Content uploaded by Sharareh Kermanshachi
Author content
All content in this area was uploaded by Sharareh Kermanshachi on Nov 03, 2022
Content may be subject to copyright.
Content uploaded by Sharareh Kermanshachi
Author content
All content in this area was uploaded by Sharareh Kermanshachi on May 26, 2022
Content may be subject to copyright.
Research Article
Transportation Research Record
2022, Vol. 2676(11) 462–478
ÓNational Academy of Sciences:
Transportation Research Board 2022
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/03611981221093632
journals.sagepub.com/home/trr
Modeling Users’ Adoption of Shared
Autonomous Vehicles Employing
Actual Ridership Experiences
Roya Etminani-Ghasrodashti
1
, Ronik Ketankumar Patel
2
,
Sharareh Kermanshachi
2
, Jay Michael Rosenberger
3
, and Ann Foss
4
Abstract
Despite the growing interest in implementing shared autonomous vehicles (SAVs) as a new mobility mode, there is still a lack
of methodologies to unpack SAV adoption by individuals after experiencing self-driving vehicles. This study aimed to fill this
gap by analyzing data collected from a users’ survey of a self-driving shuttle piloted downtown and on a university campus in
Arlington, TX. Employing structural equation modeling, the hypothesized relationships between SAV adoption and key factors
were tested. Data analyses indicated that individuals with limited access to a private vehicles, low-income people, young
adults, university students, males, and Asians were more likely to ride this new service. Furthermore, results showed that
SAV service attributes, including internal and external service performance and usual transportation mode, affected users’
willingness to continue using the service in the future. The study also highlighted the role of trip waiting time, -purpose, and -
frequency on SAV adoption. Our model simultaneously considered usual transportation mode and trip frequency as factors
that could mediate the role of vehicle ownership on SAV adoption. The results suggested that participants with greater access
to a private vehicle were strongly interested in using private vehicles and less likely to use the ridesharing alternative, conse-
quently they less frequently used the piloted SAV. The outcomes from this study are expected to inform planners with
advanced knowledge about emerging technology to help them to adjust SAV policies before autonomous vehicle services are
fully on the roads.
Keywords
Autonomous vehicles, shared self-driving, technology adoption, service attributes
Autonomous vehicles (AVs) are an emerging technology
that combines multiple sensors, computer processors,
and repositories to take over the tasks or responsibilities
otherwise undertaken by human operators (1,2). AV
technology is promising to alter the fundamentals of
human travel behavior, leading to significant social and
infrastructural changes (3). AVs could also solve major
transportation issues by providing a wide range of bene-
fits in relation to safety, increased mobility, and efficiency
(4). Enhancing the mobility options for nondrivers and
people with disabilities, and the ability to reduce car use
and vehicle crashes, traffic congestion, and fuel consump-
tion are potential advantages of AV technology (5).
Accordingly, public inclination toward AV technology
and its potential benefits will play a crucial role in the
adoption of AVs (6).
Shared autonomous vehicles (SAVs) are autonomous
taxis that allow riders to call an AV with the help of a
mobile application instead of searching for and walking
to an available car or taxi (7,8). As a result, SAVs could
become an attractive mobility option for elderly and low-
income people who have limited access to private modes
1
Center for Transportation Equity, Decisions and Dollars (CTEDD),
University of Texas at Arlington, Arlington, TX
2
Department of Civil Engineering, University of Texas at Arlington,
Arlington, TX
3
Department of Industrial, Manufacturing, and Systems Engineering,
University of Texas at Arlington, Arlington, TX
4
Office of Strategic Initiatives, City of Arlington, Arlington, TX
Corresponding Author:
Sharareh Kermanshachi, Sharareh.kermanshachi@uta.edu
(9). SAVs could also provide first- and last-mile solutions
by offering rides on lower-demand routes and comple-
menting existing public transit services. Moreover, since
users will not have to interact with the vehicle, they could
use the extra time to perform productive activities or sim-
ply to relax (9).
To predict the short- and long-term adoption of AV
technology and promote the acceptance of SAVs, it is
essential to identify potential AV user attributes and
related travel behavior and patterns (8). Recent research
has extensively explored the general public’s perceptions
of AVs in relation to being potential AV and SAV users
(10,11). Because of the limited real-life experiences of
AVs, past studies have typically relied on disaggregated
approaches and gathered required the data through sur-
veys of potential consumers (12–14). The surveys mainly
focus on travel attitudes and perceptions to identify
potential adoption of AVs/SAVs (15–17) and willingness
to pay (18–21).
Reviewing the existing studies revealed certain issues
that have not yet been addressed. First, previous research
has often been developed based on a sample of people
who have no experience of riding an AV/SAV (8,22).
Hence, there is still a gap in this area of research as a lim-
ited set of methodologies analyzed factors affecting the
adoption of SAVs by individuals who experienced this
technology. This study sought to fill this gap by evaluat-
ing users’ acceptance and adoption of SAVs by utilizing
actual SAV users’ experiences. To contribute to the
knowledge about SAVs, we identified current usage
and future adoption of a pilot project by examining a
comprehensive set of determinant factors. This study
developed a SAV adoption framework by evaluating
factors suggested by the literature, as well as innovative
factors such as service performance. Second, most of
the earlier studies were conducted in large metropolitan
areas with ample access to public transit options (12,
23,24). This study centered on the actual users’ adop-
tion of SAVs in a low-density city without access to
fixed-route transit, where most of the population relies
on private cars. Consequently, this lack of access to
conventional transit considerably affects the travel pat-
terns of the residents compared with metropolitan
areas with access to multiple public transit options.
Moreover, this research studied SAV adoption based
on a self-driving fleet integrated with an on-demand
ride service.
Given the above motivation, we analyzed data from a
self-reported survey administered to the consumers of a
SAV pilot project operating in Arlington, TX. We col-
lected the data as part of a SAV performance evaluation
7 months after its deployment in 2021. A sample of 215
users completed a comprehensive online survey after rid-
ing the SAV service. The survey asked consumers a range
of questions about their perceptions of SAVs, which pro-
vided the opportunity to examine the role of trip features,
service attributes, and socioeconomic- and demographic
attributes on the adoption of SAV in the long term. The
research findings are expected to support transportation
planners and policy makers in recognizing the determin-
ing factors in promoting the efficiency of SAV services.
Our results also provide new insights into developing pol-
icy implementation for the SAV market. Finally, we
anticipate that identifying the perceived service quality
by actual users will lead to more efficient evaluations of
SAV technology adoption and its progressive impacts.
Literature Review
SAVs have received much attention recently as a new
technology that can deal with equity and efficiency chal-
lenges in transportation. Accordingly, multiple studies
have explored AV technology (25,26) and the adoption
of SAVs (4,11,27,28) through use of stated preference
surveys, focus group discussions, and simulation meth-
ods. A rich body of literature has explored AVs and
contributed to the adoption of AVs and SAVs.
Sociodemographic factors and individuals’ attitudes, per-
ceptions, and concerns are suggested as the most preva-
lent determinant factors of AV and SAV adoption (9,
18,24,27,29,30). Therefore, we briefly review the most
relevant literature to this study as follows.
To address the acceptance and diffusion of driverless
vehicles, several studies have applied behavioral concep-
tual frameworks and explored the effects of attitudes,
beliefs, concerns, perceptions, and preferences in relation
to self-driving vehicles. For instance, Acheampong and
Cugurullo developed three conceptual models to predict
AV sharing, AV public transport services, and AV own-
ership (27). They found that 57% of the participants
believed AVs would provide utilitarian benefits like
reduced crashes, lower traffic congestion, and less envi-
ronmental pollution. Most participants (70%) were posi-
tive about using SAVs and believed they would provide
monetary and environmental benefits. Song et al. carried
out a stated preference survey in two small urban areas
of Wisconsin, to understand people’s attitudes toward
AVs (26). They found that tech-savvy people were more
likely to have a positive attitude toward integrating AVs
than those in rural areas. The results also showed a sig-
nificant difference in perception toward AV transit inte-
gration between transit and nontransit users. Jiang et al.
investigated the spatial distributions of attitudes and pre-
ferences of a population of U.S. adults toward purchas-
ing, sharing, and using privately owned and shared AVs
(28). They found that 24.2% of the participants were
willing to use a taxi service that was self-driving. In com-
parison, 21.1% of the participants were willing to use a
Etminani-Ghasrodashti et al 463
self-driving taxi that was shared with someone else.
Mohammadzadeh applied reflexive thematic analysis to
test the role of hegemonic ideology in the utilization of
SAVs and collected data from 192 residents via stated
preference surveys and three focus group interviews in
Auckland, New Zealand (31). Therefore, vehicle owner-
ship and use are suggested to have different ideological
meanings and symbolic functions for different sociode-
mographic based on gender, education, age, and ethni-
city. The results suggested that the hegemonic ideology
promotes automobility in society, which could direct pri-
vate car users toward purchasing AVs instead of using
shared mobility or SAVs. A study by Haboucha et al.
distributed a stated preference survey in Israel and North
America among 721 individuals to understand user pre-
ferences for choosing between owning and sharing AVs
(4). They found that only 75% of individuals would
choose SAVs even if the services were provided free of
charge, and 44% of individuals were inclined to use regu-
lar vehicles instead of AVs. Nazari et al. explored trave-
lers’ safety concerns about AV technology through a
recursive bivariate ordered probit model using a data set
provided by the California Energy Commission (25). The
results indicated that AV safety concerns could nega-
tively influence adoption of AVs. The results from a
study by Etzioni et al. revealed that people who do not
like using transit or sharing their seats with strangers
were less likely to choose SAVs (32). People with orga-
nized lifestyles who emphasize multitasking while riding
were more likely to choose SAVs. A study by Yuen et al.
in Da Nang, Vietnam, integrated several theoretical per-
spectives such as unified theory of acceptance and use of
technology 2 (UTAUT2) and theory of unplanned beha-
vior to identify factors affecting the adoption of SAVs
(11). The results suggested that attitude, subjective norms,
perceived behavioral control, and perceived behavioral
conditions can significantly affect intention to use SAVs.
Some studies have emphasized the importance of ser-
vice attributes in adopting AVs. According to Krueger
et al., service costs, travel, and waiting times are critical
components in adopting SAVs (9). A study by Jing et al.
demonstrated that safety, mobility, performance-to-price
value, and the value of travel time were among the signif-
icant factors affecting acceptance of AVs (33). They
found that individuals who perceived a higher value for
travel time were more likely to buy or use AVs. Similarly,
Haboucha et al. indicated that people with longer daily
travel times and travel distances were more likely to
believe in the advantages of using AVs (4).
Sociodemographic factors are of great interest in most
AV adoption studies. A rich line of research has evalu-
ated socioeconomic differences in relation to acceptance
of AV technology. Wang and Zhao used stated prefer-
ence surveys collected from 1,142 individuals in
Singapore to study the adoption of AVs and identified
how risk preference parameters are associated with the
socioeconomic characteristics of people and their adop-
tion of AVs (24). The results suggested that highly edu-
cated and higher-income individuals were risk-seeking
and more likely to adopt new technology with small
potential gains. Wang and Akar used data from 2015
and 2017 Puget Sound Regional Council surveys consist-
ing of 5,500 households to study the factors associated
with the adoption of AVs (12). According to their
results, females are less likely to commute alone in AVs
than males, and young people (18 to 34years old) are
more likely to adopt self-driving technology. They also
reported that people driving electric vehicles are more
likely to commute using AVs than people who own diesel
vehicles, whereas people using public transit and those
with flexible work schedules are more likely to adopt
SAVs. Furthermore, higher-income individuals living in
metropolitan areas and people who have had an accident
in the past are more likely to adopt AVs (34).
Although some general similarities exist in the
reviewed literature of AV and SAV adoption, the find-
ings from the disaggregated studies were produced by
surveys that did not entirely follow a homogenous pat-
tern. Underlying reasons for this might include different
sample sizes, diverse socioeconomic attributes, and varia-
tions in geographical locations (18). However, one chief
deficiency of the current literature stems from the sam-
ples’ unfamiliarity with automated technology and lack
of riding experiences. Empirical evidence is needed to
identify the behavioral responses of actual users of AVs
and SAVs. Although the earlier studies inform about the
interaction between individuals’ attitudes, preferences,
and potential adoption of self-driving services, they still
omit certain important research questions. Since the lit-
erature mainly focuses on potential riders without actual
ridership knowledge, it has neglected riders’ evaluations
and perceptions of automated service performance and
features that will affect acceptance of self-driving vehi-
cles, particularly SAVs.
Accordingly, this study sought to fill these research
gaps by providing an early glimpse of answers to the fol-
lowing questions: 1. How do people request SAV rides in
a city with a shared self-driving fleet integrated into an
existing rideshare service? 2. To what extent do factors
such as SAV service attributes, SAV trip features, usual
transportation mode, and sociodemographics shape the
future adoption of an SAV service? 3. How do SAV
riders’ responses to the adoption of the SAVs after
experiencing it correspond with the results from empiri-
cal studies about potential users? We developed our
study to identify the inclination to use SAV as a mode of
transportation in the near future, while focusing on users
learning about this technology by riding it in real life.
464 Transportation Research Record 2676(11)
Methodology
Study Area, Survey Design and Sample Preparation
In this study, we centered on a pilot project called
RAPID (Rideshare, Automation, and Payment
Integration Demonstration) that was deployed in the city
of Arlington, TX, as of March 2021. The city was
awarded an Integrated Mobility Innovation grant from
the Federal Transit Administration to link an existing
on-demand rideshare service (Via) with AV technology.
The Via rideshare service was launched in December
2017 and has since expanded the initial deployment to
serve the whole city as a fully app-based on-demand ser-
vice that provides rides to customers in one of six-
passenger vans anywhere within the service boundary
(35). By partnering with Via Transportation, May
Mobility, and the University of Texas at Arlington
(UTA), the city offers Level 4 AV service in downtown
Arlington and the UTA campus that are integrated with
the on-demand mobile application using the Via plat-
form. The service is fully on-demand, and it operates
four hybrid Lexus vehicles and one Polaris GEM vehicle
equipped to carry a wheelchair rider. Rides are provided
from 7 a.m. to 7 p.m. Monday through Friday. These
vehicles travel up to speeds of approximately 25 mph
(36). Passengers can book a ride by self-driving vehicle
to and from any destination within the service area
through the Via app. Therefore, RAPID is intended to
provide door-to-door service. RAPID is a Level 4
AV, meaning that the vehicle can sense its environment
and operate without human involvement. To meet the
safety concerns of passengers, vehicles are monitored by
fleet attendants to ensure a safe and enjoyable experience
for the riders (36). To meet the equity goals of the pilot
project, RAPID allows university students to ride the
AVs for free (35,37).
The RAPID service area includes the UTA campus
and several residential and business uses in downtown
Arlington. Targeting a transportation equity and accessi-
bility goal, the service boundary was selected in an area
with a 39% poverty rate in which 18.8% of households
have a person with a disability and 11% of households
lack access to a private vehicle—almost three times
greater than the citywide average of 4.3% of households.
Figure 1 represents the RAPID service boundary at the
time of the study.
The evaluation of SAV performance across several
metrics (e.g., efficiency, safety, accessibility, and cus-
tomer satisfaction) was defined as one of the main tasks
of the RAPID project. Accordingly, we created one short
survey and one long survey to solicit assessments of the
service by both actual and potential users. This study
focused on the sample completing the SAV short survey.
The survey was designed to collect data on the main
determinants of SAV adoption by actual riders. The sur-
vey was reviewed and approved by the university
Institutional Review Board. We developed an online sur-
vey and utilized the QuestionPro platform to create a
survey link.
We administrated the questionnaire as an online sur-
vey to the riders who had requested and completed a ride
with RAPID SAVs. The survey was distributed through
various outlets, including scannable QR-codes, question-
naire URLs, emails to UTA students, faculty, staff, and
links shared in social media networks, including
WhatsApp. To increase the participation rate, the City
of Arlington and Via Transportation assisted in distri-
buting the survey link and recruited RAPID riders from
the general public through email and social media. The
first part of the survey included a set of questions
designed to gather data related to SAV ridership and
RAPID trip features. The second was intended to
address the daily travel patterns of the respondents, and
the final section was allocated to sociodemographic
questions.
Data and Key Variables
Of the total number of riders that received the survey,
1,196 people viewed it, 388 respondents answered the sur-
vey, and 252 individuals completed it. Accordingly, our
data included 252 cases. The primary goal of RAPID
SAV was to provide service to university students and
the general public with limited or no access to a private
vehicle. Although the general public would pay for SAV
rides, RAPID is free of charge to students. The real-time
data platform of SAV provided us with the proportion of
student to nonstudent RAPID riders. Based on actual
data, about 98% of rides were taken by students, the
remainder by the general public, by January 2022.
Accordingly, the study sample represented the target and
actual population of the RAPID project.
SAV Trip Features. Since the survey was designed to recruit
riders after completing a SAV ride, the first part of the
questionnaire asked respondents about their last trip with
RAPID SAVs. SAV trip features were explored through
questions about trip purpose and trip waiting time for
the latest RAPID ride. The trip purpose was identified
through a list of destinations including using SAVs to
travel to work, school, shopping, medical places, social/
recreational places, daycare/childcare, returning home,
and other (categorical variables). The description of the
data and variables in Table 1 indicate that 50% of the
riders used SAV service for going to work and school.
Riding SAVs to return home was the other reason for
requesting RAPID SAV rides.
Etminani-Ghasrodashti et al 465
Figure 1. RAPID shared autonomous vehicle (SAV) service area.
Table 1. Description of Key Variables (n= 252)
Variables Description Frequency Percent RAPID area (%)
Age 18–24 135 53.6 40
25–34 93 36.9 30
35–44 14 5.6 10
45–54 4 1.6 7
55–64 4 1.6 6
65+ 1 .4 7
Missing 1 .4
Gender Female 71 28.2 48
Male 166 65.9 52
Other 3 1.2 NA
Prefer not to answer 7 2.8 NA
Missing 5 2.0 NA
Being a university student
a
Yes 195 77.4 85
No 53 21.0 15
Missing 4 1.6 NA
Race
b
American Indian or Alaska Native 4 1.6 0.7
Native Hawaiian or Pacific Islander 0 0.00 0
Asian 145 57.5 23.6
White 47 18.7 55.7
Black or African American 32 12.7 16.8
Others 20 7.9 3.2
Missing 4 1.6 NA
Ethnicity
c
Hispanic 29 11.5 21
Non-Hispanic 217 86.1 79
Missing 6 2.4 NA
Household income Less than $20,000 136 54.0 47.6
$20,000–$34,999 44 17.5 13.1
$35,000–$49,999 16 6.3 12.7
$50,000–$74,999 20 7.9 13.1
(continued)
466 Transportation Research Record 2676(11)
Table 1. (continued)
Variables Description Frequency Percent RAPID area (%)
$75,000–$99,999 12 4.8 7.9
$100,000 or more 9 3.6 5.6
Missing 15 6.0 NA
Vehicle ownership Zero vehicles 120 47.6 16.8
One vehicle 80 31.7 41.5
Two vehicles 25 9.9 32.4
Three vehicles or more 20 7.9 9.3
Missing 7 2.8 NA
SAV trip features
SAV trip purpose Going to work 39 15.5 NA
Going to school 89 35.3 NA
Going shopping 18 7.1 NA
Going to a medical place 7 2.8 NA
Going to social/recreational places 19 7.5 NA
Going to daycare/childcare 0 0 NA
Returning home 59 23.4 NA
Others 21 8.3 NA
SAV trip waiting time Less than 5 min 72 28.6 NA
About 5–10 min 80 31.7 NA
About 10–20 min 66 26.2 NA
About 20–30 min 27 10.7 NA
More than 30 min 7 2.8 NA
SAV trip frequency This is my first time 82 32.5 NA
This is my second time 33 13.1 NA
About once per month 11 4.4 NA
About twice per month 21 8.3 NA
About once per week 23 9.1 NA
More than two times per week 81 32.1 NA
Missing 1 .4
Ridership experiences (all apply) Via on-demand ridesharing service 225 89.3 NA
Handitran paratransit service 2 .8 NA
Milo AV Pilot 1 .4 NA
Drive.ai AV Pilot 1 .4 NA
UTA transportation service 130 51.6 NA
Public transit in another city 89 35.3 NA
I have not used public transit before 18 7.1 NA
Usual transportation mode Private vehicle 63 25.0 NA
Private app-based ride services, such
as Uber or Lyft
34 13.5 NA
Via on-demand ridesharing service 69 27.4 NA
Handitran paratransit service 2 .8 NA
University transportation service 21 8.3 NA
Walking/biking 49 19.4 NA
RAPID SAV service 5 2.0 NA
Other 7 2.8 NA
Missing 2 .8 NA
Adoption of SAV Strongly disagree 2 .8 NA
Disagree 2 .8 NA
Neutral 20 7.9 NA
Agree 64 25.4 NA
Strongly agree 162 64.3 NA
Not applicable for this ride 2 .8 NA
SAV service attributes Mean SD NA
Internal service performance Normalized factor 100 25 NA
External service performance Normalized factor 100 25 NA
a
Percentage of university students for the RAPID service area is measured based on the ‘‘in school (college/grad) population’’ divided by ‘‘total over
25 years of age population’’ in the RAPID service area.
b,c
Percentage of race and ethnicity categories in RAPID service area are only available based on the total population and not based on age categories.
Note: SAV = shared autonomous vehicles; AV = autonomous vehicles; UTA = University of Texas at Arlington; SD = standard deviation.
Etminani-Ghasrodashti et al 467
To address trip waiting time, respondents were given
five options from less than 5min to more than 30 min for
the question ‘‘how long did you wait for the SAV vehicle
after requesting the ride?’’ The results from the trip wait-
ing time indicated that approximately 63% of respon-
dents reported a trip waiting time less than 10 min,
whereas 37% of the trips were reported to have more
than 10 min waiting time (see Table 1).
SAV Service Attributes. The literature suggests that service
attributes can affect individuals’ preferences toward
adoption and acceptance of future SAVs (9).
Accordingly, the survey presented respondents with a set
of 10 statements on SAV users’ perceptions of different
service attributes after their RAPID ride. To develop the
statements about service attributes, we referred to focus
group discussions with members of the university com-
munity, people with disabilities, and members of the gen-
eral public, which were performed before deploying the
RAPID SAVs (38). These focus group discussions identi-
fied the main preferences and concerns of potential users
about the proposed SAV service, including the capacity
of the service, trip cost, safety, disability-friendly features
of the service, accessibility, and reliability (38,39). To
explore the concerns and preferences extracted from the
SAV predeployment studies, we asked respondents to
evaluate the SAV service features: ease of booking and
scheduling SAV rides through the mobile application,
waiting time, the convenience of pickup and dropoff
locations, ease of boarding, comfort of the seats, climate
control, speed of the vehicles, feeling safe in the vehicles,
and feeling safe when sharing rides with other passen-
gers. Table 2 presents these 10 variables (V1 to V10).
Response options were on a five-point Likert-type scale
(where 1 = ‘‘strongly disagree’’ to 5 = ‘‘strongly agree’’).
Because the 10 factors describing the SAV service attri-
butes were intercorrelated, we conducted a confirmatory
factor analysis (CFA) and reduced the dimensionality of
factors addressing the SAV service attributes. Statements
loaded above 0.5 were considered under one factor. To
extract the factors, we selected the principal components
method based on a fixed number of factors equal to 2.
Several methods exist to estimate factor scores including
regression, Bartlett’s test, and the Anderson–Rubin test.
We used Bartlett scores because this method estimates
scores unbiased and have zero correlations with non-cor-
responding scores (40). To assess the correlation matrix
and evaluate the appropriateness of using factor analysis
on the data set, we used the Kaiser–Meyer–Olkin (KMO)
test and Bartlett’s test of sphericity. Two factors were
loaded through factor analysis (50% variance explained,
KMO = .903), including (1) ‘‘internal performance’’ and
(2) ‘‘external performance’’ (see Table 2).
SAV Trip Frequency. The survey had a question about the
frequency of RAPID usage: ‘‘How often have you ridden
RAPID so far?’’ Responses were selected from six
options ranging from ‘‘This is my first time’’ to ‘‘More
than two times per week.’’ Exploring the responses indi-
cated that 32% and 13% of users had experienced their
first and second ride with the SAV fleet, respectively,
while 33% of respondents used it more than twice per
week.
Ridership Experiences. The survey asked about the respon-
dents’ past experiences with the available transportation
options in Arlington. The response options included
available and prior transportation: on-demand rideshar-
ing service, a paratransit service, university transporta-
tion (including on- and off-campus bus shuttles and a
late-night golf-cart service), and former AV pilot ser-
vices. We asked respondents to select all transportation
options that applied to them. The ridership experiences
Table 2. Pattern Matrix for Shared Autonomous Vehicle (SAV) Service Attributes
Please share your opinion of your last ride about the following options
(from 1 = strongly disagree to 5 =strongly agree)
Loaded factors for SAV service attributes
Internal performance External performance
V1: Booking and scheduling my RAPID trip using the Via app was easy na .527
V2: The price for riding RAPID was reasonable na .434
V3: The waiting time was reasonable na .772
V4: The pickup and dropoff locations were convenient na .508
V5: Boarding the vehicle was easy .561 na
V6: The seats in the vehicle were comfortable .759 na
V7: The climate control in the vehicle was appropriate .606 na
V8: The speed of the vehicle was reasonable .627 na
V9: I felt safe when riding RAPID .702 na
V10: I felt safe while sharing the vehicle on my RAPID ride with other passengers .690 na
Note: na = not applicable.
468 Transportation Research Record 2676(11)
of the respondents showed that about 90% of riders had
used the Via on-demand ridesharing service, 52% had
used the university transportation service, and about
36% of the respondents had ridden public transit in the
past.
Usual Transportation Mode. The survey asked respondents
to indicate their most frequent travel mode choice from
private vehicles, private app-based ride services (e.g.,
Uber or Lyft), Via Service, Handitran Paratransit service,
UTA transportation, walking/biking, and RAPID ser-
vice. The descriptive statistics of the respondents’ usual
transportation mode revealed that the Via on-demand
ridesharing service and private vehicles were the respon-
dents’ typical modes of transportation (29% and 24%,
respectively). Active transportation (walking/biking) and
private app-based ride services were the other usual travel
modes of respondents.
Adoption of SAV. The main aim of this study was to explore
the adoption of the SAV by individuals after riding and
experiencing the service. Accordingly, the survey included
a question on whether the rider ‘‘would ride RAPID
again in the future.’’ Responses were given on a five-point
Likert-type scale (where 1 = ‘‘strongly disagree’’ to
5 = ‘‘strongly agree’’). The extent to which they agreed
they would use the SAV in the future would indicate their
likely decision on adoption of this new service. As shown
in Table 1, about 89% of respondents would be interested
in requesting and using SAVs in the future, indicating sig-
nificant SAV service adoption among the current users.
Sociodemographics. The last part of the survey related to
the socioeconomic characteristics of the riders. We asked
respondents about their age group (ordinal variable),
gender (categorical variable), race and ethnicity (catego-
rical variable), household income (ordinal variable), and
vehicle ownership (continuous variable). Table 1 presents
the sociodemographic information about the sample in
comparison to the RAPID service area attributes. In gen-
eral, respondents were primarily young adults (18 to
34 years) and university students. Based on the 2019
American Community Survey (www.census.gov/pro-
grams-surveys/acs/data.html), 70% of the SAV service
area population are young adults and college students.
The gender distribution in the sample was far from the
proportions of the RAPID service area; the proportion
of female respondents (28.2%) to males (65.9) was far
from the RAPID area gender distribution. Therefore, it
seems that males were more interested in participating in
the survey. Moreover, the majority of SAV users were
Asian and non-Hispanic, whereas the White population
is prevalent in the RAPID service area. Respondents
were mostly from low-income households with less access
to private vehicles. The income categories in the sample
population represent the SAV area income.
We also ran a crosstab analysis and identified the
demographic characteristics of those using private vehi-
cles as their usual mode of transportation. Twenty-five
percent of respondents, the majority of whom were
young adults (18 to 35 years old) and university students,
declared that their usual mode of transportation was pri-
vate vehicle. About 32% of private vehicle users were
Asian, 38% were White, 13% were Black African
American, and 17% were from other races. Almost 55%
of private vehicle users resided within the SAV
service area, on- and off-campus, and downtown.
Approximately 47.5% of private vehicle users had an
annual income of less than $20,000, and 22% earned
$20,000 to 34,999 per year. Among those who reported
private vehicle use, about 13% did not have a vehicle,
52% had one vehicle, and the remainder had two, three,
or more vehicles.
Conceptual Framework and Analytical Method of SAV
Adoption
Multiple empirical studies have explored preferences and
the adoption of SAVs by individuals while considering
sociodemographic characteristics, personal attitudes
toward AV technology, and land use attributes (9,18).
However, owing to the scarcity of self-driving shuttle
demonstrations and the lack of access to ridership data,
most studies have used simulation-based fleet evolution
frameworks to predict the long-term adoption of AVs
and SAVs (8) and relied on the responses of potential-
but not actual users of SAVs (9). To address this gap, we
developed our conceptual model based on certain
assumptions.
First, we assumed that understanding the long-term
adoption of SAVs as a new technology would require
awareness of individuals’ short- and medium-term travel-
related factors (41). The extent to which technology
might affect individuals’ travel behavior will depend on
their long-term decisions (such as choice of residential
neighborhood), medium-term decisions (such as vehicle
ownership and usual travel mode choices), and short-
term decisions such as daily activities (42). Accordingly,
the conceptual framework of this study predicted SAV
adoption behavior in the long term while accounting for
users’ medium- and short-term decisions. Adoption of
SAVs in the future will depend on its users’ medium- and
short-term decisions. We classified factors such as vehicle
ownership and usual transportation mode as those that
could be shaped through individuals’ decision-making in
the medium term (dark gray boxes in Figure 2). SAV trip
features, SAV service attributes, and SAV trip frequency
Etminani-Ghasrodashti et al 469
are factors that could be formed through one’s decision
making in the short-term (light gray boxes in Figure 2).
Secondly, the utility derived from individuals’ existing
travel mode choices, use of public transport, and car
ownership has an important role in predicting adoption
of AVs as a shared and public transport mode (27).
Accordingly, past studies developed their conceptual
models based on individuals’ attitudes toward public
transport and vehicle ownership. In this study, instead of
perceptions and attitudes toward alternative transporta-
tion modes, we considered the self-reported usual trans-
portation mode, and vehicle ownership as the utility
derived from existing travel modes.
Third, we assumed that SAV trip features, SAV ser-
vice attributes, SAV trip frequency, ridership experience,
and socioeconomic status would be the factors that
directly influenced the perceptions of users toward SAV
rides and encourage them to accept and ride the service
again in the future. Conversely, the usual mode of trans-
portation and SAV trip frequency could affect adoption
of SAVs through vehicle ownership. In other words, the
usual mode of transportation and SAV trip frequency
could mediate the role of vehicle ownership on adoption
of SAVs. Since our model counts simultaneously both
direct and indirect relationships between key variables,
we utilized a structural equation model (SEM). SEM has
the ability to estimate multiple correlations between key
variables simultaneously. Basically, SEM has two parts
including the ‘‘measurement model’’ and the ‘‘structural
model.’’ The measurement model specifies the relation-
ship between the observed and the latent variables, and
the structural model identifies the relationships between
the measurement models.
SEM can comprise three sets of equations, including
1. measurement models for endogenous variables, 2.
measurement models for exogenous variables, and 3.
structural relationships between all measurements simul-
taneously. The primary framework of this study included
one measurement model for SAV service attributes. This
measurement model was constructed based on the factor
analysis presented in Table 2. As described earlier, we con-
ducted a CFA to explore the latent constructs underlying
the observed variables and identified the best patterns of
covariance among the observed variables describing SAV
service attributes. Internal SAV was a latent variable
loaded through six observed factors indicating the users’
perceptions of the quality of elements directly related to
the performance of AVs, such as ease of boarding the vehi-
cle, seat comfort, indoor climate control, speed of the vehi-
cle, riding, and ridesharing safety (see Table 2, V5 to V10).
These six factors had loadings above 0.5. The external per-
formance, evaluated through riders’ satisfaction about fea-
tures related to the service apart from the vehicle, was
loaded through four observed variables with scores over
0.4, such as ease of booking and scheduling the trip, price
of the trip, waiting time, and pickup and dropoff locations
(see Table 2 V1 to V4). The loadings of each item in each
latent variable are reported in Table 2. Thus, the latent
variables extracted through CFA were entered into the
final SEM as observed variables (43). Consequently, we
considered the extracted factors from CFA as observed
exogenous variables (44), and all other variables in the
model were treated as the observed variables, and the
structural relationships between exogenous and endogen-
ous variables were simultaneously identified,
Y=BY +GX+z
where
Y=(Ny 31) column vector of endogenous variables
(Ny = number of endogenous variables),
Trip Waiting Time
Trip Purpose
Adoption of SAV
SAV Trip
Frequency
Socio-economics of
users
Vehicle Ownership
Usual Transportation
Mode
SAV Service Attributes
SAV Trip Features
External
Performance
Internal
Performance
V1
V2
V3
V4
V5
V6
V7
V8
V9
V10
Figure 2. Structure framework relationships of the study.
Note: SAV = shared autonomous vehicle.
470 Transportation Research Record 2676(11)
X=(Nx 31) column vector of exogenous variables
(Nx = number of endogenous variables),
B=(Ny 3Ny) matrix of coefficients demonstrates the
direct effects of endogenous variables on each other,
G=(Ny 3Nx) matrix of coefficients demonstrates the
direct effects of exogenous variables on endogenous vari-
ables, and
z=(Ny 31) column vector of errors (45).
Results
Determinant Factors of Adoption of SAV
All links designated in the hypothesized conceptual
model (Figure 2) were examined empirically; however,
the construct of the structural model in the final model
was a little different from the conceptual framework we
suggested earlier. To improve the validity of the results,
we eliminated certain factors that had collinearity effects,
such as income effects on vehicle ownership, and rider-
ship experience influence on usual transportation mode.
Therefore, we dropped income and ridership experience
from the final model. We also eliminated certain catego-
rical variables that were not statistically significant to
improve the model fitness. Figure 3 illustrates the direct
effects of variables, whereas Table 3 gives the correlation
estimates (standardized direct effects) for the hypothe-
sized relationships among key variables. Ordinal vari-
ables, including the adoption of SAV, SAV trip waiting
time, SAV trip frequency, and vehicle ownership were
coded in ascending order, so we treated them as
Table 3. Path Coefficient Estimates for Standardized Direct Effects Between Key Variables and Adoption of Shared Autonomous Vehicles
(SAVs)
Estimate SE P
Standardized direct effects from key variables (#) on adoption of SAV
SAV trip waiting time 2.068 .028 .09*
SAV trip purpose 2.400 .085 .000***
Work
School 2.512 .064 .000***
Return home 2.479 .073 .000***
Base category 2.473 .070 .000***
SAV service attributes .644 .034 .000***
Internal performance
External performance .261 .037 .000***
SAV trip frequency .038 .014 .369
Usual transportation mode .138 .077 .002***
Private vehicle
Via on-demand ridesharing service .134 .070 .002***
Private app-based ride service .125 .090 .003***
Walking/biking .110 .078 .009***
Base category .043 .089 .308
Vehicle ownership 2.042 .004 .374
Age 2.001 .061 .983
18–24
25–34 2.016 .063 .696
35–44 2.090 .134 .031**
45+ 2.033 .135 .422
Gender (male) .021 .064 .621
Standardized direct effects of vehicle ownership on #
Usual transportation mode .390 .027 .000***
Private vehicle
Via on-demand ridesharing 2.226 .030 .000***
Private app-based ride service 2.108 .023 .089*
Private app-based ride service .022 .024 .725
Base category 2.090 .027 .154
SAV trip frequency 2.164 .145 .009***
Goodness-of-fit measures x
2
/df(\2) = 1.83 RMSEA
(\0.1) = .059
Standardized RMR
(0\SRMR \1) = .07
Note: SE = standard error; P=p-value; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual.
Usual transportation mode: base category = university transportation service, Handitran paratransit service, and RAPID SAV service.
SAV trip purpose: base categor y = shopping, medical, social/recreational, and other trips.
*a= 0.1; **a= 0.05; ***a= 0.01.
Etminani-Ghasrodashti et al 471
continuous variables in the SEM model. Likewise, SAV
service attributes (internal and external performances)
were extracted factors and treated as continuous vari-
ables. Conversely, usual transportation mode, SAV trip
purposes, gender, and age are categorical exogenous
variables. Accordingly, we recoded these independent
variables as binary variables. For instance, if respondents
reported the SAV trip purpose to be for work, work pur-
pose obtained a value of 1 for respondents. Likewise, if
they selected private vehicle as their usual mode of trans-
portation, the private vehicle variable acquired a value
of 1. We combined university transportation, Handitran
paratransit, and RAPID SAV services (services with
limited-service area or that were less accessible for all
people) into one variable and considered this variable as
the reference category for comparison with other cate-
gories of usual transportation mode, including private
vehicle, Via on-demand ridesharing, private app-based
ride service, and walking/biking. Similarly, we integrated
optional trip purposes (shopping, medical, social/recrea-
tional, and other trips) into one category to compare to
the effects from mandatory trips, including work, school,
and returning home, on SAV adoption with this new
variable as the base category.
This study employed the maximum likelihood (ML)
technique as the estimating method. The assumption of
using an ML estimator is the multivariate normal distri-
bution of endogenous variables in the model (46). The
endogenous variable in this study was the adoption of
SAV, which was coded on a five-point Likert-type scale.
Because our responses on this variable were coded
numerically in ascending order, we treated it as a contin-
uous variable and applied ML to estimate the model
parameters. When ordinal data have at least five cate-
gories and are approximately normal, ML does not pro-
duce severely biased results (47). We further investigated
the fitness of our model to test the difference between the
observed- and model-implied variance–covariance
matrix. First, we utilized the ratio of chi-square to its
degree of freedom to see whether it was smaller than the
value of 5 (x
2
/df = 2.9). Second, we tested the root mean
square error of approximation (RMSEA) in our model,
Adoption of SAV
External performance
Internal performance
SAV trip waiting time
.23
.54
-.05
Gender: male
.32
0.0
Age: 18-24
Age: 25-34
Age: 35-44
Age: 45+
-.13
-.29
-.02
Purpose: work
Purpose: school
Purpose: return home
Purpose: base category
-.25
-.36
-.24
-.31
Mode: private vehicle
Mode: Via on-demand
Mode: private app -based
Mode: walking/biking
Vehicle ownership
.18
-.11
-.05
-.04
-.38
-.03
.21
.27
Mode: base category
.11
.01
SAV Trip frequency
.22
.24
.01
Path Coefficient
(Unstandardized direct effects )
Figure 3. Path analysis of adoption of shared autonomous vehicles (SAVs).
472 Transportation Research Record 2676(11)
indicating the approximation error per model degree of
freedom. The RMSEA for the model was under the rec-
ommended value of less than 0.1 (RMSEA = .055).
Third, we checked the comparative fit index (CFI) for a
value greater than 0.9 (CFI = 0.955) (48,49). We also
fitted the model for standardized root mean square resi-
dual (SRMR), which varies from 0 to 1 (50).
The model results showed that service attributes in
relation to SAV internal and external performances were
strongly associated with adoption of SAV. This indicated
that those participants who had greater positive evalua-
tions of service performance were more likely to use
SAVs in the future. However, the effect of internal service
performance (b= .644) on SAV adoption was higher
than the external performance evaluation (b= .261).
Our results also revealed that usual transportation
mode influenced SAV adoption. As the usual mode of
transportation, private vehicles significantly increased
SAV adoption compared with the base category. The
same relationships were found for Via on-demand ride-
sharing, private app-based ride services, and walking/
biking modes: those who reported these three modes of
travel were more interested in adopting RAPID SAV
than the base category. The effects of private vehicles
(b= .138) and Via on-demand ridesharing (b= .134) on
SAV adoption were higher than app-based transporta-
tion (b= .125) and walking/biking (b= .110).
Following the model results, we would anticipate that
longer trip waiting times would result in slightly lower
adoption of SAVs by users (b=2.068). SAV trip pur-
pose significantly influenced the adoption of self-driving
services. For example, our analysis revealed that those
who requested a RAPID ride to go to school and return
home were less likely to adopt SAVs in the future
(b=2.512 & b=2.479) compared with the base cate-
gory that used SAV trips for optional trip purposes. For
work trips, people were still less likely to adopt SAVs in
the future, but the direct effects were a little less than the
base category.
Although the standardized direct effects of SAV trip
frequency on adoption of SAVs were positive, the rela-
tionship was not statistically significant.
Our results indicated that the effect of vehicle owner-
ship on SAV adoption was not statistically significant.
Therefore, we explored the effects of vehicle ownership
on other variables, including usual transportation mode
and SAV trip frequency. By estimating SEM, our model
was not limited to identifying relationships between SAV
adoption and key variables. In our model, usual trans-
portation mode and SAV trip frequency moderated the
effect of vehicle ownership on the usual mode of transpor-
tation. The results suggested that vehicle ownership
increased the likelihood of using the private vehicle as the
usual mode by individuals. Accordingly, both direct and
indirect effects of vehicle ownership on SAV adoption were
estimated in our model (b=2.042, b=2.005). More
access to private vehicles resulted in greater use of private
vehicles (b= .390), lower use of Via on-demand rideshar-
ing (b=2.226), and little use of private app-based ride ser-
vices (b=2.108) as the usual mode of transportation.
However, as mentioned earlier, using the private vehicle as
the usual mode of transportation significantly increased
SAV adoption. This is why vehicle ownership did not have
a significant relationship with SAV adoption.
In addition, the results indicated that greater access to
private vehicles resulted in a lower SAV trip frequency.
Those who more private vehicles were less frequent users
of the SAV service. Finally, the SEM results indicated
that age had a significant association with adoption of
SAV. Respondents aged 35 to 44 years were less likely to
adopt SAVs than other age categories.
Discussion
Based on a survey of RAPID SAV riders in Arlington
TX, this study investigated users’ adoption of self-
driving shuttles. To this end, we conceptualized the study
model based on the SAV ridership features and service
attributes, as well as riders’ travel behaviors and sociode-
mographic information, and explored the interplay
among multiple factors. Perhaps because the pilot SAV
service provides rides across the university campus and
Arlington’s downtown area with the option of free rides
to university students, the descriptive statistics indicated
that the majority of SAV users were young adults, males,
Asian, and students. This result is similar to past studies
that showed early potential SAV adopters were young
people (4), male, and those who are below the age of 35
(12,51). One possible reason for this finding is that
Asian and international students may be more likely to
request rides. In addition, SAV riders are mostly individ-
uals with a lower number of private vehicles per house-
hold, with low-income levels, and those who usually use
the Via on-demand ridesharing service and private vehi-
cles as their main mobility mode. These findings con-
firmed that the SAV pilot project could achieve its
designated goals, including 1. increasing access for stu-
dents and individuals with limited personal mobility, 2.
improving equity and accessibility to mobility options,
and 3. integrating Level 4 AVs into Arlington’s existing
public Via on-demand rideshare service (52–54).
We applied SEM to identify the influences of conven-
tional factors such as sociodemographic status, vehicle
ownership, and usual mode of transportation, as well as
novel measurements of SAV service attributes and trip
features on the adoption of SAV.
We found that SAV service attributes were among the
most important determinants of SAV adoption. This
Etminani-Ghasrodashti et al 473
finding indicated that a higher level of SAV adoption
could be generated by promoting the internal and exter-
nal performances of self-driving technology. Although
the effects of some SAV performance measures, includ-
ing travel time and travel cost, on SAV use confirmed
results from previous studies that explored preferences of
potential users of SAVs (9), other external features (e.g.,
ease of booking and scheduling trips, convenience of
pickup and dropoff locations) as well as internal features
of the self-driving vehicles (comfort of the seats, cabin
environment, safety and speed) were among the criteria
that have received less attention.
We also speculated that the typical transportation
mode (i.e., respondents’ most common daily transporta-
tion mode) would have a critical role in the use and
adoption of SAVs. Users of private vehicles and Via on-
demand ridesharing were more interested in accepting
SAVs in the future. This finding is supported by previous
research suggesting positive relationships between both
private vehicles and public transit use and the propensity
to shift to SAVs by potential users (9). A possible reason
for this positive association between using private vehi-
cles as the usual mode and SAV adoption is that individ-
uals may be more likely to complement (and not
substitute) the new SAV technology as their second
option transportation mode. These results can also indi-
cated that the reference mode of transportation may sig-
nificantly affect the adoption of SAVs, it still does not
eliminate nor substitute for the SAVs. SAVs will be able
to compete with private vehicles (i.e., reduce private vehi-
cle usage) if they provide a comparable level of flexibility
as private cars and when users feel SAV riding is more
convenient. As individuals gain more experience using
the SAV service over time, they are likely to reconsider
the service performance, that is, its flexibility and conve-
nience, meaning that they may be more likely to shift
from private car use to SAVs. Similarly, other transporta-
tion options that were reported as usual modes, including
private app-based ride services and walking/biking, could
also increase use of SAV in the future. A recent study
explored competition between SAVs, public transit, and
walking for the first-mile trip by implementing an agent-
based simulation model. Findings indicated that when
public transit supply was low, people were more likely to
walk their first-mile trip and get an AV (55). We could
employ this study to explain our findings; people may use
active travel modes to complement their SAV usage, par-
ticularly for the first-/last-mile trip. Accordingly, policies
and strategies that motivate people to relinquish private
vehicle ownership and improve the integration of walk-
ing, public transport, and on-demand transport are sug-
gested as the most beneficial approaches for engaging
automated technology in transportation systems (6).
Our results also indicated that as trip waiting times
increase, users will be less likely to use and accept SAVs
as their usual transportation mode. This result supports
the studies that suggest potential users of SAVs are those
who are more willing to pay and ride automated service
if it can improve their mobility efficiency by reducing
travel time (18). Therefore, waiting time is an essential
factor for SAV operators to consider to divert travelers
to SAVs (8,9).
The results suggested a strong relationship between
different trip purposes and using SAVs; those who
reported using SAVs to go to school and return home
seemed less likely to ride the self-driving shuttles in the
future compared with people who used SAVs for
optional trips. This result provides new insights into the
necessity of evaluating the effects of mandatory or main-
tenance trips while studying SAV adoption. This finding
suggested that since people like to follow a robust com-
mute schedule, they prefer to use a usual or more com-
mon travel mode. This confirmed findings from previous
studies that showed people who had more flexible hours
or schedules for their routine activities were more likely
to use and accept SAVs (12). Moreover, we should con-
sider that in our study the SAVs provided service to a
limited area that may not have included the desired desti-
nations of the respondents. The service area comprises
the university campus and Arlington’s downtown area,
but not long-distance destinations. Similarly, a past
study suggested that SAV acceptance by potential users
for long-distance trips may be higher than for short dis-
tances (21). In addition, the findings suggested that at
the early stage of SAV deployment, potential users may
ride the service because of an intrinsic desire to experi-
ence the technology for its own sake, rather than to
arrive at a destination (e.g., work/school/home) (56).
Although SAV trip frequency could increase the likeli-
hood of adopting SAVs in the future, the relationship
was not statistically significant in the model. Therefore,
it is anticipated that as people ride SAVs more and gain
more experience of this new technology, they will be
more likely to adopt self-driving vehicles as a usual mode
of urban transportation.
The results from estimating the interrelationships
between independent variables indicated that individuals
with greater access to private vehicle were more likely to
use private vehicle as their usual mode of transportation,
and less likely to use on-demand ridesharing. Integration
of SAVs into the current public transit system is antici-
pated to make an outstanding contribution to reducing
traffic congestion and improving safety issues through
car ownership reduction (57). To this end, it will be nec-
essary to design strategies that encourage individuals to
relinquish conventional cars once SAVs are on the road.
474 Transportation Research Record 2676(11)
Understanding this will significantly help the adoption
and the success of SAV operations in the future.
The SEM results also indicated that older adults (35
to 44 years old) were less likely to accept SAVs as a com-
mon mode of transportation compared with younger
adults (under 34 years). This finding supported previous
studies that suggested young adults have more multimo-
dal travel behavior and use more public transit modes.
Therefore, they may have the propensity to shift to self-
driving technology more smoothly than older adults (9,
12).
Conclusions
Although pilot demonstrations have been deployed in
several places, little is known about how individuals will
accept SAVs as a usual mode of transportation. SAV
pilot projects could represent great opportunities for
researchers and practitioners to fully understand the pat-
terns by which SAV adoption will be shaped.
Understanding the factors that could trigger the smooth
adoption of SAVs could also assist planners and policy
makers in adjusting the policies before more widespread
adoption of this new technology. Recent findings indi-
cate that public concerns related to AVs, including data
privacy and security, have increased over time; however,
AV adoption will also rise over time owing to increased
tech-savvy and pro-AV sentiments among individuals
(58). Therefore, a pilot demonstration of self-driving
vehicles could provide an opportunity for AV developers
and related agencies to improve public trust through
real-life experience of the technology.
This study has provided new and potentially crucial
insights into the growing body of knowledge about SAV,
but further research is needed to address the study lim-
itations. As this study was developed 7 months after
deployment of the RAPID self-driving shuttles, it will be
essential to conduct a postdeployment study of the pilot
project to investigate the SAV ridership patterns based
on a real data set platform and compare the ridership
trends, waiting times, travel times, and travel distances
over the different months of the project.
Another concern is that the research was conducted
using a cross-sectional approach, therefore, it is possible
that SAV adoption and the determinant factors will
increase over time as potential riders experience the ser-
vice more regularly and for different purposes.
Therefore, future research should be developed based on
a comprehensive survey that considers multiple variables
and focuses on a larger sample of both service users and
nonusers to evaluate the complicated associations
between factors affecting SAV technology adoption. In
addition, owing to the small number of observations,
certain associations may have resulted from the small
sample size. For instance, our findings indicated that
those who currently use the private vehicle as their usual
mode of transportation are more interested in adopting
SAVs in the future. AVs have the potential to reduce pri-
vate vehicle usage (5), but still, evidence demonstrating
how current car users would adopt AVs and SAVs in the
medium term is rare. Therefore, extending the concep-
tual study model to a larger sample size of actual SAV
users is recommended. Eventually, the cost-efficiency of
SAVs will play an essential role in SAV acceptance.
Therefore, further research should be developed to iden-
tify the potential cost-saving effects of SAVs in transpor-
tation systems.
Author Contributions
The authors confirm contribution to the paper as follows:
study conception and design: R. Etminani-Ghasrodashti, S.
Kermanshachi, J. M. Rosenberger, A. Foss; data collection: R.
Ketankumar Patel, R. Etminani-Ghasrodashti, S. Kermanshachi,
J. M. Rosenberger, A. Foss; analysis and interpretation of results:
R. Etminani-Ghasrodashti, R. Ketankumar Patel; draft manu-
script preparation: R. Etminani-Ghasrodashti, R. Ketankumar
Patel, S. Kermanshachi, J. M. Rosenberger, A. Foss. All authors
reviewed the results and approved the final version of the
manuscript.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The authors disclosed receipt of the following financial support
for the research, authorship, and/or publication of this article: The
work presented here is a part of the Arlington RAPID (Rideshare,
Automation, and Payment Integration Demonstration) project,
which is supported by the Federal Transit Administration
Integrated Mobility Innovation Program funded by the United
States Department of Transportation and City of Arlington. The
RAPID project is a collaboration among different partners includ-
ing City of Arlington, Via Transportation, May Mobility, and
UTA.
Data Accessibility Statement
The raw/processed data required to reproduce the above find-
ings cannot be shared at this time due to legal/ethical reasons.
ORCID iDs
Roya Etminani-Ghasrodashti https://orcid.org/0000-0002-
8434-7663
Sharareh Kermanshachi https://orcid.org/0000-0003-1952-
2557
Etminani-Ghasrodashti et al 475
Jay Michael Rosenberger https://orcid.org/0000-0003-4038-
1402
Ann Foss https://orcid.org/0000-0001-9060-0234
References
1. O’Toole, R. Policy Implications of Autonomous Vehicles.
Cato Institute, 2014. https://www.cato.org/sites/cato.org/
files/pubs/pdf/pa758_1.pdf.
2. Rojas-Rueda, D., M. J. Nieuwenhuijsen, H. Khreis, and
H. Frumkin. Autonomous Vehicles and Public Health.
Annual Review of Public Health, Vol. 41, 2020,
pp. 329-345.
3. Zhang, W., S. Guhathakurta, J. Fang, and G. Zhang.
Exploring the Impact of Shared Autonomous Vehicles on
Urban Parking Demand: An Agent-Based Simulation
Approach. Sustainable Cities and Society, Vol. 19, 2015,
pp. 34-45.
4. Haboucha, C. J., R. Ishaq, and Y. Shiftan. User Prefer-
ences Regarding Autonomous Vehicles. Transportation
Research Part C: Emerging Technologies, Vol. 78, 2017,
pp. 37-49.
5. Fagnant, D. J., and K. Kockelman. Preparing a Nation
for Autonomous Vehicles: Opportunities, Barriers and Pol-
icy Recommendations. Transportation Research Part A:
Policy and Practice, Vol. 77, 2015, pp. 167-181.
6. Pettigrew, S., C. Worrall, Z. Talati, L. Fritschi, and R.
Norman. Dimensions of Attitudes to Autonomous Vehi-
cles. Urban, Planning and Transport Research, Vol. 7, No.
1, 2019, pp. 19-33.
7. Fagnant, D. J., and K. M. Kockelman. The Travel and
Environmental Implications of Shared Autonomous Vehi-
cles, Using Agent-Based Model Scenarios. Transportation
Research Part C: Emerging Technologies, Vol. 40, 2014,
pp. 1-13.
8. Bansal, P., K. M. Kockelman, and A. Singh. Assessing
Public Opinions of and Interest in New Vehicle Technolo-
gies: An Austin Perspective. Transportation Research Part
C: Emerging Technologies, Vol. 67, 2016, pp. 1-14.
9. Krueger, R., T. H. Rashidi, and J. M. Rose. Preferences
for Shared Autonomous Vehicles. Transportation Research
Part C: Emerging Technologies, Vol. 69, 2016, pp. 343-355.
10. Dong, X., M. DiScenna, and E. Guerra. Transit User Per-
ceptions of Driverless Buses. Transportation, Vol. 46, No.
1, 2019, pp. 35-50.
11. Yuen, K. F., D. T. K. Huyen, X. Wang, and G. Qi. Factors
Influencing the Adoption of Shared Autonomous Vehicles.
International Journal of Environmental Research and Public
Health, Vol. 17, No. 13, 2020, p. 4868.
12. Wang, K., and G. Akar. Factors Affecting the Adoption
of Autonomous Vehicles for Commute Trips: An Analysis
With the 2015 and 2017 Puget Sound Travel Surveys.
Transportation Research Record: Journal of the Transporta-
tion Research Board, 2019. 2673: 13-25.
13. Abraham, H., C. Lee, S. Brady, C. Fitzgerald, B. Mehler,
B. Reimer, and J. F. Coughlin. Autonomous Vehicles,
Trust, and Driving Alternatives: A Survey of Consumer Pre-
ferences, Vol. 1. Massachusetts Institute of Technology,
AgeLab, Cambridge, 2016, p. 16.
14. Wu, J., H. Liao, and J.-W. Wang. Analysis of Consumer
Attitudes Towards Autonomous, Connected, and Electric
Vehicles: A Survey in China. Research in Transportation
Economics, Vol. 80, 2020, p. 100828.
15. Asmussen, K. E., A. Mondal, and C. R. Bhat. A Socio-
Technical Model of Autonomous Vehicle Adoption Using
Ranked Choice Stated Preference Data. Transportation
Research Part C: Emerging Technologies, Vol. 121, 2020,
p. 102835.
16. Bennett, R., R. Vijaygopal, and R. Kottasz. Attitudes
Towards Autonomous Vehicles Among People With Physi-
cal Disabilities. Transportation Research Part A: Policy and
Practice, Vol. 127, 2019, pp. 1-17.
17. Rahimi, A., G. Azimi, and X. Jin. Examining Human Atti-
tudes Toward Shared Mobility Options and Autonomous
Vehicles. Transportation Research Part F: Traffic Psychol-
ogy and Behaviour, Vol. 72, 2020, pp. 133-154.
18. Asgari, H., and X. Jin. Incorporating Attitudinal Factors
to Examine Adoption of and Willingness to Pay for Auton-
omous Vehicles. Transportation Research Record: Journal
of the Transportation Research Board, 2019. 2673: 418-429.
19. Liu, P., Q. Guo, F. Ren, L. Wang, and Z. Xu. Willingness
to Pay for Self-Driving Vehicles: Influences of Demo-
graphic and Psychological Factors. Transportation
Research Part C: Emerging Technologies, Vol. 100, 2019,
pp. 306-317.
20. Rahimi, A., G. Azimi, H. Asgari, and X. Jin. Adoption
and Willingness to Pay for Autonomous Vehicles: Atti-
tudes and Latent Classes. Transportation Research Part D:
Transport and Environment, Vol. 89, 2020, p. 102611.
21. Gurumurthy, K. M., and K. M. Kockelman. Modeling
Americans’ Autonomous Vehicle Preferences: A Focus on
Dynamic Ride-Sharing, Privacy & Long-Distance Mode
Choices. Technological Forecasting and Social Change, Vol.
150, 2020, p. 119792.
22. Becker, F., and K. W. Axhausen. Literature Review on
Surveys Investigating the Acceptance of Automated Vehi-
cles. Transportation, Vol. 44, No. 6, 2017, pp. 1293-1306.
23. Zhang, J. Autonomous Vehicles: Understanding Adoption
Potential in the Greater Toronto and Hamilton Area. Mas-
ter’s thesis. University of Waterloo, 2019.
24. Wang, S., and J. Zhao. Risk Preference and Adoption of
Autonomous Vehicles. Transportation Research Part A:
Policy and Practice, Vol. 126, 2019, pp. 215-229.
25. Nazari, F., M. Noruzoliaee, and A. K. Mohammadian.
(eds.). Adoption of Autonomous Vehicles With Endogen-
ous Safety Concerns: A Recursive Bivariate Ordered Probit
Model. In Annual Meeting of the Transportation Research
Board, Washington, D.C., 2019, pp. 19-02998.
26. Song, Y., M. V. Chitturi, C. McCahill, and D. A. Noyce.
People’s Attitudes Toward Automated Vehicle and Transit
Integration: Case Study of Small Urban Areas. Transporta-
tion Planning and Technology, Vol. 44, No. 5, 2021,
pp. 1-21.
27. Acheampong, R. A., and F. Cugurullo. Capturing the
Behavioural Determinants Behind the Adoption of Auton-
omous Vehicles: Conceptual Frameworks and Measure-
ment Models to Predict Public Transport, Sharing and
Ownership Trends of Self-Driving Cars. Transportation
476 Transportation Research Record 2676(11)
Research Part F: Traffic Psychology and Behaviour, Vol.
62, 2019, pp. 349-375.
28. Jiang, Z., S. Wang, A. S. Mondschein, and R. B. Noland.
Spatial Distributions of Attitudes and Preferences Towards
Autonomous Vehicles. Transport Findings, 2020, pp. 1-10.
https://doi.org/10.32866/001c.12653.
29. Manfreda, A., K. Ljubi, and A. Groznik. Autonomous
Vehicles in the Smart City Era: An Empirical Study of Adop-
tion Factors Important for Millennials. International Journal
of Information Management, Vol. 58, 2021, p. 102050.
30. Lee, J., D. Lee, Y. Park, S. Lee, and T. Ha. Autonomous
Vehicles can be Shared, but a Feeling of Ownership is
Important: Examination of the Influential Factors for
Intention to Use Autonomous Vehicles. Transportation
Research Part C: Emerging Technologies, Vol. 107, 2019,
pp. 411-422.
31. Mohammadzadeh, M. Sharing or Owning Autonomous
Vehicles? Comprehending the Role of Ideology in the
Adoption of Autonomous Vehicles in the Society of Auto-
mobility. Transportation Research Interdisciplinary Per-
spectives, Vol. 9, 2021, p. 100294.
32. Etzioni, S., R. A. Daziano, E. Ben-Elia, and Y. Shiftan.
Preferences for Shared Automated Vehicles: A Hybrid
Latent Class Modeling Approach. Transportation Research
Part C: Emerging Technologies, Vol. 125, 2021, p. 103013.
33. Jing, P., G. Xu, Y. Chen, Y. Shi, and F. Zhan. The Deter-
minants Behind the Acceptance of Autonomous Vehicles:
A Systematic Review. Sustainability, Vol. 12, No. 5, 2020,
p. 1719.
34. Shabanpour, R., N. Golshani, A. Shamshiripour, and A.
K. Mohammadian. Eliciting Preferences for Adoption of
Fully Automated Vehicles Using Best-Worst Analysis.
Transportation Research Part C: Emerging Technologies,
Vol. 93, 2018, pp. 463-478.
35. City of Arlington. Data in Action: Arlington’s via Rideshare
Service Enhancing Mobility for Citizens. Office of Commu-
nication, 2019. https://www.arlingtontx.gov/news/my_arl
ington_t_x/news_stories/datainaction_via. Accessed Novem-
ber 17, 2020.
36. City of Arlington. RAPID. Office of Strategic Initiatives,
2021. https://www.arlingtontx.gov/city_hall/departments/
office_of_strategic_initiatives/rapid.
37. City of Arlington. City of Arlington Launches First-of-its-
Kind, On-Demand Self Driving Shuttle Service With RAPID
Program. Office of Communication, 2021. https://www.ar-
lingtontx.gov/news/my_arlington_t_x/news_stories/arlington_
rapid_launch.
38. Etminani-Ghasrodashti, R., R. Ketankumar Patel, S. Ker-
manshachi, J. Michael Rosenberger, and D. Weinreich.
(eds.). Exploring Concerns and Preferences Towards Using
Autonomous Vehicles as a Public Transportation Option:
Perspectives From a Public Focus Group Study. Proc.,
International Conference on Transportation and Develop-
ment 2021, Virtually, 2021.
39. Etminani-Ghasrodashti, R., R. K. Patel, S. Kermanshachi,
J. M. Rosenberger, D. Weinreich, and A. Foss. Integration
of Shared Autonomous Vehicles (SAVs) Into Existing
Transportation Services: A Focus Group Study.
Transportation Research Interdisciplinary Perspectives, Vol.
12, 2021, p. 100481.
40. McDonald, R. P., and E. Burr. A Comparison of Four
Methods of Constructing Factor Scores. Psychometrika,
Vol. 32, No. 4, 1967, pp. 381-401.
41. Etminani-Ghasrodashti, R., and S. Hamidi. Individuals’
Demand for Ride-Hailing Services: Investigating the Com-
bined Effects of Attitudinal Factors, Land Use, and Travel
Attributes on Demand for App-Based Taxis in Tehran,
Iran. Sustainability, Vol. 11, No. 20, 2019, p. 5755.
42. Circella, G., and P. L. Mokhtarian. Impacts of Informa-
tion and Communication Technology. In The Geography of
Urban Transportation (Giuliano, G., and S. Hanson, eds.),
The Guilford Press, New York, NY, 2017, p. 86.
43. Golob, T. F. Structural Equation Modeling for Travel
Behavior Research. Transportation Research Part B: Meth-
odological, Vol. 37, No. 1, 2003, pp. 1-25.
44. Cao, X. J., Z. Xu, and F. Douma. The Interactions
Between E-Shopping and Traditional In-Store Shopping:
An Application of Structural Equations Model. Transpor-
tation, Vol. 39, No. 5, 2012, pp. 957-974.
45. Mueller, R. O. Basic Principles of Structural Equation Mod-
eling: An Introduction to LISREL and EQS. Springer Sci-
ence & Business Media, Berlin, Heidelberg, 1999.
46. Kline, R. Principles and Practice of Structural Equation
Modeling. Guilford, New York, NY, 2005, p. 366.
47. Mıˆndrila
˜, D. Maximum Likelihood (ML) and Diagonally
Weighted Least Squares (DWLS) Estimation Procedures:
A Comparison of Estimation Bias With Ordinal and Mul-
tivariate Non-Normal Data. International Journal of Digi-
tal Society, Vol. 1, No. 1, 2010, pp. 60-66.
48. Mokhtarian, P., and D. Ory. Methods: Structural Equations
Models. Entry for the International Encyclopedia of Human
Geography. Elsevier Ltd, Oxford, UK, 2008.
49. Weston, R., and P. A. Gore, Jr. A Brief Guide to Structural
Equation Modeling. The Counseling Psychologist, Vol. 34,
No. 5, 2006, pp. 719-751.
50. Byrne, B. M. Structural Equation Modeling With Mplus:
Basic Concepts, Applications, and Programming. Routledge,
New York, NY, 2013.
51. Moreno, A. T., A. Michalski, C. Llorca, and R. Moeckel.
Shared Autonomous Vehicles Effect on Vehicle-Km Tra-
veled and Average Trip Duration. Journal of Advanced
Transportation, Vol. 2018, 2018, pp. 1-10.
52. Golbabaei, F., T. Yigitcanlar, and J. Bunker. The Role of
Shared Autonomous Vehicle Systems in Delivering Smart
Urban Mobility: A Systematic Review of the Literature.
International Journal of Sustainable Transportation, Vol.
15, No. 10, 2021, pp. 731-748.
53. Papa, E., and A. Ferreira. Sustainable Accessibility and the
Implementation of Automated Vehicles: Identifying Criti-
cal Decisions. Urban Science, Vol. 2, No. 1, 2018, p. 5.
54. Khan, M. A., R. Etminani-Ghasrodashti, A. Shahmoradi,
S. Kermanshachi, J. M. Rosenberger, and A. Foss. Inte-
grating Shared Autonomous Vehicles Into Existing Trans-
portation Services: Evidence From a Paratransit Service in
Arlington, Texas. International Journal of Civil Engineering,
2022, pp. 1-18. https://doi.org/10.1007/s40999-021-00698-6.
Etminani-Ghasrodashti et al 477
55. Mo, B., Z. Cao, H. Zhang, Y. Shen, and J. Zhao. Competi-
tion Between Shared Autonomous Vehicles and Public Tran-
sit: A Case Study in Singapore. Transportation Research Part
C: Emerging Technologies, Vol. 127, 2021, p. 103058.
56. Mokhtarian, P. L., and I. Salomon. How Derived is the
Demand for Travel? Some Conceptual and Measurement
Considerations. Transportation Research Part A: Policy
and Practice, Vol. 35, No. 8, 2001, pp. 695-719.
57. Jones, E. C., and B. D. Leibowicz. Contributions of Shared
Autonomous Vehicles to Climate Change Mitigation.
Transportation Research Part D: Transport and Environ-
ment, Vol. 72, 2019, pp. 279-298.
58. Acharya, S., and P. Humagain. Public Interest in Autono-
mous Vehicle Adoption: Evidence From the 2015, 2017,
and 2019 Puget Sound Travel Surveys. Journal of Trans-
portation Engineering, Part A: Systems, Vol. 148, No. 4,
2022, p. 04022003.
478 Transportation Research Record 2676(11)
