URBAN BUILT ENVIRONMENT AND TRAVEL BEHAVIOR: UNDERSTANDING
GENDER AND SOCIO-ECONOMIC DISPARITIES IN ACCESSIBILITY AND MOBILITY
OF URBAN TRANSPORTATION
By
Jieun Lee

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Geography – Doctor of Philosophy
2014

ABSTRACT
URBAN BUILT ENVIRONMENT AND TRAVEL BEHAVIOR: UNDERSTANDING
GENDER AND SOCIO-ECONOMIC DISPARITIES IN ACCESSIBILITY AND MOBILITY
OF URBAN TRANSPORTATION
By
Jieun Lee

Urban transportation studies have shown that the ease of accessing urban opportunities in
a compact community, characterized with higher density, mixed land uses, and high
connectivity, are extremely crucial for urban residents to lead full and active lives. While
researchers have also increasingly recognized the differences in travel behavior based on gender
and different socioeconomic and ethnic/racial composition, the realities confronted by lowerincome, urban minorities, and particularly women, in neighborhoods experiencing disinvestment
and decline have been overlooked in most studies. In this context, this study focuses on the
Detroit region (SEMCOG), which is recognized as one of the most decentralized and racially
segregated metropolitan areas in the U.S. Six neighborhoods in the Detroit region were selected
based on socio-demographic and urban built environment characteristics; with two urban Detroit
neighborhoods, communities experiencing extreme disinvestment and decline, being compared
to four wealthy suburban neighborhoods.
This Ph.D. dissertation explores the linkage between specific urban built environment
characteristics and individual travel behavior, in order to identify the neighborhood typologies
defining neighborhood-level differences in travel patterns. The analysis also examines gender
differences in travel behavior after controlling for urban built environment, socioeconomic and

demographic factors. Two main datasets were used for this dissertation: (1) built environment
data from field surveys and inventories; and (2) individual or household travel data from a mail
survey.
First, the results from the multivariate regression analysis, using 1,106 road network
buffers (RNBs), confirmed the positive effects of RNBs that maintained a greater density,
diverse land uses, highly connected road networks and more bus stops in promoting nonmotorized travel while reducing motorized travel. Next, three neighborhood typologies—the
higher density urban Detroit neighborhood group, the higher density suburban neighborhood
group, and the lower density suburban neighborhood group—were defined for the within
neighborhood analysis. In addition to the significant effects of income and personal vehicle
access to a person’s travel pattern that were revealed from the OLS regression analysis, the
discriminant analysis differentiated urban Detroit neighborhoods by their lower socioeconomic
characteristics, by poor neighborhood environment quality for pedestrian activities and by their
pedestrian dominant travel patterns. Lastly, the results from the OLS regression and Analysis of
Covariance showed that the traditional gender role was still reflected in women’s daily travel in
the Detroit region in that women traveled more frequently and longer distances for household
responsibilities, and married women traveled to shop more frequently and longer distances than
married men across all three neighborhood typologies. The findings also revealed travel burdens
of the socially marginalized populations in terms of the extensive travel distances necessary to
reach daily destination due to the decentralization of urban amenities in the Detroit region.

Copyright by
JIEUN LEE
2014

This thesis is dedicated to
My parent, Hyojong Lee and Okju Kang,
and
Ivan J. Ramirez and Emmanuel Ramirez-Lee

v

ACKNOWLEDGEMENTS

Throughout my journey in the doctoral program, I greatly received special support and
guidance from some very important people. First of all, I would like to thank Dr. Igor Vojnovic
as my dissertation advisor, my mentor, and my great friend. Without his full-hearted
encouragement, this dissertation would not have come to fruition. I would also like to thank my
committee members beginning with Dr. Sue Grady, who helped me greatly with methods, and
the conceptual and methodological guidance of Dr. Joe Darden, Dr. Joe Messina, Dr. Bruce
Pigozzi, and Dr. Zenia Kotval that helped improve my research. I also thank to Dr. Heidi Harley
at New College of Florida who offered me her sincere support and opportunities. To my family, I
would like to send out a special acknowledgement. Only with their support and deep beliefs in
me, it was possible to progress my journey in the doctoral program and achieve my work. My
father, Hyo-jong Lee, is my hero since I remember and has believed in me more than myself, and
taught me how to be a responsible person with sincere humility. My ever-loving mother, Ok-ju
Kang, is my shelter and inspiration for my life. With their constant love and trust, I have become
who I am. I also thank my brother, Jung-hoon Lee, as my colleague, friend, and for all the help
he offered. Lastly but not least, I would like to thank my husband and best friend, Ivan J.
Ramirez, as his support and love has sustained my passion and confidence in me as a scientist
and the most beloved in the world. Emmanuel ‘Manu’ Ramirez-Lee, my baby, I love him so
much, and thank him for coming to us and helping me to finish this dissertation. Also, thank you,
Zeenat, Minting, Cristina, and Rita for being dear friends and colleagues. Most of all, I’d like to
thank my Lord, my God for giving me strength always and endurance in tough times to save me
from despair.

vi

TABLE OF CONTENTS
LIST OF TABLES .........................................................................................................................x
LIST OF FIGURES .....................................................................................................................xv
CHAPTER 1: INTRODUCTION .................................................................................................1
1.1. Research Focus and Dissertation Structure ....................................................................7
1.2. Research Objectives and Questions .................................................................................8
CHAPTER 2: BACKGROUND RESEARCH ..........................................................................12
2.1. Theoretical Background: The Urban Built Environment and Travel Patterns ........12
2.1.1. Transportation demand analysis ......................................................................12
2.1.2. The Urban built environment and travel behavior .........................................17
2.1.3. Transportation equity: gendered urban transportation mobility .................21
2.1.4. The urban built environment, physical activity and overweight/obesity ......27
2.2. Statement of Problem ......................................................................................................30
2.3. Study Area and Research Data .......................................................................................33
2.3.1. Study area: the Detroit region, Michigan ........................................................33
2.3.2. Research Data: Household Travel Survey .......................................................36
CHAPTER 3: Built Environments and Individual Travel Behavior ......................................68
3.1. Research Objectives, Questions, and Hypotheses .........................................................73
3.2. Built Environment and Travel Research Data and Method ........................................73
3.2.1. Built environment and travel research data ....................................................73
3.2.2. Built environment and travel behavior research methods .............................89
3.3. Urban Density...................................................................................................................94
3.3.1. Urban density and non-motorized travel .........................................................94
3.3.2. Urban density and motorized travel .................................................................96
3.4. Mixed land use..................................................................................................................98
3.4.1. Mixed land use (LUM3) and non-motorized travel ........................................98
3.4.2. Mixed land use (LUM7) and non-motorized travel ......................................100
3.4.3. Mixed land use (LUM8) and non-motorized travel ......................................102
3.4.4. Mixed land use (LUM3) and motorized travel ..............................................104
3.4.5. Mixed land use (LUM7) and motorized travel ..............................................106
3.4.6. Mixed land use (LUM8) and motorized travel ..............................................108
3.5. Street Connectivity and Bus Stops ...............................................................................111
3.5.1. Street connectivity and non-motorized travel ...............................................111
3.5.2. Street connectivity and motorized travel .......................................................112
3.5.3. Bus stops and non-motorized travel ...............................................................112
3.5.4. Bus stops and motorized travel .......................................................................113
3.6. Built environment factors and individual travel behavior .........................................116
3.6.1. Built environment factors and non-motorized travel ...................................117
vii

3.7.

3.6.2. Built environment factors and motorized travel ...........................................122
Conclusion and Discussion ............................................................................................127

CHAPTER 4: Neighborhood Typology and Travel Behavior in the Detroit region ...........136
4.1. Research Objectives, Questions, and Hypotheses .......................................................138
4.2. Neighborhood Typology Analysis Data and Method ..................................................138
4.3. Aggregate travel patterns in the Detroit region ..........................................................141
4.4. Different populations and their travel patterns in the different neighborhoods .....144
4.5. Travel behavior characteristics in three neighborhood typologies in the Detroit
region ...............................................................................................................................155
4.6. Conclusion and Discussion ............................................................................................168
CHAPTER 5: Gender difference in travel behavior in Detroit region .................................172
5.1. Research Objectives, Questions and Hypotheses ........................................................174
5.2. Disaggregate Travel Behavior Research Data and Method.......................................175
5.3. Results .............................................................................................................................180
5.3.1. Urban Detroit (Lower East Side Detroit Neighborhoods): Disaggregate
analyses on individual travel behavior in the higher density urban neighborhoods180
5.3.1.1. Gender Difference in Travel Behavior: Trips to Work ................................183
5.3.1.2. Gender Difference in Travel Behavior: Trips to Shops, Restaurants,
Personal Services, and Leisure .....................................................................................183
5.3.1.2.1. Trips to shopping ..........................................................................................183
5.3.1.2.2. Trips to restaurants ........................................................................................183
5.3.1.2.3. Trips to personal services ..............................................................................183
5.3.1.2.4. Trips for leisure .............................................................................................184
5.3.1.3. Distinctions by Gender and Marital Status in Travel Behavior: Trips to
Work ............................................................................................................................191
5.3.1.4. Distinctions by Gender and Marital Status in Travel Behavior: Trips to
Shops, Restaurants, Personal Services, and Leisure ..................................................192
5.3.1.4.1. Trips to shopping ..........................................................................................192
5.3.1.4.2. Trips to restaurants ........................................................................................193
5.3.1.4.3. Trips to personal services ..............................................................................194
5.3.1.4.4. Trips for leisure .............................................................................................195
5.3.1.5. Discussion ..........................................................................................................196
5.3.2. Higher Density Suburban (Ann Arbor and Birmingham Neighborhoods):
Disaggregate analyses on individual travel behavior in the high density suburban
neighborhoods in the Detroit region ............................................................................199
5.3.2.1. Gender Difference in Travel Behavior: Trips to Work ................................203
5.3.2.2. Gender Differences in Travel Behavior: Trips to Shops, Restaurants,
Personal Services and Leisure ......................................................................................204
5.3.2.2.1. Trips to shopping ..........................................................................................204
5.3.2.2.2. Trips to restaurants ........................................................................................204
5.3.2.2.3. Trips to personal services ..............................................................................204
5.3.2.2.4. Trips for leisure .............................................................................................205
5.3.2.3. Gender and Marital Status Distinctions in Travel Behavior: Trips to Work ..
viii

5.4.

............................................................................................................................213
5.3.2.4. Gender and Marital Status Distinctions in Travel Behavior: Trips to Shops,
Restaurants, Personal Services, and Leisure ...............................................................214
5.3.2.4.1. Trips to shopping ..........................................................................................214
5.3.2.4.2. Trips to restaurants ........................................................................................217
5.3.2.4.3. Trips to personal services ..............................................................................217
5.3.2.4.4. Trips for leisure .............................................................................................218
5.3.2.5. Discussion ..........................................................................................................219
5.3.3. Lower Density Suburban (Bloomfield Hills and West Bloomfield
Neighborhoods): Disaggregate analyses on individual travel behavior in the lower
density suburban neighborhoods in the Detroit region ..............................................223
5.3.3.1. Gender Differences in Travel Behavior: Trips to Work ..............................227
5.3.3.2. Gender Differences in Travel Behavior: Trips to Shops, Restaurants,
Personal Services and Leisure ......................................................................................227
5.3.3.2.1. Trips to shopping ..........................................................................................227
5.3.3.2.2. Trips to restaurants ........................................................................................229
5.3.3.2.3. Trips to personal services ..............................................................................229
5.3.3.2.4. Trips for leisure .............................................................................................229
5.3.3.3. Gender and Marital Status Distinctions in Travel Behavior: Trips to work
............................................................................................................................237
5.3.3.4. Gender and Marital Status Distinctions in Travel Behavior: Trips to Shops,
Restaurants, Personal Services, and Leisure ...............................................................237
5.3.3.4.1. Trips to shopping ..........................................................................................237
5.3.3.4.2. Trips to restaurants ........................................................................................238
5.3.3.4.3. Trips to personal services ..............................................................................239
5.3.3.4.4. Trips for leisure .............................................................................................240
5.3.3.5. Discussion ..........................................................................................................241
Conclusion and discussion .............................................................................................245

CHAPTER 6: CONCLUSION..................................................................................................252
6.1. Summary and Conclusion .............................................................................................252
6.2. Theoretical contributions and policy implications .....................................................262
6.3. Future research ..............................................................................................................266
APPENDIX .................................................................................................................................269
REFERENCES ...........................................................................................................................280

ix

LIST OF TABLES
Table 1 Surveys from Each Neighborhood (total: 1191) ...............................................................37
Table 2 Descriptive analysis of the survey sample by neighborhood groups (categorical
variables) ................................................................................................................................38
Table 3 Descriptive analysis of the survey sample by neighborhood groups (continuous
variables) ................................................................................................................................39
Table 4 Descriptive analysis of the survey sample by neighborhood (categorical variables) .......43
Table 5 Descriptive analysis of the survey sample by neighborhood (continuous variables) .......44
Table 6 Urban structure density and connectivity in the study area by neighborhood groups ......48
Table 7 Urban structure density and connectivity in the study area by neighborhoods ................49
Table 8 Neighborhood built environment characteristics by neighborhood in the Detroit study
region ......................................................................................................................................76
Table 9 Neighborhood built environment characteristics by neighborhood group in the Detroit
study region ............................................................................................................................77
Table 10 Total number and average area of the road network buffers (RNB) by neighborhood ..87
Table 11 Average number of structure by land use type per road network buffers (RNB) by
neighborhood ..........................................................................................................................88
Table 12 Built environment and individual travel behavior analysis data and method .................91
Table 13 Urban density and individual pedestrian travel in the Detroit region .............................95
Table 14 Urban density and individual driving travel in Detroit region........................................97
Table 15 Mixed land use (LUM3) and individual pedestrian travel in Detroit region ..................99

x

Table 16 Mixed land use (LUM7) and individual pedestrian travel in Detroit region ................101
Table 17 Mixed land use (LUM8) and individual pedestrian travel in Detroit region ................103
Table 18 Mixed land use (LUM3) and individual driving travel in Detroit region .....................105
Table 19 Mixed land use (LUM7) and individual driving travel in Detroit region .....................107
Table 20 Mixed land use (LUM8) and individual driving travel in Detroit region .....................109
Table 21 Connectivity and individual pedestrian travel in Detroit region...................................114
Table 22 Connectivity and individual driving travel in Detroit region .......................................115
Table 23 The built environment and individual pedestrian travel with LUM3 ...........................119
Table 24 The built environment and individual pedestrian travel with LUM7 ...........................120
Table 25 The built environment and individual pedestrian travel with LUM8 ...........................121
Table 26 Built environment and individual driving travel with LUM3.......................................123
Table 27 Built environment and individual driving travel with LUM7.......................................124
Table 28 Built environment and individual driving travel with LUM8.......................................125
Table 29 Variables for neighborhood typology and travel behavior models...............................140
Table 30 Descriptive statistics of socio-economic and demographic and travel pattern variables
in the Detroit region ..............................................................................................................141
Table 31 OLS regression results of aggregate travel pattern analysis in the Detroit region .......143
Table 32 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by socio-economic and demographic profile ...................................146
Table 33 Group statistics by socio-economic and demographic characteristic variables ............147

xi

Table 34 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by travel pattern ...............................................................................149
Table 35 Group statistics by aggregate travel pattern variables ..................................................150
Table 36 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by social environment ......................................................................152
Table 37 Group statistics by sense of fear variables ....................................................................153
Table 38 Canonical scores of group means with the two discriminant functions .......................154
Table 39 Descriptive analysis results of travel patterns for all activities in three neighborhood
typologies..............................................................................................................................157
Table 40 Descriptive analysis results of travel patterns for work trips in three neighborhood
typologies..............................................................................................................................158
Table 41 Descriptive analysis results of travel patterns for shopping trips in the three
neighborhood typologies ......................................................................................................161
Table 42 Descriptive analysis results of travel patterns for restaurant trips in the three
neighborhood typologies ......................................................................................................162
Table 43 Descriptive analysis results of travel patterns for personal service trips in the three
neighborhood typologies ......................................................................................................163
Table 44 Descriptive analysis results of travel patterns for leisure trips in the three neighborhood
typologies..............................................................................................................................164
Table 45 Variables for disaggregate travel behavior models.......................................................176
Table 46 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the urban Detroit neighborhoods: trip frequency ....................................181
Table 47 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the urban Detroit neighborhoods: trip distance.......................................182

xii

Table 48 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the urban Detroit neighborhoods .......................................................................185
Table 49 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—work .................................................................186
Table 50 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—shopping ..........................................................187
Table 51 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—restaurant visit ................................................188
Table 52 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—personal services .............................................189
Table 53 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—leisure ..............................................................190
Table 54 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the higher density suburban neighborhoods: travel frequency ................201
Table 55 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the higher density suburban neighborhoods: travel distance ..................202
Table 56 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the higher density suburban neighborhoods ......................................................207
Table 57 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—work ..........................................208
Table 58 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—shopping ...................................209
Table 59 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—restaurant visits ........................210
Table 60 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—personal services ......................211

xiii

Table 61 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—leisure .......................................212
Table 62 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the lower density suburban neighborhoods: travel frequency .................225
Table 63 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the lower density suburban neighborhoods: travel distance ...................226
Table 64 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the lower density suburban neighborhoods .......................................................231
Table 65 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—work ...........................................232
Table 66 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—shopping ....................................233
Table 67 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—restaurant visits .........................234
Table 68 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—personal services .......................235
Table 69 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—leisure ........................................236

xiv

LIST OF FIGURES

Figure 1 Disinvestment and decline in the city of Detroit (A). Photographed by Jieun Lee (taken
in 2010).....................................................................................................................................5
Figure 2 Disinvestment and decline in the city of Detroit (B). Photographed by Jieun Lee (taken
in 2010).....................................................................................................................................5
Figure 3 Disinvestment and decline in the city of Detroit (C). Photographed by Jieun Lee (taken
in 2010).....................................................................................................................................6
Figure 4 Disinvestment and decline in the city of Detroit (D). Photographed by Jieun Lee (taken
in 2010).....................................................................................................................................6
Figure 5 Six study neighborhoods in the Southeast Michigan (Macomb, Oakland, Washtenaw,
and Wayne). Maps produced by Jieun Lee ............................................................................45
Figure 6 Study area: The six Detroit region neighborhoods ..........................................................46
Figure 7 Land use density and types in Detroit 1 neighborhood (higher density urban). Maps
produced by Jieun Lee. ...........................................................................................................51
Figure 8 Land use density and types in Detroit 2 neighborhood (higher density urban). Maps
produced by Jieun Lee. ...........................................................................................................52
Figure 9 Land use density and types in Ann Arbor neighborhood (higher density suburban).
Maps produced by Jieun Lee. .................................................................................................53
Figure 10 Land use density and types in Birmingham neighborhood (higher density suburban).
Maps produced by Jieun Lee. .................................................................................................54
Figure 11 Land use density and types in Bloomfield Hills neighborhood (lower density
suburban). Maps produced by Jieun Lee. ...............................................................................55
Figure 12 Land use density and types in West Bloomfield neighborhood (lower density
suburban). Maps produced by Jieun Lee. ...............................................................................56
Figure 13 Land use type legend. ....................................................................................................57
xv

Figure 14 4-way intersection density in Detroit 1 and 2 neighborhoods (higher density urban).
Maps produced by Jieun Lee. .................................................................................................58
Figure 15 4-way intersection density in Ann Arbor and Birmingham neighborhoods (higher
density suburban). Maps produced by Jieun Lee. ..................................................................59
Figure 16 4-way intersection density in Bloomfield Hills and West Bloomfield neighborhoods
(lower density suburban). Maps produced by Jieun Lee. .......................................................60
Figure 17 Abandoned residential and nonresidential structures in Detroit city’s neighborhoods
(A). Photographed by Jieun Lee (taken in 2010). ..................................................................62
Figure 18 Abandoned residential and nonresidential structures in Detroit city’s neighborhoods
(B). Photographed by Jieun Lee (taken in 2010). ...................................................................62
Figure 19 Residential areas in higher density suburban neighborhoods (Ann Arbor).
Photographed by Jieun Lee (taken in 2010). ..........................................................................63
Figure 20 Residential areas in higher density suburban neighborhoods (Birmingham).
Photographed by Jieun Lee (taken in 2010). ..........................................................................63
Figure 21 Commercial structures in Detroit city’s neighborhoods (A). Photographed by Jieun
Lee (taken in 2010). ................................................................................................................64
Figure 22 Commercial structures in Detroit city’s neighborhoods (B). Photographed by Jieun Lee
(taken in 2010). .......................................................................................................................64
Figure 23 Commercial structures in higher density suburban neighborhoods (Ann Arbor).
Photographed by Jieun Lee (taken in 2010). ..........................................................................65
Figure 24 Commercial structures in higher density suburban neighborhoods (Birmingham).
Photographed by Jieun Lee (taken in 2010). ..........................................................................65
Figure 25 Pedestrian-hostile environment in Detroit city’s neighborhoods (A). Photographed by
Jieun Lee (taken in 2010). ......................................................................................................66
Figure 26 Pedestrian-hostile environment in Detroit city’s neighborhoods (B). Photographed by
Jieun Lee (taken in 2010). ......................................................................................................66
Figure 27 Pedestrian-friendly environment in higher density suburban neighborhoods
(Birmingham). Photographed by Jieun Lee (taken in 2010). .................................................67
xvi

Figure 28 Pedestrian-friendly environment in higher density suburban neighborhoods (Ann
Arbor). Photographed by Jieun Lee (taken in 2010). .............................................................67
Figure 29 Two examples of road network buffers (RNBs) in the Detroit 1 neighborhood. The
light green shaded area is the road network buffer. The blue circle with the red outline in the
middle of the road network buffer is a residential location of a survey respondent. Map
produced by Jieun Lee. ...........................................................................................................81
Figure 30 Two examples of road network buffers (RNBs) in the Detroit 2 neighborhood. The
light green shaded area is the road network buffer. The blue circle with the red outline in the
middle of the road network buffer is a residential location of a survey respondent. Map
produced by Jieun Lee. ...........................................................................................................82
Figure 31 Two examples of road network buffers (RNBs) in the Ann Arbor neighborhood. The
light green shaded area is the road network buffer. The blue circle with the red outline in the
middle of the road network buffer is a residential location of a survey respondent. Map
produced by Jieun Lee. ...........................................................................................................83
Figure 32 Two examples of road network buffers (RNBs) in the Birmingham neighborhood. The
light green shaded area is the road network buffer. The blue circle with the red outline in the
middle of the road network buffer is a residential location of a survey respondent. Map
produced by Jieun Lee. ...........................................................................................................84
Figure 33 Two examples of road network buffers (RNBs) in the Bloomfield Hills neighborhood.
The light green shaded area is the road network buffer. The blue circle with red outline in
the middle of the road network buffer is a residential location of a survey respondent. Map
produced by Jieun Lee. ...........................................................................................................85
Figure 34 Two examples of road network buffers (RNBs) in the West Bloomfield neighborhood.
The light green shaded area is the road network buffer. The blue circle with the red outline
in the middle of the road network buffer is a residential location of a survey respondent.
Map produced by Jieun Lee. ..................................................................................................86
Figure 35 Separation of groups on discriminant functions by socio-economic and demographic
profile ...................................................................................................................................147
Figure 36 Separation of groups on discriminant functions by aggregate travel pattern variables150
Figure 37 Separation of groups on discriminant functions by sense of fear variables ................153
Figure 38 Pedestrian trips for each travel activity by neighborhood groups ...............................165

xvii

Figure 39 Public transit trips for each travel activity by neighborhood groups...........................166
Figure 40 Driving trips for each travel activity by neighborhood groups ...................................167
Figure 41 Built environment factor: residential land use density ................................................270
Figure 42 Built environment factor: non-residential land use density .........................................271
Figure 43 Built environment factor: retail land use density ........................................................271
Figure 44 Built environment factor: personal services land use density .....................................272
Figure 45 Built environment factor: public institutions services land use density ......................272
Figure 46 Built environment factor: industrial land use density..................................................273
Figure 47 Built environment factor: land use mix index with 3 categories (LUM3) ..................273
Figure 48 Built environment factor: land use mix index with 7 categories (LUM7) ..................274
Figure 49 Built environment factor: land use mix index with 8 categories (LUM8) ..................274
Figure 50 Built environment factor: 4-way intersection density .................................................275
Figure 51 Built environment factor: bus stop density..................................................................275
Figure 52 Individual travel behavior variables: total pedestrian trip frequency for all activities 276
Figure 53 Individual travel behavior variables: total pedestrian trip distance for all activities ...276
Figure 54 Individual travel behavior variables: total driving trip frequency for all activities .....277
Figure 55 Individual travel behavior variables: total driving trip distance for all activities ........278
Figure 56 Socio-economic and demographic factors: age ...........................................................279
Figure 57 Socio-economic and demographic factors: income .....................................................279

xviii

CHAPTER 1: INTRODUCTION
The quality of life of urban residents is inextricably linked with that of urban areas that
offer amenities and associated opportunities, such as employment, education, retail, and personal
and public services. It is not only the distribution of urban activities, but the ease of accessing
these opportunities that is also extremely important for residents to lead full and active lives
(Boarnet & Crane, 2001; Cervero & Kockelman, 1997; Ewing and Cervero, 2001; Frank, 2000;
and Naess, 2006). According to many urban transportation researchers, it has been argued that a
dense, mixed land use and highly connected urban area is more likely to encourage the use of
more sustainable modes of transport, such as walking, cycling and public transit. These
alternative forms of transportation also potentially contribute to maintaining people’s healthier
life-styles, encouraging regular participation in moderate physical activities, such as walking and
biking (Frank, 2000, 2004; Frank et al., 2006; Handy, 1996, 2002, 2005; Saelens et al., 2003a;
Sallis et al., 2004; and Vojnovic et al., 2006). In short, the argument is that residents living in a
compact urban community—characterized by higher densities, mixed land uses and connected
street networks—with an effective and robust transportation system, tend to be healthier than
residents living in a sprawled suburban area with a singular vehicle travel mode, the private
automobile.
In recent years, transportation researchers have been also increasingly interested in
differences in travel behavior based on gender and different socio-economic and ethnic
composition (Johnston-Anumonwo, 1992, 1997; McLafferty and Preston, 1997; and Wyly, 1996,
1998). Since people make their travel decisions depending on the location of, and distance to,
various daily destinations, examining different sets of urban spatial opportunities is critical to
understanding people’s travel behavior. Suburbanization in the U.S.—along with its resulting
1

built environment topology characterized by low densities, single-use zoning, and disconnected
street networks—has created an automobile dependent urban transportation system. In addition,
with unequally distributed urban amenities across U.S. metropolitan areas (including
supermarkets and healthy restaurants options), various income, race and ethnicity, age and
gender groups face different sets of choices for their travel destinations, which may produce
distinctive travel behaviors with differential cost implications (temporal and/or monetary).
Therefore, ensuring equitable accessibility and mobility is essential to allow all people to reach
their economic, social, and recreational opportunities (Handy et al., 2005; Hanson and Hanson,
1981; Hanson and Johnston, 1985, Hanson, 1986, Johnston-Anumonwo, 1989, 1997;
MacDonald, 1999; McLafferty and Preston, 1997; Rosenbloom, 1985, 2005; and Wyly, 1996,
1998, 1999).
However, socio-economic, ethnic, and gender dynamics of accessibility and mobility
have not been explicitly understood. Longer distances between dispersing destinations (including
jobs, stores and public services) in most U.S. urban areas aggravate the conditions of the
disadvantaged, including women, minorities, and lower-income populations. For instance,
women’s travels have been greatly associated with their household responsibilities as the
traditional caregiver. Driving their children to school and various recreational activities, taking
care of their parents, and running errands means that they travel short distances, but many times
on a daily basis. That is, women tend to travel shorter distances than men, but they make more
trips than men over any given time period (Hanson, 1986; Hanson, 2010; and Turner and
Neimeier, 1997).
In this context, women who live in a household with just one, or no car, struggle with their
household responsibilities due to their constrained mobility (Hanson, 1986; Hanson, 2010; and
2

Turner and Neimeier, 1997). Particularly, low income, single mothers greatly suffer from
constrained mobility considering their dual roles as house maker and breadwinner (JohnstonAnumonwo, 1989; and McLafferty and Preston, 1997). Nevertheless, despite the social and
urban consequences of these gendered travel patterns, problems in women’s travel and their
urban opportunities have been marginalized in urban transportation research.
In order to understand gender and socio-economic disparities in accessibility and
mobility of urban transportation, six neighborhoods in the Detroit region were selected based on
socio-demographic and urban built environment characteristics. Here, the Detroit region refers to
the Southeast Michigan Council of Governments (SEMCOG) and includes Macomb, Oakland,
Wayne, St. Clair, Livingston, Washtenaw, and Monroe counties. The six neighborhood sites are
located in the counties of Wayne (two urban Detroit neighborhoods), Washtenaw (one Ann
Arbor neighborhood), and Oakland (three neighborhoods, one each in Birmingham, Bloomfield
Hills, and West Bloomfield). The two urban Detroit neighborhoods were selected particularly for
exploring individual travel behavior, and subsequent travel burdens, in communities
experiencing extreme disinvestment and decline.
With the rapid consumption of open space, as much as 13 times faster than population
growth in the area, the Detroit region has been recognized as one of the most decentralized
metropolitan centers in the U.S. (Vojnovic 2006; Vojnovic et al., 2006; and Vojnovic and
Darden, 2013b). In addition, the Detroit region has a deep history of racial segregation, a spatial
separation of the white population in the region, living largely in the suburbs, isolated from the
black population, living largely in the city (Darden et al., 1987).
The pattern of out-migration of whites into the Detroit suburbs has led to uneven
economic growth throughout the region. As Detroit’s auto industry relocated their production
3

lines into suburban Detroit, the city of Detroit lost not only its population, but also economic
opportunities, such as employment. The city of Detroit lost more than half of its population since
the 1950s, when the population peaked at about 1.85 million with a population density of 13,249
people per square mile (1950). The current population is about 700,000 (713,777 in 2010,
declining from 951,270 in 2000) with the population density of 5,170 people per square mile in
2010 (U.S. Bureau of Census, 2011) (see figures 1 to 4).
In contrast, Detroit’s suburbs have been gaining population and capital investment. For
example, in 2000, the per capita taxable assessment in Bloomfield Hills (located in Oakland
county) was $165,794, while the per capita taxable assessment in the city of Detroit was $7,573
in 2000 (SEMCOG, 2003). Hence, the suburbanization of the population resulted in the
decentralization of its tax base and a polarized fiscal capacity between the city and the wealthy
suburbs, which, subsequently, impacted opportunities for social mobility, and particularly among
the city’s minority population.

4

Figure 1 Disinvestment and decline in the city of Detroit (A). Photographed by Jieun Lee (taken
in 2010).

Figure 2 Disinvestment and decline in the city of Detroit (B). Photographed by Jieun Lee (taken
in 2010).

5

Figure 3 Disinvestment and decline in the city of Detroit (C). Photographed by Jieun Lee (taken
in 2010).

Figure 4 Disinvestment and decline in the city of Detroit (D). Photographed by Jieun Lee (taken
in 2010).

6

1.1. Research Focus and Dissertation Structure
This Ph.D. dissertation will examine the difference in travel behavior between men and
women within the context of differing urban built environment characteristics in the Detroit
region in order to evaluate women’s accessibility and mobility in urban space. As for the
research objectives, this dissertation aims to understand the linkage between urban built
environments and individual pedestrian travel behavior, to identify the neighborhood typology
that defines neighborhood-level differences in travel patterns, and lastly to examine gender
differences in travel behavior after controlling for urban built environment and socio-economic
and demographic factors.
This study involves several corresponding research hypotheses. In urban neighborhoods
characterized by higher-densities, mixed land uses, and high connectivity, people tend to utilize
non-motorized modes of travel. Next, concerning trip generation, residents living in
neighborhoods with greater income levels and automobile access tend to travel more frequently
and longer distances by car. In addition, comparing neighborhoods with similar urban density
and connectivity characteristics, residents living in neighborhoods with more diverse urban
opportunities and safer pedestrian environments (e.g. perception of fear of walking) will have
improved access to daily activities. Lastly, it is hypothesized that women’s travels are more
related to household responsibilities within the urban built environment. Within this context, in
accessing daily destinations, lower-income, urban minority women tend to rely on walking and
public transit more than other women’s groups, such as wealthier women in suburban
neighborhoods, due to lower access to a car among the lower-income Detroit women.

7

1.2. Research Objectives and Questions
In this dissertation, the research objectives are;
1) To understand the causal relationships of the urban built environment in promoting
pedestrian trips and reducing automobile use
2) To identify the neighborhood typology that helps explain differences in travel
behavior between urban and suburban neighborhoods
3) To examine gendered travel patterns and their relationship with urban built
environment characteristics and socio-economic and demographic factors
Accordingly, I have developed several research questions corresponding to the research
objectives;
1)-1. Do people utilize non-motorized travel modes such as walking and biking more in
urban neighborhoods characterized by higher-densities, mixed land uses, and high connectivity?
1)-2. Do people reduce automobile travel in urban neighborhoods characterized by
higher-densities, mixed land uses, and high connectivity?
2)-1. Which socio-economic and demographic factors significantly influence total trip
generation?
2)-2. Which neighborhood group, among urban Detroit, higher-density suburban, and
lower-density suburban neighborhoods, is differentiated against others in terms of socioeconomic and demographic characteristics, travel patterns, and neighborhood quality for
pedestrians?

8

3)-1. How do women travel differently from men within similar urban built
environments?
3)-2. How are lower-income urban minority women’s travels more disadvantaged than
other women’s groups living in the suburban Detroit neighborhoods?
Two main datasets were used for this dissertation; (1) built environment data from field
surveys and inventories in the six Detroit region neighborhoods; and (2) individual or household
travel data from a mail survey. Both datasets were collected as a part of National Science
Foundation funded study project (HSD Award # 0624263) that was active between the years
2007 and 2010.
For this PhD dissertation, first I identify the significant associations between various built
environment factors, including urban density, land use mix, street connectivity, bus stops and
individual travel behavior, both non-motorized and motorized. Multivariate regression analysis
was used to explore travel behavior among the entire data sample (n=1,106) without
neighborhood groupings. Next, the differences in the three neighborhood groupings—higher
density urban Detroit neighborhoods, higher density suburban neighborhoods and lower density
suburban neighborhoods—were examined within the context of different socio-economic and
demographic characteristics, neighborhood quality (perception of fear) and travel patterns.
Multivariate regression analysis and discriminant analysis, as well as descriptive analysis, were
used to illustrate the differences in travel.
Lastly, the study explored different travel patterns, and resulting costs, among women
themselves based on the neighborhood typology, after controlling for socio-economic and
demographic factors. For each neighborhood grouping—higher density urban Detroit
9

neighborhoods, higher density suburban neighborhoods, and lower density suburban
neighborhoods—a set of multivariate regression analyses were used to analyze travel behavior.
In the analysis, I focused on travel mode, travel distance, and travel frequency for the various
daily activities. The data set enabled the analysis into the distinctive pattern of women’s travels
from the array of trips, both work and nonwork. Also, the analysis explored the unique burdens
confronted by lower-income minority women in their daily travel activities.
I divided my dissertation into six chapters. Chapter 1 is the introduction to the
dissertation. Chapter 2 presents a background review of the travel behavior literature pertinent to
this study, along with a review of the study region, the Detroit region (Michigan). The chapter
includes a general introduction into travel demand analysis and its evolution, an overview of the
literature on urban built environment characteristics and individual travel behavior, an
assessment of gendered urban transportation mobility literature and a review of the existing work
on urban form, pedestrian travel behavior, and obesity.
Chapter 3 examines the association between urban built environment factors and
individual travel behavior across the Detroit region neighborhoods. Pedestrian and automobile
trips are explained in relation to urban density, land use mix, street connectivity and bus stop
density. In addition, the relative effects of urban built environment characteristics are tested for
promoting pedestrian trips and reducing automobile trips.
Chapter 4 identifies the neighborhood typology in terms of socio-economic and
demographic characteristics, neighborhood quality and travel patterns. Discriminant analysis
demonstrates the differences between three neighborhood groups in terms of socio-demographic
factors, travel patterns, and fear of crime and traffic. Next, descriptive analysis of travel patterns

10

explores the differences in travel patterns between neighborhood groupings, analyzing how built
environment characteristics shape travel distance, travel frequency and mode of travel.
Chapter 5 explores gendered travel behavior and the social dynamics of women’s travels.
Multivariate OLS regression results examine gendered differences in travel frequency and travel
distances for different types of travel activities, work and nonwork. Analysis of Covariance
results, in turn, add another significant explanation to gendered travel behavior by incorporating
marital status into the analysis. Exploring travel behavior within the differing neighborhood
typologies provides another dimension to the analysis, identifying the particular burdens that
urban Detroit neighborhood residents confront, and particularly urban minority women, in
accessing necessary daily destinations. Chapter 6 discusses the key findings of the study,
readdresses my research questions, and presents concluding thoughts and implications for policy
and for future research.

11

CHAPTER 2: BACKGROUND RESEARCH
2.1. Theoretical Background: The Urban Built Environment and Travel Patterns
2.1.1.

Transportation demand analysis

Specific characteristics in the urban built environment influence people’s travel patterns,
as evident in trip frequency, trip distance, mode choice, and total vehicle miles traveled (VMT)
(Cervero, 2002; Domencich and MacFadden, 1996; Frank, 2004; and Handy et al., 2006). Early
urban transportation research attempted to predict the total number of trips made using the
gravity model (Manheim, 1979; Martin, Memmott, and Bone, 1967; and Meyer and Miller,
1984). The model was used for allocating a certain amount of trips by origin-destination pairs
based on the attractive forces of a zone, such as population, employment or the intensity of land
use. To estimate a friction factor between an origin and destination, interzonal distance—such as
travel time or distance—was added to the model (Martin, Memmott, and Bone, 1967). Although
the model accounted for competition for trips between different land uses, the model was unable
to capture social and economic characteristics of zonal populations (Martin, Memmott, and
Bone, 1967; and Meyer and Miller, 1984). As an application of a simple physical law to social
behavior, a tremendous number of trips may be predicted between two neighboring zones due to
the proximity of both centroids (Martin, Memmot and Bone, 1967). The intention here is not to
criticize Martin, Memmott and Bone’s work, but to illustrate how this area of research has
evolved from this early study.
The next generation of urban transportation research expanded the model with
socioeconomic factors (Boarnet and Crane, 2001b; Domencich and McFadden, 1996; and
Hanson, 1986). The Urban Transportation Model System (UTMS), developed in the 1970s, is an

12

aggregate demand model consisting of four sequential steps, trip generation, trip distribution,
modal split and trip assignment (Manheim, 1979; Martin, Memmott, and Bone, 1967; Meyer and
Miller, 1984; Hanson, 1986; and Hanson and Giuliano, 2004). Travel behaviors are aggregated
into three dimensions of zones, peaks and homogenous groups. As the spatial dimension, zonal
characteristics are used to explain total flows of people between different parts of the urban area.
Peaks are the temporal components of travel flows that are established by converting a
continuous variable to a discrete, nominal variable. Lastly, homogeneous groups are used for
understanding the similarities in travel behavior of a specific population. Such categorizing
variables include family income, automobile ownership, and family size (Manheim, 1979; and
Meyer and Miller, 1984). Although the UTMS was the first large-scale, computer based urban
transportation demand model, the aggregate models do not allow urban transportation
researchers to understand the within-zone, or between-household, variations in trip making
(Hanson, 1986; and Meyer and Miller, 1984).
Several limitations of aggregate models were identified, based on behavioral theory, in
that the assumptions of aggregate trip generation model were biased to a specific transportation
user group. The models assumed a stereotypical multiple-occupant household headed by a male
worker commuting from the suburbs to the CBD (central business district) (Hanson, 1986; and
Meyer and Miller, 1984). Since the models were inherently biased toward a group of suburban
male commuters driving on highways to the CBD, the explanatory variables for the models
concerning household travel behavior might no longer be relevant, considering drastic
demographic changes during the past four or so decades (Hanson, 1986; Hanson and Giuliano,
2004; and Meyer and Miller, 1984).

13

In addition, it was assumed that people picked the alternative that had the maximum
value of utility for participating in various activities located at the destination points of their
journeys (Manheim, 1979; and Meyer and Miller, 1984). From the perspectives of behavioral
theory, it was argued that the fixed levels of utility maximization often confined the researchers
to seeking the underlying causes of travel choice, such as travelers’ preferences over goods, the
relative costs of those goods, and available trip making resources (Boarnet and Crane, 2001b).
Thus, more flexible variables were required to explain several other important urban travel
choices, such as trip chaining, non-home based trips, and multiple-worker household trip making
(Meyer and Miller, 1984).
Furthermore, the continuing suburbanization of U.S. metropolitan areas has brought
about new challenges for transportation planners. Rapidly increasing retail, service, office and
residential developments in suburban jurisdictions generate intrasuburban and intersuburban trip
making; trips between and within suburban nodal centers (Meyer and Miller, 1984). Thus, it is
critical to assess different populations’ accessibilities to urban amenities between municipalities
throughout the metropolitan areas in order to address critical issues of equitable urban
transportation.
Multidimensional disaggregate models were suggested to improve aggregate demand
functions by incorporating more realistic behavioral factors, models which exhibit the multiple
dimensions of choice. Various parameters for individuals or household characteristics were
required for comprehending the motivations, and relative values of trip purpose, time of day,
route, and mode in shaping trip making (Martin, Memmott, and Bone, 1967). These approaches
permitted flexibility in grouping schemes for identifying the socioeconomic and demographic
determinants of an individual trip, and so helped researchers to understand whether a specific
14

groups’ transportation experience was an expression of choice or of constraints (Hanson, 1986;
and Hanson and Guiliano, 2004).
In making travel-decisions, certain personal and household characteristics can be
constraints on an individual’s choice of trips. First, advancing age often reduces the levels of
total trips and multistop trips, and alters the purposes and timing of trips (Administration on
Aging, 1998; Alsnih and Hensher, 2003; Cao et al., 2008; and Collia et al., 2003). Second, the
availability of an automobile, which can be a function of culture or income, plays a critical role
for trip making (Vojnovic et al., 2013; forthcoming). As people travel farther by automobile than
by other modes, they can access a greater choice of workplaces and activity sites (Giuliano and
Dargay, 2005; Preston and McLafferty, 1999; and Pucher and Renne, 2003). Accordingly, trip
frequencies distinctively differ for those with access to a car versus those without access to a car
(Blumenberg, 2004; Giuliano and Dargay, 2005; and Paaswell, Recker and Milione in Hanson,
1986). Furthermore, if there is only one car available in a household, the person working will
generally take the car, while the rest of the household rely on other modes for transport (Hanson,
1986; Hanson, 2010; and Turner and Neimeier, 1997).
Third, gender is a factor that generates differences in trip making (Golob and McNally,
1997; Hanson, 1986; Hanson and Hanson, 1980; Hanson and Johnston, 1985; JohnstonAnumonwo, 1992; Kwan, 1999; Schwanen et al., 2008; and Wyly, 1998). Women tend to travel
shorter distances, work closer to home, resist devoting much time to a long commute, and
undertake more travel for shopping and personal needs, while men emphasize travel to work and
recreation (Hanson, 1986; Hanson and Hanson, 1980;and Hanson and Johnston, 1985). As
women travel to fulfill their multiple social roles of housewife, caregiver, and paid-laborer, they
tend to be more reliant on public transportation and walking, when compared to men, who make
15

greater use of the private automobile (Hanson and Hanson, 1980; Hanson and Johnston, 1985;
Hanson, 1986; Manheim, 1979; and Meyer and Miller, 1984).
Another meaningful socioeconomic and demographic factor in shaping travel behavior is
race and ethnicity. Ethnic and racial minorities living in central cities with lower income levels
use different modes of transportation and travel longer distances on the journey to work (Hanson
and Johnston, 1985; Johnston-Anumonwo, 1992; and Wyly, 1998). With implications for spacetime use, longer distances and longer trip times to work reveal significantly lower mobility levels
among racial and ethnic minorities.
Disaggregate research approaches to travel also allow studies to predict trip making
behavior by alternative travel modes at a finer-scale of analysis. At the aggregate level, at
relatively large-scales—such as traffic analysis zones (TAZ)—travel behavior by alternative
travel modes to the private automobile were often obscured and ignored since people tend to
walk or bike for closer destinations. Thus, disaggregate approaches at a finer-scale are
particularly important to identify the relationship between land use configurations and people’s
choice of alternative, and more robust, selections of travel mode (Boarnet and Crane, 2001b;
Hanson, 1986;and Manheim, 1979). Disaggregate data also help the models to detect the effects
of certain changes in the built environment on travel behavior (Hanson, 1986; Cervero and
Kockelman, 1997; Cervero, 2002; Handy, 2005a; Vojnovic, 2006; and Vojnovic et al., 2006).
Since it is only possible at the street or neighborhood level to detect changes in pedestrian
environment characteristics, these disaggregate approaches are critical to evaluate the effects of
improved pedestrian streetscape environments on people’s participation in physical activity
(Handy, 2005a; Handy et al., 2006; and Vojnovic et al., 2006).

16

2.1.2.

The Urban built environment and travel behavior

Research has shown that the pattern of urban development—whether daily activities are
spread out or more compact—has an effect on how one travels. Researchers have suggested a
diverse set of indicators of urban built environment characteristics to explain their effects on
travel. Urban transportation studies have shown a clear link between urban form and travel
mode, where people who live in higher-density, mixed-use, and connected neighborhoods drive
less, and walk and take public transit more often (Boarnet and Crane, 2001b; Cervero, 2002;
Ewing and Cervero, 2001; Frank, 2004; Handy, 2005a, 2006a, and 2006b; and Saelens et al.,
2003). Because certain urban forms alter each travel mode’s relative costs and convenience level,
people make different decision on their daily trips according to the built environment within
which they travel.
Therefore, certain characteristics in urban form, for instance, single use zoning, low
residential and employment densities, and disconnected street networks, result in a greater
dependence on private vehicle travel due to the increased distances between necessary daily
destinations. These destinations can include going to work, to shop, to school or travelling for
personal service needs, such as going to a bank, a pharmacy or a medical doctor.
Ewing and Cervero (2001) illustrated that people drive less and walk more in
communities with greater variety and greater proximity between destinations. In addition, Handy
et al. (2006) defined specific characteristics in the urban built environment that affect travel
behavior, including accessibility, proximity between destinations, physical activity options,
safety, attractiveness and the chance of socializing. Vojnovic (2006) also explored the
association between built environment characteristics and travel, concluding that characteristics
such as increased residential and employment densities, mixed land uses, increased connectivity,
17

as well as design strategies reducing pedestrian barriers and increasing pedestrian safety and
comfort, contribute to decreasing trip length, promote walking and biking, and reduce per capita
automobile ownership.
In particular, researchers have found significant causal relationship between compact
urban forms, various design strategies improving quality of the streetscape and reduced
automobile use (Frank and Pivo, 1995; Handy, 2005a; and Handy et al., 2005). Frank and Pivo
(1995) found significant relationships between residential density, land use mix and the
proportion of household trips that were in single occupant vehicles, on transit, and on foot, after
controlling for socio-economic and demographic factors.
From their study of the Seattle area, Moudon and colleagues (2005) suggest the
significantly increased likelihood of walking or biking if there was a greater presence of
destinations within a walkable distance from one’s residence. To confirm the causal link between
the built environment and travel behavior, Handy et al. (2005) used the indicators of ‘regional
accessibility.’ The study found that traditional pedestrian neighborhoods showed higher
accessibility, socializing and attractiveness characteristics for both work and nonwork travel, and
a decrease in driving distances and/or driving trips. The higher accessibility created a decrease in
the cost of walking, in terms of distance and time, and so contributed to residents substituting
driving for walking. Handy revealed that there was a statistically significant causal link between
travel behavior and neighborhood design, emphasizing the greater impact of neighborhood
design on walking trips when compared to driving trips (Handy, 2006a).
Since the characteristics of the urban built environment at different geographic scales
have different impacts on travel, it is also important to explore the complexity of travel behavior

18

at various spatial scales. Starting with the premise that walking is a neighborhood-scale mode of
travel while driving is a regional-scale mode of travel, researchers addressed the need for a
comprehensive and systematic regional approach to community design in order to promote
alternative urban transportation modes that should be competitive with driving (Handy, 2006a;
and Vojnovic, 2006).
At the regional scale, monocentric or polycentric urban spatial structures encourage
pedestrian travel activity around the urban activity centers with mixed urban functions. Also,
mixed land uses with residential, commercial, and retail, and potential employment opportunities
throughout an urban region, create different and more equitable commuting patterns. At the city
scale, urban density, mixed land uses, and greater connectivity are fundamental elements to
encourage more pedestrian activity (Vojnovic, 2006). As Handy (2006a) notes, the concentration
of functions in (neo-) traditional neighborhoods makes travel distances shorter, lowers the travel
costs and time for walking, and encourages pedestrian travel modes when compared to
automobile-oriented, modern suburban neighborhoods. Thus, at the city scale, ensuring
connectivity between and within neighborhoods becomes important in encouraging nonmotorized travel.
Finally, at the city block scale, microscale-built environment characteristics—
contributing to the comfort, safety and attractiveness of place—become important variables in
encouraging pedestrian travel (Handy, 2006a; Vojnovic, 2006; and Vojnovic et al., 2006). There
is a statistically significant effect of microscale built environment characteristics on walking
(Handy, 2006a). Certain community design elements that increase pedestrian activity include
proximity between destination end-points and a residence, the pedestrian streetscape itself

19

(formed by the building and lot width, building configuration, and building mix), and the quality
of the walking environment.
Building materials and colors, and detailed architectural elements also enhance the
quality of the streetscape for pedestrians. Large building set-backs and large, elongated building
and block widths in modern suburbs, which facilitate large parking lots, hinder pedestrians’
access between buildings and destinations, and also reduce their sense of safety. Thus, safer
street environments, facilitated by on-street parking along sidewalks and street lighting, as well
as high visual interest along a street, are critical in encouraging pedestrian travel (Vojnovic,
2006; and Vojnovic et al., 2006).

20

2.1.3.

Transportation equity: gendered urban transportation mobility

As people’s travel generating trip frequencies and purposes are closely associated with
their socioeconomic situation, how a person reaches their activities is also related to his or her
socioeconomic condition. However, mobility, the ability to reach activity sites of opportunities,
is also a fundamental component in fulfilling the travel demand as well as ‘accessibility,’ the
availability of urban opportunities within a certain distance or travel time. Certain barriers, such
as less availability of diverse transportation modes, longer distances or poor access between
necessary destinations, are well recognized transportation-related impediments (Hanson and
Giuliano, 2004). Thus, variations in transportation and transportation choices are likely to have
substantial impacts on individual access to all the necessary social and economic opportunities.
Furthermore, improved mobility can support active interactions within communities by
connecting people to diverse social and recreational points with ease. If the access to locations
becomes poorer and distances greater due to urban decentralization, people will be more
dependent on privately owned vehicles and less accessible to the destinations without a car.
Therefore, differences in mobility can be seen as a spatial consequence of urban built
environment designs.
Gender plays a key role in understanding societal trends and their impacts on
transportation demand and behavior. While the gap between women’s and men’s aggregate
travel behavior has been narrowing, it is still far from equal on a number of measures and trends,
with women having more variety in their travel patterns compared to men (Rosenbloom, 2005).
There have been several drastic changes in socio-economic structure suggesting new challenges
to urban transportation researchers. First, there has been a significant increase in the participation
of women in the paid labor force. Almost 62% of all U.S. women aged 16 and over were in the
21

labor force in 2002, which increased by 20% from 1975 (Rosenbloom, 2005). Thus, ‘trip to
work’ travel behavior is no longer applied only to men. If women work, however, the location of
their jobs tends to be closer to home when compared to those of men (Hanson and Johnston,
1985; Johnston-Anumonwo, 1992; and Wyly, 1998). As Hanson consistently illustrates in her
studies (1980, 1981 and 1986), women’s work trips are significantly shorter than men’s in both
travel distance and travel time.
However, women’s participation in paid employment has added to the burden of their
daily travel, especially for married women with children. The majority of all mothers work
outside the home, with 71% of married women with children working full time in 2003
(Rosenbloom, 2005). Accordingly, their travel demands and choices of job location tend to be
significantly different from that of their partners. Especially in explaining women’s travel
behavior, such as trip chaining for their multiple roles as homemaker and employed laborer,
requires parameters other than conventional variables used under the assumption of utility
maximization.
The differences in work trip time also reflect on women’s unique travel burdens. Women
were unwilling or unable to devote much time to a long commute due to their dual roles as
homemaker and employed laborer (Hanson and Hanson, 1980; and Hanson, 1986). Traditional
gender roles and associated trip making behaviors are also reflected in trip purposes. In
examining travel activity patterns of only full time employed women and men, male-female
differences in travel purposes are apparent in that women undertake more travel for shopping and
personal business, while men emphasize travel for work and recreation (Golob and McNally,
1997; Hanson, 1986; Kwan, 1999b; Schwanen et al., 2002; and Schwanen et al., 2008).

22

Lastly, there are a growing number of nontraditional households, single parent family
households in particular, which also generate unique travel patterns. In 2003, single parent
households accounted for almost 24% of all family households. Often, single mothers are placed
in significantly disadvantages positions within nontraditional households (Rosenbloom, 2005).
Considering women’s greater dependence on public transit and walking when compared to men,
transportation needs of women of diverse socioeconomic composition, ethnic backgrounds and
family structure should be addressed in detail (Hanson and Hanson, 1980; Hanson and Johnston,
1985; and Hanson, 1986). Therefore, a disaggregate analysis on different groups of women is
necessary in urban transportation research to account for the gender differences in travel by
socioeconomic status and ethnicity.
Sprawled urban spaces cause women to spend more time and money on their travel since
women must drive longer distances to meet their personal and household needs within these built
environments. Women in compact communities, in contrast, spend less time and money on their
travel between destinations, as their trips are characterized by shorter distances. Disaggregate
analyses using individual or household level travel behavior data have shown that women make
more trips by car, make fewer walking trips, are less likely to choose an alternative to driving for
their trips, but still drive fewer vehicle miles than men overall (Elliott and Joyce, 2004; Hanson
and Hanson, 1981; Hanson and Johnston, 1985; Hanson and Pratt, 1991; and MacDonald, 1999).
Handy et al. (2005) found that women still take on more of the responsibility for
household duties. Not only do these responsibilities restrict free time for women, but they also
necessitate additional travel needs. A report by the Surface Transportation Policy Project (1999)
found that women made two-thirds of all trips to chauffeur people around, such as driving
children to soccer practice or an older parent to the doctor. Such family responsibilities are
23

especially burdensome for women in the ‘sandwich generation’, with responsibilities for both
dependent children and aging parents. In the context of women’s frequent travel for multiple
household responsibilities, the car seems to provide more mobility in space and flexibility in
time. Thus, if a woman has no access to a private automobile and holds multiple responsibilities,
her ability is extremely limited to reach necessary destinations and meet basic daily needs.
Once disaggregated, studies have revealed that women have different travel patterns by
income, ethnicity, age, and marital status (Hanson and Hanson, 1981; Johnston-Anumonwo,
1989, 1997; McLafferty and Preston, 1997; Rosenbloom, 1989, 2005; Rosenbloom and Raux,
1985; and Wyly, 1996). Hanson and Hanson (1981) found that Swedish married women were
likely to make more shopping and domestic trips than their spouses, while they make fewer
social and recreational trips. Furthermore, Rosenbloom and Raux’s study (1985) focused on the
Netherlands, France and the Unites States found that women’s travel patterns varied significantly
with the age of their youngest child and associated needs in all three countries.
In a study of single mothers in Worcester, Massachusetts, Johnston-Anumonwo (1989)
found that women were more likely to make their work trips in cars even though they maintained
lower car ownership rates when compared to men. The study also found that single mothers had
longer work trips than comparable married women. Rosenbloom (1989) found similar patterns
with single mothers in Houston and Dallas, Texas. At all income levels except below $5,000 a
year, single mothers traveled a greater distance and more often in a car than comparable married
women.
Studies have also found a racial and ethnic dimension to travel behavior among women.
McLafferty and Preston (1992) explored the relationship between commuting time, job search,

24

and labor market segmentation, earnings, and travel modes. They found that white women had
shorter commutes to work compared to black women in the same job category. Similarly, Wyly
(1996) examined the spatial entrapment among all groups of women using work trip distance and
controlling for travel modes. He found that women made localized commutes to shorten the trip
time to their workplaces. Using the estimates of one-way work trip mileage by different travel
modes, he found that work trip costs were greater for women, and that African American women
showed a distinctive and costly commute pattern. Since urban decentralization has taken most of
the jobs to the suburbs, and yet suburban housing markets favor specific socioeconomic and
racial groups, urban racial and ethnic minorities have suffered from unemployment and longer
commutes to suburban industrial and service jobs.
It has been argued that women’s ‘spatial entrapment’ is a consequence of urban economic
restructuring and spatial shifts (Johnston-Anumonwo, 1997; McLafferty and Preston, 1992,
1996a, 1996b, 1996c, 1997; MacDonald, 1999; and Elliot and Joyce, 2004). Research by
Johnston-Anumonwo (1997) on travel differences by gender, ethnicity and socioeconomic status
was based on classifying commute types by income (more than $25,000 or less than $25,000)
and commute time (longer than 20 minutes or less than 20 minutes) using multivariate analysis
with PUMS 1990 data. In addition to the lower use of the private automobile by African
American commuters, there were clear racial and gender differences in commute times. While
both black and white women tended to have lower incomes compared to men, black women
traveled a longer time to reach their suburban jobs than suburban white women with similar
earnings and in the same occupation. MacDonald (1999) claimed that women’s commuting
patterns are constrained by limited mobility and a lack of convenient jobs due to occupational
and industrial segmentation. While a highly localized and segmented labor market results in

25

relatively short work trip distances for suburban white women, low income minority women in
the inner-city suffer from the long reverse commute to low skilled and entry-level suburban jobs.
With respect to the costs of travel, there is the added dimension of access to specific
transportation modes. Elliott and Joyce (2004) used OLS multivariate regression analysis to
identify the spatial constraints on women’s commuting in both the suburbs and the inner-city.
Using a Multi-City Survey of Urban Inequality (MCSUI) 1992-1994 and face-to-face interviews
in Los Angeles, Boston, and Atlanta, Elliott and Joyce found that single women tended to
commute longer than married women with children, and that there was significant racial
difference in commute times.
Particularly relevant to differences by ethnicity, this study showed the racial disparity in
commuting mode and time, in that black women relied more on public transit and had
significantly longer commute times than white women. In addition, black and Latina mothers of
teenagers exhibited the longest commute times. Due to relatively lower income level of single
mothers and their low car ownership rates, they suffer from living in areas with limited access to
effective public transit systems.
Within the literature on travel behavior and gender, it is important to recognize that
although nonwork travel is essential for determining a person’s quality of life, women’s nonwork
travels have not been extensively studied. Kwan (1999a, 1999b) has studied the space-time
constraints on women’s accessibility. She claims that women’s day time fixity constraints were
interrelated with their household responsibilities, employment and non-employment activities.
Utilizing a ‘space-time travel frame’ constructed from survey respondents’ travel diaries, she
found that the spatial boundaries of women’s non-employment travels were more confined than

26

men due to the effect of temporal fixity on achieving their diverse responsibilities. However,
there has been little known about women’s non-work travel patterns in detail, as well as those of
different social groups by income, race/ethnicity, and age.
Lastly, safety should be addressed as another important aspect of mobility equity. Since
walking and bicycling often becomes a necessity rather than a choice to low income and
minority households, safety in walking, cycling, or using public transit is quintessential for
women if they are to have appropriate mobility. Since women are more concerned about their
personal safety and feel less safe than men when using sidewalks and public transit, they tend to
perceive cars to be the safest mode and form of transportation (Carter in TRB 2005; Ingalls et al.,
1994; Koskela and Pain, 2000; Loukaitou-Sideris and Flink, 2009; Schulz and Gilbert, 1996; and
TRB, 2005a, 2005b). This perception may eliminate alternatives to driving from consideration
and make non-motorized modes particularly burdensome for women who do not have access to a
car. In addition, low income and minority women who live predominantly in central locations of
metropolitan areas, which are, or might be perceived as unsafe, have impaired mobility
associated with safety concerns. Thus, the comfort and the safety of a neighborhood becomes an
integral aspect of any study on gendered travel behavior.
2.1.4.

The urban built environment, physical activity and

overweight/obesity
The literature on the urban built environment and travel behavior, especially pedestrian
activity, also has implications on public health through the linkage of physical activity and
obesity (Cutts et al., 2009; and Frank et al., 2005). As the geographic separation of living,
working, shopping, and leisure destinations throughout the urban periphery has required
increased car use, the modern urban built environment has contributed to reducing opportunities
27

for active living within American metropolitan centers. From a public health perspective, there
have been adverse effects resulting from automobile-oriented built environments, such as
increased air pollution, noise, traffic congestion, road injury risk, and increased greenhouse gases
(Barla et al. 2010; Marshall et al. 2009; and Giles-Corti et al. 2011). Especially people living in
areas characterized by low density residential developments—with single land uses and
disconnected road networks—tend to get less exercise as part of their daily travel routines. A
neighborhood becomes ‘obesogenic’ when its layout prevents or discourages physical activity
(Frank and Engelke, 2001).
Half of the trips in America can be completed within a 20-minute bike ride, and a quarter
of trips are within a 20-minute walk, and yet, the vast majority of these short trips are taken by
automobile (FHA National Household Travel Survey 2001 in Hu, P. S., and T. R. Reuscher,
2004). Approximately 70% of American adults were overweight (34% obese), and 19% of all
teenagers and 17% of all children between ages 6 and 11 were overweight (CDC, 2010). Obesity
is a major risk factor for many of our most deadly diseases. The number one cause of death in the
U.S. is heart disease, and five of its six risk factors are associated with obesity: excessive weight,
inactivity, high blood pressure, high cholesterol, and diabetes.
In the United States in 2005, there were 1.5 million new diagnoses of diabetes for which
obesity is the number one risk factor (CDC, 2008). In addition, in 2007, less than half of all
Americans met the CDC’s recommendations for physical activity from work, transportation, or
leisure time exercise, and 13.5% did not get any physical activity at all (CDC, 2008). America’s
social metabolism—people’s participation in physical activity—is influenced by the built
environment in which they live, by the social environment, and by personal factors such as
gender, age, ability, and motivation (Edwards and Tsouros, 2006). In particular, the major
28

changes in American social metabolism can be explained by a reduction in physical activity in
the transportation sector. Americans made almost 90% of trips in cars, and spent on average
more than 30 miles driving every single day (FHA, 2001). Within this context, urban
transportation researchers have been advocating for an ACTIVE TRANSPORT movement,
aimed at increasing the level of physical activity by encouraging people to use non-motorized
travel modes and reversing the trends from America’s activity deficit.
There is growing evidence documenting the adverse health impacts of post-World War II,
automobile oriented land use patterns in the U.S. (Frank and Engelke, 2005; Frank et al.,2006;
Frumkin, Frank, and Jackson, 2004; and Handy, et al., 2002). People who live in “traditional” or
“walkable” neighborhoods reported that they walked about 30 minutes and more for their travel
each week (Saelens et al. 2003) and are involved more in physical activity compared to those
who live in less walkable suburban neighborhoods (Frank, et al. , 2005; King et al., 2003; and
Sealens et al., 2003a, 2003b).
Studies on walkability have examined and detailed the specific forms of neighborhoods
that are more suitable for people to walk. Lawrence Frank and colleagues (2005) showed that a
5% increase in the level of land use mix, residential density, street connectivity, and retail floor
to land area ratio was associated with a 32% increase in walking, ¼ point reduction in BMI,
6.5% increase in physical activity, and 5.5% reduction in per capita emissions of oxides of
nitrogen and volatile organic compounds (Frank et al., 2006). The study results also showed a
20% increase in the likelihood that someone walked with each addition of a recreational and
institutional facility within a 1 kilometer network distance from their residence.

29

The unprecedented national obesity epidemic and its ties to the automotive ‘monoculture’
are now far too serious to ignore. In 2007-2008, some 33.8% of American adults were obese,
with figures for American adult women (at 35.5% obese) being higher than for adult men (at
32.2% obese) (Flegal et al., 2010). With more than 60% of women in the United States not
engaging in the recommended amount of physical activity, and more than 26% of women not
being active at all, these figures are consistent with wider national trends (CDC, 2010). In
addition to this gender dimension to obesity, there is also an ethnic component to being
overweight. Between 2003 and 2006, for instance, the age-adjusted obesity rates among
American women older than 20 years varied strongly by ethnicity. While 53.3% of non-Hispanic
black women were obese, only 41.8% of Mexican-American women, and 31.6% of non-Hispanic
white women, were classified as obese (CDC, 2010).
Since community design can play an important role in increasing women’s physical
activity, providing safe and pleasing walking environments is critical in making walking a more
attractive choice for women. Indeed, studies on physical activity have shown that enjoyable
scenery is associated with increased walking activity among women. Handy et al. (2005) found
that women living in traditional walkable neighborhoods walked to the store more than twice as
often as women in suburban, low-density, automobile oriented neighborhoods. Therefore, it is
important to address equitable access and improved mobility of all social and ethnic groups, with
specific considerations provided for the needs of both men and women, in addressing issues of
socio-economic conditions and public health.
2.2. Statement of Problem

30

In order to understand people’s travel behavior, urban transportation researchers have
suggested diverse indicators to measure the effects of urban built environment characteristics on
travel, including land use density, land use diversity, proximity between destinations, and microscale design elements. Studies have found links between residents living in compact, mixed landuse, and connected neighborhoods and walking more often over shorter distances (Frank and
Pivo, 1995; Handy, 2006a; Handy et al., 2006; Saelens et al., 2003a, 2003b; Sallis et al., 2004;
Vojnovic, 2006; and Vojnovic et al., 2006). Urban transportation research, however, has focused
on major cities experiencing significant urban economic transformation into robust service
economies; cities such as Chicago, Seattle, Boston and San Francisco.
However, there has been little research on travel patterns performed in traditional
industrial cities in the Midwest and the Rustbelt, cities like Detroit, which are confronting a
rapidly declining urban core. Urban minorities in such cities find few employment opportunities
in the urban core and confront great distances to reach potential employment in the suburbs. In
addition to the economic forces, Detroit also confronts social and racial segregation, and political
fragmentation in the metropolitan area that has administratively separated the urban region along
social and racial lines. In such deteriorated inner-city neighborhoods, the implications of higher
densities, mixed land-uses, and connected neighborhoods on travel behavior and physical
activity have not been explored.
The existing literature has shown a meaningful relationship between increased
participation in physical activity and compact urban neighborhoods (characterized with higher
density, mixed land uses, and more connected road networks). However, in declining urban
cores, such as central in the city of Detroit, residents often participate less in daily physical
activity and subsequently show higher BMI levels in spite of similar compact urban built
31

environment settings. This is where the proposed research will fit into the larger literature on
urban form, travel behavior, and physical activity. The proposed dissertation intends to examine
the unsynchronized relationship between densely developed urban communities, and travel,
emphasizing gender differences while also taking into account socio-economic and ethnic
variables.
In addition, there are few studies on differential travel patterns focused on gender, and
how these variations by gender might also be affected by income, ethnicity, age, household
types, and neighborhood structure. Moreover, urban travel patterns beyond commutes to work
have not been examined in detail. Since nonwork trips tend to be more dynamic and associated
with people’s social roles and responsibilities, it is important to understand the difference in
travel activities between social groups, as they access shops, restaurants, personal services (such
as a pharmacy, doctor or bank) and various leisure destinations. While it is generally understood
that urban minorities, and women in particular, have suffered for decades from spatial constraints
of entrapment in the central city—due to racial segregation and decentralizing jobs, services, and
residential development—their full range of transportation demands and realities have not been
explored explicitly within the context of declining urban cores.
Few critical urban transportation researchers have examined urban minorities’ impaired
mobility, and even fewer have targeted a declining inner-city like Detroit (Blumenburg, 2004;
Grengs, 2010; Sanchez et al., 2004; and Turner, 1997). However, in these studies, gender
disparities along with their ethnic and socio-economic distinctions have not been explored in
detailed neighborhood level analyses, and particularly within the context of a full array of travel
destinations (from work to shopping to leisure) and associated trip constraints in terms of
distance, time, frequency and travel mode. In addition, most studies have been developed with
32

aggregate data at conventional TAZs (traffic analysis zones), such as counties or MSAs. Thus, it
has been difficult for existing studies to examine gender, along with individual socioeconomic,
and racial and ethnic variables, in shaping travel choices.
2.3. Study Area and Research Data
2.3.1.

Study area: the Detroit region, Michigan

Decentralization in the Detroit region1has been at the forefront of the discussions of U.S.
suburbanization since World War II. From 1960 to 1990, the area of developed urban land which
was converted from agricultural and natural lands has increased as much as 13 times faster than
population growth in the Detroit area (Vojnovic et al., forthcoming). In addition, the average
residential development density in the region dropped substantially during the 1990s, with new
construction being built at 1.3 housing units per acre, a significant decline from the 2.8 housing
units per acre that was the average in 1990 (SEMCOG, 2003). Considering the slow population
growth in the region, and an actual population decrease in the city of Detroit, urban
decentralization in south-east Michigan is continuously and rapidly consuming open space,
whether natural lands or agricultural lands, while at the same time abandoning existing
developed urban spaces (Vojnovic et al., 2006).
The decentralization of the Detroit region can be explained with the review of three
driving forces—economic, political and social variables. First, economic globalization and
deindustrialization has facilitated the region’s economic decline, and particularly in
manufacturing. As the region’s single dominant industry, automobile production, the locations of
1

In this dissertation, Detroit region is referred to the SEMCOG (Southeast Michigan Council of Governments)
region. Throughout the dissertation, this will make reference to as the Detroit region which consists of Macomb,
Oakland, Wayne, St. Clair, Livingston, Washtenaw, and Monroe counties.

33

the Big Three (GM, Ford, and Chrysler) automobile companies and related transportation
manufacturing have determined the economic rise and fall of metro Detroit. Since Ford started
relocating manufacturing plants from the city of Detroit during the 1940s, auto industry
producers have dispersed their production lines into the Detroit suburbs, the southern U.S. states
and abroad in order to reduce labor costs (Gallagher, 2010; and Martelle 2012). In addition, the
economic restructuring during the 1980s reinforced the uneven development of the region, as
more jobs were created in the suburban retail and office locations, following the outward
movement of the suburban population.
Second, outward decentralization was facilitated by public policy. These included such
policies as FHA loans (Federal Housing Administration), the G.I. Bill of the U.S. Department of
Veterans Affairs (The Servicemen's Readjustment Act of 1944), and IHA (The 1956 Interstate
Highway Act), all facilitating an institutionalized form of class and racial discrimination (Darden
et al., 1987; Darden et al., 2009; Thomas, 1997; Vojnovic, 1999; 2000b, 2000c; 2009). In
addition, new homebuyers appeared to favor the purchase of new housing in the suburbs over
housing in the central city. Moreover, competition between municipalities for new development
encouraged many homebuyers and businesses to locate outside of urbanized boundaries, as they
pursued lower property tax rates and growing public subsidies in undeveloped peripheral
jurisdictions (Vojnovic, 2006). In this political situation, characterized by excessive intermunicipal competition for new developments, problems associated with the building of socially
sustainable and healthy communities are often ignored by local policy makers (Vojnovic et al.,
2006).
Finally, social and racial tension has spurred the decentralization of Detroit. During the
post war years, thousands of single family homes and strip malls were built in the suburbs in an
34

attempt for the suburbanites to distance themselves from violence, crime, and the minority
populations that they considered a threat. On the other hand, the city’s black population
increased from 75 percent of the city’s entire population in 1990 to 83 percent by 2010 (U.S.
Bureau of Census, 2011). As research on Detroit region’s segregation and suburbanization have
explained, segregation and suburbanization in the region resulted from racial discrimination in
the housing market that was controlled by apartment managers, real estate brokers, and builders
(Darden et al., 1987, 2009; Sugrue, 1996; Thomas, 1997; and Vojnovic and Darden, 2013a).
As a result, Detroit has become the one of the most extreme examples of urban decline,
urban disinvestment and decentralization. The city of Detroit has lost about 25 percent of its
population during the last decade (with the population being at 951,270 in 2000 and dropping to
713,777 in 2010); a population less than half of its population highs reached in the 1950s, when
the population peaked at about 1.85 million. Accordingly, the population density in the city of
Detroit fell from 13,249 people per square mile in 1950 to 5,170people per square mile in 2010
(U.S. Bureau of Census, 2011).
In addition, the suburbanization of the population resulted in the decentralization of its
tax base and imprinted a landscape with a polarized fiscal capacity between the city and the
wealthy suburbs. The per capita income of Detroit city was $14,976 in 2008 while in the DetroitWarren-Livonia Metropolitan area the per capita income was $27,624 (U.S. Bureau of Census,
2010). In terms of the tax base, while the per capita taxable assessment in the city of Detroit was
$7,573 in 2000, the per capita taxable assessment in Bloomfield Hills, located in Oakland
County, was $165,794 (SEMCOG, 2003). Simply put, the city of Detroit has become one of the
poorest cities in the U.S. while its suburbs have been gaining population and investment, and

35

subsequently, a greater tax base. Oakland County, one of Detroit’s suburbs, is one of the richest
counties in the country (Darden et al., 2007, 2009).
The local pressure in the city’s service provision is effectively illustrated with public
education. Less than 25% of students graduated from the Detroit City School District in the
2003-2004 academic year, the lowest among the 50 largest U.S. cities (Vojnovic et al.,
forthcoming). In fact, the city of Detroit has been described as the most distressed urban core
with an increasing concentration of unemployment, poverty, and racial and ethnic minorities due
to outmigration of middle- and upper-income Whites to the suburbs, a pattern particularly
evident since the 1960s.
2.3.2.

Research Data: Household Travel Survey

For this study, 6 neighborhoods were selected within the Detroit region, specifically
falling within the boundary of SEMCOG (Southeast Michigan Council of Governments). The 6
neighborhoods were selected from Oakland, Washtenaw, and Wayne counties, southeastern
Michigan, with the purpose to isolate for specific urban built environment characteristics, and
social and ethnic population compositions (figure 5). The neighborhoods were selected based on
socio-economic and demographic characteristics, racial composition (isolating wealthy and white
communities in the suburbs and poor and black communities in the inner-city), as well as
specific conditions in the urban built environment (including urban density, land-use mix, and
connectivity).
From the six neighborhoods, there were 3200 surveys mailed-out to 1600 households
within the inner-city Detroit neighborhoods, and 1600 surveys mailed-out to 800 households for
each of the four suburban neighborhoods. Two surveys were sent to each household to capture

36

travel behavior variations, including travel variations between men and women aged over 18 in
the households. There were a total of 1191 respondents’ surveys collected in the mail survey
(table 1). Sample households were randomly selected by geographic stratification to ensure
sample representation by age, race and gender. The return rate was approximately 20% of the
actual potential sample, which is considered a good response rate for a mail survey sent out to
the general population (Sommer and Sommer, 1997).
Table 1 Surveys from Each Neighborhood (total: 1191)
Detroit 1

Detroit 2

Ann Arbor

Birmingham

Bloomfield
Hills

West
Bloomfield

128

158

297

211

201

196

F
or
the

entire survey sample, 63% of the survey respondents were women and 37% were men, 61.8%
were married and 38.2% were not married. In addition, 72.8% of the respondents were white and
27.2% were non white (table 2). The average age of the survey respondents was 53.8 years old
and the average annual personal income was $52,200. On average, the survey respondents were
responsible for less than one (0.7) dependent children per individual. With respect to automobile
ownerships, the average number of the personal automobiles was more than one car (1.2) per
individual (table 3).

37

Table 2 Descriptive analysis of the survey sample by neighborhood groups (categorical variables)
Higher density
suburban
1,191
286
508
Categorical variables

Lower density
suburban
397

Total

Urban Detroit

Women

63.0%

74.0%

61.9%

56.8%

Men

37.0%

26.0%

38.1%

43.2%

Married

61.8%

23.8%

71.3%

76.0%

Notmarried

38.2%

76.2%

28.7%

24.0%

White

72.8%

8.1%

94.7%

88.6%

N

Gender

Marital
status

Race

Non27.2%
91.9%
5.3%
11.4%
white
*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH: Bloomfield
Hills; and WB: West Bloomfield.

38

Table 3 Descriptive analysis of the survey sample by neighborhood groups (continuous variables)
Total
N

Age

Personal
income
(in $1,000)

The number
of
dependent
children
under 18
Personal
vehicle(s)
available

Min.
Max.
Mean
Median
St.Dv
Min.
Max.
Mean
Median
St.Dv
Min.
Max.
Mean
Median
St.Dv
Min.
Max.
Mean
Median
St.Dv

Urban Detroit

1,191
286
Continuous variables
18.0
18.0
94.0
94.0
52.0
53.8
52.5
53.0
16.7
15.8
3.3
3.3
125
160
52.2
23.6
52.4
15.0
31.7
20.3
0.0
0.0
8.0
8.0
0.8
0.7
0.0
0.0
1.3
1.1
0.0
0.0
5.0
6.0
0.8
1.2
1.0
1.0
0.7
0.8

39

Higher density
suburban

Lower density
suburban

508

397

19.0
93.0
51.8
51.0
15.7
7.5
160
58.1
55.0
26.6
0.0
3.0
0.6
0.0
0.9
0.0
6.0
1.3
1.0
0.6

18.0
92.0
57.3
56.0
14.8
4.5
160
63.9
65.6
32.3
0.0
5.0
0.7
0.0
1.1
0.0
4.0
1.4
1.0
0.7

Breaking down the entire sample by neighborhood groups reveals a clear socioeconomic
difference between urban Detroit neighborhoods and suburban neighborhoods, both higher
density and lower density (tables 2 and 3). The most striking difference between the urban
Detroit neighborhood group and the suburban neighborhood groups was the racial composition.
While suburban neighborhood groups were predominantly white (94.7% for the higher density
suburban neighborhood group and 88.6% for the lower density suburban neighborhood group),
the urban Detroit neighborhood group was predominantly non-white (91.9%).
Disparity in average personal income between neighborhood groups was also clearly
shown between the urban Detroit and the suburban neighborhoods. The average personal income
in the urban Detroit neighborhood group was $23,600, which was less than half of the average
personal income when compared to both the suburban neighborhood groupings ($58,100 in the
higher density suburban neighborhood group and $63,900 in the lower density suburban
neighborhood group). The lower income status of the urban Detroit neighborhood group was also
reflected in lower average automobile access per individual. There was less than one car (0.8)
that was available to a respondent in the urban Detroit neighborhoods whereas each suburbanite
had access to more than a car (1.3 for the higher density suburban neighborhood group and 1.4
for the lower density suburban neighborhood group).
The majority of urban Detroit respondents were not married (76.2%) while more than
70% of the suburban respondents from both neighborhood groups were married (71.3% for the
higher density suburban neighborhood group and 76.0% for the lower density suburban
neighborhood group). There was no significant distinction in terms of other demographic
characteristics such as gender, age, and the number of dependent children under the age of 18
between the three neighborhood groupings.
40

In sum, the urban Detroit neighborhood group was characterized as low income,
predominantly non-white with low access to a private automobile. In contrast, the two suburban
neighborhood groups, the higher density suburban and lower density suburban groups, were
characterized by higher incomes, predominantly white with greater access to a car. As seen in
tables4and 5, there is a clear distinction between the two urban Detroit neighborhoods and the
four suburban neighborhoods in terms of socioeconomic and demographic profiles.
In terms of other demographic characteristics, the average age of the survey respondents
in the urban Detroit neighborhoods were 54 and 50 (table 5).In addition, only about 25% of the
respondents were married or living with a partner (20.8% for Detroit 1 and 26.2% for Detroit 2)
(table 4). From the profile of the survey respondents and the neighborhood surveys, it was also
clear that the urban Detroit neighborhoods were shaped by years of disinvestment and decline.
Unlike the predominantly black, low-income urban Detroit neighborhoods, four suburban
neighborhoods were selected as the higher income and predominantly white communities in this
study. Except West Bloomfield (84.6%), all three suburban neighborhoods’ respondents were
over 90% white (95.1% for Ann Arbor, 94.1 % for Birmingham, and 92.4% for Bloomfield
Hills). The average annual personal incomes were well above $60,000 for the suburban
neighborhoods (about $63,000 for Birmingham and West Bloomfield, and $65,000 for
Bloomfield Hills) in 2008. Ann Arbor’s annual average personal income was slightly lower
(averaging about $55,000) (table 5). The income gap between the urban Detroit neighborhoods
and suburban neighborhoods, thus, is more than double, and almost triple between the Detroit 2
neighborhood and the Bloomfield Hills neighborhood.

41

As one might expect, the average automobiles per capita in the suburban neighborhoods
was 1.3 (except for the Bloomfield Hills neighborhood where the per capita ownership was 1.4
automobiles per individual), which is greater than that of urban Detroit respondents (0.8 for
Detroit 1 and 0.7 for Detroit 2). Demographically, the average ages of the suburban respondents
range from 49 (Ann Arbor) to 59 (Bloomfield Hills),and the majority of the respondents in the
suburbs were married or living with a partner (71.5% for Ann Arbor, 71% for Birmingham,
70.8% for Bloomfield Hills, and 81.4% for West Bloomfield) (table 4).
Between the urban Detroit neighborhoods and the suburban neighborhoods, there was no
substantial difference in terms of gender or the number of dependent children under the age of
18. However, the respondents in the urban Detroit neighborhoods had slightly more dependent
children (0.9 for Detroit 1 and 0.8 for Detroit 2) than suburban residents (0.5 and 0.6 for Ann
Arbor and Birmingham, and 0.7 for Bloomfield Hills and West Bloomfield). Also, in the survey
sample, there were more women respondents in both the urban Detroit neighborhoods (over
70%) when compared to the suburban neighborhoods (approximately 60%, ranging from 56.2%
in Bloomfield Hills to 64.2% in Ann Arbor).In sum, the residents in the urban Detroit
neighborhoods were predominantly non-white, not married, and low income, with poor access to
a personal automobile when compared to the predominantly white, married, and higher income
respondents in the suburbs.

42

Table 4 Descriptive analysis of the survey sample by neighborhood (categorical variables)
Urban Detroit

Higher density
suburban

Lower density
suburban

DT1

AA

BM

BH

WB

1,191

128
158
297
Categorical variables

211

201

196

Women

63.0%

73.1%

74.7%

64.2%

58.5%

56.2%

57.5%

Men

37.0%

26.9%

25.3%

35.8%

41.5%

43.8%

42.5%

Married

61.8%

20.8%

26.2%

71.5%

71.0%

70.8%

81.4%

Notmarried

38.2%

79.2%

73.8%

28.5%

29.0%

29.2%

18.6%

White

72.8%

4.4%

11.1%

95.1%

94.1%

92.4%

84.6%

Total
N

DT2

Gender

Marital
status

Race

Non27.2% 95.6% 88.9%
4.9%
5.9%
7.6% 15.4%
white
*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH: Bloomfield
Hills; and WB: West Bloomfield.

43

Table 5 Descriptive analysis of the survey sample by neighborhood (continuous variables)
Urban Detroit
Total
DT1
N

Age

Personal
income
(in $1,000)

The
number
of
dependent
children
under 18

1,191
Min.
Max.
Mean
Median
St.Dv
Min.
Max.
Mean
Median
St.Dv
Min.
Max.
Mean
Median
St.Dv

18.0
94.0
53.8
53.0
15.8
3.3
160
52.2
52.4
31.7
0.0
8.0
0.7
0.0
1.1

DT2

Higher density
suburban
AA

128
158
297
Continuous variables
21.0
18.0
19.0
94.0
89.0
90.0
54.4
50.1
49.4
54.0
51.0
47.0
15.8
17.1
15.4
4.5
3.3
7.5
125
105
160
24.7
22.8
54.7
15.4
15.0
52.4
20.7
20.0
24.5
0.0
0.0
0.0
5.0
8.0
3.0
0.9
0.8
0.5
0.0
0.0
0.0
1.3

1.3

0.9

Lower density
suburban

BM

BH

WB

211

201

196

23.0
93.0
55.2
54.0
15.6
9.0
160
63.0
65.0
28.6
0.0
3.0
0.6
0.0

23.0
92.0
59.1
59.0
15.8
12.5
160
64.5
67.6
30.5
0.0
4.0
0.7
0.0

18.0
84.0
55.4
54.0
13.5
4.5
160
63.1
60.0
34.2
0.0
5.0
0.7
0.0

0.9

1.0

1.1

Min.
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Max.
6.0
5.0
3.0
4.0
6.0
4.0
4.0
Personal
vehicle(s)
Mean
1.2
0.8
0.7
1.3
1.3
1.4
1.3
available Median
1.0
1.0
1.0
1.0
1.0
1.0
1.0
St.Dv
0.7
0.9
0.6
0.6
0.7
0.7
0.6
*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH: Bloomfield
Hills; and WB: West Bloomfield.

44

Figure 5 Six study neighborhoods in the Southeast Michigan (Macomb, Oakland, Washtenaw, and Wayne). Maps produced by Jieun
Lee

45

Lower density suburban: West Bloomfield (left) and Bloomfield Hills (right), MI.

Higher density suburban: Ann Arbor (left) and Birmingham (right), MI.

Higher density urban: Detroit 1 (left) and Detroit 2 (right), MI
Figure 6 Study area: The six Detroit region neighborhoods

46

The six neighborhoods were also selected to demonstrate specific conditions in the urban
built environment, including specific distinctions in urban density, land-use mix and connectivity
(figure 6). The urban Detroit neighborhoods and the higher density suburban (Ann Arbor and
Birmingham) neighborhoods were selected as the higher density neighborhoods with diverse
land uses and connected road networks. In contrast, the two lower density suburban
neighborhoods, West Bloomfield and Bloomfield Hills, were selected as typical automobile
oriented suburban neighborhoods that are characterized by low densities, single land uses
(residential) and disconnected road networks.

47

Table 6 Urban structure density and connectivity in the study area by neighborhood groups
Units per sq. km

Urban Detroit

Higher density
suburban

Lower density
suburban

Urban structure density
Single family
detached home
Semi-detached
home

501

577

375

0

1

0

Apartments

1

12

8

Townhomes/ Row
houses

3

5

9

Retail

2

5

4

Personal services

11

11

6

Public Institution

9

5

2

Industrial

1

0

0

55

0

0

16

3

Abandoned

Connectivity
Intersections

32

48

Table 7 Urban structure density and connectivity in the study area by neighborhoods
Higher density
suburban

Urban Detroit
Units per sq. km

DT1

DT2

AA

Lower density
suburban

BM

BH

WB

Urban structure density
Single family
detached home
Semi-detached
home

595

396

608

547

163

277

0

1

3

0

0

7

Apartments

1

5

3

7

10

12

Townhomes/
Row houses

1

2

17

8

7

3

Retail

2

1

2

8

0

0

Personal services

14

7

12

10

0

1

Public Institution

12

5

8

3

0

0

1

1

1

0

0

0

65

43

0

0

0

0

16

2

2

Industrial
Abandoned

Connectivity
Intersections

30

19

12

*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH: Bloomfield
Hills; and WB: West Bloomfield.

49

As seen in table 6, both the urban Detroit and the higher density suburban neighborhood
groups were characterized by higher density developments, a diverse land use mix and highly
connected road networks when compared to the low density suburban neighborhood group. Both
the urban Detroit and the higher density suburban neighborhood groups maintain more than 500
single family detached homes per square kilometer, while the low density suburban
neighborhood group maintains only 375 structures per square kilometer.
More nonresidential structures were also evident in the urban Detroit and the higher
density suburban neighborhoods (more than 20 retail, personal services, and public institutions
versus 12 nonresidential structures in the low density suburban neighborhood group). In terms of
street connectivity, there were more 4-way intersections per square kilometer in the urban Detroit
(32 per square kilometer) neighborhood group and the higher density suburban (16 per square
kilometer) neighborhood group than in the lower density suburban neighborhood group (3 per
square kilometer). Despite of higher density developments and the more connected road
networks, the urban Detroit neighborhood group was distinguished from the higher density
suburban neighborhood because of fewer retail establishments and more abandoned structures in
this neighborhood grouping. Neighborhood characteristics of urban density, land use mix, and
street connectivity are shown by individual neighborhoods in table 7.

50

Figure 7 Land use density and types in Detroit 1 neighborhood (higher density urban). Maps
produced by Jieun Lee.

51

Figure 8 Land use density and types in Detroit 2 neighborhood (higher density urban). Maps
produced by Jieun Lee.

52

Figure 9 Land use density and types in Ann Arbor neighborhood (higher density suburban).
Maps produced by Jieun Lee.

53

Figure 10 Land use density and types in Birmingham neighborhood (higher density suburban).
Maps produced by Jieun Lee.

54

Figure 11 Land use density and types in Bloomfield Hills neighborhood (lower density
suburban). Maps produced by Jieun Lee.

55

Figure 12 Land use density and types in West Bloomfield neighborhood (lower density
suburban). Maps produced by Jieun Lee.

56

Figure 13 Land use type legend.

57

Figure 14 4-way intersection density in Detroit 1 and 2 neighborhoods (higher density urban).
Maps produced by Jieun Lee.
58

Figure 15 4-way intersection density in Ann Arbor and Birmingham neighborhoods (higher
density suburban). Maps produced by Jieun Lee.
59

Figure 16 4-way intersection density in Bloomfield Hills and West Bloomfield neighborhoods
(lower density suburban). Maps produced by Jieun Lee.
60

Despite of the higher density of urban amenities, the level of abandonments among the
residential and nonresidential uses in the urban Detroit neighborhoods was also high (figures 17
to 24). On average, there were 65 visible abandoned structures per square kilometer in Detroit 1
and 43 visible abandoned structures per square kilometer in Detroit 2, whereas there were no
visible abandoned structures in either of the higher density suburban neighborhood groupings.
Comparing nonresidential landscapes leads to an even clearer understanding of differing
neighborhood qualities between the urban Detroit and the higher density suburban
neighborhoods (figures 21, 22 and 23, 24). Accordingly, the extreme level of abandonment and
the poor supply of certain commercial outlets results in a low-quality pedestrian environment in
the urban Detroit neighborhoods (figures 25, 26 and 27, 28). Therefore, it becomes increasingly
important to examine the qualitative dimensions of the neighborhood built environments, in
addition to quantifying measurements such as urban density, land use mix, and road network
connectivity, in order to understand the relationship between the urban built environment and
individual travel behavior.
Accordingly, as the first step in chapter 3,relationships between urban built environment
characteristics and individual travel behavior (pedestrian and driving trips) will be explored in
the Detroit region, in order to examine at a fine spatial scale (i.e. walkable area from an
individual’s residential location using actual road networks) the relationships between urban
form and travel. Following this analysis, in chapter 4, three neighborhood typologies will be
defined with regard to socioeconomic and demographic factors, neighborhood quality and travel
patterns. Lastly, in chapter 5, the differences in daily travel between women and men will be
analyzed after controlling for the neighborhood typology.

61

Figure 17 Abandoned residential and nonresidential structures in Detroit city’s neighborhoods
(A). Photographed by Jieun Lee (taken in 2010).

Figure 18 Abandoned residential and nonresidential structures in Detroit city’s neighborhoods
(B). Photographed by Jieun Lee (taken in 2010).

62

Figure 19 Residential areas in higher density suburban neighborhoods (Ann Arbor).
Photographed by Jieun Lee (taken in 2010).

Figure 20 Residential areas in higher density suburban neighborhoods (Birmingham).
Photographed by Jieun Lee (taken in 2010).

63

Figure 21 Commercial structures in Detroit city’s neighborhoods (A). Photographed by Jieun
Lee (taken in 2010).

Figure 22 Commercial structures in Detroit city’s neighborhoods (B). Photographed by Jieun Lee
(taken in 2010).

64

Figure 23 Commercial structures in higher density suburban neighborhoods (Ann Arbor).
Photographed by Jieun Lee (taken in 2010).

Figure 24 Commercial structures in higher density suburban neighborhoods (Birmingham).
Photographed by Jieun Lee (taken in 2010).

65

Figure 25 Pedestrian-hostile environment in Detroit city’s neighborhoods (A). Photographed by
Jieun Lee (taken in 2010).

Figure 26 Pedestrian-hostile environment in Detroit city’s neighborhoods (B). Photographed by
Jieun Lee (taken in 2010).

66

Figure 27 Pedestrian-friendly environment in higher density suburban neighborhoods
(Birmingham). Photographed by Jieun Lee (taken in 2010).

Figure 28 Pedestrian-friendly environment in higher density suburban neighborhoods (Ann
Arbor). Photographed by Jieun Lee (taken in 2010).

67

CHAPTER 3: Built Environments and Individual Travel
Behavior
In this part of the dissertation, I seek to examine how specific built environment
characteristics (such as density, land use mix and connectivity) affect individual travel behavior,
with a specific focus placed on pedestrian and driving trips in the Detroit region. The study
hypothesis is that residents in higher density neighborhoods characterized by mixed land uses
and connected road networks will travel more by non-motorized transportation modes. The
results will identify built environment characteristic(s) that are most strongly associated with
promoting pedestrian travel in the Detroit region, while discouraging driving.
Recently, with connections made to public health, there has been considerable research
interest in studying the relationship between the built environment and travel behavior, and
particularly pedestrian activity (Frank and Engelke, 2001, 2005; Frank et al., 2006; Frumkin et
al., 2004; Handy et al., 2002; Handy et al., 2005; LFC, 2005; and Saelens et al., 2003a, 2003b).
Previous studies have also investigated the potential causal links between specific built
environment characteristics and travel behavior, stressing the natural environmental impacts of
our resource intensive automobile travel patterns (Cervero and Gorham, 1995; Cervero and
Kockelman, 1997; Ewing, 1997a; Frank and Pivo, 1995; Handy, 1996; and Kockelman, 1997).
There have also been a variety of empirical works that have shown useful and rigorous ways to
associate particular urban form characteristics with travel behavior (Boarnet and Crane, 1996,
2001; Cervero, 1996, 2002a; Cervero and Kockelman, 1997; Crane and Crepeau, 1998; and
Kockelman, 1997).
The literature on the urban built environment and travel behavior shows that residents
living in neighborhoods characterized by higher densities, better accessibility, and pedestrian68

oriented designs are more likely to engage in non-motorized travel and travel less by automobile
than residents living in neighborhoods characterized by lower densities, poor accessibility, and
automobile-oriented designs (Boarnet and Crane, 2001a, 2001b; Cervero, 2002; Ewing and
Cervero, 2001; Frank and Pivo, 1995; Frank, 2004; Handy, 2005a, 2006a, and 2006b; Saelens et
al., 2003a, 2003b; Vojnovic, 2006; and Vojnovic et al., 2006; 2007; 2013). Initially, a variety of
urban density measures (such as population and employment density, dwelling units, or building
floor areas) were estimated to explain travel (Frank and Pivo, 1995; Levinson and Kumar, 1997;
Cervero, 1996; Cervero and Kockelman, 1997; Boarnet and Crane, 2000b; and Ewing and
Cervero, 2001). For example, the increased concentration in jobs within a short driving commute
time was estimated to reduce daily household VMT (vehicle miles traveled) (Cervero and
Kockelman, 1997). Cervero (1996) also found that people are less likely to drive to work and
more likely to use transit if jobs are located nearby housing, and hence neighborhoods. Boarnet
and Crane (2000) also showed that a greater concentration of commercial structures is associated
with fewer nonwork automobile trips.
While many of these studies have shown significant relationships between density and
trip frequency, average trip lengths, or travel mode split, the links between travel and density
measures should be approached cautiously. Specifically, the results of density and travel analysis
at aggregate levels of analyses—such as a city or a metropolitan area—which dominate existing
literature are limited since average density of a city masks variations in density within this large
geographic unit, and in turn, fails to reveal differences in land-use patterns and design
characteristics between places with the same density (Newman and Kenworthy, 1989; and
Handy, 1996). Moreover, as a travel measure, VMT per capita or household is often low in highdensity places (such as the city of Detroit) mainly because of lower incomes (Crane, 1996).

69

Thus, it is crucial to incorporate density measures with other urban form factors to improve our
understanding of the link between the built environment and travel behavior more accurately.
As a second set of urban form indicators, researchers have used ‘accessibility measures’
(Boarnet and Crane, 2001b; Cervero, 1996; Cervero and Kockelman, 1997; Frank and Pivo,
1995; Ewing and Cervero, 2001; Holtzclaw, 1994; Kitamura et al., 1994; and Messenger and
Ewing, 1996). Accessibility measures include analyses on the diversity of land uses, such as
retail, commercial uses, public institutions, industrial, or employment indicators. Different land
uses are represented in the count, land area, or floor area. The entropy measure is also an
important and popular diversity measure, where low values indicate single-use environments and
higher values reveal more diverse land uses (Frank et al., 2004; and Ewing and Cervero, 2001).
Destination accessibility, as an accessibility measure, represents ease of access to trip attractions.
Handy (1993), for instance, uses the distance between one’s home and the closest store as a local
accessibility measure.
Previous studies have found credible evidence in the links between accessibility and
diversity measures and travel behavior. Cervero (1996), in exploring how land use mix affects
work trip mode, found that people are less likely to travel by automobile to work and more likely
to take transit if commercial, or other non-residential uses, are nearby. Boarnet and Crane
(2001b) also explained, using regressions, the significant link between diverse commercial
concentrations and less frequent non-work automobile trips. Land use diversity has particularly
strong associations with the frequency of walking trips (Ewing and Cervero, 2001). In addition,
mixed land uses are also related to shorter VMT in Ewing and Cervero’s (2001) study.

70

Connectivity measures are another urban built environment variable that is incorporated
in travel behavior literature (Boarnet and Crane, 2001b; Cervero and Gorham, 1995; Ewing and
Cervero, 2001; Handy, 1996; Vojnovic, 2006; Vojnovic et al., 2006). Connectivity measures
represent street network characteristics within an area. Street networks vary from dense urban
grids of highly connected, straight streets to suburban networks of curvilinear streets, cul-de-sacs
and sparse intersections. Connectivity measures include average block size, proportions of four
way intersections, and densities of intersections per areal unit. Ewing and Cervero (2001), for
example, found that VMT is strongly associated with the design metrics of intersection density
and street connectivity.
Individual (or household) travel behavior, however, is not only affected by built
environment characteristics. It is also influenced by socioeconomic and demographic attributes
of individuals or households. Using disaggregate measures of both urban form and individual (or
household) socioeconomic and travel characteristics, therefore, emerges as a critical dimension
to understanding the variations in individual travel. Notice, however, that travel research using
aggregate measures for metropolitan areas, cities or zones does not allow for an exploration of
underlying factors of behavioral conditions, or the mechanisms by which urban form affects
individual decisions on trip making (Handy, 1996). Thus, disaggregate and neighborhood
analyses are more revealing of the complexities in travel behavior by incorporating urban from
measures into a transparent behavioral framework (Crane, 2000). Controlling for independent
influences, such as household income, travel demands, and mode availability, multivariate
disaggregate studies can isolate and explain more precisely the role of individual features of the
built environment on travel (Crane, 2000).

71

Frank and Pivo (1995), for instance, estimated how different aspects of urban form are
significantly related to different types of travel, while Handy (1993) used socio-economic
characteristics of households—such as household income and car ownership—as explanatory
variables to understand travel characteristics. Using regression analyses of individual travel data,
Cervero and Kockelman (1997) also addressed the significant relationship between VMT per
household, mode choice, and land use near a person’s residence after controlling for individual
socio-economic and demographic variables. Other researchers such as Kockelman (1999),
Boarnet and Crane (2001b), Boarnet and Sarmiento (1998), and Crane and Crepeau (1998)
expanded urban from and travel studies by incorporating travel cost variables as choice variables,
and found that trip demands in each travel mode decrease significantly with increasing travel
time and cost.
However, the existing literature has its limitations. First, most of the studies have focused
on specific regional areas in the U.S., and particularly West Coast cities, such as San Francisco,
San Diego, Los Angeles (California) and Seattle (Washington). Thus, the current literature on
urban form and individual travel behavior has offered a limited understanding of urban travel
within the U.S. metropolitan context. Particularly relevant for this research has been the absence
of mid-West cities, many of which are experiencing disinvestment and decline within their
neighborhoods. The Detroit region provides just such an example.
For instance, in metropolitan areas that are experiencing extreme urban decentralization,
the pattern of suburban residents’ longer commutes may be less valid since employment
opportunities increase in the adjacent suburban areas, instead of the central city location.
Furthermore, urban minorities in central cities face longer commutes to suburban employment
locations due to disappearing employment opportunities near their homes in the city (Friedman
72

et al., 1994; Crane, 2000; Kain, 1968; Vojnovic and Darden, 2013b). In addition, a proper
geographic scale is critical to understand how urban form factors influence individuals’
pedestrian travels (Crane 2000). As Utermann (1984) argued, smaller geographic scales (for
example, covering one-quarter mile or so) are the appropriate scale to explore walking trips.
Therefore, the incorporation of disaggregate built environment measures at a finer scale is
fundamental to explain how an individual’s walking trips are influenced by specific urban form
factors within a walking distance from their residential location. This is why it is important to
understand how urban density, land use mix and connectivity operate within a neighborhood
level to answer these questions.
3.1. Research Objectives, Questions, and Hypotheses
This particular chapter seeks to understand the effects of the urban built environment in
promoting pedestrian trips and reducing automobile use. It is hypothesized that people tend to
use non-motorized modes of travel more in urban neighborhoods characterized by higher
densities, mixed land uses and high connectivity. Subsequent research questions are whether
non-motorized travel modes such as walking and cycling were utilized more in the
neighborhoods that are characterized by higher-densities, mixed land uses and high connectivity
while automobile travel was reduced in such neighborhoods for daily activities.
3.2. Built Environment and Travel Research Data and Method
3.2.1.

Built environment and travel research data

To identify differing built environment characteristics in the Detroit region, three urban
form factor categories were selected. These urban form factors were used in the analysis as
independent variables to explain how individual travel behavior, and particularly pedestrian and
73

driving trips, was influenced by the surrounding urban physical environment. First, urban
development density was measured as the number of residential and non-residential structures in
the neighborhood. Next, land use mix indices were created to describe a proportion of a land use
to all land use types, ranging from three to eight categories. Lastly, connectivity was calculated
using the number of 4-way intersections and bus stops.
In the Detroit region, 2 by 2 mile neighborhoods (about a 4 sq. mile area) were selected
with differing built environment characteristics. Two Detroit urban neighborhoods and two
higher density suburban neighborhoods (one in Ann Arbor and in Birmingham) were selected as
the higher density urban development neighborhoods that were also characterized by highly
connected road networks and mixed land uses. In contrast, the lower density suburban
neighborhoods (one from Bloomfield Hills and one from West Bloomfield) were selected as the
neighborhoods representing typical low density urban developments, characterized generally by
a single dominant land use (residential) and disconnected road networks.
An added dimension of this study involved collecting and recording built environment
elements, land uses and bus stops for an area covering half a mile distance from each
neighborhood boundary. This additional process was necessary to obtain built environment
characteristics that were within an area with a walking distance of half a mile from a
respondent’s address. In total, spatial data of built environment characteristics for six 3 by 3 mile
neighborhoods, about 9 sq. miles each, in the Detroit region were created.
To configure the density and land use mix, the entire built environment within the
neighborhoods were digitally created from the 2005 US Geological Survey digital orthophoto
quarter quads datasets (DOQQs) that were obtained from the Michigan State’s Geographic Data

74

Library ( http://www.mcgi.state.mi.us/). Using ESRI ArcGIS 10.1, I digitized each single
structure for the entire area of the six neighborhoods, and I calculated X and Y coordinate for
each structure’s centroid. On average, there were 12,647 building structures in each urban
Detroit neighborhood, whereas the higher density suburban neighborhoods had 12,860 building
structures, and the lower density suburban neighborhood had 5,137 building structures.
To assign a land use type to a single structure, site surveys of the neighborhoods were
carried out to record the land uses. From these field surveys, land uses were categorized into 10
classes; single-family detached homes, semi-detached homes, apartments, townhomes/row
houses, retail, personal services, public institutions, industrial, abandoned, and others. As shown
in table 8, these land uses were grouped into 6 subsets—residential, commercial, public
institutions, industrials, abandonment, and others. Abandoned structures were included in the
dataset in order to capture a specific quality in the neighborhoods; built environments that are
experiencing urban disinvestment and decline.
In terms of the land-use mix, on average, in the urban Detroit neighborhoods, 89.1% of
the total structures were residential, while 95.9% were residential in the lower density suburban
neighborhoods and 93.3% were residential in the higher density suburban neighborhoods (table
9). This is expected given the built environment characteristics of the high density, mixed land
use urban Detroit neighborhoods. However, the Detroit figures are also influenced by the higher
proportion of abandonment (5.0%, 1,266 building structure counts) in urban Detroit; illustrating
the extent to which these neighborhoods are experiencing severe disinvestment and decline.

75

Table 8 Neighborhood built environment characteristics by neighborhood in the Detroit study
region
Higher density
Higher density
Lower density
urban
suburban
suburban
DT1
DT2
AA
BM
BH
WB
URBAN DENSITY with land use types
11,589
10,984
11,908
12,085
3,974
5,875
Residential
(86.9)
(91.9)
(90.4)
(96.4)
(92.9)
(98.0)
Single-family
11,345
10,861
11,204
11,828
3,676
5,497
detached
(85.0)
(90.9)
(85.0)
(94.3)
(85.9)
(91.7)
home
Semi2
8
31
0
21
142
detached
(0.0)
(0.1)
(0.2)
(0.0)
(0.5)
(2.4)
home
199
46
553
128
105
78
Apartments
(1.5)
(0.4)
(4.2)
(1.0)
(4.0)
(1.3)
Townhomes/
43
69
120
129
172
158
Rowhouses
(0.3)
(0.6)
(0.9)
(1.0)
(4.0)
(2.6)
583
454
1171
453
298
115
Nonresidential
(4.4)
(3.8)
(8.9)
(3.6)
(7.0)
(1.9)
108
114
100
106
42
63
Retail
(0.8)
(1.0)
(0.8)
(0.8)
(1.0)
(1.1)
Personal
245
210
787
272
156
31
service
(1.8)
(1.8)
(6.0)
(2.2)
(3.6)
(0.5)
353
324
887
378
198
94
Commercial
(2.6)
(2.7)
(6.7)
(3.0)
(4.6)
(1.6)
Public
215
121
272
75
100
21
institution
(1.6)
(1.0)
(2.1)
(0.6)
(2.3)
(0.4)
15
9
12
0
0
0
Industrial
(0.1)
(0.1)
(0.0)
(0.0)
(0.0)
(0.0)
754
512
0
0
0
2
Abandoned
(5.7)
(4.3)
(0.0)
(0.0)
(0.0)
(0.0)
415
3
100
2
7
2
Others
(3.1)
(0.0)
(0.8)
(0.0)
(0.2)
(0.0)
13,341
11,953
13,179
12,540
4,279
5,994
TOTAL
(100)
(100)
(100)
(100)
(100)
(100)
Connectivity and Bus Stops
4-way
770
574
317
399
75
57
intersection
386
321
372
108
29
51
Bus stops
1. The percentages of each land use category based on the TOTAL urban built
structures within the neighborhood.
*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH:
Bloomfield Hills; and WB: West Bloomfield.
Count
(%)1

76

Table 9 Neighborhood built environment characteristics by neighborhood group in the Detroit
study region
Higher density
Higher density
Lower density
Count
urban
suburban
suburban
1
(%)
URBAN DENSITY with land use types
22,573
23,993
9,849
Residential
(89.1)
(93.3)
(95.9)
Single-family
22,206
23,032
9,173
detached home
(87.7)
(89.6)
(89.3)
Semi-detached
10
31
163
home
(0.0)
(0.1)
(1.6)
245
681
183
Apartments
(1.0)
(2.6)
(1.8)
Townhomes/
112
249
330
Rowhouses
(0.4)
(1.0)
(3.2)
1,037
1,624
413
Nonresidential
(4.1)
(6.3)
(4.0)
222
206
105
Retail
(0.9)
(0.8)
(1.0)
455
1,059
187
Personal service
(1.8)
(4.1)
(1.8)
677
1,265
292
Commercial
(2.7)
(4.9)
(2.8)
336
347
121
Public institution
(1.3)
(1.3)
(1.2)
24
12
0
Industrial
(0.1)
(0.0)
(0.0)
1,266
0
2
Abandoned
(5.0)
(0.0)
(0.0)
424
102
9
Others
(1.7)
(0.4)
(0.1)
25,324
25,719
10,273
TOTAL
(100)
(100)
(100)
Connectivity and Bus Stops
1,299
716
132
4-way intersection
591
480
80
Bus stops
1. The percentages of each land use category based on the TOTAL urban built
structures within the neighborhood.

77

The differing neighborhood qualities become even clearer when the number, and
proportion, of retail establishments in the neighborhoods are compared. In spite of the similar
higher density characteristics, and the similar connectivity in the road network among the urban
Detroit and higher density suburban neighborhoods, there were more retail and commercial
establishments in the higher density suburban neighborhoods than in the urban Detroit
neighborhoods (tables 8 and 9). Categories of land uses are used to calculate land use mix
indices for each road network buffer in order to understand the relationship between the land use
mix and pedestrian (or driving) trips.
With regard to the street connectivity measures, information on 4-way intersections was
also collected. To obtain the number of 4-way intersections in the neighborhoods, I used the
ESRI ArcGIS 10.1 Geometric Network tool to create a shapefile of ‘real node’ points
(intersections). A ‘real node’ is a connectivity measure that represents a network point
connecting two links, as opposed to a ‘dangle node’ with a link such as a dead-end or cul-de-sac
(Dill and Voros, 2007). Of the intersecting nodes, I intentionally chose 4-way intersections as a
more strict measure of connectivity, because a 3-way intersection can be a ‘dangle node,’ which
predominantly exists in low density suburbs with disconnected road networks. There were 1,299
‘real node’ points detected–4-way (or more) intersections—in urban Detroit neighborhoods (770
in Detroit 1 and 574 Detroit 2), and 716 were evident in the higher density suburban
neighborhoods (317 in Ann Arbor and 399 in Birmingham). However, as expected, there was a
much smaller number, only 132 4-way and more intersections in the lower density
neighborhoods (75 in Bloomfield Hills and 57 in West Bloomfield) (see tables 8 and 9).
Lastly, local and regional bus stop information was included in the analysis as a
connectivity measure to public transit. In order to establish a dataset of public transit, the
78

latitudes and longitudes of bus stops in the study area neighborhoods were collected. Two bus
stops on each side of the road were counted as separate bus stops. For two urban Detroit
neighborhoods, the geographic coordinates were obtained from the Detroit Department of
Transportation (www.detroitmi.gov). Using X and Y coordinates, bus stops in the urban Detroit
neighborhoods were digitally located in ArcGIS 10.1. Geographic coordinate information of bus
stops in the Ann Arbor neighborhood was collected from the Ann Arbor Transportation
Authority (http://www.theride.org/). The list of bus stops in the lower density suburban
neighborhoods, Bloomfield Hills and West Bloomfield, were also established digitally from
Google Maps online (https://maps.google.com/), as were the geographic coordinates of each bus
stop.
Table 8 shows clear distinctions between the three neighborhood groups. The four higher
density urban and suburban neighborhoods were characterized with a higher level of built
development structures, 4-way intersections and bus stops. The two higher density suburban
neighborhoods, however, differ from the two Detroit urban neighborhoods in terms of their
greater variety of land uses in the neighborhoods. Again, this was the dimension of the research
design that allows for the assessment of travel behavior in neighborhoods experiencing
disinvestment and decline. Unlike the urban Detroit and higher density suburban neighborhoods,
the two lower density suburban neighborhoods had fewer built structures, little or no variation in
land uses (the neighborhoods were predominantly residential) and fewer 4-way intersections and
bus stops.
To demonstrate the effects of built environment characteristics on individual travel
behavior, I constructed a variety of ‘walkability’ measures within a walking distance from an
individual’s residential address. Walkability is an estimation of how conducive the built
79

environment is to pedestrian activity as an active transportation (Forsyth and Southworth, 2008;
Frank et al., 2005; Frank et al., 2009; Leslie et al., 2007; Moudon et al., 2006; Owen et al., 2004;
and Sallis et al., 2004). I defined a ‘walkable area’ of a half a mile (about 0.8 km) distance from
a survey respondent’s residential address (Ewing, 1998, 2000; O'Sullivan and Morrall, 1996; and
Sallis et al., 2004). With the ArcGIS geocoding process, 1,106 residential addresses of survey
respondents were located. Then, using the ESRI ArcGIS Network Analyst, I created ‘road
network buffers (RNBs)’, each being a service area of a half a mile distance, running in different
in-network directions from a residential location using the actual road network (see figure 29 to
34). The RNBs boundaries are determined by a series of a half mile distances in all road network
directions from each residence, with the half mile distance based on the actual in-network road
distance. The spatial linear data of the actual road network system were downloaded from the US
Census Tiger Line data. Table 10 shows how many RNBs were created for each study area
neighborhood.

80

Figure 29 Two examples of road network buffers (RNBs) in the Detroit 1 neighborhood. The light green shaded area is the road
network buffer. The blue circle with the red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

81

Figure 30 Two examples of road network buffers (RNBs) in the Detroit 2 neighborhood. The light green shaded area is the road
network buffer. The blue circle with the red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

82

Figure 31 Two examples of road network buffers (RNBs) in the Ann Arbor neighborhood. The light green shaded area is the road
network buffer. The blue circle with the red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

83

Figure 32 Two examples of road network buffers (RNBs) in the Birmingham neighborhood. The light green shaded area is the road
network buffer. The blue circle with the red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

84

Figure 33 Two examples of road network buffers (RNBs) in the Bloomfield Hills neighborhood. The light green shaded area is the
road network buffer. The blue circle with red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

85

Figure 34 Two examples of road network buffers (RNBs) in the West Bloomfield neighborhood. The light green shaded area is the
road network buffer. The blue circle with the red outline in the middle of the road network buffer is a residential location of a survey
respondent. Map produced by Jieun Lee.

86

Table 10 Total number and average area of the road network buffers (RNB) by neighborhood
Higher density
urban
DT1
DT2

Higher density
suburban
AA
BM

Lower density
suburban
BH
WB

The number of
122
107
279
195
205
198
RNBs
The average
area of RNB
0.438
0.410
0.357
0.365
0.256
0.245
(sq. mi)
*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH:
Bloomfield Hills; and WB: West Bloomfield.

Total
1,106
0.334

In this study, the total of 1,106 road network buffers (RNBs) were created from
respondents’ residential addresses that were geocoded using street address line data for the entire
survey sample in all six neighborhoods. By looking at the average areas of road network buffers
by the neighborhoods (table 10), it becomes evident that a pedestrian can reach a larger area in
densely developed neighborhoods with connected road networks. For example, the average area
per Detroit urban RNB is larger than the average area per suburban RNB. Also, the average area
of each higher density suburban RNB is larger than the average area of a lower density suburban
RNB. This means that within the RNB—the boundaries of which, once again, are determined by
walking for a half a mile along the road network in different directions from one’s home—a
resident in the Detroit urban neighborhoods can potentially cover an area of 0.44 sq. mile
(Detroit 1) and 0.41 sq. mile (Detroit 2) as opposed to covering an area of 0.36 sq. mile in Ann
Arbor and Birmingham, and 0.25 sq. mile in West Bloomfield and 0.33 sq. mile in Bloomfield
Hills.
Differences in built environment characteristics are also evident when comparing the
average numbers (counts) of structures by land use types (table 11) within these individual half
mile buffer zones. On average, in the RNBs of urban Detroit and higher density suburban
87

neighborhoods, there were substantially more residential structures as well as non-residential
structures within the individual half mile buffer zones. While there were 800-plus residential
structures on average in each RNB in the higher density neighborhoods, there were only about
300 residential structures on average in each lower density neighborhood RNB.
Table 11 Average number of structure by land use type per road network buffers (RNB) by
neighborhood
Average
count (%)
Single-family
detached
home
Semidetached
home

High density
urban
DT1

DT2

Higher density
suburban
AA

BM

Lower density
suburban
BH

WB

891
(82.3)

956
(88.1)

850
(92.8)

782
(93.9)

255
(86.8)

328
(92.7)

0 (0.0)

1 (0.2)

6 (0.6)

0 (0.0)

0 (0.0)

4 (3.2)

Apartments

3 (0.3)

5 (0.5)

21 (2.4)

8 (1.1)

15 (4.8)

0 (0.0)

Townhomes/
Rowhouses

3 (0.3)

5 (0.5)

2 (0.3)

9 (1.1)

14 (3.8)

5 (2.7)

Retail

4 (0.4)

5 (0.6)

4 (0.4)

11 (1.3)

3 (0.8)

1 (0.6)

19 (2.0)

14 (1.5)

24 (2.5)

20 (2.1)

14 (3.1)

1 (0.4)

19 (2.0)

10 (1.2)

8 (0.9)

5 (0.5)

3 (0.6)

0 (0.0)

1 (0.1)

0 (0.0)

1 (0.1)

0 (0.0)

0 (0.0)

0 (0.0)

Abandoned

91 (8.4)

74 (7.5)

0 (0.0)

0 (0.0)

0 (0.0)

0 (0.0)

Others

46 (4.4)

0 (0.0)

0 (0.0)

0 (0.0)

0 (0.0)

0 (0.0)

Personal
service
Public
Institutions
Industrial

*DT1: Detroit 1; DT2: Detroit 2; AA: Ann Arbor; BM: Birmingham; BH: Bloomfield
Hills; and WB: West Bloomfield.

In terms of the variety of housing structures, urban Detroit neighborhood RNBs were
predominantly occupied by single-family detached homes. RNBs in the higher density suburban
neighborhoods offered more diverse housing types, including semi-detached homes, apartments,
88

and townhomes, than the urban Detroit and lower density suburban RNBs. In addition, RNBs in
the lower density suburbs maintained a greater variety of housing types when compared to the
urban Detroit neighborhood RNBs. In the higher density neighborhood RNBs, there were also
more retail and personal service destinations on average. However, urban Detroit neighborhood
RNBs provided fewer commercial opportunities within walking distance of one’s residence
when compared to the higher density suburban neighborhood RNBs.
Finally, it should be noted that there was a stark difference in the presence of abandoned
structures in the RNBs between urban Detroit and both the higher density and lower density
suburban neighborhoods. On average, in the urban Detroit neighborhoods, about 8% of the
structures present within the RNB were abandoned (8.4 in Detroit 1 and 7.5 in Detroit 2),
whereas there were no abandoned structures in the RNBs in the higher density and the lower
density suburban neighborhoods.
3.2.2.

Built environment and travel behavior research methods

In order to understand the effects of different built environment characteristics on an
individual’s pedestrian trips, I utilized OLS regression analysis (table 12). Each built
environment element category (density, land use mix, and connectivity) was applied in the OLS
regression model to depict its relationship to pedestrian—and driving—trip making in the Detroit
region. After being separately estimated for each urban form category, a stepwise regression was
used to test which built environment characteristics are most strongly associated with pedestrian
or driving trips.
The dependent variable, total trip frequency per person on a weekly basis, was collected
for each mode, pedestrian and driving. For the analysis, I chose trip activities that had a specific

89

destination, such as trips to work, shopping, restaurant visits, and personal services. Active
transport has become a particularly important aspect of research in recent years because of its
positive contribution to public health.

90

Table 12 Built environment and individual travel behavior analysis data and method
Built Environment and Individual Travel Behavior Analysis
Model

Dependent variables

Independent variables (predictors)

OLS regression Pedestrian Travel
 Trip frequency
(Detroit
 Mode split
region)
 Trip distance
Driving Travel
 Trip frequency
 Mode split
 Trip distance

Density
 The # of residential structures (Z-score)
 The # of non-residential structures
(excluding abandonments, Z-score)
Mixed land use
 Land use mix index with 3 land use
categories
 Land use mix index with 7 land use
categories
 Land use mix index with 8 land use
categories
Connectivity
 The # of 4-way intersections (Z-score)
 The # of bus stops (Z-score)
Socio-economic and demographic factors
 Income (logarithm transformed)
 Age

The dependent variable, weekly total trip frequency of each mode (pedestrian and
driving), was also transformed using square-root in order to achieve as normal a distribution as
possible. Another dependent variable was cumulative trip distance per person on a weekly basis
by each mode. Again, the cumulative trip distance was transformed using logarithm to avoid the
skewedness in the distribution. In addition to trip frequency and distance, I also used the
percentage of trips of total trips by mode (walking and driving) as a dependent variable. The
‘mode share’ or ‘mode split’ figure is important to measure the effects of neighborhood types
(higher density urban, higher density suburban and lower density suburban) on how individuals
choose a mode of travel (Cervero, 1996; Cervero and Gorham, 1995; Cervero and Kockelman,

91

1997; Ewing et al., 1994; Frank and Pivo, 1995; Friedman et al., 1994; and Kain and Fauth,
1977).
As independent variables, for the urban development density category, I used the number
of residential and non-residential structures within the road network buffer (RNB). Specifically,
abandoned structures were excluded from calculating the number of non-residential structures
since they were not potential destinations to which a pedestrian or driver would travel to. Both
residential and non-residential structure counts were standardized using the z-score to measure
the variation among the samples.
To illustrate how mixed land uses affect an individual’s pedestrian—or driving—trips, I
created three separate Land Use Mix Indices. The land use mix index was developed with areas
of each land use category as the percent of the area covered by impervious surfaces in each RNB
(Frank et al., 2004). This entropy measure reflects the evenness of distribution of several landuse types within the area or region (Frank et al., 2005; and 2006). The index is calculated as
follows using the n categories of land in each region:
(Equation 1)
Land mix =

∑

×

)/

where n is the number of different land use type categories in the area and Pi is the
proportion of land in type i in the area. This land use mix index varies from 0, for homogeneous
or single land use, to 1 for a greatest variety or mixture in land uses.
First, I grouped a variety of land uses into 3 categories (LUM3); residential, commercial,
and public institutions. The residential land use group (1) includes single family homes, semi-

92

detached homes, townhouses and row houses, and apartments. The commercial land use group
(2) includes retailers and personal services. The last land use group includes public institutions
(3). In order to demonstrate whether a variety of residential structures affects individual
pedestrian—or driving—trip making, I created the second Land Use Mix Index with 7 land use
categories (LUM7); single family homes (1), semi-detached homes (2), townhouses and row
houses (3), apartments (4), retail (5), personal services (6), and public institutions (7). Finally, by
adding industrial land use structures, the model can afford to depict the significance, or
insignificance, of industrial land uses in promoting pedestrian or driving trips. Thus, the last
Land Use Mix Index is with 8 land use categories (LUM8), including industrial structures in
addition to all land uses for Land Use Mix Index 7. Here, each land use structure was counted as
frequency instead of impervious land use area. Thus, it was assumed that each land use count
was equal to land use area.
As for the connectivity measure, the number of 4-way intersections and bus stops were
used in the model. Both intersection and bus stop counts were standardized using the z-score. In
this way, each RNB in the Detroit region can be assigned with a specific position in terms of the
connectivity measure distribution in the Detroit region.
In addition to the built environment factors, two sociodemographic variables were
included in the model. Sociodemographic variables are particularly important to understand the
differences in individuals’ travel choices and demand (Boarnet and Crane, 2001b; Crane, 1996,
2000; Handy, 1993, 1996; Handy et al., 2005; and Holtzclaw, 1994). Income and age were added
to the regression analyses to control the independent influences of behavioral conditions.

93

3.3. Urban Density
3.3.1.

Urban density and non-motorized travel

With respect to travel mode, as table 13 shows, residents in the Detroit region were more
likely to travel as pedestrians for their daily activities if there were more residential and nonresidential structures within a walking distance of half a mile. While both residential and nonresidential structure densities are significantly associated with more walking trips, nonresidential structure density was more strongly related to walking trips throughout the Detroit
region (B= 2.741, p-value = 0.002 for residential density, and B= 3.725, p-value = 0.000 for nonresidential density). Also, survey respondents were likely to significantly increase the frequency
of walking if both residential and non-residential densities increased within the walking distance
buffer zone.
Similar patterns are evident between mode of travel and trip frequency. Non-residential
density (B= 0.186, p-value = 0.000) is more strongly associated with walking trip frequency than
residential density (B= 0.069, p-value = 0.046). Thus, the results illustrate that in the Detroit
region, walking trips for employment, shopping, restaurant visits, and personal services are more
influenced by non-residential density in the neighborhoods. This pattern is consistent to findings
in the existing literature which maintain that urban density is positively related to pedestrian
trips, and the number of non-residential structures in the neighborhood influences pedestrian
trips more strongly (Frank and Pivo, 1995; Levinson and Kumar, 1997; Cervero, 1996; Boarnet
and Crane, 2001a, 2001b; and Ewing and Cervero, 2001).

94

Table 13 Urban density and individual pedestrian travel in the Detroit region
Density
Dependent variables
Independent
N
Coefficient S.E
p-value
variables
(predictors)
Pedestrian Percentage
Constant
920
52.291
12.166
.000
travel
of trip
Residential3
2.741
.866
.002
Non3.725
.770
.000
residential4
Income5
-3.051
1.091
.005
Age
-.134
.050
.008
Trip
Constant
955
1.184
.489
.016
1
3
frequency
Residential
.069
.035
.046
Non.186
.032
.000
4
residential
Income5
.001
.044
.977
Age
-.010
.002
.000
Trip
Constant
269
.775
1.009
.443
2
3
distance
Residential
-.046
.066
.485
Non-.011
.051
.828
residential4
Income5
.051
.094
.589
Age
.002
.005
.609
1.

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance (mile) for work, shopping, restaurant visits,
and personal services
3. Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

Trip distances reveal another set of patterns shaped by residential and non-residential
density. Greater residential and non-residential densities decrease pedestrian trip distances. Thus,
Detroit region residents walked shorter distances to reach a potential destination if there were
more residential and non-residential structures in the road network buffer. This relationship
between shorter walking distances and increased urban density, however, were not statistically
significant. The results, thus, offer evidence that the likelihood of walking more frequently
95

increases due to shorter distances between destinations, as neighborhoods are more densely
developed, a pattern consistent with the existing literature on the built environment and physical
activity (Cervero, 1996; Cervero and Kockelman, 1997; Handy, 1993; and Ewing and Cervero,
2001).
3.3.2.

Urban density and motorized travel

In the Detroit region, survey respondents were less likely to travel by driving if there
were more residential and non-residential structures within walking distance (table 14). Both
residential and non-residential structure densities are negatively associated with percentage of
driving trips, and at a significant level (B= -4.462, p-value = 0.000 for residential density, and B=
-4.046, p-value = 0.000 for non-residential density). However, non-residential density is slightly
more associated with decreases in driving trips. Both residential and non-residential densities
also negatively influenced survey respondents’ driving trip frequency in the Detroit region.
Therefore, residents decreased the frequency of driving trips in the neighborhoods with greater
residential and non-residential structures present. Only residential density, however, was
significantly associated with a decrease in driving trip frequency (B= -0.204, p-value = 0.000).
People in Detroit region also drove fewer miles if there were more residential and nonresidential structures present in the RNB (road network buffer), which, once again, was based on
an in-networked walking distance of .5 mile. While both densities were negatively associated
with driving trip distance, only residential density significantly affected the miles that survey
respondents traveled for their daily activities (B= -0.255, p-value = 0.000). Accordingly, as
previous studies have shown, the results also displayed the pattern of less driving trips in densely
developed neighborhoods (Ewing, 1997a).

96

Table 14 Urban density and individual driving travel in Detroit region
Density
Dependent variables

Driving
travel

Percentage
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
Income5
Age
Constant
Residential3
Nonresidential4
Income5
Age
Constant
Residential3
Nonresidential4
Income5
Age

N

Coefficient

920

S.E

p-value

2.521
-4.462
-4.046

13.697
.975
.867

.854
.000
.000

955

6.908
.135
-.655
-.204
-.052

1.228
.057
.547
.039
.036

.000
.017
.231
.000
.147

864

.383
-.012
1.719
-.255
-.014

.049
.002
.636
.044
.038

.000
.000
.007
.000
.723

.219
-.016

.057
.003

.000
.000

1.

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance (mile) for work, shopping, restaurant visits,
and personal services
3. Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

97

3.4. Mixed land use
In the Detroit region, it is evident that mixed land uses were positively associated with
more frequent pedestrian trips. With respect to driving trips, mixed land uses were negatively
related to car use. However, different land use mix indices—LUM3, LUM7, and LUM8—
demonstrated the differing effects of the diverse mix of land uses on individual’s pedestrian and
driving travel patterns.
3.4.1.

Mixed land use (LUM3) and non-motorized travel

In the analysis, LUM3 was used to understand how land uses—specifically residential,
commercial, and public institutions—in the neighborhoods shape individual travel behavior. As
shown in the table 15, survey respondents were likely to significantly increase the percentage of
walking trips as part of weekly total trips if there was greater diversity in land use types,
residential, commercial, and public institutions (B= 34.037, p-value = 0.000). Respondents also
tended to walk more frequently to reach a destination in the neighborhood with a greater balance
between the three land use categories. In other words, there is a greater likelihood of walking
when the neighborhood offers more retailers, personal services, and public institutions mixed-in
with the houses (B= 1.535, p-value = 0.000). Moreover, respondents walked shorter distances
between destinations if more diverse land use types were available within walking distance.
However, the relationship between walking trip distance and LUM3 was not statistically
significant.

98

Table 15 Mixed land use (LUM3) and individual pedestrian travel in Detroit region
Mixed land use, LUM3
Dependent Independent
variables
variables
(predictors)
Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM33
Income4
Age
Constant
LUM33
Income4
Age
Constant
LUM33
Income4
Age

Coefficient S.E

920

955

269

1.

60.632
34.037
-4.031
-.161
1.428
1.535
-.034
-.011
.726
-.417
.059
.003

11.910
5.585
1.064
.050
.472
.231
.042
.002
.989
.447
.092
.005

p-value

.000
.000
.000
.001
.003
.000
.417
.000
.464
.352
.520
.577

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with three land use categories (residential, commercial, and public institutions)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

99

3.4.2.

Mixed land use (LUM7) and non-motorized travel

A land use mix index with 7 categories was also used to understand the significance of
more detailed land use categories and their impact on individual travel behavior (table 16). In
addition to commercial structures and public institutions, industrial land use was added to the
model as the seventh category. Residential structures were also separated into 4 categories,
including single family homes, semi-detached homes, townhomes and row houses, and
apartments). Similar to LUM3, LUM7 also significantly increased the pedestrian mode share in
weekly total trips that residents in the Detroit region made (B= 22.885, p-value = 0.000). Thus, a
greater variety in land uses increased the possibility of Detroit region residents walking to reach
their destinations.
LUM7 also has a significantly positive effect on walking trip frequency (B = 1.054, pvalue = 0.000). That is, the Detroit residents walked more frequently when there was a greater
balance in these 7 land use categories in the RNB. Interestingly, they walked a longer distance
with a greater LUM7 in the walkable area (RNB). It is likely because a greater variety in
‘residential’ land uses does not in itself necessarily mean a greater land use mix (residential,
commercial and industrial) that facilitates shorter walking distances to reach destinations for the
resident’s daily needs. In addition, one can also assume that the presence of industrial structures
within the RNB does not have the significant effect of shortening walking distances to most daily
destinations, including shopping, restaurant visits, and personal services. The positive effect of
LUM7 on walking trip distance, however, is not statistically significant.

100

Table 16 Mixed land use (LUM7) and individual pedestrian travel in Detroit region
Mixed land use, LUM7
Dependent Independent
variables
variables
(predictors)
Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM73
Income4
Age
Constant
LUM73
Income4
Age
Constant
LUM73
Income4
Age

Coefficient S.E

920

955

269

1.

70.807
22.885
-4.896
-.163
1.900
1.054
-.075
-.011
.586
.109
.066
.002

11.886
5.633
1.069
.051
.471
.233
.043
.002
.979
.491
.092
.005

p-value

.000
.000
.000
.001
.000
.000
.080
.000
.550
.824
.474
.614

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with seven land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, commercial, public institutions, and industrial)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

101

3.4.3.

Mixed land use (LUM8) and non-motorized travel

Lastly, the LUM8 was introduced to the model with 8 land use categories in order to
explore in more detail the impact in the diversity of ‘commercial’ land uses (table 17). Within
this analysis, retailers, and personal services were separately added into the land use mix index,
in addition to the 4 residential land uses, public institutions, and industrial structures. Similar to
the results in the analysis of LUM3 and LUM7, the Detroit residents tend to increase the
percentage of pedestrian trips for their commutes and chores when the road network buffer
offered a greater variety in the 8 land use types (B = 25.086, p-value = 0.000).
Residents were also likely to walk more frequently with the greater mixture in the 8 land
use categories in the RNB (B = 1.143, p-value = 0.000). Therefore, consistent with the effects of
LUM3 and LUM7 land use mix indices on walking trips, a greater mix of land uses in the
neighborhood increased pedestrian activity. However, residents also walked longer distances to
their daily chores when the neighborhood had a greater mixture in the 8 land use categories.
Again, as explained in the previous section on the LUM7 mix, it is assumed that a greater variety
in ‘residential’ land uses has little influence on creating a shorter walking distance to reach daily
destinations, such as shopping, going to the bank, or dropping off dry cleaning.
The results from the analyses on land use mix and walking trips revealed that people
were likely to walk more—both in terms of the percentage (mode share) of trips and the trip
frequency—if there were more diverse land uses present within walking distance of residents as
potential destinations. However, the effects of the land use mix on walking distance were rather
vague since none of the land use mix indices was significantly associated with walking distance.
Furthermore, LUM3 was the only land use mix measure that reduced walking distance whereas
increases in LUM7 and LUM8 were positively related to walking distance. This is likely because
102

a greater mixture in residential land uses, as already noted, might not necessarily reduce the
distances between destinations that are relevant for completing daily chores, such as going to
work, shopping, or reaching personal service destinations.
Table 17 Mixed land use (LUM8) and individual pedestrian travel in Detroit region
Mixed land use, LUM8
Dependent Independent
variables
variables
(predictors)
Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM83
Income4
Age
Constant
LUM83
Income4
Age
Constant
LUM83
Income4
Age

Coefficient S.E

920

955

269

1.

70.675
25.086
-4.893
-.163
1.896
1.143
-.074
-.011
.586
.125
.065
.002

11.880
5.989
1.068
.051
.471
.247
.042
.002
.979
.520
.092
.005

p-value

.000
.000
.000
.001
.000
.000
.080
.000
.599
.240
.715
.505

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with eight land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, retailers, personal services, public institutions, and industrial)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

103

3.4.4.

Mixed land use (LUM3) and motorized travel

With regard to driving trips, significant effects of increases in mixed land use on reducing
automobile trips were observed. First, with an increase in LUM3—a greater balance between
residential, commercial, and public institutions—resulted in Detroit region residents significantly
reducing the percentage of driving trips as part of weekly total trips (B = -35.658, p-value =
0.000) (table 18). They also tend to drive less frequently for their daily commutes and chores if
the neighborhood offers a greater variety in the 3 land use types (B =-0.658, p-value = 0.012)—
that is a greater variety in residential, commercial and public institution land uses.
In addition, not only did the residents drive fewer times, they drove fewer miles when
there were more diverse destinations available, such as retail outlets, personal services, and
public institutions within the RNB. In addition, the effect of LUM3 on reducing driving miles
was statistically significant (B = -0.840, p-value = 0.003). Thus, in the Detroit region
neighborhoods, consistent with findings in the exiting literature, a greater mix of land uses tends
to discourage driving trips.

104

Table 18 Mixed land use (LUM3) and individual driving travel in Detroit region
Mixed land use, LUM3
Dependent Independent
variables
variables
(predictors)
Driving
Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM33
Income4
Age
Constant
LUM33
Income4
Age
Constant
LUM33
Income4
Age

Coefficient S.E

920

955

864

1.

-12.224
-35.658
8.503
.176
1.490
-.658
.449
-.010
1.089
-.840
.279
-.014

13.549
6.354
1.211
.057
.535
.261
.048
.002
.636
.279
.057
.003

p-value

.367
.000
.000
.002
.005
.012
.000
.000
.087
.003
.000
.000

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with three land use categories (residential, commercial, and public institutions)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

105

3.4.5.

Mixed land use (LUM7) and motorized travel

The LUM72 land use mix also negatively affects driving trips of residents in the Detroit
region (table 19). Similar to LUM3 and driving trips, a greater mix in the 7 land use categories
negatively influenced the percentage of driving trips as part of weekly total trips that were made
by survey respondents (B = -20.350, p-value = 0.002). The Detroit residents were also likely to
travel less frequently by driving if the neighborhood offered a more diverse land use mix within
walking distance. In addition, residents drove fewer miles to reach their destinations for daily
chores if there was a greater evenness in the distribution of the 7 land use categories. However,
neither relationship—effects on driving trip frequency and driving trip distance—was
statistically significant.

2

Land use mix index with seven land use categories (single family homes, semi-detached homes, town homes and
row houses, apartments, commercial, public institutions, and industrial)

106

Table 19 Mixed land use (LUM7) and individual driving travel in Detroit region
Mixed land use, LUM7
Dependent Independent
variables
variables
(predictors)
Driving
Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM73
Income4
Age
Constant
LUM73
Income4
Age
Constant
LUM73
Income4
Age

Coefficient S.E

920

955

864

1.

-23.155
-20.350
9.394
.176
-1.699
-.357
.466
-.010
.846
-.424
.298
-.014

13.528
6.411
1.216
.058
.528
.261
.048
.002
.632
.279
.057
.003

p-value

.087
.002
.000
.002
.001
.171
.000
.000
.181
.129
.000
.000

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with seven land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, commercial, public institutions, and industrial)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

107

3.4.6.

Mixed land use (LUM8) and motorized travel

With a greater LUM83 land use mix, survey respondents were also likely to significantly
decrease the percentage of driving trips as part of their weekly total trips (B = -22.299, p-value =
0.001) (table 20). That is, residents in the Detroit region tend to choose driving less as a means of
transportation for their commutes and daily errands if there were more diverse land uses
available within walking distance. A greater variety in the 8 land use categories within the
walkable area (RNB) also influenced the respondents’ frequency in driving; the greater the mix
of land uses, the less frequently that respondents drove for their daily activities. The Detroit
residents also drove fewer miles for their commutes and daily chores if there was a greater
diversity in the 8 land use types within the RNB. However, the effects of LUM8 diversity on
both driving trip frequency and driving trip distance were not statistically significant.

3

Land use mix index with eight land use categories (single family homes, semi-detached homes, town homes and
row houses, apartments, retailers, personal services, public institutions, and industrial).

108

Table 20 Mixed land use (LUM8) and individual driving travel in Detroit region
Mixed land use, LUM8
Dependent Independent
variables
variables
(predictors)
Driving
Percentage
travel
of trip

Trip
frequency1

Trip
distance2

N

Constant
LUM83
Income4
Age
Constant
LUM83
Income4
Age
Constant
LUM83
Income4
Age

Coefficient S.E

920

955

864

1.

-23.039
-22.299
9.391
.176
-1.696
-.404
.466
-.010
.849
-.478
.298
-.014

13.524
6.818
1.216
.058
.528
.278
.048
.002
.632
.297
.057
.003

p-value

.089
.001
.000
.002
.001
.145
.000
.000
.179
.108
.000
.000

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance for work, shopping, restaurant visits, and
personal services
3. Land use mix index with eight land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, retailers, personal services, public institutions, and industrial)
4. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.600 for all predictors.

109

In summary, for all the land use mix indices, a greater variety in land use mix in the RNB
resulted in greater pedestrian activity as a travel mode, and also resulted in more frequent
pedestrian trips for daily activities. Additionally, the positive relationships were statistically
significant for all three LUM indices. However, with regard to walking distance, only LUM3
was negatively associated with walking trip distance. Furthermore, none of the land use mix
indices was statistically significant in explaining walking trip distance.
All three LUM indices negatively affected driving trips, in terms of percentage of driving
trips, driving trip frequency and trip distance. All LUM indices significantly decreased the
percentage of driving trips as part of weekly total trips made. However, while all three LUM
indices were negatively associated with driving trip frequency and distance, only LUM3 was
statistically significant in explaining both driving trip frequency and driving trip distance.
Subsequently, the degree of the land use mix among the three land use categories (residential,
commercial, and public institutions) demonstrated the strongest relationship between mixed land
use activity and individual travel behavior in the Detroit region.
The results from my analyses consistently support the existing literature, in that a greater
land use mix results in greater walkability, both greater pedestrian mode share and more frequent
walking trips. This indicates that there is a greater possibility to carry out pedestrian trips to
access daily activity when there are more accessible destinations in a neighborhood and more
diverse land use options within walking distance.
The results are also consistent with previous studies that argue that certain land use types
are more relevant and significant in understanding the effects of promoting pedestrian travel as a
form of ‘active transport.’ For example, the presence of industrial land use types in the walkable

110

area does not necessarily mean a greater availability of potential pedestrian destinations for daily
chores. This outcome might have been different, however, if local residents were actually
employed in these factories. Moreover, a greater variety in ‘residential’ land use types has little
influence in decreasing walking distance to reach destination. Finally, the results also found that
reduced automobile use and less driving trips in the neighborhood were associated with a greater
mix of land uses. This, once again, is a research outcome that is similar to previous studies.
3.5. Street Connectivity and Bus Stops
Across the neighborhoods in the Detroit region, the survey respondents were likely to
make more pedestrian trips when the roads were more connected and when more bus stops were
present in the neighborhood. In addition, the respondents tend to drive less with an increasing
number of 4-way intersections and bus stops within walking distance.
3.5.1.

Street connectivity and non-motorized travel

In table 13, more 4-way intersections positively influenced the percentage of pedestrian
trips as a component of total trips that the respondents made on a weekly basis. Thus, the Detroit
residents chose to walk more for daily activities if the roads are connected better with more 4way intersections in walking distance (RNB) (B = 3.159, p-value = 0.002). Residents in Detroit
region were more likely to walk frequently when the number of 4-way intersections increased in
the RNB (B = 0.075, p-value = 0.070). An increased number of 4-way intersections within the
RNB also decreased the distance that the residents traveled as pedestrians to reach a destination.
However, the relationship between 4-way intersection density and walking trip distance was not
statistically significant (table 21).

111

3.5.2.

Street connectivity and motorized travel

Unlike the relationships between 4-way intersection density and nonmotorized travel,
driving trips were discouraged with an increasing number of 4-way intersections in the road
network buffer (table 22). In the Detroit region, the percentage of driving trips as part of weekly
total trips decreased significantly with an increase in 4-way intersections within walking distance
(B = -4.348, p-value = 0.000), that is within the RNB.
Driving trip frequency, in contrast, was negatively associated with an increased number
of intersections (B = -0.177, p-value = 0.000, table 22). Driving trip distance also tends to
decrease with more connected road networks in the road network buffer. However, the
relationship between 4-way intersection density and driving trip distance was not statistically
significant (table 22).

3.5.3.

Bus stops and non-motorized travel

The density of bus stops in the road network buffer positively influenced the percentage
of pedestrian trips as part of total trips that the respondents made during a week (table 21). The
Detroit residents, therefore, were more likely to choose nonmotorized modes for daily activities
if more bus stops were present within the RNB (B = 4.770, p-value = 0.000). With more bus
stops within a walking distance, the respondents in Detroit region also tended to walk more
frequently. The number of bus stops increased significantly the likelihood of more frequent
walking for daily chores (B = 0.215, p-value = .000).
The residents in the Detroit region, however, tend to walk longer distances to a
destination for their daily activities when more bus stops were present within the RNB. This
pattern may be an outcome of fewer urban opportunities—such as retailers and personal
112

services—available in the distressed Detroit urban neighborhoods. Even though more bus stops
were available within their reach, urban Detroit residents still had to walk a longer distance, or
walk to take a bus since there were fewer commercial and institutional destination in their
neighborhood. However, the effect of bus stop density on changes in walking distance was not
statistically significant.
3.5.4.

Bus stops and motorized travel

Increasing density of bus stops, on the other hand, discouraged driving trips in the Detroit
region. As shown in table 15, the percentage of driving trips as part of weekly total trips
decreased significantly with more bus stops within the road network buffer (B = -5.775, p-value
= 0.000). The Detroit residents were likely to drive fewer times when there were more bus stops
present in the road network buffer (B =-0.110, p-value = 0.013, table 22). Not only did the
residents drive fewer times, residents also drove fewer miles to reach a daily activity destination
with more bus stops within the RNB. The effect of the number of bus stops on reducing driving
trip distance was statistically significant (B = -0.274, p-value = 0.000).

113

Table 21 Connectivity and individual pedestrian travel in Detroit region
Street connectivity and bus stops
Dependent variables
Independent
variables
(predictors)
Pedestrian Percentage
Constant
travel
of trip
4-way
intersections3
Bus stops4
Income5
Age
Trip
Constant
frequency1
4-way
intersections3
Bus stops4
Income5
Age
Trip
Constant
distance2
4-way
intersections3
Bus stops4
Income5
Age

N

Coefficient

920

955

269

1.

S.E

p-value

33.836
3.159

12.453
1.014

.007
.002

4.770
-1.493
-.087
.476
.075

.950
1.119
.050
.508
.041

.000
.183
.084
.349
.070

.215
.056
-.008
.603
-.023

.039
.046
.002
1.065
.082

.000
.220
.000
.572
.775

.033
.062
.003

.068
.100
.005

.624
.532
.537

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance (mile) for work, shopping, restaurant visits,
and personal services
3. Z-score of total number of 4-way intersections
4. Z-score of total number of bus stops
5. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

114

Table 22 Connectivity and individual driving travel in Detroit region
Street connectivity and bus stops
Dependent variables
Independent
variables
(predictors)
Driving
Percentage
Constant
travel
of trip
4-way
intersections3
Bus stops4
Income5
Age
Trip
Constant
frequency1
4-way
intersections3
Bus stops4
Income5
Age
Trip
Constant
distance2
4-way
intersections3
Bus stops4
Income5
Age

N

920

955

864

1.

Coefficient

S.E

25.071
-4.348

13.972
1.138

pval
ue
.073
.000

-5.775
5.039
.085
-.085
-.177

1.066
1.255
.056
.570
.046

.000
.000
.130
.882
.000

-.110
.322
-.012
1.969
-.013

.044
.051
.002
.652
.053

.013
.000
.000
.003
.797

-.274
.202
-.018

.047
.059
.003

.000
.001
.000

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance (mile) for work, shopping, restaurant visits,
and personal services
3. Z-score of total number of 4-way intersections
4. Z-score of total number of bus stops
5. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

115

In conclusion, Detroit residents walked more often when there were more 4-way
intersections and more bus stops present within walking distance. They walked shorter distances
to access a destination if roads were more connected. However, more bus stops influenced
positively the walking distance in the neighborhood. The Detroit residents also drove less in
RNBs with more connected road networks and bus stops. They not only drove less frequently,
but also fewer miles for their commutes and daily chores when more 4-way intersections and bus
stops were present in the RNB. This pattern is also consistent to existing literature on built
environment characteristics, connectivity and travel behavior (Cervero and Keckelman 1997; and
Ewing and Cervero, 2001, 2010).
However, the results showed a positive relationship between the number of bus stops and
walking trip distance. In this particular study, this is likely due to specific socio-economic and
physical conditions of the Detroit urban neighborhoods, which have the most bus stops. In other
words, compared to the high-income suburban neighborhoods, the low-income residents in the
urban Detroit neighborhoods were more likely to walk a longer distance, or walk to ride a bus for
their chores, since urban amenity options were limited in their neighborhoods. Furthermore, the
urban Detroit residents walked more miles to take a bus since they used public transit more often
than higher income suburban residents, due to the lower incomes and lower levels of private
automobile access.
3.6. Built environment factors and individual travel behavior
In addition to exploring the effects of each urban form attribute on individual trip
making, a stepwise regression analysis was used to test the relative importance of the different
urban form characteristics and their degree of influence on promoting motorized or non-

116

motorized trips. Therefore, urban density, connectivity measures, and land use mix indices were
incorporated in the model simultaneously. Also, each LUM was included separately into the
stepwise regression in order to identify the differing significance of the land use mix indices in
explaining pedestrian and driving trips.
3.6.1.

Built environment factors and non-motorized travel

Among urban density, land use mix, and connectivity measures, the number of bus stops
and intersections, and LUM3 significantly influenced the percentage of walking trips as part of
weekly total trips that Detroit region residents made (table 23). All three built environment
variables were positively related to increases in walking trip percentages. However, the density
of bus stops within walking distance was most strongly associated with walking trip shares as
part of weekly total trips. Then, the diversity in the three land use types (residential, commercial,
and public institutions) was the second most influential variable in increasing the percentage of
walking trips.
By incorporating LUM7 instead of LUM3, all three urban form factors were still
significantly related to the percentage of walking trips in the Detroit region (table 24). The land
use mix, however, became less important than bus stop and intersection densities in explaining
the changes in walking trip percentages as part of weekly total trips. The same pattern was also
evident in the model in explaining the walking trip percentages in the Detroit region with urban
form variables including LUM8 (table 25). Among the significant relationships, bus stop density
was most strongly associated with walking trip percentages. Next, the density of 4-way
intersections in the RNB was more influential in increasing walking trip percentages than the
land use mix index with 8 land use categories.

117

With regard to pedestrian trip frequency, only the bus stop density and the land use mix
indices were found to be significantly influential. In table 16, the number of bus stops was most
strongly associated with increasing walking trip frequency in the Detroit region, and the LUM3
was second. However, the diversity in the 7 land uses was no longer significant in explaining the
frequency of walking trips in the Detroit region (table 24). Interestingly, the incorporation of
retail and personal service land uses separately in the model reassured the importance of the land
use mix to explain the changes in walking trip frequency (table 25). Therefore, the results
confirm the relative significance of commercial land uses—retail and personal services—to
promote walking.
The analysis, however, produced no significant relationship between urban form factors
and walking trip distance. After controlling for personal income and age, none of the urban
density, land use mix indices, and connectivity measures emerged in the models as a significant
factor in changes in walking trip distance (table 23, 24, and 25).

118

Table 23 The built environment and individual pedestrian travel with LUM3
Dependent variables

Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age

N

Coefficient

920

955

S.E

p-value

33.658

12.400

.007

2.497

1.034

.016

4.237
17.795
-1.615
-.100
.621

.963
5.950
1.115
.050
.483

.000
.003
.148
.045
.199

.205
1.039
.034
-.009

.034
.241
.043
.002

.000
.000
.427
.000

.067
.002

.091
.005

.463
.602

269

1.

Square-root transformed weekly total trip frequency for work, shopping, restaurant visits, and
personal services
2. Logarithm transformed weekly cumulative trip distance (mile) for work, shopping, restaurant visits,
and personal services
3. Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with three land use categories (residential, commercial, and public institutions)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

119

Table 24 The built environment and individual pedestrian travel with LUM7
Dependent variables

Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age

N

Coefficient

920

955

269

1.
2.
3.

S.E

p-value

34.714

12.415

.005

3.041

1.012

.003

4.467
14.949
-1.722
-.098
.782

.953
5.527
1.118
.050
.484

.000
.007
.124
.049
.106

.237

.033

.000

.019
-.009
.586

.043
.002
.977

.657
.000
.550

.067
.002

.091
.005

.463
.602

Square-root transformed weekly total trip frequency for daily activities
Logarithm transformed weekly cumulative trip distance (mile) for daily activities
Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with seven land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, commercial, public institutions, and industrial)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

120

Table 25 The built environment and individual pedestrian travel with LUM8
Dependent variables

Pedestrian Percentage
travel
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age

N

Coefficie
nt

920

S.E

p-value

34.751

12.412

.005

3.038

1.011

.003

955

4.443
16.377
-1.731
-.099
.783

.954
5.886
1.118
.050
.484

.000
.006
.122
.048
.106

269

.236
.839
.019
-.009
.586

.033
.245
.043
.002
.977

.000
.001
.661
.000
.550

.067
.002

.091
.005

.463
.602

1.
2.
3.

Square-root transformed weekly total trip frequency for daily activities
Logarithm transformed weekly cumulative trip distance (mile) for daily activities
Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with eight land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, retailers, personal services, public institutions, and industrial)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

121

3.6.2.

Built environment factors and motorized travel

In the Detroit region, connectivity measures were most important in explaining changes
in driving trips. The density of bus stops and the number of 4-way intersections negatively
influenced the Detroit region residents in selecting driving to access a destination. The number of
bus stops was more strongly associated with a decrease in driving trips as part of weekly total
trips regardless of which land use mix index was incorporated in the models. In other words, the
diverse land uses in the RNB had no influence on the changes in the driving trip percentages as
part of weekly total trips (tables 26, 27, and 28).
In tables 26, 27 and 28, Connectivity was also important in understanding the changes in
driving trip frequency in the Detroit region. However, unlike the driving trip percentage, the
number of intersections was most strongly and only related to driving trip frequency, regardless
of the land use mix indices in the models. The Detroit residents were less likely to drive if there
were more intersections within walking distance from their residential address. Again, there was
no observed impact of the land use mix effect on the change in driving trip frequency in the
Detroit region.

122

Table 26 Built environment and individual driving travel with LUM3
Dependent variables

Driving
travel

Percentage
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM37
Income8
Age

N

Coefficient

920

955

864

1.
2.
3.

S.E

p-value

25.071

13.972

.073

-4.348

1.138

.000

-5.775

1.066

.000

5.039
.085
-.203

1.255
.056
.570

.000
.130
.722

-.243

.038

.000

.328
-.011
2.183
-.150

.051
.002
.635
.047

.000
.000
.001
.001

-.214

.044

.000

.183
-.018

.057
.003

.001
.000

Square-root transformed weekly total trip frequency for daily activities
Logarithm transformed weekly cumulative trip distance (mile) for daily activities
Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with three land use categories (residential, commercial, and public institutions)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

123

Table 27 Built environment and individual driving travel with LUM7
Dependent variables

Driving
travel

Percentage
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM77
Income8
Age

N

Coefficient

920

955

864

1.
2.
3.

S.E

p-value

25.071

13.972

.073

-4.348

1.138

.000

-5.775

1.066

.000

5.039
.085
-.203

1.255
.056
.570

.000
.130
.722

-.243

.038

.000

.328
-.011
2.183
-.150

.051
.002
.635
.047

.000
.000
.001
.001

-.214

.044

.000

.183
-.018

.057
.003

.001
.000

Square-root transformed weekly total trip frequency for daily activities
Logarithm transformed weekly cumulative trip distance (mile) for daily activities
Z-score of total number of residential structures (including single-family homes, semi-detached homes,
townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with seven land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, commercial, public institutions, and industrial)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

124

Table 28 Built environment and individual driving travel with LUM8
Dependent variables

Driving
travel

Percentage
of trip

Trip
frequency1

Trip
distance2

Independent
variables
(predictors)
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age
Constant
Residential3
Nonresidential4
4-way
intersections5
Bus stops6
LUM87
Income8
Age

N

Coefficient

920

955

269

1.
2.
3.

S.E

p-value

25.071

13.972

.073

-4.348

1.138

.000

-5.775

1.066

.000

5.039
.085
-.203

1.255
.056
.570

.000
.130
.722

-.243

.038

.000

.328
-.011
2.183
-.150

.051
.002
.635
.047

.000
.000
.001
.001

-.214

.044

.000

.183
-.018

.057
.003

.001
.000

Square-root transformed weekly total trip frequency for daily activities
Logarithm transformed weekly cumulative trip distance (mile) for daily activities
Z-score of total number of residential structures (including single-family homes, semi-detached
homes, townhomes and row houses, and apartments)
4. Z-score of total number of non-residential structures excluding abandonments (including retailers and
personal services, public institutions, and industrial structures)
5. Z-score of total number of 4-way intersections
6. Z-score of total number of bus stops
7. Land use mix index with eight land use categories (single family homes, semi-detached homes, town
homes and row houses, apartments, retailers, personal services, public institutions, and industrial)
8. Logarithm transformed family-structure adjusted personal income
For all models, the tolerance values were above 0.500 for all predictors.

125

The changes in driving trip distance, however, were significantly influenced by urban
density and connectivity measures. The density of bus stops was most strongly associated with a
decrease in driving trip distances, no matter which land use mix index was used in the model
(tables 26, 27, and 28).
A condition unique to driving trip distance, the residential density within the RNBs
negatively influenced driving trip distance. In other words, the Detroit residents were likely to
reduce automobile travel miles if the neighborhood was densely developed with an increasing
number of residential units. In fact, driving trip distance was the only travel behavior measure
that was influenced by urban density.
In conclusion, both pedestrian and driving trips were most influenced by connectivity, the
4-way intersection density, and by the number of bus stops within walking distance of one’s
residence. However, the respondents in the Detroit region were more likely to choose a
pedestrian mode of travel for daily activities if the neighborhood offered more diverse land uses
as potential destinations in addition to connected road networks and more bus stops. In addition,
commercial land uses—retailers and personal services—seemed more influential in increasing
the walking trip frequency. Moreover, among the land use mix indices, LUM3 was most strongly
associated with both the walking trip percentage and the walking trip frequency in the Detroit
region. However, there was no significant change in walking trip distance by urban form factors.
Driving trips were also influenced strongly by street connectivity and bus stops. The
Detroit residents were less likely to choose driving for daily activities if there were more bus
stops and more 4-way intersections present within walking distance, that is, within the RNBs.
Residents also tended to drive less frequently if road networks were connected better with more

126

4-way intersections within walking distance. A greater street connectivity and presence of bus
tops was also important in decreasing driving trip miles. Detroit residents tend to reduce driving
trip miles substantially when more bus stops were present within walking distance. They were
also likely to drive fewer miles if there were more residential structures in the adjacent
neighborhood. Unlike pedestrian trips in the Detroit region, the degree of land use mix had no
effect on the change in driving trip percentage, frequency, and/or distance.
3.7. Conclusion and Discussion
As explored in the existing literature on the built environment and transportation patterns,
potential links between urban form in the Detroit region and individual travel behavior were
examined in this chapter. Previous studies have found significant causal links between built
environment characteristics and travel behavior using multivariate analysis with strong
behavioral foundations (Boarnet and Crane, 2001a, 2001b; Cervero, 1996; Cervero and
Kockelman, 1997; Crane, 2000; Crane and Crepeau, 1998; Ewing, 1997a; Frank and Pivo, 1995;
Handy et al., 2005; Kitamura et al., 1994; Kockelman, 1999; and Messenger and Ewing, 1996).
Since the majority of the travel behavior literature has focused on areas in the U.S. West
Coast, studies have paid little to no attention to urban travel patterns in metropolitan areas that
are experiencing substantial population loss and urban disinvestment, such as in the case of
Detroit or other mid-Western cities (Friedman et al., 1994; Crane, 2000; Vojnovic, 2006; and
Vojnovic et al., 2006). Furthermore, a reconstitution of urban form factors into an appropriate
scale—such as road network buffers representing not only the unique walkable areas for each
agent, but also the unique built environment characteristics of each household—is crucial to

127

examine the effects of built environment characteristics on pedestrian travel (Crane, 2000; and
Utermann, 1984).
The other advantage of this research is the utilization of disaggregate analysis at a finer
scale of road network buffers (RNB). Disaggregate analysis permits the models to predict the
effects of the built environment on individual travel behavior more accurately by controlling for
socio-economic and demographic characteristics, variables that are also important in influencing
travel demand (Cervero and Kockelman, 1997; Crane, 2000; and Handy, 1993, 1996a).
To estimate change in travel behavior, a multivariate OLS regression analysis was
applied to answer how a variety of urban form factors influence individual’s pedestrian and
driving trips. Three trip measures, including mode split, trip frequency and trip distance
(length)—for daily activities such as commutes to work, shopping, restaurant visits, and personal
services—were used as the dependent variable. Three urban form dimensions were incorporated
in the models as the independent variables; urban density, mixed land use, and connectivity. In
order to synchronize the geographic scale of both travel measures and urban form measures, road
network buffers (RNBs) were created using ArcGIS 10.1 Network Analyst with a walking
distance of about a half a mile from a respondent’s residential address. This allowed the models
to capture the specific built environment characteristics around each of the 1,106 residence.
For the urban density measures, both residential and non-residential structure densities
within the RNB were determined. For the land use mix measures, 3 land use mix indices (LUMs)
were calculated using an entropy measure. A range of 3 to 8 land use types were separately
included in the land use mix indices to evaluate the relative importance of each land use type.
Finally, the number of 4-way intersections and the number of bus stops within the RNB was used

128

in the models. In addition to the urban form indicators, socio-economic and demographic
characteristics—personal income and age—were included as controlling factors in the analysis.
The Detroit region residents were more likely to walk to access a daily activity
destination if the neighborhood was more densely developed. Both residential and nonresidential structure densities were positively related to the percentage of walking trips as part of
weekly total trips and at a significant level. Urban density also positively influenced the
frequency of walking trips after controlling for socio-economic and demographic factors. While
both residential and non-residential densities significantly affected the walking trip share and
walking trip frequency, non-residential density was more strongly associated with walking trips.
Residents also walked shorter distances to reach their daily chores with a higher density
in urban structures; however, the relationships were not statistically significant. In the
neighborhoods that were more densely developed, the survey respondents were likely to use
automobiles significantly less for their daily activities. They not only chose to use automobiles
less, but they also drove fewer times and fewer miles to reach destinations. Among the urban
density measures, it was residential density that had the strongest influence in reducing
automobile travel.
In the Detroit region, residents were also more likely to walk for daily activities in the
neighborhoods with more diverse land uses. All 3 land use mix indices positively influenced
walking trip percentage and frequency, that is the respondents tended to choose walking more as
a transportation mode if there were more diverse destination options within walking distance of
their home. Also, they were likely to walk more frequently in such neighborhoods. All 3 land use

129

mix indices were significantly related to waking trip percentage and walking frequency after
controlling for income and age.
In terms of walking distance, however, only LUM3 (a mixture of residential, commercial,
and public institutions) was negatively associated with walking distance. In other words, there
was no effect on reduced walking distance between destinations in cases when residential land
uses were separated into different types (single-family housing, apartments, rowhouses, and so
on). Additionally, the inclusion of industrial land uses also had no impact on changes in walking
distance
Greater diversity in land uses in the road network buffer tends to also discourage
automobile use in the Detroit region. The negative relationships between the driving trip
percentage, selecting the car as mode of travel, and all 3 land use mix indices were statistically
significant. More mixed land uses within walking distance also negatively affected the frequency
and distance of driving trips. However, given that only LUM3 was statistically influential in
defining driving trip frequency and distance, it confirms that a greater balance between
residential, commercial, and public institutions, as part of the land use mix, was most important
in both reducing automobile use and promoting pedestrian trips.
Connectivity measures also increased the walking trip percentage and walking frequency
in the Detroit region. The survey respondents were more likely to choose a pedestrian mode for
daily activities if the roads were better connected with more 4-way intersections and if bus stops
were located within walking distance. Bus stop density in the RNB was more strongly associated
with an increase in walking trips, both percentage and frequency, than 4-way intersection
density. However, only the 4-way intersection density was negatively related to walking trip

130

distance. In other words, residents in the Detroit region walked longer distances to reach a
destination in the neighborhood with more bus stops. Hence, the pattern of longer walking
distance in the neighborhood with more bus stops might support the argument that there were
fewer potential destinations available within walking distance in the Detroit urban
neighborhoods. The pattern might also confirm that Detroit urban residents walked to take public
transit more often than residents in the higher income suburban neighborhoods.
Greater connectivity within the walking buffer zones also negatively influenced
automobile use. The Detroit residents were less likely to choose an automobile as a form of
transportation if the neighborhood was more connected with greater intersection densities and
had more bus stops. Residents also drove less frequently and drove fewer miles to access a daily
activity destination if there was a higher concentration of 4-way intersections and if there were
more bus stops in the ½ mile buffer zones. However, the relationship between 4-way intersection
density and driving trip distance was not statistically significant.
Finally, I used stepwise regression analysis to test the relative strength of each of the
urban form factors in influencing individual travel behavior. The Detroit residents were more
likely to choose a pedestrian mode if there were more intersections, more bus stops and diverse
land uses. The percentage of walking trips as part of weekly total trips was most strongly
associated with bus stop density. Four-way intersection density and mixed land uses were also
significantly related to walking trip share. However, unlike LUM7 and LUM8, LUM3 influenced
the walking trip percentage to a greater extent than 4-way intersection density. Walking trip
frequency was also most strongly associated with bus stop density regardless of land use mix
indices. Among the land use mix indices, only LUM3 and LUM8 were significantly related to an
increase in walking trip frequency. Thus, the results confirmed the relative importance of non131

residential land uses, and particularly commercial land uses, in increasing walking trip
frequency.
In the Detroit region, bus stops and street connectivity significantly influenced decreased
levels of automobile travel. Specifically, the number of bus stops was more strongly associated
with reduced automobile use for daily activities than the number of 4-way intersections. The
Detroit residents were also likely to drive less frequently for their daily activities if the roads
were more connected with a greater intersection density within walking distance, within their
RNBs. They also drove fewer miles if there were more bus stops present within walking distance
of their home. Driving trip distance was negatively associated with increasing residential density.
In other words, residents tend to use an automobile less frequently and they drove fewer miles if
they lived in densely developed neighborhoods with more bus stops and connected road
networks. In the Detroit region, the degree of land use mix had a relatively weak influence on
driving trips, since none of the land use mix indices emerged as significant in the stepwise
regression models to explain driving trip percentages, frequency, or distance.
In conclusion, the analysis results confirmed the significant links between built
environment characteristics and individual travel behavior. First, a densely developed
neighborhood of RNBs with diverse land uses, connected road networks, and bus stops increased
pedestrian travel. In addition, automobile use for daily activities was substantially reduced in the
densely developed RNBs with a greater land use mix, more intersections, and bus stops. People
also tend to walk more and drive less if the immediate built environment has a better balance of
land uses—and particularly a mix of residential, commercial, and public institution. Finally,
among the built environment factors, the number of bus stops and 4-way intersections were most
strongly associated with individual travel behavior.
132

The analysis results not only confirm the previously addressed relationships between
urban form characteristics and individual travel behavior, but also expand the understanding of
urban travel with more elaborate dimensions in neighborhood characteristics, such as disparities
in the neighborhoods and extreme divergences in income status. Unlike the existing studies, the
results explored the relationship between built environment and travel behavior in a region
characterized by a rapidly declining urban core, that of the Detroit region, with extreme urban
decentralization and disinvestments in the central city. Consequently, the results were able to
reveal how individual’s travel was actually affected in neighborhoods experiencing
disinvestment and decline.
For example, as confirmed in the analysis, a greater walking trip distance associated with
increasing bus stop density may imply a longer distance between fewer destination options in the
Detroit urban neighborhoods where substantially more bus stops were present than suburban
neighborhoods. The greater walking distance in the neighborhoods may also be a result of the
low income status and limited automobile access in the urban Detroit neighborhoods. To meet
their daily needs, the urban Detroit residents had to walk more and longer distances to take
public transit whereas the suburban residents would mostly drive to their more distant
destinations.
The results also reaffirmed the importance of the land use mix in promoting pedestrian
travel and reducing automobile use. Three land use indices (LUM3, 7, and 8), with different
numbers of land use types were employed in the analyses. From the different indices, the
commercial land uses—retail and personal services—were most strongly associated with greater
pedestrian activity and reduced automobile trips. A greater balance between residential,
commercial, and public institutional land uses (i.e. LUM3) was more important than a greater
133

mixture in housing types, or the presence of industrial land uses in the neighborhood (i.e. LUM7
or LUM8), in increasing pedestrian trip activity and in decreasing automobile use.
However, similar to the findings in the existing literature, the models explain poorly the
entire variation in travel behavior in that the R2 was generally low, around 0.2 or lower for the
models. Thus, other urban form measures, such as the shortest distance to a store for shopping
trips, may enhance the explanatory power. Also, in addition to built environment characteristics,
the neighborhood quality measures will also improve the model. For example, incorporating
neighborhood crime rates or street aesthetic measures will offer more elaborate explanations to
individual travel behavior and choice. Subsequently, using more socio-economic or demographic
factors may also improve the model. Household structure or car ownership, for example, may
offer a better explanation of individual or household travel demand.
Furthermore, the models explain pedestrian trip distance poorly. None of the urban from
factors was significantly associated with walking trip distance in the Detroit region. As noted
above, the incorporation of pedestrian environment quality measures may help the model
understand the variation in walking trip distance more precisely. Moreover, the scarcity of
potential destinations within walking distance probably deterred one’s desire to choose a
pedestrian mode for daily chores. Additional consideration, such as who walked to where and for
what purpose, may also be useful to understand the actual variation in walking trip distance.
Finally, in the model, there was no distinction made between work trips and nonwork
trips. Trip activities such as commutes to work, shopping, restaurant visits, and personal services,
were included in the model to capture pedestrian travel behavior. However, travel choices and
travel demand could differ between work trips and nonwork trips, since individuals have more

134

flexibility in decisions for nonwork trips compared to work trips. Also, in constantly
decentralizing metropolitan areas like Detroit, the conventional commute patterns may be less
valid in that urban minorities travel often reversely, with a longer distance to suburban
employment locations. Including locational factors (e.g. destinations) of work and nonwork trips
might improve the understanding of urban travel within the Detroit metropolitan context.
In future research approaches, therefore, work trips and nonwork trips can be explored in
separate models. Even further separation of nonwork trips into separate trip purposes could more
precisely reveal the patterns in individual travel behavior. Furthermore, individual
characteristics, such as gender and car ownership, or household characteristics such as marital
status, might also further explain individual travel behavior choices and travel behavior demands.
Some of these issues, in fact, will be addressed in subsequent chapters. Also, additional
information on travel destinations can be added in the model, such as a dummy variable
distinguishing between a central city or a suburban location. This way, the model can examine
not only the travel flows within the metropolitan region, but also exhibit the travel burdens of
racial or ethnic minorities in declining urban cores, including reverse commutes and longer
shopping trip distances to suburban retail locations.

135

CHAPTER 4: Neighborhood Typology and Travel Behavior in
the Detroit region
In chapter 4, I will discuss the value of defining the neighborhood typology in examining
individual travel behavior. Traditionally, urban transportation researchers have defined the
relationships between socio-economic and demographic characteristics of population sub-groups
and their travel patterns (e.g., trips to work or non-work related activities) aggregated at the
county, city or traffic analysis zone (TAZs)4 levels (Krizek 2003; Manheim, 1979; Martin et al.,
1967; Martin and McGuckin, 1998; McNally, 2000). While aggregate patterns of travel by
certain sub-populations provides important information about zonal averages of trip generation,
such aggregate measures obscure the actual travel behavior and subsequent demands of
particular sub-populations (Krizek, 2003; Handy 1996). Thus, in order to understand the effects
of socio-economic and demographic characteristics on individuals’ travel behavior, it is also
useful to explore travel patterns at a finer scale, such as individual-, household-, or neighborhood
scales of analysis.
In fact, a particular area of interest that is under-represented in existing travel behavior
literature is the analysis into travel for population sub-groups located in declining
neighborhoods, i.e., neighborhoods that are experiencing population loss, disinvestment and
economic decline. With a disappearing population and a loss in urban amenities, it is particularly
important to understand how certain socio-economic and demographic factors, such as gender,
income and access to a private automobile affect daily travel in terms of commutes to work,
accessing food, or reaching leisure destinations in declining neighborhoods. Even if one lives in
4

A traffic analysis zone is a spatial unit that is commonly used in conventional transportation planning models.
TAZs are constructed by census block information that contain socio-economic data such as the number of
automobiles per household, household income, and employment within these zones (Miller, Harvey J. and ShihLung Shaw. (2001) Geographic Information Systems for Transportation, Oxford University Press US. p. 248)

136

a higher density neighborhood that is highly connected and maintains a mix of land uses—a
neighborhood typology that is expected to increase accessibility to daily destinations—
individuals’ travel patterns might vary considerably depending on the destination opportunities
within the neighborhood. In addition, an individual’s options of travel modes (e.g., being forced
to walk or to take public transit or having access to a private automobile) can also shape the
nature of travel.
Exploring individuals’ travel behavior within a neighborhood—instead of, for instance,
aggregating individual trip patterns into a traffic analysis zone—is also important in identifying
how certain characteristics in the neighborhood environment influence travel, and pedestrian trip
making in particular (Hanson, 1996). For instance, perception of fear from crime or the intensity
of traffic volume varies in differing urban neighborhood environments; people in declining
inner-city neighborhoods have a greater sense of fear from crime compared to suburban
neighborhoods with similar built environment characteristics. Thus, exploring individuals’ travel
behavior within specific neighborhood characteristics can have a significant implication on urban
design and planning, if practitioners are to build neighborhoods with balanced transportation
options for their residents.
Lastly, the disaggregation of travel behavior into a comprehensive array of diverse travel
activities, and particularly differentiating by gender, offers meaningful information on the
complexity of women’s daily travel based on their socio-economic status and their multiple roles
within the household (Freedman and Kern, 1997; Hamilton and Jenkins, 1990; Hanson 1995;
Hanson and Hanson, 1980a, b; Hanson and Johnston, 1985; Hanson and Giuliano, 2004; Hu and
Young, 1999; Jones, 1990; Johnston-Anumonwo, 1992; Kwan, 1999a, 1999b, 2000; McFadden
and Reid, 1975; Schintler et al., 2000; Santos et al., 2011; Pickup, 1985, 1990). Hence,
137

measuring the easiness of reaching various destinations within the neighborhood is crucial to
identifying the travel burdens of women as they accomplish their daily household and
employment responsibilities. By disaggregating travel activities at a neighborhood level, the
differences in trip making between women and men can be more accurately explored. Similarly,
such disaggregation can allow for differences among women with different socio-economic and
demographic profiles to be explored more precisely.
4.1. Research Objectives, Questions, and Hypotheses
This chapter aims to identify the neighborhood typology that helps explain differences in
travel behavior between urban and suburban neighborhoods. First, it is hypothesized that
residents living in neighborhoods with greater income levels and automobile access tend to travel
more frequently and longer distances by car. The second hypothesis is that residents living in
neighborhoods with more diverse urban opportunities and safer pedestrian environments will
have improved access to daily activities when compared to neighborhoods with similar urban
density and connectivity characteristics. Hence, subsequent research questions are which socioeconomic and demographic factors significantly influence total trip generation and which
neighborhood group, among urban Detroit, higher-density suburban, and lower-density suburban
neighborhoods is differentiated against others in terms of socio-economic and demographic
characteristics, travel patterns, and neighborhood quality for pedestrians.
4.2. Neighborhood Typology Analysis Data and Method
In order to address the social dynamics of individuals’ travel behaviors, I performed three
different analyses: (1) OLS regression for modeling trip generation among the general population
in the Detroit region; (2) discriminant analyses for discerning neighborhood groups based on
138

socio-economic and demographic characteristics, travel patterns, and built environment
typologies; and (3) descriptive analyses for illustrating the mean difference in disaggregate travel
activities by the different neighborhood typologies (see table 29).
First, in order to identify the effects of socio-economic and demographic factors on
individual trip making, it is important to select predictors that explain the general travel patterns
among the population. Using Ordinary Least Squares (OLS) regression, I examined the specific
socio-economic and demographic variables that have significant explanatory power to explain
travel patterns. Second, three discriminant analyses showed the differing characteristics of the
urban neighborhood density groups in terms of socio-economic and demographic profile, travel
patterns and degree of sense of fear as a proxy of differing urban environment quality
characteristics. Lastly, the average trip frequencies and trip distances are discussed for each
travel activity, including trips to work, shopping, restaurant visits, personal services, and leisure
activities.
Defining the neighborhood typology is particularly important in order to understand the
travel burden of individuals throughout a metropolitan region, and particularly in neighborhoods
experiencing disinvestment and decline. Travel opportunities—such as reaching employment,
retail, or personal services—are neither equally distributed nor equally accessible across the
neighborhoods. Income disparities across the neighborhoods also create different utilization of
travel modes. For example, populations in lower-income neighborhoods often travel more by
public transit and less by car. Therefore, disaggregating individual travel behavior within
neighborhoods will provide a better understanding of travel constraints of socially marginalized
populations, and particularly lower-income women in the declining urban neighborhoods, one of
the principal areas of research interest in this dissertation.
139

Table 29 Variables for neighborhood typology and travel behavior models
Models

Dependent Variables

Independent Variables
(Predictors)

4-(1). Aggregate travel patterns
OLS regression

Total trip frequency1
Cumulative trip distance2

Income3
Age4
(All neighborhoods)
 All activities
# of personal vehicles
 Work trip
available
# of dependent children5
 Non-work trip
4-(2). Different populations and their travels in differing urban neighborhoods
Discriminant Analyses

Neighborhood density
groups

(All neighborhoods)






High Urban
High Suburban
Low Suburban

Income3
Age4
# of personal vehicles
available5
# of dependent children6
Total trip frequency1
Cumulative trip distance2
Walking and cycling
Driving

Sense of fear from crime6
Day time and Night time

Sense of fear from traffic6
Day time and Night time

4-(3). Description of travel activities in differing urban neighborhoods
Descriptive Analysis
(Within neighborhoods)

Total trip frequency7
Cumulative trip distance8








All activities
Work trip
Shopping trip
Restaurant trip
Service trip
Leisure trip

1.Weekly total trip frequency by all modes, 2.Weekly cumulative trip distance by all modes, 3.Household
structure adjusted family income, 4.Adults who are 18 years old and older, 5.Children who are younger
than 18 years old, 6.OnLikert scale, 1=none, 2=low, 3=moderate, and 4=high, 7.Weekly total trip
frequency by all modes and each mode (walking and cycling, public transit, and driving), 8.Weekly
cumulative trip distance by all modes and each mode (walking and cycling, public transit, and driving).

140

4.3. Aggregate travel patterns in the Detroit region
Across the Detroit region neighborhoods, the average age was 54 years old with less than
one dependent child under the age of 18 years old in the household. On average, the Detroit
region respondents had access to approximately one car (PVHC = 1.2) and in 2008 maintained a
household income of $52,200 USD (table 30). On average, these respondents traveled 12 times a
week for 45 miles for all activities by all modes. Specifically they traveled 5 times a week for 42
miles for work trips and 10 times per week for 27 miles for non-work trips by all modes (table
30).
Table 30 Descriptive statistics of socio-economic and demographic and travel pattern variables
in the Detroit region
N
Minimum Maximum Mean
Independent variables: socio-economic and demographic characteristics
AGE
1116
18.0
94.0
53.8
1
INCOME
1009
3,338.1 160,000.0
52,201.4
NCHILD
1075
0.0
8.0
0.7
PVHC
1068
0.0
6.0
1.2
Dependent variables: aggregate travel pattern
TTLFREQUENCY2
1087
1.0
76.0
12.1
TTLDISTANCE3
1068
0.0
349.0
45.3
4
WTTLFREQ
620
1.0
23.0
5.0
WTTLDIST5
540
0.1
965.0
42.3
NWTTLFREQUENCY6
1076
1.0
76.0
9.7
7
NWTTLDISTANCE
1058
0.3
349.1
26.6

Std. Dv.
15.8
31,721.4
1.1
0.7
8.7
51.6
2.2
60.6
8.0
33.5

1.
2.
3.
4.
5.
6.

Household structure adjusted family income, USD in 2008;
Total trip frequency per person for all activities by all modes on a weekly basis (trip count);
Cumulative trip distance to destinations for all activities by all modes on a weekly basis (mile);
Total trip frequency per person for work trips by all modes on a weekly basis (trip count);
Cumulative trip distance to destinations for work trips by all modes on a weekly basis (mile);
Total trip frequency per person for non-work trips by all modes on a weekly basis (trip count);
7. Cumulative trip distance to destinations for non-work trips by all modes on a weekly basis (mile)

OLS regression results (table 31) of the aggregate travel pattern analysis for all activities
and work and non-work trips by all travel modes revealed a strong relationship between all
socio-economic and demographic characteristics and trip frequencies (n = 848), as well as
141

cumulative distances (n=834) . Specifically income (B = 0.141, p-value = 0.056 and B = 0.253,
p-value = 0.056) and the number of personal vehicles available (B = 0.177, p-value 0.058 and B
= 0.126, p-value = 0.003) positively predicted the number of trips and distances at a significant
level. In other words, with increasing income and number of personal vehicles available, the
number of trips and cumulative trip distances also increased. With increasing age however, trip
frequency and travel distance decreased (B = -0.015, p-value 0.003 and B = -0.017, p-value =
0.003). Interestingly, with an increasing number of dependent children in the household, trip
frequency declined (B = -0.082, p-value = 0.039) but trip distances slightly increased (B = 0.007,
p-value = 0.039) for all activities by all modes.
For work trips (model II in the table 31), only income explained the number of trips
significantly (B = 0.088, p-value = 0.015); a person with greater income tended to make more
work trips (n = 502). A person with greater income was also more likely to travel longer in miles
and at a significant level (B = 0.217, p-value = 0.041). The increasing age, however, negatively
affected the trip frequencies for work and distance (only significantly for trip distance with B = 0.013, p-value = 0.008) (n = 439). The positive effect of income and the negative effect of age on
trip making continued for nonwork trips (model III in the table 31). For both trip frequency (n =
843) and distance (n = 832) for nonwork trips, older respondents were less likely to make trips
and at a significant level (B = -0.006, p-value = 0.032, and B = -.004, p-value = 0.097
respectively). However, increasing income had a positive effect only on trip distance associated
with nonwork trips at a significant level. In addition, while the effects were not significant in the
models, the number of available personal vehicles also positively affected non-work trip
frequency and distance. From the regression results, income and age were able to explain the
general travel patterns significantly at an aggregate level in the six Detroit region neighborhoods.

142

Table 31 OLS regression results of aggregate travel pattern analysis in the Detroit region
Total trip frequency1
Cumulative trip distance2
Coefficient
S.E.
Coefficient
S.E.
Aggregate travel pattern analysis I: For All activities by All modes
(Constant)
2.511***
.606
1.291**
.609
3
**
***
Income
.141
.056
.253
.056
***
***
Age
-.015
.003
-.017
.003
# of personal vehicle accessible
.117**
.058
.126**
.058
**
# of dependent children
-.082
.039
.007
.039
4
N
848
834
Aggregate travel pattern analysis II: For Work trips by All modes
(Constant)
1.306***
.388
1.119
1.137
Income3
.088**
.036
.217**
.106
***
Age
-.001
.002
-.013
.005
# of personal vehicle accessible
-.025
.033
.135
.091
# of dependent children
.023
.022
.030
.061
N4
502
439
Aggregate travel pattern analysis III: For Nonwork trips by All modes
(Constant)
2.510***
.598
1.788***
.594
3
*
Income
.068
.055
.105
.055
**
*
Age
-.006
.003
-.004
.003
# of personal vehicle accessible
.026
.057
.030
.057
# of dependent children
-.060
.038
-.003
.038
4
N
843
832
1. Total trip frequency per person on a weekly basis (trip count, square root transformed);
2. Cumulative trip distance to destinations per person on a weekly basis (mile, logarithm transformed);
3. Household structure adjusted family income (USD in 2008, logarithm transformed);
4. N = number of resident men and women;
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

143

4.4. Different populations and their travel patterns in the different neighborhoods
To address differing characteristics of neighborhood groups in the Detroit region, I used
discriminant analyses to study the socio-economic and demographic characteristics, travel
patterns, and the degree of sense of fear related to crime and traffic. Variables of fear from crime
and traffic were adopted to represent differing characteristics of the neighborhood environment
that deterred individuals from walking in the neighborhood. Discriminant analysis is used to
define differences among the neighborhood groups with three categories—urban Detroit, higher
density suburban, and lower density suburban neighborhoods. Discriminant analysis involves
identifying the linear combination of attributes, known as canonical discriminant functions,
which maximize group separation (Biles and Pigozzi, 2000; Burns and Burns, 2008; King 1970;
McConnell, 1979).Unstandardized discriminant coefficients (b in the equation 2) maximize the
distance between the means of the dependent variable and standardized discriminant coefficients
predicts the variables that have the strongest discriminatory power between groups (Burns and
Burns, 2008). An example of the discriminate equation or function used in this analysis of travel
behavior is provided below.
(Equation 2)
D = b1X1 + b2X2 + b3X3 + ∙∙∙∙∙∙ + biXi + a
Where D = discriminant function
b = the discriminant coefficient or weight for that variable
X = respondent’s score for that variable
a = a constant
i = the number of independent variables
144

The first discriminant analysis was conducted to examine whether four socio-economic
and demographic characteristic variables—age, income, the number of dependent children and
the number of personal vehicles could predict neighborhood differences (see model I in table
32). As Wilks’ Lambda is the proportion of the total variance in the discriminant scores not
explained by differences among groups, the smaller lambda (at a significant level) indicates that
group means appear to differ significantly. For example, Λ = 0.57, χ2 (8, N = 876) = 493.39, p <
0.01, indicates that overall the predictors differentiated significantly among the three
neighborhood density groups, with an unexplained variance of about 57% by the two functions
combined.
(Equation 3)
Dsocio-demographic1 = (1.474
(-0.098

Income) + (0.008

Age) + (0.564

#of personal vehicles) +

#of dependent children) – 16.699.
Dsocio-demographic2 = (-0.172 Income) + (0.054 Age) + (0.213 #of personal vehicles) +

(0.471 #of dependent children) – 1.627.
In table 32, the within-groups correlations between the predictors and the discriminant
functions were presented as well as the standardized weights. Based on these coefficients, the
income variable (0.900) demonstrates the strongest relationship with the first discriminant
function and the number of personal vehicles (0.391) was the second strongest. Age (0.823) and
the number of dependent children (0.502) showed the strongest relationship with the second
discriminant function. Accordingly, the first and second discriminant functions were named as
‘income’ and ‘life cycle’, respectively.

145

Table 32 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by socio-economic and demographic profile
Correlation
Standardized
coefficients with
Coefficients for
discriminant
discriminant
functions
functions
Function Function
Function Function
1
2
1
2
Discriminant analysis I: Socio-economic and demographic profile grouping
Income1
.905
-.116
.900
-.105
Age2
.085
.852
.115
.823
# of personal vehicle accessible
.429
.157
.391
.147
3
# of dependent children
-.075
.526
-.104
.502
Dependent Variable (categories): urban density neighborhood groups (Urban Detroit neighborhoods,
Higher density suburban neighborhoods, and Lower density suburban neighborhoods)
1. Household structure adjusted family income (USD in 2008, logarithm transformed);
2. Adults who are 18 years old and older;
3. Children who are younger than 18 years old.

Canonical scores of group means (centroids, from figure 35) show that the urban Detroit
neighborhood group (coded as 1) was separated by function 1, for the lower ‘income’ level
compared to the other two suburban neighborhood groups. By function 2, the higher density
suburban neighborhood group (coded as 2) was separated by the earlier ‘life cycle’ stage, with
younger and fewer dependent children. The discriminant functions were able to classify 76.3%
(n = 876, see table 33) of the total survey sample for socio-economic and demographic
characteristic variables.

146

Table 33 Group statistics by socio-economic and demographic characteristic variables
Urban Detroit

Income
Age
# of personal vehicle accessible
# of dependent children

Mean
(Std.Dv)
9.76
(.754)
51.76
(16.757)
0.73
(.786)
0.82
(1.228)

N
192
192
192
192

Higher density
suburban
Mean
N
(Std.Dv)
10.85
366
(.530)
51.62
366
(15.240)
1.29
366
(.655)
0.55
366
(.913)

Lower density
suburban
Mean
N
(Std.Dv)
10.93
318
(.601)
56.58
318
(14.229)
1.38
318
(.671)
0.74
318
(.119)

Figure 35 Separation of groups on discriminant functions by socio-economic and demographic
profile

147

The second discriminant analysis was conducted to examine whether travel pattern
variables—total trip frequency by walking and driving, and cumulative trip distance by
walking/cycling and driving—could predict neighborhood differences (table 34). As Wilks’
Lambda is the proportion of the total variance that is not explained by differences among groups
in the discriminant scores, a smaller lambda at a significant level indicates that group means
appear to differ significantly. Here, Λ = 0.85, χ2 (8, N = 322) = 53.501, p < 0.01, indicates that
overall the predictors differentiated significantly among the three neighborhood density groups
with an unexplained variance of about 85% by the two functions combined. Although total trip
patterns—frequency and distance—for all activities combined explained only 15% of the
variance in travel behavior in the Detroit region, the functions explained the differences between
group means at a significant level.
(Equation 4)
Dtotal travel patterns1 = (-0.920 Total pedestrian frequency) + (-0.029 Total driving trip
frequency) + (0.579 Cumulative pedestrian trip distance) + (0.348 Cumulative driving trip
distance) + 0.213.
Dtotal travel patterns2 = (0.282
frequency) + (0.791

Total pedestrian trip frequency) + (0.187

Cumulative pedestrian trip distance) + (-0.093

Total driving trip

Cumulative driving trip

distance) – 1.968.
In table 34, based on the within-groups correlations and standardized coefficients
between the predictors, the variable of total pedestrian trip frequency for all activities (-0.869)
demonstrates the strongest relationship with the first discriminant function. Both pedestrian and
driving distances for all activities (0.615 and 0.448, respectively) were also strongly associated
148

with the first function. With the second discriminant function, cumulative pedestrian trip distance
(0.840) and total pedestrian trip frequency (0.267) show the strongest relationship. Due to the
strong opposite relationship between pedestrian trip frequency and driving distance, the first
function was labeled as the ‘driving-dominant aggregate travel pattern.’ For the second
discriminant function, the label the ‘pedestrian-dominant aggregate travel pattern’ was given
accordingly.
Table 34 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by travel pattern
Correlation
coefficients with
discriminant
functions
Function Function
1
2
Discriminant analysis II: Travel pattern grouping
-.670
.645
Total trip frequency, pedestrian1
2
.252
.109
Total trip frequency, driving
3
.237
.951
Cumulative trip distance, pedestrian
4
.627
-.046
Cumulative trip distance, driving

Standardized
Coefficients for
discriminant
functions
Function Function
1
2
-.869
-.033
.615
.448

.267
.217
.840
-.119

Dependent Variable (categories): urban density neighborhood groups (Urban Detroit neighborhoods,
Higher density suburban neighborhoods, and Lower density suburban neighborhoods)
1. Total trip frequency for all activities by walking and cycling on a weekly basis;
2. Total trip frequency for all activities by driving on a weekly basis;
3. Cumulative trip distance for all activities by walking and cycling on a weekly basis;
4. Cumulative trip distance for all activities by driving on a weekly basis.

In figure 36, the lower density suburban neighborhood group (coded as 3) was separated
as the ‘driving-dominant travel pattern’ group (by function 1) for the greater canonical score than
was the case with the other two high density neighborhood groups. The urban Detroit
neighborhood group (coded as 1) was, in contrast, separated by function 2, for the greater
canonical score of the ‘pedestrian-dominant travel pattern’ compared to its suburban
counterparts. The discriminant functions, however, were able to classify only 28% (n = 322) of
the total survey sample for the aggregate travel patterns for all activities.
149

Table 35 Group statistics by aggregate travel pattern variables
Urban Detroit

Total trip frequency, pedestrian1
Total trip frequency, driving2
Cumulative trip distance,
pedestrian3
Cumulative trip distance, driving4
1.
2.
3.
4.

Mean
(Std.Dv)
2.46
(1.217)
2.94
(1.407)
1.73
(1.192)
2.67
(1.509)

N
57
57
57
57

Higher density
suburban
Mean
N
(Std.Dv)
2.15
189
(.932)
2.87
189
(1.053)
1.21
189
(.963)
2.69
189
(1.253)

Lower density
suburban
Mean
N
(Std.Dv)
1.66
76
(.718)
3.15
76
(1.211)
1.55
76
(1.189)
3.40
76
(1.178)

Total trip frequency for all activities by walking and cycling on a weekly basis;
Total trip frequency for all activities by driving on a weekly basis;
Cumulative trip distance for all activities by walking and cycling on a weekly basis;
Cumulative trip distance for all activities by driving on a weekly basis.

Figure 36 Separation of groups on discriminant functions by aggregate travel pattern variables

150

In order to exhibit different environmental characteristics between neighborhood groups,
the degree of fear from crime and traffic variables were used in the last discriminant analysis
(table 36). This measurement is particularly important in order to understand the effects of
differing neighborhood environments on individual pedestrian travel, given that fear from crime
and traffic volume discourages walking in the neighborhood. Wilks’ Lambda, Λ = 0.55, χ2 (8, N
= 1077) = 53.501, p < 0.01, denotes the fear variables significantly discerned the neighborhood
density groups by the two functions combined (unexplained variance of about 55%).
(Equation 5)
D sense of fear1 = (.821 fear from crime, day) + (.842 fear from crime, night) + (-.208 fear
from traffic, day) + (.207 fear from traffic, night) – 3.039.
D sense of fear2 = (.500 fear from crime, day) + (-1.037 fear from crime, night) + (-.449
fear from traffic, day) + (1.434 fear from traffic, night) – .764.
In table 36, based on the within-groups correlations and standardized coefficients
between the predictors, the sense of fear from crime variable during the day and night (0.618,
and 0.502, respectively) shows the strongest relationship with the first discriminant function. For
the second function, while sense of fear from crime during the night (-0.761) exhibits the greatest
magnitude to the function, traffic-related fear variables show substantially strong correlation
coefficients (0.762 for night time and 0.376 for day time). Thus, the first function was labeled as
‘fear from crime,’ and the second function was labeled as ‘fear from traffic.’

151

Table 36 Standardized coefficients and correlations of predictor variables with the two
discriminant functions by social environment
Correlation
coefficients with
discriminant
functions
Function Function
1
2
Discriminant analysis III: Social environment grouping
Sense of fear from crime1, day
.827
.118
Sense of fear from crime, night
.910
-.216
Sense of fear from traffic2, day
.379
.376
Sense of fear from traffic, night
.468
.762

Standardized
Coefficients for
discriminant
functions
Function Function
1
2
.502
.618
-.160
.176

.306
-.761
-.345
.220

Dependent Variable (categories): urban density neighborhood groups (Urban Detroit neighborhoods,
Higher density suburban neighborhoods, and Lower density suburban neighborhoods)
1. Sense of fear from crime on Likert scale, 1=none, 2=low, 3=moderate, and 4=high;
2. Sense of fear from traffic on Likert scale, 1=none, 2=low, 3=moderate, and 4=high.

Canonical scores of group means (centroids, from figure 37) show that the urban Detroit
neighborhood group (coded as 1) was separated by function 1 for the greater canonical score of
‘fear from crime’ compared to its suburban counterparts. In contrast, the lower density suburban
neighborhood group (coded as 2) was separated by function 2 for the greater canonical score of
‘fear from traffic’ compared to the other high density neighborhoods, both urban and suburban.
The discriminant functions predicted 94% (n=1077) of the variance between neighborhood
groups among the total survey sample for fear from crime and traffic as a differing
environmental characteristic that influences individual pedestrian travel.

152

Table 37 Group statistics by sense of fear variables
Urban Detroit

Sense of fear from crime, day
Sense of fear from crime, night
Sense of fear from traffic, day
Sense of fear from traffic, night

Mean
(Std.Dv)
2.29
(.951)
3.18
(.909)
2.20
(.971)
2.72
(1.002)

N
233
233
233
233

Higher density
suburban
Mean
N
(Std.Dv)
1.23 454
(.483)
1.84 454
(.698)
1.53 454
(.670)
1.75 454
(.714)

Lower density
suburban
Mean
N
(Std.Dv)
1.26
390
(.473)
1.75
390
(.655)
1.67
390
(.738)
2.07
390
(.897)

Figure 37 Separation of groups on discriminant functions by sense of fear variables

153

Table 38 Canonical scores of group means with the two discriminant functions
Function 1
Function 2
Discriminant analysis I: Socio-economic and demographic profile grouping
Urban Detroit
-1.588
.028
Higher density suburban
.364
-.177
Lower density suburban
.540
.187
Discriminant analysis II: Travel pattern grouping
Urban Detroit
-.200
.397
Higher density suburban
-.211
-.118
Lower density suburban
.675
-.004
Discriminant analysis III: Social environment grouping
Urban Detroit
1.619
.002
Higher density suburban
-.443
-.229
Lower density suburban
-.451
.266

Based on the three discriminant analyses results, distinctive characteristics of each urban
density neighborhood group were revealed (table 38). The urban Detroit neighborhood grouping
was characterized as a low income population sub-group with pedestrian-dominant travel
patterns and a greater fear from crime associated with walking in the neighborhood. The higher
density suburban neighborhood group was characterized as a high income population sub-group
at an earlier life cycle stage with walking-dominant travel patterns, but a lesser fear from crime
and traffic associated with walking in the neighborhood. Lastly, for the lower density suburban
neighborhood group, the neighborhood was characterized as a high income population sub-group
with driving-dominant travel patterns, but a greater sense of fear from traffic walking in the
neighborhood.

154

4.5. Travel behavior characteristics in three neighborhood typologies in the Detroit
region
In order to understand the effects of individual socio-economic and demographic
characteristics on trips, controlling the urban built environment is crucial to accurately
understand the effects of individual socio-economic and demographic characteristics on trip
making (Handy, 1993). Travel costs and patterns vary by neighborhood built environments, since
urban form provides different sets of potential destinations for daily activities within the
neighborhood.
First, before looking at the disaggregate individual travel behavior analysis results from
the disaggregate regression analysis and analysis of covariance in the following chapter, it is
noteworthy to examine the mean differences in travel activities between the three neighborhood
typologies. By comparing travel patterns—total trip frequency and cumulative trip distance—for
all travel activities, a clear distinction was revealed between the three urban neighborhood
density groups (table 39). On average, the respondents in the lower density suburban
neighborhoods made the least number of trips, but they traveled the greatest trip miles by all
modes for all activities. Meanwhile, the respondents in the higher density neighborhoods, both
urban and suburban, made more trips, supporting the argument that shorter neighborhood travel
distances in high density neighborhoods may spur trip generation and thereby overall travel
(Crane, 1996; Crane and Crepeau, 1998).
However, despite the similar urban densities between the higher density urban and higher
density suburban neighborhoods, the respondents in the urban Detroit neighborhoods traveled
longer distances by all modes for all activities, providing the critical cue to explore the differing
effects of neighborhood amenity characteristics on individuals’ trip making. Moreover, a pattern
155

of greater miles traveled by all modes in the lower density suburban neighborhood, both in terms
of average miles traveled and sum of miles traveled, offers empirical evidence of the extreme
automobile-dependence in urban environments characterized by low densities.
Differing travel behaviors by neighborhood type become more obvious when trip
frequencies and distances are explored separately by each travel mode (pedestrian, public transit,
and driving, see table 39). In the lower density suburban neighborhoods, there were almost no
public transit trip and the respondents in these neighborhoods made the least number of trips on
average by walking. The respondents in the lower density suburban neighborhoods also traveled
the most—both in terms of trip frequency and distance—by driving for all activities.
In the urban Detroit neighborhoods, the respondents traveled not only more frequently,
but also over greater miles, by pedestrian and public transit modes compared to respondents in
similar higher density suburban neighborhoods (figure 38 and 39). In particular, the public transit
trip distance for the urban Detroit neighborhoods was substantially greater than that of the
respondents in the higher density suburban neighborhoods. In terms of driving trips and distances
traveled to reach their destinations, respondents in the urban Detroit neighborhoods also traveled
more frequently and greater miles on average compared to those of the higher density suburban
neighborhoods (figure 40). These descriptive results suggest the greater travel burdens of the
urban Detroit neighborhoods.

156

Table 39 Descriptive analysis results of travel patterns for all activities in three neighborhood
typologies
All modes
Pedestrian
Freq.
Dist.
Freq.
Dist.
Urban Detroit neighborhood
Mean
12.44
41.08
5.96
10.63
Std. Dv.
12.303 58.261
7.191 20.853
Sum
2911
9202
673
1020
Minimum
1
0
1
0
Maximum
76
337
54
146
N
234
224
113
96
Higher density suburban neighborhood
Mean
12.31
35.88
4.78
5.20
Std. Dv.
7.023 43.209
4.417
9.362
Sum
5699 16467
1276
1164
Minimum
1
0
0
0
Maximum
36
277
24
88
N
463
459
267
224
Lower density suburban neighborhood
Mean
11.75
59.04
3.22
7.79
Std. Dv.
7.931 53.887
3.310 11.164
Sum
4584 22731
373
740
Minimum
1
1
0
0
Maximum
61
349
21
54
N
390
385
116
95

Public transit
Freq.
Dist.

Driving
Freq.
Dist.

4.43
4.685
297
0
25
67

24.41
31.949
1318
1
191
54

9.94
10.131
1939
1
76
195

37.56
53.993
7061
0
333
188

3.42
4.681
164
0
32
48

6.67
12.403
307
0
82
46

9.78
6.386
4440
0
52
454

34.43
44.468
15459
0
277
449

1.67
0.816
10
1
3
6

15.20
20.572
76
1
51
5

10.83
7.548
4202
1
61
388

56.55
52.622
21660
1
349
383

For trips to work, people in the urban Detroit neighborhoods consistently made more trips
and traveled more miles as pedestrians and by public transit than those in both the high and
lower density suburban neighborhoods (table 40). The urban respondents traveled on average 16
miles by walking, and 26 miles by public transit to work, which is a substantially longer distance
compared to 6 and 8 miles by pedestrian and public transit modes, respectively, in the similar
built form higher density suburban neighborhoods. This may imply the pattern of the reverse
commute (Kain, 1968; Johnston-Anumonu, 1992, 1997; McLafferty and Preston, 1992, 1996a, b,
c, 1997; Vojnovic and Darden, 2013b; Wyly, 1998). Among all three neighborhood density
groups, the lower density suburban neighborhood residents traveled the greatest miles to work by
157

driving, which reflects the great dependence on driving trips of the lower density suburban
neighborhood residents.
Table 40 Descriptive analysis results of travel patterns for work trips in three neighborhood
typologies
Frequency
Public
Pedestrian
transit
Urban Detroit neighborhood
Mean
3.86
3.52
Std. Dv.
1.859
2.108
85
81
Sum
1
0
Minimum
7
8
Maximum
22
23
N
Higher density suburban neighborhood
Mean
3.18
2.99
Std. Dv.
2.078
1.975
Sum
258
102
Minimum
1
1
Maximum
10
7
N
81
34
Lower density suburban neighborhood
Mean
2.92
3.00
Std. Dv.
1.978
2.828
Sum
76
6
Minimum
1
1
Maximum
7
5
N
26
2

Driving

Pedestrian

Distance
Public
transit

Driving

5.03
3.139
357
1
18
71

15.77
14.202
205
0
49
13

26.29
31.34
368
1
126
14

46.43
49.593
2507
0
220
54

4.45
1.391
1197
1
8
269

6.1
9.762
433
0
69
71

7.81
14.451
242
1
82
31

43.57
51.646
10763
0
284
247

4.81
2.135
1053
1
20
219

13.95
10.352
265
2
36
19

17.5
9.192
35
11
24
2

63.75
79.326
12240
1
965
192

158

For other daily activities, such as shopping, going to restaurants, accessing personal
service, and leisure activities, trip patterns were consistent to work trips and aggregate trips (for
all activities) (table 41, 42, 43, and 44). That is, respondents in the lower density suburban
neighborhoods were involved in the least pedestrian trips, regardless of travel activity.
Subsequently, the lower density suburban respondents traveled more frequently and traveled
more miles by driving than those in the higher density suburban neighborhoods for each activity.
However, the fact that the urban Detroit respondents traveled more miles for shopping than both
the suburban neighborhood types (lower or higher density) shows that urban minorities face
greater travel burdens from traveling between scarce destinations in the urban neighborhoods.
The greater travel burden of urban minorities in the Detroit neighborhoods is also evident
from the greater frequency and distance that these population sub-groups travel as pedestrians
and by public transit for their daily activities. For shopping, restaurant, and service trip activities,
the Detroit urban respondents made the most pedestrian trips and traveled the longest distances
when compared to all other neighborhood groupings. The Detroit urban respondents also
traveled the most miles by public transit for shopping, restaurant, service, and leisure compared
to those respondents in the similar higher density suburban neighborhoods. Higher density
suburban respondents, in fact, made the least frequent trips as pedestrians and by driving, and
traveled the least miles by any travel modes for their daily activities (except for leisure). For
leisure trips, the higher density suburban respondents made more pedestrian trips, but they
traveled fewer miles than the Detroit urban respondents. This might be attributed to after-dinner
walks or walking the dog within their neighborhood.
We can conclude from this analysis, therefore, that individual travel behaviors are a set of
choices that result from differing socio-economic and demographic characteristics, the nature of
159

available travel modes, and urban built environments in which an individual lives. It is clear that
lower income Detroit residents traveled more frequently as pedestrians and by public transit,
while the lower density suburban residents traveled more frequently by car. Furthermore, longer
trip distances were evident for the lower density suburban respondents in their daily travel
activities. However, lower income Detroit residents traveled longer distances for daily activities
when compared to the higher density suburban residents.

160

Table 41 Descriptive analysis results of travel patterns for shopping trips in the three
neighborhood typologies
Frequency
Public
Pedestrian
transit
Urban Detroit neighborhood
Mean
3.77
3.97
Std. Dv.
3.954
5.242
Sum
283
119
Minimum
0
1
Maximum
23
21
N
75
30
Higher density suburban neighborhood
Mean
2.15
1.69
Std. Dv.
1.672
1.974
Sum
215
22
Minimum
0
0
Maximum
12
8
N
100
13
Lower density suburban neighborhood
Mean
2.21
0.50
Std. Dv.
1.855
0.707
Sum
75
1
Minimum
0
0
Maximum
8
1
N
34
2

Driving

Pedestrian

Distance
Public
transit

Driving

3.47
3.567
586
0
30
169

7.70
18.888
547
0
146
71

15.00
27.563
450
1
150
30

14.36
19.950
2369
0
186
165

3.26
2.300
1371
0
16
421

2.22
3.507
213
0
29
96

2.46
2.665
32
0
9
13

6.90
6.186
2886
0
55
418

3.55
2.805
1287
0
30
363

5.79
9.834
197
0
54
34

2.00
0.000

11.81
14.151
4251
0
168
360

161

4
2
2
2

Table 42 Descriptive analysis results of travel patterns for restaurant trips in the three
neighborhood typologies
Frequency
Public
Pedestrian
transit
Urban Detroit neighborhood
Mean
2.86
1.88
Std. Dv.
3.942
1.054
Sum
120
32
Minimum
1
1
Maximum
25
5
N
42
17
Higher density suburban neighborhood
Mean
1.71
2.50
Std. Dv.
1.390
1.732
Sum
162
10
Minimum
0
1
Maximum
10
4
N
95
4
Lower density suburban neighborhood
Mean
1.48
2.00
Std. Dv.
1.238
0.000
Sum
34
4
Minimum
0
2
Maximum
5
2
N
23
2

Driving

Pedestrian

Distance
Public
transit

Driving

2.83
2.639
320
0
16
113

3.33
4.503
130
0
20
39

6.69
7.525
107
1
30
16

8.12
13.914
893
0
101
110

1.93
1.495
516
0
8
268

2.52
9.414
209
0
86
83

2.33
1.528
7
1
4
3

5.44
7.384
1332
0
61
245

2.19
1.961
558
0
12
255

6.45
9.364
142
0
43
22

5.50
6.364
11
1
10
2

9.63
11.663
2244
0
72
233

162

Table 43 Descriptive analysis results of travel patterns for personal service trips in the three
neighborhood typologies
Frequency
Public
Pedestrian
transit
Urban Detroit neighborhood
Mean
2.36
1.76
Std. Dv.
1.973
1.091
Sum
99
51
Minimum
1
1
Maximum
10
5
N
42
29
Higher density suburban neighborhood
Mean
1.83
2.38
Std. Dv.
1.365
2.774
Sum
187
19
Minimum
0
1
Maximum
7
9
N
102
8
Lower density suburban neighborhood
Mean
1.61
1.00
Std. Dv.
1.202
.
Sum
50
1
Minimum
0
1
Maximum
6
1
N
31
1

Driving

Pedestrian

Distance
Public
transit

Driving

3.10
3.017
418
1
18
135

3.93
6.434
110
0
33
28

9.58
11.217
182
0
37
19

10.54
13.247
1054
0
67
100

2.46
1.988
796
0
14
324

1.58
1.975
120
0
14
76

2.63
2.446
21
1
8
8

5.24
6.118
1304
0
43
249

2.96
2.599
936
0
30
316

2.79
3.919

5

163

.
78
0
18
28

5
5
5
1

10.53
22.339
3032
0
332
288

Table 44 Descriptive analysis results of travel patterns for leisure trips in the three neighborhood
typologies
Frequency
Public
Pedestrian
transit
Urban Detroit neighborhood
Mean
2.86
2.46
Std. Dv.
2.333
1.769
Sum
120
59
Minimum
0
0
Maximum
10
7
N
42
24
Higher density suburban neighborhood
Mean
3.13
1.63
Std. Dv.
2.780
1.408
Sum
523
13
Minimum
0
1
Maximum
15
5
N
167
8
Lower density suburban neighborhood
Mean
2.62
2.00
Std. Dv.
2.407
1.000
Sum
144
6
Minimum
0
1
Maximum
14
3
N
55
3

Driving

Pedestrian

3.23
3.655
349
1
21
108

3.63
5.304

Distance
Public
transit

Driving

69
0
21
19

12.53
11.838
213
1
42
17

20.4
28.208
1469
0
164
72

2.95
2.665
913
0
28
310

2.49
4.136
237
0
36
95

3.25
3.304
13
1
8
4

10.44
19.849
2318
0
206
222

2.99
3.497
844
0
40
282

2.81
3.257

10.00
1.414
20
9
11
2

19.78
25.574
4491
0
229
227

164

90
0
10
32

Figure 38 Pedestrian trips for each travel activity by neighborhood groups

165

Figure 39 Public transit trips for each travel activity by neighborhood groups

166

Figure 40 Driving trips for each travel activity by neighborhood groups

167

4.6. Conclusion and Discussion
First, the findings from the aggregate travel pattern analyses in the Detroit region
(n=1,148) demonstrates the significant effects of socio-economic and demographic factors
(income, age, personal automobile access, and the number of dependent children) on individual
trip making. Before separating travel modes, income and personal vehicle access had a positive
effect on trip generation (total trip frequencies and distance) for all activities, work and nonwork
travel, while age and the number of dependent children had a negative effect. Thus, a person’s
travel patterns—in terms of both trip frequency and distance—are influenced substantially by
travel modes, including whether one does, or does not have access to a personal vehicle.
In addition, individuals make different choices regarding travel options in the context of
personal inter-relationships, whether one is married, whether one is a parent with dependent
children, or whether one is a single-parent. This, in turn, implies that women in a multi-person
household form their daily travel patterns according to their specialized gender roles as paid
laborer, housewife, and mother (Jones, 1990; Hanson and Hanson, 1980a, 1980b, 1981; and
Pickup, 1985; 1990). Therefore, women’s travels, when compared to men’s travel behavior, are
affected by life cycles and household organization, and limited by spatial and temporal
constraints (Hagerstrand, 1970; Hamilton and Jenkins, 1990; Jones, 1990; Kwan, 1999a, 1999b,
2000; and Pickup, 1990). The details of these relationships will be discussed further in chapter 5.
Second, discriminant analyses were applied to confirm differing characteristics between
neighborhood groups in the Detroit region before analyzing disaggregate individual travel
behavior. By incorporating socio-economic and demographic factors (income, age, personal
vehicle access, and the number of dependent children), the discriminant analysis results

168

differentiated urban Detroit neighborhoods by their lower socio-economic characteristics
(income and personal vehicle access) and higher density suburban neighborhoods by their
younger life cycle stage characteristics (age and the number of dependent children). By
incorporating travel pattern variables (total trip frequencies and cumulative trip distances by
travel mode), the analysis discriminated the lower density suburban neighborhoods against the
other higher density neighborhood groupings by the lower density neighborhoods’ driving
dominant travel patterns. In addition, urban Detroit neighborhoods were separated by pedestrian
dominant travel patterns against the suburban neighborhood typologies. Given the differences in
income levels between suburban neighborhoods and urban Detroit neighborhoods, these results
were expected.
Moreover, using the degree of fear from crime and traffic as a proxy for different
neighborhood environment quality for pedestrian travel, the analysis results clearly discerned
urban Detroit neighborhoods with the fear from crime against both the higher density suburban
and the lower density suburban neighborhoods. This, again, can be expected given the level of
disinvestment, decline, and abandonment in the Detroit urban neighborhoods. With regard to fear
from traffic, the lower density suburban neighborhoods were differentiated against both the
urban Detroit and the higher density suburban neighborhoods. Residents in the lower density
suburban neighborhoods faced considerable fear from walking within and between their
automobile oriented neighborhoods, where particularly the connecting arterials are characterized
by high speeds and a large volume of traffic. Therefore, the discriminant analysis results confirm
the importance of structuring a typological analysis—three differing neighborhoods, higher
density urban Detroit, higher density suburban, and lower density suburban neighborhood
typologies—in order to understand the social dynamics of urban travel.

169

Lastly, the descriptive analysis results illustrated substantial differences in average trip
frequency and trip distance between the three neighborhood typologies. As argued in the
literature, populations in the lower density suburban neighborhoods made fewer trips by all
modes than in the higher density—urban or suburban—neighborhoods. Respondents in the lower
density suburban neighborhoods, however, traveled longer distances by all modes when
compared to the higher density neighborhoods, which is consistent with previous studies (Crane,
1996; and Crane and Crepeau, 1998). Nevertheless, the fact that respondents in the urban Detroit
neighborhoods traveled longer distances than the respondents in the higher density suburban
respondents supports the importance of this neighborhood typology, and its disinvestment and
decline, in understanding the travel burdens of lower-income urban minorities.
By disaggregating travel activities by travel type, the travel burdens faced by urban
minorities in the Detroit region become even clearer. Regardless of travel activities—whether
they were commutes to work, trips to shopping, restaurant visits, personal services, or leisure
destinations—the respondents in urban Detroit neighborhoods traveled longer distances not only
as pedestrians, but also by public transit, when compared to respondents who lived in the higher
density suburban neighborhoods. However, the results are still limited in depicting the full
understanding of individual travels since there were too few respondents after disaggregating by
travel activities. By driving, on average, respondents in urban Detroit neighborhoods also
traveled longer distances for their daily activities than the respondents in the higher density
suburban neighborhoods. Furthermore, to access their shopping, personal services, and leisure
destinations, Detroit urban minorities even drove longer distances than respondents in the lower
density suburban neighborhoods. Consequently, these descriptive results may offer the initial
evidence of the differing effects of neighborhood typologies on individual travel patterns, and in

170

particularly within the context of a region where low-income, minority sub-groups live in
neighborhoods experiencing disinvestment and decline. This analysis also offers the potential
insight and importance of further disaggregation of the data, and specifically reflecting on the
type of differences in travel that might be apparent by gender.

171

CHAPTER 5: Gender difference in travel behavior in Detroit
region
In Chapter 5, I will examine the differences in trip making between women and men
living in differing urban built environments—specifically the higher density urban, higher
density suburban and lower density suburban neighborhood typologies—and how these
differences vary by socio-economic and demographic characteristics. For this analysis, six
neighborhoods in the Detroit region were selected in the counties of Wayne (two urban Detroit
neighborhoods situated on the east side of the city), Washtenaw (one Ann Arbor neighborhood),
and Oakland (three neighborhoods, one each in Birmingham, Bloomfield Hills, and West
Bloomfield). Among these neighborhoods, the two urban Detroit neighborhoods and the Ann
Arbor and Birmingham neighborhoods were characterized as higher density urban built
environments, in contrast to the two suburban lower density neighborhoods located in
Bloomfield Hills and West Bloomfield.
By socio-economic and demographic characteristics, urban Detroit neighborhoods were
described as low income, predominantly non-white with low access to a private automobile. In
contrast, the two suburban neighborhood groups, the higher density suburban and lower density
suburban groups, were characterized by higher incomes, predominantly white, with greater
access to a car. In the urban Detroit neighborhoods, 91.9% of the respondents were non-white,
and the average personal income was $23,600, while the suburban neighborhood groups were
predominantly white (94.7% for the higher density suburban neighborhood group and 88.6% for
the lower density suburban neighborhood group) and higher income ($58,100 in the higher
density suburban neighborhood group and $63,900 in the lower density suburban neighborhood
group).

172

Another stark difference between the urban Detroit and the suburban neighborhoods was
the percentage of population that was married. The majority of urban Detroit respondents were
not married (76.2%) while more than 70% of the suburban respondents from both neighborhood
groups were married (71.3% for the higher density suburban neighborhood group and 76.0% for
the lower density suburban neighborhood group). Subsequently, in reflecting on socio-economic
differences from being married or living with a partner (Light, 2004; U.S. Census Bureau, 2010),
it is assumed that minority women who were not married in urban Detroit neighborhoods were at
the lowest income categories among gender and marital status groups in the survey sample.
Lastly, the lower income status of the urban Detroit neighborhood group was also reflected in the
average automobile access per capita. Average automobile access per capita was less than one
car (0.8) in urban Detroit, whereas each suburbanite had access to more than one car (1.3 for the
higher density suburban neighborhood group and 1.4 for the lower density suburban
neighborhood group). Table 3 in chapter 2 provides a more detailed break-down.
By disaggregating daily trips into trips to work, shopping, restaurants, personal services
and leisure, and exploring these differences by gender, this chapter will provide a more
systematic review of travel—including trip frequencies, trip distances, and travel modes—within
different neighborhood typologies in the Detroit region. The disaggregation of travel into a
comprehensive array of diverse travel activities can offer considerable insight into understanding
the complexity of women’s daily travel based on their socio-economic status and multiple roles
within the household (Freedman and Kern, 1997; Hamilton and Jenkins, 1990; Hanson 1995;
Hanson and Hanson, 1980a, 1980b; Hanson and Johnston, 1985; Hanson and Giuliano, 2004; Hu
and Young, 1999; Jones 1990; Johnston-Anumonwo, 1992; Kwan, 1999a, 1999b, 2000;
McFadden and Reid, 1975; Schintler et al., 2000; Santos et al., 2011; and Pickup, 1985, 1990).

173

In addition, the effects of socio-economic and demographic characteristics on travel
among women will be more clearly evident when factors that influence their travels, such as
differing built environments, are controlled for in the statistical models. Disaggregate travel
behavior analysis offers appropriate explanations on the complexity of women’s daily travels
that are highly influenced by their household structure, the nature of travel modes (whether they
have access to a car), and the immediate neighborhood environment (Freedman and Kern, 1997;
Handy, 1993, 1996; Hanson and Hanson, 1980a, 1980b; Hanson, 1982; and Srinivasan and Bhat,
2005). Hence, measuring the easiness of reaching various destinations within the neighborhood
is crucial to identify the travel burdens of women as they accomplish their daily responsibilities.
Within this context, disaggregated travel behavioral models at a neighborhood level will help to
explain the differences in trip making between women and men, and also among women with
different socio-economic and demographic profiles.
5.1. Research Objectives, Questions and Hypotheses
The goal of this chapter is to examine gendered travel patterns and their relationship with
urban built environment characteristics after controlling for socio-economic and demographic
factors. Accordingly, it is hypothesized that women’s travels are more related to household
responsibilities within the urban built environment. It is also hypothesized that lower-income,
urban minority women tend to rely on walking and public transit in accessing daily destinations
more than other women’s groups, such as wealthier women in suburban neighborhoods, due to
lower access to a car among the lower-income Detroit women. Subsequent research questions are
how women travel differently from men within similar urban built environments for daily
activities and how lower-income urban minority women’s travels are more disadvantaged than
other women’s groups living in the suburban Detroit neighborhoods.
174

5.2. Disaggregate Travel Behavior Research Data and Method
In order to address the social dynamics of an individuals’ travel behavior, particularly
between women and men, I performed two sets of disaggregated analyses: (1) OLS regression;
and (2) Analysis of Covariance (see table 45). Gender differences in travel behavior will be
examined after controlling for different urban built environments at a neighborhoods scale.
Ordinary Least Squares (OLS) regression was used to assess how women’s travel
behavior differed from those of men utilizing various travel modes—including pedestrian,
public transit, and driving—after controlling for other socio-economic and demographic
characteristics that were selected from the general trip generation models. It becomes
particularly important to understand the different utilization of travel modes by women in the
context of their travel constraints that result from differing degrees of access to a private vehicle
in the household.
Women’s travel patterns should also be understood with consideration to gender roles
and to the multifaceted household responsibilities of women. Hence, the complexity of women’s
daily travels were explored by disaggregating their travel patterns into various daily activities,
such as trips for work, shopping, restaurant visits, leisure and personal services. ANCOVA
(Analysis of Covariance) was also used to assess the mean differences between women and men
by marital status and types of activity-based travel behavior. With this method significant mean
differences between women and men were reported, taking into account whether they were notmarried or married, controlling for age and household structure and adjusted family income
(herein referred to as income). A fuller explanation of the application of these two methods in
this research follows.

175

Table 45 Variables for disaggregate travel behavior models
Models

Independent Variables
(Predictors)

Dependent Variables

Gender differences in individual travel behavior
OLS regression
(Within
neighborhoods)

Total trip frequency1
Cumulative trip distance2
 Work trip
 Shopping trip
 Restaurant trip
 Service trip
 Leisure trip

Gender (Women)3
Income4
Age5
# of personal vehicles available
# of dependent children6

Gender and marital status differences in individual travel behavior
ANCOVA
(Within
neighborhoods)

1.
2.
3.
4.
5.
6.
7.

Total trip frequency1
Cumulative trip distance2
 Work trip
 Shopping trip
 Restaurant trip
 Service trip
 Leisure trip

Gender (Women)3
Marital status7
Covariates
 Income4
 Age5

Weekly total trip frequency by each mode (walking and cycling, public transit, and driving)
Weekly cumulative trip distance by each mode (walking and cycling, public transit, and driving)
Women = 1 and men = 0
Household structure adjusted family income
Adults who are 18 years old and older
Children who are younger than 18 years old
And Married = 1 and not-married = 0

176

In order to explain the travel demand generated by individuals or households, researchers
have used disaggregated regression analyses as a function of the characteristics of these decisionmakers (White and Senior, 1983). Due to the relatively large variation in trip generations of
individuals and households between analysis groups, sometimes the coefficients of correlation
and determination are low, but this does not mean that disaggregate regressions provide inferior
fits to reality compared to other studies using regression analyses (Manheim, 1979; Martin,
Memmot, and Bone, 1967; and Meyer and Miller, 1984). Moreover, as an indicator of actual trip
making responses, disaggregate regression coefficients can aid in more fully understanding
changes in explanatory (independent) variables (White and Senior, 1983). In summary,
disaggregate methods, compared to aggregated zonal regression methods, have captured the true
behavioral variation in trip making by individuals and households more accurately and have used
survey data more efficiently. Equation 6 is an example of a disaggregate travel behavior
regression model.
(Equation 6)
Y= a + b1X1 + b2X2 + b3X3 + ∙∙∙∙∙∙ + biXi + e
Where Y = travel behavior response (dependent) variable
a = the intercept term
b = the regression coefficient (slope term) for variable X
X = socio-demographic characteristic predictors (independent variable)
i = the number of predictors (independent variables)
e = an error (disturbance) term

177

Since individual travel behavior are closely related to the role of individuals in the
household—e.g. married versus not married, households with or without dependent children, or
two income-earner households versus single income earner households—it is important to
discern the differences in trip making by household characteristics. Particularly in exploring
gender differences in individual travel behavior, the incorporation of marital status is crucial in
examining the variations in travel behavior that are differentiated by the diverse travel activities
between women and men (Hamilton and Jenkins, 1990; Hanson and Hanson, 1980a; Jones,
1990; Pickup, 1985). Here, I use the analysis of covariance (ANCOVA) to detect the mean
differences between women and men while also considering their marital status.
The ANCOVA is a statistical method that utilizes the benefits of both analysis of
variance (ANOVA) and linear regression analysis (Silk, 1979). It is also concerned with a
response or dependent variable, measured on an interval or ratio scale, and a set of predictors
(independent variables) that are measured on a nominal or ordinal scale. In addition to
categorical predictors, the ANCOVA introduces a covariate, an additional explanatory variable,
measured on a ratio or interval scale, to adjust mean values already estimated according to the
analysis of variance model on the categorical variables. The ANCOVA adjusts the mean for each
categorical group on the basis of the mean deviation for the covariate. The adjustment of
dependent variables can be found by subtracting the group deviation on the covariate that is
weighted by the regression coefficient from the unadjusted deviation. Such adjustments are
particularly important to disentangle the confounding effects between the categorical predictors
and covariates.
In this study, the ANCOVA allows for greater test sensitivity in examining the effects of
gender and marital status on determining trip characteristics of each daily activity, including
178

shopping, restaurant, and leisure trips. From household or individual travel surveys,
disaggregating individual travel behavior into a variety of daily travel activities only enables
limited observations (N) for each category—such as married women versus married men from
the interaction between gender and marital status categories. Thus, the ANCOVA can offer mean
differences as the effects of groups, while the inclusion of such categorical variables into OLS
regressions empirically results in challenges on detecting differences on individual travel
behavior at such disaggregated levels of travel activity. Equation 7 is an example of the
ANCOVA model of travel behavior that is used in this research.
(Equation 7)

Where Y = travel behavior response (dependent) variable
= deviation of dependent variable for categorical group j
b = the regression coefficient (slope term) for covariate X
X = covariate variable (income and age)
= deviation of covariate variable for categorical group j
j = the number of categorical group of predictors (independent variables)

179

5.3. Results
5.3.1.

Urban Detroit (Lower East Side Detroit Neighborhoods):

Disaggregate analyses on individual travel behavior in the higher density
urban neighborhoods
In this section, the effects of gender, socio-economic and demographic characteristics on
individual travel behavior were analyzed for disaggregate travel activities in the higher density
urban Detroit neighborhoods. First, the OLS regression was used to estimate the causal
relationships between gender and trip frequency (model 1) and distance (model 2), after
adjusting for differences in socio-economic and demographic characteristics (age, income,
number of dependent children and the number of personal vehicles available) for each of the
travel activities. Both dependent variables were transformed using square-root and logarithm
respectively. The results showing the effects of gender on trip frequency (table 46) and distance
(table 47) for daily activities are provided, for the travel activities going to work, shopping,
restaurant visits, personal service, and leisure. Travel frequency and distance were also explored
by three travel modes: pedestrian (also referred to as walking or biking), public transit and
driving. While gender (women = 1) was primarily used to understand individual travel behavior,
other independent variables (age, income, number of dependent children and personal vehicle
availability) were also employed in the models when the sample size allowed.

180

Table 46 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the urban Detroit neighborhoods: trip frequency
Total trip frequency1
N2
β
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

S.E.

df

P-value

21
22
67

.642
-.093
-.167

.188
.399
.224

3
3
3

.006
.819
.458

70
27
155

.115
.081
.221

.254
.565
.154

3
3
3

.651
.887
.155

41
15
95

-.266
-.069
-.042

.284
.197
.136

3
3
4

.356
.734
.760

40
26
122

-.028
-.408
.110

.270
.194
.161

3
3
3

.918
.051
.496

35
20
101

.059
-.926
-.052

.264
.188
.181

4
4
3

.826
.000
.777

1. Total trip frequency per person on a weekly basis (trip count, square root transformed);
2. N = number of resident men and women;
3. Gender dummy: women = 1, and men = 0, and controlled for constant, age and income (**);
4. Controlled for constant (***), age and income (**);
5. Controlled for constant, age and income;
6. Controlled for constant (*),age (**) and income;
7. Controlled for constant, age and income;
8. Controlled for constant, age and income;
9. Controlled for constant (*), age (***) and income;
10. Controlled for constant, age and income;
11. Controlled for constant (*), age (**), income and personal vehicle available (**);
12. Controlled for constant, age (***) and income;
13. Controlled for constant (***), age and income (**);
14. Controlled for constant, age (*), and income (*);
15. Controlled for constant, age, income and the number of dependent children (**);
16. Controlled for constant, age, income and the number of dependent children (**);
17. Controlled for constant, age (**) and income.
For all models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

181

Table 47 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the urban Detroit neighborhoods: trip distance
Cumulative trip distance1
N2
β
S.E.
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

df

P-value

12
14
49

.576
-.734
-1.144

1.982
.905
.489

3
3
3

.786
.441
.025

60
23
146

-.047
-1.350
-.181

.386
.714
.228

3
4
3

.904
.077
.429

32
14
96

-.459
-1.476
-.229

.428
.965
.300

3
3
3

.294
.165
.448

18
17
90

.856
-2.018
-.358

.732
.711
.298

4
3
3

.267
.018
.233

16
15
62

1.018
.017
-.589

.540
.598
.283

3
3
4

.086
.978
.043

1.
2.
3.

Cumulative trip distance to destinations per person on a weekly basis (mile, logarithm transformed);
N = number of resident men and women;
Gender dummy: women = 1, and men = 0, and controlled for constant, age and income, the tolerance
values were above 0.500 for all predictors;
4. Controlled for constant, age and income;
5. Controlled for constant (**), age, and income;
6. Controlled for constant, age, and income;
7. Controlled for constant, age, income and personal vehicle available (**);
8. Controlled for constant, age and income (*);
9. Controlled for constant, age and income (*);
10. Controlled for constant, age and income (*);
11. Controlled for constant, age and income (*);
12. Controlled for constant, age (**), income and personal vehicle available (**), the tolerance values were
above 0.500 for all predictors;
13. Controlled for constant (**), age and income (*);
14. Controlled for constant, age and income (*);
15. Controlled for constant, age and income;
16. Controlled for constant (*), age and income (**), the tolerance values were above 0.500 for all
predictors;
17. Controlled for constant (**), age (**), income and the number of dependent children (**).
For all other models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

182

5.3.1.1.

Gender Difference in Travel Behavior: Trips to Work

In the Detroit urban neighborhoods, women had significantly more trips to work via
walking or biking than men (B= 0.642, p-value = 0.005) controlling for differences in age and
income. Of those women and men who drove to work (n=49), women’s trips were significantly
shorter than men’s trips (B = -1.144, p-value = 0.025) controlling for differences in age and
income.
5.3.1.2.

Gender Difference in Travel Behavior: Trips to Shops, Restaurants,

Personal Services, and Leisure
5.3.1.2.1.

Trips to shopping

In the Detroit urban neighborhoods women overall made more frequent shopping trips
than men using all modes of transportation, however, the differences in frequency by modes
were not significant. Of women and men who traveled to shop by public transit (n=23) women
traveled shorter distances than men (B = -1.350, p-value = 0.077) controlling for age, income and
number of personal vehicles available. Men overall made longer trips to shopping than women
using all modes of transportation; however these differences in distance were not significant.
5.3.1.2.2.

Trips to restaurants

In the Detroit urban neighborhoods, women traveled less frequently and shorter distances
than men to restaurants by all modes of transportation, however these differences were not
significant controlling for age and income.
5.3.1.2.3.

Trips to personal services

In the Detroit urban neighborhoods, women traveled significantly less frequently than
men by public transit to reach personal services (B = -0.408, p-value = 0.051) controlling for
183

differences in age and income. Women also traveled significantly shorter distances to personal
services than men using public transit (B = -2.018, p-value = 0.018) controlling for differences in
age and income. In general women drove more frequently than men to personal services (B =
0.110), but their travel distance was shorter than that of men (B = -0.358). These differences in
driving and travel distances were not significant.
5.3.1.2.4.

Trips for leisure

In the Detroit urban neighborhoods, women traveled significantly fewer trips than men
by public transit for leisure activities (B = -0.926, p-value = 0.000), controlling for age, income
and number of dependent children. Women also drove significantly shorter distances than men in
their trips to leisure destinations (B = -0.589, p-value = 0.043) controlling for age and income. In
general women walked or biked longer distances than men for leisure activities, however, the
differences were marginally significant (B = 1.018, p-value = 0.086). In the high density urban
Detroit neighborhoods, differences in individual travel behavior between women and men were
also observed within the context of marital status (table 48). Particularly, from the ANCOVA
results, there were clearly observed divergences in travel patterns for disaggregated travel
activities for married and non-married women and men (tables 49-53).

184

Table 48 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the urban Detroit neighborhoods
Men

Work

Trip
Frequency
Trip
Distance

Shopping

Trip
Frequency
Trip
Distance

Restaurant
visit

Trip
Frequency
Trip
Distance

Service

Trip
Frequency
Trip
Distance

Leisure

Trip
Frequency
Trip
Distance

Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving

Notmarried
2.65
4.63
5.20
4.43
25.19
61.97
3.27
0.79
2.94
6.24
22.98
13.02
2.42
2.00
2.86
1.32
4.96
8.51
2.52
2.14
3.55
1.11
24.19
14.91
2.28
4.25
3.54
1.21
13.47
27.23

Women

Married
3.02
2.37
5.26
n/a
16.98
96.69
4.73
3.89
3.49
2.79
26.65
17.32
9.17
2.17
4.31
10.43
30.02
25.82
2.59
0.76
3.38
12.60
n/a
16.60
1.72
5.19
3.04
1.68
17.28
22.36

Notmarried
4.62
3.43
4.76
22.43
32.23
36.74
3.55
2.87
3.86
6.30
4.62
14.80
2.42
1.78
2.90
3.36
4.68
6.90
2.32
1.72
3.04
3.14
6.66
9.14
2.74
1.84
3.35
4.41
11.83
16.84

Married

*Calculated with original weekly total trip frequency and cumulative trip distance data.

185

4.72
2.59
5.97
11.57
-7.60
53.66
5.95
11.55
3.83
17.81
55.01
18.14
1.26
2.22
2.31
1.28
8.63
5.70
2.15
1.76
3.38
6.39
3.16
8.46
4.64
2.40
3.03
5.58
16.35
20.25

Table 49 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—work
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.540
(.179)
2.073
(.264)
2.224
(.187)
1.938
(.785)
3.012
(.635)
3.736
(.412)

Married
1.666
(.449)
1.387
(.428)
2.146
(.339)
n/a
3.139
(1.354)
4.433
(.662)

Women
NotMarried
2.114
(.143)
1.813
(.185)
2.081
(.115)
2.799
(.426)
2.646
(.458)
3.089
(.215)

N

Married
2.168
(.201)
1.297
(.314)
2.320
(.181)
1.966
(.653)
1.953
(1.530)
3.527
(.353)

21
22
67
12
14
49

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.937931, AGE = 43.62.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.839309, AGE = 43.45.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.148370, AGE = 45.69
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.050115, AGE = 42.75.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.879326, AGE = 41.71
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.199968, AGE = 44.04.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value
= .066).
Mean difference between marital status groups was only significant for public transit trip frequency (pvalue = .085).

186

Table 50 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—shopping
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.732
(.218)
1.213
(.690)
1.577
(.140)
1.639
(.348)
3.157
(.682)
2.260
(.205)

Women

Married
1.509
(.415)
1.894
(.668)
1.662
(.216)
0.902
(.678)
3.224
(.660)
2.473
(.325)

NotMarried
1.762
(.133)
1.537
(.220)
1.792
(.087)
1.263
(.201)
1.274*
(.217)
2.059
(.123)

N

Married
2.138
(.244)
2.969
(.495)
1.818
(.149)
1.55
(.375)
3.364
(.490)
2.402
(.206)

69
27
154
59
27
145

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.690909, AGE = 47.43.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.654437, AGE = 45.89.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.925840, AGE = 51.05.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.687254, AGE = 46.49.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.654437, AGE = 45.89.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.923101, AGE = 50.66.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value
= .066).
Mean difference between marital status groups was only significant for public transit trip frequency (pvalue = .073), and public transit trip distance (p-value = .064).
Mean difference between gender and marital status groups was only significant for public transit trip
distance (p-value = .074).

187

Table 51 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—restaurant visit
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.383
(.221)
1.367
(.332)
1.555
(.147)
.677
(.417)
1.255
(1.077)
1.257
(.267)

Women

Married
2.448***
(.329)
1.416
(.210)
1.823
(.229)
1.801**
(.447)
3.243
(.993)
2.296***
(.404)

NotMarried
1.517
(.136)
1.297
(.104)
1.596
(.087)
.903
(.192)
1.116
(.336)
1.420
(.165)

N

Married
1.213
(.224)
1.438
(.180)
1.408
(.153)
.371
(.384)
2.254
(.588)
1.416
(.278)

41
15
104
32
14
95

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.614206, AGE = 46.71.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.242150, AGE = 50.47.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.907100, AGE = 48.91.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.626288, AGE = 45.63.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.301572, AGE = 49.57.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.932552, AGE = 48.53.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value
= .035).
Mean difference between marital status groups was only significant for public transit trip distance (pvalue = .071), and driving trip distance (p-value = .079).
Mean difference between gender and marital status groups was only significant for pedestrian trip
frequency (p-value = .007), pedestrian trip distance (p-value = .041), and driving trip distance (p-value
= .074).

188

Table 52 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—personal services
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.439
(.233)
1.403
(.203)
1.708
(.149)
0.330
(.453)
2.728
(.613)
2.235
(.255)

Women

Married
1.549
(.382)
0.905
(.433)
1.66
(.235)
2.347*
(.792)
n/a
2.405
(.440)

NotMarried
1.419
(.104)
1.255
(.110)
1.602
(.086)
1.193
(.231)
1.742
(.411)
1.668
(.158)

N

Married
1.419
(.184)
1.288
(.218)
1.742
(.156)
1.180
(.292)
1.031
(.655)
1.819
(.266)

40
26
122
22
17
90

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.661107, AGE = 44.20.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.437054, AGE = 46.65.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.928739, AGE = 51.41.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.787968, AGE = 44.41.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.474528, AGE = 46.12.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.012668, AGE = 51.60.
Mean difference between gender groups was only significant for driving trip distance (p-value = .063).
Mean difference between marital status groups was only significant for pedestrian trip distance (p-value
= .051).
Mean difference between gender and marital status groups was only significant for pedestrian trip
distance (p-value = .054).

189

Table 53 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in urban Detroit neighborhoods—leisure
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.409
(.229)
1.979
(.224)
1.730
(.169)
0.417
(.625)
2.370
(.484)
2.958
(.256)

Married
1.056
(.351)
2.277
(.532)
1.587
(.274)
0.059
(1.005)
2.814
(1.117)
2.760
(.379)

Women
NotMarried
1.576
(.155)
1.341
(.168)
1.623
(.103)
1.295
(.370)
2.155
(.393)
2.104
(.164)

N

Married
1.971
(.271)
1.485
(.221)
1.635
(.181)
1.020
(.669)
1.982
(.533)
2.685
(.272)

39
20
101

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.604177, AGE = 47.46.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.423945, AGE = 38.15.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.914638, AGE = 48.69.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.847110, AGE = 44.38.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.608436, AGE = 39.67.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.985673, AGE = 48.69.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value
= .055), public transit trip frequency (p-value = .060), and driving trip distance (p-value = .096).

190

16
15
70

5.3.1.3.

Distinctions by Gender and Marital Status in Travel Behavior: Trips

to Work
More women walked or biked to work than men and this finding did not vary by marital
status (table 48). For work trips, both married and not-married women walked significantly more
than men (p-value = 0.066, table 49). Non-married women and men were more likely to take
public transit to work than married women and men; however, non-married women traveled less
frequently to work via public transit than non-married men (3.43 vs. 4.63 times per week). Both
married men and married women took public transit less frequently to work than both notmarried men and not-married women, and at a significant level (p-value = 0.085). Married
women engaged in the most frequent driving trips to work (5.97 times a week), while notmarried women engaged in the least frequent driving trips to work (4.76 times a week) when
compared to men (5.26 and 5.20).
In terms of work trip distance (Tables 48 and 49), not-married women walked the longest
distance to work (22.43 miles per week). This is almost twice as far as both married women
(11.57 miles). There was no married man in the sample who walked to work. In addition, notmarried women traveled the longest distance to work by public transit (32.23 miles per week). In
fact, married women engaged in the shortest travel by public transit to reach their work
destinations. Lastly, with regard to driving by car, married men (96.69 miles per week) and notmarried men (61.97 miles) drove longer distances to employment compared to women, both
married (53.66 miles) and not-married (36.74 miles). In summary, not-married women in the
urban Detroit neighborhoods had the least access to a car and traveled least frequently and the
shortest distances to work by driving.

191

5.3.1.4.

Distinctions by Gender and Marital Status in Travel Behavior: Trips

to Shops, Restaurants, Personal Services, and Leisure
5.3.1.4.1.

Trips to shopping

Shopping trips were a primarily women’s travel responsibility in the urban Detroit
neighborhoods. Regardless of travel modes, women made more frequent shopping trips in urban
Detroit (table 48). Married women, in particular, traveled the most frequently for shopping
activities among gender and marital status groups, and by all modes of travel. On a weekly basis,
married women walked 5.95 times for shopping, while not-married men walked least frequently
(3.27 times). The gender difference in pedestrian shopping trip frequency was statistically
significant (p-value = 0.066, table 50).
Married women also traveled for shopping most frequently by public transit (11.55 times
a week), more than twice compared to married men who traveled to shopping by public transit
(3.89 times a week). However, both married women and married men traveled more frequently
by public transit to reach shopping destinations when compared to not-married men and notmarried women, and at a significant level (p-value = 0.073, table 50). Married women also
traveled more frequently for shopping on a weekly basis by car (3.83 times per week) than
married men.
In terms of trip distance, married women traveled further distances for shopping only by
public transit. With statistically significant marital status differences in public transit distance (pvalue = 0.064), married women took public transit for the longest miles to shop (55.01 miles). In
addition, the public transit distance for shopping by not-married women (less than 5 miles)
significantly differs from those of other gender and marital status groups, who traveled distances

192

at least more than 20 miles (p-value = 0.074, table 48 and 50) to reach shopping destinations by
public transit.
In terms of walking, married women traveled the greatest miles for shopping (17.81 miles
per week, from table 48). By driving, while both married men and married women traveled the
longest distances for shopping, married women traveled more miles to shop compared to married
men (18.14 miles per week by married women versus 17.32 miles per week by married men).
Not-married women traveled the shortest distance by car (14.80 miles) to reach shopping. In fact,
not-married women engaged in the shortest distances to their shopping destinations by all modes
of travel except walking, implying their limited accessibility and mobility for shopping
opportunities in the inner-city Detroit neighborhoods (table 48).
5.3.1.4.2.

Trips to restaurants

Trips to restaurants, however, reveal very different patterns from shopping trips, and it is
evident in gender and marital status distinctions (table 48). Married men traveled the most
frequently—9.17 times a week—to visit restaurants by walking. Compared to other gender and
marital status groups, married men also traveled significantly more times to access restaurants
(p-value = 0.035, table 51). In addition, both married and not-married men walked more times on
a weekly basis to access restaurants compared to married and not-married women.
By public transit, married men and married women traveled about 2.2 times a week to
access restaurants, which was more frequent when compared to both not-married men and notmarried women. Married men also traveled most frequently by driving to reach restaurant
destinations, 4.31 times a week, while married women drove the least frequently to restaurants.

193

In terms of trip distance to access restaurants, married men traveled the farthest, and this is the
case for all modes of travel.
As pedestrians, married men traveled 10.43 miles per week to restaurants, while married
women engaged in the shortest pedestrian trip distances to restaurants (1.28 miles). In fact, the
average pedestrian trip distance by married men to a restaurant was significantly longer than the
trip distance to access restaurants by any other gender and marital status groups (p-value = 0.041,
table 51). Married men also traveled the longest distances (30.02 miles) by public transit to reach
a restaurant. Both married men and married women traveled significantly longer distances to
restaurants when compared to not-married men and not-married women (p-value = 0.071, table
51). Lastly, married men drove 25.82 miles on average to a restaurant, which significantly differs
from driving trip distances by other gender and marital status groups (p-value = 0.074, table 51).
Both married men and married women traveled significantly longer distances by a car to visit
restaurants when compared to not-married men and not-married women (p-value = 0.079).
5.3.1.4.3.

Trips to personal services

For personal service trips, married men dominated pedestrian trips, in both trip frequency
and distance. Married men traveled 2.59 times a week to reach personal service destinations,
while married women and not-married women traveled twice a week (table 48). However,
married men traveled the least frequently for personal services by public transit (0.76 time a
week), whereas not-married men made the most frequent transit trips to service destination (2.14
times a week). By driving, not-married men drove the most times (3.55 times a week) for
personal services compared to other gender and marital status groups. However, none of the
gender and marital status distinctions in personal service trip frequencies was statistically

194

significant. In terms of trip distances, married men walked significantly longer distances (10.43
miles per week, table 48) than not-married men, married women, and not-married women (pvalue = 0.054, table 52). In fact, both married men and married women walked significantly
longer distances for personal service than not-married men and not-married women (p-value =
0.051, table 52).
With regard to public transit use to access personal services, not-married men traveled
further distances (24.19 miles per week) than married women (3.16 miles per week) and notmarried women (6.66 miles per week) (table 48). There were no married men respondents using
public transit to access personal services. By driving, married men, again, traveled the greatest
miles (16.60 miles per week) to reach personal service destinations, while married women
traveled the fewest miles (8.46 miles per week) (table 48). Both married men and not-married
men in the urban Detroit neighborhoods traveled significantly longer distances by driving than
women, married and not-married (p-value = 0.063, table 52).
5.3.1.4.4.

Trips for leisure

Finally, with regard to leisure destinations, married women walked the most frequently
(4.64 times a week) while married men walked the least frequently (1.72 times a week) to access
leisure activities (table 48). Both married women and not-married women walked more
frequently for leisure than both married men and not-married men, and at a significant level (pvalue = 0.055, table 53). In contrast, both married men and not-married men took public transit
significantly more times to leisure destinations when compared to married women and notmarried women (p-value = 0.060, table 53).

195

By public transit, married men traveled most frequently (5.19 times a week) to access
leisure activities, while not-married women traveled least frequently among all gender and
marital status groups, less than twice a week (table 48). By driving, not-married men traveled
most frequently (3.54 times a week) to access leisure activities (table 48).
For leisure activities, married women walked the greatest miles (5.58 miles per week),
while not-married men walked the shortest distance (1.21 miles per week) (table 48). However,
married men traveled the longest distances by public transit (17.28 miles per week) to reach
leisure destinations. Not-married women traveled the shortest distances by public transit, among
all the groups, to reach leisure activities (11.83 miles per week). When it came to driving by car
to access leisure destinations, not-married men traveled the greatest miles (27.23 miles per
week), while the shortest driving trip distances to reach leisure destinations was by not-married
women (16.84 miles). This implies that not-married women were spatially constrained for their
leisure activities due to limited access to a car and personal time. In urban Detroit
neighborhoods, women, regardless of marital status, traveled longer distance as pedestrians for
leisure activities whereas men traveled longer distances by public transit and driving. The only
gender difference that was statistically significant in driving distances was for leisure activities
(p-value = 0.096, table 53).
5.3.1.5.

Discussion

In the urban Detroit neighborhoods, the OLS regression results illustrated that women
traveled more frequently and greater miles as pedestrians to work compared to men. However,
women traveled less frequently and fewer miles than men by motorized travel modes, both
taking public transit and driving to work. In contrast, for non-work trips related to household

196

responsibilities—such as shopping and personal services—women traveled more frequently by
car. This is consistent with findings among some of the leading gender travel studies (Hamilton
and Jenkins, 1990; Hanson and Hanson, 1980a, 1980b; Hanson and Johnston, 1985; Hanson and
Giuliano, 2004; Jones, 1990; and Pickup, 1985). Consequently, the fact that women drove more
frequently only for shopping and personal services confirms women had less access to a private
automobile for daily activities, except for household-related activities. In other words, in urban
Detroit, women suffer from greater travel burdens due to their more limited access to a car
regardless of age or income. Therefore, Detroit urban women appear to have prioritized
household affairs, given their gender role, whenever the private automobile was made available.
In addition, given the fact that women traveled shorter distances in general compared to
men reveals that women’s spatial extent is more confined compared to that of men, due to their
obligation in meeting their daily responsibilities at home and at work. This is again consistent
with prior research on the topic on gendered travel behavior (Kwan, 1999a, 1999b, 2000).
Furthermore, as shown in previous studies, women in the Detroit urban neighborhoods traveled
fewer miles due to their more limited access to a car. Detroit women’s shorter trip distances also
reveal their more limited accessibility to urban opportunities, a particularly troubling factor given
that these neighborhoods are experiencing urban decline. Notice, however, that the Detroit urban
neighborhoods—due to their built environment characteristics, the higher densities, mixed land
uses, and street connectivity—should, in fact, have higher level of accessibility to urban
amenities.
With decentralizing urban opportunities, such as employment, retail and personal
services, populations in the lower income neighborhoods face greater travel burdens in the form
of longer trip distances to suburban destinations in order to meet their daily needs. There is an
197

added burden for women compared to men within this context. Lower-income women in
Detroit’s east side neighborhoods are further constrained in their travel as a result of their greater
dependence on walking, biking and public transit regardless of their age or income status.
Furthermore, the fact that women do not travel in public transit as much as men—in
terms of trip frequency (except shopping) and distance (except leisure)—and in part likely a
result of safety related concerns, implies an added layer of burdens placed on women in their
daily travel patterns. Consequently, exploring lower-income women’s travel behavior at the
disaggregate neighborhood level contributes to understanding more accurately socially
marginalized groups’ travel constraints in a declining urban area such as Detroit. However, in
spite of these relationships between gender and individual travel behavior, there were only a few
variables, specifically age and income, which significantly explained these travel differences.
Additionally, it is possible that the results for pedestrian and public transit work trips might be
skewed due to limited Ns from a high unemployment rate and scarce employment opportunities
within the city of Detroit. Due to a lower return rate from Detroit urban neighborhoods, the
smaller numbers of respondents might have also influenced the regression results for pedestrian
and public transit trips for non-work trips.
From the ANCOVA results, more specific mean differences between gender and marital
status groups were identified for various travel activities at a disaggregate level. Overall, women
made more pedestrian trips compared to men, while men traveled more in public transit and cars.
For work, in particular, women traveled fewer miles than men even though married women
traveled most frequently as pedestrians and by driving. Therefore, married women’s work travels
were much more geographically confined, being closer to home, which is a finding consistent
with the existing literature (Kwan, 1999a, 1999b, 2000). Women’s travel for household
198

responsibilities can be also examined based on their trips for household duties. Married women
in the urban Detroit neighborhoods traveled most frequently and the farthest distances for their
shopping trips regardless of travel mode. Married men, in contrast, traveled most frequently and
the farthest distance in accessing restaurants. In addition, between the married and not-married
respondents, ones in marriage made more frequent trips for shopping, but less for leisure.
Another important finding was that not-married women traveled by public transit across
the greatest distances in reaching work, personal services, and leisure destinations, but fewer
miles for shopping and restaurant visits when compared to married women. When it came to
driving, however, not-married women traveled fewer miles when compared to married women
for all activities except restaurant visits. Since not-married women tended to have a lower
income when compared to married women, married men and not-married men (Light, 2004; and
U.S. Census Bureau 20105), their greater use of public transit is likely a result of lower personal
automobile access due to the more limited incomes.
5.3.2.

Higher Density Suburban (Ann Arbor and Birmingham

Neighborhoods): Disaggregate analyses on individual travel behavior in the
high density suburban neighborhoods in the Detroit region
In this section, I will explore the differences in travel patterns between women and men
in robust and dynamic higher density suburban neighborhoods. These neighborhoods—in Ann
Arbor and Birmingham—are characterized by higher densities, mixed land uses and connected
street networks, a similar built environment to the Detroit urban neighborhoods, but unlike the
city of Detroit neighborhoods, they are not experiencing disinvestment and decline. The OLS

5

U.S. Census Bureau, Income, Poverty, and Health Insurance Coverage In the United States: 2009, Current
Population Report, September, 2010.

199

regression analyses will examine the relationships between gender and individual travel behavior
for an array of daily activities (such as work, shopping, restaurants, personal services, and
leisure) after controlling for socio-economic and demographic factors (income, the number of
personal vehicles available, age, and the number of dependent children). In turn, the ANCOVA
will reveal the differences between gender and marital status groups in travel behavior for
disaggregate travel activities. Again, for both models, I used trip frequency and distance as
dependent variables that are transformed using square-root and logarithm respectively.

200

Table 54 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the higher density suburban neighborhoods: travel frequency
Total trip frequency1
N2
β
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

S.E.

Df

P-value

75
32
243

-.339
-.230
-.048

.139
.201
.057

3
3
3

.018
.264
.400

89
13
349

-.376
-.264
.108

.129
.336
.068

3
3
4

.005
.459
.113

85

.000

3

.999

233

-.063

.124
n/a
.079

3

.422

91
8
287

.028
-2.045
.059

.129
.249
.079

3
3
3

.829
.004
.455

154
8
258

.068
-.389
-.131

.139
.610
.095

3
3
4

.626
.558
.170

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

Total trip frequency per person on a weekly basis (trip count, square root transformed);
N = number of resident men and women;
Gender dummy: women = 1, and men = 0, and controlled for constant (*), income, and age (*);
Controlled for constant (**), age and income (*);
Controlled for constant, age (*) and income (***);
Controlled for constant (**), age and income;
Controlled for constant (**), age and income (**);
Controlled for constant (***), age (*), income, and personal vehicle access (**);
Controlled for constant (**),age and income;
Not available;
Controlled for constant, age and income;
Controlled for constant (***),age and income (*);
Controlled for constant (***),age (**) and income, the tolerance values were above 0.300 for all
predictors (n=8);
14. Controlled for constant, age (***) and income;
15. Controlled for constant (***), age and income (*);
16. Controlled for constant, age and income, the tolerance values were above 0.200 for all predictors
(n=8);
17. Controlled for constant, age, income and personal vehicle access (**).
For all other models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

201

Table 55 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the higher density suburban neighborhoods: travel distance
Cumulative trip distance1
N2
β
S.E.
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

Df

P-value

65
28
223

-.115
-.285
-.312

.256
.363
.189

3
4
3

.656
.440
.102

70
12
371

-.327
-.048
.135

.200
.501
.097

3
3
3

.107
.927
.165

57

.074

3

.785

205

-.143

.268
n/a
.143

3

.316

53
8
216

.001
-1.730
-.157

.181
1.683
.144

3
3
3

.994
.380
.276

75

-.371

3

.065

168

-.299

.197
n/a
.179

5

.096

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

Cumulative trip distance to destinations per person on a weekly basis (mile, logarithm transformed);
N = number of resident men and women;
Gender dummy: women = 1, and men = 0, and controlled for constant, income, and age (*);
Controlled for constant (**), age, income (*), and personal vehicle access (**);
Controlled for constant, age and income (**);
Controlled for constant, age and income;
Controlled for constant, age (*) and income;
Controlled for constant (***), age and income;
Controlled for constant, age and income;
Not available;
Controlled for constant, age (***) and income;
Controlled for constant, age and income;
Controlled for constant, age and income, the tolerance values were above 0.300 for all predictors
(n=8);
14. Controlled for constant, age (**) and income;
15. Controlled for constant, age (**) and income;
16. Not available;
17. Controlled for constant, age, income, personal vehicle access (**), and the number of dependent
children (**).
For all other models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

202

5.3.2.1.

Gender Difference in Travel Behavior: Trips to Work

In table 54 and 55, OLS regression results demonstrate that in the higher density
suburban neighborhoods, women made less frequent trips to work and traveled shorter distances
when compared to men, regardless of travel modes. Travelling to work, women walked
significantly fewer times than men (B = -0.339, p-value = 0.018) after controlling for socioeconomic and demographic factors, such as personal income, age, personal vehicle access, and
the number of dependent children (table 54). Women in the higher density suburban
neighborhoods also walked fewer miles to work than men. In exploring travel to work by public
transit, women respondents in the higher density suburbs traveled less frequently and shorter
distances to work compared to men. In addition, within these higher density suburban
neighborhoods, women were also less likely to travel by car to work and they also traveled fewer
miles than men by car to reach work.
These patterns can be expected, as the existing gender and transportation literature has
shown comparable outcomes, with women travelling less frequently and shorter distances to
work, and by all modes of travel (Hanson and Hanson, 1980a, 1980b; Hanson and Pratt, 1991;
Hanson and Giuliano, 2004; and Rosenbloom, 2005). The fact that women in higher density
suburbs drove fewer times and shorter distances supports a traditional pattern in gendered travel,
with work trips dominated by men. This is an important pattern to recognize since the women in
the higher density suburbs maintained higher incomes—and had greater access to a car—
compared to lower income women in urban Detroit. Despite the higher personal incomes and the
greater car access, women in the higher density suburban neighborhoods still traveled to work
less frequently and shorter distances by driving compared to women in urban Detroit.

203

5.3.2.2.

Gender Differences in Travel Behavior: Trips to Shops, Restaurants,

Personal Services and Leisure
5.3.2.2.1.

Trips to shopping

Table 54 and 55 illustrates how women travel differently to meet their daily needs,
including household responsibilities, compared to men. The OLS regression results demonstrate
that in the higher density suburban neighborhoods, women in general, made less frequent trips
and traveled shorter distances for daily activities when compared to men. For shopping, women
in the higher density suburbs traveled significantly fewer times (B = -0.376, p-value = 0.005,
table 54) and shorter distances as pedestrians than men. Women were also less likely to travel by
public transit for shopping compared to men, both in terms of frequency and distance. However,
by driving, women in the higher density suburban neighborhoods traveled more frequently and
greater miles than men for shopping.
5.3.2.2.2.

Trips to restaurants

For restaurant visits, there was no difference between men and women in trip frequency
(table 54). While women were more likely to walk to a restaurant than men, the difference was
minor (table 55). Gender differences in trips by public transit were not available since there were
no respondents who took public transit to restaurants in the higher density suburban
neighborhoods. By driving, however, women traveled less frequently and shorter distances than
men for restaurant visits, which parallels the outcomes in urban Detroit. In other words, men in
both the higher density urban and the higher density suburban neighborhoods drove more
frequently and longer distances to access restaurants.
5.3.2.2.3.

Trips to personal services

204

For travel to access personal services, women in the higher density suburbs walked more
times and longer distances, but the differences were not all that significant. In contrast, women in
the higher density suburbs took significantly fewer public transit trips to access personal services
compared to men (B = -2.045, p-value = 0.004, table 54). Women also traveled shorter distances
than men by public transit to reach personal services. However, there were only a few
respondents that utilized public transit for travel to access personal services (n=8, table 54 and
55). With regard to driving, women in the higher density suburban neighborhoods traveled more
frequently than men for personal services (table 54), however, women traveled shorter distances
by driving when compared to men (table 55).
5.3.2.2.4.

Trips for leisure

With regard to leisure destinations, women in the higher density suburban neighborhoods
traveled more frequently as pedestrians than men (table 54). However, the women respondents
living in the higher density suburban neighborhoods walked significantly shorter distances for
leisure activities (B = -0.371, p-value = 0.065, table 55). By public transit, women traveled fewer
times for their leisure activities compared to men, but as with all cases of public transit use in the
suburbs, there is a limited sample size (n=8, table 54). By driving, women made fewer trips
(table 54) and traveled significantly shorter distances (B = -0.299, p-value = 0.096) compared to
men to reach leisure destinations. Another important factor in determining driving leisure trips
was personal vehicle access (for both trip frequency and distance, table 54 and 55) and the
number of dependent children (for trip distance, table 55). Thus, in the higher density suburban
neighborhoods, a woman’s leisure opportunity was limited if she did not have personal access to
a car or if she had more dependent children. In the higher density suburban neighborhoods,
differences in individual travel behavior between women and men were also observed within the
205

context of marital status (table 56). Particularly, from the ANCOVA results, there were clearly
observed divergences in travel patterns for disaggregated travel activities for married and nonmarried women and men (tables 57-61). The mean differences between gender and marital status
groups were adjusted with two covariates, income and age.

206

Table 56 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the higher density suburban neighborhoods
Men

Work

Trip
Frequency
Trip
Distance

Shopping

Trip
Frequency
Trip
Distance

Restaurant
visit

Trip
Frequency
Trip
Distance

Service

Trip
Frequency
Trip
Distance

Leisure

Trip
Frequency
Trip
Distance

Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving

Notmarried
3.80
4.10
4.27
3.34
28.27
56.01
2.54
7.71
2.49
1.43
5.13
5.19
2.20
n/a
1.99
1.94
n/a
5.77
2.47
9.53
2.40
3.70
7.47
5.12
3.07
3.31
3.57
1.63
n/a
18.47

Women

Married
3.90
3.06
4.62
6.54
5.14
45.07
2.81
0.99
3.31
3.35
2.15
7.06
1.92
n/a
1.96
1.26
n/a
6.02
1.76
n/a
2.40
1.27
n/a
5.78
2.74
1.73
3.10
2.98
n/a
11.71

Notmarried
3.40
2.37
4.36
4.78
3.04
32.35
2.02
1.22
2.99
2.30
2.82
6.35
1.48
n/a
1.68
1.77
n/a
4.74
1.87
1.78
2.54
1.61
3.25
4.57
3.86
0.69
3.61
5.03
n/a
17.24

Married

*Calculated with original weekly total trip frequency and cumulative trip distance data.

207

2.41
2.85
4.24
6.65
5.44
38.40
1.92
1.40
3.57
1.88
1.38
7.76
1.70
n/a
2.01
3.84
n/a
4.88
1.75
1.28
2.45
1.34
1.71
5.53
3.03
1.96
2.62
1.48
n/a
6.45

Table 57 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—work
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.869
(.214)
2.008
(.269)
2.020
(.075)
1.181
(.485)
2.797**
(.450)
3.318
(.252)

Women

Married
1.866
(.110)
1.612
(.223)
2.122
(.048)
1.394
(.198)
0.957
(.351)
3.208
(.170)

NotMarried
1.761
(.204)
1.468
(.219)
2.053
(.059)
1.164
(.345)
1.207
(.379)
2.853
(.204)

N

Married
1.482
(.098)
1.581
(.158)
2.018
(.038)
1.356
(.172)
1.396
(.247)
2.902
(.128)

75
32
243
65
30
223

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.898593, AGE = 46.39.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.764266, AGE = 47.53.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.862601, AGE = 45.92.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.943127, AGE = 45.97.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.821237, AGE = 46.90.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.868839, AGE = 45.29.
Mean difference between gender groups was only significant for driving trip distance (p-value = .048).
Mean difference between marital status groups was only significant for public transit trip distance (p-value
= .037).
Mean difference between gender and marital status groups was only significant for public transit trip
distance (p-value = .010).

208

Table 58 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—shopping
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.563
(.171)
2.532*
(.486)
1.510
(.100)
0.606
(.303)
1.549
(1.030)
1.362
(.146)

Married
1.582
(.115)
1.024
(.202)
1.700
(.059)
0.869
(.177)
0.585
(.436)
1.603
(.086)

Women
NotMarried
Married
1.305
1.270
(.149)
(.084)
1.001
1.145
(.178)
(.238)
1.673
1.787
(.067)
(.045)
0.742
0.516
(.238)
(.130)
0.688
0.322
(.405)
(.490)
1.624
1.697
(.098)
(.066)

N

89
13
374
70
12
371

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.868290, AGE = 51.61.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.701677, AGE = 55.77.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.856635, AGE = 50.62.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.870293, AGE = 52.23.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.622427, AGE = 56.83.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.854712, AGE = 50.54.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value = .038),
public transit trip frequency (p-value = .038), driving trip frequency (p-value = .079), and driving trip
distance (p-value = .085).
Mean difference between marital status groups was only significant for public transit trip frequency (pvalue = .045), and driving trip frequency (p-value = .031).
Mean difference between gender and marital status groups was only significant for public transit trip
frequency (p-value = .055).

209

Table 59 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—restaurant visits
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

1.249
(.110)
n/a

Women
NotMarried
Married
1.126
1.211
(.142)
(.082)
n/a
n/a

1.331
(.115)
0.401
(.422)
n/a

1.316
(.065)
0.528
(.205)
n/a

1.171
(.082)
0.178
(.335)
n/a

1.295
(.055)
0.571
(.159)
n/a

233

1.107
(.205)

1.329
(.123)

1.140
(.157)

1.127
(.101)

205

NotMarried
1.399
(.237)
n/a

Married

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.849140, AGE = 47.41.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.581218, AGE = 57.00.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.861317, AGE = 51.71.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.751607, AGE = 47.00.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
9.819338, AGE = 58.00.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.865119, AGE = 51.80.

210

85
(3)

57
(2)

Table 60 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—personal services
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

Women

NotMarried
Married
1.525
1.214
(.195)
(.116)
3.162 n/a
(.213)
1.381
1.433
(.131)
(.067)
***
1.571
0.373
(.367)
(.167)
2.169
n/a
(1.443)
1.304
1.394
(.249)
(.129)

NotMarried
1.222
(.136)
1.283
(.164)
1.472
(.077)
0.347
(.169)
0.453
(1.111)
1.147
(.135)

Married
1.214
(.081)
1.117
(.070)
1.469
(.051)
0.379
(.114)
0.439
(.471)
1.231
(.093)

91
8
287
53
8
216

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.825470, AGE = 48.93.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.414687, AGE = 47.63.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.842210, AGE = 52.07.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.793998, AGE = 47.04.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.414687, AGE = 47.63.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.831961, AGE = 51.14.
Mean difference between gender groups was only significant for public transit trip frequency (p-value =
.004), and pedestrian trip distance (p-value = .011).
Mean difference between marital status groups was only significant for pedestrian trip distance (p-value =
.013).
Mean difference between gender and marital status groups was only significant for pedestrian trip
distance (p-value = .009).

211

Table 61 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the higher density suburban neighborhoods—leisure
Men

Trip
freq.1

Pedestrian3
Public
transit4
Driving5

Trip
dist.2

Pedestrian6
Public
transit7
Driving8

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
1.648
(.233)
1.535
(.254)
1.787
(.144)
0.764
(.441)
n/a
2.180
(.256)

Married
1.492
(.116)
1.303
(.092)
1.579
(.080)
0.812
(.166)
n/a
1.730
(.148)

Women
NotMarried
Married
1.725
1.606
(.149)
(.085)
*
0.873
1.398
(.043)
(.094)
1.783
1.507
(.091)
(.058)
0.845
0.513
(.226)
(.122)
n/a
n/a
2.159
(.174)

1.444
(.117)

N

154
8
273
75
(4)
187

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.843167, AGE = 48.05.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.728356, AGE = 55.25.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.873971, AGE = 50.16.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.775058, AGE = 48.28.
7. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.217641, AGE = 51.00.
8. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.890388, AGE = 50.55.
Mean difference between gender groups was only significant for public transit trip frequency (p-value =
.004).
Mean difference between marital status groups was only significant for driving trip distance (p-value =
.001).
Mean difference between gender and marital status groups was only significant for public transit trip
frequency (p-value = .053).

212

5.3.2.3.

Gender and Marital Status Distinctions in Travel Behavior: Trips to

Work
For their work trips, as shown in the OLS regression results, the ANCOVA results
demonstrate that women make fewer trips and travel shorter distances to work even after taking
account marital status. Incorporating marital status, however, revealed more nuanced variations
in commutes to work between men and women. First, married women traveled to work least
frequently (2.41 times a week) as pedestrians. On the other end of the spectrum, both married
and not-married men traveled to work most frequently (3.90 and 3.80 times a week) as
pedestrians (table 56). Married women preferred not to walk to work, likely due to the
inconvenience and less efficient time allocation given their household responsibilities
(Rosenbloom, 2005).
In terms of distance to work, not-married men walked the shortest distances (3.34 miles
per week) among all gender and marital status groups. Both married men and married women
walked longer distances to work than not-married men and not-married women. As pedestrians,
both married and not-married women traveled fewer times and shorter distances than both
married and not-married men, but the differences were not statistically significant.
With respect to travel to work by public transit, not-married men traveled to work most
frequently (4.10 times a week) by transit, while not-married women traveled least frequently
(2.37 times a week) by this mode of travel. Both married men and not-married men used public
transit more frequently to go to work when compared to both married women and not-married
women. When compared to all gender and marital status groups, not-married men traveled the
longest distances to get to work by public transit and at a significant level (28.27 miles per week,
p-value = 0.010 from table 57). Not-married women, in contrast, traveled the shortest distance to
213

work by public transit (3.04 miles per week). Due to the significantly longer trip distances to
work by not-married men, marital status differences in public transit trip distance was
significant; not-married men and not-married women traveled longer distances by public transit
to work than married men and married women (p-value = 0.037) (table 57).
Lastly, married men drove to work most frequently (4.62 times a week), while married
women drove to work least frequently (4.24 times a week). The gender and marital status group
who drove the greatest miles to work were the not-married men (56.01 miles per week). In
contrast, not-married women drove the shortest distances to work (32.35 miles per week) (table
56). Married and not-married men traveled significantly longer distances by driving to work
when compared to married and not-married women (p-value = 0.048) (table 57).
5.3.2.4.

Gender and Marital Status Distinctions in Travel Behavior: Trips to

Shops, Restaurants, Personal Services, and Leisure
5.3.2.4.1.

Trips to shopping

In addition to significant gender differences in shopping trips, as illustrated in the OLS
regression results, the ANCOVA results also reveal significant differences between gender and
marital status groups in non-work activities in the higher density suburban neighborhoods. For
shopping trips, married men walked most frequently (2.81 times a week), while married women
walked the fewest times (1.92 times a week) (table 56). In fact, both married and not-married
women walked fewer times to shopping when compared to married and not-married men, and at
a significant level (p-value = 0.038, table 58). This is a different pattern when compared to the
urban Detroit neighborhoods, where women walked more frequently to shop than men. This is
likely due to the fact that women in the higher density suburban neighborhoods, with a higher
personal income and greater access to a car, prefer driving to shopping destinations and their
214

greater access to resources allows them to do so. As pedestrians, married men walked the longest
distance to shop (3.35 miles per week) when married women walked shorter distance (1.88 miles
per week) than married men. Again, compared to women in urban Detroit, who walked 4.7 miles
per week for shopping, women in the higher density suburban neighborhoods had the resources,
and the choice, not to walk for their shopping trips.
By public transit, not-married men traveled the most frequently for shopping (7.71 times
a week), while married men traveled the fewest times (once a week) (table 56). Not-married
men, in fact, traveled significantly more times by public transit for shopping than any other
group (p-value = 0.055) (table 58). Therefore, both gender differences (men traveled more
frequently than women) and marital status differences (the not-married traveled more frequently
than the married) are apparent with the use of public transit to shop, with public transit shopping
trip frequencies being statistically significant (p-value = 0.038, and 0.045 respectively) (table
58).
Similar to walking trip distances for shopping, married women in the higher density
suburban neighborhoods traveled the shortest distances by public transit (1.38 miles per week).
This is significantly different from the urban Detroit neighborhoods, where married women
traveled the longest distances to shop (some 55 miles per week) by public transit. Both married
and not-married women traveled shorter distances by public transit to shop when compared to
married and not-married men. In addition, both married men and married women traveled shorter
distances by public transit to shop when compared to not-married men and not-married women.
By driving, however, married women traveled most frequently (3.57 times a week) and
over the longest distances (7.76 miles per week) for shopping. Again, this particular gender and

215

marital status group traveled the fewest miles by walking or public transit. Therefore, it confirms
that married women select to drive when shopping—when resources allow them to do so—for
convenience and efficient time allocation, and this is because of their multiple household
responsibilities (Hanson and Hanson, 1980a, 1980b; Jones, 1990; Kwan, 1999a, 1999b; 2000;
and Hanson and Giuliano, 2004). Both married and not-married women drove significantly
longer distances when compared to married and not-married men for shopping (p-value = 0.085).
Between ones who are married and not-married, both married men and married women
drove longer distances to shop than not-married men and not-married women. Not-married men
traveled by driving to shop the least frequently and the shortest distances. The married preferred
driving a car for their shopping trips due to the convenience and being able to purchase and
transport more shopping goods given the greater number of members in their household, and it
was accommodated by their greater personal income. The results, therefore, confirm
differentiated trip allocations between travel modes for shopping purposes by gender and marital
status.

216

5.3.2.4.2.

Trips to restaurants

In contrast to shopping, men traveled more frequently than women for restaurant trips in
the higher density suburban neighborhoods (table 56). As pedestrians, not-married men traveled
most frequently (2.20 times a week) to visit a restaurant while not-married women traveled least
frequently (1.48 times a week). Both married and not-married men also walked more frequently
to restaurants when compared to married and not-married women. However, married women
walked the longest distance to go to a restaurant (3.84 miles).
By driving, not-married men, again, traveled most frequently, 1.96 times a week (table
56), to a restaurant. Both married and not-married men traveled more frequently by car for
restaurant visits when compared to married and not-married women. Similar to the urban Detroit
neighborhoods, married men drove the longest distances for restaurant visits (6.02 miles per
week). However, there was no mean difference between gender and marital status groups that
was significant when it came to travel to restaurants (table 59). In addition, there were very few
respondents that travelled to restaurants by public transit—unlike in the case of the urban Detroit
neighborhoods—which meant that no actual trip patterns between the groups was evident for this
particular travel mode.
5.3.2.4.3.

Trips to personal services

It was also difficult to find significant mean differences in personal service trips (table 56
and 60). For personal services, not-married men traveled the most frequently (2.47 times a week)
and for the longest distance (3.70 miles per week) as pedestrians (table 56). Both married men
and married women walked less frequently for personal services than not-married men and notmarried women. Due to not-married men’s more frequent and farther (p-value = 0.009) walking

217

trips for personal services, men, regardless of marital status, walked significantly further
distances than women for personal services (p-value = 0.011) (table 60). Subsequently, among
marital status groups, not married groups walked significantly further distances than married
groups (p-value = 0.013) (table 60) to access personal services.
With regard to public transit trips to personal services, not-married men also traveled
most frequently (9.53 times a week) and the greatest miles (7.47 miles per week) to access
personal services. None of the married men traveled by public transit for personal services.
However, women, both married and not-married, made more frequent personal service trips
(2.45 and 2.54 times a week) by driving than men in the higher density suburban neighborhoods
(table 56). Despite the more frequent trips, both married and not-married women drove shorter
distances for personal services than married and not-married men. By driving, married men
drove the longest distance (5.78 miles per week) to reach their personal service destinations
(table 56).
5.3.2.4.4.

Trips for leisure

In the higher density suburbs, there were clear differences in leisure trips between
married and not-married respondents. As pedestrians, not-married women traveled the most
frequently (3.86 times a week) for leisure activities, while married men traveled the least
frequently (2.74 times a week). For leisure, both married and not-married women in the higher
density suburban neighborhoods walked more frequently than married and not-married men.
Not-married women also walked longer distance (5.03 miles per week) for leisure than any other
gender and marital status group. Married women, however, walked the shortest distance for
leisure (1.48 miles per week) (table 56).

218

By public transit, while not-married men traveled most frequently to reach their leisure
activity destinations (3.31 times a week, table 56), not-married women utilized public transit the
least, and significantly so (.69 time per week) among all gender and marital status groups (pvalue = 0.053, table 61). Accordingly, women’s groups in the higher density suburbs traveled
significantly fewer times by public transit than men to reach their leisure destinations (p-value =
0.004) by public transit (table 61). There were too few respondents to identify a gender and
marital status variation in public transit trip distances for leisure activities (table 61).
By driving, married women in the higher density suburban neighborhoods made the least
frequent trips to access leisure activities and traveled the shortest distance to leisure destinations
(table 56). Both not-married men and not-married women drove more than married men and
married women to reach their leisure destinations. Respondents that are married also drove
shorter distances to reach their leisure destinations when compared to respondents that are not
married. While not-married men and not-married women traveled by car 18.47 and 17.24 miles
to reach their leisure destinations, married men and married women traveled in a car 11.71 and
6.45 miles respectively. The mean difference between the married and the not-married groups
was also statistically significant (p-value = 0.001, table 61). In terms of gender differences
evident in travel for leisure, both married men and not-married men drove more frequently and
greater distances when compared to married women and not-married women. However, the
gender differences were not statistically significant.
5.3.2.5.

Discussion

First, the OLS regression results revealed that after controlling for socio-economic and
demographic factors, women traveled less frequently than men to work regardless of travel

219

modes in the higher density suburban neighborhoods. The OLS regression results also
demonstrated women’s shorter work trip distances compared to men. This is a consistent finding
with the existing literature on commuting (Hanson and Hanson, 1980a, 1980b, 1981; Hanson and
Pratt, 1991; Hanson and Johnston, 1985; Hanson and Giuliano, 2004: and Rosenbloom, 2005).
Thus, in spite of women’s increased participation in paid labor (Hanson and Giuliano, 2004;
Rosenbloom, 2005), work trips are still dominated by men.
It is important to note that, however, that unlike women respondents in urban Detroit,
women in higher density suburban neighborhoods did not walk more frequently to go to work
than men. With statistical significance, women’s less frequent pedestrian work trips can be
interpreted as the effects of greater access to a private automobile, a function of higher personal
incomes in the higher density suburban neighborhoods, when compared to urban Detroit.
Furthermore, even for shopping, women in the high density suburban neighborhoods
were less likely to walk (p-value = 0.005) or to take public transit, unlike the women in Detroit
who traveled more frequently as pedestrians and by public transit. Women in the higher density
suburban neighborhoods drove more frequently for shopping. Women also traveled longer
distances than men by driving to their shopping destinations, but shorter distances by walking
and public transit. Therefore, whether a woman does, or does not, have access to a car is an
important factor in understanding a women’s daily travel pattern, and particularly for shopping.
For trips to restaurants and personal services, there was no significant gender effect.
However, women drove fewer times for shorter distances for restaurant visits, similar to the case
in the Detroit urban neighborhoods. For personal services, women walked and drove more
frequently than men, but the differences were rather insignificant. The only statistically

220

significant gender difference was the less frequent public transit use by women to access
personal services. In their personal service travel, women in the higher density suburbs also
traveled shorter distances by public transit. Survey responses, however, were limited because so
few residents in the higher density suburban neighborhoods actually traveled by public transit for
their daily activities.
For leisure activities, women in the higher density suburban neighborhoods only traveled
more frequently than men as pedestrians. Yet they did not walk any further distance for leisure
activities than men, and at a significant level. The women in the higher density suburban
neighborhoods also drove less frequently than men in reaching their leisure destinations. They
also drove significantly shorter distances for leisure activities when compared to men.
In conclusion, the higher incomes among the suburban respondents and the more robust
nature of the outer suburban communities—and particularly when compared to the urban Detroit
neighborhoods, which are experiencing disinvestment and decline—likely played an important
role in the shorter distances and fewer frequencies traveled by public transit in the higher density
suburban neighborhoods. The effect of differing incomes on women’s travel behavior was also
evident in pedestrian trips. Whereas women in urban Detroit traveled more frequently as
pedestrians for work, shopping, and leisure, women in the higher density suburban
neighborhoods did not walk more frequently than men for these same activities.
In fact, the gender differences in pedestrian work trip frequencies in both Detroit urban
and higher suburban neighborhoods were statistically significant, but in opposing directions. In
the higher density suburban neighborhoods, both married and not-married women walked to
work less frequently than both married and not-married men whereas both married and not-

221

married women in urban Detroit neighborhoods walked more frequently to work than married
and not-married men. Therefore, gender differences in individual travel behavior should be
understood more precisely in regards to differing effects of socio-economic and demographic
characteristics, such as personal income and access to a car of particular population sub-groups.
With respect to driving, women in both urban Detroit and the higher density suburban
neighborhoods drove shorter distances than men. Despite the greater income levels and greater
automobile access when compared to the Detroit urban respondents, suburban women still drove
fewer miles than men—in all cases except shopping—after controlling for socio-economic and
demographic factors. This gendered travel pattern is consistent with the findings in existing
studies on gendered differences in travel (Kwan, 1999a, 1999b, 2000; Hanson and Hanson,
1980a, 1980b, 1981; Hanson, 1995; Hanson and Giuliano, 2004; Mclafferty and Preston, 1996 a,
b, c; Preston and Mclafferty, 1992, 1999; Rosenbloom, 2005; and Wyly, 1996, 1998). Therefore,
women, in comparison to men, still travel shorter distances for daily activities, and this is evident
across diverse income groups and neighborhood structures.
From the ANCOVA results, more specific mean differences between gender and marital
status groups were examined for the various travel activities, which were disaggregated by travel
type. Even in the dynamic and robust high density suburban neighborhoods, with a rich bundle
of travel destinations, women traveled less frequently and fewer miles in a private automobile to
work, but drove more frequently and more miles for shopping. Married women in the higher
density suburban neighborhoods drove least frequently to work while married men drove most
frequently in their commutes to work. However, married women drove most frequently and the
longest distance for shopping.

222

Married women, however, preferred not to walk or take public transit, and they, in fact,
maintained the least frequency and the shortest distances in accessing shopping and personal
services by these travel modes. The results, thus, imply that married women in the higher density
suburban neighborhoods drove predominantly for household related activities in order to
maximize the efficiency of their trips. The car allows women to travel at greater speeds, and
gives them greater flexibility, which is important given their more constrained time frame in
their multiple roles as both a worker and household maker (Hanson and Hanson, 1980a; Kwan,
1999a, 1999b, 2000; and Pickup, 1985, 1990).
Another distinctive gender and marital status difference in individual travel behavior was
found in leisure trips. Both married men and married women in the higher density suburban
neighborhoods traveled less frequently as pedestrians and by driving for leisure than not-married
men and not-married women. Higher density suburban residents that are married also traveled
shorter distances to reach their leisure destinations. In fact, married women traveled the shortest
distances regardless of travel modes (no response for public transit, however) of all groups.
Thus, married women allocated their time and resources—for example, ensuring access to a
personal automobile—in order to travel and fulfill household responsibilities, such as shopping
and personal services, instead of their own leisure activities.

5.3.3.

Lower Density Suburban (Bloomfield Hills and West Bloomfield

Neighborhoods): Disaggregate analyses on individual travel behavior in the
lower density suburban neighborhoods in the Detroit region
In order to identify how a differing urban built environment would affect individual
travel behavior—while still continuing to explore gender differences—two lower density
suburban neighborhoods, in Bloomfield Hills and West Bloomfield, were selected for the
223

analyses. Due to characteristics of the low density development patterns, and subsequently, the
limited cases (Ns) for public transit, travel by transit was omitted from this individual travel
behavior analyses. The OLS regression results were used to analyze the relationship between
gender and disaggregated individual travel behavior (all activities including trips to work,
shopping, restaurants, personal services, and leisure) after controlling for socio-economic and
demographic factors. In addition, the ANCOVA results demonstrated significant mean
differences between gender and marital status groups in the individual travel behavior for
disaggregate travel activities, when adjusting for socio-economic and demographic factors. The
dependent variables—trip frequencies and distance—were, once again, transformed using
square-root and logarithm, respectively.

224

Table 62 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the lower density suburban neighborhoods: travel frequency
Total trip frequency1
N2
Β
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

25

-.098

200

-.096

30

.018

311

.211

20

-.252

225

-.107

28

-.078

271

.069

48

.097

248

.204

S.E.

df

P-value

.253
n/a
.076

3

.702

3

.206

.209
n/a
.078

3

.932

3

.007

.270
n/a
.092

3

.365

3

.245

.163
n/a
.078

3

.637

3

.375

.191
n/a
.109

3

.614

3

.062

1. Total trip frequency per person on a weekly basis (trip count, square root transformed);
2. N = number of resident men and women;
3. Gender dummy: women = 1, and men = 0, and controlled for constant, income, and age;
4. Not available;
5. Controlled for constant, age and income;
6. Controlled for constant, age and income;
7. Not available;
8. Controlled for constant, age and income;
9. Controlled for constant, age and income;
10. Not available;
11. Controlled for constant, age and income;
12. Controlled for constant, age and income;
13. Not available;
14. Controlled for constant (***),age and income;
15. Controlled for constant (**), age and income (*);
16. Not available;
17. Controlled for constant, age and income.
For all other models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

225

Table 63 OLS regression results of gender differences in disaggregate analyses on individual
travel behavior in the lower density suburban neighborhoods: travel distance
Cumulative trip distance1
N2
β
S.E.
WORK
Women3
Pedestrian
Women4
Public transit
Women5
Driving
SHOPPING
Women6
Pedestrian
Women7
Public transit
Women8
Driving
RESTAURANT VISITS
Women9
Pedestrian
Women10
Public transit
Women11
Driving
SERVICE
Women12
Pedestrian
Women13
Public transit
Women14
Driving
LEISURE
Women15
Pedestrian
Women16
Public transit
Women17
Driving

19

-.806

176

-.389

22

-.080

291

.391

16

-.136

201

-.095

20

.473

246

-.081

22

-.087

201

.354

1.
2.
3.

df

P-value

.496
n/a
.173

4

.130

3

.026

.400
n/a
.101

3

.844

4

.000

.547
n/a
.147

3

.808

3

.522

.506
n/a
.141

3

.366

3

.567

.414
n/a
.178

3

.836

3

.048

Cumulative trip distance to destinations per person on a weekly basis (mile, logarithm transformed);
N = number of resident men and women;
Gender dummy: women = 1, and men = 0, and controlled for constant, income, age, and personal
vehicle access (**);
4. Not available;
5. Controlled for constant, age (***) and income (*);
6. Controlled for constant, age and income;
7. Not available;
8. Controlled for constant, age, income (**) and the number of dependent children (**);
9. Controlled for constant, age and income;
10. Not available;
11. Controlled for constant, age and income;
12. Controlled for constant, age and income;
13. Not available;
14. Controlled for constant, age and income;
15. Controlled for constant, age and income;
16. Not available;
17. Controlled for constant, income, and age.
For all other models, the tolerance values were above 0.600 for all predictors.
Statistical significance at the 90% (*), 95% (**), and 99% (***) level, respectively.

226

5.3.3.1.

Gender Differences in Travel Behavior: Trips to Work

In table 62, it is shown that, after controlling for socio-economic and demographic factors
(income, age, personal automobile access and the number of dependent children in the
household), women in the lower density suburban neighborhoods were less likely to travel to
work. Women not only walked less frequently to work when compared to men, but they also
drove fewer times. Similar to trip frequency, women in the lower density suburban
neighborhoods traveled shorter distances to work than men, regardless of travel modes (table
63). In addition, women in the lower density suburban neighborhoods drove significantly shorter
distances to reach their place of employment compared to men (B= -0.389, p-value = 0.026).
Therefore, the pattern of women’s less frequent and shorter trip distance to work was
consistently evident across the neighborhood groups in the Detroit region, regardless of urban
density and land use characteristics.
5.3.3.2.

Gender Differences in Travel Behavior: Trips to Shops, Restaurants,

Personal Services and Leisure
5.3.3.2.1.

Trips to shopping

In the lower density suburban neighborhoods, women also traveled more frequently for
activities that are related to household responsibilities, such as shopping and personal services.
The OLS regression results showed that women traveled more frequently as pedestrians for
shopping. Women in the lower density suburban neighborhoods, however, traveled shorter
distances as pedestrians for shopping. Both differences in shopping trip frequency and distance
were minor and statistically insignificant.

227

Women also drove significantly more times for shopping when compared to men in the
lower density suburban neighborhoods, after adjusting for socio-economic and demographic
factors (B= 0.211, p-value = 0.007) (table 62). In addition, women drove significantly longer
distances to reach their shopping destinations than men (B= 0.391, p-value = 0.000) (table 63).
This specific characteristic in shopping patterns, with women travelling longer distances to
access shops, becomes better understood with the added variable of dependent children.
The number of dependent children significantly contributes to increased shopping trip
distance, with women in the lower density suburban neighborhoods travelling further distances
when more dependent children were present in the household. Subsequently, women in such
households were more likely to travel longer distances by driving to shop for the family’s daily
needs. This is consistent with previous research on gendered travel behavior (Hanson, 1995;
Hanson and Hanson, 1980a; Hanson and Pratt, 1991; Hanson and Giuliano, 2004; Jones, 1990;
Pickup, 1990; and Rosenbloom, 2005).
In fact, the burden of women’s trips associated with shopping becomes clearer in the
context of dispersing destinations in the lower density suburbs. In particular, while women in
both the higher density and the lower density suburban neighborhoods drove more times and for
longer distances to shop when compared to men, the gender difference was only statistically
significant for the lower density suburbs.
When comparing the lower density suburban and the urban Detroit neighborhoods, the
lower-income urban minority women also drove more frequently for shopping than men, again,
similar to women in the lower density suburban neighborhoods. However, unlike the suburban
residents, the lower-income minority women in Detroit did not travel longer distances by driving

228

than men for their shopping activities. Thus, there were observable and clear distinctions by
neighborhood types and differing income levels in women’s shopping trips.
5.3.3.2.2.

Trips to restaurants

Women in lower density suburban neighborhoods were less likely to travel to visit a
restaurant as pedestrians and by driving (table 62). As pedestrians, women in the neighborhoods
traveled fewer times and shorter distances to access restaurants when compared to men. Women
in the lower density neighborhoods also travelled less frequently and shorter distances to get to
restaurants by driving (tables 62 and 63). Therefore, across all six Detroit region neighborhoods,
women traveled less frequently and shorter distances than men to access restaurants.
5.3.3.2.3.

Trips to personal services

In accessing personal services, women in the lower density suburban neighborhoods
walked less frequently, but traveled longer distances than men. However, women in these
neighborhoods drove more frequently but shorter distances then men to reach their personal
service destinations (tables 62 and 63). Thus, in the lower density suburban neighborhoods,
women chose to drive more frequently for personal services, but they travelled shorter distances
to reach these destinations by driving when compared to men.
5.3.3.2.4.

Trips for leisure

With respect to travel for leisure, women in the lower density suburban neighborhoods
walked more frequently, but shorter distances when compared to men. However, women traveled
significantly more times by driving to reach their leisure destinations than men (B= 0.204, pvalue = 0.062) (table 62). In order to reach their leisure destination, women also drove longer
distances than men and at a significant level (B= 0.354, p-value = 0.048) (table 63). This is a
229

different outcome when compared to the Detroit urban neighborhoods and the higher density
suburban neighborhoods.
This is a unique travel pattern in the lower density suburban neighborhoods, facilitated by
higher incomes, greater access to a car and longer distances between leisure destinations.
Women in the lower density suburban neighborhoods traveled more frequently and longer
distances for leisure since they had greater access to a personal automobile when compared to
lower-income urban Detroit women respondents. In addition, unlike the higher density suburban
residents, women in the lower density suburban neighborhoods drove more frequently than men
for their leisure activities and traveled longer distances to reach leisure destinations. This could
be expected given the lower density development patterns of these neighborhoods, and the
dispersed nature of destinations across this landscape. The ANCOVA results for disaggregate
travel activities examined the mean differences in individual travel behavior between gender and
marital status groupings (table 64). The detailed breakdowns of these results by trip destination
are provided below.

230

Table 64 Average weekly total trip frequency and cumulative distance (mile)* by gender and
marriage in the lower density suburban neighborhoods
Men

Work

Trip
Frequency
Trip
Distance

Shopping

Trip
Frequency
Trip
Distance

Restaurant
visit

Trip
Frequency
Trip
Distance

Service

Trip
Frequency
Trip
Distance

Leisure

Trip
Frequency
Trip
Distance

Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving
Pedestrian
Public transit
Driving

Notmarried
n/a
n/a
4.73
n/a
n/a
54.95
2.96
n/a
3.02
6.61
n/a
9.98
2.47
n/a
2.23
1.22
n/a
10.96
2.42
n/a
3.31
1.71
n/a
15.66
2.87
n/a
2.74
2.96
n/a
28.20

Women

Married
2.94
n/a
5.11
14.98
n/a
82.02
1.69
n/a
3.37
4.01
n/a
11.08
1.68
n/a
2.44
6.49
n/a
10.72
1.13
n/a
2.71
2.35
n/a
8.58
2.42
n/a
2.70
4.91
n/a
17.83

Notmarried
1.90
n/a
4.67
10.16
n/a
45.43
1.96
n/a
3.78
4.14
n/a
16.29
2.14
n/a
1.91
18.08
n/a
9.60
1.47
n/a
3.20
3.15
n/a
9.40
2.53
n/a
3.19
4.16
n/a
21.44

Married

*Calculated with original weekly total trip frequency and cumulative trip distance data.

231

2.81
n/a
4.56
13.13
n/a
57.90
1.34
n/a
3.93
2.12
n/a
12.47
0.93
n/a
2.11
4.74
n/a
9.41
1.20
n/a
2.94
3.68
n/a
9.04
2.77
n/a
3.34
2.33
n/a
19.68

Table 65 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—work
Men

Trip
freq.1

Pedestrian3
Public
transit
Driving4

Trip
dist.2

Pedestrian5
Public
transit
Driving6

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

NotMarried
n/a
n/a
2.118
(.128)
n/a
n/a
3.808
(.297)

N

1.604
(.183)
n/a

Women
NotMarried
Married
1.283
1.608
(.457)
(.193)
n/a
n/a

2.201
(.057)
2.478
(.311)
n/a

2.1
(.098)
2.29
(1.009)
n/a

2.087
(.056)
2.102
(.357)
n/a

195

3.94
(.134)

3.415
(.236)

3.536
(.129)

173

Married

24
n/a

19
n/a

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.992268, AGE = 55.25.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.011105, AGE = 51.48.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.062639, AGE = 55.58.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.041770, AGE = 51.03.
Mean difference between gender groups was only significant for driving trip distance (p-value = .066).

232

Table 66 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—shopping
Men

Trip
freq.1

Pedestrian3
Public
transit
Driving4

Trip
dist.2

Pedestrian5
Public
transit
Driving6

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

1.142
(.132)
n/a

Women
NotMarried
Married
1.356
1.155
(.231)
(.227)
n/a
n/a

1.655
(.122)
1.388
(.484)
n/a

1.697
(.065)
1.405
(.291)
n/a

1.824
(.092)
1.291
(.410)
n/a

1.898
(.056)
1.389
(.584)
n/a

305

1.906
(.162)

1.952
(.085)

2.432
(.122)

2.276
(.072)

299

NotMarried
1.676
(.273)
n/a

Married

30
n/a

22
n/a

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.920003, AGE = 52.53.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.901893, AGE = 56.23.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.876764, AGE = 56.36
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.908568, AGE = 55.75.
Mean difference between gender groups was only significant for driving trip frequency (p-value = .037),
and driving trip distance (p-value = .000).

233

Table 67 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—restaurant visits
Men

Trip
freq.1

Pedestrian3
Public
transit
Driving4

Trip
dist.2

Pedestrian5
Public
transit
Driving6

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

1.190
(.188)
n/a

Women
NotMarried
Married
1.452
0.724
(.330)
(.239)
n/a
n/a

1.380
(.167)
1.357
(1.936)
n/a

1.417
(.073)
1.827
(.410)
n/a

1.271
(.111)
2.318
(.642)
n/a

1.302
(.068)
1.198
(.559)
n/a

220

1.924
(.276)

1.906
(.120)

1.940
(.180)

1.775
(.110)

197

NotMarried
1.498
(.420)
n/a

Married

20
n/a

16
n/a

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.000764, AGE = 54.20.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.941513, AGE = 56.17.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.990896, AGE = 55.13.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.939777, AGE = 56.60.

234

Table 68 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—personal services
Men

Trip
freq.1

Pedestrian3
Public
transit
Driving4

Trip
dist.2

Pedestrian5
Public
transit
Driving6

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

1.051
(.122)
n/a

Women
NotMarried
Married
1.182
0.998
(.138)
(.142)
n/a
n/a

1.652
(.122)
0.647
(.599)
n/a

1.548
(.065)
0.620
(.355)
n/a

1.671
(.091)
1.283
(.519)
n/a

1.637
(.054)
0.711
(.511)
n/a

267

1.873
(.212)

1.839
(.120)

1.833
(.165)

1.754
(.094)

242

NotMarried
1.513
(.192)
n/a

Married

28
n/a

20
n/a

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.919367, AGE = 53.39.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.895376, AGE = 56.28.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.023041, AGE = 48.75.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.902581, AGE = 55.83.
Mean difference between gender groups was only significant for pedestrian trip frequency (p-value =
.053).

235

Table 69 ANCOVA results: average total trip frequency and cumulative distance by gender and
marriage in the lower density suburban neighborhoods—leisure
Men

Trip
freq.1

Pedestrian3
Public
transit
Driving4

Trip
dist.2

Pedestrian5
Public
transit
Driving6

Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)
Mean
(S.E.)

N

1.474
(.190)
n/a

Women
NotMarried
Married
1.478
1.507
(.251)
(.144)
n/a
n/a

1.554
(.181)
0.828
(.491)
n/a

1.459
(.090)
1.387
(.538)
n/a

1.654
(.118)
1.210
(.350)
n/a

1.657
(.072)
0.964
(.345)
n/a

243

2.704
(.276)

2.143
(.152)

2.587
(.172)

2.556
(.115)

196

NotMarried
1.563
(.291)
n/a

Married

47
n/a

21
n/a

1. Dependent variable was transformed with square-root.
2. Dependent variable was transformed with natural logarithm.
3. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.952798, AGE = 51.26.
4. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.912629, AGE = 56.16.
5. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
11.061906, AGE = 49.76.
6. Covariates appearing in the model are evaluated at the following values: LNMADJFINCOME =
10.927554, AGE = 55.78.

236

5.3.3.3.

Gender and Marital Status Distinctions in Travel Behavior: Trips to

work
With respect to work commutes, married men walked to work in the lower density
suburban neighborhoods most frequently (2.94 times a week) and the longest distance (14.98
miles per week) when compared to married men and not-married women (table 64).
By driving, it was married men again that traveled most frequently (5.11 times a week)
and the greatest distance (82.02 miles) to reach their place of employment. Married women, in
contrast, drove least frequently (4.56 times a week) to work, while not-married women drove the
shortest distance (45.43 miles per week). Accordingly, both married and not-married men made
more frequent driving trips to work, and they also traveled longer distances than both married
and not-married women. This is a consistent finding evident in previous studies (Hanson and
Giuliano, 2004; Rosenbloom, 2005), In the lower density suburban neighborhoods, the gender
difference in average driving trip distance—about 15 miles on average—was also statistically
significant (p-value = 0.066) (tables 64 and 65).
5.3.3.4.

Gender and Marital Status Distinctions in Travel Behavior: Trips to

Shops, Restaurants, Personal Services, and Leisure
5.3.3.4.1.

Trips to shopping

As consistently evident in the other neighborhood groups, both in urban Detroit and in
the higher density suburbs, married women in the lower density suburban neighborhoods also
traveled most frequently to shop. For shopping trips, not-married men in the lower density
suburban neighborhoods walked most frequently (2.96 times a week), while married women
walked least frequently (1.34 times a week). In fact, both married men and married women

237

preferred not to walk for their shopping trips when compared to not-married men and notmarried women.
Not-married men also traveled as pedestrians the longest distances (6.61 miles per week)
among all gender and marital status groups, while married women walked the shortest distances
(2.12 miles per week), see table 64. However, none of the mean differences by gender or marital
status was statistically significant. The significant difference in shopping trips was evident with
driving. To reach their shopping destinations, married women in the lower density suburban
neighborhoods drove most frequently (3.93 times a week) while not-married women drove the
longest distances (16.29 miles per week). Not-married men, in contrast, traveled least frequently
(3.02 times a week), and the shortest distances (9.98 miles) to shop. Both of the women’s groups,
married and not-married, traveled significantly more times (p-value = 0.037) and longer
distances (p-value = 0.000) by driving to shop when compared to both married and not-married
men (table 66).
5.3.3.4.2.

Trips to restaurants

One of the destinations for which married women made the least visits was restaurants.
Married women traveled least frequently (less than once a week) as pedestrians. Both notmarried men and not-married women walked more frequently for restaurant visits when
compared to married men and married women. By driving, married men traveled most frequently
to access restaurants (2.44 times a week), while not-married women made the least frequent trips
(1.91 times a week). Therefore, both married and not-married men made more frequent trips to
restaurants by driving than both married and not-married women.

238

As for the driving trip distance, married women traveled the shortest distances to
restaurants (9.41 miles per week) when compared to the other gender and marital status groups
(table 64). Similar to the frequency of pedestrian trips to restaurants, not-married men traveled
the longest distances by driving (10.96 miles per week). Thus, both not-married men and notmarried women drove further distances to access restaurants when compared to married men and
married women, even though they made fewer driving trips. None of the mean differences in
restaurant trips, however, was statistically significant.
5.3.3.4.3.

Trips to personal services

Both men and women who were not married dominated personal service trips in the
lower density suburban neighborhoods (table 64 and 68). In fact, the not-married respondents
traveled more miles for service trips, regardless of travel mode. Married men walked least
frequently (1.13 times a week) to reach their personal service destinations while not-married men
traveled most frequently (2.42 times a week) (table 64). To access personal service destinations,
however, not-married men walked the shortest distances (1.71 miles per week) while married
women walked the longest distances (3.68 miles per week) (table 64). As for the gender
differences, both married and not-married men walked more frequently for personal services,
while they walked shorter distances compared to both married and not-married women.
However, the gender difference was only statistically significant for the pedestrian trip frequency
(p-value = 0.053) (table 68).
Between the marital status groups, both not-married men and not-married women
traveled more frequently but shorter miles as pedestrians to reach personal service destinations.
By car, not-married men traveled most frequently for personal services (3.31 times a week) while

239

married men traveled the least frequently (2.71 times a week). Again, not-married men and notmarried women drove more frequently than those married to access personal service
destinations. Respondents that are not married also traveled longer distances by car for personal
services than married respondents. Married women, in fact, drove the shortest distances (9.04
miles per week) for personal services, while not-married men traveled the longest distances
(15.66 miles per week). However, in the lower density suburban neighborhoods, there was no
significant mean difference by gender and marital status in personal service travel, except for
pedestrian trip frequency.
5.3.3.4.4.

Trips for leisure

For leisure trips, there was no observed significant mean difference between the gender
and the marital status groups in the lower density suburban neighborhoods. However, notmarried men and not-married women were participating in leisure activities more frequently than
married men and married women. For example, not-married men traveled most frequently (2.87
times a week) as pedestrians to reach leisure destinations, whereas married men traveled least
frequently (2.42 times a week). Despite their low frequency count for this travel activity, married
men walked the greatest distances (4.91 miles per week) for leisure.
Married women traveled more frequently as pedestrians (2.77 times a week) and by
driving (3.34 times a week) for leisure than not-married women. However, not-married women
traveled longer distances as pedestrians (4.16 miles a week) and by driving (21.44 miles a week)
for leisure when compared to married women. Thus, it is rather unique that for leisure trips,
married women in the lower density suburbs traveled most frequently by car, while married
women in the higher density suburbs drove least frequently to access leisure. In addition, married

240

women in the lower density suburbs drove more frequently than they walked to leisure
destinations, whereas married women’s groups in both urban Detroit and the higher density
suburban neighborhoods walked more frequently than they drove.
Another important insight into travel within the lower density suburbs was associated
with married men’s leisure trips. In the lower density suburban neighborhoods, married men
drove least frequently (2.70 times a week) to reach their leisure destinations. They also drove the
shortest distances (17.83 miles per week) for leisure, while not-married men made the most
frequent leisure trips by driving. Thus, in the lower density suburban neighborhoods, married
men and women prioritized commutes to work and shopping trips in their daily trip allocations.
This is due to the longer trip distances between destinations in the lower density suburbs.
Married men and women spent more time traveling longer distances to work and to shop in the
lower density built environments. In turn, both men and women in marriage considered leisure or
restaurant trips less important compared to commutes to work or shopping trips.
5.3.3.5.

Discussion

In the lower density suburban neighborhoods in Bloomfield Hills and West Bloomfield,
there were explicit gender differences in individual travel behavior. From the OLS regression
results, women, in comparison to men, traveled more frequently and greater miles for household
related activities, but less frequently and fewer miles for work. By driving, as a single dominant
travel mode in the lower density suburban neighborhoods, women were likely to travel
significantly more times and longer distances for non-work trips, such as shopping and leisure.
Specifically for shopping trips, the number of dependent children was a significant factor in

241

increasing shopping trip distance. Thus, women with dependent children were more likely to
drive a longer distance to shop.
This confirms the nature of gendered travel behavior. Women’s daily trips are mainly for
household duties, a pattern that has been recognized in the existing gender and transportation
literature (Hanson, 1995; Hanson and Hanson, 1980a, 1980b; Hanson and Pratt, 1991; Hanson
and Giuliano, 2004; Jones, 1990; Pickup, 1990; and Rosenbloom, 2005). As pedestrians, women
also traveled more frequently for shopping and leisure, and greater miles for personal services.
However, the results were rather biased since the portion of pedestrian trips to total trips was a
minor component in the lower density suburban neighborhoods.
Unlike in both the higher density Detroit urban and suburban neighborhoods, women in
the lower density suburbs traveled by driving significantly more times and longer distances than
men to reach their leisure destinations. It is a rather unique condition found only in the lower
density suburban neighborhoods, and likely tied to the greater personal automobile access among
women living in these extreme built environments. As a function of higher income levels and the
decentralized built environment, women in the dispersed suburban neighborhoods had greater
access to an automobile. In addition, they had to travel farther to reach dispersed destinations in
the lower density suburban neighborhoods, a complete contrast to the condition of women living
in the higher density suburban neighborhoods.
The role of women in fulfilling non-work trips was effectively established from the
ANCOVA results that explored disaggregated travel behavior. Among all gender and marital
status groups, married women drove least frequently to work, but most frequently for shopping,
after adjusting for income and age covariates. As commonly evident across both the higher

242

density Detroit urban and suburban neighborhoods, married women drove predominantly for
household errands and chores, while married men traveled predominantly for work. These
findings support the results of existing travel behavior literature (Hamilton and Jenkins, 1990;
Hanson and Hanson, 1980a; Hanson and Pratt, 1991; Hanson and Giuliano, 2004; Jones, 1990;
Pickup, 1990; and Rosenbloom, 2005).
Married women in the lower density suburban neighborhoods traveled shorter distances
to work by walking and by driving, when compared to married men. Married women, however,
traveled most frequently and the longest distances by car to reach shopping destinations. In
addition, for restaurant visits, married women traveled least frequently and the shortest distances
as pedestrians and by driving compared to other marital and gender groups in the lower density
suburban neighborhoods. For leisure activities, however, married women drove more frequently
and longer distances than married men, who traveled least frequently and the shortest distances
by car. Not-married men and not-married women, in fact, drove farther to access their leisure
destinations than married men and married women in the lower density suburban neighborhoods.
It is also important to note that compared to other women’s groups in both urban Detroit
and the higher density suburban neighborhoods, married women in the lower density suburbs
walked less frequently for daily activities. This is to be expected given the dispersed urban built
environments of these communities; the low density, single use and disconnected neighborhoods
that literature has shown increase distances between destinations. Within these urban built
environments, there are few—if any—potential destinations that are accessibly by walking or
biking from the isolated and lower density residential neighborhoods. Moreover, longer distances
to travel deter married women from walking or biking to meet daily needs due to increased travel
cost, given the limited time as a result of their multiple roles at home and in the workplace.
243

Finally, married women in the lower density suburban neighborhoods prefer driving a
personal vehicle due to the greater availability of automobiles in these higher income
households. Married women in the lower density suburban neighborhoods also prefer to drive as
a result of its greater convenience in comparison to walking. Nevertheless, even in these extreme
built environments, women—regardless of marital status—still drove shorter distances to work,
while they drove longer distances to shop, in comparison to men.

244

5.4. Conclusion and discussion
While the existing literature addresses general gender differences in trip generation
across urban space, more nuanced analyses of socio-economic and demographic dynamics on
individual trip activities is necessary in order to understand the more specific effects of
population characteristics (e.g. gender and marital status). The analyses presented in this chapter
also explore an extensive array of disaggregated travel patterns, including trips to work,
restaurant visits, shopping, personal services, and leisure destinations. In particular, women’s
travels are often derived from their multiple responsibilities between paid employment and
household duties (Hanson and Hanson, 1980a, 1980b 1981; Hanson and Pratt, 1991; Jones,
1990; and Pickup, 1985, 1990). Accordingly, a disaggregate analysis of individual travel
behavior needs to include gender and marital status to examine women’s daily trips for various
travel activities, both work and non-work (Hanson, 1995; Hanson and Hanson, 1980a; Hanson
and Pratt, 1991; Hanson and Giuliano, 2004; Hamilton and Jenkins, 1990; Jones, 1990; Pickup,
1985, 1990; and Rosenbloom, 2005).
In addition to the analyses in previous studies focusing on built environment
characteristics in shaping individuals’ travel behavior, it is also important to identify differing
urban environment characteristics on individuals’ daily trips—and particularly within the context
of this research—of the socio-economically disadvantaged living in neighborhoods experiencing
disinvestment and decline. In the lower east side urban Detroit neighborhoods, racial and ethnic
minorities, and single-parent households, face greater travel burdens due to severe disinvestment
and decline within these neighborhoods, which have shifted many of the urban amenities—the
necessary daily destinations—from the immediate neighborhoods to suburban locations.

245

Due to the region’s deep history of racial segregation, Detroit’s black population is
isolated largely in the city and spatially separated from the white population in the suburbs
(Darden et al., 1987). The decentralization of Detroit has resulted from racial discrimination in
the housing market that was controlled by apartment managers, real estate brokers, and builders
(Darden et al., 1987, 2009; Sugrue, 1996; Thomas, 1997; and Vojnovic and Darden, 2013a).
Accordingly, the decentralization of its population and tax base led to a polarized fiscal capacity
between the city and the wealthy suburbs, and severely impacted opportunities for social
mobility, and particularly among the city’s minority population. As a result, the city of Detroit
has been described as the most distressed urban core with an increasing concentration of
unemployment, poverty, and racial and ethnic minorities due to outmigration of middle- and
upper-income Whites to the suburbs.
Disinvestment and decline places severe burdens on lower income populations, as they
increasingly have to travel to distant suburban locations to access basic daily necessities,
including fresh fruits and vegetables, healthy restaurant options, personal services, and leisure
destinations. In addition, with limited access to a private automobile, and greater reliance on
public transit and walking to reach daily destinations, this places an added layer of burdens on
lower income urban Detroit residents. Thus, in spite of similar built environments to the higher
density suburban neighborhoods—that is, despite having an urban form characterized by higher
density, mixed land use, with connected road networks—travel patterns among urban Detroit
respondents were vastly different from those of the wealthier higher density suburban
respondents, who had access to a rich variety of urban opportunities as potential destinations.
In order to identify how women and men travel differently in the Detroit region,
disaggregate individual travel activity analyses were used. Disaggregated analyses were
246

performed with three typologies of differing urban neighborhoods—higher density urban Detroit
neighborhoods, higher density suburban neighborhoods, and lower density suburban
neighborhoods. Disaggregated travel activity analyses—including OLS regression and Analysis
of Covariance—examined gender differences, and also the gender division of intra-household
responsibilities, in individuals’ trip making for various daily activities. The OLS regression
results suggested a relationship between gender and individual travel behavior (such as trips to
work, shopping, restaurant visits, service and leisure) after controlling for socio-economic and
demographic factors (income, age, personal vehicle access and the number of dependent
children). In turn, the ANCOVA results demonstrated the mean differences in disaggregate travel
activities in light of gender and marital status, after adjusting for income and age.
Regardless of the neighborhood typology, men traveled more frequently and over longer
distances to reach their place of employment. Women’s shorter driving trip distance to work was
consistent across all three neighborhood types, urban Detroit (at a significant level), the higher
density suburban neighborhoods (not at a significant level) and the lower density suburban
neighborhoods (at a significant level). Women, however, traveled more frequently and longer
distances for nonwork trips and particularly for trips related to household responsibilities, such as
shopping and personal services. In the multi-person household, married women tend to shop for
daily goods in the family. Subsequently married women traveled to shop more frequently than
married men and traveled longer distances, and this was consistent across all the neighborhood
typologies. However, married men consistently traveled more frequently and longer distances to
restaurant when compared to married women, and in all three neighborhood typologies.
Therefore, it is expected that access to a private automobile will critically impact a
married woman’s ability to complete her multiple responsibilities, as both homemaker and paid
247

laborer, by maximizing trip efficiency in terms of travel speed and flexibility, and also
maximizing the spatial extent of ‘reachable’ destinations. The limited access to a private
automobile will decrease a woman’s mobility and particularly if destinations are dispersed and/or
alternative modes of travels are scarce or poorly serve particular areas. This was evident among
married women and their extensive public transit trip distances to shopping in the urban Detroit
neighborhoods, where distances between destinations were increased by the disinvestment and
disappearance of necessary urban amenities within the city.
In addition, decentralizing employment resulted in extreme trip distances to work for
residents living in the Detroit urban neighborhoods. For work trips, married men in the urban
Detroit neighborhoods drove more than 80 miles per week and married women drove more than
30 miles per week to their work destinations, distances that were similar or more to that of the
lower density Detroit suburban residents. These longer commute distance patterns for urban
Detroit residents have been extensively discussed in the spatial mismatch and racial justice
literature (Kain, 1968; Vojnovic and Darden, 2013b).
While significant, the results have limitations with respect to their refinement and initial
sample groups. First, limitations were related to individual characteristics of the sample
population. The effect of increasing age in individuals’ pedestrian trip making should be
carefully treated since the observations included all adults with an age of 18 years and older. The
physical ability to walk or bike as a mode of travel, therefore, changed nonlinearly with
increasing age. For the same reason, the relationship between individuals’ trip-to-work and
socio-economic and demographic factors, such as age and income, need to be interpreted
carefully due to the fact that the sample included retirees.

248

Another limitation to the research was the ‘race’ variable, which was not included as a
separate independent variable in the models. The ideal model should incorporate a racial
breakdown between African Americans and the Whites in the neighborhoods so that white
women and men in the urban Detroit would be compared with white women and men in the
suburban neighborhoods. However, due to the scarcity of white population in the urban Detroit
neighborhoods, models were not able to capture a meaningful gender comparison between racial
groups. Thus, income and racial characteristics of each neighborhood group (urban Detroit as
predominantly lower income and African Americans, and two suburban neighborhood groups as
predominantly higher income and white) permitted to identify gender and racial differences in
individual travel behavior.
Marital status, either married or not married, might mislead the interpretation of the
results, and specifically the not-married classification, since it included both people who were
never married and people who were not currently in a marriage, including the divorced,
separated, and widowed. Therefore, in this study, there was no differentiation within the models
between never-married single-parent households and those who are not currently in a marriage.
Lastly, the differences between married men and married women did not represent exactly the
intra-household variation. Since the data did not have exact ‘paired-sets’ of husbands and wives
for each household, the inter-dependence between husband and wife, and subsequent travel
constraints, were not fully captured.
In addition, in spite of a large sample size in the region (n=1148), disaggregation of
individual travel behavior into various travel activities substantially limited the number of
observations (Ns) in each OLS regression analysis and ANCOVA—for example, public transit
trips for restaurant visits, N Detroit urban = 17 (frequency) and 16 (distance), and N high density suburban =
249

4 (frequency) and 3 (distance). Thus, results for certain daily activities by certain travel modes
were potentially skewed.
For example, the standard errors varied with relatively large extents due to limited sample
sizes. It is because the means become more sensitive to influences of individual cases in the
distribution of observed values after subgrouping the sample by gender and marital status. While
I successfully obtained interesting results for each subgroup for the travel behavior, both trip
frequency and trip distance for each travel activity, the limitations with small Ns and subsequent
large variations of S.E. are unavoidable. Also, the reality of high unemployment rates in the
urban Detroit neighborhoods limited the ability to analyze work trip behavior with meaningful
differences between groups.
Also, since the data did not capture trip chaining in daily travel activities, trip variables—
both trip frequency and distance—were assumed as single trips between home addresses and
various destinations. With potential trip chaining missing, women’s more frequent and
segmented trips between their multiple responsibilities between paid employment and household
duties could not be captured. Lastly, with respect to driving trips, the difference between a
person traveling as the driver and a passenger was not considered. As a result, a person’s driving
trips were included in the analyses even though the person might not have access to a personal
automobile.
Thus, further research using similar methods can systematically test and calibrate these
results against a more comprehensive set of data, including trip chaining components, intrahousehold variation between men and women in union, and types of vehicle travels. Also, it will
be useful to analyze the data with alternative methods, for example logistic regression, that may

250

reveal the meaningful differences between populations and allow tests with more rigorous
grouping schemes. Lastly, a longitudinal analysis will help to fully understand the changes in
urban built environments and the effects on individual travel behavior.

251

CHAPTER 6: CONCLUSION
6.1. Summary and Conclusion
Urban transportation researchers have established strong theoretical support and have
demonstrated empirically that encouraging regular participation in moderate physical activity can
encourage healthier life-styles (Frank, 2000, 2004; Frank et al., 2006; Handy, 1996, 2005a;
Hanson et al. 2002; Saelens et al., 2003a; Sallis et al., 2004, Vojnovic et al., 2006). The studies
have consistently shown that a compact community—a neighborhoods that is characterized by
higher density, mixed land uses, and high connectivity—is more likely to encourage the use of
more sustainable modes of transport, such as walking, cycling and public transit, when compared
to a low-density suburban neighborhood promoting a singular vehicle travel mode, private
automobile use.
Transportation researchers have also increasingly recognized the differences in travel
behavior based on gender and different socio-economic and ethnic/racial composition (JohnstonAnumonwo, 1992, 1997; McLafferty and Preston, 1997; and Wyly, 1996, 1998). In addition to
automobile dependent urban transportation systems, the unequal distribution of urban amenities
across U.S. metropolitan areas has resulted in differential travel costs for different income, race
and ethnicity, age and gender population subgroups that face more limited sets of choices for
their travel destinations (Handy et al., 2005; Hanson and Hanson, 1981; Hanson and Johnston,
1985, Hanson, 1986, Johnston-Anumonwo, 1989, 1997; MacDonald, 1999; McLafferty and
Preston, 1997; Rosenbloom, 1989, 2005; Rosenbloom and Raux, 1985; Vojnovic et al., 2013;
LeDoux and Vojnovic, 2013; and Wyly, 1996, 1998, 1999).

252

While studies have found links between residents living in compact, mixed land-use, and
connected neighborhoods and increased walking frequency and shorter walking distances, urban
transportation research has mainly focused on major service based cities with robust urban cores;
cities such as Chicago, Seattle, Boston and San Francisco. As a result, very little research has
focused on travel patterns in traditional industrial cities in the Midwest and the Rustbelt, cities
like Detroit. The urban minorities concentrated in such cities not only find few employment
opportunities in the urban core, but also confront great distances to reach potential employment
in the suburbs. This geographical separation between the central city and the suburbs in the
Detroit region has also been facilitated through class and racial segregation, and a politically
fragmented regional jurisdictional structure, with little inter-municipal cooperation.
This dissertation fills the gap in the literature on gender differences in travel behavior
while taking into account socio-economic and racial variables. Longer distances between
spreading destinations—such as jobs, stores and public services in most U.S. urban areas—
aggravate travel burdens of the disadvantaged, including women, minorities and lower-income
populations. Specifically women in the household with one or no car, in many cases likely due to
their lower incomes, have struggled with their multiple household responsibilities as a result of
their constrained mobility (Hanson, 1986; Hanson, 2010; and Turner and Neimeier, 1997).
Despite the significant implications of these urban and transportation stresses, women’s travel
and their access to urban opportunities have been marginalized in urban transportation research.
In addition, there is little that is understood in how these variations in travel behavior by
gender might be affected by income, ethnicity, age, household types, and neighborhood
structure. Moreover, nonwork trips have not been examined in detail, even though non-work
trips tend to be more dynamic and associated with people’s social roles and responsibilities.
253

The limitations in the existing research are also found in the scale of study. Most travel
behavior studies have attempted to understand travel demand and patterns with aggregate data at
conventional TAZs (traffic analysis zones), counties, or MSAs. The focus of studies on these
large geographic areas averages out differences in travel that can be recognized at the microscale, the neighborhood or block level. Within this context, it is important to understand the full
range of transportation demands, and specifically taking into account the realities confronted by
lower income populations, urban minorities and women within the neighborhood, and in robust
neighborhoods as well as neighborhoods experiencing disinvestment and decline.
One objective of this dissertation was to identify the effects of the built environment on
promoting non-motorized travel. A second objective was focused on women’s accessibility and
mobility in urban space within the context of differing urban built environment characteristics in
the Detroit region. First, in chapter 3, the relationship between urban density, land use mix, street
connectivity, bus stops and individual travel behavior was explored. This particular analysis
aimed to understand the causal relationships of the urban built environment in promoting
pedestrian trips and reducing automobile use. Accordingly, two main research questions were
involved: (1) do people utilize non-motorized travel modes such as walking and biking more in
urban neighborhoods characterized by higher-densities, mixed land uses, and high connectivity?
and (2) people reduce automobile travel in urban neighborhoods characterized by higherdensities, mixed land uses, and high connectivity? In this component of the research, the
hypothesis was that urban neighborhoods with higher density, mixed land uses, and high
connectivity would increase the likelihood of utilizing non-motorized travel modes.
Spatial data of urban built environment elements, including land use density and types, 4way intersection density, and the number of bus stops, were collected for six 3 by 3 mile
254

neighborhoods, about 9 sq. miles each, in the Detroit region. The additional process of boundary
data collection (an area covering half a mile distance from each neighborhood boundary) enabled
the study to capture the built environment characteristics with a walking distance of half a mile
from each respondent’s address.
Neighborhood comparative analysis results showed that there were clear distinctions
between the three neighborhood groups. Four higher density urban and suburban neighborhoods
were characterized with a greater land use density and mix, and more 4-way intersections and
bus stops than the lower density suburban neighborhoods. The two higher density urban
neighborhoods, however, differed from the two higher density suburban neighborhoods in that
there were a greater variety of both residential and nonresidential land uses in the higher density
suburban neighborhoods, due to the scale of decline and disinvestment in urban Detroit.
For the multivariate regression analysis, 1,106 road network buffers (RNBs) were created
from residential addresses of survey respondents located using the ESRI ArcGIS Network
Analyst. RNB is a service area of a half a mile distance, running in different in-network
directions of a half a mile from a residential location using the actual road network. The
dependent variables were three trip measures, including mode split, trip frequency and trip
distance (length of trip), for work and nonwork trips. As independent variables, the densities of
residential and non-residential structures, land use mix measures (entropy measures ranging from
3 to 8 land use types), the density of 4-way intersections and bus stop densities within the RNB
were determined. In addition to the urban form indicators, personal income and age were added
to the modes as controlling factors.

255

The analysis results showed the significant associations between built environment
elements and individual travel behavior. First, pedestrian travel was likely to increase in RNBs
with a greater density, diverse land uses, highly connected road networks, and more bus stops. In
addition, motorized travel for daily activities was substantially reduced in the densely developed
RNBs with a greater land use mix, more intersections and bus stops. People also tend to utilize
non-motorized travel modes more and drive less if the immediate built environment has a better
balance of land uses, and particularly a mix of residential, commercial (retail and personal
services) and public institution. Finally, I used the stepwise regression analysis to test the relative
strength of each of the urban form factors in influencing individual travel behavior. Among the
built environment factors, the number of bus stops and 4-way intersections were the most
strongly associated with individual travel behavior.
This particular analysis is important because the results not only confirm the previously
addressed relationships between urban form and individual travel behavior, but also offer the
broadened understanding of urban travel with more elaborate dimensions in neighborhood
characteristics. Unlike the existing studies, the results explored the relationship between built
environment and travel behavior in a region characterized by a rapidly declining urban core, the
Detroit region, which is experiencing extreme urban decentralization coupled with disinvestment
in the central city. The results, therefore, were able to reveal how individual’s travel was actually
affected in neighborhoods experiencing extreme disinvestment and decline.
Another important aspect of the analysis comes from the reaffirmed significance of the
land use mix in promoting pedestrian travel and reducing automobile use. The strongest
influence of commercial land uses was in increasing pedestrian trips and in decreasing
automobile trips, stressing the importance of a balance between residential, commercial, and
256

public institutional land uses (i.e. LUM3). This analysis also supports the importance of utilizing
a disaggregate analysis at a fine urban scale, the neighborhood level or lower. These
relationships between urban form factors and individual pedestrian travel were more accurately
understood from the finer scale analysis of road network buffers (RNBs), after controlling for
socio-economic and demographic characteristics. Utilizing RNBs is in itself a seldom utilized
approach to travel behavior analysis.
In chapter 4, three neighborhood typologies were defined in the exploration into trip
generation, residents’ socio-economic profile and the quality of neighborhood environments.
This chapter was to identify the neighborhood typology that helps explain differences in travel
behavior between urban and suburban neighborhoods. Subsequently, two main research
questions were involved: (1) which socio-economic and demographic factors significantly
influence total trip generation? and (2) which neighborhood group, among urban Detroit, higherdensity suburban, and lower-density suburban neighborhoods, is differentiated against others in
terms of socio-economic and demographic characteristics, travel patterns, and neighborhood
quality for pedestrians? The hypothesis was that residents in the higher income neighborhoods,
with greater automobile access, tend to travel more frequently and longer distances by motorized
travel modes. Also, residents in the neighborhoods with more diverse urban opportunities and
higher quality pedestrian environments (e.g. neighborhoods with a reduced fear of walking)
would have improved accessibility for daily activities when compared to neighborhoods with
similar urban density and connectivity.
First, the entire sample of travel survey data (n=1,148) was used to model trip generation
among the general population in the Detroit region. Findings from the OLS regression indicated
that income and personal vehicle access had a positive effect on trip generation (total trip
257

frequencies and distance) for all activities, work and non-work travel, while age and the number
of dependent children had a negative effect. Thus, the significant effect of personal vehicle
access to a person’s travel patterns was confirmed.
Second, discriminant analyses were applied to confirm differing characteristics between
neighborhood groups in the Detroit region before analyzing disaggregate individual travel
behavior. By incorporating socio-economic and demographic factors (income, age, personal
vehicle access and the number of dependent children), the discriminant analysis results
differentiated urban Detroit neighborhoods by their lower socio-economic characteristics (i.e.
lower income and personal vehicle access) and higher density suburban neighborhoods by their
younger life cycle stage characteristics (i.e. age and the fewer dependent children).
By incorporating travel pattern variables (total trip frequencies and cumulative trip
distances by travel mode), the analysis differentiated the lower density suburban neighborhoods
against the other higher density neighborhoods (both urban and suburban) by the driving
dominant travel patterns in the lower density neighborhoods. Urban Detroit neighborhoods were
also separated against the suburban neighborhoods (both higher and lower density) by their
pedestrian dominant travel patterns.
Lastly, the descriptive analysis results illustrated substantial differences in mean trip
frequency and trip distance between the three neighborhood typologies. Residents in the lower
density suburban neighborhoods traveled longer distances by all modes when compared to the
higher density neighborhoods, which is consistent with previous studies (Crane, 1996; Crane and
Crepeau, 1998). Nevertheless, the pattern of longer trip distances by the urban Detroit
neighborhood residents when compared to their higher density suburban counterparts

258

reconfirmed the importance of this neighborhood typology in understanding the unique travel
burdens of lower-income urban minorities.
Defining the neighborhood typology is particularly important in order to understand the
travel burden of individuals in neighborhoods experiencing disinvestment and decline. In
addition to the unequal distribution of travel opportunities, such as reaching employment, retail,
or personal services, income disparities across the neighborhoods result in a greater burden
placed on lower-income urban minorities with limited personal automobile access.
The neighborhood typologies also enabled the close examination of travel burdens faced
by urban minorities in the Detroit region. The fact that urban Detroit residents traveled longer
distances as pedestrians and by public transit clearly revealed their burdens in accessing urban
amenities when compared to residents in the higher density suburbs. In addition, urban Detroit
residents, for certain activities, also drove longer distances when compared to the lower density
suburban residents. Therefore, defining these neighborhood typologies provided a better
understanding of travel constraints of the socially marginalized populations, and particularly
lower-income women in the declining urban neighborhoods, one of the principal areas of this
dissertation’s research interest.
In the last analysis chapter (chapter 5), the differences in travel behavior between women
and men were examined within the full array of daily travel activities. This particular analysis
sought to examine gendered travel patterns and their relationship with urban built environment
characteristics and socio-economic and demographic factors. Two main research questions were:
(1) how women travel differently from men within similar urban built environments? and (2)
how lower-income urban minority women’s travels were more disadvantaged than other

259

women’s groups living in the suburban Detroit neighborhoods? The central hypothesis was that
women would travel more for household responsibilities regardless of neighborhood built
environment characteristics. Also, it was assumed that lower-income urban minority women
would walk more and use public transit more than women in the suburban neighborhoods, due to
lower incomes and lower automobile access.
In this chapter, the entire sample of travel surveys was divided into three neighborhood
groups, the urban Detroit neighborhoods, the higher density suburban neighborhoods and the
lower density suburban neighborhoods, in order to control the effects of the built environment
and the unique conditions of urban Detroit. Daily trips were also disaggregated into trips to
work, shopping, restaurants, personal services and leisure. In this way, the complexity of
individual travel behavior would be better understood in the context of gender effects. In the first
part of the analysis, I used OLS regression to address how women’s travel behavior would be
different from that of men in utilizing various travel modes—including pedestrian, public transit,
and driving—and after controlling for socio-economic and demographic factors (including
income and age). The socio-economic and demographic factors were selected from the general
trip generation models (chapter 4). The ‘race’ variable was excluded in this analysis since the
main focus of the research centered on gender effects of individual travel behavior.
The results showed that men traveled more frequently and over longer distances to reach
their place of employment regardless of the neighborhood typology. The gender difference in
commutes to work was also confirmed in terms of women’s shorter driving trip distance in all
three neighborhood typologies. Women, however, traveled more frequently and longer distances
for trips related to household responsibilities, such as shopping and personal services. Thus, the

260

results confirmed that the traditional gender role was still reflected in women’s daily travel in the
Detroit region.
I also used the ANCOVA (Analysis of Covariance) to detect the mean differences
between women and men while taking into account their marital status, and after controlling for
income and age. The findings revealed that married women traveled to shop more frequently
than married men and traveled longer distances across all three neighborhood typologies.
However, married men consistently traveled more frequently and longer distances to restaurants
when compared to married women, and in all three neighborhood typologies.
The results, therefore, illustrate the importance of access to a private automobile,
particularly in the context of a married woman’s ability to complete her multiple responsibilities
as a homemaker and as a workforce participant. Private automobile access critically impacts
woman’s ability to enhance trip efficiency in terms of travel speed and flexibility, and also
maximizes the spatial extent of ‘reachable’ destinations. Thus, more limited access to a private
automobile decreased a woman’s mobility in an urban area where destinations are dispersed
and/or alternative modes of travel are scarce or poorly serve particular areas.
Another important aspect of the analysis centered on the disappearing urban amenities
within declining urban neighborhoods and the associated outcomes in the accessibility of daily
destinations among urban minority populations. For instance, married women in the urban
Detroit neighborhoods traveled extensive distances by public transit for shopping due to the
increased distances between destinations, resulting from disinvestment and the disappearance of
necessary urban amenities within the city. The extreme commute distances confronted by urban

261

Detroit residents also supports the argument of decentralizing employment, spatial mismatch and
racial injustice.
6.2. Theoretical contributions and policy implications
There are several theoretical contributions advanced by this dissertation. The first comes
from the benefits of micro-scale urban travel behavior analysis where we can focus on unique
built environment characteristics that are normally just averaged out at the traditional county–
scale levels of analysis (Vojnovic, 2006; Vojnovic et al., 2008, 2013, forthcoming; Griffith,
Vojnovic, and Messina, 2012; and LeDoux and Vojnovic, 2013). While the analyses at the
county scale were able to provide significant evidence that greater urban density and diverse land
uses increased automobile travel, the traditional method was unable to capture variations of such
factors within the zone (e.g. within the county). In other words, differences in urban density or
land use mix between smaller geographical scales (e.g. a neighborhood or even smaller scale
such as a walkable area) were discarded.
For example, in this study, in the Birmingham neighborhood (a higher density suburban
neighborhood) and the two other lower density suburban neighborhoods are all located in
Oakland County, however their urban forms vary vastly as do their individual travel behavioral
characteristics. Hence, county scale analyses are unable to assess accurately different effects of
urban built environment characteristics on subpopulations’ travel behavior between
neighborhoods that are located in the same city or county. This is also particularly important to
understand how an urban neighborhood and its provision of amenities affect individual daily
travel. Despite the similar higher density urban forms, the two urban Detroit neighborhoods were
characterized with less diverse land uses, specifically, more limited nonresidential land uses

262

when compared to the two higher density suburban neighborhoods. As a result, urban Detroit
respondents traveled longer distances to meet the basic daily needs, such as shopping for food. In
addition, it is possible to evaluate the effects of urban form factors on diverse travel activities
such as trips for shopping, restaurants and personal services rather than just focusing on
commutes to work.
In addition, road network buffers (RNBs) on the existing road system permit the analysis
into the effects of the immediate built environment (within half a mile of one’s home), enabling
the exploration into the relative strength of certain urban form factors on promoting pedestrian
travel and reducing automobile travel. Measuring the relevance of each urban form factor—that
is the RNBs—becomes particularly significant since pedestrian travel, in comparison to
automobile travel, is affected to a greater extent by the immediate urban form within a walking
distance of one’s residence.
For example, the analysis results from this dissertation show the importance of the
availability of retail and personal services within walking distance of one’s home in promoting
non-motorized travel. The results also illustrate that the diversity in residential land uses and
industrial land use within walking distance of one’s residence had no significant effects on
promoting pedestrian travel. In sum, the results at a micro-scale analysis offer much greater
sensitivity to model the effects of urban built environment characteristics on individual travel
behavior.
This study also reconfirms gender differences in travel behavior, and particularly in the
higher density neighborhoods. On a weekly basis, men travel longer distances and women make
more trips for the full array of daily activities in both the higher density suburbs and higher

263

density urban neighborhoods. For example, by driving, women in both higher density urban and
suburban neighborhoods traveled more frequently for shopping and personal services. These are
findings that are consistent with the existing travel behavior literature (Hanson and Hanson,
1981; Hanson, 1986; Kwan, 1999b; and Wyly, 1998). The results also show that men in both
higher density urban and suburban neighborhoods traveled longer distances in their commute to
work. These findings, once again, are consistent with previous studies on gender and urban
transportation (Hanson and Johnston, 1985; Johnston-Anumonwo, 1992; and Wyly, 1998).
The analysis results, however, contributed to understanding another important aspect, that
is, women in the urban Detroit neighborhoods utilized pedestrian modes and public transit
significantly more to reach their daily destinations when compared to women living in the higher
density suburban neighborhoods. Not only traveling more frequently by pedestrian and public
transit modes, Detroit urban minority women also had to travel longer distances by all modes to
meet their daily needs. Accordingly, the travel patterns confirm the greater travel burdens
confronted by urban minority women in the neighborhoods experiencing extreme disinvestment
and decline.
This study, however, shows that gender differences in travel behavior in extremely low
density neighborhoods are not maintained. In the lower density suburbs, women make more
frequent trips and also travel longer distances than men. On a weekly basis, women make more
trips and travel longer distances when exploring the full array of daily activities in the lower
density suburbs (Hanson and Hanson, 1981; Hanson, 1986; Kwan, 1999b; and Wyly, 1998). The
extreme characteristics of this urban form—the low densities, the single zoned uses and the
disconnected road network—not only make everyone more automobile dependent, but they also
encourage extensive driving distances by car to reach even the basic daily necessary destinations.
264

This not only reveals the inefficiencies of these urban/suburban spaces, but speaks to the demand
for ensuring highly isolated enclaves by the wealthy as an urban design strategy to keep out the
“unwanted” (Vojnovic, 2006; Vojnovic, 2009; Vojnovic et al., 2013; and Vojnovic and Darden,
2013b).
In terms of its relevance to policymakers, this study contributes insights to urban
development and design strategies that improve walking environments and access, and in the
process, provide a greater understanding of how healthy communities can be promoted, and
especially for women, racial/ethnic minorities, and low income groups. Even though the urban
Detroit neighborhoods and the higher density suburban neighborhoods were similarly
characterized by higher densities, diverse land uses, and connected street networks, urban Detroit
respondents traveled longer distances by driving to reach daily destinations when compared to
the higher density suburban residents.
Furthermore, the respondents in the urban Detroit neighborhoods were hindered from
walking or bicycling due to greater perception of fear in the neighborhoods. However, the
implications of higher densities, mixed land-uses, and connected street networks on travel
behavior and physical activity in such deteriorated inner-city neighborhoods have not been fully
tested. The research thus also shows that while urban design is important in promoting nonmotorized travel, wider socio-economic conditions—urban disinvestment, decline, and
poverty—are even more important that design is shaping travel behavior.
This study also contributes to insights on transportation planning strategies that offer
alternative transportation options to meet our populations’ diverse travel demands. Urban Detroit
neighborhoods used public transit significantly more frequently than suburban neighborhood

265

residents, identifying the importance of the provision of appropriate public transit services in the
region. Specifically when considering the location of employment and major retail outlets in the
region, understanding the travel demand of inner-city urban residents, and the need for providing
appropriate public services, becomes extremely critical to improve not only conditions of equity,
but also basic mobility.
Lastly, this study contributes to insights on community development and planning that, in
the wider context, could potentially foster more equitable opportunities through socio-spatial
integration and economic development. The analysis results illustrate the importance of having
access to diverse nonresidential locations—such as retail and personal services—in the
immediate neighborhoods in order to encourage pedestrian travel as a part of daily physical
activity. However, longer driving distances by urban Detroit neighborhood residents also reveals
the unequal access that urban minorities face in reaching basic opportunities such as jobs,
shopping, personal services and leisure destinations. Therefore, offering a rich bundle of urban
amenities to lower-income urban minorities, as well as diverse housing options, is crucial in
order to ensure the equitable access to urban opportunities.
6.3. Future research
As for my next steps in my research endeavors, in order to reveal more precisely the
effects of the urban built environment on individual travel behavior, work trips and nonwork
trips can be explored in separate models with further separation of nonwork trips into specific
trip purposes (travel for shopping, personal services, leisure, and restaurants). Furthermore, in
addition to income and age, other individual characteristics, such as gender and car ownership, or

266

household characteristics such as marital status, can be adopted as controlling factors in the
model and might help to further explain individual travel behavior choices and travel demands.
Additional information on travel destinations can also be added in the model. For
instance, a dummy variable distinguishing between a central city or a suburban location will
allow the model to examine not only the travel flows within the metropolitan region, but also
exhibit the travel burdens of racial or ethnic minorities in declining urban cores. This will
improve the understanding of urban travel, specifically reverse commutes and longer shopping
trip distances to suburban retail locations.
The effects of pedestrian-friendly built environment factors on BMI (body mass index)
will also be explored. The models should be developed separately at a finer scale (e.g. road
network buffer for the entire sample population), and also at the neighborhood level (e.g. three
neighborhood typologies). This is particularly important since there have been unsynchronized
relationships between a higher level of BMI and densely developed urban neighborhoods with
connected road networks. Therefore, the models will examine thoroughly the effects of
neighborhood quality, such as the diversity of commercial destinations and leisure and
recreational sites, on BMI in consideration of promoting physical activity and maintaining
healthy lifestyles.
In addition, further research on gender dynamics and the complexity in individual travel
should systematically test and calibrate the previous results against a more comprehensive set of
data, including trip chaining components, intra-household variation between men and women in
union and types of vehicle travel. Since, in this study, there was no differentiation between
never-married single-parent households and those who are not currently in a marriage within the

267

models, the results may mislead the interpretation of the effects of marital status (as either
married or not married) on individual travel. Furthermore, the differences between married men
and married women did not represent exactly the intra-household variation since the data did not
specify exact ‘paired-sets’ of husbands and wives for each household. Therefore, the interdependence between husband and wife, and subsequent travel constraints, were not fully
captured.
In addition, women’s more frequent and segmented trips in their multiple responsibilities
between paid employment and household duties could not be explored since the data did not
capture trip chaining in daily travel activities. All trips were assumed as single trips between
home addresses and the various destinations. Lastly, the differentiation between a person
traveling as the driver and a passenger was not considered. As a result, driving trips of
passengers were included equally across the analyses as driving trips, even though the person
might not have access to a personal automobile.
Finally, alternative multivariate statistical procedures will be useful to analyze the data.
For example, logistic regressions may reveal the meaningful differences between populations
and allow tests with more rigorous grouping schemes. A longitudinal analysis will also enable a
better understand of changes in the urban built environment and the effects on individual travel
behavior through time.

268

APPENDIX

269

APPENDIX:
Data histograms: built environment and individual travel behavior

Figure 41 Built environment factor: residential land use density

270

Figure 42 Built environment factor: non-residential land use density

Figure 43 Built environment factor: retail land use density

271

Figure 44 Built environment factor: personal services land use density

Figure 45 Built environment factor: public institutions services land use density

272

Figure 46 Built environment factor: industrial land use density

Figure 47 Built environment factor: land use mix index with 3 categories (LUM3)
273

Figure 48 Built environment factor: land use mix index with 7 categories (LUM7)

Figure 49 Built environment factor: land use mix index with 8 categories (LUM8)

274

Figure 50 Built environment factor: 4-way intersection density

Figure 51 Built environment factor: bus stop density

275

Figure 52 Individual travel behavior variables: total pedestrian trip frequency for all activities

Figure 53 Individual travel behavior variables: total pedestrian trip distance for all activities

276

Figure 54 Individual travel behavior variables: total driving trip frequency for all activities

277

Figure 55 Individual travel behavior variables: total driving trip distance for all activities

278

Figure 56 Socio-economic and demographic factors: age

Figure 57 Socio-economic and demographic factors: income

279

REFERENCES

280

REFERENCES
Administration on Aging. 1998. Mobility and independence: Changes and challenges for older
drivers. Washington, DC: Author.
Alsnih, R. and D.A. Hensher. 2003. “The mobility and accessibility expectations of seniors in an
aging population.” Transportation Research A 37 (10):903–916.
Barla, P., L. F. Miranda-Moreno, N. Savard-Duquet, M. Theriault, and M. Lee-Gosselin. 2010.
“Diaggregated Empirical Analysis of Determinant of Urban Travel Greenhouse Gas Emissions.”
Transportation Research Record (2156):160-169.
Biles, J.J., and B.W. Pigozzi, 2000. “The Interaction of Economic Reforms, Socio-economic
Structure and Agriculture in Mexico.” Growth and Change31 (1):3-22.
Blumenberg, E. 2004. “En-gendering effective planning: spatial mismatch, low income women,
and transportation policy.” Journal of the American Planning Association70 (3):269–281.
Boarnet, M.and R.Crane. 2001a. Travel by Design: The influence of urban form on travel New
York, NY: Oxford University Press.
Boarnet, M., and R. Crane. 2001b. “The influence of land use on travel behavior: specification
and estimation strategies.” Transportation Research Part a-Policy and Practice 35 (9):823-845.
Boarnet, M. G., and S. Sarmiento. 1998. “Can land-use policy really affect travel behaviour? A
study of the link between non-work travel and land-use characteristics.” Urban Studies 35
(7):1155-1169.
Burns, R., and R. Burns. 2008. Business Research Methods and Statistics using SPSSS. Sage
publication LTD., London: UK.
Cao, X., P.L. Makhtarian and S. L. Handy. 2007. “Do changes in neighborhood characteristics
lead to changes in travel behavior? A structural equations modeling approach.” Transportation
34:535-556.
Cao, X., P.L. Makhtarian and S. L. Handy. 2008. “Neighborhood design and aging: an empirical
analysis in Northern California.” In Transportation Research Board 87th Annual Meeting,
Washington, DC.
Carter, M. 2005. “Gender differences in experience with and fear of crime in relation to public
transport.” In Research on women’s issues in transportation Vol. 2, Conference proceedings 35,
100. Washington, DC: Transportation Research Board.
Centers for Disease Control and Prevention. 1996. Physical Activity and Health: A Report of the
Surgeon General. U.S. Department of Health and Human Services, Atlanta, GA.

281

Centers for Disease Control and Prevention. 2008. Behavioral Risk Factor Surveillance System:
2007 Codebook Report. U.S. Department of Health and Human Services, Atlanta, GA.
Centers for Disease Control and Prevention. 2009. NCHS Health E-Stat, Prevalence of
overweight, obesity and extreme obesity among adults: United States, trends 1960-62 through
2005-2006. http://www.cdc.gov/nchs/data/hestat/overweight/overweight_adult.htm. Last
accessed August 2nd, 2010.
Centers for Disease Control and Prevention. Date last reviewed: January 8, 2010. U.S. Physical
Activity Statistics. http://www.cdc.gov/nccdphp/dnpa/physical/stats/index.htm. Last accessed
August 1st, 2010.
Centers for Disease Control and Prevention. Overweight and Obesity.
http://www.cdc.gov/nccphp/dnpa/obesity. Last accessed July 1st, 2010.
Cervero, R. 1996. “Mixed land uses and commuting: evidence from the American housing
survey.” Transportation Research 30A (5):361–377.
Cervero, R. 2002. “Built Environment and Mode Choice: Toward a Normative Framework.”
Transportation Research Part D 7:265-284.
Cervero, R., and R. Gorham. 1995. “Commuting in transit versus automobile neighborhoods.”
Journal of the American Planning Association 61 (2):210-225.
Cervero, R., and K. Kockelman. 1997. “Travel Demand and the 3Ds: Density, Diversity and
Design.” Transportation Research D (3):199-219.
Collia,D., J. Sharp and L. Giesbrecht. 2003. “The 2001 National Household Travel Survey: a
look into the travel patterns of older Americans.” Journal of Safety Research 34:461–470.
King, L. J. 1970. "Discrimnant analysis – review of recent theoretical contributions and
applications." Economic Geography 46 (2):367-378. doi: 10.2307/143150.
Crane, R. 1996. “Cars and drivers in the new suburbs: Linking access to travel in neotraditional
planning.” Journal of the American Planning Association 62:51-65.
Crane, R. 2000. “The influence of urban form on travel: An interpretative review.” Journal of
Planning Literature 15 (1):3-23.
Crane, R. and R. Crepeau. 1998. “Does neighborhood design influence travel?: A behavioral
analysis of travel diary and GIS data.” Transportation Research Part D: Transport and
Environmen 3 (4):225-238.
Cutts, B. B., K. L. Darby, C. G. Boone, and A. Brewis. 2009. “City structure, obesity, and
environmental justice: An integrated analysis of physical and social barriers to walkable streets
and park access.” Social Science and Medicine 69 (9):1314-1322.

282

Darden, J. T. 1987. “Choosing neighbors and neighborhoods: The role of race in housing
preference.” In Divided Neighborhoods: Changing Patterns of Racial Segregation, ed. G. A.
Tobin, p. 15-42. Beverly Hills, CA: Sage Publications.
Darden, J. T. 1989. “Blacks and other racial minorities: The significance of color in inequality.”
Urban Geography 10:562-77.
Darden, J. T., R. Hill, J. Thomas and R. Thomas. 1987. Detroit: Race and Uneven Development.
Philadelphia: Temple University Press.
Darden, J. T., and S. Kamel. 2000. “Black residential segregation in suburban Detroit: Empirical
testing of the ecological theory.” The Review of Black Political Economy 27:103-23.
Darden, J. T., and S. Kamel. 2000. “Black residential segregation in the city and suburbs of
Detroit: Does socioeconomic status matter?” Journal of Urban Affairs 22:1-13.
Darden, J. T., R. W. Thomas and C. Stokes. 2007. The State of Black Michigan: 1967-2007.
Michigan State Press: East Lansing, MI.
Darden, J., R.J. Schaetzl, and D. Brandt. D. (eds). 2009. Michigan Geography and Geology.
Pearson Custom Publishers, Boston, MA.
Darden, J. T., and R. W. Thomas. 2013. Detroit: race riots, racial conflicts, and efforts to bridge
the racial divide. East Lansing: Michigan State University Press.
Dill, J., and K. Voros. 2007. "Factors affecting bicycling demand - Initial survey findings from
the Portland, Oregon, Region." Transportation Research Record (2031):9-17. doi:
10.3141/2031-02.
Domencich, T. and D. L. McFadden. 1996. Urban Travel Demand: A Behavioral Analysis.
North-Holland Publishing Co.
Edwards, P. and A.Tsouros. 2006. The Solid Facts: Promoting physical activity and active living
in urban environments. WHO Europe: Copenhagen, Denmark.
Elliott, J.R. and M.S. Joyce. 2004. “The effects of Race and Family Structure on Women’s
Spatial Relationship to the Labor Market.” Sociological Inquiry 74(3):411-435.
Ewing, R. 1997a. Transportation and Land Use Innovations. Chicago: American Planners
Association.
Ewing, R. 1997b. “Is Los Angeles style sprawl desirable?” Journal of the American Planning
Association 63: 107–126.
Ewing, R. 1998. “Pedestrian and Transit-Friendly Design.” In Third National Conference on
Access Management. Fort Lauderdale, FL: U.S. Department of Transportation Federal Highway
Administration.

283

Ewing, R. 2000. “Pedestrian- and Transit-Friendly Design: A Primer for Smart Growth.” EPA
Smart Growth Network. P.1-22.
Ewing R. and R. Cervero. 2001. “Travel and the Built Environment: A synthesis.”
Transportation research Record: Journal of the Transportation Research Board 1870:87-113.
Ewing, R., and R. Cervero. 2010. “Travel and the Built Environment.” Journal of the American
Planning Association 76 (3):265-294.
Farley, S. Danziger and H.J. Holzer, 2000. Detroit Divided. Russell Sage Foundation, New York.
Hu, P. S., and T. R. Reuscher. 2004. Summary of Travel Trends: 2001 National Household
Travel Survey. Washington, DC: Federal HighwayAdministration, U.S. Department
ofTransportation.
Flegal, K.M., Carroll, M.D., Ogden, C.L., Curtin, L.R. 2010. “Prevalence and Trends in Obesity
Among US Adults, 1999-2008.” Journal of the American Medical Association 303(3): 235-241.
Forsyth A., and Southworth, M. 2008. “Cities afoot—pedestrians, walkability, and urban
design.” Journal of Urban Design 13(1):1–3.
Frank, L. 2000. “Land Use and Transportation Interaction: Implications on Public Health and
Quality of Life.” Journal of Planning, Education, and Research 20 (1): 6-22.
Frank, L. D. 2004. “Public Health and the Built Environment, Emerging Evidence and
Complexity.” Insert to Canadian Journal of Dietetic Practice and Research 65 (2).
Frank, L. D. and G. Pivo, G.1995. “Impacts of Mixed Use and Density on Utilization of Three
Modes of Travel: SOV, Transit and Walking.” Transportation Research Record 1466:44-55.
Frank, L. D., B. Stone, and W. Bachman. 2000. “Linking land use with household vehicle
emissions in the central Puget Sound: methodological framework and findings.” Transportation
Research Part D: Transport and Environment 5(3):173–196.
Frank, L. D., and P. Engelke. 2001. “The built environment and human activity patterns:
exploring the impacts of urban form on public health.” Journal of Planning Literature 16
(2):202–218.
Frank, L. D., M. Andresen, and T. Schmid. 2004. “Obesity Relationships with Community
Design, Physical Activity, and Time Spent in Cars.” American Journal of Preventive Medicine
27 (2):87-96.
Frank, L. D., and P. Engelke. 2005. “Multiple impacts of the built environment on public health:
walkable places and the exposure to air pollution.” International Regional Science Review 28
(2):193–216.

284

Frank, L.D., T. L. Schmid, J. F. Sallis, J. Chapman, and B. E. Saelens. 2005. “Linking
objectively measured physical activity with objec­tively measured urban form.” American
Journal of Preventive Medicine 28(2):117–125.
Frank, L.D., J. F. Sallis, T. L. Conway, J.E. Chapman, B.E. Saelens and W. Bachman. 2006.
“Many Pathways from Land Use to Health: Associations between Neighborhood Walkability
and Active Transportation, Body Mass Index, and Air Quality.” Journal of the American
Planning Association 72 (1):75-87.
Frank, L. D., B. E. Saelens, K. E. Powell, and J. E. Chapman. 2007. “Stepping towards
causation: do built environments or neighborhood and travel preferences explain physical
activity, driving, and obesity?” Social Science and Medicine 65(9):1898–1914.
Frank, L. D., J. F. Sallis, B.E. Saelens, L. Leary, K. Cain, T. Conway, and P. Hess. 2009. “The
development of a walkability index: application to the Neighborhood Quality of Life Study.”
British Journal of Sports Medicine doi:10.1136/bjsm.2009.058701
Freedman, O., and C.R.Kern. 1997. “A Model of Workplace and Residence Choice in Twoworker Households.” Regional Science and Urban Economics 27:241-260.
Friedman, B., S.P. Gordon., and J. B. Peers. 1994. “Effect of neotraditional neighborhood design
on travel characteristics.” Transportation Research Record (1466):63-70.
Frumkin, H., L. D. Frank, and R. Jackson. 2004. Urban sprawl and public health: designing,
planning, and building for healthy communities. Washington, DC: Island Press.
Frye, A. 2005. ‘Keynote Address.’ Conference Proceedings 35: Research on Women’s Issues in
Transportation: Volume 1. 2004 Nov. in Chicago, IL. Washington DC.: Transportation
Research Board.
Gallagher, J. 2010. Reimagining Detroit: opportunities for redefining an American city. Detroit,
Michigan: Wayne State University Press.
Gallagher, J. 2013. Revolution Detroit: strategies for urban reinvention. Detroit, Michigan:
Wayne State University Press.
Gibson, K. J., W. A. Darity, and S. L. Myers. 1998. "Revisiting occupational crowding in the
United States: A preliminary study." Feminist Economics 4 (3):73-95.
Giles-Corti, B., G. Wood, T. Pikora, V. Learnihan, M. Bulsara, K. Van Niel, A. Timperio, G.
McCormack, and K. Villanueva. 2011. “School site and the potential to walk to school: The
impact of street connectivity and traffic exposure in school neighborhoods.” Health & Place 17
(2):545-550.
Giuliano, G., and D. Narayan. 2003. “Another look at travel patterns and urban form: The US
and Great Britain.” Urban Studies 40 (11):2295-2312.

285

Giuliano, G., and J. Dargay. 2005. “Car ownership, travel and land use: A comparison of the US
and Great Britain.” Transportation Research Part A 40:106-24.
Golob, T. F., and M. G. McNally. 1997. “A model of activity participation and travel interactions
between household heads.'' Transportation Research B 31:177-194.
Grady, S. C. 2006. “Racial disparities in low birthweight and the contribution of residential
segregation: A multilevel analysis.” Social Science and Medicine 63:3013-29.
Grady, S.C. and S. McLafferty. 2007. “Segregation, nativity, and health: reproductive health
inequalities for immigrant and native-born black women in New York City.” Urban Geography
(Special Issue: New Directions in Segregation Research) 2S (4):377-397.
Grengs, J. 2010. “Job accessibility and the modal mismatch in Detroit.” Journal of Transport
Geography 18:42-54.
Hagerstrand, T. 1970, “What about people in regional science?” Papers of the Regional Science
Association 24: 7-21.
Hamilton, K., and L. Jenkins. 1990. “Why Women and Travel” in Gender, Transport, and
Employment: the Impact of Travel Constraints edited by Grieco, M., L.Pickup, and R.Whipp,
Oxford Transport Studies, Gower Publishing Company Limited. Hants, England, p.17-45.
Handy, S. 1993. “Regional versus local accessibility: neo-traditional development and its
implications for non-work travel.” Built Environment. 18 (4):256–267.
Handy, S. 1996. “Methodologies for Exploring the Link between Urban Form and Travel
Behavior.” Transportation Research D 1 (2):155-165.
Handy, S. 2005a. “Smart Growth and the Transportation-Land use Connection.” International
Regional Science Review 28(2):146-167.
Handy, S. 2005b. “Critical Assessment of the Literature on the Relationships Among
Transportation, Land Use, and Physical Activity.” Resource paper for Special Report 282: Does
the Built Environment Influence Physical Activity? Examining the Evidence. TRB; Committee on
Physical Activity, Health, Transportation, and Land Use, Institute of Medicine, National
Academies, Washington, D.C.http://trb.org/downloads/sr282papers/sr282Handy.pdf.
Handy, S. 2006a. “Questioning Assumptions 1: Do New Urbanists Walk More?” Planning 72
(1):36-38.
Handy, S. 2006b. “The Road Less Driven.” Journal of the American Planning Association 72
(3):274- 278.
Handy, S. M. G. Boarnet, R. Ewing, and R. E. Killingsworth. 2002. “How the built environment
affects physical activity: views from urban planning.” American Journal of Preventive Medicine
23 (2S):64–73.

286

Handy, S., X. Cao, and P. Mokhtarian. 2005. “Correlation or Causality Between the Built
Environment and Travel Behavior? Evidence from Northern California.” Transportation
Research 10:427–444.
Handy, S., X. Cao, and P. L. Mokhtarian. 2006. “Self-Selection in the Relationship between the
Built Environment and Walking.” Journal of the American Planning Association 72 (1):55-74.
Hansen, W. G., 1959. “How accessibility shapes land use.” Journal of the American Institute of
Planners 12 (2):73–76.
Hanson, S. 1982. “The determinants of daily travel-activity patterns: Relative location and sociodemographic factors.” Urban Geography 3:179–202.
Hanson. S. 1986. The Geography of urban Transportation. New York, NY, USA: Guilford.
Hanson, S.(ed.), 1995. The Geography of Urban Transportation 2nd Edition. New York:
Guilford Press.
Hanson, S. 2010. “Gender and mobility: new approaches for informing sustainability.” Gender,
Place and Culture 17 (1):5–23.
Hanson, S. and P. Hanson. 1980a. “Gender and Urban Activity Patterns in Uppsala, Sweden.”
Geographical Review 70 (3):291-299.
Hanson, S. and P. Hanson. 1980b. “The Impact of Women’s Employment on Household Travel
Patterns: A Swedish Example.” In Women’s Travel Issues: Research Needs and Priorities:
Conference Proceedings and Papers (S Rosenbloom, ed.), U.S. Department of Transportation,
U.S. Government Printing Office, Washington, D.C., 1980
Hanson, S. and P. Hanson. 1981. “The Impact of Married Women's Employment on Household
Travel Patterns.” Transportation 10:165 183.
Hanson, S. and I. Johnston. 1985. “Gender Differences in Work-trip Lengths: Explanations and
Implications.” Urban Geography 6 (3):193-219.
Hanson, S. and G. Pratt. 1991. “Job Search and the Occupational Segregation of Women.”
Annals of the Association of American Geographers 81:229-253.
Hanson, S. and G. Giuliano. 2004. The Geography of Urban Transportation 3rd Edition. New
York: The Guilford Press.
Harvey, D. 1985. The Urbanization of Capital. Basil Blackwell (Oxford) and Johns Hopkins
University Press.
Holtzclaw, J. 1994. Residential Patterns and Transit, Auto Dependence, and Costs. Resources
Defense Council, San Francisco.

287

Hu, P.S., J. Young. 1999. Summary of Travel Trends: 1995 Nationwide Personal Transportation
Survey. Federal Highway Administration, U.S. Department of Transportation. Washington, D.C.
Ingalls, G., D. Hartgen, and T. Owens. 1994. “Public fear of crime and its role in bus transit
use.” Transportation Research Record 1433:201-11.
Jargowsky, P.A. 1997. Poverty and place: ghettos, barrios, and the American city. Russell Sage
Foundation, New York.
Johnston-Anumonwo, I. 1989. “Journey to Work: A Comparison of Characteristics of Single and
Married Parents.” Journal of Specialized Transportation Planning and Practice 3 (3):219-246.
Johnston-Anumonwo, I. 1992. “The influence of household type on gender differences in work
trip distance.”Professional Geographer 44: 161-169.
Johnston-Anumonwo, I. 1997. “Race, Gender, and Constrained Work Trips in Buffalo, NY,
1990.” The Professional Geographer 49:306–317.
Jones, P. 1990. “Household organization and travel behavior”In Gender, Transport, and
Employment: the Impact of Travel Constraints edited by Grieco, M., L.Pickup, and R.Whipp,
Oxford Transport Studies, Gower Publishing Company Limited. Hants, England, p. 17-45.
Kain, J. F. 1968. “Housing segregation, negro employment, and metropolitan decentralization.”
Quarterly Journal of Economics 82 (2):175-197.
Kain, J. and R. Fauth. 1977. “The Effect of Urban Structure on Auto Ownership and Journey to
Work Modes.” Transportation Research Record 658:9-17.
Kawabata, M and Q. Shen, 2007. “Commuting inequality between cars and public transit: the
case of the San Francisco Bay Area, 1990–2000.” Urban Studies 44 (9):1759–1780.
King, L.J. 1970. “Discriminant Analysis: A Review of Recent Theoretical Contributions
andApplications.” Supplement: Proceedings. International Geographical Union, Economic
Geography 46:367-378.
Kitamura, R., P. Mokhtarian, and L. Laidet. 1994. A micro-analysis of land use and travel in five
neighborhoods in the San Francisco Bay area. University of California at Davis report prepared
for the California Air Resources Board.
Kockelman, K. M., 1999. “Application of a Utility-Theory-Consistent System-of-DemandEquations Approach to Household Travel Choice.” Transportation Research Record 1676:20-29.
Koskela, H., and R. Pain. 2000. “Revisiting fear and place:Women’s fear of attack and the built
environment.” Geoforum 31:269-80.
Krizek, K. J. 2003. “Residential relocation and changes in urban travel - Does neighborhoodscale urban form matter?” Journal of the American Planning Association 69 (3):265-281.

288

Kwan, M. P. 1999a. “Gender and Individual Access to Urban Opportunities: A Study Using
Space-Time Measures.” The Professional Geographer 51 (2):210-227.
Kwan, M. P. 1999b. “Gender, the Home-work Link, and Space-time Patterns of Nonemployment Activities.” Economic Geography 75 (4):370-394.
Kwan, M. P. 2000. Evaluating Gender Differences in Individual Accessibility: A Study Using
Trip Data Collected by the Global Positioning System. Final Report to the Federal Highway
Administration.
Lawrence Frank & Company (LFC) Inc, Mark Bradley and Keith Lawton Associates. 2005.
Travel Behavior, Emissions, & Land Use Correlation Analysis in the Central Puget Sound.
Prepared for the Washington State Department of Transportation, Olympia WA. Report no. WARD 625.1
LeDoux, T. F., and I. Vojnovic. 2013. “Going outside the neighborhood: The shopping patterns
and adaptations of disadvantaged consumers living in the lower eastside neighborhoods of
Detroit, Michigan.” Health & Place 19:1-14.
Lee, C. and A. V. Moudon. 2004. “Physical Activity and Environment Research in the Health
Field: Implication for Urban and Transportation Planning Practice and Research.” Journal of
Planning Literature 19 (2):147-181.
Leslie E., N. Coffee, L. Frank, N. Owen, A. Bauman and G. Hugo. 2007. “Walkability of local
communities: using geographic infor-mation systems to objectively assess relevant
environ­mental attributes.” Health & Place 13 (1):111–122.
Levinson, D. M., and A. Kumar. 1997. “Density and the journey to work.” Growth and Change
28 (2):147-172.
Light, A. 2004. “Gender Differences in the Marriage and Cohabitation Income Premium.”
Demography41 (2):263-284.
Litman, T. 2003. “Economic Value of Walkability.” Transportation Research Record 1828:3-11.
Lopez, R. P. and H. P. Hynes. 2006. "Obesity, physical activity, and the urban environment:
public health research needs." Environmental Health: A Global Access Science Source.
doi:10.1186/1476-069X-5-25.
Loukaitou-Sideris, A. and C. Flink. 2009. “Settings: a survey of U.S. transit agencies addressing
women’s fear of victimization in transportation.” Urban Affairs Review 44 (4):554–588.
MacDonald, H. I. 1999. “Women’s Employment and Commuting: Explaining the Links.”
Journal of Urban Planning Literature 13 (3):267-283.
Manheim, M. L. 1979. Fundamentals of Transportation Systems Analysis. The Massachusetts
Institute of Technology Press, Cambridge, Massachusetts and London, England.

289

Marshall, J.D., M. Brauer, and L.D. Frank. 2009. “Healthy Neighborhoods: Walkability and Air
Pollution.” Environmental Health Perspectives 117 (11):1752–1759.
Martelle, S. 2012. Detroit: a biography. Chicago, Ill: Chicago Review Press.
Martin, B. V., F. W. Memmott, III and A.J. Bone. 1967. Principles and techniques of predicting
future demand for urban area transportation: MIT report No.3. The Massachusetts Institute of
Technology Press, Cambridge, Massachusettsand London, England.
Martin, W.A. and N.A. McGuckin. 1998. Travel Estimation Techniques for Urban Planning,
NCHRT Report 365. Transportation Research Board, Washington D.C.
Massey, D.S. and N. A. Denton. 1993. American apartheid: Segregation and the making of the
underclass. Harvard University Press, Cambridge, Mass.
McConnell, J.E., 1979. “The Export Decision: An Empirical Study of Firm Behavior.” Economic
Geography 55 (3):171-183.
McFadden, D.L. 1974. "The Measurement of Urban Travel Demand." Journal of Public
Economics 3 (4):303-328.
McFadden, D. and F. Reid. 1975. “Aggregate Travel Demand Forecasting from Disaggregated
Behavioral Models." Transportation Research Record: Travel Behavior and Values 534:24-37.
McFadden, D.L. 2001. “Disaggregate Behavioral Travel Demand's RUM Side: A 30-Year
Retrospective.” The Leading Edge of Travel Behavior Research by David Heshner (ed.)
Pergamon Press: Oxford.
McLafferty, S. and V., Preston. 1992. “Spatial mismatch and labor market segmentation for
African-American and Latina women.” Economic Geography 68:406-431.
McLafferty, S. and V., Preston. 1996a. “Gender, race and the determinants of commuting: New
York in 1990.” Urban Geography 18 (3):192-212.
McLafferty, S. and V., Preston. 1996b. “Transportation and minority women's employment:
insights from New York.” Paper presented at the Second National Conference on Women's
Travel Issues, Baltimore, MD, 1996 October.
McLafferty, S. and V., Preston. 1996c. “Spatial mismatch and employment in a decade of
restructuring.” The Professional Geographer 48:420-431.
McNally, M.G., 2000. “The Four Step Model.” In: Handbook of Transport Modelling, ed.
Hensher, D.A., and K.J.Button, p. 35-69.
Messenger, T. and Ewing, R. 1996. “Transit-orienteddevelopment in the Sun Belt.”
Transportation Research Record 1552:145-153.

290

Meyer, J.R. and J.A. Gómez-Ibañez, 1981. Autos, Transit and Cities. Harvard University Press,
Cambridge, MA.
Meyer, M.D. and E. J. Miller. 1984. Urban Transportation Planning: a decision oriented
approach. McGraw-Hill Book Company.
Mingione, E. (ed.). 1996. Urban Poverty and the Underclass. Oxford: Basil Blackwell.
Moudon, A. V., C. Lee, A. D. Cheadle, C. Garvin, D. Johnson, and T. L. Schmid. 2006.
“Operational definitions of walkable neighborhood: theoretical and empirical insights.” Journal
of physical activity & health 3 (1):S99–S117.
Naess, P. 2006. “Accessibility, activity participation and location of activities: exploring the
links between residential location and travel behaviour.” Urban Studies 43:627–652.
Newman, P.W. G. and Kenworthy, J. R. 1989. “Gasoline consumption and cities: a comparison
of U.S. cities with a global survey.” Journal of the American Planning Association 55 (1):24–37.
Nobis, C. and L. Barbara. 2005. “Gender Differences in Travel Patterns: Role of employment
status and household structure.” In Conference Proceedings 35: Research on Women’s Issues in
Transportation: Volume 2. 2004 Nov. in Chicago, IL. Washington DC.: Transportation Research
Board.
O’Sullivan, S. and J. Morrall. 1996. “Walking Distance to and from Light-Rail Transit Stations.”
Transportation Research Record 1538:19-22.
Owen, N., N, Humpel, E. Leslie, A. Bauman, and J. F. Sallis. 2004. “Understanding
environmental influences on walking: review and research agenda.” American Journal of
Preventive Medicine 27 (1):67–76.
Paaswell, R.E., W.W. Recker and V. Milione. 1976. “Profile of a carless population.”
Transportation Research Record 578:16-28.
Parks, V. 2004. “Access to Work: the Effect of Spatial and Social Accessibility on
Unemployment for Native-Born Black and Immigrant Women in Los Angeles.” Economic
Geography 80 (2):141-172.
Pickup, L. 1985, “Women’s gender role and its influence on travel behavior.” Built Environment
10 (1):61-69.
Pickup, L., 1990, “Women’s travel requirements: employment, with domestic constraints.” In
Gender, Transport, and Employment: the Impact of Travel Constraints edited by Grieco, M.,
L.Pickup, and R.Whipp, Oxford Transport Studies, Gower Publishing Company Limited. Hants,
England, p. 17-45.
Preston, V. and S. McLafferty. 1992. “Income disparities and employment and occupational
changes in New York.” Regional Studies 27:223-235.

291

Preston, V. and S. McLafferty. 1993. “Gender differences in commuting at suburban and central
locations.” Canadian Journal of Regional Science 16:237-259.
Preston, V. and S. McLafferty. 1999. “Spatial mismatch research in the 1990s: progress and
potential.” Papers in Regional Science 78:387-402.
Pucher, J. and J. L. Renne, 2003. “Socioeconomics of urban travel: evidence from the 2001
NHTS.” Transportation Quarterly 57(3):49-77.
Rosenbloom, S. 1989. “The Transportation Needs of Single Salaried Mothers; A Critical
Analysis.” Journal of Specialized Transportation Planning and Practice 3 (3):295-310.
Rosenbloom, S. 2005. “Understanding Women’s and Men’s Travel Patterns: The research
challenge.”In Conference Proceedings 35: Research on Women’s Issues in Transportation:
Volume 12004 Nov. in Chicago, IL. Washington DC.: Transportation Research Board.
Rosenbloom, S. and C. Raux. 1985. “Employment, Childcare and Travel Behavior: France, The
Netherlands, and the United States.” In Behavioral Research for Transport Policy, VNU Science
Press, Utrecht, The Netherlands. p. 365-381.
Rutherford, B. and G. R. Wekerle. 1989. “Single Parents in the Suburbs: Journey-to-work and
Access to Transportation.” Journal of Specialized Transportation Planning and Practice III
(3):277-294.
Rutherford, B.and G. R. Wekerle .1988. “Captive Rider, Captive labor: Spatial Constraints and
Women’s Employment.” Urban Geography 9 (2):116-137.
Saelens, B. E., J. F. Sallis, and L.D. Frank. 2003a. “Environmental Correlates of Walking and
Cycling: Findings from the Transportation, Urban Design, and Planning literatures.” Annals of
Behavioral Medicine 25 (2): 80-91.
Saelens, B.E., J. F. Sallis, J. B. Black, and D. Chen. 2003b. “Neighborhood-based differences in
physical activity: An environment scale evaluation.” American Journal of Public Health
93:1552-1558.
Sallis, J. F., L. D. Frank, B. E. Saelens, and M. K. Kraft. 2004. “Active transportation and
physical activity: Opportunities for collaboration on transportation and public health research.”
Transportation Research Part A: Policy and Practice 38:249-268.
Sanchez, T.W., W. Shen, and Z.R. Peng. 2004. “Transit mobility, jobs access and low-income
labour participation in US metropolitan areas.” Urban Studies 41 (7):1313–1331.
Santos, A., N. McGucking, H.Y. Nakamoto, D. Gray, and S. Liss. 2011. Summary of Travel
Trends: 2009 National Household Travel Survey. Federal Highway Administration, U.S.
Department of Transportation. Washington, D.C.
Sassen, S. 2001. original in 1991. The global city: New York, London, Tokyo. 2nd Ed. Princeton:
Princeton University Press.
292

Schintler, L., A. Root, K. Button, and T. R. B. 2000. "Women's travel patterns and the
environment - An agenda for research." In Public and Nonmotorized Transportation in Other
Nations; Women's Issues in Transportation: Safety and Human Performance 1726: 33-40.
Schulz, D., and S. Gilbert. 1996. “Women and transit security: A new look at old issues.” In
Women’s Travel Issues Second National Conference. Washington, DC: Federal Highway
Administration.
Schwanen, T., M. Dijst, and F.M. Dieleman. 2002. “A microlevel analysis of residential context
and travel time.” Environment and Planning A 34(8):1487–1507.
Schwanen, T., M.P. Kwan and F. Ren, 2008. “How fixed is fixed? Gendered rigidity of spacetime constraints and geographies of everyday activities.” Geoforum 39 (6):2109–2121.
[SEMCOG] Southeast Michigan Council of Governments. 2003. Land use change in southeast
Michigan: Causes and consequences.
Silk, J. 1979. Analysis of covariance and comparison of regression lines (Concepts and
techniques in modern geography; no. 20). Geo Abstracts, University of East Anglia.Downloaded
from http://www.qmrg.org.uk/files/2008/11/20-analysis-of-covariance.pdf.
Sommer, B. B. and R. Sommer. 1997. A Practical Guide to Behavioral Research: Tools and
Techniques. New York: Oxford University Press.
Srinivasan, S., and C. R. Bhat. 2005. "Modeling household interactions in daily in-home and outof-home maintenance activity participation." Transportation 32 (5):523-544. doi:
10.1007/s11116-005-5329-z.
Sugrue, T. J. 1996. The Origins of the Urban Crisis: Race and Inequality in Postwar Detroit.
Princeton: Princeton University Press.
Surface Transportation Policy Project. 1999. High mileage moms. http://www.transact.org/
Thomas, J. M. 1997. Redevelopment and Race: Planning a Finer City in Postwar Detroit.
Baltimore: Johns Hopkins University Press.
Littman, T. 2003. "Economic Value of Walkability." Transportation Research Record 1828:311.
Transportation Research Board (TRB). 2005a. Research on women’s issues in transportation,
Vol. 1, Conference overview and plenary papers, Conference Proceedings 35. Washington, DC:
Author.
Transportation Research Board (TRB). 2005b. Research on women’s issues in transportation,
Vol. 2, Technical papers, Conference Proceedings 35. Washington, DC: Author.
Turner, S. C. 1997. “Barriers to a better break: employer discrimination and spatial mismatch in
metropolitan Detroit.” Journal of Urban Affairs 19 (2):123–141.
293

Turner, T., and D. Niemeier. 1997. “Travel to work and household responsibility: new
evidence.” Transportation 24:397–419.
US Census Bureau. 2002. Demographic Trends in the 20th Century.
http://www.census.gov/prod/2002pubs/censr-4.pdf
U.S. Census Bureau. 2010. Income, Poverty, and Health Insurance Coverage In the United
States: 2009, Current Population Report. U.S. Government Printing Office, Washington, DC.
Untermann, R. 1984. Accommodating the Pedestrian: Adapting Towns and Neighborhoods for
Walking and Bicycling. Van Nostrand Reinhold, New York.
Vojnovic, I. 1999. “The environmental costs of modernism: an assessment of Canadian cities.”
Cities: The International Journal of Urban Policy and Planning 16 (5):301-313.
Vojnovic, I. 2000a. “Municipal restructuring, regional planning, and fiscal accountability:
designing municipal tax zones in two Maritime regions.” Canadian Journal of Regional Science
23 (1):49-72.
Vojnovic, I. 2000b. “Shaping Metropolitan Toronto: A study of linear infrastructure subsidies.”
Environment and Planning B: Planning and Design 27:197–230.
Vojnovic, I. 2000c. “The transitional impact of municipal consolidations.” Journal of Urban
Affairs 22:385–417.
Vojnovic, I. 2003a. “Governance in Houston: Growth theories and urban pressures.” Journal of
Urban Affairs 25:589–624.
Vojnovic, I. 2003b. “Laissez-faire governance and the archetype laissez-faire city in the USA:
Exploring Houston.” Geografiska Annaler B (85):19–38.
Vojnovic, I. 2006. “Building Communities to promote physical activity: A Multiscale Analysis.”
Swedish Society for Anthropology and Geography 88 B (1):67-90.
Vojnovic, I. 2009. “Urban Settlements: Suburbanization and the Future” In Schaetzl, R., Darden,
J., and Brandt, D. (eds.) Geography and Geology of Michigan. Boston: Pearson Custom
Publishers. p.487-507.
Vojnovic, I. C. Jackson-Elmoore, J. Holtrop, and S. Bruch. 2006. “The renewed interest in urban
form and public health: Promoting increased physical activity in Michigan.” Cities 23 (1):1-17.
Vojnovic, I., Smith S., Kotval, Z., and Lee, J. 2008. “Built Environments: Planning Cities to
Encourage Physical Activity” In Dele Davies (Ed.) Obesity in Childhood and Adolescence:
Medical, Biological, and Social Issues (Vol. 1). Westport: Praeger Press. p. 93-129.
Vojnovic, I., and J. T. Darden. 2013a. “Racial Conflict, Intolerance, and Distortions in Urban
Form: Lessons for Sustainability from the Detroit Region.” Ecological Economics 96:88-98.

294

Vojnovic, I., and J. T. Darden. 2013b. The Colors That Shaped a City: The Role of Racial and
Class Tensions in Inhibiting Urban Sustainability, the Detroit Context In Urban Sustainability: A
Global Perspective edited by Vojnovic, I. Michigan State University Press, East Lansing, USA.
p. 435-474.
Vojnovic, I., J. Lee, Z. Kotval-K, A. Podagrosi, P. Varnakovida. T. LeDoux, and J. Messina.
2013. “The Burdens of Place: A Socio-economic and Ethnic/Racial Exploration into Urban
Form, Accessibility, and Travel Behavior in the Lansing Capital Region, Michigan.” Journal of
Urban Design 18 (1):1-35.
Vojnovic, I., Z. Kotval-K, J. Lee, M. Ye, T. LeDoux, P. Varnakovida, and J. Messina.
Forthcoming. “Urban Built Environments, Accessibility, and Travel Behavior in a Declining
Urban Core: An Exploration into the Detroit Region.” Journal of Urban Affairs.
Wekerle, G.R. 1981. “Women in the Urban Environment.” In C.R. Stimpson et al. (eds.), Women
and the American City. Chicago: University of Chicago Press.
Wekerle, G. R. 2005. “Gender and the Global City: Urban Restructuring, Social Exclusion and
Claims for Justice” in H. Hiller (ed) Urban Sociology in Canada. Toronto: Oxford University
Press
Wekerle, G. R., R. Peterson, and D. Morley (eds). 1980. New Space for Women. Boulder, CO:
Westview Press.
Wekerle, G. R. and B. Rutherford. 2004. “The Mobility of Capital and the Immobility of Female
Labor: Responses to Economic Restructuring.” in L. Servon and S. Fainstein (eds) Gender
Planning and Policy. New Brunswick NJ: Rutgers University Press.
White, H.P., and M. L. Senior. 1983. Transport Geography. Longman Inc.: New York, USA.
Wilson, W. J. 1987. The declining significance of race: The truly disadvantaged—the inner city,
the underclass, and public policy. Chicago: University of Chicago Press.
Wilson, W. J. 2009. More than Just Race: Being Black and Poor in the Inner City. 1st Edition.
W. W. Norton & Company.
Wyly, E. K. 1996. “Race, Gender, and Spatial Segmentation in the Twin Cities.” The
Professional Geographer 48 (4):431-444.
Wyly, E. K. 1998. “Containment and Mismatch: Gender Differences in Commuting in
Metropolitan Labor Markets.” Urban Geography 19 (5):395-430.
Wyly, E. K. 1999. “Occupational Sex Segregation and Local Labor Markets in an American
Metropolis.” Journal of Urban Affairs 21 (1):1-33.

295