Content uploaded by Yuntao Guo
Author content
All content in this area was uploaded by Yuntao Guo on Mar 21, 2023
Content may be subject to copyright.
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
1
Understanding factors that impact ridesourcing service usage frequency:
1
A case study in Shanghai
2
3
4
Feiyu Feng, Ph.D.
5
School of Transportation, Fujian University of Technology
6
No3 Xueyuan Road, University Town, Minhou, Fuzhou City, Fujian Province, China, 350118
7
Email: Fengfy92@163.com
8
9
Xinghua Li, Ph.D.
10
Professor
11
The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University
12
4800 Cao'an Road, Shanghai, China, 201804
13
Email: xinghuali@tongji.edu.cn
14
15
Yuntao Guo, Ph.D., Corresponding Author
16
Assistant Professor
17
The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University
18
4800 Cao'an Road, Shanghai, China, 201804
19
Email: yuntaoguo@tongji.edu.cn
20
21
Cheng Cheng, Ph.D.
22
Research Associate
23
The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University
24
4800 Cao'an Road, Shanghai, China, 201804
25
Email: 18608@tongji.edu.cn
26
27
28
29
30
31
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
2
ABSTRACT
1
This study investigates the influencing factors that impact ridesourcing service usage frequency
2
and explore the potential similarities and differences among groups of population based on their
3
primary usage purposes. A revealed-preference survey developed for this study was conducted
4
among 783 ridesourcing service users from Shanghai, China in September 2020. Separate random
5
parameters ordered probit models were estimated for users with different primary purposes of
6
usage to capture unobserved heterogeneity. The identified influencing factors include travelers’
7
sociodemographic characteristics, reasons to choose ridesourcing services, and other behavioral
8
characteristics. In addition, the impacts of these contributing factors were different based on their
9
primary usage purpose. The model estimation and descriptive statistics findings suggest that groups
10
of ridesourcing service users may respond differently to various types of promotional strategies.
11
The study insights may be used to design future strategies that can potentially improve the service
12
usage frequency of existing users and attract new users.
13
14
Keywords: Ridesourcing services; usage frequency; unobserved heterogeneity; trip purpose;
15
16
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
3
1. Introduction
1
Ridesourcing services offered by companies such as Uber and Didi Chuxing have increased their
2
popularity drastically since they were first introduced in 2012 (Jin et al., 2018; Zha et al., 2018;
3
Mohamed et al., 2019; Shaheen and Cohen, 2019). Such services use app-based platforms to match
4
users with drivers who typically drive their own car part-time or full-time. They offer the potential
5
to reduce user waiting time and travel costs, generate income for drivers, and improve the overall
6
transportation system efficiency (Schaller, 2021; Bansal et al., 2020, Mohamed et al., 2020). They
7
are becoming an integral part of people’s daily travel and have accumulated a tremendous number
8
of users. A recent report (Cyberspace Administration of China, 2021) showed that the number of
9
ridesourcing service users in China reached 365 million by the end of 2020, representing 26 percent
10
of China’s total population.
11
With the continuing development of ridesourcing services, they are gradually shaping the
12
travel behavior of residents in many countries (Alemi et al., 2018; Stinson et al., 2021). A recent
13
study revealed that over 15% of all trips inside the city of San Francisco were made by Uber and
14
Lyft on a typical weekday from mid-November to mid-December of 2016 (San Francisco County
15
Transportation Authority, 2017). Many policymakers, practitioners, and researchers have
16
attempted to understand the reasons behind the rapid expansion of ridesourcing services in terms
17
of mode share and the number of users by studying users’ behavior, including ‘who’ (e.g., the
18
characteristics of its current and potential user groups), ‘when’ (e.g., when people use such service),
19
and ‘why’ (e.g., possible factors affecting people’s service usage) (Alemi et al., 2018; Ghaffar et
20
al., 2020; Rizki et al., 2021, Azimi et al., 2022; Guo et al., 2022; Li et al., 2022). Usage frequency
21
of ridesourcing services is one of such behaviors that reflects the preference of travelers to use
22
ridesourcing services as usual or use them more frequently of their own volition (Dias et al., 2017;
23
Alemi et al., 2019; Aghaabbasi et al., 2020; Belgiawan et al., 2022). It can be particularly important
24
when ridesourcing services are competing with other modes of transportation such as private
25
vehicles and transit for travelers.
26
One of the first steps is to understand factors that impact ridesourcing service usage frequency.
27
There are an ample number of studies attempting to understand the impacts of various potential
28
factors on ridesourcing service usage frequency. Three noticeable types of factors are: (i)
29
sociodemographic characteristics, (ii) neighborhood-built environment, and (iii) psychological
30
factors. Regarding sociodemographic characteristics, some studies suggested that travelers who
31
were younger (Dias et al., 2017; Alemi et al., 2019; Barbour et al., 2020; Rizki et al., 2021), have
32
a higher income (Dias et al., 2017; Alemi et al., 2019; Barbour et al., 2020; Ghaffar et al., 2020;
33
Rahimi et al., 2020; Rizki et al., 2021), with higher education (Dias et al., 2017; Alemi et al., 2019;
34
Lavieri and Bhat, 2019), have a larger household size (Alemi et al., 2019; Lavieri and Bhat, 2019;
35
Ghaffar et al., 2020) were more likely to have a higher ridesourcing service usage frequency.
36
In terms of neighborhood-built environment, ridesourcing service waiting time is an important
37
factor that reflects the accessibility level to ridesourcing service. The longer the ridesourcing
38
service waiting time, the less likely that it would be used most of the time (Rayle et al., 2016; Yang
39
et al., 2021). Hence, individuals living in neighborhoods that are urban (Dias et al., 2017; Alemi et
40
al., 2018; Alemi et al., 2019; Lavieri and Bhat, 2019; Brown, 2020; Ghaffar et al., 2020; Yan et al.,
41
2020), have a high population density (Dias et al., 2017; Alemi et al., 2019; Ghaffar et al., 2020;
42
Rizki et al., 2021), have a high job to housing ratio (Alemi et al., 2019; Ghaffar et al., 2020; Rizki
43
et al., 2021), or have a low crime rate (Ghaffar et al., 2020) were more likely to have a higher
44
ridesourcing service usage frequency.
45
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
4
Regarding psychological factors, travelers who had a negative attitude toward ridesourcing
1
services such as having concerns over safety/security (de Souza Silva et al., 2018; Alemi et al.,
2
2019) and discrimination (Moody et al., 2019) were more likely to have a lower ridesourcing
3
service usage frequency. Travelers who were pro-environment (Alemi et al., 2019; Lavieri and
4
Bhat, 2019), technology-embracing (having a positive attitude towards internet-based applications)
5
(Alemi et al., 2019; Lavieri and Bhat, 2019), or variety seeking (having a positive attitude towards
6
new experiences, new products, and new cultures) (Alemi et al., 2019; Lavieri and Bhat, 2019)
7
were more likely to have a higher ridesourcing service usage frequency.
8
Despite these efforts, one of the key assumptions made in most studies was that factors affect
9
people’s ridesourcing service usage frequency similarly regardless of their primary purpose of
10
usage (primary trip purpose when using ridesourcing services). However, although limited, some
11
studies have found that people’s decision-making processes related to using ridesourcing services
12
are different based on their trip purposes (Young and Farber, 2019; Kang et al., 2021), while the
13
vast majority of other studies only focused on commute trips. For example, Young and Farber
14
(2019) suggested that younger generations were more likely to use ridesourcing services for non-
15
commute trips compared to commute ones. Kang et al. (2021) found that the same factors (age,
16
race, etc.) affect people’s preference for pooled versus private ride-hailing services differently
17
when making the commute, shopping, and leisure trips. A likely reason is that people may weigh
18
factors differently based on their trip purposes (i.e., different mode choice decision-making
19
processes) when choosing between ridesourcing services and other modes of transportation. This
20
has been identified in studies that focused on this process such as Choudhury et al. (2018), Lee et
21
al. (2020), and Kamelifar et al. (2022). Hence, it is very likely (although not directly confirmed in
22
previous studies) that the same factors may affect ridesourcing service usage frequency differently
23
among population groups based on their primary purpose of usage. Apart from these potential
24
differences, from an econometric modeling viewpoint, many studies developed different travel
25
behavior models for population groups to improve mode prediction accuracy (Guo et al., 2018; Li
26
et al., 2019; Guo et al., 2020; Guo et al., 2021a). This can be used to assist in making more tailored
27
and effective strategies to influence travel behavior. Therefore, it is important to develop separate
28
models for understanding people’s ridesourcing service usage frequency based on their primary
29
purpose of usage in order to improve the current understanding of people’s ridesourcing service
30
usage behavior and formulate more effective strategies to influence its usage.
31
This study represents one of the early efforts to understand the differences and similarities in
32
terms of the factors affecting the ridesourcing service usage frequency based on their primary
33
purpose of usage. Unlike most previous studies, three groups of users are considered based on their
34
primary trip purpose when using ridesourcing services, including commute users (primarily for
35
home-to-work trips), social/recreational users (primarily for social/recreational trips), and transfer
36
users (primarily for trips to/from airport/train station). An online survey was conducted in Shanghai,
37
China to capture these differences and similarities. Chi-square tests of independence were
38
conducted to highlight these users have statistically significant differences among them in terms of
39
key sociodemographic and behavioral characteristics which entail the heterogeneities among them
40
and possible behavioral differences. Random parameters ordered probit models were estimated to
41
understand the impacts of various factors on the ridesourcing service usage frequency for each type
42
of user. Such models can be used to capture the ordinal nature of the dependent variable and
43
unobserved heterogeneity among users (Washington et al., 2012; Sadri et al., 2013; Azimi et al.,
44
2020; Fountas et al., 2021; Guo et al., 2021). Combined likelihood ratio tests (Washington et al.,
45
2012) were also conducted to illustrate the improved prediction accuracy by estimating separate
46
models for each group compared to modeling them together. The model estimation results illustrate
47
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
5
that the factors affect ridesourcing service usage frequency differently for each user type. The
1
insights can potentially be used to design effective strategies that target different user groups and
2
have the potential to promote ridesourcing service usage.
3
The remainder of the paper is organized as follows: the next two sections present the survey
4
design and implementation, and descriptive statistics, respectively. After that, the modeling
5
framework of random parameters ordered probit models is introduced, followed by model
6
estimation results and discussion. This paper concludes with some concluding comments,
7
limitations, and future work directions.
8
2. Survey design and implementation
9
A pilot web-based survey for Shanghai’s residents in China was designed and a pilot survey was
10
first conducted among 50 graduate students of Tongji University to make sure that most of them
11
can understand the survey clearly. Based on their feedback, the formal survey was designed and
12
distributed in September 2020 through Sojump Survey Company which possesses over two million
13
potential survey participants (http://www.sojump.com/). The questionnaires are randomly sent to
14
a pool of 3,000 potential respondents via email and online social media (e.g., WeChat, QQ). It is
15
important to ensure that these potential respondents are randomly selected in order to minimize
16
sampling bias (Yang et al., 2022). The survey contained three sections: (i) sociodemographic
17
characteristics, including gender, age, education, salary, and vehicle ownership, (ii) when the
18
respondent traveled by themselves, their primary trip purpose when using ridesourcing services,
19
ridesourcing service usage frequency and the reasons to use it under their primary trip purposes,
20
and (iii) other behavioral characteristics, including preference of for-hire mobility services (taxi or
21
ridesourcing services) and the levels of satisfaction with public transit in their neighborhoods. The
22
reasons to use ridesourcing services include reasonable cost, short waiting time,
23
comfort/convenience, difficulty to get a taxi (i.e., traditional taxi), inconvenient public
24
transportation, difficulty to park, and others (e.g., bad weather and too drunk to drive). These
25
possible reasons are informed by previous studies such as Rayle et al. (2016), Beojone and
26
Geroliminis (2021), and Belgiawan et al. (2022). This is a multiple-answer question that
27
participants were allowed to choose one or more answers.
28
At the beginning of the survey, informed consent was presented to the potential respondents.
29
In the consent form, the study purpose, the targeted population, and the definition of ridesourcing
30
services are provided. Participants must be: (i) at least 18 years old, (ii) living in Shanghai at the
31
time of the survey, and (iii) using ridesourcing services at least once during the past two months.
32
Survey participation was voluntary and the participant could end it at any time. To ensure the
33
validity and quality of the survey, geolocation tracking was used to track the location of the
34
participants at the time of the survey, and two attention check questions were embedded in the
35
questionnaire. Only responses that were located in Shanghai and both attention check questions
36
were corrected were considered as valid. Each participant was paid 8 CNY (around 1.1 USD) for
37
a valid and complete survey. One CNY was equal to 0.1415 USD based on the World Bank average
38
exchange rate in September 2020. In addition, to improve the representativeness of respondents to
39
the best of the authors’ ability, the authors also introduced a quota system that ensures the
40
respondents match the potential participant pool of Sojump Survey Company in terms of their age
41
and gender characteristics. At the end of the survey, 783 valid and complete responses were
42
collected. In terms of survey sample size, it is determined based on the method provided by Gill et
43
al. (2010). Considering that there are roughly 6.7 million ridesourcing service users in Shanghai
44
(assuming 26% of the 26 million population are users), the sample size should be larger than 400
45
to ensure a confidence level of 95 and margin of error at 5 percent.
46
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
6
2.1.Descriptive statistics
1
The respondents were divided into three groups based on their primary trip purpose when using
2
ridesourcing services: commute users, social/recreational users, and transfer users. Each group
3
represents about 33 percent of total respondents and the descriptive statistics of these respondents
4
are presented in Table 1. Overall, most respondents were male, young (under 40 years old), well-
5
educated (with at least a bachelor’s degree), or had medium to high salaries. It is possible that most
6
ridesourcing service users belong to these demographics (Dias et al., 2017; Alemi et al., 2018;
7
Lavieri and Bhat, 2019). According to the Shanghai Statistical Yearbook (Shanghai Bureau of
8
Statistics, 2020), the average individual monthly salary was around 10,000 CNY, and people with
9
less than 5,000 CNY were classified as low-salary individuals in 2020. Over 60% of the
10
respondents belonged to the medium or high-salary population. More than 50% of the respondents
11
in all groups claimed that they have access to household vehicles, which is slightly higher than the
12
40% for the Shanghai average.
13
As shown in Table 1, in terms of respondents’ preference for for-hire mobility services, most
14
participants preferred ridesourcing services over taxi services. Regarding their ridesourcing
15
services usage frequency, over 70% of them used such services at least twice a month. For
16
modeling purpose, their usage frequency is classified into three levels, including low frequency (0-
17
1 times a month), medium frequency (2-3 times a month), and high frequency (over 3 times a
18
month). The top three reasons to use ridesourcing services from the first to the third are reasonable
19
cost, short waiting time, and comfort/convenience.
20
Chi-square tests of independence were used to evaluate if the participants in these three groups
21
were similar in terms of their sociodemographic characteristics, reasons to use ridesourcing
22
services, and other behavioral characteristics. The null hypothesis is that there is no difference in
23
the distribution of the outcome across comparison groups. The test results are presented in Table
24
1. All null hypotheses were rejected, which suggests that the differences in these characteristics
25
across groups were statistically significant. These results show that these groups of users are
26
different in terms of their sociodemographic characteristics, reasons to use ridesourcing services,
27
and other behavioral characteristics, which highlights the need of creating separate behavioral
28
models for these groups.
29
Table 1. Characteristics of Participants
30
Trip purpose of ridesourcing
services
Commute
users
Social/recreational
users
Transfer
users
All users
Number of respondents
261
256
266
783
Sociodemographic characteristics
Gender (percentage) *
Male
54.4
60.5
66.2
60.4
Female
45.6
39.5
33.8
39.6
Age (percentage) *
25 or younger
19.9
23.4
18.8
20.7
26-30
30.3
21.9
22.2
24.8
31-40
37.5
34.8
36.1
36.1
Older than 40
12.3
19.9
22.9
18.4
31
32
33
34
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
7
Table 1. (Continue)
1
Trip purpose of ridesourcing
services
Commute
users
Social/recreational
users
Transfer
users
All users
Education (percentage) *
Without a bachelor’s
degree
18.8
25.0
23.7
22.5
A bachelor’s degree
68.2
68.0
62.0
66.0
Post graduate degree
13.0
7.0
14.3
11.5
Average monthly salary (percentage) *
Under 5,000 CNY
15.7
23.0
16.6
18.4
5,000-10,000 CNY
48.7
43.5
38.7
43.7
Over 10,000 CNY
35.6
33.5
44.7
37.9
Vehicle ownership (percentage) *
0
27.6
34.8
20.7
27.6
1
68.2
55.9
62.8
62.3
Over 1
4.2
9.3
16.5
10.1
When the respondent traveled by themselves
Usage frequency of ridesourcing services (percentage) *
Low frequency
15.0
26.2
32.0
24.4
Medium frequency
33.3
39.4
42.5
38.4
High frequency
51.7
34.4
25.5
37.2
Reasons to use ridesourcing services (percentage) *
Comfort/convenience
56.3
53.9
40.2
50.1
Short waiting time
64.4
64.1
59.4
62.6
Reasonable cost
57.1
60.5
47.4
54.9
Hard to park
28.0
30.5
28.6
29.0
Inconvenience of public
transit
34.9
23.4
28.6
29.0
Hard to get a taxi
20.3
16.8
19.2
18.8
Others (e.g., bad weather,
and too drunk to drive)
48.7
49.3
52.2
50.7
Other behavioral characteristics
Preference pattern of for-hire mobility services (percentage) *
Ridesourcing services
85.4
84.8
78.6
82.9
Taxi
14.6
15.2
21.4
17.1
Satisfaction with public transit in their neighborhoods *
Strongly dissatisfied
3.1
3.1
1.9
2.7
Dissatisfied
9.4
3.2
1.5
4.7
Slightly dissatisfied
12.3
9.0
6.4
9.2
Neutral
21.5
17.2
23.7
20.8
Slightly satisfied
22.2
27.3
27.8
25.8
Satisfied
19.2
27.3
24.4
23.6
Strongly satisfied
12.3
12.9
14.3
13.2
* Statistically significant difference across comparison groups at 0.05 level
2
3
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
8
3. Methodology
1
Users’ ridesourcing service usage frequency is modeled as an ordered variable with low frequency,
2
medium frequency, and high frequency. To model the impacts of contributing factors
3
(sociodemographic and ridesourcing service behavioral characteristics) on service usage frequency
4
for each user group and capture heterogeneity among users, random parameters ordered probit
5
models are adopted. They are more suitable for modeling ordinal dependent variables compared to
6
mixed logit models. This method was often chosen over the ordered logit models due to the
7
underlying assumption of normality of the error term in ordered probit models and their ability to
8
bypass proportional odds assumption in ordered logit models (Washington et al., 2012; Meng et
9
al., 2018).
10
The generalized utility function of ordered probit models can be presented as follows:
11
n n n
yx
=+
(1)
where 𝑦𝑛
∗ is a latent variable determining the service usage frequency for each user n, (n = 1,…, N,
12
where N is the total number of users), 𝑥𝑛 is the vector of independent variables, 𝛽 is the vector of
13
estimable coefficients, and 𝜀𝑛 is a random error term under the assumption that it follows a standard
14
normal distribution.
15
The ridesourcing service usage frequency for user n is defined as,
16
0
01
1
1, ,
2, ,
3, ,
nn
nn
nn
y if y u
y if u y u
y if u y
=
=
=
(2)
where 𝜇 are estimable parameters or thresholds that define 𝑦𝑛 as integer ordering converted from
17
ordered responses. In this study, participants were asked to provide the exact number of times that
18
they used ridesourcing services in the last month before the survey. Most of their answers are
19
skewing towards the lower end as it is still much more expensive to use ridesourcing services
20
compared to public transit (e.g., it costs about 20 CNY to use ridesourcing service, 3 CNY to use
21
subway, or 1.5 CNY to use bus for 3-kilometer ride), their answers were classified into low,
22
medium, and high frequencies for modeling purpose. 𝜇 and 𝛽 were estimated jointly by determining
23
the probability of three specific ordered responses for each user n. Generally, the threshold
24
parameter µ
0 is set as 0. Hence, the ordered probit model results can be presented in the form of
25
ordered selection probabilities as follows,
26
1
1
( 1)=
( 2)=
(
( 3)=
)
( ) ( )
1 ( )
n n n
n n n n
n n n
P y x
P y u x x
P y u x
−
− −
− −
=
=−
=
(3)
where Φ(⋅) is the standard normal cumulative distribution function.
27
Many studies in the transportation domain used random parameters models to discover the
28
unobserved heterogeneity from responses of travelers (Zhang et al., 2014a; Guo et al., 2016; Guo
29
et al., 2018; Guo and Peeta, 2020; Guo et al., 2021b). Fixed parameters can result in inconsistent,
30
inefficient, and biased parameter estimates (Washington et al., 2012). By adding an error term that
31
correlates with the unobserved factors in 𝜀, individual heterogeneity can be translated into
32
parameter heterogeneity as follows:
33
nn
=+
(4)
where βn is a vector of estimable parameters that varies across users n, β is the vector of mean
34
parameter estimates across all users, and 𝜑𝑛 is a vector of randomly distributed terms (following
35
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
9
standard normal distribution in this study). 300 Halton draws are used in simulated maximum
1
likelihood estimation for random parameters ordered probit models (Washington et al., 2012).
2
In order to understand the changes in the probability of each usage frequency level when
3
changing the independent variables (i.e., a one-unit change of a continuous independent variable,
4
or by a change from ‘0’ to ‘1’ for an indicator variable), each variable’s marginal effects were
5
calculated. The formula of marginal effects at the sample means of each type of service usage
6
frequency can be written as:
7
1
1
( 1) ( ) '
( 2) [ ( ) ( )] '
( 3) ( ) '
Py x
x
Py xx
x
Py x
x
== − −
== − − −
== − −
(5)
where 𝜙 represents the density function of the standard normal distribution, and all other terms are
8
as previously defined.
9
Some literature (Scheiner and Holz-Rau, 2012; Guo et al., 2016, Guo et al., 2018) in terms of
10
studying travel behavior has found that their travel behavior for different sub-populations (e.g.,
11
migrants and residents) were affected by different factors. This study explores the possibility that
12
different factors affect ridesourcing service usage frequency for users with different primary usage
13
purposes (i.e., commute, recreation, and transfer). To justify this assumption, the combined
14
likelihood ratio tests proposed by Washington et al. (2012) are used to test the parameter
15
transferability across a number of groups. The formula is as follows:
16
( ) ( )
1
2M
Tm
m
LR LL LL
=
= − −
(6)
where
()
T
LL B
is the log-likelihood at the convergence of model estimated by using all user data,
17
()
m
LL B
is the log-likelihood at the convergence of model estimated by using group data, and M is
18
the total number of groups. The test statistics follow a chi-squared distribution, with degrees of
19
freedom equal to the summation of the number of parameters in all the models using the group data
20
minus the number of parameters in the overall mode.
21
22
4. Results and discussion
23
This paper used the random parameters ordered probit models to estimate the impacts of
24
contributing factors on ridesourcing service usage frequency for each user type. Potential
25
independent variables were selected from the aforementioned studies. Some of these variables were
26
removed due to multicollinearities. The remaining ones were estimated together, see Table 2. Two
27
main principles were used when determining the remaining ones: (i) it should improve the model’s
28
goodness-of-fit, and (ii) it can be used to provide valuable insights into people’s behavior and can
29
be used to assist in designing relevant policies and strategies to influence people’s behavior. Some
30
variables were newly formulated from the original survey and were not limited to their original
31
format. There are some benefits and limitations for doing so and should be judged on a case-by-
32
case basis. Taking age as an example, it was originally formulated as an ordered variable but was
33
not found to be statistically significant. Considering its importance stated in the literature (Dias et
34
al., 2017; Alemi et al., 2019; Barbour et al., 2020; Rizki et al., 2021), it was converted into a binary
35
variable. Similar conversions were made in other travel behavior related studies in the literature
36
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
10
such as Krueger et al. (2016), Haboucha et al. (2017), Lavieri et al. (2017), Zoellick et al. (2019),
1
Patel et al. (2022), and Goldbach et al. (2022) to name just a few.
2
Table 2. Results of crosstabs between independent variables and ridesourcing service usage
3
frequency
4
Independent variable
Spearman’s correlation coefficient (chi-square test)
Commute users
Recreational users
Transfer users
Sociodemographic characteristics
Young generation indicator (if traveler’s age
is below 40, 0 otherwise)
0.219**
0.264**
0.228**
High-salary indicator (if traveler’s monthly
salary is over 10,000 CNY, 0 otherwise)
0.183**
0.206**
0.056
Reasons to use ridesourcing services
Comfort/convenience indicator (if traveler
choose comfort/convenience as reason, 0
otherwise)
0.286**
0.253**
0.241**
Short waiting time indicator (if traveler
choose short waiting time as reason, 0
otherwise)
0.091
0.270**
0.213**
Reasonable cost indicator (if traveler choose
reasonable cost as reason, 0 otherwise)
0.249**
0.103
0.321**
Other behavioral characteristics
Taxi preference indicator (if traveler prefer to
use a taxi over using ridesourcing
services, 0 otherwise)
-0.035
-0.112**
-0.033
Transit satisfaction indicator (if traveler
choose slightly satisfied, satisfied,
strongly satisfied, 0 otherwise)
-0.274**
-0.271**
-0.094
** Statistically significant at 0.01 level
5
The selection of random parameters is not predetermined or arbitrarily decided but is based
6
on the model estimation results. When the variance of random parameter was significant at the 0.95
7
level of confidence or greater, the estimated coefficient of the variable was considered as the
8
random parameter. It would have a fixed parameter if otherwise. The introduction of these random
9
parameters can improve the model prediction accuracy and offer more insights of the model results
10
that fixed parameter models cannot bring (Washington et al., 2012; Sadri et al., 2013; Azimi et al.,
11
2020; Fountas et al., 2021; Guo et al., 2021).
12
In terms of modeling choice, modeling all data together was considered. It is common practice
13
to add a new variable, labeled “the primary purpose of usage”, and use one of the choices as the
14
reference (or baseline) group. This can offer valuable insights in terms of whether some users have
15
a higher or lower usage frequency based on their primary purpose of usage. However, it cannot
16
illustrate if one variable affects each group of users differently. In addition, as stated in the
17
introduction section, it is very likely that there are some behavioral differences among these users.
18
This was supported by both Chi-square test of independence and combined likelihood ratio tests
19
estimated by Eq.(6) (60.98 > F (21,0.05) =32.67). Hence, separate models were created for each
20
group of users based on their primary purpose of usage.
21
Three separate models have been estimated with same independent variables and random
22
parameters. The common utility function can be presented as follows:
23
24
1 1 2 2 3 3 4 4 5 5 6 6 7 7
y x x x x x x x
= + + + + + + + +
(7)
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
11
The meaning of independent variables and parameters are shown in Table 3, is constant, other
1
terms in equation are as previously defined. The results were estimated using Nlogit software
2
(Tables 4-6). Variables with random parameters can be found in Table 7. The estimated results for
3
commute users, social/recreational users, and transfer users are presented as Eq. (8-10) respectively,
4
the parameters with superscript r were random parameters as they had significant variance.
5
1 2 3 4 5 6 7
0.79 0.68 0.92 1.16 0.54 0.96 0.26 0.88
rr
y x x x x x x x
= + + + + + − − +
(8)
1 2 3 4 5 6 7
0.01 0.58 0.50 0.43 0.57 0.12 0.13 0.43
r
y x x x x x x x
= + + + + + − − +
(9)
1 2 3 4 5 6 7
0.20 0.67 0.36 0.57 0.53 0.78 0.21 0.17
r r r
y x x x x x x x
= + + + + + − − +
(10)
Table 3. Independent variables and parameters for three types of users
6
Parameter
Independent
variable
Description
1
1
x
Young generation indicator (if traveler’s age is below 40, 0 otherwise)
2
2
x
High-salary indicator (if traveler’s monthly salary is over 10,000 CNY, 0
otherwise)
3
3
x
Comfort/convenience indicator (if traveler choose comfort/convenience
as reason, 0 otherwise)
4
4
x
Short waiting time indicator (if traveler choose short waiting time as
reason, 0 otherwise)
5
5
x
Reasonable cost indicator (if traveler choose reasonable cost as reason, 0
otherwise)
6
6
x
Taxi preference indicator (if traveler prefer to use a taxi over using
ridesourcing services, 0 otherwise)
7
7
x
Transit satisfaction indicator (if traveler choose slightly satisfied,
satisfied, strongly satisfied, 0 otherwise)
7
4.1.Sociodemographic Characteristics
8
The young generation and high salary indicator variables were two of the most important
9
sociodemographic factors based on model estimation results. Both factors positively affect
10
ridesourcing service usage frequency and their effects were similar across all groups. Young
11
generations (under 40 years old) or high-salary (monthly salary is over Shanghai average, 10,000
12
CNY) users were more inclined to have high service usage frequency (i.e., over 3 times a month)
13
compared to their respective counterparts.
14
In terms of potential generational differences, the model estimation results show that younger
15
generations in China (under 40 years old or born after the 1980s) behave differently from older
16
generations in terms of their ridesourcing usage behavior. This difference is similar to the findings
17
in other travel behavior-related studies in China (Dias et al., 2017; Aguilera-García et al.,2022).
18
Younger generations in China grew up during a rapid economic development era compared to older
19
ones. They experienced booming car ownership and usage, and witnessed the proliferation of
20
Internet-related technology and services (i.e., e-commerce and smartphones), while most older
21
generations experienced hunger and poverty at a younger age. These experience differences
22
manifest in different travel-related behavior. Younger generations have a stronger preference for
23
using cars to travel over other modes of transportation and are more comfortable with smartphone
24
app-based services compared to older generations (Guo et al., 2018; Guo et al., 2020a). Thus, the
25
introduction of ridesourcing services may have brought a larger travel mode choice shift among
26
younger generations, particularly among those who prefer to use them for commute compared to
27
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
12
older generations. It is possible that these travelers are more susceptible to such app-based mobility
1
services using cars and many of them believe that such service is a better travel mode choice
2
compared to the ones they used before. Unlike their younger generation counterpart, older
3
generations (over 40 years of age) grew up during a period when public transportation was the
4
primary mode choice and most of them expressed various types of challenges when using app-
5
related services (Gao et al., 2015). This also highlights the importance of tapping into the potential
6
ridesourcing service market among older users (over 40 years of age) with effective and innovative
7
strategies. Possible strategies can include increasing advertising among older travelers in their
8
frequently visited locations, developing more user-friendly app interfaces for older travelers, and
9
offering customized ridesourcing services for older travelers, such as special picking-up services
10
or vehicles for physically challenged travelers, and location sharing with close relatives (e.g.,
11
children and partners).
12
Regarding the income differences, individuals with high monthly salaries (higher than the
13
Shanghai average) tend to have a higher ridesourcing services usage frequency. Similar results
14
were also observed by Dias et al. (2017) and Lavieri and Bhat (2019). It is possible that the current
15
ridesourcing services are still considered as an expensive mode of transportation compared to
16
public transportation. For example, it would take only 3 CNY to travel within 6 kilometers (around
17
6 stations in urban areas) using subway mode, which may take up to 20 CNY using ridesourcing
18
services in Shanghai. This highlights the need to develop strategies to promote ridesourcing
19
services among low-to medium-income populations. Possible strategies can include pooled
20
services (e.g., ride-splitting), fare deductible programs (e.g., customer loyalty programs for
21
discounts or rewards), and increasing service in regions where public transportation and shared
22
bike service are lacking to make ridesourcing service more accessible among low- or moderate-
23
income population.
24
4.2.Reasons to use ridesourcing services
25
In terms of reasons to use ridesourcing services, indicator variables including comfort/convenience,
26
short waiting time, and reasonable cost were found to have statistically significant impacts on some
27
or all user groups, but their impacts are different. Commute users were more likely to have a high
28
usage frequency of ridesourcing services if they chose comfort/convenience and reasonable cost as
29
a reason for using such services. It is possible that these users consider ridesourcing services as the
30
less mentally and physically taxing travel mode compared to often crowded public transportation
31
or driving by themselves in often congested peak-hour traffic of Shanghai.
32
Regarding the short waiting time indicator, it has a random parameter. According to the mean
33
and variance of random parameters submitted to the normal distribution, 61.0% of commute users
34
who chose short waiting time as a reason were more likely to have a higher level of usage frequency,
35
while the rest were less likely. This suggested that most commute users may not need to wait for a
36
long time for ridesourcing services compared to other modes of transportation, while nearly 40%
37
of them may have experienced difficulties in receiving ridesourcing services during commute hours.
38
This highlights the importance of improving ridesourcing service frequency during commute hours
39
to reduce waiting time among commuters. Possible strategies can include offering priority services
40
at a prime price, service reservation systems, and other methods to improve service reliability. It is
41
also possible to introduce pooled ridesourcing services with pick-up priorities to commuters to
42
improve vehicle utilization and shorten the waiting time (Schaller, 2021).
43
With regard to social/recreational trips, the comfort/convenience indicator is a random
44
parameter. Most of the users (about 92.2%) were more likely to have high service usage frequency
45
if they chose comfort/convenience as a reason for using such services, while a very small portion
46
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
13
of them (less than 10%) was the opposite. It is possible that a few of them may prefer to drive by
1
themselves as they may consider that it offers more comfort/convenience (e.g., when to leave and
2
when to stop) compared to using ridesourcing services. Model estimation results and marginal
3
effects also showed that if social/recreational users chose short waiting time as a reason for using
4
ridesourcing services, they were more likely to have high usage frequency for such services similar
5
to most of the commuters. These results highlight the importance of offering fast ridesourcing
6
services to potential users. Model results also found that the reasonable cost indicator is not
7
significant in the social/recreational user model. It is possible that people may not factor trip cost
8
for social/recreational trips compared to comfort/convenience and waiting time.
9
In terms of transfer users, the model estimation results and marginal effects showed that if
10
transfer users chose reasonable cost as a reason for using ridesourcing services, they were more
11
likely to have medium or high usage frequency for such services. Both comfort/convenience and
12
short waiting time indicator are random parameters, most of the users who chose short waiting time
13
(84.6%) or comfort/convenience (98.3%) as their primary reason for using ridesourcing services
14
were more likely to have medium or high ridesourcing service usage frequency, while a very small
15
portion of them was the opposite. It’s possible that a few of these users belong to low-income
16
people. Despite they believe that ridesourcing services are comfortable and convenient and can
17
reduce their waiting time, they might consider that a reasonable price is the most important factor
18
affecting their ridesourcing service usage frequency. In general, most of them were more likely to
19
have medium or high ridesourcing service usage frequency. These results suggest that most transfer
20
users consider ridesourcing services as both cheaper and more comfortable/convenient. It is
21
possible that these users want to ensure that they can arrive at their transfer locations (e.g., airport
22
and train stations) on time and the parking fee associated with driving their own vehicle can be
23
high. Ridesourcing service companies can potentially tailor some pick-up and drop-off services for
24
these travelers (e.g., cost discounts for returning trips from the transfer station and free baggage
25
pick-up or shipping service) to expand the attraction of ridesourcing services for transfer users.
26
To sum up, strategies to improve comfort/convenience, reduce waiting time, and offer cost-
27
saving services (actual or perceived) can be effective in improving the overall service usage
28
frequency. Possible strategies can be classified into three main categories, including service
29
improvement measures (e.g., improving in-vehicle conditions), time-saving instruments (e.g.,
30
encouraging ride-splitting to reduce waiting time), and price-based strategies (e.g., discounted
31
prices, subsidies, credits) (Kockelman and Kalmanje, 2005; Zhang et al., 2014b; Amirkiaee and
32
Evangelopoulos, 2018; Azhdar and Nazemi, 2020; Guo et al., 2021a). However, their effects can
33
be very different across user groups and may even be ineffective or even counterproductive for
34
some users. Additional studies are needed to understand the reasons behind these differences.
35
4.3.Other Behavioral Characteristics
36
The variable measuring taxi preference was found to have a statistically significant impact on the
37
frequency of ridesourcing service usage for all user groups, but the nature of the impact differed.
38
Users who prefer to use taxis over ridesourcing services are less likely to have a high service usage
39
frequency, particularly during commute hours. These results highlight the competition between
40
taxis and ridesourcing services, with traditional taxis experiencing declining market share as
41
observed by Kong et al. (2020), Yang et al. (2022), and Loa et al. (2023), while ridesourcing
42
services are gaining market share. Compared to taxis, ridesourcing service providers find it easier
43
to enter the market as long as they meet the requirements of qualified vehicles, three years of
44
driving experience, and no criminal record. Therefore, even though some travelers still prefer taxis
45
due to better professional quality and consumer security, many taxi users have been attracted to
46
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
14
ridesourcing services for their faster response times. This highlights the importance of preventing
1
ridesourcing services from growing disorderly and engaging in malicious competition through low
2
prices.
3
The transit satisfaction indicator variable was also found to have a negative and statistically
4
significant effect on social/recreational users based on the model estimation results and marginal
5
effects. These users who were satisfied with the transit services in their neighborhoods were less
6
likely to have high service usage frequency. However, the transit satisfaction indicator was a
7
random parameter for commute and transfer users. According to the mean and variance, it means
8
these users (85.5% commute users and 55.2% transfer users) who were satisfied with the transit
9
services in their neighborhoods were less likely to have high service usage frequency, while the
10
rest were more likely. These results suggest that transit services have a mixed relationship (mostly
11
competing) in terms of providing services for commute and transfer users, and transit services and
12
ridesourcing services may also complement each other in serving commute users and transfer users,
13
particularly for multimodal users.
14
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
15
Table 4. Model estimation results for commute users (N=261)
Independent variable
Coefficient
Z-Statistics
Average marginal effects (random parameters model)
Low frequency
Medium frequency
High frequency
Constant
0.79**
2.25
Sociodemographic characteristics
1
x
(Young generation indicator)
0.68**
2.48
-0.04
-0.22
0.26
2
x
(High-salary indicator)
0.92**
4.43
-0.03
-0.32
0.35
Reasons to use ridesourcing services
3
x
(Comfort/convenience indicator)
1.16**
(0.14)
5.60
(1.09)
-0.06
-0.38
0.44
4
x
(Short waiting time indicator)
0.54**
(1.81**)
2.67
(9.60)
-0.02
-0.19
0.21
5
x
(Reasonable cost indicator)
0.96**
(0.24)
4.91
(1.82)
-0.04
-0.33
0.37
Other behavioral factors
6
x
(Taxi preference indicator)
-0.26**
(0.01)
-3.79
(0.32)
0.01
0.08
-0.09
7
x
(Transit satisfaction indicator)
-0.88**
(0.87**)
-4.27
(4.92)
0.05
0.29
-0.34
u1
2.09**
10.05
McFadden pseudo-R2
0.16
Log-likelihood at convergence
-216.12
The figure in brackets is variance of random parameter
* Statistically significant at 0.05 level
** Statistically significant at 0.01 level
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
16
Table 5. Model estimation results for social/recreational users (N=256)
Independent variable
Coefficient
Z-Statistics
Average marginal effects (random parameters model)
Low frequency
Medium frequency
High frequency
Constant
0.01
0.02
Sociodemographic characteristics
1
x
(Young generation indicator)
0.58**
3.02
-0.18
-0.01
0.19
2
x
(High-salary indicator)
0.50**
3.10
-0.14
-0.04
0.18
Reasons to use ridesourcing services
3
x
(Comfort/convenience indicator)
0.43**
(0.21**)
2.85
(2.82)
-0.13
-0.02
0.15
4
x
(Short waiting time indicator)
0.57**
(0.46)
3.56
(0.51)
-0.17
-0.01
0.18
5
x
(Reasonable cost indicator)
0.12
(0.01)
0.78
(0.08)
-0.03
-0.01
0.04
Other behavioral characteristics
6
x
(Taxi preference indicator)
-0.13**
(0.01)
-2.17
(0.48)
0.02
0.01
-0.03
7
x
(Transit satisfaction indicator)
-0.43**
(0.07)
-2.87
(0.67)
0.13
0.02
-0.15
u1
1.24**
11.60
McFadden pseudo-R2
0.12
Log-likelihood at convergence
-242.74
The figure in brackets is the variance of random parameter
* Statistically significant at 0.05 level
** Statistically significant at 0.01 level
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
17
Table 6. Model estimation results for transfer users (N=266)
Independent variable
Coefficient
Z-Statistics
Average marginal effects (random parameters model)
Low frequency
Medium frequency
High frequency
Constant
0.20
0.75
Sociodemographic characteristics
1
x
(Young generation indicator)
0.67**
3.72
-0.23
0.09
0.14
2
x
(High-salary indicator)
0.36**
2.27
-0.11
0.02
0.09
Reasons to use ridesourcing services
3
x
(Comfort/convenience indicator)
0.57**
(0.35**)
3.57
(2.93)
-0.17
0.02
0.15
4
x
(Short waiting time indicator)
0.53**
(0.46**)
3.34
(4.45)
-0.17
0.05
0.12
5
x
(Reasonable cost indicator)
0.78**
(0.11)
4.91
(0.09)
-0.24
0.04
0.20
Other behavioral characteristics
6
x
(Taxi preference indicator)
-0.21**
(0.00)
-5.72
(0.14)
0.08
-0.02
-0.06
7
x
(Transit satisfaction indicator)
-0.17
(0.92**)
-0.10
(6.78)
0.02
-0.01
-0.01
u1
1.68**
12.25
McFadden pseudo-R2
0.12
Log-likelihood at convergence
-253.46
The figure in brackets is the variance of random parameter
* Statistically significant at 0.05 level
** Statistically significant at 0.01 level
Feiyu Feng, Xinghua Li, Yuntao Guo, & Cheng Cheng
18
Table 7. Summary of findings: effects on service usage frequency for various types of users
Independent variable
Commute users
Social/recreational users
Transfer users
Sociodemographic characteristics
1
x
(Young generation indicator)
↑
↑
↑
2
x
(High-salary indicator)
↑
↑
↑
Reasons to use ridesourcing services
3
x
(Comfort/convenience indicator)
↑
Mixed effect
(primarily↑)
Mixed effect
(primarily↑)
4
x
(Short waiting time indicator)
Mixed effect
(primarily↑)
↑
Mixed effect
(primarily↑)
5
x
(Reasonable cost indicator)
↑
—
↑
Other behavioral characteristics
6
x
(Taxi preference indicator)
↓
↓
↓
7
x
(Transit satisfaction indicator)
Mixed effect
(primarily↓)
↓
Mixed effect
(primarily↓)
↑ Positive and statistically significant
↓ Negative and statistically significant
— Not statistically significant
Mixed effect suggests the variable has a random parameter
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
19
5. Conclusions
1
This study investigated ridesourcing service users’ service usage frequency in Shanghai, China
2
and its influencing factors. Random parameters ordered probit models were estimated by users’
3
primary purpose of usage (using ridesourcing services primarily for commute, social/recreational,
4
or transfer trips) to account for unobserved heterogeneity and identify group-specific influencing
5
factors. The model estimation results show that sociodemographic characteristics (age and salary),
6
reasons to use ridesourcing services (reasons for choosing ridesourcing services included
7
comfort/convenience, short waiting time, and reasonable cost), and other behavioral characteristics
8
(preference between taxi and ridesourcing services, and level of satisfaction with transit services
9
in their neighborhoods) impact ridesourcing service usage frequency, and these impacts are
10
different among groups of users. Comfort/convenience (the reason to choose ridesourcing services)
11
is one of the more important factors that affect users’ ridesourcing service usage frequency for
12
commute users, while short waiting time and reasonable cost are more important factors that affect
13
users’ ridesourcing usage frequency of social/recreational users and transfer users, respectively.
14
First, despite the best of the authors’ efforts, there are some inherent limitations of web-based
15
survey. It is a “first come, first answer” system that cannot ensure total randomization of the
16
selection process. In addition, its participant pool is generally younger compared to the general
17
population due to its requirement for Internet familiarity and low payout rate. The authors have
18
planned a new study by combining online and in-person survey to improve the result
19
representativeness level, while increasing the survey sample size and adding additional studies
20
regions at the same time to provide a more comprehensive understanding towards ridesourcing
21
service usage frequency. This future study can provide new insights into the understanding of the
22
potential similarities and differences among travelers in cities with different sizes, travel behavior,
23
and geographic locations in terms of their service usage frequency. Second, as this study is one of
24
the early efforts that identify the potential group-specific factors affecting their service usage
25
frequency, additional psychological factors (e.g., attitude and perception related to ridesourcing
26
services), built environment factors (e.g., land use mix and accessibility), and other variables such
27
as time of day and weather condition can be included in future studies. Finally, the COVID-19
28
pandemic in Shanghai was well under control at the time of study. No domestic transmitted
29
COVID-19 case was reported in the past 7 months in Shanghai before the survey study according
30
to Shanghai Municipal Health Commission (2020-2021). It is possible that COVID-19 was and
31
still is affecting people’s usage frequency of ridesourcing services. Future studies can potentially
32
include variables such as people’s attitudes towards prevention and control measures, trust in
33
ridesourcing services, attitude toward the COVID-19 pandemic, and apprehension about pandemic
34
rebound.
35
36
Disclosure statement
37
The authors report there are no competing interests to declare.
38
39
Data availability statement
40
The data that support the findings of this study are available from the corresponding author,
41
Yuntao Guo, upon reasonable request.
42
43
Acknowledgments
44
This study is supported by the National Natural Science Foundation of China (52272322).
45
46
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
20
ORCID
1
Feiyu Feng https://orcid.org/0000-0003-0911-5712
2
Xinghua Li https://orcid.org/0000-0003-3812-4271
3
Yuntao Guo https://orcid.org/0000-0001-9649-5121
4
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
21
References
1
Aghaabbasi, Mahdi, Zohreh Asadi Shekari, Muhammad Zaly Shah, Oloruntobi Olakunle, Danial
2
Jahed Armaghani, and Mehdi Moeinaddini. 2020. “Predicting the use frequency of ride-sourcing
3
by off-campus university students through random forest and Bayesian network techniques.”
4
Transportation Research Part A: Policy and Practice 136:262-81. doi:
5
https://doi.org/10.1016/j.tra.2020.04.013.
6
Aguilera-García, Álvaro, Juan Gomez, Guillermo Velázquez, and Jose Manuel Vassallo. 2022.
7
“Ridesourcing vs. traditional taxi services: Understanding users’ choices and preferences in Spain.”
8
Transportation Research Part A: Policy and Practice 155: 161-178. doi:
9
10.1016/j.tra.2021.11.002.
10
Alemi, Farzad, Giovanni Circella, Susan Handy, and Patricia Mokhtarian. 2018. “What influences
11
travelers to use Uber? Exploring the factors affecting the adoption of on-demand ride services in
12
California.” Travel Behaviour and Society 13: 88-104. doi: 10.1016/j.tbs.2018.06.002.
13
Alemi, Farzad, Giovanni Circella, Patricia Mokhtarian, and Susan Handy. 2019. “What drives the
14
use of ridehailing in California? Ordered probit models of the usage frequency of Uber and Lyft.”
15
Transportation Research Part C Emerging Technologies 102: 233-248. doi:
16
10.1016/j.trc.2018.12.016.
17
Amirkiaee, S. Yasaman, and Nicholas Evangelopoulos. 2018. “Why do people rideshare? An
18
experimental study.” Transportation Research Part F: Traffic Psychology and Behaviour 55: 9-
19
24. doi: 10.1016/j.trf.2018.02.025.
20
Azhdar, Reihaneh, and Ali Nazemi. 2020. “Modeling of incentive-based policies for demand
21
management for the Tehran subway.” Travel Behaviour and Society 20: 174-180. doi:
22
10.1016/j.tbs.2020.03.014.
23
Azimi, Ghazaleh, Alireza Rahimi, and Xia Jin. 2022. “Exploring the attitudes of Millennials and
24
Generation Xers toward ridesourcing services.” Transportation 49 (6):1765-99. doi:
25
10.1007/s11116-021-10227-y.
26
Bansal, Prateek, Yang Liu, Ricardo Daziano, and Samitha Samaranayake. 2020. “Impact of
27
discerning reliability preferences of riders on the demand for mobility-on-demand services.”
28
Transportation Letters 12: 677-681. doi: 10.1080/19427867.2019.1691298.
29
Barbour, Natalia, Yu Zhang, and Fred Mannering. 2020. “An exploratory analysis of the role of
30
socio-demographic and health-related factors in ridesourcing behavior.” Journal of Transport &
31
Health 16: 100832. doi: 10.1016/j.jth.2020.100832.
32
Belgiawan, Prawira Fajarindra, Tri Basuki Joewono, and Muhammad Zudhy Irawan. 2022.
33
“Determinant factors of ride-sourcing usage: A case study of ride-sourcing in Bandung, Indonesia. ”
34
Case Studies on Transport Policy 10 (2):831-40. doi: https://doi.org/10.1016/j.cstp.2022.02.010.
35
Beojone, Caio Vitor, and Nikolas Geroliminis. 2021. “On the inefficiency of ride-sourcing services
36
towards urban congestion.” Transportation Research Part C: Emerging Technologies 124:102890.
37
doi: https://doi.org/10.1016/j.trc.2020.102890.
38
Brown, Anne E. 2020. Who and where rideshares? “Rideshare travel and use in Los Angeles.”
39
Transportation Research Part A: Policy and Practice 136: 120-134. doi:
40
10.1016/j.tra.2020.04.001.
41
Choudhury, Charisma F., Lang Yang, João de Abreu e Silva, and Moshe Ben-Akiva. 2018.
42
“Modelling preferences for smart modes and services: A case study in Lisbon.” Transportation
43
Research Part A: Policy and Practice 115:15-31. doi: https://doi.org/10.1016/j.tra.2017.07.005.
44
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
22
Cyberspace Administration of China. 2021. “The 47th China Statistical Report on Internet
1
Development.” Cyberspace Administration of China. Accessed February 2021.
2
http://www.cac.gov.cn/2021-02/03/1613923423079314.htm
3
de Souza Silva, Laize Andréa, Maurício Oliveira de Andrade, and Maria Leonor Alves Maia. 2018.
4
“How does the ride-hailing systems demand affect individual transport regulation?” Research in
5
Transportation Economics 69: 600-606. doi: 10.1016/j.retrec.2018.06.010.
6
Dias, Felipe F., Patrícia S. Lavieri, Venu M. Garikapati, Sebastian Astroza, Ram M. Pendyala, and
7
Chandra R. Bhat. 2017. “A behavioral choice model of the use of car-sharing and ride-sourcing
8
services.” Transportation 44: 1307-1323. doi: 10.1007/s11116-017-9797-8.
9
Fountas, Grigorios, Achille Fonzone, Adebola Olowosegun, and Clare McTigue. 2021.
10
“Addressing unobserved heterogeneity in the analysis of bicycle crash injuries in Scotland: A
11
correlated random parameters ordered probit approach with heterogeneity in means.” Analytic
12
Methods in Accident Research 32:100181. doi: https://doi.org/10.1016/j.amar.2021.100181.
13
Gao, Shang, John Krogstie, and Yuhao Yang. 2015. “Differences in the Adoption of Smartphones
14
Between Middle Aged Adults and Older Adults in China.” Paper presented at International
15
Conference on Human Aspects of IT for the Aged Population, Los Angeles, USA, August 451–
16
462.
17
Ghaffar, Arash, Suman Mitra, and Michael Hyland. 2020. “Modeling determinants of ridesourcing
18
usage: A census tract-level analysis of Chicago.” Transportation Research Part C: Emerging
19
Technologies 119:102769. doi: 10.1016/j.trc.2020.102769.
20
Gill, J., Johnson, P. and Clark. M. 2010. Research Methods for Managers. Washington DC: SAGE
21
Publications.
22
Goldbach, Carina, Jörn Sickmann, Thomas Pitz, and Tatjana Zimasa. 2022. “Towards autonomous
23
public transportation: Attitudes and intentions of the local population.” Transportation Research
24
Interdisciplinary Perspectives 13:100504. doi: 10.1016/j.trip.2021.100504.
25
Guo, Yuntao, Shubham Agrawal, Srinivas Peeta, and Sekhar Somenahalli. 2016. “Impacts of
26
Property Accessibility and Neighborhood Built Environment on Single-Unit and Multi-Unit
27
Residential Property Values.” Transportation Research Record, 2568(1): 103–112. doi:
28
10.3141/2568-15.
29
Guo, Yuntao, Yaping Li, Panagiotis Ch. Anastasopoulos, Srinivas Peeta, and Jian Lu. 2021a.
30
“China’s millennial car travelers’ mode shift responses under congestion pricing and reward
31
policies: A case study in Beijing.” Travel Behaviour and Society 23: 86-99. doi:
32
10.1016/j.tbs.2020.11.004.
33
Guo, Yuntao, and Srinivas Peeta. 2020a. “Impacts of personalized accessibility information on
34
residential location choice and travel behavior.” Travel Behaviour and Society 19: 99-111. doi:
35
10.1016/j.tbs.2019.12.007.
36
Guo, Yuntao, Srinivas Peeta, Shubham Agrawal, and Irina Benedyk. 2021b. “Impacts of Pokémon
37
GO on route and mode choice decisions: exploring the potential for integrating augmented reality,
38
gamification, and social components in mobile apps to influence travel decisions.” Transportation
39
49: 395-444. doi: 10.1007/s11116-021-10181-9.
40
Guo, Yuntao, Jian Wang, Srinivas Peeta, and Panagiotis Ch Anastasopoulos. 2018. “Impacts of
41
internal migration, household registration system, and family planning policy on travel mode
42
choice in China.” Travel Behaviour and Society 13: 128-143. doi: 10.1016/j.tbs.2018.07.003.
43
Guo, Yuntao, Jian Wang, Srinivas Peeta, and Panagiotis Ch Anastasopoulos. 2020b. “Personal and
44
societal impacts of motorcycle ban policy on motorcyclists’ home-to-work morning commute in
45
China.” Travel Behaviour and Society 19: 137-150. doi: 10.1016/j.tbs.2020.01.002.
46
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
23
Guo, Yuntao, Xinwu Qian, Tian Lei, Shuocheng Guo, and Lei Gong. 2022. “Modeling the
1
preference of electric shared mobility drivers in choosing charging stations.” Transportation
2
Research Part D: Transport and Environment 110:103399. doi: 10.1016/j.trd.2022.103399.
3
Haboucha, Chana J., Robert Ishaq, and Yoram Shiftan. 2017. “User preferences regarding
4
autonomous vehicles.” Transportation Research Part C: Emerging Technologies 78:37-49. doi:
5
10.1016/j.trc.2017.01.010.
6
Jin, Scarlett T., Hui Kong, Rachel Wu, and Daniel Z. Sui. 2018. “Ridesourcing, the sharing
7
economy, and the future of cities. Cities 76: 96-104.” doi: 10.1016/j.cities.2018.01.012.
8
Kamelifar, Mohammad Javad, Behzad Ranjbarnia, and Houshmand Masoumi. 2022. “The
9
Determinants of Walking Behavior before and during COVID-19 in Middle-East and North Africa:
10
Evidence from Tabriz, Iran.” Sustainability 14 (7):3923. https://doi.org/10.3390/su14073923.
11
Kang, Shuqing, Aupal Mondal, Aarti C. Bhat, and Chandra R. Bhat. 2021. “Pooled versus private
12
ride-hailing: A joint revealed and stated preference analysis recognizing psycho-social factors.”
13
Transportation Research Part C: Emerging Technologies 124:102906. doi:
14
https://doi.org/10.1016/j.trc.2020.102906.
15
Kockelman, Kara M., and Sukumar Kalmanje. 2005. “Credit-based congestion pricing: a policy
16
proposal and the public's response.” Transportation Research Part A 39: 671-690. doi:
17
10.1016/j.tra.2005.02.014.
18
Kong, Hui, Xiaohu Zhang, and Jinhua Zhao. 2020. "Is ridesourcing more efficient than taxis?"
19
Applied Geography 125:102301. doi: 10.1016/j.apgeog.2020.102301.
20
Krueger, Rico, Taha H. Rashidi, and John M. Rose. 2016. “Preferences for shared autonomous
21
vehicles.” Transportation Research Part C: Emerging Technologies 69:343-55. doi:
22
10.1016/j.trc.2016.06.015.
23
Lavieri, Patrícia, Venu Garikapati, Chandra Bhat, Ram Pendyala, Sebastian Astroza, and Felipe F.
24
Dias. 2017. “Modeling Individual Preferences for Ownership and Sharing of Autonomous Vehicle
25
Technologies.” Transportation Research Record: Journal of the Transportation Research Board
26
2665:1-10. doi: 10.3141/2665-01.
27
Lavieri, Patrícia S., and Chandra R. Bhat. 2019. “Investigating objective and subjective factors
28
influencing the adoption, frequency, and characteristics of ride-hailing trips.” Transportation
29
Research Part C: Emerging Technologies 105: 100-125. doi: 10.1016/j.trc.2019.05.037.
30
Lee, Yongsung, Giovanni Circella, Patricia L. Mokhtarian, and Subhrajit Guhathakurta. 2020.
31
“Are millennials more multimodal? A latent-class cluster analysis with attitudes and preferences
32
among millennial and Generation X commuters in California.” Transportation 47 (5):2505-28.
33
doi: 10.1007/s11116-019-10026-6.
34
Lei, Tian, Shuocheng Guo, Xinwu Qian, and Lei Gong. 2022. “Understanding charging dynamics
35
of fully-electrified taxi services using large-scale trajectory data.” Transportation Research Part
36
C: Emerging Technologies 143:103822. doi: https://doi.org/10.1016/j.trc.2022.103822.
37
Li, Yaping, Yuntao Guo, Jian Lu, and Srinivas Peeta. 2019. “Impacts of congestion pricing and
38
reward strategies on automobile travelers’ morning commute mode shift decisions.”
39
Transportation Research Part A: Policy and Practice 125: 72-88. doi: 10.1016/j.tra.2019.05.008.
40
Li, Xinghua, Guanhua Xing, Xinwu Qian, Yuntao Guo, Wei Wang, and Cheng Cheng. 2022a.
41
“Subway Station Accessibility and Its Impacts on the Spatial and Temporal Variations of Its
42
Outbound Ridership.” Journal of Transportation Engineering Part A Systems 148. doi:
43
10.1061/JTEPBS.0000766.
44
Loa, Patrick, Sk Md Mashrur, and Khandker Nurul Habib. 2023. "What influences the substitution
45
of ride-sourcing for public transit and taxi services in Toronto? An exploratory structural equation
46
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
24
model-based study." International Journal of Sustainable Transportation 17 (1):15-28. doi:
1
10.1080/15568318.2021.1978018.
2
Patel, Ronik Ketankumar, Roya Etminani-Ghasrodashti, Sharareh Kermanshachi, Jay Michael
3
Rosenberger, and Ann Foss. 2022. “Exploring willingness to use shared autonomous vehicles.”
4
International Journal of Transportation Science and Technology. doi: 10.1016/j.ijtst.2022.06.008.
5
Sadri, Arif Mohaimin, Satish V. Ukkusuri, and Pamela Murray-Tuite. 2013. “A random parameter
6
ordered probit model to understand the mobilization time during hurricane evacuation.”
7
Transportation Research Part C: Emerging Technologies 32:21-30. doi:
8
10.1016/j.trc.2013.03.009.
9
San Francisco County Transportation Authority. 2017. “TNCs Today: A Profile of San Francisco
10
Transportation Network Company Activity.” San Francisco County Transportation Authority.
11
Accessed June 2017.
12
http://www.sfcta.org/sites/default/files/content/Planning/TNCs/TNCs_Today_112917.pdf
13
Schaller, Bruce. 2021. “Can sharing a ride make for less traffic? Evidence from Uber and Lyft and
14
implications for cities.” Transport Policy 102: 1-10. doi: 10.1016/j.tranpol.2020.12.015.
15
Scheiner, Joachim, and Christian Holz-Rau. 2012. “Gendered travel mode choice: a focus on car
16
deficient households.” Journal of Transport Geography 24:250-61. doi:
17
10.1016/j.jtrangeo.2012.02.011.
18
Shaheen, Susan, and Adam Cohen. 2019. “Shared ride services in North America: definitions,
19
impacts, and the future of pooling.” Transport Reviews 39: 427-442. doi:
20
10.1080/01441647.2018.1497728.
21
Shanghai Municipal Health Commission. 2020. “No indigenous COVID-19 case in Shanghai on
22
March 1”, Shanghai Municipal Health Commission. Accessed March 2 2020.
23
http://wsjkw.sh.gov.cn/tfggwssj-yjcz/20200401/f1e2fb20566c431494dcee30f5e66d9d.html
24
Shanghai Municipal Health Commission. 2021. “Adding 2 indigenous diagnosis cases in Shanghai
25
on August 2, 2021.” Shanghai Municipal Health Commission. Accessed August 3 2021.
26
http://wsjkw.sh.gov.cn/xwfb/20210803/6b9999c3bf564dfe8950a55bc09b0ec9.html
27
Shanghai Bureau of Statistics. 2020. Shanghai Statistical Yearbook .
28
http://tjj.sh.gov.cn/tjnj/20210303/2abf188275224739bd5bce9bf128aca8.html.
29
Stinson, Monique, Bo Zou, Diana Briones, Alicia Manjarrez, and Abolfazl Mohammadian. 2021.
30
“Vehicle ownership models for a sharing economy with autonomous vehicle considerations.”
31
Transportation Letters 4: 1-17. doi: 10.1080/19427867.2021.2007681.
32
Rahimi, Alireza, Ghazaleh Azimi, and Xia Jin. 2020. “Examining human attitudes toward shared
33
mobility options and autonomous vehicles.” Transportation Research Part F: Traffic Psychology
34
and Behaviour 72: 133-154. doi: 10.1016/j.trf.2020.05.001.
35
Rayle, Lisa, Danielle Dai, Nelson Chan, Robert Cervero, and Susan Shaheen. 2016. “Just a better
36
taxi? A survey-based comparison of taxis, transit, and ridesourcing services in San Francisco.”
37
Transport Policy 45: 168-178. doi: 10.1016/j.tranpol.2015.10.004.
38
Rizki, Muhamad, Tri Basuki Joewono, Prawira Fajarindra Belgiawan, and Muhammad Zudhy
39
Irawan. 2021. “The travel behaviour of ride-sourcing users, and their perception of the usefulness
40
of ride-sourcing based on the users' previous modes of transport: A case study in Bandung City,
41
Indonesia.” IATSS Research 45: 267-276. doi: 10.1016/j.iatssr.2020.11.005.
42
Meng, Fanyu, Yuchuan Du, Yuen Chong Li, and S. C. Wong. 2018. “Modeling heterogeneous
43
parking choice behavior on university campuses.” Transportation Planning and Technology 41:
44
154-169. doi: 10.1080/03081060.2018.1407518.
45
Xinghua Li, Feiyu Feng, Yuntao Guo, & Cheng Cheng
25
Mohamed, Mohamed Jama, Tom Rye, and Achille Fonzone. 2019. “Operational and policy
1
implications of ridesourcing services: A case of Uber in London, UK.” Case Studies on Transport
2
Policy 7: 823-836. doi: 10.1016/j.cstp.2019.07.013.
3
Mohamed, Mohamed Jama, Tom Rye, and Achille Fonzone. 2020. “The utilisation and user
4
characteristics of Uber services in London.” Transportation Planning and Technology 43: 424-
5
441. doi: 10.1080/03081060.2020.1747205.
6
Moody, Joanna, Scott Middleton, and Jinhua Zhao. 2019. “Rider-to-rider discriminatory attitudes
7
and ridesharing behavior.” Transportation Research Part F: Traffic Psychology and Behaviour 62:
8
258-273. doi: 10.1016/j.trf.2019.01.003.
9
Washington, Simon. P., Matthew. G. Karlaftis, and Fred Mannering. 2012. Statistical and
10
Econometric Methods for Transportation Data Analysis. Boca Raton, FL: CRC Press.
11
Yan, Xiang, Xinyu Liu, and Xilei Zhao. 2020. “Using machine learning for direct demand
12
modeling of ridesourcing services in Chicago.” Journal of Transport Geography 83: 102661. doi:
13
10.1016/j.jtrangeo.2020.102661.
14
Yang, Hongtai, Yuan Liang, and Linchuan Yang. 2021. “Equitable? Exploring ridesourcing
15
waiting time and its determinants.” Transportation Research Part D: Transport and Environment
16
93:102774. doi: 10.1016/j.trd.2021.102774.
17
Yang, Hongtai, Jinghai Huo, Renbin Pan, Kun Xie, Wenjia Zhang, and Xinggang Luo. 2022.
18
"Exploring built environment factors that influence the market share of ridesourcing service."
19
Applied Geography 142:102699. doi: 10.1016/j.apgeog.2022.102699.
20
Young, Mischa, and Steven Farber. 2019. “The who, why, and when of Uber and other ride-hailing
21
trips: An examination of a large sample household travel survey.” Transportation Research Part
22
A: Policy and Practice 119:383-92. doi:10.1016/j.tra.2018.11.018.
23
Yang, Hongtai, Guocong Zhai, Linchuan Yang, and Kun Xie. 2022. “How does the suspension of
24
ride-sourcing affect the transportation system and environment?” Transportation Research Part
25
D: Transport and Environment 102:103131. doi: 10.1016/j.trd.2021.103131.
26
Zha, Liteng, Yafeng Yin, and Zhengtian Xu. 2018. “Geometric matching and spatial pricing in
27
ride-sourcing markets.” Transportation Research Part C: Emerging Technologies 92: 58-75. doi:
28
10.1016/j.trc.2018.04.015.
29
Zhang, Dapeng, David José Ahouagi Vaz Magalhães, and Xiaokun Wang. 2014a. “Prioritizing
30
bicycle paths in Belo Horizonte City, Brazil: Analysis based on user preferences and willingness
31
considering individual heterogeneity.” Transportation Research Part A: Policy and Practice 67:
32
268-278. doi: 10.1016/j.tra.2014.07.010.
33
Zhang, Zheng, Hidemichi Fujii, and Shunsuke Managi. 2014b. “How does commuting behavior
34
change due to incentives? An empirical study of the Beijing Subway System.” Transportation
35
Research Part F: Traffic Psychology and Behaviour 24: 17-26. doi: 10.1016/j.trf.2014.02.00
36
Zoellick, Jan C., Adelheid Kuhlmey, Liane Schenk, Daniel Schindel, and Stefan Blüher. 2019.
37
“Amused, accepted, and used? Attitudes and emotions towards automated vehicles, their
38
relationships, and predictive value for usage intention.” Transportation Research Part F: Traffic
39
Psychology and Behaviour 65:68-78. doi: 10.1016/j.trf.2019.07.009.
40
