You are on page 1of 15

Transport Policy 99 (2020) 314–328

Contents lists available at ScienceDirect

Transport Policy
journal homepage: http://www.elsevier.com/locate/tranpol

The associations between older adults’ daily travel pattern and objective
and perceived built environment: A study of three neighbourhoods
in Singapore
Yuting Hou a, *, Winston Yap a, Rochelle Chua b, Siqi Song c, Belinda Yuen a
a
Lee Kuan Yew Centre for Innovative Cities, Singapore University of Technology and Design, Singapore
b
Singapore-ETH Centre for Global Environmental Sustainability, Singapore
c
National University of Singapore, Singapore

A B S T R A C T

The increasing preference for ageing-in-place highlights the pivotal role of the neighbourhood environment in meeting the mobility needs of older adults. This paper
examines the combined associations of objective and subjective measures of built environment features with older adults’ travel pattern in Singapore. A multivariate
ordered probit (OP) modelling framework is applied to examine older adults’ daily trip frequency by four transport modes: walking, public bus, Mass Rapid Transit/
Light Rapid Transit (MRT/LRT), and private motorized modes, while controlling for the correlations among the use of different modes. The results show that while
perceived access to recreational facilities exert independent positive effects on older adults’ daily walking trip frequency and enhance the effects of comparable
objective measures, other subjective measures do not significantly moderate the effects of objective features on walking trip-making, including those of actual access
to utilitarian destinations, the availability of sheltered walkways and unsafe pedestrian crossings near residence. The results also find independent positive effects of
perceptions of transit proximity and comfortable and safe pedestrian connections on older adults’ daily transit trip-making.

1. Introduction transport-related physical activity such as walking and cycling (Shige­


matsu et al., 2009; Cerin et al., 2013a; Hoehner et al., 2005). Though the
The influence of land use and the design of neighbourhoods on in­ perceived measures of environmental features are likely to be influenced
dividual travel behaviour is an important aspect of transportation by the actual environment, previous studies suggest that perceived
research. Many empirical studies have indicated that various dimensions (subjective) and objective measures of the built environment can be
of built environment characteristics, such as land use density and di­ weakly associated with each other and their impacts on physical activ­
versity, street connectivity, distance to public transit service and ities differ (Orstad et al., 2017; Jack and Mccormack, 2014; Nyunt et al.,
accessibility to essential destinations, significantly influence people’s 2015). While many studies have incorporated both types of measures,
travel pattern (e.g. travel tendency, mode choice, and trip duration/ only a few of them have explored how perceptions of the built envi­
length) (e.g. Cervero and Kockelman, 1997; Ewing and Cervero, 2001; ronment affect the relationship between the objectively-assessed built
Ewing and Cervero, 2010). Older adults’ travel characteristics are ex­ environment and walking/cycling (e.g. Ma and Cao, 2019, Jack and
pected to be more sensitive to the influences of the neighbourhood built Mccormack, 2014), and fewer studies have extended to look at other
environment as they are less likely to travel far from their homes and non-active travel modes (e.g. Ma and Cao, 2019; Cao et al., 2009; Lee,
have higher demands for the safety and convenience of the built envi­ 2013).
ronment (Böcker et al., 2017). While the travel impacts of the built This paper aims to address these knowledge gaps and contribute to
environment has been widely studied, much less is known about the the empirical literature by exploring older adults’ interactions with the
impacts of the built environment on different transportation options built environment in accomplishing their daily travel demands and ac­
among older adults (Figueroa et al., 2014). tivities, using Singapore as a case study. Specifically, we employ both
Recent studies on physical activities and health have discussed the the self-reported survey data as well as GIS data on transportation and
importance of people’s perceptions of the built environment in affecting land use to examine how different built environment features influence

* Corresponding author. Lee Kuan Yew Centre for Innovative Cities, Singapore University of Technology and Design, 8 Somapah Road, Buiding 3, Level 2, 487372,
Singapore.
E-mail addresses: yuting_hou@sutd.edu.sg, houyuting86@gmail.com (Y. Hou), winston_yap@sutd.edu.sg (W. Yap), chua92rochelle@gmail.com (R. Chua),
siqigracesong@gmail.com (S. Song), belinda_yuen@sutd.edu.sg (B. Yuen).

https://doi.org/10.1016/j.tranpol.2020.06.017
Received 15 August 2019; Received in revised form 26 March 2020; Accepted 27 June 2020
Available online 29 July 2020
0967-070X/© 2020 Elsevier Ltd. All rights reserved.
Y. Hou et al. Transport Policy 99 (2020) 314–328

older adults’ daily trip frequency by different transportation modes (e.g. et al., 2008; Frank et al., 2010; Cain et al., 2014; Etman et al., 2014, Van
walking, public transit, private vehicle), and how the effects of the Holle et al., 2014, Procter-Gray et al., 2015). Proximity to neighbour­
objective and perceived built environment measures agree and enhance hood amenities such as social facilities (Etman et al., 2014; Cerin et al.,
with each other. Currently, we do not have much knowledge on the 2013a, Procter-Gray et al., 2015), transportation facilities (Figueroa
travel impacts of subjective built environment beyond active modes, et al., 2014), retail facilities (King et al., 2003) and recreational facilities
especially for older people, nor do we know much about the combined (Choo et al., 2016; Moniruzzaman et al., 2015) are associated with
effects of the two measures of built environment on individual travel. higher transport related walking among older people. A good urban mix
This study is the first to address these issues in the Singapore context. of amenities encourages trip chaining where older adults visit multiple
Singapore as a city-state provides a good example of integrated land destinations in one trip, extending the amount of time spent walking per
use and transport planning with much emphasis on public transport trip.
network as the backbone for promoting active mobility since the 1970s In terms of the design dimension, street characteristics including
(Centre for Liveable Cities and Urban Land Institute, 2014). By focusing connectivity, aesthetics, safety and building diversity are found to in­
on the older population in Singapore, this study aims to provide insights fluence transportation for older adults. Frank et al. (2010) and Van Holle
on the types of pedestrian-friendly urban fabrics that encourage walking et al (2014) found that denser streets lead to more walking and shorter
and the use of public transport—the two more often used modes of car trip lengths. Also, traffic intersection control and pleasant aesthetics
transport with ageing (Shrestha et al., 2017; Webber et al., 2010; are found to have a positive association with transport related walking
Wretstrand et al., 2009). A secondary aim is to suggest urban planning for older adults (Cain et al., 2014). Böcker et al. (2017) found that
and design interventions that may promote healthy travel and active building diversity along streets positively affect the probability of
living for older adults. These implications would be useful for Singapore transport related walking for older adults.
as well as other cities that are facing ever-increasing land use constraints
and a rapidly ageing population. 2.2. Travel impacts of perceived measures of the built environment
The paper is organized as follows. Section 2 conducts a brief review
of the empirical studies involving the relationship between the built In addition to objective measures of the built environment, many
environment and older adults’ travel behaviour. Section 3 introduces studies from physical activity literature have incorporated perceived
the research design and methods, while section 4 presents the main built environment measures when investigating the travel impacts of
findings from the regression analysis. Section 5 concludes with a dis­ living environment. As suggested by the ecological model created by
cussion of the main findings and their implications for urban planning Sallis et al. (2006), individuals’ perceptions of environments are
and design. distinguished from more objective aspects of environments and both are
likely to be significant influences.
2. Literature review Findings from various literature show that perceived built environ­
ment measures are commonly associated with transport related walking
This section reviews recent empirical studies on the relationship and cycling, but not with driving (Ding et al., 2014; Owen et al., 2004;
between older adults’ travel behaviour and the built environment, Saelens and Handy, 2008). This makes intuitive sense as pedestrians
which is defined as composing of both objective and perceived measures experience the urban environment on a perceptual scale whereas drivers
(Davison and Lawson, 2006). We review studies in both transportation do not and hence may be less perceptive of the built environment. Ding
and health literature that looks into transportation-related physical ac­ et al. (2014) found that older adults with driving capacities tend to be
tivities and summarize the impacts of objective and subjective measures more aware of park locations compared to non-drivers; however, no
of the built environment separately. We also discuss related empirical indication of association with total physical activity was found.
studies that look into the structural relationship between objective and Previous studies examined associations between perceived built
perceived measures of the built environment to identify the potential environment measures and transport related walking for older adults
interactive effects of the two measures on the use of different transport (Koh et al., 2015; Pelclová et al., 2012; Cerin et al., 2013b). At the
options in older age. planning scale, higher perceived residential density is found to be
positively associated with transport related walking (Shigematsu et al.,
2.1. Travel impacts of objective measures of the built environment 2009; Pelclová et al., 2012). Similarly, residents who perceive their
residential neighbourhoods to be more urbanized engage in more
The theoretical framework of land use-travel correlation widely used transport related walking than rural residents (Van Cauwenberg et al.,
in transportation research is derived from the random utility foundation 2012). Perceived proximity to amenities is also found to be associated
(Boarnet and Crane, 2001; Boarnet and Sarmiento, 1998; Cervero and with transport related walking for older adults. Cerin et al. (2013a,b)
Kockelman, 1997). Individual travel demands for a given travel mode and Corseuil et al. (2017) found that perceived proximity to social and
are determined by time and costs of travel and their socio-demographic recreational facilities lead to higher probabilities of trip-making for
characteristics, while land use and built environment features (e.g. older people. Furthermore, perceived proximity to retail amenities is
density, diversity, and design) can be factored into the travel demand associated with higher levels of transport related walking for older
model by altering the time and costs of travel. adults (Inoue et al., 2011; Chudyk et al., 2015).
A wide range of studies present the associations between the objec­ At the design scale, perceived street connectivity, aesthetics and
tively determined built environment and transportation for older adults. availability of pedestrian infrastructure are found to be positively
At the planning scale, residents in areas with higher population density associated with transport related walking (Inoue et al., 2011; Corseuil
are less car-dependant (Mercado and Páez, 2009; Moniruzzaman et al., Giehl et al., 2017; Shigematsu et al., 2009; Cerin et al., 2013b). Streets
2015; Figueroa et al., 2014) and more likely to engage in active travel that are well-maintained and connected to a wider range of destinations
(Moniruzzaman et al., 2013; Frank et al., 2010; Van Holle et al., 2014; can help to encourage older people to walk more. On the other hand, the
Choo et al., 2016). Similar results are observed when the density of relationship between perceived safety and transport related walking is
household units was measured (Böcker et al., 2017; Moniruzzaman more complex. Cerin et al. (2013b) found perceived crime to be posi­
et al., 2015). These findings reflect the consensus that areas with high tively related to transport related walking. This is possibly due to the
population density generally have greater access to services, which need to take detours around unsafe areas. Other studies found perceived
reduce the need for car travel. safety to be negatively associated with physical activity (Van Cauwen­
In addition, many studies show that urban areas with greater land berg et al., 2012; Corseuil Giehl et al., 2017). Depending on the urban
use mix are associated with higher levels of transport-related walking (Li structure, the effects of safety on walking is ambiguous.

315
Y. Hou et al. Transport Policy 99 (2020) 314–328

2.3. Relationship between objective and perceived measures 1) pedestrian-friendly built environment features (e.g. density, street
connectivity, pedestrian infrastructure, access to transit and other ser­
Many prior studies have examined both objective and perceived as­ vice destinations) are positively associated with older adults’ daily
pects of built environment effects on individual transport in the same walking and transit travel and negatively associated with car travel; 2)
model (Ding et al., 2014; Van Dyck et al., 2012; Jack and Mccormack, the associations between walkable features and individual daily travel
2014; Cerin et al., 2013a). The findings indicate that the two measures are enhanced by higher levels of corresponding perceived measures of
do not correspond well with each other and their effects on active travel those features. Both the objective aspect of built environment charac­
and physical activity should be considered as independent of each other teristics and their perceived measures are included in the model and
instead of approximating each other (Nyunt et al., 2015; Orstad et al., their direct and combined effects on individual travel are explored.
2017). However, only a few studies posit a relationship between
objective and perceived measures of the built environment when
3.1. Conceptual model
examining their travel impacts (Handy et al., 2006, Foster and
Giles-Corti, 2008). Some studies suggest that perceptions of the built
A disaggregate framework is developed to examine older adults’
environment mediate the association between objectively-determined
daily trip frequency by four classes of transport modes: walking, public
built environment and individual transport-related activities (Jack and
bus, Mass Rapid Transit/Light Rapid Transit (MRT/LRT) trains, and
Mccormack, 2014). For example, Ma and Cao, 2019 applied a structural
private motorized modes (including car driver, car passenger, motor­
equation modelling approach to examine the structural relationship
cycle driver, motorcycle passenger). Following both the utility-based
between the two measures on individual travel and found that access to
travel demand theories and the ecological model, we consider individ­
specific destinations, including religious/civic buildings and shops,
ual travel demand for a given travel mode as a function of their socio-
significantly affects the propensity to walk or bike to those destinations
demographic characteristics, neighbourhood built environment and its
through their influences on people’s perceived access to those destina­
perception, which also moderates the effects of actual environment
tions. Van Dyck et al. (2013) focused on women living in
features. The conceptual model is indicated in Fig. 1.
socio-economically disadvantaged neighbourhoods and found that their
Here we assume that the way that older adults’ perceptions towards
perceived physical activity environment significantly mediate the effects
the built environment alter the association between the objective mea­
of objectively-assessed destination access/street connectivity in the
sures and individual travel follows a step-function. That is, when older
neighbourhood on transport walking.
adults’ self-reported ratings of the built environment features reach
Other studies suggest that the two built environment measures
some level, the effects of corresponding objective measures on travel
moderate each other’s effects on individual travel (Jack and Mccormack,
behaviour is more pronounced than those with lower levels of perceived
2014; Ding et al., 2012; Bracy et al., 2014), which is also supported by
built environment measures. This assumption is tested by dichotomizing
the ecological model (Sallis et al., 2006). The revealed interactive effects
the moderating variable (i.e., perceived built environment measures) at
on walking produce mixed evidence. For example, Jack and McCormack
the point where the step in the function is likely to occur (Baron and
(2014) found that higher walkability of the neighbourhood environ­
Kenny, 1986) (see detailed discussions in Section 3.5).
ment, measured in objective terms, is associated with higher propensity
and longer transport walking duration. Meanwhile, the effects of
perceived environment features such as mix of utilitarian land use and 3.2. Analysis methods: A multivariate ordered probit model
pedestrian infrastructure on transport walking are enhanced by higher
objectively-assessed walkability. Bracy et al. (2014), however, found no Given that the travel outcome variable is ordinal and discrete in
significant interactive effects of objective walkability and perceived nature, we apply an ordered probit (OP) approach (Badoe, 2007;
traffic and pedestrian safety on total walking time for transportation Boarnet and Sarmiento, 1998; Páez et al., 2007; Roorda et al., 2010). As
purposes per week, though their direct independent associations with discussed by Train (2003), the observed number of trips (or classes of
transport walking are significant. Koohsari et al. (2015), on the other trip frequency) made by individual i per day using travel mode m (Yim )
hand, examined how the mismatch between objective and subjective can be considered to be associated with some levels of utility (Uim )
measures of street connectivity and land use mix impacts individual through threshold cutoffs:
walking propensity. Their results indicate that higher perceptions of the
environment features compensate the negative impacts of the actual low Uim = Zim βm + εim , Yim = j if Cutij < Uim ≤ Cutij+1 , j = 0, .., Jm (1)
neighbourhood walkability on transport walking, while lower percep­
where Zim represents the set of explanatory variables of individual i
tions of walkability would reverse the impacts of high walkable features
associated with trip-making by mode m, βm is the corresponding vector
on transport walking (Koohsari et al., 2015). j j+1
Evidence on the combined effects of the two measures on travel by of coefficients to be estimated, εim is the error term, and Cu ti and Cut i
other modes is very limited. For example, Lee (2013) looked at both the are the lower and upper bound for j number of trip-making by mode m
independent effects of perceived neighbourhood environment and their (Cuti0 = − ∞, Cut Ji m +1 = +∞ for each mode type m). The probability that
interactive effects with the actual neighbourhood walkability on public individual i makes j number of trips (or trip frequency classes) per day
transport usage. Her results indicate that in mixed-land-use environ­ using mode m is (Badoe, 2007):
ment, the propensity of using public transport would be negatively ( )
Pr(Yim = j ) = Pr Cutj < Uim ≤ Cut( j+1
affected by unfavourable perceptions of personal safety on crime and )
= Pr( Cutj < Zim βm + εim ≤ Cutj+1
violence. Meanwhile negative perceptions of pedestrian safety in = Pr( Cutj − Zim βm <) εim ≤
)
low-density environment (e.g. unsafe crossings, fast traffic) would ( Cutj+1 − Zim)βm
= ∅ Cutj+1 − Zim βm − ∅ Cutj − Zim βm
further reduce transit use. j = 0, …, Jm + 1
The mixed findings on walking travel and limited findings on other (2)
modes of travel suggest that the interactions of objective built envi­
ronment features and their comparable perceived measures in relation where ∅( ⋅) is the cumulative standard normal distribution function.
to individual travel need to be further explored. In the usual OP models, the εim terms are standard normally
distributed and assumed to be independent and identically distributed
3. Research approach across individuals for each mode type m. This assumption implies that
the frequencies of daily trip-making by different travel modes are in­
Based on the literature review, two basic hypotheses are developed: dependent of each other. In reality, however, this assumption may not

316
Y. Hou et al. Transport Policy 99 (2020) 314–328

Fig. 1. The conceptual mode of individual travel modelling.

hold and individuals’ decisions on the level of trip-making across travel they were coded as “no one at home”. The 1002 interviews were
modes are likely to be “complements” or “substitutes” to each other due completed in March 2018 and were evenly distributed across the 3
to the unobserved factors. For example, individuals who choose a more neighbourhoods, achieving a 40% response rate. To ensure that the re­
active lifestyle may have a higher propensity in undertaking more trips sults of the survey represent the population profile of the three study
by both walking and public bus/MRT and fewer trips by cars in a day, neighbourhoods, the survey data were weighted based on the age,
even after observed effects such as environmental factors are controlled. gender and ethnic information of each neighbourhood obtained from the
This implies a complementary relationship between walking and transit 2015 General Household Survey of Singapore (Singapore Department of
trip-makings and a substitution relationship between walking and car Statistics, 2018). Weighted data were used for all statistical tests in this
trip-making. study.1
To account for the possible unobserved factors/determinants of those Except for the socio-demographic characteristics (e.g. age, gender,
travel outcome variables, this study follows previous studies and extend ethnicity, dwelling unit type), the survey also collected information on
the standard OP model to a multivariate OP model system (Ferdous older adults’ perceptions of the neighbourhood environment and their
et al., 2010; Bhat and Srinivasan, 2005; Greene and Hensher, 2009). trip-making by different transportation modes in a typical day during
Using our notation in Equation (1), the εim terms in the system of m the past week of the survey period. Specifically, a single trip in our
underlying latent seemingly unrelated equations representing the or­ survey is identified as any point-to-point travel in a whole journey (from
dered choices of trip frequency by m travel modes for individual i are an origin to a final destination) made by an individual. For example, a
allowed to be correlated with each other. Defining εi = (εi1 , …, εim )T , εi long-distance travel made by MRT usually includes multi-modal seg­
is multivariate normal distributed with a mean vector of zeros and a ments of trips, such as walking from home to a bus stop, taking a bus
correlation matrix R as follows (Ferdous et al., 2010): from the original stop to a bus interchange, and walking from the
⎡⎛ ⎞ ⎛ ⎤ interchange to a MRT station or a final destination. This means that
0 1 ρ12 … ρ1m ⎞ individuals’ choices of the number of trip-making by different transport
⎢⎜ 0 ⎟ ⎜ ρ21 1 … ρ2m ⎟⎥
εi ∼ N ⎢ ⎜ ⎟
⎣⎝ ⋮ ⎠, ⎝
⎥ or εi ∼ N[0, R] (3) modes per day, as defined here, are naturally correlated, which further
⋮ ⋮ ⋱ ⋮ ⎠⎦
justifies the multivariate OP modelling approach. We excluded those
0 ρm1 ρm2 … 1
infrequent travellers (who have not travelled during the past week) from
Positive elements in the correlation matrix reflect that the presence the regression analyses, as these people are less exposed to the neigh­
of common unobserved factors that influence the propensity of indi­ bourhood and more likely to have misperceptions of the built environ­
vidual trip-making by two modes in the same direction, while negative ment characteristics (Koohsari et al., 2015). This yields a total sample of
elements reflect the effects of common unobserved factors on the cor­ 900 older adults from the 3 neighbourhoods. Table 1 shows that the
responding two modes to be opposite (Ferdous et al., 2010; Greene and average trip-making per day for older adults in the 3 study neighbour­
Hensher, 2009). If all the non-diagonal elements in R are zero, the modal hoods are around 4 trips by walking, 2 trips by public bus or MRT/LRT,
system in Equation (1) collapse to independent OP models of and 1 trip by private motorized modes. Compared with the 2012 na­
trip-making by each type of travel mode. tional results based on the Household Interview Travel Survey (HITS)
The multivariate OP model system is estimated by the full- conducted by the Land Transport Authority of Singapore, older adults of
information maximum likelihood approach and is implemented by the 3 neighbourhoods in the survey year 2018 seemed to prefer making
Roodman’s (2011) conditional mixed-process (CMP) estimator (in more trips by public bus and MRT/LRT per day and slightly lower
STATA) (Roodman, 2011). To facilitate comparison of effects of built number of daily trips by walking and private modes.2
environment features on trip-making by different modes, the same set of
explanatory variables is applied in all regression analyses. 3.3.1. Objective measures of the built environment
Following the literature review, we constructed GIS measures of the
built environment features by focusing on the following 6 dimensions:
3.3. Data and construction of variables
(a) residential density, (b) street connectivity, (c) pedestrian/cycling
infrastructure, (d) distance to public transport hubs (i.e. bus stops/in­
The main data used in this study are drawn from our resident survey
terchanges, MRT/LRT stations), (e) distance to retail and other service
of 1002 senior citizens (aged 55+) in 3 pre-selected neighbourhoods of
destinations (e.g. wet markets, shops, community centres, senior activity
Singapore as part of the ‘‘Innovative planning and design of age-friendly
centres), and (f) distance to parks, open spaces, and other recreational/
neighbourhoods in Singapore’’ research project. Respondents were
exercise facilities. To examine the variations in the built environment
selected from a Singapore Department of Statistics supplied sample of
2500 household addresses with one household representative aged 55
years or older residing in the 3 neighbourhoods. Each neighbourhood 1
The mean value of sampling weights is 1. We also conducted statistical
was allocated with an equal number of selected addresses, which have a
analyses using the survey data without the weight but did not find the estimates
distribution close to the actual distribution of housing unit type in that to differ significantly from those using weighted data.
neighbourhood. Each selected household address was sent an invitation 2
Based on the 2012 HITS data, the average daily trip frequency for Singa­
letter describing the purpose of the study about 1–2 weeks before the porean residents aged 55 and over is about 4.62 trips by walking (including
surveyor visited the address. To mitigate possible selection bias, each walking-only trips and first and last mile walk trips), 0.73 trips by public bus,
selected household was visited three visits across times and days before 0.33 trips per day by MRT/LRT, and 1.54 trips per day by private vehicle/taxi.

317
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 1
Descriptive statistics of objective and subjective measures of built environment characteristics.
Variable Description Expected Sign on N Mean Std. Min Max
walking trip frequency Dev.

Subjective measures of the built environment

Perceptions of connecting spaces/pedestrian facilities


High level + 900 0.57 0.49 0 1
Low level (reference) N/A 0.43 0.49 0 1
Perceptions of access to transit hubs
High level + 900 0.94 0.24 0 1
Low level (reference) N/A 0.06 0.24 0 1
Perceptions of access to utilitarian facilities
High level + 900 0.53 0.50 0 1
Low level (reference) N/A 0.47 0.50 0 1
Perceptions of access to recreational facilities
High level + 900 0.49 0.50 0 1
Low level (reference) N/A 0.51 0.50 0 1
Objective measures × Subjective measures of the built environment
Residential density
den400_du Density of dwelling units within a 400 m buffer zone1 of a residence + 900 9160.1 2951.3 678.4 14039.5
(number per square kilometres)
___ × High perceptions of pedestrian connection + 517 9307.9 2926.1 678.4 14039.5
___ × Low perceptions of pedestrian connection (reference) N/A 383 8960.6 2977.2 2120.7 14039.5
___ × Perceived transit access + 844 9294.5 2856.5 678.4 14039.5
___ × No perceived transit access (reference) N/A 56 7134.7 3585.8 2120.7 14039.5
___ × High perceptions of access to utilitarian facilities + 477 8873.7 3158.4 1132.0 14039.5
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 9414.0 2733.1 678.4 14039.5
___ × High perceptions of access to recreation facilities + 438 9289.8 3030.8 678.4 14039.5
___ × Low perceptions of access to recreation facilities (reference) N/A 462 9023.3 2862.2 1022.6 14039.5
Street connectivity
cnt400_3wayJun Number of road intersections that are at least 3-way within a 400 m + 900 8.18 4.63 2 38
buffer zone of a residence
___ × High perceptions of pedestrian connection + 517 8.14 4.60 3 38
___ × Low perceptions of pedestrian connection (reference) 383 8.24 4.67 2 37
Pedestrian infrastructure
cnt400_rdcwpedlght Number of road crossing with pedestrian/cyclists traffic signals within a + 900 11.19 5.29 1 23
400 m buffer zone of a residence
___ × High perceptions of pedestrian connection + 517 11.68 5.27 1 23
___ × Low perceptions of pedestrian connection (reference) N/A 383 10.52 5.25 1 23
cnt400_rdcnolght Number of road crossing without pedestrian/cyclists traffic signals + 900 1.93 1.58 0 5
within a 400 m buffer zone of a residence
___ × High perceptions of pedestrian connection + 517 2.07 1.56 0 5
___ × Low perceptions of pedestrian connection (reference) N/A 383 1.74 1.60 0 5
cnt400_pedpass Number of pedestrian overhead bridge & underpass within a 400 m – 900 0.84 1.03 0 4
buffer zone of a residence
___ × High perceptions of pedestrian connection + 517 0.77 0.98 0 4
___ × Low perceptions of pedestrian connection (reference) N/A 383 0.94 1.08 0 4
len400_covftpth Length of covered walkway within a 400 m buffer zone of a residence + 900 194.26 160.45 0 710.64
(m)
___ × High perceptions of pedestrian connection + 517 200.09 160.23 0 710.64
___ × Low perceptions of pedestrian connection (reference) N/A 383 186.38 160.62 0 710.64
len400_uncovftpth Length of unsheltered footpath within a 400 m buffer zone of a – 900 5189.22 837.96 2664.02 9363.21
residence (m)
___ × High perceptions of pedestrian connection + 517 5227.25 829.3 2664.02 9363.21
___ × Low perceptions of pedestrian connection (reference) N/A 383 5137.88 847.9 2664.02 8424.17
nrdist_cycnw Distance to nearest cycling path or park connector path (from each – 900 481.26 490.39 14.24 1602.53
respondent’s residential address)
___ × High perceptions of pedestrian connection – 517 501.11 495.39 17.96 1602.53
___ × Low perceptions of pedestrian connection (reference) N/A 383 454.48 482.92 14.24 1602.53
Distance to public transit (from each respondent’s residential address, unit: meter)
nrdist_busstp Distance to nearest bus stop – 900 123.02 54.79 26.84 320.69
___ × Perceived transit access – 844 122.58 54.96 26.84 320.69
___ × No perceived transit access (reference) 56 129.61 52.15 33.62 223.37
nrdist_mrtstn Distance to nearest transit station – 900 361.55 204.68 31.38 1040.02
___ × Perceived transit access – 844 361.36 204.58 31.38 1040.02
___ × No perceived transit access (reference) N/A 56 364.30 207.99 31.38 890.13
Distance to utilitarian destinations (from each respondent’s residential address, unit: meter)
nrdist_wetmkt Distance to nearest wet market – 900 1183.01 626.36 51.3123 2138.36
___ × High perceptions of access to utilitarian facilities – 477 1247.69 559.19 51.31 2083.06
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 1110.07 687.76 51.31 2138.36
nrdist_comm Distance to nearest community service facilities (e.g. Community – 900 147.49 115.08 0 621.09
centres/clubs, Senior Activity Centres/Daycare Centre Facilities)
___ × High perceptions of access to utilitarian facilities – 477 131.45 92.69 0 621.09
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 165.57 133.81 0 621.09
nrdist_hlcare Distance to nearest health care facilities (e.g. clinic/polyclinic, – 900 170.52 97.96 0 468.73
registered pharmacy)
___ × High perceptions of access to utilitarian facilities – 477 172.38 94.69 0 468.73
(continued on next page)

318
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 1 (continued )
Variable Description Expected Sign on N Mean Std. Min Max
walking trip frequency Dev.

Subjective measures of the built environment

___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 168.42 101.59 0 466.27
nrdist_otherretail Distance to nearest (non-food) retail establishment – 900 322.56 373.65 0 1266.58
___ × High perceptions of access to utilitarian facilities – 477 265.04 358.22 0 1266.58
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 387.42 380.44 0 1266.58
nrdist_food Distance to nearest food/beverage service facility (e.g. hawker centres – 900 136.14 69.70 0 411.89
and food shops)
___ × High perceptions of access to utilitarian facilities – 477 139.83 69.99 0 411.89
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 131.99 69.22 0 371.00
nrdist_plwrshp Distance to nearest place of worship (e.g. Chinese temples, mosques, – 900 349.31 126.70 19.23 581.89
Indian temples)
___ × High perceptions of access to utilitarian facilities – 477 348.63 119.39 19.23 581.89
___ × Low perceptions of access to utilitarian facilities (reference) N/A 423 350.08 134.62 49.31 581.89
Distance to recreational destinations (from each respondent’s residential address, unit: meter)
nrdist_sptfac Distance to nearest sports facilities (e.g. gyms, swimming pools, sports – 900 498.29 278.30 89.64 1331.94
fields)
___ × High perceptions of access to recreation facilities – 438 459.30 262.72 89.64 1331.94
___ × Low perceptions of access to recreation facilities (reference) N/A 462 535.25 287.74 89.64 1331.94
nrdist_open Distance to nearest open space and playground – 900 324.70 154.12 42.69 728.27
___ × High perceptions of access to recreation facilities – 438 278.53 152.59 42.69 728.27
___ × Low perceptions of access to recreation facilities (reference) N/A 462 368.48 142.49 43.82 701.81
nrdist_parks Distance to nearest park (national park or neighbourhood park) – 900 288.27 136.16 42.33 690.00
___ × High perceptions of access to recreation facilities – 438 282.96 137.38 42.33 637.27
___ × Low perceptions of access to recreation facilities (reference) N/A 462 293.30 134.95 59.43 690.00

measures within a neighbourhood as well as across the 3 neighbour­ CFA because of their variable formats. Instead, three indices are created:
hoods, all measures are applied to an immediate zone surrounding each (1) presence of public transport hubs (bus stops, interchanges and MRT/
respondent’s 6-digit postal code address. LRT stations); (2) utilitarian land use mix that includes the self-reported
The main data sources for developing the GIS measures are gathered count of destinations (i.e., supermarket, shop, shopping centre, bank,
from Onemap.sg, which provides point-based GIS layers for various post office, senior activity/daycare centre, wet/dry markets, hawker
service facilities; DataMall which provides geographic and attribute in­ centre/food court, community club/centre, place of worship, clinics and
formation on transportation infrastructure and facilities such as foot­ pharmacy); (3) recreational land use mix that includes the self-reported
path, cycling path, road crossing, and traffic lights and land use data like count of destinations (open public space, parks, indoor sports facilities
housing, retail, and social health facilities. Apart from the 3 data sour­ such as gyms, swimming complex and recreation club).
ces, we also employ other public data sources such as Google map to As discussed in the conceptual model, interactions terms between
collect information on the distribution of wet markets in Singapore. objective measures of built environment features and conceptually
matched subjective measures of the features need to be generated. To
3.3.2. Perceived measures of the built environment create greater variability of self-ratings of the built environment, orig­
The perceived measures of the neighbourhood environment are inal variables generated from NEWS scores were further converted into
derived using items in an abbreviated version of Neighbourhood Envi­ categorical variables. Considering that the non-land-use subscales of
ronment Walkability Scale (NEWS-A) (Cerin et al., 2006), which was NEWS measures (the 5 dimensions derived from CFA) are related with
modified in order to capture particular features of the built environment the comfort, convenience and safety of spaces connecting between
in Singapore. Our survey collected information on respondents’ per­ destinations, these measures were further classified through a two-step
ceptions of the following dimensions/subscales of built environment K-means cluster analysis (Makles, 2012).4 The result detects a
characteristics: 1) land-use mix; 2) street connectivity; (3) pedes­ two-class optimal solution to the participants’ overall ratings of the
trian/cycling infrastructure (Hanibuchi et al., 2011); (4) aesthetics of neighbourhoods’ connecting spaces (i.e. high vs low), which implies
neighbourhood surroundings; (5) neighbourhood safety; and (6) ame­ that the effects of non-land-use subscales of NEWS measures are repre­
nities of the neighbourhood. The first dimension was assessed by 16 sented by a categorical variable with dichotomous values. This cate­
survey questions that rate respondents’ perceptions on the presence of gorical perception variable is interacted with all the objective measures
various destinations (e.g. public transport hubs, shops, parks) within of pedestrian infrastructure and connections (street, footpath, park
10–15 min walk from home in a binary way (Yes = 1, No = 0).3 The connectors, crossings, etc.). In line with the results for the non-land-use
other 5 subscales were assessed by 23 questions capturing respondents’ measures, perceived measures of both utilitarian and recreational land
ratings of pedestrian infrastructure and connecting spaces between use mix indices are also dichotomized using median splits. The dichot­
neighbourhood destinations based on a 5-point Likert scale from 1 omized land use indices and transit access perception variables are
(strongly disagree) to 5 (strongly agree). Following Cerin et al. (2006), a interacted with the related objective measures of land use and transit
confirmatory factor analysis (CFA) was conducted on the items for the access, respectively.5 Moreover, the objective measure of residential
latter 5 dimensions of the NEWS and the results show that the mea­ density is interacted with all the dichotomized perception variables
surement model with the 5 defined factors yields a good fit to the data
(Cerin et al., 2006) (see Table 1a in the appendix).
The perceived land use mix measures were not incorporated into the 4
Our preliminary analysis also used all the five non-land-use measures of
NEW items to generate interaction terms with each objective measure of street/
pedestrian infrastructure and included these interactive variables in the
regression analysis. However, the results were unstable and difficult to interpret
3
We did not collect respondents’ perceptions on density because Singapore is because of the high correlations among the interaction terms.
5
characterized by high-density development that is well recognized by its local The descriptive statistics of original subjective built environment variables
residents and its perceived measure would yield little variations. by the dichotomous categories are shown in Table 2a in the appendix.

319
Y. Hou et al. Transport Policy 99 (2020) 314–328

since high-density places are usually more likely to have good transit Table 2
service and various destinations nearby with well-connected streets and Descriptive statistics of socio-demographic characteristics and travel outcome
pedestrian infrastructure (Sallis et al., 2012). measures (n = 900).
Table 1 shows the definition and descriptive statistics of subjective Variables Mean Std. Min Max
and objective variables and their interaction terms. Dev.

Dependent variables (trip frequency)


3.3.3. Socio-demographic characteristics of individuals Number of walking trips per day 3.59 2.44 0 14
Individual and households’ socio-demographic characteristics are Number of public bus trips per day 1.70 1.78 0 14
Number of MRT/LRT trips per day 1.56 1.89 0 14
controlled as background variables in this study. These data are also
Number of trips per day by private motorized 0.99 1.95 0 20
extracted from our resident survey. The socio-demographic variables are modes (car/motorcycle)
found to be weakly correlated with the perceived built environment Socio-demographic characteristics
variables defined above (the polychoric correlation coefficients ranging Age
between − 0.28 and 0.29).6 This implies that individual perceptions of 55-64 (reference) 0.43 0.50 0 1
65–74 0.32 0.47 0 1
the neighbourhood environment may play a relatively independent role 75–84 0.21 0.41 0 1
from individual/household characteristics in influencing travel 85+ 0.03 0.18 0 1
behaviour. Gender
The descriptive statistics for the chosen individual characteristics Male (reference) 0.4 0.49 0 1
Female 0.6 0.49 0 1
and travel outcome variables are illustrated in Table 2.
Ethnicity
Chinese 0.83 0.38 0 1
4. Results Malay 0.10 0.30 0 1
Indian 0.06 0.23 0 1
Table 3 reports the regression results on the number of trips per day Other (reference) 0.02 0.12 0 1
Education
by different modes of transport. The likelihood ratio (LR) statistics is No qualification 0.15 0.35 0 1
significant at p < 0.001 level, indicating that our model as a whole is Primary 0.33 0.47 0 1
statistically significant, as compared to the null model with no pre­ Lower secondary 0.28 0.45 0 1
dictors. We also find that for the same individuals, the cross-equation Post-secondary (non-tertiary) 0.12 0.32 0 1
Polytechnic 0.03 0.16 0 1
error correlations are all significant (at p < 0.05 level). The positive
Professional Qualification and other Diploma 0.02 0.14 0 1
correlations between walk, bus, and train trip-making suggest that the University 0.05 0.22 0 1
three modes are complements to each other, with the complementary Others (reference) 0.03 0.17 0 1
effects being the strongest between bus and train trips (ρ = 0.405). Work status
Meanwhile, the relationships between the above three modes and car/ Working full-time 0.25 0.43 0 1
Working part-time 0.15 0.35 0 1
motorcycle travel are likely to be substitutes and the most negative Working without pay in a family or other 0.002 0.05 0 1
correlation coefficient is found between bus and private motorized trip- organizations
making (ρ = − 0.345). The threshold parameters are found to be Not working (but able to work) 0.03 0.17 0 1
significantly different from each other in each equation, suggesting that Not working & not able to work (reference) 0.02 0.14 0 1
Retired 0.31 0.46 0 1
the number of daily trips is appropriately categorized to represent
Homemaker 0.24 0.43 0 1
different levels of utility (not reported in the table). Occupation
Because the signs of the estimated coefficients do not directly Professional, Manager, Executive, 0.09 0.28 0 1
represent the direction of effects of the corresponding explanatory var­ Other white collar workers 0.11 0.31 0 1
iables on the intermediate outcomes (Wooldridge, 2002), we also esti­ Sales or service workers 0.08 0.27 0 1
Blue collar worker 0.27 0.44 0 1
mate the (average) marginal probability effects on daily trip frequency
Others/Not know (reference) 0.45 0.50 0 1
at different levels of travel outcomes. The estimated marginal effects on Household income
the probability of travel outcome at levels higher than the average Below $2000 per month 0.43 0.50 0 1
values are found to have the same signs as the ordered probit co­ $2000 to $3999 per month 0.19 0.39 0 1
$4000 to $5999 per month 0.11 0.31 0 1
efficients, while those effects on trip-making below the average levels
$6000 to $9999 per month 0.06 0.24 0 1
usually have the opposite signs. Thus, to ease explanation, the following $10,000 and above 0.04 0.19 0 1
discussions mainly focus on the marginal effects on travel outcomes Declined/Don’t known (reference) 0.17 0.37 0 1
close to the average levels, which are 3–4 trips per day by walking, 2 Type of dwelling
trips per day by bus or MRT/LRT, and 1 trip per day by private motor­ HDB 1-Room 0.02 0.15 0 1
HDB 2-Room 0.06 0.24 0 1
ized modes. We set the significance level at α = 0.05 and focus on those
HDB 3-Room 0.49 0.50 0 1
significant variables. HDB 4-Room 0.25 0.44 0 1
HDB 5-Room or Executive Flat 0.08 0.27 0 1
4.1. Individual characteristics Condominium or Private Flat 0.06 0.24 0 1
Landed Property 0.02 0.14 0 1
Other (reference) 0.01 0.11 0 1
Results for age groups suggest that the oldest-old residents (85 years Dwelling owned/rented
and above) are the least active age group in terms of daily walking and Owned 0.95 0.22 0 1
train trip-making, followed by the mid-old group (75–84 years old). Not owned (reference) 0.05 0.22 0 1
Time living in the neighbourhood
Compared to the emerging-old (55–64) group, the oldest-old (85+) and
Less than 5 years (reference) 0.07 0.25 0 1
the mid-old (75–84) groups are significantly less likely to undertake an 5–19 years 0.29 0.46 0 1
above average level of walking trips per day (Y = 4) by 7.6% and 3.7%, 20–29 years 0.20 0.40 0 1
respectively, all else being equal. As such, it could be inferred that older 30–39 years 0.26 0.44 0 1
people’s ability to take frequent walks may be inhibited significantly 40 or more years 0.19 0.39 0 1
Household size 2.89 1.50 1 9
Household structure
Lives alone 0.17 0.38 0 1
6
See Uebersax (2015) on the calculation of polychoric correlation (continued on next page)
coefficients.

320
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 2 (continued ) favour bus travel that their probability of making 2 and more trips is on
Variables Mean Std. Min Max average 8.9% higher than other groups.
Dev. Among older residents of all types of dwelling units, those living in 1-
Lives with spouse 0.55 0.50 0 1
room HDB flats7 are least likely to be train travellers that their pro­
Living with children but without spouse 0.17 0.38 0 1 pensity of making any train trips per day is on average lower than other
Other 0.10 0.30 0 1 residents by 33.7%. Those living in 2-room and 3-room HDB flats are on
Needs aids or equipment when move outside average less likely to make any trips by car/motorcycle per day than
Yes (reference) 0.09 0.28 0 1
other groups by 34.9% and 26.5%, respectively. Those living residing
No 0.91 0.28 0 1
within the landed property, by contrast, are found to be infrequent bus
travellers that their chances of making no bus trip per day are higher
after the age of 75, possibly due to the worsening of their physical health than HDB or other flat dwellers.
conditions. There is also a steady decrease in the daily train trip fre­ The number of years residing in the current dwelling units is
quency as older adults’ age increases: for the 65–74, 75–84, and 85+ observed to be negatively associated with train trip-making frequency.
groups, their respective probability of making 2 trips by train per day are Compared with those living within 5 years of their residence, those with
4.3%, 9.2% and 15.5% lower than that for the 55–64 group (at p < longer living years have significantly lower propensity of making 2 or
0.001). However, no significant differences in bus or car/motorcycle more train trips per day, with the estimated marginal probability
daily trip frequency have been found among various age groups. ranging between 11% and 15%. It is possible that those who have lived
Significant gender differences in daily bus and private motorized in their neighbourhood for longer years are more familiar with the
trip-making are observed. On average, older females have 9.3% lower neighbourhood surroundings and less often utilize services and facilities
chances of making at least 1 trip by car/motorcycle per day than older outside their neighbourhoods for daily purposes by MRT/LRT, which are
males, all else equal, while their chances of making at least 2 bus trips usually for long-distance travel. Moreover, those living 30 years or more
per day are 9.1% higher. No significant differences are observed in the in their current residence are found with significantly lower propensity
daily trip-making by various modes of transportation among older of making frequent bus trips than their cohorts living less than 30 years
adults of different educational qualifications. of their current residence.
Different ethnic groups are not found to differ in their daily trip- Household size is negatively associated with the probability of
making by walking, bus, or private motorized vehicles. In terms of making more bus trips per day (Y ≥ 2), with an estimated average
train travel, the Malays are significantly more likely to make no trip per marginal effect of 3.5%. However, its effects on no daily trip-making by
day by MRT/LRT than their cohorts of other ethnic groups by 21.4%, all other travel modes are not observed.
else equal.
Employment status is found to impose significant impact on walking 4.3. Perceptions of built environment
and car/motorcycle daily travel. Compared with those who are not able
to work, the average chances for those who have working abilities but As discussed in Section 3, four aspects of individual perceptions of
are not working, those retired, and homemakers to make 3 or more the residential environment are summarized: convenience and safety of
walking trips per day are significantly higher by 21.6%, 25.9%, and pedestrian connections, availability of public transport hubs, access to
21.8%, respectively. Older adults who are working full time, on the utilitarian facilities (e.g. shops, community centres) and recreational
other hand, are more likely to make at least 1 trip per day by private facilities (gyms, parks and open spaces). The estimated results indicate
motorized modes by 18.3% than those of other employment status that the four aspects of perceived built environment measures exert
groups. independent effects on daily trip-making that vary across travel modes.
Differences in daily travel demands are also observed across older Older adults who perceived their neighbourhoods’ pedestrian con­
people of different occupations. Those who are professionals, managers, nections to be highly walkable and safe are on average more likely than
executives, or businessmen (PMEBs) exhibit lower preferences for bus those with lower perceptions of their neighbourhood connecting spaces
travel that their probability of making 2 or more bus trips per day is to undertake bus and train trips at an above-average level (Y ≥ 2) by
lower by 12.9%, compared with other occupation groups (at p < 0.05). 2.8% and 2.5%, respectively. Those who perceive the availability of
This could potentially be explained by the tight work schedules of those transit services within walking distance are 13% more likely to make at
PMEB jobs that do not allow for frequent motorized travel with slower least 2 trips per day by train than those who do not perceive the avail­
speeds. ability of nearby transit services. Higher perceptions of access to utili­
Compared with those who rely on walking aids when travelling tarian facilities seem to render frequent daily train trips less important,
outside home, older respondents who do not need walking aids exhibit which suggests that older people’s reliance on long-distance travel by
significantly higher likelihood of undertaking more frequent daily trips train for reaching daily services and facilities outside their neighbour­
by walking (Y ≥ 4), bus and train (Y ≥ 2), with the estimated marginal hoods would be reduced if they perceive nearby these facilities as
effects for the three modes being 18.6%, 20.9%, and 19.7%, respec­ accessible. Moreover, higher perceived access to recreational facilities
tively. The differences in travel by car/motorcycle between the two plays an important role in increasing older adults’ propensity of making
groups of older people are not significant. more frequent walking trips per day.

4.2. Household characteristics 4.4. Objective measures of the built environment and their interactions
with corresponding perceived measures
Older adults with the lowest household income level (below $2000
per month) seem not to be able to afford frequent car/motorcycle trips, The effects of built environment characteristics are examined
which are likely to be more expensive than other modes. On average, through five dimensions: residential density, street connectivity and
their probability of making a trip by car/motorcycle per day is 8% lower design, walking infrastructure, and access to retail and service facilities.
than other groups. In contrast, these older adults are more likely to Except for their independent travel impacts, interactive effects between

7
These are public housing units built by the Housing and Development Board
(HDB) and popularly referred to as ‘HDB flats’. They comprise different flat
types, size and rooms.

321
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 3
Results on daily trip frequency.
Walk trips Bus trips MRT/LRT trips Private motorized trips

Socio-demographic characteristics of individuals


Age bands (reference: 55–64)
65–74 0.009 0.116 ¡0.262** − 0.0404
(0.09) (1.18) (-2.59) (-0.37)
75–84 ¡0.413** − 0.103 ¡0.494*** 0.0814
(-3.24) (-0.79) (-3.66) (0.55)
85+ ¡0.763** − 0.431 ¡0.756** 0.115
(-3.21) (-1.71) (-2.76) (0.41)
Gender (reference: Male)
Female − 0.0818 0.261** 0.147 ¡0.286**
(-0.92) (2.86) (1.56) (-2.80)
Ethnicity (reference: others)
Chinese − 0.287 − 0.377 − 0.352 0.390
(-0.94) (-1.22) (-1.07) (1.09)
Malay − 0.465 − 0.309 ¡0.688* 0.567
(-1.44) (-0.94) (-1.97) (1.51)
Indian − 0.175 − 0.255 − 0.461 0.620
(-0.52) (-0.74) (-1.26) (1.58)
Work Status (reference: not working because of disability)
Working full-time 0.557 − 0.0525 0.0116 0.596*
(1.81) (-0.17) (0.04) (2.01)
Working part-time 0.520 0.0974 0.0465 0.386
(1.66) (0.30) (0.14) (1.01)
Working without pay in a family business − 0.260 − 0.878 − 1.191 1.378
(-0.31) (-0.91) (-1.31) (1.67)
Not working (but able to work) 0.798* − 0.266 − 0.208 0.400
(2.05) (-0.74) (-0.57) (0.95)
Retired 0.792** 0.0833 0.107 0.376
(2.70) (0.28) (0.35) (1.05)
Homemaker 0.673* − 0.181 − 0.378 0.362
(2.31) (-0.60) (-1.24) (1.02)
Occupation (reference: others)
Professional, Manager, Executive, Businessman 0.0611 ¡0.366* − 0.280 − 0.0277
(0.35) (-2.03) (-1.53) (-0.14)
Other white collar workers 0.0369 − 0.055 − 0.067 0.182
(0.23) (-0.34) (-0.41) (1.02)
Sales or service workers 0.303 0.0607 − 0.198 0.142
(1.72) (0.34) (-1.06) (0.72)
Blue collar worker 0.0124 − 0.132 − 0.259 − 0.244
(0.09) (-0.95) (-1.80) (-1.54)
Walking aid (reference: walking aid required)
No_aids needed 0.585*** 0.583*** 0.579*** − 0.252
(3.97) (3.75) (3.47) (-1.50)
Household characteristics
Income per month (reference: unknown/decline to supply)
Below $2000 − 0.120 0.252* 0.104 − 0.247*
(-0.97) (1.97) (0.80) (-2.05)
$2000 to $3999 − 0.174 0.196 − 0.0634 0.0591
(-1.24) (1.35) (-0.43) (0.37)
$4000 to $5999 − 0.020 0.150 0.101 0.0333
(-0.13) (0.93) (0.61) (0.19)
$6000 to $9999 − 0.276 0.271 0.0703 − 0.0571
(-1.45) (1.38) (0.35) (-0.27)
$10,000 and above abovemonth − 0.213 0.0524 − 0.444 0.174
(-0.81) (0.19) (-1.57) (0.63)
Dwelling units type (reference: others)
HDB 1-Room 0.245 − 0.734 ¡1.072*** − 0.727
(0.52) (-1.49) (-3.43) (-1.35)
HDB 2-Room − 0.227 − 0.093 0.244 ¡1.051*
(-0.57) (-0.23) (0.58) (-2.51)
HDB 3-Room in a family business − 0.0386 − 0.202 0.231 ¡0.774*
(-0.11) (-0.54) (0.61) (-2.10)
HDB 4-Room − 0.00068 − 0.218 0.0747 − 0.590
(-0.00) (-0.59) (0.20) (-1.65)
HDB 5-Room or Executive Flat 0.176 − 0.052 0.291 − 0.551
(0.47) (-0.13) (0.74) (-1.45)
Condominium or Private Flat 0.553 − 0.291 0.786 − 0.715
(1.08) (-0.54) (1.47) (-1.33)
Landed Property − 0.731 ¡1.323* − 0.923 − 0.386
(-1.23) (-2.16) (-1.43) (-0.61)
Years living in the house
5–19 years 0.0715 − 0.332 ¡0.414* 0.0791
(0.43) (-1.95) (-2.37) (0.42)
20–29 years 0.234 − 0.308 ¡0.364* 0.0874
(continued on next page)

322
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 3 (continued )
Walk trips Bus trips MRT/LRT trips Private motorized trips

(1.33) (-1.72) (-1.99) (0.44)


30–39 years 0.106 ¡0.374* ¡0.415* − 0.0215
(0.61) (-2.14) (-2.31) (-0.11)
40 or more years 0.225 ¡0.375* ¡0.474* − 0.348
(1.24) (-2.03) (-2.51) (-1.61)
Household size − 0.0212 ¡0.0997** − 0.00983 0.036
(-0.66) (-2.97) (-0.29) (1.00)
Walking aid (reference: require walking aid)
No_aids needed 0.585*** 0.583*** 0.579*** − 0.252
(3.97) (3.75) (3.47) (-1.50)
Subjective measures of the built environment
Perceptions of connecting spaces/pedestrian facilities (reference: low level)
High leve 0.344 1.634* 1.576* − 0.807
(0.49) (2.21) (2.00) (-0.99)
Perceptions of access to transit hubs (reference: low level)
High level − 0.322 1.540 2.850** 0.458
(-0.36) (1.59) (2.70) (0.44)
Perceptions of access to utilitarian facilities (reference: low level)
High level − 0.827 − 0.713 ¡1.532* 1.228
(-1.39) (-1.18) (-2.45) (1.78)
Perceptions of access to recreational facilities (reference: low level)
High level 1.076* − 0.180 0.512 0.528
(2.14) (-0.35) (0.98) (0.96)
Objective measures × Subjective measures of the built environment
Street connectivity
cnt400_3wayJun 0.0208 0.0505* 0.0508* − 0.0199
(0.88) (2.07) (2.00) (-0.74)
___ × high perceptions of pedestrian connection 0.00541 − 0.00648 − 0.0237 0.00362
(0.21) (-0.24) (-0.84) (0.12)
Pedestrian infrastructure
cnt400_ rdcnolght ¡0.244*** 0.00421 0.0831 − 0.109
(-3.39) (0.06) (1.09) (-1.34)
___ × high perceptions of pedestrian connection 0.0397 0.0411 0.00988 − 0.0181
(0.48) (0.49) (0.11) (-0.19)
len400_covftpth 0.00149** − 0.000372 0.0000837 − 0.00111
(2.65) (-0.62) (0.14) (-1.72)
___ × high perceptions of pedestrian connection − 0.00015 − 0.000177 − 0.00013 0.000621
(-0.29) (-0.32) (-0.23) (1.02)
Distance to public transit hubs
nrdist_busstp 0.00394 0.00628 0.0143** − 0.004
(0.99) (1.44) (2.93) (-0.87)
___ × perceived availability of transit access − 0.00348 − 0.00781 ¡0.0135** 0.00354
(-0.87) (-1.78) (-2.75) (0.76)
Distance to utilitarian destinations
nrdist_wetmkt ¡0.000893** ¡0.000497* ¡0.000858** 0.000126

(-3.24) (-2.02) (-2.91) (0.38)


___ × high perceptions of access to utilitarian facilities 0.0000617 0.0000294 0.000435* − 0.000296
(0.32) (0.15) (2.13) (-1.37)
nrdist_plwrshp ¡0.00188* − 0.000238 − 0.000367 0.000551
(-2.54) (-0.31) (-0.46) (0.63)
___ × high perceptions of access to utilitarian facilities 0.000626 0.000423 0.00121 − 0.000221
(0.74) (0.49) (1.34) (-0.23)
Distance to recreational facilities
nrdist_ sptfac − 0.000513 − 0.000782 − 0.000888 0.0000515
(-1.12) (-1.66) (-1.82) (0.10)
___ × high perceptions of access to recreation facilities ¡0.00115** 0.000268 − 0.0000877 0.000544
(-3.17) (0.72) (-0.23) (1.36)
nrdist_open 0.000858 0.000862 0.000416 − 0.000387
(1.51) (1.47) (0.68) (-0.60)
___ × high perceptions of access to recreation facilities ¡0.00211*** 0.0000171 0.00027 − 0.000198

(-3.47) (0.03) (0.42) (-0.30)


nrdist_park ¡0.00194** − 0.000455 0.0000401 − 0.0000342
(-2.93) (-0.67) (0.06) (-0.05)
___ × high perceptions of access to recreation facilities − 0.00111 − 0.000501 − 0.00116 0.000183
___ * high perceptions of access to recreational facilities (-1.53) (-0.68) (-1.51) (0.23)
Correlation matrix of the residuals
Walk trips Bus trips MRT/LRT trips Private motorized trips
Walk trips 1

Bus trips 0.282*** 1
(7.71) –
MRT/LRT trips 0.224*** 0.405*** 1
(5.83) (11.41) –
Private motorized trips − 0.13*** − 0.345*** − 0.246*** 1
(-2.95) (-7.73) (-5.19) –

323
Y. Hou et al. Transport Policy 99 (2020) 314–328

N = 900; Log-likelihood = − 4541.24; LR chi2(340) = 857.61.


T statistics in parentheses; *, p < 0.05; **, p < 0.01; ***, p < 0.001.
1
The 400-metre buffer zone is applied for most measures of transportation infrastructure because 400-metre is the current planning service radius in Singapore.

relatively matched objective and subjective measures of the built envi­ purposes by, for example, protecting them from adverse weather
ronment are also expected on older adults’ daily trip frequency. The conditions.
estimation results again indicate differential effects of various di­ The estimated effects of access to bus stops on travel by MRT/LRT
mensions of the built environment on daily trip frequency by different show unexpected signs and vary by individuals’ perceptions of transit
modes. access. However, we find that most of the surveyed respondents are
Residential density and its interaction terms with the four perceived actually living within a 200-metre buffer of each bus stop/interchange
measures of the built environment are not found to have significant and the difference in the actual proximity to bus stops/interchanges
travel impacts. This might be because the residential density within the from homes between those who perceive the availability of transit ser­
3 study neighbourhoods is already high and do not have much variation. vice nearby and who do not perceive as such is not statistically signifi­
A more grid-like street pattern within 400 m of residences is found to cant as indicated by the t-test (result not shown). Thus, the result is
increase older adults’ probability of making bus or train trips at an explained in another way as how the gaps in rail transit usage between
above-average level (Y ≥ 2), but the estimated marginal effects are very the two groups of different perceived accessibility to transit are likely to
small (less than 2%). Given the positive and significant correlations be narrowed as distance to bus stops from home increases. For example,
between walking and transit trips, it might be inferred that bus and train for those living 100 metres away from bus stations, the differences in the
trips are more favoured by older people when the street network nearby predicted probability of making 2 train trips per day between those who
their homes give them more direct route choices for walking to the perceive the availability of transit service and who do not is as high as
transit stations. The effects of street connectivity do not differ signifi­ 15.7%, while for those living at the 130 metres distance from the nearest
cantly by individual perceptions of connecting spaces in their bus services, the gap in the train trip-making propensity at 2 trips per
neighbourhoods. day was reduced to 7.2% (see Fig. 2). Beyond the 140-m buffer of bus
Pedestrian infrastructure examined here includes pedestrian cross­ stations, however, the difference in train usage at an average level of 2
ings with or without traffic lights for pedestrians, footpaths with or trips per day between the two groups is still significant but with a
without shelters (covers), and proximity to cycling path. Results indicate magnitude of less than 5%. The results imply that increasing people’s
that older adults’ propensity of making more frequent walking trips per awareness of their neighbourhoods’ transit access would promote their
day are hindered by pedestrian crossings without any pedestrian traffic propensity of MRT/LRT to a considerable degree (e.g. by more than 5%)
lights within 400 m of their homes; every additional crossing of that kind only if they live within close proximity of bus stops/interchanges.
are associated with 1.3% increase in the probability of making zero walk Proximity to wet markets is significantly associated with an
trip per day and 6.1% increase in the probability of making only 1–2 increased propensity to make more frequent walking and bus trips per
walking trips per day, all else equal. The insignificant coefficient for the day among older adults of all levels of perceived accessibility to retail
interaction term implies that negative effects of those unsafe pedestrian and service facilities. Assuming that wet markets are “brought closer” to
crossings on walking are not moderated by individuals’ positive evalu­ the older residents’ homes by 100 metres, for example, from the mean
ations of the pedestrian facilities in their neighbourhood. value of 1183 to 1083 metres, the predicted probability of making 4 and
Covered walkways are also found to play positive roles in older more walking trips per day would on average increase by about 2.8%
adults’ daily walking travel regardless of individuals’ perceptions of and the probability of making 2 and more bus trips per day would in­
their neighbourhoods’ pedestrian connections. On average, every 100 m crease by about 1.6%.
increase of those walkways within a 400-m buffer of each residence is In terms of daily trip frequency by MRT/LRT, the effects of proximity
associated with 5% increase in the propensity of making more frequent to wet markets are also found to be significantly positive but vary in
walking trips per day (Y ≥ 4). This result implies that in Singapore where magnitude based on older residents’ perceived measures of access to
its tropical climate may pose frequent challenges such as hot tempera­ local utilitarian destination. For example, as distance to the nearest wet
ture and heavy rain, an extended network of footpath with covers can be market changes in the same range as above, the predicted probability of
important factors in encouraging older people to walk for transport making 2 and more train trips per day on average increases by 2.8% for

Fig. 2. Predicted probability of making 2 trips by MRT/LRT per day vs. distance to nearest bus stops/interchanges (m).

324
Y. Hou et al. Transport Policy 99 (2020) 314–328

older adults with higher perceptions of accessibility to nearby utilitarian built environment characteristics moderate each other’s effects on
facilities and 1.4% for those with lower perceptions. However, the gaps older adults’ travel by different modes, this study adds to existing
in the propensity of making 2 MRT/LRT trips per day between older literature on the built environment-travel links and older people’s
adults of higher and lower levels of perceived access to local utilitarian mobility.
destinations are very small in general (less than 5%) and do not change Using a multivariate OP model to account for the correlations of
very much with distance to wet markets (see Fig. 3). travel modes within each person, our results indicate that different di­
Places of worship seem to be another favoured walking trip desti­ mensions of perceived and objective built environment characteristics
nations for all older adults. For example, if the average distance between exert both independent and interactive effects on older adults’ daily trip-
homes and religious places is reduced by 100 metres, the average in­ making and their effects vary by travel modes. While better perceptions
crease in the probability of making 4 and more walking trips is about of access to recreational facilities significantly increases older adults’
5%. The positive effects of access to religious destinations on transport daily walking trip-making and enhances their travel responses to actual
walking do not vary by individuals’ subjective measure of accessibility proximity to those facilities by walking, the positive role of access to
to utilitarian destinations. The effects of proximity to other retail and utilitarian destinations such as wet markets and religious establishments
service establishments on trip frequency by various modes of trans­ in older adults’ walking trip frequency is not moderated by individual
portations are not found to be statistically significant for either perceptions of access to these destinations. Better perceived access to
perception groups. transit hubs is positively associated with older adults’ daily trip fre­
Access to recreational facilities is examined in relation to the dis­ quency by rail transit. Moreover, perceptions of barrier-free and safe
tance to three types of amenities, including the parks/green spaces, open pedestrian connections encourage daily trip-making by public bus and
spaces, and sports facilities. Consistent with the corresponding subjec­ MRT/LRT. This result complements previous findings of the negative
tive measures, proximity to parks and green spaces are observed to be role of undesirable perceptions of personal and traffic safety in transit
associated with more frequent walking trips per day for older people of usage (Lee, 2013). The actual provision of comfortable and safe pedes­
both high and low perception groups. With the average distance be­ trian infrastructure is found to play important roles in influencing older
tween homes and the nearest parks/green spaces decreasing by 100 adults’ daily travel choices: covered walkways near residence that pro­
metres, the average increase in the probability of making at least 4 vide shade and protection from wet weather significantly increases older
walking trips per day is expected to be as high as 8.1%. adults’ trip frequency by walking and decreases their car trip-making. In
On the other hand, the positive effects of access to open spaces and contrast, more unsignalized pedestrian crossings near residence signif­
indoor sports facilities are only evident for those who report higher icantly reduces daily walking travel, implying that these facilities would
perceptions of accessibility to the local recreational facilities. For these be a potential barrier to traffic safety for older pedestrians.
older adults, the estimated marginal probability of a 100-metre decrease The results of this study provide meaningful insights for designing
in the distances between homes and the two types of recreational fa­ and implementing healthy travel-promoting interventions (e.g. walking
cilities on walking trip frequency is 4.2% and 5.5%, respectively, at the and public transit) among older adults. In addition to the suite of
mean value of the two distance variables. For those who do not consider frequently cited interventions such as the provision of comfortable and
local recreational facilities as easily accessible, the actual proximity to attractive pedestrian infrastructure (e.g. sheltered walkways and safe
open spaces and sports facilities would not have significant impacts on road crossings), mixed land use and daily amenities (e.g. markets)
them. within a walkable distance (Cerin et al., 2013a; Ewing and Cervero,
2010; Sallis et al., 2012), our findings also highlight the potential of a
5. Conclusion more efficient and maybe easier way to enhance older adults’ travel
experience by fostering positive perceptions towards the existing phys­
In light of the common trend among older adults to become inactive ical environment and amenities within their neighbourhoods. For
and sedentary with age, there is renewed interest on the neighbourhood example, urban planners and designers could consider introducing
built environment, a place that people regularly interact with across walking maps, wayfinding signage, directional markings, or street
their lifespan. Using neighbourhood level data, this study examines paintings on the first and last mile pathways to public transit hubs to
older adults’ interactions with the built environment in their daily travel improve older adults’ perceptions of the safety, comfort and aesthetics of
activities. By investigating how perceived and objective-determined those walking connections and consequently increase their transit usage.

Fig. 3. Predicted probability of making 2 trips by MRT/LRT per day vs. distance to the nearest wet market (m).

325
Y. Hou et al. Transport Policy 99 (2020) 314–328

Based on previous findings (Cheng and Chen, 2015), maintaining a city-state of Singapore as well as in multi-country investigations of
safe walking environment would also be important to enhance older metropolitan areas with different development patterns to clarify the
residents’ perceived access to available facilities (e.g. transit hubs, relationships between neighbourhood environment and older adults’
open/green space), which encourages more walking and transit trips for travel behaviour.
them. Another important determinant of perceived ease of access is
neighbourhood walking conditions. Design inclusivity through adoption Ethics approval
of barrier-free standards (manageable stair riser elevation, slip-resistant
tiling, supporting handrails, etc.) can ameliorate perceived transport The research protocol was approved by the Singapore University of
disadvantages and help improve perceived access to transit and other Technology and Design Institutional Review Board.
facilities for older adults (Hine and Mitchell, 2001). Other initiatives
and/or design strategies aiming at enhancing the legibility of transit CRediT authorship contribution statement
hubs and other facilities in the neighbourhood would further promote
older adults’ perceptions of local accessibility and more healthy travel. Yuting Hou: Conceptualization, Methodology, Software, Formal
However, future research is needed to better understand the causal analysis, Investigation, Data curation, Writing - original draft, Writing -
relationship between the built environment and travel behaviour among review & editing, Visualization. Winston Yap: Writing - original draft,
older adults using longitudinal data incorporating older residents’ travel Writing - review & editing. Rochelle Chua: Writing - original draft. Siqi
and residential preferences in the analysis. Due to the cross-sectional Song: Validation, Writing - review & editing. Belinda Yuen: Resources,
nature of the analysis, the results indicate only an association instead Writing - review & editing, Supervision, Project administration, Funding
of causality. Given the complexity of the subject being studied, it is acquisition.
difficult to attribute the effects of built environmental factors as causing
the travel outcome without sequential evidence. The potential effects of Declaration of competing interest
residential self-selection, where older adults who prefer walking tend to
live in neighbourhoods that are more walkable and therefore walk more, The authors have no conflict of interest to declare.
cannot be ruled out either. Furthermore, the findings and implications
are drawn from only 3 local neighbourhoods of Singapore. While every Acknowledgements
effort has been made to ensure a heterogeneous and comparable selec­
tion, the diversity of different communities and urban environments This research is supported by the Singapore Ministry of National
across cities means that other locational factors (either regional or Development and the National Research Foundation, Prime Minister’s
community-based) could play a role in influencing older adults’ travel Office under the Land and Liveability National Innovation Challenge (L2
pattern. Further research is needed to test this proposition. For other NIC) Research Programme (L2 NIC Award No. L2NICTDF1-2017-2).
metropolitan areas with car-oriented infrastructure design and mono- Any opinions, findings and conclusions or recommendations
functional land use, we expect that the interactive effects of built envi­ expressed in this material are those of the authors and do not reflect the
ronment features on daily travel pattern could be different. It would be views of the Singapore Ministry of National Development and National
interesting to perform similar analyses at national level, for the whole Research Foundation, Prime Minister’s Office, Singapore.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.tranpol.2020.06.017.

Appendix B

Table 1a
Confirmatory factor analysis for perceptions towards the neighbourhood build environment.

Factors Questions Loadings

Well-connected street network The streets in my neighbourhood are well-connected 0.720


The distance between intersections in my neighbourhood is usually short 0.675
There are many alternative routes for getting place to place in my neighbourhood 0.703

Pedestrian- & bicycle-friendly The footpaths in my neighbourhood are free of barriers 0.584
infrastructure The footpaths in my neighbourhoods are well-maintained 0.640
There are bicycle and/pedestrian trails (e.g. park connector networks) in or near my neighbourhood that are easy to get to 0.437
There are pedestrian crossings to help people cross busy streets in my neighbourhood 0.490
There are covered walkways in my neighbourhood that provide shade and protection from the weather 0.498
The street signs in my neighbourhood are readable making it easy for me to find my way between places 0.435

Clean & well-maintained surrounding There are trees along the footpaths in my neighbourhood that provide shade 0.526
There are many interesting things to look at while walking in my neighbourhood 0.416
My neighbourhood is generally free from litter and rubbish 0.562
My neighbourhood is generally free from bad smells and odours 0.516
My neighbourhood is generally free of loud noises 0.533
There are many attractive nature sights in my neighbourhood 0.434
There are landmarks in my neighbourhood that help me find my way between places 0.441

Safe neighbourhood There is not much traffic on streets near my home that it makes it difficult or unpleasant to walk in my neighbourhood 0.303
The speed of traffic on most streets near my home is usually slow 0.413
No crime in my neighbourhood makes it safe to go outdoors 0.310
(continued on next page)

326
Y. Hou et al. Transport Policy 99 (2020) 314–328

Table 1a (continued )
Factors Questions Loadings

I see and speak to other people when I am walking in my neighbourhood 0.381

Street furniture There are many places to meetup with friends within easy walking distance of my home 0.509
There are convenient public toilets that I can use when I am out and about in my neighbourhood 0.477
There are many seats in my neighbourhood, making it easy to sit and rest when I am out and about 0.592

Fit statistics
RMSEA, 0.047; CFI, 0.905; SRMR, 0.048

Table 2a
Descriptive statistics of (original) subjective built environment variables by dichotomized categories.

Variable All High level Low level

Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.

Perceptions of connecting spaces/pedestrian facilities (NEWS scores)

N=900 N=517 N=383

Well-connected street network 11.81 1.69 12.52 1.23 10.84 1.75


Pedestrian- & bicycle-friendly infrastructure 22.84 3.35 24.61 2.47 20.45 2.87
Clean & well-maintained surrounding 24.01 4.26 26.23 3.27 21.01 3.56
Safe neighbourhood 14.50 2.17 15.27 2.05 13.45 1.86
Street furniture 10.27 2.28 11.40 1.80 8.75 1.94

Perceived availability of retail/service facilities within 10-15 minute walking distance

N=900 N=477 N=423

Supermarket 0.77 0.42 0.93 0.26 0.59 0.49


Shopping/Entertainment centres 0.16 0.37 0.22 0.41 0.10 0.30
Transaction Services (e.g. bank, post office) 0.29 0.46 0.33 0.47 0.25 0.43
Senior Activity Centres / Daycare Centre Facilities 0.47 0.50 0.71 0.45 0.19 0.39
Wet/Dry Markets 0.94 0.25 0.99 0.12 0.88 0.33
Community Club/Centre 0.62 0.49 0.91 0.28 0.29 0.45
Place of Worship (e.g. Church, Temple, Mosque, etc.) 0.40 0.49 0.66 0.47 0.12 0.32
Medical facilities (e.g. Hospital/Polyclinic/Pharmacy) 0.48 0.50 0.76 0.43 0.17 0.38
Hawker centres / Coffee houses / Food courts 0.94 0.25 1.00 0.06 0.87 0.34
Fast-food Restaurants 0.55 0.50 0.80 0.40 0.26 0.44
Total number of utilitarian facilities perceived as available 5.62 2.15 7.29 1.15 3.72 1.25
Perceived availability of recreational facilities within 10-15 minute walking distance

N=900 N=438 N=462

Indoor fitness facility (e.g. Gym, Recreation Club, Swimming Complex, etc.) 0.31 0.46 0.61 0.49 0.03 0.17
Open public spaces (e.g. void decks, senior citizen corners, playgrounds) 0.68 0.47 0.96 0.20 0.42 0.49
Green spaces (public parks) 0.44 0.50 0.84 0.36 0.06 0.24
Total number of recreational facilities perceived as available 1.43 1.07 2.41 0.49 0.51 0.50

References sectional study on what destinations matter and when. Int. J. Behav. Nutr. Phys.
Activ. 10, 78.
Cerin, E., Macfarlane, D., Sit, C., Ho, S., Johnston, J., Chou, K.L., Chan, W., Cheung, M.,
Badoe, D.A., 2007. Forecasting travel demand with alternatively structured models of
Ho, K., 2013b. Effects of built environment on walking among Hong Kong older
trip frequency. Transport. Plann. Technol. 30, 455–475.
adults. Hong Kong Med. J. 19, S39–S41.
Baron, R.M., Kenny, D.A., 1986. The moderator–mediator variable distinction in social
Cerin, E., Saelens, B.E., Sallis, J.F., Frank, L.D., 2006. Neighborhood Environment
psychological research: conceptual, strategic, and statistical considerations. J. Pers.
Walkability Scale: validity and development of a short form. Med. Sci. Sports Exerc.
Soc. Psychol. 51, 1173–1182.
38, 1682–1691.
Bhat, C.R., Srinivasan, S., 2005. A multidimensional mixed ordered-response model for
Cervero, R., Kockelman, K., 1997. Travel demand and the 3Ds: density, diversity, and
analyzing weekend activity participation. Transp. Res. Part B Methodol. 39,
design. Transport. Res. Transport Environ. 2, 199–219.
255–278.
Centre for Liveable Cities and Urban Land Institute, 2014. Creating Healthy Places
Boarnet, M.G., Crane, R., 2001. The influence of land use on travel behavior:
Through Active Mobility. Retrieved from. https://www.clc.gov.sg/docs/default-sou
specification and estimation strategies. Transport. Res. Pol. Pract. 35, 823–845.
rce/books/active-mobility.pdf.
Boarnet, M.G., Sarmiento, S., 1998. Can land-use policy really affect travel behaviour? A
Cheng, Y., Chen, S., 2015. Perceived accessibility, mobility, and connectivity of public
study of the link between non-work travel and land-use characteristics. Urban Stud.
transportation systems. Transport. Res. Part A 77, 386–403.
35, 1155–1169.
Choo, S., Sohn, D., Park, M., 2016. Mobility characteristics of the elderly: a case for Seoul
Böcker, L., Van amen, P., Helbich, M., 2017. Elderly travel frequencies and transport
metropolitan area. KSCE J. Civ. Eng. 20, 1023–1031.
mode choices in Greater Rotterdam, The Netherlands. Transportation 44, 831–852.
Chudyk, A.M., Winters, M., Moniruzzaman, M., Ashe, M.C., Gould, J.S., Mckay, H., 2015.
Bracy, N.L., Millstein, R.A., Carlson, J.A., Conway, T.L., Sallis, J.F., Saelens, B.E., Kerr, J.,
Destinations matter: the association between where older adults live and their travel
Cain, K.L., Frank, L.D., King, A.C., 2014. Is the relationship between the built
behavior. J. Transp. Health 2, 50–57.
environment and physical activity moderated by perceptions of crime and safety?
Corseuil Giehl, M.W., Hallal, P.C., Brownson, R.C., D’orsi, E., 2017. Exploring
Int. J. Behav. Nutr. Phys. Activ. 11, 24.
associations between perceived measures of the environment and walking among
Cain, K.L., Millstein, R.A., Sallis, J.F., Conway, T.L., Gavand, K.A., Frank, L.D., Saelens, B.
Brazilian older adults. J. Aging Health 29, 45–67.
E., Geremia, C.M., Chapman, J., Adams, M.A., 2014. Contribution of streetscape
Davison, K.K., Lawson, C.T., 2006. Do attributes in the physical environment influence
audits to explanation of physical activity in four age groups based on the Microscale
children’s physical activity? A review of the literature. Int. J. Behav. Nutr. Phys.
Audit of Pedestrian Streetscapes (MAPS). Soc. Sci. Med. 116, 82–92.
Activ. 3, 19.
Cao, J., Mokhtarian, P., Handy, S., 2009. The relationship between the built environment
Ding, D., Sallis, J.F., Conway, T.L., Saelens, B.E., Frank, L.D., Cain, K.L., Slymen, D.J.,
and nonwork travel: A case study of Northern California. Transport. Res. Pol. Pract.
2012. Interactive effects of built environment and psychosocial attributes on
43 (5), 548–559. https://doi.org/10.1016/j.tra.2009.02.001.
physical activity: a test of ecological models. Ann. Behav. Med. 44, 365–374.
Cerin, E., Lee, K.-Y., Barnett, A., Sit, C.H., Cheung, M.-C., Chan, W.-M., Johnston, J.M.,
Ding, D., Sallis, J.F., Norman, G.J., Frank, L.D., Saelens, B.E., Kerr, J., Conway, T.L.,
2013a. Walking for transportation in Hong Kong Chinese urban elders: a cross-
Cain, K., Hovell, M.F., Hofstetter, C.R., 2014. Neighborhood environment and

327
Y. Hou et al. Transport Policy 99 (2020) 314–328

physical activity among older adults: do the relationships differ by driving status? relationships with transportation physical activity among older persons. Int. J.
J. Aging Phys. Activ 22, 421–431. Behav. Nutr. Phys. Activ. 12, 108.
Etman, A., Kamphuis, C.B., Prins, R.G., Burdorf, A., Pierik, F.H., Van lenthe, F.J., 2014. Orstad, S.L., Mcdonough, M.H., Stapleton, S., Altincekic, C., Troped, P.J., 2017.
Characteristics of residential areas and transportational walking among frail and A systematic review of agreement between perceived and objective neighborhood
non-frail Dutch elderly: does the size of the area matter? Int. J. Health Geogr. 13, 7. environment measures and associations with physical activity outcomes. Environ.
Ewing, R., Cervero, R., 2001. Travel and the built environment: a synthesis. Transport. Behav. 49, 904–932.
Res. Rec.: J. Transport. Res. Board 87–114. Owen, N., Humpel, N., Leslie, E., Bauman, A., Sallis, J.F., 2004. Understanding
Ewing, R., Cervero, R., 2010. Travel and the built environment: a meta-analysis. J. Am. environmental influences on walking: review and research agenda. Am. J. Prev.
Plann. Assoc. 76, 265–294. Med. 27, 67–76.
Ferdous, N., Eluru, N., Bhat, C.R., Meloni, I., 2010. A multivariate ordered-response Páez, A., Scott, D., Potoglou, D., Kanaroglou, P., Newbold, K.B., 2007. Elderly mobility:
model system for adults’ weekday activity episode generation by activity purpose demographic and spatial analysis of trip making in the Hamilton CMA, Canada.
and social context. Transp. Res. Part B Methodol. 44, 922–943. Urban Stud. 44, 123–146.
Figueroa, M.J., Nielsen, T.A.S., Siren, A., 2014. Comparing urban form correlations of the Pelclová, J., Frömel, K., Bláha, L., Zając-gawlak, I., Tlučáková, L., 2012. Neighborhood
travel patterns of older and younger adults. Transport Pol. 35, 10–20. environment and walking for transport and recreation in Central European older
Foster, S., Giles-Corti, B., 2008. The built environment, neighborhood crime and adults. Acta Gymnica 42, 49–56.
constrained physical activity: an exploration of inconsistent findings. Prev. Med. 47, Procter-Gray, E., Leveille, S.G., Hannan, M.T., Cheng, J., Kane, K., Li, W., 2015.
241–251. Variations in community prevalence and determinants of recreational and utilitarian
Frank, L., Kerr, J., Rosenberg, D., King, A., 2010. Healthy aging and where you live: walking in older age. J. Aging Res. 2015.
community design relationships with physical activity and body weight in older Roodman, D., 2011. Fitting fully observed recursive mixed-process models with cmp.
Americans. J. Phys. Activ. Health 7, S82–S90. STATA J. 11, 159–206.
Greene, W.H., Hensher, D.A., 2009. Modeling Ordered Choices. Roorda, M.J., Páez, A., Morency, C., Mercado, R., Farber, S., 2010. Trip generation of
Handy, S., Cao, X., Mokhtarian, P.L., 2006. Self-selection in the relationship between the vulnerable populations in three Canadian cities: a spatial ordered probit approach.
built environment and walking: empirical evidence from Northern California. J. Am. Transportation 37, 525–548.
Plann. Assoc. 72, 55–74. Saelens, B.E., Handy, S.L., 2008. Built environment correlates of walking: a review. Med.
Hanibuchi, T., Kawachi, I., Nakaya, T., Hirai, H., Kondo, K., 2011. Neighborhood built Sci. Sports Exerc. 40, S550.
environment and physical activity of Japanese older adults: results from the Aichi Sallis, J.F., Cervero, R.B., Ascher, W., Henderson, K.A., Kraft, M.K., Kerr, J., 2006. An
Gerontological Evaluation Study (AGES). BMC Publ. Health 11. ecological approach to creating active living communities. Annu. Rev. Publ. Health
Hiine, J., Mitchell, F., 2001. Better for everyone? Travel experiences and transport 27, 297–322.
exclusion. Urban Stud. 38, 319–332. Sallis, J.F., Floyd, M.F., Rodríguez, D.A., Saelens, B.E., 2012. Role of built environments
Hoehner, C.M., Ramirez, L.K.B., Elliott, M.B., Handy, S.L., Brownson, R.C., 2005. in physical activity, obesity, and cardiovascular disease. Circulation 125, 729–737.
Perceived and objective environmental measures and physical activity among urban Shigematsu, R., Sallis, J.F., Conway, T.L., Saelens, B.E., Frank, L.D., Cain, K.L.,
adults. Am. J. Prev. Med. 28, 105–116. Chapman, J.E., King, A.C., 2009. Age differences in the relation of perceived
Inoue, S., Ohya, Y., Odagiri, Y., Takamiya, T., Kamada, M., Okada, S., Oka, K., neighborhood environment to walking. Med. Sci. Sports Exerc. 41, 314–321.
Kitabatake, Y., Nakaya, T., Sallis, J.F., 2011. Perceived neighborhood environment Shrestha, B., Millonig, A., Hounsell, N., Mcdonald, M., 2017. Review of public transport
and walking for specific purposes among elderly Japanese. J. Epidemiol. 21, needs of older people in European context. J. Popul. Ageing 10, 343–361.
481–490. Singapore Department of Statistics, 2018. Statistics Singapore - general household survey
Jack, E., Mccormack, G.R., 2014. The associations between objectively-determined and 2015. viewed 4 April 2018. https://www.singstat.gov.sg/publications/publications
self-reported urban form characteristics and neighborhood-based walking in adults. -and-papers/GHS/ghs2015.
Int. J. Behav. Nutr. Phys. Activ. 11, 71. Train, K.E., 2003. Discrete Choice Methods with Simulation. Cambridge University Press,
King, W.C., Brach, J.S., Belle, S., Killingsworth, R., Fenton, M., Kriska, A.M., 2003. The Cambridge.
relationship between convenience of destinations and walking levels in older Uebersax, J., 2015. Introduction to the tetrachoric and polychoric correlation
women. Am. J. Health Promot. 18, 74–82. coefficients. https://www.john-uebersax.com/stat/tetra.htm.
Koh, P., Leow, B., Wong, Y., 2015. Mobility of the elderly in densely populated Van Cauwenberg, J., Clarys, P., De bourdeaudhuij, I., Van Holle, V., Verté, D., De
neighbourhoods in Singapore. Sustain. Cities Soc. 14, 126–132. witte, N., De donder, L., Buffel, T., Dury, S., Deforche, B., 2012. Physical
Koohsari, M.J., Badland, H., Sugiyama, T., Mavoa, S., Christian, H., Giles-corti, B., 2015. environmental factors related to walking and cycling in older adults: the Belgian
Mismatch between perceived and objectively measured land use mix and street aging studies. BMC Publ. Health 12, 142.
connectivity: associations with neighborhood walking. J. Urban Health 92, 242–252. Van Dyck, D., Cerin, E., Conway, T.L., De bourdeaudhuij, I., Owen, N., Kerr, J.,
Lee, J., 2013. Perceived neighborhood environment and transit use in low-income Cardon, G., Frank, L.D., Saelens, B.E., Sallis, J.F., 2012. Perceived neighborhood
populations. Transport. Res. Rec. 2397, 125–134. environmental attributes associated with adults’ transport-related walking and
Li, F., Harmer, P.A., Cardinal, B.J., Bosworth, M., Acock, A., Johnson-Shelton, D., cycling: findings from the USA, Australia and Belgium. Int. J. Behav. Nutr. Phys.
Moore, J.M., 2008. Built environment, adiposity, and physical activity in adults aged Activ. 9, 70.
50–75. Am. J. Prev. Med. 35, 38–46. Van Dyck, D., Veitch, J., De bourdeaudhuij, I., Thornton, L., Ball, K., 2013.
Ma, L., Cao, J., 2019. How perceptions mediate the effects of the built environment on Environmental perceptions as mediators of the relationship between the objective
travel behavior. Transportation 46, 175–197. https://doi.org/10.1007/s11116-017- built environment and walking among socio-economically disadvantaged women.
9800-4. Int. J. Behav. Nutr. Phys. Activ. 10, 108.
Makles, A., 2012. Stata tip 110: how to get the optimal k-means cluster solution. STATA Van Holle, V., Van Cauwenberg, J., Van Dyck, D., Deforche, B., Van de Weghe, N., De
J. 12, 347–351. Bourdeaudhuij, I., 2014. Relationship between neighborhood walkability and older
Mercado, R., Páez, A., 2009. Determinants of distance traveled with a focus on the adults’ physical activity: results from the Belgian Environmental Physical Activity
elderly: a multilevel analysis in the Hamilton CMA, Canada. J. Transport Geogr. 17, Study in Seniors (BEPAS Seniors). Int. J. Behav. Nutr. Phys. Activ. 11, 110.
65–76. Webber, S.C., Porter, M.M., Menec, V.H., 2010. Mobility in older adults: a comprehensive
Moniruzzaman, M., Páez, A., Habib, K.M.N., Morency, C., 2013. Mode use and trip length framework. Gerontol. 50, 443–450.
of seniors in Montreal. J. Transport Geogr. 30, 89–99. Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. The MIT
Moniruzzaman, M., Páez, A., Scott, D., Morency, C., 2015. Trip generation of seniors and Press, Cambridge, Massachusetts.
the geography of walking in Montreal. Environ. Plann. 47, 957–976. Wretstrand, A., Svensson, H., Fristedt, S., Falkmer, T., 2009. Older people and local
Nyunt, M.S.Z., Shuvo, F.K., Eng, J.Y., Yap, K.B., Scherer, S., Hee, L.M., Chan, S.P., Ng, T. public transit: mobility effects of accessibility improvements in Sweden. J. Transp.
P., 2015. Objective and subjective measures of neighborhood environment (NE): Land Use 2, 49–65.

328

You might also like