers, followed by highlighting the research gap
that this study addresses.
Evolution of TNC services
In recent years, transportation has undergone an
unprecedented transformation due to the rapid
deployment of new technologies such as the
internet and smartphones (Taylor et al. 2015). These
advancements are the main drivers of the explosive
growth of TNCs. As of 2016, ride-hailing services
were active in almost 500 cities in the U.S. (Murphy
2016). However, a 2015 survey of 4,787 American
adults by the Pew Research Center found that only
3% and 12% of TNC riders use these services on a
daily or weekly basis, respectively (Smith 2016).
Characteristics of TNC users
Several correlation studies have identied
characteristics of travelers with a higher propensity
to use ride-hailing services. A survey conducted
across seven major cities in the U.S. found that
college-educated individuals adopt TNC services
at almost twice the rate of those without a college
degree (Clewlow and Mishra 2017). Furthermore,
travelers younger than 29 years and older than 65
years were shown to be the most and least frequent
users of ride-hailing services, respectively. Kooti et al.
(2017) drew similar conclusions from data provided
by Uber for 59 million rides and 4.1 million riders
collected over seven months. The authors observed
that younger riders are more likely to take frequent,
shorter rides, while older travelers are more inclined
toward infrequent, longer rides. In another study,
Alemi et al. (2018) modeled individuals’ lifestyles
using the California Millennials Dataset to identify
factors affecting the adoption of ride-hailing services.
The results of this study indicated that highly
educated, independent millennials who live in core
urban areas without owning personal vehicles and
who do not have children have the highest adoption
rate. Individuals with variety-seeking and technology-
embracing attitudes are more likely to use ride-hailing
services.
Geographic context and built environment factors
also play important roles in determining the usage
frequency of ride-hailing. Clewlow and Mishra (2017)
found that the adoption rate of these services is
higher in urban areas (29%) than in suburban ones
(15%). Alemi et al. (2018) also observed a positive
association between the demand for TNC services
and the urbanization of the neighborhood. Along
the same lines, the Pew Research Center study
found that users of these services most frequently
live in urban (21% of respondents) or suburban
(15%) areas (Smith 2016). In a more recent study,
Yu and Peng (2019) investigated the relationship
between characteristics of the built environment
and ride-hailing demand using 2016-2017 trip data
from RideAustin, a local TNC company. The results
support the ndings of previous studies and also
indicate that population density is a good predictor
of ride-hailing demand. Moreover, areas with greater
road and sidewalk densities are more likely to have
a higher demand for TNC services.
8
Eliciting Preferences of Ride-Hailing Users and Drivers
Impact of TNCs on vehicle
ownership and public transit
ridership
The availability of TNC services is likely to decrease
household vehicle ownership, but the extent of
the reduction is not clear. The American Public
Transportation Association (2016) reported that
ride-hailing users are more likely to own fewer
cars. Similarly, Conway et al. (2018) used National
Household Travel Survey data to examine the
expansion of ride-hailing services within the U.S.
and concluded that users are more likely to be
multimodal, own fewer cars and use alternative
modes of transportation. This is supported by a
survey conducted in Austin, Texas by Hampshire et
al. (2017), who found that 9% of ride-hailing users
purchased a vehicle after a suspension of these
services, and 45% of the TNC trips were replaced
by driving.
The impact of TNC services on public transit is also
unclear. Sadowsky and Nelson (2017) implemented
a regression discontinuity design to measure the
effect of TNCs on public transportation across
28 major U.S. cities. The authors found that the
introduction of Uber led to an increase in the use of
public transportation. However, the introduction of
Lyft after a few months had a negative impact on
the use of public transportation. They hypothesized
that the competition between these TNCs led to
a decrease in cost and wait time, causing more
users to prefer these services to public transport.
Dias et al. (2018) analyzed around one million trips
by RideAustin and found that, although individuals
living in neighborhoods with relatively poor access
to public transit are more inclined to use ride-
hailing services, there is a synergy between public
transport and ride-hailing in other areas. In another
study, Barbar et al. (2017) used a difference-in-
difference design to quantify the impact of ride-
hailing services and observed a signicant decrease
in road-based public transit services, especially
in areas with poor public transport coverage, but
an increase in the use of subways and commuter
rail. Hall et al. (2018) also adopted a difference-
in-difference design and found that Uber is, on
average, complementary to public transit; however,
the impact on public transport ridership is positive in
larger cities but negative in smaller ones.
Preferences of TNC drivers
Studies have touched upon driver safety (Feeney
2015), wages (Berger and Frey 2017), and
sociodemographic characteristics of TNC drivers
(Kooti 2017; Hall and Krueger 2018), as well as the
willingness of individuals to become ride-hailing
drivers (Berliner and Tal 2018). Hall and Kruger
(2018) analyzed data from two surveys of Uber
drivers in the U.S. that were conducted in December
2014 (N=601) and November 2015 (N=632). They
found that 30% were aged between 30 and 39,
47.7% had college or graduate degrees, and 14%
were women. The authors also observed that more
Uber drivers were single than married, and married
drivers tended to have children. Berliner and Tal
(2018) estimated the willingness of individuals
to drive for TNCs using stated preference data
collected in Irvine, California in 2015. They found
that the desire to earn extra income and a fondness
for driving were the two most common motivations
among respondents willing to drive for ride-hailing
services. Berliner and Tal also conclude that age,
number of children, vehicle ownership, gender, and
positive attitudes toward ride-hailing are signicant
predictors of a willingness to become a TNC driver.
Research gaps and contributions
This research builds upon prior studies on the
preferences of riders and drivers for TNC services
using revealed-preference, representative data for
Introduction
9
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Introduction
TNC-served areas in the U.S. First we expand the
literature on understanding the socio-demographic
characteristics of TNC riders and drivers. Second,
we determine the demographic segments of TNC
users that are most likely to be interested in ride-
pooling, which has received limited attention in
the literature (Lavieri and Bhat, 2018). However,
it is crucial to the timely and effective deployment
of pooling services, which make TNCs more
environmentally viable. Third, we identify TNC
drivers with a high propensity to switch to fuel-
efcient cars, to further understand the impact of
ride-hailing services on greenhouse gas emissions.
Finally, we elicit preferences of TNC drivers to buy,
rent or lease new vehicles with driving for TNCs as
a major purchase consideration, a phenomenon that
has not been explored in the existing literature.
Eliciting Preferences of Ride-Hailing Users and Drivers 10
Eliciting Preferences of Ride-Hailing Users and Drivers
Data and Summary Statistics
Data collection and weight
computation
We use data from a survey conducted by Strategic
Vision Incorporated in 2017 of 11,902 ride-hailing
riders, drivers, and non-users in TNC-served
areas in the United States. Figure A1 in the
Appendix shows the geographical distribution of
the respondents across the contiguous 48 states.
The survey data provides information on household
characteristics and attitudes, as well as on ride-
hailing usage and preferences. The former includes
age, gender, marital status, education level,
household income, ethnicity, residential location,
mode(s) of commuting, and household size.
The sample under- or over-represents some
demographic groups. For example, women above
the age of 49 living in metropolitan or urban areas
with annual incomes under $100,000 who use public
or non-motorized transport are under-represented
in the sample, while men above 54 living in small
towns with annual incomes greater than $100,000
are over-represented. To address this concern, we
estimate person-level weights using the iterative
proportional tting (IPF) technique (Bergmann 2011).
IPF matches the joint probability distribution of
various demographic characteristics in the collected
sample and the population-level datasets on urban
and rural housing units (2016 American Community
Survey data and 2010 U.S. Census Bureau data).
In other words, we compute weights by scaling the
survey sample proportions, in four demographic
classes and 32 categories (four gender- and age-
based, two income-based, two travel mode-based,
and two residence location-based groups), relative
to the corresponding class-specic proportions in
the population-level data. We implement this IPF
method using the ‘ipfweight’ package in Stata. The
estimated weights vary between 0.15 and 4.65. All
results presented in this paper are based on this
weighted sample.
Explanatory and response
variables
Table 1 summarizes the statistics of the population-
weighted explanatory variables used in all logistic
regression models of this study. The sample
statistics are consistent with those of the population.
Explanatory variables NMean Median SD Min. Max.
Male indicator 8,791 0.48 00.50 0 1
Single indicator 8,791 0.35 00.48 0 1
Age (in years) 8,791 46.77 47 15.73 18 100
Post-graduation indicator 8,791 0.31 00.46 0 1
Annual income ($) 8,791 94,909 72,500 90,915 15,000 1,000,000
Metropolitan resident indicator 8,791 0.30 00.46 0 1
Household size 3+ indicator 8,025 0.32 00.47 0 1
Total vehicle ownership 8,791 2.15 21.17 0 6
Early adopter indicator 8,791 0.19 00.39 0 1
Drive daily indicator 1,541 0.25 00.43 0 1
Table 1. Key explanatory variables.
Source: KAPSARC.
11
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Data and Summary Statistics
Response variables (indicators) NMean SD
Model 1
TNC driver 8,791 0.29 0.45
TNC user 8,791 0.29 0.45
TNC non-user 8,791 0.43 0.49
Model 2
Ride-pooling user 2,365 0.13 0.33
Model 3
TNC drivers who would prefer to switch to fuel-efcient vehicles 1,533 0.53 0.50
Model 4
TNC drivers who considered driving for ride-hailing services while buying or
leasing a new vehicle 1,540 0.47 0.50
Table 2. Key response variables.
Source: KAPSARC.
For example, in the weighted sample, the fraction
of male respondents, average age, and average
annual income are 48%, 46.77 years and $94,909,
respectively; this compares with 49%, 45.75 years
and $81,346 in the general population. Around 19%
of respondents are early adopters of these ride-
hailing services. Among 1,541 TNC drivers in the
sample, around 25% drive daily.
Table 2 gives the key summary statistics of the
response variables. The results indicate that the
sample proportions of TNC riders, drivers, and
non-users are 29%, 29%, and 42%, respectively.
Among ride-hailing riders, 13% had used ride-pooling
services. Among drivers, 53% indicated that they
would be inclined to shift toward more fuel-efcient
vehicles, and 47% showed a high propensity toward
buying, leasing or renting a new vehicle as a result of
driving for TNCs.
Mobility patterns of ride-hailing
riders
Table 3 depicts the association between the mobility
option most used by respondents and their TNC
usage frequencies. We label those who use TNCs
at least once per week as ‘frequent TNC users’
and the remainder as ‘infrequent TNC users.’
The results suggest that for frequent TNC users,
personal vehicles (53%-61%) and ride-hailing
(22%-32%) are the two most used travel modes,
followed by public transportation (4%-10%). For
infrequent TNC users, personal vehicles (79%-
87%) are dominant, with few respondents using
either public transit (3%-6%) or ride-hailing (3%-
5%). We note a marginal increase in the share of
public transit among the most used travel modes
for frequent versus infrequent TNC users. These
patterns indicate that ride-hailing replaces personal
vehicle use but does not signicantly impact public
transport demand.
In Table 4, we analyze how the unavailability
of one’s preferred travel mode affects mobility
patterns. Among those who normally favor ride-
hailing, 66% are likely to switch to driving personal
vehicles and 14% to public transit, respectively.
Hampshire et al. (2017) observed a similar trend:
when ride-hailing services were suspended in
12
Eliciting Preferences of Ride-Hailing Users and Drivers
Austin, 41% of users switched to driving. For
respondents who prefer to drive personal vehicles,
31% and 46% would shift to ride-hailing and car-
pooling/car-sharing, respectively, while those who
most frequently use public transit mainly change to
ride-hailing (29%) and personal driving (43%) rather
than car-pooling/car-sharing (4%). Although personal
vehicle drivers and public transport users switch to
ride-sharing at similar rates, the former accounts
for 86% of the sample and therefore would have a
much larger impact on ride-hailing demand.
The above ndings strengthen our earlier conclusion
that ride-sharing and car-sharing services are
mainly capturing personal-driving demand,
without substantially affecting the use of public
transportation. This nding has implications for
other countries, including Saudi Arabia, that, like
the U.S., have high rates of personal vehicle usage
and limited public transit options. Early evidence
suggests that recent investments in Riyadh’s transit
infrastructure (Riyadh Development Authority 2019;
Nanji 2018) are unlikely to be affected by the growth
Data and Summary Statistics
Frequency of using ride-hailing
services or taxis
User type What mobility options do you use most often?
Ride-hailing Driving
personal
vehicle
Public
transit
Walk or bike Car
sharing or
car-pooling
Once or more a day Frequent TNC
users
32% 53% 4% 6% 5%
Once or more per week 22% 61% 10% 5% 2%
Once or more per month; once or more
in 3 months Infrequent TNC
users
5% 79% 6% 6% 3%
Once or more a year; once or more
every few years or never 3% 87% 3% 5% 2%
Mobility option used most
often
How would you make your most frequent trip if your most frequently used option was not available?
NRide-hailing Drive
personal
vehicle
Public
transit
Walk or bike Car
sharing or
car-pooling
Would not
make the trip
Ride-hailing 256 0% 66% 14% 7% 13% 0%
Drive personal vehicle 6,369 31% 0% 9% 8% 46% 6%
Public transit 243 29% 43% 0% 17% 4% 2%
Walk or bike 372 12% 60% 17% 0% 6% 5%
Car sharing or car-pooling 151 30% 52% 13% 2% 0% 2%
Table 3. Most often used mobility options of the ride-hailing users (N = 11,902).
Table 4. Mobility option most used in the absence of their current travel mode.
Source: KAPSARC.
Source: KAPSARC.
13
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Data and Summary Statistics
of TNC services. In contrast to Clewlow and Mishra
(2017), we nd evidence that TNC services do not
induce travel demand—only 0.45% of frequent
ride-hailing users would not have made their
trips if these services were unavailable. These
discrepancies can be attributed to different target
samples in the two studies; in particular, the
analysis by Clewlow and Mishra (2017) had an
oversampling of respondents in San Francisco and
Los Angeles.
We present other insightful statistics about ride-
hailing preferences. Around 10% of TNC users
reported postponing the purchase of a new car.
This result is similar to that of Hampshire et al.
(2017) that 9% of the ride-hailing users purchased
a vehicle after the suspension of TNC services in
Austin. In terms of trip purpose, a high proportion
(46%) of respondents selected ride-hailing, car-
sharing or car-pooling as their preferred means
of travel to and from social events. Furthermore,
21% of respondents chose “didn’t want to drive
after drinking” as their primary motivation for using
ride-hailing services, making it the second most
common reason behind “convenience” (24%).
These results align with studies that found these
two factors to be the most prevalent motivations
for preferring ride-hailing (Rayle et al. 2016; Alemi
et al. 2018; Conway et al. 2018; Young and Farber
2019). They are also consistent with previous
studies that found the most common ride-hailing
trips are for recreational activities (Alemi et al.
2018; Lavieri and Bhat 2018; Young and Farber
2019).
We further analyze the 13% of ride-hailing users
who have used ride-pooling services at least once.
These respondents opted to ride-pool for slightly
over a third (34%) of their ride-hailing trips. Among
the ride-hailing users that never used ride-pooling
services, when asked for their reasons for not
using ride-pooling services, half of them indicated
that they had not heard of these services. Another
dominant reason, chosen by 22%, is a preference
for “private rides.”
Preferences of non-users
The leading reason, cited by 36% of non-users, for
not utilizing TNC services is their preference for
driving a personal vehicle. Another 21% of non-
users replied that they had never needed taxi or
ride-hailing services. Non-users were also asked
about utilizing TNC services for so-called ‘last
mile’ travel—i.e., how to reach the nearest station
for trips on public transportation. Approximately
41.3% reported their willingness to make this switch,
suggesting that ‘last mile’ trips offer high potential to
boost the adoption of TNC services among current
non-users.
Preferences of TNC drivers
Drivers were asked about their level of satisfaction
regarding working for TNCs. Most reported their
experience to be excellent (54%), followed by neutral
(28%) and unsatisfactory (18%). As expected, TNC
drivers who work more frequently drive more miles.
Those working daily and every other day average
42 miles and 31 miles per week, respectively;
those who work less than once per month drive an
average of 16 miles per week.
Table 5 summarizes the relationship between the
decision to rent/lease/purchase a vehicle and driving
frequency. Among drivers who work daily, 65%
indicated that driving for TNCs was a consideration
when acquiring a new vehicle. As expected, this
declines to 51% and 40%-46% for those who
drive every other day and once a week or less,
respectively. Moreover, 93% of drivers use their
primary vehicles to drive for ride-hailing services, with
little variation across different driving frequencies.
Eliciting Preferences of Ride-Hailing Users and Drivers 14
Eliciting Preferences of Ride-Hailing Users and Drivers
Data and Summary Statistics
Frequency of driving for ride-
hailing services
Number of
responses
Was driving for ride-hailing service a consideration in your decision to rent/lease/
purchase your primary vehicle?
Yes No
Daily 478 65% 35%
Every other day 343 51% 49%
Once in a week 646 43% 57%
Once in a month 298 46% 54%
Less than once per month 353 40% 60%
Table 5. Decision to rent/lease/purchase car based on the frequency of driving for ride-hailing.
Source: KAPSARC.
We then asked about the downtime activities of
drivers, i.e., activities in-between rides. Around
29% reported that they drive to the busy parts of
the city to get more rides, inducing extra vehicle
miles traveled (VMT). Picking up passengers takes
an average of nine and 10 minutes during peak and
off-peak hours, respectively, and adds around
2-3 miles per trip. For future vehicle purchases,
26% of TNC drivers who work more than 20 hours
per week would prefer diesel vehicles, due to their
higher fuel economy, but this falls to 8%-11% among
those who drive fewer hours. Around 25% would
prefer to buy a hybrid electric vehicle.
15
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Results and Discussion
We report parameter estimates of the link
function of the logit models and relative
risk or odds-ratio estimates with 95%
condence intervals. We explore non-linear effects
of continuous variables by plotting the predicted
choice probabilities and marginal effect estimates
over the support of the covariate. To compute
these quantities at a given level of the covariate of
interest, we x all other covariates at their sample
mean and make a unit change in the specic
covariate of interest.
Model 1—Preference for being a
TNC rider or a driver
Table 6 shows the parameter estimates and the
relative risk ratios of the multinomial logistic model,
which explain the association between individuals’
sociodemographic characteristics and being a TNC
rider, driver, or non-user.
The predicted choice probabilities and the marginal
effect plots of age (see Appendix, Figure A2)
indicate that the preference for being a TNC driver
decreases, and being a non-user increases,
with an increase in age, all else being equal.
This result is consistent with the ndings of Hall
and Krueger (2018) who also observed a higher
proportion of younger people among Uber drivers.
Perhaps younger people nd working through a
platform such as Uber attractive because they
value a exible work schedule and are willing to
take multiple part-time jobs. However, we observe
a non-linear effect of age on the propensity of
being a TNC user. The likelihood of being a user,
compared to being a driver or non-user, increases
with age for someone younger than 44 years, but
the pattern is reversed for those over 44 years.
Finally, the higher inclination of younger people to
be TNC users or drivers aligns with the education
and psychology studies that have established
that younger people are more likely to adopt new
information and communication technologies
(Helsper and Eyon 2010; Milojev and Sibley 2017).
Higher income individuals are more inclined to
use TNC services but have a lower propensity to
become TNC drivers. These results are consistent
across all income values (see Appendix, Figure A3)
and align with previous studies (Rayes et al. 2014;
Clewlow and Mishra 2017). A probable explanation
is that wealthier individuals are more able to pay
for TNC rides and value the convenience and
time savings they offer. Similarly, they are less
motivated to earn additional income by becoming
TNC drivers. After controlling for key covariates
including education, age, marital status, and
gender, individuals with higher household vehicle
ownership are less likely to use TNCs or become
drivers. However, the variation in the predicted
choice probabilities due to a change in vehicle
ownership is very small (see Appendix, Figure A4).
TNC services might serve as a convenient travel
option rather than a frequent mode of transport
for households with high vehicle ownership. Bhat
and Lavieri (2018) found a decrease in ride-hailing
frequency with the increase in vehicle availability.
Early adopters of technologies and residents
of metropolitan areas are more inclined toward
riding and driving ride-hailing services. In fact,
magnitudes of relative risk ratios indicate that these
two covariates have the most practically signicant
relationship to an individual’s preference to ride or
drive using TNC services.
Early adopters are 4.81 and 1.36 times more likely
to be drivers and riders, respectively, compared to
being non-users. For metropolitan area residents,
these gures are 1.94 and 1.53 times for drivers
and riders, respectively.
16
Eliciting Preferences of Ride-Hailing Users and Drivers
These ndings are consistent with the literature.
Alemi et al. (2018) found that early adopters of TNC
services are likely to be “technology-oriented” and
thus tend to adopt such services in bundles as a
part of their modern lifestyles. Similarly, Hall and
Krueger (2018) and Alemi et al. (2018) observed
a higher inclination among metropolitan residents
toward using TNC services. Lavieri and Bhat
(2018) speculated three possible reasons for this
tendency: parking restrictions in urban areas, lower
costs for ride-hailing trips due to shorter distances,
and the higher reliability of TNC services compared
with taxis and public transit. In addition, higher
travel demand in metropolitan areas also explains
residents’ higher propensities to be drivers.
Finally, postgraduate degree holders and single
individuals, everything else constant, are more
likely to be TNC riders but less prone to be drivers.
These results align with the ndings of Lavieri and
Bhat (2018).
Model 2—Preference to ride-pool
Table 7 shows the parameter estimates and
odds ratios of a binary logistic model reecting
the preferences of TNC riders for ride-pooling
(compared to the base category of non-users of
ride-pooling). As the literature on ride-pooling is
still at a nascent stage, we highlight some of the
similarities between our results and the results
of car-pooling/car-sharing studies, including new
insights on the non-linear relationships between
TNC rider preferences for ride-pooling and their
gender and levels of education.
The predicted probability and marginal effect plots
of age (see Appendix, Figure A5) indicate that
proclivity to use ride-pooling is inversely related
to the age of TNC users. Lavieri and Bhat (2018)
observed a similar trend, which they attribute to
tech-savviness and the variety-seeking behavior
of younger people. Traditional car-pooling has also
been found to be common among individuals aged
between 25 and 55 (Shaheen et al. 2016). However,
the relationship between age and propensity to ride-
pool varies with sociodemographic characteristics
such as education level (Appendix, Figure A6) and
gender (Appendix, Figure A7). The inclination of
male and postgraduate TNC users to ride-pool is
less affected by age than that of females and those
with below postgraduate education, keeping all
other characteristics the same. A young female TNC
user is much more likely to ride-pool than a young
male. More specically, a female TNC user who is
younger than 54 years has a higher probability of
pooling than a male user of the same age, but the
trend reverses for users older than 54 years. Riders
with less than postgraduate education, and who are
younger than 34 years, have a higher tendency to
use ride-pooling than postgraduate TNC users of
the same age, but the pattern is reversed for those
older than 34 years.
We nd that household vehicle ownership is
negatively associated with ride-pooling: the odds of
TNC riders opting to ride-pool decrease by a factor
of 0.87 with the addition of a household vehicle. The
predicted probability plot in the Appendix, Figure
A8 shows the linear nature of this relationship. This
is consistent with the ndings of Lee et al. (2018) in
their study on car-pooling.
Similar to the results of Model 1 above, metropolitan
residents and early adopters of technology are more
inclined to ride-pool, and males are less likely, all
else being equal. A few studies in the environmental
psychology literature (e.g., Glover et al. 1997) argue
that females take a stronger stance on ethical,
environmental and pro-social behavior compared
to males; this may explain the higher proclivity of
females to use ride-pooling services. Residential
Results and Discussion
17
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Explanatory variables Parameter estimates Relative risk ratios
Estimate Std. err. z-stat Estimate LB (95% CI) UB (95% CI)
Ride-hailing driver
Male indicator -0.530 0.100 -5.28 0.589 0.484 0.717
Single indicator -0.026 0.121 -0.22 0.974 0.769 1.234
Age -0.134 0.005 -27.76 0.875 0.866 0.883
Postgraduation indicator -0.620 0.120 -5.15 0.538 0.425 0.681
Annual income (US$) -5.07E-06 1.55E-06 -3.27 0.999995 0.999992 0.999998
Metropolitan resident indicator 0.665 0.112 5.93 1.945 1.561 2.423
HH size 3+ indicator 0.485 0.108 4.49 1.625 1.315 2.008
Total vehicle ownership -0.116 0.049 -2.38 0.890 0.809 0.980
Early adopter indicator 1.571 0.117 13.46 4.813 3.828 6.050
Constant 5.693 0.261 21.83
Ride-hailing user
Male indicator -0.286 0.072 -3.97 0.751 0.652 0.865
Single indicator 0.292 0.092 3.17 1.339 1.118 1.604
Age -0.045 0.003 -15.70 0.956 0.950 0.961
Postgraduation indicator 0.141 0.077 1.83 1.152 0.990 1.340
Annual income (US$) 7. 35E- 0 6 5.82E-07 12.63 1.000007 1.000006 1.000008
Metropolitan resident indicator 0.425 0.088 4.83 1.529 1.287 1.817
HH size 3+ indicator -0.160 0.083 -1.93 0.852 0.724 1.002
Total vehicle ownership -0.086 0.035 -2.47 0.918 0.857 0.982
Early adopter indicator 0.309 0.109 2.83 1.362 1.100 1.687
Constant 1.235 0.195 6.33
N8,086
Loglikelihood -6109.5
Pseudo R-square 0.292
Note: “Ride-hailing non-user” is a base category. Std. err. = standard error; LB= lower bound; UB= upper bound; CI= condence interval.
Table 6. Multinomial logistic parameter estimates and relative risk ratios (Model 1).
Source: KAPSARC.
Results and Discussion
location is the most practically signicant predictor:
TNC users living in metropolitan areas are more
likely to ride-pool by a factor of 1.73 compared
to those living in suburban areas. This result is
consistent with the ndings of Almei et al. (2018)
and Lavieri and Bhat (2018).
18
Eliciting Preferences of Ride-Hailing Users and Drivers
Results and Discussion
Model 3—Preferences of TNC
drivers for fuel-efcient vehicles
Table 8 summarizes the results of a binary logistic
model that identies the characteristics of TNC
drivers who would prefer to switch to fuel-efcient
vehicles (compared to the base category of
drivers who prefer not to switch to fuel-efcient
vehicles). Since there is no previous study on such
preferences of TNC drivers, we highlight some
similarities and differences between our results and
those of studies concerned with the characteristics
of electric vehicle (EV) buyers.
The predicted probability plot shows a negative
correlation between the ages of TNC drivers and
their propensity to switch to fuel-efcient vehicles
(see Appendix, Figure A9). Hidrue et al. (2011)
observe a similar trend among EV buyers. The
relationships between the inclination of drivers to
switch to fuel-efcient vehicles and their residential
locations (see Appendix, Figure A10), education
levels (see Appendix, Figure A11), and marital status
(see Appendix, Figure A12) vary across driver ages.
We nd that postgraduate education and
metropolitan residency are the two most signicant
predictors of propensities to switch to more fuel-
efcient vehicles. Drivers with postgraduate
education who live in metropolitan areas are more
prone than their counterparts to make this change
if their age is below 48 years, all else being equal,
Explanatory variables Parameter estimates Odds ratio
Estimate Std. err. z-stat Estimate LB (95% CI) UB (95% CI)
Male indicator -1.341 0.635 -2.11 0.262 0.075 0.908
Age -0.047 0.014 -3.43 0.955 0.929 0.980
Postgraduation indicator -0.614 0.662 -0.93 0.541 0.148 1.981
Metropolitan resident indicator 0.550 0.165 3.34 1.733 1.255 2.393
Total vehicle ownership -0.138 0.084 -1.64 0.871 0.739 1.027
Early adopter indicator 0.321 0.201 1.6 1.378 0.930 2.042
Male indicator X age 0.026 0.014 1.89 1.026 0.999 1.054
Postgrad indicator X age 0.016 0.014 1.13 1.016 0.988 1.045
Constant 0.207 0.601 0.34
N2,504
Loglikelihood -853.9
Pseudo R-square 0.048
Note: “Ride-pooling non-user” is a base category and parameter estimates are for “Ride-pooling user.” Std. err. = standard error; LB= lower
bound; UB= upper bound; CI= condence interval.
Table 7. Binary logistic parameter estimates and odds ratios (Model 2).
Source: KAPSARC.
19
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Results and Discussion
although this pattern reverses for postgraduate
drivers older than 48 years. These results are
consistent with the previous studies showing that
younger and more highly educated individuals
are more prone to buy alternative fuel vehicles
(Dütschke et al. 2013; Hackbarth and Madlener
2013). Individuals with higher levels of education
tend to have a greater awareness of the fuel-saving
benets and lower environmental impacts from
improved fuel economy. Additionally, hybrid electric
vehicles are more suited to improving fuel-savings
in the stop-and-go trafc conditions of congested
metropolitan areas (Romm and Frank 2006).
Married drivers are more inclined to switch to fuel-
efcient vehicles than their single counterparts
among drivers below 60 years of age. These results
are consistent with those of Peter et al. (2011), who
found that households with children are more likely
to be EV buyers. In the context of this study, married
drivers might be more conscious about fuel-efciency
because they may need to drive their vehicles more
than their single counterparts do for other household
activities. A recent report by Ipsos (2017) also shows
that married households are more inclined toward
buying electric vehicles.
Drivers who are early adopters of technologies
have greater odds of switching to a fuel-efcient
vehicle by a factor of 1.47. This observation makes
intuitive sense because early adopters are likely
to have a greater afnity toward technology. The
results of a survey conducted by CleanTechnica are
also consistent with our ndings that 38% of drivers
surveyed selected “love for new technology” as the
reason for switching to the fuel-efcient vehicles.
TNC drivers who drive daily are also more inclined
toward fuel-efcient vehicles than those who drive
less frequently. Daily drivers are likely more sensitive
to fuel prices due to more frequent driving and higher
VMT. It is well established in EV literature that early
adopters of EVs accumulate more VMT (Plotz et
al. 2014). Hidrue et al. (2011) have shown that an
individual’s willingness to pay a premium for an EV
is primarily driven by savings in fuel costs, which is
further associated with VMT.
Drivers from households that own more vehicles
also have a higher tendency to switch to fuel-
efcient vehicles. The probability of switching
increases linearly with the increase in the number
of vehicles (see Appendix, Figure A13). Since
the income of drivers is not controlled in the
specication (because it was not statistically
signicant), perhaps the missing income effect is
also reinforcing the pro-fuel-efciency behavior of
drivers with higher vehicle ownership. We observe
similar relationships in EV literature. Previous
studies have shown that higher-income consumers
are more likely to buy EVs (Erdem et al. 2010;
Saarenpää et al. 2013). Moreover, EVs are typically
owned by high-income households with more than
one car (Hjorthol 2013).
Explanatory variables Parameter estimates Odds ratio
Estimate Std. err. z-stat Estimate LB (95% CI) UB (95% CI)
Drive daily indicator 0.346 0.147 2.35 1.413 1.059 1.887
Single indicator -1.026 0.431 -2.38 0.359 0.154 0.834
Age -0.015 0.009 -1.64 0.985 0.967 1.003
Table 8. Binary logistic parameter estimates and odds ratios (Model 3).
20
Eliciting Preferences of Ride-Hailing Users and Drivers
Results and Discussion
Model 4—Preference of TNC
drivers to buy a new vehicle
We use a binary logistic regression to estimate
whether driving for a ride-hailing service was a
consideration in TNC drivers’ decisions to buy, rent
or lease a new vehicle. Results are shown in Table
9 (compared to the base category of no change in
the preference to buy a new vehicle as a result of
driving for TNCs).
The predicted probability and marginal effect plots
(in Appendix, Figure A14) indicate that the tendency
for driving for TNCs to be a major consideration
when buying a new vehicle is inversely related
to the age of the TNC driver, keeping everything
else constant. Among drivers below 55 years
of age, those with postgraduate education and
who are married have a higher propensity than
their counterparts to buy a car to drive for TNCs,
but these patterns reverse for older drivers (see
Appendix, gures A15 and A16) as the negative
effect of the increase in age on the probability of
buying a car to drive for TNCs becomes much
higher.
Higher income drivers have a lower inclination
to buy vehicles with driving for TNCs as a
consideration (see the predicated probability and
marginal effect in Appendix, Figure A17). Further
investigation of this income effect in single and
married drivers indicates that the increase in income
of married drivers increases their probability of
buying cars to drive for TNCs, but the reverse
is seen for single drivers (see Appendix, Figure
A18). The combined negative income effect is a
manifestation of the steeper rate of decline in the
probability for single drivers as compared to the
rate of increase for married drivers. For example,
for single and married drivers with annual incomes
Explanatory variables Parameter estimates Odds ratio
Estimate Std. err. z-stat Estimate LB (95% CI) UB (95% CI)
Postgrad indicator 1.447 0.573 2.53 4.250 1.384 13.054
Metropolitan resident indicator 1.392 0.441 3.16 4.024 1.695 9.554
Total vehicle ownership 0.054 0.055 0.97 1.055 0.947 1.176
Early adopter indicator 0.385 0.129 2.98 1.470 1.140 1.895
Single indicator X age 0.016 0.013 1.3 1.017 0.992 1.042
Postgrad indicator X age -0.032 0.016 -2.02 0.969 0.940 0.999
Metropolitan resident indicator X age -0.032 0.013 -2.5 0.969 0.945 0.993
N1,534
Loglikelihood -994.3
Pseudo R-square 0.06
Note: “No preference of ride-hailing drivers to switch to fuel efcient vehicles” is a base category and parameters are estimated for
“Preference for ride-hailing…” LB (95% CI) and UB (95% CI) imply lower and upper bounds of a 95% condence interval. Std. err. =
standard error; LB= lower bound; UB= upper bound; CI= condence interval.
Source: KAPSARC
21
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Results and Discussion
below $10,000 the predicted probability is the same
(around 0.48), but at $100,000 income it increases
to 0.51 for married drivers and falls to 0.38 for single
ones, keeping all other characteristics the same.
The high propensity of low-income single drivers
to buy, rent or lease a vehicle with driving for TNCs
being a major consideration in part explains why
some TNCs offer renting and leasing services to
attract drivers from this demographic group.
Male drivers residing in metropolitan areas, who are
early adopters, have higher vehicle ownership and
drive daily, have a higher inclination to buy vehicles
with driving for TNCs being a major purchase
consideration than those with the opposite prole,
all else being equal. In terms of the strength of these
relationships, being an early adopter of technologies
increases the odds by a factor of 2.59, driving daily
by a factor of 2.09, and metropolitan residency by a
factor of 1.90.
Explanatory variables Parameter estimates Odds ratio
Estimate Std. err. z-stat Estimate LB (95% CI) UB (95% CI)
Drive daily indicator 0.736 0.159 4.620 2.09 1.53 2.85
Male indicator 0.459 0.120 3.840 1.58 1.25 2.00
Single indicator -0.518 0.529 -0.980 0.60 0.21 1.68
Age -0.032 0.010 -3.150 0.97 0.95 0.99
Postgraduation indicator 1.122 0.588 1.910 3.07 0.97 9.72
Annual income (US$) 1.17E-06 6.18E-07 1.900 1.000001 11.000002
Metropolitan resident indicator 0.639 0.138 4.620 1.90 1.44 2.49
Total vehicle ownership 0.163 0.056 2.920 1.18 1.05 1.31
Early adopter indicator 0.953 0.135 7.060 2.59 1.99 3.38
Single indicator X age 0.020 0.016 1.250 1.02 0.99 1.05
Postgrad indicator X age -0.024 0.017 -1.400 0.98 0.95 1.01
Single indicator X annual income -5.64E-06 2.42E-06 -2.330 0.999994 0.999990 0.999999
Constant -0.699 0.376 -1.860
N1,539
Loglikelihood -914.5
Pseudo R-square 0.139
Note: “No change in preference to buy a new vehicle due to being a ride-hailing driver” is a base category and parameters are estimated
for “change in preference…” LB (95% CI) and UB (95% CI) imply lower and upper bounds of a 95% condence interval. Std. err. = standard
error; LB= lower bound; UB= upper bound; CI= condence interval.
Table 9. Binary logistic parameter estimates and odds ratios (Model 4).
Source: KAPSARC
Eliciting Preferences of Ride-Hailing Users and Drivers 22
Eliciting Preferences of Ride-Hailing Users and Drivers
Concluding Remarks and Implications
for Stakeholders
This research provides new insights into
behaviors related to TNC services by
identifying the relationship between
individuals’ sociodemographic characteristics and
their preferences to use ride-hailing services (as
riders or drivers) and willingness to ride-pool. The
analysis uses calibrated multinomial and binary
logistic models based on a dataset taken from a
2017 survey (N=11,902) in TNC-served U.S. cities.
We make a unique contribution to ride-hailing
literature by estimating the proclivity of TNC drivers
to buy new vehicles with the consideration of driving
for TNCs and shifting to fuel-efcient vehicles as
inuencing factors, neither of which have been
explored in the existing literature. We also observe
non-linear relationships by accounting for the
interaction effects of continuous covariates (e.g., age
and income) with binary covariates (e.g., gender)
while estimating preferences.
The ndings from this study can inform key
stakeholders (such as transportation planners,
government agencies, automakers, and TNCs) as
they develop policies to encourage ride-pooling
and the deployment of high fuel economy vehicles,
leading to systemic environmental benets. For
instance, automotive manufacturers and auto-
leasing companies could partner with TNCs to offer
leasing plans for high fuel efciency and electric
vehicles to the identied pro-fuel-efciency drivers.
This would not only benet drivers by reducing their
operating costs but would also help automakers
meet zero emission vehicle requirements and
federal eet fuel economy targets. The lease options
offered to Uber and Lyft drivers by General Motors
on its all-electric Chevrolet Bolt is a case-in-point
(Kurczewski 2017). In addition, the deployment of a
higher number of fuel-efcient vehicles would help
TNCs to balance supply and demand for their new
environmentally friendly initiatives, such as Lyft’s
‘green mode’ (Price 2019), which allow riders to
specically call ‘green’ vehicles.
Targeted campaigns could be organized to
spread awareness of ride-pooling services and
their environmental benets to the ride-hailing
users identied as having a high propensity to
ride-pool. Our research showed that younger
female passengers, who are known to be more
environmentally conscious, prefer ride-pooling. To
ensure passenger safety while pooling, automakers
could be encouraged to provide tailor-made vehicles
for pooling that include partitions.
It is important to note that this study explores
correlation patterns only: readers should not
interpret the results as identifying causal
relationships between individuals’ sociodemographic
characteristics and their preferences. Collaborations
with TNCs to conduct randomized experiments in
order to disentangle causal effects is a potential
avenue for future research.
23
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
References
Alemi, Farzad, Giovanni Circella, Susan Handy, and
Patricia Mokhtarian. 2018. "What Inuences Travelers
to Use Uber? Exploring the Factors Affecting the
Adoption of On-demand Ride Services in California."
Travel Behaviour and Society 13: 88-104. https://doi.
org/10.1016/j.tbs.2018.06.002
Babar, Yash, and Gordon Burtch. 2017. "Examining
the Heterogeneous Impact of Ridehailing Services
on Public Transit Use." Minnesota: Carlson School of
Management, University of Minnesota. Available at
SSRN 3042805. https://papers.ssrn.com/sol3/papers.
cfm?abstract_id=3042805
Bergmann, Michael. 2011. "ipfweight: Stata Module to
Create Adjustment Weights for Surveys." Statistical Soft-
ware Components S457353. Department of Economics,
Boston College. http://econpapers.repec.org/software/
bocbocode/s457353.htm
Berliner, Rosaria M., and Gil Tal. 2018. "What Drives
Your Drivers: An In-Depth Look at Lyft and Uber Drivers."
January. Davis, California: Institute of Transportation
Studies, University of California. https://itspubs.ucdavis.
edu/wp-content/themes/ucdavis/pubs/download_pdf.
php?id=2851
CleanTechnica. 2016. "Electric Car Drivers: Desires,
Demands & Who They Are." Accessed March 25, 2019.
https://cleantechnica.com/les/2017/05/Electric-Car-Driv-
ers-Report-Surveys-CleanTechnica-Free-Report.pdf
Clewlow, Regina R., and Gouri S. Mishra. 2017. "Dis-
ruptive Transportation: The Adoption, Utilization, and
Impacts of Ride-hailing in the United States." Davis,
California: Institute of Transportation Studies, University
of California. https://itspubs.ucdavis.edu/wp-content/
themes/ucdavis/pubs/download_pdf.php?id=2752
Conway, Matthew, Deborah Salon, and David King.
2018. "Trends in Taxi Use and the Advent of Ridehailing,
1995–2017: Evidence from the US National Household
Travel Survey." Urban Science 2(3): 79. https://www.mdpi.
com/2413-8851/2/3/79
Dias, Felipe F., Patrícia S. Lavieri, Taehooie Kim, Chan-
dra R. Bhat, and Ram M. Pendyala.2019. "Fusing
Multiple Sources of Data to Understand Ride-Hailing
Use." Transportation Research Record no. 2673 (6):214-
224. https://doi.org/10.1177/0361198119841031
Dütschke, Elisabeth, Uta Schneider, and Anja Peters.
2013. "Who Will Use Electric Vehicles?" Working Paper
Sustainability and Innovation (6). Fraunhofer Institute for
Systems and Innovation Research. https://www.isi.fraun-
hofer.de/content/dam/isi/dokumente/sustainability-inno-
vation/2013/WP06-2013_Electric_Vehicles.pdf
Erdem, Cumhur, İsmail Şentürk, and Türker Şimşek.
2010. "Identifying the Factors Affecting the Willingness
to Pay for Fuel-efcient Vehicles in Turkey: a Case of
Hybrids." Energy Policy 38(6): 3038-3043. https://doi.
org/10.1016/j.enpol.2010.01.043
Franzen, Axel, and Dominikus Vogl. 2013. "Two Dec-
ades of Measuring Environmental Attitudes: A Compar-
ative Analysis of 33 Countries." Global Environmental
Change 23(5): 1001-1008. https://doi.org/10.1016/j.
gloenvcha.2013.03.009
Greenblatt, Jeffery B., and Susan Shaheen. 2015.
"Automated Vehicles, On-demand Mobility, and
Environmental Impacts." Current Sustainable/Renewa-
ble Energy Reports 2(3):74-81. https://doi.org/10.1007/
s40518-015-0038-5
Glover, Saundra H., Minnette A. Bumpus, John E. Logan,
and James R. Ciesla. 1997. "Re-examining the Inuence
of Individual Values on Ethical Decision Making." In From
the Universities to the Marketplace: The Business Ethics
Journey, 109-119. Springer, Dordrecht.
Hall, Jonathan D., Craig Palsson, and Joseph Price.
2018. "Is Uber a Substitute or Complement for Public
Transit?" Journal of Urban Economics 108:36-50. https://
doi.org/10.1016/j.jue.2018.09.003
Hall, J. V., and A. B. Krueger. 2018. An Analysis of the
Labor Market for Uber’s Driver-partners in the U.S."
Industrial and Labor Relations Review 71(3): 705-732.
https://doi.org/10.1177/0019793917717222
Hackbarth, André, and Reinhard Madlener. 2013.
"Consumer Preferences for Alternative Fuel Vehicles:
A Discrete Choice Analysis." Transportation Research
Part D: Transport and Environment 25:5-17. https://doi.
org/10.1016/j.trd.2013.07.002
Hampshire, Robert, Chris Simek, Tayo Fabusuyi, Xuan
Di, and Xi Chen. 2017. "Measuring the Impact of an
Unanticipated Disruption of Uber/Lyft in Austin, TX."
Available at SSRN: https://ssrn.com/abstract=2977969 or
http://dx.doi.org/10.2139/ssrn.2977969
24
Eliciting Preferences of Ride-Hailing Users and Drivers
Hawkins, Andrew. 2018. "Uber Express Pool Offers
the Cheapest Fares Yet in Exchange for a Little
Walking." The Verge, February 21. Accessed April 4,
2019. https://www.theverge.com/2018/2/21/17020484/
uber-express-pool-launch-cities.
Helsper, Ellen Johanna, and Rebecca Eynon. 2010.
"Digital Natives: Where is the Evidence?" British Edu-
cational Research Journal 36(3): 503-520. https://doi.
org/10.1080/01411920902989227
Hidrue, Michael K., George R. Parsons, Willett Kempton,
and Meryl P. Gardner. 2011. "Willingness to Pay for Elec-
tric Vehicles and their Attributes." Resource and Energy
Economics 33(3): 686-705. https://doi.org/10.1016/j.
reseneeco.2011.02.002
Hjorthol, Randi. 2013. "Attitudes, Ownership and Use
of Electric Vehicles—A Review of Literature." Institute
of Transport Economics, Norwegian Centre for Trans-
port Research. Oslo. https://www.toi.no/publications/
attitudes-ownership-and-use-of-electric-vehicles-a-re-
view-of-literature-article31833-29.html
Ipsos. 2017. "Is There a Target Market for Electric
Vehicles?" Accessed February 20, 2019. https://www.
ipsos.com/sites/default/les/2017-04/ipsos-marketing-tar-
get-market-electric-vehicles.PD__0.pdf.
Kooti, Farshad, Mihajlo Grbovic, Luca Maria Aiello,
Nemanja Djuric, Vladan Radosavljevic, and Kristina
Lerman. 2017. "Analyzing Uber,s Ride-sharing
Economy." April. In Proceedings of the 26th International
Conference on World Wide Web Companion, 574-582.
International World Wide Web Conferences Steering
Committee. https://doi.org/10.1145/3041021.3054194
König, Alexandra, Tabea Bonus, and Jan Grippenkoven.
2018. "Analyzing Urban Residents’ Appraisal of Ride-
pooling Service Attributes with Conjoint Analysis."
Sustainability 10(10): 3711. https://doi.org/10.3390/
su10103711
Kurczewski, Nick. 2017. "GM Announces Bolt Lease for
Uber, Lyft Drivers." Accessed April 10, 2019. https://www.
cars.com/articles/gm-announces-bolt-lease-for-uber-lyft-
drivers-1420695330355
Lee, Jae Hyun, and Konstadinos G. Goulias. 2018. "A
Decade of Dynamics of Residential Location, Car Owner-
ship, Activity, Travel and Land Use in the Seattle Metro-
politan Region." Transportation Research Part A: Policy
and Practice 114B: 272-287. https://doi.org/10.1016/j.
tra.2018.01.029
Lavieri, Patrícia S., Chandra R. Bhat. 2018. "MaaS in
Car-Dominated Cities: Modeling the Adoption,
Frequency, and Characteristics of Ride-hailing Trips in
Dallas, Texas." Department of Civil, Architectural and
Environmental Engineering, The University of Texas at
Austin.
Martin, Elliot W., and Susan A. Shaheen. 2011. Green-
house Gas Emission Impacts of Carsharing in North
America. Transactions on Intelligent Transportation
Systems 12(4): 1074-1086. The Institute of Electrical
and Electronics Engineers. https://doi.org/10.1109/
TITS.2011.2158539
Milojev, Petar, and Chris G. Sibley. 2017. "Normative
Personality Trait Development in Adulthood: A 6-year
Cohort-sequential Growth Model." Journal of Personality
and Social Psychology 112(3): 510.
Feigon, Sharon, and Colin Murphy. 2016. "Shared Mobil-
ity and the Transformation of Public Transit." Chicago:
American Public Transportation Association. https://www.
apta.com/wp-content/uploads/Resources/resources/
reportsandpublications/Documents/APTA-Shared-Mobil-
ity.pdf
Nanji, Noor. 2018. "Riyadh Metro Mega-project to be Fully
Operational by 2021." The National, March 8. https://www.
thenational.ae/business/travel-and-tourism/riyadh-met-
ro-mega-project-to-be-fully-operational-by-2021-1.711522
Neoh, Jun Guan, Maxwell Chipulu, and Alasdair Marshall.
2017. "What Encourages People to Carpool? An Evalua-
tion of Factors with Meta-analysis." Transportation 44(2):
423 - 4 47. https://doi.org/10.1007/s11116 - 015- 9661-7
Peters, Anja, Raphael Agosti, Mareike Popp, and
Bettina Ryf. 2011. "Electric Mobility: A Survey of Dif-
ferent Consumer Groups in Germany with Regard to
Adoption." Summer Study Proceedings Series, Euro-
pean Council for an Energy Efcient Economy. https://
www.eceee.org/library/conference_proceedings/
eceee_Summer_Studies/2011/4-transport-and-mobili-
ty-how-to-deliver-energy-efciency160/electric-mobil-
ity-a-survey-of-different-consumer-groups-in-germa-
ny-with-regard-to-adoption/
Plötz, Patrick, Uta Schneider, Joachim Globisch, and
Elisabeth Dütschke. 2014. "Who Will Buy Electric
Vehicles? Identifying Early Adopters in Germany."
Transportation Research Part A: Policy and Practice 67:
96-109. https://doi.org/10.1016/j.tra.2014.06.006
References
25
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Price, Emily. 2019. "Lyft Will Allows Riders to Request
an EV Via a New ‘Green Mode’." Fortune, Febru-
ary 26. Accessed on April 10, 2019. http://fortune.
com/2019/02/06/lyft-will-allows-riders-to-request-an-ev-
via-a-new-green-mode
Rayle, Lisa, Susan A. Shaheen, Nelson Chan, Danielle
Dai, and Robert Cervero. 2014. "App-based, On-demand
Ride Services: Comparing Taxi and Ridesourcing Trips
and User Characteristics in San Francisco." University
of California, Berkeley Transportation Center. https://
www.its.dot.gov/itspac/dec2014/ridesourcingwhitepa-
per_nov2014.pdf
Riyadh Development Authority. 2019. "King Abdulaziz
Project for Riyadh Public Transport." http://www.ada.gov.
sa/ADA_e/DocumentShow_e/?url=/res/ADA/En/Projects/
RiyadhMetro/index.html
Romm, Joseph J., and Andrew A. Frank. 2006. "Hybrid
Vehicles Gain Traction." Scientic American 294(4):
72-79.
Saarenpää, Jukka, Mikko Kolehmainen, and Harri
Niska. 2013. "Geodemographic Analysis and Estima-
tion of Early Plug-in Hybrid Electric Vehicle Adoption."
Applied Energy 107: 456-464. https://doi.org/10.1016/j.
apenergy.2013.02.066
Schaller, Bruce. 2018. "The New Automobility: Lyft, Uber
and the Future of American Cities." Accessed March 1,
2019. http://www.schallerconsult.com/rideservices/auto-
mobility.pdf
Shaheen, Susan, and Nelson Chan. 2016. "Mobility
and the Sharing Economy: Potential to Facilitate the
First- and Last-mile Public Transit Connections." Built
Environment 42(4): 573-588.
Shaheen, Susan A., Nelson D. Chan, and Teresa Gay-
nor. 2016. "Casual carpooling in the San Francisco Bay
Area: Understanding user characteristics, behaviors, and
motivations." Transport Policy 51: 165-173. https://doi.
org/10.1016/j.tranpol.2016.01.003
Smith, Aaron. 2016. "Shared, Collaborative and on
Demand: The New Digital Economy." May 19, Pew
Research Center. https://www.pewresearch.org/
internet/2016/05/19/the-new-digital-economy/
Sadowsky, Nicole, and Erik Nelson. 2017. "The Impact
of Ride-hailing Services on Public Transportation Use: A
Discontinuity Regression Analysis." Economics
Department Working Paper Series 13. Bowdoin Col-
lege. https://digitalcommons.bowdoin.edu/econpapers/13
Taylor, Brian D., Ryan Chin, Melanie Crotty, Jennifer Dill,
Lester A Hoel, Michael Manville, Steve Polzin, Bruce
Schaller, Susan Shaheen and Dan Sperling. 2016.
"Between Public and Private Mobility: Examining the Rise
of Technology-Enabled Transportation Services."
Transportation Research Board Special Report 319.
Washington: The National Academies Press. https://doi.
org/10.17226/21875
Yu, Haitao, and Zhong-Ren Peng. 2019. "Exploring the
spatial variation of ridesourcing demand and its rela-
tionship to built environment and socioeconomic factors
with the geographically weighted Poisson regression."
Journal of Transport Geography. 75: 147-163. https://doi.
org/10.1016/j.jtrangeo.2019.01.004
References
26
Eliciting Preferences of Ride-Hailing Users and Drivers
Appendix 1: Figures
Figure A1. Geographical distribution of respondents at state-level (N=11,902).
Figure A2. Predicted probability and the marginal effect of age (Model 1).
Source: KAPSARC.
Source: KAPSARC.
22
Appendix 1: Figures
Figure A1. Geographical distribution of respondents at state-level (N=11,902)
Source: KAPSARC
0.79
0.40
0.00
20 40 60
Probability
Age
Predicted Probability with 95% CIs (Driver)
80
-5.0e-04
-1.2e-02
-2.4e-02
20 40 60
Probability Diff.
Age
Age
Marginal Effects with 95% CIs (Driver)
80
0.37
0.26
0.14
20 40 60
Probability
Age
Predicted Probability with 95% CIs (User)
80
1.1e-02
1.5e-03
-7.8e-03
20 40 60
Probability Diff.
Marginal Effects with 95% CIs (User)
80
Age
0.86
0.45
0.05
20 40 60
Probability
Age
Predicted Probability with 95% CIs (Non-user)
80
1.9e-02
1.2e-02
5.5e-03
20 40 60
Probability Diff.
Marginal Effects with 95% CIs (Non-user)
80
27
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Figure A3. Predicted probability and the marginal effect of income (Model 1).
Figure A4. Predicted probability and the marginal effect of total vehicle ownership (Model 1).
Source: KAPSARC.
Source: KAPSARC.
0.38
0.21
0.03
0 100,000 200,000 300,000
Probability
Income
Predicted Probability with 95% CIs (Driver)
-2.9e-03
-5.0e-03
-7.1e-03
Probability Diff.
0 100,000 200,000 300,000
Income
Marginal Effects with 95% CIs (Driver)
0.73
0.44
0.14
0 100,000 200,000 300,000
Probability
Income
Predicted Probability with 95% CIs (User)
1.2e-02
8.6e-03
5.3e-03
Probability Diff.
0 100,000 200,000 300,000
Income
Marginal Effects with 95% CIs (User)
0.51
0.36
0.21
0 100,000 200,000 300,000
Probability
Income
Predicted Probability with 95% CIs (Non-user)
3.0e-04
3.2e-03
-6.6e-03
Probability Diff.
0 100,000 200,000 300,000
Income
Marginal Effects with 95% CIs (Non-user)
0.32
0.27
0.23
012345
Probability
Total Vehicles Total Vehicles
Predicted Probability with 95% CIs (Driver)
2.4e-03
-7.7e-03
-1.8e-02
Probability Diff.
0 1 2 3 4
Marginal Effects with 95% CIs (Driver)
0.33
0.28
0.23
012345
Probability
Total Vehicles Total Vehicles
Predicted Probability with 95% CIs (User)
3.3e-03
-8.5e-03
-2.8e-02
Probability Diff.
0 1 2 3 4
Marginal Effects with 95% CIs (User)
0.52
0.45
0.37
012345
Probability
Total Vehicles Total Vehicles
Predicted Probability with 95% CIs (Non-user)
2.8e-02
1.7e-02
5.5e-03
Probability Diff.
0 1 2 3 4
Marginal Effects with 95% CIs (Non-user)
Appendix 1: Figures
28
Eliciting Preferences of Ride-Hailing Users and Drivers
Figure A5. Predicted probability and the marginal effect of age (Model 2).
Figure A6. Interaction effect of age and postgrad dummy (Model 2).
Source: KAPSARC.
Source: KAPSARC.
.3
.2
.1
0
Predicted Probability (with 95% CIs)
Probability
Age
Probability Difference
20 40 60 80
0
-.002
-.004
-.006
-.008
Marginal Effects (with 95% CIs)
Age
20 40 60 80
0.25
.2
.15
.1
.05
20 40 60 80
Probability
Age
Below Postgrad
Postgrad
Appendix 1: Figures
29
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Figure A7. Interaction effect of age and male dummy (Model 2).
Figure A8. Predicted probability and the marginal effect of total vehicle ownership (Model 2).
Source: KAPSARC.
Source: KAPSARC.
.3
.25
.2
.15
.1
.05
20 40 60 80
Probability
Age
Female
Male
.2
.15
.1
.05
Predicted Probability (with 95% CIs)
Probability
Total Vehicles Total Vehicles
Probability Difference
0 0 1 2 3 41 2 3 4 5
.01
0
-.01
-.02
-.03
-.04
Marginal Effects (with 95% CIs)
Appendix 1: Figures
30
Eliciting Preferences of Ride-Hailing Users and Drivers
Figure A9. Predicted probability and the marginal effect of age (Model 3).
Figure A10. Interaction effect of “age” and “metropolitan resident” dummy (Model 3).
Source: KAPSARC.
Source: KAPSARC.
.7
.6
.5
.4
.3
.2
Predicted Probability (with 95% CIs)
Probability
Age
Probability Difference
20 40 60 80
Age
20 40 60 80
-.002
-.004
-.006
-.008
-.01
Marginal Effects (with 95% CIs)
.8
.6
.4
.2
20 40 60 80
Probability
Age
Not Metropolitan
Metropolitan
Appendix 1: Figures
31
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Appendix 1: Figures
Figure A11. Interaction effect of age and postgrad dummy (Model 3).
Figure A12. Interaction effect of age and single dummy (Model 3).
Source: KAPSARC.
Source: KAPSARC.
.8
.6
.4
.2
20 40 60 80
Probability
Age
Not Postgrad
Postgrad
.7
.6
.5
.4
.3
.2
20 40 60 80
Probability
Age
Not Single
Single
32
Eliciting Preferences of Ride-Hailing Users and Drivers
Appendix 1: Figures
Figure A13. Predicted probability and the marginal effect of total vehicle ownership (Model 3).
Figure A14. Predicted probability and the marginal effect of age (Model 4).
Source: KAPSARC.
Source: KAPSARC.
.65
.6
.55
.5
.45
Predicted Probability (with 95% CIs)
Probability
Total Vehicles
012345
.04
.03
.02
.01
0
1.01
Marginal Effects (with 95% CIs)
Probability Difference
Total Vehicles
0 1 2 3 4
.6
.5
.4
.3
.2
.1
Predicted Probability (with 95% CIs)
Probability
Age
20 40 60 80
-.002
-.004
-.006
-.008
-.01
Marginal Effects (with 95% CIs)
Probability Difference
Age
20 40 60 80
33
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
Appendix 1: Figures
Figure A15. Interaction effect of age and postgrad dummy (Model 4).
Figure A16. Interaction effect of age and single dummy (Model 4).
Source: KAPSARC.
Source: KAPSARC.
.7
.6
.5
.4
.3
.2
20 40 60 80
Probability
Age
Not Postgrad
Postgrad
.6
.5
.4
.3
.2
20 40 60 80
Probability
Age
Not Single
Single
Eliciting Preferences of Ride-Hailing Users and Drivers 34
Eliciting Preferences of Ride-Hailing Users and Drivers
Appendix 1: Figures
Figure A17. Predicted probability and the marginal effect of total vehicle ownership (Model 3).
Figure A18. Predicted probability and the marginal effect of age (Model 4).
Source: KAPSARC.
Source: KAPSARC.
.5
.45
.4
.35
.3
Predicted Probability (with 95% CIs)
Probability
Income
0 100,000 200,000 300,000
Probability Difference
.001
0
-.001
-.002
-.003
-.004
Marginal Effects (with 95% CIs)
Income
0 100,000 200,000 300,000
.6
.5
.4
.3
.2
0100,000 200,000 300,000
Probability
Income
Not Single
Single
35
Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers Eliciting Preferences of Ride-Hailing Users and Drivers
About the Project
The future of light-duty vehicle demand project aims to explore the impact of policy, technology
advancements and consumer attitudes on personal mobility choices and energy demand. It focuses
on the impact of the rise of mobility-on-demand services on key aspects of travel behavior. The
research combines psychographic and demographic data with mobility data to understand the
motivations of mobility-on-demand drivers and users and how this will shape the future of energy in
both developed and developing economies.
About the Authors
Prateek Bansal
Akanksha Sinha
Rubal Dua
Ricardo Daziano
Prateek Bansal is a postdoctoral research fellow at Imperial College
London, working primarily on Bayesian Machine Learning methods and
causal inference models with applications in transport systems. He holds
a Ph.D. from Cornell University, an M.Sc. degree from The University of
Texas at Austin, and a B.Tech. from the Indian Institute of Technology (IIT)
Delhi.
Akanksha Sinha is a transportation engineering assistant at DKS
Associates where she is working on a variety of transportation projects.
She holds an M.Sc. from Cornell University and Colorado State University
and a B.Arch from Pune University.
Rubal is a research fellow at KAPSARC working on vehicle
regulatory policy and shared mobility research from the consumer
perspective. He holds a Ph.D. from KAUST, Saudi Arabia, an M.Sc. from
the University of Pennsylvania, and a B.Tech. from the Indian Institute of
Technology, Roorkee.
Ricardo is an associate professor at the School of Civil and Environmental
Engineering and Systems Engineering at Cornell University. His
research goal is to better understand the interplay of consumer behavior
with engineering, investment, and policy choices for energy-efcient
technologies. He holds a Ph.D. in Economics from Laval University,
Canada.
36
Eliciting Preferences of Ride-Hailing Users and Drivers
www.kapsarc.org
