University of Southern California Dissertations and Theses

Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

(USC Thesis Other)

Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ACTIVE TRAVEL, OUTDOOR LEISURE, AND NEIGHBORHOOD
ENVIRONMENT: PATH ANALYSIS, LOS ANGELES COUNTY

by

Yong-Jin Ahn

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POLICY, PLANNING, AND DEVELOPMENT)

May 2014

Copyright 2014 Yong-Jin Ahn

ii
DEDICATION

To

My parents, wife, and newborn baby for their support, sacrifice, and love

iii
TABLE OF CONTENTS
DEDICATION ...................................................................................................................... II
LIST OF TABLES .................................................................................................................V
LIST OF FIGURES ........................................................................................................... VII
ACKNOWLEDGEMENTS ............................................................................................. VIII
ABSTRACT ..................................................................................................................... IX
CHAPTER 1. INTRODUCTION ....................................................................................... 12
1.1. RESEARCH BACKGROUND ...................................................................................... 12
1.2. RESEARCH GAPS AND PROBLEMS ........................................................................... 14
1.3. RESEARCH STATEMENT .......................................................................................... 17
1.4. ORGANIZATION OF THE THESIS .............................................................................. 21
CHAPTER 2. LITERATURE REVIEW ........................................................................... 23
2.1. THEORETICAL FOUNDATION .................................................................................. 23
2.2. FACTOR AFFECTING ACTIVE TRAVEL AND OUTDOOR LEISURE .............................. 27
2.3. CAUSALITY AND RESIDENTIAL SELF-SELECTION ................................................... 41
CHAPTER 3. CONCEPTUAL FRAMEWORK .............................................................. 51
CHAPTER 4. DATA AND METHODOLOGY ................................................................ 56
4.1. DATA ...................................................................................................................... 56
4.2. VARIABLES ............................................................................................................. 61
4.3. METHODOLOGY...................................................................................................... 80
CHAPTER 5. NEIGHBORHOOD ENVIRONMENT AND ACTIVE TRAVEL ......... 86
5.1. DESCRIPTIVE ANALYSIS ......................................................................................... 86
5.2. REGRESSION RESULTS ............................................................................................ 93
5.2.1. Active travel time .......................................................................................... 93
5.2.2. Active travel frequency ............................................................................... 106

iv
CHAPTER 6. NEIGHBORHOOD ENVIRONMENT AND OUTDOOR LEISURE . 124
6.1. DESCRIPTIVE ANALYSIS ....................................................................................... 124
6.2. REGRESSION RESULTS .......................................................................................... 129
6.2.1. Outdoor leisure activity time ....................................................................... 129
6.2.2. Outdoor leisure activity frequency .............................................................. 142
CHAPTER 7. CONCLUSIONS AND DISCUSSION .................................................... 160
7.1. MAJOR FINDINGS AND CONTRIBUTIONS ............................................................... 160
7.2. POLICY IMPLICATIONS.......................................................................................... 164
7.3. LIMITATION AND FUTURE STUDY ......................................................................... 168
REFERENCES .................................................................................................................. 173
APPENDIX ................................................................................................................... 188

v
LIST OF TABLES
Table 2-1 Summary of studies using SEM model ................................................................. 47
Table 4-1 Location types of leisure destinations (Outdoor/indoor) ....................................... 63
Table 4-2 Composites of neighborhood types for active travel module ................................ 71
Table 4-3 Summary of descriptive statistics of variables ...................................................... 75
Table 4 4 Definition of variables for active travel model ..................................................... 79 79
Table 4-5 Definition of variables for outdoor leisure activity mode ..................................... 80 80Table 5-1 Mean of frequency/duration of active travel by purpose (pooled samples) 87
Table 5-1 Mean of frequency/duration of active travel by purpose (pooled samples) .......... 87
Table 5-2 Mean of frequency of active travel by mode and purpose (income groups) ........ 88
Table 5-3 Mean of duration of active travel by mode and purpose (income groups) ............ 88
Table 5-4 Distribution of travel time by active travel modes ................................................ 90 90
Table 5-5 Distribution of travel distance by active travel modes .......................................... 90 90
Table 5-6 Frequency and mean distribution of active travel (pooled samples) ..................... 91 91
Table 5-7 Frequency and mean distribution of active travel (income groups) ...................... 92 92
Table 5-8 Results from path analysis on active travel time (pooled samples) ....................... 94 94
Table 5-9 Results from path analysis on active travel time by income groups ................... 103 103
Table 5-10 Results from path analysis on walking time by income groups ........................ 104 104
Table 5-11 Results from path analysis on walking time (binary) by income groups .......... 105 105
Table 5-12 Results from path analysis on active travel frequency ...................................... 107 107
Table 5-13 Results from path analysis on active travel frequency by income groups ......... 117 117
Table 5-14 Results from path analysis on walking frequency by income groups ............. 118 118

vi
Table 5-15 Results from path analysis on walking frequency (binary) by income groups . 119 119
Table 5-16 Key predictors of active travel (pooled samples) .............................................. 120 120
Table 5-17 Key predictors of active travel (income groups) ............................................... 123 123Table 6-1 Time spent in outdoor leisure activity by income groups 125
Table 6-1 Time spent in outdoor leisure activity by income groups ................................... 125
Table 6-2 Frequency of outdoor leisure activity by income groups .................................... 126
Table 6-3 Outdoor leisure activity (time and frequency) by neighborhood types ............... 127
Table 6-4 Descriptive statistics of outdoor leisure trips ...................................................... 127
Table 6-5 Frequency distribution of outdoor leisure travel time (duration) ........................ 128
Table 6-6 Frequency distribution of outdoor leisure travel distance ................................... 128
Table 6-7 Results from path analysis for outdoor leisure activity time ............................... 136
Table 6-8 Results of outdoor exercise/sport time by income groups ................................... 139
Table 6-9 Result of outdoor exercise/sport (including park activity) time by income
groups ................................................................................................................ 140
Table 6-10 Results of outdoor leisure (exercise/park/active leisure travel) time by
income ............................................................................................................... 141
Table 6-11 Results from path analysis for outdoor leisure activity frequency .................... 148
Table 6-12 Results from path analysis for outdoor exercise/sport frequency by income .... 152
Table 6-13 Results of outdoor exercise/sport (including park activity) frequency by
income ............................................................................................................... 153
Table 6-14 Results of outdoor exercise/park/active leisure travel frequency by income .... 154
Table 6-15 Key predictors of outdoor leisure (pooled samples) .......................................... 155
Table 6-16 Key predictors of outdoor leisure (income groups) ........................................... 159

vii
LIST OF FIGURES
Figure 3-1 Conceptual framework for daily decision on active travel/outdoor leisure
activity ................................................................................................................. 52
Figure 4-1 Study area (Los Angeles County: 3,381 households) .......................................... 60
Figure 4-2 Four domains of physical activity ........................................................................ 61
Figure 4-3 Neighborhood types (Urban vs. Suburban neighborhoods) ................................. 72
Figure 4-4 Neighborhood types (High-leisure vs. Low-leisure neighborhoods) ................... 73

viii
ACKNOWLEDGEMENTS

I am deeply indebted to Professor Genevieve Giuliano, my dissertation chair and
academic advisor. She is an exceptional scholar, teacher, and mentor who generously
provided me with unwavering support and excellent guidance throughout my graduate
studies at University of Southern California. I have tremendous respect for Professor
Giuliano, and I am blessed to have met her as my advisor.
I would also appreciate Professor Marlon Boarnet who is my committee member
and director of graduate programs in Urban Planning and Development, Sol Price School of
Public Policy at USC. He is a brilliant and nationally reputed scholar in the field of
transportation and land use. Due to his thoughtful and detailed comments on my dissertation
topic and proposal, I could develop my research and finally complete it. I also wish to thank
the outside committee member, Professor Donna Spruijit-Metz in Department of Preventive
Medicine, Keck School of Medicine at USC, for her helpful advice and insightful comments
on my work.
I owe special thanks to the faculty and staff in Price School. METRANS
Transportation Center where I have been working as a research assistant provided the great
chance to have lots of research experiences, while offering financial support during my
graduate studies. Lusk Center for Real Estate also gave me the fellowship for my dissertation
work. I would also like to thank Caltrans and SCAG for kindly providing private version of
NHTS California add-on dataset as well as various spatial datasets in Los Angeles County
that I employed in this dissertation research.

ix
ABSTRACT

In response to increasingly sedentary lifestyles, many policymakers have suggested
making neighborhood design more conducive to active lifestyles, emphasizing the role of
physical elements, such as high density, mixed land use, street design, and easy access to
local destinations. However, a neighborhood not only provides opportunities for active living
but also contains various types of environmental barriers. While the role of environmental
opportunity is well acknowledged, little is known about the deterrent role of environmental
barriers.
Drawing on a fairly large number of adult samples obtained from NHTS California
data while simultaneously developing a more fine-grained objective measurement of
neighborhood environment, this dissertation explores what aspects of neighborhood
environment facilitate and/or constrain active travel and outdoor leisure on a daily basis.
Identifying the potential gap between low- and high- income groups in terms of
environmental opportunities and barriers, this study intends to better understand physical
inactivity among subpopulations more exposed to those barriers. Using path analysis which
allows the analyst to isolate the impact of neighborhood environment from that of residential
sorting, I address the complex relationships between personal/household backgrounds,
neighborhood characteristics, and behavioral decisions on active travel/outdoor leisure
activity.

x
Empirical results reveal not only significant supportive (e.g., activity density, street
design, and parks) but also deterrent (e.g., violent crimes) roles of neighborhood
environment in the duration and frequency of daily active travel. However, active travel
mainly depends on individual/household characteristics. In particular, household travel
options (i.e., car availability) consistently indicated the most profound and negative
influence at the significant level. A clear gap between low- and high- income groups in
environmental correlates of active travel was also found. In terms of the duration, the low-
income group is more responsive to activity density and neighborhood (violent) crimes than
the high-income group. On the other hand, both groups reported a consistently deterrent role
of violent crimes in the frequency measured by a continuous scale. Interestingly, however,
such a deterrent role was not significant in determining whether low-income groups walk or
not.
Analytic models for outdoor leisure activity consistently presented the positive
impact of local facilities for exercise/sport, confirming the supportive role of neighborhood.
However, other variables yielded inconsistent results, suggesting that the function of outdoor
leisure is mostly subject to specific types of leisure activity and household income. For low-
income groups, a leisure-friendly neighborhood had a significant positive association with
the duration of outdoor exercise/sport. Given the limited travel options of low-income
groups, this might imply the importance of neighborhood amenities for leisure. High-income
groups are more sensitive to perceived safety concern when they engage in outdoor leisure,
whereas low-income groups are more responsive to exposure to traffic incidents objectively
measured. Surprisingly, it is found that the amount of time spent on active travel is the most
profound factor affecting outdoor leisure activity. Such trade-off relationships between

xi
different domains of physical activity cast doubt on the conventional wisdom that engaging
in active travel increases the overall physical activity.
To effectively promote active behavior and life styles, policy should be devoted to
injury prevention designed to reduce neighborhood barriers, in tandem with supportive
neighborhood design. Moreover, this study finds that active-living campaigns will achieve
better performance in combination with transportation planning (e.g., more pricing on
vehicle use and improving mobility of low-income groups). However, despite such
combined policy endeavors, increasing overall physical activity is a much bigger challenge if
the total budget of physical activity is fixed or controlled by other factors beyond the policy
and environmental strategies.

12
CHAPTER 1.
INTRODUCTION

1.1. RESEARCH BACKGROUND

The ultimate goal of planning is to improve people’s quality of life and overall well-being.
Creating sustainable/healthy communities (or neighborhoods) has been recognized as an
important planning strategy for achieving this goal. It is frequently emphasized in the
planning/transportation/health field that neighborhood which is walking-friendly and/or
highly accessible to local amenities can provide opportunities for residents to increase
physical activity and hence contribute to improving their health status (Booth et al., 2001;
Frank et al., 2005; Sallis and Owen, 1999).
More specifically, the proponents of New Urbanism have maintained that people
who live in a compact, traditional neighborhood are more likely to walk and less likely to
drive than those living in other neighborhoods especially in suburban areas. Based on
supportive evidence for this argument, they conclude that urban design plays a crucial role in
shaping or governing individual behavior and activity patterns (Calthorpe, 1993; Katz, 1994).
In this discourse, both active travel and outdoor leisure are often considered as a
means of incorporating physical activity into daily routines. For instance, in transportation
studies, walking is a common form of moderate physical activity but also a popular “man-
powered” mode of travel which allows people to attain easily the recommended level of
daily physical activity (i.e., 30 minutes of brisk walking or 15 minutes of jogging) (Roberts

13
et al., 1996; Sallis et al., 2004). In addition, leisure/recreation studies also document health
benefits from either being outdoors or outdoor leisure activities (Sallis et al., 2000; Coleman
et al., 1993). Being outdoors increases the likelihood of physical activity over being indoors
(Godbey, 2009). Participation in outdoor leisure activities (e.g., playing sports, visiting parks,
jogging and strolling, etc) contributes substantially to improving physical health status as
well as reducing the prevalence and severity of mental disorders (e.g., depression and stress)
through contact with nature and greenness (Siegenthaler, 1997; Tarrant et al., 1994).
However, a sedentary life style is still dominant among the U.S. population. People
rarely engage in active transportation or spend their leisure time in outdoor places.
According to the 2001 NHTS, the market share of active transportation is 3.9% in work-
related trips and 14% in social and recreation trips (Pucher and Renne, 2003). On average,
Americans spend 17 minutes in sports, exercise, and recreation per a day, whereas they
spend 216 minutes in sedentary leisure activities such as watching TV, reading, relaxing, and
using computers (BLS, 2006). Furthermore, even though no leisure-time physical activity
has decreased over two decades in all age/racial groups, it is found that 25% of U.S. adults
do not participate in any physical activities or exercise (Moore et al., 2012).
Given the prevalence of sedentary life styles, there has been much concern about
low levels of physical activity (French et al., 2001). The lack of physical activity has been
highlighted as a major factor which explains the increase in obese populations (Owen and
Bauman, 1992; Owen et al., 2000). As noted in public health literature, more than one-third
of U.S. adults are obese, and the estimated health cost related to obesity was about $75
billion in 2003 (Flegal et al., 2002; Frinkelstein et al., 2004). Those backgrounds directly

14
motivated academic researchers to pay a great deal of attention to physical activity, while
integrating discussions derived from diverse disciplines into the movement for active living.

1.2. RESEARCH GAPS AND PROBLEMS

A growing body of literature has introduced research frameworks over the past decades to
shed light on how people make behavioral decision on various types of physical activity,
based on several theoretical foundations from the perspectives of micro-economics, human
psychology, and environment.
Previous studies in the field of planning and public health have often explored the
potential link between the built environment and physical activity, highlighting the
supportive role of the built environment (i.e., urban form features and the access to local
resources) in governing active living (Sallis et al., 2000; Humpel et al., 2002; Owen et al.,
2004; Giles-Corti et al., 2005; Frank and Engleke, 2000; Saelens et al., 2003). However,
findings from previous studies are mixed, and the interplay of human behavior and
environment is still elusive (Craig et al., 2002; Boarnet et al., 2005). Some studies conclude
empirically that the role of the built environment is either insignificant or marginally
significant, and that the built environment alone rarely explains the complex pathway to
active travel/leisure activity (Handy, 2004; Boarnet, 2006; Forsyth et al., 2007; Wendel-Vols
et al., 2007).
Meanwhile, current literature on ‘active living’ has focused more on the
medium/high-income group within suburban contexts (Hynes and Lopez, 2009; Sallis et al.,
2006) although low-income populations more rarely meet the recommended level of
physical activity when compared with high-income populations (ALD, 2010; Gidelow et al.,

15
2007; Pucher and Renne, 2003). One potential reason behind physical inactivity of low-
income populations is that, in general, they are more expose to environmental barriers and
constraints that may deter such activities.
Neighborhoods where individuals are embedded through a way of living, working,
playing, and socializing not only provide fundamental resources for health and well-being
but also contain various types of risk factors (Hanna and Coussens, 2001; Fitzpatrick and
Gory, 2000). The point here is that both positive and negative factors for health are “spatially
and socially constructed” beyond any individual’s capacity (Fitzpatrick and Gory, 2000: 8).
Social inequality produces different levels of access to local resources and of exposure to
local hazards, and hence can finally lead to health disparities. If there is a clear link between
social inequality and health disparities, and between the physical inactivity of low-income
populations and the penalty of the unhealthy place where low-income populations live, it is
more imperative to reduce the gap in local barriers between neighborhoods than it is to
reduce the gap in local resources. However, earlier studies have normally placed their foci
on the facilitating role of local resources, whereas the deterrent role of local barriers is
relatively unknown in the literature.
Give the literature gap, we need to address 1) what aspects of neighborhood
environment may prevent people from engaging in active travel/outdoor leisure, even though
they are willing to do so; 2) whether the deterrent role of neighborhood environment is more
important than the facilitating role; and 3) how such influence varies across different
subpopulations (especially, low- vs. high-income groups). Scientific knowledge on these
research questions not only helps us to better understand the dynamic role of neighborhoods

16
in active living but also informs what kinds of policies should be designed to promote active
living and under what condition these tools can work effectively.
Physical activity is a very broad context, including various types of activity and
locations/domains where such activities are embedded (Berrigan and Toriano, 2002).
Despite the diversity in physical activity, most studies which explored environmental
correlates of physical activity have been interested in “total” physical activity regardless of
type and location (Handy, 2004). The simplification of “total” physical activity may not fully
reflect the different nature of physical activity in different domains (in particular, active
travel and outdoor leisure), while it sometimes leads to misunderstandings on different
environmental roles in specific types of physical activity. Likewise, previous studies have
often ignored the different natures of indoor and outdoor leisure physical activities in
investigating environmental correlates. Again, the simplification of ‘total’ leisure-time
physical activity can yield misinterpretations in the observed environmental influence.
As another research problem, the difficulties in determining the effect of the built
environment on active travel/outdoor leisure activity can be mainly explained by the
uncertainty arising from lack of information as well as the complex mechanism of behavioral
decisions. The one is closely related to data/measurement issues, but the other is more
associated with methodological issues. In the case of data/measurement issues, the
researcher rarely knows the specific locations of households and of travel/activity
destinations. Either the absence of the specific locations’ geo-codes or the presence of loose
geographic information also makes it difficult to define a geographic unit, raising a
modifiable areal unit problem. Thus, one can hardly disaggregate the characteristics of the
built environment with more fine-grained measurements.

17
In addition, behavioral decisions on active travel/outdoor leisure activities can be
explained by various individual, household, and environmental forces (Hawley, 1988).
Among those factors, intrapersonal values (e.g., motivation, attitude, and preference) plays a
crucial role in addressing spurious relationships between built environment and behavioral
decisions on active travel/outdoor leisure activities. However, the uncertainty of
intrapersonal values also makes it difficult to determine the role of the built environment.
The complex mechanism by which human behavioral decisions are made leads to
critical methodological issues. More specifically, causality has often been considered as a
pivotal challenge. Correlation does not necessarily imply causality, since causal inference
additionally requires non-spuriousness and time precedence between ‘cause’ and ‘effect’
(Cao et al., 2008). However, observational research using cross-sectional data provides
limited information on causal relationships. In addition, a spurious relationship created by
confounding factors is another source of methodological challenges. For example, people
who prefer active transportation or outdoor leisure may choose neighborhoods which support
their preference. Given the residential self-selection issue, the effect of the built environment
of a neighborhood can be overestimated and the true effect might be distorted (Handy, 2004).

1.3. RESEARCH STATEMENT

This study aims to inform effective policy strategies for a sustainable/healthy
community by testing several hypotheses on the complex interplay of human behavior and
neighborhood environment. Not only filling in research gaps but also addressing research
problems mentioned before, the research explores the mechanism through which daily active
travel/outdoor leisure activity is shaped or governed in behavioral settings.

18
Focusing on specific domains of physical activity, this study identifies what aspects
of neighborhood environment may facilitate or constrain active travel/outdoor leisure
activities. In addition, this study investigates how those impacts of neighborhood
environment differ across income groups and spatial contexts. In this process, different
functions of active travel and leisure activity will be tested in an advanced framework that 1)
incorporates individual/household characteristics, attitudinal factors, and neighborhood
types; 2) measures various aspects of neighborhood environment at the highly disaggregated
level; and 3) addresses the potential impact of residential self-selection. Obviously, this
approach allows for identifying the relative importance of multiple factors as well as for
isolating the impact of neighborhood environment from that of residential sorting. More
importantly, the potential gap between high- and low-income groups in terms of
environmental correlates of active travel/outdoor leisure activity will be illuminated.
One essay quantifies the environmental role in daily active travel, focusing on two
measurements: frequency (i.e., the number of incidents of active travel) and duration (i.e.,
time spent in active travel). Both are common measurements of demand for active travel,
specifying the likelihood of choice and the extent of participation, respectively. As generally
assumed, a built environment with a certain type of physical configurations (e.g., high
density, mixed use, well-connected street patterns, and access to public transit) can
encourage people to walk more. However, this essay argues that the role of the built
environment can be overestimated in two ways. One is that, as often mentioned in previous
studies, people (especially, high-income groups) who prefer to walk (or bike) can choose
neighborhoods supportive to their travel preference (i.e., residential self-selection). The other
is that people (especially, low-income groups) who live in urban neighborhoods due to

19
housing affordability and job opportunities are more likely to engage in active travel since
they have limited options for motorized travel rather than because they live in neighborhood
supportive to active travel.
Revisiting the role of the built environment among different income groups, this
study highlights another aspect of neighborhood environment at a highly disaggregated level
which discourages daily active travel. The deterrent role of environmental constraints (i.e.,
risk-scape) in a neighborhood might provide reasonable clues to the likelihood that people
who live in a neighborhood with a walking-friendly urban form engage less in daily active
travel. To better understand the role of various neighborhood environmental characteristics,
this study will investigate the complex relationship between neighborhood environment and
active travel, while controlling for individual/household characteristics, attitudinal factors,
and residential sorting (e.g., neighborhood type). The specific research questions to be
answered are as follows.
• What aspects of neighborhood environmental characteristics may facilitate or
constrain daily behavioral decisions on active travel?
• Does the influence of neighborhood environment vary across different income
groups? If so, which factor is the most important in influencing active travel of high-income
groups and of low-income groups?
•How does attitudinal factor affect active travel? Are there both direct and indirect
effects of attitudinal factor on walking or/and biking?
• Does the built environment still matter in explaining the variation in active travel
time and frequency, when residential self-selection is addressed?

20
The other essay explores environmental correlates of outdoor leisure activity. It is
often suggested that neighborhoods with highly accessible local resources (e.g., recreational
parks) can encourage people to participate in outdoor leisure activity. As in empirical studies
of travel behavior, however, the residential self-selection issue is also problematic in
figuring out the role of the built environment in outdoor leisure activity.
Focusing on another aspect of neighborhood influence, this essay speculates that
low-income groups more exposed to environmental barriers and constraints engage less in
outdoor leisure activity even though they live in neighborhoods with more access to parks.
Furthermore, the relationship between outdoor leisure activity and daily active travel will be
investigated to see whether different domains of physical activity in the outdoors are
substitute or complementary. Potentially, people who engage more in daily active travel are
less (or more) likely to participate in outdoor leisure activity while considering active travel
as another type of outdoor exercise (or reflecting the propensity for outdoor activity).
To better understand behavioral decisions on outdoor leisure, this essay will
investigate the complex relationship between the built environment and outdoor leisure
activity, while controlling for individual/household characteristics, attitudinal factors, daily
active travel, and residential sorting (e.g., neighborhood type). The specific research
questions to be answered are as follows.
• What aspects of neighborhood environmental characteristics facilitate or
constrain daily behavioral decision on outdoor leisure activity?
• Does the influence of neighborhood environment vary across different income
groups? If so, which factor is the most important in influencing outdoor leisure activity of
high-income groups and of low-income groups?

21
• How does attitudinal factor affect the participation in outdoor leisure activity?
Are there both direct and indirect effects of attitudinal factor on outdoor leisure activity?
• Does the built environment still matter in explaining the variation in outdoor
leisure activity time and frequency, when residential self-selection is addressed?
• How does daily active travel affect outdoor leisure activity (i.e., substitute or
complement)?

1.4. ORGANIZATION OF THE THESIS

As an introduction to this dissertation, I have demonstrated the overall background of the
research topic, major gaps/problems in previous studies, and current statements in Chapter 1.
The remaining parts are organized in five chapters.
Chapter 2 reviews the literature on environmental correlates of active travel/outdoor
leisure activity, starting with several theoretical foundations devoted to the topic. Along with
the summary of previous findings, I also briefly discuss causality and residential self-
selection given the importance of methodological issues. In Chapter 3, I suggest the
conceptual framework for the prediction of active behavioral outcomes (i.e., daily active
travel/outdoor leisure activity), developing a socio-ecological theory that explains the
interplay between human behavior and environment. Before empirical analysis, Chapter 4
elaborates how a cross-sectional/quantitative research approach was designed in terms of
data and method. Both data sources/structure and study area will be described, followed by
the definition/measurement of variables. The descriptive statistics of variables will be
summarized prior to model specification of path analysis.

22
The following two chapters are the core of this dissertation. In Chapter 5, I explore
the relationship between neighborhood environment and duration/frequency of daily activity
travel, comparing the results of different income groups (i.e., high- and low income groups).
In Chapter 6, I investigate the relationship between neighborhood environment and
duration/frequency of outdoor leisure activity, specifying three types (e.g., exercise/sport,
park activity, and leisure active travel) associated with physical activity. Finally, Chapter 7
concludes this dissertation with major findings in previous chapters, land-use/transportation
policy implications for active living, and further discussion of contributions, limitations, and
recommendations for future study.

23
CHAPTER 2.
LITERATURE REVIEW

2.1. THEORETICAL FOUNDATION

Previous studies have relied on several theoretical foundations from the view of micro-
economics, human psychology, and environmental behavior so as to understand how people
make behavioral decisions on travel/activity. First of all, utility-maximizing theory, from the
micro-economic perspective, provides a crucial clue to how people choose travel/activity as
behavioral outcomes. In general, the choice of travel/activity is determined by the
individual’s own feasibility and cost-benefit calculation, as well as self-interest given the
choice set (Ben-Akiva and Lerman, 1985; McFadden, 1974). It is assumed that people
assess alternatives and finally choose the best option under budget restrictions, on the basis
of economic rationality that maximizes utility and minimizes cost (Dela, 1989).
However, utility-maximizing theory has a limit in that it rarely considers both the
nature of travel demand for activities and the role of other factors that can affect behavioral
decision: attitudes of economic agents and behavioral feedback in the process of decision
(Garling et al., 1998; Maat et al., 2005). People never move for travel per se; rather, they
change their location in order to engage in unique activities and amenities that the
destination offers (Maat et al., 2005). In addition, “remembered (or memorized) utility” as
an example of behavioral feedback proves that an irrational method of making travel
decisions is also significant (Kahneman et al., 1997).

24
By extension, activity-based approach plays as a complementary platform for utility-
maximizing theory. The basic assumption is that travel demand is derived from the need or
desire to participate in various activities and amenities that are spatially allocated and
temporally varied (Bowman and Ben-Akiva, 2000; Kitamura, 1988). In activity-based frame,
the cost-benefit calculation is still valid: that is, the trade-off between the benefit of the
activity and the cost of travel is a major determinant of the choice of activity (Goodwin and
Hensher, 1978). The built environment factors into this process not only by providing
amenities for activity but also by affecting travel cost. The concept of accessibility explains
how urban form (e.g., high density, mixed land use, street network, and proximity to
destinations) affects travel cost as well as travel mode choice (Boarnet and Crane, 2001;
Cervero and Kockelman, 1997).
Distinguishing travel for activity, positive utility of travel explains the possibility of
traveling for its inherent value, raising a question about a travel demand “purely” derived
from something (Handy, 2004: 12, Handy et al., 2002). For example, attractive scenery can
yield a relatively long distance for travel “as an end in its own right” (e.g., driving for
sightseeing) (Handy, 2004: 12). A tenet of utility-maximization theory (i.e., minimizing
travel cost to maximize travel benefit) might be modified to suit this framework. It is
through choosing alternative modes slower than driving that we can expect an increase in
travel time, but non-motorized travel (e.g., walking and bike) can generate other types of
positive utilities. In this context, minimizing travel time does not necessarily mean
maximizing utility.
From the psychological perspective, planned-behavior theory focuses on the role of
belief, arguing that belief is a unique factor that induces travel/activity as a behavioral

25
outcome (Ajzen, 1991). There are three types of belief: behavioral, normative, and control
beliefs, which affect attitudes, subjective norms, and perceived behavioral control
respectively (ibid). In explaining behavioral outcomes, Ajzen proposed a framework that
stresses the mediating role of intention between attitude, subjective norms, perceived control,
and behavior. However, planned-behavior theory has some limitations in operationalizing
the conceptual frame: how to measure and distinguish specific types of psychological factors
and how to incorporate the environment as a behavioral setting into this frame. Even though
positioning the role of environment might go beyond the scope of the planned-behavior
theory, apparently the built environment can play its role in directly formulating perceived
control and attitude thereby indirectly affecting intention.
From the environmental-behavior perspective, the theory of environment
enhancement emphasizes the role of environment in promoting well-being and health,
employing environmental frames at meso-macro scale such as “environment stress”,
“neighborhood disorder”, and “urban imageability” (King et al., 2002: 18). For example,
SRT (stress-reduction theory) pays more attention to the relationship between exposure to
natural elements, affective response, and relaxed attention (Ulrich, 1981). A physical
environment with green spaces plays a restorative role by providing a motivation to exercise
as well as the opportunity to reduce stress (Olmsted, 1952). The restorative role of
environment has been empirically supported by the evidence of positive impacts of the
exposure (or access) to greenness on physical activity (Giles-Corti et al., 2002). In addition,
the theory of defensible space pinpoints physical design features, such as high-density, high-
rise residential buildings with poor site planning, which decrease residents’ sense of
defensible space (Newman, 1973). The theory of environmental incivilities notes that

26
neighborhood disorder, as indicated by environmentally negative signs such as broken
windows and graffiti, can decrease use of public space for socializing and recreation within
neighborhoods (Perkins et al., 1993).
Drawing psycho-sociological perspective into human behavioral decisions, many
behavioral theories explain a behavioral change and a life style modification: social learning
(Bandura, 1969), social cognitive (Bandura, 1991), social marketing (Andreasen, 1995), self-
efficacy (Bandura, 1997), health belief model or planned behavior (BecKer, 1974; Ajzen,
1991), and so on. Despite their different foci, behavioral theories have in common the tenet
that change in behavior derives from motivations of change found at the interpersonal or
intrapersonal levels (e.g., attitudes, beliefs, self-confidence, perceptions, and other internal
values). However, both environmental approaches and behavioral approaches have
limitations in that the former usually ignores individual (or group) differences in response to
exposure, and that the latter also fails to consider various environmental influences (e.g.,
economic, social, and cultural constraints and facilitators) (Stokols, 1996).
The theory of social ecology suggests both the holistic view of the human-
environment and some dialectic unification between behavioral - and environmental
approaches, inviting the concept of “ecology” which generally implies the interrelationship
of organisms and environments (Hawley, 1950; Stokols, 1992; Stott, 2000). The core
assumption of the social-ecological approach is that individual behaviors are influenced by
multiple facets of environment (e.g., physical, social, political, and cultural) as well as by
multilevel system settings (micro-, meso-, exo-, and macro system) (McLeroy et al., 1988;
Reidpath et al., 2002). In addition, social-ecological theory provides a crucial platform for
health studies, viewing human health as outcomes driven by dynamic interplays between

27
factors at the individual level and contextual values at the environment level. However, this
perspective rarely offers explicit explanation of the pathway through which intrapersonal
components compound the relationship between built environment and behavior (Bauman et
al., 2002). Moreover, the theory does not fully address the possibility that a certain type of
choice (e.g., behavioral choice for travel/activity) is subject to other choices (e.g., residential
location choice) within a decision hierarchy (Handy, 2004; Van Acker et al., 2010).
Even though many theories have suggested a unique angle that helps us to better
understand the mechanism of travel/activity decision making, in general, it appears that no
single theoretical foundation clearly explains why, how, and under what condition people
engage in active travel/outdoor leisure activities, in particular. Given the absence of
agreement on an explicit framework, a more comprehensive conceptual approach will be
necessary to avoid a confused interpretation and to fill the gap in our understanding of
behavioral decision.

2.2. FACTOR AFFECTING ACTIVE TRAVEL AND OUTDOOR LEISURE

Individual/household characteristics
It is well-established that the demand for travel/activity participation generally
reflects a variety of personal factors and household characteristics, leading to different
behavioral patterns (Moss et al., 1969; Chapin, 1974; Ferge, 1972). Some studies especially
devoted to activity-based approach have shown that travel behavior is the result of daily
decisions on activity participation as a function of socio-demographic factors (Chapin, 1974;
Lu and Pas, 1999). Such ideas justify controlling for multiple personal/household
backgrounds in exploring the environmental correlates of active travel/outdoor leisure.

28
While conducting empirical analysis, many studies employed specific populations
distinguished by age, gender, ethnicity, income, car-ownership, and household composition
to see whether each subgroup has different correlates (Agrawal and Schimek, 2007; Arnold
and Shinew, 1998; Wen et al., 2007; Turrell et al., 2013). As often pointed out, individual
correlates of active travel/outdoor leisure vary widely in their effects. Some variables are
similarly affecting, but others are not.
More specifically, among demographic factors, age is consistently found to have a
negative influence on active travel/outdoor leisure in adult samples (Greenwald and Boarnet,
2001; Clifton and Dill, 2005; Marans, 1972; Gobster, 2005), confirming that the demand for
activity decreases and inactivity rates rise as people get older (Fredman et al., 2011; Van
Stralen et al., 2009; Bolen et al., 2000; Payne et al., 2002). However, such a relationship is
not apparent in a certain type of active travel (e.g., for leisure), showing paradoxically that
older people engage more in leisure walking (Saelens et al., 2012).
Educational attainment is another variable that consistently and positively affects
active travel/outdoor leisure (Pucher et al., 2011; Agrawal and Schimek, 2007; Buehler et al.,
2011; Yang et al., 2012, King et al., 2000; Dishman et al., 1985). Although a negative
association with walking trips was also observed (Clifton and Dill, 2005), a leisure/active
life style is fairly dominant among people with high levels of education, which makes them
participate more frequently in active travel (Cerin et al., 2009).
On the other hand, both gender and racial composition affects active travel and
outdoor leisure. While a female is likely to utilize more active travel than a male, the
prevalence of outdoor leisure activities (in general, leisure-time physical activity) appears to

29
be higher among males (Agrawal and Schimek, 2007; Pucher and Renne, 2003; Buehler et
al., 2011; Sylvia-Bobiak and Caldwell, 2006; Kemperman and Timmermans, 2008).
The findings on the influence of racial groups are slightly mixed, but a racial/ethnic
gap is repeatedly reported. Non-White groups are positively associated with active travel (in
particular, walking for transportation) and negatively associated with leisure-time physical
activity including walking for recreation (Wen et al., 2007; Pucher et al., 2011; Saelens et al.,
2012; Marans, 1972; Gobster, 2005). However, a confused result also indicated that non-
White groups are less likely to engage in utility-walk trips than Whites after controlling for
individual/household variables (e.g., income and car-ownership) (Agrawal and Schimek,
2007). The mixed evidence suggests that racial composition may confound the role of the
factor that reflects socio-economic status.
Socio-economic factors such as household income and employment are often
acknowledged as a key predictor of the likelihood of engaging in active travel/outdoor
leisure (Marans, 1972; Cerin et al., 2009; Van Acker et al., 2010). The evidence is fairly
consistent that people with lower SES engage more in active travel (in particular, walking
trips for transportation), whereas people with higher SES participate more in leisure-time
physical activity including leisure walking or outdoor leisure (Ford et al., 1991; Bauman et
al., 2012).
Even though physical activity in the domain of occupation, household, and
transportation is commonly observed among low income households, some studies notified
that low-SES groups were often categorized as sedentary or inactive groups because they
rarely achieve the recommended level of physical activity, and because no leisure-time
activity is highly prevalent compared with high-income households (Giles-Corti and

30
Donovan, 2002; ALD, 2010). Notably, a study using samples from Glasgow in Scotland
found that economically advantaged respondents record more active travel but less leisure-
time physical activity (Ogilvie et al., 2008). Those observations are somewhat contradictory
to previous findings, but this might represent the underlying differences in spatial/cultural
context between the U.S. and Europe, or perhaps indicate that high-income groups might
demand less recreation activity because the opportunity cost of earnings is relatively high
(Buehler et al., 2011; Fredman et al., 2011).
The opportunity cost of earnings provides a clue to the negative association between
employment and active travel/outdoor leisure activity, explaining the lack of available time
to walk or enjoy recreation among the employed populations (Pucher et al., 2011; Van Dyck
et al., 2010; Kirk and Rhodes, 2011). However, the evidence on the influence of employment
is still mixed. It is found that the employed take higher percentages in any walking but also
engage more in leisure physical activity than people who are unemployed or not in the
workforce (Buehler et al., 2011; Van Stralen et al., 2009).
Personal characteristics in the psychological dimension, such as motivation, self-
efficiency, perceived control, preference, and attitude toward specific travel/activity, are
often conceptualized in previous studies as a key variable which facilitates or constrains
active living based on planned-behavior theory (Driver and Knopf, 1977; Dishman et al.,
1985; Tsai and Coleman, 2009; Panter and Jones, 2010; Degenhardt et al., 2011; Wyrwich et
al., 2006; Bauman et al 2012; Saelens et al., 2012).
More specifically, attitudinal factors may confound the relationship between the
built environment and walking/leisure physical activity. Attitudinal factors related to
travel/activity not only affect directly behavioral outcomes but also provide some clues to

31
residential location choice which allows people to live in the built environment with
different characteristics (Cao et al., 2007). In addition, prior research suggests that the role of
the built environment becomes less significant when individual attitudinal factors are
incorporated (Bagley and Mokhtarian, 2002).
Notably, the evidence shows that attitudinal factors are the strongest predictor of
active travel/physical activity among various explanatory variables (Kitamura et al., 1997;
Wyrwich et al., 2006; Lee and Moudon, 2006). Another recent study that highlighted the
importance of attitudinal factors found that the impact of environmental factors varies across
the levels of personal walking attitudes, concluding that the built environment (e.g., urban
form variables) matters more for people with positive attitude to walking, whereas the social
environment (e.g., violent crime rates) matters more for people with negative attitudes to
walking (Joh et al., 2012).
Along with the psychological factors mentioned above, physical health problems are
commonly cited as a main reason for inactivity (Trost et al., 2002; Sugiyama and Thompson,
2007; Van Stralen et al., 2009; Bauman et al., 2012). Many studies rely on self-rated or
doctor-diagnosed measurements for the level of disability related to mobility, employing
categorical scales (Caspi et al., 2013; Spivock et al., 2008; Taylor et al., 1998; Christensen et
al., 2010). The general findings confirm that physical disability reduces the
ability/opportunity to engage in active living, consistently suggesting that adults with a
disability are less likely to meet the recommended level of physical activity and that the
prevalence rates of arthritis leads to limited activity, compared with adults without such
physical barriers (Rimmer et al., 2008; Hootman et al., 2003).

32
Two household characteristics are often highlighted as key variables: car availability
and the presence of children. Research consistently finds a negative association between car
availability and active travel, suggesting that people with available cars in a household are
less likely to engage in active travel (Buehler et al., 2011; Turrell et al., 2013). According to
NHTS data, the largest increase in active travel has been observed among people without the
access to cars over the last decade (Pucher et al., 2011), and they also indicated a strong
environmental correlate on active travel compared with people who have available cars
(Ogilvie et al., 2008). This evidence might recall the theory of travel behavior based on
utility maximization which underlines the feasibility of travel options and the calculation of
their costs/benefits (Burbidge and Goulias, 2009).
Unlike the strong influence of car availability on active travel for utility, the
relationship to leisure active travel is less clear or not significant (Cerin et al., 2009; Agrawal
and Schimek, 2007). Moreover, empirical findings on the relationship between the access to
private vehicles and outdoor leisure consistently revealed an opposite result, implying that
the possibility of making certain discretionary travel choices (in particular, for leisure) may
increase with the availability of cars (i.e., high mobility) (Ogilvie et al., 2008; Sugiyama and
Thompson, 2007).
Furthermore, the presence of a child in a household was often incorporated as an
explanatory variable in prior leisure or physical activity studies but less considered in active
travel research. Some studies found that having children is positively associated with the
likelihood of participation in leisure activity, confirming that the presence of children serves
as an opportunity for a family gathering (Marans, 1972; Schipperijn et al., 2010). On the
contrary, however, there are also empirical results which demonstrate a negative influence

33
on outdoor leisure, as well as on overall physical activity (Fredman et al., 2011; Godbey et
al., 2005; Bauman et al., 2002). In interpreting the negative relationship, several studies
suggested that it is not only physical burden but also the amount of time spent on taking care
of children that might reduce the opportunity to engage in leisure activity (Harrington et al.,
1992; Brown et al., 2001).
The current literature from activity-based/time-use perspectives also points out that
lack of available time is a major constraint explaining non-participation in activity/travel
(Dunn et al., 1999; Pate et al., 1995; Trost et al., 2002). The concept of space-time prism in
the field of time geography offers a crucial theoretical basis, highlighting the importance of
space- and time components in behavioral decisions on daily activity/travel (Hägerstrand,
1970). In this framework, the argument is that time allocation to some activities can be
determined by time allocation to others since each individual has a limited time budget
(Pendyala and Goulias, 2002).
Given such coordination, prior studies paid much attention to time allocation for
working- or travel time in explaining behavioral decisions on participation in out-of-home
activities including leisure (Golob, 2000; Godbey et al., 2005; Christian, 2009). The findings
consistently confirmed a trade-off between travel and activity, and between different types of
activities, implying that the amount of travel time decreases activity demand and that
households maximize utility by properly combining working and leisure hours (Maat et al.,
2005; Vaara and Matero, 2011).
Furthermore, several studies investigated how the propensity for active travel affects
leisure physical activity, and vice versa, in order to identify the relationships between
domains of physical activity (Meloni et al., 2009; Forsyth et al., 2007) as well as the relative

34
contribution to total physical activity. Some studies found a positive association between
active travel and physical activity, suggesting that active travel contributes to greater
physical activity both self-reported and accelerometer-measured (Thommen et al., 2007;
Sisson and Tudor-Locke, 2008). It is also reported that the likelihood that active travel users
will meet the guideline for leisure-time physical activity is higher than that for non-active
travel users (Moore et al., 2012), and that either the difference in transportation-related
walking or the propensity for active modes of travel can be explained by the participation in
leisure physical activity (Cerin et al., 2009).
In the context of total physical activity, a study using a sample of adults in the UK
consistently demonstrated that total physical activity is also higher among people who
reported high levels of active travel, concluding that substantial physical activity can be
accumulated by active travel (Sahlqvist et al., 2012). Such evidence might bolster the
argument that the provision of neighborhood environment conducive to active travel can
serve as a promising approach to improving the level of physical activity (Hoehner et al.,
2005).
However, the contribution of active travel to overall physical activity is still
inconclusive. At least, it appears to be reasonable when the degree of physical activity in
other domains is fixed. Pointing to this condition, several recent studies cast a skeptical eye
on a dose-response relationship between active travel and recreational/total physical activity
(Merom et al., 2003; De Nazelle et al., 2011; Ogilvie et al., 2008). Rather than the
commensurate increase in total physical activity, they suggested substitute relationship
assuming that the increase in active travel might be translated into some compensation for
the reduced physical activity in other domains.

35
Neighborhood environment
A large volume of review papers and empirical studies has explored the relationship
between active travel (in particular, walking) and the built environment, highlighting the role
of land-use patterns, transportation systems, urban design, and other neighborhood
characteristics (Saelens et al., 2003; Owen et al., 2004; Saelens and Handy, 2008).
More specifically, comparative studies that are more centered on walking behavior
of residents in different types of neighborhood found that walking is more prevalent in
neighborhoods based on design elements advocated by New Urbanists (Saelens et al., 2003;
Handy, 2004; Devlin and Frank 2009). However, some findings reveal no variation in
walking for exercise between high- and low-walkable neighborhoods (Handy, 1992; Handy,
1996), implying that the observed difference in walking behavior results mainly from the
gap in walking for utilitarian purposes rather than for exercise (Saelens et al., 2003).
On the other hand, correlation studies have investigated the relative magnitude of
design elements conducive to walking behavior (Badland and Schofield, 2005; Saelens et al.,
2003; Saelens and Handy, 2008; Frank and Engleke, 2000). More specifically, as a common
proxy for either accessibility or closer proximity to destinations, both density (i.e.,
population and employment) and diversity (i.e., the degree of mixed use) revealed fairly
sufficient evidence on the positive association with walking (Frank and Pivo, 1995; Cevero,
1996; Kockelman, 1997). Street connectivity, which represents more direct routes and
shorter distance to potential destinations, has been recognized as a significant factor that
encourages people to walk (Greenwald and Boarnet 2001). Both sidewalk and transit access
also have a positive influence on walking trips (Kitamura et al., 1997).

36
Consistent with comparison studies, major findings from correlation studies indicate
that different walking purposes have different functions and correlates with the built
environment (Owen et al., 2004), suggesting that a factor affecting walking for transport is
not necessarily related to walking for recreation and total walking (Lee and Moudon, 2006;
Saelens and Handy, 2008). Those studies also conclude that the built environment somewhat
supports walking, but the link between specific elements of the built environment and
specific types of walking is still mixed.
Along with walking behavior, the findings from physical activity studies also
demonstrate that a supportive environment appears necessary to increase levels of physical
activity, but its contribution might be conditional and insufficient on its own (Giles-Corti
and Donovan, 2003; Brownson et al., 2001). Understanding this insufficiency leads
researchers to look at another aspect of the built environment as well as other factors beyond
the built environment that affect individual behavioral decisions on active living.
It is well acknowledged that the built environment not only provides opportunities
but also contains various barriers that discourage people from engaging in walking and other
types of activities. However, a large volume of literature puts an emphasis on the facilitating
role of the built environment (Ewing and Cervero 2001; Ewing and Cervero 2011), and the
deterrent role is less explored. The reason behind this scant attention is that the context of
environmental barriers is broad and related to psychological processes.
Raising the issue of safety and security, several studies have identified specific
barrier sources embedded in physical/social environments (Loukaitou-Sideris, 2004;
Loukaitou-Sideris, 2006). The sources commonly cited in literature include crime, physical
disorder, social disorder, unattended dogs, traffic-related accidents and injuries, and air

37
pollution (Foster and Giler-Corti, 2008; FHA, 1994; Cutts et al., 2009; King, 2000; Sallis et
al., 1997; Wilcox et al., 2000; Frumkin et al., 2004; Marshall et al., 2009). Lack of safety
arising from crime is considered a significant barrier to walking and outdoor physical
activity for leisure in adults and urban youth (Hawthorne, 1989; Gomez et al., 2004). From
the perspective of criminology, physical disorder (e.g., vandalism, litter, graffiti, abandoned
buildings, and disrepair) and social disorder (e.g., loitering, drug use, and drunkenness) can
increase the level of fear of crime (LaGrange et al., 1992). While there is a relatively clear
link between neighborhood safety and (either perceived or objectively measured)
neighborhood disorder, the link between safety and walking/physical activity is still elusive
(Loukaitou-Sideris, 2006; Loukaitou-Sideris, 2004).
Traffic-related safety concerns have been recognized as a potential barrier that
reduces the ability to support street activities (Williams 1995). Major hazard sources that
directly increase safety concerns are traffic collisions, pedestrian injuries, and lack of
pedestrian facilities (Cutts et al., 2009; Romero et al., 2001). Other traffic conditions, such as
speed, volume, parked vehicles, and busy intersections, may also contribute to decreasing
the level of perceived safety (Loukaitou-sideris, 2004). In general, an area with a high
volume of activity attracts traffic along with pedestrian volume and thus leads to high
exposure to collision risk and pedestrian injury (Levine et al., 1995; Roberts et al., 1995). As
a result, it appears that road capacity might be negatively related to recreation utility in
streets (Williams 1995). However, it is still less established whether there is a clear link
between traffic-related barriers and physical activity, including walking (Wilcox et al., 2000;
King et al., 2000).

38
In line with traffic-related barriers, exposure to air pollution is also often viewed as
an environmental constraint that prevents people from engaging in activities in local streets
(Frumkin et al., 2004; Marshall et al., 2009). A potential link between sedentary behaviors
and urban outdoor air pollution has been highlighted as a pathway to health impairment
(Ezzati et al. 2002; Hill et al. 2003). Noting the gap in the air quality between neighborhoods
with different levels of walkability, research argues that active travel in pedestrian-friendly
environments brings about negative health impacts by increasing pedestrians’ level of
exposure to pollutants, and that there is a trade-off between the risks and benefits of
walkable neighborhood design (Frank and Engelke, 2005; Marshall et al., 2009; De Nazelle
et al., 2009). However, so far, there is only limited evidence to support this argument.
Contrary to forecasts, earlier studies failed to find strong evidence of the relationship
between neighborhood safety and physical activity including walking. The deterrent role of
various environmental hazards such as crime, traffic incidents, and air pollution is still
unclear (Romero et al., 2001; Murray et al., 1995; Sallis et al., 1997). Heterogeneity in
research design offers possible explanations for inconsistent findings. As frequently noted,
different populations, subject areas, dependent variables, and measurement of built
environments lead to mixed results (Loukaitou-sideris, 2004).
Environmental association varies across individuals with different demo-socio-
economic characteristics as well as neighborhoods where individuals with similar
characteristics are clustered. Safety concern is related to walking for White people but not
for Black people (Hooker et al., 2005). Neighborhood safety affects differently residents’
walking, depending on spatial contexts (i.e., urban vs. suburban; low-income vs. high-
income neighborhood) (Hovell et al., 1999; Sallis et al., 1989). For example, as often

39
observed, there are a limited number of well-maintained sidewalks with nice scenery in low-
income communities. Given these neighborhood characteristics, people who live in such
neighborhoods have limited opportunity to walk for leisure (King 2000; Brownson et al.,
2001). There is limited pedestrian access to local amenities as trip destinations in suburban
areas where high-income communities predominate. Bypass roads characterized by wide
street patterns and fast traffic in such neighborhoods rarely allow for safe pedestrian crossing.
This physical condition can lead to pedestrian injuries (Southworth, 2005; Ewing and
Dumbaugh, 2009).
Different classifications and measurements of the dependent variable also explain
inconsistent results. For example, literature found that safety is positively associated with
walking for leisure (Ball et al., 2007; Suminiski et al., 2005), but not significantly associated
with walking for transport and walking frequency (Pikora et al., 2006; Doyle et al., 2006). In
addition, there are few associations between safety and total physical activity. Safety
concerns arising from crime, traffic accidents, and air-pollution are more applicable to
outdoor activities than to indoor activities (Foster and Giler-Corti, 2008). However, in
general, total leisure physical activity aggregates activities in various behavioral settings
(both indoor and outdoor), and this might mask the significant influence of environmental
barriers on outdoor activities.
The variation in safety measurement provides another source of mixed findings.
Neighborhood safety has often been either subjectively measured by individual
perception/assessment or objectively measured by spatial and social data sources. The two
measurements allow researchers to capture neighborhood safety in different ways, but there

40
is a discrepancy in levels of significance between perceived safety and objective safety
(Sooman and Macintyre, 1995).
In sum, there are several points that require more detailed identification and
assessment. First of all, environmental correlates of active travel/outdoor leisure might be
more dynamic among low-income/minority groups than other subpopulation groups, as their
neighborhoods generally offer both environmental opportunities and barriers. The question
regarding the relative importance of the supportive- and deterrent roles of neighborhood
environment remains unanswered. To address this, researchers should simultaneously
consider environmental opportunities and barriers to inform a more comprehensive and
promising policy strategy to remedy the physical inactivity of low-income populations.
Given the psychological process by which behavioral decisions on active
travel/outdoor activity are constrained by individual perceptions of risk, some studies argue
that subjective measurement may predict behavioral outcome better than objective
measurements (Foster and Giler-Corti, 2008). However, there is still a need to quantify the
role of objective risk since high exposure to neighborhood risk (e.g., crime incidents) can
substantially decrease the willingness of low-income groups to engage in active travel even
though they live in highly walkable neighborhoods (Doyle et al., 2006; Sallis et al., 2009;
Loukaitou-Sideris and Eck, 2007).
Moreover, previous studies fail to properly capture local environmental risk by
incorporating the aggregate at an upper level or predefined geographic unit (e.g., county
level or city level) (Piro et al., 2006; Doyle et al., 2006; Van Lenthe et al., 2005; Joh et al.,
2012). Since local outdoor environments (including neighborhood streets) are a popular
behavioral setting where 66% of physical activity (including walking) occurs (Loukaitou-

41
Sideris, 2004), it appears reasonable to measure various barrier sources at the highly micro
level.
Physical activity includes various types and locations/domains where such activities
are embedded. However, most studies have investigated environmental correlates of
(aggregate) physical activity without the specification. Ignoring types and locations of
physical activity can yield misunderstandings on the role of environment. To properly
identify environmental influence, research should suggest a behavior-specific approach that
disentangles specific types of physical activity in different domains (e.g., active
travel/outdoor leisure). In addition, a location-specific approach that sorts out different
settings (e.g., outdoor/indoor activities) and distinguishes outdoor activity destinations may
be helpful to figure out the spatial match between the built environment (predictive variable)
and physical activity (outcome variable).
Furthermore, the relationship between active travel and other domains of physical
activity ought to be tested for a better understanding of the big picture of total physical
activity on a daily basis. As several scholars stress, in general, people with similar socio-
economic status engage in the same (total) amount of physical activity (Rodriguez et al.,
2006; Forsyth et al., 2008; Krizek et al., 2009). Under the condition that the individual
budget of physical activity is fixed, the contribution of active travel to increasing overall
physical activity may not be as straightforward as advocates suggest.

2.3. CAUSALITY AND RESIDENTIAL SELF-SELECTION

There is much evidence of the significant relationship between the built environment
and active travel (including physical activity) in many previous studies, but the issue of

42
causality and residential self-selection should be addressed to make sure that the observed
relationship is relevant and valid. More specifically, both the lack of information based on a
longitudinal dataset and a hypothesis derived from a less explicit theoretical frame makes it
difficult to establish a causal link (Boarnet, 2004; Handy et al., 2006; Guo, 2010).
Residential self-selection might also bring another challenge to behavioral studies. That is,
the observed association between the built environment and human behavior (i.e., active
travel/outdoor leisure) might be spurious, while reflecting the relationship between the built
environment and residents’ characteristics (Bhat and Guo, 2007).
To properly address methodological issues and hence make valid estimates of the
relationship, previous studies suggested several analytic methods including 1) simply
controlling for the attributes of individual decision-makers (Kitamura et al., 1997; Lund
2003; Khattak and Rodriguez, 2005); 2) constructing multiple pathways between variables in
one regression model (Bagley and Mokhtarian, 2002; Acker and Witox, 2010); 3) employing
instrumental variables that represent potential endogeneity in residential choice (Boarnet and
Sarmiento, 1998; Greenwald and Boarnet, 2001); 4) developing sample selection by
Heckman or 5) designing a (quasi-) longitudinal frame that takes into account attitudinal
factors before and after household moving (Krizek, 2000).
However, as often acknowledged, incorporating control variables that directly reflect
attitudinal factors might have a limited capacity to expose the true effect of the built
environment. Furthermore, there are also methodological difficulties in adequately
employing strong instrumental variables that are fairly correlated with one variable (i.e.,
built environment) but not the other (i.e., behavioral outcomes) (Mokhtarian and Cao, 2008).

43
Neither the experimental design nor the longitudinal approach is feasible under the cross-
sectional, secondary data structure.
Given non-experimental, cross-sectional, quantitative, or correlational data, path
analysis as a specific type of SEM approach can allow researchers to deal with mutual
directions among multiple variables, while providing some clues to causal influence with
global assessments of model fit (Buhi, 2007). In this process, the direct, indirect, and total
impact of a specific variable denoted in hypothetical paths can be identified. Most studies
using path analysis have incorporated attitudinal variables into hypothetical frameworks for
the mechanism through which individuals make decisions on their residential locations,
thereby addressing the role of residential self-selection. Specific variables in human
psychological dimensions are lifestyle, travel-related attitude (e.g., specific travel modes,
congestion, air pollution, and so on), and preference for housing and neighborhood (Bagley
and Mokhtarian, 2002; Scheiner and Holz, 2007; Scheiner, 2010; Cao et al., 2007; Bohte et
al., 2010).
More specifically, Bagley and Mokhtarian (2002) investigated the multiple
directions of causality between neighborhood types and travel outcomes (e.g., number of
miles travelled by vehicle, transit, and walk/bike) to capture the confounding role of
residential self-selection. Individual characteristics including demo-socio-economic status
and car availability are also considered as exogenous variables. Attitudinal variables were
defined not only as endogenous variables that indicate the preference for neighborhood and
travel (e.g., pro-high density, pro-driving and pro-transit), but also as exogenous variables
that explain both lifestyle and individual predispositions regarding travel and environment
(e.g., pro-pricing, pro-drive alone, pro-growth, and pro-environment).

44
Neighborhood type reflecting the characteristics of residential location is categorized
into traditional and suburban neighborhoods. The nature of a traditional neighborhood was
loaded as a composite value with continuous/disaggregate measurements rather than a single
dimension. This is because the same neighborhood might have different degrees of
traditional characteristics such as population density and convenience of public transit. As a
result, they found that lifestyle and attitudinal factors are the greatest influence on travel
behavior, whereas the built environment (i.e., neighborhood type) has little influence on
travel behavior and attitude.
Focused on the study area in Germany, Scheiner and Holz (2007) explored the
complex relationship between individual/household values, the built environment (e.g.,
neighborhood type and quality of public transportation), and mode choice (e.g., sharing of
trips via car, non-motorized, and public transit)/VKT. Specific individual compositions fall
into three factors: 1) demo-socio-economic status; 2) lifestyle presented by leisure
preference, life aims, aesthetic taste, and social contacts; and 3) location attitudes
operationalized by subjective importance of neighborhood attributes. Neighborhood
attributes incorporated in this study are the quality of public transportation and the density of
local opportunities for retail, service, and leisure. Results from six analytic models including
car availability revealed that both individual SES and lifestyle directly and indirectly affect
mode choice by mediating the roles of location attitude and neighborhood type. They also
found a significant direct/indirect effect of location attitude on mode choice mediated by the
built environment (i.e., the quality of public transportation). Based on these findings, they
concluded that residential self-selection plays a significant role in the connection between
the built environment and travel behavior.

45
Using the same data and case areas in Scheiner and Holz (2007), Scheiner (2010)
tested whether residential self-selection has a significant impact on trip distances with
different purposes: work, maintenance, and leisure. Results revealed that all types of trip
distance are strongly influenced by SES, but there was no evidence to support substantial
direct/indirect impact of several attitudinal values (e.g., lifestyles and location attitude) on
trip distance, except for leisure trips. In other words, unlike the results from analytic models
for mode choice, residential self-selection did not play a significant role in this study. This
finding implies that the effect of residential self-selection varies across travel outcomes (i.e.,
mode choice vs. trip distance) as well as trip purposes.
Bohte et al (2010) addressed the complex mutual directions among the built
environment, attitude, and travel behavior (i.e., VKT and share of car trip), controlling car
availability and individual SES. Assuming that both residential location choice and travel
behavior can also influence individual attitude, they explored “reverse” causality of the built
environment and travel outcome on attitude. To do this, attitudinal factors toward both travel
and neighborhood were incorporated in analytic models that target homeowner samples
living in three municipalities in the center of the Netherlands.
In the empirical framework, attitude toward each travel mode (i.e., car use, cycling,
and public transportation) was separately loaded, and attitude toward neighborhood
characteristics, representing the subjective importance of distance to activity locations, was
measured by a 5-point scale. The built environment was measured by distance from home to
activity locations such as green areas, railway stations, and city center. As a result, the study
identified the magnitude influence of built environment on attitude, but a small effect of
residential self-selection on travel mode and VKT was found. Moreover, under the reverse

46
effect of travel behavior and the built environment on attitudinal factors, the effect of
residential self-selection disappeared. Controlling for residential self-selection, the
investigators confirmed that the built environment has a direct impact on travel behavior as
well as indirect impact via attitude. They also argue that ignoring reverse influence can
overestimate the effect of residential self-selection.
Studies have often employed cross-sectional data, but those observational studies
failed to establish a causal inference based on time precedence. That is, it is still questionable
whether change in the built environment leads to change in travel behavior. Cao et al (2007)
suggested a more advanced method with a quasi-longitudinal design to overcome this
limitation. In this approach, a study tried to disentangle the independent role of the built
environment on travel behavior (changes in driving and walking), controlling for auto
ownership and attitude toward residential area and travel.
Employing residents who moved into a different neighborhood, the investigator
measured the difference in neighborhood design between previous and current residential
neighborhoods. They found that both attitudinal factors (i.e., preferred neighborhood
characteristics and travel preference) have an influence on the change in travel mode through
the change in the built environment. It was also found that the change in the built
environment was associated with change in travel behavior. Based on those findings, this
study confirmed the significant independent effect of the built environment as well as the
role of residential self-selection on travel behavior.
Studies using path analysis have aimed to identify the role of residential self-
selection on the link between the built environment and travel behavior, taking mediating
and confounding factors into account. However, the evidence from those studies is mixed,

47
Table 2-1 Summary of studies using SEM model
Authors Study area Data/method Variables (Endogenous) Findings
Bagley and Mokhtarian SF Bay Area Cross sectional Distance travelled by modes The greatest influence of LS/ATT on TB
(2002) (vehicle, transit, and walk/bike) Little impact of BE (NT) on ATT/TB
Preference for neighborhood and travel
(pro-high density, pro-driving and pro-transit)
Scheiner and Holz Germany Cross sectional % of trips by mode and VKT Significant impact of ATT(NT) on TB
(2007) (car, non-motorized, and public transit) BE mediates the impact of ATT on TB
Attitude toward travel/neighborhood (ATT->BE->TB)
Scheiner Germany Cross sectional Trip distances by different purposes No sig. impact of LS/ATT(NT) on TB
(2010) (work, maintenance, and leisure) (except for leisure trips)
LS and ATT(NT)
Bohte et al Netherlands Cross sectional VKT and % of car trip BE->TB; BE->ATT->TB
(2010) Aattitude toward each travel mode Reverse influence of BE/TB on ATT
(car use, cycling, and public transportation) Under reverse influence, RSS disappears
Cao et al Northern Quasi-longitudinal Change in mode (driving and walking) BE->ATT->TB
(2007) California Attitude (scale measure/factor-loaded) BE->TB; ATT->BE->TB
(pro-bike/walk, car dependency, safety of car)
Auto ownership/ATT(NT) also controlled

Note: LS=Life styles; ATT=Attitude variable;
TB=Travel behavior;
BE=Built environment; NT=Neighborhood type;
RSS=Residential self-selection

48
and it is still inconclusive. Some studies suggested that residential self-selection plays an
important role in travel behavior (Bagley and Mokhtarian, 2002; Scheiner and Holz, 2007;
Scheiner, 2010), while others concluded that the built environment still matters when
residential self-selection is addressed (Cao et al., 2007; Bohte et al., 2010).
As usual, mixed evidence is generally created by heterogeneity in research design,
with different levels of simplification on multiple relationships, different ways of
conceptualizing mutual dependences, and different ways of operationalizing variables that
constitute multiple relationships and mutual dependences. More specifically, there are
various factors potentially or substantially involved in residential self-selection: lifestyle,
attitude toward travel and neighborhood, habit, belief, and car availability. However,
researchers rarely take all of these factors into account due to the limited availability of data.
Given this uncertainty, ignoring the influence of some factors or simplifying complex
relationships can lead to an underestimation of the effect of residential self-selection (Bohte
et al., 2010).
Howe to define causal direction (i.e., one direction or bi-direction) and the nature of
variables (i.e., endogenous or exogenous) is another crucial point that explains the mixed
results. Some studies suggest bi-directional relationships among attitudinal factors, the built
environment, and travel behavior, considering attitudinal factors as endogenous variables
(Bohte et al., 2010; Bagley and Mokhtarian, 2002). On the other hand, others provide
unidirectional relationships between attitudinal factors and the built environment, and
between attitudinal factors and travel behavior, seeing attitudinal factors as exogenous
variables (Cao et al., 2007; Scheiner and Holz, 2007; Scheiner, 2010). However, no exact

49
agreement has been reached on the underlying nature of mutual causalities and
interrelationships among those factors.
Different methods of operationalizing attitudinal factors make it difficult to draw
general conclusion from previous empirical studies. By and large, attitudinal factors are
latent and entwined with psychological dimensions. In this vein, the measurement is indirect
and varies across the interests of the researcher and the availability of data. For example, in
addressing attitudinal factors, studies either separately incorporate multi-item attitudes in
various dimensions or cluster multi-item attitudes into one composite value. While
composite values can simplify the nature of multi-items and provide a more generalized role
for clustered attitudinal factors, they can also reduce the ability to explain individual role of
multi-items.
Although a substantial amount of literature suggested an SEM approach framed by
diverse research designs, there are still several points less explored in travel behavior
research. More specifically, the majority of research has often focused on travel demand
modeling (Golob, 2003), while the amount of research paying much attention to daily
decisions on travel/activity participation (in particular, active travel and outdoor leisure) is
relatively small (Boone-Heinonen et al., 2011). Conceptually, it is well-documented that
such short-term decisions are normally conditional on other behavioral decisions (e.g.,
residential location choice), but little is known empirically about how the relationship
between decisions in a hierarchy is defined in the specific context of active travel/outdoor
leisure activity.
It is possible that residential self-selection might be influential among some people,
but not among others (Pinjari et al., 2008). In other words, high-income groups have many

50
options in choosing their residential locations. Preference for a certain type of residential
environment or travel mode is one factor among many to be considered, and this makes them
self-select residential locations where they want to live. On the other hand, since low-income
groups have very limited options in choosing their residential locations due to various
reasons (e.g., housing prices, job location, and mobility issues), it is less likely that they will
self-select their residential areas based entirely on the preference. Given this difference
between high- and low-income groups in the choice of residential area, previous findings on
the influence of residential self-selection might be mixed. Therefore, a further study that
separates subpopulation groups in exploring the influence of residential self-selection can
contribute to spreading more robust scientific knowledge on this issue.

51
CHAPTER 3.
CONCEPTUAL FRAMEWORK

Figure 3-1 displays a conceptual framework that incorporates several variables affecting
behavioral decisions on active travel/outdoor leisure activities. As demonstrated in the
literature review, relevant factors derived from socio-ecology theory are generally
categorized into three dimensions: personal, household, and neighborhood physical
environment.
Several relationships marked by “hypothesized path (A)” explain the direct
influence on daily active travel/outdoor leisure activity. The specific components in each
dimension can be considered as either an opportunity that drives people to engage in daily
travel/activity or a barrier that provides some reason for non-participation (King et al., 1995;
Raymore 2002; Crawford et al., 1991). The relative importance of opportunities- and barriers
can be simultaneously investigated in this framework.
It is hypothesized that age, White race, male gender, employment, income, car
availability, safety concern, and crime factors are negatively associated with daily active
travel (i.e., operating as a constraint), whereas education, attitudinal factor, activity density,
street design, and park availability are positively associated (i.e., operating as an
opportunity). Unlike the correlates of active travel, it is postulated that age, White race,
male gender, income, number of children, car availability, and the availability of
sports/exercise facilities are positively associated with outdoor leisure activity, whereas

52
Figure 3-1 Conceptual framework for daily decision on active travel/outdoor leisure activity

53
physical disability, limited available time, and traffic accidents are negatively associated. In
addition, two external factors beyond neighborhood environment (e.g., season and trip date)
can also affect directly outdoor leisure activity. Such influences are based on the assumption
that participation in outdoor leisure is sensitive to weather condition but also related to
weekly scheduling of travel/activity. The specific hypothesized direction is that outdoor
leisure is positively associated with warm weather and weekend days.
More critically, travel behavior theory suggests that the choice of daily
travel/activity as a short-range decision is conditional on long-range decisions (e.g.,
residential/job location choice) (Domencich and McFadden, 1975; Ben-Akiva and Atherton,
1977; Handy, 2004; Van Acker et al., 2010). However, ecological framework rarely
accounts for the dynamic nature of decisions in a hierarchy, often ignoring the relationship
between choices at different levels. This shortcoming can be overcome by incorporating
residential location choice into the daily travel/activity decision model. Such combined (or
joint) choices will be established by multiple paths with the time order.
As a sequential decision, “hypothesized path (B)” speculates that people first choose
their residential location based on individual/household characteristics, and then make
decisions on travel/activity (Bagley and Mokhtarian, 2002; Maat and Timmermans, 2009;
Brownstone and Golob, 2009). In other words, it is assumed that residential location choice
is only a function of exogenous individual/household characteristics. Several factors at the
individual/household level indirectly affect travel/activity decisions via residential location
choice (e.g., neighborhood type) (Bohte et al., 2009). The influence of residential self-
selection can be captured in this process.

54
Presumably, people with high incomes, low propensities for active travel, and car
availability are more likely to choose their residential locations in suburban neighborhoods
(Brownstone and Golob, 2009). Given this choice of residential location, people are less
likely to engage in active travel. The opposite case (i.e., urban neighborhoods) can also be
hypothesized: people with low incomes, high propensities for active travel, and the limited
access to cars are more likely to choose their residential locations in urban neighborhoods,
and hence they are more likely to engage in active travel. Likewise, people with high
incomes, high propensities for outdoor leisure, and car availability are more likely to select
residential areas that offer many leisure-friendly local amenities. Such neighborhood types,
in turn, increase the likelihood that residents will participate in outdoor leisure activity. The
opposite case can also be hypothesized.
Furthermore, “hypothesized path (C)” represents the relationship between different
domains of physical activity (in particular, the transportation- and leisure domain in this
study). As mentioned in the previous chapter, either positive or negative association is
expected in this framework. Reflecting individual active life-styles, the positive association
suggests that people who spend much time engaging in active travel are also more likely to
participate in outdoor leisure activity. On the other hand, the negative association implies
that people who spend much time engaging in active travel are less likely to participate in
outdoor leisure activity, at least where the amount of active travel replaces the opportunity
for outdoor leisure.
It is also notable that several studies suggest reciprocal relationships between
individual/household characteristics (e.g., attitudinal factor), the built environment, and daily
behavioral outcomes (e.g., active travel/outdoor leisure) (Mokhtarian and Cao, 2008; Cao et

55
al., 2007; Handy et al., 2006). For example, people who had a high propensity for active
travel/outdoor leisure in the past choose residential locations where habitual patterns are
supported (Handy, 2004). Neighborhood type possibly influences attitudinal factor toward
travel/activity as a consequence of individual feedback on residential environment (Chatman
2009).
It would be more ideal to suggest a conceptual framework that fully addresses
endogeneity problems stemming from unobserved factors, but observations at two time
points (before/after moving the current residential location) are required to properly capture
the reverse relationship (Bohte et al., 2010). Such information is not available in cross-
sectional data, which allow for the observation in a small time window. Rather, it is more
plausible, in the short run, to conceptualize unidirectional relationships. Furthermore,
reciprocal relationships may exist between car-ownership and neighborhood type. As often
mentioned in previous studies, car-ownership can be influenced by the built environment
(Giuliano and Dargay, 2006) and hence it can mediate the role of the built environment on
travel behavior (Bhat and Guo, 2007). Ignoring this mediating role might overestimate the
influence of the built environment on travel behavior (Van Acker et al., 2010).
However, car-ownership in itself is generally determined by socio-economic
variables at the individual/household level (in particular, income) (Kockelman, 1997). Both
household demographics and the built environments matter in explaining car-ownership
choice, but a more dominant role belongs to household demographics (Bhat and Guo, 2007;
Zegras, 2010). Given this finding, this study focuses more on residential location choice than
on car-ownership choice in exploring the relation to daily behavioral decisions on active
travel/outdoor leisure, while controlling for a rich set of individual/household characteristics.

56
CHAPTER 4.
DATA AND METHODOLOGY

4.1. DATA

The basic analytic model for daily travel/activity was established by merging two major
sources
1
: the NHTS (National Household Travel Survey: 2008-2009) California add-on and
other GIS datasets spatially targeted in Los Angeles County. First of all, as a key secondary
data source, the NHTS California add-on file provides lots of information on person,
household, vehicle, and trip. The first three inventories were collected from a telephone
survey that includes 206 questions on individual demo-social-economic status, household
compositions, personality (e.g., attitudes and perception), and vehicle. Respondents were
asked to complete a one-day travel diary that reports trip information, such as trip-purpose,
mode-choice, travel time/distance, and the location of origin-destinations (x-y coordinates).

1
Potential sources of secondary data for analysis on travel/activity patterns are ATUS
(American Time Use Survey) and NHANES (National Health and Nutrition Examination Survey).
However, both have several crucial challenges in analyzing active travel/outdoor leisure in a daily
context. First, those have limited information on daily travel. ATUS data are focused on individual
time use (i.e., how, where, and with whom Americans spend their time) rather than on travel behavior.
It contains some trip information but is limited in the number, length, and purpose of all trips taken in
a particular month. NHANES data also provide some walking information, but it is more concentrated
on leisure purpose and solely covers walking during the last month. Second, the measurement of data
is based on the (retrospective) self-reported method rather than on travel dairy with objective
measurement. Third, those are more centered on samples nationally representative, so that we can
hardly capture travel/activity patterns in a localized context. Last, there is little information on
household locations (geo-coded) as well as destinations of activities, so that we can hardly measure
highly disaggregated characteristics of neighborhood but also identify specific types (e.g., outdoor or
indoor) of leisure activity.

57
A personalized set of information allows us not only to objectively measure the
outcomes of daily active travel/outdoor leisure activity but also to directly capture
individual/household backgrounds including attitudinal variables often considered as crucial
factors in making the decision on such behaviors. More importantly, the specific geo-coded
household location and trip origin/destination provides the unique platform for measuring
the attributes of various behavioral settings where travel/activity is observed at the micro
level, allowing for a link between individual travel/activity data and spatial data.
Despite those benefits, the main drawback of the NHTS data is that it does not cover
completely all types of active travel as well as outdoor leisure activity. For instance, the data
often ignore walking from parking space (or garage) to indoor place (or the final trip
destination), stairway walking inside buildings, and playing in a front/rear yard within a
housing lot. In addition, some people might argue that one-day travel diary does not fully
capture weekly behavioral patterns which are regular and routine.
However, there is no secondary data which can overcome this limitation, so I finally
considered the NHTS California samples as a major data source for this empirical work.
Unlike travel survey at the national level, this add-on version serves the information on the
objectively measured travel/activity patterns within a fairly localized spatial context with
large sample sizes for a robust inference. The estimated number of households is 3,381
households (6,161 persons and 21,062 trips) located in Los Angeles County (Figure 4-2).
Among those observations, this study targets adult samples
2
(3,528 persons whose age

2
Due to missing values in personal/household variables, in total 3,486 adults were finally
included in analytic models in Chapter 5 and 6.

58
ranges from 18 to 64 years old) since they might have totally different patterns in active
travel/outdoor leisure behavior from other age groups (e.g., youth and elder).
Next, various types of spatial data are also incorporated in this analytic framework
to measure the characteristics of neighborhood. The primary spatial information was made
by SCAG GIS data (2008) that include land use attributes at the parcel level and street
networks. The information of local destinations such as park, public transit, and
sport/recreation facility was obtained from ESRI park GIS data (2010), LA metro data
(2008), and Info USA (2008). The information of risk-scape (e.g., crimes and traffic
incidents) was sourced from Los Angeles County sheriff (2011) and SWITRS (2008-2009)
which provides the specific location of relevant events.
Neighborhood is one of behavioral settings where daily travel/activity originates
from. The point here is how to define the exact spatial boundary of neighborhood or the
salience walkable distance for local trips. Indeed, no agreement on operationalizing the
concept of neighborhood is reached, but I created a ¼ mile buffer from household in order to
measure the characteristics of neighborhood. It is still arguable that people not only walk
more than ¼ mile but also engage in outdoor leisure activity beyond that distance. However,
a ¼ mile buffer area has been often employed as spatial boundary in community design,
transportation, and health studies (Boarnet and Greenwald, 1999; Beimborn et al., 2003;
Clifton and Kreamer-Fults, 2007; Gordon et al., 2011). More specifically, it is supported by
the proponents for New Urbanism who suggest that less than ¼ mile from housing units is
the acceptable distance of pedestrian local travel for both non-work (shopping)- and
recreational (parks) purpose (Calthorpe, 1989; Lund, 2003; Ryan and McNally, 1995).
Furthermore, one study, which dealt with the spatial dimension of daily life in Los Angeles,

59
found that half of adult samples in L.A.FANS data reported “several blocks or streets” as a
neighborhood size (Sastry et al., 2002). When simply converting one block length into 1/10
to 1/8 mile, a ¼ mile buffer around household areas appears reasonable in operationalizing
neighborhood spatial dimension.
Last, even though the samples from California region are available, I confined study
areas in Los Angeles County. The main reason is that Los Angeles is the most populous
county
3
in the United States and contains the City of Los Angeles which is the largest city in
California. More importantly, according to the NHTS California add-ons, Los Angeles
revealed the greatest number of (weighted) walk/bike origins
4
and the highest percentage of
concern about barriers to active travel (e.g., streets too wide, too many cars, unsafe street
crossings, and fast traffic) among the counties in California (McGuckin, 2012).
Acknowledging this situation, such policies as the PLACE
5
program, the Safe
Routes to School strategic plan in the City of Los Angeles, Long Beach Bike Path, and
Walking in LA, have been adopted in Los Angeles County to support the development of
healthy, safe, and active environment. Given the limited ability to measure the micro-level
spatial data across the entire state regions, it is valuable to provide objective/scientific
knowledge and evidence that informs and guides the effective policy and investments
specifically targeted on the confined study area.

3
According to US Census (2010), the population of Los Angeles County is 9,818,605.
4
In total, 3,338,540,000 (weighted) walk/bike origins are recorded in Los Angeles, and those
origins take 48% of the total numbers in California.
5
Policies for Livable, Active Communities and Environments

60
Figure 4-1 Study area (Los Angeles County: 3,381 households)

61
4.2. VARIABLES

Measurement of variables (dependent variables)
In general, as illustrated in Figure 4-2, there are four domains of individual physical
activity: transportation, recreation, household, and occupation. Each domain has a unique
type of physical activity, but sometimes the boundary of domain is less explicit since a
certain activity in one domain is related to another domain. For example, recreation domain
shares specific types of activity (e.g., leisure active travel) with transportation domain.
Recreational physical activity also occurs in various locations including indoor places which
intersect with other domains (e.g., a workout at household or occupation).
Figure 4-2 Four domains of physical activity
PA domain: Total PA
Transportation
(active travel)
Daily
active travel
Recreation
(Outdoor) (Indoor)
Household Occupation
Utilitarian
Recreational
Leisure
destinations (A)
Leisure
destinations (B)
Home
maintenance
Job-related
PA
Outdoor
leisure PA

Acknowledging the difference between transportation and recreation PA domain,
and between outdoor- and indoor leisure destinations, this study focuses on daily active
travel and outdoor leisure physical activity. More specifically, active travel in this study
covers any type of walking/bike trip reported in the NHTS data, but dependent variables in
regression model (Chapter 5) were defined by specific categories: total active travel (both
walking and biking), daily total walking, and daily walking recommended for health (binary

62
scale)
6
. Even though NHTS data provide specific information of active travel in terms of
trip-purposes, this study disregards those categories since it is more interested in total active
travel (e.g., both utilitarian- and recreation).
Meanwhile, I generated regression models that disentangle walking from active
travel since walking is a dominant mode of active travel and it has a function slightly
different from biking. It would be better to separate different modes of active travel because
both walking and bike have different functions and trip performance. However, the model
devoted to daily total biking was not developed into this analytic framework due to very
limited numbers of bike trips recorded in travel survey data.
This analytic model employed different measurements of daily active travel:
frequency and duration which are key measurements in the IPAQ (International Physical
Activity Questionnaire) (Giles-Corti et al., 2006). Frequency generally represents the
likelihood (e.g., habitual choice of participation) of active travel, but it does not effectively
reflect the accumulation of activity (e.g. the extent of participation) that might suggest some
implications for health impact. In this vein, the duration of active travel is also incorporated
in the analysis. It is notable that both frequency and duration observed in travel diary (one-
day survey) are dominated by a large number of zero values.
7
Given this condition, this
study considered both a continuous- and binary scale in measuring active travel.
Next, the NHTS (2009) data provide several subcategories as specific purposes for
leisure trip: 1) ‘Go to gym/exercise/play sports’, 2) ‘Rest or relaxation/vacation’, 3) ‘Visit

6
In regression models for the duration, the value of outcome is 1 if daily total walking time
spent by respondents is 30 minutes or more (recommended for health). On the other hand, in
regression models for the frequency, the value of outcome is 1 if respondents engage in any types of
walking.
7
2,606 samples (approximately, 74% of respondents) did not record any active travel on a
given day.

63
friends/relatives’, 4) ‘Go out/hang out: entertainment/theater/sports event/go to bar’, and 5)
‘Visit public place: historical site/museum/park/library’. Among those categories, this study
focuses on ‘exercise/play sports (Type_1)’ and ‘park activity (Type_5)’ since such types of
leisure activity are closely related to ‘active leisure’ and more likely to be linked to active
living environment than others.
More specifically, total 1,466 trips for both ‘Type1’ and ‘Type5’ were observed in
the NHTS data, but those leisure trips include trip destinations both indoor and outdoor. To
sort out leisure activity in outdoor venues, I exported specific geo-code information of
destinations on satellite map (Google Earth). As a result, total 947 outdoor leisure activities
were identified and finally incorporated in analytic process, and 519 trips for indoor leisure
activity were dropped (Table 4-1). Among those outdoor leisure activities, outdoor
exercise/sports reported 733 trips, and park activity observed was 214 trips. The number of
outdoor exercise/sports in public open space (including parks) is 100.

Table 4-1 Location types of leisure destinations (Outdoor/indoor)
Specific destinations Number %
Public open space park (state, local, and community), plaza, square, historic open space 314 21.419
Playgrounds track, baseball, basketball, football, tennis, pool, golf 354 24.147
Residential areas greenways/trails, sidewalk, porket yard, community green space 179 12.21
Other amenities beach, pier, lake, garden, mountain 100 6.8213
fitness (gym, karade, taekwondo, boxing)
clubs (dance, yoga, spa, ballet, billiard, bowling, ice skate)
YMCA, senior center, community center, church
museum, art gallery, library, shopping center
1466 100
192
327
Total
Location
Outdoor
Indoor
Sport facilities
Others
13.097
22.306

Based on this identification, dependent variables in regression model (Chapter 6)
were defined as three types of outdoor leisure activity. One dependent variable was
measured by aggregating outdoor leisure activity devoted to exercise/sport only (i.e.,

64
Type_1) in a survey day. The other two variables were measured by adding park activity and
leisure active travel to outdoor exercise/sport, respectively. The reason for the inclusion of
leisure active travel is that, in a broad sense, it can also be considered as a kind of outdoor
activity for leisure.

Measurement of variables (individual/household characteristics)
Control variables employed in this analysis consist of various individual/household
backgrounds including demo-socio-economic status (e.g., age, gender, race, education,
income, and employment), personality (e.g., attitudinal factors), time-constraint, physical
disability, and household composition (e.g., access to household vehicles and the presence of
children).
More specifically, ‘age’ was measured by a continuous scale which ranges from 18
to 64 years old, and ethnic groups were categorized as 5 groups: ‘Hispanic (reference)’,
‘White’, ‘Black’, ‘Asian/PI’, and ‘Others’. ‘Gender’ was defined by a dummy variable
(Male=1). Variables for economic status represent household income, education level, and
employment. In the NHTS data, household income was classified by 18 categories, and I
generated three income groups (i.e., high, low, and medium groups), based on the quartiles.
Education levels were also categorized into three groups: ‘high (bachelor degree or more)’,
‘low (less than high school)’, and ‘medium (others)’. ‘Employment’ was defined as a
dummy variable (Employed=1).
As other personal characteristics relevant to psychological factors, two variables
were employed in analytic models: attitude toward active travel and outdoor leisure,
respectively. The concept of attitude is broadly defined by “a (psychological) tendency that

65
is expressed by evaluating a particular entity with some degree of favour or disfavour”
(Eagley and Chaiken, 1993:1). Operationalizing a psychological concept with affective
responses might require certain types of measurement. However, it should be acknowledged
that the measurement is indirect and not completely verifiable since it is rarely observable in
a direct way (Bohte et al., 2009). Moreover, there is the limited information on attitudinal
factors in the secondary dataset. Given this data limitation, this study incorporates a proxy
variable that represents the personality on active travel and outdoor leisure.
More specifically, ‘attitude toward active travel’ was derived from survey question
regarding travel issues
8
. Respondents who considered ‘lack of walkways or sidewalks’ as
important travel issues were defined as individuals who are more oriented on active travel.
Even though this survey question does not fully capture personal psychological dimensions,
it can reflect some aspects of personality regarding different modes of travel. In other words,
people who raised the issue of ‘lack of walkways/sidewalks’ might have high
propensity/tendency of active travel than respondents who raised other travel issues
regarding highway, public transit, and travel price (e.g., gasoline/tolls fee).
Meanwhile, ‘attitude toward outdoor leisure’ was derived from survey question
regarding the reason for active travel
9
. Respondents who considered ‘exercise’ as the main
reason for active travel were defined as individuals with attitude toward outdoor leisure. This

8
More specifically, the question goes to “Of the following issues, please tell me which one is
the most important to you?: 1) highway congestion, 2) access to or availability of public transit, 3)
lack of walkways or sidewalks, 4) the price of travel including things like transit fees, tolls and the
cost of gasoline, 5) aggressive or distracted drivers, and 6) safety concerns, like worrying about being
in a traffic accident”.
9
Focused on respondents who mentioned that they walked outsides or engaged in biking in
the past week, survey question asked the main reason for active travel among the following: 1) To
walk or exercise the dog, 2) on the way to or from work, 3) on the way to or from public
transportation, 4) escorting children to or from school, 5) running errands or shopping, 6) for exercise,
and 7) for any other reasons.

66
definition is relatively straightforward, directly reflecting individual propensity for active
travel associated with outdoor leisure activity.
Given the limited daily time budget, the decision on a certain type of daily activities
can be affected by other activities/travels engaged in the same day. To identify the potential
relationship between outdoor leisure activity and other types of activity/travel, two variables
were incorporated in this analysis: daily time-constraint and daily active travel time. Based
on the record of travel diary, daily time-constraint was measured by aggregating the time
spent in work place and commuting. Daily time allocation for active travel was measured by
aggregating the amount of time spent in active travel for any purposes.
Two variables for household characteristics were employed: ‘car availability’ and
‘children’. ‘Car availability’ was measured by a continuous (i.e., the number of vehicles in
household) and binary scale (i.e., own or not), and ‘children’ was defined as a dummy
variable (having children=1). Other dummy variables are made by ‘physical disability
(disable=1)’, ‘weather condition (warm season=1)’, and ‘trip date (weekend days=1)’.

Measurement of variables (neighborhood characteristics)
The attributes of neighborhood potentially play as a facilitator or constraint for
active travel and outdoor leisure activity. Neighborhood opportunities include several
physical urban form features and local destinations (e.g., public transit, parks, and
sport/recreation facility). On the other hand, neighborhood barriers reflect (perceived) safety
concern and various types of risk-scape including crime, traffic incidents, and exposure to air
pollution.

67
Using GIS data at the parcel level, I objectively measured physical configurations of
urban form (e.g., density, diversity, design, and destinations). As specified below, ‘density’
(net activity density) was calculated by total number of population and employment divided
by developed areas; ‘diversity’ which represents the level of land use mixed was measured
by entropy index (N=4; residential, commercial, industrial, and public or semi-public use);
and ‘design’ that means the pattern of street connectivity was measured by walkability index
(i.e., the number of 3, 4 ways intersection).
As a local destination, public transportation is also recognized as a neighborhood
facility that encourages active travel by providing alternative travel options. The availability
of public transportation was measured by the number of bus stops and rail stations weighted
by the available bus/rail lines within a ¼ mile buffer area. Esri park GIS data (2010) were
used for measuring the number of park that intersects with a ¼ mile buffer area. Park data
cover both national- and local/neighborhood park areas. The availability of sport/recreational
facility was measured by counting the number of facility within a ¼ mile buffer. This
information was obtained from Info USA (2008)
10
, and the indoor facilities were excluded in
counting the number. In the final model, this variable was defined as a dummy variable since
there was little variation in the number of facilities observed in a ¼ mile buffer area.
(Value=1 if 1 or more facilities are located within a ¼ mile buffer area).

 Density
o Net activity density = Activity (number) /developed land (acre)
o Activity = Total job (number) + total population (number)

10
The specific types of facility include recreation centers (SCI Code: 799701), swimming
pools (799969), tennis courts (799971), and golf (799201) in public use.

68
o Developed land (acre) = Acreages of all parcels except vacant, recreational,
agricultural, and transportation and utilities

 Diversity
o Land use mixed (Entropy index) =
N
P P
i i
n
i
ln
ln
1


 (N= 4: land use types
include residential, commercial, industrial, and public or semi-public; ‘P’
represents the proportion of land in the ‘i’ use type).

 Design
o Street connectivity (Walkability index) =
i
n
i
I

1
(I= 0.5 for 3 way
intersections; 1 for 4 way intersections, manually deleting intersection
points along highways and interchanges)

 Destination: public transit
o Rail station: number of transit stations (metro, commuter rail and light rail)
within a ¼ mile buffer.
o Bus stop: number of unique bus stops (the same stop used by different
routes counted multiple times) within a ¼ mile buffer.

 Destination: park
o The number of park which intersects with a ¼ mile buffer. Esri park GIS
data (2010) were used for this measurement which covers both national- and
local/neighborhood parks.

 Destination: sport/recreation facility
o The number of sport/recreation facility with a ¼ mile buffer. In final model,
this variable is dummy (value=1 if 1 or more facilities are located within a ¼
mile buffer neighborhood).
Environmental barriers can be identified by two types of measurements: subjective
and objective. Barriers subjectively measured are the perceived safety concern. This variable

69
was derived from survey question regarding ‘safety concerns’ in the NHTS dataset and
measured by a 3-scale point: ‘A little issue’, ‘A moderate issue’, and ‘A big issue’, but this
study employed a dummy variable (1: if respondents raised safety issue) to make it easy to
interpret the results.
Complementing the perceived safety concern, three barriers were objectively
measured: crimes, traffic incidents, and exposure to air-pollution. The variable for crimes
was measured by aggregating the number of various types of violent crimes (e.g., arson,
assault, burglary, drugs/alcohol violation, robbery, sex crimes, vandalism, vehicle break-in,
and weapons) occurred within a ¼ mile buffer area from May to November, 2011. The
information on the location and type of crimes was obtained from Los Angeles County
sheriff.
11

As another type of risk-scape at the street level, traffic incidents were measured by
aggregating the number of collisions that relates to pedestrian and bike-users injuries in
order to evaluate pedestrian-level safety concern. The identification of traffic incidents was
obtained from SWITRS (2008-2009) that provides the information on the type and geo-
coded location of collision. In this process, traffic incidents on freeways were excluded since
those are irrelevant to active travel and outdoor leisure activity.
The exposure to air-pollution was indirectly measured by the proximity of freeways.
According to previous studies, pollutant particle concentration decreases exponentially with
downwind distance from freeways (Zhu et al., 2004; Sioutas et al., 2005). The empirical test

11
The reason of the discrepancy in time period between NHTS dataset (2008-2009) and
crime data (2011) is the limited access to historical crime data. The information over the past 6
months was available when I initially requested the data. Acknowledging that general distribution
patterns of crime incidents appear to be similar across different time periods, I employed crime data
(2011) in this analysis.

70
on freeways located in Los Angeles County revealed that a threshold where the
concentration of particle drops sharply is a 90m distance from freeways (Zhu et al., 2002).
Based on this finding, I created a 100m-buffer from freeways to identify the neighborhood
located in air-pollution hot spot. In the final model, a dummy variable was employed to
simply sort out such neighborhoods that intersect with a 100m-buffer from freeways.

 Crimes
o The number of violent crimes (e.g., arson, assault, burglary, drugs/alcohol
violation, robbery, sex crimes, vandalism, vehicle break-in, and weapons)
within a ¼ mile buffer from May to November, 2011.

 Traffic incidents
o The number of traffic incidents which involve pedestrian and bike-users
injury within a ¼ mile buffer. Traffic incidents on freeways were excluded
since those are irrelevant to active travel and outdoor leisure activity.

 The exposure of air-pollution
o A 100m-buffer from freeway was created to identify the neighborhood
located in air-pollution hot spot. In final model, a dummy variable was
employed to simply sort out such neighborhoods that intersect with a 100m-
buffer from freeways.
Along with environmental opportunities and barriers, neighborhood types (i.e.,
urban- vs. suburban neighborhood; high-leisure vs. low-leisure neighborhood) were
employed to explore how different types of residential location influence the
duration/frequency of active travel/outdoor leisure activity. On the basis of the levels
physical attributes supporting active travel and outdoor leisure activity, different
neighborhood types were categorized.

71
Regarding active travel, each household has different levels of activity density and
walkability index around home location. As illustrated in Table 4-2, total observations are
divided into four-groups by the combination of highest- and lowest quartile of activity
density and walkability index. Given this category, three dummy variables were generated:
urban, suburban, and mixed neighborhoods. Urban neighborhoods refer to the neighborhood
with characteristics of high activity density-high walkability; suburban neighborhoods
represent the neighborhood with characteristics of low activity density-low walkability; and
others were defined by mixed neighborhoods. As illustrated in Figure 4-3, urban
neighborhoods (high-high) are more concentrated in inner-city areas, while suburban
neighborhoods are scattered outside.

Table 4-2 Composites of neighborhood types for active travel module
High Low
High Urban Mixed
Low Mixed Suburban
Composites
(Net) Activity density
Walkability

To identify the levels of leisure-friendly neighborhood, I calculated the proportion of
park/recreational land use within a ¼ mile buffer area. Based on this value, I categorized
three groups of neighborhood: ‘high-leisure (high quartile)’, ‘low-leisure (low quartile)’, and
‘medium-leisure (others)’. As displayed in Figure4-4, high-leisure friendly neighborhoods
are scattered outside the inner-city and some of those neighborhoods are located in west
coast areas, whereas low-leisure friendly neighborhoods are mainly located in inner city
areas.

72
Figure 4-3 Neighborhood types (Urban vs. Suburban neighborhoods)

Note: Total number of buffers (urban neighborhoods) = 411; Total number of buffers (suburban neighborhoods) = 488

73
Figure 4-4 Neighborhood types (High-leisure vs. Low-leisure neighborhoods)

Note: Total number of buffers (low-leisure neighborhoods) = 886; Total number of buffers (high-leisure neighborhoods) = 879

74
Descriptive statistics of variables
Table 4-3 summarizes descriptive statistics of variables in both pooled- and sub-
samples.
12
Summary statistics from pooled samples indicated that, in general, respondents
are fairly middle aged (on average, 45 years old) but senior-oriented
13
, white-dominant (61%
of respondents), well educated (42% of respondents fall under ‘College graduate’), and
mostly employed (71%). Very limited numbers of respondents were found in physical
disability (6%), and most respondents were fairly accessible to private vehicles (96%).
About one third of the respondents have children (38%).
Regarding other individual characteristics, about one fifth of the respondents
reported the positive attitude toward active travel, but the percent of respondents who have
the positive attitude toward outdoor leisure activity was relatively high (45.5%). Furthermore,
on average, the amount of individual daily time-constraints measured by the time spent in
work place and commuting was 232.89 minutes. This implies that respondents generally
spend about 16.2% of daily total time budget (1,440 minutes) in both working and
commuting. Among exogenous variables, most trips were observed on weekdays (74%), but
there was no difference between seasons (i.e., cold/warm) in pooled samples.
Descriptive statistics also revealed that there are variations of neighborhood
environmental features. More specifically, activity (both population and employment)
density ranges from 0 to 591.6 (mean=34.3). On average, ‘entropy index’ for the diversity
(land use mixed) was 0.38 (‘value=1’ denotes the ideal/perfect balance), which implies

12
Descriptive statistics of dependent variables (i.e., the duration/frequency of active travel
and outdoor leisure activity) will be summarized in Chapter 5 and 6, respectively, on the basis of two
analytic approaches: aggregate (focused on ‘person’) and disaggregate (focused on ‘trip’ or ‘activity’).
13
More specifically, age groups ranging 35 to 64 years old take 78.3% of total adult samples.

75
Table 4-3 Summary of descriptive statistics of variables
Sample (Num)
Variable Mean Std Dev Min Max Mean Std Dev Min Max Mean Std Dev Min Max Mean Std Dev Min Max
AGE 44.933 12.539 18 64 40.777 12.726 18 64 51.374 9.99 19 64 44.062 12.369 18 64
GENDER (MALE=1) 0.48 0.5 0 1 0.473 0.5 0 1 0.455 0.498 0 1 0.494 0.5 0 1
WHITE 0.613 0.487 0 1 0.362 0.481 0 1 0.791 0.407 0 1 0.657 0.475 0 1
BLACK 0.065 0.246 0 1 0.085 0.279 0 1 0.062 0.24 0 1 0.056 0.231 0 1
ASIAN 0.103 0.304 0 1 0.063 0.244 0 1 0.086 0.281 0 1 0.13 0.336 0 1
HISPANIC (Ref.) 0.149 0.356 0 1 0.374 0.484 0 1 0.025 0.155 0 1 0.094 0.292 0 1
OTHERS 0.071 0.256 0 1 0.116 0.321 0 1 0.037 0.189 0 1 0.063 0.243 0 1
EDUC_LOW 0.095 0.293 0 1 0.298 0.458 0 1 0.006 0.078 0 1 0.035 0.183 0 1
EDUC_HIGH 0.429 0.495 0 1 0.12 0.325 0 1 0.622 0.485 0 1 0.494 0.5 0 1
HHINCOME_LOW 0.254 0.435 0 1 1 0 1 1 0 0 0 0 0 0 0 0
HHINCOME_HIGH 0.233 0.423 0 1 0 0 0 0 1 0 1 1 0 0 0 0
EMPLY (JOB=1) 0.713 0.452 0 1 0.572 0.495 0 1 0.804 0.397 0 1 0.742 0.438 0 1
DISAB (DISABLE=1) 0.062 0.241 0 1 0.107 0.31 0 1 0.042 0.2 0 1 0.048 0.214 0 1
DATE (WEEKEND=1) 0.267 0.442 0 1 0.247 0.432 0 1 0.232 0.423 0 1 0.292 0.455 0 1
WEATHER (WARM=1) 0.49 0.5 0 1 0.503 0.5 0 1 0.483 0.5 0 1 0.487 0.5 0 1
CAR_DUMMY 0.963 0.188 0 1 0.888 0.315 0 1 0.993 0.086 0 1 0.987 0.113 0 1
CAR_NUM 2.394 1.252 0 12 1.634 1.222 0 12 2.759 1.096 1 7 2.533 1.204 0 8
CHILD_DUMMY 0.382 0.486 0 1 0.444 0.497 0 1 0.398 0.49 0 1 0.333 0.471 0 1
DENSITY 34.305 27.934 0 591.6 46.718 31.518 0 216.8 32.135 24.987 0.129 218.5 29.147 25.329 0 591.6
DIVERSITY 0.384 0.251 0 0.985 0.464 0.243 0 0.985 0.377 0.253 0 0.953 0.348 0.246 0 0.955
DESIGN 17.085 10.154 0 92.6 19.085 9.914 0.3 78.9 17.177 10.444 0.3 92.6 16.054 9.992 0 86.3
RAIL_DUMMY 0.01 0.1 0 1 0.012 0.111 0 1 0.012 0.11 0 1 0.008 0.088 0 1
BUS_DUMMY 0.747 0.435 0 1 0.85 0.358 0 1 0.726 0.446 0 1 0.706 0.456 0 1
PARK_NUM 0.369 0.602 0 4 0.36 0.575 0 3 0.392 0.643 0 3 0.363 0.595 0 4
SPORT_FACILITY 0.104 0.305 0 1 0.081 0.274 0 1 0.133 0.34 0 1 0.102 0.303 0 1
SAFETY (CONCERN=1) 0.121 0.327 0 1 0.159 0.366 0 1 0.089 0.284 0 1 0.117 0.322 0 1
FREEWAY_DUMMY 0.138 0.345 0 1 0.202 0.402 0 1 0.128 0.334 0 1 0.111 0.314 0 1
CRIME_Q 12.803 18.26 0 182 18.435 21.666 0 150 11.765 18.894 0 178 10.488 15.307 0 182
TRAFFIC ACC_Q 2.706 4.252 0 41 4.4 5.656 0 41 2.439 3.632 0 32 1.988 3.386 0 40
NT_URBAN 0.118 0.323 0 1 0.223 0.416 0 1 0.108 0.311 0 1 0.07 0.256 0 1
NT_SUBUR 0.14 0.347 0 1 0.054 0.227 0 1 0.181 0.385 0 1 0.164 0.37 0 1
NT_LEISURE_HIGH 0.248 0.432 0 1 0.25 0.433 0 1 0.251 0.434 0 1 0.247 0.431 0 1
NT_LEISURE_LOW 0.254 0.439 0 1 0.259 0.44 0 1 0.255 0.44 0 1 0.252 0.434 0 1
ATTI_ACTIVE TRAVEL 0.22 0.414 0 1 0.258 0.438 0 1 0.187 0.39 0 1 0.216 0.412 0 1
ATTI_OUTDOOR LEISURE 0.455 0.498 0 1 0.391 0.488 0 1 0.513 0.5 0 1 0.46 0.499 0 1
Pool (3486) Low income (885) High income (813) Medium income (1788)
76

relatively homogenous distribution of land use. The average walkability index was 17,
ranging from 0 to 92.6. Regarding local destinations, few respondents are accessible to rail
stations within their neighborhoods (1%), whereas most respondents have bus stations within
their neighborhoods (74.7%). The average number of park is 0.36, and 10% of respondents
have a sport/recreation facility within their neighborhoods. In terms of potential barriers for
active travel and outdoor leisure activity, on average, 12% of respondents raised (perceived)
safety concern. The average number of violent crimes and traffic incidents within a ¼ mile
buffer area is 12.8 and 2.7, respectively. About 13.8% of respondents live in neighborhoods
closely located in freeways, which implies that the number of respondents more exposed to
air pollution is relatively small.
Results also confirmed that, in general, low-income groups live in neighborhoods
with high density, diverse land use, and well-connected street patterns, compared with high-
income groups
14
. Among local destinations, descriptive statistics indicated different patterns
of public transit between rail stations and bus stops. For instance, while there was no or little
difference in rail stations (t-value=0.02) between low- and high-income groups, low-income
groups reported a slightly high level of access to bus stops (t-value=6.34) in neighborhood
(85%) compared with high-income groups (73%). Notably, the evident gap in the number of
park was not found in different income groups (t-value= -1.08), but the level of access to
sport/recreation facilities was higher in high-income groups than low-income groups (t-
value= -3.45).

14
Results from t-test revealed that statistically significant differences between low- and high
income respondents were observed for urban form (3D) variables: activity density (t-value=10.5),
mixed land use (t-value=7.29), and street connectivity (t-value=3.86).
77

A substantial gap in potential barriers for active travel/outdoor leisure activity was
found among different income groups. Regarding subjective measure, 15% of low-income
groups perceived safety concern, whereas 8% of high-income groups revealed the concern
(t-value= 4.42). Consistently, low-income groups were more exposed to three risk factors
objectively measured, compared with high-income groups. Difference in each risk factor
between low- and high-income groups is statistically significant (crime: t-value= 8.55; traffic
incidents: t-value= 8.54; proximity of freeways (100 meter): t-value= 4.12).
Table 4-4 and 4-5 display the variables employed in the final regression models for
active travel and outdoor leisure activity, respectively. Two issues should be notified here.
As often acknowledged, some of built environmental features are highly correlated with one
another, which may lead to collinearity problem in the regression model. To identify those
variables, simple correlations between explanatory variables at the neighborhood level was
conducted among pooled samples (for more information, refer to Appendix). As a result,
several variables revealed strong correlations (|r| >0.6)
15
. Based on this, three variables (i.e.,
‘diversity’, ‘access to public transit’, and ‘traffic incidents’) were excluded in the final
regression model for active travel to avoid collinearity problem
16
.
Next, I dropped urban form features (i.e., 3D: density, diversity, and design) from
the final model for outdoor leisure activity since those factors appear theoretically less

15
Fairly consistent with our expectation, activity density was highly correlated with diversity
(i.e., land use mixed: r = 0.64), bus stops (r = 0.63), and traffic incidents (r = 0.69) at the 5%
significance level in pooled samples. Furthermore, the significant positive correlation was also
reported between bus stops and traffic incidents (r = 0.61) and between traffic incidents and crimes (r
= 0.61), whereas the significant negative correlation was observed between low-leisure neighborhoods
and park availability (i.e., No. of park: r = -0.61).
16
Along with pairwise correlation, I also tested VIFs to detect collinearity problem. As a
result, explanatory variables with a VIF value of less than 10) were incorporated in the final
regression.
78

relevant to outdoor leisure activity. To make some justification on the lack of theoretical
relevance, I conducted the simple regression model (Table 1 in Appendix). The result
showed that urban form features (i.e., 3D: density, diversity, and design) were not
significantly associated with outdoor leisure activity. Empirical evidence also supports that
neighborhood design factors are not related to leisure-time physical activity under the
control for individual/household characteristics (Saelens et al., 2003). In addition, park
availability (i.e., the number of parks) as a local destination was also dropped from the final
model due to collinearity with neighborhood type. Regarding deterrent factors most relevant
for the topic at hand, the variable of violent crimes was dropped when the variable of traffic
incidents was included in regression models since it is highly correlated with traffic incidents.
79

Table 4-4 Definition of variables for active travel model
Dimension Variable Description Data source
Dependent variables
Daily total active travel Duration Time spent in daily total active travel NHTS (2008,9)
Frequency Number of daily total active travel NHTS (2008,9)
Daily total walking (only ) Duration Time spent in daily total walking travel (continuous/binary) NHTS (2008,9)
Frequency Number of daily total walking travel (continuous/binary) NHTS (2008,9)
Independent variables
Demography Age Age (adult: 18 to 64) NHTS (2008,9)
Gender Male=1; Female (Ref.) NHTS (2008,9)
Race NHTS (2008,9)
- Race1 Hispanic (Ref.)
- Race2 White
- Race3 Black
- Race5 Asian&PI
- Race5 Others
Socio-Economic S. Education NHTS (2008,9)
- Educ1 Less than High School
- Educ2 HS & some college (Ref.)
- Educ3 BA+
Income Household income NHTS (2008,9)
- High High quartile
- Low Low quartile
- Medium Others (Ref.) NHTS (2008,9)
Employment Employed (Yes=1); Unemployed (Ref.) NHTS (2008,9)
Other individual factors Attitude Travel issues: lack of sidewalk (Yes=1) NHTS (2008,9)
Household variables Vehicle access Number of cars in household NHTS (2008,9)
Neighborhood attributes
Neighborhood types Urban form Density/Design composition (2*2 matrix) SCAG GIS (2008)
- Urban Density (high quartile)/Design (high quartile)
- Suburban Density (low quartile)/Design (low quartile)
- Mixed Others (Ref.)
Built environment Density Activity density=(population+employment)/developed land SCAG parcel data
Resident population SCAG TAZ (2008)
Total employment Info US (2008)
Design Street connectivity (walkability index: 3,4 ways intersections) SCAG GIS (2008)
Destination Number of urban park ESRI GIS data (2010)
Number of public transit SCAG GIS (2008)
Riskscape Safety_Perceived Safety concerned (Yes=1) NHTS (2008,9)
Violent crimes Number of arson, assault, burglary, robbery, vandalism etc LAC Sheriff (2011)
Air pollution Proximity of freeway (100M buffer) SCAG GIS (2008)
(1/4 miles buffer from household location)

80

Table 4-5 Definition of variables for outdoor leisure activity model
Dimension Variable Description Data source
Dependent variables
Outdoor leisure activity Duration NHTS (2008,9)
Frequency NHTS (2008,9)
Independent variables
Demography Age Age (adult: 18 to 64) NHTS (2008,9)
Gender Male=1; Female (Ref.) NHTS (2008,9)
Race NHTS (2008,9)
- Race1 Hispanic (Ref.)
- Race2 White
- Race3 Black
- Race5 Asian&PI
- Race5 Others
Socio-Economic S. Education NHTS (2008,9)
- Educ1 Less than High School
- Educ2 HS & some college (Ref.)
- Educ3 BA+
Income Household income NHTS (2008,9)
- High High quartile
- Low Low quartile
- Medium Others (Ref.) NHTS (2008,9)
Employment Employed (Yes=1); Unemployed (Ref.) NHTS (2008,9)
Physical disability Disable (Yes=1); Ability (Ref.) NHTS (2008,9)
Other individual factors Attitude Reason for active travel: Exercise (Yes=1) NHTS (2008,9)
Time constraints Number of minutes for working activity and commuting NHTS (2008,9)
Active travel Duration of daily total active travel NHTS (2008,9)
Household variables Vehicle access Number of cars in household NHTS (2008,9)
Children Having chilred (Yes=1) NHTS (2008,9)
Weather condition Season Warm (April to September=1) NHTS (2008,9)
Trip date Weekend/days Weekend(Yes=1); Weekdays(Ref.) NHTS (2008,9)
Neighborhood attributes
Neighborhood types Leisure-friendly Proportion of park/recreational land cover (parcel level) SCAG parcel data
- High leisure High quartile
- Low leisure Low quartile
- Mixed Others (Ref.)
Built environment Destination Availability of outdoor sport/recreation facility (Yes=1) Info US (2008)
Riskscape Safety_Perceived Safety concerned (Yes=1) NHTS (2008,9)
Traffic incident Number of collision (pedestrian and bike users injury) SWITRS (2008-2009)
Air pollution Proximity of freeway (100M buffer) SCAG GIS (2008)
Outdoor exercise/sport, park activity, and leisure active travel
(1/4 miles buffer from household location)

4.3. METHODOLOGY

As a special case of SEM approach, path analysis was applied to properly test the complex
hypothetical relationships illustrated in conceptual framework (Figure 3-1). Unlike OLS
81

model specifying a default model (i.e., the impact of independent variables on dependent
variable), a recursive path analysis establishes hypothesized relationships a priori and tests
structural links, typically employing multiple endogenous/exogenous variables (MacDonald,
2012). In this regard, path analysis can be considered as a confirmative method rather than
an exploratory method (Golob, 2003).

The relationships between endogenous and exogenous variables are defined by the
matrices, and coefficient matrices of endogenous/exogenous variables determine structure of
path analysis. A column vector of endogenous/exogenous variables is denoted by Y (L*1
matrix) and X (K*1 matrix), respectively. A matrix of coefficients of endogenous variable is
denoted by is B (L*L), whereas a matrix of coefficients of exogenous variables is denoted by
(K*K). ‘ ’ (L*1matrix) denotes a column vector of error (Hancock and Mueller, 2006).
Covariance-based analysis is applied to estimate parameter. The model-implied
covariance matrixes of the observed endogenous/exogenous variables are reproduced in
specific functions of unknown parameters. When the value for unknown parameter is
inserted, the model-implied covariance matrixes are obtained, and then the difference
between the reproduced covariance matrix and the sample covariance matrix is measured.
Such difference is minimized by estimation process, and finally analytic model assumes that
the model-implied covariance matrix is equal to the sample covariance matrix (Raykov and
Marcoulides, 2000).
82

The assumption of the multivariate normal distribution is generally used in
estimating procedure for continuous endogenous variables. However, when endogenous
variables are measured by binary, ordered categorical (ordinal), and unordered categorical
(nominal) scale, statistic inferences will not be properly estimated due to the violation of the
assumption of normal distribution (Muthén and Muthén, 2007). In the case, an alternative
estimator (i.e., weighted least square parameter estimator) is required to yield robust
standard errors.
It should also be noted that both collinearity problem among exogenous variables
and insufficient sample sizes can lead to a lack of stability or less accuracy in estimations
(Petraitis et al., 1996; Lleras, 2005). Acknowledging these issues, this study excluded several
variables highly correlated with other variables to correct collinearity issue, based on the
results of pair correlation and VIFs (refer to Appendix). Furthermore, for a more stable and
accurate estimation, fairly sufficient sample sizes (pooled samples=3,486) were employed in
this study, and the software package M-plus that offers the procedures for weighted least
square parameter estimator was utilized so as to model endogenous variables not measured
on a continuous scale.
Given this background of analytic method, I suggest a recursive path model based
on the conceptual framework, while simultaneously specifying a function of residential
location choice and daily behavioral choice. More specifically, three endogenous variables
include behavioral outcome (i.e., active travel) and two residential location choices (NT:
urban- and suburban neighborhood type). Other exogenous variables consist of
individual/household characteristics (X), attitudinal factor (A), and a set of neighborhood
environment (B) including neighborhood type (NT). Categorical variables for neighborhood
83

type were considered not only as endogenous variables in residential location choice model
but also as exogenous variables in predicting daily active travel.

Y = Active travel (measured by the duration and frequency)
X = Attributes of individual/household
B = Attributes of neighborhood environment
NT = Categorical variables of neighborhood type (i.e., urban, suburban, and mixed)
EF = Attributes of environmental facilitators
EC = Attributes of environmental constraints (barriers)
A = Dummy variable of attitudinal factor (toward active travel)

To test several hypotheses mentioned in Chapter 3, three analytic models for active
travel time were developed, employing different types of dependent variables: daily total
active travel time (adding up walking and biking), daily total walking time, and daily
walking time recommended for health, respectively. Different measurements for dependent
variables need different analytic approaches. In predicting daily active travel time, simple
OLS regression might provide biased results since the unobserved active travel time have a
positive non-zero value. Tobit regression was adopted to address this nature of travel
outcome, while considering daily active travel time as a censored variable. On the other hand,
logistic regression model was adopted to test the probability of total walking time
84

recommended for health since dependent variable is a discrete and measured by a binary
scale (i.e., the value is 1 if daily total walking time is 30 minutes or more).
Along with active travel time, three analytic models for active travel frequency were
also developed. This is because a function of frequency might differ from the duration.
Unlike active travel time, the frequency can be considered as an event that follows a discrete
probability distribution. Likewise, in predicting active travel frequency, simple OLS method
might also lead to biased and inconsistent results. Poisson regression was adopted to address
this nature of a count variable. On the other hand, logistic regression model was also
developed to test the probability of daily walking frequency (i.e., the value is 1 if daily
walking trip was reported).
Analytic models for the prediction of outdoor leisure also include three endogenous
variables: behavioral outcomes (i.e., outdoor leisure) and two residential location choices
(NT: high- and low leisure-friendly neighborhood type). Other exogenous variables are
individual/household characteristics (X), attitudinal factor (A), and a set of neighborhood
environment (B) including neighborhood type (NT). Categorical variables for neighborhood
type were considered not only as endogenous variables in residential location choice model
but also as exogenous variables in predicting outdoor leisure.

Y = Outdoor leisure activity (measured by the duration and frequency)
X = Attributes of individual/household
85

B = Attributes of the neighborhood environment
NT = Categorical variables of residential location
(e.g., levels of outdoor leisure environment: high, low, and medium level)
EF = Attributes of environmental facilitators
EC = Attributes of environmental constraints (barriers)
A = Dummy variable of individual attitude toward outdoor leisure

Three analytic models for outdoor leisure activity were developed, employing both a
narrow- and broad concept of outdoor leisure activity. Like in analytic models for active
travel, the duration of outdoor leisure activity was predicted by tobit regression model to
address the unobserved time value since daily outdoor leisure activity time as a censored
variable with a positive non-zero value. On the other hand, for the prediction of the
frequency of outdoor leisure activity, Poisson regression model was adopted to address this
nature of a count variable.
86

CHAPTER 5.
NEIGHBORHOOD ENVIRONMENT AND ACTIVE TRAVEL

5.1. DESCRIPTIVE ANALYSIS

Before presenting the results from path analysis based on the analytic framework mentioned
in Chapter 4, the basic descriptive analysis of daily active travel helps us not only gain
general ideas about travel behavior but also identify the difference in travel patterns among
income groups.
Person-based analysis offers the basic statistic summary of daily average frequency
and duration (3,486 adults), including individuals who did not engage in active travel given a
survey day. Both person and daily trip files were merged in this process to calculate the
prevalence of active travel per capita. Table 5-1 indicates the summaries of active travel
sorted by trip purpose, focusing on the frequency (i.e., how many trips people take) and
duration (i.e., how many minutes people spend). On average, people took 0.43 trips per day
and 4.43 trips per week, spending 6.81 minutes per day and 155.7 minutes per week in active
travel (both walk and bike). Daily trips per capita are slightly lower than those
17
of samples
from national travel survey in 2009 (Pucher et al., 2011), informing that people living in
study area engage less in active travel than national samples.

17
In the NHTS (2009), the number of trips (active travel) per capita per day is 0.549 and the
time spent in active travel per capita per day is 6.955 minutes.
87

Table 5-1 Mean of frequency/duration of active travel by purpose (pooled samples)
Frequency Time Frequency Time Frequency Time
Work-trip 0.043 0.548 0.006 0.103 0.049 0.651
Non-work trip 0.215 2.605 0.014 0.245 0.229 2.85
Trip to leisure 0.137 2.88 0.014 0.425 0.151 3.305
Daily (total) 0.395 6.033 0.033 0.773 0.428 6.806
Weekly (total) 4.106 139.82 0.328 15.862 4.434 155.682
Pool
Walk Bike Active travel

Note: Daily total active travel was calculated by combining work, non-work, and leisure trips. Those
specific trip purposes are based on travel diary (objective measurement). Instead of travel diary,
weekly total active travel is based on the simple survey question
18
(subjective measurement). Those
questions do not include the information on specific trip purposes.

Regarding trip purpose, travel data showed that non-work trips take a large amount
of frequency and time, whereas work-trips have very small portion of total active travel as
expected. The amount of leisure active travel in Los Angeles County samples is greater and
longer than that in national samples. As expected, the frequency of active travel for
recreation is smaller than that of active travel for non-work, but the duration (i.e., travel
time) of active travel for recreation is longer. Potential reason of this result is that non-work
trips have a relatively short distance compared with leisure trips. In other words, people are
generally willing to spend more time in travel for recreational purpose than for utilitarian
purpose due to some reasons (e.g., jogging for health as a circular trip). In this vein, it can be
suggested that active travel is generally considered as derived demand but also particularly
recognized as a unique activity with intrinsic values.

18
For weekly active travel (walking or biking) frequency, questions were asked to
respondent: “In the past week, how many times did take a walk outside including walking the dog and
walks for exercise?” and “In the past week, how many times did ride a bicycle outside including
bicycling for exercise?” For weekly active travel (walking or biking) duration, respondent answered
to the questions of “In the past week, how much total time did spend walking?” and “In the past week,
how much total time did spend biking?”
88

Table 5-2 Mean of frequency of active travel by mode and purpose (income groups)
High_income Low-income High_income Low-income High_income Low-income
Work-trip
0.031 0.064 0.004 0.011 0.035 0.074
Non-work trip
0.193 0.328 0.006 0.037 0.199 0.364
Trip to leisure
0.152 0.118 0.012 0.02 0.164 0.138
Daily (total)
0.376 0.51 0.021 0.067 0.398 0.577
Weekly (total)
4.214 4.561 0.285 0.465 4.499 5.026
Active travel
Trip frequency
Walk Bike

Table 5-3 Mean of duration of active travel by mode and purpose (income groups)
High_income Low-income High_income Low-income High_income Low-income
Work-trip
0.593 0.726 0.078 0.172 0.671 0.897
Non-work trip
2.056 4.424 0.076 0.666 2.132 5.09
Trip to leisure
3.217 2.211 0.384 0.664 3.601 2.875
Daily (total)
5.866 7.36 0.538 1.502 6.404 8.862
Weekly (total)
136.505 152.346 14.431 20.183 150.936 172.529
Active travel
Travel time
Walk Bike

As shown in Table 5-2 and 3, different income groups reported different levels of
active travel. According to trip purposes, low-income groups engage more frequently in both
work- and non-work purpose than high-income groups
19
. The findings are consistent with
the results from previous research that informs the negative association between income
level and utility walk trips as well as the positive association between income level and
recreational walk trips (Agrawal and Schimek, 2007). Regarding the duration of active
travel, low-income groups generally spend more time in active travel than high-income
groups, but this pattern differs across trip purposes: high-income groups spend more time in
active travel for recreational purpose, whereas low-income groups spend more time in active
travel for utilitarian purpose
20
.

19
The difference in trip frequency between low- and high income groups is statistically
significant (t-value= 8.51 in non-work trip; t-value= 8.42 in total active travel).
20
The difference in trip duration between low- and high income groups is also statistically
significant (t-value= 10.35 in non-work trip; t-value= 9.64 in total active travel).
89

Trip-based analysis (total 3,417 trips
21
) provides a snap shot of total active travel
(3,417 trips) in terms of travel time and distance. More specifically, the frequency and mean
distribution of each active travel will be analyzed by specific trip purposes (e.g., a work,
non-work, and leisure trip). To better understand the variation in daily active travel between
subpopulations, individual trip data will be separated by different income groups (i.e., high-
income vs. low-income).
I analyzed the characteristics of active travel in terms of time and distance. To
capture the distribution patterns of active travel time, I divided travel time into 3 types: less
than 5 minutes, 6-29 minutes, and 30 minutes or more. The reason of this category is that 5
minutes is often considered as the time people spend in walking around a ¼ mile catchment
area, and that 30 minutes is the recommended daily active travel time for health. In addition,
to identify the distribution patterns of active travel distance, I also divided into 4 categories
measured by block size: less than 0.25 miles, 0.25 to 1 mile, 1 to 3 miles, and 3 miles or
more.
As shown in Table 5-4, the prevalence of active travel decreases as travel time
increases. About 25% of all daily active travel was recorded as ‘less than 5minutes’. This
pattern was also found in walk trips but not in bike trips. Not surprisingly, compared to the
duration of walk trips, the percentage of a short travel time is low but the share of a long
travel time is high in bike trips. Nearly one-third of bike trips reported 30 minutes or more.
This result reflects that a walk trip not only covers a relatively short distance but also is more
sensitive to travel time than a bike trip.

21
The number of walk trips is3,149 trips; the number of bike trips is 268 trips
90

Table 5-5 indicates the distribution of active travel distance. As observed in travel
time, there is the negative association between the frequency of active travel and travel
distance. Over 80% of all daily active travel show a relatively short distance (i.e., ‘less than
1mile’), whereas active travel with a long distance have a small portion. Walk trips
presented the same distribution patterns to active travel, but bike trips did not.
Table 5-4 Distribution of travel time by active travel modes
Travel time N % N % N %
Less than 5mins 812 25.8 40 14.9 852 24.9
6 to 29 mins 1865 59.2 153 57.1 2018 59.1
30 mins or more 472 15 75 28 547 16
Total 3149 100 268 100 3417 100
Walk Bike Active travel

Table 5-5 Distribution of travel distance by active travel modes
Travel distance N % N % N %
Less than .25 1320 41.9 51 19 1371 40.1
.25 to 1 miles 1416 45 87 32.5 1503 44
1 to 3 miles 310 9.8 58 21.6 368 10.8
3 miles or more 103 3.3 72 26.9 175 5.1
Total 3149 100 268 100 3417 100
Walk Bike Active travel

To identify whether different segments of active travel have different distribution
patterns in time and distance, I separated total samples into 2 segments. As conducted in
previous analysis, I also sorted out active travel into 4 categories: home bound, work-trips,
non-work trips, and leisure trips. Given those types of active travel, the basis statistics (e.g.,
mean and standard deviation) of travel time and distance among different income groups are
compared. Before the comparison of distribution patterns between separated samples,
general description of active travel in pooled samples (3,417 trips) is provided in Table 5-6.
91

It is shown that, in general, the frequency, duration, and distance differ across trip
purposes. In terms of frequency, among specific trip purposes of active travel, both home
bounce and non-work trips account for a large share of active travel frequency, presenting
more than one-third of total numbers of active travel (1,308 home bounce trips; 1,160 non-
work trips). The remnant portion is taken by leisure trip (about 25%) and work-trip (about
5%). This distribution pattern of frequency is fairly consistent with both walk and bike trips.

Table 5-6 Frequency and mean distribution of active travel (pooled samples)
N % Mean Std Dev Mean Std Dev
Walk_home 1193 37.89 16.655 24.424 0.7336 0.9671
Walk_work 169 5.37 12.76 10.966 0.5271 0.5987
Walk_non work 1103 35.03 12.698 10.546 0.5607 0.7455
Walk_leisure 684 21.72 21.675 18.435 0.9249 0.9956
Total (Walk) 3149 100 15.947 16.093 0.6866 0.8267
Bike_home 115 42.91 23.983 20.969 3.4435 5.4258
Bike_work 21 7.84 17.333 11.038 2.3651 2.2394
Bike_non work 57 21.27 17.158 16.254 1.9347 4.0869
Bike_leisure 75 27.99 26.213 22.499 3.9081 5.3911
Total (Bike) 268 100 21.172 17.69 2.9129 4.2858
Active travel_home 1308 38.28 17.412 24.069 1.0768 2.4829
Active travel_work 190 5.56 14.089 11.91 1.3675 3.4234
Active travel_non work 1160 33.95 13.071 11.082 0.7253 2.4951
Active travel_leisure 759 22.21 22.661 21.812 2.194 14.444
Total (Active travel) 3417 100 16.808 17.218 1.3409 5.7113
Pool sample
Active travel mode Frequency Duration Distance

The descriptive statistics of active travel time revealed that leisure trips have a
relatively long travel time, whereas utilitarian trips take a relatively short travel time. The
amount of travel time for recreation almost doubles in that for utilitarian purpose. Likewise,
the distribution pattern of active travel distance also found that leisure trips have a relatively
long distance, whereas utilitarian trips took a relatively short distance. This result is fairly
consistent with general finding from travel behavior studies, confirming that utilitarian
92

active travel is more sensitive to travel cost (i.e., time and distance) than recreational active
travel.

Table 5-7 Frequency and mean distribution of active travel (income groups)
N % Mean S.D. Mean S.D. N % Mean S.D. Mean S.D.
Walk_home 499 41.9 16.65 13.15 0.69 1.12 274 34.68 19.03 28.85 0.84 1.01
Walk_work 92 7.72 21.43 16.71 0.85 0.91 36 4.56 11.64 10.1 0.56 0.59
Walk_non work 441 37.03 14.14 11.24 0.66 1.02 279 35.32 11.64 10.15 0.48 0.43
Walk_leisure 159 13.35 18.69 16.55 0.74 1.1 201 25.44 24.01 17.04 1.09 0.99
Total (Walk) 1191 100 17.73 14.41 0.74 1.03 790 100 16.58 16.53 0.74 0.76
Bike_home 53 42.74 23.38 18.58 2.64 4.42 27 41.54 29.41 25.37 5.43 6.92
Bike_work 8 6.45 15.63 12.66 1.89 2.64 8 12.31 21.13 10.7 3 1.93
Bike_non work 30 24.19 17.27 11.97 1.7 2.36 13 20 19.77 26.44 3.47 7.7
Bike_leisure 33 26.61 26.88 22.25 2.99 5.14 17 26.15 24.12 20.23 4.94 4.67
Total (Bike) 124 100 20.79 16.36 2.3 3.64 65 100 23.6 20.68 4.21 5.3
Active travel_home 552 41.98 17.49 14.08 0.98 2.5 301 35.2 19.87 26.86 1.31 2.64
Active travel_work 100 7.6 20.72 16.14 1.3 1.86 44 5.15 13.31 11.55 0.99 1.36
Active travel_non work 471 35.82 14.41 11.31 0.75 1.24 292 34.15 12.35 12 0.89 4.64
Active travel_leisure 192 14.6 19.96 17.74 1.13 2.47 218 25.5 23.98 17.43 2.64 18.29
Total (Active travel) 1315 100 18.14 14.82 1.04 2.02 855 100 17.38 16.96 1.46 6.73
Active travel mode
Low income
Frequency Duration Distance Frequency Duration Distance
High income

In the subsequent analysis, the gap between different income groups was displayed
in terms of frequency, duration, and distance of active travel sorted by specific trip purposes.
As indicated in first two columns in Table 5-7, active travel was more frequent in low
income groups than in high income groups. As mentioned before, leisure trip is more
prevalent in high-income groups, whereas non-work trip is more frequent in low-income
groups. However, unlike the distribution of frequency, there was no gap in the mean of
duration between those groups. This result is consistent with both total walk and total bike
trips.

93

5.2. REGRESSION RESULTS

5.2.1. Active travel time

Unstratified model (pooled samples)
Table 5-8 summarizes the results from analytic models for active travel time.
Censored variables were regressed in Model 1 (daily total active travel time) and Model 2
(daily total walking travel time), whereas binary measurement of walking time (daily total
walking time: 30 minutes or more) was regressed in Model 3. As often observed in other
travel behavior studies, the explanatory power (r-square) of each pooled model was small,
ranging from 0.025 to 0.037, but several indexes for general model fit of path analysis fall
under the acceptable levels.
Like the results from previous descriptive analysis, the consistent results were found
in both Model 1 and Model 2. There are several variables at the individual/household level
that present statistical significance in predicting active travel time as well as walking travel
time. For example, high levels of education positively affect active travel time, while both
employment and the number of vehicles in household were negatively associated at the 5%
significant level. Individual personality is often considered as an important individual
characteristic in making the decision on active travel. Supporting the significant role of
personality on active travel, regression models indicated the positive influence on both active
travel time and walking travel time at the 10% significant level. Those findings are fairly
consistent with the evidence from the previous travel behavior research. Based on results,
we confirmed that people with high levels of education or personality on active travel are

94

Table 5-8 Results from path analysis on active travel time (pooled samples)
Coef. β t p Coef. β t p Coef. O/R z p
Direct effect
Intercepts 10.168 0.611 5.075 *** 8.717 0.593 5.109 *** 0.069 1.071436 1.293
AGE -0.024 -0.018 -0.923 -0.016 -0.014 -0.723 0 1 0.655
GNDR (MALE=1) -0.041 -0.001 -0.068 -0.475 -0.016 -0.864 -0.019 0.981179 -1.783 *
WHITE 0.614 0.018 0.603 0.511 0.017 0.567 0.034 1.034585 1.937 *
BLACK -2.238 -0.033 -1.465 -2.107 -0.035 -1.626 -0.021 0.979219 -0.722
ASIAN -1.594 -0.029 -1.217 -1.56 -0.032 -1.277 0.032 1.032518 1.488
OTHERS -0.939 -0.014 -0.585 0.232 0.004 0.178 -0.003 0.997004 -0.115
EDC_LOW 1.72 0.03 1.503 1.152 0.023 1.116 0.017 1.017145 0.932
EDC_HIGH 1.735 0.052 2.419 ** 1.482 0.05 2.358 ** 0.013 1.013085 1.101
HHINC_LOW 0.342 0.009 0.318 0.301 0.009 0.332 0.047 1.048122 2.297 **
HHINC_HIGH 0.57 0.014 0.744 0.731 0.021 1.098 0 1 0.018
EMPLY -2.428 -0.066 -3.471 *** -2.289 -0.07 -3.823 *** -0.018 0.982161 -1.563
NUM_CAR -1.446 -0.109 -4.018 *** -1.239 -0.105 -4.049 *** -0.015 0.985112 -2.159 **
ACT_DENSITY 0.041 0.069 4.841 *** 0.036 0.069 4.764 *** 0 1 -0.119
STREET_DESIGN 0.042 0.026 1.438 0.051 0.036 2.052 ** 0.001 1.001001 2.16 **
NUM_PARK 0.958 0.035 2.392 ** 0.82 0.034 2.376 ** 0.001 1.001001 0.066
SAFTY_CONCERN -1.29 -0.025 -1.055 -1.28 -0.028 -1.187 -0.007 0.993024 -0.328
FREEWAY_100M -0.453 -0.009 -0.496 -0.069 -0.002 -0.088 -0.005 0.995012 -0.332
CRIME_QMILE -0.073 -0.075 -3.217 *** -0.069 -0.081 -2.982 *** -0.001 0.999 -1.592
NT_URBAN -0.063 -0.004 -0.136 -0.094 -0.007 -0.216 0.015 1.015113 1.302
NT_SUBUR -0.296 -0.019 -0.231 -0.05 -0.004 -0.049 -0.025 0.97531 -0.997
ATTITUDE 1.638 0.041 1.663 * 1.518 0.043 1.761 * 0.018 1.018163 0.975
HHINC_LOW 0.43 0.18 3.507 *** 0.43 0.18 3.504 *** 0.43 3.505 ***
HHINC_HIGH 0.267 0.109 2.116 ** 0.267 0.109 2.117 ** 0.269 2.128 **
NUM_CAR -0.129 -0.156 -3.173 *** -0.13 -0.157 -3.178 *** -0.13 -3.174 ***
ATTITUDE 0.014 0.006 0.095 0.012 0.005 0.081 0.013 0.087
HHINC_LOW -0.434 -0.179 -1.751 * -0.423 -0.175 -1.707 * -0.423 -1.706 *
HHINC_HIGH 0.086 0.035 0.383 0.092 0.037 0.407 0.091 0.405
NUM_CAR 0.152 0.18 2.123 ** 0.151 0.179 2.112 ** 0.151 2.119 **
ATTITUDE -0.27 -0.106 -1.103 -0.272 -0.107 -1.108 -0.274 -1.116
Indirect effect
HHINC_LOW -0.156 -0.004 -0.261 -0.062 -0.002 -0.129 -0.004 -0.308
HHINC_HIGH 0.009 0 0.048 -0.02 -0.001 -0.136 0.006 0.889
NUM_CAR 0.053 0.004 0.259 0.02 0.002 0.119 0.002 0.408
ATTITUDE -0.081 -0.002 -0.228 -0.015 0 -0.053 -0.007 -0.706
SAMPLE SIZE 3486 3486 3486
R-SQUARE 0.035 0.037 0.025
Chi-sq/df 7.13761 7.1397 7.604625
CFI (=1.0) 0.872 0.891 0.873
TLI (=1.0) 0.743 0.792 0.815
RMSEA (<.08) 0.085 0.085 0.087
Model 2 Model 1
Daily active travel time
NT_URBAN
NT_SUBUR
Daily active travel time
Daily walking time
NT_URBAN
NT_SUBUR
Daily walking time
Model 3
Daily walking time (30 mins)
NT_URBAN
NT_SUBUR
Daily walking time (30 mins)
Variables

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

likely to spend more time in active travel, whereas people with a job or available cars are
less devoted to active travel.
95

Age, race, and income level are frequently assumed as a significant predictor of
active travel in previous studies. By and large, non-White or low-income groups are more
likely to engage in walking than White or high-income groups (Pucher et al., 2012).
However, this study found no significant association between those variables and active
travel time. The reason for different results goes to different definitions of active travel. For
instance, other studies often classified and separated utilitarian trips from recreational trips in
analytic frameworks, whereas this analysis employed active travel time regardless of trip
purpose. In addition, the results indicated that active travel time was positively related to
White or high-income groups even though the relationship was found statistically
insignificant. This observation is opposed to findings from other studies, but it reflects, as
demonstrated in the basic descriptive analysis on daily active travel, that both White and
high-income groups are willing to spend more time in active travel for recreational purpose
rather than for utilitarian purpose. Another potential reason for inconsistent results is that
those samples were overrepresented in this data and hence, potentially, they might report
more active travel time than others.
On the other hand, Model 3 that incorporates a binary measurement of walking time
found different results from Model 1 and 2. Among control variables at the
individual/household level, gender, race, and low-income were significant associated with
walking time (30 minutes or more), suggesting the likelihood that female, White, or low-
income populations engage in the recommended levels of walking is higher than other
groups. However, both employment and education that indicated the statistical significance
in Model 1 and 2 were no longer significantly related to walking time. Notably, given the
96

binary measure of walking time, the direct effect of car availability still remains statistically
significant in predicting walking time.
Among the attributes of built environment, both activity density (measured by
population and employment) and park availability (measured by the number of parks nearby
home) revealed a significantly positive association with active travel time as well as walking
time. Street design measured by road connectivity showed a significant and positive impact
on walking time only. Those are generally consistent with extensive literature that highlights
the substantial role of the built environment as the opportunity set for active travel behavior.
Interestingly, however, the results conflict with the one study with similar measurements
(Forsyth et al., 2007). They found no significant association between the total amounts of
walking in one-day travel diary and population/residential density within a ¼ mile buffer
area, suggesting zero-sum game between “high density-more travel walking” and “low
density-more leisure walking”.
Another aspect of neighborhood environment can be reflected by the factor which
may deter individual decision on active travel. As expected, all deterrent factors consistently
showed the negative influence on both active travel and walking time, but the statistical
significance differs across deterrent factors. For example, violent crimes objectively
measured indicated the significant association, confirming that people living in
neighborhoods highly exposed to violent crimes are less likely to spend their time in active
travel including walking.
However, unlike results from other studies that highlight the evident role of
(subjective) neighborhood safety concern on active travel, this model did not indicate any
significant impact. Although the objectively measured crime factor was a significant
97

predictor, it is still unclear what relationship exists between subjectively measured
neighborhood safety, objectively measured neighborhood safety, and active travel behavior.
Theoretically, perceived neighborhood safety can mediate the role of objectively measured
neighborhood safety on active travel outcomes, but the model employed here is not intended
to provide the explicit information on such complex mechanism.
Regarding the relative importance among variables, standardized coefficients
informed that car availability has the most profound effect on active travel time (Beta: -
0.109). This finding supports that the decision on active travel largely depends on household
travel options. Meanwhile, it is worth noting that some of neighborhood characteristics, such
as activity density (Beta: 0.069) and violent crimes (Beta: -0.075), also indicated the relative
importance, compared to other control variables, such as employment (Beta: -0.066), high
levels of education (Beta: 0.052), and attitude toward active travel (Beta: 0.041).
Surprisingly, both environmental opportunities and barriers did not report any
significant association with walking time measured by a binary scale (i.e., 30 minutes or
more). More specifically, the variable of violent crimes was no longer considered as a
significant factor in a binary model. Remarkably, street design was positively associated
with walking time at the recommended level, suggesting that people who live in
neighborhoods with well-connected street patterns are more likely to spend 30 minutes or
more in daily walking trips. However, the size in effect of street design (Beta: 0.04) was
relatively marginal, compared to the significant socio-economic variables (e.g., low levels of
income and car availability).
Neighborhood types (i.e., urban - and suburban neighborhood) were employed in the
path model to test whether residential location (i.e., where people live) affects active travel
98

time. However, there was no significant association between neighborhood types and active
travel time. Furthermore, it is often mentioned that the choice of residential location can
mask the relationship between the built environment and travel behavior, particularly active
travel. To address this issue, neighborhood type as a proxy for residential location is
additionally incorporated in regression models, in tandem with other urban form features. In
a subsequent model which considers neighborhood type as an endogenous variable, two
neighborhood types were regressed on the level of household income, the number of
vehicles in household, and personality on active travel, respectively.
As theoretically assumed, urban neighborhood type was positively associated with
low-income groups, but it was negatively associated with the number of vehicles in
household at the significant level. Surprisingly, high-income groups also revealed positive
and significant association with urban neighborhood type. This finding suggests that high-
income groups are likely to live in urban neighborhoods, compared with medium-income
groups. Not surprisingly, the results from the separated regression model for suburban
neighborhood type revealed opposite directions to urban neighborhood type. For example,
there was a significant negative relationship between low-income groups and suburban
neighborhoods, and positive relationship between the number of vehicles in household and
suburban neighborhoods. In line with theoretical prediction, personality on active travel was
positively related to urban neighborhoods and negatively related suburban neighborhoods.
However, no significant influence of individual attitudinal factor on each neighborhood type
was found.
More importantly, this study explores a chain of links between individual/household
characteristics, neighborhood type, and active travel time so as to identify an indirect effect
99

of individual/household characteristics on active travel time via neighborhood type.
However, the results from path analysis did not reveal any significant effect of residential
self-selection on active travel time. In other words, even though household income level and
car availability were found to have a significant effect on neighborhood type, the data
employed in this analysis did not support the mediating role of neighborhood type.

Stratified model (income levels)
Analytic models separated by three income levels were employed in order to offer
adequate explanation on whether subpopulations have different functions and predictors of
daily active travel time, walking time, and recommended walking time (binary), respective.
Before the interpretation on results from stratified models, regression models separated by
different income groups explain better the variance of daily active travel time than the model
with pooled samples, presenting the increase in the explanatory power (r-square) by 0.12.
This is mainly because that some of individual/household characteristics (e.g., employment
and car availability) in separated models yield a more pronounced impact on active travel
time, compared with the model with pooled samples. In addition, the separated models also
produce the improved model fit indexes at the acceptable level.
As shown in Table 5-9, 10, and 11, the results from an array of models indicated
highly divergent patterns across income groups, displaying key factors that significantly
affect daily active travel time. Regarding individual/household characteristics, low-income
groups found a consistently significant impact of gender on various types of daily active
travel time, demonstrating that a female spends more time in active travel than a male.
Notably, the evident gender gap in active travel time was solely observed in low-income
100

groups. Consistent with the evidence from previous studies, low-income groups also
reported that both employment and car availability were negatively associated with daily
active travel time at the significant level.
High-income groups did not report any significant influence of gender, employment
and car availability, whereas both high education and personality on active travel yielded
significant and positive relationship to walking time at the recommended level. These
findings imply that high levels of education and attitudinal factor can increase walking time
only for high-income groups.
The different results between high- and low-income groups were also reported in
various neighborhood attributes. Low-income groups revealed that both activity density and
violent crimes are crucial factors that affect positively and negatively daily active travel time,
respectively, whereas high-income groups did not show any significant impact of such
variables. This inconclusive finding is indeed not surprising because high-income groups are
more likely to choose a motorized travel mode to reach the destination within a short
distance. Another reason behind this difference is that high-income groups are less exposed
to violent crimes than low-income groups, and hence criminal activities in neighborhoods
might not be a critical issue that precludes high-income populations from engaging in active
travel.
Furthermore, it is also notable that there was no significant association between
street design and daily active travel time in low-income groups, whereas high-income groups
revealed a consistently significant impact of street design on active travel time as well as on
walking time with the positive direction. Even though it is not entirely clear why street
design has a profoundly significant influence on active travel time in high-income groups but
101

not in low-income groups, this finding confirms that urban form features differentially affect
the duration of active travel depending on income levels.
Unlike the results from a pooled model that revealed a consistently significant and
positive impact of the availability of park (i.e., the number of parks) on active travel time,
the relationship was not significant in stratified models. This finding suggests that the role of
neighborhood parks varies across different income groups, demonstrating that it is not
necessarily translated into the increase in duration of active travel when the samples were
sorted by income levels. Inconclusive evidence may reflect different active travel patterns
between high- and low-income groups.
More specifically, as demonstrated in descriptive analysis, low-income groups spend
more time in active travel for non-leisure purposes, whereas high-income groups take a large
amount of active travel time for leisure purposes. In this sense, the role of neighborhood
parks seems to be less evident for low-income groups since the available park is not directly
linked to non-leisure trips. High-income groups might also tend to be less responsive to
neighborhood parks since high levels of mobility can allow them to reach park areas with
more attractive leisure amenities beyond the neighborhood boundary.
Regarding neighborhood type, the results indicated that low-income groups living in
urban neighborhoods spend more time in active travel time, whereas low-income groups
living in suburban neighborhoods revealed the opposite results. Those findings are fairly
consistent with conventional wisdom that active travel is more prevalent in urban
neighborhoods, but such associations were not statistically significant. The direct impact of
neighborhood type was also insignificant for high-income groups, and those samples did not
support the positive relationship between urban neighborhoods and active travel time that
102

was found among low-income groups. In addition, the stratified models did not show any
significant indirect impact of individual/household characteristics on active travel time via
neighborhood type. In other words, as indicated in a pooled model, there was no evidence of
residential self-selection among sub-populations.

103

Table 5-9 Results from path analysis on active travel time by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 14.906 0.861 4.449 *** 15.055 0.915 2.403 ** 6.171 0.397 2.303 **
AGE -0.056 -0.041 -1.045 -0.091 -0.055 -1.249 0.032 0.025 0.92
GNDR (MALE=1) -3.057 -0.088 -2.222 ** 1.465 0.044 1.19 0.929 0.03 1.152
WHITE 1.306 0.036 0.83 -1.299 -0.032 -0.287 0.375 0.011 0.24
BLACK -3.253 -0.052 -1.142 -2.597 -0.038 -0.494 -2.46 -0.037 -0.987
ASIAN 0.775 0.011 0.293 -4.635 -0.079 -0.906 -1.524 -0.033 -0.848
OTHERS -1.109 -0.021 -0.483 5.083 0.058 0.92 -2.33 -0.036 -0.798
EDC_LOW 1.04 0.027 0.712 -1.596 -0.008 -0.023 0.738 0.009 0.286
EDC_HIGH -1.537 -0.029 -0.669 2.292 0.068 1.643 1.779 0.057 1.954 *
HHINC_LOW
HHINC_HIGH
EMPLY -4.597 -0.131 -3.098 *** -1.792 -0.043 -1.066 -1.182 -0.033 -1.236
NUM_CAR -2.893 -0.242 -1.648 * -0.95 -0.053 -0.442 -0.636 -0.049 -1.255
ACT_DENSITY 0.066 0.121 3.47 *** 0.003 0.004 0.082 0.005 0.009 0.223
STREET_DESIGN 0.032 0.018 0.489 0.148 0.094 2.617 *** 0.014 0.009 0.311
NUM_PARK 0.636 0.021 0.644 1.374 0.054 1.425 1.024 0.039 1.905 *
SAFTY_CONCERN 1.195 0.025 0.428 -0.659 -0.011 -0.245 -2.054 -0.043 -1.073
FREEWAY_100M -1.469 -0.034 -0.773 -1.52 -0.031 -0.617 1.382 0.028 1.241
CRIME_QMILE -0.086 -0.108 -2.193 ** -0.101 -0.098 -1.554 -0.003 -0.003 -0.08
NT_URBAN 0.901 0.053 1.001 -1.81 -0.111 -1.139 0.946 0.062 0.984
NT_SUBUR -1.975 -0.16 -1.134 -2.641 -0.178 -1.036 1.063 0.069 0.647
ATTITUDE 0.407 0.01 0.067 2.458 0.058 0.713 1.272 0.034 0.952
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -0.168 -1.791 * -0.116 -0.106 -0.583 -0.16 -0.188 -2.44 **
ATTITUDE 0.101 0.044 0.363 0.29 0.112 0.674 -0.003 -0.001 -0.011
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.632 -0.652 -0.933 0.512 0.427 0.925 0.103 0.121 1.177
ATTITUDE 0.585 0.183 0.213 -0.03 -0.011 -0.03 0.275 0.112 0.89
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 1.141 0.095 0.67 -1.142 -0.064 -0.577 -0.261 -0.02 -1.002
ATTITUDE -1.064 -0.027 -0.193 -0.444 -0.011 -0.158 0.29 0.008 0.482
SAMPLE SIZE 885 813 1788
R-SQUARE 0.102 0.096 0.028
Chi-sq/df 3.0871 2.01092 9.4156
CFI (=1.0) 0.912 0.953 0.869
TLI (=1.0) 0.873 0.909 0.764
RMSEA (<.08) 0.041 0.035 0.072
Variables
Medium income Low Income High income
Daily active travel time
Neighborhood Type_URBAN
Neighborhood Type_SUBUR
Daily active travel time

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
104

Table 5-10 Results from path analysis on walking time by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 12.666 0.841 4.536 *** 14.728 0.954 2.603 *** 5.267 0.399 2.374 **
AGE -0.027 -0.023 -0.591 -0.079 -0.051 -1.159 0.031 0.029 1.09
GNDR (MALE=1) -3.363 -0.112 -2.618 *** 1.272 0.041 1.08 0.38 0.014 0.562
WHITE 0.824 0.026 0.611 -2.091 -0.055 -0.515 0.725 0.026 0.549
BLACK -2.747 -0.051 -1.162 -2.429 -0.038 -0.508 -1.629 -0.028 -0.81
ASIAN 0.945 0.015 0.467 -4.757 -0.087 -1.04 -1.94 -0.049 -1.142
OTHERS -0.114 -0.002 -0.064 4.905 0.06 0.98 -1.256 -0.023 -0.555
EDC_LOW 0.391 0.012 0.296 -1.16 -0.006 -0.021 0.389 0.005 0.184
EDC_HIGH -1.006 -0.022 -0.562 2.304 0.072 1.692 * 1.304 0.049 1.758 *
HHINC_LOW
HHINC_HIGH
EMPLY -3.618 -0.119 -2.939 *** -2.203 -0.057 -1.411 -1.384 -0.046 -1.802 *
NUM_CAR -2.468 -0.238 -1.594 -0.742 -0.044 -0.337 -0.573 -0.052 -1.324
ACT_DENSITY 0.054 0.113 2.631 *** 0.013 0.02 0.388 -0.011 -0.02 -0.529
STREET_DESIGN 0.038 0.025 0.661 0.14 0.095 2.823 *** 0.026 0.02 0.648
NUM_PARK 0.167 0.006 0.206 1.446 0.06 1.658 * 0.895 0.04 1.944 *
SAFTY_CONCERN 0.512 0.012 0.218 -1.535 -0.028 -0.623 -1.214 -0.03 -0.77
FREEWAY_100M -0.208 -0.006 -0.142 -1.398 -0.03 -0.605 1.288 0.031 1.358
CRIME_QMILE -0.093 -0.133 -2.198 ** -0.096 -0.1 -1.613 0.011 0.012 0.334
NT_URBAN 0.988 0.067 1.329 -1.633 -0.107 -1.194 0.465 0.036 0.536
NT_SUBUR -1.675 -0.155 -1.03 -3.11 -0.223 -1.351 1.106 0.085 0.851
ATTITUDE 0.503 0.015 0.098 2.855 0.072 0.774 0.989 0.031 0.843
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -0.168 -1.791 * -0.114 -0.104 -0.571 -0.16 -0.188 -2.446 **
ATTITUDE 0.101 0.044 0.363 0.291 0.112 0.678 -0.003 -0.001 -0.013
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.625 -0.651 -0.923 0.518 0.431 0.936 0.103 0.121 1.176
ATTITUDE -0.54 -0.17 -0.197 0.024 0.008 0.024 -0.276 -0.112 -0.891
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.931 0.09 0.618 -1.426 -0.085 -0.688 -0.188 -0.017 -0.85
ATTITUDE -0.804 -0.023 -0.171 -0.402 -0.01 -0.125 0.303 0.009 0.603
SAMPLE SIZE 885 813 1788
R-SQUARE 0.101 0.12 0.029
Chi-sq/df 3.185 2.01262 9.33386
CFI (=1.0) 0.893 0.959 0.879
TLI (=1.0) 0.865 0.922 0.783
RMSEA (<.08) 0.041 0.035 0.072
Variables
Daily walking time
Low Income High income Medium income
Daily walking time
Neighborhood Type_SUBUR
Neighborhood Type_URBAN

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

105

Table 5-11 Results from path analysis on walking time (binary) by income groups
Coef. O/R z p Coef. O/R z p Coef. O/R z p
Direct effect
Intercepts 0.212 1.236148 2.863 *** 0.075 1.077884 0.537 0.025 1.02532 0.323
AGE -0.001 0.999 -1.198 0.001 1.001001 0.583 0.001 1.001 1.435
GNDR (MALE=1) -0.081 0.922194 -2.877 *** -0.005 0.995012 -0.234 0.004 1.00401 0.313
WHITE 0.028 1.028396 0.957 0.032 1.032518 0.315 0.026 1.02634 0.949
BLACK -0.043 0.957911 -0.679 -0.053 0.94838 -0.396 0 1 -0.004
ASIAN 0.159 1.172338 3.811 *** 0.054 1.055485 0.498 -0.012 0.98807 -0.343
OTHERS -0.016 0.984127 -0.382 0.093 1.097462 0.794 -0.037 0.96368 -0.678
EDC_LOW 0.026 1.026341 0.93 -0.096 0.908464 -0.116 -0.007 0.99302 -0.179
EDC_HIGH 0 1 -0.007 0.004 1.004008 0.159 0.016 1.01613 1.179
HHINC_LOW
HHINC_HIGH
EMPLY -0.013 0.987084 -0.525 -0.044 0.956954 -1.546 -0.011 0.98906 -0.679
NUM_CAR -0.089 0.914846 -1.037 -0.022 0.97824 -1.094 -0.003 0.997 -0.466
ACT_DENSITY 0 1 1.021 0 1 -0.52 0 1 -0.775
STREET_DESIGN 0.001 1.001001 1.045 0.002 1.002002 1.678 * 0.001 1.001 1.087
NUM_PARK -0.003 0.997004 -0.16 -0.015 0.985112 -0.784 0.01 1.01005 0.951
SAFTY_CONCERN 0.065 1.067159 1.271 -0.063 0.938943 -1.269 -0.026 0.97434 -0.867
FREEWAY_100M 0.021 1.021222 0.765 -0.051 0.950279 -1.104 -0.01 0.99005 -0.436
CRIME_QMILE -0.002 0.998002 -2.413 ** 0.001 1.001001 0.952 0 1 0.095
NT_URBAN 0.018 1.018163 0.833 -0.001 0.999 -0.025 0.022 1.02224 1.313
NT_SUBUR -0.106 0.899425 -1.511 0.011 1.011061 0.764 -0.015 0.98511 -0.606
ATTITUDE 0.026 1.026341 0.086 0.067 1.069295 1.788 * 0.015 1.01511 0.646
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -1.79 * -0.115 -0.578 -0.16 -2.443 **
ATTITUDE 0.101 0.364 0.291 0.677 -0.003 -0.015
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.632 -0.933 0.508 0.918 0.103 1.172
ATTITUDE -0.521 -0.19 0.021 0.021 -0.276 -0.891
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.065 0.762 0.006 0.583 -0.002 -0.468
ATTITUDE -0.054 -0.182 0 -0.031 -0.004 -0.434
SAMPLE SIZE 885 813 1788
R-SQUARE 0.166 0.025 0.023
Chi-sq/df 3.895 2.72 9.94424
CFI (=1.0) 0.858 0.916 0.825
TLI (=1.0) 0.821 0.863 0.747
RMSEA (<.08) 0.05 0.046 0.072
Variables
Low Income High income Medium income
Daily walking time (30 mins or not)
Neighborhood Type_URBAN
Neighborhood Type_SUBUR
Daily walking time (30 mins or not)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

106

5.2.2. Active travel frequency

Unstratified model (pooled samples)
Table 5-12 summarizes results from analytic models for active travel frequency.
Count variables were regressed in Model 1 (daily active travel frequency) and Model 2
(daily walking frequency), whereas a binary variable was regressed in Model 3 (walking or
not). Compared to previous models computed for active travel time, the explanatory power
(r-square) of analytic models with different measurements slightly improves but still remains
small, ranging from 0.048 to 0.052. In other words, this analytic model better explains the
variation in the frequency of active travel than in the time spent in daily active travel.
Several indexes for general model fit of path analysis also fall under the acceptable levels.
By and large, regression models for active travel frequency revealed the result fairly
consistent with active travel time. For example, among individual/household characteristics,
high levels of education, employment, and car availability revealed a consistently significant
association with various types of active travel frequency at the 5% significant level. In
addition, personality on walking was also found a significant predictor contributing to the
increase in the frequency of active travel. In line with the evidence from extensive literature,
those findings confirm that people with high levels of education and attitude toward walking
engage more frequently in active travel, whereas people with a job and available cars
participate less frequently in active travel.

107

Table 5-12 Results from path analysis on active travel frequency
Coef. β t p Coef. β t p Coef. O/R z p
Direct effect
Intercepts 0.426 0.749 6.038 *** 0.563 0.7 5.969 *** 0.435 1.544963 6.233 ***
AGE -0.001 -0.03 -1.624 -0.002 -0.032 -1.808 * -0.001 0.999 -1.516
GNDR (MALE=1) -0.015 -0.013 -0.778 -0.041 -0.025 -1.429 -0.014 0.986098 -0.703
WHITE 0.039 0.033 1.174 0.052 0.032 1.096 0.037 1.037693 1.119
BLACK -0.106 -0.046 -2.048 ** -0.096 -0.029 -1.339 -0.092 0.912105 -1.786 *
ASIAN -0.044 -0.024 -1.006 -0.091 -0.034 -1.34 -0.053 0.94838 -1.184
OTHERS 0.055 0.025 1.23 0.055 0.018 0.883 0.049 1.05022 1.084
EDC_LOW 0.089 0.046 2.36 ** 0.137 0.05 2.587 ** 0.089 1.093081 2.364 **
EDC_HIGH 0.061 0.053 2.65 *** 0.103 0.063 3.154 *** 0.058 1.059715 2.529 **
HHINC_LOW 0.008 0.006 0.241 0.05 0.027 0.994 -0.019 0.981179 -0.55
HHINC_HIGH 0.03 0.023 1.156 0.008 0.004 0.222 0.027 1.027368 1.022
EMPLY -0.092 -0.074 -3.934 *** -0.099 -0.056 -3.087 *** -0.102 0.90303 -4.317 ***
NUM_CAR -0.054 -0.12 -4.361 *** -0.081 -0.126 -4.652 *** -0.056 0.945539 -4.482 ***
ACT_DENSITY 0.001 0.074 4.359 *** 0.002 0.077 5.474 *** 0.002 1.002002 4.586 ***
STREET_DESIGN 0.003 0.054 2.93 *** 0.004 0.052 3.333 *** 0.003 1.003005 3 ***
NUM_PARK 0.017 0.018 1.088 0.018 0.014 0.844 0.019 1.019182 1.201
SAFTY_CONCERN -0.048 -0.028 -1.21 -0.06 -0.024 -1.043 -0.055 0.946485 -1.386
FREEWAY_100M -0.012 -0.007 -0.417 -0.032 -0.014 -0.842 -0.019 0.981179 -0.666
CRIME_QMILE -0.002 -0.046 -2.293 ** -0.002 -0.049 -2.523 ** -0.002 0.998002 -2.331 **
NT_URBAN 0.007 0.014 0.388 0.011 0.014 0.432 0.009 1.009041 0.462
NT_SUBUR -0.007 -0.013 -0.181 -0.012 -0.015 -0.202 -0.008 0.992032 -0.208
ATTITUDE 0.083 0.06 2.45 ** 0.079 0.04 1.69 * 0.082 1.085456 2.431 **
HHINC_LOW 0.43 0.18 3.505 *** 0.43 0.181 3.508 *** 0.43 3.505 ***
HHINC_HIGH 0.269 0.11 2.128 ** 0.27 0.11 2.141 ** 0.269 2.128 **
NUM_CAR -0.13 -0.156 -3.175 *** -0.13 -0.157 -3.179 *** -0.13 -3.174 ***
ATTITUDE 0.013 0.005 0.086 0.013 0.005 0.087 0.013 0.087
HHINC_LOW -0.427 -0.177 -1.723 * -0.424 -0.175 -1.708 * -0.424 -1.709 *
HHINC_HIGH 0.091 0.036 0.404 0.094 0.038 0.419 0.091 0.406
NUM_CAR 0.151 0.18 2.118 ** 0.151 0.18 2.114 ** 0.151 2.118 **
ATTITUDE -0.273 -0.107 -1.112 -0.272 -0.107 -1.111 -0.273 -1.114
Indirect effect
HHINC_LOW 0 0 0.01 0 0 -0.01 0 0.01
HHINC_HIGH 0.003 0.002 0.405 0.004 0.002 0.44 0.003 0.469
NUM_CAR 0 0 0.016 0 0.001 0.037 0 0.021
ATTITUDE -0.002 -0.001 -0.169 -0.003 -0.002 -0.189 -0.002 -0.193
SAMPLE SIZE 3486 3486 3486
R-SQUARE 0.05 0.048 0.052
Chi-sq/df 7.60513 9.6048 7.604625
CFI (=1.0) 0.816 0.808 0.873
TLI (=1.0) 0.732 0.648 0.815
RMSEA (<.08) 0.08 0.079 0.087
NT_URBAN
NT_SUBUR
Daily active travel frequency
Model 1
Daily active travel frequency
NT_URBAN
NT_SUBUR
Daily walking frequency
Model 2
Daily walking frequency Daily walking or not
NT_URBAN
NT_SUBUR
Daily walking or not
Model 3
Variables

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

108

Interestingly, unlike the result of active travel time, there was a significant
association between low education and active travel frequency. It suggests that people less
educated also engage more frequently in active travel. When considering that education is
closely related to income, this finding appears reasonable since they have limited options for
daily travel. Gender was significantly associated with active travel time, but such association
was not significant in regression model for active travel frequency. This different result
implies that gender is not a factor explaining the variation in the frequency of active travel in
this sample.
It is generally observed in previous studies that demo-socio-economic variables,
such as age, race, and income, play a profound role in individual decision on active travel
decision. However, analytic models employed here indicate that those factors did not affect
significantly active travel frequency except for Black. The negative association between
Black and active travel frequency was found to be significant, which presents that the Black
ethnic group is less likely to engage in active travel than Hispanic. As mentioned before, the
outcomes of active travel measured by different mode types and trip purposes provide some
clues behind inconclusive evidence.
The influence of some of neighborhood characteristics on active travel frequency
was fairly similar to active travel duration, but others showed somehow different results.
More specifically, consistent with the results of active travel time, both activity density and
street design produced a significant positive association with active travel frequency. Those
findings support the link between the built environment and active travel, demonstrating that
people living in areas with high density and well-connected street patterns not only spend
more time in active travel but also engage more in active travel. However, unlike results of
109

active travel time, no significant impact of park availability (e.g., the number of parks nearby
home) was found in active travel frequency. Based on this result, park availability may
contribute to increasing the time spent in daily active travel, but it does not necessarily lead
to the increase in the frequency of daily active travel.
As reported in active travel time, there are also a fairly consistent and negative
association between various types of deterrent factors in neighborhoods and active travel
frequency. The negative association was evident in three models of active travel frequency,
and these findings suggest that people modify individual active travel choice and reduce the
demand for active travel (especially, discretionary one) given the exposure to a certain type
of deterrent factors. However, the statistical significance was solely found in violent crimes
in neighborhoods, confirming that people living in a neighborhood highly exposed to violent
crimes are less likely to engage in daily active travel.
It is still unclear why some factors play a significant deterrent role in individual
active travel but others do not. This inconclusive evidence might reflect the complex
psychological process through which individual decision on active travel is made or deterred.
Furthermore, it can also be acknowledged that people might respond more or less sensitively
to deterrent factors depending on the characteristics of active travel demand (e.g., mandatory
or optional) and individual propensity (or preferences) for active travel.
Standardized coefficients allow us to compare the relative importance among
variables by adjusting the standard deviations, and hence to identify the variables with a
pronounced impact on active travel frequency. Consistent with the finding from active travel
time, car availability revealed the most profound effect on active travel frequency (Beta: -
0.12 to -0.126) across all three models. This finding generally supports that the available
110

options for travel (i.e., car availability) is a crucial factor which affects the observed daily
active travel decision. Other individual/household variables, such as employment (Beta: -
0.056 to -0.081), education (Beta: 0.046 to 0.063), and personality on active travel (Beta:
0.04 to 0.06) also indicated the relatively high magnitude, compared to neighborhood
environmental factors which did not yield any significant association.
Meanwhile, several neighborhood factors significantly affecting active travel
frequency consistently reported a pronounced impact. Those variables include activity
density (Beta: 0.074 to 0.077), street design (Beta: 0.052 to 0.055), and violent crimes (Beta:
-0.046 to -0.049). Confirming the theory of travel behavior, those findings verify that the
likelihood people engage in active travel can be mainly determined by individual/household
backgrounds as well as various aspects of neighborhood opportunities and barriers. However,
not surprisingly, the magnitude of impact of neighborhood environment on active travel
frequency is smaller than that of car availability in household.
Regarding the role of residential location, no direct impact of neighborhood type on
active travel frequency was found, and this result was consistent with the finding from the
analytic model for active travel time. A subsequent regression model that defines
neighborhood type as a function of household income levels, the number of vehicles in
household, and personality on active travel also revealed that both household income and the
number of vehicles available in household were observed as a significant predictor, but the
individual personality was not considered as a significant predictor explaining whether
people choose urban or suburban neighborhoods.
Even though it is well established that neighborhood type can be explained by
individual/household characteristics, there was no indirect effect of household income and
111

car availability on active travel frequency via neighborhood type. In other words, this model
did not observe any significant chain of links between individual/household characteristics,
neighborhood type, and active travel frequency, while providing little evidence for the role
of residential self-selection in individual decision on active travel.

Stratified model (income levels)
Three analytic models separated by different levels of household income were
developed to see whether a function of daily active travel frequency varies across different
income groups, and how subpopulations are responsive to various neighborhood
environmental factors. Before specific interpretation on results, the stratified models explain
better the variance of daily active travel frequency than a pooled model, presenting the
increase in the explanatory power by 0.14. The results from path analysis also indicated
several indexes for general model fit at the statistically acceptable level.
The more robust explanatory power might be attributable to key variables with a
pronounced impact on active travel frequency in stratified models. Not surprisingly, like in
regression models for active travel time, those significant predictors of active travel
frequency vary across different income groups. More specifically, as summarized in Table 5-
13, 14, and 15, there are highly divergent patterns between low- and high income groups in
explaining the variation in the frequency of daily active travel.
Among various individual/household backgrounds, low-income groups reported that
gender, employment, and car availability have a consistently significant impact on active
travel frequency measured by a continuous and binary scale, both. These results prove that
low-income respondents who are female, unemployed, or have a limited number of private
112

vehicles engage more frequently in active travel than the counterparts. However, surprisingly,
those significant associations were not found in high-income groups. The result informs us
that, for high-income groups, gender and car availability might not be significant predictors
of active travel frequency. In particular, gender did not show any significant influence on
active travel frequency in a pooled model. This finding suggests that the gender gap in active
travel behavior is more manifest in low-income groups, reflecting that low-income women
often encounter the limited transportation opportunities.
Meanwhile, high-income groups found a significant negative association between
age and active travel frequency, but such association was not significant in low-income
groups at the 5% significance level. As often mentioned in active travel literature, senior
groups are less likely to engage in active travel due to physical disability. The result here
also supports the previous finding but suggests that the significant impact of age might be
the case of high-income groups rather than of low-income groups. The reasoning behind this
difference is that low-income seniors rarely have travel options to replace active travel,
compared with high-income seniors, and hence they may engage more frequently in active
travel.
Furthermore, low-income groups reported the significant and negative association
between education and daily walking frequency, whereas high-income groups found the
significant and positive association. It is often reported in previous findings that there is a
negative association between education and active travel, presenting the likelihood people
less educated engage frequently in walking. However, this result demonstrates that high-
income populations more educated also engage more frequently in walking trips. Potentially,
this finding might reflect the prevalent discretionary active travel in high-income populations
113

and non-discretionary active travel in low-income populations. This opposite direction in the
relation of education to walking frequency verifies different functions of walking trips
among different income groups, illuminating the interaction effect of education and income
on active travel behavior.
Among the individual characteristics, attitudinal factor indicated different results
between low- and high-income groups. More specifically, high-income populations found a
significant positive impact of attitudinal factor on total active travel frequency and walking
measured by a binary scale at the 10% significant level, but such association was not
significant in low-income populations. Unlike previous studies which yielded significant role
of attitudinal factor in individual behavioral decision on active travel, this result suggests
that, for low income groups, attitude toward active travel may be not necessarily translated
into the increase in frequency of active travel. Potentially, it appears that, regardless of
attitude, they are more likely to rely on active travel due to some reasons, such as the limited
travel options and the burden of travel costs for vehicle use.
Regarding the role of neighborhood environment, the result showed that some
variables had a consistently significant influence on active travel frequency in both low- and
high-income groups, but others found highly divergent patterns among different income
groups. For instance, activity density has a significantly positive impact on active travel
frequency and walking frequency measured by a binary scale. This finding was consistently
observed in both income groups. However, notably, high-income groups did not find any
significant association between activity density and daily walking frequency measured by a
continuous scale. This finding seems to be reasonable when considering the propensity of
high-income groups for motorized travel to reach the destination with a short distance. Based
114

on this, it can be suggested that low-income groups are more responsive to activity density in
neighborhoods than high-income groups.
On the other hand, there was a significant positive association between street design
and active travel frequency in high-income groups. This association was consistently
observed in all measurements of active travel frequency, so that we can expect that affluent
people living in neighborhoods with well-connected street patterns are more likely to engage
in active travel. However, low-income groups did not report any significant influence of
street design on active travel frequency. It is not entirely clear why low-income groups lost
the significance in the impact of street design, but this finding suggests that high-income
groups appear to be more responsive to street design pattern in neighborhoods than low-
income groups.
In addition, as found in a pooled model, the relationship between parks nearby home
and the frequency of active travel was not significant in both low- and high-income groups.
Unlike the analytic models for active travel time that produced a consistently significant and
positive impact of neighborhood parks, this model suggests that the accessibility to parks
might not be translated into the increase in frequency of active travel.
Regarding the impact of neighborhood barriers, both low- and high-income groups
reported a consistently significant and negative influence of a violent crime on active travel
frequency as well as walking frequency measured by a continuous scale. As mentioned in
regression model for active travel time, neighborhood violent crimes emerged as a profound
factor that may decrease the frequency of active travel. However, such deterrent role was
not found significant in a binary regression model that predicts the probability of daily
walking frequency of low-income groups. This is not surprisingly because active travel can
115

be often considered as a mandatory daily activity among low-income populations who are
more exposed to neighborhood violent crimes. On the other hand, the negative relationship
between neighborhood violent crimes and daily walking frequency (binary) still remains
significant in high-income groups. Therefore, it can be concluded that the exposure to
neighborhood violent crimes is a crucial factor to determine whether high-income groups
walk or not, but it is not the case of low-income groups.
Consistent with the result from a pooled model, the stratified models by income
levels revealed that neighborhood safety subjectively measured and neighborhood barriers
objectively measured were negatively associated with active travel frequency in both income
groups, but those relationships were not statistically significant. Unlike previous studies
highlighting the significant role of neighborhood safety in active travel behavior, this finding
informs that the impact of perceived neighborhood safety may not be pronounced when
other variables (e.g., neighborhood barriers objectively measured) are incorporated in the
analytic model.
Regarding neighborhood type, the results revealed that low-income groups living in
urban neighborhoods engage more frequently in active travel, whereas low-income groups
living in suburban neighborhoods indicated the opposite results. Those findings are fairly
supporting theoretical assumption that presumes the positive influence of urban
neighborhoods as well as the negative influence of suburban neighborhoods on active travel,
but such associations were not statistically significant.
The lack of significance in neighborhood type was also found in high-income groups.
Unexpectedly, they yielded inconclusive and mixed impact of neighborhood type on various
measurements of active travel frequency, which seems to contradict general expectation. For
116

instance, high-income groups reported that urban neighborhood type was negatively
associated with daily total active travel frequency and daily walking measured by a binary
scale, but it was positively associated with daily walking frequency measured by a
continuous scale. In addition, the separated models did not find any significant indirect
impact of individual/household characteristics on active travel frequency via neighborhood
type. Based on this result, it is confirmed that there was no significant role of residential self-
selection in population subgroups.

117

Table 5-13 Results from path analysis on active travel frequency by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.621 1.028 5.717 *** 0.597 0.989 2.83 *** 0.299 0.559 2.671 ***
AGE -0.002 -0.048 -1.411 -0.005 -0.083 -2.087 ** 0 0.003 0.134
GNDR (MALE=1) -0.09 -0.075 -2.15 ** 0.03 0.025 0.705 -0.003 -0.003 -0.128
WHITE 0.045 0.036 0.907 -0.01 -0.007 -0.067 0.034 0.03 0.689
BLACK -0.086 -0.04 -1.082 -0.133 -0.053 -0.78 -0.105 -0.045 -1.281
ASIAN -0.027 -0.011 -0.283 -0.139 -0.065 -0.84 -0.039 -0.025 -0.657
OTHERS 0.007 0.004 0.115 0.284 0.089 1.559 0 0 0.005
EDC_LOW 0.078 0.059 1.62 -0.014 -0.002 -0.036 0.069 0.024 0.972
EDC_HIGH 0.003 0.002 0.047 0.068 0.055 1.51 0.064 0.06 2.125 **
HHINC_LOW
HHINC_HIGH
EMPLY -0.159 -0.131 -3.47 *** -0.105 -0.069 -1.874 * -0.047 -0.038 -1.499
NUM_CAR -0.109 -0.262 -2.292 ** -0.064 -0.098 -1.163 -0.033 -0.074 -2.175 **
ACT_DENSITY 0.002 0.129 3.019 *** 0.002 0.086 1.978 ** 0 0.011 0.441
STREET_DESIGN 0.001 0.011 0.315 0.009 0.16 3.654 *** 0.001 0.02 0.753
NUM_PARK 0 0 0.013 0 0 0.001 0.021 0.024 1.014
SAFTY_CONCERN -0.06 -0.036 -0.742 -0.138 -0.065 -1.354 -0.002 -0.001 -0.03
FREEWAY_100M -0.017 -0.011 -0.34 -0.084 -0.047 -1.131 0.056 0.033 1.395
CRIME_QMILE -0.002 -0.063 -1.703 * -0.005 -0.139 -2.505 ** 0.001 0.018 0.64
NT_URBAN 0.043 0.072 1.034 -0.014 -0.023 -0.3 -0.004 -0.007 -0.114
NT_SUBUR -0.049 -0.114 -1.04 0.031 0.056 0.354 0.005 0.01 0.114
ATTITUDE 0.08 0.058 0.519 0.133 0.086 1.649 * 0.064 0.05 1.449
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -0.168 -1.79 * -0.115 -0.105 -0.578 -0.16 -0.188 -2.443 **
ATTITUDE 0.101 0.044 0.363 0.291 0.112 0.677 -0.003 -0.001 -0.014
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.632 -0.656 -0.933 0.508 0.424 0.918 0.103 0.121 1.175
ATTITUDE -0.521 -0.163 -0.19 0.022 0.008 0.021 -0.276 -0.112 -0.891
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.026 0.063 0.59 0.017 0.026 0.36 0 0 0.005
ATTITUDE -0.021 -0.015 -0.155 -0.005 -0.003 -0.136 0.001 0.001 0.113
SAMPLE SIZE 885 813 1788
R-SQUARE 0.127 0.098 0.024
Chi-sq/df 3.7942 2.18958 9.4156
CFI (=1.0) 0.846 0.901 0.858
TLI (=1.0) 0.825 0.872 0.753
RMSEA (<.08) 0.045 0.04 0.078
Variables
Daily active travel frequency
Neighborhood Type_SUBUR
Neighborhood Type_URBAN
Low Income High income Medium income
Daily active travel frequency

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

118

Table 5-14 Results from path analysis on walking frequency by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.827 1.014 5.563 *** 0.84 1.019 2.931 *** 0.336 0.439 2.532 **
AGE -0.004 -0.06 -1.724 * -0.006 -0.074 -1.88 * 0.001 0.011 0.46
GNDR (MALE=1) -0.209 -0.128 -3.344 *** 0.022 0.013 0.354 0.005 0.003 0.128
WHITE 0.038 0.023 0.569 -0.084 -0.041 -0.448 0.036 0.022 0.477
BLACK -0.029 -0.01 -0.291 -0.205 -0.06 -0.919 -0.14 -0.042 -1.156
ASIAN 0.012 0.003 0.092 -0.299 -0.102 -1.325 -0.095 -0.042 -0.999
OTHERS 0.025 0.01 0.301 0.068 0.016 0.257 -0.002 -0.001 -0.022
EDC_LOW 0.135 0.076 2.196 ** -0.115 -0.011 -0.117 0.075 0.018 0.675
EDC_HIGH -0.045 -0.018 -0.449 0.12 0.071 1.732 * 0.108 0.071 2.557 **
HHINC_LOW
HHINC_HIGH
EMPLY -0.234 -0.142 -3.936 *** -0.129 -0.062 -1.595 0.005 0.003 0.103
NUM_CAR -0.125 -0.221 -2.342 ** -0.051 -0.057 -0.572 -0.059 -0.092 -2.723 ***
ACT_DENSITY 0.004 0.173 5.469 *** 0.002 0.066 1.508 0 0.011 0.42
STREET_DESIGN 0 -0.003 -0.085 0.011 0.136 4.459 *** 0.003 0.035 1.421
NUM_PARK 0.008 0.006 0.18 -0.007 -0.005 -0.136 0.026 0.02 0.923
SAFTY_CONCERN -0.013 -0.006 -0.108 -0.077 -0.026 -0.55 -0.075 -0.031 -0.981
FREEWAY_100M 0.04 0.02 0.542 -0.049 -0.02 -0.534 -0.053 -0.022 -0.907
CRIME_QMILE -0.004 -0.099 -2.407 ** -0.005 -0.101 -2.225 ** 0.001 0.019 0.694
NT_URBAN 0.044 0.055 0.904 0.006 0.007 0.1 0.009 0.013 0.227
NT_SUBUR -0.053 -0.091 -0.922 -0.067 -0.09 -0.493 0.017 0.023 0.276
ATTITUDE 0.003 0.001 0.015 0.1 0.048 0.803 0.098 0.053 1.726 *
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -0.168 -1.79 -0.115 -0.105 -0.577 -0.16 -0.188 -2.444 **
ATTITUDE 0.101 0.044 0.364 0.291 0.112 0.677 -0.002 -0.001 -0.008
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.632 -0.656 -0.933 0.508 0.424 0.917 0.103 0.121 1.174
ATTITUDE -0.519 -0.163 -0.189 0.012 0.004 0.012 -0.279 -0.113 -0.903
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.028 0.05 0.564 -0.035 -0.039 -0.443 -0.003 -0.005 -0.349
ATTITUDE -0.023 -0.012 -0.155 0.003 0.001 0.037 0.005 0.003 0.261
SAMPLE SIZE 885 813 1788
R-SQUARE 0.14 0.072 0.029
Chi-sq/df 7.16746 3.16746 11.5027
CFI (=1.0) 0.832 0.895 0.774
TLI (=1.0) 0.674 0.705 0.613
RMSEA (<.08) 0.053 0.047 0.077
Variables
Neighborhood Type_URBAN
Daily walking frequency
Low Income High income Medium income
Daily walking frequency
Neighborhood Type_SUBUR

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

119

Table 5-15 Results from path analysis on walking frequency (binary) by income groups
Coef. O/R z p Coef. O/R z p Coef. O/R z p
Direct effect
Intercepts 0.62 1.858928 5.809 *** 0.574 1.775354 2.625 *** 0.296 1.34447 2.61 ***
AGE -0.002 0.998002 -1.24 -0.005 0.995012 -2.045 ** 0 1 0.141
GNDR (MALE=1) -0.088 0.915761 -2.127 ** 0.028 1.028396 0.651 0 1 -0.013
WHITE 0.033 1.033551 0.673 0.052 1.053376 0.316 0.038 1.03873 0.751
BLACK -0.075 0.927743 -0.958 -0.05 0.951229 -0.268 -0.095 0.90937 -1.17
ASIAN -0.051 0.950279 -0.522 -0.074 0.928672 -0.406 -0.043 0.95791 -0.709
OTHERS -0.008 0.992032 -0.122 0.348 1.416232 1.735 * 0.002 1.002 0.032
EDC_LOW 0.077 1.080042 1.634 -0.022 0.97824 -0.057 0.066 1.06823 0.923
EDC_HIGH -0.002 0.998002 -0.031 0.062 1.063962 1.381 0.061 1.0629 2.029 **
HHINC_LOW
HHINC_HIGH
EMPLY -0.177 0.83778 -3.831 *** -0.129 0.878974 -2.256 ** -0.049 0.95218 -1.562
NUM_CAR -0.109 0.89673 -2.315 ** -0.061 0.940823 -1.195 -0.032 0.96851 -2.133 **
ACT_DENSITY 0.002 1.002002 2.802 *** 0.002 1.002002 1.979 ** 0 1 0.89
STREET_DESIGN 0 1 0.105 0.009 1.009041 3.637 *** 0.002 1.002 1.097
NUM_PARK 0.005 1.005013 0.142 0.004 1.004008 0.123 0.02 1.0202 0.961
SAFTY_CONCERN -0.061 0.940823 -0.762 -0.139 0.870228 -1.375 -0.013 0.98708 -0.257
FREEWAY_100M -0.019 0.981179 -0.393 0.054 1.055485 0.757 -0.055 0.94649 -1.36
CRIME_QMILE -0.001 0.999 -1.198 -0.005 0.995012 -2.525 ** 0 1 0.284
NT_URBAN 0.041 1.041852 0.995 -0.016 0.984127 -0.341 0.007 1.00702 0.211
NT_SUBUR -0.049 0.952181 -1.034 0.005 1.005013 0.06 0.002 1.002 0.047
ATTITUDE 0.082 1.085456 0.535 0.132 1.141108 1.804 * 0.061 1.0629 1.381
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.118 -1.79 * -0.115 -0.578 -0.16 -2.444 **
ATTITUDE 0.101 0.364 0.291 0.677 -0.004 -0.015
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.633 -0.934 0.508 0.917 0.103 1.172
ATTITUDE -0.522 -0.191 -0.021 -0.021 0.276 0.891
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.026 0.594 0.004 0.101 -0.001 -0.186
ATTITUDE -0.021 -0.157 0.005 0.294 -0.001 -0.045
SAMPLE SIZE 885 813 1788
R-SQUARE 0.13 0.097 0.024
Chi-sq/df 3.895 2.760417 10.1546
CFI (=1.0) 0.858 0.942 0.836
TLI (=1.0) 0.821 0.775 0.667
RMSEA (<.08) 0.05 0.047 0.072
Variables
Neighborhood Type_URBAN
Neighborhood Type_SUBUR
Daily walking frequency (walk or not)
Low Income High income Medium income
Daily walking frequency (walk or not)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

120

Table 5-16 Key predictors of active travel (pooled samples)
Direct. Sign. Direct. Sign. Direct. Sign. Direct. Sign.
Individual/household level
Age (--) *
Gender (--) *
White (+) *
Black (--) ** (--) *
Education_low (+) ** (+) **
Education_high (+) ** (+) ** (+) **
Employment (--) ** (--) ** (--) **
Income_low (+) **
Income_high
Car availability (#. car) (--) ** (--) ** (--) ** (--) **
Intrapersonal (Yes=1) (+) * (+) ** (+) **
Neighborhood environment
Activity density (+) ** (+) ** (+) **
Street design (+) ** (+) ** (+) ** (+) **
Park availability (#. park) (+) **
Safety concern (Yes=1)
Crime (#. violent) (--) ** (--) ** (--) **
Neighborhood type
NT_Urban
NT_Suburb
Indirect effect
Mediating role of NT
Continuous Binary Continuous Binary
Duration of active traval Frequency of active traval
Key variables

Note: ‘(+)’ positive impact; ‘(--)’ negative impact; ‘**’ p<0.05; ‘*’ p<0.1;

Table 5-16 summarizes the results from analytic models (with pooled samples) that
predict the frequency and duration of active travel. The influence of some of key variables in
both models was consistently similar, but the influence of others was slightly different. More
specifically, education, employment, and car availability were observed as a significant
predictor for active travel frequency at the 5% level, and this result was consistently found in
all types of active travel frequency. Meanwhile, individual personality was a more profound
predictor for the frequency than for the duration.
121

Both activity density and street design produced a significant positive association
with active travel frequency, supporting the link between the built environment and active
travel. The negative association between neighborhood crime factors and active travel
frequency was also consistently found in all pooled models with various types of
measurements, confirming the deterrent role of violent crimes in neighborhoods. Consistent
with active travel time, car availability revealed the most profound effect on active travel
frequency across all three models. Again, this finding generally supports that the available
travel options is a crucial factor affecting the observed daily active travel patterns.
As displayed in Table 5-17, however, the separated models by different income
levels indicated slightly different results from a pooled model, demonstrating that there is a
clear gap between low- and high-income groups in explaining the variation in frequency of
active travel. More specifically, highly divergent patterns between different income groups
indicate that low-income groups are more responsive to activity density than high-income
groups, whereas high-income groups are more responsive to street design than low-income
groups. This finding is in line with other studies concluding that walkability is strongly
associated with walking for transport in high-income groups (Sallis et al., 2009).
Meanwhile, both low- and high-income groups reported a consistently deterrent role
of violent crimes on total active travel frequency and walking frequency with a continuous
scale. Interestingly, however, a binary regression model that predicts the probability of daily
walking frequency of low-income groups did not produce any significant negative impact of
violent crimes. The implication made by this finding is that the exposure to neighborhood
violent crimes is crucial factor to determine whether high-income groups walk or not, but it
122

is not the case of low-income groups. This may reflect that active travel of high-income
groups is far more likely to be discretionary than that of low-income groups.
Last, none of regression models reported the significant association between
neighborhood type and active travel time/frequency. Although descriptive analysis basically
informs us of a clear difference between urban- and suburban residents in terms of the
average time (duration) and frequency of active travel in a day, there was no significant
direct influence of neighborhood type on active travel in regression models where other
explanatory variables were incorporated. Furthermore, a subsequent model exploring
interdependent relationships indicated that the link between individual/household
characteristics and neighborhood type is fairly evident, but there was little or no indirect
impact of individual/household backgrounds on active travel via neighborhood type. That is,
path model did not find any significant mediate role of neighborhood type that represents the
effect of residential self-selection.
123

Table 5-4 Key predictors of active travel (income groups)
W+Bike Walk Walk(bin) W+Bike Walk Walk(bin) W+Bike Walk Walk(bin) W+Bike Walk Walk(bin)
Gender (--) ** ** ** ** ** **
Employment (--) ** ** ** ** * **
Car availability (#. car) (--) * ** ** **
Intrapersonal (Yes=1) (+) * * **
Activity density (+) ** ** ** ** ** ** **
Street design (+) ** ** * ** ** **
Park availability (#. park) (+) *
Crime (#. violent) (--) ** ** ** * ** ** ** **
NT_Urban
NT_Suburb
Mediating role of NT
High income
Duration Frequency
Direction Key variables Low income High income Low income

Note: ‘(+)’ positive impact; ‘(--)’ negative impact; ‘**’ p<0.05; ‘*’ p<0.1;
W+Bike refers to walking or bike mode (total active travel)
Walk (bin) refers to walking measured by binary scale (i.e., duration: 30 minutes or more; frequency: walk or not)
124

CHAPTER 6.
NEIGHBORHOOD ENVIRONMENT AND OUTDOOR LEISURE

6.1. DESCRIPTIVE ANALYSIS

Before presenting the results from path analysis based on the analytic framework mentioned
in Chapter 4, the basic descriptive analysis of outdoor leisure activity helps us not only gain
a general idea about individual behavioral decision on outdoor leisure but also identify the
difference in daily patterns of outdoor leisure activity among different income groups and
spatial contexts.
Person-based analysis offers a simple statistic summary of daily outdoor leisure
frequency and duration based on total samples (3,486 adults), including individuals who did
not engage in outdoor leisure activity given a travel survey day. Merging person and daily
trip files, the analysis captures the prevalence of outdoor leisure activity per capita. To better
understand the variation in daily outdoor leisure activity, the data will be separated by
different income groups (e.g., high vs. low-income).
Table 6-1 indicates the statistic summaries of the time spent in outdoor leisure
activity sorted by different income levels. As defined in Chapter 4, three types of outdoor
leisure activity were employed in measuring the time spent in outdoor leisure activity. A
narrow concept of outdoor leisure activity is made by outdoor exercise/sport only, whereas a
broader concept is measured by aggregating outdoor exercise/sport, park activity, and leisure
active travel.
125

Table 6-1 Time spent in outdoor leisure activity by income groups
Mean Std Dev Mean Std Dev Mean Std Dev
Outdoor exercise 5.588 31.727 5.01 28.959 6.124 32.594
Outdoor exercise+park 8.506 41.325 6.847 34.337 9.801 44.322
Outdoor exercise+park+leisure trip 11.811 43.807 10.025 37.054 13.215 46.76
Pool Low income High income

Sample sizes: pooled samples (3,486), high income (813), and low income (885)

The results showed that, people spent 5.59 minutes in outdoor exercise/sport only,
8.51minutes in outdoor exercise/sport including park activity, and 11.81 minutes in outdoor
exercise/sport including park activity and leisure active travel. Compared with daily total
active travel time (6.8 minutes per day), those amounts are larger than our expectation.
According to the finding from ATUS in 2011, on average, people spend 18 minutes (0.3
hour) per day in participating in sports, exercise, and recreation. When considering that the
duration includes leisure activities occurred in both indoor and outdoor places, the result
observed here appears to be fairly reasonable and consistent with the pattern of national
samples. Not surprisingly, different income groups produced different levels of outdoor
leisure time at statistically significant level. This finding is fairly supportive of the general
assumption that high-income groups are more willing to spend their time for leisure activity
in outdoor places than low-income groups.
Table 6-2 summarizes the basic statistic results of the frequency of daily outdoor
leisure activity among different samples. A pooled sample found that the average frequency
of outdoor leisure activity is about 0.125(outdoor exercise/sport), 0.155 (including park
activity), and 0.305(including park activity and leisure active travel). Given that the average
frequency of active travel is 0.42 per day, this finding informs that people rarely engage in
126

outdoor leisure activity.
22
Consistent with the duration of outdoor leisure activity, different
income groups produced different levels of outdoor leisure frequency. However, the gap in
the frequency of outdoor leisure activity between different samples is relatively small.
Whereas the difference in outdoor exercise between different income groups is not
statistically significant, there is the statistical significance in difference in two broad types of
outdoor leisure activity.

Table 6-2 Frequency of outdoor leisure activity by income groups
Mean Std Dev Mean Std Dev Mean Std Dev
Outdoor exercise 0.125 0.401 0.118 0.377 0.127 0.375
Outdoor exercise+park 0.155 0.449 0.144 0.418 0.175 0.475
Outdoor exercise+park+leisure trip 0.305 0.786 0.279 0.734 0.308 0.79
Pool Low income High income

Sample sizes: pooled samples (3,486), high income (813), and low income (885)

As indicated in Table 6-3, different spatial contexts (i.e., high vs. low leisure-
friendly neighborhoods) also yielded a different amount of duration and frequency of
outdoor leisure activity. As assumed, people living in high leisure-friendly neighborhoods
spend more time in outdoor leisure activity than those who live in low leisure-friendly
neighborhoods, and this gap was statistically significant. Moreover, it is also consistently
founded that people living in high leisure-friendly neighborhoods engage more frequently in
outdoor leisure activity than those who live in low leisure-friendly neighborhoods, but the
gap was not statistically significant. Those findings might provide some insights on
supportive role of neighborhood environment as an opportunity setting for outdoor leisure
activity.

22
2,867 samples (approximately, 82% of respondents) did not record any outdoor leisure
activity (i.e., exercise/sport or park activity) on a given day.
127

Table 6-3 Outdoor leisure activity (duration and frequency) by neighborhood types
Duration Frequency Duration Frequency
Outdoor exercise 7.152 0.147 5.165 0.114
Outdoor exercise+park 11.873 0.186 7.408 0.14
High_leisure neighborhood Low-leisure neighborhood

Trip-based analysis provides a snap shot of (outdoor) leisure trips in terms of travel
time to reach outdoor leisure destinations and travel distance. In this process, exercise/sport
and park activity are considered as a specific type of outdoor leisure activity. The NHTS
data reported 947 trips for such activities (733 trips for outdoor exercise/sports; 214 trips for
park activity) among 1,466 trips for recreational purpose. Like in Chapter 5, the frequency
distribution of travel time for outdoor leisure is sorted by three categories: less than 5minutes,
6 to 29 minutes, and 30 minutes or more. The frequency distribution of travel distance to
outdoor leisure is also sorted by four categories: less than .25 miles, .25 to 1 miles, 1 to 3
miles, and 3 miles or more.

Table 6-4 Descriptive statistics of outdoor leisure trips
Variable N Mean Std Dev Min Max
Travel time 733 21.4 19.6808 1 179
Travel distance 733 3.44 12.0844 0 220
Travel time 947 27.8 44.7777 1 560
Travel distance 947 11.4 89.3628 0 2500

Note: Outdoor exercise/sport (N=733); Outdoor exercise/sport including park activity (N=947)

I analyzed the characteristics of leisure trip to outdoor leisure destinations in terms
of time and distance using the disaggregated data. As summarized in Table 6-4, there were
differences in travel time and distance among the types of outdoor leisure activity.
Interestingly, when park activity was added in outdoor leisure activity, the mean of travel
time and distance dramatically increases. This result implies that people usually take longer
128

leisure trips for park activity than outdoor exercise/sport, as well as spend more time to reach
such leisure destinations.
Table 6-5 and 6 present the distribution of trip duration and distance for outdoor
leisure activity. Regarding trip duration, about 14% of leisure trips for outdoor exercise/sport
recorded ‘less than 5minutes’. When park activity was included, there were some changes in
distribution patterns of trip duration. The percent of the longest trip duration (‘30 minutes or
more’) increased, whereas the percent of the shortest trip duration (‘less than 5minutes’)
slightly decreased.
Table 6-5 Frequency of outdoor leisure by travel time
Travel time Frequency (N) Percent (%) Frequency (N) Percent (%)
Less than 5mins 103 14.05 123 12.99
6 to 29 mins 440 60.03 537 56.71
30 mins or more 190 25.92 287 30.31
Total 733 100 947 100
Outdoor leisure (exercise) (exercise or park)

Table 6-6 Frequency of outdoor leisure by travel distance
Travel distance Frequency (N) Percent (%) Frequency (N) Percent (%)
Less than .25 126 17.19 159 16.79
.25 to 1 miles 264 36.02 297 31.36
1 to 3 miles 179 24.42 216 22.81
3 miles or more 164 22.37 275 29.04
Total 733 100 947 100
Outdoor leisure (exercise) (exercise or park)

The result of trip distance was also consistent with the distribution pattern of trip
duration. About 53% of outdoor exercise/sport was ‘less than 1 mile’, but notably leisure
trips with the shortest distance (‘less than 0.25 miles’) recorded a relatively small portion
(17% of outdoor exercise/sport). Like in the duration of outdoor leisure trip, when park
activity was included, the percent of outdoor leisure trips with a relatively long distance (‘3
miles or more’) increased, whereas the percent of outdoor leisure trips with a relatively short
129

distance (‘less than 1 mile’) decreased. An interesting point is that the evidence of duration
decay function is well established, whereas distance decay function is less evident in outdoor
leisure trips (especially for park activity).
Based on those findings, it can be concluded that both duration and distance of
leisure trips for park activity are substantially longer than those for outdoor exercise/sport. In
other words, it is likely that people choose local facilities at close range for outdoor
exercise/sport, but it is not the case for park activity. A trade-off between trip duration to
reach outdoor leisure places and activity duration at such destinations might provide a clue
to this conclusion. Potentially, people might choose adjacent outdoor places for outdoor
exercise/sport in order to reduce travel time, thereby spending more time in outdoor leisure
activity at destination. On the other hand, people are willing to spend more time on the road
to reach a high-quality park facility if local areas do not offer such amenities.

6.2. REGRESSION RESULTS

6.2.1. Outdoor leisure activity time

Unstratified model (pooled samples)
Table 6-7 outlines the results from analytic models that predict the duration of
outdoor leisure activity. Three count variables measured by the time spent in outdoor leisure
activity were regressed in Model 1 (outdoor exercise/sport only), Model 2 (outdoor
exercise/sport including park activity), and Model 3 (outdoor exercise/sport including park
activity and leisure active travel time), respectively. Before the specific interpretation, the
130

explanatory power (r-square) of each pool model was extremely small, but several indexes
for general model fit of path analysis fall under the acceptable levels.
First of all, the duration of outdoor exercise/sport was predicted in Model 1.
Regarding individual/household characteristics, gender (a male), White, and high levels of
education/income were positively associated with the time spent in outdoor exercise/sport,
whereas employment and low levels of education/income revealed the negative association.
Even though those findings are in line with conventional wisdom that outdoor recreation is
more prevalent among people with higher SES, such connections were not statistically
significant in this data.
As another part of individual/household backgrounds, both car availability (i.e., the
number of vehicles in household) and household with children were positively related to the
time spent in outdoor exercise/sport. This finding seems to be reasonable because available
cars can improve individual mobility to reach leisure destinations, and because the children
might provide some motivation to involve in outdoor recreation with family. On the other
hand, individual time constraints (e.g., time spent at workplace as well as on the road) and
physical disability revealed the negative relationship to the duration of outdoor
exercise/sport, as expected. It is seemingly self-evident that the likelihood people spend the
time in outdoor exercise/sport is very low when they have the limited time or lack physical
ability, but this analytic model only observed a pronounced influence of individual time
constraints on outdoor exercise/sport at the 10% of statistical significance level.
More importantly, individual personality on outdoor leisure, which is often
considered as a strong predictor, also yielded a significant impact. This finding is fairly
supportive to theoretical foundation that puts an emphasis on the role of human
131

psychological dimensions, such as attitude, preference, and motivation, in behavioral
decision, confirming that the duration of outdoor exercise/sport can increase when people
have a strong personality on such activities.
Furthermore, there was the negative association between the amount of daily total
active travel time and the duration of outdoor exercise/sport. This finding suggests that the
time spent in outdoor exercise/sport can decrease as the amount of active travel time
increases. Given a large amount of time spent in active travel, people might be less willing to
allocate their spare times to outdoor exercise/sport. Although the argument for a substitute
relationship is fairly plausible, there was no significant impact of the duration of active travel
on the time spent in outdoor exercise/sport.
Regarding other exogenous variables, both seasonal and date effect are often
considered as a crucial factor affecting behavioral decision on outdoor activity. As expected,
the result revealed that both factors were positively associated with outdoor exercise/sport.
This finding is fairly consistent with the evidence from previous leisure studies, suggesting
that people are more likely to enjoy outdoor leisure activity on weekends or during warm
weather. However, no statistical significance was found in both factors.
Among the explanatory variables at the neighborhood level, the presence of sport
facility in outdoor reported a significantly positive influence on the duration of outdoor
exercise/sport. Based on this result, it is confirmed that the provision of sport/recreation
facilities within a neighborhood can contribute to increasing the time people spend in
outdoor exercise/sport, all else being equal. When considering that the destination for
outdoor exercise/sport can be located beyond the defined spatial boundary of neighborhood,
this finding is somewhat interesting. Even though the observed relationship does not
132

necessarily represent the causation, a strong connection between neighborhood resources for
leisure opportunity and outdoor leisure activity might support the argument “If you build it,
they will come”.
Along with the facilitating role in outdoor exercise/sport, several neighborhood
barriers that may deter outdoor exercise/sport were employed in this analytic model. The
results found that perceived safety concern, proximity to freeways, and traffic accidents were
negatively associated with the duration of outdoor exercise/sport, as expected. However,
unlike the results from analytic models for active travel, such deterrent factors were not
statistically significant.
Two variables for neighborhood type with different levels of leisure amenities (i.e.,
high- and low-leisure friendly neighborhoods) were employed in the path model to test
whether residential location (i.e., where people live) has a direct impact on the duration of
outdoor exercise/sport. Consistent with the expectation, the result showed that people living
in high leisure-friendly neighborhoods spend more time in outdoor exercise/sport, while
residents in low leisure-friendly neighborhoods spend less time. However, the direct impact
of neighborhood type was not statistically significant.
Furthermore, it can be argued that the choice of residential location can mask the
relationship between the built environment and individual recreational behavior, particularly
outdoor exercise/sport. To address this issue, neighborhood type as a proxy for residential
location was additionally incorporated in subsequent regression models, in tandem with
other neighborhood attributes. More specifically, neighborhood type as an endogenous
variable was regressed on household income levels, the number of vehicles in household,
and individual personality on outdoor leisure, respectively.
133

Basically, this approach is to explore a chain of links between individual/household
characteristics, neighborhood type, and outdoor leisure activity time, and hence to identify
an indirect effect of individual/household characteristics on outdoor leisure time via
neighborhood type. However, the results from the path model did not reveal any significant
effect of residential self-selection on the duration of outdoor exercise/sport. In other words,
the data employed in this analysis did not support the mediating role of neighborhood type.
Regarding the relative importance among variables, standardized coefficients
present that the impact of employment (Beta: -0.057), attitudinal factor (Beta: 0.052), and
time-constraint (Beta: -0.043) was relatively magnitude. It is worth noting that the
availability of outdoor sport facilities as an opportunity setting in neighborhoods also yielded
a profound impact (Beta: 0.043). This finding supports that the decision on outdoor leisure
largely depends on individual characteristics (e.g., available time and psychological
dimensions), as well as on neighborhood characteristics (e.g., leisure amenities/resources).
The second column in Table 6-7 displays the results of analytic model which focuses
on the duration of outdoor exercise/sport including park activity. In general, the regression
model revealed the similar results to previous model, but some factors produced different
patterns. More specifically, employment, time-constraint, and the availability of sport
facilities had a consistently significant impact on the duration of outdoor exercise/sport
including park activity.
Notably, this model also found that age, weather condition, and household with
children were significantly related to such leisure outcomes with expected (positive)
direction. Those findings are somewhat reasonable when considering that the older is more
likely to have available times for leisure than the younger, that warm-season seems to be a
134

more feasible time to enjoy outdoor leisure amenity, and that children might generate some
demands for outdoor leisure.
More interestingly, when park activity was included in outdoor exercise/sport, the
amount of daily total active travel time revealed a negative impact at the 5% significant level.
This finding confirmed that the time spent in outdoor leisure activity can decrease as people
spend more time in active travel. As mentioned before, the potential reason behind this
substitute relationship between active travel and outdoor leisure activity is that people might
be less willing to allocate their spare times to outdoor leisure activity, while regarding active
travel (both utilitarian and recreational) as another type of outdoor activity.
Another factor leading to different results is individual personality on outdoor
exercise. While the previous model (outdoor exercise/sport only) reported a significant
profound impact of attitudinal factor, this model including park activity did not find any
significant influence. The inconclusive finding on the role of attitudinal factor is not
surprising since this variable was measured by whether or not respondents engage in active
travel mainly for exercise. In other words, this measurement might be explicitly linked to
behavioral intention of outdoor exercise/sport, but it is not the case for park activity.
Last, the third column in Table 6-7 indicates the results of regression model
computed for a broader sense of outdoor leisure activity. When park activity and leisure
active travel were added up to outdoor exercise/sport, not only the negative impact of active
travel time but also the positive impact of the availability of sport facility persists and
remains statistically significant. Neighborhood type also reported a significant direct impact,
suggesting that people living in low leisure-friendly neighborhoods spend less time in
outdoor leisure activity.
135

However, unlike the finding from previous models, there was no significant
pronounced effect of individual/household backgrounds (e.g., employment, time constraint,
and household with children) in this model. It is not entirely clear why some factors carry on
a consistently significant role in outdoor leisure time but others do not. Potentially, this
inconclusive evidence is due in part to the broad definition of dependent variable. In other
words, the mixture of outdoor leisure activity and active travel for leisure might obscure a
unique function of specific types of outdoor leisure activity.

136

Table 6-7 Results from path analysis for outdoor leisure activity time
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 2.299 0.075 0.446 4.777 0.142 1.091 5.71 0.133 0.894
AGE 0.054 0.022 0.997 0.113 0.042 1.831 * 0.104 0.03 1.265
GNDR (MALE=1) 0.557 0.009 0.414 1.081 0.016 0.789 0.208 0.002 0.126
WHITE 3.187 0.051 1.275 0.061 0.001 0.029 0.536 0.006 0.191
BLACK 2.242 0.018 0.704 0.012 0 0.004 4.646 0.027 1.316
ASIAN -2.205 -0.022 -0.519 -4.369 -0.039 -1.368 0.189 0.001 0.054
OTHERS 2.724 0.023 0.836 1.198 0.009 0.351 1.531 0.009 0.438
EDC_LOW -3.029 -0.029 -0.819 -0.138 -0.001 -0.046 4.485 0.031 1.4
EDC_HIGH 0.85 0.014 0.659 0.431 0.006 0.285 -0.292 -0.003 -0.153
HHINC_LOW -1.37 -0.02 -0.74 -1.258 -0.016 -0.657 -1.535 -0.016 -0.626
HHINC_HIGH 1.059 0.015 0.546 2.366 0.03 1.257 2.175 0.021 0.933
EMPLY -3.837 -0.057 -2.469 ** -3.295 -0.044 -1.748 * -0.695 -0.007 -0.322
DISABLE -7.001 -0.055 -1.283 -3.813 -0.027 -1.109 -0.504 -0.003 -0.135
WEEKEND 0.65 0.009 0.379 0.089 0.001 0.055 1.747 0.018 0.832
WEATHER (WARM=1) 0.677 0.011 0.52 2.254 0.033 1.649 * 0.644 0.007 0.353
CHILD_OWN 2.286 0.036 1.484 2.665 0.038 1.68 * 1.547 0.017 0.759
TIME_CONSTRAINT -0.005 -0.043 -1.757 * -0.007 -0.053 -2.261 ** -0.004 -0.026 -1.1
NUM_CAR 0.137 0.006 0.223 0.406 0.015 0.594 0.698 0.02 0.983
ACTIVE TRAVEL -0.116 -0.064 -0.888 -0.195 -0.098 -1.992 ** -0.254 -0.1 -2.046 **
SPORT_FACILITY 4.251 0.043 2.394 ** 4.103 0.037 2.19 ** 5.228 0.037 1.943 *
SAFTY_CONCERN -0.264 -0.003 -0.131 -0.469 -0.005 -0.239 -2.885 -0.022 -0.982
FREEWAY_100M -1.28 -0.014 -0.551 -1.266 -0.013 -0.65 -1.808 -0.015 -0.798
TRAFFIC ACC -0.012 -0.002 -0.08 -0.036 -0.004 -0.227 -0.41 -0.041 -1.364
NT_LEISURE HIGH 0.555 0.018 0.878 0.748 0.022 1.081 0.921 0.022 1.155
NT_LEISURE LOW -0.469 -0.015 -0.895 -0.759 -0.023 -1.31 -1.531 -0.036 -2.004 **
ATTITUDE 3.192 0.052 2.336 ** 1.841 0.027 1.208 1.451 0.017 0.834
HHINC_LOW 0.05 0.022 0.731 0.05 0.022 0.732 0.05 0.022 0.733
HHINC_HIGH -0.066 -0.028 -1 -0.066 -0.028 -1.005 -0.065 -0.028 -0.997
NUM_CAR -0.066 -0.083 -2.929 *** -0.066 -0.083 -2.929 *** -0.066 -0.083 -2.932 ***
ATTITUDE 0.036 0.018 0.754 0.036 0.018 0.751 0.036 0.018 0.75
HHINC_LOW -0.002 -0.001 -0.03 -0.002 -0.001 -0.036 -0.003 -0.001 -0.04
HHINC_HIGH 0.063 0.027 1.038 0.064 0.027 1.042 0.063 0.027 1.03
NUM_CAR 0.013 0.016 0.674 0.013 0.016 0.671 0.013 0.016 0.669
ATTITUDE -0.019 -0.009 -0.427 -0.019 -0.009 -0.427 -0.019 -0.009 -0.426
Indirect effect
HHINC_LOW -0.027 0 -0.64 0.039 0.001 0.398 0.05 0.001 0.324
HHINC_HIGH 0.007 0 0.093 -0.098 -0.001 -0.833 -0.157 -0.002 -0.919
NUM_CAR 0.031 0.001 0.649 -0.059 -0.002 -1.007 -0.081 -0.002 -1.069
ATTITUDE 0.011 0 0.32 0.041 0.001 0.568 0.062 0.001 0.559
SAMPLE SIZE 3486 3486 3486
R-SQUARE 0.021 0.023 0.02
Chi-sq/df 6.37574 6.36963 5.90321
CFI (=1.0) 0.988 0.992 0.997
TLI (=1.0) 0.976 0.984 0.994
RMSEA (<.08) 0.039 0.039 0.038
Variables (Duration)
(LEISURE_HIGH)
(LEISURE_LOW)
Exercise/Park Activity
Model 1
Exercise/Sport only
(LEISURE_HIGH)
Model 3 Model 2
Exercise/Park Activity
(LEISURE_LOW)
Exercise/Park/Leisure AT
(LEISURE_HIGH)
(LEISURE_LOW)
Exercise/Park/Leisure AT Exercise/Sport only

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
137

Stratified model (income levels)
Analytic models separated by three income levels were employed in order to offer
adequate explanation on whether subpopulations have a unique function of outdoor leisure
time. Before specific interpretation on the results, stratified regression models present that
the overall explanatory power (r-square) was slightly small. In addition, compared with a
model with pool samples, the separated path models indicated poor but statistically
acceptable model fit indexes.
As shown in Table 6-8, 9, and 10, the separated models yielded very different results
than a pool model. Unexpectedly, several factors that were significantly associated with the
duration of outdoor leisure activity in a pool model lost their significance. Employment,
time-constraint, active travel, and attitudinal factor did not report any significant association
in separated models. The potential reason behind insignificant results is that the reduced
sample size in separated models might lead to insufficient variation in dependent variable.
Meanwhile, as indicated in Table 6-9 and 10, no significant predictor was found when
dependent variable includes park activity and leisure active travel. As mentioned before, it
can be interpreted that a broad concept of outdoor leisure activity might obscure a unique
function of specific types of outdoor leisure activity.
Despite this methodological issue, some variables indicated a significant impact on
the duration of outdoor leisure activity. More specifically, low-income groups found the
positive association between neighborhood type and time spent in outdoor exercise/sport at
the 5% significance level, presenting that low-income groups living in leisure-friendly
neighborhoods are likely to spend more time in outdoor exercise/sport. On the other hand,
138

high-income groups did not find any significant relationship between neighborhood type and
the duration of outdoor exercise/sport.
The different results might reflect mobility issue that low-income groups encounter.
In general, low-income groups have a limited ability to reach leisure destinations beyond
their neighborhoods, compared with high-income groups. Given the limited leisure options
in neighborhoods, low levels of mobility of low-income groups can decrease the amount of
time spend in outdoor leisure activity even though they want to enjoy outdoor leisure. In this
case, leisure-friendly neighborhoods can allow low-income groups to participate in outdoor
leisure activity by providing leisure destinations easily accessible.
139

Table 6-8 Results of outdoor exercise/sport time by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 4.798 0.195 0.662 8.338 0.283 0.205 -3.026 -0.1 -0.265
AGE 0.051 0.027 0.435 0.044 0.015 0.216 0.098 0.04 1.217
GNDR (MALE=1) 1.889 0.038 0.77 -0.391 -0.007 -0.109 1.359 0.023 0.598
WHITE 1.526 0.03 0.607 5.505 0.076 0.141 3.062 0.048 0.354
BLACK 1.036 0.012 0.258 6.973 0.057 0.177 3.313 0.025 0.354
ASIAN 0.269 0.003 0.037 2.183 0.021 0.054 0.097 0.001 0.009
OTHERS 1.521 0.02 0.346 9.597 0.062 0.243 3.594 0.029 0.385
EDC_LOW -2.522 -0.047 -0.749 -4.913 -0.013 -0.04 -2.573 -0.016 -0.193
EDC_HIGH -2.231 -0.029 -0.335 2.326 0.038 0.546 -0.867 -0.014 -0.405
HHINC_LOW
HHINC_HIGH
EMPLY -2.915 -0.059 -1.016 -6.446 -0.087 -1.379 -2.439 -0.035 -0.945
DISABLE -1.209 -0.015 -0.203 -5.018 -0.034 -0.435 -3.544 -0.025 -0.339
WEEKEND 2.412 0.042 0.549 0.871 0.013 0.192 -0.584 -0.009 -0.232
WEATHER (WARM=1) 1.15 0.023 0.392 1.644 0.028 0.395 0.048 0.001 0.021
CHILD_OWN 1.854 0.037 0.714 -7.912 -0.025 -0.063 3.506 0.058 1.617
TIME_CONSTRAINT -0.007 -0.075 -1.527 -0.004 -0.032 -0.398 -0.002 -0.021 -0.537
NUM_CAR -0.362 -0.021 -0.32 -1.523 -0.048 -0.675 0.273 0.011 0.247
ACTIVE TRAVEL -0.073 -0.053 -0.387 -0.143 -0.084 -0.615 -0.112 -0.06 -0.669
SPORT_FACILITY 0.326 0.004 0.057 2.837 0.033 0.582 6.921 0.07 2.81 ***
SAFTY_CONCERN -0.904 -0.013 -0.186 -0.604 -0.006 -0.093 1.39 0.015 0.52
FREEWAY_100M 0.568 0.009 0.175 -0.965 -0.011 -0.165 -2.24 -0.023 -0.539
TRAFFIC ACC -0.153 -0.035 -0.411 0.04 0.005 0.088 0.186 0.021 0.657
NT_LEISURE HIGH 1.394 0.057 1.979 ** 0.049 0.002 0.059 0.067 0.002 0.091
NT_LEISURE LOW 1.008 0.041 1.312 -0.067 -0.002 -0.082 -1.126 -0.037 -1.814 *
ATTITUDE 4.532 0.09 1.107 5.476 0.093 1.697 * 2.112 0.035 1.071
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.022 -0.104 -0.124 -3.338 ***
ATTITUDE 0.076 0.037 0.756 -0.018 -0.009 -0.182 0.048 0.024 0.702
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.062 -1.229 -0.011 -0.01 -0.214 0.067 0.08 2.402 **
ATTITUDE -0.069 -0.034 -0.74 -0.009 -0.004 -0.098 0.003 0.001 0.046
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.048 0.003 0.554 0.001 0 0.066 -0.068 -0.003 -0.626
ATTITUDE 0.175 0.003 0.744 -0.002 0 -0.124 0.006 0 0.086
SAMPLE SIZE 885 813 1788
R-SQUARE 0.035 0.031 0.02
Chi-sq/df 6.79103 4.06759 10.1245
CFI (=1.0) 0.876 0.843 0.753
TLI (=1.0) 0.848 0.821 0.717
RMSEA (<.08) 0.075 0.062 0.08
NeighborhoodType_LEISURE LOW
Exercise/Sport only
Low Income High income
Exercise/Sport only
NeighborhoodType_LEISURE HIGH
Medium income
Variables (Duration)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
140

Table 6-9 Result of outdoor exercise/sport (including park activity) time by income groups
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 8.164 0.298 1.336 4.128 0.11 0.252 0.453 0.014 0.075
AGE 0.048 0.022 0.429 0.145 0.038 0.794 0.114 0.045 1.221
GNDR (MALE=1) 2.044 0.037 0.72 -0.145 -0.002 -0.031 0.952 0.015 0.475
WHITE 2.473 0.043 0.853 0.989 0.011 0.09 -1.425 -0.021 -0.502
BLACK -3.244 -0.033 -0.393 -1.355 -0.009 -0.097 2.763 0.02 0.722
ASIAN -0.692 -0.006 -0.119 -5.17 -0.039 -0.393 -3.286 -0.035 -0.795
OTHERS 1.569 0.018 0.36 -5.679 -0.028 -0.275 0.736 0.006 0.165
EDC_LOW -1.462 -0.024 -0.365 -10.74 -0.022 -0.081 7.697 0.045 1.914 *
EDC_HIGH -1.074 -0.013 -0.279 5.139 0.066 1.236 -0.182 -0.003 -0.087
HHINC_LOW
HHINC_HIGH
EMPLY -3.088 -0.056 -0.848 -1.854 -0.02 -0.378 -2.658 -0.037 -1.05
DISABLE -1.784 -0.02 -0.347 -5.488 -0.029 -0.291 -0.77 -0.005 -0.177
WEEKEND 3.656 0.058 1.317 -1.264 -0.014 -0.202 -0.437 -0.006 -0.19
WEATHER (WARM=1) 0.445 0.008 0.166 2.485 0.033 0.624 2.311 0.037 1.197
CHILD_OWN 0.856 0.015 0.274 3.956 0.01 0.241 4.467 0.071 2.404 **
TIME_CONSTRAINT -0.004 -0.038 -0.686 -0.005 -0.034 -0.519 -0.007 -0.063 -1.747 *
NUM_CAR -0.473 -0.025 -0.454 -0.03 -0.001 -0.01 1.279 0.049 1.47
ACTIVE TRAVEL -0.146 -0.096 -0.842 -0.233 -0.107 -1.189 -0.194 -0.099 -1.645
SPORT_FACILITY 6.623 0.066 1.676 * 3.017 0.027 0.49 1.424 0.014 0.475
SAFTY_CONCERN 0.732 0.01 0.172 -8.902 -0.067 -1.532 1.146 0.012 0.387
FREEWAY_100M 0.827 0.012 0.244 -4.435 -0.039 -0.773 -0.141 -0.001 -0.044
TRAFFIC ACC -0.277 -0.057 -0.741 -0.443 -0.043 -0.397 0.647 0.069 1.298
NT_LEISURE HIGH 0.99 0.036 1.461 0.816 0.022 0.536 0.45 0.014 0.509
NT_LEISURE LOW -0.319 -0.012 -0.457 -1.706 -0.045 -1.716 * -0.21 -0.007 -0.28
ATTITUDE 3.374 0.06 1.036 2.402 0.032 0.579 0.337 0.005 0.17
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.023 -0.104 -0.124 -3.339 ***
ATTITUDE 0.076 0.037 0.758 -0.018 -0.009 -0.18 0.048 0.024 0.703
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.061 -1.228 -0.011 -0.01 -0.214 0.067 0.08 2.403 **
ATTITUDE -0.069 -0.034 -0.743 -0.009 -0.005 -0.099 0.003 0.001 0.046
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.051 -0.003 -0.927 0.019 0.16 -0.061 -0.002 -0.45
ATTITUDE 0.097 0.002 0.609 0.001 0.003 0.021 0 0.358
SAMPLE SIZE 885 813 1788
R-SQUARE 0.035 0.037 0.029
Chi-sq/df 6.7562 4.06771 10.0824
CFI (=1.0) 0.886 0.858 0.761
TLI (=1.0) 0.878 0.827 0.738
RMSEA (<.08) 0.75 0.062 0.08
NeighborhoodType_LEISURE HIGH
Low Income High income
Exercise/Sport including Park Activity
Variables (Duration)
Medium income
NeighborhoodType_LEISURE LOW
Exercise/Sport including Park Activity

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
141

Table 6-10 Results of outdoor leisure (exercise/park/active leisure travel) time by income
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 12.987 0.376 1.42 2.71 0.062 0.133 1.34 0.031 0.139
AGE -0.02 -0.008 -0.142 0.251 0.057 1.09 0.142 0.04 1.091
GNDR (MALE=1) 0.301 0.004 0.094 2.791 0.032 0.693 -2.643 -0.03 -0.922
WHITE -2.327 -0.032 -0.532 3.365 0.031 0.262 -0.191 -0.002 -0.042
BLACK 6.443 0.052 1.164 0.705 0.004 0.039 9.92 0.052 1.85 *
ASIAN 0.481 0.003 0.065 5.72 0.037 0.392 0.322 0.002 0.061
OTHERS 5.919 0.055 1.282 -1.855 -0.008 -0.122 -2.934 -0.016 -0.375
EDC_LOW 3.642 0.048 0.909 -32.99 -0.059 -0.185 12.42 0.052 2.229 **
EDC_HIGH -1.285 -0.012 -0.224 1.338 0.015 0.326 -0.001 0 0
HHINC_LOW
HHINC_HIGH
EMPLY -2.175 -0.031 -0.561 1.675 0.015 0.304 -0.511 -0.005 -0.146
DISABLE -1.628 -0.015 -0.279 3.698 0.017 0.42 1.373 0.007 0.238
WEEKEND 0.854 0.011 0.224 5.155 0.05 1.004 2.019 0.021 0.632
WEATHER (WARM=1) 3.536 0.051 1.135 0.214 0.002 0.053 1.42 0.016 0.498
CHILD_OWN -2.786 -0.04 -0.878 -13.248 -0.028 -0.282 4.302 0.049 1.538
TIME_CONSTRAINT -0.005 -0.041 -0.804 0.003 0.019 0.334 -0.008 -0.049 -1.319
NUM_CAR 1.137 0.048 1.317 -1.399 -0.029 -0.484 1.357 0.037 1.231
ACTIVE TRAVEL -0.168 -0.088 -0.786 -0.315 -0.124 -1.633 -0.258 -0.095 -1.389
SPORT_FACILITY 6.074 0.048 1.152 -1.15 -0.009 -0.181 3.735 0.026 0.909
SAFTY_CONCERN -2.595 -0.028 -0.437 1.91 0.012 0.197 -3.811 -0.028 -0.806
FREEWAY_100M 1.293 0.015 0.308 -8.442 -0.064 -1.896 * -2.015 -0.014 -0.47
TRAFFIC ACC -0.699 -0.114 -1.345 -0.37 -0.031 -0.363 0.123 0.009 0.29
NT_LEISURE HIGH 0.405 0.012 0.412 2.556 0.058 1.68 * 1.211 0.028 1.182
NT_LEISURE LOW -1.249 -0.036 -1.195 -2.87 -0.065 -1.905 * -1.503 -0.034 -1.402
ATTITUDE 3.695 0.052 0.927 0.329 0.004 0.081 1.528 0.017 0.543
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.024 -0.104 -0.124 -3.339 ***
ATTITUDE 0.076 0.037 0.757 -0.018 -0.009 -0.18 0.048 0.024 0.703
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.061 -1.228 -0.011 -0.01 -0.214 0.067 0.08 2.402 **
ATTITUDE -0.069 -0.034 -0.738 -0.009 -0.004 -0.098 0.003 0.001 0.046
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.027 0.001 0.364 0.034 0.001 0.131 -0.227 -0.006 -1.245
ATTITUDE 0.117 0.002 0.588 -0.021 0 -0.043 0.054 0.001 0.317
SAMPLE SIZE 885 813 1788
R-SQUARE 0.046 0.043 0.025
Chi-sq/df 6.79092 4.06839 10.1259
CFI (=1.0) 0.891 0.873 0.784
TLI (=1.0) 0.882 0.848 0.736
RMSEA (<.08) 0.075 0.062 0.08
NeighborhoodType_LEISURE LOW
NeighborhoodType_LEISURE HIGH
Exercise/Sport/Park/Leisure Active Travel
Exercise/Sport/Park/Leisure Active Travel
Low Income High income Medium income
Variables (Duration)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
142

6.2.2. Outdoor leisure activity frequency

Unstratified model (pooled samples)
Table 6-11 summarizes the results from analytic models that predict the frequency
of outdoor leisure activity. Before the specific interpretation, the explanatory power (r-
square) of each pool model was overall extremely small, but several indexes for general
model fit of path analysis fall under the acceptable levels.
As indicated in Model 1(outdoor exercise/sport only), by and large, statistic
prediction for the frequency of outdoor exercise/sport was fairly consistent with that for the
duration of exercise/sport. However, there are also inconclusive results between the duration
and the frequency of outdoor exercise/sport, which informs us that a function or predictor of
frequency differs from the duration. More specifically, among individual/household
characteristics, age, gender (a male), White, and high levels of income reported the positive
association with the frequency of outdoor exercise/sport, whereas employment and low
levels of income indicated the negative association. However, among those variables,
employment and age solely revealed a statistically significant impact. This finding suggests
that adults who were employed are less likely to engage in outdoor exercise/sport and older
people are more likely to participate in such activities than the counterparts.
On the other hand, unlike the prediction for the duration, car availability (i.e., the
number of vehicles in household) revealed a significant and positive influence on the
frequency of outdoor exercise/sport at the 10% significance level. As mentioned before, this
finding appears reasonable because available cars can provide the opportunity for outdoor
leisure by improving individual mobility to get the access to outdoor destinations beyond
143

their neighborhoods. Moreover, the result found that individual time constraints (e.g., time
spent at workplace as well as on the road), which had a significant impact on the time spent
in outdoor exercise/sport, was not significantly associated with the frequency of outdoor
exercise/sport. This finding suggests that the available time might not be a crucial predictor
of the frequency of outdoor leisure.
More importantly, the amount of daily total active travel time did not find any
significant impact on the duration, but it was negatively associated with the frequency of
outdoor exercise/sport at the 5% significance level. This finding informs that people who
spent more time in active travel may participate less frequently in outdoor exercise/sport
than people who spent less time in active travel. Given a large amount of time spent in active
travel, this finding supports the possibility that people might be less willing to engage in
outdoor exercise/sport. Potentially, time spent in active travel can be considered as a certain
type of outdoor leisure activity for them.
As another important aspect of individual characteristics, this model found a
pronounced impact of attitudinal factor on the frequency at the 5% significance level. This
result is fairly consistent with finding from previous model for the duration of outdoor
exercise/sport. Verifying a strong role of attitude in behavioral choice for outdoor leisure
activity, this finding suggests that people with personality on outdoor exercise/sport can
spend more time and engage more frequently in such activities.
Regarding other exogenous variables, both seasonal and date effect were also
positively associated with outdoor exercise/sport as expected. This finding is fairly
consistent with the evidence from previous leisure studies, suggesting that people are more
likely to enjoy outdoor leisure activity on weekends or during warm weather. However, as
144

shown in previous models for the duration of outdoor exercise/sport, both exogenous factors
were not statistically significant in explaining the variation in the frequency of outdoor
exercise/sport.
Fairly similar results between the duration and frequency of outdoor exercise/sport
were also found in neighborhood attributes. For instance, the availability of outdoor exercise
amenities reported a significant and pronounced influence on the frequency of outdoor
exercise/sport with the expected (positive) direction. Based on this result, it is confirmed
that, all else being equal, the provision of sport/recreation facilities within a neighborhood
can contribute to increasing not only the amount of time people spend in outdoor
exercise/sport but also the likelihood people engage in such activities. Even though the
strong association was observed in this model, it does not necessarily mean the causality per
se. This is mainly because this observation based on cross-sectional analysis does not
provide any clue to time precedence between the provision of sport/recreation facilities and
the increase in outdoor leisure activity.
Along with the facilitating role in outdoor exercise/sport, several neighborhood
barriers for outdoor exercise/sport were employed in this analytic model. Consistent with the
results from the duration model, the negative associations between deterrent factors (e.g.,
perceived safety concern, proximity to freeways, and traffic accidents) and the frequency of
outdoor exercise/sport were not statistically significant.
Two variables for neighborhood type with different levels of leisure amenities (i.e.,
high- and low leisure friendly neighborhoods) were employed in the path model to test
whether residential location (i.e., where people live) has a direct impact on the frequency of
outdoor exercise/sport. Consistent with the expectation, there was the positive association
145

between high leisure-friendly neighborhoods and the frequency of outdoor exercise/sport;
and the negative association between low leisure-friendly neighborhoods and the frequency
of outdoor exercise/sport. However, those associations were not statistically significant.
Furthermore, to investigate the impact of residential self-selection that may mask the
relationship between the built environment and individual outdoor exercise/sport, subsequent
regression models additionally incorporated neighborhood type as a proxy for residential
location, in tandem with other neighborhood attributes. In this analytic framework, each
neighborhood type was also considered as an endogenous variable that has a function of
household income levels, the number of vehicles in household, and personality on outdoor
leisure. However, like in path model for the duration of outdoor exercise/sport, the data
employed in the frequency model did not support the mediating role of neighborhood type,
reporting that there was no significant effect of residential self-selection on the frequency of
outdoor exercise/sport.
Regarding the relative importance among variables indicated by standardized
coefficients, there was a clear difference between prediction for the duration and for the
frequency. More specifically, the magnitude of socio-economic (i.e., employment: Beta: -
0.057), attitudinal factor (Beta: 0.052), and time-constraint (Beta: -0.043) was relatively high
in explaining the variation in the time spent in outdoor exercise/sport. On the other hand,
daily total active travel time (Beta: -0.118) was the most important factor affecting the
number of participation in outdoor exercise/sport, followed by other significant predictors,
such as age (Beta: 0.05), employment (Beta: -0.035), attitudinal factor (Beta: 0.036), car
availability (Beta: 0.032), and the availability of outdoor sport facilities (Beta: 0.04). This
finding provides some evidence supporting that the prevalence of outdoor leisure activity
146

can be largely determined by the amount of time spent in active travel, highlighting a
substitute relationship.
Next, Model 2 (the second column) in Table 6-11 indicates the results of analytic
model devoted to outdoor exercise/sport including park activity. In general, regression
models revealed the similar results to previous model (Model 1), but some factors produced
different statistic results. More specifically, car availability, daily total active travel, and the
availability of outdoor sport facilities reported a consistently significant association with the
frequency of outdoor leisure activity.
In addition, this model also found several significant predictors, which did not yield
statistical significance in the duration of outdoor leisure activity. For instance, both gender (a
male) and high levels of income have a positive impact on the frequency of outdoor
exercise/sport including park activity, suggesting that a male or high-income person is more
likely to engage in such outdoor leisure activities. Notably, the exposure of traffic accident
revealed the negative influence on the frequency of outdoor exercise/sport including park
activity at the 5% significance level. The finding offers the supportive evidence on the
deterrent role of neighborhood traffic accidents, confirming that the likelihood people
engage in outdoor leisure activity decreases when they live in neighborhoods more exposed
to traffic accidents.
Moreover, there was a negative association between neighborhood type (low leisure-
friendly neighborhoods) and the frequency of outdoor exercise/sport including park activity
at the 5% significance level. This finding also informs us of the importance of neighborhood
environment, suggesting that people who live in low leisure-friendly neighborhoods engage
less frequently in outdoor leisure activity, in part, due to limited resources for outdoor leisure.
147

As found in the analytic model for the duration of outdoor leisure, the impact of
personality on outdoor leisure is somewhat inconclusive. While the previous model (outdoor
exercise/sport only) reported a significant effect of attitudinal factor, this model including
park activity did not find any significant influence. Apparently, the mixed finding is not
surprising since this variable can be explicitly linked to behavioral intention of outdoor
exercise/sport rather than to general park activity.
Last, Model 3 (the third column) in Table 6-11 indicates the results from regression
model computed for a broader sense of outdoor leisure activity. When park activity and
leisure active travel were added up to outdoor exercise/sport, not only the negative impact of
active travel time but also the positive impact of car availability and sport facility persists
and remains statistically significant. However, unlike the finding from previous models,
there was no significant pronounced effect of individual/household backgrounds in this
model.
148

Table 6-11 Results from path analysis for outdoor leisure activity frequency
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.047 0.121 0.974 0.047 0.108 0.904 0.198 0.259 2.263
AGE 0.002 0.05 2.604 *** 0.001 0.023 1.203 0.001 0.021 1.115
GNDR (MALE=1) 0.002 0.003 0.151 0.029 0.033 1.886 * 0.016 0.01 0.564
WHITE 0.014 0.018 0.593 0.039 0.043 1.483 0.041 0.026 0.899
BLACK 0.044 0.028 1.37 0.048 0.027 1.363 0.083 0.027 1.303
ASIAN -0.022 -0.017 -0.648 0.004 0.003 0.108 0.036 0.014 0.591
OTHERS 0.031 0.02 0.931 0.053 0.031 1.556 0.028 0.009 0.421
EDC_LOW 0.018 0.013 0.605 0.013 0.009 0.404 0.088 0.034 1.768 *
EDC_HIGH 0.014 0.018 0.89 0.001 0.001 0.063 0.021 0.013 0.629
HHINC_LOW -0.012 -0.013 -0.629 -0.024 -0.024 -1.061 -0.011 -0.006 -0.282
HHINC_HIGH 0.005 0.006 0.258 0.045 0.043 1.982 ** 0 0 0.002
EMPLY -0.03 -0.035 -1.762 * -0.012 -0.013 -0.596 -0.016 -0.009 -0.424
DISABLE -0.051 -0.032 -1.298 -0.037 -0.02 -0.979 -0.073 -0.023 -1.038
WEEKEND 0.005 0.005 0.265 0.018 0.018 0.964 0.014 0.008 0.439
WEATHER (WARM=1) 0.012 0.015 0.846 0.005 0.006 0.303 0.009 0.006 0.32
CHILD_OWN 0.017 0.021 1.047 0.028 0.031 1.541 0.039 0.025 1.238
TIME_CONSTRAINT 0 -0.027 -1.308 0 -0.033 -1.506 0 -0.002 -0.08
NUM_CAR 0.01 0.032 1.798 * 0.012 0.035 1.817 * 0.02 0.032 1.692 *
ACTIVE TRAVEL -0.003 -0.118 -2.832 *** -0.003 -0.115 -4.171 *** -0.007 -0.153 -4.489 ***
SPORT_FACILITY 0.051 0.04 2.485 ** 0.067 0.047 2.834 *** 0.149 0.059 3.884 ***
SAFTY_CONCERN -0.01 -0.009 -0.472 -0.007 -0.005 -0.289 -0.021 -0.009 -0.479
FREEWAY_100M -0.015 -0.013 -0.66 -0.005 -0.004 -0.225 -0.016 -0.007 -0.388
TRAFFIC ACC -0.002 -0.026 -1.156 -0.005 -0.045 -2.059 ** -0.002 -0.009 -0.445
NT_LEISURE HIGH 0.003 0.008 0.38 0.014 0.031 1.513 -0.006 -0.008 -0.332
NT_LEISURE LOW -0.004 -0.011 -0.563 -0.018 -0.041 -2.078 ** -0.002 -0.002 -0.102
ATTITUDE 0.028 0.036 2.009 ** 0.023 0.026 1.408 0.042 0.027 1.519
HHINC_LOW 0.05 0.022 0.732 0.05 0.021 0.727 0.05 0.022 0.73
HHINC_HIGH -0.066 -0.028 -0.998 -0.066 -0.028 -0.998 -0.066 -0.028 -1
NUM_CAR -0.066 -0.083 -2.931 *** -0.066 -0.083 -2.93 *** -0.066 -0.083 -2.93 ***
ATTITUDE 0.036 0.018 0.753 0.036 0.018 0.751 0.036 0.018 0.748
HHINC_LOW -0.002 -0.001 -0.032 -0.002 -0.001 -0.035 -0.002 -0.001 -0.033
HHINC_HIGH 0.063 0.027 1.037 0.063 0.027 1.037 0.064 0.027 1.038
NUM_CAR 0.013 0.016 0.671 0.013 0.016 0.672 0.013 0.016 0.671
ATTITUDE -0.019 -0.009 -0.427 -0.019 -0.009 -0.427 -0.019 -0.009 -0.427
Indirect effect
HHINC_LOW 0 0 -0.319 0.001 0.001 0.365 0 0 -0.3
HHINC_HIGH 0 0 -0.067 -0.002 -0.002 -0.948 0 0 0.138
NUM_CAR 0 0.001 0.248 -0.001 -0.003 -1.259 0 0.001 0.278
ATTITUDE 0 0 0.076 0.001 0.001 0.587 0 0 -0.198
SAMPLE SIZE 3486 3486 3486
R-SQUARE 0.027 0.03 0.033
Chi-sq/df 6.33506 5.64928 6.33523
CFI (=1.0) 0.889 0.992 0.918
TLI (=1.0) 0.76 0.983 0.822
RMSEA (<.08) 0.065 0.037 0.059
Model 1
Exercise/Sport only
Exercise/Park Activity
(LEISURE_HIGH)
(LEISURE_LOW)
Exercise/Park Activity
Exercise/Sport only
(LEISURE_HIGH)
(LEISURE_LOW)
Exercise/Park/Leisure AT
(LEISURE_LOW)
Variables (Frequency)
Model 3
Exercise/Park/Leisure AT
(LEISURE_HIGH)
Model 2

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
149

Stratified model (income levels)
Analytic models separated by three income levels were employed in order to offer
adequate explanation on whether subpopulations have different functions or predictors of
outdoor leisure frequency. Before specific interpretation on results from stratified models,
compared with a pool model, separated regression models present that the explanatory power
(r-square) slightly increased but it still remains small. In addition, the separated path models
indicate poor but statistically acceptable fit indexes of path analysis.
As shown in Table 6-12, 13, and 14, the separated models by three income levels
produced very different results from a pool model. Surprisingly, both car availability and the
availability of sport facility consistently reported a pronounced influence on how frequently
people engage in outdoor leisure activity in a pool model, but those factors did not produce
any significant impact in separated models. Partially, this inconclusive result might attribute
to the reduced sample size in separated models, thereby leading to insufficient variation in
dependent variable.
The analytic model devoted to outdoor exercise/sport only (Table 6-12) reported that,
for high-income groups, active travel time was significantly related to the frequency with
expected (negative) direction, whereas no significant impact of active travel time was found
in low-income groups. This finding might imply that high-income groups are more
responsive to active travel time in making the decision on outdoor exercise/sport than low-
income groups. On the other hand, the positive impact of personality on outdoor
exercise/sport persists in both income groups at the significant level, supporting the
consistently pronounced role of attitudinal factor on outdoor exercise/sport.
150

The different results across household income levels were also observed in analytic
model where park activity was included in outdoor exercise/sport (Table 6-13). Consistent
with the result of a narrow concept of outdoor leisure (Table 6-12), high-income groups
revealed a significant negative association between the amount of time spent in active travel
and outdoor leisure frequency, whereas such association was not significant in low-income
groups. On the other hand, unlike the result from the previous model, low-income groups
consistently found a significant positive role of attitudinal factor in frequency of outdoor
exercise/sport including park activity, whereas such influence was not significant in high-
income groups.
More interestingly, perceived safety concern had a negative impact on the frequency
of outdoor leisure activity in high-income groups, but low-income groups did not find any
significant influence of safety concern. It is also notable that low-income groups reported a
significant negative impact of traffic accidents on outdoor exercise/sport including park
activity, but such deterrent role of risk factor was found insignificant in high-income group.
Those findings suggest that high-income groups are more sensitive to safety concern
subjectively measured when they engage in outdoor exercise/sport or park activity, whereas
low-income groups are more responsive to exposure to traffic incidents objectively measured.
Incorporating a broader concept of outdoor leisure activity measured by time spent
in outdoor exercise/sport including park activity and leisure active travel, the analytic model
revealed a significant negative association between the duration of daily active travel and the
frequency of outdoor leisure activity in both income groups. Among the variables, the
duration of daily active travel produced the most magnitude impact (Beta: -0.152 in low-
income groups; -0.179 in high-income groups). In addition, individual personality on
151

outdoor leisure also had a significant and profound impact on the frequency of outdoor
leisure activity in high-income group impact (Beta: 0.078), but such influence was found
insignificant in low-income groups. However, there was no significant deterrent role of
perceived safety concern and traffic incidents in the analytic model devoted to a broader
concept of outdoor leisure activity.
152

Table 6-12 Results from path analysis for outdoor exercise/sport frequency by income
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.17 0.485 2.55 ** 0.046 0.138 0.317 -0.018 -0.045 -0.24
AGE 0 0.012 0.314 0.002 0.063 1.211 0.002 0.058 2.207 **
GNDR (MALE=1) 0.002 0.003 0.078 0.002 0.003 0.08 0.007 0.009 0.327
WHITE -0.001 -0.001 -0.023 0.02 0.024 0.245 0.025 0.03 0.593
BLACK -0.011 -0.008 -0.201 -0.024 -0.017 -0.244 0.105 0.061 2.019 **
ASIAN 0.062 0.043 1.119 -0.04 -0.034 -0.41 -0.013 -0.011 -0.242
OTHERS 0.009 0.008 0.216 0.057 0.032 0.584 0.055 0.034 1.001
EDC_LOW 0.022 0.029 0.682 -0.125 -0.03 -0.108 0.059 0.027 0.977
EDC_HIGH -0.006 -0.006 -0.134 0.043 0.063 1.522 -0.004 -0.004 -0.155
HHINC_LOW
HHINC_HIGH
EMPLY -0.012 -0.017 -0.345 -0.032 -0.039 -0.904 -0.028 -0.031 -1.067
DISABLE -0.015 -0.013 -0.3 -0.058 -0.035 -0.686 -0.059 -0.032 -0.78
WEEKEND 0.022 0.027 0.607 0.011 0.014 0.326 0.013 0.014 0.52
WEATHER (WARM=1) 0.005 0.008 0.199 0.01 0.016 0.414 0.001 0.001 0.047
CHILD_OWN -0.036 -0.051 -1.268 0.029 0.008 0.227 0.046 0.057 2.02 **
TIME_CONSTRAINT 0 -0.022 -0.473 0 0.04 0.868 0 -0.043 -1.464
NUM_CAR 0.007 0.028 0.793 0.004 0.012 0.306 0.01 0.029 1.084
ACTIVE TRAVEL -0.002 -0.104 -1.188 -0.003 -0.149 -2.134 ** -0.003 -0.113 -2.086 **
SPORT_FACILITY 0.026 0.02 0.486 -0.002 -0.002 -0.052 0.073 0.056 2.476 **
SAFTY_CONCERN -0.065 -0.067 -1.367 0.043 0.037 1.065 -0.006 -0.005 -0.177
FREEWAY_100M -0.023 -0.026 -0.649 -0.008 -0.008 -0.188 0 0 0
TRAFFIC ACC -0.004 -0.059 -1.169 -0.002 -0.022 -0.388 0.001 0.01 0.387
NT_LEISURE HIGH 0.013 0.036 0.791 -0.007 -0.021 -0.5 -0.002 -0.006 -0.201
NT_LEISURE LOW 0.011 0.032 0.8 -0.008 -0.023 -0.577 -0.007 -0.017 -0.633
ATTITUDE 0.072 0.099 2.302 ** 0.046 0.07 1.805 * 0.001 0.001 0.056
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.023 -0.104 -0.124 -3.338 ***
ATTITUDE 0.076 0.037 0.759 -0.018 -0.009 -0.18 0.048 0.024 0.703
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.061 -1.228 -0.011 -0.01 -0.214 0.067 0.08 2.402 **
ATTITUDE -0.069 -0.034 -0.743 -0.009 -0.004 -0.098 0.003 0.001 0.045
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0 0.001 0.315 0 0 0.127 -0.001 -0.002 -0.39
ATTITUDE 0.002 0.002 0.591 0 0 0.044 0 0 -0.12
SAMPLE SIZE 885 813 1788
R-SQUARE 0.045 0.049 0.029
Chi-sq/df 7.185 4.38362 11.2363
CFI (=1.0) 0.873 0.825 0.704
TLI (=1.0) 0.781 0.755 0.634
RMSEA (<.08) 0.078 0.064 0.085
Low Income High income
NeighborhoodType_LEISURE HIGH
Exercise/Sport only
Medium income
Exercise/Sport only
NeighborhoodType_LEISURE LOW
Variables (Frequency)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
153

Table 6-13 Results of outdoor exercise/sport (including park activity) frequency by income
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.167 0.42 2.083 ** 0.016 0.036 0.067 -0.013 -0.03 -0.182
AGE -0.001 -0.025 -0.67 0.002 0.05 1.228 0.001 0.032 1.177
GNDR (MALE=1) 0.005 0.006 0.146 0.059 0.064 1.609 0.027 0.031 1.206
WHITE 0.028 0.033 0.73 0.111 0.099 0.541 0.009 0.01 0.253
BLACK 0.097 0.068 1.865 * 0.019 0.01 0.085 0.048 0.026 0.962
ASIAN 0.024 0.015 0.345 0.052 0.032 0.239 -0.009 -0.007 -0.186
OTHERS 0.083 0.067 1.962 * 0.05 0.021 0.216 0.032 0.018 0.574
EDC_LOW 0.017 0.019 0.432 -0.201 -0.034 -0.129 0.062 0.027 1.028
EDC_HIGH -0.016 -0.013 -0.318 0.035 0.037 0.901 -0.011 -0.013 -0.466
HHINC_LOW
HHINC_HIGH
EMPLY -0.013 -0.016 -0.331 -0.011 -0.01 -0.227 -0.015 -0.015 -0.472
DISABLE -0.072 -0.056 -1.042 -0.062 -0.027 -0.631 0.024 0.012 0.496
WEEKEND 0.032 0.034 0.916 0.015 0.014 0.344 0.017 0.018 0.63
WEATHER (WARM=1) 0.053 0.067 1.757 * 0.01 0.011 0.275 0.014 0.017 0.637
CHILD_OWN -0.017 -0.021 -0.547 -0.009 -0.002 -0.042 0.054 0.063 2.545 **
TIME_CONSTRAINT 0 -0.061 -1.387 0 0.039 0.795 0 -0.058 -1.833 *
NUM_CAR 0.008 0.029 0.798 -0.02 -0.04 -0.937 0.026 0.073 2.747 ***
ACTIVE TRAVEL -0.002 -0.093 -1.499 -0.004 -0.152 -2.516 ** -0.003 -0.108 -3.139 ***
SPORT_FACILITY 0.076 0.052 1.408 0.06 0.045 1.159 0.077 0.055 2.506 **
SAFTY_CONCERN 0.004 0.004 0.096 -0.113 -0.071 -2.175 ** 0.018 0.013 0.465
FREEWAY_100M 0.012 0.013 0.326 -0.037 -0.027 -0.751 -0.014 -0.01 -0.394
TRAFFIC ACC -0.007 -0.106 -1.787 * -0.009 -0.075 -1.064 0.002 0.016 0.666
NT_LEISURE HIGH 0.028 0.07 1.85 * 0.007 0.015 0.35 0.009 0.021 0.735
NT_LEISURE LOW -0.026 -0.066 -1.709 * -0.041 -0.09 -2.344 ** -0.006 -0.013 -0.485
ATTITUDE 0.08 0.098 2.331 ** 0.007 0.008 0.201 0.005 0.006 0.217
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.023 -0.104 -0.124 -3.338 ***
ATTITUDE 0.076 0.037 0.759 -0.018 -0.009 -0.18 0.048 0.024 0.703
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.061 -1.228 -0.011 -0.01 -0.214 0.067 0.08 2.402 **
ATTITUDE -0.069 -0.034 -0.743 -0.009 -0.004 -0.098 0.003 0.001 0.045
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.001 -0.002 -0.353 0.001 0.001 0.193 -0.001 -0.004 -0.669
ATTITUDE 0.004 0.005 0.746 0.001 0.001 0.057 0 0 0.393
SAMPLE SIZE 885 813 1788
R-SQUARE 0.065 0.061 0.029
Chi-sq/df 6.7562 4.06771 10.0824
CFI (=1.0) 0.886 0.858 0.761
TLI (=1.0) 0.878 0.827 0.738
RMSEA (<.08) 0.75 0.062 0.08
NeighborhoodType_LEISURE LOW
NeighborhoodType_LEISURE HIGH
Exercise/Sport including Park Activity
Exercise/Sport including Park Activity
Low Income High income Medium income
Variables (Frequency)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
154

Table 6-14 Results of outdoor exercise/park/active leisure travel frequency by income
Coef. β t p Coef. β t p Coef. β t p
Direct effect
Intercepts 0.347 0.531 2.712 *** 0.04 0.058 0.144 0.112 0.143 0.904
AGE 0 0.009 0.249 0.005 0.066 1.47 0.002 0.034 1.346
GNDR (MALE=1) 0.052 0.039 1.086 -0.024 -0.017 -0.418 0.003 0.002 0.082
WHITE 0.016 0.011 0.258 -0.036 -0.021 -0.199 0.013 0.008 0.183
BLACK 0.107 0.045 1.273 -0.071 -0.025 -0.36 0.073 0.022 0.747
ASIAN 0.164 0.061 1.717 * -0.155 -0.063 -0.732 -0.008 -0.004 -0.092
OTHERS -0.004 -0.002 -0.049 -0.02 -0.005 -0.09 -0.024 -0.008 -0.227
EDC_LOW 0.032 0.022 0.544 -0.295 -0.033 -0.14 0.167 0.039 1.938 *
EDC_HIGH -0.067 -0.033 -0.805 0.105 0.073 1.829 * -0.008 -0.005 -0.19
HHINC_LOW
HHINC_HIGH
EMPLY -0.015 -0.011 -0.234 0.069 0.039 0.845 -0.012 -0.007 -0.223
DISABLE -0.08 -0.038 -0.942 -0.009 -0.003 -0.07 -0.06 -0.016 -0.442
WEEKEND -0.002 -0.001 -0.034 -0.01 -0.006 -0.144 -0.003 -0.002 -0.07
WEATHER (WARM=1) 0.021 0.016 0.437 0.033 0.024 0.629 -0.002 -0.001 -0.057
CHILD_OWN -0.044 -0.034 -0.838 -0.041 -0.005 -0.143 0.094 0.06 2.183 **
TIME_CONSTRAINT 0 0.006 0.123 0 0.02 0.445 0 -0.027 -0.971
NUM_CAR -0.005 -0.011 -0.268 0.026 0.034 0.909 0.026 0.04 1.522
ACTIVE TRAVEL -0.006 -0.152 -2.202 ** -0.007 -0.179 -2.856 *** -0.007 -0.146 -3.28 ***
SPORT_FACILITY 0.108 0.045 1.303 0.025 0.012 0.287 0.188 0.073 3.506 ***
SAFTY_CONCERN -0.115 -0.065 -1.391 0.117 0.048 1.345 0 0 0.008
FREEWAY_100M -0.024 -0.015 -0.375 -0.003 -0.002 -0.038 -0.026 -0.01 -0.384
TRAFFIC ACC -0.006 -0.049 -1.128 0 -0.001 -0.026 0.002 0.008 0.321
NT_LEISURE HIGH -0.047 -0.073 -1.486 0.04 0.057 1.388 -0.007 -0.009 -0.314
NT_LEISURE LOW 0.032 0.05 1.203 -0.039 -0.056 -1.513 -0.007 -0.009 -0.331
ATTITUDE 0.083 0.062 1.552 0.108 0.078 2.021 ** 0.002 0.002 0.06
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.065 -0.094 -1.536 0.001 0.001 0.023 -0.104 -0.124 -3.338 ***
ATTITUDE 0.076 0.037 0.759 -0.018 -0.009 -0.18 0.048 0.024 0.703
HHINC_LOW
HHINC_HIGH
NUM_CAR -0.043 -0.062 -1.229 -0.011 -0.01 -0.214 0.067 0.08 2.402 **
ATTITUDE -0.069 -0.034 -0.743 -0.009 -0.004 -0.098 0.003 0.001 0.045
Indirect effect
HHINC_LOW
HHINC_HIGH
NUM_CAR 0.002 0.004 0.532 0 0 0.123 0 0 0.077
ATTITUDE -0.006 -0.004 -0.719 0 0 -0.053 0 0 -0.327
SAMPLE SIZE 885 813 1788
R-SQUARE 0.055 0.062 0.036
Chi-sq/df 7.19089 4.38362 11.2363
CFI (=1.0) 0.898 0.839 0.71388
TLI (=1.0) 0.847 0.827 0.6289
RMSEA (<.08) 0.78 0.0625 0.08
NeighborhoodType_LEISURE HIGH
NeighborhoodType_LEISURE LOW
Exercise/Sport/Park/Leisure Active Travel
Exercise/Sport/Park/Leisure Active Travel
Medium income Low Income High income
Variables (Frequency)

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1
155

As displayed in Table 6-15, by and large, the participation in outdoor leisure activity
can be partially explained by individual characteristics (in particular, employment, time-
constraint, daily active travel, and attitudinal factor) and neighborhood environments.
However, it was observed that different types of outdoor leisure activity lead to different
functions. A strong predictor varies across specific types of outdoor leisure activity,
presenting inconsistent/inconclusive results. For instance, a narrow concept of outdoor
leisure (i.e., outdoor exercise/sport only) was profoundly influenced by personality on
outdoor exercise/sport, whereas a broad concept of outdoor leisure activity including park
Table 6-15 Key predictors of outdoor leisure (pooled samples)
Direct. Sign. Direct. Sign. Direct. Sign. Direct. Sign.
Individual/household level
Age (+) * (+) **
Gender (+) *
Education_low
Employment (--) ** (--) * (--) *
Income_high (+) **
Physical disability (Yes=1)
Weather (Warm=1) (+) *
Children (Yes=1) (+) *
Time constraint (--) * (--) **
Car availability (#. car) (+) * (+) *
Active travel time (--) ** (--) ** (--) **
Intrapersonal (Yes=1) (+) ** (+) **
Neighborhood level
Sport/exercise facility (Yes=1) (+) ** (+) ** (+) ** (+) **
Safety concern (Yes=1)
Traffic incident (#. pedestrian) (--) **
Neighborhood type
NT_Leisure_high
NT_Leisure_low (--) **
Indirect effect
Mediating role of NT
Variables
Duration of outdoor leisure Frequency of outdoor leisure
Exercise/sport (+) Park activity Exercise/sport (+) Park activity

Note: ‘(+)’ positive impact; ‘(--)’ negative impact; ‘**’ p<0.05; ‘*’ p<0.1;
156

activity and leisure active travel was acutely affected by the amount of time spent in active
travel. Those findings provide supportive evidence on a clear link between individual
personality and outdoor leisure activity, as well as on a substitute relationship between active
travel and outdoor leisure activity.
Unlike inconsistent/inconclusive results found in individual characteristics, the
availability of sport facility revealed a consistently significant and positive influence on all
types of outdoor leisure activity. Highlighting a supportive role of neighborhood
environment in outdoor leisure activity, this finding suggests that the provision of
sport/recreation facilities within a neighborhood can contribute to increasing the time people
spend in outdoor leisure activity, all else being equal.
The analytic models (with pool samples) predicting the frequency of outdoor leisure
activity indicated fairly consistent results with the duration models, but some variables
produced slightly different results. More specifically, the availability of sport facility in
outdoor consistently reported a significant positive influence on the frequency of all types of
outdoor leisure activity. The persistent results were also found in individual personality. Not
only the duration but also the frequency of outdoor exercise/sport was positively associated
with personality on outdoor exercise/sport at the 5% significance level. This finding suggests
that attitudinal factor can play a crucial role in behavioral choice for outdoor leisure activity
among individual/household characteristics.
Meanwhile, car availability (i.e., the number of vehicles in household) revealed a
consistently significant and positive influence on the frequency of all types of outdoor
leisure activity at the 10% significance level, but it is notable that those associations were
found insignificant in analytic models for the duration of outdoor leisure. This finding
157

suggests that household travel options improve individual mobility to reach leisure
destinations beyond the neighborhoods, thereby increasing the likelihood people participate
in outdoor leisure.
Individual time constraint (e.g., time spent at workplace as well as on the road) was
also considered as a significant predictor of the time spent in outdoor exercise/sport
including park activity, but it was not significantly associated with the frequency. Daily total
active travel time was observed as a key variable with a significantly profound effect on the
frequency of all types of outdoor leisure, whereas such profound effect was not found in the
duration of a narrow concept of outdoor leisure activity (i.e., exercise/sport only).
Notably, the exposure of traffic accident revealed the negative influence on the
frequency of outdoor exercise/sport including park activity at the 5% significance level, but
such significant impact was not found in the duration of outdoor leisure activity. This finding
provides scientific evidence on deterrent roles of neighborhood environment in behavioral
decisions on outdoor leisure activity, informing us that the likelihood people engage in
outdoor exercise/sport including park activity decreases when they live in a neighborhood
exposed to traffic accidents.
As summarized in Table 6-16, the separated models by different income levels
indicated slightly different results from a pool model, demonstrating that there is a clear gap
between low- and high income groups in explaining the variation in duration/frequency of
outdoor leisure activity. For example, low income groups reported a significant positive
association between neighborhood type and time spent in outdoor exercise/sport at the 5%
significance level, whereas high income groups did not find any significant relationship
between neighborhood type and outdoor exercise/sport.
158

This finding might reflect mobility issue that low-income groups encounter.
Compared with high-income groups, in general, low-income groups have the limited travel
options to reach leisure destinations beyond their neighborhoods. Given this issue, leisure-
friendly neighborhoods can allow low-income groups to participate in outdoor leisure
activity by providing leisure destinations easily accessible.
Furthermore, high-income groups reported that active travel time was significantly
related to the frequency of outdoor exercise/sport, whereas no significant impact of active
travel time was found in low-income groups. Low-income groups consistently found a
significant positive role of attitudinal factor in frequency of outdoor exercise/sport including
park activity, whereas such influence was not significant in high-income groups. Regarding
neighborhood attributes, high-income groups are more sensitive to safety concern when they
engage in outdoor exercise/sport including park activity, whereas low-income groups are
more responsive to exposure to traffic incidents objectively measured when they engage in
such outdoor leisure activities.
Regression models reported the positive (or negative) association between high (or
low) leisure-friendly neighborhoods and outdoor leisure duration/frequency, but statistical
significance of such connections varies across sample sizes (i.e., pool samples or sub-
samples) as well as specific types (i.e., a narrow or broad) of outdoor leisure activity.
Furthermore, a subsequent model exploring the mediating role of neighborhood type
indicated that there was little or no indirect impact of individual/household backgrounds on
outdoor leisure activity via neighborhood type. That is, path model did not find any
significant mediating role of neighborhood type that represents the effect of residential self-
selection.
159

Table 6-16 Key predictors of outdoor leisure (income groups)
Exercise (+)Park (+)Travel Exercise (+)Park (+)Travel Exercise (+)Park (+)Travel Exercise (+)Park (+)Travel
Car availability (#. car) (+)
Time constraint (--)
Active travel time (--) ** ** ** **
Intrapersonal (Yes=1) (+) ** ** * **
Sport/exercise facility (Yes=1) (+) *
Safety concern (Yes=1) (--) **
Traffic incident (#. pedestrian) (--) *
Freeway(100M) (--) *
NT_Leisure_high (+) ** * *
NT_Leisure_low (--) * * * **
Mediating role of NT
Direction
Duration Frequency
Low income High income Low income High income Key variables

Note: ‘(+)’ positive impact; ‘(--)’ negative impact; ‘**’ p<0.05; ‘*’ p<0.1;
‘(+) Park’ refers to dependent variable measured by adding up outdoor exercise and park activity
‘(+) Travel’ refers to dependent variable measured by adding up outdoor exercise, park activity, and leisure active travel

160
CHAPTER 7.
CONCLUSIONS AND DISCUSSION

7.1. MAJOR FINDINGS AND CONTRIBUTIONS

From the results from Chapter 5 and Chapter 6, it is found that no single factor determines
behavioral decision on daily active travel/outdoor leisure activity. Rather, multiple factors
conceptualized by the opportunities or/and constraints at the multiple dimensions play
independent roles. More specifically, the evidence shows a link between environmental
opportunities and active living, confirming that the provision of walking-friendly urban form
features and local resources contributes to an increase in both transport- and outdoor leisure-
related physical activity of general populations. This finding supports the claims made by
active-living campaigns.
Along with the facilitating role of neighborhoods, the evidence also indicates a
significant association between environmental barriers and active living, suggesting that
people exposed to unsafe sources within a neighborhood might not participate in active
living even though they live in a neighborhood conducive to such activities. Even though it
is difficult to clarify the relative importance of neighborhood opportunities and barriers due
to mixed results, the finding on deterrent role tells us that mitigating barriers is another
practical way of governing active living.
However, the relationship remains complex and inconclusive. The significance of
association varies across the measurements (i.e., continuous/binary scale), specific types of

161
activity (e.g., active travel/walking only; exercise/sports or park activity), and different
populations (i.e., sub-populations sorted by income level). Despite the mixed results, several
explanatory variables reveal a consistently significant correlation with active travel/outdoor
leisure activity.
Among those significant predictors, household travel option (i.e., car availability) is
the most crucial factor that negatively affects observed patterns in daily active travel. The
amount of active travel time is also the most crucial factor that negatively affects observed
daily outdoor leisure activity. The magnitude of those crucial factors is greater than that of
neighborhood environmental factors. In other words, even though some of environmental
opportunities/barriers are consistently significant, the decision on active behavior is about
more than neighborhood environment. This point challenges the claims made by active-
living campaigns.
More importantly, there is a clear gap in the correlate of behavioral outcomes
between different income groups. This reflects different levels of travel option and
environmental exposure among subpopulations: 99% of respondents who fall into high-
income groups have one or more private vehicles; low-income groups normally live in areas
with high density, compared with high-income groups. Both limited travel options and
environmental opportunities (e.g., high activity density) explain why low-income groups
generally engage more in active travel than high-income groups.
Given that low-income groups have limited leisure options in their neighborhoods,
public open space and local streets would be a common behavioral setting for leisure
physical activity in the outdoors. The prevalence of environmental barriers (e.g., traffic
accidents) provides crucial clues to the low levels of outdoor leisure (including leisure active

162
travel). On the other hand, it appears reasonable to assume that high-income groups are less
responsive to both environmental opportunities and barriers not only because they have
travel options that reduce the demand for active travel and reach leisure destinations beyond
their neighborhoods, but also because they have leisure options/resources that allow for other
types of leisure physical activity, particularly in indoor venues.
Furthermore, the evidence of a trade-off between active travel and outdoor leisure
activity deserves to be highlighted. Given the small time window in a day, people who
engage more in active travel are less willing to allocate their spare time to outdoor leisure
activity since they consider active travel (both utilitarian and recreational) as another type of
outdoor leisure activity. As often emphasized, active travel per se is an important source of
physical activity in a daily routine, but it can yield a decrease in physical activity in other
behavioral settings (e.g., outdoor leisure in this research).
This finding delivers the message: enhancing active travel doesn’t necessarily
translate into an increase in overall physical activity. It remains to be determined how much
active travel contributes to promoting physical activity more generally. Moreover, even
though this study found several explanatory variables significantly affecting active
travel/outdoor leisure activity, the ability of those factors to explain the variation in observed
patterns is extremely low. There might be other crucial factors not included here but
substantially shaping behavioral decisions. However, it is still uncertain about what exactly
it is.
In terms of research design, several contributions of this study can be summarized as
follows. Most importantly, this study has objectively measured various types of
neighborhood barriers at a highly disaggregate level, utilizing the specific locations (x-y

163
coordinates) of households, crimes, and traffic incidents. Previous studies have often
employed a loose or coarse geographic unit in delineating neighborhood boundaries as well
as incorporated the aggregate data in measuring the prevalence of crime/traffic incidents,
based on a predefined geographic unit (e.g., administrative district or census tract) which
does not match well to the individual units of actual neighborhoods.
Second, this study has highlighted a substantial gap in neighborhood barriers
between low- and high-income groups, especially informing the deterrent role of
neighborhood environment in daily active travel/outdoor leisure activity of low-income
groups living in urban neighborhoods where environmental barriers are more concentrated.
Previous studies have often investigated the connection between the built environment
(mainly, physical urban form features) and physical activity, focusing on high/medium-
income groups living in the suburban context.
Third, to better understand how to govern individual daily active living in highly
localized contexts, this study has investigated two different venues for active living: active
travel and outdoor leisure activity. To properly capture the patterns of travel/activity, this
study employed private versions of national travel surveys with a fairly large set of samples
and an objective measurement. Previous studies have often incorporated either travel data at
the national/ regional level or other secondary data based on a retrospective, self-reported
measurement of walking or outdoor leisure activity.
Finally, this study has developed an advanced analytic method to adequately address
the potential residential self-selection in the observed relationship between neighborhood
environment and active travel/outdoor leisure activity, given a daily travel/activity schedule.
While isolating the impact of neighborhood environment from that of residential sorting, this

164
study has assessed the relative importance or unique contributions of environmental
opportunities/barriers among various individual/household backgrounds, including car
availability, personality (e.g., attitudinal factor), and daily time allocation.

7.2. POLICY IMPLICATIONS

Based on major findings demonstrated before, several policy implications for active
travel/outdoor leisure activity can be suggested as follows. First of all, land-use policy might
be able to achieve the goal of promoting active travel and/or outdoor leisure activity by
designing neighborhoods with favorable physical layouts and providing supportive
infrastructure in neighborhoods (e.g., sidewalks, bike paths, and park/sport/recreation
facilities). However, even though a facilitating role was found for the built environment, this
study delivers the message to practitioners that such policy tools might not ensure substantial
increases in active travel/outdoor leisure or behavioral change into active lifestyles when
people are exposed to neighborhood barriers. Since those barriers can affect individual
discretionary choices, additional policy should also be devoted to reducing neighborhood
safety concerns such as violent crime or pedestrian traffic accidents, in tandem with land-use
policy that attempts to enhance environmental opportunities.
Moreover, neighborhood barriers are not evenly distributed but heavily concentrated
in certain areas (i.e., hot spots) where low-income groups predominantly live. As the finding
from this study shows, the deterrent impact of neighborhood barriers is more pronounced
among low-income groups. In this vein, tailoring the intervention for reducing neighborhood
barriers to low-income groups can contribute to realizing the deferred/latent demand for
active travel/outdoor leisure activity.

165
Another point is that both neighborhood opportunities and barriers are generally
mixed in the same neighborhood. For instance, urban neighborhoods are more walkable but
also more crime-prone/traffic hazard-ridden. While it is often observed that physical layout
for active travel (especially, walking trips) has been well established in urban neighborhoods,
the prevalence of risk factors still remains higher. Given this context, the overall contribution
of neighborhood environment to total active travel might be a zero-sum game, at best, or at
worst the positive influence of favorable urban form features may be countervailed by the
negative influence of neighborhood violent crimes/traffic accidents. The finding from this
study supports such possibilities
23
. Thus, the effort to enhance neighborhood safety by
reducing risk factors appears to be effective in promoting active travel among urban
residents, while ensuring the supportive role of favorable urban forms.
As a practical way of mitigating such barriers, there is a certain appeal to both
“crime prevention by environmental design (CPTED)” and traffic engineering tools designed
to reduce pedestrian-vehicle crashes. For example, various physical characteristics including
broken windows, vandalism, litter, graffiti, and abandoned buildings can affect
crime/disorder in neighborhood. Fear arising from those sources can be mitigated by creating
a more defensible space that improves street surveillance. Even though this study does not
directly address the impact of urban design strategy (e.g., shared open space, windows facing
the street, lighting design, and other elements of self-policing neighborhoods) on street
surveillance, such an environmental design approach can contribute to increasing pedestrian
activity in neighborhoods (Sallis et al., 1998).

23
A potential trade-off between the advantage of favorable urban form features and the
disadvantage of neighborhood barriers is verified in this study. More specifically, the relative
magnitude of impact of neighborhood barrier (e.g., violent crime) on the duration of daily active
travel is higher than that of neighborhood opportunities (e.g., activity density and street design).

166
Traffic-calming is also a popular form of injury prevention that aims to improve
street-level pedestrian safety by modifying the built environment. Acknowledging that a
vehicle crash involving pedestrian injuries is mainly a function of traffic speed, traffic-
calming is primarily focused on speed control, particularly in behavioral settings (e.g.,
residential areas, school zones, and park/playgrounds) where many street activities are
embedded. Even though empirical results are mixed, traffic-calming devices (e.g., speed
humps, traffic circles, and curb extensions) can be considered as effective traffic engineering
measures that prevent traffic-related injuries and hence enhance traffic safety (Ewing and
Dumbaugh, 2009; Jacobsen et al., 2009; Bunn et al., 2003). Along with speed management,
other traffic engineering schemes such as the separation of pedestrians from vehicles and
increase of the visibility/conspicuity of pedestrians might make pedestrian activities in
neighborhoods/local streets more appealing (Retting et al., 2003).
Next, current efforts to promoting active travel by modifying neighborhood
environments (both maximizing environmental opportunities and minimizing environmental
barriers) are not likely to be convincing in the absence of transportation intervention that
attempts to directly control private vehicle use. As the theory of travel behavior (i.e., utility
maximization) explains, people finally choose one mode of travel based on cost-benefit
calculation among many options. It appears that people are more likely to choose private
vehicles rather than active travel when motorized travel provides great benefits at a low cost.
In the U.S. context, where the cost of motorized travel (both time and monetary) is
relatively low and the rate of mobility by private vehicle is fairly high, the high access to
such modes can be a substantial barrier to active travel. The finding from this study supports

167
this expectation, indicating that the relative magnitude of positive impact of neighborhood
environment is smaller than that of negative impact of car availability.
Given the attractiveness of competing modes, people are not likely to engage in
active travel, even though they live in neighborhoods with favorable urban form features. In
this sense, several restrictions that make vehicle use more difficult/inconvenient can be
effective in promoting active travel. The question here is whether reducing private vehicle
use directly leads to increasing active travel. It appears that active travel has a limited ability
to substitute for motorized travel due to distance sensitivity, but transportation policy
devoted to more pricing on gasoline and/or parking will complement land-use policy
mentioned earlier.
As demonstrated in previous chapters, the likelihood that people will engage in
outdoor leisure activity may increase when they live in neighborhoods with local amenities
for recreation (e.g., sport/recreation facility). Given that low-income groups who have
limited travel options are likely to live in neighborhood with limited leisure options,
providing leisure destinations within walking distance will encourage them to engage in
outdoor leisure activity. However, it might be a challenge to create new leisure amenities by
demolishing existing buildings. Thus, enhancing residents’ mobility to reach leisure
destinations beyond their neighborhoods is another practical way of improving access to
leisure amenities.
Potentially, allocating leisure destinations around public transit stations can promote
outdoor leisure participation among low-income groups. However, such efforts might be
less effective if a user/membership fee is charged for using leisure amenities. In this case,
either the presence of or spatial proximity to a recreational facility does not necessarily

168
translate into easy access since poor people cannot afford it. Thus, it is essential to provide
leisure amenities without any monetary cost.
Furthermore, the number of nearby leisure destinations might not affect how often
people visit or how long people stay there. If many recreational facilities are located within a
walkable distance but they are poorly maintained or less attractive, people rarely use them.
Park characteristics (e.g., park management, maintenance, design, safety, cleanliness, and
aesthetics) can be a crucial factor that ensures substantial increases in outdoor leisure activity,
far more than the accessibility and proximity of park areas (Godbey et al., 2005; Boslaugh et
al., 2004). Thus, more tailored policies to improve the quality of recreational resources will
be required.
It should be noted that, to effectively promote active behavior/life styles, policy
should be devoted to injury prevention that intends to reduce neighborhood barriers, in
tandem with supportive neighborhood design. Moreover, active-living campaigns will
achieve better performance in combination with transportation planning (e.g., more pricing
on vehicle use and improving mobility of low-income groups). However, despite such
combined policy endeavors, increasing overall physical activity is a much bigger challenge if
the total budget of physical activity is fixed or controlled by other factors beyond the policy
and environmental strategies.

7.3. LIMITATION AND FUTURE STUDY

While the findings from this study are relevant and produce several policy implications, the
limitations created by data/measurement issues should also be noted to improve our

169
understanding of behavioral decisions on active travel/outdoor leisure activity and of their
relationship to neighborhood environment.
First of all, even though travel survey data offer the actual daily patterns of
travel/activity, several types of active travel/outdoor leisure activity might be still ignored in
this data. Moreover, a one-day travel diary might be inadequate to properly capture weekly-
basis behavioral patterns that are regular and routine. The underreported/unobserved patterns
of active travel/outdoor leisure activity can underestimate the role of some explanatory
variables and hence yield misunderstandings. A more comprehensive travel survey method
that not only allows for measuring the underreported travel/activity but also reports multi-
day travel diaries can overcome this limitation.
Second, this study measured several environmental attributes, based on a ¼ mile
buffer area delineating the spatial boundaries of a neighborhood. However, it is possible that
people often walk more than ¼ mile and engage in outdoor leisure activity beyond that
distance. Given that there is no agreement on a precise theoretical scale of neighborhood, it
is necessary to operationalize various geographic scales (e.g., more or less than ¼ mile) and
see whether different scales produce consistent results. Furthermore, using the quartile
measurement of physical dimensions for walkability and park availability, this study
identified neighborhood types which support active travel and outdoor leisure activity,
respectively. However, the advanced measurements (e.g., cluster/factor analysis) that deal
with multiple/latent factors can improve the ability to sort out neighborhood type in a more
comprehensive manner.
Third, this study employed two personality variables regarding attitudinal factor
toward active travel and outdoor leisure, relying on simple, indirect survey questions (i.e.,

170
single item-measures). This approach might be imperfect in capturing the complex personal
psychological dimensions. Given the limited measurement of personal attitude, both direct -
and indirect impacts of attitude on behavioral decisions might be over- or underestimated.
This point provides some clues to the insignificant impact of residential self-selection. In
addition, the measurement of outcome variables should also be notified. This study
measured two outcomes of active travel/outdoor leisure activity: its duration (i.e., how long
do people take active travel or stay at leisure destinations in the outdoors?) and frequency
(i.e., how often/many times do people engage in such travel/activity?) However, technically,
those variables do not directly represent the level of intensity. It is perhaps more reasonable
to measure the intensity of such activities when discussing their substantial impact on health
(i.e., physical activity).
Another limitation can be explained by methodological challenges. Even though
path analysis is a unique analytic method that addresses potential endogeneity problems and
offers some hints for causal process, the results from path analysis do not ensure a true
meaning of causal inference. More definitive evidence on causal inference can be found in a
more advanced research framework devoted to experimental design which involves random
assignment and comparison of treatment/control groups. Given the imperfect process of
operationalizing a true experimental design, a longitudinal approach using panel data can
provide a feasible way of improving our understanding on causal inference. Implementing a
pre/post-test that accounts for time precedence might be one solution to this challenge.
Furthermore, the hypothetical model developed in this study was simplified to some degree,
while assuming no feedback loops between variables. However, there is still the dynamic
nature of the potential mutual relationship between attitude and the built environment. To

171
address this limitation, future studies must identify the change in attitude by measuring the
attitude before/after people chose their current residential locations.
Finally, analytic models in this study explain a small portion of the variation in daily
active travel/outdoor leisure activity. One potential reason for this result is that a large
portion of adult samples employed in analytic models did not report any active
travel/outdoor leisure activity on a survey day. The insufficient variation in outcome
variables attributed to the underreported/unobserved daily travel/activity might decrease the
explanatory power of analytic models. The other reason is that analytic models might
exclude some crucial factors that explain the variation in active travel/outdoor leisure
activity. One factor, crucial but omitted in this analysis, would be variables derived from
psychological process. It is expected that the approach deeply devoted to psychological
mechanisms through which human behavioral decisions are finally made can improve the
explanatory power. Further study is needed to develop the conceptual/empirical framework
based on planned behavior theory, using a well-defined measurement of attitudinal factors
and the perception of environmental opportunities/barriers.
There are also several points not addressed in this study that deserve further scrutiny.
This study focused on adult samples and excluded other age groups since different age
groups have different functions of active travel/outdoor leisure activity. Given the
importance of physically active behavioral habits in youth and the limited travel/activity of
seniors, it is valuable to identify the factors affecting active travel/outdoor leisure activity
among those groups and explore how and to what extent neighborhood opportunities/barriers
play a role in the behavioral decision process of both groups. Additional research that
focuses on specific subgroups (e.g., different ethnic/gender groups) with different

172
travel/activity/time-use patterns will contribute to obtaining more generalized conclusion on
environmental correlates.
The complex relationship between perceived neighborhood safety and neighborhood
barriers, objectively measured, is also worthy of revisiting. Previous studies revealed
inconclusive findings on whether perceived dimensions of neighborhood environment have a
more pronounced influence than objective dimensions, yielding incongruity between
subjective- and objective measurement in terms of significance and magnitude. Theoretically,
perceived safety concerns can be shaped by the various sources of neighborhood barriers. In
other words, both subjective- and objective neighborhood characteristics are potentially
related to each other rather than playing an independent role.
Moreover, there might be a relationship between different neighborhood
environments objectively measured. Physical urban form features (e.g., high density, mixed
land use, and street connectivity) can affect neighborhood barriers (e.g., crime, traffic
accidents, and air pollution). Thus, future studies will need to explicitly define the
relationship between neighborhood environmental factors in predicting daily travel/activity,
while properly addressing the mediating role and confounding effects.

173
REFERENCES

ALD(ActiveLivingbyDesign) (2010). "Low Income Populations and Physical Activity: An
overview of issues related to active living." www.activelivingbydesign.org
Agrawal, A. W. and P. Schimek (2007). "Extent and correlates of walking in the USA."
Transportation Research Part D: Transport and Environment 12(8): 548-563.
Ajzen, I. (1991). The Theory of Planned Behavior. Organizational Behavior and Human
Decision Processes, 50:179–211.
Amemiya, T. (1973). "Regression analysis when the dependent variable is truncated
normal." Econometrica: Journal of the Econometric Society: 997-1016.
Andreasen, A. (1995). Marketing social change: Changing behavior to promote health, social
development, and the environment, Jossey-Bass Inc Pub.
Arnold, M. L. and K. J. Shinew (1998). "The role of gender, race, and income on park use
constraints." Journal of Park and Recreation Administration 16(4): 39-56.
Babey, S. H., T. A. Hastert, et al. (2008). "Physical Activity Among Adolescents:: When Do
Parks Matter?" American journal of preventive medicine 34(4): 345-348.
Badland, H. and G. Schofield (2005). "Transport, urban design, and physical activity: an
evidence-based update." Transportation Research Part D: Transport and
Environment 10(3): 177-196.
Bagley, M. N. and P. L. Mokhtarian (2002). "The impact of residential neighborhood type
on travel behavior: A structural equations modeling approach." The Annals of
Regional Science 36(2): 279-297.
Ball, K., A. Timperio, et al. (2007). "Personal, social and environmental determinants of
educational inequalities in walking: a multilevel study." Journal of Epidemiology
and Community Health 61(2): 108-114.
Bandura, A. (1969). Social-learning theory of identificatory processes. Handbook of
socialization theory and research. Chicago: Rand McNally 213: 262.
Bandura, A. (1991). Social cognitive theory of moral thought and action. Handbook of moral
behavior and development 1: 45-103.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Barnes, P. (2007). "Physical activity among adults: United States, 2000 and 2005."
Hyattsville, MD: US Department of Health and Human Services, CDC.
Bauman, A. E., R. S. Reis, et al. (2012). "Correlates of physical activity: why are some
people physically active and others not?" The Lancet.
Bauman, A., F. Wallner, et al. (1996). "No ifs no buts Illawarra physical activity project:
Baseline research report." Warrong, NSW: Commonwealth Department of Health
and Family Services.
Bauman, A., J. Sallis, et al. (2002). Toward a better understanding of the influences on
physical activity: The role of determinants, correlates, causal variables, mediators,
moderators, and confounders. American journal of preventive medicine 23(2): 5-14.
Becker, M. (1974). The health belief model and personal health behavior, Slack.

174
Beimborn, E. A., M. J. Greenwald, et al. (2003). "Accessibility, connectivity, and captivity:
impacts on transit choice." Transportation Research Record: Journal of the
Transportation Research Board 1835(-1): 1-9.
Ben-Akiva, M. and S. Lerman (1985). Discrete choice analysis: theory and application to
travel demand, The MIT Press.
Berrigan, D. and R. P. Troiano (2002). "The association between urban form and physical
activity in US adults." American journal of preventive medicine 23(2): 74-79.
Bhat, C. R. and J. Y. Guo (2007). "A comprehensive analysis of built environment
characteristics on household residential choice and auto ownership levels."
Transportation Research Part B: Methodological 41(5): 506-526.
BLS. (2006). Time-Use Survey- First Results Announced by BLS and Technical Notes. U.S.
Department of Labor, Washington, D.C.
Boarnet, M. (2004). The Built Environment and Physical Activity: Empirical Methods and
Data Resources. Paper prepared for the Transportation Research Board and the
Institute of Medicine Committee on Physical Activity, Health. Transportation, and
Land Use.
Boarnet, M. (2006). About This Issue: Planning's Role in Building Healthy Cities: An
Introduction to the Special Issue. Journal of the American Planning Association
72(1): 5-9.
Boarnet, M. and R. Crane (2001). The influence of land use on travel behavior: specification
and estimation strategies, Elsevier. 35: 823-845.
Boarnet, M. G. and M. J. Greenwald (1999). "Land Use, Urban Design, and Non-Work
Travel: Reproducing for Portland, Oregon, Empirical Tests from Other Urban
Areas." Center for Activity Systems Analysis, Institute of Transportation Studies
(UCI), UC Irvine.
Boarnet, M. G., A. Forsyth, et al. (2011). "The Street Level Built Environment and Physical
Activity and Walking." Environment and Behavior 43(6): 735-775.
Boarnet, M. G., C. L. Anderson, et al. (2005). Evaluation of the California Safe Routes to
School legislation: Urban form changes and children's active transportation to school.
American Journal of Preventive Medicine 28(2): 134-140.
Bohte, W. (2010). Residential self-selection and travel: The relationship between travel-
related attitudes, built environment characteristics and travel behaviour, IOS Press.
Bohte, W., K. Maat, et al. (2009). "Measuring attitudes in research on residential self-
selection and travel behaviour: a review of theories and empirical research."
Transport Reviews 29(3): 325-357.
Bolen, J. C., L. Rhodes, et al. (2000). "State-specific prevalence of selected health behaviors,
by race and ethnicity?-Behavioral Risk Factor Surveillance System, 1997." MMWR
CDC Surveill Summ 49(2): 1-60.
Boone-Heinonen, J., P. Gordon-Larsen, et al. (2011). "Environment and physical activity
dynamics: the role of residential self-selection." Psychology of Sport and Exercise
12(1): 54-60.
Booth, M. L., N. Owen, et al. (2000). "Social-cognitive and perceived environment
influences associated with physical activity in older Australians." Preventive
medicine 31(1): 15-22.

175
Booth, S. L., J. F. Sallis, et al. (2001). "Environmental and societal factors affect food choice
and physical activity: rationale, influences, and leverage points." Nutrition reviews
59(3): S21-S36.
Boslaugh, S. E., D. A. Luke, et al. (2004). "Perceptions of neighborhood environment for
physical activity: Is it “who you are” or “ where you live?" Journal of Urban Health
81(4): 671-681.
Bowman J, & Ben-Akiva M (2000) Activity-based disaggregate travel demand model
system with activity schedules. Transportation Research A 35: 1–28.
Brown, P. R., W. J. Brown, et al. (2001). "Perceived constraints and social support for active
leisure among mothers with young children." Leisure Sciences 23(3): 131-144.
Brownson, R. C., E. A. Baker, et al. (2001). "Environmental and policy determinants of
physical activity in the United States." Journal Information 91(12).
Bruijn, G.., Kremers, S., Lensvelt-Mulders, G.., de Vries, H., van Mechelen, W. and Wrug,
J.,(2006) Modeling individual and physical environmental factors with adolescent
physical activity. American Journal of Preventive Medicine 30, 507-512.
Buehler, R., J. Pucher, et al. (2011). "Active travel in Germany and the US: contributions of
daily walking and cycling to physical activity." American journal of preventive
medicine 41(3): 241-250.
Buhi, E., Goodson, P. and Neilands, T. (2007). Structural Equation Modeling: A Primer for
Health Behavior Researchers. American Journal of Health Behavior 31, 74-85.
Bunn, F., T. Collier, et al. (2003). "Traffic calming for the prevention of road traffic injuries:
systematic review and meta-analysis." Injury Prevention 9(3): 200-204.
Burbidge, S. and K. Goulias (2009). "Active travel behavior." Transportation Letters: the
International Journal of Transportation Research(2): 147-167.
Calthorpe, P. and M. Mack (1989). "Pedestrian pockets: new strategies for suburban
growth." The Pedestrian Pocket Book-A New Suburban Design Strategy, Princeton
Architectural Press, New York: 110.
Cao, X., P. L. Mokhtarian, et al. (2007). "Do changes in neighborhood characteristics lead to
changes in travel behavior? A structural equations modeling approach."
Transportation 34(5): 535-556.
Cao, X., P. L. Mokhtarian, et al. (2009). "Examining the impacts of residential self-selection
on travel behaviour: A focus on empirical findings." Transport Reviews 29(3): 359-
395.
Caspi, C. E., I. Kawachi, et al. (2013). "The social environment and walking behavior among
low-income housing residents." Social science & medicine 80: 76-84.
Cerin, E., E. Leslie, et al. (2009). "Explaining socio-economic status differences in walking
for transport: An ecological analysis of individual, social and environmental
factors." Social science & medicine 68(6): 1013-1020.
Cervero, R. and K. Kockelman (1997). Travel demand and the 3Ds: density, diversity, and
design. Transportation Research Part D: Transport and Environment 2(3): 199-219.
Cervero, R. and M. Duncan (2003). "Walking, bicycling, and urban landscapes: evidence
from the San Francisco Bay Area." American journal of public health 93(9): 1478
Cervero, R. and R. Gorham (1995). "Commuting in transit versus automobile
neighborhoods." Journal of the American planning Association 61(2): 210-225.
Cevero, R., (1996) Mixed land-uses and commuting: evidence from the American housing
survey. Transportation Research A 30 5, pp. 361–377.

176
Chapin, F. S. and F. S. Chapin (1974). Human activity patterns in the city: things people do
in time and in space, Wiley New York.
Christensen, K. M., J. M. Holt, et al. (2010). "Effects of perceived neighborhood
characteristics and use of community facilities on physical activity of adults with
and without disabilities." Prev Chronic Dis 7(5): A105.
Christian, Thomas J. (2009). Opportunity Costs Surrounding Exercise and Dietary
Behaviors : The Effect of Commuting Time Costs on Health-Related Activities.
Social Science Research Network Working paper: 1-45
Clifton, K. J. and J. Dill (2005). Women's Travel Behavior and Land Use: Will New Styles
of Neighborhoods Lead to More Women Walking? Transportation Research Board
Conference Proceedings.
Clifton, K. J. and K. Kreamer-Fults (2007). "An examination of the environmental attributes
associated with pedestrian-vehicular crashes near public schools." Accident Analysis
& Prevention 39(4): 708-715.
Cohen, D. A., T. L. McKenzie, et al. (2007). "Contribution of public parks to physical
activity." Journal Information 97(3).
Coleman, D. and S. E. Iso-Ahola (1993). "Leisure and health: The role of social support and
self-determination." Journal of leisure research 25: 111-111.
Corless, J. and G. Ohland (1999). "Caught in the crosswalk: pedestrian safety in California."
Surface Transportation Policy Project Report (San Francisco: Surface Transportation
Policy Project, 1999)(www. transact. org/ca/caught99/caught. htm [September 26,
2005]).
Corti, B., R. Donovan, et al. (1996). "Factors influencing the use of physical activity
facilities: results from qualitative research." Health Promotion Journal of Australia
6: 16-21.
Crawford, D. W., E. L. Jackson, et al. (1991). "A hierarchical model of leisure constraints."
Leisure Sciences 13(4): 309-320.
Crawford, D., A. Timperio, et al. (2008). "Do features of public open spaces vary according
to neighbourhood socio-economic status?" Health & Place 14(4): 889-893.
Cutts, B. B., K. J. Darby, et al. (2009). "City structure, obesity, and environmental justice: an
integrated analysis of physical and social barriers to walkable streets and park
access." Social science & medicine 69(9): 1314-1322.
De Nazelle, A., D. A. Rodriguez, et al. (2009). "The built environment and health: impacts
of pedestrian-friendly designs on air pollution exposure." Science of the total
environment 407(8): 2525-2535.
De Nazelle, A., M. J. Nieuwenhuijsen, et al. (2011). "Improving health through policies that
promote active travel: a review of evidence to support integrated health impact
assessment." Environment international 37(4): 766-777.
Degenhardt, B., J. Frick, et al. (2011). "Influences of personal, social, and environmental
factors on workday use frequency of the nearby outdoor recreation areas by working
people." Leisure Sciences 33(5): 420-440.
Devlin, A. and L. Frank (2009). Physical Activity and Transportation Benefits of Walkable
Approaches to Community Design in British Columbia, Vancouver
Dishman, R. K., J. F. Sallis, et al. (1985). "The determinants of physical activity and
exercise." Public health reports 100(2): 158.

177
Doyle, S., A. Kelly-Schwartz, et al. (2006). "Active community environments and health:
the relationship of walkable and safe communities to individual health." Journal of
the American planning Association 72(1): 19-31.
Driver, B. and R. C. Knopf (1977). "Personality, outdoor recreation, and expected
consequences." Environment and Behavior 9(2): 169-193.
Dunn, A. L., B. H. Marcus, et al. (1999). "Comparison of lifestyle and structured
interventions to increase physical activity and cardiorespiratory fitness." JAMA: the
journal of the American Medical Association 281(4): 327.
Eagly, A. H. and S. Chaiken (1993). The psychology of attitudes, Harcourt Brace
Jovanovich College Publishers.
Ewing, R. and R. Cervero (2001). "Travel and the built environment: a synthesis."
Transportation Research Record: Journal of the Transportation Research Board
1780(-1): 87-114.
Ewing, R. and E. Dumbaugh (2009). "The Built Environment and Traffic Safety A Review
of Empirical Evidence." Journal of Planning Literature 23(4): 347-367.
Eyler, A. A., E. Baker, et al. (1998). "Physical activity and minority women: a qualitative
study." Health Education & Behavior 25(5): 640-652.
Ezzati, M., A. D. Lopez, et al. (2002). "Selected major risk factors and global and regional
burden of disease." The Lancet 360(9343): 1347-1360.
Federal Highway Administration. (1994). "Traffic calming, auto-restricted zones and other
traffic management techniques." The National Bicycling and Walking Study:
Transportation Choices for a Changing America, Final Report. U.S. Department of
Transportation.
Ferge, S. (1972). "Social differentiation in leisure activity choices." The Use of Time: Daily
Activities of Urban and Suburban Population in Twelve Countries, Mouton, Den
Haag: 213-227.
Finkelstein, E. A., I. C. Fiebelkorn, et al. (2004). "State-Level Estimates of Annual Medical
Expenditures Attributable to Obesity*." Obesity research 12(1): 18-24.
Fitzpatrick, K. and M. La Gory (2000). Unhealthy places: The ecology of risk in the urban
landscape, Routledge.
Flegal, K. M., M. D. Carroll, et al. (2002). "Prevalence and trends in obesity among US
adults, 1999-2000." JAMA: the journal of the American Medical Association
288(14): 1723-1727.
Floyd, M. F., J. O. Spengler, et al. (2008). "Park-Based Physical Activity in Diverse
Communities of Two US Cities:: An Observational Study." American journal of
preventive medicine 34(4): 299-305.
Floyd, M. F., W. C. Taylor, et al. (2009). "Measurement of park and recreation environments
that support physical activity in low-income communities of color: Highlights of
challenges and recommendations." American journal of preventive medicine 36(4):
S156-S160.
Ford, E. S., R. K. Merritt, et al. (1991). "Physical activity behaviors in lower and higher
socioeconomic status populations." American Journal of Epidemiology 133(12):
1246-1256.
Forsyth, A., J. Mary Hearst, et al. (2008). "Design and Destinations: Factors Influencing
Walking and Total Physical Activity" Urban Studies 45(9): 1973-1996.

178
Forsyth, A., J. Oakes, et al. (2007). Does residential density increase walking and other
physical activity? Urban Studies 44(4): 679-697.
Forsyth, A., J. Oakes, et al. (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.
Foster, S. and B. Giles-Corti (2008). "The built environment, neighborhood crime and
constrained physical activity: An exploration of inconsistent findings." Preventive
medicine 47(3): 241-251.
Frank, L. D. and P. Engelke (2005). "Multiple impacts of the built environment on public
health: Walkable places and the exposure to air pollution." International Regional
Science Review 28(2): 193-216.
Frank, L. D., and G. Pivo. (1995). Impacts of Mixed Use and Density on Utilization of Three
Modes of Travel: Single-Occupant Vehicle, Transit, and Walking. In Transportation
Research Record 1466, TRB, National Research Council, Washington, D.C., pp. 44–
52.
Frank, L. D., and P. O. Engelke. (2000). How Land Use and Transportation Systems Impact
Public Health: A Literature Review of the Relationship Between Physical Activity
and Built Form. Centers for Disease Control, Atlanta, Ga.
Frank, L. D., T. L. Schmid, et al. (2005). "Linking objectively measured physical activity
with objectively measured urban form: findings from SMARTRAQ." American
journal of preventive medicine 28(2): 117-125.
Fredman, P., U. Romild, et al. (2011). "Latent demand and time contextual constraints to
outdoor recreation in Sweden." Forests 3(1): 1-21.
Frumkin, H., Frank, L. D., & Jackson, R. (2004). The public health impacts of sprawl.
Washington, DC: Island Press.
Garling, T., T. Laitila, et al. (1998). Theoretical foundations of travel choice modeling: An
introduction. Theoretical foundations of travel choice modeling: 1-25.
Gidlow, C., L. H. Johnston, et al. (2006). "A systematic review of the relationship between
socio-economic position and physical activity." Health Education Journal 65(4):
338-367.
Giles-Corti, B. and R. J. Donovan (2002). "Socioeconomic status differences in recreational
physical activity levels and real and perceived access to a supportive physical
environment." Preventive Medicine: An International Journal Devoted to Practice
and Theory; Preventive Medicine: An International Journal Devoted to Practice and
Theory.
Giles-Corti, B. and R. J. Donovan (2003). "Relative influences of individual, social
environmental, and physical environmental correlates of walking." Journal
Information 93(9).
Giles-Corti, B., A. Timperio, et al. (2005). "Understanding physical activity environmental
correlates: increased specificity for ecological models." Exercise and sport sciences
reviews 33(4): 175.
Giles-Corti, B., A. Timperio, et al. (2006). "Development of a reliable measure of walking
within and outside the local neighborhood: RESIDE's Neighborhood Physical
Activity Questionnaire." Preventive Medicine 42(6): 455-459.

179
Giuliano, G. and J. Dargay (2006). "Car ownership, travel and land use: a comparison of the
US and Great Britain." Transportation Research Part A: Policy and Practice 40(2):
106-124.
Gobster, P. H. (2005). "Recreation and leisure research from an active living perspective:
taking a second look at urban trail use data." Leisure Sciences 27(5): 367-383.
Godbey, G. (2009). "Outdoor recreation, health, and wellness: Understanding and enhancing
the relationship." RFF Discussion Paper No. 09-21.
Godbey, G. C., L. L. Caldwell, et al. (2005). "Contributions of leisure studies and recreation
and park management research to the active living agenda." American Journal of
Preventive Medicine 28(2): 150-158.
Golob, T. F. (2000). "A simultaneous model of household activity participation and trip
chain generation." Transportation Research Part B: Methodological 34(5): 355-376.
Golob, T. F. (2003). "Structural equation modeling for travel behavior research."
Transportation Research Part B: Methodological 37(1): 1-25.
Gomez, J. E., B. A. Johnson, et al. (2004). "Violent crime and outdoor physical activity
among inner-city youth." Preventive medicine 39(5): 876-881.
Goodwin, P. and D. Hensher (1978). The transport determinants of travel choices: An
overview. Determinants of travel choice, edited by D. Hensher and Q. Dalvi.
Westmead, UK: Saxon House.
Gordon, C., M. Purciel-Hill, et al. (2011). "Measuring food deserts in New York City's low-
income neighborhoods." Health & place 17(2): 696-700.
Greenwald, M. J. and M. G. Boarnet (2001). "Built environment as determinant of walking
behavior: analyzing nonwork pedestrian travel in Portland, Oregon." Transportation
Research Record: Journal of the Transportation Research Board 1780(-1): 33-41.
Hägerstraand, T. (1970). "What about people in regional science?" Papers in regional
science 24(1): 7-24.
Hahn, Anderew, and Emma Craythorn. (1994) "Inactivity and the physical environment in
two regional centers." Health Promotion Journal of Australia 4(2): 43-45.
Hancock, G. R. and R. O. Mueller (2006). Structural equation modeling: A second course,
IAP.
Handy, S. (1992). "Regional versus local accessibility: neo-traditional development and its
implications for non-work travel." Built Environment 18(4): 253-267.
Handy, S. L. (1996). "Urban form and pedestrian choices: study of Austin neighborhoods."
Transportation Research Record: Journal of the Transportation Research Board
1552(-1): 135-144.
Handy, S. L. (2004). Critical assessment of the literature on the relationships among
transportation, land use, and physical activity. Davis, CA: Department of
Environmental Science and Policy, University of California, Davis.
Handy, S. L., M. G. Boarnet et al. (2002). How the Built Environment Affects Physical
Activity: Views from Urban Planning. In American Journal of Preventive Medicine,
23(2 supplement): 64–73.
Handy, S., X. Cao, et al. (2006). "Self-selection in the relationship between the built
environment and walking: Empirical evidence from Northern California." Journal of
the American Planning Association 72(1): 55-74.
Hanna, K. and C. Coussens (2001). Rebuilding the unity of health and the environment: a
new vision of environmental health for the 21st Century, National Academies Press.

180
Harrington, M., D. Dawson, et al. (1992). "Objective and subjective constraints on women's
enjoyment of leisure." Loisir et soci 챕 t 챕/Society and Leisure 15(1): 203-221.
Hawley, A. (1950). Human ecology: A theory of community structure, Ronald Press Co.
Hawley, A. H. (1988). Human ecology: a theory of community structure, Ronald Press C
Hawthorne, W. (1989). Why Ontarians Walk, Why Ontarians Don't Walk More: A Study
into the Walking Habits of Ontarians, Energy Probe Research Foundation.
Henderson, K. A. and M. D. Bialeschki (2005). "Leisure and active lifestyles: Research
reflections." Leisure Sciences 27(5): 355-365.
Henderson, K. A., S. Hodges, et al. (2002). "Context and dialogue in research on women and
leisure." Journal of Leisure Research 34(3): 253-271.
Hill, J. O., H. R. Wyatt, et al. (2003). "Obesity and the environment: where do we go from
here?" Science 299(5608): 853-855.
Hoehner, C. M., L. K. Brennan Ramirez, et al. (2005). "Perceived and objective
environmental measures and physical activity among urban adults." American
journal of preventive medicine 28(2): 105-116.
Hooker, S. P., D. K. Wilson, et al. (2005). "Perceptions of environmental supports for
physical activity in African American and white adults in a rural county in South
Carolina." Prev Chronic Dis 2(4): A11.
Hootman, J. M., C. A. Macera, et al. (2003). "Physical activity levels among the general US
adult population and in adults with and without arthritis." Arthritis Care & Research
49(1): 129-135.
Hovell, M. F., J. F. Sallis, et al. (1989). "Identifying correlates of walking for exercise: an
epidemiologic prerequisite for physical activity promotion." Preventive medicine
18(6): 856-866.
Humpel, N., N. Owen, et al. (2002). "Environmental factors associated with adults'
participation in physical activity* 1:: A review." American journal of preventive
medicine 22(3): 188-199.
Hynes, H. and R. Lopez (2009). Urban health: readings in the social, built, and physical
environments of US cities, Jones & Bartlett Learning.
Jackson, E. L. and D. Scott (1999). "Constraints to leisure." Leisure studies: Prospects for
the twenty-first century: 299-321.
Jacobsen, P. L., F. Racioppi, et al. (2009). "Who owns the roads? How motorised traffic
discourages walking and bicycling." Injury Prevention 15(6): 369-373.
Jerrett, M., R. McConnell, et al. (2010). Automobile traffic around the home and attained
body mass index: A longitudinal cohort study of children aged 10-18 years.
Preventive medicine 50: S50-S58.
Joh, K., M. T. Nguyen, et al. (2012). "Can Built and Social Environmental Factors
Encourage Walking among Individuals with Negative Walking Attitudes?" Journal
of Planning Education and Research.
Kaczynski A.T, et al. (2011). Variation in observed park physical activity intensity level by
gender, race, and age: individual and joint effects. Journal of Physical Activity and
Health, 8(s2): 151-160.
Kaczynski, A. T. and K. A. Henderson (2007). "Environmental correlates of physical
activity: a review of evidence about parks and recreation." Leisure Sciences 29(4):
315-354.

181
Kahneman, D., P. Wakker et al. (1997). Back to Bentham? Explorations of experienced
utility. InQuarterly Journal of Economics, 112(2):375–406.
Kaplan, R. (2001). "The nature of the view from home." Environment and Behavior 33(4):
507-542.
Katz, P., V. J. Scully, et al. (1994). The new urbanism: Toward an architecture of
community, McGraw-Hill New York.
Kemperman, A. D. A. M. and H. J. P. Timmermans (2008). "Influence of socio-
demographics and residential environment on leisure activity participation." Leisure
Sciences 30(4): 306-324.
King, A. C., C. Castro, et al. (2000). "Personal and environmental factors associated with
physical inactivity among different racial-ethnic groups of US middle-aged and
older-aged women." Health Psychology 19(4): 354.
King, A. C., Stokols, D., Talen, E., Brassington, G. S., & Killingsworth, R. (2002).
Theoretical Approaches to the Promotion of Physical Activity: Forging a
Transdisciplinary Paradigm. American Journal of Preventive Medicine, 23(02), 15-
25.
Kirk, M. A. and R. E. Rhodes (2011). "Occupation correlates of adults' participation in
leisure-time physical activity: a systematic review." American journal of preventive
medicine 40(4): 476-485.
Kitamura, Ryuichi. (1988) An evaluation of activity-based travel analysis. Transportation
15(1-2): 9-34.
Kockelman, K. M. (1997). Travel Behavior as Function of Accessibility, Land Use Mixing,
and Land Use Balance: Evidence from San Francisco Bay Area, Transportation
Research Board of the National Academies. 1607: 116-125.
Krizek, KJ, Handy, et al. (2009). "Explaining changes in walking and bicycling behavior:
challenges for transportation research." Environment and Planning B: Planning and
Design 36: 725-740.
LaGrange, R. L., K. F. Ferraro, et al. (1992). "Perceived risk and fear of crime: Role of
social and physical incivilities." Journal of research in crime and delinquency 29(3):
311-334.
Lee, C. and A. V. Moudon (2006). "Correlates of walking for transportation or recreation
purposes." Journal of Physical Activity & Health 3: 77.
Levine, N., K. E. Kim, et al. (1995). "Spatial analysis of Honolulu motor vehicle crashes: I.
Spatial patterns." Accident Analysis & Prevention 27(5): 663-674.
Lleras, C. (2005). "Path analysis." Encyclopedia of social measurement 3: 25-30.
Loukaitou-Sideris, A. (2004). "Transportation, Land Use, and Physical Activity: Safety and
Security Considerations" TRB Special Report 282.
Loukaitou-Sideris, A. (2006). "Is it Safe to Walk? 1 Neighborhood Safety and Security
Considerations and Their Effects on Walking." Journal of Planning Literature 20(3):
219-232.
Lu, X. and E. I. Pas (1999). "Socio-demographics, activity participation and travel
behavior." Transportation Research Part A: Policy and Practice 33(1): 1-18.
Lund, H. (2003). "Testing the claims of new urbanism: Local access, pedestrian travel, and
neighboring behaviors." Journal of the American Planning Association 69(4): 414-
429.

182
Maat, K. and H. J. Timmermans (2009). "A causal model relating urban form with daily
travel distance through activity/travel decisions." Transportation Planning and
Technology 32(2): 115-134.
Maat, K., B. van Wee, et al. (2005). Land use and travel behaviour: expected effects from
the perspective of utility theory and activity-based theories. Environment and
Planning B: Planning and Design 32(1): 33-46.
Marans, R. W. (1972). "Outdoor recreation behavior in residential environments."
Environment and the social sciences 293: 217-232.
Marshall, J. D., M. Brauer, et al. (2009). "Healthy neighborhoods: walkability and air
pollution." Environmental health perspectives 117(11): 1752.
Masse, L. Dassa, C., Gauvin, L., et al. (2002). Emerging measurement and statistical
methods in physical activity research. American Journal of Preventive Medicine
23(Suppl 2): 44-55.
MacDonald, K. (2012). "Structural Equation Modeling Using the sem Command and SEM
Builder." Stata Conference, San Diego,
http://fmwww.bc.edu/repec/san2012/macdonald.san2012.pdf
McFadden, D. (1974). The measurement of urban travel demand. Journal of Public
Economics 3(4): 303-328.
McGuckin, N. (2012). Walking and Biking in California: Analysis of the CA-NHTS,
University of California, Davis. Institute of Transportation Studies. Research report.
McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An Ecological Perspective on
Health Promotion Programs. Health Education & Behavior, 15(4), 351-377.
McNeill, L. H., M. W. Kreuter, et al. (2006). "Social environment and physical activity: a
review of concepts and evidence." Social science & medicine 63(4): 1011-1022.
Meloni, I., A. Portoghese, et al. (2009). "Effects of physical activity on propensity for
sustainable trips." Transportation Research Record: Journal of the Transportation
Research Board 2134(1): 43-50.
Merom, D., A. Bauman, et al. (2003). "An environmental intervention to promote walking
and cycling?-The impact of a newly constructed Rail Trail in Western Sydney."
Preventive Medicine 36(2): 235-242.
Mokhtarian, P. L. and X. Cao (2008). "Examining the impacts of residential self-selection on
travel behavior: A focus on methodologies." Transportation Research Part B:
Methodological 42(3): 204-228.
Moore, L. V., A. V. Diez Roux, et al. (2008). "Availability of recreational resources in
minority and low socioeconomic status areas." American journal of preventive
medicine 34(1): 16-22.
Moore, L. V., C. D. Harris, et al. (2012). "Trends in No Leisure-Time Physical Activity?:
United States, 1988-2010." Research quarterly for exercise and sport 83(4): 587-591.
Murray, N.G., S.H. Kelder, P.K. Orpinas, R. Marshall. (1995). Environmental safety and
level of physical activity in low-income urban Hispanic youth. Med. Sci. Sports
Exerc, 27: S166
Muthén, L. and B. Muthén (2007). "Mplus." Statistical analysis with latent variables.
Version 3.
Neighbors, C. J., D. X. Marquez, et al. (2008). "Leisure-time physical activity disparities
among Hispanic subgroups in the United States." Journal Information 98(8).
Newman, O. (1973). Defensible space: people and design in the violent city.

183
Ogilvie, D., R. Mitchell, et al. (2008). "Personal and environmental correlates of active
travel and physical activity in a deprived urban population." International Journal of
Behavioral Nutrition and Physical Activity 5(1): 43.
Olmsted, F.L. (1952) The value and care of parks. Landscape Architecture, 20-23.
Owen, N., N. Humpel, et al. (2004). "Understanding environmental influences on walking::
Review and research agenda." American journal of preventive medicine 27(1): 67-
76.
Panter, J. R. and A. Jones (2010). "Attitudes and the environment as determinants of active
travel in adults: what do and don't we know?" Journal of physical activity & health
7(4): 551.
Pas, Eric I, and Frank S Koppelman. 1986. "An examination of the determinants of day-to-
day variability in individuals' urban travel behavior." Transportation 13:183-200.
Pate, R. R., M. Pratt, et al. (1995). "Physical activity and public health." JAMA: the journal
of the American Medical Association 273(5): 402-407.
Payne, L. L., A. J. Mowen, et al. (2002). "An examination of park preferences and behaviors
among urban residents: the role of residential location, race, and age." Leisure
Sciences 24(2): 181-198.
Pendyala, R. M. and K. G. Goulias (2002). "Time use and activity perspectives in travel
behavior research." Transportation 29(1): 1-4.
Perkins, D., A. Wandersman, et al. (1993). The physical environment of street crime:
Defensible space, territoriality and incivilities1. Journal of Environmental
Psychology 13(1): 29-49.
Peter, C. (1993). The next American metropolis: ecology, community, and the American
dream, New York: Princeton Architectural Press.
Petraitis, P., A. Dunham, et al. (1996). "Inferring multiple causality: the limitations of path
analysis." Functional Ecology: 421-431.
Pikora, T. J., B. Giles-Corti, et al. (2006). "Neighborhood environmental factors correlated
with walking near home: using SPACES." Medicine & Science in Sports & Exercise
38(4): 708.
Pikora, T., B. Giles-Corti, et al. (2003). "Developing a framework for assessment of the
environmental determinants of walking and cycling." Social science & medicine
56(8): 1693-1703.
Pinjari, A. R., N. Eluru, et al. (2008). "Joint model of choice of residential neighborhood and
bicycle ownership: accounting for self-selection and unobserved heterogeneity."
Transportation Research Record: Journal of the Transportation Research Board
2082(1): 17-26.
Powell, L. M., S. Slater, et al. (2004). "The relationship between community physical
activity settings and race, ethnicity and socioeconomic status." Evidence-Based
Preventive Medicine 1(2): 135-44.
Powell, L. M., S. Slater, et al. (2006). "Availability of physical activity-related facilities and
neighborhood demographic and socioeconomic characteristics: a national study."
Journal Information 96(9).
Pucher, J. and J. L. Renne (2003). "Socioeconomics of urban travel: evidence from the 2001
NHTS." Transportation Quarterly 57(3): 49-77.

184
Pucher, J., R. Buehler, et al. (2011). "Walking and cycling in the United States, 2001-2009:
evidence from the National Household Travel Surveys." Journal Information
101(S1).
R. Messent, C. B. C., Jonathon Long, P (1999). "Primary and secondary barriers to
physically active healthy lifestyles for adults with learning disabilities." Disability &
Rehabilitation 21(9): 409-419.
Raykov, T. and G. A. Marcoulides (2000). "A method for comparing completely
standardized solutions in multiple groups." Structural equation modeling 7(2): 292-
308.
Raymore, L. A. (2002). "Facilitators to leisure." Journal of Leisure Research 34(1): 37-51
Reidpath, D. D., C. Burns, et al. (2002). An ecological study of the relationship between
social and environmental determinants of obesity. (Elsevier), 8: 141-145.
Retting, R. A., S. A. Ferguson, et al. (2003). "A review of evidence-based traffic engineering
measures designed to reduce pedestrian-motor vehicle crashes." American Journal of
Public Health 93(9): 1456-1463.
Rimmer, J., L. Wolf, et al. (2008). Physical activity among adults with a disability-United
States, 2005 (Reprinted from MMWR, vol 56, pg 1021-1024, 2007), AMER
MEDICAL ASSOC 515 N STATE ST, CHICAGO, IL 60610-0946 USA. 299:
1255-1256.
Roberts, I., H. Owen, et al. (1996). "Pedalling Health: Health Benefits of a modal transport
shift. 1996." ISBN: 0 7308: 0741.
Roberts, I., R. Norton, et al. (1995). "Effect of environmental factors on risk of injury of
child pedestrians by motor vehicles: a case-control study." Bmj 310(6972): 91-94
Romero, A. J., T. N. Robinson, et al. (2001). "Are perceived neighborhood hazards a barrier
to physical activity in children?" Archives of Pediatrics and Adolescent Medicine
155(10): 1143.
Ryan, S. and M. G. McNally (1995). "Accessibility of neotraditional neighborhoods: A
review of design concepts, policies, and recent literature." Transportation Research
Part A: Policy and Practice 29(2): 87-105.
Saelens, B. E. and S. L. Handy (2008). "Built environment correlates of walking: a review."
Medicine and science in sports and exercise 40(7 Suppl): S550.
Saelens, B. E., J. F. Sallis, et al. (2012). "Neighborhood environment and psychosocial
correlates of adults' physical activity." Med Sci Sports Exerc 44(4): 637-646.
Saelens, Brian E., Sallis, James F., and Frank, Lawrence D. (2003). Environmental correlates
of walking and cycling: Findings from the transportation, urban design, and planning
literatures. Annals of Behavioral Medicine 25(2): 80-91.
Sahlqvist, S. L., Y. Song, et al. (2012). "Is active travel associated with greater physical
activity? The contribution of commuting and non-commuting active travel to total
physical activity in adults." Preventive Medicine 55(3): 206-211.
Sallis, J., A. Bauman, et al. (1998). "Environmental and policy interventions to promote
physical activity." American journal of preventive medicine 15(4): 379-397.
Sallis, J. F., B. E. Saelens, et al. (2009). "Neighborhood built environment and income:
examining multiple health outcomes." Social science & medicine 68(7): 1285-1293.
Sallis, J. F., C. Richard Hofstetter, et al. (1989). "A multivariate study of determinants of
vigorous exercise in a community sample." Preventive medicine 18(1): 20-34.

185
Sallis, J. F., J. J. Prochaska, et al. (2000). "A review of correlates of physical activity of
children and adolescents." Medicine & Science in Sports & Exercise 32(5): 963.
Sallis, J. F., L. D. Frank, et al. (2004). "Active transportation and physical activity:
opportunities for collaboration on transportation and public health research."
Transportation Research Part A: Policy and Practice 38(4): 249-268.
Sallis, J. F., M. F. Johnson, et al. (1997). "Assessing perceived physical environmental
variables that may influence physical activity." Research quarterly for exercise and
sport 68(4): 345.
Sallis, J. F., R. B. Cervero, et al. (2006). "An ecological approach to creating active living
communities." Annu. Rev. Public Health 27: 297-322.
Sastry, N., A. R. Pebley, et al. (2002). "Neighborhood definitions and the spatial dimension
of daily life in Los Angeles."
Scheiner, J. (2010). "Social inequalities in travel behaviour: trip distances in the context of
residential self-selection and lifestyles." Journal of Transport Geography 18(6): 679-
690.
Scheiner, J. and C. Holz-Rau (2007). "Travel mode choice: affected by objective or
subjective determinants?" Transportation 34(4): 487-511.
Schipperijn, J., U. K. Stigsdotter, et al. (2010). "Influences on the use of urban green space:
A case study in Odense, Denmark." Urban forestry & urban greening 9(1): 25-32.
Shaw, S. M., A. Bonen, et al. (1991). "Do more constraints mean less leisure? Examining the
relationship between constraints and participation." Journal of Leisure Research.
Shinew, K. J., M. F. Floyd, et al. (1996). "Class polarization and leisure activity preferences
of African Americans: Intragroup comparisons." Journal of Leisure Research 28:
219-232.
Siegenthaler, K. (1997). "Health Benefits of Leisure. Research Update." Parks and
Recreation.
Sioutas C, Delfino RJ, Singh M. (2005). Exposure assessment for atmospheric ultrafine
particles (UFPs) and implications in epidemiologic research. Environmental Health
Perspectives 113:947
Sisson, S. B. and C. Tudor-Locke (2008). "Comparison of cyclists' and motorists' utilitarian
physical activity at an urban university." Preventive medicine 46(1): 77-79.
Sooman, A. and S. Macintyre (1995). "Health and perceptions of the local environment in
socially contrasting neighbourhoods in Glasgow." Health & Place 1(1): 15-26.
Spivock, M., L. Gauvin, et al. (2008). "Promoting active living among people with physical
disabilities: evidence for neighborhood-level buoys." American journal of
preventive medicine 34(4): 291-298.
Stokols, D. (1992). Establishing and maintaining healthy environments. American
Psychologist 47(1): 6-22.
Stokols, D. (1996). Translating social ecological theory into guidelines for community health
promotion. American Journal of Health Promotion 10: 282-298.
Stott, R. (2000). The ecology of health, Resurgence Books.
Sugiyama, T. and C. W. Thompson (2007). "Outdoor environments, activity and the well-
being of older people: conceptualising environmental support." Environment and
Planning A 39(8): 1943-1960.
Suminski, R. R., W. S. C. Poston, et al. (2005). "Features of the neighborhood environment
and walking by US adults." American journal of preventive medicine 28(2): 149-155.

186
Sylvia-Bobiak, S. and L. L. Caldwell (2006). "Factors related to physically active leisure
among college students." Leisure Sciences 28(1): 73-89.
Tarrant, M. A., M. J. Manfredo, et al. (1994). "Recollections of outdoor recreation
experiences: A psychophysiological perspective." Journal of Leisure Research 26:
357-357.
Taylor, W. C., T. Baranowski, et al. (1998). "Physical activity interventions in low-income,
ethnic minority, and populations with disability." American journal of preventive
medicine 15(4): 334-343.
Thommen Dombois, O., C. Braun-Fahrl 채 nder, et al. (2007). "Comparison of adult
physical activity levels in three Swiss alpine communities with varying access to
motorized transportation." Health & place 13(3): 757-766.
Troped, P. J., R. P. Saunders, et al. (2001). "Associations between self-reported and
objective physical environmental factors and use of a community rail-trail."
Preventive medicine 32(2): 191.
Trost, S. G., N. Owen, et al. (2002). "Correlates of adults' participation in physical activity:
review and update." Medicine & Science in Sports & Exercise 34(12): 1996.
Tsai, E. H.-L. and D. J. Coleman (2009). "The influence of constraints and self-efficacies on
participation in regular active recreation." Leisure Sciences 31(4): 364-383.
Turrell, G., M. Haynes, et al. (2013). "Can the built environment reduce health inequalities?
A study of neighbourhood socioeconomic disadvantage and walking for transport."
Health & place 19: 89-98
Ulrich, R. (1981). Natural versus urban scenes. Environment and Behavior 13(5): 523.
Vaara, M. and J. Matero (2011). "Modeling daily outdoor recreation participation and time-
use as a two-stage choice process: a Finnish example." Leisure Sciences 33(4): 269-
289.
Van Acker, V., B. Van Wee, et al. (2010). "When transport geography meets social
psychology: toward a conceptual model of travel behaviour." Transport Reviews
30(2): 219-240.
Van Dyck, D., G. Cardon, et al. (2010). "Neighborhood SES and walkability are related to
physical activity behavior in Belgian adults." Preventive medicine 50: S74-S79.
Van Stralen, M. M., H. De Vries, et al. (2009). "Determinants of initiation and maintenance
of physical activity among older adults: a literature review." Health Psychology
Review 3(2): 147-207.
Wells, N. M. (2000). "At home with nature." Environment and Behavior 32(6): 775-795.
Wen, M., N. R. Kandula, et al. (2007). "Walking for transportation or leisure: what
difference does the neighborhood make?" Journal of General Internal Medicine 22:
1674-1680.
Wendel-Vos, W., M. Droomers, et al. (2007). "Potential environmental determinants of
physical activity in adults: a systematic review." Obesity Reviews 8(5): 425-440.
Wilcox, S., C. Castro, et al. (2000). "Determinants of leisure time physical activity in rural
compared with urban older and ethnically diverse women in the United States."
Journal of Epidemiology and Community Health 54(9): 667-672.
Williams, S. (1995). Outdoor recreation and the urban environment, Routledge.
Wolch, J., J. P. Wilson, et al. (2005). "Parks and park funding in Los Angeles: an equity-
mapping analysis." Urban Geography 26(1): 4-35.

187
Wyrwich, K. W., R. C. Brownson, et al. (2006). "Individual, social environmental, and
physical environmental influences on physical activity among black and white
adults: a structural equation analysis." Annals of Behavioral Medicine 31(1): 36-44.
Yang, Y. and A. V. Diez-Roux (2012). "Walking distance by trip purpose and population
subgroups." American journal of preventive medicine 43(1): 11-19.
Zegras, C. (2010). "The built environment and motor vehicle ownership and use: evidence
from Santiago de Chile." Urban Studies 47(8): 1793-1817.
Zakarian, J. M., M. F. Hovell, et al. (1994). "Correlates of vigorous exercise in a
predominantly low SES and minority high school population." Preventive medicine
23(3): 314-321.
Zhu Y, Hinds WC, Kim S, Sioutas C. (2002). Concentration and size distribution of ultrafine
particles near a major highway. Journal of the air & waste management association
52:1032-42
Zhu Y, Hinds WC, Shen S, Sioutas C. (2004). Seasonal Trends of Concentration and Size
Distribution of Ultrafine Particles Near Major Highways in Los Angeles Special
Issue of Aerosol Science and Technology on Findings from the Fine Particulate
Matter Supersites Program. Aerosol Science and Technology 38:5-13

188
APPENDIX

Table 1 Results of simple regression model (pooled samples; N=3,486)
Time Freq Time Freq Time Freq Time Freq Time Freq Time Freq
Age -0.00378 -0.01341 0.00089 -0.0139 0.01893 -0.0071 0.02067 0.0445 0.04179 0.04105 0.03402 0.01721
*** ** ** **
Gender -0.02364 -0.01553 -0.0401 -0.0414 -0.0428 -0.0361 0.01339 0.01112 0.0222 0.02459 0.01304 0.01196
** ** ** **
R_White -0.03184 -0.02907 0.03312 0.02513 -0.0277 -0.0251 0.04348 0.01576 0.01155 -0.0017 -0.0077 0.0088
* * * **
R_Black 0.03078 0.02672 -0.0276 -0.0132 0.02905 0.03232 0.00357 0.01667 -0.0005 0.03561 0.03604 0.01041
* * * ** **
R_Asian 0.03058 0.03473 -0.0389 -0.0456 0.00691 0.03997 -0.0346 -0.0231 -0.0117 -0.0034 -0.0005 0.00569
* ** ** *** ** **
R_Hispanic 0.00591 0.00203 0.00139 0.00098 0.01093 0.00554 -0.0331 -0.0179 -0.0051 -0.0185 -0.0141 -0.0179
*
R_Others 0.00293 0.01438 0.00776 0.01759 0.01778 0.02295 0.0009 0.00624 -0.0005 -0.0012 0.00009 -0.0087
Educ_low 0.04701 0.04559 0.04078 0.06258 0.03307 0.06632 -0.0308 -0.0059 -0.0118 -0.0082 -0.0011 0.0037
*** *** ** *** * *** *
Educ_high 0.03804 0.03754 0.03801 0.04715 0.01278 0.02883 0.02226 0.01594 0.01644 0.00692 -0.0021 0.01526
** ** ** *** *
Employment -0.0647 -0.04418 -0.0701 -0.057 -0.042 -0.0815 -0.0103 0.0015 0.00573 0.02683 0.0197 0.00881
*** *** *** *** ** ***
Income_low 0.02443 0.01234 0.02447 0.01728 0.04831 0.03124 -0.0106 -0.011 -0.0225 -0.0222 -0.0238 -0.0194
*** *
Income_high 0.02265 0.0208 0.02981 0.02977 0.01094 0.03087 0.00933 0.00223 0.02862 0.01907 0.01769 0.0016
* * * *
Disability -0.03833 -0.04067 0.04141 0.03367 -0.0007 -0.0514 -0.0284 -0.0324 -0.0242 -0.031 -0.0142 -0.0313
** ** ** ** *** * * * *
Trip date -0.01849 -0.01309 -0.0136 -0.0344 -0.0358 -0.0131 -0.025 -0.0135 -0.0155 0.00316 0.00958 -0.0048
** **
Weather 0.01087 -0.00629 0.00695 -0.0106 -0.0167 -0.0111 0.00359 0.00752 -0.0357 0.00299 -0.0009 0.00466
**
Car (Dummy) -0.10549 -0.09788 -0.1107 -0.1147 -0.0623 -0.1288 0.00669 0.03049 0.02711 0.02318 0.02635 0.02148
*** *** *** *** *** *** *
No. of Car -0.12237 -0.12474 -0.1246 -0.1537 -0.0652 -0.1542 0.01939 0.0477 0.01822 0.04097 0.03211 0.04555
*** *** *** *** *** *** *** ** * ***
Children -0.0113 -0.0097 -0.026 -0.0018 -0.036 -0.0088 0.01281 0.00483 0.0099 0.00837 0.00106 0.01438
**
Variables Exercise (ty3)
Daily active traval Outdoor leisure activity
Active travel Walk Walk (dummy) Exercise (ty1) Exercise (ty2)

189
Time Freq Time Freq Time Freq Time Freq Time Freq Time Freq
NT_urb 0.05657 0.06179 0.05269 0.08019 0.04905 0.08446
*** *** *** *** *** ***
NT_sub -0.00026 -0.0228 -0.0114 -0.0321 -0.0068 -0.0289
* *
NT_Leisure_high -0.0124 -0.0055 0.01601 0.02259 0.02572 -0.0045
NT_Leisure_low -0.0098 -0.0059 -0.0234 -0.0322 -0.0352 0.00084
**
Density 0.07176 0.0876 0.07221 0.11072 0.01654 0.10922 -0.0084 -0.0264 -0.017 -0.0208 -0.0216 -0.0175
*** *** *** *** ***
Diversity 0.03937 0.05185 0.04997 0.08145 0.01974 0.08277 -0.0155 -0.0088 0.0066 0.00972 0.01169 0.01019
** *** *** *** ***
Design 0.05033 0.05977 0.05865 0.08143 0.03605 0.08398 -0.0117 -0.0259 -0.0226 -0.0208 -0.0147 -0.0362
*** *** *** *** ** *** **
No. of Park 0.05167 0.04565 0.05234 0.04763 0.0128 0.04512 -0.0029 0.00837 -0.0078 0.01233 0.02499 -0.0072
** *** *** *** ***
Sport/recreation 0.03229 0.0225 0.02729 0.03153 0.01944 0.05761
* * ***
RailStops -0.00348 0.00398 -0.0003 -0.0014 -0.0128 0.0119
BusStops 0.0429 0.04212 0.04796 0.07848 0.01344 0.06893
** ** *** *** ***
Safety concern -0.00136 -0.01372 -0.0018 -0.0021 -0.0094 -0.0106 -0.0092 -0.013 -0.0013 -0.0185 -0.0194 -0.0124
Crime_Q -0.03338 -0.04342 -0.0342 -0.069 -0.0017 -0.0649 0.00933 -0.0123 -0.0053 0.00569 0.00593 0.00248
** *** ** *** ***
Traffic acc_Q -0.09201 -0.09455 -0.0937 -0.1298 -0.0292 -0.126 -0.0138 -0.0412 -0.018 -0.0519 -0.0454 -0.0263
*** *** *** *** * *** ** *** ***
Freeway (100M) -0.00126 0.02018 0.00539 0.02443 -0.0021 0.02408 -0.0199 -0.0211 0.00429 -0.0029 0.00538 -0.0136
Variables
Daily active traval Outdoor leisure activity
Active travel Walk Walk (dummy) Exercise (ty1) Exercise (ty2) Exercise (ty3)

Note: In outdoor leisure activity, ‘ty1’ refers to exercise/sport only; ‘ty2’ indicates exercise/sport or
park activity; and ‘ty3’ points to exercise/sport, park activity, or leisure active travel
(‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1)

190
Table 2 Results of simple correlation between variables (pooled samples; N=3,486)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
NT_urb 1
1
NT_sub -0.1475 1
2 ***
NT_Leisure_high 0.01832 -0.05401 1
3 ***
NT_Leisure_low 0.00919 0.06186 -0.40895 1
4 *** ***
Activity density 0.45116 -0.35761 0.01587 0.00593 1
5 *** ***
Diversity (mixed use) 0.29839 -0.40369 0.08344 -0.0881 0.63995 1
6 *** *** *** *** ***
Street design 0.47687 -0.43181 0.03055 -0.03148 0.3758 0.321 1
7 *** *** * * *** ***
No. of Park 0.11275 -0.07443 0.46376 -0.6105 0.08566 0.08907 0.15922 1
8 *** *** *** *** *** *** ***
Sport facility 0.13165 -0.1186 0.05178 -0.01917 0.13841 0.18338 0.18222 0.08271 1
9 *** *** *** *** *** *** ***
Rail stops 0.03127 -0.02856 -0.02713 0.00518 0.05559 0.09071 0.02145 -0.00933 0.00329 1
10 * * *** ***
Bus stops 0.34607 -0.19764 0.01236 -0.00734 0.62827 0.36159 0.31603 0.03355 0.14731 0.14513 1
11 *** *** *** *** *** ** *** ***
Safety concern 0.03304 -0.04359 0.0039 -0.00572 0.03213 0.03426 0.01948 -0.00025 -0.00301 0.02372 0.01479 1
12 * ** * **
Violent crimes 0.30327 -0.22075 -0.03836 0.06032 0.49678 0.38731 0.26958 -0.00844 0.13412 0.05918 0.49133 0.02161 1
13 *** *** ** *** *** *** *** *** *** ***
Traffic incidents 0.44713 -0.22896 -0.02032 0.02234 0.69198 0.44955 0.36971 0.06278 0.14975 0.05417 0.61125 0.03296 0.61861 1
14 *** *** *** *** *** *** *** *** *** * ***
Freeway_100M 0.06523 -0.12066 -0.01057 -0.00199 0.12811 0.15444 0.00487 0.01026 0.00793 0.00447 0.11149 -0.01112 0.12635 0.05255 1
15 *** *** *** *** *** *** ***

Note: ‘***’ p<0.01; ‘**’ p<0.05; ‘*’ p<0.1

Abstract (if available)

Abstract In response to increasingly sedentary lifestyles, many policymakers have suggested making neighborhood design more conducive to active lifestyles, emphasizing the role of physical elements, such as high density, mixed land use, street design, and easy access to local destinations. However, a neighborhood not only provides opportunities for active living but also contains various types of environmental barriers. While the role of environmental opportunity is well acknowledged, little is known about the deterrent role of environmental barriers. ❧ Drawing on a fairly large number of adult samples obtained from NHTS California data while simultaneously developing a more fine-grained objective measurement of neighborhood environment, this dissertation explores what aspects of neighborhood environment facilitate and/or constrain active travel and outdoor leisure on a daily basis. Identifying the potential gap between low- and high- income groups in terms of environmental opportunities and barriers, this study intends to better understand physical inactivity among subpopulations more exposed to those barriers. Using path analysis which allows the analyst to isolate the impact of neighborhood environment from that of residential sorting, I address the complex relationships between personal/household backgrounds, neighborhood characteristics, and behavioral decisions on active travel/outdoor leisure activity. ❧ Empirical results reveal not only significant supportive (e.g., activity density, street design, and parks) but also deterrent (e.g., violent crimes) roles of neighborhood environment in the duration and frequency of daily active travel. However, active travel mainly depends on individual/household characteristics. In particular, household travel options (i.e., car availability) consistently indicated the most profound and negative influence at the significant level. A clear gap between low- and high- income groups in environmental correlates of active travel was also found. In terms of the duration, the low- income group is more responsive to activity density and neighborhood (violent) crimes than the high-income group. On the other hand, both groups reported a consistently deterrent role of violent crimes in the frequency measured by a continuous scale. Interestingly, however, such a deterrent role was not significant in determining whether low-income groups walk or not. ❧ Analytic models for outdoor leisure activity consistently presented the positive impact of local facilities for exercise/sport, confirming the supportive role of neighborhood. However, other variables yielded inconsistent results, suggesting that the function of outdoor leisure is mostly subject to specific types of leisure activity and household income. For low- income groups, a leisure-friendly neighborhood had a significant positive association with the duration of outdoor exercise/sport. Given the limited travel options of low-income groups, this might imply the importance of neighborhood amenities for leisure. High-income groups are more sensitive to perceived safety concern when they engage in outdoor leisure, whereas low-income groups are more responsive to exposure to traffic incidents objectively measured. Surprisingly, it is found that the amount of time spent on active travel is the most profound factor affecting outdoor leisure activity. Such trade-off relationships between different domains of physical activity cast doubt on the conventional wisdom that engaging in active travel increases the overall physical activity. ❧ To effectively promote active behavior and life styles, policy should be devoted to injury prevention designed to reduce neighborhood barriers, in tandem with supportive neighborhood design. Moreover, this study finds that active-living campaigns will achieve better performance in combination with transportation planning (e.g., more pricing on vehicle use and improving mobility of low-income groups). However, despite such combined policy endeavors, increasing overall physical activity is a much bigger challenge if the total budget of physical activity is fixed or controlled by other factors beyond the policy and environmental strategies.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The built environment, tour complexity, and active travel

PDF

The demand for reliable travel: evidence from Los Angeles, and implications for public transit policy

PDF

Healthy mobility: untangling the relationships between the built environment, travel behavior, and environmental health

PDF

Urban air pollution and environmental justice: three essays

PDF

Location of warehouses and environmental justice: Three essays

PDF

The long-term impact of COVID-19 on commute, employment, housing, and environment in the post-pandemic era

PDF

The flexible workplace: regional tendencies and daily travel implications

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

Children’s travel behavior in journeys to school

PDF

Environmental justice in real estate, public services, and policy

PDF

Choice neighborhoods: a spatial and exploratory analysis of Housing Authority City of Los Angeles public housing

PDF

Social construction of the experience economy: the spatial ecology of outdoor advertising in Los Angeles

PDF

A view from below: the development and role of organizational social capital in neighborhood regeneration in Los Angeles

PDF

Reshaping Los Angeles: housing affordability and neighborhood change

PDF

Governance networks: internal governance structure, boundary setting and effectiveness

PDF

Affordable south Los Angeles: survival, support, and different futures

PDF

Collaboration: is it worth it? The Magnolia Community Initiative from the perspective of initiative partner participants

Asset Metadata

Creator Ahn, Yong-Jin (author)

Core Title Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

School School of Policy, Planning and Development

Degree Doctor of Policy, Planning & Development

Degree Program Policy, Planning, and Development

Publication Date 01/17/2016

Defense Date 12/19/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag daily travel/activity,GIS,healthy neighborhood,neighborhood opportunity/barrier,OAI-PMH Harvest,outdoor leisure,path analysis,Walking

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Giuliano, Genevieve (committee chair), Boarnet, Marlon (committee member), Spruijt-Metz, Donna (committee member)

Creator Email dadaist21@gmail.com,yongjina@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-360773

Unique identifier UC11295179

Identifier etd-AhnYongJin-2230.pdf (filename),usctheses-c3-360773 (legacy record id)

Legacy Identifier etd-AhnYongJin-2230.pdf

Dmrecord 360773

Document Type Dissertation

Format application/pdf (imt)

Rights Ahn, Yong-Jin

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA