Upload
khangminh22
View
1
Download
0
Embed Size (px)
Citation preview
Structural Analysis on Activity-travel Patterns, Travel Demand, Socio-demographics, and
Urban Form: Evidence from Cleveland Metropolitan Area
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University
By
Yu-Jen Chen
Graduate Program in City and Regional Planning
The Ohio State University
2017
Dissertation Committee:
Professor Gulsah Akar, Advisor
Professor Zhenhua Chen
Professor Jean-Michel Guldmann
ii
Abstract
Research on travel behavior continues to be one of the most prominent areas in
the transportation area. Planners and policymakers try to understand and manage travel
behavior. Making and implementation of travel demand management (TDM) policies
greatly rely on the understanding of the determinants of activity-travel patterns and travel
demand. Among the activity-travel patterns, trip chaining and joint travel have received
much research interest. Trip chaining is typically viewed as a home-based tour that
connects multiple out-of-home activities. Joint travel is commonly defined as traveling
with others. Travel demand is generally measured by trip generation and travel distances.
Investigating different aspects of travel behavior helps us better understand the links
between activity participation and mobility, and improves the evaluation of the
transportation infrastructure investments and policies such as high occupancy vehicle
(HOV) lanes and vehicle miles traveled (VMT) reduction programs.
Several studies have regarded trip chaining, joint travel, trip generation, and travel
distances as different dimensions of travel behavior to be examined in terms of various
socio-demographics and urban form factors. However, limited work has been done to use
activity-travel patterns as mediating variables and analyze how trip chaining and joint
travel shape the resulting travel demand. Furthermore, relationships between travel
behavior and urban form factors at out-of-home activity locations remain unclear. Based
iii
on the 2012 travel survey data from the Cleveland Metropolitan Area, this study first
investigates the relationships among trip chaining, joint travel, home-based tour
generation, and travel distances at three different levels: tour, individual, and household
levels. Second, the influences of socio-demographics and urban form factors at tour
origins and destinations on travel behavior are examined simultaneously. Lastly, while
using trip chaining and joint travel as mediating variables, this study estimates the
mediating effects of socio-demographics and urban form via activity-travel patterns on
travel demand. The Structural Equation Modelling (SEM) approach is applied.
The study reveals the existence of significant relationships between activity-travel
patterns and travel demand. Trip chaining is negatively associated with joint travel. While
it increases travel distances, this effects gets weakened through its indirect effect via
decreased tour generation. Joint travel appears to increase tour generation but decrease
the travel distances. Most socio-demographics have significant effects with expected
signs on travel behavior. The analysis suggests that urban form factors at tour origins and
destinations play important roles on the resulting travel demand. Some urban form factors
may not have direct effects on travel demand but have significant indirect effects on tour
generation or travel distances through activity-travel patterns. This research presents how
activity-travel patterns shape travel demand and concludes that trip chaining and joint
travel should be taken into consideration while analyzing travel demand. The findings on
socio-demographics and urban form factors can be used as inputs to improve the future
evaluation of transportation projects and help planners integrate land-use strategies as
iv
tools to change people’s travel behavior. This will further mitigate the negative
externalities associated with our travel patterns.
vi
Acknowledgments
Countless people have been with me and offered great help during my doctoral
study. I deeply thank to my advisor, Dr. Gulsah Akar, for her endless support, continuous
encouragement, and valuable assistance on both my academic studies and my life
throughout this long journey at The Ohio State University. She guided me to be a
professional researcher, problem slover, and independent thinker. I am lucky to have her
as my advisor.
I am truly grateful to my committee members, Dr. Jean-Michel Guldmann and Dr.
Zhenhua Chen for their insightful suggestions on my dissertation work and future
research directions.
I would like to express my thank to Dr. Rachel Garshick Kleit, for her planning
theory lectures that motivated me to explore theoretical foundation of transportation
planning. I am also thankful to Dr. Philip A. Viton for his advice on the quantitative
methodology and techniques. Many thanks to Dr. Mark Moritz, for the research
oppurtunities in the Department of Anthopology at The Ohio State Univeristy. I want to
acknowledge the project manager Ms. Rebekah Anderson of Ohio Department of
Transportation for providing the datasets of my research and dissertation.
vii
I am thankful to my friends and colleague in the doctoral program, Chih-Hao
Wang, Mi Namgung, Miyoung Park, Emre Tepe, Na Chen, Hyeyoung Kim, Daisuke
Nagase, Seungbeom Kang, Burcu Ozdemir, Youngbin Lym, Seunghoon Kim, Kailai
Wang, and Yujin Park, for their friendship and support.
I am indebted to to my family, my dad Ching-Sen Chen, my mom Yu-Fen Lin,
and my sister Yi-Ling Chen. They have always been helping me with their constant love
and concern. Lastly and importantly, my special appreciation goes to my girlfriend, Chia-
Yu Lin. Without her encouragement and support, I would never have been able to
accomplish this long journey.
viii
Vita
2006 ...............................................................B.S. Civil Engineering, National Central
University
2011 ...............................................................M.S. Civil Engineering, The Ohio State
University
2011 to 2012 .................................................Graduate Research Assistant, City and
Regional Planning, the Ohio State
University
2011 to 2013 .................................................Graduate Research Assistant, Department of
Anthropology, the Ohio State University
2012 to 2013 .................................................Graduate Research Assistant, , Department
of Veterinary Preventive Medicine, the Ohio
State University
2013 to 2016 .................................................Graduate Teaching Associate, City and
Regional Planning, the Ohio State
University
Publications
Y. Chen, & G. Akar. (2017). Using Trip Chaining and Joint Travel as Mediating
Variables to Explore the Relationships among Travel Behavior, Socio-demographics and
Urban Form. Forthcoming, Journal of Transport and Land Use, 11(1), 1-16
M. Moritz, S. Handa, Y. Chen, & N. Xiao. (2015). Herding Contracts and Pastoral
Mobility in the Far North Region of Cameroon. Human Ecology, 43(1), 141-151
M. Moritz, I. M. Hamilton, Y. Chen, & P. Scholte. (2014). Mobile pastoralists in the
Logone Floodplain distribute themselves in an Ideal Free Distribution. Current
Anthropology, 55(1), 115-122.
ix
M. Moritz, I. M. Hamilton, P. Scholte, & Y. Chen (2014). Ideal Free Distributions of
Mobile Pastoralists in Multiple Seasonal Grazing Areas. Rangeland Ecology &
Management, 67(6), 641-649
Fields of Study
Major Field: City and Regional Planning
x
Table of Contents
Abstract ............................................................................................................................... ii
Dedication ........................................................................................................................... v
Acknowledgments.............................................................................................................. vi
Vita ................................................................................................................................... viii
Table of Contents ................................................................................................................ x
List of Tables .................................................................................................................... xv
List of Figures ................................................................................................................. xvii
Chapter 1: Introduction ....................................................................................................... 1
1.1 Activity-based Approach and Activity-travel Patterns ................................................. 3
1.2 Problem Definition........................................................................................................ 6
1.3 Research Objectives ...................................................................................................... 7
1.4 Conceptual Framework ................................................................................................. 9
1.5 Organization ................................................................................................................ 11
Chapter 2: Literature Review ............................................................................................ 13
2.1 Travel Behavior and Socio-demographics .................................................................. 14
2.1.1 Trip Chaining and Socio-demographics .................................................................. 14
2.1.2 Joint Travel and Socio-demographics ...................................................................... 18
xi
2.1.3 Trip/Tour Generation and Socio-demographics ...................................................... 22
2.1.4 Travel Distances and Socio-demographics .............................................................. 26
2.2 Travel Behavior and Urban Form Factors .................................................................. 30
2.2.1 Trip Chaining and Urban Form Factors ................................................................... 31
2.2.2 Joint Travel and Urban Form Factors ...................................................................... 35
2.2.3 Trip/Tour Generation and Urban Form Factors ....................................................... 36
2.2.4 Travel Distances and Urban Form Factors .............................................................. 41
2.3 Relationships among Different Aspects of Travel Behavior ...................................... 46
2.4 Discussion of the Existing Literature .......................................................................... 47
Chapter 3: Research Design and Data .............................................................................. 50
3.1 Research Design.......................................................................................................... 50
3.2 Data ............................................................................................................................. 53
3.2.1 Datasets of Activity-travel Patterns and Travel Demand ........................................ 53
3.2.2 Travel Demand: Tour Generation and Travel Distances ......................................... 55
3.2.3 Activity-travel Patterns: Trip Chaining ................................................................... 55
3.2.4 Activity-travel Patterns: Joint Travel ....................................................................... 56
3.2.5 Classification of Tours and the Primary Activity for a Home-based Tour .............. 58
3.2.6 Datasets for Urban Form Factors ............................................................................. 59
3.3 Variables of Interest and Descriptive Statistics Analysis ........................................... 60
xii
3.3.1 Endogenous and Exogenous Variables .................................................................... 60
3.3.2 Descriptive Statistics of the Sample Data ................................................................ 66
3.3.3 Travel Behavior Comparison Analysis .................................................................... 76
Chapter 4: Methodology ................................................................................................... 83
4.1 Path Analysis and Structural Equation Modeling ....................................................... 84
4.2 Model Estimation and Assumptions ........................................................................... 86
4.3 Model Fit Indices ........................................................................................................ 87
4.3.1 Chi-square (χ2) ......................................................................................................... 89
4.3.2 Root-mean-square Error of Approximation (RMSEA) ........................................... 90
4.3.3 Comparative Fit Index (CFI) ................................................................................... 90
4.3.4 Standardized Root Mean Square Residual (SRMR) ................................................ 91
Chapter 5: Results and Discussions for the Tour-based Model ........................................ 92
5.1 Model Fit Indices in the Tour-level Model ................................................................. 93
5.2 Links Between Activity-travel Patterns and Travel Demand in the Tour-level Model
........................................................................................................................................... 94
5.3 Direct Effects of Socio-demographics on Travel Behavior in the Tour-level Model . 95
5.4 Direct Effects of Urban Form Factors on Travel Behavior in the Tour-level Model . 97
5.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Tour-level Model
........................................................................................................................................... 99
xiii
5.6 Total Effects of Exogenous Variables on Travel Behavior in the Tour-level Model 100
Chapter 6: Results and Discussions of the Individual-level Model ................................ 104
6.1 Model Fit Indices for the Individual-level Model..................................................... 105
6.2 Links Between Activity-travel Patterns and Travel Demand in the Individual-level
Model .............................................................................................................................. 106
6.3 Direct Effects of Socio-demographics on Travel Behavior in the Individual-level
Model .............................................................................................................................. 107
6.4 Direct Effects of Residential Urban Form Factors on Travel Behavior in the
Individual-level Model.................................................................................................... 109
6.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Individual-level
Model .............................................................................................................................. 111
6.6 Total Effects of Exogenous Variables on Travel Demand in the Individual-level
Model .............................................................................................................................. 112
Chapter 7: Results and Discussions for Household-level Model.................................... 117
7.1 Model Fit Indices for the Household-level Model .................................................... 118
7.2 Links Between Activity-travel Patterns and Travel Demand in the Household-level
Model .............................................................................................................................. 119
7.3 Direct Effects of Socio-demographic on Travel Behavior in the Household-level
Model .............................................................................................................................. 120
xiv
7.4 Direct Effects of Urban Form Factors on Travel Behavior in the Household-level
Model .............................................................................................................................. 122
7.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Household-level
Model .............................................................................................................................. 123
7.6 Total Effects of Exogenous Variables on Travel Demand in the Household-level
Model .............................................................................................................................. 124
Chapter 8: Conclusions ................................................................................................... 129
8.1 Summary and Contributions ..................................................................................... 130
8.1.1 Relationships among Activity-travel Patterns and Travel Demand ....................... 131
8.1.2 Relationships Between Socio-demographics and Travel behavior ........................ 133
8.1.3 Relationships Between Urban Form Factors and Travel behavior ........................ 134
8.2 Policy Implications ................................................................................................... 136
8.3 Limitations and Future Directions ............................................................................ 139
Reference ........................................................................................................................ 142
xv
List of Tables
Table 1. Variables of Interest ............................................................................................ 64
Table 2. Endogenous and Exogenous Variables in Each Model ...................................... 65
Table 3. Descriptive Statistics of Sample Data Used for Tour-level Model .................... 67
Table 4. Descriptive Statistics of Travel Behavior in Sample Data Used for Individual-
level and Household-level Models.................................................................................... 70
Table 5. Descriptive Statistics of Socio-demographics in the Sample Data Used for
Individual-level and Household-level Models .................................................................. 75
Table 6. Descriptive Statistics of the Travel Behavior at the Tour and Individual Levels
Across the Binary Personal Characteristics ...................................................................... 79
Table 7. Kruskal-Wallis Test Results of the Travel Behavior at the Tour and Individual
Levels Across the Binary Personal Characteristics .......................................................... 80
Table 8. Descriptive Statistics of the Travel Behavior at the Tour and Individual Levels
Across the Presence and Age Stages of the Youngest Child in the Household ................ 81
Table 9. Kruskal-Wallis Test Results of the Travel Behavior at the Tour and Individual
Levels Across the Presence and Age Stages of the Youngest Child in the Household .... 82
Table 10. Model Fit Indices and Acceptable Fit Criteria.................................................. 89
Table 11. Model Fit Indices of the Tour-level Model ...................................................... 94
Table 12. Estimation Results for the Tour-level Model (Standardized Effects) ............. 101
Table 13. Model Fit Indices of the Individual-level Model ............................................ 106
xvi
Table 14. Estimation Results for the Individual-level Model (Standardized Effects) .... 114
Table 15. Model Fit Indices of the Household-level Model ........................................... 119
Table 16. Estimation Results for the Household-level Model (Standardized Effects) ... 126
xvii
List of Figures
Figure 1. Conceptual Framework ..................................................................................... 10
Figure 2. Tour-level Model Framework ........................................................................... 51
Figure 3. Individual-level Model Framework ................................................................... 52
Figure 4. Household-level Model Framework .................................................................. 53
Figure 5. Study Area – the Cleveland Metropolitan Area ................................................ 54
Figure 6. Histogram of Travel Distances at the Tour Level ............................................. 68
Figure 7. Histogram of Joint Travel at the Tour Level ..................................................... 68
Figure 8. Histogram of Trip Chaining at the Tour Level .................................................. 69
Figure 9. Histogram of VMT at the Individual Level ....................................................... 70
Figure 10. Histogram of the Number of Auto Tours at the Individual Level ................... 71
Figure 11. Histogram of Joint Travel at the Individual Level .......................................... 71
Figure 12. Histogram of Trip Chaining at the Individual Level ....................................... 72
Figure 13. Histogram of VMT at the Household Level .................................................... 73
Figure 14. Histogram of the Number of Auto Tours at the Household Level .................. 73
Figure 15. Histogram of Joint Travel at the Household Level ......................................... 74
Figure 16. Histogram of Trip Chaining at the Household Level ...................................... 74
Figure 17. Illustration of direct, indirect, and total effects ............................................... 86
Figure 18. Model Framework at the Tour Level............................................................... 93
Figure 19. Illustration of Direct Effects in the Tour-level Model ................................... 103
xviii
Figure 20. Model Framework at the Individual Level .................................................... 105
Figure 21. Illustration of Direct Effects in the Individual-level Model .......................... 116
Figure 22. Model Framework at the Household Level ................................................... 118
Figure 23. Illustration of Direct Effects in the Household-level Model ......................... 128
1
Chapter 1: Introduction
Recent studies have placed much attention on travel behavior because the
negative outcomes of travel, primarily vehicle miles traveled (VMT) and single
occupancy vehicle (SOV) use, have increased dramatically. Based on data from the U.S.
Department of Transportation (DOT), VMT increased almost 150% in urban areas from
1980 to 2014. This increase is quite significant, particularly when compared with the
population growth of approximately 30% (Frey, 2012; United States Census Bureau,
2016; United States Department of Transportation; Bureau of Transportation Statistics,
2016). The increasing VMT has been viewed as one of the key contributors to various
negative externalities. For example, The U.S. Environmental Protection Agency (EPA)
reports that transportation produces about 25% of the total carbon dioxide, the primary
contributor to the greenhouse effects in 2014 (United States Environmental Protection
Agency, 2016b). Furthermore, other negative externalities such as energy consumption,
traffic congestion, and crashes have been regarded as results of the increases in VMT and
SOV use (Akar, Chen, & Gordon, 2016; Brownstone & Golob, 2009; Rentziou, Gkritza,
& Souleyrette, 2012). In order to seek solutions to mitigate the negative externalities
associated with our transportation patterns, planners and policy makers try to understand
and manage individuals’ travel behavior through their activity-travel patterns and travel
demand. Instead of expanding the existing transportation infrastructure, many
2
metropolitan areas have been shifting their policies to reduce SOV travel, make driving
less attractive, and increase the share of alternative travel modes such as public transport,
bicycles or walking (Gärling et al., 2002; Loukopoulos, Jakobsson, Gärling, Schneider, &
Fujii, 2004).
One of the well-known policy sets is the transportation demand management
(TDM) strategies, which refers to various actions employed to increase the transportation
system efficiency and to achieve traffic congestion reduction and energy conservation
while focusing on the safety and mobility of the road users. These actions are adopted to
modify individuals’ travel behavior by congestion pricing, encouraging off-peak travel,
and reducing the number of single-occupancy vehicles (SOV) trips (S. Bricka, Moran,
Miller, & Hudson, 2015; Obermann, 2012). In addition to the aforementioned strategies,
the transportation authorities and professionals have developed programs to improve air
quality by reducing vehicular emissions since the 1990s (S. G. Bricka, 2008). Fixing
America’s Surface Transportation (FAST) Act was designed and passed in 2015 to
provide long-term funding certainty for surface transportation infrastructure planning and
investment. FAST also provides more than $11 billion federal funding for Congestion
Mitigation and Air Quality Improvement Program (CMAQ) from 2016 to 2020. The
CMAQ was established in 1991 to help the State and local governments improve their
transportation projects to meet the requirements of the Clean Air Act. The Clean Air Act
became effective in 1963 and has been one of the essential environmental laws in U.S.
requiring regulatory controls for air pollution with several major amendments in 1970,
1977, and 1990 (United States Environmental Protection Agency, 2016a). These policies
3
encourage individuals to maintain their vehicles, consolidate trips, and make more trips
by transit, carpools, bicycles, and by walking (McGuckin, Zmud, & Nakamoto, 2005).
Making of informed strategies, the implication of policies, and investments on
improvement projects greatly depend on the understanding of different dimensions of
travel behavior including people’s activity-travel patterns and travel demand. From the
academic and research standpoint, different aspects of travel behavior are modeled
separately against various determinants to identify the influential factors of daily travel
patterns. The factors usually include socio-demographics and residential built
environment variables. The findings on these factors help planners and decision makers
to forecast travel demand under changing social, cultural, economic conditions, and land-
use patterns.
1.1 Activity-based Approach and Activity-travel Patterns
The need for a deeper understanding of travel outcomes led to the development of
activity-based approach. One of the fundamental concepts of the approach is that travel
demand is derived from individuals’ desire to pursue activities distributed in time and
space (Bhat & Koppelman, 2003; Kitamura, 1988; M. G. McNally & Rindt, 2007).
Consequently, the accuracy of travel demand modeling and forecasting rely on a better
understanding of activity-travel patterns. This calls for a closer look at the way people
organize their travel for desired activities, including the time they travel, the locations of
out-of-home activities, the transportation modes they choose, and the companions they
travel with.
4
Among various activity-travel patterns, trip chaining and joint travel are the two
patterns that have received much research interest in recent decades. Several different
definitions of trip chaining have been developed since the 1970s. The definition that
Adler and Ben-Akiva (1979) propose is that a trip tour (trip chain) begins and ends at an
individual’s home. In addition, a tour should consist of a set of consecutive trip links (trip
segments) that link at least one sojourn (out-of-home activity). Holzapfel (1986) states
that a trip chain should include three or more trip segments and originate and end at
home. Goulias and Kitamura (1991) describe trip chaining as a sequence of trip segments
that exist between a pair of locations such as home, work, and school. After reviewing the
related literature, Primerano, Taylor, Pitaksringkarn, and Tisato (2008) propose a
generalized definition of trip chaining: “a trip chain is a multi-segmented trip originating
and ending at the home and containing primary and secondary activities in between”
(Primerano et al., 2008). The “home-based tour” became a widely-used definition of trip
chaining in previous research. Several studies point out that trip chaining has been a
growing trend and has become a significant part of individuals’ daily life. S. G. Bricka
(2008) indicates that 61% of U.S. working-age adults have chained trip segments and
generated 68% of average daily VMT in 2001. One of the reasons working age adults
have such high percentages of chained trips is that workers are more likely to consolidate
non-work activities into their commute tours to maintain personal and household needs.
Some other researchers found that travelers who chain their trip segments tend to travel
by private cars and drive alone or with household companions (McGuckin et al., 2005;
Pendyala, Yamamoto, & Kitamura, 2002; Wallace, Barnes, & Rutherford, 2000).
5
The focus on studying joint travel making has been supported by the recognition
that an individual’s activity-travel decisions are not only made by him or herself. The
decisions may be affected by the other members of the same household. Joint travel is
therefore commonly defined as traveling with household members in existing studies (Ho
& Mulley, 2013a, 2013b, 2015; Vovsha, Petersen, & Donnelly, 2003). Traveling with
household companions has been found to comprise a significant proportion of a region’s
travel. The home-based tours which include joint trip segments represent a substantial
percentage (around 40 to 50%) of total tours generated in many metropolitan areas, such
as Ohio and New York in the United States, Sydney in Australia, and Hong Kong in
China (Ho & Mulley, 2013a, 2013b, 2015; Lin & Wang, 2014; Vovsha et al., 2003). As a
result, such intra-household dependencies and interactions should be explicitly
accommodated in travel demand analysis for improving the accuracy of forecasting. The
decisions on travel companion, which are motivated by engaging in activities jointly or
alone, may affect the vehicle type choice and occupancy. The number of vehicles and the
nature of vehicle types in the network significantly influence the level of traffic
congestion and emissions (Srinivasan & Bhat, 2008). Therefore, understanding the
motivation for joint travel and the circumstances under which it takes place are critical
for proposing transportation policy as well as evaluating projects that promote non-SOV
travel options and introduce high occupancy vehicle (HOV) lanes and high occupancy
toll (HOT) lanes (Ho & Mulley, 2015; Vovsha et al., 2003).
Trip chaining and joint travel are two major activity-travel patterns which capture
the complexity of home-based tours, interactions among household members, and travel
6
schedule coordination under the spatial-temporal constraints. From travelers’ perspective,
consolidating trip segments and coordinating travel patterns with other household
members can help reduce the use of travel resources and achieve more efficient time use
while participating in out-of-home activities. Therefore, investigating the behavioral
mechanisms that lead to trip chaining and joint travel provides a deeper understanding of
people’s activity and travel patterns, which in turn improves the overall quality of travel
demand models. In light of this recognition, the findings of this research can be used as
inputs to evaluate the potential success of transportation projects and various policy
interventions.
1.2 Problem Definition
Several researchers have considered land-use and transportation planning as a
prominent strategy for decreasing car-oriented behaviors and achieving sustainable
transportation patterns. Empirical evidence has shown the existence of strong
associations between urban form factors and travel behavior. For example, people who
live in low-density suburban areas may have greater auto dependency. On the other hand,
neighborhoods with compact development and greater land-use mixes may provide
shorter commutes, cyclists and pedestrian friendly environments, and better accessibility
to the amenities for the residents. These all increase the use of public transit and non-
motorized transportation modes. Existing research unravels the links between activity-
travel patterns and the resulting travel demand (S. G. Bricka, 2008; Duncan, 2015;
Gopisetty & Srinivasan, 2013; Vovsha et al., 2003). For example, the total travel
7
distances and the number of home-based tours may substantially depend on the
complexity of the individuals’ home-based tours and their decisions on traveling with
others. Limited work has been done to use activity-travel patterns as mediating variables
and analyze how trip chaining and joint travel shape the resulting travel demand. While
many studies have examined the influence of urban from around residential
neighborhoods on travel behavior, the effects of out-of-home locations where people
pursue their activities remain unclear.
Within this context, this dissertation aims to investigate how activity-travel
patterns shape the resulting travel demand. This research uses trip chaining and joint
travel as two measurements of activity-travel patterns, and tour generation and travel
distances as two outcomes of travel demand. The analysis aims to address the
interrelationships among four different dimensions of travel behavior (trip chaining, joint
travel, tour generation, and travel distances), identify the contribution of activity-travel
patterns on travel demand, and investigate the influence of urban form factors at
residence and out-of-home locations on travel behavior while controlling for personal and
household characteristics.
1.3 Research Objectives
The existing studies that examine the relationships among travel behavior, socio-
demographics, and urban form factors provide limited evidence on how trip chaining and
joint travel affect tour generation and travel distances at the same time. In this regard, the
objective of this dissertation is to contribute to the empirical understanding of the
8
interrelationships among activity-travel patterns, travel demand, socio-demographics, and
urban form. The basis of this research hinges on answering the following main questions.
First, do activity-travel patterns affect the resulting travel demand? If so, how do
trip chaining and joint travel influence tour generation and travel distances? Specifically,
what are the links among trip chaining, joint travel, tour generation, and travel distances?
Are there any differences in the relationships among travel behavior at different levels
(e.g., at the home-based tour level, at the individual level, and at the household level)?
Second, to what extent do urban form factors affect activity-travel patterns and
travel demand directly? In addition, how does the urban form factors at residential and
out-of-home locations influence travel behavior? Are there any differences across urban
form factors effect at different locations? If so, what types of locations have greater
impacts?
Third, to what extent socio-demographics and urban form factors indirectly affect
the resulting travel demand through intermediate activity-travel patterns?
While many empirical studies have examined trip chaining, joint travel, tour
generation, and travel distances as separate functions of socio-demographics and urban
form factors, this dissertation investigates the interconnections among multiple aspects of
travel behavior simultaneously while accounting for personal characteristics, household
attributes, and urban form factors at three different levels: home-based tour level,
individual level, and household level.
9
1.4 Conceptual Framework
By using the household travel survey data collected by the Ohio Department of
Transportation (ODOT) from the Cleveland Metropolitan Area in 2012, the purpose of
this dissertation is to analyze the interrelationships among trip chaining, joint travel, tour
generation, travel distances, socio-demographics and urban form factors at three different
levels. Figure 1 illustrates the conceptual framework, presenting the links among several
primary elements included in the study.
Built on the previous studies, the conceptual model proposes to extend the current
travel demand modeling framework by incorporating two indicators of activity-travel
patterns (trip chaining and joint travel). The framework consists of three key variable
sets, together with their measurements and associations with each other. The variable sets
included in this framework are generally used in existing research and can be listed as: (i)
travel behavior, (ii) socio-demographics, and (iii) urban form factors. Travel behavior is
captured by two activity-travel patterns (trip chaining and joint travel) and two travel
demand measurements (tour generation and travel distances). Socio-demographics
include individual’s gender, age, driver’s license ownership, employment status,
household size, household vehicle ownership, household income, and household
structure. Urban form factors describe the neighborhood characteristics by residential
density, retail employment density, non-retail employment density, intersection density,
bus stop density, and job-population index.
Within this framework, this research hypothesizes that the way individuals
organize their activity-travel patterns affects their travel demand. The connections among
10
the different aspects of travel behavior are of particular interest. Socio-demographics and
urban form factors are assumed to be important determinants of travel behavior. This
study develops comparable behavioral models at three levels: home-based tour level,
individual level, and household level. The basic unit of analysis and variable sets used in
each model are adjusted to each level. The travel behavior variables are trip chaining,
joint travel, and travel distances for each home-based tour in the tour-level model. The
travel behavior variables include trip chaining, joint travel, home-based tour generation,
and VMT for each individual and household in the individual-level model and household-
level model respectively. The socio-demographics used in the tour-level and individual-
level models are personal characteristics and household attributes. The socio-
demographics included in the household-level model are only household attributes. The
urban form factors examined in the tour-level model are built environment variables
measured at home and out-of-home locations. The urban form factors investigated in the
individual and household levels only capture the residential built environments.
Figure 1. Conceptual Framework
11
1.5 Organization
This dissertation is organized as follows. The next chapter discusses the existing
literature related to socio-demographics, urban form factors, and travel behavior. Chapter
2 first presents the effects of personal and household characteristics on travel behavior.
Then it reviews how urban form factors affect individuals’ activity-travel patterns and
travel demand. The third section provides the associations among different dimensions of
travel behavior. This is followed by the limitation of the previous studies and how this
research adds to the existing knowledge.
Chapter 3 presents the research design and the data used in this research. The
research design section details the measurements of activity-travel patterns and travel
demand at each level. The data for this study come from ODOT’s household travel
surveys. The data were collected from the Cleveland Metropolitan Area in 2012. This
section presents the study area, variables of interest, and descriptive analysis of variables.
The modeling approach applied in this dissertation is introduced in Chapter 4. In
order to disentangle the complex relationships among socio-demographics, urban form,
and travel behavior simultaneously, Structural Equation Modeling (SEM) approaches are
used in this research to estimate the interconnections among activity-travel patterns and
travel demand as well as the effects of socio-demographics and urban form factors on
travel behavior.
Chapter 5 discusses the model results based on home-based tours. The socio-
demographics are represented by personal and household characteristics, the urban form
12
factors consist of built environment variables at home and out-of-home locations, and
travel behavior includes trip chaining, joint travel, and travel distances for each home-
based tour.
Chapters 6 and 7 provide analyses at the individual and household levels,
respectively. In Chapter 6, travel behavior is presented by trip chaining, joint travel, the
number of auto tours traveled by individuals, and individual VMT. Chapter 7 presents the
estimated model at the household level, examining the influence of household attributes
and urban form factors on household activity-travel patterns, the number of auto tours by
households, and household VMT.
Chapter 8 concludes with the research findings and main contributions of this
dissertation in terms of policy implications. The limitations of this study and future
research directions are also discussed in Chapter 8.
13
Chapter 2: Literature Review
This chapter presents the existing research related to activity-travel patterns and
travel demand. Activity-travel patterns discussed in this chapter include trip chaining and
joint travel. Trip chaining represents the complexity of a home-based tour via the
number of activity episodes within one home-based tour (Golob, 1986; Liu, 2012; Noland
& Thomas, 2007; Wallace et al., 2000; Wang, 2014). Some researchers have considered
joint trip making as an alternative pattern to solo trip making (Chandrasekharan &
Goulias, 1999; Ho & Mulley, 2013a, 2013b, 2015; Srinivasan & Bhat, 2008; Vovsha et
al., 2003). Trip/tour generation and travel distances stand out as two critical indicators of
travel demand and both factor into many travel demand models (Akar et al., 2016;
Kitamura, Mokhtarian, & Laidet, 1997; Lu & Pas, 1999; Zhang, Hong, Nasri, & Shen,
2012). Researchers have developed various analytical methods to explain trip chaining,
joint travel, trip/tour generation, and travel distances in terms of various determinants.
Two sets of variables, socio-demographics and urban form factors, appear most
often in existing travel behavior literature. The socio-demographics consist of travelers’
information, including personal characteristics and household attributes. The urban form
factors describe the land-use patterns and built environment of a given neighborhood.
This chapter provides an overview of the relationships among travel behavior, socio-
14
demographics, and urban form factors. This chapter contains four sections. The first
section discusses the effects of socio-demographics on travel behavior. The second
section reviews the links between urban form factors and travel behavior. The third
section presents the associations between activity-travel patterns and travel demand. The
last section summarizes the limitations of existing research and opportunities for future
research.
2.1 Travel Behavior and Socio-demographics
Previous research has revealed that socio-demographic characteristics play
important roles in explaining travel behavior. Investigating the influence of socio-
demographics on travel behavior helps planners evaluate the market response to
transportation policies. Socio-demographics usually consist of personal characteristics,
including an individual’s gender, age, employment status, and driver’s license ownership,
while household attributes commonly include household size, income, vehicle
ownership/number of vehicles, lifestyle, and household structure.
2.1.1 Trip Chaining and Socio-demographics
Statistical modeling approaches for count data (e.g., Poisson regression and
negative binomial regression) have been frequently used in existing studies to examine
trip chaining, a variable often operationalized as the number of stops within a home-
based tour. Using data from the 1990 Household Travel Survey of Puget Sound Regional
Council (PSRC), Wallace et al. (2000) apply a negative binomial regression approach to
identify the potential factors affecting the propensity of trip chaining. Their model
includes the socio-demographic variables of gender, age, employment status, household
15
income, number of household vehicles, number of children in the household, and number
of adults in the household. The authors note that women have a greater likelihood of
chaining trip segments than men. They find that the necessity of driving to work
decreases the tendency to consolidate trip segments. Conversely, the necessity of driving
to pick up children increases the propensity for trip chaining. Employment status also
stands out as an influential variable: as the number of workdays per week increases, the
likelihood of chaining trips decreases. On the other hand, people who work at home have
a higher tendency to consolidate trip segments. With regard to household attributes,
household size and household income correlate negatively with trip chaining.
Using the 2001 National Household Travel Survey (NHTS) Data, Noland and
Thomas (2007) develop an ordered probit model and a negative binomial model to
explain the tour complexity in terms of socio-demographics, land-use patterns, and tour-
specific variables. The personal characteristics in their analysis are gender, age, race, and
employment. The household attributes include household size, household income, and
household structure. The authors find that being female and age both positively correlate
with tour complexity. The model results reveal that different household structures have
different influences on trip chaining. As compared to the households without children,
those with children less than 15 years old tend to make complex tours.
Some research has investigated trip chaining by estimating the probability of
consolidating trip segments. Using the 2001 National Household Travel Survey (NHTS)
Data, S. G. Bricka (2008) develops logit models to examine trip chaining patterns of
working age adults. The author uses gender, age, education status, employment status,
16
working settings (working duration, occupation, and the auto requirement for work) to
describe the individuals’ characteristics. The number of adults in the household, the
number of children in the household, the number of household vehicles, household
income, and the ratio of vehicles to workers in the household are household attributes in
the analysis. Some of the socio-demographics trends identified in this research aligned
with those found by Noland and Thomas (2007) and Wallace et al. (2000). For example,
females have a greater tendency to make complex tours than males. Middle-aged adults
(ages 35 to 44) are more likely to chain trips than those in other age groups. People with
higher education levels and non-workers tend to consolidate trip segments. The number
of adults and the number of children in the household affect the chance of combining
trips in opposite directions: while the number of adults in the household is negatively
correlated to trip chaining, the number of children in the household is positively
associated with the complexity of tours. It is interesting to note that increasing the
number of household vehicles increases the probability of trip chaining in the analysis. S.
G. Bricka (2008) suggests that it is likely that the study considers trip chaining
propensities by the working age adults, not all adults workers and non-workers.
Ma, Mitchell, and Heppenstall (2014) investigate the propensity to chain trip
segments by applying discrete choice approaches based on the 2008 Activity Daily
Survey Data collected from the workers in Beijing, China. Six socio-demographic
variables are used in the analysis: gender, age, occupation, personal income, household
size, and the presence of children in the household. Most socio-demographics trends
match those identified in prior research: women are inclined to make complex tours than
17
men, workers with larger households tend to make simple tours, and high-income
workers are more likely to make complicated tours (S. G. Bricka, 2008; Kitamura &
Susilo, 2005; McGuckin & Murakami, 1999; Noland & Thomas, 2007; Wallace et al.,
2000).
Several studies consider trip chaining an endogenous variable along with other
travel behavior in structural analysis. Kitamura and Susilo (2005) propose several
structural equation models to investigate travel time, the number of trip chains, the
number of trip segments in trip chains, and non-work activity participation
simultaneously among non-workers in Osaka, Japan. They use gender, age, driver’s
license ownership, household size, the number of household vehicles, and household
structure to characterize the socio-demographics. The analysis shows that female non-
workers with children tend to consolidate more trip segments than those without children.
Consistent with the findings of other studies (S. G. Bricka, 2008; Wallace et al., 2000),
the authors point out that the adults between ages of 35 and 44 are more likely to make
complex tours. Non-workers from larger households have higher propensities to make
simple tours.
Yang et al. (2010) use the Activity-travel Survey Data from Shangyu, China, to
develop a structural analysis of the relationships among socio-demographics, activity
participation, and trip chaining. Gender, age, employment status, and education level are
included to describe household heads’ characteristics. The number of workers in the
household and household income are two household attributes examined in the analysis.
The model results show that age is positively associated with tour complexity. The
18
authors suggest that employment status has negative effects on tour complexity, that
household heads with higher education levels tend to make complex home-based tours,
and that the number of household workers is negatively associated with tour complexity,
which all match the findings of S. G. Bricka (2008). Household income has positive
effects on trip chaining patterns, which also echoes the findings from Wallace et al.
(2000).
Van Acker and Witlox (2011) implement an SEM approach and model the links
among land use patterns, car availability, car use for commuting, trip chaining, and
commuting distances while controlling for socio-demographics and land-use patterns at
home locations and workplaces. The analysis is based on the 2000 to 2001 Ghent
(Belgium) Travel Behavior Survey Data, in which the socio-demographics contain
gender, age, marital status, whether auto is needed during work hours, household size, the
number of children aged below 6 years in the household, household income, and the ratio
of autos and drivers in the household. The authors find that people who always need
autos during work hours are more likely to make complex tours. In agreement with the
findings of Wallace et al. (2000) and S. G. Bricka (2008), the analysis indicates that
household size negatively affects tour complexity, whereas the number of children in the
household positively affects trip chaining.
2.1.2 Joint Travel and Socio-demographics
The studies of joint travel originated from the development of models explaining
the intra-household interactions and time allocations within a household during the
1980s. The first conceptual framework is developed by Townsend (1987), which includes
19
the interdependencies within a household while analyzing household activity behavior.
van Wissen (1991) uses the Household Travel Survey Data from Netherlands and
proposes structural models to analyze time allocation for out-of-home activities for
household members. The authors conclude that people travel jointly mostly for shopping,
recreation, and visits.
Some studies consider a trip as a joint trip if it is taken in the company of one or
more household members. They feature various discrete choice models to estimate the
relationship between explanatory variables and the probability of choosing solo versus
joint travel (Chandrasekharan & Goulias, 1999; Ho & Mulley, 2013a, 2013b, 2015;
Srinivasan & Bhat, 2008; Vovsha et al., 2003). Chandrasekharan and Goulias (1999) use
the Puget Sound Transportation Panel (PSTP) Data and investigate the probability of
traveling jointly. The explanatory variables for describing individuals’ characteristics are
gender, age, driver license ownership, and employment status. The household attributes
used in the analysis are household size, household income, the number of adults in the
household, the number of children in the household, and the number of household
vehicles. The authors report that those between the ages of 35 and 65 are less likely to
travel together with household members. The model results reveal that an individual’s
occupation has some impact on joint travel outcomes. Professionals make more joint trips
than those with other occupations, and those who classify themselves as a “driver” tend
to make fewer joint trips. As for household attributes, travelers from young households
(multi-member households under 35) are more likely to travel solo. Furthermore, the
20
authors indicate that the number of household vehicles is negatively correlated with joint
travel.
Using data from the mid-Ohio and the New York region, Vovsha et al. (2003)
examine decisions of traveling solo versus jointly for non-work activities at the
household level in terms of household income, household auto ownership, and household
size and composition. The authors find that low-income (less than $30,000 per year)
households have a strong tendency to implement joint travel for shopping. Full-time
workers have greater tendency to travel jointly. In disagreement with Chandrasekharan
and Goulias (1999), the analysis shows that households with higher auto ownership (i.e.,
a greater number of autos than workers in the household) are more likely to travel jointly
for eating out and shopping activities. This finding is surprising because owning multiple
cars generally facilitates more individual activities, while a shortage of cars may motivate
joint traveling. The authors explain that joint travel planning occurs at the household
level and may not result from car shortage. The model results reveal that different
household structures may result in different effects on joint travel making. Households
with more workers make fewer joint trips, while households with more non-workers tend
to travel jointly for maintenance and discretionary activities. In addition, increasing
number of students in the household increases the propensity of traveling solo, whereas
the number of children in the household is a strong factor favoring joint trips for
shopping and maintenance activities.
Ho and Mulley (2013a) focus on joint travel patterns of the car-negotiating
households (households that have fewer autos than driver's license holders) in Sydney,
21
Australia. The authors proposed nine types of joint home-based tours in their study: fully
joint tours, drop-off tours, pickup tours, shared tides tours, joint drop-off tours, joint
pickup tours, joint drop-off and pickup tours, joint in middle tours, and individual tours.
The study applies discrete choice models to examine the joint travel patterns and travel
mode choices against household size, household composition, household income, and
time and cost of different travel modes. Household size and composition have significant
influence on joint travel. The authors find that individuals in larger households (i.e., with
more than five members) are more likely to travel with other household members.
Mothers have a greater propensity for involvement in drop-off tours, pick-up tours, and
fully joint tours. The model results reveal that the presence of children in the household
has negative effects on traveling solo. The authors conclude that the tours for education
are less likely to be solo tours and tours for dropping off household members,
maintenance activities, and discretionary activities are more likely to be partially joint
tours.
Lin and Wang (2014) examine the individuals’ joint travel patterns based on the
2010 Hong Kong Activity-travel Survey Data. They use an SEM approach and address
the roles of individuals’ socio-demographics and social-network factors in shaping joint
activity and travel participation and their choice of companion. The personal socio-
demographics include gender, age, education attainment, and employment status.
Household attributes comprised household auto ownership, the presence of children in
the household, the presence of elderly in the household, household income, and
household size. The model results suggest that an individual’s age is negatively
22
associated with traveling with friends. In accordance with findings by Vovsha et al.
(2003), the authors indicate that students and full-time workers are more likely to travel
alone and the presence of children in the household increases the probability of traveling
jointly with family members. On the contrary, the presence of elderly people in the
household increases the chance of traveling solo. The analysis reveals that socio-network
attributes (emotional support, instrumental support, social companionship, and contacts
with family or friends) have significant influence on joint activity and travel patterns.
People who get emotional support from their family members are less likely to travel solo
but more likely to travel with household companions. Similarly, people who get
emotional support from their friends tend to travel with their friends.
2.1.3 Trip/Tour Generation and Socio-demographics
Various statistical models have been used to examine the factors that influence
trip/tour generation, a variable generally defined as the number of trips or tours traveled
in a given time period (Cao, Mokhtarian, & Handy, 2009; Cervero & Gorham, 1995;
Chatman, 2009; Goulias & Kitamura, 1991; Goulias, Pendyala, & Kitamura, 1990;
Kitamura et al., 1997). Based on travel data from the Detroit Metropolitan Area, Goulias
et al. (1990) explain household trip generation in terms of socio-demographics and
residential built environment. Gender and age describe the personal characteristics of
household heads. The household attributes consist of household size, household income,
the number of children in the household, the number of adults in the household, the
number of adult males and females in the household, household lifecycle stages, the
number of household vehicles, and the number of licensed drivers in the household. The
23
authors find that these socio-demographics are strong explanatory variables of trip
generation. For example, increasing number of adults in the household increases the
number of work and school trips. The number of licensed drivers in the household has
positive effects on the number of work and social trips. The household size is positively
correlated with the frequency of school trips. The number of children between the ages of
5 and 15 in the household positively affects the number of school and social trips. The
number of adult females in the household increases the generation of shopping and
school trips.
Using the data from the Puget Sound Transportation Panel (PSTP), Wallace,
Mannering, and Rutherford (1999) employ a Poisson model and a negative binomial
model to evaluate the effects of socio-demographics on different types of trip generation
(home-based work trips, home-based shopping trips, home-based other trips, work-based
other trips, and other-based other trips). Household size, household income, the number
of household vehicles, the number of adults in the household, and the number of working
days per week serve as the socio-demographics. The model results show that larger
households tend to make more home-based trips for shopping and other activities, more
work-based trips, and more other-based other trips. Wealthier households make more
home-based work and work-based other trips. The number of household vehicles is a
strong explanatory variable as it is positively correlated with all types of trips. Similar to
Goulias et al. (1990), increasing number of adults in the household increases home-based
work trip generation.
24
Jang (2005) develop a negative binomial model, a zero-inflated Poisson model,
and a zero-inflated negative binomial model to analyze household trip generation based
on the data from Jeonju City, Korea. The author investigates how travel demand is
influenced by eight socio-demographic characteristics at the household level: household
size, the number of workers in the household, household income, building type,
household size, the number of household vehicles, and household head’s occupation. The
results indicate that household head’s occupation affects household travel behavior: blue-
collar workers make fewer trips than white- collar workers. Supporting the findings of
Wallace et al. (1999), Jang’s (2005) findings suggest household size, household income,
and the number of household vehicles are positively correlated with household trip
generation. Consistent with this trend, the number of workers in the household is
positively associated with the frequency of household trips.
Several studies feature SEM approach to examine trip generation along with other
endogenous travel behavior variables. Using the data from the 1994/95 Oregon-
Southwest Washington Activity and Travel Survey, Lu and Pas (1999) propose several
SEMs and look into the interconnections among socio-demographics, activity
participation, and travel behavior. There are eight socio-demographic variables in the
model: age, gender, driver’s license ownership, employment status, household income,
the number of household vehicles, the number of workers in the household, and the
number of children in the household. Regarding personal characteristics, being female,
age, and holding a driver’s license are positively associated with the number of home-
based tours. In contrast, workers make fewer tours as compared to the counterparts. As
25
for the household attributes, the number of children and the number of workers in the
household are positively correlated with home-based tour generation.
Kuppam and Pendyala (2001) implement SEM approach and analyze the
relationships among socio-demographics, activity engagement, and the number of
complex work tours based on data from the Washington, D.C., Metropolitan Area.
Personal characteristics include gender, age, and employment status. Household attributes
consist of household income, the number of commuters in the household, the number of
household bicycles, and the number of household vehicles. Consistent with the findings
from Lu and Pas (1999), being female is positively associated with trip generation. They
also found that commuters over 60 years old undertake fewer complex work trip chains,
presumably due to their semi-retired status. The findings of household attributes show
that household size and the number of commuters in the household negatively affect the
number of complex home-based work tours. In addition, the household income and the
number of household bicycles are positively associated with the number of complex work
tours.
Silva, Morency, and Goulias (2012) utilize the 2003 Travel Survey Data from the
Greater Montreal Area and propose a SEM to analyze the workers’ trip generation by
three different modes (auto, transit, and non-motorized transportation). The exogenous
socio-demographics consist of gender, age, household total annual income, household
size, the average age of household members, the number of workers in the household,
and household structure. The authors indicate that increasing individual age increases the
number of auto trips but decreases the number of transit trips. As compared to females,
26
males generate fewer transit trips. The number of workers in the household is positively
related to the frequency of transit trips. The model results show that older households and
households with two individuals make fewer auto trips. Following a similar trend,
households with greater income levels, households with teenagers, and households with
two individuals make fewer non-motorized trips.
2.1.4 Travel Distances and Socio-demographics
Travel distance stands out as another important travel outcome and has received
much research attention in the past few decades. Stead (1999) applies multiple regression
analysis to explore the relationships among built environment, socio-demographics, and
individual travel patterns by using the data from four National Travel Surveys (1979-81,
1985-86, 1989-91, and 1991-93) in Great Britain. The author selects gender, age,
employment status, and driver’s license ownership as personal characteristics. The
household attributes in the analysis are household size, household composition and
structure, household auto ownership, the number of licensed drivers in the household, and
the number of workers in the household. As for the effects of personal characteristics, the
analysis reveals that men travel longer distances than women. Individuals aged between
30 and 39 travel more than those in other age groups. People with drivers’ licenses travel
further than those without. Full-time workers and students travel further than others do.
Household size has negative effects on individual travel distances. Increasing number of
household vehicles increases the travel distances. The model results show that household
composition is strongly associated with travel distances. For instance, residents of
households with one employee often travel further than those of other household types.
27
Individuals from households with one or two adults travel further than those from
households with more adults.
Bento, Cropper, Mobarak, and Vinha (2005) employ the 1990 Nationwide
Personal Transportation Survey (NPTS) Data and investigate travel mode choice, driving
decision (whether or not to drive to work), vehicle choice, and vehicle miles traveled
(VMT) of commuters in 114 U.S. urban areas. The authors use ordinary least squares
(OLS) regression to examine the annual VMT as a function of socio-demographics,
residential land-use patterns, weather attributes, and travel cost for one-, two-, and three-
or-more-vehicle households. Income, education level, the number of male and female
adults, the number of non-working adults, the number of children, and the presence of the
elderly constitute the household socio-demographics in the analysis. The authors find that
the one-vehicle and two-vehicle households with elderly members make fewer VMT than
the counterparts. The number of children between the ages of 17 and 21 in one- and
three-or-more-vehicle households has positive effects on VMT. The household income
and the number of working male adults positively affect annual VMT for all types of
households.
Using the 1999 Household Travel Survey Data from the Mid-Ohio region, Akar et
al. (2016) explain mean (per trip) and total (per day) travel distances at the individual
level. Spatial error models are developed to investigate the effects of socio-
demographics, land-use patterns, and transportation infrastructure. In their study, gender,
age, driver’s license status, education level, race, and employment status are categorized
as personal characteristics. Household income and household vehicle ownership are
28
introduced as household attributes. The model results indicate that being female is related
to shorter mean trip distances. Consistent with Stead (1999), travelers in other age groups
(0 to 16 years, 17 to 22 years, and over 65 years) travel shorter distances on average than
those between the ages of 36 to 50. Owning a valid driver’s license, having a bachelor’s
degree or above, and being employed are positively correlated to average and total travel
distances. The same trend emerges for the household attributes: household income and
the ratio of vehicles to drivers in the household are positively correlated with average and
total travel distances.
Feng, Dijst, Wissink, and Prillwitz (2017) explore the effects of socio-
demographics and built environment on travel distances based on the 2008 Nanjing
Residents Travel Survey Data. The authors include gender, age, education level,
household auto ownership, household income, and household types as socio-demographic
variables in the regression analysis. They find that travelers between the ages of 20 and
29 travel further than those in other age groups (40 to 49 and more than 50). Some model
results echo the findings from the former studies (Akar et al., 2016; Bento et al., 2005).
For instance, being female is negatively correlated with travel distances. Education level
and household income have positive effects on travel distances. With respect to
household types, households of single members and households of couples travel further
than core family households do. Feng et al. (2017) explain this finding as the emergence
of a consumption society and an upsurge in out-of-home activities.
Some studies apply the SEM approach and include travel distances as one of
endogenous variables to investigate the links among gas prices, transport emissions,
29
vehicle usage, and travel behavior (Dillon, Saphores, & Boarnet, 2015; Liu, 2012; Liu &
Shen, 2011; Song, Diao, & Feng, 2016; Van Acker & Witlox, 2011). Based on the travel
survey data from the Baltimore Metropolitan Area, Liu (2012) investigates the influence
of socio-demographics and built environment on travel behavior and energy consumption
by implementing the SEM approach. The author first develops a structural model based
on an all home-based tour sample. The all tour sample is further disaggregated by travel
purposes (work, mixed of work and non-work, and non-work) and travel modes (auto and
transit). Finally, the author estimates the structural models separately based on these five
sub-samples (work tours, mixed of work and non-work tours, non-work tours, auto tours,
and transit tours). The socio-demographics consist of gender, age, race, education level,
employment status, household income, and household vehicle ownership. The model
results indicate that males travel farther than females in work-tour model and auto-tour
model. Employment status is a strong explanatory determinant of travel distances: it has
positive effects on travel distances in every model except for the non-work-tour model.
As for household attributes, the number of household vehicles is positively associated
with travel distances in the all-tour model, work-tour model, and mixed-of-work-and-
non-work-tour model.
By using a Southern California subsample of the 2009 National Household Travel
Survey (NHTS) Data, Dillon et al. (2015) develop a SEM and analyze the relationships
among gas prices, vehicle fuel efficiency, residential built environment, household socio-
demographics, and household travel demand. Different models are estimated based on the
sample of household VMT for all trips, household VMT for work trips, and household
30
VMT for non-work trips. The socio-demographics examined in the analysis are
household income, household size, the number of workers in the household, the number
of children under 16 in the household, the number of adults between the ages of 16 and
24 in the household, and the ratio of vehicles to drivers in the household. The model
results reveal that household income and the number of workers in the household are
strong explanatory variables as they positively affect household VMT in every model.
Household size appears to increases household VMT for all trips and work trips. The
number of adults between the ages of 16 and 24 is positively correlated with household
VMT for work trips but negatively associated with household VMT for non-work trips.
2.2 Travel Behavior and Urban Form Factors
The links of urban form and travel behavior have been investigated and reviewed
since the 1980s with the increasing new urbanism and smart growth studies (Cao, 2007;
Cervero & Kockelman, 1997). Urban form factors generally refer to the built
environment characteristics for describing a given neighborhood. The well-known three
dimensions (3Ds) of built environment (density, diversity, and design) are widely
integrated into the existing travel behavior research (Cao, 2007; Cervero & Kockelman,
1997; Ewing & Cervero, 2001b, 2010). S. L. Handy, Boarnet, Ewing, and Killingsworth
(2002) argue that the built environment consists of three major components: urban
design, land use, and transportation systems. Based on these conceptual frameworks,
urban form factors commonly include population density, employment density, housing
density, distance to the central business district (CBD), intersection density, public transit
31
density, and land-use mix entropy. The investigation of the relationships between travel
behavior and urban form helps planners and policy makers better assess the transportation
projects and land-use policies in different areas. The following section addresses the
effects of different urban form factors on activity-travel patterns (e.g., trip chaining and
joint travel) and travel demand (e.g., tip generation and travel distances) in previous
literature.
2.2.1 Trip Chaining and Urban Form Factors
Many existing studies have examined the associations between trip chaining and
the urban form factors at residences. Based on the Puget Sound Transportation Panel
(PSTP) Data from 1989 to 1998, Krizek (2003) addresses how urban form settings
contribute to the travel behavior. The author applies principal component analysis to
create a single measure of Neighborhood Accessibility (NA) by combining three built
environment variables for each 150-meter grid cell: the number of housing, the number of
retail employees, and average block area. NA correlates negatively with average block
area but is positively associated with the number of housing and the number of retail
employees. The results of regression analysis show that NA has negative effects on trip
chaining. This finding indicates that individuals living in more assessable residential
areas make complex tours less often.
In another study, Noland and Thomas (2007) consider population density as a
proxy of land-use mix, urban design, and regional accessibility. They examine the effects
of population density on tour complexity using the 2001 National Household Travel
32
Survey (NHTS) Data. Being in line with the findings of Krizek (2003), the authors
indicate that increasing population density decreases the number of stops within a tour.
Frank, Bradley, Kavage, Chapman, and Lawton (2008) utilize the 1999
Household Travel Survey of Puget Sound Regional Council (PSRC) Data and analyze the
relationships among trip chaining, transportation mode choice, and land-use patterns
around home locations and workplaces. The analysis includes residential density,
intersection density, land-use mix, and density of retail building floor area as the built
environment determinants of four types of trip chaining: home-based work (HBW) tours,
home-based non-work/other (HBO) tours, and work-based other (WOW) tours. The
authors find that the retail densities at residences and workplaces are negatively
correlated with the complexity of HBW and WOW tours. People live in high intersection
density and high land-use mix areas are more likely to make complex HBO tours.
Applying the data from 1997 Baton Rouge National Personal Transportation
Survey (BRNPTS), Antipova and Wang (2010) examine the effects of land-use types on
the propensity of trip chaining by gender and employment status (workers and non-
workers). The analysis develops five land-use classifications based on different land uses
and housing structures: Agricultural/rural areas are the land devoted to agriculture and
generally located outside of urban areas. Commercial/office areas consist of office,
commercial, and public or semi-public land uses. Single-family detached houses
characterize low-density residential areas, while smaller-lot single-family houses
characterize medium-density residential areas. Condominiums and apartment complexes
dominate high-density residential areas. The model results are generally consistent with
33
previous studies (Krizek, 2003; Noland & Thomas, 2007). The authors find that living in
agricultural/rural areas increases the probability of making complex tours for non-
workers. Male non-workers residing in commercial/office neighborhoods are less likely
to make complex tours. On the other hand, male non-workers living in high residential
areas are more likely to make complex tours.
Ma et al. (2014) include population density, retail employment density, and
service facility accessibility (the average distance to service facilities) at residences and
workplaces in their analysis of the effects of urban form factors on trip chaining behavior
of workers in Beijing, China. Similar to existing literature (Krizek, 2003; Noland &
Thomas, 2007), they find that residents in the areas with higher population density tend to
travel simpler tours. Additionally, the model results suggest that increasing retail
employment density and service facility accessibility at workplaces increases the
tendency of making complex tours.
Several additional studies involve an SEM approach for estimating the
interrelationships between various urban form factors and different aspects of travel
behavior. Van Acker and Witlox (2011) use an SEM approach to examine the influence
of built environment in residential neighborhoods and workplaces on commuting travel
patterns. The built environment variables of both residences and workplaces in the
analysis include employment density, built-up index (percentage of built-up surface),
land-use mix, distance to the nearest public transit stations, distance to the CBD, and auto
accessibility (the number of jobs that can be reached by auto within 15 and 30 minutes).
With respect to the built environment effects at home locations, the employment density
34
and the distance to the nearest public transit are negatively associated with trip chaining.
As for the influence of built environment at the workplaces, the built-up index is
positively correlated with trip chaining. Similar to the effects at the residences, increasing
distances to public transit decreases the tour complexity. Auto accessibility within 15
minutes and 30 minutes at the workplaces have opposite effects: auto accessibility within
15 minutes has positive effects on trip chaining, whereas auto accessibility with 30
minutes has negative influence on trip chaining.
Liu (2012) estimates the effects of urban form factors at tour origins and
destinations on travel behavior and energy consumption. The structural analysis uses
population density, employment density, intersection density, land-use mix, accessibility
index (derived by the Gravity Calibration function, which involves the number of jobs,
travel time, and travel flow matrixes), and distance to the nearest public transit stop to
describe the neighborhood characteristics. Consistent with the findings of Van Acker and
Witlox (2011), the model results show that distance to the nearest public transit stop at
the origins has negative effects on trip chaining for all tours. Furthermore, transit
accessibility at the destinations is negatively associated with trip chaining for all tours.
The author finds that the distance to the nearest public transit at the destinations
positively affects the complexity of auto tours. In disagreement with previous literature
(Antipova & Wang, 2010; Krizek, 2003; Noland & Thomas, 2007), the population
density at tour origins is negatively associated with trip chaining of the non-work tours.
This may be due to the model being based on a sub-sample of the non-work tours.
35
2.2.2 Joint Travel and Urban Form Factors
Although the empirical work looking at the relationships between urban form
factors and joint travel is limited compared with those investigating the effects on trip
chaining, trip generation, and travel distances, some studies have reported the potential
influence of neighborhood characteristics on join travel decisions. Using the Puget Sound
Transportation Panel (PSTP) Data, Chandrasekharan and Goulias (1999) apply the factor
analysis to determine the tract-level accessibility measures and examine the effects on
joint travel. Each measure is associated with population density, household income, travel
time for transit by auto, and travel time for transit by walk respectively. Based on the
model results, most of the accessibility measures do not have significant influence on
joint travel. Only the accessibility measure associated with auto travel time for transit has
negative effects on the probability of joint trip making. The authors indicate that people
living in areas with better accessibility tend to drive alone and avoid carpooling with
others.
Vovsha et al. (2003) investigate the effects of residential location on the joint
travel making and the travel companions based on the Travel Survey Data from the Mid-
Ohio region and the New York region. They find that people living in urban areas tend to
travel jointly. Furthermore, individuals living in the urban areas are less likely to travel
with other adults, whereas people residing in urban areas tend to travel with the
companions of other adults and children.
Some studies focus on parents’ and children’s travel patterns to schools. Based on
the Atlanta Household Activity-travel Survey Data, Vovsha and Petersen (2005) examine
36
the joint travel arrangements that arise when adult household members escort children to
school. The model results show that households located in urban areas are more likely to
chauffeur children to schools than those living in suburban or rural areas. McDonald
(2005) utilizes the 2002 NHTS data and explains the joint travel patterns of mothers and
children. Consistent with the existing studies (Vovsha & Petersen, 2005; Vovsha et al.,
2003), the author finds that children living in urban areas are more likely to travel with
their mothers as compared to those living in rural areas.
2.2.3 Trip/Tour Generation and Urban Form Factors
A considerable amount of literature has examined the effects of urban form
factors on trip/tour generation. The investigation is mainly motivated by the goal of
reducing auto dependency (Boarnet & Crane, 2001; Etminani-Ghasrodashti & Ardeshiri,
2016). M. McNally and Kulkarni (1997) use data from Orange County, California, and
analyze the differences in household trip generation among three types of residential
locations. Applying the K-means clustering algorithm, the authors create three new
neighborhood types: traditional and neo-traditional neighborhoods (TND), planned unit
development neighborhoods (PUD), and neighborhoods with mixed characteristics of
TND and PUD (MIX). These types are classified based on population density,
intersection density, the ratio of commercial area to total area, the ratio of cul-de-sacs to
total intersections, the ratio of four-way intersections to total intersections, and the ratio
of access points to the develop perimeter. TND represents few or no cul-de-sacs, many
access points, high population density, smaller-than-average residential areas, and
greater-than-average values of commercial areas. PUD is characterized by circuitous
37
transportation networks with many cul-de-sacs, a limited number of access points,
isolated land uses, and low population density. MIX integrates the characteristics of the
PUD and the TND. For example, MIX features many cul-de-sacs and single-family
housings like the PUD, while maintaining an overall grid structure on major arterials like
the TND. The population density of MIX is between that of the PUD and the TND. The
authors indicate that households in TND have the lowest number of total trips, followed
by MIX and then PUD areas. ANOVA results further show that the differences in trip
generation among neighborhood types are significant.
Based on the 1986 Travel Behavior Survey Data from San Diego, Boarnet and
Crane (2001) use ordered probit regressions to estimate the land-use effects on the
number of non-work auto trips. The urban form factors included in the analysis consist of
two street network variables (a mixture of connected and cul-de-sac street network and
connected street network), three land-use mix variables (proportion of residential land-
use area, proportion of commercial land-use area, and proportion of vacant area), distance
from trip origin to CBD, and street network density. The authors find that the street
network variables are positively correlated with the number of non-work auto trips. The
model results show that the distance to CBD and the distance to CBD squared are in
opposite directions: the distance to CBD has positive effects on non-work auto trip
generation, while the distance to CBD squared has negative effects on the number of non-
work auto trips. The findings indicate that households living further from the downtown
tend to make more non-work trips up to a limit, as indicated by the negative coefficient of
the squared term.
38
Shay and Khattak (2007) address how auto ownership and travel behavior vary
across different neighborhood types as well as in relation to household characteristics by
using the 2001 Travel Survey Data from Charlotte, North Carolina. They apply factor and
cluster analysis to create five factor indices and seven neighborhood clusters based on 34
built environment features, such as areas of different land uses, densities of employment
and intersection, the number of bus stops, and median value and area of parcels. The five
factor indices (walkability, accessibility, agglomeration, industry, and property value)
and seven clusters (CBD, urban, inner suburban ring, middle suburban ring, outer
suburban ring, rural mixed, and greenfields) are regarded as independent variables to
explain the household auto trip generation in the negative binomial regression analysis.
As for the effects of factor indices, accessibility and property value are positively
correlated with the frequency of auto trips, whereas the industry index has negative
effects on the number of auto trips. The findings suggest that households in
neighborhoods with better accessibility and greater property value make more auto trips.
On the other hand, households in high industrial density areas make fewer auto trips.
With regard to the influence of clusters, the authors find the households residing in CBD,
inner, and outer suburban rings generate more auto trips than those living in greenfields.
Using 2003 and 2004 Travel Survey Data from the San Diego and the San
Francisco Bay Areas, Chatman (2009) applies negative binomial regression analysis to
examine the links between residential built environment features and the number of non-
work trips by auto, transit, and walk or bike. The following built environment features are
measured with circular buffers at residential locations: the number of retail employees
39
within a 0.25-mile radius and 1-mile radius, residents per road mile within 1 mile, the
number of intersections within 0.25 mile, the presence of light-rail and heavy-rail station
within 0.5 miles, distance to CBD, and the presence of a sidewalk on both sides of the
street. These features are modeled in the analysis as a function of non-work trips by auto,
transit, and walking or biking. The author finds that increasing the number of residents
per road mile increases the city’s number of non-work auto trips. The number of retail
employees within a 1-mile buffer positively affects the number of non-work transit trips.
The number of intersections within 0.25 mile is positively associated with the frequency
of non-work trips made by walking or biking. Consistent with Boarnet and Crane (2001),
the model results indicate that people live further from the downtown generate more non-
work trips by transit and non-motorized transportation modes.
Some recent research has emphasized the causal links between the urban form
factors and the outcomes of travel behavior, such as trip generation, travel distances,
VMT, and travel mode choice (e.g., Bagley & Mokhtarian, 2002; Cao, 2007; Cao et al.,
2009; S. Handy, Cao, & Mokhtarian, 2005). These studies find that the associations
between the built environment and travel behavior are partially explained by residential
self-selection, the tendency of individuals to select their residential locations based on
their travel preferences (Cao, 2007; S. Handy, Cao, & Mokhtarian, 2005; Mokhtarian &
Cao, 2008; Namgung, 2014). The effects of urban form factors on travel behavior would
be overstated and overestimated without controlling for residential self-selection (Cao et
al., 2009). Several studies apply a SEM approach to examine multiple aspects of travel
behavior simultaneously while accounting for the residential self-selection bias. For
40
instance, Silva and Goulias (2009) employ a SEM framework with the 2000 Puget Sound
Transportation Panel (PSTP) travel survey data to address the relationships between built
environment and travel behavior. The authors create three land-use factors (central and
denser areas, bus supply, and a mix of land uses) characterizing both the residence and
employment locations based on the urban form variables (densities of residential and
employment population, intersection density, bus availability, and land-use entropy). The
model results show that people residing and working in more central and denser areas
make more transit and non-motorized trips. The analysis suggests that the mix of land-use
is negatively associated with the auto trip generation, while it positively affects the
frequency of transit and non-motorized trips.
Etminani-Ghasrodashti and Ardeshiri (2016) use 2014 Travel Survey Data from
Shiraz, Iran, to examine the relationships between residential urban form, the number of
home-based work (HBW) trips, and home-based non-work (HBN) trips by auto, transit,
and walking. The authors include land-use entropy, distance to CBD, and densities of
residents, employees, streets, intersections as urban form factors in their structural
analysis. As for the HBW trips, the model results indicate that residential employment
densities have positive effects on the number of non-motorized trips. Intersection density
is negatively correlated with auto trips but positively correlated with non-motorized trips.
With regard to the HBN trips, the authors find that employment density has positive
influence on the number of transit and non-motorized trips. The densities of streets and
intersections have positive effects on auto trips but have negative effects on non-
motorized trips. In agreement with the existing findings (Boarnet & Crane, 2001;
41
Chatman, 2009; Silva & Goulias, 2009), increasing distance to CBD increases auto trip
generation but decreases the number of transit trips.
2.2.4 Travel Distances and Urban Form Factors
As one of the critical proxies of the resulting travel demand, the links between
travel distances and urban form have been examined in a considerable amount of existing
literature. Stead (1999) use data from four National Travel Surveys (1979-81, 1985-86,
1989-91, and 1991-93) in Great Britain and examines the effects of built environment on
individual travel distances. The built environment in the regression analysis consists of
population density, bus frequency, proximity to local facilities (e.g., post office,
drugstore, and grocery store), distance to primary business street stores, and proximity to
a public transit station. The author finds two factors have consistent and significant
effects on individual travel distances across the four datasets: population density and bus
supply. People who live in low-density areas (less than 10 people per hectare) travel
longer distances than those live in other areas, while residents of neighborhoods with
greater bus supplies (more than one bus per hour) travel shorter distances.
Bento et al. (2005) estimate the effects of public transit supply and urban form
factors on annual household VMT by using the 1990 Nationwide Personal Transportation
Survey (NPTS) Data. Several urban form features are examined as a function of the
household VMT: whether the city is circular or not, road density, population density, bus
transit supply (the ratio of miles traveled by bus and urban areas), and rail transit supply
(the ratio of miles traveled by rail and urban areas). The authors find that that road
density positively affects the VMT for the households with one auto. Increasing rail
42
supply decreases the VMT for households with two cars. Households with one auto and
those in circular cities tend to have fewer VMT.
Based on the travel data of four regions (Seattle, WA; Richmond-Petersburg and
Norfolk-Virginia Beach, VA; Baltimore, MD; and Washington, DC.), Zhang et al. (2012)
develop a Bayesian multilevel model to examine relationships between urban form and
VMT while controlling for socio-demographics. Five urban form factors are included in
the model: residential density, employment density, land-use, average block size (as a
measurement of transit/walking friendliness), and distance to city center. In agreement
with Stead (1999), the residential density has negative effects on VMT for all four
regions. The employment density is negatively associated with VMT only in the Seattle
and Baltimore areas. Land-use entropy has a negative impact on VMT for all regions
except the Virginia area. People living in more compact and mixed-development
neighborhoods tend to drive less. A smaller block size is considered to have better street
connectivity and walkability in the analysis. The model results indicate that the
increasing block size increases the VMT for all of the study regions. The distance to CBD
is positively associated with VMT for all neighborhoods except for the Virginia area,
which suggests that people living further away from the city center tend to drive more.
Hong, Shen, and Zhang (2014) use the 2006 Household Activity Survey Data
from the Seattle Metropolitan Area and examine the effects of urban form factors on
VMT of work tours and non-work tours. The authors employ Bayesian hierarchical
models with the following urban form factors: residential density, non-residential density,
land-use entropy, intersection density, distance to the nearest bus station, and distance to
43
CBD. Consistent with the existing studies (Stead, 1999; Zhang et al., 2012), the model
results show residential density is negatively correlated with VMT for non-work tours
and the non-residential density is negatively related to VMT for work tours. Increasing
distance to the nearest bus station increases VMT of non-work tours. People living in
more mixed and higher intersection density neighborhoods drive less for non-work
activities. Similar to the findings of Zhang et al. (2012), the analysis indicates that the
distance to CBD has positive effects on VMT of work tours.
Akar et al. (2016) investigate how mean (per trip) and total (per day) travel
distances are affected by residential neighborhood types, which are identified based on
the urban form factors through the K-means clustering approach. The authors create
seven neighborhood types (high-density and mixed-use central neighborhoods, medium-
density and mixed land-use central neighborhoods, CBD, high employment urban
neighborhoods, new dense residential neighborhoods, medium-density suburban
neighborhoods, and low-density single-family neighborhoods) by using five built
environment measures (population density, employment density, intersection density,
median age of buildings, and percentage of single detached house). They find that, as
compared to people who live in high-density and mixed-use central neighborhoods, those
living in high employment urban, new dense residential, medium-density suburban, and
low-density single-family neighborhoods make more trip distances on average. With
respect to the total travel distances, the analysis shows that individuals who reside in new
dense residential and low-density single-family areas travel further in total than those live
in high-density and mixed-use central areas.
44
Jiang, Gu, Chen, He, and Mao (2016) examine the effects of land-use patterns and
street characteristics on household car ownership and use based on a travel survey from
the city of Jinan in China. Principle Component Analysis (PCA) is applied to create four
land-use components (land density, street diversity, street design, and street density) from
fifteen urban form features. The land density component measures the population density,
housing density, facilities point of interests, and street network density. The street
diversity component captures the right of way for transit users, cyclists, and pedestrians.
The street design component represents the street walkability in terms of street-wall
continuity, building coverage, and street facade. The street density component places
high loadings on densities of ground-floor retails and crossing facilities. The model
results show that the land density and street density components have negative effects on
household VMT. These outcomes are consistent with previous literature, which find that
areas with greater population, denser street networks, and more crossing facilities exhibit
less auto use (Hong et al., 2014; Stead, 1999; Zhang et al., 2012). In addition, the street
design component is negatively associated with household VMT, implying that walkable
streets with continuous street facades and small setbacks reduce the household VMT.
Silva, Golob, and Goulias (2006) argue that employing the SEM technique allows
researchers to estimate the effects of urban form characteristics on travel behavior while
controlling for the bias of self-selection at residences and workplaces. Based on the travel
survey data from the Lisbon Metropolitan Area in Portugal, the authors apply structural
analysis to investigate the influence of urban form at residences and workplaces on
multidimensional travel behavior, such as travel time, travel distances, and trip
45
generation. They use factor analysis to reduce twenty-four urban form measures to eight
land-use factors (residence/workplaces in traditional urban areas, workplaces in compact
and central urban areas, road supply, freeway supply in the residence/workplaces, and
residence/workplaces in specialized areas). The residing/workplaces in traditional urban
areas factors have high loadings in population density, land-use mix, auto accessibility,
and public transit supply. The factor of workplaces in compact and central urban areas
shares similar characteristics with the factor of workplaces in traditional urban areas and
has additional high loadings on the distance to CBD. The road supply factor has high
loadings on the distances of roads per person at both residences and workplaces. The
population density around freeway nodes at residences and workplaces respectively
characterize the freeway supply in the residence/work areas factors. The
residence/workplaces in specialized areas factors are highly correlated with the land-use
entropy at residences and workplaces respectively. Supporting the findings of Zhang et
al. (2012) and Hong et al. (2014), the model results reveal that people living and working
in more compact, mixed, and traditional neighborhoods travel shorter auto distances.
Additionally, the authors find that freeway supply in the residence/workplaces are
positively associated with the auto travel distances.
Dillon et al. (2015) propose a structural analysis and consider residential urban
form a latent variable with high loadings on population density, housing density,
employment density, the percentage of renter occupied housing, distance to the city
center, land-use entropy, and public transit supply. The model results show that this latent
variable has negative effects on VMT of total trips, VMT for work trips, and VMT for
46
non-work trips. This finding can be explained by the fact that households locate in
compact neighborhoods with more rental residences and better transit supply tend to
drive less.
2.3 Relationships among Different Aspects of Travel Behavior
As compared to the studies looking at the links among socio-demographics, urban
form factors, and travel behavior, empirical work investigating the relationships among
different aspects of travel behavior has been more limited. Utilizing the 1994 Activity
and Travel Survey Data from Portland, Oregon, Golob (2000) propose a SEM that jointly
forecasts three sets of endogenous variables (activity participation, travel time, trip
chaining, and trip generation) as separate functions of household characteristics and
accessibility indices. The model results reveal that there are significant relationships
among the four household trip generation variables (the number of simple work tours, the
number of complex work and non-work tours, the number of simple non-work tours, and
the number of complex non-work tours). The author finds that the number of complex
non-work tours is negatively related to the number of simple non-work tours. The number
of simple work tours is negatively correlated with the frequencies of complex work and
non-work tours, simple non-work tours, and complex non-work tours. Interestingly, the
number of complex work and non-work tours has negative effects on the number of
simple non-work tours but has positive effects on the frequency of complex non-work
tours.
47
Based on the 2005 Travel Data from Chennai city, India, Gopisetty and
Srinivasan (2013) explain household trip generation in terms of socio-demographics,
built environment, and travel attributes. Trip chaining is included as one of the
explanatory variables in the regression analysis. The model results show that the trip
chaining has positive effects on the number of household trips. The authors explain that
the households with high levels of activity and travel requirements may chain their trips
to save time.
Duncan (2015) uses Bay Area Travel Survey Data of 2000 and estimates the
potential VMT reduction by trip chaining patterns. The author addresses how much trip
chaining can reduce VMT per activity stop. The model outputs indicate that trip chaining
can lead to significant VMT reduction. Duncan (2015) finds that weekday tours generally
generate fewer VMT than weekend VMT. This can be explained by the fact that people
are more flexible during weekends for traveling to distant destinations. Based on the
analysis, the author points out that adding one non-work activity to a simple commute
tour is expected to reduce VMT per destination by 20 to 30%. Furthermore, adding two
non-work activities to a work tour results in an expected reduction of 25 to 50% of VMT
per destination.
2.4 Discussion of the Existing Literature
Many previous studies have confirmed that socio-demographics are influential
determinants of trip chaining, joint travel, trip/tour generation, and travel distances, as
summarized above. As for personal characteristics, females, non-workers, and individuals
48
with higher education levels tend to generate simple home-based tours and more tours (S.
G. Bricka, 2008; Kitamura & Susilo, 2005; Kuppam & Pendyala, 2001; Lu & Pas, 1999;
Ma et al., 2014; Silva et al., 2012; Wallace et al., 2000; Yang et al., 2010). Individuals
between the ages of 35 and 65, students, and full-time workers are less likely to travel
jointly with other household members (Chandrasekharan & Goulias, 1999; Lin & Wang,
2014). Men, licensed drivers, full-time workers, and people with higher education levels
travel further than their counterparts (Akar et al., 2016; Feng et al., 2017; Liu, 2012;
Stead, 1999). With regard to household attributes, people from larger households tend to
make simpler tours (Kitamura & Susilo, 2005; Ma et al., 2014; Van Acker & Witlox,
2011; Wallace et al., 2000). Increasing household size and the presence of children and
elderly in the household increase the probability of joint travel making (Ho & Mulley,
2013a; Lin & Wang, 2014). The number of household vehicles, the number of workers in
the household, and the household income are all positively associated with trip generation
and travel distances (Akar et al., 2016; Bento et al., 2005; Dillon et al., 2015; Jang, 2005;
Liu, 2012; Silva et al., 2012; Stead, 1999; Wallace et al., 1999).
Various urban form factors also contribute significantly to activity-travel patterns
and travel demand. People living in neighborhoods characterized by high residential and
employment densities make simple tours more often (Frank et al., 2008; Noland &
Thomas, 2007; Van Acker & Witlox, 2010). Some studies point out that people residing
in urban areas increase likelihood of travel with household members (McDonald, 2005;
Vovsha et al., 2003). Individuals living in compact and mixed-development
neighborhoods with high intersection density and proximity to the city center make fewer
49
auto trips and have fewer VMT (Chatman, 2009; Etminani-Ghasrodashti & Ardeshiri,
2016; Hong et al., 2014; Shay & Khattak, 2007; Silva et al., 2012; Stead, 1999; Zhang et
al., 2012). While many existing studies have examined the relationships between travel
behavior and urban form features at residences and workplaces, the effects of urban form
around other out-of-home activity locations remain unclear. Similarly, the relationship
between trip chaining and joint travel and their contribution to the resulting travel
demand warrants further research.
With the above considerations in mind, the present study addresses the following
concerns where the research gaps remain. First, this study considers trip chaining and
joint travel as two mediating variables and analyzes how they shape the resulting tour
generation and travel distances. Second, this study aims to examine the effects of urban
form factors not only at residential neighborhoods but also at out-of-home activity
neighborhoods. Third, this study looks at the intermediating effects of exogenous
variables (socio-demographics and urban form factors) on travel demand outcomes (tour
generation and travel distances) channeled via activity-travel patterns (trip chaining and
joint travel) through several structural analysis frameworks.
50
Chapter 3: Research Design and Data
This chapter is composed of three sections: Research Design, Data, and Variables
of Interest. The first section presents the analysis framework for the models at each level
(tour level, individual level, and household level). The second section introduces the
study area, the data used in the analysis, and the different aspects of travel behavior to be
investigated in this study. The variables of interest and the descriptive statistics are
included in the last section of this chapter.
3.1 Research Design
As discussed in previous chapters, the activity-based approach has been used in a
great amount of travel behavior research over the past few decades. One of the key
concepts of this well-established approach is that trip making is a means to satisfy the
need to pursue the activities distributed in different places (Bhat & Koppelman, 2003;
Goulias et al., 1990; Kitamura, 1988). Therefore, this study postulates that the
individuals’ decisions of whether or not travel jointly with household companions depend
on the activities that they participate in. Trip chaining is assumed to affect joint travel
based on this principle. Both activity-travel patterns are hypothesized to influence two
travel demand outcomes: tour generation and travel distances. According to the
discussion in previous chapters, this analysis assumes that socio-demographics and urban
form factors affect trip chaining, joint travel, tour generation, and travel distances.
51
Based on the aforementioned assumptions, this study applies analytical
frameworks with limited adjustments to examine the interconnections among travel
behavior, socio-demographics, and urban form factors at three levels: tour level,
individual level, and household level. The tour-level model uses home-based tour as the
basic unit of analysis. As depicted in Figure 2, the tour-level model considers trip
chaining and joint travel as two mediating variables of the resulting travel distances at the
tour level. Several studies indicate that using a tour as the unit of analysis instead of a trip
in empirical work provides a better description of travel decisions because a tour
consolidates individual trip segments, both departing and return trips, and all the activity
episodes made along the way (Frank et al., 2008; Liu, 2012). Therefore, the tour-based
analysis may more closely match the mechanism that travel decisions are made as
compared to the trip-based analysis. In addition, this model investigates the effects of
socio-demographics and urban form factors at tour origins and destinations on activity-
travel patterns and travel demand simultaneously.
Figure 2. Tour-level Model Framework
52
The individual-level model framework in Figure 3 shows how individual trip
chaining and joint travel shape the resulting individual tour generation and VMT. The
socio-demographics including personal characteristics and household attributes are used
as explanatory variables in the model. The urban form factors in this model only capture
the built environment features around residential neighborhoods.
Figure 3. Individual-level Model Framework
The third model is at the household level, which employs a household as the basic
unit of the analysis. The model framework is presented in Figure 4. The household-level
model applies the similar model framework at the individual-level model and investigates
the links among trip chaining, joint travel, tour generation, and VMT at the household
level. The socio-demographics in the household-level model are only composed of
53
household attributes. Similar to the individual-level model, the urban form factors in the
household-level model include residential urban form factors.
Figure 4. Household-level Model Framework
3.2 Data
This section presents the sources of data used in this study, demonstrates the study
area, and introduces the definitions of activity-travel patterns and travel demand to be
examined in the analysis. The data used in this dissertation include two main datasets:
GIS-based travel survey data and the corresponding built environment data, which are
derived from different sources.
3.2.1 Datasets of Activity-travel Patterns and Travel Demand
The primary travel survey data are the 2012 Northeast Ohio Regional Travel
Survey (NORTS) data provided by the Ohio Department of Transportation (ODOT). This
survey was conducted between February 2012 and March 2103 in five counties
(Cuyahoga County, Geauga County, Lake County, Lorain County, and Medina County)
54
around the Cleveland Metropolitan Area (Wilhelm, Wolf, Kang, & Taylor, 2014). The
travel survey data include trip segments and activity location information, along with
individuals’ personal and household characteristics. The final travel survey data sample
contains 70,333 trip segments made by 10,066 individuals from 4,540 households. Figure
5 shows the map of the study area.
Figure 5. Study Area – the Cleveland Metropolitan Area
55
3.2.2 Travel Demand: Tour Generation and Travel Distances
This study examines two travel demand outcomes: tour generation and travel
distances. The resulting travel demand is the distances of a home-based tour in the tour-
level model. The individual-level model and household-level model focus on the use of
vehicles by Cleveland households. Tour generation at the individual and household levels
are measured by the number of auto tours, in which all trip segments are traveled by car.
The individual and household travel distances are captured by the distances traveled by
the vehicles (VMT). The resulting travel demand at the individual level are the number of
auto tours and VMT traveled by an individual, respectively. Similarly, the resulting travel
demand at the household level is represented by the number of auto tours and VMT
traveled by a household, respectively.
3.2.3 Activity-travel Patterns: Trip Chaining
The trip segments in this travel survey dataset are combined into home-based
tours, which are defined as the sequence of trips starting and ending at an individual’s
residence (Primerano et al., 2008). Most previous studies use this definition and regard
trip chaining as a measurement of tour complexity. A home-based tour becomes more
complex as the number of stops within the tour increases. In this study, trip chaining at
the tour level is defined as the number of stops within a home-based tour. Equation 1
presents the formula of trip chaining in the tour-level model.
𝑇𝐶𝑇𝐿 = 𝑆𝑖 (1)
Where 𝑇𝐶𝑇𝐿 is the trip chaining each tour at the tour level, and
56
𝑆𝑖 is the number of out-of-home activities of home-base tour i
Trip chaining at the individual-level model is calculated as the number of stops
per auto tour traveled by an individual. Equation 2 shows the formula of trip chaining at
the individual level.
𝑇𝐶𝐼𝐿 =
∑ 𝑆𝑖𝑚𝑖
𝑇𝑚 (2)
Where 𝑇𝐶𝐼𝐿 is the trip chaining of each individual at the individual level,
∑ 𝑆𝑖𝑚𝑖 is the total number of stops of all auto tours traveled by individual m, and
𝑇𝑚 is the total number of auto tours traveled by individual m.
In the household-level model, trip chaining measures the number of stop per auto
tour traveled by a household. The following equation demonstrates the formula of trip
chaining at the household level.
𝑇𝐶𝐻𝐿 =
∑ 𝑆𝑖𝑛𝑖
𝑇𝑛 (3)
Where 𝑇𝐶𝐻𝐿 is the trip chaining of each household at the household level,
∑ 𝑆𝑖𝑛𝑖 is the total number of stops of all auto tours traveled by household n travel,
and
𝑇𝑛 is the total number of auto tours traveled by household n travel.
3.2.4 Activity-travel Patterns: Joint Travel
Some existing studies have regarded joint travel making as an alternative for
travelers. These studies have considered a trip to be joint if it is traveled jointly with one
or more household members (Chandrasekharan & Goulias, 1999; Ho & Mulley, 2013b;
Vovsha et al., 2003). This research considers joint travel as an indicator to better capture
57
the level of intra-household interactions. Adopted from the previous research, this study
considers a trip segment of a home-based tour as joint if it is traveled jointly with one or
more household members. The joint travel in the analysis is measured as a ratio of the
number of joint trip segments and the number of total trip segments for each home-based
tour. For two different home-based tours with the same tour complexity (both tours have
the same number of stops), the one with greater a joint travel ratio has higher intra-
household interaction and household coordination.
Joint travel is represented by the ratio of the number of joint trip segments to the
number of total trip segments for each tour in the tour-level model. Equation 4 shows the
formula of joint travel at the home-based tour level.
𝐽𝑇𝑇𝐿 =
𝐽𝑇𝑆𝑖
𝑇𝑇𝑆𝑖 (4)
Where 𝐽𝑇𝑇𝐿 is the joint travel of each home-based tour i at the home-based tour level,
𝐽𝑇𝑆𝑖 is the number of joint trip segments of each home-based tour i, and
𝑇𝑇𝑆𝑖 is the number of trip segments of each home-based tour i.
Joint travel at the individual level is calculated by the ratio of joint travel per auto
tour traveled by an individual and is presented in Equation 5
𝐽𝑇𝐼𝐿 =∑
𝐽𝑇𝑆𝑖𝑚
𝑇𝑇𝑆𝑖𝑚𝑖
𝑇𝑖𝑚
(5)
Where 𝐽𝑇𝐼𝐿 is the joint travel of each individual m at the individual level,
𝐽𝑇𝑆𝑖𝑚 is the number of joint trip segments in each auto tour i traveled by
individual m,
58
𝑇𝑇𝑆𝑖𝑚 is the number of total trip segments in each auto tour i traveled by
individual m and,
𝑇𝑖𝑛 is the total number of auto tours traveled by individual m.
Similar to the joint travel calculation at the individual level, joint travel at the
household level is calculated by the ratio of joint travel per auto tour traveled by a
household. Equation 6 presents the joint travel included in the household-level model.
𝐽𝑇𝐻𝐿 =∑
𝐽𝑇𝑆𝑖𝑛
𝑇𝑇𝑆𝑖𝑛𝑖
𝑇𝑖𝑛
(6)
Where 𝐽𝑇𝐻𝐿 is the joint travel of each household m at the household level,
𝐽𝑇𝑆𝑖𝑛 is the number of joint trip segments in each auto tour i traveled by
household n,
𝑇𝑇𝑆𝑖𝑛 is the number of total trip segments in each auto tour i traveled by
household n and,
𝑇𝑖𝑛 is the total number of auto tours traveled by household n.
3.2.5 Classification of Tours and the Primary Activity for a Home-based Tour
This study aims to examine the effects of urban form factors at tour origins and
destinations on trip chaining, joint travel, and travel distances while controlling for socio-
demographics in the tour-level model. The tour origins are set to be individuals’
residential locations because all home-based tours in the sample data start from home
locations. With regard to tour destinations, this study considers the primary activity
locations. Some existing research classifies the tours by activity purposes in order to
decide which activity is the primary activity of a home-based tour. For example, some
studies only focus on work tours and non-work tours (Frank et al., 2008; Ma et al., 2014;
Van Acker & Witlox, 2011). Some classify the tours into three types. For instance, Golob
59
(2000) and Liu (2012) categorize the tours into three types: work, non-work, and mixed
work tours. Krizek (2003) and Ho and Mulley (2015) also classify the tours into three
groups: work, maintenance, and discretionary tours.
Adapted from Frank et al. (2008) and Liu (2012), the classification scheme used
in this study categorizes the tours into three types based on the duration and purpose of
each activity stop within a tour. The mandatory tours are composed of trips covering only
two types of mandatory activities: work and school. The mandatory-mixed tours consist
of the trips for mandatory and non-mandatory activities. The non-mandatory tours only
include trips for non-mandatory activities. Among the three tour categories, the primary
activity of a mandatory tour and a non-mandatory tour is the activity with the longest
duration within the tour. The primary activity for a mandatory-mixed tour is the
mandatory activity with longest activity duration. Hence, the tour destination of a
mandatory tour or a non-mandatory tour is the location where the mandatory or non-
mandatory activity with the longest duration take place. The tour destination of a
mandatory-mixed tour is the location where the mandatory activity with the longest
duration take place.
3.2.6 Datasets for Urban Form Factors
The location information of activities, trip origins, and trip destinations in the
primary household travel survey data are released at the transportation analysis zone
(TAZ) level due to privacy concerns. The corresponding urban form data are derived
from the following sources: Census Transportation Planning Products (CTPP), Google
General Transit Feed Specification (GTFS), and Topologically Integrated Geographic
60
Encoding and Referencing (TIGER). The urban form factors that characterize the built
environment features at tour origins and tour destinations are discussed in detail in the
next section.
3.3 Variables of Interest and Descriptive Statistics Analysis
This section presents detailed information on endogenous and exogenous
variables used in this study, followed by the descriptive statistics of these variables and
the Kruskal-Wallis Test of trip chaining, joint travel, tour generation, and travel distances
on selected socio-demographics.
3.3.1 Endogenous and Exogenous Variables
Table 1 presents the endogenous and exogenous variables used in the structural
equation models. The endogenous variables of interest are the two activity-travel patterns
(trip chaining and joint travel) and the two travel demand outcomes (tour generation and
travel distances). Trip chaining, joint travel, and travel distances for each home-based
tour are the three endogenous variables included in the tour-level model. The individual-
level model contains four endogenous variables: individual trip chaining, individual joint
travel, the number of auto tours traveled by an individual, and individual VMT. Similar
to the individual-level model, household-level model has four endogenous variables:
household trip chaining, household joint travel, the number of auto tours traveled by a
household, and household VMT. The individual VMT and household VMT variables
included in the analysis are log transformed.
61
The exogenous variables used in this research consist of socio-demographics and
urban form factors. Socio-demographics include various personal characteristics and
household attributes. The analysis uses individual’s gender, age, driver’s license
ownership, employment status, and transit pass ownership to describe the personal
characteristics. The household size, number of household vehicles, household income,
percentage of household licensed drivers, and presence and age stages of the youngest
child in the household are used to account for the household attributes in this study.
The other set of exogenous variables is the urban form factors. The functional
uses of land and its physical characteristics are generally termed urban form factors.
Previous studies have classified various urban form factors to several categories. For
instance, Cervero and Kockelman (1997) suggest that built environment influences travel
behavior along three principal dimensions (3Ds): density, diversity, and design.
“Density” reflects how intensively land is used for residence, employment, and other
purposes. “Design” represents the physical transportation infrastructure including street
connectivity and accessibility of public transit. “Diversity” measures the degree of land-
use mixtures (Cao, 2007; Cervero, 2002; Cervero & Kockelman, 1997; Namgung, 2014).
S. L. Handy et al. (2002) argue that the built environment consists of three major
components: urban design, land use, and transportation system. They illustrate that the
“urban design” component refers to “the design of the city and the physical elements
within it, including both their arrangement and their appearance, and is concerned with
the function and appeal of public spaces”. The “land use” component characterizes “the
distribution of activities across space, including the location and density of different
62
activities, where activities are grouped into relatively coarse categories”. The authors
describe the “transportation system” component as “the physical infrastructure of roads,
sidewalks, bike paths, railroad tracks, bridges, and so on, as well as the level of service
provided as determined by traffic levels, bus frequencies, and the like” (S. L. Handy et
al., 2002).
Based on the data availability and previous literature, this study selects six urban
form factors to outline the built environment features and neighborhood characteristics:
residential density, retail employment density, non-retail employment density,
intersection density, bus stop density, and job-population index. The index is first used by
Ewing et al. (2011). It is adopted by other studies in recent years (Akar et al., 2016; Chen
& Akar, 2016). The equation of job-population index included in this study is presented
as follows.
Job − population index𝑘 = 1 −
(|𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 − 0.7×𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛|)
(𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 + 0.7×𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛) (7)
Where Job − population index𝑘 means the job-population index for TAZ k. The
parameter 0.7 represents a balance of employment and population in the study area.
This index ranging from 0 to 1 captures the balance between employment and
residential population. Only jobs or residents exist in the TAZ when the index is close to
0. On the other hand, the ratio of jobs and residents is optimal from travel behavior
perspective when the index value ranges to 1. The parameter is a region-specific factor
and varies based on different empirical facts of study areas in existing research. In order
to maximize the explanatory power of the variable, Ewing et al. (2011) use 0.2 and Akar
63
et al. (2016) use 0.5 in their studies. This study tests the parameter ranging from 0.1 to
0.9. The value of 0.7 is then chosen because it generates significant coefficients with best
model fit across the three models.
Table 2 summarizes the endogenous and exogenous variables used in each model
(tour-level model, individual-level model, and household-level model). In the tour-level
model, the endogenous variables are trip chaining, joint travel, and travel distances. The
exogenous variables used in the tour-level model are personal characteristics, household
attributes, and urban form factors at tour origins and destinations. The endogenous
variables at the individual level are trip chaining, joint travel, number of auto tours, and
individual VMT. The exogenous variables in the individual-level model consist of
personal characteristics, household attributes, and urban form factors at tour origins. The
endogenous variables in the household-level model are trip chaining, joint travel, number
of auto tours, and household VMT. At the household level, the exogenous variables are
household attributes and urban form factors at tour origins.
64
Table 1. Variables of Interest Variables Description
Travel Demand and Activity-travel Patterns
Tour travel distances Distances traveled at the tour level
Individual VMT VMT at the individual level
Household VMT VMT at the household level
Individual auto tour generation Number of auto tours at the individual level
Household auto tour generation Number of auto tours at the household level
Tour joint travel Joint travel ratio of tour at the tour level
Individual joint travel Joint travel ratio per tour at the individual level
Household joint travel Joint travel ratio per tour at the household level
Tour trip chaining The number of stops of tour at the tour level
Individual trip chaining The number of stops per tour at the individual level
Household trip chaining The number of stops per tour at the household level
Personal characteristics
Female 1, if an individual is female; 0, otherwise
Age Age (in years) of an individual
Driver’s license status 1, if an individual has a driver’s license; 0, otherwise
Employment status 1, if an individual is employed; 0, otherwise
Transit pass ownership 1, if an individual has transit pass; 0, otherwise
Household (HH) attributes
HH size The number of household members
HH vehicles The number of household vehicles
HH income Household income (in thousand US dollar)
Percentage of licensed drivers in HH The percentage of licensed drivers in household
No children presence 1, if no child in the household; 0, otherwise
Youngest child’s age is up to 5 1, if the age of youngest child is up to 5; 0, otherwise
Youngest child’s age is 6 to 10 1, if the age of youngest child is 6 to 10; 0, otherwise
Youngest child’s age is 11 to 15 1, if the age of youngest child is 11 to 15; 0, otherwise
Youngest child’s age is 16 to 18 1, if the age of youngest child is 16 to 18; 0, otherwise
Urban form factors
Residential density Residential population per square mile
Retail density Retail employment per square mile
Non-retail density Non-retail employment per square mile
Intersection density The number of intersection nodes per square mile
Bus stop density The number of bus stops per square mile
Job-population index Index that measures the employment and population
balance Source: Authors’ illustration
65
Table 2. Endogenous and Exogenous Variables in Each Model
Models Endogenous variables Exogenous variables
Travel Behavior Socio-demographics Urban form factors
Activity-travel patterns
& travel demand
Personal characteristics &
Household (HH) attributes At tour origins At tour destinations
Tour-level
Model
Trip chaining Female HH income Residential density Residential density
Joint travel Age Percentage of licensed drivers Retail density Retail density
Travel distances Driver’s license status No children presence Non-retail density Non-retail density Employment status Youngest child’s age is up to 5 Intersection density Intersection density Transit pass ownership Youngest child’s age is 6 to 10 Bus stop density Bus stop density HH size Youngest child’s age is 11 to 15 Job-population index Job-population index HH vehicles Youngest child’s age is 16 to 18
Individual-level
Model
Trip chaining Female HH income Residential density
Joint travel Age Percentage of licensed drivers Retail density
Number of auto tours Driver’s license status No children presence Non-retail density
Individual VMT Employment status Youngest child’s age is up to 5 Intersection density
Transit pass ownership Youngest child’s age is 6 to 10 Bus stop density
HH size Youngest child’s age is 11 to 15 Job-population index
HH vehicles Youngest child’s age is 16 to 18
Household-level
Model
Trip chaining HH size Residential density
Joint travel HH vehicles Retail density
Number of auto tours HH income Non-retail density
Household VMT Percentage of licensed drivers Intersection density
No children presence Bus stop density
Youngest child’s age is up to 5 Job-population index
Youngest child’s age is 6 to 10
Youngest child’s age is 11 to 15
Youngest child’s age is 16 to 18
Source: Authors’ illustration
65
66
3.3.2 Descriptive Statistics of the Sample Data
For the interest of this study, households with one person are dropped from the
sample data as these households would not generate any joint trips with other household
members. After data screening and cleaning, the data used for the tour-level analysis
consist of 7,390 tours, which are made by 4,649 individuals from 2,440 households.
Among these 7,390 tours, 6,598 tours are auto tours in which all trip segments are driven.
The sample data used for analysis at the person/household level are these 6,598 auto
tours, which are made by 4,184 individuals from 2,297 households.
Table 3 summarizes the descriptive statistics of the sample data used in the tour-
level model. Among 7,390 tours, the average travel distances are 17 miles; the mean of
stops for each tour is 2.43, and around 25% of trip segments within each tour are made
jointly. Figures 6 through 8 show the histograms of travel distances, joint travel, and trip
chaining at the tour level. Figure 6 shows that the distances of most tours traveled are
within 20 miles. Figure 7 indicates that most tours are solo tours (joint travel ratio is 0)
and fully joint tours (joint travel ratio is 1). The frequency distribution of tour complexity
demonstrated in Figure 8 suggests that most tours consist of less than four stops.
Based on Table 3, women are slightly more represented than men. Around 90% of
participants have driver’s licenses, about 7.5% of individuals have transit passes, the
average age of participants is approximately 50, and nearly 60% of survey respondents
are employed. The average household size is approximately three. Each household has
two vehicles on average, and the average annual household income is around $70,000.
The percentage of licensed drivers is roughly 77% per household. About 65% of
67
households do not have any children. Around 10% of households have a child five years
of age or younger, 7% of households’ youngest child is between the ages of six and ten,
10% of the households’ youngest child is between the ages of 11 and 15, and 6% of the
households’ youngest child is between the ages of 16 and 18.
Table 3. Descriptive Statistics of Sample Data Used for Tour-level Model Variables Sample Percentage Mean Std. Dev.
Travel behavior at tour level
Travel distances (miles) 17.107 17.485
Joint travel 0.249 0.395
Trip chaining 2.427 1.904
Personal characteristics
Female 51.97%
Age 49.947 17.458
Has driver’s license 88.363%
Has transit pass 7.40%
Employed 58.01%
Household (HH) attributes
Household size 2.873 1.189
The number of household vehicles 2.016 1.052
Household income (in thousand) 68.131 48.276
The percentage of licensed drivers 76.692 29.1768
No children presence 65.25%
Youngest child’s age up to 5 9.75%
Youngest child’s age 6 to 10 7.58%
Youngest child’s age 11 to 15 10.74%
Youngest child’s age 16 to 18 6.68% Source: Authors’ calculation
68
Figure 6. Histogram of Travel Distances at the Tour Level
Figure 7. Histogram of Joint Travel at the Tour Level
69
Figure 8. Histogram of Trip Chaining at the Tour Level
Table 4 presents the descriptive statistics of the person- and household-level
travel behavior. As for the travel behavior at the individual level, the average VMT is
about 28 and the number of auto tours is approximately 1.58 on average. The mean ratio
of joint travel is about 25% and the average number of stops per auto tour is 2.41 at the
individual level. The Figures 9 through 12 are histograms of VMT, the number of auto
tours, joint travel, and trip chaining at the individual level, respectively. Figure 9 shows
that most individual VMT are between 5 to 30 miles. According to Figure 10, most
travelers make less than three auto tours. The frequency distribution of the average joint
travel ratio per tour indicates that most individuals tend to travel solo for the entire tour.
Figure 12 depicts that a large proportion of auto tours are fewer than four stops.
70
Table 4. Descriptive Statistics of Travel Behavior in Sample Data Used for Individual-
level and Household-level Models Variables Mean Std. Dev.
Travel behavior at the individual level (The number of individuals = 4,189)
VMT 27.911 23.100
Number of auto tours 1.576 0.801
Joint travel 0.241 0.356
Trip chaining 2.453 1.689
Travel behavior at the household level (The number of households = 2,298)
VMT 47.499 38.010
Number of auto tours 2.812 1.674
Joint travel 0.190 0.288
Trip chaining 2.467 1.470 Source: Authors’ calculation
Figure 9. Histogram of VMT at the Individual Level
71
Figure 10. Histogram of the Number of Auto Tours at the Individual Level
Figure 11. Histogram of Joint Travel at the Individual Level
72
Figure 12. Histogram of Trip Chaining at the Individual Level
As for household travel behavior,
Table 4 reveals that the mean VMT is 47.50, and each household generates 2.81
auto tours on average. The average of joint travel ratio is around 19% per auto tour. There
are approximately 2.5 stops for each auto tour. The histograms of VMT, the number of
auto tours, joint travel, and trip chaining at the household level are presented in Figures
13 through 16. Figure 13 shows most household VMT are between the ranges of 10 to 40
miles. According to Figure 14, most households make one to four auto trip chains. Figure
15 shows that most households tend to make solo tours (joint travel ratio is 0). The figure
also indicates that the frequencies of fully joint tours (joint travel ratio is 1), half joint
73
tours (joint travel ratio is 0.5), and partially joint tours (joint travel ratio is 0.35) are
similar. Figure 16 indicates that most households make complex tours with two stops.
Figure 13. Histogram of VMT at the Household Level
Figure 14. Histogram of the Number of Auto Tours at the Household Level
74
Figure 15. Histogram of Joint Travel at the Household Level
Figure 16. Histogram of Trip Chaining at the Household Level
75
Table 5 provides the descriptive statistics of socio-demographic data at the
individual and household levels. The proportion of women in the sample data is greater
than that of men. The average age of participants is 50.64 and nearly 60% are employed.
The ratio of holding a valid driver’s license is higher than 90%, while the transit pass
ownership ratio is only 5.3%.
The households in the sample data have 2.89 members and 2.10 vehicles on
average. The average percentage of licensed drivers in the household is almost 80% and
the mean of household income is around $70,000. The table shows that roughly 65% of
households do not have any children. Approximately 10% of households have a child 5
years old or younger, around 7% of households’ youngest child is between the ages of 6
to 10, about 11% of households have a child between the ages of 11 and 15, and 6% of
the households’ youngest child is between the ages of 16 and 18.
Table 5. Descriptive Statistics of Socio-demographics in the Sample Data Used for
Individual-level and Household-level Models
Variables Sample Percentage Mean Std. Dev.
Personal characteristics
Female 51.864%
Age 50.642 17.190
Has driver’s license 93.044%
Has transit pass 5.306%
Employed 59.465%
Household (HH) attributes
HH size 2.886 1.192
The number of HH vehicles 2.101 0.992
Percentage of licensed driver in the HH 79.173% 26.622%
HH income (in thousand) 70.630 48.005
No children presence in the HH 65.216%
Youngest child’s age up to 5 in the HH 9.621%
Youngest child’s age 6 to 10 in the HH 7.578%
Youngest child’s age 11 to 15 in the HH 10.927%
Youngest child’s age 16 to 18 in the HH 6.661% Source: Authors’ calculation
76
3.3.3 Travel Behavior Comparison Analysis
This section provides the descriptive statistics of trip chaining, joint travel, tour
generation, and travel distances across the binary and categorical socio-demographics
(gender, driver’s license ownership, employment status, transit pass ownership, and the
presence and age stages of children in the household). The Kruskal-Wallis one-way
analysis of variance is performed to identify whether the differences in these four aspects
of travel behavior across the selected socio-demographics are statistically significant. The
Kruskal-Wallis test is a non-parametric test that does not assume a normal distribution. It
can be used to determine whether there are statistically significant differences between
two or more groups of an independent variable on a continuous or ordinal dependent
variable (Kruskal & Wallis, 1952). This method is chosen due to the fact that some travel
behavior variables do not follow a normal distribution in the sample data. Tables 6 to 9
provide the descriptive statistics and the corresponding results of the comparison
analysis.
Table 6 demonstrates the descriptive statistics of trip chaining, joint travel, tour
generation, and travel distances at the tour and individual levels across the binary
personal characteristics. Table 7 shows the Kruskal-Wallis test results for these selected
personal characteristics. At the tour level, there are statistically significant differences of
trip chaining for driver’s license ownership and transit pass ownership. The results show
that people without driver’s licenses and people with transit passes tend to make complex
tours. At the tour and individual levels, significant differences of joint travel exist
between females and males, between people with driver’s licenses and people without
77
driver’s licenses, and between workers and non-workers. The findings indicate that
females, people without driver’s licenses, and non-workers tend to travel with household
members. The differences of travel distances at the tour level and VMT at the individual
level are statistically significant across all selected personal characteristics. The analysis
shows that males, people with driver’s licenses, workers, and transit pass owners are
more likely to travel further.
Table 8 presents the descriptive statistics of trip chaining, joint travel, tour
generation, and travel distances at the tour, individual, and household levels across five
household structure groups (Households without children, households with youngest
child’s age up to 5, households with youngest child’s age between 6 to 10, households
with youngest child’s age between 11 to 15, and households with youngest child’s age
between 16 to 18). The Kruskal-Wallis test results of four aspects of travel behavior
across these groups are shown in Table 9. The findings show that the differences in joint
travel, number of auto tours, and travel distances/VMT at the tour, individual, and
household levels are statistically significant among the groups of the presence and age
stages of youngest child in the household. The results indicate that households with
youngest child’s age up to 5 are more likely to travel jointly than others at the tour and
household levels. Households with youngest child’s age between 11 to 15 generate more
auto tours at individual and household levels. The findings of travel distances at the tour,
individual, and household levels across the five groups are mixed. Households without
children travel further at the tour level. People from households with youngest child’s age
between 6 to 10 generate more individual VMT. Households with youngest child’s age
between 11 to 15 drive further than others at the household level.
78
These findings suggest that the differences in travel behavior vary across different
socio-demographics. However, the comparison results are based on binary or categorical
variables. The results may be biased and the effects may be over or under estimated if
other socio-demographics and urban form factors are also relevant to the travel behavior.
In addition, the comparison analysis does not examine the mediating effects of the
variables through activity-travel patterns on travel demand. Therefore, further
investigation of the links among travel behavior, socio-demographics, and urban form
factors is needed.
79
Table 6. Descriptive Statistics of the Travel Behavior at the Tour and Individual Levels Across the Binary Personal
Characteristics
Gender Driver’s license
ownership Employ status
Transit pass
ownership Female Male With Without Employed Unemployed With Without
Travel behavior at the tour level
Trip
chaining
Mean 2.458 2.294 2.376 2.901 2.427 2.426 3.071 2.383
S.D. 1.927 1.880 1.838 2.386 1.908 1.901 2.501 1.848
Joint travel Mean 0.229 0.267 0.238 0.352 0.198 0.360 0.256 0.248
S.D. 0.401 0.386 0.389 0.430 0.316 0.427 0.387 0.395
Travel
distances
Mean 16.298 17.989 17.717 11.434 19.492 13.922 15.234 17.236
S.D. 16.527 18.438 17.838 11.451 18.951 14.726 16.321 17.558
Travel behavior at the individual level
Trip
chaining
Mean 2.414 2.489 2.438 2.656 2.451 2.457 2.646 2.442
S.D. 1.726 1.647 1.656 2.067 1.696 1.679 1.876 1.677
Joint travel Mean 0.262 0.217 0.223 0.470 0.177 0.334 0.266 0.239
S.D. 0.367 0.343 0.343 0.442 0.304 0.403 0.374 0.355
Number of
auto tours
Mean 1.588 1.563 1.600 1.251 1.552 1.610 1.369 1.587
S.D. 0.819 0.780 0.811 0.547 0.768 0.846 0.651 0.807
VMT Mean 26.601 29.321 28.671 17.734 30.685 23.841 23.718 28.145
S.D. 21.747 24.400 23.339 16.561 23.533 21.824 20.622 23.210 The bold statistics indicate that the difference between the given travel behavior is statistically significant at the 95% level
Source: Authors’ calculation
79
80
Table 7. Kruskal-Wallis Test Results of the Travel Behavior at the Tour and Individual Levels Across the Binary Personal
Characteristics
Gender Driver’s license
ownership Employ status
Transit pass
ownership
Female VS Male With VS
Without
Employed VS
Unemployed
With VS
Without
Travel behavior at the tour level
Trip chaining chi-square 3.231 22.887 0.004 27.254
p-value 0.072 0.001 0.950 0.001
Joint travel chi-square 22.834 65.784 147.995 1.596
p-value 0.001 0.001 0.001 0.206
Travel
distances
chi-square 6.293 127.899 181.402 6.880
p-value 0.012 0.001 0.001 0.010
Travel behavior at the individual level
Trip chaining chi-square 2.880 0.259 0.316 1.128
p-value 0.090 0.617 0.573 0.288
Joint travel chi-square 18.096 94.992 163.080 1.120
p-value 0.001 0.001 0.001 0.290
Number of
auto tours
chi-square 0.434 60.603 2.324 17.225
p-value 0.509 0.001 0.127 0.001
VMT chi-square 7.559 82.426 124.984 8.192
p-value 0.006 0.001 0.000 0.004 The bold statistics indicate that the difference between the given travel behavior is statistically significant at the 95% level.
Source: Authors’ calculation
80
81
Table 8. Descriptive Statistics of the Travel Behavior at the Tour and Individual Levels Across the Presence and Age Stages of
the Youngest Child in the Household
The bold statistics indicate that the difference between the given travel behavior is statistically significant at the 95% level. Source: Authors’ calculation
Presence and age stages of the youngest child in the household
Without
children
Youngest
child’s age up to 5
Youngest
child’s age 6 to 10
Youngest
child’s age 11 to 15
Youngest
child’s age 16 to 18
Travel behavior at the tour level
Trip chaining Mean 2.459 2.443 2.389 2.297 2.428
S.D. 1.890 1.989 1.975 1.778 2.013
Joint travel Mean 0.228 0.297 0.278 0.288 0.247
S.D. 0.391 0.404 0.394 0.406 0.386
Travel
distances
Mean 17.577 16.102 16.736 16.612 16.166
S.D. 17.529 17.235 18.092 17.509 16.820
Travel behavior at the individual level
Trip chaining Mean 2.466 2.476 2.571 2.322 2.416
S.D. 1.672 1.685 1.860 1.482 1.918
Joint travel Mean 0.221 0.272 0.273 0.279 0.266
S.D. 0.356 0.353 0.350 0.356 0.359
Number of
auto tours
Mean 1.502 1.635 1.791 1.793 1.564
S.D. 0.725 0.847 0.969 0.961 0.790
VMT Mean 27.243 28.006 31.378 30.318 26.453
S.D. 22.646 23.598 24.534 24.382 23.396
Travel behavior at the household level
Trip chaining Mean 2.487 2.414 2.579 2.289 2.355
S.D. 1.496 1.427 1.673 1.161 1.460
Joint travel Mean 0.162 0.260 0.242 0.259 0.184
S.D. 0.289 0.312 0.279 0.268 0.247
Number of
auto tours
Mean 2.469 2.982 3.299 3.968 3.477
S.D. 1.319 1.902 1.972 2.092 2.052
VMT Mean 43.895 46.577 52.912 60.502 56.626
S.D. 35.684 39.829 39.317 42.078 42.298
81
82
Table 9. Kruskal-Wallis Test Results of the Travel Behavior at the Tour and Individual
Levels Across the Presence and Age Stages of the Youngest Child in the Household
Presence and age stages of the
youngest child in the household
Travel behavior at the tour level
Trip Chaining chi-square 10.813
p-value 0.029
Joint Travel chi-square 67.783
p-value 0.000
Travel Distances chi-square 23.825
p-value 0.001
Travel behavior at the individual level
Trip Chaining chi-square 6.718
p-value 0.152
Joint Travel chi-square 47.574
p-value 0.000
Number of auto tours chi-square 63.441
p-value 0.000
VMT chi-square 17.217
p-value 0.002
Travel behavior at the household level
Trip Chaining chi-square 2.425
p-value 0.658
Joint Travel chi-square 105.568
p-value 0.000
Number of auto tours chi-square 160.704
p-value 0.000
VMT chi-square 55.525
p-value 0.000 The bold statistics indicate that the difference between the given travel behavior is statistically significant at the 95%
level.
Source: Authors’ calculation
83
Chapter 4: Methodology
As discussed in Chapter 2, many previous studies have explained trip chaining,
joint travel, trip/tour generation, and travel distances by the exogenous variables (e.g.,
socio-demographics and urban form factors). A number of studies have used activity-
travel patterns (e.g., trip chaining) as one of the exogenous variables to explain the travel
outcomes (e.g., trip/tour generation and travel distances). These findings indicate that
some exogenous variables may affect trip/tour generation and trip distances either
directly or indirectly through the activity-travel patterns. Therefore, using traditional
regression methods may not be sufficient to examine the direct and indirect relationships
at the same time. Structural Equation Model (SEM) is a good alternative to measure the
interrelationships among the multiple endogenous variables and the links between
endogenous and exogenous variables simultaneously (Cao, 2007). The path analysis, a
special case of the SEM that only includes observed variables (van den Berg, Arentze, &
Timmermans, 2013), is applied in this study.
The first section of this chapter briefly introduces the SEM, path analysis, and the
concept of direct, indirect, and total effects. The second section presents the assumptions
of the model. Finally, the indices commonly used for model fit evaluation are discussed.
84
4.1 Path Analysis and Structural Equation Modeling
SEM is a statistical modeling technique that is used to deal with several directions
of influence among multiple endogenous and exogenous variables. The SEM has been
widely used in different fields including education, psychology, business, and sociology
(Liu, 2012). Golob (2003) provides a comprehensive review of studies using SEM to
examine different aspects of travel behavior. These studies cover dynamic travel demand
models, activity-based travel analysis, driver behavior and travelers’ attitudes and
perceptions toward transportation mode choices. As compared to traditional multiple
regression models, the SEM has the following advantages. First, it accomdates multiple
dependent and independent variables and any number of equations. Second, it corrects for
measurement error in all observed variables. Third, it provides an approach of modeling
the mediation effects by estimating the direct, indirect, and total effects among variables
(Byrne, 2013; Golob, 2003; Kline, 2015). A typical form of path analysis in SEM with
“G” endogenous variables can be presented as follows (Kuppam & Pendyala, 2001; Lu &
Pas, 1999):
[ 𝑌1
𝑌2
.
.
.𝑌𝐺]
= [𝑌 𝑋] [BГ] +
[ 𝜀1
𝜀2
.
.
.𝜀𝐺]
(8)
The above matrix equation system can be rewritten as
𝑌 = (𝐼 − 𝐵)−1(Γ𝑋 + 𝜀) (9)
(or)
𝑌 = 𝐵𝑌 + Γ𝑋 + 𝜀 (10)
85
Where 𝑌 is a column vector of observed endogenous variables,
B is a matrix of coefficients associated with right-hand-side endogenous variables,
𝑋 is a column vector of observed exogenous variables,
Г is a matrix of coefficients associated with observed exogenous variables, and
𝜀 is a p×1 vector of residuals of the endogenous variables.
This research employs path analysis to disentangle the interconnections among
travel behavior, socio-demographics, and urban form factors. By using activity-travel
patterns as mediating variables, this study estimates the direct, indirect, and total effects
of socio-demographics and urban form factors on travel behavior. Direct effects are the
links that go directly from one variable to another. Indirect effects exist between two
variables that are mediated by one or more intervening variables. Indirect effects are
therefore calculated as the sum of all the effects of all intervening variables. Total effects
are the sum of direct effects and indirect effects. Figure 17 illustrates a simple example
presenting the concept of direct effects, indirect effects, and total effects.
The example assumes that variable C is affected directly by variable A (with the
coefficient Pac) and variable B (with the coefficient Pbc). Variable B is influenced directly
by variable A at the same time (with coefficient Pab). Therefore, the direct effects of
variable A on variable B and variable C are Pab and Pac respectively. The direct effects of
variable B on variable C are represented as Pbc. The indirect effects of variable A on
variable C are calculated as Pab × Pbc. The total effects of variable A on variable C are
calculated as Pac + Pab × Pbc.
86
Assume the coefficients Pac and Pab are both positive, but the coefficient Pbc is
negative. The indirect effects of variable A on variable C are due to the effects on
variable B and the effects through variable B. The direct effects of variable A on variable
B are positive, but because variable B affects variable C negatively, the magnitude of the
indirect effects of variable A on variable C decreases. In other words, variable A may
indirectly bring about a decrease in variable C by increasing variable B. In this case, the
total effects of variable A on variable C (calculated as Pac + Pab × Pbc) are smaller than the
direct effects of variable A on variable C (Pac). This implies that the mediating variable B
weakens the direct effects of variable A on variable C.
Figure 17. Illustration of direct, indirect, and total effects
4.2 Model Estimation and Assumptions
The general approach to estimating SEM coefficients is the covariance structure
analysis (Liu, 2012). Assume that the covariance matrix of the input data is represented
as 𝛴 and the model-implied covariance matrix is denoted as �̂�. All the parameters from
Equation 8 can be represented as a parameter vector of 𝛺. The estimation goal is to find
the set of estimated parameters �̂� that minimizes the discrepancy between 𝛴 and �̂�. In
other words, the coefficients are estimated by applying covariance analysis to make the
87
variance covariance matrix implied by the model as similar as possible to the input
variance covariance matrix (Bowen & Guo, 2011; Golob, 2003).
The commonly-used estimation approach for the SEM is maximum likelihood
(ML), which assumes that the observed variables follow a multivariate normally
distribution. Given the fact that the sample data employed in this study involve non-
normal variables, the bootstrapping procedure with maximum likelihood (ML) estimator
is applied for coefficient estimation. Bootstrapping procedure, a resampling method, is
applied when assumptions of multivariate normality may not hold. Applying this
procedure also allows for confirming the mediation effects and assessing the stability of
parameter estimates (Byrne, 2013; Cheung & Lau, 2008; Lin & Wang, 2014).
4.3 Model Fit Indices
The model fit indices of SEM determine how close the implied variance-
covariance matrix �̂� is to the observed variance covariance matrix 𝛴. Commonly used
model fit indices are summarized in
89
Table 10. Model Fit Indices and Acceptable Fit Criteria
Model Fit Indices Cut-off Value Description
Chi-square (χ2) p > 0.05
Measuring the discrepancy between
observed covariance matrix and
model-implied covariance matrix
Root-mean-square Error of
Approximation (RMSEA) <0.05
Measuring the amount of error of
approximation per model degree of
freedom.
Comparative Fit Index
(CFI) >0.90
Assessing the improvement in fit of
the researcher’s hypothesized model to
the independence model
Standardized Root Mean
Square Residual (SRMR) <0.05
Measuring the standardized residuals
of observed covariance matrix and
model-implied covariance matrix Source: Authors’ summary
4.3.1 Chi-square (χ2)
Equation 11 shows the formula for calculating the χ2 statistic. A nonsignificant χ2
indicates that two variance-covariance matrices are statistically similar and the model
cannot be rejected. However, the χ2 value is sensitive to sample size. As sample size
increases (generally above 200), the χ2 value increases and, thus, the model may be
rejected even though the difference between the input sample variance-covariance matrix
and the model-implied variance-covariance matrix may be small. (Schumacker & Lomax,
2004; Van Acker & Witlox, 2010).
𝜒2 = (N − 1)×𝐹𝑀𝐿 (11)
Where 𝜒2 is chi-square value,
N is sample size and,
𝐹𝑀𝐿 is the maximum likelihood fit function
90
4.3.2 Root-mean-square Error of Approximation (RMSEA)
The formula for calculating the RMSEA is shown in Equation 12. RMSEA is a
parsimony-adjusted index which measures the amount of error of approximation per
model degree of freedom. Smaller RMSEA value refers to better model fit. A RMSEA
value less than 0.05 is considered acceptable (Browne, Cudeck, & Bollen, 1993;
Schumacker & Lomax, 2004).
𝑅𝑀𝑆𝐸𝐴 =√
max [(𝜒2
𝑑𝑓) − 1
𝑁 − 1, 0]
(12)
Where 𝜒2 is chi-square value,
𝑑𝑓 is degree of freedom and,
N is sample size
4.3.3 Comparative Fit Index (CFI)
The CFI is an incremental fit index which assesses the relative improvement in fit
of the researcher’s hypothesized model over a baseline model. The baseline model
assumes zero population covariance among the observed variables. The CFI value ranges
from 0 to 1, and the formula to calculate CFI is presented as Equation 13. A model with
CFI value greater than 0.90 is considered as an acceptable model (Kline, 2015).
𝐶𝐹𝐼 = 1 −
max[(𝜒𝑀2 − 𝑑𝑓𝑀) , 0]
max[(𝜒𝐵2 − 𝑑𝑓𝐵) , 0]
(13)
91
Where 𝜒𝑀2 is the chi-square value for the researcher’s hypothesized model,
𝑑𝑓𝑀 is the degree of freedom for the researcher’s hypothesized model,
𝜒𝐵2 is the chi-square value for the baseline model ,and
𝑑𝑓𝐵 is the degree of freedom for the baseline model.
4.3.4 Standardized Root Mean Square Residual (SRMR)
SRMR measures the standardized residuals of input sample variance-covariance
matrix and model-implied variance-covariance matrix. The value of SRMR ranges from 0
to 1, with well-fitting models obtaining values less than 0.05 (Byrne, 2013; Hu & Bentler,
1999).
92
Chapter 5: Results and Discussions for the Tour-based Model
The tour-level model uses a home-based tour as the basic unit of analysis and
investigates how trip chaining and joint travel contribute to travel distances at the home-
based tour level. The model regards trip chaining and joint travel as mediating variables
of travel distances. Figure 18 presents the model framework at the tour level. Trip
chaining is captured by the number of stops and joint travel is measured by the ratio of
joint trips for each tour at the tour level. The travel distances are the distances in miles of
each home-based tour. The analysis examines the effects of personal characteristics,
household attributes, and urban form factors at tour origins and destinations on the three
aspects of travel behavior simultaneously. This chapter discusses the results of the tour-
level model. Table 11 reports model fit indices. Table 12 shows standardized effects of
travel behavior, socio-demographics, and urban form factors of the tour-level model.
Figure 19 demonstrates the paths with standardized direct effects between endogenous
and exogenous variables at the tour level. All coefficients in the model are examined and
only those with p-values less than 0.1 are kept and presented.
93
Figure 18. Model Framework at the Tour Level
This chapter is organized as followed. The first section reports the model fit
indices. The relationships among trip chaining, joint travel, and travel distances are
presented in the second section. The direct effects of socio-demographics and urban form
factors are then discussed. The final section discusses the indirect and total effects of the
exogenous variables.
5.1 Model Fit Indices in the Tour-level Model
Based on the indices shown in Table 11, the tour-level model has a χ2 of 21.886
with 32 degrees of freedom, which brings a probability of 0.911. This implies that the
model fits the data very well and cannot be rejected. The ratio of χ2 and degrees of
freedom is 0.683, which is an indicator of a good fit (Schermelleh-Engel, Moosbrugger,
& Müller, 2003). The value of RMSEA measuring the approximation error is 0.001. The
value CFI that examines the discrepancy between the data and the hypothesized model is
0.990. The SRMR capturing the standardized difference between the observed correlation
94
and predicted correlation is 0.003. The three indices are within the range of the cut-off
criteria, which indicate that the tour-level model is considered as acceptable (Hu &
Bentler, 1999; Mueller & Hancock, 2010; Schumacker & Lomax, 2004).
Table 11. Model Fit Indices of the Tour-level Model
Model fit indices Value Recommended cut-off values
χ2 21.866
degree of freedom 32.000
χ2 / degree of freedom 0.683
χ2 P-value 0.911 >0.05
RMSEA 0.001 <0.08
CFI 0.990 >0.90
SRMR 0.003 <0.05
5.2 Links Between Activity-travel Patterns and Travel Demand in the Tour-level Model
Based on the standardized coefficients presented in Table 12, the direct effects
among trip chaining, joint travel, and travel distances are statistically significant. Trip
chaining has negative direct effects on joint travel. This finding indicates that tour
complexity may decrease the tendency of sharing the same travel patterns with other
household members. As expected, increasing trip chaining directly increases travel
distances. This may be due to the fact that as the tours become more complex, travelers
tend to take indirect routes to their desired destinations. All else being equal, joint travel
ratio has negative direct effects on the resulting travel distances. These results show that
tours with higher joint travel ratios are more likely to be short tours. The interpretation of
this finding could be that traveling longer distances requires greater spatial-temporal
coordination among family members than traveling shorter distances. In addition, the
95
home-based tours with high joint travel ratio tend to be short and local, such as tours of
chauffeuring children to school. The positive indirect effects of trip chaining on travel
distances are because of the negative directs effects of trip chaining to joint travel and
from joint travel to travel distances. This implies that the tour complexity also increases
travel distances indirectly through joint travel. The total effects of trip chaining to travel
distances are the sum of its direct and indirect effects. The results show that the positive
total effects on travel distance are reinforced by the positive direct and indirect effects.
5.3 Direct Effects of Socio-demographics on Travel Behavior in the Tour-level Model
According to Table 12, most socio-demographic variables have significant direct
effects on travel behavior. Consistent with the recent research (Akar et al., 2016; Elldér,
2014; Jahanshahi, Jin, & Williams, 2015), females are found to make shorter tours than
males. The positive effects of being a female on joint travel mean that women are more
likely to travel with household members as compared to men. This is probably due to the
fact that females are more dependent on others in terms of traveling and participating in
out-of-home activities. It may be also because females are more apt to be involved in
parenting activities. Age poses a negative influence on trip chaining, showing that older
travelers tend to make simple home-based tours.
The possession of a driver’s license has significant influence on trip chaining,
joint travel, and travel distances. The model results show that driver’s license ownership
is negatively related to trip chaining, which indicates that licensed drivers tend to make
simple tours. The driver’s license ownership negatively affects joint travel. This can be
96
explained by the fact that people without driver’s licenses are more reliant on those with
licenses in an auto-oriented environment. In addition, people with driver’s license are
inclined to travel longer distances, which is similar to the findings in existing studies
(Akar et al., 2016; Mercado & Páez, 2009; Morency, Paez, Roorda, Mercado, & Farber,
2011; Schmöcker, Quddus, Noland, & Bell, 2005).
Consistent with the findings of Lin and Wang (2014), the negative effects of
being employed on joint travel reveal that workers have higher propensities to travel
alone. Due to workers’ less flexible daily schedules, they may have fewer intra-household
interactions on traveling with their household members. In agreement with previous
literature (Akar et al., 2016; Liu, 2012; Mercado & Páez, 2009; Schmöcker et al., 2005),
workers are more likely to travel longer distances. The positive effects of transit pass
ownership on trip chaining reveal that pass owners tend to make complex tours. This is
probably because they tend to use public transit, which may require transfers to arrive at
travelers’ desired destinations. Susilo and Kitamura (2008) argue that transit pass holders
have greater propensities to visit non-work places near transit stops on their commute
routes.
With respect to the direct effects of household attributes, Table 12 reports that
household size negatively affects tour complexity and travel distances. Similar to the
findings of Kitamura and Susilo (2005) and Susilo and Kitamura (2008), people from
larger households tend to make shorter and simpler tours. Wallace et al. (2000) indicate
that larger households may have greater numbers and variety of destinations, which
decrease the propensity of making complex home-based tours. As expected, household
97
size appears to be positively associated with the ratio of joint travel. This shows that
individuals from large households are more likely to travel with household members,
which is consistent with the findings of Chandrasekharan and Goulias (1999). Household
income is positively associated with travel distances, which implies that household
income may be viewed as a resource constraint while traveling. Since wealthier families
may be less sensitive to travel costs, they may be less likely to combine trip segments to
save money.
The number of household vehicles affects joint travel and travel distances
differently: increasing number of household vehicles decreases joint travel but increases
travel distances. Households are prompted to coordinate their travel schedules if fewer
household vehicles are available. The findings could be explained by the fact that
household vehicle ownership is also considered as a resource constraint for traveling.
5.4 Direct Effects of Urban Form Factors on Travel Behavior in the Tour-level Model
The model results in Table 12 reveal that the urban form factors are influential
determinants of travel behavior at tour origins (i.e., household locations) as well as the
tour destinations (i.e., out-of-home primary activity locations). Similar to the findings of
Kitamura and Susilo (2005) and Antipova and Wang (2010), high residential density at
residence neighborhoods is found to increase tour complexity. The estimated results show
residential density at tour origins and destinations negatively affect travel distances,
implying that the tours people start and end in compact residential areas tend to be short.
The retail density at tour destinations is positively associated with joint travel. This
98
implies that people may tend to travel jointly when their primary activities are located in
retail intensive areas. At tour origins, it is reasonable to see that the effects of retail
density and non-retail density on activity-travel patterns are opposite. The retail density
influences trip chaining and joint travel negatively, whereas the non-retail density affects
both activity-travel patterns positively.
Intersection density and bus stop density capture the street connectivity and public
transit accessibility respectively. The results demonstrate that intersection density at tour
destinations decreases the ratio of joint travel. As anticipated, the negative effects of
intersection density at tour origins imply that people who live in neighborhoods with
better street connectivity are more likely to make short tours. Increasing bus stop density
at tour destinations increases tour complexity and travel distances. This could be
explained by the fact that people traveling to areas with better transit facilities tend to
make complex tours and travel further.
The job-population index measures the balance of employment and residential
population. It affects travel distances negatively at both tour origins and destinations. The
index increases the propensity of traveling jointly at origins and destinations. The results
show that people are inclined to travel with family companions and make shorter tours if
they live in or travel to areas with a well-balanced ratio of employment and population,
such as commercial and residential mix areas.
99
5.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Tour-level Model
Two activity-travel patterns, trip chaining and joint travel, are used as mediating
variables of the resulting travel distances in the tour-level model. The indirect effects in
Table 12 show that some of the exogenous variables affect travel distances through
activity-travel patterns indirectly. The model results suggest that while the indirect effects
of some socio-demographics are channeled through either one of the patterns, others are
channeled through both. For example, the negative indirect effects of being a female on
travel distances are through joint travel only. The positive indirect effects of employment
status and the number of household vehicles on travel distances are only channeled via
joint travel. On the other hand, driver’s license ownership and household size have
indirect effects on travel distances via both activity-travel patterns. These two exogenous
variables have negative indirect effects on travel distances.
The analysis reveals that the indirect effects of most urban form variables at tour
origins and destinations are statistically significant. Most urban form factors affect travel
distances indirectly through either one of activity-travel patterns. For instance, the
positive indirect effects of bus stop density at tour destinations are only channeled
through trip chaining. On the other hand, the job-population index at tour origins and
destinations has negative indirect effects through joint travel. Furthermore, the model
results indicate that some urban form factors affect the travel distances indirectly only.
The retail density at tour destinations decreases travel distances indirectly through
increasing the propensity of joint travel. In contrast, increasing non-retail density at tour
origins increases travel distances indirectly via both activity-travel patterns. These
100
findings suggest that some urban form variables may not directly affect travel distances,
but may have indirect effects on tour distances through joint travel and/or trip chaining.
5.6 Total Effects of Exogenous Variables on Travel Behavior in the Tour-level Model
The total effects are measured as the sum of direct and indirect effects. Table 12
shows that most total effects of exogenous variables on travel distances keep the same
signs as the corresponding direct effects or indirect effects. Nearly all socio-demographic
variables affect travel distances significantly at the 95 percent confidence level. The
absolute magnitudes of standardized total effects help determine which variables have
greater influences on travel distances. The results suggest that employment status has the
strongest effect on travel distances among personal characteristics. Among household
attributes, household size influences the resulting travel distances the most.
With regard to urban form factors, all total effects of the measures at tour origins
and destinations on travel distances are statistically significant except for intersection
density at tour destinations. In addition, residential density, non-retail density, bus stop
density, and job-population index affect travel distances at destinations more than they do
at tour origins. This finding indicates that significant influence of urban form on travel
distances exists not only at household locations but also at primary activity locations.
Based on the absolute magnitudes of standardized total effects, travel distances are
affected by job-population index the most at tour destinations and by intersection density
at tour origins.
101
Table 12. Estimation Results for the Tour-level Model (Standardized Effects)
Travel distances Joint travel Trip chaining
Types of Effects Direct Indirect Total Direct Indirect Total Direct Total
Endogenous variables: Joint travel and Trip chaining
Joint travel -0.023 (0.014)
-0.023 (0.014)
Trip chaining 0.426
(<0.01) 0.001
(0.031) 0.427
(<0.01) -0.022 (0.021)
-0.022 (0.021)
Exogenous variables: Personal Socio-demographics
Female -0.044 (<0.01)
-0.001 (0.018)
-0.045 (<0.01)
0.037 (<0.01)
0.037 (<0.01)
Age
-0.014 (0.019)
-0.014 (0.019)
0.001 (0.035)
0.001 (0.035)
-0.032 (0.012)
-0.032 (0.012)
Driver’s license ownership 0.058
(<0.01) -0.02
(<0.01) 0.037
(<0.01) -0.032 (0.029)
0.001 (0.022)
-0.031 (0.030)
-0.049 (<0.01)
-0.048\9 (<0.01)
Employed status 0.108
(<0.01) 0.012 (0.02)
0.120 (<0.01)
-0.133 (<0.01)
-0.133 (<0.01)
Transit pass ownership
0.026 (0.01)
0.026 (<0.01)
-0.001 (0.022)
0.060 (<0.01)
0.060 (<0.01)
Exogenous variables: Household Socio-demographics
Household size -0.057 (<0.01)
-0.023 (<0.01)
-0.080 (<0.01)
0.106 (<0.01)
0.001 (0.022)
0.107 (<0.01)
-0.048 (<0.01)
-0.048 (<0.01)
Household income 0.043
(<0.01) -0.015 (0.012)
0.028 (0.049)
0.001 (0.023)
0.001 (0.023)
-0.036 (<0.01)
-0.036 (<0.01)
Household vehicle 0.041
(<0.01) 0.003
(0.018) 0.044
(<0.01) -0.134 (<0.01)
-0.134 (<0.01)
No children presence (base case)
Youngest child’s age up to 5
Youngest child’s age 6 to 10
Youngest child’s age 11 to 15
-0.001a (0.063)
-0.001a (0.063)
0.026 a (0.051)
0.026 a (0.051)
Youngest child’s age 16 to 18
(CONTINUED)
100
102
Table 12. (cont.) Estimation Results for the Tour-level Model (Standardized Effects) Travel distances Joint travel Trip chaining
Types of Effects Direct Indirect Total Direct Indirect Total Direct Total
Exogenous variables: Urban form at tour destinations
Residential density -0.056 (<0.01)
-0.056 (<0.01)
Retail density
-0.001 (0.016)
-0.001 (0.016)
0.036 (<0.01)
0.036 (<0.01)
Non-retail density 0.067
(<0.01)
0.067 (<0.01)
Intersection density
0.001 (0.014)
0.031a (0.090)
-0.058 (<0.01)
-0.058 (<0.01)
Bus stop density 0.043
(0.021) 0.023
(<0.01) 0.066
(<0.01)
-0.001 (0.021)
-0.001 (0.021)
0.054 (<0.01)
0.054 (<0.01)
Job-population index -0.068 (<0.01)
-0.001 (0.042)
-0.069 (<0.01)
0.028 (0.032)
0.028 (0.032)
Exogenous variables: Urban form at tour origins
Residential density -0.059 (0.012)
0.014 (0.020)
-0.044 (0.034)
-0.001 (0.04)
-0.001 (0.04)
0.033 (0.020)
0.033 (0.020)
Retail density -0.025 (<0.01)
-0.044 (<0.01)
-0.059 (<0.01)
0.001 (0.021)
-0.057 (<0.01)
-0.061 (<0.01)
-0.061 (<0.01)
Non-retail density
0.018 (0.027)
0.018 (0.027)
0.04 (<0.01)
-0.001 (0.040)
0.039 (0.011)
0.045 (0.022)
0.045 (0.022)
Intersection density -0.056 (<0.01)
-0.056 (<0.01)
Bus stop density -0.038 (<0.01)
0.001 (0.033)
-0.038 (<0.01)
-0.024 (0.023)
-0.024 (0.023)
Job-population index -0.046 (<0.01)
-0.001 (0.031)
-0.047 (<0.01)
0.029 (0.021)
0.029 (0.021)
Note: Coefficients in bold type are statistically significant at the 95% level; p-values are shown in parenthesis; superscript a means coefficients are statistically significant
at the 90% level
101
104
Chapter 6: Results and Discussions of the Individual-level Model
The individual-level model investigates the relationships between activity-travel
patterns, travel demand, socio-demographics, and urban form at the individual level. The
sample data used in the structural analysis include 6,599 auto tours traveled by 4,189
individuals from 2,298 households in the Cleveland Metropolitan Area. Figure 20 shows
the framework of the individual-level model. Trip chaining and joint travel are two
activity-travel patterns at the individual level, which are measured as the number of stops
per auto tour and ratio of joint travel per auto tour. The number of auto tours and VMT
traveled by an individual are considered as two individual travel demand outcomes.
Personal and household attributes characterize socio-demographics in the analysis. The
urban form factors only include residential built environment features in the model. This
chapter discusses the results derived from the individual-level model. Table 13 shows the
model fit indices and Table 14 reports the relationships among travel behavior, socio-
demographics, and urban form factors. Figure 21 presents the paths with standardized
direct effects between endogenous and exogenous variables at the individual level. All
coefficients in the individual -level model are investigated and only those with p-values
less than 0.1 are kept and discussed
105
Figure 20. Model Framework at the Individual Level
The first section presents the model fit indices. The effects among different
aspects of travel behavior are shown in the second section. The third and fourth sections
discuss the direct effects of socio-demographics and residential urban form factors. The
indirect and total effects of exogenous variables are reported in the last two sections.
6.1 Model Fit Indices for the Individual-level Model
The commonly used model fit indices in the individual-level model and the
recommended cut-off values are presented in Table 13. The χ2 value is 36.009 with 34
degrees of freedom, which yields the p-value for χ2 to be 0.375 and the ratio of χ2 to
degrees of freedom is 1.059. These values suggest that the model fits the data very well
and cannot be rejected (Schermelleh-Engel et al., 2003). Other indices such as RMSEA,
CFI, and SRMR are also within the range of recommended cut-off values (RMSEA =
0.004, CFI = 0.990, TLI = 0.990, and SRMR = 0.005). All the reported model fit indices
show that this individual-level model fits the data well (Hu & Bentler, 1999; Mueller &
Hancock, 2010).
106
Table 13. Model Fit Indices of the Individual-level Model
Model fit indices Value Recommended cut-off values
χ2 36.009
degree of freedom 34
χ2 / degree of freedom 1.059
χ2 P-value 0.375 >0.05
RMSEA 0.004 <0.08
CFI 0.990 >0.90
SRMR 0.005 <0.05
6.2 Links Between Activity-travel Patterns and Travel Demand in the Individual-level
Model
The standardized direct, indirect, and total effects among activity-travel patterns
and travel demand are shown in Table 14. The model results reveal that while trip
chaining does not have significant direct effects on joint travel, its direct effects on other
travel outcomes are significant as expected: trip chaining has negative direct effects on
the number of auto tours and positive direct effects on VMT. The findings indicate that
people with greater tour complexity tend to make fewer auto tours but they are expected
to generate more VMT. On the other hand, joint travel has positive direct effects only on
auto tour generation. This could be explained by the fact that people who have greater
travel demand may have a greater level of joint travel to achieve greater efficiency of
time use. Intuitively, the model results suggest that increasing the number of auto tours
increases individual VMT.
107
Both trip chaining and joint travel have significant indirect effects on VMT
through the number of auto tours. These two activity-travel patterns have opposite
indirect effects on VMT because the signs of their direct effects on auto tour generation
are in different directions. The model outcomes indicate while trip chaining has negative
indirect effects on VMT, the indirect effects of joint travel on VMT are positive. This
means that increasing tour complexity indirectly reduces VMT by decreasing the number
of tours. In contrast, joint travel ratio indirectly increases VMT through increasing the
number of auto tours.
The total effects of the number of auto tours, joint travel, and trip chaining on
VMT are positive. The total effects of tour generation on VMT keep the same signs and
magnitudes due to its direct effects on VMT and lack of the indirect effects. The total
effects of trip chaining on VMT have the same sign as its direct effects. The magnitude of
total effects is weakened through the negative indirect effects. This implies that although
tour complexity has overall positive effects on VMT, a small decrease brought about by
the number of auto tours is observed.
6.3 Direct Effects of Socio-demographics on Travel Behavior in the Individual-level
Model
Several key socio-demographics greatly explain activity-travel patterns and the
resulting travel demand in the individual-level model. The results show that most socio-
demographics have significant direct effects on joint travel, the number of auto tours, and
108
VMT. Among personal characteristics, most effects on joint travel are in the same
direction as at the tour-level model. For example, females tend to travel with other family
members as compared to males. On the other hand, driver’s license ownership, being
employed, and having a transit pass are negatively correlated with joint travel.
The influence of most exogenous personal attributes on travel demand is
consistent with existing research. For instance, the model results indicate that women
generate fewer VMT than men, which echoes the findings from existing research (Akar et
al., 2016; Sun, Ermagun, & Dan, 2016; Zhang et al., 2012). Consistent with the existing
literature (Akar et al., 2016; Lu & Pas, 1999), driver’s license holders tend to drive more
and further. Workers generate fewer auto tours, which may be because of their less
flexible schedules. Workers tend to travel further than non-workers do, which is
consistent with previous studies (Akar et al., 2016; Heres-Del-Valle & Niemeier, 2011;
Liu, 2012; Lu & Pas, 1999). The transit pass holders tend to have fewer auto tours. Such
finding is expected, considering they are more willing to use public transit for travel than
their counterparts.
The household characteristics have strong associations with joint travel, the
number of auto tours, and VMT in this study. The direct effects of some household
characteristics on joint travel are in the same direction as in the tour-level model:
household size is positively associated with joint travel and the number of household
vehicles is negatively related to the propensity of traveling with family members.
Echoing with the findings of previous studies (Hong et al., 2014; Zhang et al., 2012),
109
people from larger households are more likely to have fewer VMT. Travelers from
wealthier households tend to drive further, which is consistent with previous studies
(Akar et al., 2016; Cervero & Murakami, 2010; Hong et al., 2014).
The household structure is a critical factor that influences different aspects of
individual travel behavior. The negative sign of percentage of household drivers indicates
as the proportion of licensed drivers in the household increases, individuals drive less
frequently. The presence and age stage of children in the household affect trip chaining
and tour generation. The model results show that the direct effects of households with the
youngest child between the ages of 6 and 10 and households with the youngest child
between the ages of 11 and 15 on auto tour generation are significant and positive. This
finding corresponds to several previous studies (Golob, 2000; Manoj & Verma, 2015;
Noland & Thomas, 2007), which may be explained by the fact that the households with
children have greater need of assisting the children in participating in out-of-home
activities than the counterparts.
6.4 Direct Effects of Residential Urban Form Factors on Travel Behavior in the
Individual-level Model
The results in Table 14 indicate that most residential urban form factors have
significant direct effects on the individual travel behavior. Consistent with the results of
prior studies (Cervero & Murakami, 2010; Ding & Lu, 2016; Hong et al., 2014; Liu &
Shen, 2011; Zhang et al., 2012), people who live in high residential density areas tend to
110
generate fewer VMT. The direct effects of retail density on individual VMT and joint
travel are negative, implying that individuals who reside in high retail density
neighborhoods are less likely to travel with household members and travel further.
Consistent with Frank et al. (2008) and Van Acker and Witlox (2010), the retail density is
also negatively related to trip chaining. The positive effects of non-retail density on joint
travel indicate that people living in high non-retail density areas tend to travel with
household members.
Intersection and bus stop density variables measure how street connectivity and
transit facilities contribute to individuals’ auto travel behavior. Intersection density is
positively correlated with the complexity of auto tours, which is consistent with the
findings of Frank et al. (2008). A high intersection density neighborhood may represent a
pedestrian friendly environment (Hong et al., 2014). As expected, the analysis finds that
intersection density poses negative influence on VMT. This finding is consistent with
previous literature indicating that increasing intersection density reduces the use of
private vehicles and VMT (Etminani-Ghasrodashti & Ardeshiri, 2016; Hong et al., 2014).
Bus stop density is positively associated with the ratio of joint travel, implying that better
transit proximity provides more transportation alternatives and thus decreases the need
for auto joint travel. The job-population index captures land-use mix by measuring the
balance of employment and population in this analysis. Supporting the findings of many
previous studies (Ewing et al., 2011; Hong et al., 2014; Jiang et al., 2016; Kuzmyak,
2009; Zhang et al., 2012), the index is negatively related to individual VMT. This
indicates that people living in well-mixed neighborhoods tend to drive less.
111
6.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Individual-level
Model
The analysis results in Table 14 show that some socio-demographics and urban
form factors have significant effects on the number of auto tours and VMT indirectly
through trip chaining and joint travel. Based on the model outcomes, all personal
characteristics have significant indirect effects on auto tour generation and VMT. In
addition, the direct and indirect effects of some exogenous variables have different signs,
whereas others have the same. For example, owning a driver’s license has positive direct
effects but negative indirect effects on auto tour generation. Being employed has positive
direct effects but negative indirect effects on VMT. On the other hand, the direct and
indirect effects of driver’s license ownership on VMT are both positive. The direct and
indirect effects of employment status and transit pass ownership on auto tour generation
are negative. As compared to personal characteristics, fewer household attributes have
significant indirect effects on the frequency of auto tours and VMT. Household size and
having a child between the ages of 6 to 10 have positive indirect effects on the number of
auto tours and VMT respectively. The number of household vehicles and percentage of
household licensed drivers have negative indirect effects on the number of auto tours and
VMT respectively.
As for the indirect effects of residential urban form factors, Table 14 shows that
the direct and indirect effects of intersection density and job-population index on VMT
112
have opposite signs. In contrast, the direct and indirect effects of retail density are both
negative on VMT. Non-retail density and intersection density have significant indirect
effects on the number of auto tours channeled via different activity-travel patterns. While
increasing non-retail density increases auto tour generation indirectly through joint travel,
increasing intersection density decreases auto tour generation indirectly via trip chaining.
6.6 Total Effects of Exogenous Variables on Travel Demand in the Individual-level
Model
Consistent with the trends of the tour-based model, most total effects of socio-
demographics and urban form on individual travel demand are statistically significant and
keep the same signs as the corresponding direct effects. The magnitudes of some total
effects are different from the corresponding direct effects because they are either
reinforced or weakened through the direct and indirect effects. For instance, the
magnitude of total effects of driver’s license ownerships on VMT is greater than the
corresponding direct effects. It is due to the fact that the total effects are reinforced
through the direct and indirect effects. On the contrary, the magnitude of total effects of
employment status on VMT is smaller than its direct effects as the total effects are
weakened through the corresponding positive direct effects and negative indirect effects.
Among socio-demographics, the total effects of driver’s license ownerships on VMT and
the total effects of employment status and transit pass ownership on the number of auto
tours are reinforced through their direct and indirect effects. The total effects of driver’s
113
license ownership on the number of auto tours and the total effects of employment status
on VMT are weakened through their opposite direct and indirect effects. Similar trends
are shown for the effects of residential urban form. The total effects of retail density on
VMT are reinforced, whereas the total effects of job-population index are weakened.
Based on the absolute magnitudes of standardized total effects, auto tour
generation is affected by the driver’s license ownership the most among the socio-
demographics and by intersection density the most among urban form factors.
Employment status has the strongest effect on VMT among socio-demographics and
retail density affects VMT the most among urban form variables.
114
Table 14. Estimation Results for the Individual-level Model (Standardized Effects)
Ln(Individual VMT) Number of Auto Tours Join Travel Trip Chaining
Direct Indirect Total Direct Indirect Total Direct Total Direct Total
Endogenous variables:
Number of Auto Tours 0.393
(<0.01)
0.393 (<0.01)
Joint travel
0.026 (<0.01)
0.026 (<0.01)
0.065 (<0.01)
0.065 (<0.01)
Trip chaining 0.371
(<0.01) -0.048 (<0.01)
0.322 (<0.01)
-0.123 (<0.01)
-0.123 (<0.01)
Exogenous variables: Personal Socio-demographics
Female -0.036 (0.037)
0.001 (<0.01)
-0.034 (0.036)
0.003 (<0.01)
0.003 (<0.01)
0.051 (<0.01)
0.051 (<0.01)
Driver’s license ownership 0.044
(<0.01) 0.062
(<0.01) 0.106
(<0.01) 0.165
(<0.01) -0.007 (<0.01)
0.158 (<0.01)
-0.106 (<0.01)
-0.106 (<0.01)
Employed status 0.157
(<0.01) -0.031 (<0.01)
0.126 (<0.01)
-0.066 (<0.01)
-0.012 (<0.01)
-0.078 (<0.01)
-0.188 (<0.01)
-0.189 (<0.01)
Transit pass ownership
-0.024 (<0.01)
-0.024 (<0.01)
-0.059 (<0.01)
-0.002 (0.036)
-0.061 (<0.01)
-0.035 (0.040)
-0.035 (0.036)
Exogenous variables: Household Socio-demographics
Household size -0.057 (<0.01)
0.018 (<0.01)
-0.039 (<0.01)
0.038 (0.034)
0.008 (<0.01)
0.045 (0.046)
0.118 (<0.01)
0.118 (<0.01)
Household income 0.074
(<0.01)
0.074 (<0.01)
Household vehicle
-0.04 (<0.01)
-0.04 (<0.01)
-0.01 (<0.01)
-0.01 (<0.01)
-0.155 (<0.01)
-0.155 (<0.01)
Driver percentage
-0.034 (<0.01)
-0.034 (<0.01)
-0.086 (<0.01)
-0.086 (<0.01)
No children presence (base case)
Youngest child’s age up to 5
Youngest child’s age 6 to 10 0.016
(0.039)
0.016
(0.039)
0.042
(0.039)
0.042
(0.039)
Youngest child’s age 11 to 15
0.019 (0.049)
0.019 (0.049)
0.070 (<0.01)
0.005 (0.023)
0.075 (<0.01)
0.030 a (0.082)
0.030 a (0.082)
-0.028 a (0.07)
-0.028 a (0.07)
Youngest child’s age 16 to 18
(CONTINUED)
113
115
Table 14. (cont.) Estimation Results for the Individual-level Model (Standardized Effects)
Ln(Individual VMT) Number of Auto Tours Join Travel Trip Chaining
Direct Indirect Total Direct Indirect Total Direct Total Direct Total
Exogenous variables: Urban form at tour origins
Residential density -0.062
(0.015)
0.013 a
(0.51) -0.049
(0.038) 0.032 a
(0.051) 0.032 a
(0.051)
Retail density -0.060
(<0.01)
-0.014
(0.016)
-0.074
(<0.01) -0.057
(<0.01)
-0.057
(<0.01)
-0.038
(0.025)
-0.038
(0.025)
Non-retail density 0.022
(<0.01) 0.004
(<0.01)
0.004
(<0.01)
0.058
(<0.01)
0.058
(<0.01)
Intersection density -0.065
(0.018)
0.022
(<0.01)
-0.044 a
(0.067) -0.008
(<0.01)
-0.008
(<0.01) 0.067
(<0.01)
0.067
(<0.01)
Bus stop density -0.01
(0.019) -0.002
(0.019) -0.002
(0.019) -0.038
(0.019)
-0.038
(0.019)
Job-population index -0.066
(<0.01)
0.01
(<0.01)
-0.062
(<0.01) 0.003
(<0.01) 0.003
(<0.01) 0.051
(<0.01)
0.051
(<0.01)
Note: Coefficients in bold type are statistically significant at the 95% level; p-values are shown in parenthesis; superscript a means coefficients are statistically significant
at the 90% level
114
117
Chapter 7: Results and Discussions for Household-level Model
The household-level model uses households as the unit of analysis and examines
the connections among household characteristics, residential urban form factors, and
household travel behavior. The sample data used in this model are the 6,599 auto tours
that are traveled by 2,298 households. Figure 22 depicts the model framework at the
household level. The household travel behavior includes activity-travel patterns and
travel demand at the household level, which are considered as endogenous variables. The
activity-travel patterns, trip chaining and joint travel, are calculated as the number of
stops per tour and the ratio of joint travel per auto tour respectively for each household.
Two travel demand outcomes in the model are the number of auto tours and VMT
generated by households. The exogenous variables in the model are household attributes
and residential urban form factors. This chapter discusses the results of the household-
level model. Table 15 demonstrates the indices of model fit. Table 16 reports the analysis
results, including the direct, indirect, and total effects.
Figure 23 shows the paths with standardized direct effects between endogenous
and exogenous variables at the household level. All effects in the model are estimated
and only those with p-values less than 0.1 are kept and reported.
118
Figure 22. Model Framework at the Household Level
Chapter 7 starts with the discussion of model fit indices. The second section
presents the relationships among trip chaining, join travel, number of auto tours, and
VMT at the household level. The third and forth sections show the direct effects of
household attributes and urban form factors on the household travel behavior. Finally, the
indirect and total effects of exogenous variables are discussed.
7.1 Model Fit Indices for the Household-level Model
Table 15 shows several model fit indices of the home-level model and the
corresponding recommended cut-off values. The value of χ2 and degrees of freedom are
15.218 and 25 respectively. The p-value for the χ2 is 0.936, which is greater than 0.05.
This indicates that the model cannot be rejected. The ratio between χ2 and degrees of
freedom is 0.609, which is an indicator of a good fit (Schermelleh-Engel et al., 2003).
Other measures of fit, such as root-mean-square error of approximation (RMSEA =
0.001), comparative fit index (CFI=0.99), and standardized root mean square residual
(SRMR=0.006) are found to be in the acceptable range. The reported indices indicate that
119
the household-level model is a good-fitting model (Hu & Bentler, 1999; Mueller &
Hancock, 2010).
Table 15. Model Fit Indices of the Household-level Model
Model fit indices Value Recommended cut-off values
χ2 15.218
degree of freedom 25.000
χ2 / degree of freedom 0.609
χ2 P-value 0.936 >0.05
RMSEA 0.001 <0.08
CFI 0.995 >0.90
SRMR 0.006 <0.05
7.2 Links Between Activity-travel Patterns and Travel Demand in the Household-level
Model
Based on Table 16, most effects of trip chaining and joint travel on the number of
auto tours and VMT are statistically significant with expected signs. The positive direct
effects of trip chaining on the number of auto tours and VMT implies that as tour
complexity increases, the number of auto tours is expected to decrease and household
VMT is anticipated to increase. Similar to the findings of the individual-level model, trip
chaining has opposite direct effects on two travel demand outcomes at the household
level. The results may be explained by the fact that households with greater diversity of
activities tend to combine more trip segments to reduce the additional home-based tours.
The other activity-travel pattern, joint travel, also has opposite direct effects on the
number of auto tours and VMT. The joint travel is positively associated with the number
of auto tours and negatively related to VMT, showing that households with greater
propensity to travel jointly are more likely to make more tours but less likely to drive
further. One possible explanation is that households with greater travel demand may have
120
greater levels of inter-household coordination to achieve efficient time use. The results
are consistent with the findings in the tour-level model revealing that tours with high joint
travel ratio tend to be short and local tours. As anticipated, the number of auto tours has
positive direct effects on household VMT.
The indirect effects of trip chaining and joint travel on VMT are both statistically
significant. While trip chaining has negative indirect effects on VMT through the number
of auto tours, joint travel poses positive indirect effects on VMT. Trip chaining brings
about a decrease in VMT by decreasing the number of auto tours. On the other hand, joint
travel brings about an increase in VMT by increasing the number of auto tours. The total
effects of trip chaining and joint travel on household VMT are calculated by summing up
their direct effects and indirect effects. The structural analysis finds that the total effects
of trip chaining and joint travel are weakened by the corresponding direct effects. The
total effects of the number of auto tours on VMT are the same as its direct effects since
the absence of indirect effects. These mixed effects suggest planners the importance of
the potential impacts of activity-travel patterns on the resulting travel demand.
7.3 Direct Effects of Socio-demographic on Travel Behavior in the Household-level
Model
The socio-demographics in the model are the household attributes. Based on
Table 16, some household attributes have significant influence on trip chaining and joint
travel, while most household attributes significantly affect the number of auto tours and
VMT. Consistent with the findings of existing studies (Chung, Kim, Lee, & Choi, 2004;
Noland & Thomas, 2007; Wallace et al., 1999), larger households generate more auto
121
tours. As for the effects on activity-travel patterns, household size poses positive effects
on joint travel. These findings are consistent with previous studies, which indicate that
the tours from larger households are more likely to be joint tours (Chandrasekharan &
Goulias, 1999; Wallace et al., 2000). The possible explanation could be that larger
households may have greater intra-household interactions and greater varieties of
activities to participate in, which increase the joint travel propensity.
The model results show that household income is positively correlated with the
number of auto tours and VMT, implying that the wealthier families tend to make more
auto tours and drive further. The findings echo the existing literature which suggest that
the households with higher income levels may have higher propensities to participate in
out-of-home recreational and leisure activities, resulting in more auto tours and VMT
(Jang, 2005; Wallace et al., 1999). Consistent with previous studies (Liu & Shen, 2011;
Manoj & Verma, 2015; Stead, 1999), the number of household vehicles is positively
associated with VMT. The analysis shows that the number of household vehicles has
negative effects on joint travel. Chandrasekharan and Goulias (1999) point out that
households with fewer vehicles may be motivated to make joint trips. As expected, the
percentage of licensed drivers in the household has significant effects on the number of
auto tours and VMT: as the percentage of licensed drivers in the household increases, the
number of auto tours and VMT increases.
With respect to the effects of presence and age stages of children in the
household, the analysis finds that households with the youngest child between the ages of
11 to 15 are more likely to make simple tours and travel jointly with other household
members than those without children. The model results indicate strong associations
122
between the number of auto tours and the presence and age stages of children in the
household. All categorical variables of age stages are positively associated with auto tour
generation except for households with youngest child’s age up to five as compared to
those without children.
7.4 Direct Effects of Urban Form Factors on Travel Behavior in the Household-level
Model
The model results in Table 16 report that residential urban form factors have
significant effects on household travel behavior. The findings on residential density, retail
density, and non-retail density are consistent with the findings of previous two models
and studies (Frank et al., 2008; Liu, 2012; Van Acker & Witlox, 2010; Zhang et al.,
2012). Residential density and non-retail density have positive effects on trip chaining.
Households residing in high retail density neighborhoods tend to make simple tours and
generate fewer VMT. Some research indicates that high retail density neighborhoods
have better walkability. People may drive less and walk more if they reside in areas with
greater retail employment density (Boarnet & Crane, 2001; Cho & Rodríguez, 2015;
Zhang et al., 2012).
The intersection density and bus stop density have significant effects on the
household travel behavior. Increasing intersection density is found to decrease household
VMT. This finding implies that households living in areas with better street connectivity
are less likely to drive longer distances and echoes the findings of Hong et al. (2014). The
model results show that as bus stop density increases, the tendency of joint travel
decreases. As expected, job-population index is negatively correlated with household
123
VMT. As this index captures land-use mix and the balance of employment and resident
population, the analysis suggests that households residing in well-mixed neighborhoods
tend to drive less.
7.5 Indirect Effects of Exogenous Variables on Travel Behavior in the Household-level
Model
The indirect effects of household attributes and residential urban form factors on
the number of auto tours and VMT are provided in Table 16. The model results indicate
that the indirect effects of most household attributes on the number of auto tours and
VMT are statistically significant. In addition, while some household attributes have
significant indirect effects on both travel demand outcomes, others have indirect effects
on either the number of auto tours or VMT. For example, household size, household
income, the number of household vehicles, and households with the youngest child
between the ages of 11 to 15 have indirect effects on both the number of auto tours and
VMT. The findings are due to the fact that these variables have significant direct effects
on trip chaining and/or joint travel. In addition, trip chaining and joint travel have
significant direct effects on the number of auto tours and VMT. Conversely, the
percentage of licensed drivers in the household, households with the youngest child
between the ages of 6 to 10, and households with the youngest child between the ages of
16 to 18 affect either the number of auto tours or VMT indirectly.
The indirect effects of residential urban form variables are presented in Table 16.
According to the analysis outcomes, only residential density and retail density have
significant indirect effects on household VMT. Residential density has negative indirect
124
effects on tour generation through trip chaining. The negative indirect effects of retail
density on VMT are mediated via trip chaining.
7.6 Total Effects of Exogenous Variables on Travel Demand in the Household-level
Model
Table 16 reports that most exogenous variables have statistically significant total
effects on the number of auto tours and VMT. Similar to the trends in the tour-level
model and the individual-level model, the total effects of some exogenous variables in the
household-level model are either reinforced or weakened by the corresponding direct and
indirect effects. For example, the total effects of household income, the number of
household vehicles, and the percentage of licensed drivers in the household on VMT are
reinforced through their direct and indirect effects. On the contrary, the positive total
effects of household income on the number of auto tours are weakened by its negative
indirect effects. The total effects of some other household attributes are equal to either
direct or indirect effects. For instance, the total effects of household size on VMT are the
same as its indirect effects. The total effects of the percentage of licensed drivers in the
household on the number of auto tours are equal to its direct effects. The findings are due
to the lack of direct effects on household VMT or indirect effects on the number of auto
tours. The absolute magnitudes of the coefficients tell which of household attributes have
greater influence on the number of auto tours and VMT. The standardized coefficients
reveal that household size is the strongest variable affecting auto tour generation and
VMT.
125
As for the total effects of residential urban form factors, the residential density has
negative total effects on the number of auto tours, whereas retail density, intersection
density, and job-population index have negative total effects on VMT. Among residential
urban form factors, the absolute magnitudes of standardized total effects indicate that
residential density affects the number of auto tours the most. The retail density is the
strongest variable influencing household VMT in the model.
126
Table 16. Estimation Results for the Household-level Model (Standardized Effects)
Ln(Household VMT) Number of Auto Tours Join Travel Trip Chaining
Direct Indirect Total Direct Indirect Total Direct Total Direct Total
Endogenous variable
Number of Auto Tours 0.538
(<0.01)
0.538
(<0.01)
Joint travel -0.079
(<0.01)
0.031
(<0.01)
-0.048
(0.016)
0.058
(<0.01)
0.058
(<0.01)
Trip chaining 0.310
(<0.01)
-0.052
(<0.01)
0.257
(<0.01)
-0.097
(<0.01)
-0.097
(<0.01)
Exogenous variables: Household Socio-demographics
Household size
0.174
(<0.01)
0.174
(<0.01)
0.360
(<0.01)
0.013
(<0.01)
0.373
(<0.01)
0.139
(<0.01)
0.139
(<0.01)
-0.051a
(0.058)
-0.051a
(0.058)
Household income 0.051
(0.013)
0.067
(<0.01)
0.118
(<0.01)
0.108
(<0.01)
-0.01
(<0.01)
0.098
(<0.01)
-0.176
(<0.01)
-0.176
(<0.01)
Household vehicle 0.105
(<0.01)
0.003
(0.018)
0.108
(<0.01)
-0.003
(<0.01)
-0.003
(<0.01)
-0.053
(<0.01)
-0.053
(<0.01)
Driver percentage 0.038
(0.032)
0.085
(<0.01)
0.123
(<0.01)
0.183
(<0.01)
0.188
(<0.01)
No children presence (base case)
Youngest child’s age up to 5
Youngest child’s age 6 to 10
0.031
(0.019)
0.031
(0.019)
0.057
(0.019)
0.057
(0.0190
Youngest child’s age 11 to 15
0.088
(<0.01)
0.088
(<0.01)
0.186
(<0.01)
0.007
(<0.01)
0.193
(<0.01)
0.052
(<0.01)
0.052
(<0.01)
-0.038
(0.041)
-0.038
(0.041)
Youngest child’s age 16 to 18
0.044
(<0.01)
0.044
(<0.01)
0.081
(<0.01)
0.081
(<0.01)
(CONTINUED)
125
127
Table 16. (cont.) Estimation Results for the Household-level Model (Standardized Effects)
Ln(Household VMT) Number of Auto Tours Join Travel Trip Chaining
Direct Indirect Total Direct Indirect Total Direct Total Direct Total
Residential density
0.020
(0.018)
-0.008
(0.017)
-0.008
(0.017)
0.079
(<0.01)
0.079
(<0.01)
Retail density -0.070
(<0.01)
-0.016
(<0.01)
-0.086
(<0.01)
-0.050 a
(0.071)
-0.054 a
(0.071) -0.072
(0.019)
-0.072
(0.019)
Non-retail density
0.062 a
(0.064)
0.058 a
(0.064) 0.061
(0.033)
0.061
(0.033)
Intersection density -0.079
(0.022)
-0.079
(0.022)
Bus stop density
0.002 a
(0.089)
0.002 a
(0.089)
-0.003 a
(0.073)
-0.003 a
(0.073)
-0.049 a
(0.074)
-0.049 a
(0.074)
Job-population index -0.067
(<0.01)
-0.002 a
(0.073) -0.069
(<0.01)
0.002 a
(0.082)
0.002 a
(0.082)
0.042 a
(0.064)
0.042 a
(0.064)
Note: Coefficients in bold type are statistically significant at the 95% level; p-values are shown in parenthesis; superscript a means coefficients are statistically significant
at the 90% level
126
129
Chapter 8: Conclusions
The studies of travel behavior have received much research attention due to the
increasing VMT in recent years, which is one of the key contributors to negative
externalities associated with the transportation system. In light of these concerns, many
planners and decision-makers focus on transportation demand management (TDM)
policies designed to improve the transportation system efficiency and mitigate the
negative externalities. These policies include encouraging trip chaining, reducing single-
occupant vehicle (SOV) trips, and promoting carpooling programs. The making of TDM
strategies therefore greatly depends on the understanding of different dimensions of
travel behavior including activity-travel patterns and travel demand. Using the 2012
Travel Survey Data collected from the Cleveland Metropolitan Area, Ohio, this study
uses activity-travel patterns (trip chaining and joint travel) as mediating variables of
travel demand (tour generation and travel distances) and analyzes how socio-
demographics and urban form factors affect different aspects of travel behavior
simultaneously. The Structural Equation Modeling (SEM) approach is applied in this
dissertation to disentangle the complex links among activity-travel patterns, travel
demand, socio-demographics, and urban form factors. Understanding these links would
help planners and policy makers better understand individuals’ travel behavior and
propose strategies to mitigate the negative externalities related to travel.
130
This chapter concludes this dissertation as follows. The first section summarizes
the major findings of this research and discusses its contribution. The policy implications
are presented in the second section. The last section discusses the research limitations and
future research directions.
8.1 Summary and Contributions
One of the fundamental principles of the activity-based approaches is that
individuals’ travel demand is derived from participating in out-of-home activities
distributed in different places (Bhat & Koppelman, 2003) . Based on this principle, this
study assumes that activity-travel patterns (trip chaining and joint travel) affect travel
demand (tour generation and travel distances), and socio-demographics (personal
characteristics and household attributes) and urban form factor influence these four
aspects of travel behavior at the same time. The research framework in this study
considers activity-travel patterns and travel demand as exogenous variables and regards
socio-demographics and urban form factors as exogenous variables. Three models at
different levels are proposed. These models follow a similar research framework. This
study starts with the model at the home-based tour level, which use the home-based tour
as a basic unit in the analysis. The endogenous variables in the tour-level model are trip
chaining, joint travel, travel distances at the tour level. The exogenous variables in the
tour-level are personal characteristics, household attributes, and urban form factors at
tour origins and destinations. The second model is at the individual level, which
considers an individual as the basic unit. The endogenous variables in the individual-level
131
model include individual trip chaining per auto tour, individual joint travel per auto tour,
the number of auto tours traveled by the individual, and individual VMT. The exogenous
variables are personal characteristics, household attributes, and residential urban form
factors. The third model is at the household level, which uses a household as the basic
unit of analysis. The endogenous variables in the household-level model are household
trip chaining per auto tour, household joint travel per auto tour, the number of auto tours
traveled by the household, and household VMT. The exogenous variables are household
attributes and residential urban form factors.
8.1.1 Relationships among Activity-travel Patterns and Travel Demand
This study examines the links among two activity-travel patterns (trip chaining
and joint travel) and two travel demand outcomes (tour generation and travel distances).
Trip chaining is captured as the number of stops within a home-based tour, which depicts
the tour complexity. Joint travel is measured by the ratio of joint trip segments to total
trip segments for each tour, which illustrates the level of travel coordination among
household members. Two travel demand outcomes, tour generation, and travel distances,
are included in this study. The number of auto tours for each individual and household
captures tour generation in the individual-level model and household-level model,
respectively. The tour distances, individual VMT, and household VMT measure the
travel distances in this study.
The analysis results show the existence of significant associations among activity-
travel patterns and the resulting travel demand. With respect to the direct effects, the tour-
132
level model results find that trip chaining is negatively related to joint travel, which
implies that people are less likely to cooperate their tours with other household
companions when the tours become more complex. The analysis indicates that the effects
of trip chaining and joint travel on the resulting travel demand are in opposite directions.
Trip chaining is positively related to travel distances, while joint travel is negatively
associated with travel distances at the tour level. The individual-level model results show
that trip chaining has negative effects on the number of auto tours but positive effects on
VMT. Similarly, the household-level analysis outcomes suggest that trip chaining is
negatively correlated with the number of auto tours but is positively correlated with the
number of auto tours. On the other hand, joint travel is positively related to the number of
auto tours but is negatively related to VMT.
The model results show that trip chaining and joint travel have adverse indirect
effects on travel distances. Trip chaining has negative indirect effects on travel distances
at the tour, individual, and household levels. Joint travel has positive indirect effects on
travel distances at the individual and household levels. Therefore, the analysis reveals
that the total effects of trip chaining and joint travel on household VMT are weakened.
These findings provide some detailed insights regarding how activity-travel patterns
contribute to the resulting travel demand, which improves the overall quality of travel
demand modeling.
133
8.1.2 Relationships Between Socio-demographics and Travel behavior
The influence of travelers’ personal and household socio-demographics on travel
behavior is analyzed in this study. Five personal characteristics included in the models
are gender, age, driver’s ownership, employment status, and transit pass ownership.
Household size, household income, the number of household vehicles, the percentage of
licensed drivers in the household, and the presence and age stages of children in the
household capture household attributes in the analysis.
The results suggest that the direct effects of most socio-demographics on travel
behavior are statistically significant and consistent with many existing studies. For
example, females tend to travel shorter distances and have less VMT. Driver’s license
holders tend to make simple tours and have greater travel distances but they are less
likely to travel with other family members. Similarly, workers generate more VMT but
have less joint travel ratio. The transit pass ownership is positively related to trip chaining
but negatively associated with joint travel and number of auto tours. With regard to the
effects of household attributes, people from larger households tend to make simple tours
and travel jointly. Household income and the number of household vehicles are viewed as
resources as they have negative effects on joint travel but have positive effects on travel
distances. The percentage of licensed drivers in the household inversely affects the
number of auto tours at the individual level and the household level. The increasing
percentage of licensed drivers in the household decreases individual auto tour generation
but increases the household auto tour generation.
134
The results show that the indirect effects of most socio-demographic variables are
statistically significant. The signs of indirect and direct effects of some socio-
demographics may be in opposite directions because of the different mediating effects.
For instance, the direct and indirect effects of driver’s license ownership on individual
VMT are both positive. On the other hand, the indirect effects of employment status on
individual VMT are negative, whereas the corresponding direct effects are positive.
Therefore, the total effects of socio-demographics on travel demand are reinforced or
weakened due to the corresponding indirect and direct effects in the models. These mixed
effects of socio-demographics in different models help planners and decision makers
better evaluate the market response to the transportation policies and forecast the travel
demand at different scales.
8.1.3 Relationships Between Urban Form Factors and Travel behavior
Based on the research frameworks of “3Ds” (density, design, and diversity) and
“three built environment components” (urban design, land use, and transportation system)
from previous literature (Cervero & Kockelman, 1997; S. L. Handy et al., 2002), this
study includes six urban form factors to describe built environment characteristics:
residential density, retail density, non-retail density, intersection density, bus stop
density, and job-population index. The tour-level model estimates the effects of urban
form factors at tour origins and destinations. The primary activity location of a home-
based tour is considered as the tour destination. The overall results show that the direct
effects of the residential urban form are consistent across the models in this study and
135
with the findings in previous studies. People living in compact neighborhoods with high
retail density tend to generate simple tours and fewer VMT. The non-retail density at
residences is positively associated with trip chaining and joint travel. Individuals living in
areas of high intersection density tend to drive less. The job-population index at home
locations negatively affects travel distances in the tour-level model, individual-level
model, and household-level model. The model results show that the urban form factors at
tour destinations affect travel behavior significantly. While some urban form factors
affect travel behavior in the same direction, some affect travel behavior in opposite
directions. For example, residential density and job-population index at tour origins and
destinations have negative effects on travel distances at tour origins and destinations. On
the other hand, retail density at tour origins has positive effects on joint travel, but at tour
destinations it has negative effects on joint travel.
The model results indicate that while some urban form factors may not affect
travel demand outcomes directly, they may have indirect effects through trip chaining
and/or joint travel. For instance, non-retail density at residences has negative indirect
effects on travel distances via both activity-travel patterns in the tour-level model.
Consistent with the findings on socio-demographics, the total effects of urban form
factors are reinforced or weakened due to the corresponding indirect and direct effects.
The total effects of residential retail density on VMT are reinforced in the individual-
level model and the household-level model. The links between urban form factors and
travel behavior can be used to assess the potential success of transportation projects in
neighborhoods with different built environment characteristics.
136
8.2 Policy Implications
Several researchers have found that trip chaining and joint travel represent a
substantial percentage of household travel patterns. For example, S. G. Bricka (2008)
indicates that about 60% of household trip segments are chained into household tours in
the United States. Vovsha et al. (2003) find that roughly half of mid-Ohio tours and more
than one-third of New York tours are joint tours. Lin and Wang (2014) show that 32% of
trips are traveled jointly with family and relatives in Hong Kong. Ho and Mulley (2015)
demonstrate about half of home-based tours are joint tours in Sydney. According to this
study, around 60% of home-based tours are complex tours and more than 30% of the
tours have at least one joint trip segment in the Cleveland Metropolitan Area. The
number of VMT, which has been regarded as the primary negative outcome of travel
behavior, has increased almost 150% in metropolitan areas from 1980 to 2014 (Frey,
2012; United States Census Bureau, 2016; United States Department of Transportation;
Bureau of Transportation Statistics, 2016). Therefore, a better understanding of different
aspects of travel behavior is necessary for the development of transportation demand
management (TDM) strategies. The TDM strategies are a set of policies aiming to
decrease VMT, increase the efficiency of the urban transportation system, and reduce
auto dependency while providing reliable transportation choices to meet growing
demand.
By investigating the links among trip chaining, joint travel, tour generation, and
travel distances while controlling for socio-demographics and urban form factors, this
137
research is directly relevant to planners attempting to improve the existing TDM policies
and suggests several policy implications. First, the findings shed light on the importance
of integrating the activity-travel patterns in travel demand modeling and examining
different aspects of travel behavior simultaneously. The model results improve the
understanding of the links among activity-travel patterns and travel demand, which
suggest that transportation policies may need to take into account individuals’ trip
chaining and joint travel while proposing travel demand forecasts. The travel demand
models without considering trip chaining and joint travel may lead to biased estimates of
travel demand.
Second, the analysis results confirm that socio-demographics are important
explanatory variables of travel behavior. The analysis presents that most personal and
household attributes have significant direct effects on activity-travel patterns and indirect
effects on the travel demand. The outcomes of socio-demographics help policy makers
identify the travel behavior of population segments and improve the effectiveness of the
transportation policies. For example, the model results show that household size has
strongest positive effects on joint travel among the household attributes. Therefore,
marketing the non-SOV options to individuals from larger households may be more
effective. This study reveals that most total effects of socio-demographics on the
resulting travel demand are weakened or reinforced by the intermediating effects. Hence,
the detailed information can improve the overall accuracy of travel demand modeling and
the evaluation of the market responses when implementing transportation policies on
targeted population segments.
138
Finally, the model results reveal that the existence of strong associations between
travel behavior and urban form factors not only at tour origins but also at tour
destinations. According to the standardized effects of urban form factors in the tour-level
model, this study indicates that the urban form factors at tour destinations also play
critical roles in explaining travel behavior. Most urban form factors at tour destinations
have greater effects on travel distances than those at tour origins (e.g., residential density,
non-retail density, bus stop density, and job-population index). Some urban form factors
at tour origins and destinations may have opposite effects on travel behavior. For
example, bus stop density at tour origins has negative effects on travel distances but bus
stop density at tour destinations has positive effects on travel distances. Retail density has
a similar trend: it is negatively related to joint travel at tour origins but is positively
associated with joint travel at tour destinations. These findings indicate that transportation
and land-use policies intended to manage travel behavior by changing urban form should
focus on the effects at destinations of home-based tours. In addition, some urban form
factors only have indirect effects on the resulting travel demand channeled via the
mediating activity-travel patterns. The insights gained from the models can be
implemented as inputs for assessing various transportation projects. The findings call for
planners’ attention regarding the policymaking and assessment of transportation and
land-use investments. These complex relationships between travel behavior and urban
form factors suggest that any single transportation project or land-use regulation cannot
offer a complete checklist of multiple TDM policies and achieve the goal of VMT
reduction. Instead, decision makers should be aware that a mix of different technologies,
139
policies, and strategies are necessary (Rubin & Noland, 2010). Mixed policies involving
land-use plans and transportation strategies will help to reduce auto travel demand, to
improve the overall efficiency of transportation systems, and to promote transportation
sustainability.
8.3 Limitations and Future Directions
Although this study provides an in-depth understanding of the relationships
among activity-travel patterns, travel demand, socio-demographics, and urban form, this
empirical effort is subject to several data and methodological limitations. These
limitations suggest directions for future research.
The first limitation is related to residential self-selection: people tend to choose
residential locations based on their travel needs and abilities (Mokhtarian & Cao, 2008).
The effect of the urban form factors on travel behavior would be biased without
accounting for residential self-selection. Previous studies suggest that including
attitudinal variables in the models is one of the solutions to avoid self-selection bias
(Brownstone, 2008; Cao et al., 2009; Mokhtarian & Cao, 2008). However, since this
study uses cross-sectional household travel survey data without any individual attitudinal
information, the ability to estimate the impacts of self-selection at tour origins and
destinations is limited.
The second limitation is the accuracy of the urban form effects. Because of the
restricted availability of specific longitudes and latitudes of all locations, this study
140
assumes that homes and primary activities are located at the centroid of TAZs. The urban
form factors used in this study are measured at the TAZ level. If all the specific
addresses of residence and activities are known, this study can improve the accuracy of
estimating the urban form effects on activity-travel patterns and travel demand.
The third limitation is the lack of data regarding the activity-travel patterns. For
instance, the household travel survey data do not include the intended tour destinations
and the alternative travel plans for each home-based tour. The individuals’ intended tour
destinations may better explain the decision-making mechanism of trip chaining. The
alternative travel plans may provide specific evidence on the travel cost saved by
consolidating multiple trip segments. Furthermore, the data used in this study do not
include information on travel dependence and social networks of each individual and
household. The travel dependence of one individual on other household members and
social network attributes may affect joint travel significantly. As these aspects are not
covered by the travel survey data, the effects on activity-travel patterns are not addressed
in this study.
Based on the aforementioned research limitations, this study can be extended
along different directions. First, pending data availability, a natural follow up would be
the investigating the influence of individuals’ activity preference, alternative travel plans,
travel dependence, and social interactions on trip chaining, joint travel. The integration of
these variables may lead to better measures of the relationships among travel behavior,
socio-demographics, and urban form.
141
The second extension is to apply the research frameworks for other transportation
modes. This study shows that the Cleveland Metropolitan Area is an auto-oriented
environment with high driver’s license ownership, high proportion of auto tours, and low
public transit and non-motorized trips/tours. Pending data availability, the future research
can examine the relationships among travel behavior, socio-demographics, and urban
form factors for transit and non-motorized modes.
Finally, this study uses the urban form factors separately in models to explain
travel behavior. However, some previous literature points out that single measures of the
built environment do not represent overall neighborhood characteristics, which may
result in ambiguous influence on travel behavior and inconsistent results (Ewing &
Cervero, 2001a; Namgung, 2014). Studies applying cluster analysis and creating distinct
neighborhood typologies based on combining individual built environment variables may
produce consistent findings (Akar et al., 2016; Chen & Akar, 2016; Clifton, Currans,
Cutter, & Schneider, 2012; Manaugh, Miranda-Moreno, & El-Geneidy, 2010; Namgung
& Akar, 2015). Therefore, future studies can be extended by creating new neighborhood
clusters based on built environment features and examining their effects on different
aspects of travel behavior while controlling for socio-demographics.
142
Reference
Adler, T., & Ben-Akiva, M. (1979). A theoretical and empirical model of trip chaining
behavior. Transportation Research Part B: Methodological, 13(3), 243-257.
Akar, G., Chen, N., & Gordon, S. I. (2016). Influence of neighborhood types on trip
distances: Spatial error models for Central Ohio. International Journal of
Sustainable Transportation, 10(3), 284-293. doi:10.1080/15568318.2014.903447
Antipova, A., & Wang, F. (2010). Land use impacts on trip-chaining propensity for
workers and nonworkers in Baton Rouge, Louisiana. Annals of GIS, 16(3), 141-
154.
Bento, A. M., Cropper, M. L., Mobarak, A. M., & Vinha, K. (2005). The effects of urban
spatial structure on travel demand in the United States. Review of Economics and
Statistics, 87(3), 466-478.
Bhat, C. R., & Koppelman, F. S. (2003). Activity-Based Modeling of Travel Demand. In
R. W. Hall (Ed.), Handbook of Transportation Science (pp. 39-65). Boston, MA:
Springer US.
Boarnet, M., & Crane, R. (2001). The influence of land use on travel behavior:
specification and estimation strategies. Transportation Research Part A: policy
and practice, 35(9), 823-845.
Bowen, N. K., & Guo, S. (2011). Structural equation modeling: Oxford University Press.
Bricka, S., Moran, M., Miller, K., & Hudson, J. (2015). The Future of TDM: Technology
and Demographic Shifts and Their Implications for Transportation Demand
Management. Retrieved from
http://static.tti.tamu.edu/tti.tamu.edu/documents/PRC-15-25F.pdf
Bricka, S. G. (2008). Trip chaining: linking the influences and implications. (PhD
Dissertation), University of Texas at Austin.
Browne, M. W., Cudeck, R., & Bollen, K. A. (1993). Alternative ways of assessing
model fit. Sage Focus Editions, 154, 136-136.
Brownstone, D. (2008). Key relationships between the built environment and VMT.
Draft paper prepared for the Transportation Research Board panel on
143
Relationships Among Development Patterns, Vehicle Miles Traveled, and Energy,
October.
Brownstone, D., & Golob, T. F. (2009). The impact of residential density on vehicle
usage and energy consumption. Journal of Urban Economics, 65(1), 91-98.
Byrne, B. M. (2013). Structural equation modeling with AMOS: Basic concepts,
applications, and programming. New York: Routledge.
Cao, X. (2007). The Causal Relationship between the Built Environment and Personal
Travel Choice: Evidence from Northern California. (PhD Dissertation),
University of California Davis. Retrieved from
http://www.escholarship.org/uc/item/1n90z8h8
Cao, X., Mokhtarian, P. L., & Handy, S. L. (2009). The relationship between the built
environment and nonwork travel: A case study of Northern California.
Transportation Research Part A: policy and practice, 43(5), 548-559.
doi:http://dx.doi.org/10.1016/j.tra.2009.02.001
Cervero, R. (2002). Built environments and mode choice: toward a normative framework.
Transportation Research Part D: Transport and Environment, 7(4), 265-284.
doi:http://dx.doi.org/10.1016/S1361-9209(01)00024-4
Cervero, R., & Gorham, R. (1995). Commuting in Transit Versus Automobile
Neighborhoods. Journal of the American Planning Association, 61(2), 210-225.
doi:10.1080/01944369508975634
Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: density, diversity,
and design. Transportation Research Part D: Transport and Environment, 2(3),
199-219.
Cervero, R., & Murakami, J. (2010). Effects of built environments on vehicle miles
traveled: evidence from 370 US urbanized areas. Environment and Planning A,
42(2), 400-418.
Chandrasekharan, B., & Goulias, K. G. (1999). Exploratory longitudinal analysis of solo
and joint trip making using the Puget Sound transportation panel. Transportation
Research Record: Journal of the Transportation Research Board, 1676, 77-85.
Chatman, D. G. (2009). Residential Choice, the Built Environment, and Nonwork Travel:
Evidence Using New Data and Methods. Environment and Planning A, 41(5),
1072-1089. doi:10.1068/a4114
Chen, N., & Akar, G. (2016). Effects of neighborhood types & socio-demographics on
activity space. Journal of Transport Geography, 54, 112-121.
144
Cheung, G. W., & Lau, R. S. (2008). Testing Mediation and Suppression Effects of
Latent Variables: Bootstrapping With Structural Equation Models. Organizational
Research Methods, 11(2), 296-325. doi:10.1177/1094428107300343
Cho, G.-H., & Rodríguez, D. A. (2015). Neighborhood Design, Neighborhood Location,
and Three Types of Walking: Results from the Washington DC Area.
Environment and Planning B: Planning and Design, 42(3), 526-540.
doi:10.1068/b130222p
Chung, J.-H., Kim, S., Lee, Y.-K., & Choi, Y.-S. (2004). Multilevel structural equation
model for activity participation and travel behavior: Data from the Puget sound
transportation panel. Transportation Research Record: Journal of the
Transportation Research Board, 1898(1), 52-60.
Clifton, K. J., Currans, K. M., Cutter, A. C., & Schneider, R. (2012). Household Travel
Surveys in Context-Based Approach for Adjusting ITE Trip Generation Rates in
Urban Contexts. Transportation Research Record: Journal of the Transportation
Research Board, 2307(1), 108-119.
Dillon, H. S., Saphores, J.-D., & Boarnet, M. G. (2015). The impact of urban form and
gasoline prices on vehicle usage: Evidence from the 2009 National Household
Travel Survey. Research in Transportation Economics, 52, 23-33.
Ding, Y., & Lu, H. (2016). Activity participation as a mediating variable to analyze the
effect of land use on travel behavior: A structural equation modeling approach.
Journal of Transport Geography, 52, 23-28.
doi:http://dx.doi.org/10.1016/j.jtrangeo.2016.02.009
Duncan, M. (2015). How much can trip chaining reduce VMT? A simplified method.
Transportation, 1-17. doi:10.1007/s11116-015-9610-5
Elldér, E. (2014). Residential location and daily travel distances: the influence of trip
purpose. Journal of Transport Geography, 34, 121-130.
Etminani-Ghasrodashti, R., & Ardeshiri, M. (2016). The impacts of built environment on
home-based work and non-work trips: An empirical study from Iran.
Transportation Research Part A: policy and practice, 85, 196-207.
doi:http://dx.doi.org/10.1016/j.tra.2016.01.013
Ewing, R., & Cervero, R. (2001a). Travel and the built environment: a synthesis.
Transportation Research Record: Journal of the Transportation Research Board,
1780(1), 87-114.
Ewing, R., & Cervero, R. (2001b). Travel and the Built Environment: A Synthesis.
Transportation Research Record: Journal of the Transportation Research Board,
1780, 87-114. doi:doi:10.3141/1780-10
145
Ewing, R., & Cervero, R. (2010). Travel and the Built Environment. Journal of the
American Planning Association, 76(3), 265-294.
doi:10.1080/01944361003766766
Ewing, R., Greenwald, M., Zhang, M., Walters, J., Feldman, M., Cervero, R., . . .
Thomas, J. (2011). Traffic generated by mixed-use developments—six-region
study using consistent built environmental measures. Journal of Urban Planning
and Development, 137(3), 248-261.
Feng, J., Dijst, M., Wissink, B., & Prillwitz, J. (2017). Changing travel behaviour in
urban China: Evidence from Nanjing 2008–2011. Transport Policy, 53, 1-10.
doi:http://dx.doi.org/10.1016/j.tranpol.2016.08.011
Frank, L., Bradley, M., Kavage, S., Chapman, J., & Lawton, T. K. (2008). Urban form,
travel time, and cost relationships with tour complexity and mode choice.
Transportation, 35(1), 37-54.
Frey, W. H. (2012). Population Growth in Metro America since 1980.
Gärling, T., Eek, D., Loukopoulos, P., Fujii, S., Johansson-Stenman, O., Kitamura,
R., . . . Vilhelmson, B. (2002). A conceptual analysis of the impact of travel
demand management on private car use. Transport Policy, 9(1), 59-70.
Golob, T. F. (1986). A nonlinear canonical correlation analysis of weekly trip chaining
behaviour. Transportation Research Part A: General, 20(5), 385-399.
Golob, T. F. (2000). A simultaneous model of household activity participation and trip
chain generation. Transportation Research Part B: Methodological, 34(5), 355-
376.
Golob, T. F. (2003). Structural equation modeling for travel behavior research.
Transportation Research Part B: Methodological, 37(1), 1-25.
Gopisetty, V., & Srinivasan, K. K. (2013). Joint models for analysis of household trip
frequency and vehicle ownership in Chennai city. International Journal of
Advances in Engineering Sciences and Applied Mathematics, 5(2-3), 129-144.
Goulias, K. G., & Kitamura, R. (1991). Recursive model system for trip generation and
trip chaining. UC Berkeley: University of California Transportation Center.
Goulias, K. G., Pendyala, R. M., & Kitamura, R. (1990). Practical method for the
estimation of trip generation and trip chaining. UC Berkeley: University of
California Transportation Center.
Handy, S., Cao, X., & Mokhtarian, P. (2005). Correlation or causality between the built
environment and travel behavior? Evidence from Northern California.
146
Transportation Research Part D: Transport and Environment, 10(6), 427-444.
doi:http://dx.doi.org/10.1016/j.trd.2005.05.002
Handy, S. L., Boarnet, M. G., Ewing, R., & Killingsworth, R. E. (2002). How the built
environment affects physical activity: views from urban planning. American
journal of preventive medicine, 23(2), 64-73.
Heres-Del-Valle, D., & Niemeier, D. (2011). CO2 emissions: Are land-use changes
enough for California to reduce VMT? Specification of a two-part model with
instrumental variables. Transportation Research Part B: Methodological, 45(1),
150-161. doi:http://dx.doi.org/10.1016/j.trb.2010.04.001
Ho, C., & Mulley, C. (2013a). Incorporating Intrahousehold Interactions into a Tour-
Based Model of Public Transport Use in Car-Negotiating Households.
Transportation Research Record: Journal of the Transportation Research
Board(2343), 1-9.
Ho, C., & Mulley, C. (2013b). Tour-based mode choice of joint household travel patterns
on weekend and weekday. Transportation, 40(4), 789-811.
Ho, C., & Mulley, C. (2015). Intra-household Interactions in tour-based mode choice:
The role of social, temporal, spatial and resource constraints. Transport Policy,
38, 52-63. doi:http://dx.doi.org/10.1016/j.tranpol.2014.12.001
Holzapfel, H. (1986). Trip Relationships in Urban Areas. Aldershot, U.K.: Gower
publication.
Hong, J., Shen, Q., & Zhang, L. (2014). How do built-environment factors affect travel
behavior? A spatial analysis at different geographic scales. Transportation, 41(3),
419-440.
Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural equation
modeling: a multidisciplinary journal, 6(1), 1-55.
Jahanshahi, K., Jin, Y., & Williams, I. (2015). Direct and indirect influences on employed
adults’ travel in the UK: New insights from the National Travel Survey data
2002–2010. Transportation Research Part A: policy and practice, 80, 288-306.
doi:http://dx.doi.org/10.1016/j.tra.2015.08.007
Jang, T. Y. (2005). Count data models for trip generation. Journal of transportation
engineering, 131(6), 444-450.
Jiang, Y., Gu, P., Chen, Y., He, D., & Mao, Q. (2016). Influence of land use and street
characteristics on car ownership and use: Evidence from Jinan, China.
147
Transportation Research Part D: Transport and Environment.
doi:http://dx.doi.org/10.1016/j.trd.2016.08.030
Kitamura, R. (1988). An evaluation of activity-based travel analysis. Transportation,
15(1-2), 9-34.
Kitamura, R., Mokhtarian, P. L., & Laidet, L. (1997). A micro-analysis of land use and
travel in five neighborhoods in the San Francisco Bay Area. Transportation,
24(2), 125-158.
Kitamura, R., & Susilo, Y. O. (2005). Is travel demand insatiable? A study of changes in
structural relationships underlying travel. Transportmetrica, 1(1), 23-45.
doi:10.1080/18128600508685640
Kline, R. B. (2015). Principles and practice of structural equation modeling: Guilford
publications.
Krizek, K. J. (2003). Residential relocation and changes in urban travel: does
neighborhood-scale urban form matter? Journal of the American Planning
Association, 69(3), 265-281.
Kruskal, W. H., & Wallis, W. A. (1952). Use of Ranks in One-Criterion Variance
Analysis. Journal of the American Statistical Association, 47(260), 583-621.
doi:10.2307/2280779
Kuppam, A. R., & Pendyala, R. M. (2001). A structural equations analysis of commuters'
activity and travel patterns. Transportation, 28(1), 33-54.
Kuzmyak, R. (2009). Estimates of point elasticities'. Phoenix, AZ: Maricopa Association
of Governments.
Lin, T., & Wang, D. (2014). Social networks and joint/solo activity–travel behavior.
Transportation Research Part A: policy and practice, 68, 18-31.
Liu, C. (2012). Exploring the influence of urban form on travel and energy consumption,
using structural equation modeling. (PhD Dissertation), Univeristy of Maryland,
College Park.
Liu, C., & Shen, Q. (2011). An empirical analysis of the influence of urban form on
household travel and energy consumption. Computers, Environment and Urban
Systems, 35(5), 347-357.
Loukopoulos, P., Jakobsson, C., Gärling, T., Schneider, C. M., & Fujii, S. (2004). Car-
user responses to travel demand management measures: goal setting and choice of
adaptation alternatives. Transportation Research Part D: Transport and
Environment, 9(4), 263-280.
148
Lu, X., & Pas, E. I. (1999). Socio-demographics, activity participation and travel
behavior. Transportation Research Part A: policy and practice, 33(1), 1-18.
Ma, J., Mitchell, G., & Heppenstall, A. (2014). Daily travel behaviour in Beijing, China:
An analysis of workers' trip chains, and the role of socio-demographics and urban
form. Habitat International, 43, 263-273.
doi:http://dx.doi.org/10.1016/j.habitatint.2014.04.008
Manaugh, K., Miranda-Moreno, L. F., & El-Geneidy, A. M. (2010). The effect of
neighbourhood characteristics, accessibility, home–work location, and
demographics on commuting distances. Transportation, 37(4), 627-646.
Manoj, M., & Verma, A. (2015). A structural equation model based analysis of non-
workers’ activity-travel behaviour from a city of a developing country.
Transportation, 1-29.
McDonald, N. C. (2005). Does residential density affect the travel “gender gap”. Paper
presented at the Conference Proceedings 35: Research on Women’s Issues in
Transportation.
McGuckin, N., & Murakami, E. (1999). Examining trip-chaining behavior: Comparison
of travel by men and women. Transportation Research Record: Journal of the
Transportation Research Board, 1693(1), 79-85.
McGuckin, N., Zmud, J., & Nakamoto, Y. (2005). Trip-chaining trends in the United
States: understanding travel behavior for policy making. Transportation Research
Record: Journal of the Transportation Research Board, 1917(1), 199-204.
McNally, M., & Kulkarni, A. (1997). Assessment of Influence of Land Use-
Transportation System on Travel Behavior. Transportation Research Record:
Journal of the Transportation Research Board, 1607, 105-115. doi:10.3141/1607-
15
McNally, M. G., & Rindt, C. R. (2007). The activity-based approach Handbook of
Transport Modelling: 2nd Edition (pp. 55-73): Emerald Group Publishing
Limited.
Mercado, R., & Páez, A. (2009). Determinants of distance traveled with a focus on the
elderly: a multilevel analysis in the Hamilton CMA, Canada. Journal of Transport
Geography, 17(1), 65-76.
Mokhtarian, P. L., & Cao, X. (2008). Examining the impacts of residential self-selection
on travel behavior: A focus on methodologies. Transportation Research Part B:
Methodological, 42(3), 204-228. doi:http://dx.doi.org/10.1016/j.trb.2007.07.006
149
Morency, C., Paez, A., Roorda, M. J., Mercado, R., & Farber, S. (2011). Distance
traveled in three Canadian cities: Spatial analysis from the perspective of
vulnerable population segments. Journal of Transport Geography, 19(1), 39-50.
Mueller, R., & Hancock, G. (2010). Structural equation modeling. The reviewer’s guide
to quantitative methods in the social sciences, 371-383.
Namgung, M. (2014). The Relationship between Attitudes, Neighborhood Types, and
Travel Behavior: Implications for Public Transportation. (PhD Dissertation), The
Ohio State University. Retrieved from
http://rave.ohiolink.edu/etdc/view?acc_num=osu1417699048
Namgung, M., & Akar, G. (2015). Influences of Neighborhood Characteristics and
Personal Attitudes on University Commuters’ Public Transit Use. Transportation
Research Record: Journal of the Transportation Research Board(2500), 93-101.
Noland, R. B., & Thomas, J. V. (2007). Multivariate analysis of trip-chaining behavior.
Environment and Planning B: Planning and Design, 34(6), 953.
Obermann, W. R. (2012). Travel Demand Management: A Toolbox of Strategies to
Reduce Single-occupant Vehicle Trips and Increase Alternate Mode Usage in
Arizona (No. FHWA-AZ-12-654). Retrieved from
https://ntl.bts.gov/lib/44000/44300/44358/AZ654.pdf
Pendyala, R. M., Yamamoto, T., & Kitamura, R. (2002). On the formulation of time-
space prisms to model constraints on personal activity-travel engagement.
Transportation, 29(1), 73-94. doi:10.1023/a:1012905110686
Primerano, F., Taylor, M. A., Pitaksringkarn, L., & Tisato, P. (2008). Defining and
understanding trip chaining behaviour. Transportation, 35(1), 55-72.
Rentziou, A., Gkritza, K., & Souleyrette, R. R. (2012). VMT, energy consumption, and
GHG emissions forecasting for passenger transportation. Transportation Research
Part A: policy and practice, 46(3), 487-500.
Rubin, J. D., & Noland, R. B. (2010). Transportation and Climate Change: Developing
Technologies, Policies, and Strategies. TR News(268).
Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of
structural equation models: Tests of significance and descriptive goodness-of-fit
measures. Methods of psychological research online, 8(2), 23-74.
Schmöcker, J.-D., Quddus, M. A., Noland, R. B., & Bell, M. G. (2005). Estimating trip
generation of elderly and disabled people: analysis of London data.
Transportation Research Record: Journal of the Transportation Research Board,
1924(1), 9-18.
150
Schumacker, R. E., & Lomax, R. G. (2004). A beginner's guide to structural equation
modeling: Psychology Press.
Shay, E., & Khattak, A. (2007). Automobiles, Trips, and Neighborhood Type:
Comparing Environmental Measures. Transportation Research Record: Journal
of the Transportation Research Board, 2010, 73-82. doi:doi:10.3141/2010-09
Silva, J. d. A. e., Golob, T., & Goulias, K. (2006). Effects of Land Use Characteristics on
Residence and Employment Location and Travel Behavior of Urban Adult
Workers. Transportation Research Record: Journal of the Transportation
Research Board, 1977, 121-131. doi:doi:10.3141/1977-17
Silva, J. d. A. e., & Goulias, K. (2009). Structural Equations Model of Land Use Patterns,
Location Choice, and Travel Behavior. Transportation Research Record: Journal
of the Transportation Research Board, 2135, 106-113. doi:doi:10.3141/2135-13
Silva, J. d. A. e., Morency, C., & Goulias, K. (2012). Using structural equations modeling
to unravel the influence of land use patterns on travel behavior of workers in
Montreal. Transportation Research Part A: policy and practice, 46(8), 1252-
1264.
Song, S., Diao, M., & Feng, C.-C. (2016). Individual transport emissions and the built
environment: A structural equation modelling approach. Transportation Research
Part A: policy and practice, 92, 206-219.
doi:http://dx.doi.org/10.1016/j.tra.2016.08.005
Srinivasan, S., & Bhat, C. R. (2008). An exploratory analysis of joint-activity
participation characteristics using the American time use survey. Transportation,
35(3), 301-327.
Stead, D. S. J. (1999). Planning for less travel: identifying land use characteristics
associated with more sustainable travel patterns. University of London.
Sun, B., Ermagun, A., & Dan, B. (2016). Built environmental impacts on commuting
mode choice and distance: Evidence from Shanghai. Transportation Research
Part D: Transport and Environment.
doi:http://dx.doi.org/10.1016/j.trd.2016.06.001
Susilo, Y. O., & Kitamura, R. (2008). Structural changes in commuters’ daily travel: The
case of auto and transit commuters in the Osaka metropolitan area of Japan,
1980–2000. Transportation Research Part A: policy and practice, 42(1), 95-115.
doi:http://dx.doi.org/10.1016/j.tra.2007.06.009
Townsend, T. A. (1987). The Effects of Household Characteristics on the Multi-day Time
Allocations and Travel/activity Patterns of Households and Their Members:
Northwestern University.
151
United States Census Bureau. (2016). Population and Housing Unit Estimates.
Retrieved from http://www.census.gov/programs-surveys/popest.html
United States Department of Transportation; Bureau of Transportation Statistics. (2016).
National Transportation Statistics 2016. Washington, DC Retrieved from
http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/NTS_Entire_16Q1.pdf.
United States Environmental Protection Agency. (2016a). Evolution of the Clean Air Act.
Retrieved from https://www.epa.gov/clean-air-act-overview/evolution-clean-air-
act
United States Environmental Protection Agency. (2016b). Fast facts : U.S. transportation
sector greenhouse gas emissions 1990-2014. (EPA-420-F-16-020). Retrieved
from http://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P100ONBL.pdf.
Van Acker, V., & Witlox, F. (2010). Car ownership as a mediating variable in car travel
behaviour research using a structural equation modelling approach to identify its
dual relationship. Journal of Transport Geography, 18(1), 65-74.
Van Acker, V., & Witlox, F. (2011). Commuting trips within tours: how is commuting
related to land use? Transportation, 38(3), 465-486.
van den Berg, P., Arentze, T., & Timmermans, H. (2013). A path analysis of social
networks, telecommunication and social activity–travel patterns. Transportation
Research Part C: Emerging Technologies, 26, 256-268.
van Wissen, L. J. (1991). A model of household interactions in activity patterns.
University of California Transportation Center.
Vovsha, P., & Petersen, E. (2005). Escorting children to school: statistical analysis and
applied modeling approach. Transportation Research Record: Journal of the
Transportation Research Board, 1921(1), 131-140.
Vovsha, P., Petersen, E., & Donnelly, R. (2003). Explicit modeling of joint travel by
household members: statistical evidence and applied approach. Transportation
Research Record: Journal of the Transportation Research Board, 1831, 1-10.
Wallace, B., Barnes, J., & Rutherford, G. (2000). Evaluating the Effects of Traveler and
Trip Characteristics on Trip Chaining, with Implications for Transportation
Demand Management Strategies. Transportation Research Record: Journal of the
Transportation Research Board, 1718(-1), 97-106. doi:10.3141/1718-13
Wallace, B., Mannering, F., & Rutherford, G. (1999). Evaluating effects of transportation
demand management strategies on trip generation by using Poisson and negative
binomial regression. Transportation Research Record: Journal of the
Transportation Research Board(1682), 70-77.
152
Wang, R. (2014). The stops made by commuters: evidence from the 2009 US National
Household Travel Survey. Journal of Transport Geography.
Wilhelm, J., Wolf, J., Kang, E., & Taylor, D. (2014). The Cleveland GPS Household
Travel Survey: Survey Design, Imputation of Trip Characteristics, and Secondary
Uses of the Data. Paper presented at the Transportation Research Board 93rd
Annual Meeting.
Yang, M., Wang, W., Ren, G., Fan, R., Qi, B., & Chen, X. (2010). Structural Equation
Model to Analyze Sociodemographics, Activity Participation, and Trip Chaining
Between Household Heads. Transportation Research Record: Journal of the
Transportation Research Board, 2157(1), 38-45.
Zhang, L., Hong, J. H., Nasri, A., & Shen, Q. (2012). How built environment affects
travel behavior: A comparative analysis of the connections between land use and
vehicle miles traveled in US cities. Journal of Transport and Land Use, 5(3), 40 -
52. doi:10.5198/jtlu.v5i3.266