Content uploaded by Linchuan Yang
Author content
All content in this area was uploaded by Linchuan Yang on May 21, 2022
Content may be subject to copyright.
Transactions in GIS. 2022;26:1339–1354. wileyonlinelibrar y.com/journal/tgis
|
1339
© 2022 John Wiley & Sons Ltd
DOI : 10.1111/tgi s.129 16
RESEARCH ARTICLE
Using a system of equations to assess the
determinants of the walking behavior of older
adults
Linchuan Yang1,2 | Xianglong Tang1 | Hongtai Yang3 | Fanyu Meng4 |
Jixiang Liu2,5
1Department of Urban and Rural Planning,
School of Architec ture, Southwest Jiaotong
University, Chengdu, China
2Shanghai Key Laboratory of Urban
Renewal a nd Spatial Optimization
Technology, Tongji University, Shanghai,
China
3School of Transport ation and Logistics,
Southwe st Jiaoton g University, Chengdu,
China
4Academy for Advanced Interdisciplinary
Studies , Southern University of Scie nce and
Technology, Shenzhen, China
5Depar tment of Urban Planni ng, School of
Architecture and Civil Engineering, Xiamen
University, Xiamen, China
Correspondence
Jixiang Liu, Shang hai Key Laboratory of
Urban Renewal and Spatial Optimization
Technology, Tongji University, Shanghai,
China.
Email: u3004679@hku.hk
Funding information
Shanghai Key Laboratory of Urban Renewal
and Spatial Optimization Technology,
Tongji University, China 20210202;
Sichuan Science and Technology P rogram:
2022JDR0178
Abstract
As the most prevalent physical activity and transporta-
tion mode for older people, walking is considered to have
multiple health and well- being benefits. Previous studies
used separate models to assess the built- environment de-
terminants of a battery of walking behavior measures, such
as walking frequency and duration. In a departure from
them, this study develops a system of equations, which is
estimated by seemingly unrelated regression, to determine
the built- environment factors that significantly influence
two correlated walking behavior measures (including walk-
ing frequency and duration) of older adults in Xiamen (a
medium- sized Chinese city) based on data from the Travel
Survey of Xiamen Residents 2015 and built- environment
geo- data. The results show the following: (1) the walking
frequency and duration of older adults are affected by the
built environment and socio- demographic characteristics;
(2) land- use mix, intersection density, and bus route density
positively influence older adults’ walking frequency and
duration; (3) distance to the commercial center adversely
impacts the walking frequency and duration; and (4) the
built environment has similar effects on the two measures.
This study offers a worthwhile reference for policy inter-
vention to promote older adults' walking activities, thereby
contributing to active and healthy aging.
1340
|
YANG et A l.
1 | INTRODUCTION
Most countries worldwide are experiencing population aging, owing to a surging life expectancy and a declining
fertility rate (the former is the major contributor) (Christensen, Doblhammer, Rau, & Vaupel, 2009). According to
the United Nations, the global population of those aged 65 years or above is projected to reach 1.00, 1.30, and
2.46 billion by 2030, 2050, and 2100, respectively (United Nations, 20 19). Apar t from the rise in absolute quantity,
the percentage of this age group is growing substantially as well. In 2019, older people (aged 65 years or above)
in Japan, Italy, Germany, Greece, Finland, and Portugal accounted for at least 20% of their national populations
(28%, 23%, 22%, 22%, 22%, and 22%, respectively) (United Nations, 2 019). In many developing countries with
rapid economic growth, issues related to population aging are as serious as in the developed world (Pettersson &
Schmöcker, 2010). For example, in China, with the largest older population globally, people aged 65 years or above
totaled 191 million (accounting for 13.5%) at the end of 2020, repor ted by the latest data of the Seventh National
Census. The population in China is aging at an unprecedented rate, which has triggered many social and economic
threats (e.g., medical care, social burden, and public welfare financial expenditure) (Cheng et al., 20 19). Therefore,
it is indispensable for government officials, urban planners, and policymakers to take steps to cater for the future
vigorous expansion in the quantity and proportion of the older population in China.
The expeditious increase in the older population may have a compelling effect on multiple facets. The exist-
ing literature has long focused on the influences of aging on economic development, labor markets, family and
intergenerational transfers, health and long- term care provisions, senior housing offerings, social welfare pro-
grams, endowment insurance, disease prevention, treatment, and rehabilitation (Bloom, C anning, & Fink, 2008;
Ha rp er, 2014). However, recently, the influences of aging on the transportation system have attracted substantial
attention from academia (Buehler & Nobis, 2010; Jing, Zhi, Yang, & Wang, 2021).
As a prerequisite for active aging, mobility is closely related to the quality of life and subjective well- being
of older adults. A low mobility level significantly reduces overall life satisfaction (Kim, 2 011; Kovac s, McLeod, &
Curtis, 2020; Metz, 2000; Wong, Szeto, Yang, Li, & Wong, 2018). Maintaining mobility has paramount significance
in keeping older people's bodies in good condition. However, older adults are often inactive and make fewer non-
discretionary trips (Kovacs et al., 2020). The reason is the gradual degradation of their physical functions, such as
muscle atrophy, osteoporosis, and cognitive impairment, which makes their daily travel more limited and makes
them more likely to experience social isolation and loneliness (Cao & Zhang, 2016; Jamal & Newbold, 2020).
Hence, the mobility of older adults should be promoted to ensure better integration into social life, which im-
proves their physical and mental health (Yang et al., 2020).
As a significant constituent of physical activity and a popular travel mode, walking has the advantages of
low injury risk, low intensity, low cost, and it is easy to carry out with no external device required (Hanson &
Jones, 2015; Lee & Buchner, 2008; Winters et al., 2015). Walking helps promote healthy and active aging. On the
one hand, walking has a wide array of known health benefits. The World Health Organization recommends that a
60 to 150- min weekly walk can effectively reduce the incidence of myocardial disease, psychosis, senile dementia,
heart disease, osteoporosis, diabetes, gout, Parkinson, and hypertension (Michel, Leonardi, Martin, & Prina, 2021;
Rudnicka et al., 2020). As such, walking contributes to the attainment of the “healthy aging” goal. On the other
hand, walking is conducive to increasing interpersonal communication, promoting social interaction, cultivating
weak social ties, facilitating activity participation, and decreasing societal costs (Leyden, 2003; Lund, 2002), which
are essential for older adults. Hence, walking contributes to the achievement of the “active aging” goal (Cheng
et al., 2019; Pettersson & Schmöcker, 2010). In addition, older adults generally have limited access to cars (Chudyk,
McKay, Winters, Sims- Gould, & A she, 2017; Forsyth, Oakes, Lee, & Schmitz, 2009), which makes walking one of
their oft- used travel modes. Hence, promoting walking activities for older adults is significant to improving their
overall quality of life and well- being.
The built environment refers to various artificial buildings and places and the environment that can be changed
through human activities (Chen et al., 2022; Ewing & Cer vero, 2001, 2010; Frank & Engelke, 2001; Wang &
|
1341
YANG et Al.
Lin, 2019). It has attracted the attention of numerous fields, such as geography, GIS, urban planning, transpor-
tation, public administration, and public health (Royal, 2013; Dao & Thill, 2018; Li, Santi, Courtney, Verma, &
Ratti, 2018). A great deal of empirical research shows that the built environment promotes or inhibits people's
travel behavior significantly ( Van Cauwenberg et al., 2011; Day, 2016; Tsunoda et al., 2012; Wang & Lin, 2019).
In addition, compared with young people, the same built environment characteristic (e.g., uneven road surface)
may pose greater challenges to older adults and more significantly affect their daily walking behavior (Forsyth
et al., 2009; Wang & Yang, 2019). Therefore, focusing on improving the neighborhood- level built environment of
older adults is necessary to enhance their life satisfaction and quality of life in the community.
Walking behavior can be evaluated by various measures, such as walking frequency and duration (or time).
Previous studies used separate equations to assess the built- environment determinants of multiple walking behav-
ior measures (or walking outcomes). However, the measures are often correlated, and some interactions between
the individual equations exist (Srivastava & Giles, 2020). For example, generally, walking frequency is positively
associated with walking duration. Therefore, developing a system of equations (or a set of simultaneous equa-
tions), each explaining some phenomenon, to describe walking behavior is indispensable. In light of the above, this
study establishes a seemingly unrelated regression equations (SURE) model (a type of a system of equations) to
analyze the determinants of two walking outcomes (daily walking frequency and daily walking duration) of older
adults in Xiamen (a medium- sized Chinese city, an understudied location) to address the above issues and fill the
gap left by previous research. The Travel Survey of Xiamen Residents 2015 (TSXR 2015) is used as the walking
behavior dataset. This Xiamen- based study can provide a valuable reference for other cities possessing similar
characteristics, particularly medium- sized cities.
The primary contributions of this study are the following. (1) Applying a novel method to the research prob-
lem for the first time. In a departure from existing studies using separate models for diverse (though correlated)
outcomes, this study uses a system of equations to accurately reveal the relationship between various correlated
outcomes (walking frequency and duration in this study) and their determinants. Methodologically, the linked
nature of walking behavior measures gives rise to the SURE model specification, which triggers the application
of the model in transportation research. (2) Identifying the determinants of the walking behavior of older adults
in a medium- sized cit y (a location receiving scant scholarly attention), which is in a departure from most studies
focusing on large cities (e.g., Tokyo, Hong Kong, Beijing, Shanghai, and Nanjing). (3) Comparing the determinants
of two correlated walking outcomes. The two outcomes are evidently correlated. A large walking frequency often
leads to a long walking duration. However, their determinants are expected to be correlated but not necessarily
the same. Underst anding the similarities and differences of the determinants of the two walking outcomes is es-
sential for age- friendly planning and design.
The remainder of this ar ticle is arranged as follows. Section 2 present s the data. Section 3 introduces the
methodology of the SURE model. Section 4 sheds light on the modeling results and illustrates the key factors
affecting older adults' walking frequency and duration. Section 5 discusses the results. Section 6 summarizes the
major conclusions and reveals research limitations.
2 | DATA
2.1 | Study area
The study area is Xiamen (see Figure 1), a special economic zone and sub- provincial city in Fujian Province. Xiamen
(a.k.a. Amoy) is located in the southeast coastal region of China, beside the Taiwan Strait. Xiamen governs six
districts. Among the six districts, Siming and Huli Districts occupy Xiamen Island, facing the other four districts
across the sea. They are the most prosperous and the highest urbanized areas in Xiamen. Given its pleasant living
environment and high walkability, Xiamen has become one of the most popular travel destinations for tourists. As
1342
|
YANG et A l.
FIGURE 1 Study area and the sampled and unsampled communities
|
1343
YANG et Al.
TAB LE 1 Descriptions and summar y statistics of the predicted and predictor variables
Variable Description Mean/percentage SD Min Max
Predicted variables
Walking frequency Daily walking frequency 1.11 1.43 011
Walking duration Daily walking duration (unit: min) 21.81 38. 51 0940
Predictor variables: socio- demographics
Male Dummy variable = 1 for male, 0 for female 0.50 - 0 1
Age Age of a person (unit : year) 67. 61 6.94 60 103
Education Ordered variable; 1 = middle school and below, 2 = high
school or vocational college, 3 = undergraduate and
above
2.06 1.17 1 3
Household size Number of f amily numbers 3.18 1.34 110
Driver's license Dummy variable =1 for a person with driver's license, 0
otherwise
0.10 - 0 1
Automobile Dummy variable =1 for a person with a car, 0 otherwise 0.32 - 0 1
Predictor variables: built environment
Population density Neighborhood- level population density (unit: 1/ha) 186.8 0 16 7.9 9 2.01 8 97. 78
Land- use mix Entropy for land uses within the neighborhood =
∑i(pi × ln pi)/ln N, where pi is the proportion of t he ith
land use and N is the number of land use c ategories
0.68 0.15 00.93
Intersection density Densit y of street intersections within the neighborhood
(unit: 1/ha)
0.22 0.25 0.01 3.21
Distance to commercial center Distance from the centroid of the community to the
nearest commercial center (unit: km)
9.0 6 9.3 1 0.13 38 .59
Bus route density Count of bus routes per unit of area within the
neighborhood (unit: 1/ha)
0.68 0.74 03.20
Sample size 11,73 2
1344
|
YANG et A l.
of 2019, Xiamen had an area of 1701 km2 (urban area = 334 km2) and a population of 4.29 million. It is a typical
medium- sized city in China.
2.2 | Data
Travel behavior dat a are extracted from TSXR 2015. TSXR 2015 is a long- term travel behavior survey. The survey
was completed from June 13 to 19, 2015, by district governments, sub- district offices, towns, and committees
in cooperation with Jimei University and Beijing Jingzhong Intelligent Transportation Technology Co., Ltd. TSXR
2015 determines the sampling strategy and method according to residents’ spatial distribution, family structure,
age, and gender information obtained from the local census report. TSXR 2015 records the detailed travel activ-
ity information of respondents in the past 24 h, including the starting and ending point, departure and arrival
time, travel purpose, and main travel mode. In total, 120,603 questionnaires were distributed, and 96,010 were
recovered (excluding 15,025 children under the age of 6 years and about 5000 collective households). Moreover,
93,861 were valid, representing an efficient recovery rate of 97.8% and a sampling rate of 3.1% (a very high sam-
pling rate for a city- scale travel survey). In this study, people aged 60 years or above in TSXR 2015 were selected
for analysis.
The neighborhood- level built environment is measured in the ArcGIS 10.6 software with original data col-
lected from OpenStreetMap (www.opens treet map.org). We collected various data, for example, POIs (points of
interest), AOIs (areas of interest), bus networks, and boundaries. The neighborhood is chosen as the community
(shequ), the smallest area unit we can identify in TSXR 2015 data.
2.3 | Variables
Table 1 shows a summary of the predicted and predic tor variables. This study concentrates on two dimen-
sions of walking behavior, namely walking frequency and duration. They are two predicted variables of a SURE
model.
The predictor variables of this study consist of two types: individual or family socio- demographic variables
and built environment variables. Among them, the socio- demographic variables of individuals or families, such as
gender, age, education, household size, driver's license, and automobile, are the control variables.
The most commonly used method of built environment assessment is the “3Ds” model (Cervero &
Kockelman, 1997 ). The model consists of density, diversity, and design. Since then, the “3Ds” model has been
further expanded into the “5Ds” model (incorporating destination accessibility and distance to transit) and the
“7Ds” model (incorporating demand management and demographics) (Ewing & Cervero, 2001). Following the
“5Ds” model and the existing literature and considering data availability, this study chooses population densit y,
land- use mix (diversity), intersec tion density, distance to the commercial center, and bus route density as the built
environment variables of interest (Cheng et al., 2 019; Feng, 2017; Yang, Ao, Ke, Lu, & Liang, 2021).
As shown in Table 1, the means of daily walking frequency and duration of older adults are 1.11 (SD = 1.43)
and 21.81 (SD = 38.51), respectively. The gender distribution of older adults is very symmetrical (50% for males
and 50% for females). Moreover, the average household size of older adults is 3. Notably, only 10% of older adults
hold a driver's license.
In short, taking the walking frequency and duration of older adults in Xiamen as predicted variables, this study
controls for the socio- demographic differences to single out the effect of a neighborhood- level built environment
on the walking behavior of older people. After excluding older adult observations with incomplete information,
we use the remaining 11,732 older adults residing in 316 communities for subsequent analysis. The distribution of
the 316 communities is shown in Figure 1.
|
1345
YANG et Al.
3 | METHODOLOGY
A system of equations includes multiple equations involving various predicted variables instead of one equation
for a single predicted variable. It has two types. One is the simultaneous equations model, which contains endog-
enous and exogenous regressors. The other is the SURE model, which only has exogenous regressors. The SURE
model can be traced back to Zellner (1962), who coined the term “seemingly unrelated regressions.” It can also be
viewed as a specific case of the simultaneous equations model, with no endogenous right- hand- side regressors.
Considering the jointness of the individual equations, which is an inherent property of the SURE model, is likely to
enhance the sharpness of inferences about the estimated parameters. The SURE model has been used in numer-
ous fields, such as economics and behavioral analysis (Srivastava & Giles, 2020).
The SURE model includes a set of equations with errors correlated across equations for a particular individual
but uncorrelated across individuals. The SURE model with M equations and T observations can be expressed as
follows (Srivastava & Giles, 2020):
where
yj
is a (
T×1
) vector of observations on the jth predicted variable,
Xj
is a (
T
×
lj
) matrix with rank
lj
of observations
on predictors (with
lj
the number of predictors for the jth predicted variable),
𝛽j
is a (
lj
×
1
) vector of regression coeffi-
cients, and
uj
is a (
T×1
) vector of errors of the jth equation.
The system of equations can also be written as:
or
where
y
≡[y
1
y
2
⋯y
M
]
T
,
𝛽
≡[𝛽
1
𝛽
2
⋯𝛽
M
]
T
,
u
≡[u
1
u
2
⋯u
M
]
T
, and
X
is the diagonal block matrix on the right- hand side of
the above equation.
The error vector
u
has the following variance– covariance matrix:
where I is a (
T×T
) unit matrix and
𝜎𝜇𝜃
=
E(u𝜇tu𝜃t)
for
t
=
1, 2,
…
,T
and
𝜇,𝜃
=
1, 2,
…
,M
.
The optimal estimator is the generalized least squares (GLS) estimator instead of the ordinary least squares
(OLS) estimator, which is commonly used in linear regression models. The GLS estimator is expressed as follows:
(1)
yj=Xj𝛽j+uj
(2)
⎡⎢⎢⎢⎢⎢⎢⎣
y1
y2
⋮
yM
⎤
⎥⎥⎥
⎥⎥⎥⎦
=
⎡
⎢⎢⎢
⎢⎢⎢⎣
X10⋯0
0X2⋮
⋮⋱0
0⋯0XM
⎤
⎥⎥⎥
⎥⎥⎥⎦
⎡
⎢⎢⎢
⎢⎢⎢⎣
𝛽1
𝛽2
⋮
𝛽M
⎤
⎥⎥⎥
⎥⎥⎥⎦
+
⎡
⎢⎢⎢
⎢⎢⎢⎣
u1
u2
⋮
uM
⎤⎥⎥⎥⎥⎥⎥⎦
(3)
y
=
X𝛽
+
u
(4)
Σ=⎡
⎢⎢⎢
⎢⎢⎢⎣
𝜎11I𝜎12 I⋯𝜎1MI
𝜎21I𝜎22 I⋯𝜎2MI
⋮⋮ ⋮
𝜎M1I𝜎M2I⋯𝜎MMI
⎤
⎥⎥⎥
⎥⎥⎥⎦
=
⎡
⎢⎢⎢
⎢⎢⎢⎣
𝜎11 𝜎12 ⋯𝜎1M
𝜎21 𝜎22 ⋯𝜎2M
⋮⋮ ⋮
𝜎M1𝜎M2⋯𝜎MM
⎤
⎥⎥⎥
⎥⎥⎥⎦
⊗I=Σ
c⊗
I
(5)
𝛽
GLS =
(
XT
−1
∑
X
)−1
XT
−1
∑y
1346
|
YANG et A l.
4 | RESULTS
Before the regression analysis, this study first conduc ts a pair wise correlation test to diagnose multicollinearity
among the predictor variables. Figure 2 reveals the results. It illustrates the absence of the multicollinearity issue.
As noted above, this study applies a novel method, namely the SURE model, to scrutinize the influence of built
environment attributes on two correlated walking behavior measures of older adult s. Two equations are estimated
simultaneously. The first equation examines walking frequency, while the second equation investigates walking
duration. Table 2 reports the SURE modeling results. The correlation between errors in the two equations is posi-
tive and ver y high (0.68), indicating that uncaptured variables are likely to influence the two variables similarly. For
example, a rainy day decreases walking frequency and duration for Adult A, whereas it increases the frequency
and duration for Adult B. The result of the Breusch– Pagan test for error independence shows a chi- square test
statistic equaling 5420.29 and a p value smaller than 0.001, indicating that a statistically significant correlation
exists between the errors in the two equations. These observations strongly justify the use of the SURE model
instead of separate models ignoring such a correlation.
Older adults' socio- demographic attributes are the control variables. They perform consistently in the two
equations in terms of significance and influencing direction. Most socio- demographic variables are significant
FIGURE 2 The correlation test results of the predictor variables
|
1347
YANG et Al.
determinants of the two walking outcomes. Specifically, age, being female, low educational attainment, living with
few people, and possessing no driver's license are significantly positively connec ted to walking frequency and
duration. Females walk more than males (having higher walking frequency and duration). A possible explanation
for this observation is that women take on more out- of- home household responsibilities than men (Feng, Dijst,
Prillwit z, & Wissink, 2013; Ghani, Rachele, Washington, & Turrell, 2016).
Age is negatively associated with the two walking outcomes. A convincing explanation is that walking usually
decreases with age, mainly due to the decline in physical fitness and the loss of functional ability. In other words,
age is a good predictor of the walking frequency and duration of older adults. The oldest people often have phys-
ical diseases, performing few out- of- home walking activities.
Household size negatively impacts older adults' walking frequency and duration. A possible explanation is
household responsibility sharing (Wong, Szeto, Yang, Li, & Wong, 2018). In China, older adults commonly receive
a substantial body of material and spiritual support (Wang & Lin, 2019). Older adults who live with more family
members, particularly their adult children, are less likely to walk out to complete family tasks (e.g., going out to
purchase daily necessities and drugs). In other words, they can easily delegate tasks to other family members,
thereby reducing opportunities for participating in out- of- home activities.
Driver's license ownership significantly decreases both walking frequency and duration of older adults at
the 1% level, indicating its prominent role in shaping their walking trips. It is reasonably inferred that older
adults with a driver's license are more willing to travel by car rather than walk. This outcome aligns with the
literature (Barnett, Barnett, Nathan, Van Cauwenberg, & Cerin, 2017; Delbosc & Nakanishi, 2017). Even more
TAB LE 2 SURE modeling results
Variable
First equation: walking frequency
Second equation: walking
duration
Coefficient z Value Coefficient z Value
Predictor variables: socio- demographics
Male −0.225** −8.31 −2.1 34 *−2. 93
Age −0.014** −7.4 8 −0.182** −3. 51
Education −0.025*−2. 82 – –
Household size −0.026*−2.4 0 −1.517** −5.14
Driver's license −0.345** −7.4 8 −9.0 81** −7. 2 6
Automobile −0.009 −0. 28 1.025 1.20
Predictor variables: built environment
Population density −0.001 −0.9 −0.006 −1.90
Land- use mix 0.523** 5.85 9.467 ** 3.89
Intersection density 0.202*3.30 3.666*2.20
Distance to commercial center −0.016** −9.27 −0.389** −8 .19
Bus route density 0.068*3 .41 1.488*2 .76
Constant 2.081 13.43 36. 850 8.79
Root mean square error 1.400 38 .031
Chi square 519. 82 295 .1
Correlation of errors across
equations
0.68
Sample size 11,732
*Significant at the 5% level; **Significant at the 1% level.
1348
|
YANG et A l.
interesting is that household car ownership has no significance in determining either older adults' walking
frequency or duration. That is, car ownership appears to be too weak to influence older adults' walking be-
havior, which may be due to the low car ownership of older adult s in China. This characteristic differs from
car- dominant Western countries (e.g., the United States and Australia), where owning a private car is the norm
instead of the exception.
Figuring out how the built environment factors affect older adult s’ walking frequency and duration is of pre-
dominant interest here. First, population density, which is commonly regarded as a facilitator of walking (more
broadly, active travel) in previous research, seems to play a weak role in shaping older adults’ walking behavior. A
possible explanation is that the relationship between population density and the walking behavior of older adults
is non- linear. For example, the positive link between population density and walking behavior exclusively exists
in a particular interval (e.g., population density <50,000 persons/km2), and a higher risk of getting hurt exists in
ultra- dense locations (Yang, Ao, et al., 2021). Second, the performance of land- use mix is highly consistent in the
first and second equations. They are significant at the 1% level, which means that older adults residing in neigh-
borhoods with a higher land- use mix (more diverse) are more likely to walk. This outcome resonates with previous
studies (Yang, Ao, et al., 2021). A convincing explanation is that a high land- use mix is usually characterized by
abundant types of destinations or activities, meaning adequate service facilities nearby are available to satisfy
various demands of older adults (e.g., catering, leisure, recreation, and shopping). Therefore, older adults residing
in a highly mixed neighborhood (in terms of land use) prefer to walk more frequently and for longer.
Moreover, intersection density positively influences older adults' walking frequency and duration. Given that
older people are likely to face osteoporosis and physical decline, a high road intersection densit y, which means
strong street connectivit y (Bentley et al., 2018), is beneficial for them to perform more frequent and longer dura-
tion walking. As expected, proximity to the commercial center is shown to significantly increase the two walking
outcomes, indicating that those who live closer to commercial facilities are more inclined to walk. This outcome is
largely consistent with prior studies (Feng et al., 2013; Gomez et al., 2010 ). Finally, bus route density appears to be
positively connected to the walking frequency and duration of older adults. This finding is because older adults'
need to walk to/from bus stops when using buses (Hess, 2012), thereby increasing their walking trips. This result
streng thens empirical findings from other regions (e.g., Japan) (Tsunoda et al., 2012).
To enhance the plausibility of this study, we choose to take a robustness check test by shuf fling the age thresh-
old from 60 to 65 years (redefining older adults) and re- estimating the SURE model. That is, the SURE model is
developed for older people aged 65 years or above (N decreases from 11,732 to 6974).
Table 3 presents the results, which are largely consistent with those presented before (see Table 2). Four built
environment variables significantly influence walking frequency of people aged 65 years or above. This obser va-
tion confirms the credibility of our core findings.
5 | DISCUSSION
Addressing the challenge of population aging should be considered comprehensively in an extensive way and
along multiple dimensions. The unprecedented population aging in China and many other countries has prompted
policymakers, urban planners, and researchers to pay much attention to related issues. Therefore, figuring out
how the built environment can promote active travel (e.g., walking activity) of older adults is a top priority in urban
and transport planning. Most existing research demonstrated that subjective and objective attributes of the built
environment are closely related to the walking behavior of older adults (Barnett et al., 2017; Gomez et al., 2010).
Considering that built environment determinants may connect distinctively to older adults' walking frequency
and duration, this study focuses on this issue in the context of a medium- sized but developed city in China. As we
know, this study is a pioneer that illustrates the links between the built environment and walking behavior among
the older population in a medium- sized Chinese city.
|
1349
YANG et Al.
It is indispensable to establish an age- friendly and pleasant built environment through precise policy inter-
vention or planning/design so that older adults can easily per form outdoor activities and keep physically active
(Winters et al., 2015), which is of great benefit for their health and well- being. However, inevitably, older adults
suffer from a natural decline in daily mobility with advancing age, which hinders their walking trips to a great ex-
tent. Hence, developing a supportive built environment (more broadly, living environment) to satisfy the walking
demands of older adult s is the focus of future research. To this end, this stud y attempts to identify the community-
level built environment characteristics af fecting the walking behavior of older adults. The study also identifies the
indicators suitable for active travel to provide a decision- making basis for age- friendly spatial design and planning.
This study not only provides us with an enriched understanding of and profound scientific value in the issue
of older adults' walking behavior, but also offers practical implications for policymakers and urban planners/de-
signers to determine community environmental planning methods and optimize the spatial environment for the
promotion of older adults' walking trips. We found that land- use mix, intersection density, bus route density, and
proximit y to the commercial center positively affect the walking frequency and duration. Hence, in the population
aging era, building a community environment with complex and diverse land functions, good access to transit
and service facilities, and high street connectivity seems essential, particularly in the areas where older adults
assemble. Notably, commercial accessibility appears to have a positive correlation with the two kinds of walking
behavior, which evidently echoes previous research (Boakye- Dankwa et al., 2019; Wang & Lin, 2 019). However,
some discrepancies exist (Barnett et al., 2017; Chudyk et al., 2017). This finding implies that even in cities which
are celebrated for their high walkability, some effective steps still need to be adopted to promote older people's
TAB LE 3 Robustness check results
Variable
First equation: walking frequency
Second equation: walking
duration
Coefficient z Value Coefficient z Value
Predictor variables: socio- demographics
Male −0.150** −4.40 −0.805 −0.89
Age −0.021** −7.8 5 −0.346** −4.68
Education −0.020 −1. 8 9 – –
Household size −0.061** −4.33 −2 .169** −5.69
Driver's license −0.250** −3.4 8 −6 .767** −3.50
Automobile 0.021 0. 51 0.632 0.56
Predictor variables: built environment
Population density 0.001 0.13 −0.004 −1. 02
Land- use mix 0.635** 5.49 11.0 06** 3. 52
Intersection density 0.215*2.83 3.575 1.74
Distance to commercial center −0.013** −5.79 −0.336** −5.47
Bus route density 0.053*2.11 0.74 5 1.10
Constant 2.548 11. 23 49.173 8.07
Root mean square error 1.382 3 7.311
Chi square 268 .77 155.18
Correlation of errors across
equations
0.70
Sample size 6974
*Significant at the 5% level; **Significant at the 1% level.
1350
|
YANG et A l.
walking trips. For example, urban planners should consider the heterogeneity of commercial facilities in future
planning and design.
In addition, we found a weak correlation between population density and older adults' walking behavior. As
some previous studies have pointed out (Cheng et al., 2 019; Wang & Lin, 2019), our findings echo var ying cor-
relations between the built environment and older people's walking trips. Nevertheless, this outcome is widely
divergent from the results of other research, particularly that conducted in the West (Bentley et al., 2018; Boakye-
Dankwa et al., 2019). A possible reason is that Xiamen is a densely populated area. In short, when optimizing the
built environment, policymakers and urban planners should not blindly increase/decrease the value of one or
several variables. The most suit able variable interval may exist ( Yang, Ao, et al., 2021). Only in this way can we
effectively promote the walking activities of older adults under the restriction of limited investment.
6 | CONCLUSIONS
Nowadays, population aging has become a prominent socio- demographic problem, which is the product of eco-
nomic, social, scientific, and technological development. This topic has elicited the attention of the United Nations
and governments worldwide, including China ( Yang, Liu, Liang, Lu, & Yang, 2021). Walking is one of the most prev-
alent transportation means for older Chinese people, which is distinct from that of the developed world (Barnett
et al., 2017; Herrmann- Lunecke, Mora, & Vejares, 2021; Larrañaga, Rizzi, Arellana, Strambi, & Cybis, 2016). Given
the multiple health and well- being benefits of walking for active aging, determining the influencing factors of older
adults' walking behavior to offer an age- friendly physical environment in China is of great priority. Dealing with
older adults, especially their walking activities, was the terra incognita to traditional urban planners and designers.
Fortunately, many recent studies have devoted themselves to this issue, enriching our understanding and offering
valuable practical insights and guidance.
Existing studies scrutinized the correlation between the built environment and older adults' walking behavior
(Bentley et al., 2018; Chan, Schwanen, & Banister, 2021; Gomez et al., 2010 ). However, few have discussed the
built environment attributes that correlate with various walking behavior measures in medium- sized cities using
a system of equations. Therefore, this study integrates the walking behavior data of Xiamen residents and the
built- environment geo- data. Then, it establishes SURE models to identify the built environment factors that sig-
nificantly affec t two walking outcomes (i.e., walking frequenc y and duration) of older adults. The results show the
following: (1) the walking frequency and duration of older adults are affected by the built environment and socio-
demographic characteristics; (2) land- use mix, intersec tion density, bus route density, and accessibility to the
commercial center are positively associated with older adults' walking frequency and duration; and (3) population
density plays a limited role in shaping the walking frequency and duration of older adults in this high- density city.
However, this study is definitely not exempt from limitations. First, we merely cover two common walking
transport indicators, namely walking frequency and duration. The reason is that a system of equations can only
include linear regression models, which makes us exclude other types of predicted variables (e.g., binary vari-
able for walking propensity). Admittedly, fitting more measures (e.g., walking propensity) is conducive to re-
flecting a broad pic ture of the walking behavior. Second, many potential predictors of walking behavior, such as
attitudes, preferences, and weather, cannot be modeled because of TSXR 2015 data limit ations. Additionally,
age- related health conditions and symptoms, which play an essential role in older adults’ walking behavior,
are not recorded in TSXR 2015 data. Hence, carrying out rigorous research and conducting a survey to collect
first- hand data on more aspects (e.g., attitudes or age- related health conditions and symptoms) are necessary
for future studies (He et al., 2021). Third, this study evaluates the built environment using overly simplistic
approaches. Using the virtual geographical environment (VGE) analysis tool, which represents an actual geo-
graphic feature, to accurately assess the built environment c an be explored in the future (Fu, Zhu, Li, You, &
Hua, 2021; Li et al., 2020, 2021). We feel that VGEs, which are adept at representation, visualization, analysis,
|
1351
YANG et Al.
simulation, and prediction, can play a larger role in activity behavior– environment research (Lin et al., 2013).
Last but not least, in terms of study methods, we applied a novel model, namely the SURE model, to explore
what factors promote or hinder two correlated walking outcomes. However, the model is linear in nature and
cannot handle the non- linearity issue. Therefore, multiple machine learning techniques (e.g., machine learn-
ing, random forest, AdaBoost decision tree, and extreme gradient boosting) are recommended (Xiao, Lo, Liu,
Zhou, & Li, 2021; Lu et al., 2021, 2022; Lu and Chen, 2022) to unravel the complicated relationship between
the built environment and the walking behavior of older people to draw more accurate and persuasive conclu-
sions. Moreover, incorporating simultaneous estimation thinking into machine learning techniques (for two or
more predicted variables) is to be applauded. We believe that it is a methodological improvement worthy of
investigation.
ACKNOWLEDGMENTS
This study was suppor ted by Shanghai Key Laboratory of Urban Renewal and Spatial Optimization Technology,
Tongji University, China (No. 20210202) and Sichuan Science and Technology Program (No. 2022JDR0178). The
authors are grateful to the three reviewers for their helpful comments.
CONFLICT OF INTEREST
The authors declare they have no conflict of interest.
DATA AVAI LAB IL IT Y S TATEM EN T
The data that support the findings of this study are available from the corresponding author upon reasonable
request.
REFERENCES
Barnet t, D. W., Barnett, A., Nathan, A ., Van Cauwenberg, J., & Cerin, E. (2017). Built environment al correlates of older
adults' total physical activity and walking: A systematic review and meta- analysis. International Journal of Behavioral
Nutrition and Physical Activity, 14(1), 1– 24. https://doi.org/10.1186/s1296 6- 017- 0558- z
Bentley, R., Blakely, T., Kavanagh, A., Aitken, Z., King, T., McElwee, P., … Turrell, G. (2018). A longitudinal study examining
changes in street connectivity, land use, and densit y of dwellings and walking for transport in Brisbane, Australia.
Environmental Health Perspectives, 126(5), 057003. https://doi.org/10.1289/EHP2080
Bloom, D. E., Canning, D., & Fink, G. (2008). Population aging and economic growth (Program on the Global Demography of
Aging Working Paper No. 31). Cambridge, MA: Harvard University.
Boakye- Dankwa, E., Nathan, A., Barnett, A ., Busija, L., Lee, R. S. Y., Pachana, N., … Cerin, E. (2019). Walking be-
haviour and patterns of perceived access to neighbourhood destinations in older adult s from a low- density
(Brisbane, Australia) and an ultra- dense cit y (Hong Kong, China). Cities, 84, 23– 33. ht tps://doi .org /10.1016/j.
cities.2018.07.002
Buehler, R., & Nobis, C . (2010). Travel behavior in aging societies: Comparison of Germany and the United States.
Transportation Research Record, 2182(1), 62– 70. https://doi.org/10.3141/2182- 09
Cao, J., & Zhang, J. (2016). Built environment, mobility, and quality of life. Travel Behaviour and Society, 5, 1– 4. h t t p s: //d o i .
org/10.1016/j.tbs.2015.12.001
Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: Density, diversit y, and design. Transportation Re search
Part D: Transpor t and Environment, 2(3), 199– 219. https://doi.org/10.1016/S1361 - 9209(97)000 09 - 6
Chan, E. T., Schwanen, T., & Banis ter, D. (2021). The role of perceived environment, neighbourhood characteristics, and
attitudes in walking behaviour: Evidence from a rapidly developing city in China. Transportation, 48(1), 431– 454.
https: //doi.o rg/10.10 07/s1111 6- 0 19- 10062 - 2
Chen, S., Wang, T., Bao, Z., &Lou, V. (2022). A path analysis of the effect of neighborhood built environment on
public health of older adults: A Hong Kong study. Frontiers in Public Health, 10, https://doi.org/10.3389/
fpubh.2022.861836
Cheng, L ., Chen, X., Yang, S., C ao, Z., De Vos, J., & Witlox, F. (2019). Active travel for ac tive ageing in China: The role of
built environment. Journal of Transport Geography, 76, 142– 152. https://doi.org/10.1016/j.jtran geo.2019.03.010
Christensen, K., Doblhammer, G., Rau, R ., & Vaupel, J. W. (2009). Ageing populations: The challenges ahead. The Lancet,
374 (9696), 1196– 1208. https://doi.org/10.1016/S0140 - 6736(09)61460 - 4
1352
|
YANG et A l.
Chudyk, A. M., McKay, H. A., Winters, M., Sims- Gould, J., & Ashe, M. C. (2017). Neighborhood walkability, physical activ-
ity, and walking for transport ation: A cross- sectional study of older adults living on low income. BMC Geriatrics, 17(1),
1– 14 . https://doi.org/10.1186/s1287 7- 017- 0469- 5
Dao, T. H. D., & Thill, J. C. (2018). Three- dimensional indoor network accessibility auditing for floor plan design.
Transactions in GIS, 22(1), 288– 310. https://doi.o rg/10.1111/t gis .12310
Day, K. (2016). Built environmental correlates of physic al activity in China: A review. Preventive Medicine Repor ts, 3, 303–
316. https://doi.org/10.1016/j.pmedr.2016.03.007
Delbosc, A., & Nakanishi, H. (2017). A life course perspective on the travel of Australian millennials. Transportation
Research Part A: Policy and Practice, 104, 319– 336. https://doi.org/10.1016/j.tra.2017.03.014
Ewing, R., & Cervero, R. (2001). Travel and the built environment: A synthesis. Transportation Research Record, 1780 (1),
87– 1 14 . htt ps://doi.or g/10.3141/178 0- 10
Ewing, R., & Cervero, R. (2010). Travel and t he built environment: A meta- analysis. Journal of the Americ an Planning
Association, 76 (3), 265– 294. https://doi.org/10.1080/01944 36100 3766766
Feng, J. (2017). The influence of built environment on travel behavior of the elderly in urban China. Transportation
Research Part D: Transport and Environm ent, 52, 619– 633. https://doi.org/10.1016/j.trd.2016.11.003
Feng, J., Dijst, M., Prillwitz, J., & Wissink, B. (2013). Travel time and distance in international perspective: A compari-
son between Nanjing (China) and the Randst ad (The Netherlands). Urban Studies, 50(14), 2993– 3010. ht t p s :// do i .
org /10.1177/0042 0 98 013 4 82504
Forsyth, A., Oakes, J. M., Lee, B., & Schmitz, K. H. (2009). The built environment, walking, and physical activity: Is the
environment more important to some people than others? Transportation Research Part D: Transport and Environment,
14(1), 42– 49. https://doi.org/10.1016/j.trd.2008.10.003
Frank, L . D., & Engelke, P. O. (2001). The built environment and human activity pat terns: Exploring the impac ts of
urban form on public health. Journal of Planning Literature, 16(2), 202– 218. https://doi.org/10.1177/08854 12012
2093339
Fu, L., Zhu, J., Li, W. L., You, J. G., & Hua, Z. Y. (2021). Fast estimation method of volumes of landslide deposit by
the 3D reconstruc tion of smar tphone images. Landslides, 18, 3269– 3278. http s://doi.org/10.1007/s1034 6- 021-
01702 - 9
Ghani, F., Rachele, J. N., Washington, S., & Turrell, G. (2016). Gender and age differences in walking for tr anspor t and
recreation: Are the relationships the same in all neighborhoods? Preventive Medicine Reports, 4, 75– 80. ht t p s : //d oi .
org/10.1016/j.pmedr.2016.05.001
Gomez, L. F., Parra, D. C ., Buchner, D., Brownson, R . C., Sarmiento, O. L., Pinzon, J. D., … Lobelo, F. (2010). Built envi-
ronment attributes and walking patter ns among the elderly population in Bogota. American Journal of Preventive
Medicine, 38(6), 592– 599. https://doi.org/10.1016/j.amepre.2010.02.005
Hanson, S., & Jones , A. (2015). Is there ev idence that walking groups have health benefits? A sys tematic review and m eta-
analysis. British Journal of Sports Medicine, 49(11), 710– 715. https://doi.org/10.1136/bjspo rts- 2014- 094157
Harper, S. (2014). Economic and social implications of aging societies. Science, 346(6209), 587– 591. h t tp s : //d o i .
org/10.1126/scien ce.1254405
He, B.- J., Zhao, D., Xiong, K ., Qi, J., Ulpiani, G., Pignatta, G., Prasad, D., & Jones, P. (2021). A framework for addressing
urban heat challenges and associated adaptive behavior by the public and the issue of willingness to pay for heat
resilient infrastructure in Chongqing. China. Sustainable Cities and Society, 75, 103361. http s://doi.o rg/10.1016/j.
scs.2021.103361
Herrmann- Lunecke, M. G ., Mora, R ., & Vejares, P. (2021). Perception of the built environment and walking in peri-
central neighbourhoods in Santiago, Chile. Travel Behaviour and Society, 23, 192– 206. http s://doi.o rg/10.1016/j.
tbs.2021.01.002
Hess, D. B . (2012). Walking to th e bus: Perceive d versus act ual walking d istance to bu s stops for old er adults. Transportation,
39(2), 247– 266. ht tps ://doi.org /10.10 07/s1111 6- 011- 9341- 1
Jamal, S., & Newbold, K. B. (2020). Factors associated with travel behavior of millennials and older adults: A scoping
revi ew. Sustainability, 12(19), 8236. https://doi.org/10.3390/su121 98236
Jing, C., Zhi, J., Yang, S., & Wang, W. (2021). Impact of driver age and experience in sof tware us age on driving
safety and usability of car- sharing software. Journal of Advan ced Transportation, 2 021, 6633379. h t t ps : //d o i .
org/10.1155/2021/6633379
Kim, S. (2011). Assessing mobility in an aging society: Personal and built environment fac tors associated with older peo-
ple’s subjective transport ation deficiency in the US. Transpor tation Research Part F: Traffic Psychology and Behaviour,
14(5), 422– 429. https://doi.org/10.1016/j.trf.2011.04.011
Kovacs, F. S., McLeod, S., & Curtis, C . (2020). Aged mobilit y in the era of transportation disruption: Will autonomous ve-
hicles address impediments to the mobility of ageing populations? Travel Behaviour and Society, 20, 122– 132. h t tp s : //
doi.org/10.1016/j.tbs.2020.03.004
|
1353
YANG et Al.
Larrañaga, A . M., Rizzi, L. I., Arellana, J., Strambi, O., & Cybis, H. B. B. (2016). The inf luence of built environment and travel
attitudes on walking: A case s tudy of Por to Alegre. Brazil. International Journal of Sustainable Transportation, 10 (4),
332– 34 2. https://doi.org/10.1080/15568 318.2014.933986
Lee, I. M., & Buchner, D. M. (2008). The import ance of walking to public health. Medicine & Scie nce in Spor ts & Exercise,
40(Suppl. 7), S512– S518. https://doi.org/10.1249/MSS.0b013 e3181 7c65d0
Leyden, K . M. (2003). Social capital and the built environment: The importance of walkable neighborhoods. American
Journal of Public Health, 93(9), 1546– 1551. htt ps://doi. org/10.2105/AJPH .93.9.154 6
Li, W., Zhu, J., Fu, L., Zhu, Q., Guo, Y., & Gong, Y. (2021). A rapid 3D reproduction system of dam- break floods con-
strained by post- disaster infor mation. Environmental Modelling & Software, 139, 104994. https://doi.or g/10.1016/j.
envso ft.2021.104994
Li, W., Zhu, J., Zhang, Y., Fu, L., Gong, Y., Hu, Y., & Cao, Y. (2020). An on- demand construction method of disaster scenes
for multilevel users. Natural Hazards, 101(2), 409– 428. https://doi.org/10.1007/s1106 9- 020- 03879 - z
Li, X., Santi, P., Courtney, T. K., Verma, S. K., & Ratti, C . (2018). Investigating the association between street scapes and
human walking activities using Google Street View and human trajector y data. Transactions in GIS, 22(4), 1029– 1044.
https: //doi.o rg/10.1111/t gis .12472
Lin, H., Chen, M., Lu, G., Zhu, Q., Gong, J., You, X ., … Hu, M. (2013). Virtual geographic environments (VGEs): A
new generation of geographic analysis tool. Earth- Science Reviews, 126, 74– 84. https://doi.org/10.1016/j.earsc
irev.2013.08.001
Lu, W., Chen, J., & Xue, F. (2022). Using computer vision to recognize composition of construction waste mixtures: A se-
mantic segmentation approach. Resources, Conservation and Recycling, 178 , 106022. https://doi.org/10.1016/j.resco
nrec.2021.106022
Lu, W., & Chen, J. (2022). Computer vision for solid waste sorting: A critic al review of ac ademic research. Wa ste
Management, 142, 29– 43. https://doi.org/10.1016/j.wasman.2022.02.009
Lu, W., Lou, J., Webster, C., Xue, F., Bao, Z., & Chi, B. (2021). Estimating construction waste generation in the
Greater Bay Area, China using machine learning. Waste Management, 134, 78– 88. https://doi.org/10.1016/j.
wasman.2021.08.012
Lund, H . (2002). Pedest rian environments and sen se of community. Journal of Planning Education and Research, 21(3), 301–
312. https://doi.org/10.1177/07394 56X02 02100307
Metz, D. H. (2000). Mobilit y of older people and their qualit y of life. Transpor t Policy, 7(2), 149– 152. h t t ps : //d o i .
org/10.1016/S0967 - 070X(00)00004 - 4
Michel, J.- P., Leonardi, M., Mar tin, M., & Prina, M. (2021). WHO’s repor t for the decade of healthy ageing 2021– 30 sets
the stage for globally comparable data on healthy ageing. The Lancet Healthy Longevity, 2(3), e121– e122. h t t p s: // do i .
org /10.1016/S26 66 - 7568 (21) 00 002 - 7
Pettersson, P., & Schmöcker, J.- D. (2010). Active ageing in developing countries? Trip generation and tour complex-
ity of older people in Metro Manila. Journa l of Transport Geography, 18(5), 613– 623. https://doi.org/10.1016/j.jtran
geo.2010.03.015
Royal, N. (2013). Dracunculiasis, proximity, and risk: Analyzing the location of G uinea worm disease in a GIS. Transactions
in GIS, 17(2), 298– 312. htt ps: //doi.o rg/10.1111/t gis .120 03
Rudnicka, E., Napierała, P., Podfigurna, A., Męczekalski, B., Smolarczyk, R., & Grymowicz, M. (2020). The World
Health Organization (WHO) approach to healthy ageing. Maturitas, 139, 6– 11. http s://doi.org/10.1016/j.matur
itas.2020.05.018
Srivastava, V. K., & Giles, D. E. (2020). Seemingly unrelated regression equations: Models, estimation and inference. Boca
Raton, FL : CRC Press.
Tsunoda, K., Tsuji, T., Kitano, N., M itsuishi, Y., Yoon, J. Y., Yoon, J., & Okura, T. (2012). Associ ations of physical activit y with
neighborhood environments and transportation modes in older Japanese adults. Preventive Medicine, 55(2), 113 – 118.
https://doi.org/10.1016/j.ypmed.2012.05.013
United Nations. (2019). World population prospects 2019: Highlights. New York, NY: United Nations Depar tment of
Economic and Social Affairs, Population Division.
Van Cauwenberg, J., De Bourdeaudhuij, I., De Meester, F., Van Dyck, D., Salmon, J., Clarys, P., & Deforche, B. (2011).
Relationship between the physical environment and physical activity in older adults: A systematic review. Health &
Place, 17(2), 458– 4 69. https://doi.org/10.1016/j.healt hplace.2010.11.010
Wang, D., & Lin, T. (2019). Built environment, travel behavior, and re sidential self- selection: A study based on panel data
from Beijing. China. Transportation, 46(1), 51– 74. htt ps: //doi.o rg/10.1007/s1111 6- 017- 9783- 1
Wang, H., & Yang, Y. (2019). Neighbourhood walkability: A review and bibliometric analysis. Cities, 93, 43– 61. h t t p s: //d o i .
org/10.1016/j.cities.2019.04.015
Winters, M., Barnes, R., Venners, S., Ste- Marie, N., McK ay, H., Sims- Gould, J., & Ashe, M. (2015). Older adults' outdoor
walking and the built environment: Does income matter? BMC Public Health, 15(1), 1– 8. https://doi.org/10.1186/
s1288 9- 015- 2224- 1
1354
|
YANG et A l.
Wong, R. C . P., Szeto, W. Y., Yang, L ., Li, Y. C., & Wong, S. C. (2018). Public transport policy measures for improving elderly
mobility. Transport Policy, 63, 73– 79. https://doi.org/10.1016/j.tranp ol.2017.12.015
Xiao, L., Lo, S., Liu, J., Zhou, J., & Li, Q. (2021). Nonlinear and synergistic effec ts of TOD on urban vibrancy: Applying local
explanations for gradient boosting decision tree. Sustainable Cities and Society, 72, 103063. https://doi.org/10.1016/j.
scs.2021.103063
Yang, L., Ao, Y., Ke, J., Lu, Y., & Liang, Y. (2021). To walk or not to walk? Examining non- linear effects of streetscape green-
ery on walking propensity of older adults. Journ al of Transport Geography, 94, 103099. https://doi.org/10.1016/j.jtran
geo.2021.103099
Yang, L., Liu, J., Liang, Y., Lu, Y., & Yang, H. (2021). Spatially varying effects of street greenery on walking time of older
adults. ISPRS International Journal of Geo- Information, 10(9), 596. https://doi.org/10.3390/ijgi1 0090596
Yang, L., Liu, J., Lu, Y. I., Ao, Y., Guo, Y., Huang, W., … Wang, R. (2020). Global and local associations between urban
greenery and travel propensity of older adults in Hong Kong. Sustainable Cities and Society, 63, 102442. h t t ps : //d o i .
org/10.1016/j.scs.2020.102442
Zellner, A. (1962). An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias.
Journal of the American Statistical Association, 57(298), 348– 368. htt ps://doi.org/10.1080/01621 459.1962.104 80664
How to cite this article: Yang, L., Tang, X., Yang, H., Meng, F., & Liu, J. (2022). Using a system of equations to
assess the determinants of the walking behavior of older adults. Transactions in GIS, 26, 1339– 1354. h t t ps : //
doi .or g/10 .1111/t gis .12916