IATSSR-00244; No of Pages 7

IATSS Research xxx (2020) xxx

Contents lists available at ScienceDirect

IATSS Research

Research Article

Attitude toward physical activity as a determinant of bus use intention:

A case study in Asuke, Japan
Yen Tran a,⁎, Toshiyuki Yamamoto b, Hitomi Sato c, Tomio Miwa b, Takayuki Morikawa c
a
Graduate School of Engineering, Nagoya University, Japan
b
Institute of Materials and Systems for Sustainability, Nagoya University, Japan
c
Institutes of Innovation for Future Society, Nagoya University, Japan

a r t i c l e i n f o a b s t r a c t

Article history: This study examines the effect of people's attitudes toward physical activity on their bus use intentions in rural
Received 9 September 2019 areas in Japan. We utilized the theory of planned behavior and designated three variables—attitude toward bus
Received in revised form 10 March 2020 use, subjective norms, and perceived behavioral control—as mediators for the hypothesized effect. Analysis re-
Accepted 13 March 2020
sults showed that attitude toward physical activity had a significant effect on bus use intention. For transport pol-
Available online xxxx
icy, this implies that improving attitudes toward physical activity can increase bus use and reduce physical
Keywords:
inactivity, partly caused by car dependence. Results from a multiple-group analysis, for the hypothesized
Attitude toward physical activity model, revealed that this finding is particularly valid for young people, car drivers, and people living within
Extended theory of planned behavior (TPB) five minutes of the nearest bus stop. Finally, the effect of attitude toward physical activity on bus use intention
Bus use intention is higher with the presence of mediators; these mediators help to increase the model's predictability for the var-
iance of bus use intention from 8.6% to 64.4%.
© 2020 International Association of Traffic and Safety Sciences. Production and hosting by Elsevier Ltd. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction Contextualized in Japan, this paper studies how to improve public

Public transport systems play an important role in providing accessi-
bility in rural areas in developed countries.1 However, these systems are
generally of lower quality than systems in urban areas, due to limited in-
vestments and subsidies. Consequently, there exists a large gap in acces-
sibility between rural and urban areas [1,2]. Inequality in accessibility is
even more serious for population groups heavily dependent on public
transport—such as the elderly, the unemployed, and the disabled—as in-
adequate accessibility can affect social inclusion and, hence, social jus-
tice [3]. Economically, low accessibility in rural areas also reduces the
agglomeration benefits brought about by their geographic proximity
to urban areas [4]. Thus, improving accessibility in rural areas should
be given special importance, as a large portion of the world population
still lives in rural communities [2,5].
⁎ Corresponding author.
E-mail addresses: tranvietyen@gmail.com (Y. Tran), yamamoto@civil.nagoya-u.ac.jp
(T. Yamamoto), sato@trans.civil.nagoya-u.ac.jp (H. Sato), miwa@nagoya-u.jp (T. Miwa),
morikawa@nagoya-u.jp (T. Morikawa).
Peer review under responsibility of International Association of Traffic and Safety Sciences.
transport services in rural areas. The continuous decline in local bus ser-
vice (LBS)2 ridership in many rural areas is an urgent issue for transport
sectors; for example, in 2007, levels of local bus ridership in Japan were
nearly half that of 1968 levels [6]. Because low bus ridership cannot at-
tract private bus operators—whose main interests are profitable opera-
tions—and local authorities must reduce subsidies for local bus systems
due to budget deficits [7], the termination of LBS programs are a foresee-
able threat [8,9]. The failure of LBS programs would cause considerable
inconvenience for those heavily dependent upon public transport and
exacerbate the situation of disadvantaged groups—as is occurring, not
only in Japan's rural areas, but also in those of other developed coun-
tries, such as the UK [10], the Czech Republic [1], and Sweden [11]. For
the survival of LBS, in this context, increasing bus ridership appears to
be a necessary and effective solution.
Asuke, a small Japanese town now part of Toyota City, was chosen as
our case study. The town has approximately 8000 inhabitants, with
37.5% of the population aged 65 and over. Similar to other rural areas,
the LBS in Asuke is inadequate and bus ridership is low. Community
buses (CBs), the main form of public transport in Asuke, serve 13 fixed

2
In remote and small-scale areas in Japan, the LBS is preferred over the traditional pub-
lic transport system due to its low investment cost. LBS uses small vehicles and often
serves a limited number of routes infrequently. This kind of bus is sometimes called a com-
1
Note that public transport is rare in rural areas in developing countries [1]. munity bus.

https://doi.org/10.1016/j.iatssr.2020.03.002
0386-1112/© 2020 International Association of Traffic and Safety Sciences. Production and hosting by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
2 Y. Tran et al. / IATSS Research xxx (2020) xxx
routes; however, this service is only available for each route on a specific
day of the week. On that day, one bus will run once in the morning (in
one direction) and once in the afternoon (in the opposite direction).
Data from the local transport department shows that CB ridership in
Asuke has been halved in eight years, from 11,619 in 2007 to 6664 in
2016. To support the CB system, people of all ages may use the school
buses that serve students on 11 routes; however, this system only
operates during periods when students attend school.3 As part of a mo-
bility management (MM) campaign conducted in Asuke, we attempted
to find factors that influence people's bus use intention (BUI) to raise
bus ridership. For example, cars far outperform buses in terms of certain
attributes—such as travel cost, comfort, and flexibility—hence, raising
bus ridership by improving these attributes is not cost-effective.
This study explores a causal relationship between attitude toward
physical activity (APA) and bus use intention. This causality is intuitive
as Japanese people generally show great affinity for physical activities;
therefore, we can expect that APA can have an effect on various subse-
quent behaviors that contain physical activities. In fact, bus use can be
considered a physically active behavior due to time spent walking and
cycling to and from bus stops, which may bring benefits in terms of
physical health. Evidence on the association between physical activity
and public transport use can be found in research [12]. While this fact
makes it possible to hypothesize a positive effect of APA on BUI, we
did not find any previous studies that confirm this effect. A study on
teenagers' travel-to-school mode choice [13] is the only case in the liter-
ature that considers APA in a travel context, but this study is concerned
with mode choice behaviors, rather than intentions. We further argue
that transport policies should focus on physical activity—as increasing
this object is related to several national policy domains and, hence, ben-
efits cross-sectoral policy designs. For instance, raising physical activity
levels has been recognized as a primary objective of the health sector
[14], and making the built environment more walkable is associated
with an increase in active transport behaviors [15]. Inactivity caused
by car dependence [16]—one of the negative consequences of transpor-
tation for human health—can, thus, be mitigated both by improving
people's APA and by promoting more physically active transport
modes, such as riding the bus, walking, and cycling. Thus, the confirma-
tion of a relation between APA and BUI can be valuable, not only for
transport policies, but also for other public policies.
In summary, this study analyzes APA as a determinant of BUI in
Asuke, Japan. We also attempt to determine how both individuals'
sociodemographic characteristics and built environment variables mod-
erate the hypothesized effect from APA to BUI. We describe the theoret-
ical basis for our model in Section 2; the development of our study
method is presented in Section 3; and the results, discussion, and con-
clusions are presented in Sections 4, 5, and 6, respectively.
2. Theoretical basis
Generally, attitudes can be defined as an individual's tendency to
evaluate an entity with a certain level of favor/disfavor [17]. Depending
on the types of entities, attitudes can be specific (i.e. when the entities
are specific behaviors, such as traveling by bus or car) or general
(i.e. when the entities are general objects, such as physical activity).
To facilitate the hypothesized causal relation between APA and BUI,
we follow Ajzen's suggestion [18] that exogenous factors, unrelated to
specific behaviors—such as personalities or situational factors—can
only affect these behaviors through specific attitudes and perceived
normative expectations. This is also in line with Bamberg's strong
criticism [19] regarding the incorrect assumption about a direct
effect from environmental concerns—a form of general attitude—to
3
School buses in Asuke operate twice daily every weekday except during the three sea-
sonal vacations–the summer, winter, and spring holidays, which generally last for two
weeks or one month.
Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
environmentally related behaviors. The literature includes examples of
models of the form attitude-mediator-intention [18–21].
Accordingly, we designated three variables in the theory of planned
behavior (TPB) [20] as mediators for the hypothesized effect of APA on
BUI. The original TPB model with specific attitudes, subjective norms,
perceived behavioral control, and intention, was extended to include
an additional factor, APA. The rationale for this extension of the original
TPB model can be described through the process in which specific atti-
tudes are formed. Specifically, Fazio [22] pointed out that in the TPB,
the process of constructing an attitude toward a specific behavior may
be affected by an “automatic component.” This component is, in fact, a
process where certain chronic beliefs related to the behavior are acti-
vated automatically [22]. As attitudes can be treated as beliefs [20,23],
in this case APA can be considered to be the belief that “physical activi-
ties are beneficial for health.” This belief can create an association be-
tween physical activities (e.g. walking and cycling) and health
benefits, and the strength of this association depends on the strength
of the belief. As walking and cycling are related to bus use, an association
between bus use and health benefits is also possible. Thus, for a person
who has associated bus use with health benefits, bus use may be evalu-
ated as more favorable. The influence of APA on the formation of atti-
tudes toward bus use is illustrated in Fig. 1.
This automatic process also applied to the normative component of
the TPB model [22]. We extended this notion by assuming that APA in-
directly affects BUI through three TPB variables: attitude toward bus
use, subjective norm (SN), and perceived behavioral control (PBC). In
addition, we compared the extended TPB model with a base model, in
which APA was allowed to directly cause BUI to examine how the
TPB's variables enhance the effect of APA on BUI. If the enhancement
was observed, then our psychological basis would validated. Our frame-
work is depicted in Fig. 2.
3. Method
3.1. Questionnaire design and data collection
We used a questionnaire designed to capture the sociodemographic
characteristics of respondents and their evaluations through questions
aimed at their specific attitude, SN, PBC, APA, and BUI.
We used a single indicator (“I love to use the CB”) designed to assess
their specific attitudes toward bus use. The SN for bus use was measured
by three indicators that reflected the individual's perceived societal
pressures to use the bus. These pressures can come from the perception
that public transport benefits both the community and the individual
citizen. For the indicator of PBC for bus use, we asked respondents
how well they know the bus system's operations. Given the low fre-
quency and limited number of routes for LBS in Asuke, we assumed
that the lack of knowledge about bus fares, timetables, and routes dis-
courages people from intentions to use the bus. To measure APA, two in-
dicators, that represent the respondent's evaluation on the importance
for their health of walking and going out (e.g. not staying at home)
were used. The BUI was measured by three indicators in both direct
and indirect ways (e.g. by asking respondents how their schedules
will be affected by bus use).
Accordingly, we used a five-point Likert scale ranging from 1
(strongly disagree) to 5 (strongly agree) for all the indicators, except
the question about the knowledge of bus use. To assess how well people
knew the bus timetables and routes, the respondents were required to
select one among five statements, arranged in a decreasing order of
knowledge. Then, their answers were reversed so that the modified re-
sponses followed an increasing level of bus use knowledge.
In the MM program, questionnaires (in Japanese) were distributed
to a total of 2838 households in Asuke in September 2017, and 1009
households responded by October 2017. In total, 2352 residents an-
swered the questions. However, as the questionnaires were mailed,
many people did not complete all the required items. After excluding
 Y. Tran et al. / IATSS Research xxx (2020) xxx
Fig. 1. The effect of APA on BUI. Note: the solid arrows represent the one-way effects, and the double-headed dashed arrows show the associations between objects.
the cases with incomplete information, a total of 1604 questionnaires
were analyzed.
3.2. Data analysis
To estimate the effects in the proposed models, the structural equa-
tion modeling (SEM) technique was used, following several steps. First,
an exploratory factor analysis (EFA) was conducted to verify the inter-
nal consistency of indicators within each factor. Second, if all the indica-
tors satisfied the consistency condition, they were submitted to a
confirmatory factor analysis (CFA) to test the suitability for SEM. To con-
duct the EFA and CFA, we divided the total sample (1604 respondents)
into two equal parts of 802 respondents. The EFA was conducted with
the first sample, and the CFA was done with the second sample. Finally,
the hypothesized effects, among the proposed constructs, were esti-
mated by SEM with the full sample of 1604 respondents.
To make SEM models identifiable, we placed several constraints fol-
lowing the instructions in Bollen [24] and Kenny et al. [25]. Specifically,
the error terms of the observed variables were assumed to be uncorre-
lated with each other and latent variables. The scales of latent variables
were specified by setting both variances of exogenous latent variables
and variances of error terms of endogenous latent variables to 1 (in
case of standardized solution) or by constraining the factor loading of
the first indicator of every construct to 1 (in case of unstandardized so-
lution). Note that we did not treat the two constructs, attitude toward
bus use (ATT) and perceived behavioral control (PBC), with single indi-
cators as directly measured, such as in Shen & Takeuchi [26] and Donald
et al. [27]. Instead, we formulated these constructs as single-indicator
latent variables. In addition, the residual variances of these single
Fig. 2. The extended TPB model (a) and the base model (b) for explaining BUI in Asuke. The notation follows [24].
Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
3
indicators were not constrained to any fixed values. In this way, the
score reliabilities of these single indicators were estimated by
the model rather than being fixed [24]. This step helps to overcome
the weakness of using single indicators in SEM.
In addition, the mediating effects of several exploratory variables, as
shown in Table 1, on the hypothesized causal effects in our models,
were investigated by employing a multiple-groups SEM framework
[24,28]. Although multiple-group analysis enabled the test for several
levels of invariance in SEM (e.g. form invariance, regression coefficient
invariance, factor loading invariance, and so on), we restricted this anal-
ysis to the test of equality in parameters (regression coefficients) across
groups. The multiple-groups analysis was carried out in which the sam-
ple grouping is based on characteristics listed in Table 1 (e.g. car driver
versus non-car driver).
4. Results
4.1. Sample statistics
The statistics of the sociodemographic characteristics of the respon-
dents are summarized in Table 1.
The average age of the sample was 60.14 years. Compared with the
average age of 52.81 of the Asuke population, this statistic indicates a
high portion of elderly people in the sample. A noticeable statistic is
the distance (in time) to the nearest bus stop. Nearly half of the respon-
dents live within five minutes of a bus stop, and approximately 80% of
them are within ten minutes of a bus stop. The sample shows an
equal distribution between males and females, similar to the Asuke
population, while the occupation statistic identifies two main
4 Y. Tran et al. / IATSS Research xxx (2020) xxx

Table 1 found no difference in the means of APA of male and female groups.
Sociodemographic characteristics of respondents (N = 1604). However, the mean of APA of the elderly male group (≥70 years old)
Feature Category Percentage Mean SD was significantly larger than that of the young male group (b70 years
Gender Male 49.69% – –
old)—the difference was 3.27%. This implies that for this male group,
Female 50.31% an increase in APA was associated with an increase in their ages. The
Age (years) b20 2.93% 60.14 18.50 mean of BUI of the female group (3.14) was significantly larger than
20–29 4.99% that of the male group (3.02), and this fact is intuitive as, generally, fe-
30–39 7.42%
males have lower access to cars than males have. In both the male and
40–49 10.6%
50–59 16.08% female groups, the means of BUI of elderly respondents (3.38 and
60–69 27.93% 3.47) were significantly larger than those of younger respondents
70–79 14.03% (2.87 and 2.99). In other words, an increase in the respondents' ages
80–89 13.03% was associated with an increase in their BUI. Finally, the mean of BUI
N90 2.99%
Time to the nearest bus ≤5 76.37% 8.36 50.63
of the female non-driver group (3.34) was significantly larger than
stop (minutes) 6–10 14.28% that of the female self-driver group (3.07).
11–15 3.3%
16–20 2.56% 4.2. The measurement model
21–25 0%
N25 3.49%
Usual transport mode Car (Driven by me) 57.63% An EFA was conducted using the R system for statistical computing
Car (Driven by family, etc.) 26.74% [29] on groups of indicators of BUI, SN, and APA using the first half of
Motorcycle or moped 4.42% the total sample (N = 802). In every case, only one factor emerged.
Bicycle 6.07%
The results of EFA with principle component analysis (PCA) are pre-
Three-wheeled car 0.85%
Other 4.29% sented in Table 2, in which the internal consistencies of factors were
Occupation Company employee, 24.19% assessed by Cronbach's alpha test (see [28] for the alpha formulation).
company management All the Cronbach's alphas are greater than 0.5, showing potential for
Public servant/group staff 7.11% good internal consistency among the indicators [28]. Thus, the next step
Student 3.30%
Part-time job 15.15%
was to verify the measurement model using a CFA approach.
Housewife/househusband 13.03% Before being submitted to the CFA, the input data was checked to en-
Agriculture/forestry 6.80% sure that it was suitable for the CFA. We employed maximum likelihood
Unemployed 25.56% estimation (MLE) for CFA (and SEM) as MLE is asymptotically unbiased
Other 4.86%
in large samples and is a consistent and efficient estimator [24]. Addi-
tionally, the Mardia's test [30] for the multinormality of the observed
variables was conducted by using the MVN package [31] in the R system
groups—company staff (24.19%) and unemployed people (25.56%). The for statistical computing [29]. The results show that the test statistic
portion of respondents with a job, in our sample, accounts for 53.25% of for Mardia skewness is 1133.4 (P-value = 0), and for kurtosis it is
the total sample and this is similar to the occupation distribution of the 14.51 (P-value = 0). The observed variables were, thus, statistically
Asuke population. The car is the most common mode of the sample. The not multivariate normally distributed. To account for the non-
total portion of people who drive themselves or who are driven by a rel- normality of the input data, bootstrapping was embedded into the
ative was approximately 84.38%. The information collected in the sam- MLE estimation process to calculate the standard errors [32].
ple reflects the typical situation of the low effectiveness of the bus The estimation for CFA was done by using a SEM package [33] in the
system in Asuke. R system for statistical computing [29]. Note that, in this study, we con-
The means of respondents' scores in psychological indicators is given sidered the indicators as continuous variables. The CFA models (run
in Table 2. Generally, respondents in the sample had fairly high APA but with the second half of the total sample) showed quite a good fit with
not strong BUI. The t-tests of unequal sizes and unequal variances, com- the observed data. All the factor loadings were significant, with the
paring the means of APA and BUI of several sociodemographic groups, values ranging from 0.54 to 0.91 (standardized solution) for both the
revealed some interesting facts. At the 95% level of confidence, we base and the extended TPB models. All the fit indexes were acceptable:

Table 2
EFA results for the indicators.

No Factor Indicators Corresponding question Mean (Standard PCA factor Cronbach's

deviation) loading alpha

1 Attitude toward physical activity (APA) APA1 Walking is good for your health. 4.47 (0.81) 0.86 0.63
APA2 Going out (e.g. not staying at home) is useful 4.27 (0.89) 0.86
for maintaining your health.
2 Attitude toward bus use (ATT) ATT1 I love to use the CB. 2.44 (1.13) – –
3 Subjective norms (SN) for bus use SN1 Improving public transport, more than now, will 3.53 (1.23) 0.89 0.79
enrich your family's life.
SN2 The town will flourish as public transport 3.74 (1.13) 0.85
improves.
SN3 I feel inconvenienced by daily traffic. 3.09 (1.49) 0.77
4 Perceived behavioral control (PBC) PBC1 How well do you know the bus fares, routes, 2.37 (1.00)
of bus use and timetable?a
5 Bus use intention (BUI) BUI1 I plan to use CB more. 2.94 (1.2) 0.87 0.78
BUI2 My life is inconvenient if CBs are unavailable. 3.38 (1.37) 0.84
BUI3 I can reorganize my daily schedule according to 2.95 (1.27) 0.79
public transport schedules.
a
The answers for this question were numbered from 1 to 5 and arranged in a decreasing level of knowledge about buses' timetables and routes, such as “I'm familiar with the routes and
timetable” or “I know little of the fare, routes, and timetable.”

Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
 Y. Tran et al. / IATSS Research xxx (2020) xxx 5

Table 3
Estimates of the extended TPB model and the base model for the whole sample and for the compared groups.

Extended TPB model Base model

From APA APA APA ATT SN PBC APA APA

↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ a) ↓
To ATT SN PBC BUI BUI BUI BUI BUI
All sample 0.59/0.32 (0.006)** 0.99/0.5 (0)** 0.47/0.31 (0.004)** 0.59/0.98 (0)** 0.3/0.51 (0)** 0.3/0.36 (0)** 0.79/0.68 0.65/0.31(0)**

≥70 years old 2.76/1.83 (0.074) 2.34/0.75 (0.007)** 2.19/1.01 (0.064) 0.76/0.8 (0)** 0.11/0.21 (0.355) 0.32/0.2 (0)** N/A
b70 years old 0.4/0.22 (0)** 0.9/0.45 (0)** 0.33/0.24 (0)** 0.58/0.93 (0)** 0.32/0.55 (0)** 0.2/0.26 (0.004)** 0.58/0.52
Unemployed 2.26/1.26 (0.15) 2.11/0.85 (0.03)* 1.86/0.86 (0.136) 0.62/0.78(0.001)** 0.24/0.37(0.052) 0.29/0.26(0.01)** N/A These groups were
Employed 0.48/0.26 (0)** 0.96/0.45 (0)** 0.38/0.25 (0)** 0.63/1.04(0)** 0.29/0.54(0)** 0.27/0.33(0)** 0.69/0.6 found as statistically
Car userb) 0.45/0.24 (0.001)** 0.93/0.46 (0)** 0.35/0.27 (0)** 0.58/0.9 (0)** 0.31/0.55(0)** 0.26/0.33(0.01)** 0.64/0.56 equal in effect from
Non-car user 1.42/0.77 (0.098) 1.39/0.73 (0.017)* 1.26/0.67 (0.152) 0.68/1.02(0)** 0.2/0.28 (0.094) 0.34/0.36(0.02)* N/A APA to BUI
N5 min c) 0.31/0.2 (0.64) 0.64/0.39 (0.319) 0.5/0.46 (0.297) 0.8/1.05 (0)** 0.22/0.37(0.021)* 0.41/0.54(0.065) N/A
≤5 min 0.69/0.34 (0.006)** 1.11/0.52 (0)** 0.44/0.25 (0.017)* 0.54/0.94(0)** 0.33/0.56(0)** 0.26/0.3 (0)** 0.85/0.68

Note: ** p-value b.01; * 0.01 b p-value b.05; a) The total indirect effects of APA on BUI, in the extended TPB models, are calculated from significant (p-value b.05) direct effects; b) Respon-
dents who selected the car as their usual transport mode, are considered car users; c) This denotes respondents who reported a distance of more than five minutes from the nearest bus
stop; N/A: Calculation for this indirect effect of APA on BUI was not possible due to the insignificance (p-value N.05) of the direct effects; From left to right: unstandardized estimates, stan-
dardized estimates (in italics), and p-values (in the parentheses); The standard errors (which determine the corresponding p-values) were calculated by using unstandardized scores of
observed variables [24].

the chi-square (with the degrees of freedom in brackets) χ2(25) = 109.43
(P-value = 0); χ2/df = 4.38; RMSEA = 0.06; GFI = 0.97; NFI = 0.95;
and CFI = 0.96 for the extended TPB model and χ2(4) = 10.89 (P-value =
.03); χ2/df = 2.72; RMSEA = 0.05; GFI = 0.99; NFI = 0.99; and CFI =
0.99 for the base model (see the note under Table 4 for the fit index ab-
breviations listed above).
4.3. The structural equation model (SEM)
After the measurement models were verified with CFA and showed
suitability, the proposed models were then estimated by SEM with the
full dataset (1604 respondents). The estimates and the fit indexes for
the extended TPB model and base model are presented in Tables 3
and 4. Note that only the results from the multiple-groups analysis, for
the compared groups in which their effects of APA on BUI are statisti-
cally different at 95% level of confidence, are presented.
In the extended TPB model, the three TPB variables—ATT, SN, and
PBC—were found to positively affect BUI with significant regression co-
efficients ranging from 0.36 to 0.98 (standardized). In addition, the un-
standardized effect (in terms of regression coefficient) of APA to BUI in
the extended TPB model was 0.79. Compared with the unstandardized
effect, in the case of direct relation between APA and BUI in the base
model at 0.65, the moderators helped to increase the regression coeffi-
cient for this relation by 21.5%. Another advantage of employing TPB
variables, as mediators for the effect of APA on BUI, is the improvement
in the model predictability for BUI. The percentage of variance of BUI,
explained by the extended TPB model, is 64.4% while the base model
can only account for 8.6% of the variance.
The estimates of the extended TPB models, with data from compared
groups, revealed some interesting facts. Regarding the estimates by age
group, only the differences in the estimates of the extended TPB model
for respondents aged 70 years and older, and those for the younger re-
spondents, are significant. Observing the estimates of the extended TPB
models for these two groups more deeply, APA does not seem to be a
strong motivation for the BUI for the older group—an indirect link be-
tween these two factors could not be established at the 95% level of con-
fidence. However, if a less restrictive criterion, based on the threshold of
p-value considered to be significant (e.g. less than 0.1 rather than 0.05)
is applied, then the indirect effect of APA on BUI can still be found for
this group (mediated by ATT and PBC, not by all the TPB's variables as
with the younger groups). For the younger group, the highly significant
estimates strongly suggest that this group views APA as a real motiva-
tion for the intention to use the bus. This view can also be found for
the unemployed group and the employed group. While the effect of
APA on BUI is only mediated by SN for the former group, if a threshold
of p-value at 0.052 is considered significant, the latter group shows
the highly significant role of APA in causing the intention to use the
bus. The final comparisons reveal a similar pattern for car users and

Table 4
Model fit indexes of the extended TPB model and the base model for the whole sample and for the compared groups.

Sample/group % variance explained for BUI Group χ2 Total χ2 Df χ2/df RMSEA GFI NFI CFI Sample size

Extended TPB model

All sample 64.4% 370.05 29 12.76 0.09 0.95 0.92 0.93 1604

≥70 years old 80.1% 148 379.28 58 6.54 0.08 1.00 0.92 0.93 482
b70 years old 60.1% 231.27 1122

Unemployed 75.7% 169.88 385.55 58 6.65 0.08 1.00 0.92 0.93 410
Employed 64.7% 215.68 1194

Car user 60.5% 269.31 368.64 58 6.36 0.08 1.00 0.92 0.93 1291
Non-car user 73.3% 99.33 313

N5 min 65.5% 152.03 427.48 58 7.37 0.09 0.99 0.91 0.92 379
≤5 min 63.6% 275.45 1225

Base model
All sample 8.6% 11.43 4 2.9 0.03 1.00 0.99 1.00 1604

Note: From left to right: % variance explained for BUI: The ability of the model to account for the variance of BUI, calculated by the standardized solution; Group χ2: The amount of chi-
square for each group of the estimated two-groups SEM; Total χ2: The chi-square test statistic to assess the overall goodness of fit of both the two-groups SEM and normal SEM; Df:
The degree of freedom of the model; RMSEA: The root mean square error of approximation; GFI: The goodness-of-fit statistic; NFI: The normed fit index; CFI: The comparative fit.

Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
6 Y. Tran et al. / IATSS Research xxx (2020) xxx
people living within five minutes of the nearest bus stop, both groups
have strong physical activity motivation for BUI. However, the link be-
tween APA and BUI for non-car users can only be established at the
90% level of confidence and is unavailable for people who live more
than five minutes from the nearest bus stop. One important fact must
be noted here—the sample sizes of the compared groups. Most of the
cases of lower significances of the estimates are with groups of smaller
sample size. This finding implies that small sample sizes can be a cause
of the insignificant estimates from SEM and, thus, a larger sample size
would strengthen the belief that APA has an effect on BUI. Apart from
the above four cases, we found no statistically significant differences
in the estimates of the extended TPB models for other pairs of groups,
such as models for males versus females.
5. Discussion
To support transport policies in rural areas in Japan by promoting in-
creased bus use, this study examines (1) the potential effects of APA and
other factors on BUI, and (2) the mediating effects of some exploratory
variables on the hypothesized relations. As some hypothesized effects
were observed to be significant, we discuss potential policy and meth-
odological implications below.
Regarding the first issue, our study shows that both the three TPB
variables and APA were found to influence BUI. Apart from confirmation
of the frequently reported predictability of the TPB in the travel domain,
the findings related to APA deserve special attention. The results from
the base model and the extended TPB model provide supportive evi-
dence for the causal effect of APA on BUI. In addition, the improvement
in modeling the effect of APA on BUI, brought about by the introduction
of the TPB's variables as mediators, is in line with the suggestion in [19]
that general attitudes influence subsequent behaviors indirectly. From a
transport policy perspective, this finding suggests that BUI can be in-
creased if promotion campaigns increase people's APA, attitude toward
bus use, SN, and PBC for bus use. As intention is the most immediate and
important predictor of behavior [34], the increased BUI can lead to in-
creased bus use, which is important for ensuring accessibility in rural
areas and for mitigating physical inactivity, partly caused by car depen-
dence. This finding can also be seen as a supportive argument for the
idea of combining transport policies with health polices—the existing
literature suggests that combining mobility and physical activity offers
a cheap, feasible strategy for increasing a large population's physical
activities [35].
For the second issue, we also found significant differences in the pat-
terns of effects among several pairs of groups. First, for younger people,
preferences in physical activity are likely to prevail in their subsequent
physically active behaviors; in the case of bus use, these interests
(along with other factors such as SN and PBC) can positively bias their
attitudes toward bus use as they seek to maintain physical health. For el-
derly people, although these causal relations are not as clear as for youn-
ger people, and the effect of SN on BUI is strongly insignificant, the same
expectation can be made with a little relaxation of the significance
threshold. This implies that promotion campaigns for raising APA, in
order to increase BUI, can be effective with the elderly. Another notice-
able finding is the strong effect of APA on BUI of car users. As car depen-
dency is generally high in rural areas in developed countries, and
motivating public transport use is not straightforward, this finding sug-
gests an alternative measure of intervention—promoting modal shifts
for car users by increasing awareness of the benefits of physical activity.
Finally, as only respondents living close to a bus stop are associated with
significant effects of APA on attitudes toward bus use, SN, and PBC—as
opposed to respondents of longer distance—transport policy interven-
tions, to promote bus use through changing the built environment,
should be evaluated. That is, efforts to increase bus ridership by improv-
ing people's attitudes toward physical activity can be effective only for
respondents living near bus stops. Thus, we suggest that among the ef-
forts in improving the built environment for walking and cycling,
Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
solutions for reducing walking distances to bus stops (e.g. by placing
more bus stops along bus routes) should be given higher priority. This,
combined with promotion campaigns to improve people's attitudes
toward physical activity, will potentially lead to an increase in bus rider-
ship, along with the expected improvements in accessibility to bus use.
This study has limitations that could be addressed by further re-
search. First, more studies are needed before we can arrive at a firm
conclusion that APA, either directly or indirectly, causes BUI. However,
our study's findings lead us to be optimistic about the existence of
such a causal effect. Second, using other forms of mediators for the
attitude-intention relation may help determine whether attitudes
such as APA affect specific behavioral intentions. Third, we suggest
future studies apply SEM with single-indicator latent variables, when-
ever single indicators are used, and the analyst is uncertain about their
reliabilities. Finally, given our study area's poor local bus service, we
expect the effect of APA on BUI may differ in areas with better local
bus service. Thus, further studies are needed to arrive at a clearer
conclusion about the effect of APA on BUI.
6. Conclusions
In this study, we found a significant effect, both direct and indirect, of
APA on BUI, and also found that the effect varies between groups. There-
fore, we suggest that improving people's APA can produce multiple out-
comes, both of improved physical health and of increased bus usage,
and this effort can either be focused on the general population or on po-
tential groups of bus users, such as young people and car users. We also
suggest that improvements in the built environment, such as more bus
stops, would potentially lead to more bus usage. Finally, in terms of the
framework for investigating bus use intention, our study provides sup-
portive evidence that general attitudes should be modeled to affect sub-
sequent behaviors, indirectly instead of directly.
Declaration of Competing Interest
None.
Acknowledgements
The data used in this study was provided by the Mobility Manage-
ment campaign for Asuke conducted by Nagoya University and funded
by the TOYOTA Mobility Foundation and JST COI Grant Number
JPMJCE1317. We would like to thank two anonymous reviewers for
their helpful comments that have contributed to the improvements in
this study.
References
[1] M. Šťastná, A. Vaishar, The relationship between public transport and the progres-
sive development of rural areas, Land Use Policy 67 (2017) 107–114, https://doi.
org/10.1016/J.LANDUSEPOL.2017.05.022.
[2] E. González-González, S. Nogués, Long-term differential effects of transport infra-
structure investment in rural areas, Transp. Res. Part A Policy Pract. (2018)
https://doi.org/10.1016/J.TRA.2018.01.026.
[3] J. Farrington, C. Farrington, Rural accessibility, social inclusion and social justice: to-
wards conceptualisation, J. Transp. Geogr. 13 (2005) 1–12, https://doi.org/10.1016/J.
JTRANGEO.2004.10.002.
[4] E. Meijers, D. van der Wouw, Struggles and strategies of rural regions in the age of
the ‘urban triumph’, J. Rural. Stud. 66 (2019) 21–29, https://doi.org/10.1016/J.
JRURSTUD.2019.01.027.
[5] J. Eckhardt, L. Nykänen, A. Aapaoja, P. Niemi, MaaS in rural areas-case Finland, Res.
Transp. Bus. Manag. 27 (2018) 75–83, https://doi.org/10.1016/J.RTBM.2018.09.005.
[6] H. Sakai, K. Shoji, The effect of governmental subsidies and the contractual model on
the publicly-owned bus sector in Japan, Res. Transp. Econ. 29 (2010) 60–71, https://
doi.org/10.1016/J.RETREC.2010.07.009.
[7] Y. Takahashi, Provision of local bus services in Japan: Focusing on the roles for local
governments and nonprofit organisations, Proceedings of the 10th International
Conference on Competition and Ownership in Land Passenger Transport, 2007 .
https://ses.library.usyd.edu.au/handle/2123/6052 , (accessed March 29, 2019).
 Y. Tran et al. / IATSS Research xxx (2020) xxx
[8] A. Taniguchi, S. Fujii, F. Hara, Promotion of a public transport by mobility manage-
ment and verification of its quantitative effect : a case study for community bus in
Obihiro city, Proc. East. Asia Soc. Transp. Stud. 5 (2005) 316–324.
[9] F. Mizutani, T. Urakami, A private-public comparison of bus service operators, ERSA
Conf. Pap. Ersa02p146, Eur. Reg. Sci. Assoc, 2002 . http://www-sre.wu.ac.at/ersa/
ersaconfs/ersa02/cd-rom/papers/146.pdf , (accessed December 19, 2018).
[10] C. Wang, M. Quddus, M. Enoch, T. Ryley, L. Davison, Exploring the propensity to
travel by demand responsive transport in the rural area of Lincolnshire in En-
gland, Case Stud. Transp. Policy 3 (2015) 129–136, https://doi.org/10.1016/J.
CSTP.2014.12.006.
[11] J. Berg, J. Ihlström, The importance of public transport for mobility and everyday ac-
tivities among rural residents, Sociol. Sci. 8 (2019) 58, https://doi.org/10.3390/
socsci8020058.
[12] C. Rissel, N. Curac, M. Greenaway, A. Bauman, Physical activity associated with pub-
lic transport use—a review and modelling of potential benefits, Int. J. Environ. Res.
Public Health 9 (2012) 2454–2478, https://doi.org/10.3390/ijerph9072454.
[13] M. Kamargianni, S. Dubey, A. Polydoropoulou, C.R. Bhat, Investigating the subjective
and objective factors influencing teenagers’ school travel mode choice–an inte-
grated choice and latent variable model, Transp. Res. Part A Policy Pract. 78
(2015) 473–488, https://doi.org/10.1016/J.TRA.2015.06.011.
[14] G.W. Heath, D.C. Parra, O.L. Sarmiento, L.B. Andersen, N. Owen, S. Goenka, F. Montes,
R.C. Brownson, Lancet physical activity series working group, Evidence-based inter-
vention in physical activity: lessons from around the world, Lancet (London, En-
gland) 380 (2012) 272–281, https://doi.org/10.1016/S0140-6736(12)60816-2.
[15] M. Smith, J. Hosking, A. Woodward, K. Witten, A. MacMillan, A. Field, P. Baas, H.
Mackie, Systematic literature review of built environment effects on physical activ-
ity and active transport – an update and new findings on health equity, Int. J. Behav.
Nutr. Phys. Act. 14 (2017) 158, https://doi.org/10.1186/s12966-017-0613-9.
[16] F.C. Bull, L. Gauvin, A. Bauman, T. Shilton, H.W. Kohl, A. Salmon, The Toronto charter
for physical activity: a global call for action, J. Phys. Act. Health 7 (2010) 421–422.
[17] A.H. Eagly, S. Chaiken, The Psychology of Attitudes, 1993.
[18] I. Ajzen, M. Fishbein, Attitudinal and normative variables as predictors of specific be-
havior, J. Pers. Soc. Psychol. 27 (1973) 41–57, https://doi.org/10.1037/h0034440.
[19] S. Bamberg, How does environmental concern influence specific environmentally
related behaviors? A new answer to an old question, J. Environ. Psychol. 23
(2003) 21–32, https://doi.org/10.1016/S0272-4944(02)00078-6.
[20] I. Ajzen, The theory of planned behavior, Organ. Behav. Hum. Decis. Process. 50
(1991) 179–211.
Please cite this article as: Y. Tran, T. Yamamoto, H. Sato, et al., Attitude toward physical activity as a determinant of bus use intention: A case study
in Asuke, Japa..., IATSS Research, https://doi.org/10.1016/j.iatssr.2020.03.002
7
[21] M. Fishbein, I. Ajzen, Predicting and Changing Behavior: The Reasoned Action
Approach, Psychology Press, New York, 2009 https://doi.org/10.4324/9780203
838020.
[22] R.H. Fazio, Multiple processes by which attitudes guide behavior: the MODE model
as an integrative framework, Adv. Exp. Soc. Psychol. 23 (1990) 75–109.
[23] M.P. Zanna, J.K. Rempel, Attitudes: A new look at an old concept, in: D. Bar-Tal, A.W.
Kruglanski (Eds.), Soc. Psychol. Knowl, Cambridge University Press, New York, NY,
US 1988, pp. 315–334.
[24] K.A. Bollen, Structural Equations with Latent Variables, Wiley, New York, 1989.
[25] D.A. Kenny, D.A. Kashy, N. Bolger, Data analysis in social psychology, Handb. Soc.
Psychol 1998, pp. 233–265.
[26] B.-J. Shen, D.T. Takeuchi, A structural model of acculturation and mental health sta-
tus among Chinese Americans, Am. J. Community Psychol. 29 (2001) 387–418,
https://doi.org/10.1023/A:1010338413293.
[27] I.J. Donald, S.R. Cooper, S.M. Conchie, An extended theory of planned behaviour
model of the psychological factors affecting commuters’ transport mode use, J. En-
viron. Psychol. 40 (2014) 39–48.
[28] R.B. Kline, Principles and Practice of Structural Equation Modeling, The Guilford
Press, 2011.
[29] R Core Team, R: A language and environment for statistical computing, R Founda-
tion for Statistical Computing, Vienna, Austria, 2018. https://www.r-project.org/.
[30] K.V. Mardia, Measures of multivariate skewness and kurtosis with applications,
Biometrika. 57 (1970) 519–530.
[31] S. Korkmaz, D. Goksuluk, G. Zararsiz, MVN: an R package for assessing multivariate
normality, R J. 6 (2014) 151–162.
[32] S.J. Finney, C. DiStefano, Non-normal and categorical data in structural equation
modeling, Quant. Methods Educ. Behav. Sci. Issues, Res. Teaching. Struct. Equ.
Model. A Second Course, IAP Information Age Publishing, Charlotte, NC, US 2013,
pp. 439–492.
[33] R. Yves, lavaan: an R package for structural equation modeling, J. Stat. Softw. 48
(2012) 1–36.
[34] P. Sheeran, Intention—behavior relations: a conceptual and empirical review, Eur.
Rev. Soc. Psychol. 12 (2002) 1–36.
[35] T. Götschi, J. Garrard, B. Giles-Corti, Cycling as a part of daily life: a review of health
perspectives, Transp. Rev. 36 (2016) 45–71, https://doi.org/10.1080/01441647.
2015.1057877.