Research on the Mode Choice Intention of the Elderly for Autonomous Vehicles Based on the Extended Ecological Model

Abstract

Due to the elderly’s limited physical ability, their mode choice behavior with particular demand for the traffic system is significantly distinguished compared to young people. The emergence of Autonomous Vehicles (AVs) and Shared Autonomous Vehicles (SAVs) will allow the elderly to travel independently and offer more mode choices. However, emerging vehicles will continue to coexist with other traditional modes such as public transport. This paper aims to explore the internal mechanism of the elderly’s choice behavior among public transport, AVs, and SAVs. We integrated the relevant factors by expanding the ecological model and used the Multiple Indicators and Multiple Causes (MIMIC) model to analyze the constructs’ relationship. The results show that the elderly believe that public transport, AVs, and SAVs are useful and convenient travel modes for themselves, affecting intention significantly. In addition, the elderly’s well-being and social influence during travel are also significant constructs for their behavioral intention. The research could provide academic supports for the traffic management departments when making relevant policies and measures for the elderly.

Full text

sustainability
Article
Research on the Mode Choice Intention of the Elderly
for Autonomous Vehicles Based on the Extended
Ecological Model
Huiqian Sun, Peng Jing * , Mengxuan Zhao, Yuexia Chen, Fengping Zhan and Yuji Shi
School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang 212013, China;
sun_huiqian@yeah.net (H.S.); zhaomengxuango@163.com (M.Z.); cyx08@126.com (Y.C.); fpzhan@126.com (F.Z.);
yujishi@ujs.edu.cn (Y.S.)
*Correspondence: jingpeng@ujs.edu.cn
Received: 16 November 2020; Accepted: 17 December 2020; Published: 20 December 2020


Abstract:
Due to the elderly’s limited physical ability, their mode choice behavior with particular
demand for the traffic system is significantly distinguished compared to young people. The emergence
of Autonomous Vehicles (AVs) and Shared Autonomous Vehicles (SAVs) will allow the elderly to
travel independently and offer more mode choices. However, emerging vehicles will continue to
coexist with other traditional modes such as public transport. This paper aims to explore the internal
mechanism of the elderly’s choice behavior among public transport, AVs, and SAVs. We integrated
the relevant factors by expanding the ecological model and used the Multiple Indicators and Multiple
Causes (MIMIC) model to analyze the constructs’ relationship. The results show that the elderly
believe that public transport, AVs, and SAVs are useful and convenient travel modes for themselves,
affecting intention significantly. In addition, the elderly’s well-being and social influence during travel
are also significant constructs for their behavioral intention. The research could provide academic
supports for the traffic management departments when making relevant policies and measures for
the elderly.
Keywords: elderly; travel mode; choice behavior; extended ecological model; autonomous vehicles
1. Introduction
China is prospecting a larger proportion of elderly people in the total population, and this trend is
expected to increase in coming years, which has brought a challenge to society. The travel problem
related to the quality of life for the elderly has especially attracted researchers [
1
]. Traffic travel
behavior is the fundamental guarantee for the elderly to maintain their daily activities. The physical
condition of the elderly and the current traffic modes have become obstacles to their travel activities.
The regulations concerning drivers stipulated that older people over 70 years cannot apply for a license
before 2020. Some elderly people report problems such as “squeeze, unstable, and not to come” on the
bus, which weakens their mobility. Mobility is closely related to the physical condition and quality of
life of the elderly [2,3] and promotes society’s overall development [4].
Hu et al. found that an average of 80% of the elderly in China prefer to use walk and public
transport, which would increase more than 90% with age, while only less than 3% choose driving [
5
].
On the contrary, older travelers are almost 90% car-oriented, with low public transport usage levels,
less than 2% in North America [
6
]. That indicates there are enormous different travel patterns for the
elderly between China and North America. Therefore, we excluded the mode of driving in this study.
However, with Chinese economic development and increased individual income, more people are
accustomed to driving. In 2020, the Chinese elderly, more than 70 years old, could still apply for their
Sustainability 2020,12, 10661; doi:10.3390/su122410661 www.mdpi.com/journal/sustainability

Sustainability 2020,12, 10661 2 of 22
license if they have a good physical condition. We could anticipate that the percent of the elderly
choosing to drive a car will increase shortly in China. Therefore, the exclusion of driving from the
elderly’s travel mode choice set is the limitation of this study. We will introduce the driving option in
our future research as a benchmark.
The effect of Autonomous Vehicles (AVs, we listed all the Glossary) on the transport system, such as
travel behavior, traffic safety, and congestion, has been much discussed in the past half-decade [
7
].
Harper et al. proposed that the emergence of AVs and Shared Autonomous Vehicles (SAVs) will
provide more choices for the elderly, enhance their travel convenience and mobility [
8
]. AVs will
replace traditional cars and become an emerging travel choice, bringing the ultimate experience of
comfort, safety, and convenience to travelers in the future. AVs are faster and can improve road capacity
and alleviate urban traffic congestion; they are more environmentally-friendly and can improve fuel
efficiency and reduce exhaust pollution [
9
]. At the same time, Nuzzolo et al. found AVs could provide
an opportunity to reduce air pollution in the cities’ central areas [
10
]. Tan et al. also demonstrated the
driverless cars’ potential implications on sustainable tourism [
11
]. Liu et al. reported that the public
would intend to accept AVs’ risk when AVs are more environmentally-friendly [
12
]. The integration of
AVs and “shared travel” will be realized with the development of autonomous driving technology and
the sharing economy. Experts predict that SAVs will also become an emerging travel mode, which will
replace traditional taxis and net cars in the future [13].
However, whether AVs or SAVs may take a long time from the current development stage to
mature promotion and full popularity stage. Technology developments should be accompanied
by researching individuals’ habits and consumption motivations in the population [
14
]. When the
technology becomes mature, the implementation of legislation and social mentality adjustment should
be solved to enable travelers to enjoy safe, comfortable, and free travel on their own [
15
]. For the
elderly, the emergence of AVs and SAVs is indeed a boon to meeting the need for independent travel,
but the ability to accept and adapt to new things is still unknown, the long-term use of buses will
not change immediately. Levin and Stephen proposed that AVs and SAVs will coexist with public
transport for a long time in the future transportation system [
16
]. To our best knowledge, there are few
literatures concerning AVs’ impact on the elderly’s mode choice behavior, which is essential for the
elderly’s later life quality in the future. Therefore, it is meaningful to investigate the elderly’s mode
choice between public transport, AVs, and SAVs.
In recent years, the ecological model has been gradually applied to the elderly’s travel behavior.
Hough, Cao, and Handy proved the ecological model’s applicability in the elderly’s travel behavior [
17
].
Mifsud, Attard, and Ison found that three types of independent variables in the ecological model:
personal factors, social factors, and environmental factors affect older people’s travel behaviors [
18
].
Several articles indicated that the ecological model has the explanation strength in the elderly’s travel
behaviors. However, researchers have focused on the impact of observable variables without further
considering their psychological state. In particular, observable variables may not be sufficient to predict
future older populations’ travel mode choice behavior, given their escalating physical and mental level
demands in modern society. Therefore, one of the major gaps in the elderly’s travel behaviors is the lack
of psychological variables in applying the ecological model. In this study, an extended ecological model
involving psychological variables and travel attributes was proposed to comprehensively analyze the
elderly’s mode choice behavior.
Furthermore, elders’ travel modes vary due to the differences between individual countries
regarding social systems and cultural backgrounds. To our knowledge, significant development
and experiments of the ecological model about the elderly’s travel behaviors have been obtained in
developed countries, including travel behavior of older women in the USA [
13
], public transport
use among older adults in Australia [
19
], and the elderly’s mode choice between cars and buses in
Malta [
18
]. However, the study of the elderly’s travel behavior through ecological modeis rarely
conducted, leading to another gap in this field.

Sustainability 2020,12, 10661 3 of 22
The majority of current research into the elderly’s travel behavior was realized by examining
travel mode choices related to buses, walking, cycling, and cars. As a forthcoming advanced transport,
autonomous driving technology would provide great convenience to the elderly or disabled [
8
].
The authors also proposed that AVs and SAVs could be used as an alternative to traditional cars,
and eventually become one of the most dominating mode choices among older adults. Surprisingly,
AVs and SAVs were very few in the elderly travel mode choices domain.
Overall, this study attended to supplement research on elderly travel behavior. It may be the first
to take the Chinese elderly as the research object and put forward an extended ecological model to
explore their mode choices among public transport, AVs, and SAVs.
The objective of this study includes: (1) exploring the adaptability and application of the extended
ecological model in the elderly’s mode choice behavior in the future; (2) making a deeper understanding
of the elderly’s attitude towards the AVs and SAVs, and discovering the mechanism concerning the
relationship among constructs affecting travel mode choice behavior; (3) proposing suggestions
for manufacturers to develop AVs and SAVs to serve for the elderly’s travel need and providing
support for the government to introduce policies and measures to promote the application of AVs
and SAVs in the market. We found that the elderly believe that public transport, AVs, and SAVs
are useful and convenient travel modes for themselves, affecting intention significantly. In addition,
the elderly’s well-being and social influence during travel are also significant constructs for their mode
choice intention.
The remainder of this paper proceeds as follows. Section 2reviews the background; Section 3
illustrates the extended ecological model; Section 4presents the data survey and models, and Section 5
describes the results and discussions. Finally, Section 6reveals the limitation and conclusion.
2. Background
This study aims to analyze the mode choice behaviors of the elderly towards public transport,
AVs, and SAVs, which are based on the elderly’s travel needs and the extended ecological model.
Before the analysis, this paper reviewed the research of predecessors.
2.1. Travel Mode Choice of the Elderly According to Autonomous Vehicle
With the development of transportation, travel modes are becoming more and more diversified in
recent years. Although the elderly appear to be more mobile than earlier generations, their mobility
levels lag behind young people [
20
]. Smeltzer mentioned that the elderly’s travel demand goes
unsatisfied under an aging population’s gradually grave background [
21
]. Several studies focused on
the elderly’s travel demand and tried to improve their travel conditions [
17
,
18
,
22
,
23
]. We listed the
related research in Table 1. The mode choice set of these studies include bus, private car, walking, etc.
Table 1. Summary of studies on the travel behavior of the elderly.
Authors (Year) Theoretical Model Travel Mode Explanatory Variables
Hough, Cao, and Handy 2008 [
17
]
The ecological model Private car
Individual factors, Social environment,
Physical environment
Schmoecker et al. 2008 [23] - Bus, Private car Personal travel characteristic, Family
travel characteristic, Travel attribute
Amen 2014 [24] - Bus, Private car
Personal characteristic, Family
characteristic, Weather conditions,
Built environment
Rahman et al. 2016 [25] - Private car Socioeconomic variables
Liu et al. 2016 [26] - Bus, Walking Personal characteristic,
Family characteristic
Mifsud, Attard, and Ison 2017 [
18
]
The ecological model Bus, Private car
Individual factors, Social environment,
Physical environment
Payyanadan and Lee 2018 [22] - Sharing car Socioeconomic variables,
Travel characteristic
Note: The sign “-” means no theoretical model is used in the literature.

Sustainability 2020,12, 10661 4 of 22
AVs technology has made rapid advances in recent years, which are likely to change the future of
mobility [
27
]. Harper et al. suggested that AVs technology may provide an excellent convenience for
the elderly who cannot drive due to poor physical function [
8
]. Hence, AVs may become one of the
elderly’s travel modes in the future. However, most of the studies considering the AVs travel mode
choice mainly focus on commuters (see Table 2). There is little literature that focuses on the elderly
mode choice among AVs or SAVs. It may be the initial attempt to predicate the impact of AVs and
SAVs’ emergence on the elderly travel behavior intention.
Table 2. The research for travel mode choice behavior of autonomous vehicles.
Authors (Year) Subject Travel Mode Explanatory Variables
Levin and Boyles 2015 [16] commuter Public transport,
AVs, SAVs Travel attribute
Lamondia et al. 2016 [27] traveler
Bus, Private car, AVs, Bus
Socioeconomic variables,
Travel attribute
Yap, Correia,
and van Arem 2016 [28]commuter Public transport, Cycle,
Private car, AVs
Socioeconomic variables,
Latent attitude variables
J. Liu et al. 2017 [29] urban resident Private car, SAVs Socioeconomic variables,
Travel attribute
Haboucha, Ishaq,
and Shiftan 2017 [30]commuter Bus, AVs, SAVs
Socioeconomic variables,
Travel attribute,
Psychological variables
Moreno et al. 2018 [13] traveler AVs, SAVs
Socioeconomic variables,
Travel attribute,
Psychological variables
Nazari, Noruzoliaee,
and Mohammadian 2018 [31]traveler Autonomous taxi,
AVs, SAVs
Socioeconomic variables,
Travel attribute,
Psychological variables,
environmental variables
Pakusch et al. 2018 [32] traveler
Traditional car,
Traditional car-sharing,
AVs, SAVs, Bus
Socioeconomic variables,
Travel attribute,
Psychological variables
Menon et al. 2019 [15]
Private car owner
Traditional car, SAVs Socioeconomic variables,
Psychological variables
Pettigrew, Dana,
and Norman 2019 [33]traveler AVs, SAVs
Socioeconomic variables,
Travel attribute,
Psychological variables
Acheampong and
Cugurullo 2019 [34]traveler Bus, AVs, SAVs Socioeconomic variables,
Psychological variables
2.2. Application of the Ecological Model in Elderly Travel Mode Choice
Since Hough introduced the ecological model into older adults’ travel behavior [
17
]. Truong and
Somenahalli explored the elderly’s use of public transport through the ecological model and verified
its strong explanatory power [
19
]. Mifsud, Attard, and Ison used the ecological model to analyze older
people’s travel mode choice in Malta, indicating that the ecological model affects the elderly’s travel
behaviors through three independent variables: personal factors, social factors, and environmental
factors [
18
]. Considering specific cultural and social contexts may differ in Chinese elderly travel
patterns with abroad counterparts [
20
], there is an urgent need to construct a theoretical framework
to understand elderly travel behavior in China, improving their life quality [
35
]. Therefore, we first
attempt to introduce the ecological model into the Chinese elderly’s travel mode choice.
The ecological model takes less account of psychological factors and travel attributes [
17
].
Xia believed it is necessary to focus on older adults’ travel attributes and psychology to understand

Sustainability 2020,12, 10661 5 of 22
their mode choice behavior [
36
]. On the one hand, exploring travel attribute variables could provide a
clearer understanding of the elderly’s mode choice preferences [1]. On the other hand, psychological
factors have an essential role in predicting and understanding the elderly’s travel behavior [
1
,
37
].
Therefore, this research might first extend the ecological model by the variables related to travel
attributes and psychology.
3. Extended Ecological Model
3.1. Extended Travel Attribute Variable
Travel attributes refer to travel modes’ inherent characteristics, including quantitative (travel cost,
travel time) and qualitative factors (comfort, reliability, timeliness). Travel time has always been a
crucial variable in measurement. Truong and Somenahalli reported that too long travel time would
bring fatigue to the elderly, reduce the initiative of the elderly, and affect their quality of life [
19
].
However, the length of travel time often contradicts the travel cost. Rahman et al. showed that public
transport’s preferential policies would attract some elderly people with lower retirement wages [
25
].
Therefore, travel cost could determine the elderly’s Behavioral Intention (BI).
When traffic congestion becomes a severe social problem, travel time variability would greatly
influence the private car’s choice behavior. Due to the same reason, public transport’s travel time
reliability also needs to be improved, although there are bus priority lanes. Krueger, Rashidi, and Rose
have pointed out that AVs’ emergence will alleviate traffic congestion because artificial intelligence
technology can keep AVs a safe distance from the front car at a steady speed [
38
]. However, it is
unknown whether the travel time’s variability on the way can be eliminated in the future, so we
introduced the variable of travel variability time. What is more, Olawole and Aloba proposed that
travel comfort plays a vital role in the elderly’s mode choice behavior and travel mobility [
4
]. Therefore,
travel comfort is also used in our study.
To sum up, we select four travel attributes to analyze the elderly’s mode choice behavior: travel
time, travel variability time, travel cost, and travel comfort. There will be mutual constraints among
the four variables for AVs, SAVs, and public transport, which depend on the actual demand and
self-choice preference of the elderly. Therefore, it is of practical significance to study travel attributes”
influence on choice intention in improving the service quality of travel mode itself and meeting the
travel demand of the elderly.
3.2. Theoretical Framework
We extend the original ecological model by introducing PU, PE, AT, SWB, and SI. Additionally,
the four travel attributes, including travel time, travel variability time, travel cost, and travel comfort,
as shown in Figure 1.
Sustainability 2020, 12, x FOR PEER REVIEW 6 of 23
Figure 1. The theoretical framework of mode choice behavior for the elderly.
4. Methodology
4.1. Questionnaire Survey and Implementation
We use a survey questionnaire to obtain empirical data. The questionnaire contains three parts.
The first one is the ecological model's original constructs: personal factors, social factors, and
environmental factors. The second part includes the expanded constructs: PU, PE, AT, SWB, and SI.
The third part is the extended travel attributes: travel time, travel variability time, travel cost, and
travel comfort. In this study, we use three question items to measure the extended latent variable for
the elderly, using the Likert five-scale. We use the orthogonal design method to set three effective
selection scenarios for each mode. The respondents need to choose among the three options for each
scenario. We presented the question items concerning constructs, as shown in Table 3.
Table 3. Sources of constructs and items used in the research.
Constructs Item Source
Perceived
Usefulness
(PU)
I expect that the following travel modes are useful
for my life in the future.
I expect that the following travel modes are helpful
for me in the future.
I expect that the following travel modes are
functional for me in the future.
(Davis et al., 1989) [39]
Perceived Ease
of use
(PE)
I think the following travel modes are easy for me in
the future.
(Davis et al., 1989) [39]
I think the operation of the following travel modes is
easy for me in the future.
I think it is easy for me to use the facilities and
services of the following travel modes.
Attitude
(AT)
I prefer to using the following travel modes in the
future.
I think it is a good idea to use the following travel
modes in the future.
I support the use of the following travel modes in the
future.
(Davis et al., 1989; Zhang et al.,
2019) [39,40]
Behavioral
Intention
(BI)
I will intend to use following travel modes in the
future.
I will use the following travel modes in the future.
(Panagiotopoulos and
Dimitrakopoulos, 2018) [41]
Travel time
Travel
variability time
Travel cost
AT
SN
PT、AV 、SAV
Extended constructs
（Latent variables）
SWB
PC
Travel comfort
Individual
factors
Social
environment
Physical
environment
Ecological model
(Observable variable)
Extended travel
attribute variable
(Observable variable)
Figure 1. The theoretical framework of mode choice behavior for the elderly.

Sustainability 2020,12, 10661 6 of 22
4. Methodology
4.1. Questionnaire Survey and Implementation
We use a survey questionnaire to obtain empirical data. The questionnaire contains three
parts. The first one is the ecological model’s original constructs: personal factors, social factors,
and environmental factors. The second part includes the expanded constructs: PU, PE, AT, SWB,
and SI. The third part is the extended travel attributes: travel time, travel variability time, travel cost,
and travel comfort. In this study, we use three question items to measure the extended latent variable
for the elderly, using the Likert five-scale. We use the orthogonal design method to set three effective
selection scenarios for each mode. The respondents need to choose among the three options for each
scenario. We presented the question items concerning constructs, as shown in Table 3.
Table 3. Sources of constructs and items used in the research.
Constructs Item Source
Perceived Usefulness (PU)
I expect that the following travel modes
are useful for my life in the future.
(Davis et al., 1989) [39]
I expect that the following travel modes
are helpful for me in the future.
I expect that the following travel modes
are functional for me in the future.
Perceived Ease of use (PE)
I think the following travel modes are
easy for me in the future.
(Davis et al., 1989) [39]
I think the operation of the following
travel modes is easy for me in the future.
I think it is easy for me to use the
facilities and services of the following
travel modes.
Attitude (AT)
I prefer to using the following travel
modes in the future.
(Davis et al., 1989; Zhang et al., 2019) [39,40]
I think it is a good idea to use the
following travel modes in the future.
I support the use of the following travel
modes in the future.
Behavioral Intention (BI)
I will intend to use following travel
modes in the future.
(Panagiotopoulos and Dimitrakopoulos, 2018) [
41
]
I will use the following travel modes
in the future.
I will recommend my relatives and
friends to use the following travel
modes in the future.
Subjective Well-Being (SWB)
I believe the following travel modes are
close to my ideal in the future.
Bergstad et al. (2011) [42]
I am satisfied with the following travel
modes in the future.
When I choose one of the following
travel modes to use in the future,
I believe I have the best one.
We collected data in the urban area of Suzhou, Jiangsu Province, China, from July 4 to July 8,
in 2018. Suzhou lies in the south of Jiangsu Province, east of China, with an area of 8657 km
2
, and a
population of 10.72 million. A pre-investigation was conducted among 23 older people (more than
60 years old) randomly before a formal survey. We corrected the wording and sorting errors of
the items and re-edited the questionnaire according to their comments and feedback. We randomly
sampled and distributed questionnaires in the neighborhood, small squares, and other crowded
places. The participants were asked to complete the questionnaire if they gave consent to accept.
If the participants had problems during the survey, the investigators would use objective words to
describe and explain the questionnaire. A total of 600 questionnaires were distributed, and 524 were
collected. The recovery rate was 87.33%. The questionnaires with apparent incomplete and wrong
data (younger than 60 years old) were excluded. A total of 438 participants constituted our final

Sustainability 2020,12, 10661 7 of 22
sample; the effective rate was 83.59%. We believe that the response rate and efficiency of this survey
are relatively satisfactory.
4.2. The Models
4.2.1. Multiple Indicators and Multiple Causes (MIMIC) Model
Multiple indicators and multiple causes (MIMIC) model is a simplified structural equation model,
which can be used to analyze the relationship among latent variables, and between observed and latent
variables [
43
]. The MIMIC model can deal with multiple latent variables and endogenous indicators
simultaneously, and allow the existence of measurement errors, so its theoretical framework is more
flexible than other indirect measurement methods.
4.2.2. A Generalized RRM Model (G-RRM)
We applied Biogeme to estimate the generalized random regret minimization model (G-RRM)
and analyze travel attributes’ sensitivity to the elderly’s mode choice behavior. We also compared the
goodness of fit with the RUM model.
The random regret minimization model (RRM) is based on the regret theory to minimize the regret
of travelers’ mode choice [
44
]. Loomes and Sugden found that RRM is a decision-maker to compare
the selected scheme with other options and to minimize the expected regret, that is, to evade the
psychology [
45
]. If another alternative has higher utility, the decision-makers will regret, and vice versa.
Chorus investigated that the G-RRM model is more suitable for travelers’ decision-making behavior
because the model considers that when decision-makers choose a scheme that minimizes their regrets,
some unobservable variables affect the regret value [
46
]. The elderly have bounded rationality when
making travel decisions. G-RRM model could capture the regret avoidance psychology of the elderly
more realistically and accurately. Therefore, we choose the G-RRM model to analyze travel attributes.
5. Results and Discussions
5.1. Extended Psychological Latent Variable
5.1.1. Perceived Characteristic and Attitudes
Davis proposed that the Technology Acceptance Model (TAM) ‘s core concept is that users’
acceptance of emerging information technology will affect users’ intentions [
39
]. Figure 1shows the
framework of TAM. Perceived Usefulness (PU) and Perceived Ease of use (PE) in TAM influence users’
intentions for emerging information technologies. Ma et al. found that Perceived Characteristics (PC)
could represent PU and PE for emerging travel modes [
47
]. Once users have specific knowledge of
emerging information technology through external variables, they will have a particular perception of
their characteristics. PU and PE act on the users’ Attitude (AT), which could affect Behavioral Intention
(BI), and actual behavior occurs, as shown in Figure 2.
Sustainability 2020, 12, x FOR PEER REVIEW 8 of 23
Davis proposed that the Technology Acceptance Model (TAM) 's core concept is that users'
acceptance of emerging information technology will affect users' intentions [39]. Figure 1 shows the
framework of TAM. Perceived Usefulness (PU) and Perceived Ease of use (PE) in TAM influence
users' intentions for emerging information technologies. Ma et al. found that Perceived
Characteristics (PC) could represent PU and PE for emerging travel modes [47]. Once users have
specific knowledge of emerging information technology through external variables, they will have a
particular perception of their characteristics. PU and PE act on the users' Attitude (AT), which could
affect Behavioral Intention (BI), and actual behavior occurs, as shown in Figure 2.
Figure 2. Technology Acceptance Model Framework.
In summary, the PC measurement shows high predictive power in applying new technologies,
so the elderly's characteristics and ease of use affect their attitude toward travel modes when AVs are
emerging in the traffic system.
5.1.2. Subjective Well-Being
Happiness refers to individuals' subjective emotional feelings in the travel process, namely
Subjective Well-Being (SWB). The elderly's SWB is significant in studying individual behavior and
producing positive feedback on social development and health benefits [37]. Little empirical research
concerns the elderly's travel well-being in China, which may cause the policy-makers' support for the
elderly's quality of life is not enough.
5.1.3. Social network
Watts proposed the famous "small world model" in nature, claiming the individual is not only a
single existence but a member of a social group [48]. The social network (SN) between people can be
seen as a small world model, which indicates that interactions between people can affect their social
behavior, as shown in Figure 3. Montazemi and Conrath pointed out that a social network can
increase or decrease consumer trust in new technologies and affect usage intentions [49]. Kim,
Rasouli, and Timmermans used SN to explore the impact of travel behavior for SAVs through Social
Influence (SI) [50].
Figure 3. Schematic diagram of the social network.
In the process of socialization, the elderly will inevitably absorb or reject some of the other's
views. Therefore, from the perspective of SN, we extend the ecological model using SI.
5.1.4. Hypothesis
External
variables
PU
PE
AT BI Actual
behavior
Figure 2. Technology Acceptance Model Framework.

Sustainability 2020,12, 10661 8 of 22
In summary, the PC measurement shows high predictive power in applying new technologies,
so the elderly’s characteristics and ease of use affect their attitude toward travel modes when AVs are
emerging in the traffic system.
5.1.2. Subjective Well-Being
Happiness refers to individuals’ subjective emotional feelings in the travel process, namely
Subjective Well-Being (SWB). The elderly’s SWB is significant in studying individual behavior and
producing positive feedback on social development and health benefits [
37
]. Little empirical research
concerns the elderly’s travel well-being in China, which may cause the policy-makers’ support for the
elderly’s quality of life is not enough.
5.1.3. Social Network
Watts proposed the famous “small world model” in nature, claiming the individual is not only a
single existence but a member of a social group [
48
]. The social network (SN) between people can
be seen as a small world model, which indicates that interactions between people can affect their
social behavior, as shown in Figure 3. Montazemi and Conrath pointed out that a social network
can increase or decrease consumer trust in new technologies and affect usage intentions [
49
]. Kim,
Rasouli, and Timmermans used SN to explore the impact of travel behavior for SAVs through Social
Influence (SI) [50].
Sustainability 2020, 12, x FOR PEER REVIEW 8 of 23
Davis proposed that the Technology Acceptance Model (TAM) 's core concept is that users'
acceptance of emerging information technology will affect users' intentions [39]. Figure 1 shows the
framework of TAM. Perceived Usefulness (PU) and Perceived Ease of use (PE) in TAM influence
users' intentions for emerging information technologies. Ma et al. found that Perceived
Characteristics (PC) could represent PU and PE for emerging travel modes [47]. Once users have
specific knowledge of emerging information technology through external variables, they will have a
particular perception of their characteristics. PU and PE act on the users' Attitude (AT), which could
affect Behavioral Intention (BI), and actual behavior occurs, as shown in Figure 2.
Figure 2. Technology Acceptance Model Framework.
In summary, the PC measurement shows high predictive power in applying new technologies,
so the elderly's characteristics and ease of use affect their attitude toward travel modes when AVs are
emerging in the traffic system.
5.1.2. Subjective Well-Being
Happiness refers to individuals' subjective emotional feelings in the travel process, namely
Subjective Well-Being (SWB). The elderly's SWB is significant in studying individual behavior and
producing positive feedback on social development and health benefits [37]. Little empirical research
concerns the elderly's travel well-being in China, which may cause the policy-makers' support for the
elderly's quality of life is not enough.
5.1.3. Social network
Watts proposed the famous "small world model" in nature, claiming the individual is not only a
single existence but a member of a social group [48]. The social network (SN) between people can be
seen as a small world model, which indicates that interactions between people can affect their social
behavior, as shown in Figure 3. Montazemi and Conrath pointed out that a social network can
increase or decrease consumer trust in new technologies and affect usage intentions [49]. Kim,
Rasouli, and Timmermans used SN to explore the impact of travel behavior for SAVs through Social
Influence (SI) [50].
Figure 3. Schematic diagram of the social network.
In the process of socialization, the elderly will inevitably absorb or reject some of the other's
views. Therefore, from the perspective of SN, we extend the ecological model using SI.
5.1.4. Hypothesis
External
variables
PU
PE
AT BI Actual
behavior
Figure 3. Schematic diagram of the social network.
In the process of socialization, the elderly will inevitably absorb or reject some of the other’s views.
Therefore, from the perspective of SN, we extend the ecological model using SI.
5.1.4. Hypothesis
The Relationship between PU, PE, and AT
An individual’s behavioral intention is the critical driver of actual behavior [
39
]. In this study,
behavioral intention means the elderly’s usage intentions towards travel modes.
In TAM, PE and PU are considered two crucial factors that directly impacted persons’ attitudes
to the new technology [
39
,
51
]. PU refers to the degree to which individuals believe that using a
particular system would enhance their performance, and PE is defined as the degree to which a person
believes that using a particular system would be free of effort [
39
]. Chen and Chao investigated the
travelers’ public transport choice behavior through PU and PE, which significantly affects the AT [
52
].
Especially focusing on the elderly, we assume that the more the elderly feel easy and useful about
travel modes, the more they have a positive attitude to use it in our research, therefore, hypothesis 1 is
proposed. In addition, Shang-Yu Chen discovered perceived green usefulness plays an intermediary
role between perceived green ease of use and AT for public bicycles [
53
]. While in the field of the

Sustainability 2020,12, 10661 9 of 22
elderly’s travel mode choice, especially for public transport, AVs and SAVs, few studies concentrate on
the relationships between PU and PE, which seems necessary. Hence, we proposed that PE has a direct
effect on PU and have hypothesis 2. Moreover, the elderly’s attitude toward travel modes reflects
their general evaluation and inclination toward this behavioral intention [
54
,
55
]. Thus, hypothesis 3
is formulated. Above all, to investigate the association of PU, PE and AT, the following hypotheses
are drawn:
Hypothesis 1 (H1). PU and PE have direct positive effects on AT for the elderly’s choice behavior.
Hypothesis 2 (H2). PE has a direct positive effect on PU for the elderly’s choice behavior.
Hypothesis 3 (H3). AT has a direct positive effect on BI for the elderly’s choice behavior.
The Role of SWB in the Extended Ecological Model
Subjective well-being, one of the most vital factors in this study, is defined as the elderly’s emotional
well-being during the travel mode choice process. The more the elderly experience happiness via their
engagement in travel by AVs and SAVs, the greater their motivation to make relevant decisions [
56
].
Therefore, we propose hypothesis 4 supporting that SWB has a positive influence on the elderly’s
mode choice intention. In addition, while making choices, AT and PE are positively correlated with
happiness and satisfaction [
31
,
57
]. It indicated that if the elderly have a positive attitude and feel
easy, they enjoy traveling more. Hence, hypothesis 5 is proposed. According to previous studies,
the following hypotheses may be correct:
Hypothesis 4 (H4). SWB has a significant positive correlation with BI for the elderly’s choice behavior.
Hypothesis 5 (H5).
AT and PE have significant positive correlations with SWB for the elderly’s choice behavior.
Social Influence (SI)
Social influence refers to the interactions between the elderly and other people, affecting the
elderly’s choice behavior. Kim et al. found that individuals tend to change to some extent with the
group’s general phenomenon [
50
]. When an individual realizes that other people think they should
choose, they will have the motivation to make relevant decisions [
58
]. Therefore, it could be inferred
that SI would positively impact the AT and BI toward the elderly’s modes choice behavior [
50
]. Thus,
we proposed hypothesis 6 and hypothesis 7:
Hypothesis 6 (H6). SI has a positive correlation with BI for the elderly’s choice behavior.
Hypothesis 7 (H7). SI has a positive correlation with AT for the elderly’s choice behavior.
5.2. Descriptive Statistical Analysis
As shown in Table 4, we use descriptive statistics on the empirical data collected from
the survey, including the ecological model’s original variables: personal factors, social factors,
and environmental factors.

Sustainability 2020,12, 10661 10 of 22
Table 4. Descriptive statistical analysis of sample characteristics.
Personal Particulars Group Sample Size Percentage (%)
Gender Male 225 51.37%
Female 213 48.63%
Age
60–69 years 287 65.53%
70–79 years 110 25.11%
80 years or above 41 9.36%
Education
Below primary 12 2.74%
Primary or junior 206 47.03%
Senior 153 34.93%
College or above 67 15.30%
Income
0 68 15.53%
<3000 124 28.31%
3000–6000 153 34.93%
>6000 93 21.23%
Physical health
Terrible 57 13.01%
Bad 106 24.20%
Average 150 34.25%
Well 82 18.72%
Excellent 43 9.82%
Mental health
Terrible 35 7.99%
Bad 52 11.87%
Average 90 20.55%
Well 174 39.73%
Excellent 87 19.86%
License Yes 239 54.57%
No 199 45.43%
Living status
Alone 42 9.59%
Couple living together 217 49.54%
Living alone with children 83 18.95%
Couples live together with children
96 21.92%
Rely on Yes 198 45.21%
No 240 54.79%
Participate in
social activities
Never 74 16.89%
Occasionally 195 44.52%
Frequently 169 38.59%
Station distance
<500 m 82 18.72%
500–1000 m 157 35.85%
1000–2000 m 101 23.06%
>200 m 98 22.37%
Frequency of using
public transport
Never 18 4.11%
Several times a year 69 15.75%
Several times a month 142 32.42%
Two or three times a week 107 24.43%
Everyday 102 23.29%
As seen in the table, the proportion of men and women in our sample is relatively uniform, and the
age is mostly from 60 to 69 years old, accounting for 65.53%. Nearly half of the elderly’s education
is in primary or lower secondary schools, and 34.93% of older persons graduate from high school.
One-third participants of the sample have monthly income from RMB 3000 to 6000. A total of 37.21% of
the elderly are not in good physical condition, and nearly half of the elderly’s mobility depends on their
families. We can also see that most elderly still like to participate in social activities; 44.52% occasionally
take part in social activities, and 38.59% often participate in social activities. According to the sample,
47.72% of the elderly use public transport every week, even every day. At present, buses are still an
essential mode for the elderly.

Sustainability 2020,12, 10661 11 of 22
5.3. Measurement Model
To examine the survey items’ scale reliability, we calculated Cornbrash’s
α
, as shown in Table 5.
The minimum value of Cronbach’s
α
is 0.760, and all values are higher than the recommended
minimum value of 0.70. We used the Average Variance Extracted (AVE) values and the construct’s
standardized factor loading to test the measurement model’s convergent validity. Convergence validity
reflects whether the items used to measure the same structure have internal reliability. The AVE of
constructs range from 0.514 to 0.781, and the standard factor loading values were between 0.696 and
0.934, meeting the recommended minimum value of 0.5 [
59
]. Therefore, the measurement model has
good performances in reliability and validity terms.
Table 5. Reliability and validity test results of constructs.
Constructs Items Standardized
Factor Loading Cronbach’s αAVE
Public transport
Perceived
usefulness (PU)
PT_PU1 0.835
0.864 0.679
PT_PU2 0.859
PT_PU3 0.795
Perceived ease of
use (PE)
PT_PE1 0.827
0.872 0.695
PT_PE2 0.812
PT_PE3 0.879
Attitude (AT)
PT_AT1 0.794
0.876 0.701
PT_AT2 0.841
PT_AT3 0.883
Subjective
well-being (SWB)
PT_SWB1 0.916
0.915 0.781
PT_SWB2 0.908
PT_SWB3 0.848
Social influence (SI)
PT_SI1 0.795
0.867 0.685
PT_SI2 0.886
PT_SI3 0.813
Behavioral
intention (BI)
PT_BI1 0.825
0.854 0.736
PT_BI2 0.872
PT_BI3 0.796
AVs
Perceived
usefulness (PU)
AV_PU1 0.742
0.821 0.604
AV_PU2 0.827
AV_PU3 0.775
Perceived ease of
use (PE)
AV_PE1 0.904
0.906 0.764
AV_PE2 0.836
AV_PE3 0.898
Attitude (AT)
AV_AT1 0.755
0.801 0.573
AV_AT2 0.794
AV_AT3 0.733
Subjective
well-being (SWB)
AV_SWB1 0.882
0.840 0.638
AV_SWB2 0.748
AV_SWB3 0.778
Social influence (SI)
AV_SI1 0.705
0.815 0.597
AV_SI2 0.856
AV_SI3 0.764
Behavioral
intention (BI)
AV_BI1 0.748
0.794 0.576
AV_BI2 0.762
AV_BI3 0.712

Sustainability 2020,12, 10661 12 of 22
Table 5. Cont.
Constructs Items Standardized
Factor Loading Cronbach’s αAVE
SAVs
Perceived
usefulness (PU)
SAV_PU1 0.696
0.760 0.514
SAV_PU2 0.746
SAV_PU3 0.725
Perceived ease of
use (PE)
SAV_PE1 0.802
0.878 0.707
SAV_PE2 0.893
SAV_PE3 0.834
Attitude (AT)
SAV_AT1 0.727
0.821 0.605
SAV_AT2 0.825
SAV_AT3 0.792
Subjective
well-being (SWB)
SAV_SWB1
0.934
0.912 0.776
SAV_SWB2
0.844
SAV_SWB3
0.876
Social influence (SI)
SAV_SI1 0.888
0.881 0.713
SAV_SI2 0.792
SAV_SI3 0.863
Behavioral
intention (BI)
SAV_BI1 0.863
0.854 0.685
SAV_BI2 0.847
SAV_BI3 0.803
5.4. Multiple Indicators and Multiple Causes (MIMIC) Model
5.4.1. Model Construction and Fitting Effect Analysis
Based on the seven hypothetical relationships proposed above, a path analysis diagram for the
decision-making model of the elderly’s mode choice behavior, as shown in Figure 4. The upper and
lower rows of rectangular boxes represent the personal factors, social factors, and travel attributes.
Ellipses represent latent variables, and rectangles indicate corresponding measurement variables of
each latent variable.
Sustainability 2020, 12, x FOR PEER REVIEW 12 of 23
Subjective
well-being
(SWB)
AV_SWB1
AV_SWB2
AV_SWB3
0.882
0.748
0.778
0.840 0.638
Social
influence (SI)
AV_SI1
AV_SI2
AV_SI3
0.705
0.856
0.764
0.815 0.597
Behavioral
intention
(BI)
AV_BI1
AV_BI2
AV_BI3
0.748
0.762
0.712
0.794 0.576
SAVs
Perceived
usefulness
(PU)
SAV_PU1
SAV_PU2
SAV_PU3
0.696
0.746
0.725
0.760 0.514
Perceived ease
of use (PE)
SAV_PE1
SAV_PE2
SAV_PE3
0.802
0.893
0.834
0.878 0.707
Attitude (AT)
SAV_AT1
SAV_AT2
SAV_AT3
0.727
0.825
0.792
0.821 0.605
Subjective
well-being
(SWB)
SAV_SWB1
SAV_SWB2
SAV_SWB3
0.934
0.844
0.876
0.912 0.776
Social
influence (SI)
SAV_SI1
SAV_SI2
SAV_SI3
0.888
0.792
0.863
0.881 0.713
Behavioral
intention
(BI)
SAV_BI1
SAV_BI2
SAV_BI3
0.863
0.847
0.803
0.854 0.685
5.4. Multiple Indicators and Multiple Causes (MIMIC) Model
5.4.1. Model Construction and Fitting Effect Analysis
Based on the seven hypothetical relationships proposed above, a path analysis diagram for the
decision-making model of the elderly's mode choice behavior, as shown in Figure 4. The upper and
lower rows of rectangular boxes represent the personal factors, social factors, and travel attributes.
Ellipses represent latent variables, and rectangles indicate corresponding measurement variables of
each latent variable.
PU
PE
AT
SI
SWB
BI
PU1
SWB2
PU2
PU3
PE3
PE2
PE1
SWB3SWB1
AT1 AT 1AT1
SI1 SI3SI2
BI3
BI2
BI1
Gender Age Edu Income Career Physical Health Mental Health License
Living Status Rely On Social Activity Station Distance Frequency
Figure 4. The path analysis of the MIMIC model.

Sustainability 2020,12, 10661 13 of 22
The evaluation of the model fitting effect is a prerequisite for the analysis of the results. If the
fitting index is not in a reasonable interval, the model needs to be modified. The fitting indicators of
the MIMIC model usually include the following: Chi-square degree of freedom (
χ2
/df ), Comparative
Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Squared Error of Approximation (RMSEA) and
Standardized Root Mean Squared Residual (SRMR). Of course, each indicator has a reasonable interval.
Table 6shows the fit of the structural equation model for the three modes.
Table 6. Evaluation of model fitting indicators.
Fitting Index χ2/df CFI TLI RMSEA SRMR Fitting
Effect
Reasonable range
<3, Good; <5, Acceptable
>0.90 >0.90
<0.05, Excellent; <0.08, Good
<0.08, Acceptable –
PT 3.473 0.912 0.921 0.057 0.041 √
AVs 3.241 0.917 0.918 0.053 0.044 √
SAVs 3.059 0.915 0.920 0.049 0.043 √
5.4.2. The Relationship between Observed and Latent Variables
In this paper, the effects of observed factors on latent variables are discussed according to three
modes, such as public transport, AVs, and SAVs. The model analysis results are shown in Tables 7and 8.
Table 7. Relationship between personal factors with latent variables.
Variables Gender Age Edu Income Career Physical
Health
Mental
Health License
PT
PU 0.035 −0.071 * 0.038 −0.013 0.113 0.108 *−0.009 −0.015 *
PE 0.014 0.179 1.372 0.004 0.004 −0.056 * 0.010 0.019
AT 0.006 −0.280 0.062 0.001 0.517 1.004 −0.064 * 0.431
SWB −0.013 −0.059 0.053 −0.226 0.226 0.182 ** −0.103 *−0.055 *
SI 0.042 0.061 −0.007 0.043 0.093 0.087 0.066 0.049
AVs
PU 0.011 0.049 0.194 0.162 ** 0.056 0.604 0.092 0.007
PE 0.013 *−0.017 ** 0.272 ** 0.099 0.113 0.075 −0.611 0.084 **
AT 0.247 0.101 0.205 0.041 0.039 *0.049 *** −0.055 * 1.052
SWB 0.406 0.308 0.177 0.118 * 0.001 0.125 ** 0.412 0.010
SI 0.088 −0.052 0.005 0.010 0.022 0.037 −0.043 0.002
SAVs
PU 0.042 * 0.507 0.065 *0.056 * 0.107 0.904 0.370 −0.226 *
PE 0.033 −0.026 0.047 * 0.787 0.004 0.220 0.011 0.003
AT 0.274 0.904 0.663 0.065 1.073 0.031 ** 0.558 0.060
SWB 0.180 0.085 * 0.330 0.539 0.006 −0.319 ** 0.036 −0.017
SI 1.005 −0.013 0.007 −0.044 0.809 0.542 0.031 0.029
Note: *: p<0.05; **: p<0.01; ***: p<0.001.
Table 8. Relationship between social factors and travel attributes with latent variables.
Variables Living Status Rely On Social Activity Station Distance Frequency
PT
PU 0.074 0.240 0.015 −0.032 ** 0.013 *
PE 0.009 −0.189 0.004 −0.107 * 1.057
AT 0.015 0.004 0.080 −0.040 0.379 *
SWB −0.063 0.011 0.331 −0.551 0.002
SI 0.550 −0.061 1.027 0.033 0.010

Sustainability 2020,12, 10661 14 of 22
Table 8. Cont.
Variables Living Status Rely On Social Activity Station Distance Frequency
AVs
PU 0.044 *0.028 ** 0.099 0.005* 0.018
PE 0.107 −0.937 0.012 0.046 −0.044
AT 0.028 0.011 * 0.225 0.119 0.006
SWB 0.003 0.050 0.100 0.260 0.057 **
SI 0.083 −0.228 0.147 ** 0.007 0.050
SAVs
PU −0.009 0.105 * 0.339 1.005 −0.032
PE 0.201 −0.016 0.060 0.049 0.030
AT 0.018 * 0.499 0.247 0.020 −0.011
SWB −0.056 0.070 0.023 0.177 0.409
SI 0.037 0.002 0.021 ** 0.204 0.032
Note: *: p<0.05; **: p<0.01.
Table 7shows the effects of personal factors on latent variables. In the MIMIC model of the
elderly’s travel patterns, not all observed variables have significant effects on latent variables. Overall,
the physical health of the elderly has a significant impact on most latent variables; mental health
has a significant effect on latent variables in models for public transport and AVs; age and license
are significantly related to individual latent factors for three modes. The three variables of gender,
education, and income have a conspicuous influence on latent variables in the MIMIC model of the
AVs and SAVs. It is worth noting that career only affects latent variables in the AVs model.
Table 8shows the influence of personal characteristics and social factors on the latent variables.
The following is the discussion of each observed variable.
(1) Social factors: The living status variable showed a significant positive correlation for perceived
usefulness and attitude in the MIMIC models of AVs and SAVs, respectively. Whether to rely on others
for daily travel has a significant correlation with perceived usefulness in the MIMIC models of AVs
and SAVs, indicating the diversification and convenience of elderly people who depend on others to
travel. Besides, this variable plays a positive role in the attitude of the AVs model. The frequency of
participation in social activities has a positive impact on the social impact in the MIMIC model of AVs
and SAVs, implying that older people often participate in social activities and are exposed to other
people’s perceptions of travel modes, which affect their mode choice pattern, especially for emerging
modes. The elderly will share with and influence each other before using emerging travel mode.
(2) Travel attributes: The variable of station distance negatively correlates with perceived
usefulness and perceived ease of use in the MIMIC model for public transport, which indicates that the
further the bus stop is from home, the less likely the elderly will use the bus. Station distance has a
positive correlation with perceived usefulness in AVs. When the bus station is far away, the AVs will
meet their travel needs and improve travel convenience. The variable of the frequency of using buses
has played a positive role in the perceived usefulness and attitude in the MIMIC model for public
transport, showing that the higher the frequency of using the bus, the more positive they are likely to
travel. It also has a significant positive impact on AVs’ SWB, which may be due to the fatigue of the
elderly who use buses often, and they may prefer a comfortable and free way of travel.
5.4.3. The Relationship between Latent Variables
We apply MIMIC models to analyze choice behavior among three travel modes: public transport,
the AVs, and SAVs. The path relations among the constructs in the MIMIC models for different
modes are shown in Figures 5–7. The value in parentheses is the z-value, and the
*
on the numerical
superscript indicates the degree of significance (
***
means p<0.001,
**
means p<0.01,
*
means p<0.05).
R
2
indicates the MIMIC coefficient’s judgment to predict the elderly’s intention to choose a specific

Sustainability 2020,12, 10661 15 of 22
mode. The goodness of fit based on model construction is good. We present the model estimation
results as follows.
Sustainability 2020, 12, x FOR PEER REVIEW 15 of 23
(1) In the MIMIC model, the variance of intention interpretation for the choice intention of the
public transport, AVs, and SAVs was 74, 71, and 75%, respectively, which indicates that the extended
ecological model has a relatively high explanatory power for the elderly.
Figure 5. Standardized path coefficients between latent variables for public transport. Note: *: p < 0.05;
**: p < 0.01; ***: p < 0.001.
Figure 6. Standardized path coefficients between latent variables for AVs. Note: *: p < 0.05; **: p < 0.01;
***: p < 0.001.
Figure 7. Standardized path coefficients between latent variables for SAVs. Note: *: p < 0.05; **: p <
0.01; ***: p < 0.001.
(2) Perceived usefulness has a significant positive impact on attitude, with standardized path
coefficients of 0.43 (p < 0.01), 0.32 (p < 0.01), and 0.09 (p < 0.05) for public transport, AVs, and SAVs,
respectively. Perceived ease of use also showed a significant positive impact on attitude, with
standardized path coefficients of 0.64 (p < 0.001), 0.53 (p < 0.01), and 0.44 (p < 0.01). The model results
indicate that the elderly are concerned with whether an emerging mode is easy to operate, which will
affect their attitude towards travel behavior, thus supporting hypothesis 1. The figures above show
PU
PE
AT
SI
SWB
BI
0.35
*
(5.27)
0.24
*
(2.09)
0.15
**
(3.04)
0.23
*
(3.17)
0.44
**
(7.10)
0.25(2.09)
0.46
*
(6.80)
0.43
**
(3.27)
0.64
***
(10.41)
R
2
=0.74
0.51
*
(1.57)
PU
PE
AT
SI
SWB
BI
0.39
* * *
(5.13)
0.47
***
(7.05)
0.45
*
(2.07)
0.17
*
(2.98)
0.51
**
(1.09)
0.25
*
(2.43)
0.26
*
(3.95)
0.32
**
(4.60)
0.53
**
(5.32)
R
2
=0.71
0.21 (1.57)
PU
PE
AT
SI
SWB
BI
0.54
* *
(7.05)
0.32
**
(3.69)
0.53
***
(10.27)
0.30
*
(2.47)
0.26
*
(3.95)
0.16
*
(1.50)
0.03(1.10)
0.09
*
(1.69)
0.44
**
(4.76)
R
2
=0.75
0.27
*
(1.44)
Figure 5.
Standardized path coefficients between latent variables for public transport. Note: *: p<0.05;
**: p<0.01; ***: p<0.001.
Sustainability 2020, 12, x FOR PEER REVIEW 15 of 23
(1) In the MIMIC model, the variance of intention interpretation for the choice intention of the
public transport, AVs, and SAVs was 74, 71, and 75%, respectively, which indicates that the extended
ecological model has a relatively high explanatory power for the elderly.
Figure 5. Standardized path coefficients between latent variables for public transport. Note: *: p < 0.05;
**: p < 0.01; ***: p < 0.001.
Figure 6. Standardized path coefficients between latent variables for AVs. Note: *: p < 0.05; **: p < 0.01;
***: p < 0.001.
Figure 7. Standardized path coefficients between latent variables for SAVs. Note: *: p < 0.05; **: p <
0.01; ***: p < 0.001.
(2) Perceived usefulness has a significant positive impact on attitude, with standardized path
coefficients of 0.43 (p < 0.01), 0.32 (p < 0.01), and 0.09 (p < 0.05) for public transport, AVs, and SAVs,
respectively. Perceived ease of use also showed a significant positive impact on attitude, with
standardized path coefficients of 0.64 (p < 0.001), 0.53 (p < 0.01), and 0.44 (p < 0.01). The model results
indicate that the elderly are concerned with whether an emerging mode is easy to operate, which will
affect their attitude towards travel behavior, thus supporting hypothesis 1. The figures above show
PU
PE
AT
SI
SWB
BI
0.35
*
(5.27)
0.24
*
(2.09)
0.15
**
(3.04)
0.23
*
(3.17)
0.44
**
(7.10)
0.25(2.09)
0.46
*
(6.80)
0.43
**
(3.27)
0.64
***
(10.41)
R
2
=0.74
0.51
*
(1.57)
PU
PE
AT
SI
SWB
BI
0.39
* * *
(5.13)
0.47
***
(7.05)
0.45
*
(2.07)
0.17
*
(2.98)
0.51
**
(1.09)
0.25
*
(2.43)
0.26
*
(3.95)
0.32
**
(4.60)
0.53
**
(5.32)
R
2
=0.71
0.21 (1.57)
PU
PE
AT
SI
SWB
BI
0.54
* *
(7.05)
0.32
**
(3.69)
0.53
***
(10.27)
0.30
*
(2.47)
0.26
*
(3.95)
0.16
*
(1.50)
0.03(1.10)
0.09
*
(1.69)
0.44
**
(4.76)
R
2
=0.75
0.27
*
(1.44)
Figure 6.
Standardized path coefficients between latent variables for AVs. Note: *: p<0.05; **: p<0.01;
***: p<0.001.
Sustainability 2020, 12, x FOR PEER REVIEW 15 of 23
(1) In the MIMIC model, the variance of intention interpretation for the choice intention of the
public transport, AVs, and SAVs was 74, 71, and 75%, respectively, which indicates that the extended
ecological model has a relatively high explanatory power for the elderly.
Figure 5. Standardized path coefficients between latent variables for public transport. Note: *: p < 0.05;
**: p < 0.01; ***: p < 0.001.
Figure 6. Standardized path coefficients between latent variables for AVs. Note: *: p < 0.05; **: p < 0.01;
***: p < 0.001.
Figure 7. Standardized path coefficients between latent variables for SAVs. Note: *: p < 0.05; **: p <
0.01; ***: p < 0.001.
(2) Perceived usefulness has a significant positive impact on attitude, with standardized path
coefficients of 0.43 (p < 0.01), 0.32 (p < 0.01), and 0.09 (p < 0.05) for public transport, AVs, and SAVs,
respectively. Perceived ease of use also showed a significant positive impact on attitude, with
standardized path coefficients of 0.64 (p < 0.001), 0.53 (p < 0.01), and 0.44 (p < 0.01). The model results
indicate that the elderly are concerned with whether an emerging mode is easy to operate, which will
affect their attitude towards travel behavior, thus supporting hypothesis 1. The figures above show
PU
PE
AT
SI
SWB
BI
0.35
*
(5.27)
0.24
*
(2.09)
0.15
**
(3.04)
0.23
*
(3.17)
0.44
**
(7.10)
0.25(2.09)
0.46
*
(6.80)
0.43
**
(3.27)
0.64
***
(10.41)
R
2
=0.74
0.51
*
(1.57)
PU
PE
AT
SI
SWB
BI
0.39
* * *
(5.13)
0.47
***
(7.05)
0.45
*
(2.07)
0.17
*
(2.98)
0.51
**
(1.09)
0.25
*
(2.43)
0.26
*
(3.95)
0.32
**
(4.60)
0.53
**
(5.32)
R
2
=0.71
0.21 (1.57)
PU
PE
AT
SI
SWB
BI
0.54
* *
(7.05)
0.32
**
(3.69)
0.53
***
(10.27)
0.30
*
(2.47)
0.26
*
(3.95)
0.16
*
(1.50)
0.03(1.10)
0.09
*
(1.69)
0.44
**
(4.76)
R
2
=0.75
0.27
*
(1.44)
Figure 7.
Standardized path coefficients between latent variables for SAVs. Note: *: p<0.05;
**: p<0.01
;
***: p<0.001.
(1) In the MIMIC model, the variance of intention interpretation for the choice intention of the
public transport, AVs, and SAVs was 74, 71, and 75%, respectively, which indicates that the extended
ecological model has a relatively high explanatory power for the elderly.
(2) Perceived usefulness has a significant positive impact on attitude, with standardized
path coefficients of 0.43 (p<0.01), 0.32 (p<0.01), and 0.09 (p<0.05) for public transport, AVs,
and SAVs, respectively. Perceived ease of use also showed a significant positive impact on attitude,
with standardized path coefficients of 0.64 (p<0.001), 0.53 (p<0.01), and 0.44 (p<0.01). The model
results indicate that the elderly are concerned with whether an emerging mode is easy to operate,
which will affect their attitude towards travel behavior, thus supporting hypothesis 1. The figures
above show that attitude positively influences behavioral intention, which is significant for AVs with
the standardized path coefficient 0.51 (p<0.01), so hypothesis 3 is proved.

Sustainability 2020,12, 10661 16 of 22
(3) We can see from Figures 5–7that SWB has a significant positive correlation with behavioral
intentions, the standardized path coefficient for the AVs is 0.47 (p<0.001), indicating that AVs can
make the elderly experience happiness when their travel demand satisfaction. Therefore, hypothesis 4
proposed above is verified. Attitude also plays a significant positive correlation with SWB, implying
that the subjective attitude of the elderly is closely related to their happiness in later life. Perceived ease
of use for SWB shows a significant positive correlation. In the model of the SAVs, the standardized
path coefficient is 0.53 (p<0.001), indicating that the ease of use of travel mode has an impact on the
mood of the elderly and indirectly affects the intention to use, thus supporting hypothesis 5.
(4) Social influence shows a positive correlation with behavioral intentions in the three models,
with path coefficients of 0.35 (p<0.05), 0.39 (p<0.001), and 0.54 (p<0.01), respectively. It presents that
the elderly’s travel behavior is affected by the social network and people’s behaviors and attitudes,
thus verifying hypothesis 6.
(5) As shown in Table 9, when we remove the perceived usefulness in all three structural equation
models, the perceived ease of use has significantly increased the normalized path coefficient and
z-value of behavioral intention. This shows that the existence of perceived usefulness weakens the
direct impact of perceived ease of use on behavioral intention, thus supporting hypothesis 2—PU plays
a mediating role between PE and AT in the elderly’s mode choice behavior. Similarly, in the absence
of attitudes, the role of social influence on behavioral intentions increases significantly, indicating
that attitudes share the direct effect of the social influence on the behavioral intentions. Therefore,
hypothesis 7 can be verified.
Table 9. Mediation variable.
Mediating Variable Public Transport AVs SAVs
The mediating effect of the PU between PE and AT
PU (included) 0.64 *** (10.41) 0.53 ** (5.32) 0.44 ** (4.76)
PU (excluded) 0.79 *** (16.94) 0.69 ** (11.05) 0.61 ** (8.23)
The mediating effect of the AT between SI and BI
AT (included) 0.35 * (5.27) 0.39 *** (5.13) 0.54 ** (7.05)
AT (excluded) 0.59 * (9.14) 0.62 *** (12.70) 0.73 ** (14.09)
Note: *: p<0.05; **: p<0.01; ***: p<0.001.
5.5. A Generalized RRM Model (G-RRM)
5.5.1. Parameter Calibration and Model Estimation
Before setting up the G-RRM model, we sorted out the data of travel attributes in the questionnaire.
The values of travel time, travel variability time, and travel costs were all taken the unit value according
to the scenario’s situation. The following rules apply to the value of the travel comfort: “0” indicates
the comfort seat of AVs and SAVs; “1” implies the “front seat” of the bus; “2” represents the “back seat”
of the bus; “3” labels the “station” of the bus, as shown in Table 10.
Table 10. Travel attribute variables.
Modes Travel Time Travel Variability Time Travel Cost Travel Comfort
Public Transport
30 8 1 1
45 14 2 2
60 18 3 3
AVs
10 2 20 0
20 5 25 0
34 8 35 0
SAVs
15 5 15 0
25 10 20 0
40 15 30 0

Sustainability 2020,12, 10661 17 of 22
The value of the travel attributes is encoded and replaced in the G-RRM model. The weights of
each travel attributes can be normalized to obtain the standardized coefficients, as Table 11 shows.
Table 11. Travel attributes weights for the elderly.
Indicators Travel Time Travel Variability Time Travel Cost Travel Comfort
Weights 3.59 3.51 3.75 3.82
Normalized coefficients 0.247 0.244 0.254 0.255
We could infer from Figure 8that the elderly attach the highest importance to travel comfort,
followed by travel cost, travel time, and travel variability time related to the elderly’s leisure time
after retirement.
Sustainability 2020, 12, x FOR PEER REVIEW 17 of 23
60 18 3 3
AVs
10 2 20 0
20 5 25 0
34 8 35 0
SAVs
15 5 15 0
25 10 20 0
40 15 30 0
The value of the travel attributes is encoded and replaced in the G-RRM model. The weights of
each travel attributes can be normalized to obtain the standardized coefficients, as Table 11 shows.
Table 11. Travel attributes weights for the elderly.
Indicators Travel Time Travel Variability Time Travel Cost Travel Comfort
Weights 3.59 3.51 3.75 3.82
Normalized coefficients 0.247 0.244 0.254 0.255
We could infer from Figure 8 that the elderly attach the highest importance to travel comfort,
followed by travel cost, travel time, and travel variability time related to the elderly's leisure time
after retirement.
Figure 8. Standardized weight values.
To better reflect the explanatory strength and applicability of the G-RRM model for the travel
attributes of the elderly, we also estimate the RUM model. Table 12 compares the results of the two
models.
In the questionnaire, the elderly need to choose one of three scenarios. The sample size was 438,
so the model's final sample size was 1314 (438×3=1314). The four travel attribute variables
significantly affect the elderly's choice behavior, implying that the model has high explanatory power
and credibility. LL(0) represents the logarithmic likelihood estimation value with parameter 0. LL(β)
represents the logarithmic likelihood estimate with parameter β. ρ
P
2
P
represents the model's fitting
degree. Both models' fitting effect is greater than 0.4, indicating that the model fitting effect is good.
When the parameter is β, the LL(β) of the G-RRM model is larger than that of the RUM model. The fit
of the G-RRM model is 0.425, while that of the RUM model is 0.417. Therefore, the G-RRM model
may be more suitable for analyzing travel attribute variables' influence on the elderly's travel
behavior.
Table 12. Parameter estimation of elderly travel attribute variables.
Variables
G-RRM RUM
Paramete
r
Estimation T-test Paramete
r
Estimation T-test
Figure 8. Standardized weight values.
To better reflect the explanatory strength and applicability of the G-RRM model for the travel
attributes of the elderly, we also estimate the RUM model. Table 12 compares the results of the
two models.
Table 12. Parameter estimation of elderly travel attribute variables.
Variables
G-RRM RUM
Parameter
Estimation t-Test Parameter
Estimation t-Test
Travel time −0.0194 −4.83 ** −0.0465 −4.92 **
Travel variability time −0.0325 −5.61 ** −0.0650 −5.64 **
Travel cost −0.0169 −2.37 ** −0.0329 −2.32 **
Travel comfort −0.0083 −1.96 ** −0.0167 −2.08 **
Samples 1314 1314
LL(β)−2,283,143 −2,314,908
LL(0) −3926.427 −3,970,683
ρ20.425 0.417
Note: **: p<0.01.
In the questionnaire, the elderly need to choose one of three scenarios. The sample size was
438, so the model’s final sample size was 1314 (438
×
3=1314). The four travel attribute variables
significantly affect the elderly’s choice behavior, implying that the model has high explanatory
power and credibility. LL(0) represents the logarithmic likelihood estimation value with parameter 0.
LL(
β
) represents the logarithmic likelihood estimate with parameter
β
.
ρ2
represents the model’s fitting
degree. Both models’ fitting effect is greater than 0.4, indicating that the model fitting effect is good.

Sustainability 2020,12, 10661 18 of 22
When the parameter is
β
, the LL(
β
) of the G-RRM model is larger than that of the RUM model. The fit
of the G-RRM model is 0.425, while that of the RUM model is 0.417. Therefore, the G-RRM model may
be more suitable for analyzing travel attribute variables’ influence on the elderly’s travel behavior.
As can be seen from Table 12, the maximum estimated coefficients are travel comfort, indicating
that the elderly are concerned about comfort during the travel. The emergence of AVs and SAVs can
provide a comfortable travel environment for the elderly and meet their travel needs. Travel cost has a
significantly negative effect on the elderly’s mode choice behavior, which means that the sensitivity for
the travel fare is relatively high. Travel time and travel variability time are also significantly correlated
with the travel behavior of the elderly because too long travel time will lead to the elderly’s fatigue,
thus reducing the possibility of choosing the mode. Since the travel process is dynamic, human and
road factors will make travel time changes. The arrival of AVs may bring about a qualitative change in
the accessibility of the entire transportation system.
5.5.2. Sensitivity Analysis
The sensitivity analysis is to obtain the corresponding change of the probability of the elderly
for a specific travel mode when the attribute changes and can also be viewed as the elasticity value
corresponding to the G-RRM model. Sensitivity analysis is usually carried out through direct elasticity
and indirect elasticity. Direct elasticity refers to the change of the choice probability of the selected
travel mode when a particular travel attribute changes by 1%. We analyze the estimated results through
direct elasticity and indirect elasticity, as shown in Table 13.
Table 13. Direct elasticity and indirect elasticity based on G-RRM model.
Selected
Modes
Direct Elasticity
Unselected
Modes
Indirect Elasticity
Travel
Time
Travel
Variability
Time
Travel
Cost
Travel
Comfort
Travel
Time
Travel
Variability
Time
Travel
Cost
Travel
Comfort
Public
transport −
0.201
−0.174 −0.421 0.927 AVs 0.090 0.089 0.210 −0.461
SAVs 0.097 0.082 0.217 −0.469
AVs −
0.217
−0.192 −0.475 0.948
Public
transport 0.102 0.091 0.224 −0.472
SAVs 0.113 0.098 0.239 −0.479
SAVs −
0.208
−0.187 −0.468 0.935 PT 0.102 0.087 0.224 −0.468
AVs 0.110 0.090 0.230 −0.462
Take the travel time as an example to understand the elastic value in the above table. When bus
travel time increases by 1%, travelers’ choice probability of bus decreases by 0.201%. When AVs and
SAVs’ travel time increase by 1%, respectively, the probability of choosing buses increases by 0.090%
and 0.097%, respectively. It can be seen that the elderly are the most sensitive to changes in travel
comfort, especially for the AVs. When the travel comfort increases by 1%, the probability of the elderly
choosing AVs will increase by 0.948%. However, when buses and SAVs’ travel comfort increase by 1%,
the probability of the elderly choosing AVs decreases by 0.472% and 0.479%, respectively. Similarly,
the sensitivity of the elderly to travel time and travel variability time is not very high.
6. Limitations and Conclusions
Some limitations need to be resolved in our future work. Firstly, we analyze the elderly’s mode
choice behavior among three modes: public transport, AVs, and SAVs. Although less than 3% of
the elderly in China drive [
5
], we could predict that the percentage of driving for the elderly would
increase sharply due to the elderly’ license with no age limitation and growth in the living standard.
Therefore, we will introduce driving into our research and explore the potential constructs affecting
the elderly’s driving behavior in the future study. Secondly, most participants did not ride AVs and
had no field experience with AVs’ benefits or drawbacks. The participants’ knowledge concerning

Sustainability 2020,12, 10661 19 of 22
AVs from the internet and other channels may mislead their perspectives or attitudes. In our future
research, we will use a video or AVs’ simulator to make the participants have a more real feeling
for AVs. Thirdly, the research sample comes from Suzhou survey data, and the result may not be
generalizable to the entire population. We may pay attention to the difference among various regions
for the elderly’s travel behavior.
Based on the elderly’s travel needs and the extended ecological model, we analyze the mode
choice behaviors of the elderly towards public transport, AVs, and SAVs in the future. The research
conclusions are as follows:
(1) We integrate the relevant factors that affect the travel of the elderly and expand the ecological
model by introducing the constructs, which provide a theoretical framework for the elderly. Moreover,
we analyze the modeling results of empirical data and verify the theoretical framework’s applicability
to the elderly’s travel behavior.
(2) The MIMIC model is used to analyze the relationship between the extended latent variables
and the original observed variables of the ecological model. We can see that the observed variables
have different degrees of influence on the constructs in the MIMIC model for public transport, AVs,
and SAVs. The interaction and internal mechanism of expanded constructs were analyzed with the
MIMIC model. Seven hypotheses proposed in this paper were proved to understand the elderly’s
travel psychology.
(3) Through the analysis and discussion of relevant influencing factors, we can have a deeper
understanding of the views and acceptability of the elderly towards AVs and SAVs. Given the
psychological characteristics and elderly’s behaviors, enterprises can develop corresponding AVs to
better meet and serve the travel needs in the future.
We also proposed practical suggestions according to our research results.
(1) We suggest that the local government could make policies or measures to create a good living
environment and safeguard the elderly’s physical health, which significantly influences their attitude
towards AVs and SAVs. In other words, policies concerning the elderly’s physical health have a
potential opportunity to increase their acceptance of AVs and SAVs.
(2) SWB has a significant positive correlation with BI. We advise enterprises to install an
entertainment system with a large screen, making the elderly enjoy leisure and happiness, which could
improve their SWB during the travel in AVs and SAVs.
(3) Travel variability time is significantly correlated with the travel behavior of the elderly.
To improve travel time reliability and decrease the elderly’s travel variability time, we suggested that
the public transport agencies could improve bus system efficiency, such as optimizing bus routes net,
increasing departure frequency, and reducing waiting time.
(4) In our study, PU refers to the degree to which the elderly believe that using a transport
mode would enhance their travel performance. Meanwhile, PE indicates the degree to which the
elderly believe that using public transport, AVs, and SAVs would be free of physical and mental effort.
The study results show that PU and PE significantly impact attitude for the three travel modes, and PU
influences BI significantly for SAVs. To accelerate the extensive application of AVs and SAVs for the
elderly, the enterprises could install a small slide and provide enough space to help disabled elderly
get on or offto improve their PU and PE.
Author Contributions:
Writing and editing—original draft preparation, H.S.; conceptualization and supervision,
funding acquisition, P.J.; data analysis, M.Z. and Y.C.; validation and literature review, F.Z. and Y.S. All authors
have read and agreed to the published version of the manuscript.
Funding:
This study was supported by the National Natural Science Foundation of China (Grant No. 71871107).
Conflicts of Interest: The authors declare no conflict of interest.

Sustainability 2020,12, 10661 20 of 22
Glossary
AVs Autonomous Vehicles
SAVs Shared Autonomous Vehicles
TAM Technology Acceptance Model
PU Perceived Usefulness
PE Perceived Ease of use
PC Perceived Characteristics
AT Attitude
BI Behavioral Intention
SWB Subjective Well-Being
SN Social Network
SI Social Influence
MIMIC Multiple Indicators and Multiple Causes
G-RRM Generalized Random Regret Minimization Model
RRM Random Regret Minimization Model
CFI Comparative Fit Index
TLI Tucker-Lewis Index
RMSEA Root Mean Squared Error of Approximation
SRMR Standardized Root Mean Squared Residual
AVE Average Variance Extracted
References
1.
Wong, R.C.P.; Szeto, W.Y.; Yang, L.; Li, Y.C.; Wong, S.C. Public transport policy measures for improving
elderly mobility. Transp. Policy 2018,63, 73–79. [CrossRef]
2.
Banister, D.; Bowling, A. Quality of life for the elderly: The transport dimension. Transp. Policy
2004
,11,
105–115. [CrossRef]
3.
Newbold, K.B.; Scott, D.M.; Spinney, J.E.L.; Kanaroglou, P.; P
á
ez, A. Travel behavior within Canada’s older
population: A cohort analysis. J. Transp. Geogr. 2005,13, 340–351. [CrossRef]
4.
Olawole, M.O.; Aloba, O. Mobility characteristics of the elderly and their associated level of satisfaction with
transport services in Osogbo, Southwestern Nigeria. Transp. Policy 2014,35, 105–116. [CrossRef]
5.
Hu, X.; Wang, J.; Wang, L. Understanding the travel behavior of elderly people in the developing country:
A case study of changchun, China. Procedia Soc. Behav. Sci. 2013,96, 873–880. [CrossRef]
6.
Boschmann, E.E.; Brady, S.A. Travel behaviors, sustainable mobility, and transit-oriented developments:
A travel counts analysis of older adults in the Denver, Colorado metropolitan area. J. Transp. Geogr.
2013
,33,
1–11. [CrossRef]
7.
Muhammad, T.; Kashmiri, F.A.; Naeem, H.; Qi, X.; Chia-Chun, H.; Lu, H. Simulation study of autonomous
vehicles’ effect on traffic flow characteristics including autonomous buses. J. Adv. Transp.
2020
,2020, 1–17.
[CrossRef]
8.
Harper, C.D.; Hendrickson, C.T.; Mangones, S.; Samaras, C. Estimating potential increases in travel with
autonomous vehicles for the non-driving, elderly and people with travel-restrictive medical conditions.
Transp. Res. Part. C Emerg. Technol. 2016,72, 1–9. [CrossRef]
9.
Tan, L.; Ma, C.; Xu, X.; Xu, J. Choice behavior of autonomous vehicles based on logistic models. Sustainability.
2020,12, 54. [CrossRef]
10.
Nuzzolo, A.; Persia, L.; Comi, A.; Polimeni, A. Shared autonomous electrical vehicles and urban mobility:
A vision for rome in 2035. In Data Analytics: Paving the Way to Sustainable Urban. Mobility; Nathanail, E.G.,
Karakikes, I.D., Eds.; Springer International Publishing: Berlin/Heidelberg, Germany, 2019; pp. 772–779.
[CrossRef]
11.
Tan, W.-K.; Lin, C.-Y. Driverless car rental at tourist destinations: From the tourists’ perspective. Asia Pac. J.
Tour. Res. 2020,25, 1153–1167. [CrossRef]
12.
Liu, P.; Ma, Y.; Zuo, Y. Self-driving vehicles: Are people willing to trade risks for environmental benefits?
Transp. Res. Part. A Policy Pract. 2019,125, 139–149. [CrossRef]
13.
Moreno, A.T.; Michalski, A.; Llorca, C.; Moeckel, R. Shared autonomous vehicles effect on vehicle-Km
traveled and average trip duration. J. Adv. Transp. 2018,2018, 1–10. [CrossRef]

Sustainability 2020,12, 10661 21 of 22
14.
Asmussen, K.E.; Mondal, A.; Bhat, C.R. A socio-technical model of autonomous vehicle adoption using
ranked choice stated preference data. Transp. Res. Part C Emerg. Technol. 2020,121, 102835. [CrossRef]
15.
Menon, N.; Barbour, N.; Zhang, Y.; Pinjari, A.R.; Mannering, F. Shared autonomous vehicles and their potential
impacts on household vehicle ownership: An exploratory empirical assessment. Int. J. Sustain. Transp.
2019
,
13, 111–122. [CrossRef]
16.
Levin, M.W.; Boyles, S.D. Effects of autonomous vehicle ownership on trip, mode, and route choice. Transp.
Res. Rec. 2015,2493, 29–38. [CrossRef]
17.
Hough, J.A.; Cao, X.; Handy, S.L. Exploring travel behavior of elderly women in rural and small urban North
Dakota: An ecological modeling approach. Transp. Res. Rec. 2008,2082, 125–131. [CrossRef]
18.
Mifsud, D.; Attard, M.; Ison, S. To drive or to use the bus? An exploratory study of older people in Malta.
J. Transp. Geogr. 2017,64, 23–32. [CrossRef]
19.
Truong, L.T.; Somenahalli, S.V.C. Exploring frequency of public transport use among older adults: A study
in Adelaide, Australia. Travel Behav. Soc. 2015,2, 148–155. [CrossRef]
20.
Feng, J. The influence of built environment on travel behavior of the elderly in urban China. Transp. Res.
Part. D Transp. Environ. 2017,52, 619–633. [CrossRef]
21.
Srnettzer, S.R. Constraints on Elderly Daily Travel Behaviour. Ph.D. Thesis, The University of Western
Ontario, London, ON, Canada, 1997.
22.
Payyanadan, R.P.; Lee, J.D. Understanding the ridesharing needs of older adults. Travel Behav. Soc.
2018
,13,
155–164. [CrossRef]
23.
Schmöcker, J.-D.; Quddus, M.A.; Noland, R.B.; Bell, M.G.H. Mode choice of older and disabled people:
A case study of shopping trips in London. J. Transp. Geogr. 2008,16, 257–267. [CrossRef]
24.
Amen, P.A.M.V. Exploring Elderly Mobility in the Greater Rotterdam Area: Assessing the Influence of
Personal Characteristics and Weather Conditions on Mode Choice & the Number of Trips. Master’s Thesis,
Faculty of Geosciences Theses, Utrecht University, Utrecht, The Netherlands, 2014.
25.
Rahman, M.M.; Strawderman, L.; Adams-Price, C.; Turner, J.J. Transportation alternative preferences of the
aging population. Travel Behav. Soc. 2016,4, 22–28. [CrossRef]
26.
Liu, S.; Chen, W.; Chi, Q.; Yan, H. Comparison of the elderly travel behavior in the south and the north of
China. J. Guizhou Univ. (Nat. Sci.). 2016,33, 604–608. [CrossRef]
27.
Haboucha, C.J.; Ishaq, R.; Shiftan, Y. User preferences regarding autonomous vehicles. Transp. Res. Part C
Emerg. Technol. 2017,78, 37–49. [CrossRef]
28.
LaMondia, J.J.; Fagnant, D.J.; Qu, H.; Barrett, J.; Kockelman, K. Shifts in long-distance travel mode due
to automated vehicles: Statewide mode-shift simulation experiment and travel survey analysis. Transp.
Res. Rec. 2016,2566, 1–11. [CrossRef]
29.
Yap, M.D.; Correia, G.; van Arem, B. Preferences of travellers for using automated vehicles as last mile public
transport of multimodal train trips. Transp. Res. Part A Policy Pract. 2016,94, 1–16. [CrossRef]
30.
Liu, D.; Du, H.; Southworth, F.; Ma, S. The influence of social-psychological factors on the intention to choose
low-carbon travel modes in Tianjin, China. Transp. Res. Part A Policy Pract. 2017,105, 42–53. [CrossRef]
31.
Nazari, F.; Noruzoliaee, M.; Mohammadian, A.K. Shared versus private mobility: Modeling public interest
in autonomous vehicles accounting for latent attitudes. Transp. Res. Part C Emerg. Technol.
2018
,97, 456–477.
[CrossRef]
32.
Pakusch, C.; Stevens, G.; Boden, A.; Bossauer, P. Unintended effects of autonomous driving: A study on
mobility preferences in the future. Sustainability 2018,10. [CrossRef]
33.
Pettigrew, S.; Dana, L.M.; Norman, R. Clusters of potential autonomous vehicles users according to propensity
to use individual versus shared vehicles. Transp. Policy 2019,76, 13–20. [CrossRef]
34.
Acheampong, R.A.; Cugurullo, F. Capturing the behavioural determinants behind the adoption of
autonomous vehicles: Conceptual frameworks and measurement models to predict public transport,
sharing and ownership trends of self-driving cars. Transp. Res. 2019,62, 349–375. [CrossRef]
35. Davey, J.A. Older people and transport: Coping without a car. Ageing Soc. 2007,27, 49–65. [CrossRef]
36.
Xia, X. Study on Travel Behavior of the Elderly. Ph.D. Thesis, Beijing University of Technology, Beijing,
China, 2015.
37.
Shergold, I. Taking part in activities, an exploration of the role of discretionary travel in older people’s
wellbeing. J. Transp. Health 2019,12, 195–205. [CrossRef]

Sustainability 2020,12, 10661 22 of 22
38.
Krueger, R.; Rashidi, T.H.; Rose, J.M. Preferences for shared autonomous vehicles. Transp. Res. Part C
Emergy Technol. 2016,69, 343–355. [CrossRef]
39.
Davis, F.D. Perceived usefulness, perceived ease of use, and user acceptance of information technology.
MIS Q. 1989,13, 319–340. [CrossRef]
40.
Zhang, T.; Tao, D.; Qu, X.; Zhang, X.; Lin, R.; Zhang, W. The roles of initial trust and perceived risk in public’s
acceptance of automated vehicles. Transp. Res. Part. C Emerg. Technol. 2019,98, 207–220. [CrossRef]
41.
Panagiotopoulos, I.; Dimitrakopoulos, G. An empirical investigation on consumers’ intentions towards
autonomous driving. Transp. Res. Part C Emerg. Technol. 2018,95, 773–784. [CrossRef]
42.
Bergstad, C.J.; Gamble, A.; Gärling, T.; Hagman, O.; Polk, M.; Ettema, D.; Olsson, L.E. Subjective well-being
related to satisfaction with daily travel. Transportation 2011,38, 1–15. [CrossRef]
43.
Chen, X.; Li, X.; Zhu, H. Condition evaluation of urban metro shield tunnels in Shanghai through multiple
indicators multiple causes model combined with multiple regression method. Tunn. Undergr. Space Technol.
2019,85, 170–181. [CrossRef]
44. Levy, H. Regret theory: State dominance and expected utility. J. Math. Psychol. 2017,79, 1–12. [CrossRef]
45.
Loomes, G.; Sugden, R. Regret theory: An alternative theory of rational choice under uncertainty. Econ. J.
1982,92, 805–824. [CrossRef]
46.
Chorus, C.G. A generalized random regret minimization model. Transp. Res. Part B Methodol.
2014
,68,
224–238. [CrossRef]
47.
Ma, L.; Zhang, X.; Ding, X.; Wang, G. Bike sharing and users’ subjective well-being: An empirical study in
China. Transp. Res. Part A Policy Pract. 2018,118, 14–24. [CrossRef]
48.
Watts, D.J. Networks, dynamics, and the small-world phenomenon. Am. J. Sociol.
1999
,105, 493–527. [CrossRef]
49.
Montazemi, A.R.; Conrath, D.W. The use of cognitive mapping for information requirements analysis. MIS Q.
1986,10, 45–56. [CrossRef]
50.
Kim, J.; Rasouli, S.; Timmermans, H.J. Investigating heterogeneity in social influence by social distance in
car-sharing decisions under uncertainty: A regret-minimizing hybrid choice model framework based on
sequential stated adaptation experiments. Transp. Res. Part C Emerg. Technol. 2017,85, 47–63. [CrossRef]
51.
Davis, F.D. User acceptance of information technology: System characteristics, user perceptions and
behavioral impacts. Int. J. Man-Machine Stud. 1993,38, 475–487. [CrossRef]
52.
Chen, C.-F.; Chao, W.-H. Habitual or reasoned? Using the theory of planned behavior, technology acceptance
model, and habit to examine switching intentions toward public transit. Transp. Res. Part F Traffic
Psychol. Behav. 2011,14, 128–137. [CrossRef]
53.
Chen, S.-Y. Using the sustainable modified TAM and TPB to analyze the effects of perceived green value on
loyalty to a public bike system. Transp. Res. Part A Policy Pract. 2016,88, 58–72. [CrossRef]
54.
Liu, J.; Kockelman, K.M.; Boesch, P.M.; Ciari, F. Tracking a system of shared autonomous vehicles across the
Austin, Texas network using agent-based simulation. Transportation 2017,44, 1–18. [CrossRef]
55.
Stark, J.; Hössinger, R. Attitudes and mode choice: Measurement and evaluation of interrelation. Transp. Res.
Procedia 2018,32, 501–512. [CrossRef]
56.
Kim, M.J.; Lee, C.-K.; Preis, M.W. The impact of innovation and gratification on authentic experience,
subjective well-being, and behavioral intention in tourism virtual reality: The moderating role of technology
readiness. Telemat. Inform. 2020,49, 101349. [CrossRef]
57.
Adam, Z.; Walasek, L.; Meyer, C. Workforce commuting and subjective well-being. Travel Behav. Soc.
2018
,
13, 183–196. [CrossRef]
58.
Ajzen, I.; Fishbein, M. Understanding Attitudes and Predicting Social Behavior; Prentice-Hall: Englewood Cliffs,
NJ, USA, 1980.
59.
Bagozzi, R.P. Evaluating structural equation models with unobservable variables and measurement error:
A Comment. J. Mark. Res. 1981,18, 375–381. [CrossRef]
Publisher’s Note:
MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.
©
2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).