University of Southern California Dissertations and Theses

The built environment, tour complexity, and active travel

(USC Thesis Other)

The built environment, tour complexity, and active travel

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

THE BUILT ENVIRONMENT, TOUR COMPLEXITY, AND
ACTIVE TRAVEL

by

Jeongwoo Lee

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)

December 2013

Copyright 2013 Jeongwoo Lee

iii

DEDICATION

This dissertation is dedicated to
my husband, YeongGon Bae and my son, Jae-O Bae.

iv

ACKNOWLEDGEMENTS

This dissertation was funded by Sol Price School’s Research Assistantship and
the Los Angeles Women’s Transportation Coalition Scholarship.
This dissertation was made possible due to the generous support of many
individuals, and I owe my sincerest gratitude to all of them. First, I am most grateful to
Dr. Genevieve Giuliano, the chair of my committee, for her patient and incisive guidance
throughout this research process. Without her detailed reviews for accuracy and precision,
I could not have completed this effort. Most importantly, she taught me how to question
established knowledge like a true scholar. Her perspective was quite different from the
planners’ perspective I acquired through my engineering training in an architectural firm.
She was always available and asked the most interesting questions. I believe her teaching
and mentorship will genuinely help me with my own teaching and will continue to inspire
me as a scholar.
My sincerest gratitude also goes to Dr. Marlon Boarnet, who provided me with
wonderful opportunities for intriguing research. My research has greatly benefited from
his generous financial support and insightful comments. I admire his capacity to criticize,
make sense of, and give additional value to projects. In addition, Dr. Lisa Schweitzer has
served as a good mentor throughout the years. She has given me numerous chances to
contribute to several research projects and provided me with several good ideas for my
research during its earliest stages. I have been impressed by how warm-hearted she is and
how much she truly cares about her students’ success.
v

I would also like to acknowledge Dr. James Moore for joining my dissertation
committee as an outside member. My conversations with him during the research process
were always very pleasant, and he helped me develop my research skills during this
process. I have consistently been amazed by how responsive he has been to every email
that I sent and to all of my questions.
I gratefully acknowledge that the data collection during this research was
supported by the California Department of Transportation and the Southern California
Association of Governments (SCAG). Special thanks go to Dr. Simon Choi, Dr. Hsi-Hwa
Hu, and Sungsu Yoon from SCAG, who helped me complete this project.
A broad thank you goes out to the faculty and staff members at USC who helped
me conduct this research, demonstrating how to think and write critically while enjoying
the entire process. In particular, I would like to thank Dr. Tridib Banerjee, Dr. Peter
Gordon, Dr. Eric Heikkila, MaryAnn Murphy, and Victoria Valentine. I am also very
thankful to my classmates and research colleagues at USC.
Finally, I feel very fortunate to have the support of a wonderful family. My
husband, YeongGon Bae, has consistently been the biggest supporter for my doctoral
project. Without his tireless and unconditional love and support, I would not have been
able to complete this work. I would also like to thank my parents for encouraging me to
complete my studies and my sister-in-law, Hyosung Bae, who helped take care of my son
with love while I was away from home. Without their sacrifice and patience, I would not
have achieved all that I have today.
vi

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ............................................................................................................................... iv
TABLE OF CONTENTS .................................................................................................................................... vi
LIST OF TABLES ................................................................................................................................................. ix
LIST OF FIGURES ................................................................................................................................................. x
ABSTRACT .............................................................................................................................................................. xi
CHAPTER 1. INTRODUCTION ..................................................................................................................... 1
1.1. BACKGROUND AND MOTIVATION ..................................................................................................... 1
1.2. RESEARCH STATEMENT ........................................................................................................................ 2
1.2.1. Access to Transit and the Built Environment ...................................................................... 2
1.2.2. Mode Choice and the Built Environment .............................................................................. 5
1.3. ORGANIZATION OF THE DISSERTATION ........................................................................................... 8
CHAPTER 2. LITERATURE REVIEW .................................................................................................... 10
2.1. THEORIES OF TRAVEL BEHAVIOR ................................................................................................... 10
2.1.1. Utility-Maximization Theory .................................................................................................. 10
2.1.2. Value of Time .............................................................................................................................. 12
2.1.3. Constraint Approach .................................................................................................................. 15
2.1.4. Activity-Based Approach ......................................................................................................... 18
2.2. THE ROLE OF THE BUILT ENVIRONMENT IN TRANSIT USE ..................................................... 20
2.3. THE ROLE OF THE BUILT ENVIRONMENT IN THE CHOICE OF TRAVEL MODE ................... 24
2.4. DISCUSSION ............................................................................................................................................ 26
CHAPTER 3. RESEARCH APPROACH, DATA, AND MEASURES ......................................... 30
3.1. RESEARCH APPROACH ........................................................................................................................ 30
3.1.1. Determinants of Transit Use ................................................................................................... 30
3.1.2. Mode Choice Analysis: Complex tour ................................................................................. 32
3.2. DATA AND MEASURES ...................................................................................................................... 36
3.2.1. Data ................................................................................................................................................. 36
3.2.2. Measures of the Built Environment ...................................................................................... 38
CHAPTER 4. PERCEIVED BUILT ENVIRONMENT AND TRANSIT USE IN ................... 43
vii

LOW-INCOME POPULATIONS ................................................................................................................. 43
4.1. INTRODUCTION ...................................................................................................................................... 43
4.2. DESCRIPTIVE ANALYSIS ..................................................................................................................... 46
4.2.1. General Patterns of Transit Use ............................................................................................. 46
4.2.2. Perceptions about Walking Environment ............................................................................ 48
4.3. METHODOLOGY .................................................................................................................................... 50
4.3.1. Analysis Framework .................................................................................................................. 51
4.3.2. Key Variables .............................................................................................................................. 53
4.4. ANALYSIS AND FINDINGS .................................................................................................................. 59
4.4.1. Principal Component Analysis ............................................................................................... 59
4.4.2. Cluster Analysis .......................................................................................................................... 61
4.5. ESTIMATION RESULTS ......................................................................................................................... 64
4.6. DISCUSSION ............................................................................................................................................ 70
CHAPTER 5. TOUR COMPLEXITY AND TRAVEL MODE CHOICE: THE ROLE OF 73
NEIGHBORHOOD WALKABILITY ......................................................................................................... 73
5.1. INTRODUCTION ...................................................................................................................................... 73
5.2. DATA AND METHODOLOGY .............................................................................................................. 74
5.2.1. Data ................................................................................................................................................. 75
5.2.2. Measuring the Walkability Index .......................................................................................... 76
5.2.3. Methodology ................................................................................................................................ 78
5.3. DESCRIPTIVE ANALYSIS ..................................................................................................................... 85
5.3.1. Distribution of Mode of Travel by Neighborhood Walkability ................................... 86
5.3.2. Trip Chaining and Mode Choice ........................................................................................... 88
5.4. RESULTS OF HYPOTHESES TESTED ................................................................................................. 90
5.4.1. Determinants of Tour Complexity and VMT .................................................................... 90
5.4.2. Determinants of the Choice of Tour Mode ......................................................................... 96
5.5. DISCUSSION ......................................................................................................................................... 105
CHAPTER 6. CONCLUSION ...................................................................................................................... 108
6.1. SUMMARY OF FINDINGS .................................................................................................................. 108
6.2. CONTRIBUTION TO THE LITERATURE .......................................................................................... 111
6.3. POLICY AND PLANNING IMPLICATIONS ...................................................................................... 113
viii

6.3.1. Access to Transit by Low-Income Households .............................................................. 113
6.3.2. Neighborhood-Level Interventions to Promote Active Travel .................................. 114
6.4. DIRECTIONS FOR FUTURE RESEARCH ......................................................................................... 116
BIBLIOGRAPHY .............................................................................................................................................. 119

ix

LIST OF TABLES

Table 4. 1. Variable definitions and data sources ..................................................................................... 57
Table 4. 2. Descriptive statistics of individual‐level variables ............................................................ 58
Table 4. 3. Descriptive statistics of neighborhood‐level variables .................................................... 58
Table 4. 4. Factor loading matrix (Principal Component analysis of perception variables) .. 6 0
Table 4. 5. T‐test to compare low‐ and higher‐income traveler’s factor score ............................ 61
Table 4. 6. Mean values by neighborhood type (2,339 household neighborhoods clustered
by Euclidean distance) ......................................................................................................................................... 63
Table 4. 7. The results of ordinal logit model 1 ......................................................................................... 6 8
Table 4. 8. The results of ordinal logit model 2 ......................................................................................... 69
Table 5. 1. Description of variables ................................................................................................................ 78
Table 5. 2. Descriptive statistics of categorical variables ...................................................................... 84
Table 5. 3. Descriptive statistics of continuous variables ..................................................................... 84
Table 5. 4. T‐tests of mode choice and tour type for work and non‐work tours ......................... 89
Table 5. 5. Regression results for tour generation ................................................................................... 94
Table 5. 6. Tobit regression results for VMT .............................................................................................. 95
Table 5. 7. Multinomial logit model results (pooled‐sample model) ............................................ 10 2
Table 5. 8. Multinomial logit model results (work tour model) ...................................................... 10 3
Table 5. 9. Multinomial logit model results (non‐work tour) .......................................................... 10 4
Table 5. 10. Elasticity estimates for active travel mode ..................................................................... 105

x

LIST OF FIGURES

Figure 3. 1. Conceptual model of the study: factors influencing transit use ................................. 32
Figure 3. 2. Conceptual model of the study: factors influencing tour mode choice ................... 36
Figure 3. 3. Measuring land‐use mix within a ¼‐mile area around each household location 40
Figure 3. 4. Measuring walking distance from the household location to the nearest bus stop
....................................................................................................................................................................................... 41
Figure 4. 1. Transit use patterns by income groups ................................................................................ 45
Figure 4. 2. Access mode to get to public transit ...................................................................................... 47
Figure 4. 3. Number of transit use per month by access mode ........................................................... 47
Figure 4. 4. Ranked order of the reasons people in Los Angeles County do not walk more... 49
Figure 4. 5. Geographical clustering of neighborhood types in Los Angeles County ................. 63
Figure 5. 1. Proportion of trip chaining within a work / non‐work tour ........................................ 85
Figure 5. 2. Distribution of modes of travel by different types of tours and neighborhood
walkability (a) simple tour, (b) complex tour, (c) work tour, (d) non‐work tour ...................... 87
xi

ABSTRACT
The central question posed in this dissertation is concerned with the role that the built
environment has in people’s travel behavior. The built environment is examined in terms
of psychological and objective aspects of the quality of the walking environment that
might affect travel patterns. To examine the relationship between these two aspects of the
built environment and travel behavior, this dissertation analyzes people’s transit-use
patterns and choice of travel modes using the Los Angeles County sample which is a
subset of the data from the 2009 National Household Travel Survey California Add-on.
This study first examines whether the perceptions that low-income people have
about the walking environment affect their willingness to use public transportation by
analyzing self-reported frequency of transit use and measured neighborhood attributes. A
principal component analysis is used to reduce many overlapping perceptional variables
to latent factors that are used in subsequent models of transit use. Four perceptional
attributes are identified that affect regular transit use, i.e., physical safety, personal safety,
amenities, and perceived isolation. The results of this study show that unfavorable
perceptions of environmental conditions are independently associated with decreased
transit use; however, these effects vary among different types of neighborhoods. Notably,
low-income travelers view the conditions of their walking environment as problematic
more often than higher-income travelers, and it appears that their transit use is more
likely to be affected by safety concerns than by other urban design factors of their
neighborhoods. The findings suggest that safety concerns precede amenity concerns;
xii

therefore, enhancing the safety of neighborhoods is the necessary first step to increasing
the utility of transit for low-income people.
The second part of the empirical analysis is to test whether a compact
neighborhood design has any association with tour complexity and the choice of tour
mode. The study develops activity-based analysis that considers the linked nature of most
travel and tests the choice of tour mode in one integrated framework considering the
characteristics of the built environment at both the origin and the destination of a tour and
the detailed, mode-specific travel time for each leg of the trip within a tour. The results
suggest that neighborhood walkability has a strong influence on both distance traveled
and mode choice. People tend to economize their travel by traveling shorter distances, by
decreasing the total number of tours, and by reducing the number of trip chaining when
they live in a neighborhood with compact urban form characteristics. In particular, the
built environment has more effect on non-work trips in terms of encouraging active
transportation modes. Conversely, there is some evidence of induced travel demand
related to the trips in a compact built environment of the work location −hence the
evidence on reduced car travel for commuting trips is not as robust as the evidence for
non-work trips.

1

CHAPTER 1. INTRODUCTION

1.1. BACKGROUND AND MOTIVATION

The context of land use and transportation planning in California has changed
dramatically over the past few years due to the enactment of the California Global
Warming Solutions Act (AB 32) and State Bill 375, which set goals for the reduction of
statewide greenhouse gas emissions to 1990 levels by 2020. There was almost unanimous
agreement among policy makers and practitioners that the highest priority for achieving
this goal should be a reduction in vehicle miles traveled (VMT) per capita (Shaheen,
2009). Strategies for the reduction of VMT that are addressed most commonly by policy
makers generally involve smart growth and transit-oriented development. These land use
policies focuses on raising population densities and increasing the development of mixed
land use to facilitate the use of transit and to encourage non-motorized modes such as
walking and bicycles. However, it is recognized that this change in behavior is likely to
be very difficult to accomplish in the implementation of strategies for the reduction of
VMT.
Changing people’s behaviors is a complex undertaking. There are many factors
that can be motivators or barriers to the accomplishment of this goal, and the factors vary
depending on individual characteristics and preferences. Although many studies have
addressed the role of the built environment in explaining behavior, previous studies still
2

lack the sound methodological framework required for a study of the relationship
between the built environment and travel behavior. Major issues that can be found in
previous studies include: 1) lack of the development of appropriate conceptual models, 2)
lack of micro-level data that link the built environment and daily travel patterns, 3) lack
of the identification of key variables, such as psychological aspects related to the
environment, and 4) lack of the incorporation of the effect of activity patterns and
lifestyles.
To fill these gaps in previous research, this dissertation examines how the built
environment influences travel behavior by focusing on the roles of the perceived built
environment in transit use and the roles of the observed (i.e., directly measured) built
environment in the choice of tour modes.

1.2. RESEARCH STATEMENT

1.2.1. Access to Transit and the Built Environment

The analysis of data from the 2009 National Household Travel Survey-California Add-
On (2009 NHTS-CA) showed that 91.6% of transit users in Los Angeles County access
their transit stops by walking. The data further showed that people who walked to their
transit stops were more likely to use transit regularly than people who drove their cars to
the transit stops. This raises a question concerning whether the experiences of walking to
transit have an effect on people’s willingness to use public transportation. Previous
3

research on the determinants of transit use has addressed this issue and investigated the
impact that the ease of access to transit stops has on the use of transit (Agrawal et al.,
2008; Alshalafah and Shalaby, 2007; Cervero, 2001; El-Geneidy et al. 2010;
Loutzenheiser, 1997).
A large portion of the studies on access to transit systems has been focused on
how the built environment influences transit use. These studies have shown that
pedestrian access is related to the distance people must walk to get to the transit stop,
residential density, land-use mix, transit supply and other urban-design factors, such as
sidewalks and landscaping. However, only a few studies have investigated the perceived
environment empirically by focusing on transit passengers’ perceptions of the transit
environment (Loukaitou-Sideris, 1999; Wallace et al., 1999). Such studies have made a
vast improvement in our understanding of passengers’ perceptions of safety; yet, the
research results offered little insight into the role of perceived environment on actual
travel demand; The significant gaps in the research included determining the kinds of
perceptions that may prevent potential riders from using the transit system; determining
how the perceptions and their effects differ by the type of neighborhood, and determining
the perceptions of transit use among different racial and income groups.
In this study, three hypotheses are proposed to address these gaps. The first
hypothesis is that there are differences in the perceptions of the transit environment
between those who do not use the transit system, occasional users, and regular users. The
purpose of this study was to examine whether different perceived environmental
4

characteristics prevent people from using the transit system. Previous research has relied
on a one-day travel diary focused on the choices travelers made about their trips.
However, in this study, I made full use of available travel-behavior data to make a more
complete estimation of the actual market for public transit by accounting for people who
do not use the transit system every day or in a regular pattern. Thus, this research
attempted to examine the differences in perceptions of the transit environment among
three classifications of travelers, i.e., travelers who did not use the transit system,
occasional users, and regular users. Clearly, any relationships that can be identified
between traveler’s perceptions of the walking environment and their use of transit system
would be helpful in making policies concerning effective transit system.
The second hypothesis is that traveler’s perceptions of the walking environment
are related to the type of neighborhoods in which they live. Therefore, the different
perceptions in different types of neighborhoods would have a role determining the extent
of transit use in different types of neighborhoods. In this study, models were developed to
determine how the perceived environmental qualities actually affect transit use and to
determine the differences in these effects based on the different types of neighborhoods
after controlling for the characteristics of the travelers’ households and the quality of the
transit service.
The third hypothesis is that the association between the perceived environment
and transit use is stronger and more significant for low-income travelers than higher-
income travelers. If the elasticity of perceived environmental variable is statistically
5

significant, the built environment might play an important role in promoting transit use
among low-income people. As was identified in the descriptive analysis of the 2009
NHTS-CA data, low-income travelers viewed the conditions of their walking
environment as problematic more often than higher-income travelers. Consequently, low-
income travelers might encounter environmental barriers when walking to the transit stop.
In addition, other factors might impact their use of the transit system, including the
availability of transit service, the type of neighborhood in which they reside, and the
characteristics of their households. This study attempted to offer an accurate evaluation
of the use of transit by low-income travelers for comparison with the use by higher-
income travelers.
The approach that was taken should help policy makers better understand how
perception effects interact within the context of different neighborhoods. In particular,
our examination of these effects among low-income people should help researchers gain
an accurate understanding of transit-dependent markets. Thus, testing these hypotheses
will be useful for decision makers in their efforts to define and develop policy tools that
serve poor and transit-dependent populations more effectively.

1.2.2. Mode Choice and the Built Environment

The large body of research on travel behavior has established differences in travel mode
based on various types of land use. These studies have shown that travel behavior is
6

related to accessibility, density, mixed land uses, and other household characteristics
(Kockelman, 1997; Kitamura et al., 1997; Zhang, 2004). By comparison, there have been
very few research efforts that have investigated the relative impact of land use on mode
choice using the tour-based approach. The previous research related to mode choice has
made a vast improvement in our understanding of the influence of land use on travel
demand; yet this research failed to account for true behavioral causality, which can be
explained using the activity-based approach only when the linkages between trips are
properly incorporated into the analysis.
A particular interest of the second empirical study of this dissertation is complex
tour, that is, trip-chained travel connecting more than one activity. Previously researchers
have found that the complexity of travel constrains mode conversion, and complex tours
are more likely to be taken by private vehicle (de Nazelle, 2010). Similarly, Krygsman et
al. (2007) found a relationship between transport mode and chaining behavior, but
evidence of this relationship is indeterminant when location factors of intermediate
activities are considered within a tour. While in general, activity-based studies have
addressed this tour complexity issue, very few studies have actually used the tour-based,
mode choice framework to examine the relationship between mode choice, travel
complexity, and land-use factors.
As there is little empirical evidence about the link between land use and chain
behavior, this study focused on the effect of activity demand on the travel formation and
on the mode choice of trip-chained travel.
7

The three following hypotheses were proposed in this study. First, this research
hypothesized that the compact neighborhood design and household characteristics have
association with travel complexity and generation. In, this dissertation, walkability index
was used as the proxy of the compact urban form. Walkability attributes, in terms of
density, balance of land uses, street network patterns, and intensity of retail employment
are surrogate measures of land use. The purpose of this test was to determine whether the
residents economize their travel by traveling shorter distances, by decreasing the total
number of trips per day, and by increasing the opportunities for trip chaining when they
live in a neighborhood with compact urban form characteristics.
Second, when the cost and time of travel, the travelers’ socio-economic factors,
and tour characteristics are controlled, residential and destination walkability still matter
in choosing the travel mode. I assumed that most travelers might jointly consider both
origin and destination locations of the tour when they decide their travel mode. This
study tested the choice of mode in one integrated framework, considering both the built
environment at the origin and destination of the tour and the detailed level-of-service and
price for an individual trip within the tour. The test of this hypothesis will be helpful to
tell us how many more people will choose to travel by active travel
1
mode if we improve
an existing walking environment.

1
Active travel refers to destination-oriented travel by non-motorized mode (i.e., walking
and biking) and transit use would also count in this dissertation. Transit use is classified
as active travel because, in most case, it requires walking to and from transit system.
8

Third, the effect of walkability on the choice of tour mode varies by different
types of tours. Past research has shown that the choice of tour mode is affected by the
complexity of tours and by the purpose of the travel. It was hypothesized that
neighborhood walkability has a more significant and stronger influence on the choice of
mode in a simple tour than in a chained tour. In other words, land use might have less
influence on the formation of the set of choices of travel modes available to the traveler if
the tour consists of a sequence of trips. It also was expected that neighborhood
walkability has a stronger effect on the choice of tour mode for a non-work-related tour
than in a work-related tour. One can expect that increased walkability at the destination
of a non-work-related tour would result in a greater likelihood of choosing the active
travel mode. The outcome of the tests of these hypotheses will give us a comprehensive
understanding of the various relationships between the built environment and mode
choice for different types of tours. Testing these hypotheses will tell us whether land-use
policies can increase the use of the active travel mode by improving the ease with which
people can walk to gain access to their destination or a transit system.

1.3. ORGANIZATION OF THE DISSERTATION

This dissertation comprises one literature review and two empirical studies, some of
which have been already published in academic journals or presented in academic
9

conferences. Thus, each main chapter includes introduction and discussion that partially
overlap with corresponding chapters of dissertation.
The dissertation is organized in six chapters. Chapter 2 reviews literature on travel
behavior theories and empirical studies on how the built environment and travel
complexity affect travel demand. Chapter 3 elaborates research approach and data used in
this dissertation. In this chapter, I explain the conceptual frameworks that are used in two
main empirical studies of this dissertation.
The following two chapters are major empirical studies that test the research
hypotheses. Chapter 4 explores the determinants of transit use. Although much has been
written about local access to public transport, few studies have examined the role of
perceived environment in promoting transit by considering how the effects of these
perceptions differ by neighborhood type. This study examines how travelers’ perceptions
of walking environment affect their travel behavior by analyzing self-reported frequency
of transit use and measured neighborhood type.
Chapter 5 presents a study on the built environment impacts on mode choice of
the tour. Using a refined, tour-based, mode choice model, this study investigates how
neighborhood walkability is associated with mode choice and tour types. Finally, in
Chapter 6, I summarize major findings of the dissertation and discuss their contribution to
the literature and implications on policy and planning practice.
10

CHAPTER 2. LITERATURE REVIEW

2.1. THEORIES OF TRAVEL BEHAVIOR

This section reviews a number of theories from different disciplines that provide the
foundation for conceptual models describing travel behavior. I start with the utility-
maximization theory that has dominated the research in this field until recent years. I then
review the extensions of utility-maximization theory, which describes the behavior of an
individual faced with the multiple constraints of different social, economic, spatial and
personal circumstances. These theories provide important insights into the development
of conceptual models for the study exploring the relationship between the built
environment and travel behavior.

2.1.1. Utility-Maximization Theory

Travel behavior can be analyzed using utility-maximization theory from the field of
economics. This theory posits that individuals make rational choices based on utility—in
this case, the best tradeoff between the benefits and the costs of traveling to an activity
location. In this framework, most travel is a demand derived from travelers’ need to do
something at a location other than their current one. Typically, travel is considered
inconvenient and counterproductive, so it is assumed that people try to reduce the time
11

and cost associated with reaching their destinations for daily activities, such as work,
shopping, and school. Therefore, the theory traditionally has considered the minimization
of monetary cost and travel time (Domencich and McFadden, 1975; Ben-Akiva and
Lerman, 1985; Train, 1986; Maat et al., 2005; Blumenberg et al., 2007).
This concept of “generalized cost” has been extending the range of factors
included in the model as determinants that potentially influence the utility of travel
(Handy, 2005). Thus, the theory can predict behaviors that are economically rational
under a wide range of circumstances. For example, it can forecast the demand for
walking trips by specifying a systematic choice model, which hypothesizes that
environmental characteristics act as an incentive for walking travel. Besides the standard
measure of generalized costs, the theory accounts for the cost of one mode relative to the
cost of other modes (e.g., walking mode vs. driving mode) (Boarnet and Crane, 2001).
That is, the theory can accommodate the simultaneous choice of different modes that are
associated with travel time and cost in addition to various attributes influencing
individual’s travel choice and it assumes that the most advantageous mode will be chosen
for each time of a day.
Although utility theory of travel behavior has dominated previous research in this
field until recent years, the theory has encountered some criticism. First, travel is not
always a derived demand. It has been argued that people sometimes travel for the mere
sake of travel, which has “positive utility” (Mokhtarian and Salomon, 1999; McMillan,
2005). Redmond and Mokhtarian’s (2001) study demonstrated that commute time could
12

carry positive benefits; hence, travel is not necessarily always a source of “disutility” to
be minimized. Another empirical study consistently argued that travelers are not always
cost minimizers. For example, Mokhtarian and Salomon’s (1999) study showed that
people like to travel and most travelers are not cost minimizers in that they are actually
satisfied with their current commute time, which is longer than the ideal commute time
drawn from the theoretical expectations. In addition, the notion of rationality cannot
identify individual responses to choice situations limited by their constraints, self-
interests and psychological states, which are not revealed in the market. For example, if
people feel psychological attractions to adopting transit travel for any reason relating to
personal and contextual factors, they may use transit not because of travel time and cost,
but because of a greater enjoyment from reading a book in a train or habits from past
experiences (Brown et al., 2003; Matthies et al., 2002).
Although critics of the utility-maximization theory are increasingly numerous and
vocal, it is still the most robust approach in practice and is the basis of most models of
travel mode choice. Furthermore, the theory provides a variety of extensions to enhance
its applicability—such as the notion of value of time and activity-based approach, as will
be reviewed in following sections.

2.1.2. Value of Time

Value of time (sometimes called “value of travel time savings”) is a crucial parameter in
the analysis of travel behavior. Becker’s theory of the allocation of time (1965) provides
13

theoretical foundation of the study on the uses of time. The author considered time as a
resource and assumed consumers are subject to a time constraint. Thus, he incorporated
time into utility framework and suggested that “time can be converted into goods by
using less time at consumption and more at work” (Becker, 1965: p. 497). Under this
approach, individuals are assumed to organize their consumption of goods and their
allocation of activity time so as to maximize their utility. Following this logic, Becker’s
theory explains that people will decide how much to work based on their utility between
wage rate and leisure time. That is, marginal wage rate of an hour of work is what people
are trading off in terms of losing their one hour leisure; and thus, wage rate is the best
proxy for time value.
Taking this traditional approach, DeSerpa (1971) extended time-valuation
models to all activities. He was the first person who assigned time value for each activity
explicitly and proposed three types of the value of time. The first is the value of time “as
a resource,” which is a consistent view with Becker (1965). Total amount of available
time is fixed by the total time constraint. The second is the value of time “as a
commodity,” which refers to the value of time allocated to a certain activity. DeSerpa
(1971) viewed time itself as a possible source of utility, not a simple factor used in the
production of other goods. In other words, he noted the importance of substitution
between time and money, arguing that time and money are not in fixed proportions in
producing commodities. The third type is the value of saving time in a certain activity. In
example of mode choice, an individual would be willing to pay more money for saving
travel time by using faster mode (Hess et al., 2005).
14

These theories have shown the fundamental nature of travel mode choice chosen
in the functions of consumer production. Under this conventional approach,
transportation mode is assumed to be chosen based on travel time and goods (e.g., cost or
service) required for the trip to participate activities. It has been argued that the time cost
of travel varies by different socioeconomic and demographic characteristics of a traveler.
In other words, trip makers will compare alternative-specific cost of the goods and time
reflecting their personal value of time (de Donnea, 1972). For example, higher income
people tend to have a higher value of time and thus they are more likely to select the
faster mode and route to avoid travel time (Jou et al, 2005).
The concept of value of time has been used in formulating several empirical
analyses (Hensher, 2001; Greenwald and Boarnet, 2001; Lam and Small, 2001;
Brownstone and Small, 2005; Small et al., 2005; Tseong and Verhoef, 2008). Most of
these studies have consistently shown that the cost of travel time varies by different
activities and by different value of time for a person. Especially, progress has been made
in the area of congestion pricing. For example, Small et al. (2005) examined the
distribution of values of travel-time savings and reliability (i.e., “the predictability of
travel time”). They collected a sample of motorists who participated in a value-pricing
experiment and analyzed their choice between express and regular lanes as conditional
variations in tolls and other factors including residential location, travel mode, time of
day, and car occupancy. They found that motorists value differently their travel time and
reliability according to their preference variation. Tseng and Verhoef (2008) examined
time-varying values of travel time savings and schedule delay costs, using the data of
15

survey respondents’ departure time choices for the morning commute. They found that
value of travel time savings is strongly time-dependent (e.g., morning peak) and that
value of time varies by socio-economic characteristics. Findings reveal the higher income
groups tend to have higher value of travel time savings, such that they are willing to pay
more to avoid travel time, while the lower income groups face tighter scheduling
constraints. Although the results produced by a separation by gender are less clear-cut,
the study shows that women tend to typically have higher values of time than male
drivers, all else equal.

2.1.3. Constraint Approach

Advocates of the constraint approach (e.g., Hagerstrand, 1970; Kwan, 2000) have
suggested that individual travel behavior is not only based on utility-maximization
theory, but also constrained by spatial, time and other individual constraints. Hagerstrand
(1970) developed the concept of space-time constraint of human activity, based on the
idea that both space and time are scarce resources and constrain daily activity patterns.
He identified three major constraints described as “capability constraints,” “coupling
constraints,” and “authority constraints.” First, “capability constraints” are biological
needs-related (e.g., sleeping, eating) or ability-related (e.g., means of transport available).
For example, people’s activity is limited at regular intervals for the necessity of sleeping,
eating, and any other physiological reasons. In addition, people’s activity is restricted by
ability to use transportation technology. A person who is able to drive has less spatial or
16

time constraints than those who must travel on foot. Second, “coupling constraints” refer
to that a person needs to undertake certain activities at a particular place with other
people during a certain period of time. For example, daytrips of employed people are
restricted because they are required to undertake subsistence activities at their work
places during a day time of weekday. Another example is that parents need to adjust their
activity schedule for driving their children to school or picking them up. Third, “authority
constraints” are related to the “social, political, and legal restrictions” placed on the
subject’s activity (Hanson, 2004). For instance, a lower income household tends to have
access to “fewer or inferior domains” compared to a higher income household. If they are
not available to afford a rent close to a place of work, they cannot avoid long commuting
times, which will result in reducing available times for other activities (Hagerstrand,
1970).
As viewed in the space-time perspective, these three constraints (i.e. capability
constraints, coupling constraints, authority constraints) affect people’s travel behavior. As
“space-time autonomy” increases, people get greater accessibility to places and have
more freedom and time for activities (Hanson, 2004). In this vein, travel outcomes should
be interpreted within the context of space-time constraint approach.
The past accumulation of empirical findings has provided consistent indications
that coupling and resource constraints are the primary determinants associated with travel
behavior. Gender and employment status have been found as important factors that
influence activity-travel patterns. Kwan (1999) observed gender differences in space-time
17

constraints, using activity-travel data of European-American population. She compared
the space-time patterns of the out-of-home, non-employment activities of three groups:
women employed full-time, women employed part-time, and men employed full time.
While facing similar “time-budget constraints” (i.e. limited time available for other
activities), women employed full-time had less space-time flexibility for pursuing out-of-
home activities during the day, compared to the men employed full-time. As Kwan
(1999) found women tend to face higher level of space-time constraints, she further
examined the effects of “fixty constraint” on employment status and commuting
distances. The results revealed that there was a reciprocal causation between women’s
daytime fixty constraints and their employment status. That is, women with higher levels
of fixty constraints tend to have part-time job, whereas part-time employment status is
likely to increase the level of fixty constraints. For example, women employed part-time
tend to have a shorter commute distance than full-time employed women and men.
Similarly, Pas and Koppelman (1987) found that those who face more stringent coupling
and resource constraints tend to have lower variability in travel patterns, when compared
to people who are relatively free of such constraints. For instance, the influence of the
presence of children in the household on female’s travel behavior is stronger and more
significant than that of males, possibly due to the gender-related role expectations and the
scheduling constraints associated with children’s care. The study also found that
employed people have much lower variability in trip generation rates, compared to those
who are not employed.

18

2.1.4. Activity-Based Approach

Advocates of the activity-based approach (e.g., Bhat and Koppelman, 1999; Kitamura,
1988; Bowman and Ben-Akiva, 2001) has argued that individuals tend to optimize their
entire activity patterns, rather than just to maximize utility for separate travel choices
(Maat, 2005). While the trip-based approach provides a useful framework for a
straightforward means of predicting the outcome of a decision for mode choice by
examining each trip in isolation, the activity-based approach tries to illuminate the
complex interactions in travel behavior and activities, focusing on the sequential decision
structure (Bhat and Koppelman, 1999).
First, activity-based travel theory views most travel as a demand derived from
demand for activities (Bowman and Ben-Akiva, 2001). If trips are induced by needs for
activities, then the conventional trip-based approach which looks at each trip in isolation
will provide at best an imperfect setting for travel behavior analysis (Kitamura, 1988).
Extensive analysis has been made on the association between activity and travel patterns.
One aspect of activity patterns concerns the association between activity duration and
travel time. Schwanen and Dijst (2002) found that commuting time is related to duration
time for work activity. That is, people are likely to balance the commuting time and the
time spent at the employment place. Similarly, Kitamura and Fujii’s (1997) research
analyzed activity-travel patterns focusing on the relationship between trip generation and
commuting time, and the study found that reduction in commute time tend to increase the
number of trip chains for non-work activities after returning home from work. Several
19

other empirical studies on activity-travel patterns showed that destination and mode
choice are dependent upon the choice of a daily activity pattern (e.g., Kitamura, 1984 a;
1984 b). These empirical results call for more rigorous examination of the
interdependencies among activity patterns and travel patterns.
Second, the activity-based travel theory highlights changes in travel behavior
over time. The advocators of the activity-based approach have asserted that people’s
activity-travel patterns are likely to be affected by temporal-spatial constraints, because
they tend to perform different activities in different spaces at different points in time
(Hagerstrand, 1970). Thus, individual travel behavior is the interaction between the
motivations of activity participation and the constraints on activities within time-space
lens (Bowman and Ben-Akiva, 2001).
Empirical investigation has been made in the area of constraints on activity and
travel behavior. Analyses of multi-day travel behavior have shown that individual travel
behaviors vary across days (Pas and Koppelman; 1987; Kitamura, 1988). Hanson and
Huff (1981) measured travel-activity patterns and compared patterns from different days,
using multi-day behavior data. They found a noticeable amount of variability in
individual daily travel patterns, although there was a great deal of repetition for a full-
time employed person on weekdays. Pas and Koppelman (1987) looked specifically at
the day-to-day variability in individual trips in order to investigate the impact of
constraints on activity on travel behavior. The findings of their study showed that
employment status, presence of children, and other relevant resources together influence
20

day-to-day variability in individual travel behavior. These empirical studies demonstrate
that analyzing people’s travel behavior for a single day cannot explain variations in their
travel patterns; and therefore understanding people’s complex travel-activity patterns
necessitates an examination of the day-to-day variability in travel (Hanson and Huff,
1981).
As reviewed above, all of these travel behavior theories shed light on the
important aspect of travel decision. An ideal approach to analyze travel behavior would
account for all these notions of utility, value of time, activity, and constraint approach.
Accordingly, travel behavior can be explained as the outcome of the interaction between
utilities associated with demand for activities, opportunities that activity sites provide,
and space-time constraints on activities.

2.2. THE ROLE OF THE BUILT ENVIRONMENT IN TRANSIT USE

Theories suggest that the choice of transit use can be influenced by travel times and
personal preferences, characteristics, and other constraints. Time cost is related to the
quality of transit service that includes in-vehicle time, transit access, transit transferring
and transit waiting time. However, the differences in travel time cannot be completely
explained by differences in transit service quality. The cost of travel time is a generalized
time cost from the person’s origin location to all possible destinations. Thus, it has to do
with accessibility, which can be influenced by the characteristics of the built environment
21

near both the origins and the destinations of trips (Boarnet and Crane, 2001; Maat et al.,
2005).
From the perspective of utility theory, the built environment can affect travel
behavior in two ways. First, the built environment has to do with the spatial distribution
of the activity places people encounter. A greater concentration of destinations induces
travel because activity places are opportunities that can satisfy typical needs such as
work, shopping and leisure. Second, the built environment affects the ease of access to
destinations. By improving the spatial constraints along the access routes to destinations
(e.g., transit stops or potential destinations), travelers can reduce their travel costs. For
example, more direct and comfortable walking paths connected to transit stops can
increase the utility of travel, thus potentially influencing the choice to take transit. In this
context, increasing appropriate access to transit systems has been seen as a means of
attracting more pedestrians to the transit system.
The majority of studies on transit access have focused on defining the walking
catchment areas—the areas where potential riders live and work (Agrawal et al., 2008;
Alshalafah and Shalaby, 2007; El-Geneidy et al., 2010; Jiang et al., 2012; O’Sullivan and
Morral, 1996). In general, catchment areas for bus stops and rail stops are considered to
be ¼-miles and of ½-miles from the stops, respectively. Most of the pertinent studies
have indicated that, in addition to the distances from transit stops, the built environment
that exists between the potential user’s location and the transit stop is a key factor that
affects use. The phrase “built environment” encompasses many factors, including
22

residential density, land-use mix, street connectivity (Cervero and Radisch, 1996;
Cervero and Kockelmann, 1997; Crane, 2000; Cervero, 2001; Loutzenheiser, 1997;
Messenger and Ewing, 1996; Moudon et al., 1997; Ryan and Frank, 2009), the condition
of the sidewalk, route characteristics (Kitamura et al., 1997; Hess, 2009; Boarnet and
Crane, 2001; Greenwald and Boarnet, 2001; Saelens et al., 2003; Leslie et al., 2005), and
other urban-design factors, such as barriers to car use and the presence of attractive
buildings, trees, and landscaping (Agrawal et al., 2008; Estupinan and Rodriguez, 2008).
These objectively measured environmental variables are important in
understanding what determines an individual’s decision about travel. However, some
recent literature on physical activity has discussed whether the relevant analyses should
focus on people’s perceptions of the environment or on objective measurements of the
built environment (Wen et al., 2006). This literature has shown that a traveler’s
perceptions of the environment have a strong effect that often limits the practical range of
individual choice (Mota et al., 2005; Evenson et al., 2007). For example, King et al.
(2000) found that the lack of a safe place to exercise was the top-ranked barrier to
physical activity for specific groups of minority women, e.g., African American women;
also, personal safety was the major concern of elderly people in their physical activity.
Similarly, Loukaitou-Sideris (2006) found that women and elderly people are more likely
to feel afraid and that they are at risk in neighborhoods with high levels of inactivity, but
despite their fear, they have to walk anyway for utilitarian purposes.
23

Although researchers have studied the effect of people’s perceptions of the
neighborhood environment on their physical activity (Evenson et al., 2007; Gidlow et al.,
2010; King et al., 2000; Leslie et al., 2005; Mota et al., 2005), their findings have seldom
been reflected in the design and development of public transportation systems or people’s
access to them. Only a few empirical studies have addressed travelers’ perceptions as
significant factors that influence the use of transit systems. One study conducted a survey
in which 328 pedestrians were asked how they chose the route they used to access rail
transit stations in the San Francisco Bay Area and in Portland, Oregon (Agrawal et al.,
2008; Stradling et al., 2007). This study showed that the primary factors that influenced
route choices were minimizing time and distance, safe areas away from busy streets,
sidewalks that were in good condition, and aesthetic issues, such as the presence of
attractive buildings, trees, and landscaping. Jiang’s study (2012) found that more direct,
comfortable, and pleasant walking corridors are associated with increased walking
distance to transit stations. This result demonstrates that improved spatial constraints
reduce transit users’ travel costs by lowering their real or perceived walking access time
to stations.
Past research has, therefore, highlighted the potential importance of the perceived
environment’s influence on transit use. Yet many questions remain about the effect of
that perception on transit demand. First, most of the pertinent studies have relied on
internal surveys of transit passengers, and have been mainly concerned with how riders
perceive their transit environment once they have decided to take public transportation;
these studies have rarely accounted for the role of perceived environment in promoting
24

the choice to take transit. Additionally, although some of these studies (Loukaitou-Sideris
and Fink, 2009) have addressed gender differences in perceptions of safety on transit, no
research has focused on low-income travelers’ perceptions of access environment and
transit use. Examining these effects among low-income groups helps researchers to
understand transit-dependent markets correctly, given that low-income travelers have low
mobility and often rely on public transit as a primary means of transportation.

2.3. THE ROLE OF THE BUILT ENVIRONMENT IN THE CHOICE
OF TRAVEL MODE

Extensive research has been conducted on disaggregate mode choice models, but the
majority of such models involve trip-based analyses, and they rely heavily on traditional
utility-maximization theory, which considers trips discretely. The discrete-choice model
provides a useful framework for the straightforward prediction of the outcome of a mode
choice decision under the irrelevant independent alternatives (IIA) property assumption,
i.e., that the relative probabilities of choosing alternatives are independent across other
alternatives. Advocates of the activity-based approach argue that individuals tend to
optimize their entire activity pattern, rather than just maximizing the utility for separate
travel choices (Bhat and Frank, 1999; Kitamura, 1988; Bowman and Ben-Akiva, 2001).
These theoretical notions imply that a simple distance and trip-oriented analysis based on
a utility-maximization concept cannot account fully for the complexities of travel
behavior, while the activity-based approach can explain more clearly the ways travel
25

decisions are actually made by considering activity schedules instead of trips (Maat et al.,
2005).
Most tour-based, mode choice models have been developed by focusing on mode
attributes (i.e., travel time and cost) and the traveler’s demographic and socioeconomic
characteristics (Miller et al., 2005). Land-use attributes often have been left out of the
mode choice equations under the assumption that the attributes of land use are related to
people’s long-term, residential-location decisions, which are fully captured in travel times
and costs (Zhang, 2004). Recently, it has been recognized in a few studies that these
omitted land-use variables could bias model parameters, so they have included these
variables in the tour-based, mode choice framework to investigate the impact of land-use
policy on travel behavior.
Frank et al.’s study (2008) is one of the few studies that have investigated the
effects of land-use patterns on mode choices using the tour-based approach. Their study
examined trip-chaining mode choice in the Central Puget Sound region and found that
land-use characteristics of work places were important factors in determining mode
choice for travel to work and mid-day travel. The results of this tour mode choice study
demonstrate that true behavioral causality can be explained adequately by the activity-
based approach only when the linkages between trips are properly incorporated into the
analysis. Another empirical study that estimated the effects of land use on tour-based
mode choice was conducted to analyze traveler’s chain behavior in Portland, Oregon. The
research showed that various land-use policies had a marginal influence on regional
26

travel and that activity-based models have an advantage over trip-based models (Shiftan,
2008).
A limitation of the above studies is that the analyses did not explicitly address
how the land-use factors are related significantly to the different levels of tour
complexity. Frank et al. (2008) controlled for tour complexity when testing the
relationship between land use and mode split. From the analysis, the authors found that
land use attributes are associated with the frequency and complexity of tour patterns, but
they did not test the variations of land-use effects on mode choice for different tour types.
Another limitation of the previous study is related to the lack of comprehensive, micro-
level data linking land-use attributes and travel-behavior outcomes. Most of these studies
neglected the role of the destination environment for non-work activities. In addition,
many of the previous studies did not provide detailed level-of-service and price variables
of each mode even though they are essential for a mode choice study.

2.4. DISCUSSION

This chapter reviewed the theoretical and empirical literature on the nature of travel
behavior. As reviewed from the literature, travel outcome reflects the individual’s activity
demands, spatial opportunities, and space-time constraints (Giuliano, 2003). First, taking
the utility perspective, individuals are not primarily interested in just minimizing travel
costs, but, rather, in maximizing utility. Thus, improving proximity to transit station is
not an only determinant of transit mode choice. If the more variety is offered by a more
27

distant destination accessible by a car, a person may drive to a more distant shop in order
to get a cheaper and higher quality of product. Second, the costs of travel time vary by
different trip purpose and by different value of time for a person. For example, higher
income groups may be willing to pay more to avoid travel time by using faster modes.
Third, individual travel behavior is not only based on the cost and benefit of travel, but is
also constrained by social, political, and resource restrictions. Lower income groups may
face more stringent constraints on their choice of living locations and transport modes,
and thus they may have more difficulty to optimize their travel choices. Fourth,
individuals try to schedule activities in a daily pattern based on the consideration of time
available for activity participation and the interaction with the next activity. Thus,
individuals may choose their mode choice depending on their entire activity patterns,
rather than for the utility of the separate trips.
In summary, travel mode choice does not just have to do with the built
environment, but it has to do with what kind of trips people are taking, what their
resources are, how much they value travel time savings, what physical, psychological
states they are, what social, political constraints they face and so on. Travel outcome is
the result of the all of these factors. Therefore, an ideal measure would control all of these
factors to isolate the impact of the built environment on travel demand. Unfortunately, in
reality, there is no data available to build such a measure.
However, a number of additional developments could be made to better represent
the impact of the built environment. Generally, the measurable role of the built
28

environment on mode choice is expected to be small (Hanson, 1982; Kockelman, 1997).
However, the magnitude of the effect of the built environment may vary depending on
trip purpose (e.g., commute, non-work trip), on service areas (e.g., urban, rural), and on
different demographic characteristics (e.g., gender, race). Separate models for various
sub-groupings might provide us a clearer picture of the effect of the built environment on
travel demand.
Most importantly, the scale of the data on the built environment should have
relevance to the spatial scale that individual travel decisions are actually made. The scale
of the neighborhood data as well as the scale of the regional data must be considered,
because, for example of mode choice, much transit access activity is undertaken near
individual’s home and workplace and transit activity is reflected by regional accessibility.
In addition, the accurate location information on the home and workplace would more
sharply represent the effect of the built environment on mode choice.
Although objectively measured environment variables are important to include in
a model, individual’s perceptions of environment may be a more direct modifier which
limits the practical range of individual choice. The subjective factors, such as perceptions
of neighborhood safety with respect to street crime and traffic danger, and the quality of
amenities (e.g., pleasant and comfortable environment for pedestrian), are important
factors that complete the objective measures of the built environment (Loukaitou-Sideris
et al., 2001, 2006; National Research Council, 2005).
29

In addition, personal attitudes and preferences are critical factors to influence
travel behavior. Some empirical studies commonly argue that individual’s personality,
lifestyle preferences, and travel-related attitudes are more influential factors in travel
decision than land use configuration (Clay and Mokhtarian, 2004; Cao and Mokhtarian,
2005). The results of these studies imply that the association between the built
environment and travel behavior is insufficient to establish causality. Therefore, it is
important to control travel-related attitudes with person’s socio-demographic
characteristics that may impact his/her selection of neighborhood location.
Finally, the majority of studies exploring the relationship between the built
environment and travel behavior have used trip-based approach, although these trip-
oriented analyses cannot fully account for the complexities of travel behavior. To address
this concern, the latter part of this dissertation focuses on activity-based approach, and
develops a comprehensive model that examines the effects of the built environment on
mode choices by looking at activity patterns over the day instead of by looking at trip
patterns. This approach will provide a holistic view to analyze travel and explain how the
built environment influences traveler’s behavior by determining constraints and
opportunities for activity participation.

30

CHAPTER 3. RESEARCH APPROACH, DATA, AND
MEASURES

Upon review of the gaps in the previous studies, I examine the impact of the built
environment on travel behavior in two ways: 1) in terms of transit use; and 2) in terms of
mode choice of complex tour. This chapter presents research approach and data used in
each empirical analysis of this dissertation.
3.1. RESEARCH APPROACH

3.1.1. Determinants of Transit Use

The debate over how the built environment affects transit access activities has recently
been reinvigorated (Agrawal et al., 2008; Cervero, 2007; Werner et al., 2009). Walking
catchment areas are often defined as the area from which potential riders are drawn. A
quarter mile around bus stops and a half mile around rail stops are commonly used to
identify walking catchment areas. These simplified catchment areas assume that all
transit users and trips are similar so that people are willing to use transit system if a stop
is built within a given walking distance (O’Sullivan et al., 1996). However, only creating
transit stops does not guarantee that nearby residents will take public transportation.
Beyond the distance to a transit stop, in reality there are many reasons that people do not
take public transportation. These decisions are based on a set of factors relating to
individual mobility, time and cost of service, reliability of transit service, and travel
31

barriers along pedestrian routes. Among these factors, pedestrian access has been
identified as an important factor for integrating land-use with transportation planning
(Hsiao et al., 1997).
Among the studies on pedestrian access to transit systems, a large body of
research has shown that transit use is related to walking distance to stops, residential
density, land-use mix, transit supply and other urban-design factors such as sidewalks and
landscaping. By comparison, few studies have examined the role of perception. The first
empirical study of this dissertation examines how the built environment influences the
use of public transportation, with particular emphasis on the role of perceptions of the
walking environment near household location. To frame empirical explorations in this
study, I develop a simple conceptual model of transit use.
Figure 3. 1 illustrates that the decision to take public transportation is based on a
set of factors relating to transit service, neighborhood type, traveler’s demographic
characteristics and perceptions about the environment. Personal- and household-specific
characteristics such as age, race, and income can influence transit use. Transit
accessibility is also an important factor that directly influences transit use. Transit
accessibility can be measured using both the transit supply through the transit route
network and the service quality of the routes. The attributes of neighborhood
environment constitute a third factor that can determine the possibility of using transit
and traveler’s perception about the environment, so play an important role in transit use.
Finally, this study focuses on perception factors that directly influences travel behavior.
32

Perception factors include convenience, esthetics, and safety concerns such as whether
people perceive nearby paths or trails, whether street-crossing is seen as safe, whether
people perceive nearby shops or interesting places to go on the way to access transit stops,
whether there is a fear of street crime, whether they perceive any difficulty in walking
that is caused by air pollution, weather, hill and so on.

Figure 3. 1. Conceptual model of the study: factors influencing transit use

3.1.2. Mode Choice Analysis: Complex tour

In the second empirical study of this dissertation, I attempt to address the gaps of existing
literature by using a refined, tour-based, mode choice framework to investigate how the
built environment is related to the choice of tour modes and tour types. Importantly, the
Individual Constraints
Neighborhood Type
Density
Land Use
Housing Structure
Crime rate
N
Perception of
Environment
Safety
Amenity
Pedestrian Accessibility
Transit User Type
Transit Accessibility
Network Connection
 Transit Service Quality
Socio‐demographic/
economic Characteristics
P
T
S
U
Environmental Constraints
Non user
Occasional user
Regular user
Age
Race
Income
33

unit of this analysis is a tour, which is defined as a ‘home-to-home loop.’ Based on the
previous literature, the author formulated the following three hypotheses: 1)
Characteristics of the built environment and household structure have association with
tour complexity and generation; 2) The built environments at trip origin and destinations
affect the mode that is chosen for a tour, when travel costs and time, travelers’ socio-
economic factors, and tour characteristics are controlled; and 3) The effect of the built
environment on the choice of tour mode varies by tour types and purposes.
These hypotheses are based on the research that examines the relationship
between travel-mode choice and land use at both the origin and the destination of a tour.
In general, residents are more likely to use transit or non-motorized modes if they live in
a highly-walkable, and mixed land-use area. The main focus of this paper is to determine
whether this pedestrian-friendly, community design encourages individuals to choose
active travel mode −i.e., transit, walking, and biking and how these travel patterns are
affected by different types of tours.
Figure 3. 2 shows the conceptual model of the study. This study builds on
Bowman and Ben-Akiva’s (2001) study that views tour patterns are conditioned by the
selection of the day’s activity scheduling. Activity scheduling involves a set of choices
including the choice of a certain activity agenda, the choice of activity priority, the choice
of arranging activities in a sequence, and the choice of activity locations and times spent
in those locations. In the reverse way, the choice of activity patterns is also affected by
34

available tour alternatives. The propensity to chain multiple trips in a single tour may
alter the individual activity patterns.
These activities and tour patterns together directly influence the mode choice of
the tour. This study assumes that individual activity agendas and tour patterns are
predetermined and then mode choice is formed. One may argue that mode choice may
precede and lead to form tour patterns. The findings from previous literature showed
bidirectional causality between tour complexity and mode choice, but the literature found
that the marginal effects of complex tour on mode choice is much bigger than the reverse
one (Ye et al., 2007). This supports the notion that tour complexity contributes to the
choice of tour mode. In addition, such a model would be more appropriate to explain
behavioral changes that may be associated with a system change (e.g., the improvement
of the built environment). If we assume the opposite causal structure is valid, that is, if
we assume tour formation is driven by mode choice, then the model can easily
overestimate the effects of the built environment improvement by altering the nature of
individual tour formation correspond to the mode shift (Ye et al., 2007).
Upon reviews of the previous literature, the study also assumes that personal and
household characteristics influence activity and tour patterns, and tour mode choice. The
variations in tour mode share are tested according to the variations in individual and
household characteristics. As Hagerstrand (1970) and Roosenbloom (1989) pointed out
gender differences in the travel patterns, this study hypothesizes that there are differences
in terms of the choice of trip chaining and travel modes by different personal or
35

household responsibilities. Other personal and household-related attributes that must be
explored in a model include age, race, income, family life cycle, and possession of driver
license.
The primary purpose of this study is to test whether the built environments at tour
origin and destination play an important role in travel mode share to perform activities.
As shown in Figure 3. 2, this study assumes that travelers jointly consider both origin and
destination locations of the tour when they decide their travel mode; and thus the study
develops one integrated framework considering the characteristics of the built
environment at both tour origin and destination and the detailed level-of-service and price
for an individual trip within the tour. This approach will be useful to provide information
on how many more people will choose to travel by transit or non-motorized mode if we
enhance the quality of the built environment at the local level.
The second purpose is to identify tour and activity attributes that influence mode
choice. If the built environment influences mode choice, are the influences different for
different types of tours? To answer this question, I establish the mode choice models to
analyze different types of tours. First, two separate sub-models are developed for the
allocation of work activity since there may be differences between people who are
conducing only work activities and those who are combining work activities with non-
work activities. The identical structure is used to develop sub-models for non-work-
related tours.

36

Figure 3. 2. Conceptual model of the study: factors influencing tour mode choice

3.2. DATA AND MEASURES

3.2.1. Data

This paper uses the 2009 National Household Travel Survey-California Add-On (NHTS-
CA), collected from April 2008 through May 2009. The NHTS-CA is a household-based
travel survey that the Federal Highway Administration conducted via computer-assisted
telephone interviews. Households were randomly selected for participation. The survey
data include four datasets: household, individual, and vehicle information, plus a one-day
travel diary for each person. Types of the data collected provide information on the
Activities
Tour Patterns
Tour Mode Choice
Household/Personal
Characteristics
Destination
Environment
Home
Environment
37

demographic data on household, vehicle, and individuals, and detailed information on
daily travel that was conducted by all modes of transportation including transit and non-
motorized mode. Each household was randomly assigned the “travel day,” and was asked
to record all travel by household members over a 24-hour period. Then interviewers
followed up with a phone call to collect the detailed information about the travel of each
household member. “Travel day” was assigned for all seven days of the week. The data
from travel diary contain information on the miles of travel, trip purpose, mode, transit
time, trip length, and the location and the time trip begins and ends and other related
attributes of daily trips taken within a “travel day” (Santos et al., 2011).
The NHTS-CA also contains information on usual travel behaviors not collected
from a travel diary. The data contains information on usual commute mode, distance to
commute, frequency of transit use during the past month, vehicle mileage during the past
year, and total time spending walking/biking in the past week.
The individual is the unit of analysis for this dissertation. Personal variables of
travel behavior originate from the Los Angeles County sample, which is a subset of the
NHTS-CA. The original dataset of the Los Angeles County sample include 2,989
surveyed households and 5,861 persons. The first empirical study of this dissertation
analyzes people’s usual travel behavior regarding public transportation by using
frequency with which a person uses public transit during the past month as a key variable.
The second empirical study of this dissertation examines people’s choice of travel mode
by using travel diary data collected within a specific 24-hour period, their travel day. By
38

examining these two aspects of travel behaviors, this dissertation presents a more holistic
view of the actual travel patterns, using both the information from usual, general
observation as well as the information from one day observation.
The confidential NHTS-CA data provide information about home and visited
location with the coordinates (latitude and longitude) of surveyed respondents. With the
use of the point-level location info, trip origins and destinations were geocoded in the
geographic information system (GIS). Land-use and transit-service attributes were
assigned to the geocoded location of every household’s home and visited location. All
environmental variables were measured within a ¼-mile Euclidian buffer around each
geocoded location using GIS. The land-use and transportation attributes used in this study
came from two archived sources: census files and the Southern California Association of
Governments (SCAG)’s roadway network and land-use data for the year 2008.

3.2.2. Measures of the Built Environment

The built environment can be broadly defined as human-made spaces that we live and
work in. This dissertation builds on Handy (2005) that define three attributes of the built
environment: land use patterns, transportation systems, and design features. Land use
patterns refer to the spatial arrangement of land (i.e., location where human activities take
place) devoted to different purpose, for example, industries and residential homes.
Transportation system refers to physical infrastructure and the services that provide the
fluid network system and connectivity among activities. Design refers to the aesthetic
39

dimension of the urban environment including building and street design, and natural
landscape. The quality of urban design has been addressed that relates to pedestrian’s
perceptions in terms of sense of comfort, sense of interest, and sense of safety. This study
uses perceived measures of the environment as the proxy of the design factors. These
measures include perceptions about safety, traffic, people, park, store, sidewalk, and path.
In this study, the characteristics of the built environment proposed to be relevant
to active travel − i.e., transit, walking, and riding a bike are defined, including population
density, walking distance to the transit stops, land-use mix, street connectivity, and the
intensity of retail employment. The next section explains how these variables were
measured.

Land Use Mix
Land use mix variable was measured with an entropy measure using equation (1). Six
types (e.g., residential, commercial, office, industrial, educational, and open space uses)
of land development were measured. This widely used metric presents the level of
integration of different types of land uses, such as residential and commercial uses. The
value of higher mixed diversity will be close to 1, and zero indicates single land use
within a ¼-mile radius around each household location.
Land-use mix diversity = - ∑
· J
J

(1)
Where,
40

P
j
= proportion of land development type of the j
th
parcel; and
J = number of different types of land development type

Figure 3. 3. Measuring landuse mix within a ¼mile area around each household location

Walking Distance
Walking distance from each household to the nearest bus stop was measured as an
indicator of pedestrian accessibility. It was calculated following the street network by
using ArcView Network Analyst in the GIS. In the second empirical study of this
dissertation, walking and bicycle travel times were estimated based on network distance
between origin and destination, applying the average walking speeds of 1.22 meter per
second (m/s) and the average cycling speeds of 5.36 m/s (Rodriguez and Joo, 2004).
1/4 mile
41

Figure 3. 4. Measuring walking distance from the household location to the nearest bus stop

Street Connectivity
Research in planning and urban design has addressed that a more connected street
network system potentially influences active travel by providing travelers with a greater
number of route options, by providing with more direct routes to destinations, and by
making walking and riding a bicycle more feasible. The grid pattern is viewed as the
archetype of a highly-connected street pattern. In that regard, this study uses the measures
of intersection density to measure street connectivity. The number of intersections (4-way
nodes) within a ¼-mile area was calculated around each household location. Multiple
steps were taken to measure and improve accuracy of the data. Since the study focuses on
42

only local street network that are possibly used by pedestrians, all other non-local streets
were removed from the dataset including highways, freeways, expressways and access
ramps. Nodes were created at the 4-way intersections using the data of centerlines of
street network, after removing the nodes created at the street segments that do not directly
connect to other portions of local street network.

Intensity of Retail Employment
Retail stores and commercial activities within walking distance have been identified as
one of important factors believed to stimulate pedestrian activity. To measure the
intensity of retail stores and commercial services in the study area, this study used the
employment data by industry sector based on the NAICS classification code from
InfoUSA 2008. Using the geocoded location of the business establishments, this study
mapped the distribution of retailers and calculated total number of retail employment
within a ¼-mile radius around each household location. The retail trade sector includes
establishments engaged in retailing merchandise and providing after-sales services (e.g.,
automobile dealers, supplies stores). This value of retail employee density was used to
measure walkability variable in this dissertation, instead of using retail floor-area ratio
that has been included as one of components of the walkability index in previous studies
(Frank et al., 2010; Sundquist et al., 2011).

43

CHAPTER 4. PERCEIVED BUILT ENVIRONMENT
AND TRANSIT USE IN
LOW-INCOME POPULATIONS

4.1. INTRODUCTION

The analysis of the data from 2009 NHTS-CA (2009 National Household Travel Survey-
California Add-On) shows that the majority (91.6%) of transit users, in Los Angeles
County, access their transit stops by walking and they are likely to use transit more
regularly than those who drive to the stops. This raises the question: do the experiences of
walking to transit have an effect on people’s willingness to use public transportation?
Previous research on the determinants of transit use has addressed this issue and has
investigated the ease of access to transit stops and its impact on transit choice (Agrawal et
al., 2008; Alshalafah and Shalaby, 2007; Cervero, 2001; El-Geneidy et al. 2010;
Loutzenheiser, 1997).
A large body of research on access to transit systems has focused on how the built
environment influences transit use. These studies have shown that pedestrian access is
related to walking distance to stops, residential density, land-use mix, transit supply and
other urban-design factors such as sidewalks and landscaping. By comparison, only a few
studies have empirically investigated perceived environment, by focusing on transit
passengers’ perceptions of the transit environment (Loukaitou-Sideris 1999; Wallace et
al., 1999). Such literature has failed to address the role of perceived environment in
promoting transit use, with significant gaps in the research: what kind of perceptions may
44

prevent potential riders from taking transit; how the effects of perceptions differ by
neighborhood types, given differences in neighborhood context; and how the effects of
low-income travelers’ perceptions differ (or not) from those of higher (not low)-income
travelers.
To address these gaps, this study aims to explore the role of perceived environment
in influencing the choice to take transit by considering both users’ and non-users’
perceptions and other individual characteristics. I undertake this analysis in two income
groups: low-income and higher-income travelers. In this paper, “low-income population”
is defined using HUD’s (Department of Housing and Urban Development) definition of
low income, which is adjusted both for household size and geographic region. HUD
defines low income as less than 80% of the median household income for the county or
metropolitan area. Using this definition, 43.8% of the respondents of the sample of Los
Angeles County that is a subset of the 2009 NHTS-CA are classified as low-income; the
remaining 56.2% are classified as higher-income. Despite the comparable sizes of the two
samples, these two income groups differ widely in their transit use, as expected. Regular
transit users who used transit at least four times a month make up a much higher
proportion (16.4%) of the low-income group than of the higher-income-group (8.3%)
(Figure 4. 1). Thus, these two groups serve as two comparative cases to help us answer
the following questions:
1) Do perceptions about walking environment play a role in transit use if we
control for differences in neighborhood type, transit service, and socio-economic
status?
45

2) Given that perceptions vary by neighborhood type, do those variations have an
impact on transit use?
The approach presented helps policy makers better understand how perception
effects interact within the context of neighborhood. In particular, examining these effects
among low-income groups helps researchers to understand transit-dependent markets
correctly. Thus, the answers to these questions will be useful for decision makers to find
policy tools that more effectively serve the poor and transit-dependent populations.

Figure 4. 1. Transit use patterns by income groups

46

4.2. DESCRIPTIVE ANALYSIS

4.2.1. General Patterns of Transit Use

The 2009 NHTS-CA data were used to examine the roles of perceptions in transit use.
The 2009 survey of Los Angeles County included 2,989 households and 5,861 persons.
Among the 5,861 travelers, 321(5.48%) used public transportation, and 294 (5.02%)
travelers accessed to transit stops by walking during the reported travel day. In this paper,
an “occasional user” is defined as a person who uses transit at least once per month, and
the “regular user” refers to a person who uses transit at least four times per month. A
“non-user” is a person who never uses the public transportation system or uses it less
frequently than once a month. About 20% of the entire respondents in Los Angeles
County use public transportation at least occasionally and 11% of respondents use it
regularly.
Figure 4. 2 shows that the majority of transit trips are accessed by walking
(85.3%), while the other access mode share (e.g., private vehicle and bike) is 8.3% in Los
Angeles County. In Figure 4. 3, the data further show that people who walked to their
transit stops were more likely to use transit regularly than people who drove their cars to
the transit stops. Interestingly, hardly anybody access transit by bike, but those who do
are very frequent transit riders. It might be the result of transit dependency due to the
limited resources (e.g., car) or the trip purpose (e.g., commute trip). If these people have
no other choice, they will use transit even if it is perceived to be uncomfortable and
47

inconvenient. Given the fact that the most transit users access stops by walking and they
are likely to be a regular users, it’s reasonable to consider whether the experiences of
walking to transit have an effect on people’s propensity to use public transportation.

Figure 4. 2. Access mode to get to public transit
(Source: the sample of Los Angeles County, 2009 NHTSCA)

Figure 4. 3. Number of transit use per month by access mode
(Source: the sample of Los Angeles County, 2009 NHTSCA)
48

4.2.2. Perceptions about Walking Environment

Attitudinal and perceptual data can be used as proxies for socio-psychological
factors that influence travel activity behavior. Personal variables of perception come from
the sample of Los Angeles County that is a subset of the 2009 NHTS-CA. The NHTS
looked at people’s perceptions of their neighborhoods’ problems regarding walking. The
respondents were asked about whether they consider a certain factor as a problem that
keeps them from walking in their neighborhoods. Since this study focuses on perceived
local environment, certain answers related to the personal business (e.g., too busy, things
to carry) were excluded, leaving 13 environmental factors related to barriers to walking.
Therefore, this study used 13 perception variables that indicate whether or not each
respondent cares about the given environmental factor; these variables were recoded as 1
for “no”, 2 for “neither mentioned”, and 3 for “yes.”
Figure 4. 4 shows the results on measures of answers to this question. “Not
enough light at night” was the biggest problem following the “no shops or other
interesting places to go.” Slightly more low-income persons gave these reasons than any
other. “Fast traffic” and “fear of street crime” were the third- and fourth-ranked reasons
for not walking, with more than twice the percentage of low-income respondents citing
these reasons as compared with higher-income respondents. In most cases, low-income
travelers viewed the conditions of their walking environment as problematic more often
than higher-income travelers did. This makes sense given that low-income residents are
more likely to live in dangerous places if they cannot afford to rent in a safe area. A
49

somewhat larger proportion of low-income people also expressed amenities concerns
about a lack of shops, parks, and paths nearby and considered each of these a serious
deterrent to walking.
Above descriptive statistics show that perceptions that low-income people have
about the walking environment are more negative than that of higher-income people. To
further investigate to what extent these negative perceptions that low-income travelers
have about their walking environment affect their propensity to use public transportation,
I develop the analysis framework and explain the model approach in the next section.

Figure 4. 4. Ranked order of the reasons people in Los Angeles County do not walk more
(Source: The sample of Los Angeles County, 2009 NHTSCA)
50

4.3. METHODOLOGY

The NHTS-CA contains information about usual travel behaviors. This study uses the
frequency with which a person uses public transit as a key variable to examine that
person’s usual behavior regarding public transportation. Most previous studies of the
demand for public transit have analyzed the trip-choice behavior of travelers, focusing on
how travelers choose transit among other available modes, but relatively few studies have
examined the frequencies with which these travelers use a public-transit system
(Giuliano, 2005; Wibowo, 2008). While the literature has dealt with how people choose
the mode of transportation for their daily travel, it has tended to undercount people’s use
of transit because so many people do not use transit every day. By comparison, this study
uses the frequencies with which people use transit systems in the past month to capture
both regular and occasional users. These data will allow us to make a more complete
estimation of the actual market for public transit by accounting for people who do not use
public transportation every day or in a regular pattern.
Another contribution of this study is to measure land-use attributes in a combined
dimension and to use the most detailed unit of analysis. Previous literature has usually
measured the independent variable in a single measurement such as density and land-use
mix. Those environmental variables are highly correlated with each other, and such a
multicollinearity problem often occurs in a model. To avoid this problem, this study uses
combined measurements such as principal component analysis (PCA) and cluster
analysis. Moreover, this study uses the most detailed unit of analysis available.
51

Household locations are geocoded based on residence location. Land-use and transit-
service attributes are assigned to each household, using the GIS. All environmental
independent variables are measured within a ¼-mile Euclidian buffer around each
household location. The land-use and transportation attributes used in the research come
from two archived sources: census files and the SCAG’s roadway network and land-use
data for the year 2008.

4.3.1. Analysis Framework

This study tests the hypothesis that perceptions of a traveler’s neighborhood influence the
probability that he or she will use transit frequently. A conceptual framework was
developed for analyzing individual travel behavior relating to four dimensions: transit
service, socio-economic characteristics, neighborhood type, and perception.
I conducted a principal component analysis (PCA) with a Varimax rotation to
reduce the multiple overlapping perceptional variables to their underlying components
(also called factors). The components created from PCA are continuous standardized
scores with a mean of zero and they were used to represent perception factors in
subsequent analysis. The heart of this analysis aims to understand what particular types of
neighborhoods are making a difference with regard to the effect of perceptions on transit
use. Therefore, neighborhood types were measured in continuous dimensions that consist
of physical characteristics of land use measured within a ¼-mile Euclidian buffer around
each household location. To ascertain neighborhood type, a two-step cluster analysis was
52

employed, using seven physical characteristics of neighborhood variables that were
assessed objectively.
These component variables resulting from PCA and two-step cluster analysis
were then used in regression models of transit use. In order to examine the predictive
power of perceptional and neighborhood-type variables on public transportation use, two
separate ordinal logit models were developed. Model 1 estimates the overall effect of
perception on transit use. Model 2 estimates the interaction effects between perception
and neighborhood type, controlling for other factors that could potentially affect transit
use. These models were run separately for the two income groups: low-income and
higher-income travelers. The equation is presented as follows:
U = f (N, P, T, S, I) (2)
where
U = Type of transit user (1=non-user; 2=occasional user; 3=regular user)
N = Neighborhood type
P = Perceptions
T = Transit Service
S = Socio-demographic/economic characteristic
I = Interaction terms (Neighborhood Type * perceptions)

An ordinal logit model was used to obtain the probability that a traveler is one of
the three categories of the dependent variable (i.e., non-user, occasional user, and regular
user) and tests the hypothesis that people are more likely to use public transportation
regularly if they perceive their neighborhood environments as safe and destination-rich
53

places to walk. By adding interaction terms to the ordinal logit model, this paper tests
how these effects of perception vary by neighborhood type.

4.3.2. Key Variables

The individual is the unit of analysis for this research. Individual variables of travel
behavior and perception originate from the Los Angeles County sample, which is a subset
of the NHTS-CA. Originally the dataset included 5,861 interviewed persons. Because of
missing values, due largely to perception and land-use variables, some observations had
to be dropped from the original sample. The final sample used for the estimations
contains 3,498 persons. The models of this study are based on the assumption that
missing values occur randomly.
Table 4. 1 shows definitions of the dependent and the independent variables as
well as their sources. Data from these sources were compiled to analyze the effect of the
built environment on transit use in the year of 2009. The attributes of household and
individual characteristics including perception variables were obtained from the NHTS-
CA. Population density and housing structure were drawn from U.S. Census 2010. Land
use and street density variables were calculated in GIS and their source were obtained
from the SCAG’s parcel-level land use data and street network data for the year 2008.
The crime data were collected from the website
2
available.

2
www.crimemapping.com
54

Table 4. 2 shows the descriptive statistics for the individual and household
characteristics of respondents. The original data for the dependent variable was made up
of continuous values based on the quantity of transit use in the past month. To avoid the
excess-zero problem (i.e., a substantial fraction of the observations are not transit users), I
used categorical variables instead of original count variables. I divided the data into three
categories to define the type of transit user based on the distribution of the data. A regular
user is identified as a member of the first decile of travelers, one who reports using transit
at least four times a month. An occasional user is defined as a person who uses transit at
least once a month; they are located between the first and second deciles of travelers. A
non-user is a person who uses public transportation less than once a month or not at all.
The variable was coded as 1 for a non-user, 2 for an occasional user, and 3 for a regular
user. The variable is a categorical value and clearly ordinal. Therefore, instead of the
linear estimation or usual discriminant model, an ordinal logit model was used. The
distribution of this categorical dependent variable is presented in.
In this study, neighborhood type and perception factors were the focus; transit
accessibility and other socio-demographic and economic variables were used as control
variables. A set of eight of variables related to the neighborhood environments was
measured in the unit of ¼-mile radius around each household location. Table 4. 3 shows
the descriptive statistics of neighborhood-level variables including density, land-use,
housing structure, crime rate, and transit accessibility. To measure transit accessibility,
the travel times of local buses were analyzed rather than rail travel times, in light of the
fact that the majority (85%) of transit trips, in Los Angeles County, are by bus. Transit
55

accessibility measures both the transit supply in terms of transit time that was measured
through the local bus network and the size and distribution of employment served by the
routes. Therefore, a TAZ with good transit accessibility should have local buses that
connect easily to many destinations with large employment, which indicates high
opportunities for working, shopping, and leisure activity. Transit accessibility was
measured in a gravity-based model. The transit accessibility of household location i was
estimated by summing up the number of employments in TAZ
j
and travel impedance f
(C
ij
):
T
i
= ∑ Ej fCij / Max T
i
(3)
f (C
ij
) = exp (-bC
ij
) (4)
where,
T
i
= transit accessibility index for household location i;
E
j
= number of jobs located in TAZ j
C
ij
= travel time (in-vehicle time + walk access time + transfer wait time + initial
wait time) from household i to centroid j
b = the impedance factor = 0.0994

Transit accessibility score was weighted by the maximum value of T
i
and ranges
from zero to 1. A score of zero indicates no access and 1 indicates the best access to
employment. The impedance factor for trips was calculated using the inverse of the
average commuting distance for the Los Angeles region in 2009. The data on network
observed travel times for local buses were provided by the SCAG. Peak-hour travel time
was measured through the local bus network. By using these data, travel impedance was
56

estimated by summing up the in-vehicle time, walk time to bus stops, transfer wait time,
and initial wait time.
The data on crime counts collected over six months of 2012 contain the type of
crime (including robbery, theft/larceny, and burglary) and street address and time of each
reported crime. The crime data were geocoded and aggregated to the neighborhood level
(one-mile radius around each household location). Other control variables include the
traveler’s socio-demographic and economic characteristics from the 2010 Census and the
2009 NHTS-CA. Gender and race were included in the initial models, but they were
insignificant and dropped from the specifications in the final models.

57

Table 4. 1. Variable definitions and data sources
58

Table 4. 2. Descriptive statistics of individuallevel variables

Std. Dev. = standard deviation

Table 4. 3. Descriptive statistics of neighborhoodlevel variables

Std. Dev. = standard deviation
59

4.4. ANALYSIS AND FINDINGS

4.4.1. Principal Component Analysis

A principal component analysis (PCA) was used to reduce the multiple overlapping
perceptional variables to a smaller number of underlying factors that were used in
subsequent analysis of transit use. Among thirteen perception variables of individuals, ten
variables were subjected to PCA, because three variables from the questionnaire
produced no factors and were dropped from the analysis. Table 4. 4 shows that Varimax
rotation produced four factors (also called components) that account for 45% of the
variance in the dataset.
The number of factors was chosen based on the eigenvalues greater than 1, which
represents a substantial amount of variation. Each eigenvalue indicates the relative
contribution of its associated factor to the total variance explained by all factors. A factor
score was calculated as the weighted sum of an individual’s scores on ten perception
variables, where the weights were the loadings for that factor. Therefore, the scores
represent the relative position on the latent component variables.
Factor 1 accounts for 11.9% of the variance and is labeled concern for “physical
safety”. It includes four (out of a possible 10 items) variables: “no sidewalks”, “exposure
to air pollution”, “fast traffic” and “unsafe street crossing”. Factor 2 accounts for 11.3%
of the variance, and is termed “personal safety”. It includes: “fear of street crime” and
“not enough people”. Factor 3 accounts for 11.14% of the variance and is labeled
60

“amenities”. It includes: “no shops or interesting places to go” and “no nearby paths or
trails”. Factor 4 accounts for 10.8% of the variance and is labeled “isolation”. It includes:
“too-wide streets” and “no one to walk with”.
Table 4. 4. Factor loading matrix (Principal Component analysis of perception variables)

In this study, a range of variables related to transit use were analyzed by income
group. To test for the statistical significance of differences between mean values of
individual perception factor scores within each income group, a T-test was employed and
differences were considered significant if t-test P values were below 0.05. Table 4. 5
shows that low-income travelers have significantly higher concerns about their
neighborhood problems than do higher-income travelers. The most significant gap
61

between the two income groups was observed for concerns about physical safety and
personal safety.
Table 4. 5. Ttest to compare low and higherincome traveler’s factor score

SD
a
: Standard Deviation, SE
b
: Standard Error, P(Sig.): Significance level

4.4.2. Cluster Analysis

To classify neighborhood type, a two-step cluster analysis used the eight variables of
urban forms listed in Table 4. 6. Variables of population density and housing structure
were obtained from the Census data for the year 2010. The variable of street density
correlates with the degree of land-use mixing. In New Urbanist design theory, these
factors typify the important design characteristics of a neo-traditional neighborhood
(Calthorpe 1993; Handy 1992). Studies have shown that this type of neighborhood can
help people walk more, and it is associated with pedestrian access to transit stops. To
analyze the spatial layout and the land-use pattern of each neighborhood, this study used
parcel-based land-use and street network data on Los Angeles County. All these
62

neighborhood variables were measured in the unit of ¼-mile Euclidian buffer around
each household location. Population and housing-structure variables were aggregated
variables at the census tract level. Using an areal interpolation method, these values were
re-measured in the GIS. The proportion of each tract area was calculated and summed up
in a quarter-mile radius around each household location.
The distributions of all variables were standardized to normal distributions before
conducting cluster analysis. Among a variety of techniques for cluster analysis, two-step
cluster analysis was chosen because this method can automatically select the optimal
number of clusters by using an agglomerative hierarchical method. Euclidean distance
was used to assign a neighborhood to the nearest cluster. Through a two-step cluster
analysis, three different types of neighborhood emerged. Table 4. 6 presents the cluster
mean value of each variable. The neighborhood types are listed in order of their average
population density. This neighborhood type includes much more information than
population density alone. Type 1 is referred to as a “high-density neighborhood”, as the
location of the neighborhood is in an inner-city area close to the CBD. Type 1 makes up
31.5% of the sample and covers the neighborhoods with the highest population and street
density, but relatively low-levels of mixed land use. Type 2 is referred to as a “mixed
land-use neighborhood”. Type 2 makes up 37% of the sample and includes
neighborhoods with a high proportion of commercial land use and multi-family housing.
Type 3 is referred to as a “low-density neighborhood” and makes up 31.5% of the
sample. Type 3 is identified as a neighborhood with low density and high proportions of
single-family dwellings.
63

Table 4. 6. Mean values by neighborhood type (2,339 household neighborhoods clustered by
Euclidean distance)

a
Pop density: Population density,
b
INDS: The proportion of industrial land-use,
c
COMM: The proportion of
commercial land-use,
d
Park: The proportion of park land-use,
e
SFD: The proportion of single family
dwelling among total housings,
f
MFD: The proportion of multi-family dwelling among total housings

Figure 4. 5. Geographical clustering of neighborhood types in Los Angeles County
Neighborhood Type 1 (dense, mixed-use neighborhood); Neighborhood Type 2 (medium-dense, single-
residential neighborhood); Neighborhood Type 3 (low-dense, single-residential neighborhood)

64

4.5. ESTIMATION RESULTS

An ordinal logit model was run using the three categories of the dependent variable. As
mentioned earlier, the dependent variable is user type: “non-user”, “occasional user” and
“regular user”, corresponding to the values from 1 to 3. Table 4. 7 and Table 4. 8 present
the results of the ordinal logit models that estimate the determinants of a traveler who is a
non-user, an occasional user, or a regular transit user. Model 1 estimates the overall effect
of perception on transit use. Model 2 estimates the interaction effects between perception
and neighborhood type, controlling for other factors that could potentially affect transit
use. These models were run separately for the two income groups: low-income and
higher-income travelers.
The substantive results are consistent across all models; that is, five primary
factors appear to determine travelers’ propensity to use public transportation: income,
vehicle/person ratio, age, driver license, and transit accessibility. The significance of
these factors is to be expected, as they have been identified in previous studies (Chapin,
1974; Kitamura et al., 1997). Young age has significant positive effects on likelihood of
being a regular user, while being elderly has negative effects. Income and vehicle/person
ratio has negative effects on transit use, while transit accessibility has positive impacts.
To clarify the interpretation of the size of these effects that have practical
importance, marginal effects were estimated at the mean of each variable sample by using
the following formula:
65

P
X
= P

[θ
‐θ
] where, P

= the coefficient for X
k
(5)

The effect sizes of some of these variables are large (Table 4. 7). As the marginal
effects and the elasticity of these variables show, vehicle/person ratio is particularly
important element for low-income group. A 10% increase in the vehicle/person ratio
decreases the likelihood of using transit regularly by 9.2%, while an identical increase in
total household-income decreases the probability of being a regular user by 4.4%.
Regardless of a traveler’s socio-economic and demographic status, transit accessibility
appears to be universally important in frequent transit use. The variable of transit
accessibility is still significant for higher-income travelers, but the elasticity decreases
from 0.33 to 0.24. This result indicates that increasing transit accessibility will be more
effective for low-income than for higher-income travelers. The marginal effects cannot be
provided for neighborhood and perception variables that are created from the principal
component and cluster analyses, due to the difficulty in interpreting practical change in
component score.
From the results displayed in Table 4. 7 and Table 4. 8, it is clear that travelers’
neighborhood type influence their decisions to use public transportation. Type 2 is high-
density neighborhood, and the reference type for neighborhood type dummies. The
likelihood of being a regular user increases with the mixed land-use neighborhood type.
This means that a person who lives in a mixed land-use neighborhood is more likely to
use public transportation regularly, all other factors being equal. Notably, neighborhood
type has a significant effect only for higher-income travelers; this is consistent with the
66

expectation that most higher-income travelers use transit by choice, and thus they are
more likely to be attracted to gain a relative advantage as neighborhood becomes more
mixed land-use.
The results of models show that perception of neighborhood environment plays an
important role in determining travelers’ propensity to use public transportation.
Specifically, the pooled model indicates that an increase in concerns about physical
safety decreases the probability of being a regular user. This means that if a traveler has a
high level of fear concerning physical safety, he or she is less likely to use public
transportation regularly. However, the coefficient of concern for physical safety is no
longer significant in the segmented model by income group. This shows that the change
in the physical-safety concern exhibits less variability in the segmented model than the
change in the physical-safety concern across all income groups. Most importantly, the
variable of personal safety is significant at the 95% level for all travelers, regardless of
their income level; higher concerns about personal safety from street crime and perceived
lack of surveillance tend to decrease the probability of being a regular user. Notably, the
perceptions about isolation have very different effects across the income groups. The
negative impact of isolation concerns is significant for low-income travelers, while this is
not significant for higher-income travelers. The variables of amenities concerns are not
statistically significant for any group in Model 1.
This study also tests the hypothesis that a certain type of neighborhood is making
a difference with regard to these effects of perceptions. To test this hypothesis, the second
67

model includes the interaction effects between perception factors and neighborhood
types. From the results displayed in Table 4. 8, the perceptions have very different effects
across neighborhood types and income groups. The concerns about physical safety and
isolation have a significant effect only for poor people in low-density areas. They seem to
feel threatened by fast traffic, unsafe street crossings, and isolated environments when
they walk the wide streets of the suburban areas which lie in the large urban fabric. These
negative perceptions appear to be a significant deterrent to using public transportation.
On the other hand, in mixed land-use areas, personal safety concerns were more
important to low-income travelers than physical safety concerns and they have significant
negative impact on their transit use. However, for higher-income travelers in mixed land-
use areas, both personal safety and amenities concerns (lack of interesting places and
walking trails nearby) matter and they decrease the probability of taking public
transportation. This finding may indicate that low-income travelers are reluctant to use
public transportation because of perceived safety concerns, while higher-income travelers
are more likely to be affected by the physical attributes or amenities factors of their
neighborhoods.

68

69

70

4.6. DISCUSSION

In an era of growing interest in strategies for increasing transportation sustainability, it is
vital that planners and policymakers understand how to provide the quality of mobility
and accessibility via transit systems to everyone.
This study examines the effect of perceived environment on transit use by means
of an ordinal logit model while controlling for neighborhood type, individuals’ socio-
economic and demographic factors, and transit accessibility. It empirically distinguishes
among transit users according to three ordinal categorical values: non-user, occasional,
and regular user.
While the results are mixed, evidence does exist to suggest that environmental
perception plays a role in determining travelers’ propensity to use public transportation.
In terms of regular transit use, there are significant differences between those who
perceive a threat to their physical or personal safety and those who do not. Higher
concerns about personal safety from street crime are likely to decrease regular transit use.
This effect is universally significant for all travelers, regardless of their income level.
Similarly, high levels of fear concerning physical safety from traffic danger and
perceived isolation are associated with a decrease in people’s willingness to use public
transportation. Actual crime rate of the neighborhood is particularly important for the
low-income travelers, probably because low-income residents are more likely to live in
dangerous places than any other. However, for higher-income travelers, it is personal
safety concerns that decrease the probability of taking public transportation rather than
71

the actual crime rate. The evidence suggests that personal attitude toward environment is
a stronger predictor of travel behavior than is the actual environment. This study also
reinforces some of the existing literature that explains individual socioeconomic
characteristics and transit service quality are associated with transit use. Variables such as
age, vehicle/person ratio, driver license, and transit accessibility are all found to be useful
predictors of user type, and the inclusion of these controls in the model further reinforces
the significance of the findings.
The results presented in this paper demonstrate that the relationship between
traveler’s perceptions, neighborhood type, and transit use is complex. Clearly there are
similarities between income groups, but differences are more evident. Perceived lack of a
safe place to walk seems to be the top-ranked barrier to use of public transportation by
low-income travelers, while physical environmental features and amenities appear to
encourage transit use by higher-income travelers—those who are more likely to use
transit by choice. This might reflect that higher-income travelers are attracted to gain
relative advantage as environments become more amenable to walking, while low-
income residents are more susceptible to risk exposure in neighborhoods rather than to
urban design amenities. This confirms that safety needs precede amenities needs. That is,
if an individual’s need for safety is unmet, a very pleasant, clean, or comfortable
environment setting would not necessarily compel him or her to use public transportation
(Alfonzo, 2005).
72

These findings suggest that unfavorable perceptions of environmental conditions
are a significant deterrent to using public transportation. An important implication for
policy is that policy makers can increase the utility of transit for low-income people by
enhancing neighborhood safety and by providing better quality in local access to transit.
Much of what we discovered in this study is not simply within the control of transit
agencies. This suggests that transit agencies have to work with public entities. If we want
to solve neighborhood deterrence −whether it be an unsafe walking environment resulting
from street crime or one resulting from traffic danger, we need an active policy decision
to enhance public safety and to make the neighborhood more walkable.
The contribution of this study lies primarily in the empirical demonstration of the
relationship between low-income travelers’ perceptions and their transit uses, and
secondly in its approach and methods. However, some limitations of this study should be
highlighted as well. First, it does not contain objective measures of the sidewalks or
topographic characteristics that might be important for neighborhood attributes. Second,
the crime data was collected in 2012 so there is time inconsistency with the main NHTS
data, which was collected in 2009. The effect of this data limitation on the analysis is
unclear. Finally, this study relies on cross-sectional data and thus does not necessarily
describe causal relationships among variables. Future studies should include longitudinal
studies to better understand the causal structure of perceived environment and transit use.

73

CHAPTER 5. TOUR COMPLEXITY AND TRAVEL
MODE CHOICE: THE ROLE OF
NEIGHBORHOOD WALKABILITY

5.1. INTRODUCTION

The large body of research on travel behavior has established differences in travel mode
based on various types of land use. These studies have shown that travel behavior is
related to accessibility, density, mixed land uses, and other household characteristics
(Kockelman, 1997; Kitamura et al., 1997; Zhang, 2004). By comparison, there have been
very few research efforts that have investigated the relative impact of land use on mode
choice using the tour-based approach. The previous research related to mode choice has
made a vast improvement in our understanding of the influence of land use on travel
demand; yet this research failed to account for true behavioral causality, which can be
explained using the tour-based approach only when the linkages between trips are
properly incorporated into the analysis.
Previously researchers have found that the complexity of tours constrains mode
conversion, and complex tours are more likely to be taken by private vehicle (de Nazelle,
2010). Similarly, Krygsman et al. (2007) found a relationship between transport mode
and chaining behavior, but evidence of this relationship is indeterminant when location
factors of intermediate activities are considered within a tour. While general, tour-based
74

studies have addressed this tour complexity issue, few studies have actually used the tour-
based, mode choice framework to determine the causal relationship between mode choice,
tour complexity, and land-use factors.
Since there is little empirical evidence about the relationship between land use
and mode choice of chained travel, in this study, I used a tour-based, mode choice
framework to analyze how travelers decide the mode of their tours. In doing so, I
considered their activity patterns and tour formation, which may differ from household to
household, by lifestyle, and by residential location. In the study, I also analyzed how the
effects of land use vary by tour types as well as whether land use influences the
residential environment or the destination environment for chained activities.
In the next section, related, tour-based, mode choice research is reviewed,
focusing mainly on empirical studies. Then the mode choice patterns by different tour
types and household characteristics of the surveyed travelers are described, and this is
followed by the empirical results of the tour generation models and mode choice models.

5.2. DATA AND METHODOLOGY

In this study, I attempted to address the gaps in the literature by using a refined, tour-
based, mode choice framework to investigate how land-use patterns are related
significantly to mode choices and tour types. Based on the previous literature, the author
formulated the following three hypotheses: 1) Neighborhood walkability and household
characteristics have association with tour complexity and vehicle miles traveled (VMT );
75

2) Residential walkability and destination walkability matter to mode choice when travel
costs and time, travelers’ socio-economic factors, and tour characteristics are controlled;
and 3) The effect of walkability on the choice of tour mode varies for work tour vs. non-
work tour and for a simple tour vs. a complex tour that links multiple trips.
These hypotheses are based on the research that examines the relationship
between travel mode choice and land use both for the origins and the destinations of the
tours. In general, residents are more likely to use non-motorized and transit modes if they
live in highly-walkable, and mixed-use area. The main focus of this paper was to
determine whether this pedestrian-friendly, community design encourages individuals to
choose the mode of transit, walking, and riding a bike (from here mentioned as ‘active
travel’) and how these effects vary by different types of tours.

5.2.1. Data

Table 5. 1 shows the list of variables considered in the model. The individual was the unit
of analysis. The data of individual travel behavior and any household variables came
from the Los Angeles County sample, a subset of the 2009 National Household
Transportation Survey for California Add-on (NHTS-CA). The sub-sample of this data
were used for the analysis, including only persons who were 16 years of age and older
and who responded to the travel survey on a weekday (Monday through Friday). This
resulted in a total of 6,834 tours performed by 4,231 respondents residing in 2,171
households that were used for pooled-sample models.
76

The employment data originally came from Info USA 2008. The parcel-level land
use data and street network data were provided by the SCAG. Other land-use attributes
used in this study came from Census files. Mode choice models were estimated based on
the explicit travel time and cost for each leg of the trip within a tour to reflect the
popularity of the alternatives. The main data source was extracted from the origin-
destination matrices of travel time and cost by all modes at the level of the Traffic
Analysis Zones (TAZ), obtained from the SCAG’s Regional Transportation Model.
Walking and bicycle travel times were estimated based on street network distances, and it
was assumed that average travel speeds were 1.22 m/s and 5.36 m/s for walking and
bicycle use, respectively (Rodriguez and Joo, 2004).

5.2.2. Measuring the Walkability Index

The author used the walkability index as the proxy of the compact urban form.
Walkability attributes in terms of density, balance of land uses, street network patterns,
and intensity of retail employment are surrogate measures of land use.
In this study, methods presented in a previous study (Frank, 2010; Sundquist et al.,
2011) were used to calculate the walkability index. The concept of walkability relies on a
combination on four variables, i.e., residential density, street connectivity, land-use mix,
and the retail floor-area ratio. Land-use data, street centerline data, employment data, and
census data were integrated spatially using the GIS method to create the walkability
index as the composite variable separately for the origin and destination of the tours. To
77

do this, a walkability index was measured based on a radius of ¼-mile Euclidian radius
around the respondents’ residential areas and around the respondents’ tour destinations
(For work tours, the destination was the respondents’ place of employment). Previous
studies have used the variable of retail floor-area ratio as one of components of the
walkability index. However, due to the limited data, it was decided in this study to use
the retail employment data obtained from infoUSA to measure the intensity of retail
stores and commercial services. The walkability index was calculated as the sum of the
‘Z’ scores for the four factors, and the following formula was used:
Walkability Index = 2 * Z (street connectivity) + Z (residential density) + Z (land use mix)
+ Z (retail and commercial services) (6)
The walkability index was used in the models to avoid multi-collinearity among
the objective measures of the built-environment variables and to characterize land-use
characteristics as ratio variables rather than simple, discrete variables, which tend to have
a “predictive disadvantage” due to much smaller variation over the control variables,
such as income and trip attributes (Cervero and Kockelman, 1997; Srinivasan and
Ferreira, 2002).
78

Table 5. 1. Description of variables

5.2.3. Methodology

To test the relationship between neighborhood walkability and tour complexity,
generation, and mode shares, this study developed activity-based analysis that considers
the linked nature of most travel. Different regression models were specified for two types
of models: tour generation model and tour mode choice model.

79

a. Tour Generation Models
The hypothesis of tour generation models are neighborhood walkability has
association with tour complexity and VMT. The author estimated a series of regression
models to test hypotheses with regard to tour characteristics. I associated tour complexity,
frequency, and distance with household and geographic characteristics.
Three outcome variables were estimated in separate regression models developed
with identical structures to ensure consistency between each of the models. A poisson
regression model was used to predict the number of tours due to the nature of the count
data, while linear regression models were used to predict the average number of trips
within a tour. Tobit regressions were estimated for the measures of daily travel distance
by private vehicle because those are left censored (i.e., several respondents reported no
driving). The models were run separately for two different types of tours: work tour and
non-work tour. Followings are tour characteristic regression models to test the
significance of neighborhood walkability:
The number of trips per tour = f (NW, CBD, DISTOWK, HH, PS) (7)
The travel distance per day = f (NW, CBD, DISTOWK, HH, PS) (8)
The number of tours per day = f (NW, CBD, DISTOWK, HH, PS) (9)
Where NW indicates walkability of residential and work location; control
variables include the CBD which denotes the distance to the CBD from the household
location, DISTOWK which represents the distance to the work place from home location,
80

HH which denotes household-level travel demand variables including socio-economic
status and life cycle, and PS which denotes person-level attributes such as age, gender,
race, work status, and possession of driver license.
The unit of this analysis was a person. Explanatory variables included household-
level variables such as socio-economic status, lifecycle, and household structure, and the
variables of place of residence (i.e., walkability of home neighborhood, the distance to
the CBD) and place of employment (i.e., walkability of employment location, distance to
work). The study assumes that travelers jointly consider both home and work location of
the tour when they decide their travel patterns. All variables of land use and tour
attributes entered the regression as continuous variables. I assigned midpoints to
categories for income variables in order to create continuous measures.
Some observations from the original data sets were excluded. First, the regression
models included only persons who were 16 years of age and older and who responded to
the travel survey on a weekday (Monday −Friday). Second, trips longer than 100 miles
were also excluded from the regression samples so as to eliminate atypical cases. Third, I
excluded observations for which trip time/mode/distance or any values of independent
variables were unknown.

b. Tour Mode Choice Models
In the analysis of tour mode choice, I tested the likelihood of driving alone versus
sharing a ride, taking transit, or walking/riding a bike. Explanatory variables related to
tour attributes (tour purpose/time/type/mode) and person-level travel demand, in addition
81

to the variables used in tour generation analyses. Basic statistics of the categorical and
continuous variables were summarized in Table 5. 2 and Table 5. 3.
In contrast to the tour generation models, the unit of mode choice analysis was a
tour, which was defined as a ‘home-to-home loop.’ A complex tour includes more than
two trips with at least two intermediate stops. All persons, households, and their
neighborhood characteristics were merged into the dataset of the tours.
The mode choice models were categorized by four types of tours. This study
distinguished work tours from non-work tours. Chains that included a journey to work
were classified as work tours, and all others were classified as non-work tours. These
work and non-work tours also were divided into simple and complex categories. Simple
tours involved only one stop, whereas complex tours involved two or more intermediate
stops. Complex work tours combined work and non-work trips, while complex non-work
tours were comprised of only non-work trips.
In this study, to assign a primary purpose to a multi-trip tour, the primary purpose
of a tour was defined as the purpose of a tour that relates to the longest distance of all
legs of the trip in the tour. In case there are two equal longest distances in the tours, the
primary purposes were determined based on the following hierarchy: work > shopping >
leisure > discretionary. This is the method that previous studies commonly have used to
define the primary purpose of a tour (Ye et al., 2007; Frank et al., 2008). The primary
destination of non-work tours was decided following the same method as for the primary
purpose.
82

Five modes of travel were considered in this study and the primary mode of the
tour was determined based on the following priority sequence: transit > carpool > drive
alone > bike > walk. In other words, if the transit mode were used for any trip legs within
a tour, the primary purpose of the tour was defined as transit mode. The ranking was
decided in this way to reflect the anticipated consideration that travelers who view their
travels strategically would use in evaluating their mode of travel (Frank, 2008). Because
very few trip records for bike mode (1.3%) were available, separate bike mode analyses
were not carried out. Bike and walk modes fell into one category in mode choice models
to test the impact on the likelihood of using non-motorized mode for the tour.
The mode choice was estimated using the multinomial logit (MNL) model
regression. The utility function used in this analysis is the sum of the utility of person i
choosing a mode j for the tour:
U
ij
= V
ij
+ ij
,

(10)

V
ij
= α +

j
c (11)

=
+ + + (12)

U
ij
is the utility of the traveler i’s choosing option j for the tour;
V
ij
is the systematic utility of the traveler i’s choice for the tour;
is the non-systematic part of the utility for the choice;
α is alternative specific constant;
is the vector of coefficients for the utilities;
T is the attributes of the tour person i traveled (i.e., time, purpose, distance);
83

P is person i’s attributes (i.e., age, gender, ethnicity);
H is person i’s household variables (i.e., income, vehicle number per person, family life);
RW is the walkability variable of residential neighborhood where person i reside;
DW is the walkability variable of primary destination where person i traveled (in case of work
tour, the destination is person’s employment place);

P(ij) =
exp
∑ exp
(13)

P(ij) is the probability of person i choosing a dominant mode j from a feasible choice set c.

is a vector of explanatory variables and is the parameter to be estimated.
To ensure consistency with the theoretical discussion in the methodology, this
study took the utility function approach that assumes the marginal effect of travel time
and cost on mode choice varies by different mode (Ben-Akiva and Lerman, 1985).
Therefore, the model estimated alternative-specific travel time variables reflecting the
attractiveness of the alternative mode.
The author estimated multi-nomial logit models for both a pooled sample and for
samples divided into four groups according to the types of tours. It should be noted that
sampling of the work tour was different with that of the non-commute tour. Work tour
included only for samples of workers, while non-work tour models were estimated both
for workers and non-workers.

84

Table 5. 2. Descriptive statistics of categorical variables

Table 5. 3. Descriptive statistics of continuous variables

85

5.3. DESCRIPTIVE ANALYSIS

The following sections present descriptive and statistical analyses of the samples and the
findings. The first descriptive discussion includes the distribution of mode of travel,
tabulated by trip purpose and neighborhood walkability. In particular, I classified all tours
into four categories according to their purpose (work/non-work) and complexity
(simple/complex).
Figure 5. 1 illustrates basic descriptive statistics that show why studying trip
chaining is useful. Almost 21% of all tours involved commuting, and about 45% of
commuting trips were complex tours that were linked to non-work trips. In the case of
non-work tours, which accounted for about 79% of all tours, 35% of the tours were
complex tours that had multiple, non-work stops. The proportions of trip-chaining were
slightly greater among commute tours than non-commute tours.

Figure 5. 1. Proportion of trip chaining within a work / nonwork tour

86

5.3.1. Distribution of Mode of Travel by Neighborhood Walkability

The author conducted a cross sectional comparison of low- and high-walkability
neighborhood in four different tour types to see the distribution of mode of travel. The
neighborhoods were categorized by two types based on the distribution of the walkability
index scores. A low-walkability neighborhood is identified as a neighborhood within the
first, second, third, and fourth deciles of the walkability index scores, while a those
within the seventh, eighth, ninth, and tenth deciles were defined as a high-walkability
neighborhood. This approach is consistent with the previous studies of Owen et al. (2007),
Sallis et al. (2009), and Sundquist et al. (2011).
Figure 5. 2 compares a low-walkability neighborhood to a high-walkability
neighborhood in the distribution of mode share by tour type. Neighborhood residents in
high-walkability neighborhoods made more walking, biking, and transit trips, as well as
fewer auto trips than residents in low-walkability neighborhoods. These differences were
more evident in the case of simple tours. Examining the row-based percentages indicated
that 37% of simple tours were taken by transit, walking, or bike modes in high-
walkability neighborhoods. In contrast, only 18% of simple tours were not auto-based
trips in low-walkability neighborhoods. Interestingly, if the tours had multiple stops
(referred to as complex tours), the travel pattern of high-walkability neighborhoods
became similar to that of low-walkability neighborhoods. Most complex tours were auto
trips (91%), regardless of the walkability of the neighborhoods.

87

Figure 5. 2. Distribution of modes of travel by different types of tours and neighborhood
walkability (a) simple tour, (b) complex tour, (c) work tour, (d) nonwork tour

Different travel patterns also were evident between work and non-work tours.
First, it was confirmed that the travelers were more likely to drive alone if they were on
tours that included work stops. In contrast, a traveler was more likely to carpool than to
drive alone for non-work activities. It was conceivable that non-work tours were more
88

likely to be inter-household trips, and thus, in many cases, carpools were used for the
tours. Second, a traveler was more likely to use a non-motorized mode if the tour
contained only non-work-related stops. This was more prevalent in high-walkability
neighborhoods than low-walkability neighborhoods. This is likely because the above-
mentioned, non-work tours contained maintenance (e.g., shopping) or leisure activities.
These activities are often satisfied in the range of services in highly walkable and mixed-
use neighborhoods, whereas the commuting tour is less likely to be used locally (Krizek,
2003).

5.3.2. Trip Chaining and Mode Choice

The second part of the descriptive analysis focused on the relationship between the
complexity of trip chaining patterns and mode choice, particularly focusing on auto and
transit mode. Because the transit may have less flexibility to chain them in complex tours,
transit use was expected to be negatively affected by the formation of a complex tour. By
contrast, auto tour was expected to have a higher probability of being chosen for a
complex tour and a longer distance tour. The reasons for the selection of key variables are
explained based on the findings of these descriptive analyses in the following section.
Regarding the travel mode allocations by different travel patterns, public
transportation are more likely to be used in simple travel patterns, probably because
transit users may have less flexibility to chain trips in complex tour. There was evidence
89

that suggested the existence of these differences between tours taken by public
transportation and tours taken by private vehicle. To statistically measure the magnitude
of the difference between transit tours and auto tours, t-tests were applied.
Table 5. 4 shows that transit tours involved fewer numbers of stops for work tours.
That is, the use of public transportation was utilized to a greater degree in the context of
simple, fewer-stop, trip chains. Conversely, non-work transit tours were shorter than
tours for which private vehicles were used as the primary mode of transportation.
Therefore, it appeared that transit tours tended to be simpler and shorter than auto tours.
Table 5. 4. Ttests of mode choice and tour type for work and nonwork tours

SD
a
: Standard Deviation, SE
b
: Standard Error

Above descriptive statistics consistently show that tour types and neighborhood
walkability may be an important factor in the choice of tour modes. To further understand
to what extent the tendency to chain trips and the choice of tour mode can be explained
by neighborhood attributes, I estimated two types of regression models: tour generation
model and tour mode choice model in the next section.

90

5.4. RESULTS OF HYPOTHESES TESTED

5.4.1. Determinants of Tour Complexity and VMT

Hypothesis 1. Neighborhood walkability and household characteristics have
association with tour complexity and frequency.
Hypothesis 2. Neighborhood walkability and household characteristics have
association with VMT.

The results of tour generation models were summarized in Table 5. 5 and Table 5. 6. First,
household socio-economic and demographic characteristics help to explain persons'
travel behavior. As expected, household income strongly affected the numbers of tours
and chains and the travel distance by private vehicle. Higher-income households traveled
longer distances and linked multiple stops in a single tour. Also, they tended to make
more tours per day. Vehicle/person ratio was a particularly important element for number
of chains for non-work tours. Those who had more vehicles per person in the household
were likely to travel longer distance and to involve more number of stops for non-work
tours. This is consistent with the expectations that the use of private vehicles is utilized to
a greater degree in the context of complex, multiple-stop, trip chains. Households in
which children were present traveled longer distances and made more tours for non-work
activities, due to children's mobility dependency. This supports the view that travel
demand is derived from the demand for daily activities (Bowman and Ben-Akiva, 2001).
91

Several individual characteristics also had the association with individual travel
outcome. Young travelers were less likely to chain the trips and they tended to generate
small number of non-work tours during a weekday, although the estimated coefficients
for young travelers showed a less clear pattern for their daily travel distance. The effect
of gender on tour formation was mixed. For non-work tour, women tended to link
multiple stops for non-work tours while the estimated coefficients for gender showed a
less clear pattern for trip chaining for work tour. This may be because women made more
intra-household trips for non-work activities and shared more of the responsibility for
childcare (Rosenbloom, 2006). Moreover, the numbers of tours and chains for non-work
tour were affected significantly by the possession of driver license and work status.
Drivers tended to chain trips and to take greater numbers of non-work tours. Workers
also made more frequent trip, but showed a less clear pattern for the trip chaining and
distance for non-work activities. This may be because travel time budget might constrain
the worker's non-work activities during a weekday.
The geographic characteristics also had an important effect on the tour formation
and distance. As expected, travel distances were affected significantly by the distance to
the work place. As the distance to work increases daily commute distance increases. In
case of non-work tours, the longer distance to the CBD was significantly associated with
the longer travel distance by private vehicle. That is, central city households tended to
take significantly shorter trips than households in the suburbs, all else being equal.
92

The primary purpose of this study is to test whether the built environments at
home and work locations play an important role in trip chaining behavior. As shown in
Table 5. 5 and Table 5. 6, local land-use effects occurred after household characteristics
and geographic effects were controlled. People who were located in high-walkability
neighborhoods took significantly shorter trips and they were less likely to link stops for
both work and non-work tours. This results seem to be in line with previous studies in
that the findings show higher-levels of neighborhood access matters, in particular as they
related to tour formation (Krizek, 2003).
Interestingly, the effect of home walkability on tour formation was mixed. The
results showed that people in higher-levels of neighborhood walkability tended to chain
fewer numbers of intermediate stops but they had a greater number of non-work tours per
day. In other words, people made less complex tours but they tended to leave home more
often for non-work activities when they live in a highly walkable area. Thus one may
argue that VMT savings of walkability appears to be ambiguous in that the results
represent a fraction of non-work activities. Given results from the previous descriptive
analyses which indicated that a large portion of the non-work tours were short, many of
these frequent non-work tours may have been conducted local to one's neighborhood.
However, what remains relatively unclear is whether the majority of these tours were
pursued by non-motorized (i.e., walking and biking) or transit mode.
For work tour, walkability of the work place positively influenced the number of
chains, while the walkability of the residence negatively affected tour complexity.
93

Intuitively, whether or not the land use around work location offers more opportunities
for non-work activities is relevant when travelers chain their trips to draw advantages
from the affluent environment of employment place. The results of Tobit regression
model of VMT consistently indicated that a high walkability of work location
significantly contributes to the increase of daily commute distance.
All in all, while the results are mixed, evidence does exist to suggest that
neighborhood walkability of both residential and employment location is associated with
VMT. Residential neighborhoods' walkability could potentially reduce VMT because it
tends to decrease travel distances and number of chains in a tour. However, increasing
walkability might also induce more trip-making as persons living in a central area take
more trips for grocery shopping, personal errands and entertainment. Similarly, a high
degree of walkability of employment location could influence VMT, because it induces
the opportunities for trip chaining and thus increases total commute distance.

94

Table 5. 5. Regression results for tour generation

Note: Coef.=coefficient; P(Sig)=significance
95

Table 5. 6. Tobit regression results for VMT

  Note: Coef.=coefficient; P(Sig)=significance

96

5.4.2. Determinants of the Choice of Tour Mode

This section presents the results of the estimations provided by the mode choice models
that were developed as described in this paper. The focus of the work was on following
two hypotheses that have both policy and methodological implications for the study of
the relationship between land use and travel behavior.
Hypothesis 3: Neighborhood walkability matters in choosing the transit mode when
travel costs and time, travelers’ demographic and socio-economic factors, and tour
characteristics are controlled.
In the pooled-sample models (Table 5. 7), the author analyzed the relationship
between the modes used for various types of tours and the full range of factors that might
affect the choice of mode. Nagelkerke’s R-squared model fit the data reasonably well.
Notably, the inclusion of walkability variables improved the model’s explanatory power
from 0.348 in the base model to 0.371 in the extended model. The results supported the
hypothesis that neighborhood walkability is the strongest predictor of the mode that will
be chosen for the tour.
All household characteristics proved to be significant for a number of modes. As
expected, those who had drivers’ licenses and had more vehicles per person in the
household were more likely to drive alone. Young travelers were more likely to use
transit, walk, or ride a bike, whereas female travelers were more likely to carpool and less
likely to walk than males. In general, households with children were more likely to
97

engage in carpool trips, but less likely to use transit. These findings were consistent with
previous studies of the choice of mode at the trip-segment level (Cervero, 2002;
Kockelman, 1997; Zhang, 2004). The effects of people’s races on the choice of mode
were consistent across all of the models. The shares of Black and White people had
negative impacts on carpools and transit use, which suggests that they were more likely
to drive independently.
As shown in Table 5. 7, there were considerable changes in the significance level
of the estimated coefficients when the walkability variables were included in the model.
The level of significance of the A.M. peak variable increased from 90% to 98% for
transit mode when walkability variables were included in the model. That is, travelers
with identical demographics and neighborhood characteristics were more likely to use
public transportation when their tours began during the A.M. period of peak traffic.
Similarly, the significance level of walking time required to access public transit
increased from 89% to 94% when the walkability variables were included, suggesting
that the measure of neighborhood walkability had an independent influence on the choice
of tour mode. Interestingly, the results showed that increased walking time to access
public transit encouraged more carpool use, and its significance level increased from 88%
to 96% when walkability variables were included. This confirmed that the major barrier
to enhanced use of public transportation was the excessive walking time required to
access public transit, a barrier that contributes to increased travel by automobile. The
correlation between the walkability index of residence and the distance to the CBD was
moderate (r > 0.4). These two variables were used as predictors in initial versions of the
98

analyses of tour generation. The distance to the CBD was considered to be a redundant
variable, so it was eliminated from the mode choice models.
The actual choice of travel mode was predicted on the basis of the estimates of
travel cost associated with alternative modes. While the driving-time coefficient for
motorized modes was statistically insignificant, the transit in-vehicle time coefficient for
transit mode was negative and significant. The coefficient for the walking travel time was
consistently negative and significant for non-motorized modes across all models.
The complexity of tours had significant and positive impacts on the choice of tour
mode for shared rides, but it was not significant for the use of transit or non-motorized
travel modes. In other words, travelers with identical socioeconomic characteristics and
land-use attributes made essentially the same choices of travel modes irrespective of
whether their tours were chained trips or not. However, it does not necessarily mean that
the complexity of a tour does not have an effect on the choice of tour mode for all tour
segments. Detailed sub-section analysis in the following section will account for how
different choices of travel mode might result from different types of tours.

Hypothesis 4: The effect of walkability on the choice of tour mode varies for work vs.
non-work travel and for a simple tour vs. a complex tour.
To test the robustness of the effects of neighborhood walkability in the previous
mode choice models, four sub-section models were developed specifically to test whether
the impacts of neighborhood walkability were consistent across different types of tours.
99

Two separate sub-models were developed for the allocation of work activity since it was
expected that there might be differences between people who were conducting only work
activities and those who were combining work activities with non-work activities. The
identical structure was used to develop sub-models for non-work tours.
The results from the mode choice models showed the varying effects of
walkability attributes on the choice of tour mode for different types of tours. As shown in
Table 5. 8, for a work tour, walkability measures at the residence no longer mattered if
the tour were a complex commuting tour, although these measures did matter for a simple
commuting tour. In contrast, the residential walkability consistently influenced the use of
the active transportation mode for non-work tours, no matter whether the tour was a
simple tour or a complex tour (Table 5. 9). Notably, walkability at the destination became
important to the decision of non-motorized mode for complex non-work tours. Intuitively,
whether or not the land use at the destination is pedestrian-friendly is relevant when the
travelers chain their trips to draw advantages from the destination’s environment.
The roles of several household characteristics also were changed by the
complexity of tours. For a work tour, households with no children no longer preferred
transit to other modes for complex commuting tours. The variable of non-driver status
became insignificant for the transit mode, while the presence of children became
significant for carpool use if the tour was a chained commute to work. Two tour
characteristics also changed. First, distance to work changed from statistically
insignificant for simple commuting to significant for complex commuting by carpool and
100

transit use. That is, the demand for shared rides and transit use decreased in association
with increases in distance to work if the tour required chained trips. Second, the
probabilities of using transit during the A.M. peak hours became insignificant for
complex work travel, although they were significant for simple commuting travel.
Interestingly, the demands for shared rides were associated with increase in transit
walking access time. This implies that many trips may have to be in carpool modes when
transit access is neither convenient nor accessible.
Unlike the work-tour models, age and gender had a significant influence on the
non-work tours (Table 5. 9). Young travelers were more likely to walk or ride bikes and
less likely to carpool. Gender difference also was seen in activity associated with
carpooling for simple, non-work tours. Female travelers were more likely to carpool for
simple, non-work tours, although the results were insignificant for their commuting tours.
This may be because women made more intra-household trips and shared more of the
responsibility for childcare (Rosenbloom, 2006). Further, households with children were
more likely to carpool, because of the mobility dependence of children.
With respect to the tour characteristic-related variables for non-work tours, it was
found that those beginning in the morning peak period of 7:00 – 9:00 A.M. were more
prone to use public transportation and carpools for chained trips. The purpose of the tour
was an extremely important predictor of the choice of transit mode as well. The analysis
showed that travelers were less likely to use transit if the primary purpose of the tour was
101

shopping or recreation, while non-motorized modes were more likely to be used for
recreation and leisure activities.
The analysis also showed that transit riders were sensitive to changes in walking
time to access public transit for a simple, non-work tour. The results suggested that a
considerable growth in transit ridership could be achieved by reducing the walking time
required to access transit. This result was consistent with the measured, land-use
variables that showed a positive association with the use of active transportation for non-
work tours. All of these results confirmed that the major contributor to increased use of
the active transportation mode was the ease with which people can walk to gain access to
their destination or a transit system.
The magnitude of the elasticity estimates of the walkability was very small
because the walkability variables represent combined measures of land use. Although
these combined measures do not tell us the elasticity with respect to each of land use
attributes, it is still meaningful to compare the elasticity of combined measures of land
use with that of other travel time and socio-economic variables given that land use
attributes could change simultaneously through planning and design; for example, higher
density is likely to be along with more balanced land use and more improved network
connectivity. Therefore, the results show that neighborhood walkability has only
marginal effects on active travel. The elasticity indicates that the share of walking/biking
for non-work trips is very sensitive to increase in walk time. Similarly, the share of transit
trips for non-work trips is sensitive to increase in transit walking access time.
102

103

104

105

Table 5. 10. Elasticity estimates for active travel mode

5.5. DISCUSSION

In this study, the influence of land-use factors on the behavioral aspects of choosing tour
modes was investigated with refined methodologies. Understanding the motives behind
such behaviors is important for policy makers to encourage use of active transportation
and to reduce automobile travel.
106

The results of this study indicated that travelers make strategic decisions about
their daily tour activities, not just about their single trips to work or to non-work activities
(Lee et al., 2006). Travelers seem to make different decisions concerning the modes of
their tours, and these decisions are affected by household structure, lifestyle, and the
locations of their residences. The statistical tests showed that neighborhood walkability,
complexity of tours, household characteristics, and the attractiveness of available
transport modes collectively influenced VMT and the choice of the travel mode that
would be used for the tours.
Canvassing the results of tour generation models, residential land use has a strong
influence on both tour complexity and VMT ―a result that is robust across the tour
purposes studied in this paper. People tended to economize their travel by traveling
shorter distances, by reducing the number of trip chaining when they live in a
neighborhood with compact urban form characteristics. This confirms the idea that new
urban designs with higher density, mixed land uses, grid-pattern street network, and the
proximity to the commercial activity locations will decrease miles traveled by private
vehicles. However, there is some evidence of induced automobile travel related to the
trips in a compact built environment of work location ―hence the evidence on reduced
automobile trip generation for commuting trips was not as robust as the evidence for non-
work trips.
To further investigate the behavioral aspect of individual travel decision, this
study developed behavioral frameworks and considered built environment factors in
107

shaping mode choice while generalizing cost of travel. The results of mode choice
models showed that neighborhood walkability, considering density, mixed use, and street
connectivity, and retail employment appeared to increase the use of active transportation,
i.e., transit, walking, and riding a bike, for a simple commute tour or a non-work tour.
However, this was not the case for a complex work tour. That is, choices of tour modes
for complex commute trips are less likely to be influenced by land-use patterns, and most
commuters tend to rely on private vehicles when they need to accomplish non-work,
intermediate activities on the way to work.
This supports the view that the effects of transportation policies aimed at reducing
automobile travel between home and work will be limited, because a large share of the
commuting trips are complex tours that often are not served adequately by relatively slow,
inflexible transit services (Krygsman et al., 2007). Conversely, the results suggested that
land use has a stronger influence on non-work-related tours than on work-related tours. In
particular, transit riders were more sensitive to changes in walking time to access public
transit than to changes in the time spent in the transit vehicle or the cost of the transit.
This confirmed the idea that new urban designs with higher density, mixed land uses,
grid-pattern street network, and the proximity to the commercial activity locations will
increase non-motorized travel and transit use, thereby decreasing automobile travel. The
effect of the destination walkability on modes used to reach the subsequent destinations
was remarkable in that it appeared to affect the complexity of non-work trips. This
suggests that land use has a strong influence on both the residential environment and the
destination’s environment for a chained, non-work activity.
108

CHAPTER 6. CONCLUSION

The central question posed in this dissertation was concerned with the role that the built
environment has in people’s travel behavior. The built environment was examined in
terms of psychological and objective aspects of the quality of the walking environment
that might affect travel patterns. The relationship between these two aspects of the built
environment and travel behavior was assessed in this dissertation by analyzing people’s
transit-use patterns and choice of tour modes.

6.1. SUMMARY OF FINDINGS

In this research, first, I examined whether the perceptions that people have about
the walking environment affected their willingness to use public transportation. Four
perceptional attributes were identified that affect regular transit use, i.e., physical safety,
personal safety, amenities, and perceived isolation. The results of this study showed that
unfavorable perceptions of environmental conditions were independently associated with
decreased regular transit use; however, these effects varied among different types of
neighborhoods. Personal safety related to crime and violence was the major concern that
was associated with decreased transit use in mixed land-use neighborhoods, whereas, in
low-density neighborhoods, isolation from the street environment and physical safety
concerns, including dangerous crosswalks, were the significant deterrents to the use of
109

public transportation. Notably, it was found that the association between the perceived
environment and transit use was stronger for low-income travelers than for higher-
income travelers. Low-income travelers viewed the conditions of their walking
environment as problematic more often than higher-income travelers, and it appeared that
their transit use was more likely to be affected by safety concerns than by other urban
design factors of their neighborhood. The findings suggest that safety concerns are more
important to low-income travelers than amenity concerns; therefore, enhancing the safety
of neighborhoods is the necessary first step to increasing the utility of transit for low-
income people.
The second part of the empirical analysis was to test whether a compact
neighborhood design has any association with tour complexity and the choice of tour
mode. The results of this study indicated that travelers make strategic decisions about
their daily tour activities, not just about their single trips to work or to non-work activities.
Travelers seem to make different decisions concerning the modes of their tours, and these
decisions are affected by household structure, lifestyle, and the locations of their
residences.
The differences in the impacts of walkability on choice of travel mode by the
types of tours were most relevant in the context of this study. Neighborhood walkability,
considering density, mixed land-use, street connectivity, and the intensity of retail
employment appeared to increase active travel, i.e., transit, walking, and riding a bike, for
a simple commute or a non-work tour. However, this was not the case for a complex
110

work tour because the choices of travel modes for complex commutes were less likely to
be influenced by land-use patterns, and most commuters tended to rely on private
vehicles when they needed to accomplish non-work, intermediate activities on the way to
work. This supports the view that the effects of transportation policies aimed at reducing
automobile travel between home and work will be limited, because a large share of the
commuting trips are complex tours that often are not served adequately by inflexible
transit services.
Conversely, the results suggested that the built environment had a stronger
influence on mode choice for non-work tours than for work tours. It appeared non-work
tours were more likely to be shorter in distance, more likely to be completed in simple
travel patterns, and were more likely to be taken by walking, biking, or using transit
compared to commute tours. Therefore, non-work related tours are the most likely
candidates to have mode choices affected by the built environment improvement.
Consistent with the effect of the measured neighborhood walkability, the results
showed that transit riders were more sensitive to changes in walking time to access public
transit than to changes in the time spent in the transit vehicle or the cost of the transit for
a non-work tour. The results suggested that considerable growth in transit ridership could
be achieved by reducing the walking time required to access transit. Notably, the effect of
the destination walkability on non-motorized modes used to reach the subsequent
destinations was remarkable in that it appeared to affect the complexity of non-work trips.
111

The results from these two empirical studies of the dissertation consistently
confirmed that the major contributor to increased use of the active transportation mode
was the ease with which people can walk to gain access to their destination or a transit
system.

6.2. CONTRIBUTION TO THE LITERATURE

This dissertation complements the literature on travel behavior in several important ways.
To the author’s current knowledge, it is the first analysis that examines the role of the
perceived environment in promoting transit by considering different types of transit users.
Since previous studies only captured the perceptions of passengers, little was known
about what kinds of perceptions prevent potential riders from taking transit. This
dissertation extends previous research by estimating differences in perceptions about the
walking environment between non-transit users, occasional users, and regular users.
In addition, unlike past research on access to transit, this dissertation recognizes
that the effects of perception on transit use can vary between different types of
neighborhoods. By observing the dynamic interaction of different types of neighborhoods
with travelers’ perceptions about the walking environment, the study provides a better
understanding concerning the particular types of neighborhoods that make a difference in
the influences of people’s perceptions on transit use.
112

The second empirical study also advanced past research in three important ways.
First, in this study, a conceptual model framework was developed in which travel
decisions are activity based, and it was used to assess the choice of tour mode in one
integrated framework considering the characteristics of the built environment and the
detailed, mode-specific travel time and cost for each leg of the trip within a tour. This
approach permits the modeling of the interaction among trip decisions that involve
spatially-displaced activities, while simultaneously incorporating alternative-specific
travel times and costs, thereby reflecting the attractiveness of the alternative mode of
travel. Second, the other central component of this study that advanced previous research
was that it included tests to indicate whether the built environment at trip destinations had
a measurable influence on the mode that was chosen for a tour. Unlike previous studies
that only considered the built environment at the origin of the trip, this study incorporated
analyses of the characteristics of the built environment at both the origin and the
destination of the tour.
Finally, to the author’s knowledge, there are no published studies that have tested,
in one integrated framework, the variations of effects of the built environment on mode
choice for different tour purposes as well as for different tour complexities. This study
made progress in overcoming these limitations, and it used a refined methodology to
contribute to the continuing effort to examine the relationship between the built
environment and the choice of tour mode.

113

6.3. POLICY AND PLANNING IMPLICATIONS

Contemporary cities and metropolitan regions face increasingly complex issues related to
transportation and land use, such as urban sprawl, traffic congestion, auto-dependence,
and the associated environmental and health risks. Investigating the spatial characteristics
of individual travel behavior provides important insights to policy makers who seek to
promote environmentally-sustainable travel choices and mobility patterns.
The results presented in this dissertation can help planners and policy makers
answer important questions, such as 1) Which policy will enhance access to transit for
low-income households?; 2) What kinds of policies are required to promote active travel?;
and 3) Which planning tools can support the development of walkable communities
throughout the urban and suburban neighborhoods of Los Angeles County?

6.3.1. Access to Transit by Low-Income Households

With respect to the access to transit by low-income households, the results of this study
showed that unfavorable perceptions of environmental conditions are a significant
deterrent to the use of public transportation. The perceived lack of a safe place to walk
was the top-ranked barrier to the use of public transportation by low-income travelers.
This suggests that safety needs are more important than amenity needs. Thus, a very
pleasant, clean, or comfortable environment is not likely to compel people to use public
transportation if they feel unsafe in doing so.
114

An important implication of this finding is that policy makers can increase transit
use by low-income people by enhancing neighborhood safety and by providing the
quality of local access to transit. Much of what was discovered in this study simply is not
within the control of transit agencies, which suggests that the transit agencies and other
public entities must work together to increase the use of public transportation. If the
neighborhood deterrence problem is to be solved, whether it relates to an unsafe walking
environment resulting from street crime or the dangers associated with fast traffic, active
policy decisions must be made and implemented to enhance public safety and to make
neighborhoods more walkable.

6.3.2. Neighborhood-Level Interventions to Promote Active Travel

Policy makers and practitioners have been interested in how the planner’s tools can
influence travel behavior in ways that can promote active travel, since active travel has
been recognized as an important factor of a healthy lifestyle, incorporating physical
activity into daily life. While the results are mixed, this study showed that the choice of
active transportation mode can be influenced by different approaches to urban design.
The evidence in this study supports a number of neighborhood-level interventions, i.e.,
increasing the densities of residential and commercial areas, increasing street
connectivity and route choices, balancing land-use mix, creating more sidewalks and
pedestrian paths, making it safer to cross the street, and enhancing public safety.
115

The findings also suggest that future interventions that are designed to increase
active travel should target non-work activities. The evidence here shows that
neighborhood design could have potential to increase active travel for non-work activities,
but it could have limited impact in the context of commute tours. In contrast to most non-
work activities, work activities are relatively fixed and spatially-constrained for workers.
As shown in the research, commute trips are less likely to be conducted locally. As
people travel longer distances to commute to work, they might be faced with many
intermediate activity requirements, which are not adequately served by relatively slow,
inflexible transit systems. These trends have continued as employment has become
decentralized or suburbanized over the past few decades in the Los Angeles area. In that
regard, transportation policies aimed at reducing automobile travel between home and
work could be limited in the context of neighborhood-scale initiatives.
This study provides innovative evaluation methods that will be useful as cities and
communities attempt to design and implement bottom-up, neighborhood-based
interventions that can enhance people’s quality of life at the local level. In particular, it is
critical for policy makers to developing strategies to estimate travel outcomes of ongoing
programs and community-based initiatives, such as Complete Streets, which have not
been evaluated for their effectiveness in enhancing safer and easier walking, biking, and
transit use.

116

6.4. DIRECTIONS FOR FUTURE RESEARCH

There are several possible extensions of this dissertation that can be pursued in the future.
First, this dissertation calls for additional scholarly attention on the connection among
multiple tours taken in an entire day. A better understanding of such connection will be
critical because people tend to organize their activities within a day schedule. Thus, there
may be substantial interactions across tour decisions when they make decisions about
scheduling in order to manage the time and space constraints associated with making
trips. A model that incorporates the duration of activities and the time of day will help
broaden our understanding of how tour decisions are constrained by the activity schedule.
The findings of this study provide some important guidance concerning the link between
tours and activity patterns, and one such finding was that people’s tour patterns are
affected by the time their trips begin (e.g., the A.M. peak period). However, additional
research on the interaction between tour decisions and the choice of activity patterns will
be valuable.
Second, since the characteristics of the built environment at both the origin and
the destination of a tour were incorporated in the analyses in this dissertation, it was
evident that the characteristics of the destination of the tour have a measurable influence
on choice of travel mode for non-work activities. However, the models used in the
dissertation did not explicitly treat the choice of tour mode as endogenous to selection of
the destination (Greenwald, 2006). For example, if travelers select the mode and the
117

destination simultaneously, then the error term of the model may be correlated with the
outcome of travel demand. Similarly, one may argue that the choice of the time their trips
begin may be endogenous to mode choice. Within the context of this study, the choice of
destination and the choice of departure times of trips are assumed exogenous to the model
system. However, future analyses must address these endogenous issues, and examine the
simultaneous causal relationships among mode choice, departure times of trips, and
destination choice.
Third, this dissertation relied on cross-sectional data that limit causal relations and
may be affected by unexpected situational variables at the time of assessment. In
particular, it is important to acknowledge that modeling the effects of land use on travel
demand can always give rise to the issue of self-selection. Travelers can choose
neighborhoods that have attributes consistent with their needs, travel abilities, and
mobility preferences. For example, if a household does not have a vehicle, then the
members of the household may choose to live in a neighborhood with good transit
facilities and hence use transit more often. Beyond a household’s economic constraints,
the members’ attitudes and preferences about travel modes also can influence their
tendency to choose locations. For example, if a household has a preference for walking
activity, then the household may decide to live in a walkable neighborhood. Since multi-
variate analysis was used in this dissertation to account for the effect of socio-
demographic characteristics and individual perception variables, it might partially
alleviate the effect of residential selection. In addition, self-selection may be less relevant
for low-income households who probably have more limited location choice. However,
118

to completely filter out the self-selection effect, future studies should expand on these
findings to include perspective longitudinal data collected at the individual level, which
would help to address appropriate time order, non-spuriousness, and causal mechanisms.

119

BIBLIOGRAPHY

Agrawal, Asha Weinstein, Marc Schlossberg, and Katja Irvin. 2008. “How Far, by Which
Route and Why? A Spatial Analysis of Pedestrian Preference.” Journal of Urban
Design 13 (1):81 - 98.

Alfonzo, Mariela A. 2005. “To Walk or Not to Walk? The Hierarchy of Walking Needs.”
Environment and Behavior 37 (6):808-836.

Alshalalfah, B. W., and Amer S. Shalaby. 2007. “Case Study: Relationship of Walk
Access Distance to Transit with Service, Travel, and Personal Characteristics.”
Journal of Urban Planning and Development 133 (2):114-118.

Bagley, Michael N., and Patricia 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.

Becker, Gary S. 1965. “A Theory of the Allocation of Time.” The Economic Journal 75
(299):493-517.

Ben-Akiva, Moshe E. Ben, and Steven R. Lerman. 1985. Discrete choice analysis: theory
and application to predict travel demand. Vol. 9. The MIT press.

Bhat, C. R, and F. S. Koppelman. 1999. "Activity-Based Modeling of Travel Demand."
In Handbook of Transportation Science, 35-61: Springer.

Blumenberg, Evelyn, Moira Donahue, Susan L Handy, Kristin Lovejoy, Caroline J
Rodier, Susan Shaheen, and James Volker. 2007. “Travel of Diverse Populations:
Literature Review.”

Boarnet, Marlon Gary, and Randall Crane. 2001. Travel by Design the Influence of
Urban Form on Travel.

Bowman, J. L., and M. E. Ben-Akiva. 2001. “Activity-Based Disaggregate Travel
Demand Model System with Activity Schedules.” Transportation Research Part A:
Policy and Practice 35 (1):1-28.

120

Brown, Barbara B., Carol M. Werner, and Naree Kim. 2003. “Personal and Contextual
Factors Supporting the Switch to Transit Use: Evaluating a Natural Transit
Intervention.” Analyses of Social Issues and Public Policy 3 (1):139-160.

Brownstone, David, and K A Small. 2005. “Valuing Time and Reliability: Assessing the
Evidence from Road Pricing Demonstrations.” Transportation Research Part A-
Policy and Practice 39 (4):279-293.

Calthorpe, Peter. 1993. The Next American Metropolis: Ecology, Community, and the
American Dream: Princeton Architectural Press.

Cao, Xinyu, and Patricia L. Mokhtarian. 2005. “How Do Individuals Adapt their Personal
Travel? A Conceptual Exploration of the Consideration of Travel-Related
Strategies.” Transport Policy 12 (3):199-206.

Cervero, Robert, and Carolyn Radisch. 1996. “Travel Choices in Pedestrian Versus
Automobile Oriented Neighborhoods.” Transport Policy 3 (3):127-141.

Cervero, Robert, and Kara Kockelman. 1997. “Travel Demand and the 3Ds: Density,
Diversity, and Design.” Transportation Research Part D: Transport and
Environment 2 (3):199-219.

Cervero, Robert. 2001. “Walk-and-Ride: Factors Influencing Pedestrian Access to
Transit.” Journal of Public Transportation 7 (3):1-23.

Cervero, Robert. 2002. "Built Environments and Mode Choice: Toward a Normative
Framework." Transportation Research Part D: Transport and Environment 7
(4):265-284.

Cervero, Robert. 2007. “Transit-Oriented Development's Ridership Bonus: a Product of
Self-Selection and Public Policies.” Environment and Planning A 39 (9):2068-
2085.

Chapin, Francis Stuart, and F. Stuart Chapin. 1974. Human Activity Patterns in the City:
Things People Do in Time and in Space. New York: Wiley.

121

Clay, Michael J., and Patricia L. Mokhtarian. 2004. “Personal Travel Management: the
Adoption and Consideration of Travel-Related Strategies.” Transportation
Planning and Technology 27 (3):181 - 209.

Crane, Randall. 2000. “The Influence of Urban Form on Travel: An Interpretive Review.”
Journal of Planning Literature 15 (1):3-23.

de Donnea, François Xavier. 1972. “Consumer Behaviour, Transport Mode Choice and
Value of Time: Some Micro-Economic Models.” Regional and Urban Economics 1
(4):355-382.

de Nazelle, Audrey, Brian J. Morton, Michael Jerrett, and Douglas Crawford-Brown.
2010. "Short Trips: An Opportunity for Reducing Mobile-Source Emissions?"
Transportation Research Part D: Transport and Environment 15 (8):451-457.

DeSerpa, Allan C. 1971. “A Theory of the Economics of Time.” The Economic Journal
81 (324):828-846.

Domencich, Thomas A., and Daniel McFadden. 1975. Urban Travel Demand-A
Behavioral Analysis.

El-Geneidy, A. M., P. Tetreault, and J. Surprenant-Legault. 2010. "Pedestrian Access to
Transit: Identifying Redundancies and Gaps Using a Variable Service Area
Analysis." Transportation Research Board 89th Annual Meeting. No. 10-0837.

Estupiñán, Nicolás, and Daniel A. Rodriguez. 2008. “The Relationship between Urban
Form and Station Boardings for Bogota’s BRT.” Transportation Research Part A:
Policy and Practice 42 (2):296-306.

Evenson, Kelly R., Molly M. Scott, Deborah A. Cohen, and Carolyn C. Voorhees. 2007.
“Girls' Perception of Neighborhood Factors on Physical Activity, Sedentary
Behavior, and BMI.” Obesity 15 (2):430-445.

Frank, Lawrence, Mark Bradley, Sarah Kavage, James Chapman, and T Keith Lawton.
2008. “Urban Form, Travel Time, and Cost Relationships with Tour Complexity
and Mode Choice.” Transportation 35 (1):37-54.

122

Frank, Lawrence, James F Sallis, Brian E Saelens, Lauren Leary, Kelli Cain, Terry L
Conway, and Paul M Hess. 2010. “The Development of a Walkability Index:
Application to the Neighborhood Quality of Life Study.” British Journal of Sports
Medicine 44 (13):924-933.

Gidlow, Christopher, Thomas Cochrane, Rachel C. Davey, Graham Smith, and Jon
Fairburn. 2010. “Relative Importance of Physical and Social Aspects of Perceived
Neighbourhood Environment for Self-Reported Health.” Preventive Medicine 51
(2):157-163.

Giuliano, Genevieve. 2003. “Travel, Location and Race/Ethnicity.” Transportation
Research Part A: Policy and Practice 37 (4):351-372.

Giuliano, Genevieve. 2005. "Low Income, Public Transit, and Mobility." Transportation
Research Record: Journal of the Transportation Research Board 1927 (1): 63-70.

Greenwald, Michael, and Marlon 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.

Greenwald, Michael J. 2006. “The Relationship between Land Use and Intrazonal Trip
Making Behaviors: Evidence and Implications.” Transportation Research Part D:
Transport and Environment 11 (6):432-446.

Hägerstraand, Torsten. 1970. “What about People in Regional Science?” Papers in
Regional Science 24 (1):7-24.

Handy, Susan. 1992. “Regional Versus Local Accessibility: Neo-Traditional
Development and Its Implications for Non-work Travel.” Built Environment 18
(4):253-267.

Handy, Susan. 2005. "Critical Assessment of the Literature on the Relationships among
Transportation, Land Use, and Physical Activity." Department of Environmental
Science and Policy, University of California, Davis. Prepared for the Committee on
Physical Activity, Health, Transportation, and Land Use. Washington:
Transportation Research Board.

123

Hanson, Susan, and James O. Huff. 1981. “Assessing Day-to-Day Variability in Complex
Travel Patterns.” Transportation Research Record 891:18-24.

Hanson, Susan. 1982. “The Determinants of Daily Travel-Activity Patterns: Relative
Location and Sociodemographic Factors.” Urban Geography 3:179-202.

Hanson, Susan. 2004. “The Context of Urban Travel: Concepts and Recent Trends.” in
chapter 1 in Hanson, S. and G. Giuliano. The Geography of Urban Transportation.
3rd edition. New York: Guilford Press.

Hensher, David A. 2001. “Measurement of the Valuation of Travel Time Savings.”
Journal of Transport Economics and Policy 35 (1):71-98.

Hess, Daniel Baldwin. 2009. “Access to Public Transit and its Influence on Ridership for
Older Adults in Two U.S. Cities.” Journal of Transport and Land Use 2 (1):3-27.

Hess, Stephane, Michel Bierlaire, and John W. Polak. 2005. “Estimation of Value of
Travel-Time Savings Using Mixed Logit Models.” Transportation Research Part
A: Policy and Practice 39 (2-3):221-236.

Hsiao, Shirley, Jian Lu, James Sterling, and Matthew Weatherford. 1997. “Use of
Geographic Information System for Analysis of Transit Pedestrian Access.”
Transportation Research Record: Journal of the Transportation Research Board
1604 (1):50-59.

Jiang, Yang, P. Christopher Zegras, and Shomik Mehndiratta. 2012. “Walk the Line:
Station Context, Corridor Type and Bus Rapid Transit Walk Access in Jinan,
China.” Journal of Transport Geography 20 (1):1-14.

Jou, Rong-Chang, Soi-Hoi Lam, Yu-Hsin Liu, and Ke-Hong Chen. 2005. "Route
Switching Behavior on Freeways with the Provision of Different Types of Real-
Time Traffic Information." Transportation Research Part A: Policy and
Practice 39 (5): 445-461.

King, Abby C, Cynthia Castro, Sara Wilcox, Amy A Eyler, James F Sallis, and Ross C
Brownson. 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):354-365.
124

Kitamura, Ryuichi. 1984a. “Incorporating Trip Chaining into Analysis of Destination
Choice.” Transportation Research Part B: Methodological 18 (1):67-81.

Kitamura, Ryuichi. 1984b. “Sequential Model of Interdependent Activity and Destination
Choices.” Transportation Research Record 987:81-89.

Kitamura, Ryuichi, and Lidia P. Kostyniuk. 1986. “Maturing Motorization and
Household Travel: The Case of Nuclear-Family Households.” Transportation
Research Part A: General 20 (3):245-260.

Kitamura, Ryuichi. 1988. "An Evaluation of Activity-Based Travel Analysis."
Transportation 15 (1): 9-34.

Kitamura, Ryuichi, Satoshi Fujii, and Eric I. Pas. 1997. "Time-Use Data, Analysis and
Modeling: Toward the Next Generation of Transportation Planning
Methodologies." Transport Policy 4 (4): 225-235.

Kitamura, Ryuichi, Patricia L. Mokhtarian, and Laura Daidet. 1997. “A Micro-Analysis
of Land Use and Travel in Five Neighborhoods in the San Francisco Bay Area.”
Transportation 24 (2):125-158.

Kockelman, Kara. 1997. “Travel Behavior as Function of Accessibility, Land Use
Mixing, and Land Use Balance: Evidence from San Francisco Bay Area.”
Transportation Research Record: Journal of the Transportation Research Board
1607 (1):116-125.

Krizek, KevinJ. 2003. “Neighborhood Services, Trip Purpose, and Tour-Based Travel.”
Transportation 30 (4):387-410.

Krygsman, Stephan, Theo Arentze, and Harry Timmermans. 2007. "Capturing Tour
Mode and Activity Choice Interdependencies: A Co-Evolutionary Logit Modelling
Approach." Transportation Research Part A: Policy and Practice 41 (10):913-933.

Kwan, Mei-Po. 1999. “Gender, the Home-Work Link, and Space-Time Patterns of
Nonemployment Activities.” Economic Geography 75 (4):370-394.

125

Kwan, Mei-Po. 2000. “Gender Differences in Space-Time Constraints.” Area 32 (2):145-
156.

Lam, Terence C., and Kenneth A Small. 2001. “The Value of Time and Reliability:
Measurement from a Value Pricing Experiment.” Transportation Research Part E
37:231-251.

Lee, Bumsoo, Peter Gordon, James E Moore, and Harry W Richardson. 2006.
“Residential Location, Land Use and Transportation: The Neglected Role of Non-
Work Travel.” Paper read at Western Regional Science Association 45th Annual
Conference.

Leslie, Eva, Brian Saelens, Lawrence Frank, Neville Owen, Adrian Bauman, Neil Coffee,
and Graeme Hugo. 2005. “Residents’ Perceptions of Walkability Attributes in
Objectively Different Neighbourhoods: A Pilot Study.” Health & Place 11 (3):227-
236.

Loukaitou-sideris, Anastasia. 1999. “Hot Spots of Bus Stop Crime.” Journal of the
American Planning Association 65 (4):395-411.

Loukaitou-Sideris, Anastasia, Robin Liggett, Hiroyuki Iseki, and William Thurlow. 2001.
“Measuring the Effects of Built Environment on Bus Stop Crime.” Environment
and Planning B: Planning and Design 28 (2):255-280.

Loukaitou-Sideris, Anastasia. 2006. “Is it Safe to Walk? Neighborhood Safety and
Security Considerations and Their Effects on Walking.” Journal of Planning
Literature 20 (3):219-232.

Loukaitou-Sideris, Anastasia, and Camille Fink. 2009. “Addressing Women's Fear of
Victimization in Transportation Settings.” Urban Affairs Review 44 (4):554-587.

Loutzenheiser, David. 1997. “Pedestrian Access to Transit: Model of Walk Trips and
Their Design and Urban Form Determinants around Bay Area Rapid Transit
Stations.” Transportation Research Record: Journal of the Transportation
Research Board 1604 (1):40-49.

126

Maat, Kees, Bert van Wee, and Dominic Stead. 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.

Matthies, Ellen, Silke Kuhn, and Christian A. Klöckner. 2002. “Travel Mode Choice of
Women.” Environment and Behavior 34 (2):163-177.

McMillan, Tracy E. 2005. “Urban Form and a Child’s Trip to School: The Current
Literature and a Framework for Future Research.” Journal of Planning Literature
19 (4):440-456.

Messenger, Todd, and Reid Ewing. 1996. “Transit-Oriented Development in the Sun
Belt.” Transportation Research Record: Journal of the Transportation Research
Board 1552 (1):145-153.

Miller, Eric J, Matthew J Roorda, and Juan Antonio Carrasco. 2005. "A Tour-Based
Model of Travel Mode Choice." Transportation 32 (4):399-422.

Mokhtarian, Patricia L, and Ilan Salomon. 1999. “Travel for the Fun of It.” Access 15
(Fall): 26-31.

Mokhtarian, Patricia L, and Xinyu 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.

Mota, Jorge, Mariana Almeida, Paula Santos, and José Carlos Ribeiro. 2005. “Perceived
Neighborhood Environments and Physical Activity in Adolescents.” Preventive
Medicine 41 (5–6):834-836.

Moudon, Anne, Paul Hess, Mary Snyder, and Kiril Stanilov. 1997. “Effects of Site
Design on Pedestrian Travel in Mixed-Use, Medium-Density Environments.”
Transportation Research Record: Journal of the Transportation Research Board
1578 (1):48-55.

National Research Council (US). Committee on Physical Activity, et al. 2005. “Does the
Built Environment Influence Physical Activity?: Examining The Evidence.” No.
282. Transportation Research Board.

127

O'Sullivan, Sean, and John Morrall. 1996. “Walking Distances to and from Light-Rail
Transit Stations.” Transportation Research Record: Journal of the Transportation
Research Board 1538 (1):19-26.

Owen, Neville, Ester Cerin, Eva Leslie, Lorinne dutoit, Neil Coffee, Lawrence D Frank,
Adrian E Bauman, Graeme Hugo, Brian E Saelens, and James F Sallis. 2007.
"Neighborhood Walkability and the Walking Behavior of Australian Adults."
American Journal of Preventive Medicine 33 (5):387-395.

Pas, Eric I., and Frank S. Koppelman. 1987. “An Examination of the Determinants of
Day-to-Day Variability in Individuals' Urban Travel Behavior.” Transportation 14
(1):3-20.

Redmond, Lothlorien S., and Patricia L. Mokhtarian. 2001. “The Positive Utility of the
Commute: Modeling Ideal Commute Time and Relative Desired Commute
Amount.” Transportation 28 (2).

Rodríguez, Daniel A., and Joonwon Joo. 2004. "The Relationship between Non-
Motorized Mode Choice and the Local Physical Environment." Transportation
Research Part D: Transport and Environment 9 (2):151-173.

Rosenbloom, Sandra. 1989. “Trip Chaining Behaviour: a Comparative and Cross Cultural
Analysis of the Travel Patterns of Working Mothers.” Gender, Transport and
Employment.

Rosenbloom, Sandra. 2006. “Understanding Women’s and Men’s Travel Patterns.” Paper
read at Research on Women’s Issues in Transportation. Report of a Conference:
Transportation Research Board: Washington, DC.

Ryan, Sherry, and Lawrence F. Frank. 2009. “Pedestrian Environments and Transit
Ridership.” Journal of Public Transportation 12 (1):39-57.

Saelens, Brian, James Sallis, and Lawrence Frank. 2003. “Environmental Correlates of
Walking and Cycling: Findings from the Transportation, Urban Design, and
Planning Literatures.” Annals of Behavioral Medicine 25 (2):80-91.

Sallis, James F, Brian E Saelens, Lawrence D Frank, Terry L Conway, Donald J Slymen,
Kelli L Cain, James E Chapman, and Jacqueline Kerr. 2009. “Neighborhood Built
128

Environment and Income: Examining Multiple Health Outcomes.” Social Science
& Medicine 68 (7):1285-1293.

Santos, Adelia, Nancy McGuckin, Hikari Yukiko Nakamoto, Danielle Gray, and Susan
Liss. 2011. Summary of Travel Trends: 2009 National Household Travel Survey.

Schwanen, Tim, and Martin Dijst. 2002. “Travel-Time Ratios for Visits to the Workplace:
the Relationship between Commuting Time and Work Duration.” Transportation
Research Part A: Policy and Practice 36 (7):573-592.

Shaheen, Susan A, Jade Bejamin-Chung, Denise Allen, and Linda Howe-Steiger. 2009.
“Achieving California’s Land Use and Transportation Greenhouse Gas Emission
Targets under AB 32: An Exploration of Potential Policy Processes and
Mechanisms.”

Shiftan, Yoram. 2008. “The Use of Activity-Based Modeling to Analyze the Effect of
Land-Use Policies on Travel Behavior.” The Annals of Regional Science 42 (1):79-
97.

Small, Kenneth A., Clifford Winston, and Jia Yan. 2005. “Uncovering the Distribution of
Motorists' Preferences for Travel Time and Reliability.” Econometrica 73 (4):1367-
1382.

Srinivasan, Sumeeta, and Joseph Ferreira. 2002. "Travel Behavior at the Household Level:
Understanding Linkages with Residential Choice." Transportation Research Part D:
Transport and Environment 7 (3):225-242.

Stradling, Stephen, Michael Carreno, Tom Rye, and Allyson Noble. 2007. “Passenger
Perceptions and the Ideal Urban Bus Journey Experience.” Transport Policy 14
(4):283-292.

Sundquist, Kristina, Ulf Eriksson, Naomi Kawakami, Lars Skog, Henrik Ohlsson, and
Daniel Arvidsson. 2011. "Neighborhood Walkability, Physical Activity, and
Walking Behavior: The Swedish Neighborhood and Physical Activity (SNAP)
Study." Social Science & Medicine 72 (8):1266-1273.

Train, Kenneth. 1986. Qualitative Choice Analysis: Theory, Econometrics, and an
Application to Automobile Demand. Vol. 10: The MIT press.
129

Tseng, Yin-Yen, and Erik T. Verhoef. 2008. “Value of Time by Time of Day: A Stated-
Preference Study.” Transportation Research Part B: Methodological 42 (7-8):607-
618.

Wallace, Richard, Daniel Rodriguez, Christopher White, and Jonathan Levine. 1999.
“Who Noticed, Who Cares? Passenger Reactions to Transit Safety Measures.”
Transportation Research Record: Journal of the Transportation Research Board
1666 (1):133-138.

Wen, Ming, Louise C. Hawkley, and John T. Cacioppo. 2006. “Objective and Perceived
Neighborhood Environment, Individual SES and Psychosocial Factors, and Self-
Rated Health: An Analysis of Older Adults in Cook County, Illinois.” Social
Science & Medicine 63 (10):2575-2590.

Werner, Carol M., Barbara B. Brown, and Jonathan Gallimore. 2009. “Light Rail Use is
More Likely on ‘Walkable’ Blocks: Further Support for Using Micro-Level
Environmental Audit Measures.” Journal of Environmental Psychology.

Wibowo, Sony Sulaksono. 2008. “Frequency of Transit Use and Access Characteristics:
Case Study of Metro Manila.” In Transportation Research Board 87th Annual
Meeting, Washington D.C.

Ye, Xin, Ram M Pendyala, and Giovanni Gottardi. 2007. “An Exploration of the
Relationship between Mode Choice and Complexity of Trip Chaining Patterns.”
Transportation Research Part B: Methodological 41 (1):96-113.

Zhang, Ming. 2004. "The Role of Land Use in Travel Mode Choice: Evidence from
Boston and Hong Kong." Journal of the American Planning Association 70
(3):344-360.

Abstract (if available)

Abstract The central question posed in this dissertation is concerned with the role that the built environment has in people's travel behavior. The built environment is examined in terms of psychological and objective aspects of the quality of the walking environment that might affect travel patterns. To examine the relationship between these two aspects of the built environment and travel behavior, this dissertation analyzes people's transit-use patterns and choice of travel modes using the Los Angeles County sample which is a subset of the data from the 2009 National Household Travel Survey California Add-on. ❧ This study first examines whether the perceptions that low-income people have about the walking environment affect their willingness to use public transportation by analyzing self-reported frequency of transit use and measured neighborhood attributes. A principal component analysis is used to reduce many overlapping perceptional variables to latent factors that are used in subsequent models of transit use. Four perceptional attributes are identified that affect regular transit use, i.e., physical safety, personal safety, amenities, and perceived isolation. The results of this study show that unfavorable perceptions of environmental conditions are independently associated with decreased transit use

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

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

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

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

Children’s travel behavior in journeys to school

PDF

Spatial and temporal expenditure-pricing equity of rail transit fare policies

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

A capability-based approach to defining performance characteristics of the built environment

PDF

The impact of demographic shifts on automobile travel in the United States: three empirical essays

PDF

Household mobility and neighborhood impacts

PDF

Lessons from TAP implementation: obstacles and solutions to improve the transit users experience

PDF

Effects of the perceived and objectively assessed environment on physical activity in adults and children

PDF

Environmental justice in real estate, public services, and policy

PDF

Spatial analysis of urban built environments and vehicle transit behavior

PDF

Congestion pricing with an unpriced time period and with heterogeneous user groups

PDF

Household carbon footprints: how to encourage adoption of emissions‐reducing behaviors and technologies

PDF

Location of warehouses and environmental justice: Three essays

PDF

Walkability as 'freedom': the ecology of school journey in inner city Los Angeles neighborhoods

PDF

Productive frictions and urbanism in transition: planning lessons from traffic flows and urban street life in Ho Chi Minh City, Vietnam

Asset Metadata

Creator Lee, Jeongwoo (author)

Core Title The built environment, tour complexity, and active travel

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Policy, Planning, and Development

Publication Date 11/11/2015

Defense Date 09/25/2013

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

Tag active travel,built environment,OAI-PMH Harvest,perception,transit use,travel behavior,walkability

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Giuliano, Genevieve (committee chair), Boarnet, Marlon (committee member), Moore, James Elliott, II (committee member), Schweitzer, Lisa (committee member)

Creator Email jeongwol@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-346757

Unique identifier UC11296245

Identifier etd-LeeJeongwo-2148.pdf (filename),usctheses-c3-346757 (legacy record id)

Legacy Identifier etd-LeeJeongwo-2148.pdf

Dmrecord 346757

Document Type Dissertation

Format application/pdf (imt)

Rights Lee, Jeongwoo

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