THE BUILT ENVIRONMENT, TRAVEL PATTERNS AND ENVIRONMENTAL BURDENS:
A STUDY OF SIX NEIGHBORHOODS IN THE DETROIT, MICHIGAN REGION
By
Zeenat Kotval-Karamchandani

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Geography - Doctor of Philosophy
2013

ABSTRACT
THE BUILT ENVIRONMENT, TRAVEL PATTERNS AND ENVIRONMENTAL BURDENS:
A STUDY OF SIX NEIGHBORHOODS IN THE DETROIT, MICHIGAN REGION
By
Zeenat Kotval-Karamchandani
The transportation industry is a significant contributor of greenhouse gasses. Within the
transportation industry, light duty vehicles are the largest source of these emissions. Light duty
vehicle emissions result from travel behavior and vehicle technology. Consequently there has
been a plethora of studies that look into the travel patterns of residents living in various built
environments. Research to date has also studied the various factors that impact travel patterns
and travel behavior of residents with varying socio-economic characteristics. Linking travel
patterns with pollutant emissions has not been as thoroughly researched. Emissions have been
researched taking various built environment characteristics into consideration: these include
congestion spots, highways, road conditions, and topography. Also, the effect of emissions on
resident health has also been studied. However, research on the socio-economic characteristics
of residents that affect environmental burdens, as measured by fuel consumption and vehicular
emissions, is sorely lacking in the geography, planning and transportation literature. This study
explores how the socio-economic composition of residents impacts travel patterns and
environmental burdens within six neighborhoods in the Detroit, Michigan region.
The dataset for this study includes primary travel behavior data from a mail survey, and
secondary data on vehicles from the Michigan Secretary of State and on specific gasoline
consumption and vehicle emissions from the Environmental Protection Agency. Methods of
analysis include Discriminant Analysis, K-Means Cluster Analysis and Regression Analyses to

derive the impacts of travel behavior and actual vehicle characteristics on travel patterns and
environmental burdens. The six study neighborhoods were combined into three neighborhood
typologies: Higher Density Urban (HDU), Higher Density Suburban (HDS) and Lower Density
Suburban (LDS). This enabled the study to look at the effect of varying socio-economic
characteristics of residents within each typology and how these contributed to environmental
burdens.
Results of the study show that holding the built environment constant, the wealthier
residents drive more and are responsible for higher levels of environmental burdens. This is the
case even after taking into consideration that lower income residents have older, less
technologically advanced vehicles that would typically consume more gas and emit higher levels
of emissions. In addition, disinvestment can transform neighborhoods traditionally viewed as
highly accessible (higher density, mixed land use and connected built environments) into
neighborhoods that maintain the accessibility characteristics of low density, single-use zoned and
disconnected suburbs. Despite this, it is evident that that environmental burdens associated with
travel are shaped by class, with higher income earners disproportionately contributing to
negative anthropogenic environmental impacts.

To my sister, my best friend, and my guardian angel Zenia, for her constant love, encouragement
and guidance.
To the love of my life, my husband Shelly, for his endless support, patience and love.

iv

ACKNOWLEDGEMENTS

I received a great deal of support and encouragement from some very special people in my life
throughout my years in the doctoral program.
First of all, I am honored to have had Dr. Igor Vojnovic as my mentor and advisor. This
dissertation has come to fruition due to his guidance, support and encouragement. I have gained
an invaluable friend and colleague in Igor. I would also like to thank my committee members:
Dr. Joe Darden, Dr. Joe Messina and Dr. Trish Machemer for their help and comments that
always improved my work.
I would like to send out a special acknowledgement to my family who has done everything they
could to help me out and support me as I progressed in my doctoral program. Specifically, I’d
like to my sister, Zenia, for always being there to help and encourage me in any and all ways
possible, my mom, Maharukh, for being the de facto babysitter (which was a big help), my
mother-in-law, Champa, for her constant encouragement to finish writing my dissertation, and
my brother, Zubin, for patiently going through vehicle characteristics and explaining “car” facts
to help me through with my study. Most of all, I’d like to thank my husband, Shelly, and my
son, Riley, as their care, support and love has been a phenomenal factor in the pursuance of my
career goals.
Lastly, having friends around to always talk to and share ideas, problems and achievements has
been a blessing. Thank you Jieun, Minting, Cristina and Rita for always being there for me.

v

TABLE OF CONTENTS

LIST OF TABLES

x

LIST OF FIGURES

xv

Chapter I – Introduction
The Research Question
Data Sources
Method of Analysis
General Hypothesis
Main Contribution and Implications

1
5
5
6
6
7

Chapter II – Literature Review
Suburbanization
Effects of Suburbanization
Travel Patterns/Behavior
Built Environment and Travel Demand
Socioeconomic Characteristics and Travel Behavior
Built environment and socioeconomic factors and travel behavior
Vehicle emissions and travel behavior
Driving Conditions
Vehicle Technology
Costs associated with travel and emissions
Scale and gap in research
Gap in Research, focusing on Detroit, MI

10
10
13
17
18
21
25
27
27
29
30
32
36

Chapter III – The Detroit Region
Introduction
The Growth Period
Economic/Industrial factors
Municipal Expansion
Social factors and related mobility
Transition Period
Economic/Industrial factors
The Role of Roads, Highways and the Automobile
Social factors and related mobility
Period of Decline
Economic/Industrial factors
Built environment factors and related mobility
Social factors and related mobility

40
40
45
45
45
46
50
50
50
52
57
57
58
60

vi

The Outlook: Dismal, yet hopeful!
Commentary

62
65

Chapter IV – Study Neighborhoods
Introduction
The Study Region
The Inner-City Detroit Neighborhoods
Built Environment: Detroit
Socio-Economic and Demographic Characteristics: Detroit
Travel Characteristics: Detroit
Neighborhood Specifics: Detroit
The Ann Arbor Neighborhood
Built Environment: Ann Arbor
Socio-Economic and Demographic Characteristics: Ann Arbor
Travel Characteristics: Ann Arbor
Neighborhood Specifics: Ann Arbor
The Birmingham Neighborhood
Built Environment: Birmingham
Socio-Economic and Demographic Characteristics: Birmingham
Travel Characteristics: Birmingham
Neighborhood Specifics: Birmingham
The Bloomfield Hills Neighborhood
Built Environment: Bloomfield Hills
Socio-Economic and Demographic Characteristics: Bloomfield
Hills
Travel Characteristics: Bloomfield Hills
Neighborhood Specifics: Bloomfield Hills
The West Bloomfield Neighborhood
Built Environment: West Bloomfield
Socio-Economic and Demographic Characteristics: West
Bloomfield
Travel Characteristics: West Bloomfield
Neighborhood Specifics: West Bloomfield
Commentary

67
67
67
73
73
74
75
76
82
82
83
84
87
90
90
91
92
94
98
98

Chapter V - Research Hypotheses, Data and Methodology
Introduction
Research Question
Objective
Hypotheses
OLS Regression Model
Data
Methodology
Rationale for the Neighborhood Typology Grouping
Quantitative Analyses
Discriminant Analysis (DA)

115
115
116
116
118
120
121
125
125
128
129

vii

100
101
103
106
106
109
109
111
114

Cluster Analysis (CA)
Regression Analyses
Caution
Commentary

132
133
135
136

Chapter VI – Results and Discussion
Analysis of Variance
Vehicle Characteristics
Demographic Characteristics
Travel Behavior
Discriminant Analysis (DA)
Discussion
Higher Density Urban – City of Detroit Neighborhoods
Cluster Analyses: The Higher Density Urban Neighborhood
Typology
Regression Analyses: The Higher Density Urban (HDU)
Neighborhood Typology
A Discussion by Income Groupings: The Higher Density
Urban (HDU) Neighborhood Typology
Discussion
Higher Density Suburban – Ann Arbor and Birmingham
Neighborhoods
Cluster Analyses: The Higher Density Suburban Neighborhood
Typology
Regression Analyses: The Higher Density Suburban (HDS)
Neighborhood Typology
A Discussion by Income Groupings: The Higher Density
Suburban (HDS) Neighborhood Typology
Discussion
Low Density Suburban – Bloomfield Hills and West Bloomfield
Neighborhoods
Cluster Analyses: The Low Density Suburban Neighborhood
Typology
Regression Analyses: The Low Density Suburban (LDS)
Neighborhood Typology
A Discussion by Income Groupings: The Low Density
Suburban (LDS) Neighborhood Typology
Discussion
Commentary
Work-based Trips
Non-Work Trips
Environmental Burdens

137
137
137
141
142
145
148
150

Chapter VII – Conclusion
Future Research Possibilities

228
241

viii

153
157
163
167
169
172
178
186
190
193
196
200
209
213
214
220
222
224

APPENDICES
Appendix A – Data Validation
Appendix B – Example of EPA Window Sticker
Appendix C – Example of Fuel Economy from the Department
of Energy
Appendix D – Code Book
Appendix E – Distributions of Variables (for Discriminant Analysis)
Appendix F - Distributions of Variables (for Regression Analyses)

244
245
247

BIBLIOGRAPHY

260

ix

249
250
253
254

LIST OF TABLES

Table 1: Per Capita Income by SEMCOG County

42

Table 2: Median Family Income by SEMCOG County

42

Table 3: Population Density of Census Tracts Encompassing the Detroit
Study Neighborhoods

78

Table 4: Population Density of Census Tracts Encompassing the Ann Arbor
Study Neighborhood

87

Table 5: Population Density of Census Tracts Encompassing the Birmingham
Study Neighborhood

94

Table 6: Population Density of Census Tracts Encompassing the Bloomfield
Hills Study Neighborhood

103

Table 7: Population Density of Census Tracts Encompassing the West
Bloomfield Study Neighborhood

111

Table 8: Built Environment characteristics by Neighborhood

126

Table 9: Built Environment characteristics by Neighborhood Type

126

Table 10: Variables used in a Stepwise DA

130

Table 11: Summary of PCA and Cluster Analyses

133

Table 12: Analysis of Variance Table across the three Neighborhood Types

138

Table 13: Variables used in the Discriminant Analysis

145

x

Table 14: Functions at Group Centroids

145

Table 15: Structure Matrix

146

Table 16: Group Statistics

147

Table 17: HDU Breakdown of Categorical Variables

150

Table 18: HDU - Descriptive Statistics

152

Table 19: Cluster Analysis on Demographic Variables – HDU

156

Table 20: Cluster Analysis on All Variables – HDU

156

Table 21: HDU Regression Results – Mode

159

Table 22: HDU Regression Results – Annual Miles

160

Table 23: HDU Regression Results – Vehicle Characteristics

161

Table 24: HDU Respondents split into three Household Income Groups

164

Table 25: HDU Vehicle Ownership

165

Table 26: HDU Demographic Characteristics

165

Table 27: HDU Mode of Travel

165

Table 28: HDU Average Annual Miles to End-Point Destinations by Trip
Purpose Based on Usual Weekly Travel Patterns

167

Table 29: HDU Average Environmental Burdens Based on Usual Weekly
Travel Patterns

167

Table 30: HDS Breakdown of Categorical Variables

170

xi

Table 31: HDS Descriptive Statistics

171

Table 32: Cluster Analysis on Demographic Variables – HDS

173

Table 33: Cluster Analysis on All Variables – HDS

173

Table 34: HDS Regression Results – Mode

181

Table 35: HDS Regression Results – Annual Miles

182

Table 36: HDS Regression Results – Vehicle Characteristics

183

Table 37: HDS Respondents split into three Household Income Groups

186

Table 38: HDS Vehicle Ownership

186

Table 39: HDS Demographic Characteristics

187

Table 40: HDS Mode of Travel Based on Usual Weekly Travel Patterns

187

Table 41: HDS Average Annual Miles to End-Point Destinations by Trip
Purpose Based on Usual Weekly Travel Patterns

188

Table 42: HDS Average Environmental Burdens Based on Usual Weekly
Travel Patterns

189

Table 43: LDS Breakdown of Categorical Variables

193

Table 44: LDS Descriptive Statistics

195

Table 45: Cluster Analysis on Demographic Variables – LDS

197

Table 46: Cluster Analysis on All Variables – LDS

197

Table 47: LDS Regression Results – Mode

204
xii

Table 48: LDS Regression Results – Annual Miles

205

Table 49: LDS Regression Results – Vehicle Characteristics

206

Table 50: LDS Respondents split into three Household Income Groups

209

Table 51: LDS Vehicle Characteristics

209

Table 52: LDS Demographic Characteristics

210

Table 53: LDS Mode of Travel Based on Usual Weekly Travel Patterns

210

Table 54: LDS Average Annual Miles to End-Point Destinations by
Trip Purpose Based on Usual Weekly Travel Patterns

211

Table 55: LDS Average Environmental Burdens Based on Usual Weekly
Travel Patterns

212

Table 56: Summary of Travel Behavior by Income Group and
Neighborhood Typology

216

Table 57: Data Validation – Detroit

245

Table 58: Data Validation – Ann Arbor

245

Table 59: Data Validation – Birmingham

245

Table 60: Data Validation – Bloomfield Hills

246

Table 61: Data Validation – West Bloomfield

246

Table 62: Code Book

250

Table 63: Results of Stepwise DA

253

xiii

Table 64: Transformed Variables used in DA

253

Table 65: Original Distribution of Variables – HDU Neighborhoods

254

Table 66: Transformed Distribution of Variables – HDU Neighborhoods

255

Table 67: Original Distribution of Variables – HDS Neighborhoods

256

Table 68: Transformed Distribution of Variables – HDS Neighborhoods

257

Table 69: Original Distribution of Variables – LDS Neighborhoods

258

Table 70: Transformed Distribution of Variables – LDS Neighborhoods

259

xiv

LIST OF FIGURES

Figure 1: The Detroit Region

41

Figure 2: Population Trends by SEMCOG County

43

Figure 3: Percent White Population Trend by SEMCOG County

43

Figure 4: Percent Black Population Trend by SEMCOG County

44

Figure 5: Percent ‘Other’ Population Trend by SEMCOG County

44

Figure 6: Detroit Annexation Map

46

Figure 7: The Detroit Tri-County Area

68

Figure 8: The Detroit-Ann Arbor-Flint CMSA

69

Figure 9: The Detroit Region/SEMCOG Jurisdiction

69

Figure 10: Population densities in the SEMCOG Region

71

Figure 11: Detroit Study Neighborhood Map

76

Figure 12: Detroit Department of Transportation Bus Routes

77

Figure 13: Detroit Neighborhood 1 – Land Use Map

79

Figure 14: Detroit Neighborhood 2 – Land Use Map

80

Figure 15a: Industrial Deterioration in Detroit

81

Figure 15b: A Residential Street in the Detroit Study Neighborhood

81

Figure 16: Ann Arbor Study Neighborhood Map

83

Figure 17: Ann Arbor Transportation Authority Bus Routes

86

Figure 18: Ann Arbor Neighborhood – Land Use Map

88

Figure 19a: Ann Arbor Main Street

89

xv

Figure 19b: Typical view of the Ann Arbor Study Neighborhood

89

Figure 20: Birmingham Study Neighborhood Map

91

Figure 21: SMART Bus Routes Map with Birmingham Inset

93

Figure 22: Birmingham Neighborhood – Land Use Map

95

Figure 23a: Downtown Birmingham

96

Figure 23b: Typical Birmingham Neighborhood

97

Figure 24: Bloomfield Hills Study Neighborhood

98

Figure 25: Bloomfield Hills Neighborhood – Land Use Map

100

Figure 26: SMART Bus Routes Map with Bloomfield Hills Inset

102

Figure 27a: Typical Bloomfield Hills Residential Street

104

Figure 27b: Typical Bloomfield Hills Residential Neighborhood

105

Figure 28: West Bloomfield Study Neighborhood

107

Figure 29: SMART Bus Routes Map with West Bloomfield Inset

110

Figure 30: West Bloomfield Neighborhood – Land Use Map

112

Figure 31a: Typical West Bloomfield Residential Neighborhood

113

Figure 31b: Typical West Bloomfield Residential Street

113

Figure 32: Study Area

122

Figure 33: Canonical Functions Plot

148

Figure 34: Example of an EPA Window Sticker

247

Figure 35: Example of a Fuel Economy Sheet

249

xvi

Chapter I - Introduction
Suburbanization has dominated the primary settlement pattern and preference of many
Americans since the mid twentieth century (Jackson, 1987). The impacts have been researched
and debated upon extensively by scholars and professionals. Initial catalysts of the
decentralization of American cities included the extensive highway system, a decline in real
energy prices, subsidized infrastructure and land use and zoning practices (Baum and Snow,
2007b; Brown, Morris and Taylor, 2009; Harrison, 1974; Kopecky and Suen, 2010; McGrath,
2005; Vojnovic, 2000b). These factors, among others, have encouraged people to move further
away from inner cities to less dense suburban areas, as the more affordable commutes became
desirable and reasonable trade-off to ensuring a large home, a large lot, and privacy. The
dependence on the automobile has been a cause and an effect of suburbanization. The
transportation needs, patterns, and impacts of the developments have been researched under two
general themes: travel patterns and behaviors, and the environmental concerns related to
vehicular emissions (Ahn and Rakha, 2008; Bento et al, 2005; Brownstone and Golob, 2009;
Cervero and Murakami, 2010; Crane and Crepeau, 1998; Ewing, 2001, 2010; Liu and Shen,
2011; Newman and Kenworthy, 1989, 1999; Vojnovic et al, forthcoming).
There have been a plethora of studies on commuting and travel patterns done at different
levels of analysis with a general lack of smaller, individual and neighborhood levels of analyses
to capture truer, more nuanced constraints and travel decisions (Holtzclaw, 2002; Horner and
Murray, 2002; Lin and Long, 2008; Niedzielski, 2006). Existing studies have recognized the
under-representation of research that allow for greater control of built environment
characteristics and socio-economic conditions at the neighborhood scale. The result has been an
increased effort to conduct studies at finer scales of analysis. Also, the attempt to capture both
1

larger analysis levels (like the Census tract or the TAZ) with the smaller levels (like blocks and
block groups) has been undertaken (Antipova et al, 2011). In studies of travel, the use of
accessibility as a measure in neighborhood level analyses for travel behavior was adopted widely
after Hansen’s (1959) seminal piece on the topic (Ewing and Cervero, 2010; Liu and Shen, 2011;
Zolnik, 2011).
There is also a notable absence of research on travel behavior focused on disadvantaged
communities, i.e. neighborhoods experiencing disinvestment and decline. Within the existing
built environment and travel behavior literature, the relationships between socio-economic
conditions, the urban built environment and travel are recognized as underrepresented in research
on urban planning, urban design, and urban and transportation geography (USDHHS, 2000; Day,
2003; 2006; Vojnovic, 2006; Vojnovic et al., 2006).
The importance of devoting more attention to research on the disadvantaged is also
accentuated by a new understanding that the built environment and human behaviors, including
travel, have a separate and distinct set of parameters that are influenced by not only class but also
culture (Vojnovic et al., 2013). This lack of research on disadvantaged communities has led to
the growing calls for more attention to be paid to the condition of lower income populations and
their neighborhoods by both academic researchers and government agencies (Day, 2003; 2006;
USDHHS, 2000; Vojnovic, 2006; Vojnovic et al, 2006).
In fact, it can be argued that while the number of disadvantaged has been on the rise,
public officials, policymakers, and academics have been increasingly disinterested in the
condition of marginalized neighborhoods. The concern for the poor and minority populations
was a critical aspect of urban research and policy interest during the 1960s and 1970s, yet the
focus on the disadvantaged has been diminishing since the 1980s (Slater, 2006; Podagrosi &
2

Vojnovic, 2008; Podagrosi et al., 2011). This is particularly evident in the research on the built
environment, accessibility and travel behavior, an area of study that has almost exclusively
focused on relatively dynamic and robust communities.
A similar under-representation of research is also evident in the literature on emissions.
Emissions inventory and modeling have been addressed under the broader topic of the
transportation industry as a key contributor to greenhouse gasses (GHGs) and climate change
(Brown et al, 2008; Kromer et al, 2010; Lankao et al, 2009; Lutsey and Sperling, 2009;
McCollum and Yang, 2005; Meyer, 2010; Moore et al, 2010; Plotkin, 1999; Schmalensee, 1998;
Wheeler, 2008; Yang et al, 2009). The literature on the socio-economic dimension of gasoline
consumption and pollutant emissions is sorely lacking. Studies on environmental burdens have
been conducted at larger geographic scales, such as cities and metropolitan areas, and have also
even been compared across countries. However, neighborhood level studies to understand the
characteristics of those subgroups most responsible for environmental burdens, in terms of
gasoline consumption and pollutant emissions, from travel within their immediate physical
environment are notably absent. The existing literature has also not explored the relationships
between physical and socio-economic conditions of neighborhoods and environmental burdens
(gasoline consumption and pollutant emissions).
Within the context of the existing literature—and in an attempt to develop a better
understanding of the effects of class on travel behavior, gasoline consumption and pollutant
emissions at the neighborhood scale—this dissertation will focus on the intra-neighborhood
analyses of three neighborhood typologies in the Detroit region. The six research sites are
located within the SEMCOG (SouthEast Michigan Council of Governments) Region. This
Region includes the counties of St. Clair, Livingston, Oakland, Macomb, Washtenaw, Wayne,
3

and Monroe. The analyses will also consider the unique condition of disadvantaged
communities—neighborhoods experiencing decline and disinvestment—with a focus on two
lower eastside Detroit neighborhoods. The exploration of travel patterns, and resulting
environmental burdens, including in communities experiencing severe disinvestment and
decline, is a topic that has not been explored in the existing literature.
The research involves three neighborhood typologies: 1) higher density urban (HDU); 2)
higher density suburban (HDS), and 3) lower density suburban (LDS). The higher density
suburban and lower density suburban neighborhoods maintain similar socio economic
characteristics, but they have very different urban built environments. These neighborhood
typologies tend to be the traditional focus of travel behavior studies, exploring differences in
travel between higher density, mixed land use, and highly connected neighborhoods and low
density, generally single use, and disconnected neighborhoods. Beyond the built environment
differences, these neighborhoods tend to be relatively robust, with regard to socio-economic
composition.
The higher density urban neighborhoods are similar to the higher density suburban
neighborhoods, in terms of urban form, but the socio economic characteristics of the urban
neighborhoods are very different, and unique to travel behavior research. The Detroit urban
neighborhoods are experiencing severe disinvestment and decline, despite the fact that they offer
relatively high densities, a mix of land uses, and highly connected street networks.
Using built environment objective data and primary travel survey data, this research will
focus on examining the impacts of socio-economic variables on travel, gasoline consumption,
and pollutant emissions, specifically carbon monoxide emissions. The inclusion of the Detroit

4

urban neighborhoods will also enable an assessment into the nature of travel behavior in
neighborhoods characterized by urban disinvestment and decline, yet neighborhoods that still
maintain higher urban densities, a concentrated land use mix, and a high degree of street
connectivity. As already noted, within the existing travel behavior and urban form literature, this
relationship between the built environment, class variables, and travel behavior is recognized as
being underrepresented, and particularly in the context of exploring environmental burdens, here
specifically gasoline consumption and pollutant emissions.

The Research Question
The primary research question then, is “Are there significant impacts of individual socioeconomic characteristics on travel behavior and environmental burdens?” The main subquestion that would help answer the primary research question is: What socio-economic
characteristics have a significant effect on travel patterns, gasoline consumption and vehicular
emissions when holding the built environment constant?

Data Sources
Primary data, collected through mail-out surveys to six neighborhoods, has been used for
this study. Secondary data from the U.S. Environmental Protection Agency (EPA) are also used
to supplement the survey data in the analysis. These neighborhoods include two in inner city
Detroit, and one each in the suburban areas of Ann Arbor, Bloomfield Hills, West Bloomfield
and Birmingham. These neighborhoods were chosen to represent a mix of built environment,
social, and economic factors.

5

Along with the survey data, emission estimates from the Environmental Protection
Agency (EPA) have been used to get to best case scenarios for emissions specific to each vehicle
owned by respondents to the survey. Key categories that the collected data variables would fall
into include Demographic (age, sex, educational attainment, income), Travel Behavior (mode of
travel, frequency of travel, and exact origin and destinations for travel to work, school, and
optional purposes), and Vehicle Characteristics (make, model, year, miles driven, emissions, fuel
economy).

Method of Analysis
Discriminant Analysis will help in identifying the factors that are significant in
discriminating between the three neighborhood typologies: Higher Density Urban (HDU),
Higher Density Suburban (HDS) and Lower Density Suburban (LDS). Cluster Analysis would
help understand the groups that the data would form into within each typology. Lastly, using
regression analysis, factors significantly affecting gasoline consumption and emissions will be
derived across three neighborhood typologies. All methodologies are discussed in greater detail
in Chapter V.

General Hypothesis
It is hypothesized that higher income residents will consume higher amounts of gasoline
due to their larger vehicles and stronger engine horsepower. Even though they are expected to
have newer vehicles and thus better fuel economy, their reliance on the private vehicle for

6

transportation and the distance that they travel in larger vehicles would overpower the positive
impacts of the higher fuel economy and result in higher amounts of gasoline consumption.
Lower income residents on the other hand are more dependent on public transportation and
walking and so will consume less gasoline and emit fewer emissions. When considering only the
built environment (density, land use mix and connectivity), low density suburban residents are
expected to consume more gasoline due to the longer commutes from their suburban homes to all
trip destinations, while higher density suburban residents and inner city residents are expected to
consume less gasoline due to the high connectivity, density and land use mix of their built
environment.

Main Contribution and Implications
In essence, this study will analyze the nature of environmental burdens across the six
neighborhoods in the Detroit metropolitan region. These results have policy implications for
planners and transportation professionals. This is especially important at a time when the State is
looking at three legislations that either propose to increase gas taxes or State sales taxes to help
pay for road maintenance and repair (Eggert, 2013). Also, this study would be useful to the issue
of the Gas Guzzler Tax (1978) that is a tax on manufacturers for those vehicles sold that do not
meet minimum fuel efficiency standards. This tax however, applies only to light duty vehicles
and exempts SUVs, minivans, and pick-up trucks. Further, this study would lend more detail
and uncover nuances inherent in research on travel behavior and emissions modeling. As such
the research on this topic lacks insight into the socio-economic impacts on environmental
burdens.

7

This dissertation is divided into seven chapters. The next chapter (Chapter II) is the
literature review, which will discuss research done on suburbanization to first get at the causes
and effects of the different urban built environments. Then travel patterns will be discussed to
get at research done on vehicle miles traveled (VMT) and travel behavior of people. Lastly, the
literature review will discuss emissions in an effort to show the disconnect between a physical
phenomenon (emissions/pollutants) and its very anthropogenic source.
Chapter III is a brief overview of Detroit and journey through its growth years and the
subsequent decline. It is important to know where this City has been and what were the factors
in its decline to truly understand its uniqueness and the difference from other large declining
cities. Chapter IV goes into the details of the neighborhoods included in this study. The
reasoning behind the various neighborhoods selected for this study and an objective assessment
of the physical character of these neighborhoods is laid out to help the reader get a complete look
at built environment characteristics.
Chapter V is on the Research Hypotheses, Data and Methodology. This chapter
discusses the data gathered, including the primary and secondary data used in the study. This
study uses a database that is very detailed, both in scale and scope, which is a big part of the
unique contribution this study offers to the existing literature. The chapter also lays out the
research hypotheses to be tested and the analyses conducted. Apart from the usual regressions
used to test the hypotheses, the study has used a blend of quantitative and qualitative assessments
to tease out the differences in travel patterns and environmental burdens.
The sixth chapter discusses the results of the analyses. Along with the results of the
hypotheses tests conducted, I present a complete picture of the differences in travel behavior and

8

environmental burdens across the neighborhoods and within similar neighborhood types.
Following the quantitative analyses results, nuances are extracted with the use of qualitative
assessments that drive the main argument of this study further. This chapter ends with a
concluding commentary on all the results and what they mean when we look at the whole issue
of the socio economics and the built environment impacts of travel behavior and environmental
burdens in a combined manner. The seventh chapter forms the conclusion of the entire
dissertation, ties everything together and presents a complete package – a holistic view of the
travel behavior and environmental burdens research conducted for this dissertation.

9

Chapter II - Literature Review
Suburbanization
Suburbanization, best understood as a process of residential and commercial dispersion
from the central city to outer, fringe areas, has been researched for decades. It is so engrained in
the American culture and way of life that it has been labeled as the “manifestation of the
fundamental characteristic of American society” (Jackson, 1987: Pg 4). Shay and Khattak
(2012) point out that while the urban population has grown by 80% between 1960 and 2000, the
urbanized land area has grown by 130%. While studies date the beginnings of suburbanization
th

to the 18 century (Lewis, 2002; Walker & Lewis, 2001), most of the literature recognizes
increasing levels of suburbanization as occurring from the mid 1900s and onwards, a result of
increased economic prosperity and rapid population growth. For instance, the 1950s were
marked by mortgage issuance policy initiatives that encouraged the purchase of large singlefamily homes in the suburbs. This decade was also marked by the construction of the Interstate
highway system that further facilitated the rapid move outward of the urban population.
Most of the studies also attribute suburbanization to be an effect of different
transportation modes. For example, Warner (2004) shows how the streetcar led to
suburbanization in Boston in the late 1800s. Similarly, the streetcar system had been a
significant force in the growth of the twin cities of Minneapolis and St Paul (Adams and
VanDrasek, 1993). Thus, even before the private automobile, other transportation modes were
facilitating early American urban decentralization (Xie and Levinson, 2010).
Then came the Interstate Highway System. Baum-Snow (2007b) studied the effects of
highway construction on central city and Metropolitan Statistical Area population density for all
10

large MSAs in the country between 1950 and 1990. He estimated that with every addition of a
highway ray, i.e. a highway that connects the CBD to a region outside, population density falls
by about 1%. In terms of the population itself, each additional ray of highway causes a 9%
decline in central city population. Baum-Snow also estimated that highway construction
accounts for a third of the population density difference between the metro area as a whole and
the central cities between 1950 and 1990. Had the interstate highway system not been built, the
central city population, between the years 1950 and 1990, would have increased by 8%, instead
of declining by 17% (Baum-Snow, 2007b).
Urban highways that existed before the Interstate highway system were designed to flow
with the urban fabric and increase the speed of vehicles (transit and private auto) with the least
possible disruption. With the building of the Interstate Highway system, and the shift of
transportation control and funding from local to state and federal agencies, the focus on the
fabric of the neighborhoods lost priority and the primary focus became simply speed, as that was
the sign of being progressive (Brown, Morris, and Taylor, 2009).
Kopecky and Suen (2010) assert that rising incomes and falling automobile prices were
the key drivers of suburbanization. Furthermore, into the second half of the 20th century, city
centers became increasingly marked by crime, lack of educational opportunities, and high
poverty rates, all of which became a strong ‘push” factor for residents that could afford the
suburbs (Mieszkowski & Mills, 1993). This, in turn, overburdened the central city capacity to
provide public services, leading to higher taxes and lower quality of services, further facilitating
suburbanization.

11

While there are various causes that lead to suburbanization, government policies are seen
as an important variable driving decentralization. These policies relate to building transportation
network infrastructure, allowing fragmentation of metropolitan areas, and handing over political
control to local governments that are under pressure from their majority constituents to keep
certain ethnic or income groups out of their jurisdictions. The under-pricing of infrastructure by
local governments has led to an oversupply and overconsumption of these services by residents
(Vojnovic, 2000b). Cities have continually subsidized services such as road infrastructure, water
and sewer, and residents have not had to pay the true costs of these amenities.
Overconsumption, in turn, has led to further decentralization as people pay less to move to
further places and also purchase houses on large lots. On a similar note, McGrath (2005) asserts
that had the costs of private transportation been more reflective of actual costs, U.S. metro areas
would not be that spread out; the higher costs would have resulted in more compact urban
regions.
Harrison (1974) maintains that institutional inflexibility furthers the financial difficulties
of central cities. Failure of policy is the main cause for the condition that inner cities are in now
and most attention needs to be given to these policies if any remedy is to be sought. Similarly,
Adams et al (1996) show that a strong central city benefits the metro area as a whole and policies
and institutional arrangements (such as tax base sharing, metro-wide economic development, and
disbursement of low income families through mixed income neighborhoods) should be in place
to discourage central city decline.
In essence, various theories and hypotheses explaining growth in central cities and their
suburbs have evolved, ranging from “push” to “pull” factors. The “push” factor suggests that
growth in suburbs is the result of the flight of the population from central cities to avoid crime,
12

congestion, taxes, poverty and social and economic problems (Mieszkowski and Mills 1993).
The “pull” factor suggests that the search for better amenities, larger housing, and a better quality
of life by residents, and the search for a larger and more affluent customer base by businesses,
drives the growth in suburban locations compared to central cities (Leichenko, 2001). Whatever
the actual impetus may be for particular families, households, or businesses to suburbanize, the
one clear pattern in the process is that urban decentralization has been rapidly increasing.

Effects of Suburbanization
With the decentralization of cities, new social costs emerged across the American urban
landscape. One social cost became evident with the loss of community and street life (Jacobs,
1961). With suburbanization, increasing segregation also crystallized (Darden & Kamel, 2000).
Residential segregation by race is persistent in many urban areas that have been vacated by the
white, moderate income population that moved into suburbs. Most often, the African American
population is the one that has been associated with being segregated, that is, living in enclaves in
the inner city with a majority percentage of residents of the same race. Segregation affects a
variety of outcomes that deal with health, education, income, and occupation. Ananat (2011)
refers to segregation as being the main cause of economic inequality. She finds that there is
greater inequality and higher poverty rates within the black, segregated population, while lower
inequality and poverty rates exist among the white population.
Kain’s (1968) seminal piece on spatial mismatch has received generous attention in the
academic literature on the topic. Together with the Kerner Commission’s Report (1968), these
works stress the critical nature of accessibility to jobs, especially for the African American
13

population. The general explanation for the spatial mismatch theory is that African-American
employment is lower in suburbs because the job location is far from their residential location,
which is usually in the inner cities. More importantly, many African-Americans do not have
automobiles and there is a lack of robust public transit systems connecting the urban core to the
suburbs, which results in higher costs (monetary and/or temporal) for them to reach jobs in the
suburbs. The Kerner Commission’s Report on the riots in Los Angeles in the 1960s stressed the
access problems for the African American residents and the resultant unemployment as being the
main underlying causes for the rioting. Cohn and Fosselt (1998) put forth a further dimension to
the issue of spatial “mismatch” by proposing that minority urban residents hesitate to take up
jobs in areas where they perceive discrimination against them, the suburban jurisdictions.
While numerous dimensions of inequality have emerged from resulting suburbanization
patterns, education has been a particular point of concern. Baum-Snow and Lutz (2011)
conducted a study of 92 large urban school districts and found that whites made up a
disproportionate section of movers, as the authors show that the white population declined by 13
percent and black population increased by 54 percent between 1960 and 1990. With the exodus
from the inner city, came deteriorating properties, lower taxes, and lower budgets for urban
school jurisdiction. Card and Krueger (1996) assert that an increase in residential segregation
increases beneficial school resources for the white community, while lowering those for the
black community. The residential move to suburban neighborhoods, along with the move of
upper class white students to suburban schools, has left inner city schools with a racehomogenous student population (African-American students) and funded by a class-homogenous
resident population (lower-income African-American urban residents). In essence, school

14

segregation has been an important aspect of suburbanization that has had a substantial effect on
the racial fabric of central cities.
In his book on Urban Economic Development, Harrison (1974) shows that about four
thousand to nine thousand jobs were in suburban Detroit in 1952, which African-Americans were
unable to take up due to residential segregation. These suburbanizing jobs required low skills
while those left in the central city required higher skills. Thus African-Americans lost out on the
jobs left in the central city to the suburbanites who were competing for and were better suited to
those central city jobs. The author also talks about the “exploitation hypothesis,” which states
that the benefits that a central city gives its non-resident population are borne by taxes paid by
the resident population. Workers coming in to the central city use parks, roads, and
entertainment venues which all need increased public services. The costs of these services are
borne by the poor residents who live in the region alone. Even though some have argued that the
problem with central city’s poverty rates is one of the main forces behind the flight of the white
population to the suburbs, and not race per se, Harrison believes that there is something beyond
financial resources and economic systems in force in these decentralization patterns.
Recent research has shown that segregation is decreasing, even though it is a slow
process (Clark and Blue 2004; Frey 2000; Clark 2007). The main point is that with increasing
incomes, the Black population has been able to move to the suburbs and live in integrated
neighborhoods. Blacks in central cities are still segregated from whites (Clark and Blue, 2004).
Also, the influx of minorities as immigrants, who choose to move directly to suburbs to settle,
has been influencing segregation in U.S. metropolitan areas. Although Clark (2007) shows this
apparent integration as being disproportionately visible in the suburbs, and less so in central
cities, Fischer (2008) shows that Black segregation has declined mainly in the inner cities (at
15

least for the Northeast and Midwest Regions). In addition, Fischer stresses that between 1980
and 2000, increases in Black suburbanization has been met with segregation patterns within the
suburbs. In those areas where a large portion of Blacks moved to suburban locations, the
threshold of tolerance of whites already living in such locations was breached; whites started
moving out, the neighborhood became increasingly black, and this resulted in segregated
neighborhoods. As long as Blacks remained the minority and did not tip the neighborhood
balance, their movement into suburban neighborhoods was welcomed, or perhaps more
appropriately, tolerated.
Through the long standing history of research and inquiry on suburbanization, it is clear
that government policies have been at the crux of the urban decentralization experienced by
American cities. The result of these urban development processes have been evident in intended
and unintended consequences. While quality of life factors, personal choices, and prosperity
have been achieved for those who could afford suburban lifestyles, many argue that segregation,
blight, economic drain and environmental degradation have also been resulting outcomes that
have produced extensive costs to American urban regions. Moving on from this, we can now
look at the various and contrasting environments, patterns and behavioral elements that get
adopted by residents living in different areas: inner-city and suburban, and how these impact
travel patterns and behavior.

16

Travel Patterns/Behavior
Over the last several decades, there have been numerous studies focusing on the various
dimensions of travel patterns and vehicle miles traveled (henceforth referred to as VMT) within
North American cities. In essence, studies try to determine what factors impact VMT and what
policy implications should stem from such impacts. Some studies focus on built environment
designs and conditions that influence travel patterns (Bento et al, 2005; Brownstone & Golob,
2009; Vojnovic, 2006; Cervero & Murakami, 2010; Dunphy & Fisher, 1996; Kockelman, 1996;
Schimek, 1996; Gordon & Richardson; Newman & Kenworthy, 1989, 1999; Cervero & Radisch,
1996; Krizek, 2003; Lin & Long, 2008) while others attribute varying travel patterns to personal
and household characteristics (Weber & Kwan, 2003; Crane & Crepeau, 1998; Rodriguez et al,
2006; Vojnovic et al, 2013; forthcoming). Some of the studies have analyzed vehicle ownership
and travel patterns of residents (Goetzke and Weinberger, 2012; Holtzclaw, 2002; Shay and
Khattak, 2012; Taylor and Mauch, 1998) while others have focused on emissions modeling
(Lindsey et al, 2007; McCollum and Yang, 2009; Stone et al, 2007). The following section of
this chapter will explore the nature of relationships established between travel behavior, gasoline
consumption, pollutant emissions, socio-economic conditions, and the urban built environment.
Another set of studies focus on issues of inequality between residents of different density
areas, different races, built up levels of areas (urban/suburban) and distance and time spent in
travel for a typical resident (Pucher & Renne, 2003; Blumenberg, 2004; Sultana, 2002).
Research on factors that impact vehicle ownership and miles of travel has typically dealt with
two broad, overlapping aspects, neighborhood characteristics and personal/household
socioeconomic characteristics.

17

Built Environment and Travel Demand
Neighborhood physical and design characteristics and their impact on vehicle ownership
and VMT have been an important aspect of travel behavior research. Within this line of
research, density as a built environment characteristic, and its impact on travel, has been the
focus of many works (Bhat and Guo, 2007; Cao et al, 2007; Holtzclaw et al, 2002; Kahn, 2006;
Su, 2010; Cervero and Murakami, 2010). The research shows that the greater the density, the
less vehicles owned and fewer miles traveled by private automobile, as accessibility to different
destinations is improved, and public transit systems (if present) would carry much of the travel
load. In addition, an increased ownership of light duty trucks has been associated with a higher
number of vehicles per household, a higher number of household members, and with suburban,
low density, residential neighborhoods (Potoglou 2008; Cao et al 2006; Zhao and Kockelman
2000). Bento et al (2005) used the 1990 NPTS data and studied the impact of urban form and
transit availability on Vehicle Miles Traveled (VMT) and commute mode choice. They find that
urban sprawl measures and public transit supply have marginal effects on travel demand,
however, taken together (i.e. simultaneously), these measures have a considerable impact on
travel. In essence, they caution against policy measures that only focus on one variable (like
density) to influence VMT. There are a varied number of measures of urban form (like
population and employment density, road density, jobs-housing balance, streetscape texture,
population centrality and availability of public transit) that work together to affect t ravel
behavior. Policies should be cognizant of this complexity in the relationship between urban built
environment characteristics and travel.
This line of thought has been the bases for many studies that look at commute times and
distances as an effect of suburbanization. The general idea is that commute times and distances
18

th

have increased in the late 20 century as a result of the sprawling patterns of residential
locations (Rossetti and Eversole, 1993; McCuckin and Srinivasan, 2003; Reschovsky, 2004).
The increasing VMT is related to increasing local and global pollution, energy consumption,
open space consumption and that of other scarce resources (Cervero & Murakami, 2010).
Curbing this trend, compact development could reduce vehicle miles traveled in the next four
decades by about 17% (Bartholomew and Ewing, 2009). Shay and Khattak (2012) find that
characteristics of neighborhoods that encourage walkability and non-motorized forms of travel
translate into fewer cars per household and a greater propensity to use those non-motorized
forms of travel. However, these neighborhood characteristics would entail a greater number of
trips due to the shorter trip lengths.
Other studies have found more ambiguous results between the association of
suburbanization and auto ownership, commutes times and commute distances. For example,
Kain (1967) implied that the relationship between density and car ownership is such that both
influence each other and that there is no unidirectional influence. Kahn (2006) found that low
density suburbs had longer commute distances to work but lower commute times, due to the
speed of travel, when compared to higher density suburbs. Ewing (2002; 2003a) found that there
was no clear indication that commute times had increased for suburban residents He attributed
this to the fact that although an imbalance between jobs and housing increased travel distances to
work, the decentralization process reduced commute times in the sense that highway travel was
faster and people could cover longer distances in less time.
Zolnik (2011), on the other hand, found that higher residential density, the degree of
centering and accessibility (street system) had a small, but significant negative impact on

19

commuting to work distances and times. Similarly, Liu and Shen (2011) find that amongst the
urban form characteristics, accessibility is the only characteristic that has a significant negative
impact on annual household VMT. Holtzclaw (2002), in a study that modeled vehicles per
household and VMT per household as they would relate to residential densities, concluded that
increased densities were associated with lower VMT.
The complex relationships between different neighborhood characteristics and travel
demand complicate the issue further. For instance, Cervero and Murakami (2010) studied some
370 urbanized areas throughout the country to determine the effects of the built environment on
VMT. They find that although there is a direct, strong, negative association between population
density and VMT per capita of -.604 (indicating that a doubling of population density would
reduce VMT by 60%), there are also other positive, indirect associations. Road density,
urbanized area size and retail accessibility are three factors that yield in increases in VMT. Their
combined positive association reduces the net negative association between population density
and VMT per capita to -.381, indicating that a doubling of population density would reduce
VMT by 38%.
Similarly Su (2010) studies 85 urban areas over a 20 year period (from 1982 to 2003)
across the nation and finds that road density and urban spatial size have a positive and significant
impact on travel demand, while population density and congestion have a negative and
significant impact on travel demand. Vojnovic et al (2006) studied urban form and public health
in Michigan and stressed that even though increased densities will not ensure that people will go
shopping at the nearest store or eat out at the nearest restaurant, the increased densities will
shorten distances between destinations and give residents the “choice” to use non-motorized
modes of travel, and over the long run, relieve the dependence on the automobile. In 2010,
20

Ewing and Cervero did a review of previous research on the built environment and travel
patterns and found that VMT was most strongly associated with destination accessibility,
followed by distance to downtown (which they associated with density and accessibility), and
then design characteristics; with the three maintain the strongest association with VMT.
In sum, a majority of the existing research agrees that the built environment does have an
influence on travel patterns. Density has a positive association with number of trips, which have
a greater likelihood to be by non-motorized means, but a negative association with the distance
of travel, vehicle ownership, and size of vehicle. Similarly, accessibility and land-use mix have
been shown to have a positive association with alternatives (including non-motorized) travel
modes and a negative association with travel distances (VMT).

Socioeconomic Characteristics and Travel Behavior
Socio economic characteristics relate to travel behavior based on cultural and social
norms and class factors. Researchers have identified endogenous and contextual effects (termed
as “social spillover” effects) on auto ownership. These effects translate into the notion that an
individual’s actions and behavior change with the attitudes and behaviors of his/her peers
(Goetzke and Weinberger, 2012; Manski 2000). Consequently policies aiming to reduce VMT
through car ownership should recognize that there is a self-reinforcing social multiplier that
comes into play when goals for reduced car ownership are formed (Chen et al, 2008; Weinberger
and Goetzke, 2010; Weinberger et al, 2009; Shoup, 2005).
While neighborhood physical characteristics are important, some researchers believe that
pushing for higher densities would have a negligible impact on VMT due to a higher influence of
21

individual driver characteristics (Brownstone & Golob, 2009). Weber & Kwan (2003) have
worked on multiple levels to gain an understanding and differentiate between
neighborhood/location effects and personal/household characteristics when analyzing timeaccessibility in Portland, Oregon. They conclude that it is not the location characteristic but the
individual characteristic that has an influence on travel and accessibility, and that further studies
on the topic should approach it from an individualistic perspective. Similar findings were
presented by Crane & Crepeau (1998) and Rodriguez et al (2006).
The sensitivity of travel patterns have also been linked to socio-economic traits in more
recent research. Paulley et al (2006) have shown that increasing incomes are significantly
related to increased auto ownership and VMT. Ewing and Cervero (2001) find that some aspects
of travel, such as trip generation rates, are more sensitive to socio economic characteristics,
while mode choice is affected more by the local built environment.
Several researchers have also shown an increasing effect of socio demographic
characteristics of the minority and disadvantaged population subgroups (low income, women and
the elderly) on mode choice and VMT (Helling, 2000; McDonald, 2005; Cao et al, 2010; Kim
and Ulfarsson, 2004; Matthies et al, 2002). As different scales give different results to research
on travel behavior (Vojnovic, 2006; Sultana, 2002), different ethnic and population groups
would also have different impacts on VMT (Pucher & Renne, 2003; Blumenberg, 2004). These
studies suggest that the marginalized populations (females, racial minorities, poor) depend on
travel via private automobile to a great extent and largely because of limited alternative travel
options. Even though this may be a burden on their finances, the lack of effective public transit
leaves lower-income sub-groups with no other choice than to drive to be able to reach their
destinations, including employment. Running daily errands and necessity shopping can be
22

handled with ease when driving around in a private vehicle. Trip chaining, a concept that
suggests that people might get to multiple destinations in one trip (for example, going to the
grocery store, drycleaners and the post office on your way home from work) is favored by
women travelers, especially as this seems a more practical use of time and resources. However,
trip chaining also reduces the attractiveness of alternative transportation modes to the private
automobile, and especially public transit (Taylor and Mauch, 1998).
The subject of ‘self-selection’ has been explored in many studies, however, some
researchers feel that self-selection is manifest in the socio-demographic characteristics of the
residents and when these are taken into control, self-selection is also controlled (Shay and
Khattak, 2012). The caveat being that the built environment of these neighborhoods allows for
that choice (self-selection) to be exercised and similar to controlling for socio-demographic
characteristics, controlling for the built environment deals effectively with this issue.
Glaeser et al (2008) contend that the poor live in inner cities mainly because they need to
be closer to public transportation. Public transit plays a major role in the initial tendency of the
poor to locate in inner cities as they need to be closer to work destinations. Once they settle in
these inner city areas, problems due to the concentration of the poor in small areas exacerbate
other social and financial problems that prove to be the impetus for others to move out. The
main point in their national study of MSAs being that the housing market alone is not responsible
for the sorting of population by class in the suburbanization process, but that transportation mode
choice plays a critical role in explaining income sorting (Glaeser et al, 2008). On a similar note,
Vojnovic et al (2007), in their study on the built environment and its effects on physical activity
and health impacts state that eradicating racial and class discrimination is critical to the success
of policies that encourage compact development. In turn, the compact developments help in
23

increasing physical activity and can promote positive health outcomes by reducing the
dependence on the automobile.
Konduri et al (2011) in their study on the relationship between vehicle type choice and
trip length, found that the former influenced the latter. They found that the use of vans as the
mode of travel was associated with longer distances, while the use of a SUV was associated with
shorter distances, even though the SUV seemed to be the first choice for a household with more
than one vehicle in its fleet. The car, as the vehicle of choice, was used for longer trips than trips
by SUV or pickup truck, but slightly shorter trips than those undertaken with a van. The authors
caution against policies that have the intent of encouraging the use of smaller vehicles
(specifically the car), as people exercise trade-offs in their use of vehicles and might use their
smaller vehicles for longer distances, which would negate the positive effect of the policy.
Although not specifically related to any socio-economic characteristics, this is a derivative of it
as it is related to actions and choices by people about the types of vehicles they buy and travel in
for different trips.
Liu and Shen (2011) also examine the impacts of urban form and socio economic
characteristics in their study on VMT, energy consumption and carbon dioxide emissions. They
find that although urban form is only indirectly associated with VMT, socio-economic
characteristics, such as number of vehicles owned, income, gender, race age and family structure
have significant impacts on VMT and energy consumption.

24

Built environment and socioeconomic factors and travel behavior
Incorporating the built environment and socio economic characteristics, Vojnovic et al
(2013) study the impacts on travel behavior and health of residents in six Lansing area,
Michigan, neighborhoods. Their results show that neighborhoods in urban and inner suburban
areas had a greater mix of land uses and higher connectivity characteristics. Residents of outer
suburban areas were far more reliant on the automobile than those in inner suburban and urban
neighborhoods. Even so, socio economic conditions dictated the extent to which residents would
use non-motorized/motorized modes of travel. The lack of stores with healthy food options in
the declining urban cores and the proliferation of these in the outer, wealthy suburban areas made
average distances to these stores shorter for higher-income suburban residents and longer for
lower-income urban residents, indicating that neighborhood investment was a critical factor in
deriving distances to destinations and affecting travel patterns. The built environment
characteristics facilitated walkability and lower levels of reliance on the automobile, however,
socio economic factors turned out to be the main dictator of travel behavior.
A review of research on the various associations with VMT by Ewing and Cervero in
2001 showed that although mode of travel choice depended on both the built environment and
socioeconomic characteristics, trip frequency was primarily associated with socioeconomic
characteristics and secondarily with the built environment. In contrast, trip length was primarily
associated with the built environment and secondarily with socioeconomic characteristics. This
is understandable, as areas that are denser, more accessible and have mixed uses, would require
shorter trip lengths to get to final destination points. However, due to the short trip
characteristic, people are willing to get out more often to get to their destinations, especially

25

since those destinations are closer, resulting in a larger number of trips. These shorter trips,
however, have a greater likelihood of being by non-motorized means.
Grengs (2010) takes the notion of spatial mismatch further, agreeing with researchers
such as Blumenberg and Manville (2004), and shows that it is not the case of distance between
job availability and prospective employees’ residence, but that the lack of a private vehicle to get
to the available jobs is what is hindering greater employment levels for poor African Americans,
at least in the Detroit metropolitan context. In his study, he used data from the 2000 Census
Transportation Planning Package (CTPP), 2000 Census, and data from SEMCOG. He shows
that the inner city of Detroit is where accessibility to low-wage jobs is the poorest, compared to
the rest of the metropolitan region. That said, there are considerable variations in the different
neighborhoods in the inner city that relate to accessibility, indicating that not all neighborhoods
face the problem of spatial mismatch.
However, on closer investigation, he shows that since Detroit’s public transit is so
inferior compared to other large metropolitan regions, that accessibility is more a function of
owning a private vehicle than that of modal choice: residents in Detroit simply do not have
enough of choice in selecting travel mode. His suggestion is to increase access to car ownership
for poor inner city residents, stressing the fact that the benefits from automobile ownership
would far outweigh the burdens of pollution, congestion, costs of ownership and driving, and the
time and costs involved in strengthening the existing public transit system. What is missing in
this study is the nature of actual vehicle owners (their socio economic characteristics), their
actual travel behavior, and resulting emissions estimations.

26

Vehicle emissions and travel behavior
There is an added dimension of travel behavior research that has focused on
environmental burdens, such as gasoline consumption and/or pollutant emissions. Automobile
dependent cities in the US consume excessive amounts of energy, much more so than cities in
other countries (Schiller et al, 2010). Schiller and colleagues reveal that “an average American
city of 400,000 inhabitants uses as much energy for private passenger transportation as an
average Chinese city of 10 million people” (Pg. 8). Although residents make their travel choices
taking congestion, time required and routes into consideration, rarely do they consider the
environmental and/or energy requirements into this decision-making process.
Vehicle emissions are often researched as Greenhouse Gas (GHG) emissions in the
transportation industry. The transportation industry is the fastest growing source of GHGs
(Fitzgerald, 2010: Pg 145). Meyer (2010) states that transportation emissions account for over
28% of the overall emissions in the country. In the most basic and generic sense, emissions from
vehicles are a factor of speed, acceleration, road grade, vehicle weight, rolling resistance and
aerodynamic drag (Boriboonsomsin and Barth 2009). While some of these emission factors fall
under the ‘driving condition’ aspects, some of them also fall under the ‘vehicle technology’
aspects. Both dimensions of emissions are discussed in more detail below.

Driving Conditions
McCollum & Yang (2009) investigate the ways in which we can reduce total GHG
emissions in the long term, into 2050. They are of the opinion that the transportation sector has
the most potential for emission reductions and that further emphasis should be placed on utilizing
27

the many options that are available to reduce this sector’s emissions. A study done by Ahn and
Rakha (2008) used three emission models: 2 microscopic and 1 macroscopic. The microscopic
models are supposed to be better as they use more detailed and have realistic data, as far as
driving conditions are concerned. The microscopic models show that travel on arterial roads
resulted in lower emissions and energy requirements compared to travel on highways, even if
highway travel resulted in time savings. On average, energy cost savings of 18%-23% resulted
with a travel time increase of about 4.3 minutes.
Lankao (2009) studied the impact that cities and urban transportation have on the
atmosphere. They found that population size, density, and GDP per capita were the main
influences on emissions. Cities, due to their density and land use characteristics, have 70%
fewer emissions than suburbs (Fitzgerald 2010). Similarly, Stone (2007) looks at the land use
patterns, specifically sprawl, and air quality in major cities in the U.S. He shows that land use
has a direct relationship with air quality (measured through the ozone layer) as well as an indirect
relationship through transportation, industry and power generation facilities.
Lindsey et al (2011) add the further dimension of topography to their analysis. They find
that emissions are not strictly a linear function of VMT. Usually with uphill and downhill
sections, the relative fuel consumption increases and decreases balance out. In their study on the
effect of road grades on fuel consumption and emissions for light duty vehicles, they find that
relatively flat routes increase the overall fuel efficiency by 15% to 20% over the uphill and
downhill routes. Therefore, a longer route that is more “flat” may be preferable over hilly routes
for the purposes of fuel consumption and emissions (provided the length increase is not more
than 15% over that of the hilly route).

28

Vehicle Technology
There is also a new technology adaption dimension to the study of VMT, gasoline
consumption and pollutant emissions. Plotkin (1999) discusses various technological
advancements that can effectively lower GHG emissions from light duty vehicles. He is of the
opinion that there are ways to improve emissions, but due to the low cost of fuel, these are likely
not to have a great impact. Sperling and Gordon (2008) note that “if vehicle performance and
size had been frozen in the early 1980s, current vehicles in the U.S. would consume 25% to 30%
less fuel” (pg. 67). They talk about the changes in car technology and how we came to the point
at where we are now. They assert that even though vehicle technology has improved over the
years, this improvement has been offset by changing vehicle characteristics. Vehicles between
1980 and 2005 have increased in size, power, and energy-consuming features, as evident in the
extensive adoption of four-wheel drive cars and trucks. The increase in size is evident in the
increase in ownership of SUVs, minivans and pickup trucks. The increase in power is evident in
the increase of 6 and 8 cylinder and turbo engines. And energy consuming features, such as four
wheel drives, require more power and are heavier, and that means lower fuel economy (Sperling
and Gordon, 2008).
In their study of the Chicago metropolitan area, Sperling and Gordon also find that the
type of vehicle used in travel is important in determining emissions. Even though a majority of
the vehicle fleet in the metro area was of the moderate emissions type, the minority of lowefficiency vehicles in the suburban areas—where most of the long distance travel was generated
and undertaken—increased the overall emissions. Lutsey & Sperling (2009) have come up with

29

a “supply curve” that shows the effects of various technological and fuel efficiencies on
transportation emissions. They find that in order to achieve the climate reduction goal of 80%
reduction in climate emissions by 2050, a substantial reduction in travel demand, advanced fuel
technologies (decarbonized fuels) and vehicle technologies (minimal GHG emissions) will be
needed by all sectors associated with the industry.

Costs associated with travel and emissions
Another line of research has focused on the costs associated with driving and related
emissions. Morrow et al (2010) study the effects of various policy scenarios in reducing
emissions. They are of the opinion that increasing the cost of driving is essential to the change in
travel patterns, vehicle purchases, residential location and the main way to reduce transportation
emissions. Similarly Carroll-Larson and Caplan (2009) estimate that a tax on the annual VMT
would encourage people to reduce their travel. A tax of $0.006 and $0.02 per mile for passenger
cars and light-duty trucks respectively, would reduce emissions of Particulate Matter (PM2.5) by
12% and 23%. Thinking of costs, Mashayekh et al (2011) did a study on the total costs of
driving and emissions in 86 US Metropolitan Areas. The total costs included external costs of
air pollution from personal vehicles in normal and congested driving conditions in these urban
areas and this estimate amounted to $53 billion annually. Factors such as ownership,
maintenance, fuel, insurance and depreciation were included in total costs of vehicle use.
External costs would be represented by factors such as health impacts and damages to the
environment.

30

Similarly, the National Research Council estimated in 2007 that the health costs of
emissions averaged to about 1.3 to 1.4 cents per VMT for gasoline vehicles. The pollution costs
form a small fraction of 2.5% of total driving costs, yet when aggregated, they amount to a
substantial figure. When adding in congestion, the costs are even higher. The 2009 Urban
Mobility Report by the Texas Transportation Institute (TTI) estimated that costs of congestion in
the U.S. in 2007 amounted to about $87.2 billion, which resulted in a time delay of 4.2 billion
hours and extra fuel usage of 2.8 billion gallons. Mashayekh et al’s (2011) results indicate that
there was a total external emissions cost of $145 million per day, averaging to about $1.7 million
per day per urban area in the US, further resulting in a cost of 64 cents per person per day or 3
cents per VMT. Part of this total external emissions cost is the cost associated with congestion,
and this amounts to $24 million per day.
These costs signify that there is both an economic and social dimension to policies that
aim to reduce VMT and emissions. Fitzgerald (2010) stresses the importance of public
transportation as a factor in reducing emissions, while also being an important element of
economic progress and local urban sustainability. In comparison to private vehicles, public
transportation produces only 5% of the total carbon monoxide, 10% of the total volatile organic
compounds, and 50% of the carbon dioxide and nitrogen oxide per passenger mile. Not only is it
a more sustainable option for travel, it also has economic benefits. Fitzgerald (2010) states that
the $47 million budget on public transportation in the U.S. helps support 1.7 million jobs in that
industry on an annual basis. There are savings associated with this mode of travel as well.
Public transit ridership saves 1.4 billion gallons of gasoline a year, an amount that translates into
$6,200 annual savings for a two-worker household.

31

On a slightly different dimension, Moore, Staley and Poole (2010) emphasize that
policies aimed at reducing VMT will have economic and social burdens of reduced mobility,
access to jobs, and housing choice. They are of the opinion that climate change policies, in fact,
should aim straight at reducing GHGs, instead of attempting to alter behavioral aspects of
residents. York (2003) did a study that looked at the demographic, economic and socio political
impacts on national car fleet data. He said that we need to look at different perspectives in order
to get a full understanding of the transportation issue. Emissions from the transportation sector
are influenced by three factors: 1) fuel consumption, 2) carbon content of emissions, and 3)
vehicle miles traveled. No one method is strong enough to reduce emissions to meet targets set
by various states throughout the U.S. A combined effort is the only way that current goals can
be achieved.

Scale and gap in research
Research has often been done at various scales, from Schmalensee (1998), who does an
international level comparison of emissions, to Drummond (2010), who looks at the state-level
policy effects on GHG emissions. Wheeler (2008) assesses climate action plans implemented or
initiated by each state in the U.S., while Yang and McCollum (2008) use California as a case
study to analyze different scenarios where emissions may be reduced in the transportation sector.
Environmental burden studies, analyzing fuel consumption for instance, have usually focused on
scales beyond the neighborhood, mostly due to the availability of data (Dahl and Sterner, 1991a;
1991b). As an example, Su (2011) studies, at the urban/city scale, the effect of population
density, congestion and road network densities on average household gasoline consumption for

32

households in urban areas across the country. The author finds that households in cities with low
population densities, higher freeway densities, or areas with higher levels of congestion,
consume more gasoline.
Karathodorou et al (2010) analyze data from the 1999 Millennium Cities Database for
Sustainable Transport, which is a compilation of information at the city scale for about 100 cities
around the world, to study the effects of density on fuel consumption. They find that density
does negatively affect fuel consumption indirectly through the number of cars owned and the
distance traveled, but the analysis is at the city scale, which means that neighborhood level
characteristics of spatial structure within the cities are simply averaged out.
The studies done at various scales are representative of the different components that
make up emissions. Carbon Monoxide (CO) is usually researched at the micro level due to its
immediate impacts on health, while Hydro Carbons (HC) and Nitrogen Oxides (NOx) are studied
at regional levels as they contribute to ozone formation (Bachman et al 2000). Carbon dioxide
(CO2) emissions have been increasing faster than any other type of emission, approximately 29%
from 1990 to 2007 (Mashayekh et al 2011).
The neighborhood level of analysis, where the most varied and detailed information is
found, is underrepresented in emissions and travel behavior research. Emissions, by their
specific characteristics, are a dynamic phenomenon. Total emissions measured in any area are a
combination of household, industrial, commercial, and vehicular sources. Vehicular emissions,
as a source, are inherently dynamic, i.e. although vehicles can emit emissions when stationary,
most of the emissions are created when the vehicles are in motion. So not only is the source
dynamic, the pollutants, once emitted, are also subject to movement due to wind and natural

33

atmospheric systems. Rarely do emissions that are created in one neighborhood stay within
those neighborhood boundaries.
A disadvantage that many researchers have pointed out in this field of study is that they
are restricted to data that usually cannot be obtained at smaller scales, specifically neighborhood
levels). Data at lower levels of scale, such as the National Household Travel Survey are useful,
however the sampling levels are quite low and that poses an issue. Brownstone and Golob
(2009) use the 2001 National Housing Travel Survey (NHTS) California subsample database to
study the effects of density on VMT and fuel usage at the household scale. They find that the
effect of density alone on VMT is too small to warrant policies as a tool to constrain VMT or
greenhouse gas emissions from personal vehicular travel.
Holtzclaw (2002), in his study involving the 2001 NPTS, suggests that the abundance of
variables that affect auto ownership and use at a regional scale and the “lack of adequate
neighborhood-level data has frustrated definitive analysis of these impacts. ….[The] NPTS data
is insufficient in the number of respondents to allow analysis by individual metropolitan areas
with statistical accuracy…. The nature of the NPTS data handicaps its use for the analysis of
density, transit and other neighborhood impacts on vehicle availability and VMT” (Pg 5).
Similarly, Lin and Long (2008) used the 2000 CTTP and NHTS data for their study and admit
that “…as better data (e.g. finer spatial scale, more attributes) become available, finer
representations of neighborhoods will result.” (Pg 749). They conclude their analysis with the
point that “..some of the issues are unlikely to be addressed adequately in regional transportation
planning. Additional effort must warrant…local and neighborhood level [research]” (Pg 757).
In short, researchers recognize the absence of neighborhood scales of analysis—which enables
greater control of built environment characteristics—in gas consumption, emissions and VMT,
34

and acknowledge the fact that a study at this finer scale would be beneficial to overcome the
shortcomings of traditional, higher-scale data.
The general trend in research on this topic has been to compare and contrast
neighborhoods with different built environments to see the differences in travel behavior and fuel
consumption, while taking the differences in socio-economic characteristics of the residents
living in these different areas into account. The gap in the research is two-fold: First, there is a
paucity of research that explores how residents in similar built environments behave differently
due to socio-demographic factors, save a few studies (LeDoux and Vojnovic, 2013; Scott et al,
2009; Williams, 2005). Vojnovic et al (2013) look at neighborhood disinvestment and travel
behavior among residents within neighborhoods based on their socio-demographic
characteristics. They study six neighborhoods in the Lansing, Michigan region and find that
socio-economic conditions affected by neighborhood disinvestment in the declining urban core
are critical to shaping travel behavior.
Similarly, LeDoux and Vojnovic (2013) study access to food stores by residents in two
compact, pedestrian oriented neighborhoods in Detroit, Michigan, which are experiencing severe
disinvestment and decline. They effectively show that residents of lower income neighborhoods
shop for food supplies in ways that are contrary to popular thought. Rather than shopping at the
closest store, they consistently shop at farther stores, thus exercising a choice that is shaped
largely by social preferences rather than convenience factors. The study also shows that there is
considerable variability in travel within the neighborhood that is based on socio-economic
differences.

35

Second, there is a notable absence of research not only on the neighborhood scale
analysis of environmental burdens (such as gasoline consumption and pollutant emissions), but
also a general absence of studies that concentrate on the environmental burdens of residents in
varying, but controlled, built environments based on different socio-economic characteristics.
This forms the main aim of this analysis and this dissertation. The studies mentioned earlier,
Brownstone and Golob (2009), Karathodorou et al (2010), Mindali et al (2004), and Su (2011),
have all tried to reflect on the relationships between the built environment and fuel consumption
at larger geographical scales. There has been no study that takes this notion further: exploring
the socio-economic characteristics of the residents in varied built environments at the
neighborhood level, in order to see more specifically the characteristics that impact fuel
consumption and associated pollutant emissions.

Gap in Research, focusing on Detroit, MI
Another layer of research shortcomings is evident with the intended study area, the
Detroit metropolitan region. Much has been written about the growth and decline of the City.
However, detailed studies, at the neighborhood level, are few and far in between. Vojnovic et al
(forthcoming) look at the travel patterns of residents in six different neighborhoods in the Detroit
region. Similar to the above studies, the research finds the residents in similar built
environments have different travel patterns that vary with their socio-economic characteristics.
The Detroit Area Studies are a series of studies based on an annual survey (done through
a combination of interviews and mail-out surveys) carried out by the Sociology Department and
the Institute for Social Research at the University of Michigan. The survey was conducted
36

annually from 1951 to 2004 and focused on different themes and issues of public and personal
life (ICPSR, n.d.). Although these studies are at the neighborhood level, and the data is rich and
detailed, none of the studies have focused on transportation patterns and/or emissions. Research
studies on emissions in Detroit are in line with broader emissions research discussed earlier.
They look at emissions as affecting air quality and health of residents. For example, Duvall et al
(2012) did a study on the sources of the fine particulate matter (PM2.5) in various residential
outdoor locations in Detroit, MI. Since PM concentrations have been known to be a causal factor
in increased mortality, the US EPA establishes standards for air quality that aim to protect the
public from the negative effects of PM concentrations. Detroit, MI is designated as an area that
is in “non attainment” of these standards. This group of researchers found that residential areas
differed in the measurements of the main sources of PM (particulate matter). Residential areas in
downtown Detroit had road dust as the primary contributor, while areas near a highway had
motor vehicles as the primary source. In summary, this study looked at the spatial variability in
the source of one of the emission factors of vehicles. These studies, however, do not distinguish
between contributions of these emissions or the socio-economic attributes of polluters.
Lindhjem et al (2012) conducted a study on the emissions and air quality estimations in
the Detroit and Atlanta metropolitan regions. They created a new model (CONCEPT MV) that
uses data from EPA’s current emissions estimation model, MOVES2010, and estimates
emissions on a road link by link basis, allowing for a more detailed estimation of vehicular
emissions (light duty vehicles and heavy duty vehicles). What they find is that the new model,
by using more fine grained data and inputs (at the road link level), estimates emissions that are
much more accurate. They call for the importance of detailed data that would lead to truer
estimations of emissions. Although this study is important for detailed emissions inventory, just
37

like most other emissions studies, it does not touch upon the social characteristics and built
environment as contributors to environmental burdens, that is gasoline consumption and
pollutant emissions.
In general, emissions studies in Detroit, MI and most other places have concentrated on
the sources and spatial and temporal variability in different emission factors. This leaves a
considerable gap in current research. Just as it is important to know what the specific sources of
emissions and air pollutants are, so it is important to know what the physical and socio economic
characteristics are that pertain to the contributors of these sources. Research in Detroit is missing
the combination of the social characteristics of the residents of different neighborhoods in the
region (including the City) and their relationship with travel behavior and subsequent vehicle
emissions, thereby establishing detailed “environmental burden” by specific neighborhood
physical traits and socio-economic characteristics. Reiterating the conceptual framework,
differences in the built environment do have an impact on travel patterns and behaviors of its
residents. However, social and economic characteristics of the residents also have an important
role in shaping travel patterns, vehicle characteristics, gasoline consumption, and pollutant
emissions. This work will determine whether travel patterns and behavior, although influenced
by the built environment, are determined more by the preferences and actions of residents, and
these are derived from their specific socio-economic characteristics.
This study is important in two ways. First, it is conducted at a very disaggregated level—
a neighborhood scale of research—indeed one of the lowest levels of analyses ever undertaken in
exploring environmental burdens. Second, this analysis focuses on socioeconomic conditions in
affecting travel and environmental burdens. The linkage of actual vehicle characteristics to
specific household characteristics in travel behavior (and environmental burdens) is a critical
38

addition to the exiting literature. An important added dimension to this research is that it not
only focuses on residents that live in robust communities, but also includes two neighborhoods
experiencing extreme disinvestment and decline in the City of Detroit. This is a unique
dimension to travel behavior research and a significant contribution to the existing literature.

39

Chapter III - The Detroit Region
Introduction
The study area for this dissertation is the Detroit Region in southeast Michigan. This
Region, akin to the jurisdiction of the Southeast Michigan Council of Governments (SEMCOG),
includes the counties of St. Clair, Livingston, Oakland, Macomb, Washtenaw, Wayne, and
Monroe. Detroit City is the largest urbanized area in this Region and as such, has had a
tremendous influence on the morphology of the regional spatial structure. The purpose of this
chapter is to give the reader an overview and highlight certain aspects of the history of Detroit.
The review will provide a background for Detroit’s growth and expansion in the years leading up
to its current distressed state. This, in turn, will lead to a clearer understanding of why this
region, and certain neighborhoods within it, were chosen as the focus for this dissertation. The
City has been written about extensively. Various aspects of Detroit’s past, and in greater detail,
can be gained by reading the many excellent works on this subject (Darden et al, 1987; Darden &
Thomas, 2013; Farley, Danzinger & Holzer, 2000; Gallagher, 2010; Galster, 2012; Sugrue, 1996;
Thomas, 1997; Williams, 2009). This chapter however, gives an overview of Detroit’s
morphology through its growth, transition and decline periods, and provides a link to various
economic, political and social forces that have shaped the city’s built environment. In addition
to providing a review of Detroit’s history this chapter will also provide a discussion of the wider
region and various aspects that lead to the selection of the six neighborhoods included in this
research.
The Detroit region is unique in the sense that the city of Detroit itself is in one of the
poorest counties (Wayne County, MI) in the country, even though its surrounding counties

40

(for example, Oakland County, MI) are amongst the wealthiest in the nation. Tables 1 and 2 show the historic trends in per
capita income and median family income by county for the Southeast Michigan Council of Governments (SEMCOG) Region.
Figure 1: The Detroit Region

*For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this
dissertation

41

Table 1 – Per Capita Income by SEMCOG County
Per Capita Income
State of Michigan
City of Detroit
Wayne
Oakland
Macomb
Livingston
Monroe
St. Clair
Washtenaw

1970
$3,373
$3,227
$3,505
$4,515
$3,605
$3,320
$3,213
$3,048
$3,767

1980
$7,688
$6,215
$7,608
$10,658
$8,655
$8,323
$7,356
$7,080
$8,703

1990
$14,154
$9,443
$13,016
$21,125
$16,187
$17,327
$13,893
$13,257
$17,115

2000
$22,168
$14,717
$20,058
$32,534
$24,446
$28,069
$22,458
$21,582
$27,173

2010
$25,482
$15,261
$22,351
$36,314
$26,661
$31,751
$25,774
$23,960
$32,529

Source: U.S. Census Bureau

Table 2 Family Income
Median – Median Family Income by SEMCOG County
State of Michigan
City of Detroit
Wayne
Oakland
Macomb
Livingston
Monroe
St. Clair
Washtenaw

1950
$3,519
$3,955
$3,989
$4,031
$3,733
$3,156
$3,450
$3,245
$3,435

1960
$6,256
$7,715
$6,597
$7,576
$7,091
$5,775
$5,892
$5,546
$6,890

1970
$11,032
$10,045
$11,351
$13,826
$13,110
$11,551
$11,398
$10,125
$12,294

1980
$22,107
$17,033
$22,134
$28,803
$26,666
$26,339
$23,281
$21,119
$25,465

1990
$36,652
$22,566
$34,099
$50,980
$44,586
$49,910
$40,532
$35,678
$47,308

2000
$53,457
$33,853
$48,805
$75,540
$62,816
$75,284
$59,659
$54,450
$70,393

2010
$60,895
$33,445
$53,004
$84,580
$67,454
$82,888
$66,284
$59,952
$84,770

Source: U.S. Census Bureau

The stark inequalities in the physical and social fabric of the region, visible at the county
scale, show even more nuance and detail when viewed from a lower scale of analysis, such as a
municipality or neighborhood. There is a visible, disproportionate percentage of the population
segregated within the inner city, which is predominantly black and of lower socio-economic
status than residents in the surrounding counties/suburbs (Darden and Kamel, 2000; Farley et al,
2000). Figures 2, 3, 4, and 5 depict these concentrations and population trends by race for the
counties in the SEMCOG region. Detroit city itself has pockets of hope, evident in beautiful
42

houses in a few neighborhoods, the riverfront area, and in the downtown employment cluster and
entertainment venues, such as the sports stadiums, casino, and ethnic enclaves. Mostly,
however, the city is saddled with abandoned houses, boarded up manufacturing plants, empty
industrial lots and general physical deterioration.
Figure 2 – Population Trends by SEMCOG County

Figure 3: Percent White Population Trend by SEMCOG County

43

Figure 4: Percent Black Population Trend by SEMCOG County

Figure 5: Percent ‘Other’ Population Trend by SEMCOG County

44

The Growth Period
Economic/Industrial factors
At the beginning of the twentieth century, Detroit was a magnet for people and new
development because of its rich manufacturing base. Manufacturing had long been the dominant
industry of the region, with rail cars and stoves being important early industries (Voyles and
Rodrique, 2012). The automobile industry expanded after World War I under the leadership of
pioneers that had the economic strength to take advantage of Michigan’s manufacturing base and
its location and resources (Voyles and Rodrique, 2012). With automobile manufacturing
expanding at unprecedented rates, the ‘car’ began to dominate the city, and particularly after the
process of manufacturing became standardized. Henry Ford pioneered the assembly line process
at his Highland Park plant (Galster, 2012). The city itself, nicknamed the ‘Motor City’, emerged
as the center for automobile production and the Detroit metropolitan area grew because of the
allied industries that fueled automobile manufacturing.

Municipal Expansion
As the population in and around the City of Detroit grew, annexation was used to
increase the boundaries of the City so as to accommodate the growth and keep the taxes and
businesses within the jurisdiction of the City. The figures below show Detroit’s growth through
the years. From 1815 to 1891, the City grew from 0.330 square miles in size to 28.350 square
miles. The largest amount of land during that period, just under 7 square miles, was annexed in
th

1857. In the first quarter of the 20 Century, from 1906 to 1926, Detroit grew exponentially due

45

to annexations (Galster, 2012). It expanded from 28.350 square miles in 1906 to its current size
of 139 square miles by 1926, with the largest amount of land being annexed in 1925 (over 42
square miles) (Board of County Auditors, 1926).

Figure 6: Detroit Annexation Map (showing the years in which different land parcels were
annexed by the City)

Social factors and related mobility
By the 1920s, the principle mode of travel had changed from horse-driven carriages to
the automobile. Labor demands at the time were primarily for blue collar jobs that required
repetitive, physical work to be performed (like working on an assembly line). Immigrants and
46

migrants were flooding the city to be a part of Detroit’s industrial success. The Ford Motor
Company was the largest employer at the time and Henry Ford wanted to give his employees
(mostly immigrants) a chance at a higher standard of living so that they could be assimilated into
the American mainstream (Bates, 2012). He started the $5 wage per day, which is attributed to
the movement that created the middle class (Counts, 2011). This wage level also meant that the
employees could afford to buy a vehicle for themselves. More households owned cars and its
ownership was more utilitarian than luxury based, fueling consumerism throughout the region.
th

Into the mid-20 century, the influx of people was a response to the growing demand for
workers at manufacturing plants that catered to the automobile industry and eventually to the
needs of the Second World War. The increases in population lead to an increase in the demand
for housing. However, the housing supply was unable to keep up with the demand by the new
waves of migrants entering the City. Established Black residential neighborhoods saw
overcrowding and deteriorating conditions as more Blacks, attracted by the available jobs, tried
to settle in the City (Bates, 2012). Some new segregated housing projects were constructed to
accommodate the increasing Black population. For example, the Sojourner Truth Housing
Project was built in the beginning of the 1940s to accommodate black residents (Farley, 2009).
Black families did occupy the housing in Sojourner Truth, but only after violent protests by
white residents in surrounding neighborhoods was controlled by military and police presence.
Into the 1950s, people, mostly whites, who could afford to move started leaving the City.
They located in less crowded areas outside of the City and spurred the movement and growth of
the Detroit suburbs. Preferences for larger homes on larger lots had a big impact on
suburbanization and development in the surrounding region. The rapid growth and movement of

47

the population into the suburbs, along with economic prosperity in the form of increased
disposable incomes and lower energy costs, produced a development preference for single
family, detached housing in open and less dense areas. These decentralization patterns also had
clear racial imprints. The white population was suburbanizing and distancing itself from
population sub-groups they considered a threat. The suburbs have been quite effective in
keeping the Black population out of their jurisdictions through house pricing structures,
gatekeeping institutions, and local development policies (Thomas, 1997). Steadily the
segregation of the population became increasingly and clearly demarcated. Blue and white collar
workers moved out from the central city to the surrounding periphery. Warren was one of these
towns that was revamped from a rural area to a bustling blue collar town. The white collar
residents moved more to the north and west of the City. Black inner-city neighborhoods
developed very differently from the new white suburban neighborhoods. The City was slowly
becoming an area for the black working class with lower incomes and fewer urban amenities.
Detroit had long been a major stop on the Underground Railroad system, a metaphor for
the network that was used extensively to shift black populations from the southern states into the
northern areas to free them from slavery. These populations would then either stay in the City
and take advantage of the booming manufacturing industry and its employment opportunities, or
continue north to Canada (Counts, 2011). Many of these migrants found work at the auto
manufacturing plants such as the Ford Motor Company. Henry Ford had a big hand in
increasing the morale and employment opportunities for black workers. Blacks had a larger
range of jobs open to them at the Ford River Rouge plant than at any other plant (Ford or any
other company). The black workers at the Rouge plant were in positions that were reserved for
whites in other manufacturing companies (Meier & Rudwick, 1981). At the same time,
48

management tried their best to avoid organized labor and unions amongst its workers. With Ford
employees joining the United Auto Workers union in the 1940s, Ford moved operations to other
jurisdictions to take advantage of new technologies and to get away from labor union powers
(Sugrue, 2010).
A large part of the increase in population was due to the migration of black workers from
the South. These workers came to the north in search of a good work environment, a decent
wage, and to escape the persistent discrimination so entrenched in southern culture (Galster,
2012). The quick increase in the black population was being dealt with by segregating them into
an area on the east side along Hastings Street, an area called ‘black bottom’ (which later got
redeveloped into Lafayette park) (Williams, 2009). Interestingly, the name ‘black bottom’ was
originally given by the French indicating the dark, rich and fertile topsoil of the area (Binelli,
2012).
North of Hastings Street was an area called Paradise Valley, which was the entertainment
and business hub of the black residents. These bustling areas enjoyed strength and prosperity
even in the Prohibition years, when the illegal sale of liquor was a dominant ‘underground
economic activity’ (Bates, 2012). Money was flowing through the community, either through
employment at Ford plants or through the circulation of money within the community through
legal and illegal establishments. Segregation into certain areas was not the only form of
discrimination practiced by landlords and housing authorities. The discrimination was also
evident in the rents. Blacks shared their residential blocks with other eastern European
immigrants, but they paid as much as about 50% more in rent for the same housing facilities,
when compared to the new immigrants (Bates, 2012).

49

Transition Period
Economic/Industrial factors
The Great Depression years, during the early 1930s, witnessed decreased economic
activity and incomes. In order to sustain the consumerism that was sweeping the country, new
ideas and alternatives that replaced the iconic all-black utilitarian Model T with colorful and
stylish vehicles were implemented by General Motors (Bates, 2012). Ford began suffering
losses and market share of its products. Its workers were hit hard. Unemployment, reduced
work days, and mounting home payments added to the woes of the working class, not only
employed at Ford, but at Chrysler and General Motors as well.

The Role of Roads, Highways and the Automobile
Fueling the mobility associated with economic prosperity was the Federal Highway
Administration’s Interstate Highway Act of 1956, which substantially increased the movement of
American people across urban regions and beyond (Weingroff, 1996). Part of this very
ambitious project was the construction of the Lodge freeway and the later the Edsel Ford
expressway that was purported to improve the travel of people from the outer areas of the City
into the Central Business District (CBD). Travel by highways and expressways was encouraged
by policies and automobile companies that tried their best to discourage the use of public transit
(Whitt and Yago, 1985). The ‘Big Three’, Ford, General Motors and Chrysler had used all their
powers to keep the automobile the primary vehicle for mobility and dissuade public transit
permeation within the Detroit region. The Interstate Highway system was moving along so
rapidly and with such intensity that support for public transit, and also its federal funding, faded

50

away. This affected the service and customer satisfaction levels for whatever transit was
available at the time. The city that brought the automobile industry to life was left with a dismal
transit system (Vojnovic and Darden, 2013).
th

In the early 20 century, making the best of whatever they had, Detroit’s African
Americans took control of Black Bottom and Paradise Valley, making these areas highly
successful. Businesses, theaters, services and facilities meant solely for the black residents were
mushrooming through the area during the 1920s and into the Prohibition period (Bates, 2012).
However, this success was not meant to last long. Both the black dominated areas of the City
(Black Bottom and Paradise Valley) were targeted during the period of Urban Renewal in the
late 1950s. Housing projects and freeways replaced Detroit’s thriving Black neighborhoods.
Housing projects took out the poor residents with the promise of giving them better housing once
the redevelopment was completed (Thomas, 1997). Instead they were displaced altogether and
had to find other areas of the city to call home.
Similarly, other government policies have also had a hand in the decline of the inner city
area of the Region. Investment in and subsidization of suburban infrastructure has been a big
factor in suburbanization (Vojnovic, 2009). During the post World War II period, the Federal
Housing Authority (FHA) and the Veteran’s Administration (VA) insured and subsidized
mortgages to suburban residential areas created an attractive incentive for people to move to the
suburbs and realize their ‘American Dream’.
Initially, the mechanism for coping with growth was annexation. However, the state of
Michigan used the ‘home rule’ power in the state constitution, which subsequently allowed
places to get incorporated so as to create new municipalities, with the result that segregated
51

fragmentation occurred without any control. Consequently, the city boundaries stopped
expanding in 1926. Incorporation of municipalities has led to fragmented governments and
policies (Thomas 1997). Between 1950 and 1980, fifty incorporated places were added to the
Detroit metropolitan area (Darden, 2009). This fragmentation harbored a sense of control for
suburbanites as they demarcated their area and consequently fostered pro-sprawl growth,
imprinting and stereotyping the socio-economic inequalities (Galster, 2012).

Social factors and related mobility
Rising racial issues were not only felt through housing segregation, but permeated into
the public school system as well. In 1954, the US Supreme Court case of Brown vs. Board of
Education ruled that segregation within public schools was illegal and it was against the
Fourteenth Amendment that gave equal rights to all citizens (Brown v. Board of Education,
1954). However, the 1974 landmark case of Bradley v. Milliken in the Supreme Court ruled that
since the suburban municipalities were not segregating students but were open to all students of
the municipality (that essentially had primarily white residents), they were not in fact promoting
segregation (Bradley v. Milliken, 1974). This ensured that suburban, fragmented municipalities
could effectively keep the inner city children out of their schools (just as they had kept the inner
city residents from entering their suburban communities). The Detroit Public Schools were
subsequently left with a segregated, concentrated pool of black and poor students. The strain on
racial relations between the two groups has been stressed upon by researchers who posit that the
blacks have been continuously discriminated by whites in all areas of everyday life choices, be it
housing, employment, or education (Darden et al., 2007; Thomas, 1992; Sugrue, 1996).

52

During the 1950s and 1960s, huge tracts of land needed to be cleared for the newly
sanctioned highways and the black neighborhoods were the ones that bore the brunt of these
projects. Just like in urban renewal projects, the established black neighborhoods (residential
and commercial) were condemned and razed so that highways could be built and people could
travel into the city at high speeds. Mayor Jeffries had condemned 129 acres of Black Bottom so
that the Chrysler freeway could be built (Binelli, 2012). The rest of Black Bottom was later
replaced by Lafayette Park. In total, about 5,000 buildings were demolished, most of which
belonged to the black community (Williams, 2009).
The widely known incentives for home-buyers in the growing suburban areas were
attracting middle class residents out of troubled Detroit and into the newly built residential
suburbs. However, these incentives were only available to the white population (Darden, 2009).
The black population would be overtly rejected for suburban home mortgages. Appraisers would
be allowed to mark primarily black neighborhoods with red ink, indicating high risk areas where
government subsidies and mortgage financing should not be given, hence the term ‘redlining’
(Darden, 2009). The practice went on for three decades, from the mid-1930s to the mid-1960s
(Renolds, Danziger and Holzer, 2000). If federal programs, such as these would be challenged
and overruled, local governments had other ways to keep racial segregation going – namely
zoning measures and housing restrictions (Galster, 2012). Not only the governments, but real
estate agents played their role in fostering segregation as well. Keeping the interests of their
white clients in mind, they would ‘steer’ black potential home buyers to areas that were
predominantly occupied by the same race. After all, “…whites did not care to live with blacks
and could not get FHA/VA-backed mortgages if they did; blacks…would not get FHA/VAbacked mortgages in any event” (Galster, 2012; pp. 148).
53

The situation was so severe that a developer, in order to build a white neighborhood near
Eight Mile road (which had black neighborhoods in close proximity), had to erect a concrete
wall, six feet high and half mile long along the black neighborhood so as to physically separate
the black and white communities (Rhodes and Jeffries, 2010). This was done because the
Federal Housing Administration would not have backed mortgages in the white neighborhood
because it was so close to the ‘redlined’ black neighborhood. Mortgages were approved in the
area after the wall was erected.
As if it was not enough that the government and the market were doing their best to keep
residential neighborhoods segregated, the homeowners themselves used harassment and
restrictive covenants to do their part in keeping the unwanted out of their neighborhoods
(Galster, 2012). Restrictive covenants were like clauses added to property deeds that restricted
future uses and certain kinds of occupants within a neighborhood. These covenants were used
widely until 1948, when the Supreme Court ruled against them in the Shelley vs. Kraemer case
saying that they were not enforceable by law as they were private agreements and as such they
were against the Fourteenth Amendment to the Constitution that gave equal protection to all
citizens (Shelley v. Kraemer).
The exodus from the City was primarily by white residents. Housing policies,
gatekeeping institutions and real estate markets would effectively keep the black population
within the city and segregated in certain areas, and would steer them away from or refuse to help
them attain the quality of life and residential choice that their white counterparts were demanding
and receiving. Apart from such formal and structural methods of racial segregation, social
effects such as racial solidarity have also had an impact on residential preferences (Bledsoe et al,
1995). In a survey administered to black residents in inner city Detroit and suburban Detroit, in
54

segregated neighborhoods as well as integrated neighborhoods, Bledsoe and colleagues (1995)
find that there is greater solidarity among Blacks in segregated neighborhoods when compared to
their counterparts living in integrated neighborhoods, forming one of the reasons why Blacks
may choose to live in neighborhoods that are more segregated/homogenous. However, social
and economic conditions of those living in the suburbs were found to be much higher than those
living in the inner city, resulting in a dilemma between increasing racial integration or the
advancement of the Black population based on the strong bonds of solidarity (Pg 453).
Any discussion of the Detroit region is practically incomplete if the racial tensions and
riots are not brought up. Racial tensions had been simmering since the beginning of the 20

th

century. The early 1940s witnessed a race riot that started on Belle Isle and engulfed the entire
th

City, starting on June 20 1943 and lasting for 36 hours. The end came after six thousand
federal troops were called in to help end the disturbance that took the lives of twenty-five black
and nine white residents (Binelli, 2012). Although the 1950s saw growing population and
suburbanization, it also witnessed increased racial hostility. Darden and Thomas (2013) describe
the racial resentment taking life in the form of hostility by white residents towards black families
that tried to move into a predominantly white neighborhood. The exacerbated results of failed
urban renewal programs that uprooted and displaced entire black neighborhoods added to the
frustration felt by this dislodged and mistreated community. Hostility and segregation efforts
towards the black families had the effect of them living in “…overcrowded, poor ghettos, which
became breeding grounds for angry black youth waiting for an opportunity to explode.” (Darden
and Thomas, 2013; Pg 5)

55

Two decades later, the riots of 1967 did worse damage. Following a raid on a ‘blind pig’
(an after-hours drinking establishment) by the police, what was expected to include a small
number of arrests (about 20 people) turned out to be an issue that got out of control, resulting in
violence, death and a section of the city being burnt down. The riots went on for five days, over
forty-three people died, about seven thousand people were arrested, and about three thousand
buildings were burnt (Binelli, 2012). Only after the influx of the police, the Michigan National
Guard and the U.S. Army to contain the rioting did the rampage die down. Subsequently,
President Johnson created an eleven member commission, the National Advisory Commission on
Civil Disorders, popularly known as the Kerner Commission - to look into urban riots. They
stated in their 1968 report that “..our nation is moving toward two societies, one black, one white
– separate and unequal.” (Kerner Commission, 1968). They indicted the white society for
isolating the African American community and urged the government to take action to aid the
black community so that they can avoid further racism from the dominant race. However, the
social effects of these outbursts resulted in people, and generally whites, continuing to be
encouraged to move out of the city and into the surrounding suburban areas to avoid being in the
midst of such violent conflicts.

Period of Decline
Economic/Industrial factors

56

White flight was not the only exodus out of the city during the second half of the 20

th

Century. Retail, commercial, and industrial investment also moved out of Detroit. In addition,
the ‘big three’ automobile companies (Ford, Chrysler and General Motors) overpowered the
other automobile manufacturers in the area and drove them out of business (Sugrue, 2010). In
addition, the ‘the big three’ moved to expand their businesses, but did so in the outskirts of the
City. Following them, smaller allied industries moved to the outskirts as well. With the Ford
employees joining the United Auto Workers union in the 1940s, Ford moved operations to other
jurisdictions to take advantage of new technologies and to get away from labor union powers
(Sugrue, 2010). In the 1950s, about 150 new plants (manufacturing and allied industries) were
built in the suburbs of Detroit, with one third of these relocating out from the City to make this
move (Galster, 2012). Detroit witnessed “…a dramatic intraregional shift in the location of all
economic activity – manufacturing and nonmanufacturing alike – in the last half century”
(Galster, 2012; pp. 51).
th

In the last few decades of the 20 century, the economic downturn facing the automobile
industry crippled the City that was built around it. Competition from international companies,
and particularly from Germany and Japan, with new manufacturing procedures left the Big Three
in dire straits and they suffered significant losses (Galster, 2012). To keep up with the
competition, the auto companies employed more technology driven procedures that required
fewer workers. The economic crisis exacerbated the white flight and the City was increasingly
left with an unskilled, lower income, segregated population base and a suffering local
government. With jobs moving out of the City, and unemployment rising within the City,

57

property prices and taxes declined. This affected the local finances and public services suffered,
and particularly the Detroit public school system (Mirel, 1999).

Built environment factors and related mobility
The second half of the century witnessed a decline in Detroit like no other city in the
country had experienced. Commuting was not an issue to suburban residents and was preferred
over living in the inner city. Although some studies have shown that the move to the suburbs by
residents and commerce alike helped move commutes out of the city center and into the suburbs,
thus alleviating congestion, other studies showed that suburbanization did not move commutes
out of the inner city, but only increased commute times as people traveled longer distances into
the inner city areas for work (Kirby and LeSage 2009; McGuckin and Srinivasan 2003).
The election of the City’s first black Mayor, Coleman Young, in 1973 marked an
optimistic hope for the residents. Young has done many things to improve the City, its economic
slump, racial tensions and housing markets. One of these efforts was targeted at rebuilding
residential neighborhoods that had abandoned and vacant lots (Farley, Couper, and Krysan,
2007). These efforts from Young’s office and his successors (Archer in 1993 and Kilpatrick in
2003) focused on giving subsidies to developers to build in the City. Developers were initially
reluctant to build in Detroit because no one wanted small-lot and small-sized housing, together
with the fact that these would be expected to be bought by middle class black residents (as the
white residents were reluctant to move back into Detroit). However, these subsidies did attract
some developers to reinvest and rebuild certain parts of Detroit.
58

th

st

In the last decade of the 20 century and into the first decade of the 21 century,
improvements in the housing market saw private redevelopment efforts in the downtown and
surrounding areas, made possible largely through heavy subsidies provided by government.
Ryan (2006) reported that all developments during the 1990s and early-2000s included some
form of government subsidy. Subsidies ranged from creating Neighborhood Enterprise Zones (a
tax abatement for 10 years on a new home purchase), to large and costly subsidies for site
remediation and provision of foundation-ready lots (Ryan, 2012). For example, the Victoria
Park redevelopment in the Jefferson-Chalmers area received subsidies to the tune of $125,000
per house (Ryan, 2012). Most of the other deteriorated parts of the City received little incentives
for developers to build as the market was uninterested in purchasing property located in
‘troublesome’ areas. Renewal efforts targeted at many of the neighborhoods were abysmal. As
effectively expressed by Ryan (2012), “the growth machine reproduced urban renewal’s worst
features – involuntary displacement of low-income homeowners for a mostly failed attempt to
attract “high-end” homes – without bringing any of the stabilizing influences that large-scale
government intervention in an abandoned neighborhood might have delivered” (Pg. 88).
The Detroit suburbs, on the other hand, continue to prosper with vibrant residential and
commercial activity. Suburban home values are among the highest in the country, educational
facilities (schools and universities) are among the top ranked and quality of life of the residents
keeps drawing more people away from the City (Woosley, 2008). Infrastructure, public services,
and entertainment are top notch. This has resulted in a blatant divide between the city and the
suburbs. The population in the suburbs consists mostly of whites, along with an increasing
presence of Asians, with the socio-economic composition of these residents consisting
extensively of middle and upper income earners.
59

Both the city and the suburban residents are eager to hold on to their assets, be it financial
or physical. The city refuses to delve into land banks, has declined to hand over the Cobo Center
to a regional governing authority, or to enter into joint regional control over its water system,
being reluctant to give up any more of its ‘assets’ to the white majority in the region (Galster,
2012). On the suburban front, Oakland County turned down a proposal in 2005 for increased
taxes to rehabilitate cultural attractions in the City that would benefit both Wayne and Oakland
counties. Similarly, suburban residents, until recently, refused to be a part of regional
transportation agreements offered through the Suburban Mobility Authority for a Regional
Transportation (SMART) bus system. They do not want increased taxes to foster a bus transit
system that mostly poor blacks would make use of (Galster, 2012). However, a Regional Transit
Authority was created in March 2013 and work plans for a bus rapid transit system for four
counties in the SEMCOG region (Wayne, Oakland, Macomb, and Washtenaw) are currently
being created. This cooperation was mandated by the federal government, as a precondition for
receiving funds for a light rail line that has been proposed and declined 23 times in the past
(MDOT, 2013).

Social factors and related mobility
While Glaeser, Kahn and Rappaport (2008) state that more than 19 percent of the central
city population in general are poor, the 2010 US Census figures show that over 37 percent of the
population in Detroit City are poor. Home values are at an all-time low and residents are locked
into deteriorating neighborhoods (Green, 2013). Increased city taxes have not been able to

60

improve the condition of public services, while they have added an extra burden on Detroit’s
poor.
Racial tensions as well as class conflicts have been vital in the decentralization of Detroit
(Darden et al 1987). As the boundaries of urban Detroit grew, so did segregation, racial conflict
and a distinct ‘hollowness’ within the central city. Residents continued to move to the suburbs, a
condition that has only worsened, as the most recent census shows a decrease in population in
the city that has brought population levels down to that of 1920 (Census, 2010). The Detroit
metropolitan area has seen slightly declining inner city segregation and increased
suburbanization of black residents, however, blacks in suburban Detroit are still highly
segregated (Darden et al, 2007). It is a frustrating time for Detroit residents who have to deal
with unemployment, discrimination, high taxes, crime, and property disrepair and devaluation.
The extreme scale of disinvestment and decline in the city has received extensive national and
international media coverage. In the words of Chris Hansen (2010), who recently described the
condition of Detroit on Dateline NBC as follows:
They litter the landscape, thousands and thousands of abandoned homes. And just like
these buildings, Detroit is a shell of its former self. One third of the people here live in
poverty. Almost half the adults are illiterate, and about 75 percent of kids drop out of
school. I could be describing some ravaged foreign nation, but this is the middle of
America.

The Outlook: Dismal, yet hopeful!

61

st

The 21 Century sees little improvement in the situation facing the City and its people.
Darden et al (2007) show that the inequalities present between the black and white residents in
the state of Michigan has only increased in the four decades since the 1967 riots. Residents are
still leaving, with the 2010 Census showing the City of Detroit as having plummeted in
population from about 950,000 in 2000 to 714,000 in 2010. The City that was once one of the
largest in the country in terms of population (ranking in fourth place) is now smaller than Austin,
Texas and Charlotte, North Carolina, ranking only in eighteenth place by size (Seelye, 2011).
Two-thirds of the City’s children under the age of five years live in poverty and the
unemployment rate has topped 15% (Data driven Detroit, 2012). Moreover, the City has just
filed for bankruptcy, the ex-mayor went on trial for embellishment of public money, and the big
three auto makers had to be bailed out by the government in order to avoid collapse (Cho and
Marr, 2009; Davey and Walsh, 2013; McClam, 2013;).
The public school system is inferior in the quality of education, with both the level of
segregation and the dropout rates being among the highest in the nation. Greene and Winters
(2006) show that the Detroit public school system has the highest dropout rate in the nation in
2003, at 58%. The dropout rate has been decreasing in recent years, although it still remains
high in national comparisons (Detroit Public Schools, 2011). Crime rates continue to remain
high. The number of fires in the City is still high, with about 30 structure fires being reported
each day (Schwartz and Gold, 2012). The landscape is dotted with abandoned homes and
businesses and almost one third of the City lies vacant.
The vacant landscape is considered the biggest challenge and also the biggest
opportunity. The dilemma of how to utilize these vast amounts of vacant land that are starting to

62

look like rural or natural landscapes is what many public officials are contemplating. Detroit has
been symbolized through nicknames several times throughout history. Starting off with it being
th

th

the ‘industrial boom town’ in the late 19 century to the early 20 century (Maynard, 2011),
followed by it being the ‘arsenal of democracy’ throughout the second world war (Detroit
th

Historical Society, 2013). It embraced planning in the second half of the 20 century and
adopted the ‘model city’ icon, and after that things went downhill (Thomas, 1997). ‘Murder
capital’ (Mach, 2013), ‘ghost town’ (Hitchens, 2011), ‘shrinking city’ (Independent Lens, 2013)
– these have all been names describing the city as it has been journeying through its troubled
times. Galster (2012) offers four “cold statistics” that summarize Detroit’s self-propelling
decline: “From 1950 to 2005, Detroit lost 29 percent of its homes, 52 percent of its people, 55
percent of its jobs, and 60 percent of its property tax revenue” (pp. 238).
Even amidst all this, there is an undying hope and struggle to reinvent the City. The
devastation has left Detroit in such a state that it could start off as a clean slate for new inventive
and creative ideas. The City still has a relatively high density – averaging over 5,000 people per
square mile as reported in the 2010 Census (down from about 13,000 people per square mile in
the 1950s and 6,500 residents per square mile in the 2000 Census). Comparatively, cities such
Phoenix with a density of 2,900 residents per square mile and Dallas with a density of 3,400
residents per square mile are much more sparse (Gallagher, 2010). The Heidelberg Project, a
block of residential area turned into an art space, has gathered attention in the past and still does
(The Heidelberg Project, 2013). The Project attracts fifty thousand visitors annually and has
extended to a few blocks of the surrounding residential area.

63

In other areas of Detroit, volunteers work tirelessly to maintain City parks, demolish
abandoned homes so as to curtail crime activity, induce community involvement, and help keep a
vigil out on neighborhoods (Blight Busters, 2013). The food desert label that has been given to
the City is also on the verge of being remedied. The owner of Hantz Farms has been purchasing
land in the city to farm fresh fruits and vegetables for the residents of Detroit (Huffstutter, 2009).
Whole Foods has been given a subsidy to open a mini-store in the Midtown portion of the City,
thereby introducing a space for fresh and healthy produce to be sold to the residents (Gallagher,
2013). Meijers is also opening a store at Woodward and Eight Mile, which is poised to open in
July 2013 (CBS Detroit, 2013).
In addition, Detroit is a major-league city, with the Red Wings, Tigers and Lions all
having their arenas in the downtown. There are three casinos and many restaurants and bars
attracting a bustling life in the city’s core. Many of the city’s residential neighborhoods still
have beautiful homes, are well-kept and command a high price, and most of them are on the
National Register of Historic Places. These include neighborhoods such as the Indian Village,
Bush Park Historic District, Woodbridge Historic District, Palmer Woods, Rosedale Park and the
East English Village (National Park Service). Jazz music venues, artist homes, and urban
agriculture are all on the table for this struggling city. How the City charts its future, is yet to be
seen.

Commentary

64

The Detroit region has gone through extreme conditions over the last century. It has
cradled the automotive industry and raised it to levels that have prompted innovation and
unprecedented strategic practices. However, just as with any product and its business cycle, the
automotive industry experienced a wave of decline, which coupled with other racial issues of the
time, affected the Detroit Region in many and devastating ways.
The Region has been a poster child for showing the effects of concentrated poverty and
high levels of segregation. The exodus of the middle class white population to the suburbs has
left the city with vast amounts of vacant land, abandoned properties, inferior public services and
a poor and struggling workforce. The situation is changing, albeit at a slow pace. Reinvestment
is being welcomed. However, race and class is still a major issue, with the majority black
population, coupled with high rates of poverty, living in this city characterized by unprecedented
disinvestment and decline.
With disinvestment and decline, mobility and access to basic amenities in the city has
been also affected and in dramatic ways. . Mobility has been particularly shaped by the lack of
an extensive public transit system for a city of this size and vintage, leading to over-dependence
on the automobile (Vojnovic and Darden, 2013). Deteriorating conditions of streets and
sidewalks, lack of vibrant shopping areas and uninviting public spaces are also built environment
factors that have impacted mobility. Social factors have also played a role in determining
transportation through this city. The perception of safety and preferences of the residents to go to
certain destination over others has been important in shaping mobility. This is a city that has
been traditionally built to induce transportation in all forms (motorized and non-motorized) and
has had the text book qualities of high density, high connectivity, mixed land uses to promote
high accessibility throughout the once active and vibrant urban core. However, disinvestment
65

and decline in many of its neighborhoods—which have removed many necessary daily
destinations from the city—coupled with a deteriorating built environment have adversely
affected travel in this city. In addition, social influences and prejudices have taken on a more
dominant role in how and where the residents travel.
The next chapter in the dissertation will discuss the study areas in greater detail. General
descriptions of the cities and/or towns in which the six study neighborhoods are located, as well
as objective built environment data on the neighborhoods themselves, will be provided. The
rationale for the selection of the areas and neighborhoods for this research will also be explained.

66

Chapter IV - Study Neighborhoods
Introduction
This chapter delves into the details of the study area and the specific neighborhoods
included in the research. Beginning with a description of the larger general region, the Detroit
Region and the SEMCOG jurisdiction, the chapter progressively focuses in on the five
cities/towns included in the study. The discussion then turns to the characteristics of the very
neighborhoods within the selected cities/towns that are part of the analysis. Appendix A shows
the comparison of some socio-economic and demographic variables between the City/Town that
each study neighborhood is located in, the Census tracts that encompass the study neighborhood
and finally the study neighborhood itself. In this way the characteristics of the residents in each
neighborhood will be viewable in light of the larger geographic region that they live in. In all,
six neighborhoods are included in this study, which are then grouped in twos to form three
distinct neighborhood typologies on which the analyses are conducted. This chapter describes
and details the six neighborhoods that are the basis of this research.

The Study Region
The City of Detroit is nestled within Wayne County. The urban area is delineated by the
Tri-county Region of Wayne, Oakland and Macomb. A larger reference area, the DetroitWarren-Livonia Metropolitan Statistical Area, adds the counties of Lapeer, Livingston and St.
Clair to the Tri-County area. The Census Bureau designates the Detroit-Ann Arbor-Flint area as
the Consolidated Statistical Area (Figure 8). The CMSA adds the counties of Genesee, Monroe
and Washtenaw to the MSA. Lenawee was originally included in the CMSA, however, the
67

Census dropped it from the CMSA in the year 2000. The Southeast Michigan Council of Governments (SEMCOG) is a
regional governmental entity that has jurisdiction over Monroe, Washtenaw, Wayne, Livingston, Oakland, Macomb and St. Clair
counties. The SEMCOG region is slightly smaller than the CMSA region, and is a regional association of governments that aims to
solve the issues and challenges of the Southeast area of Michigan.
Figure 7: The Detroit Tri-County Area

68

Figure 8: The Detroit-Ann Arbor-Flint CMSA

Figure 9: The Detroit Region/SEMCOG Jurisdiction

69

The six neighborhoods chosen (Figure 32) for this study fall within the jurisdiction of the
Southeast Michigan Council of Governments (SEMCOG), which is referenced throughout the
rest of this dissertation as the Detroit region. The 2000 Census data and objective land use
neighborhood data (street networks and land uses) were used in the original neighborhood
selection process. Traditional travel behavior studies have focused on generally robust and
dynamic urban neighborhoods—characterized by higher densities, a mix of land uses, and high
connectivity—and compared them to newer, suburban neighborhoods characterized by lower
density, predominantly single-use neighborhoods with curvilinear street patterns.
The built environments selected for comparisons were generally upscale and dynamic
and little focus was placed on trying to capture travel behavior among residents of lower income
neighborhoods, which might be characterized by higher densities, mixed land uses, and
connected street networks, but that might also be experiencing disinvestment and decline. Within
the existing built environment and travel behavior literature, the relationship between socioeconomic conditions (and particularly in neighborhoods experiencing urban decline and
disinvestment), urban form and travel are recognized as underrepresented in urban research
(USDHHS, 2000; Day, 2003; 2006; Vojnovic, 2006; Vojnovic et al., 2006; Vojnovic et al., 2013;
Vojnovic et al, forthcoming; LeDoux and Vojnovic, 2013).
Four of the neighborhoods in this research focus on the traditional urban/suburban
comparisons of generally wealthy and robust communities. Added to the mix are two lower
income urban neighborhoods (higher density, mixed land use, and highly connected) that are
experiencing disinvestment and decline. The two higher density and dynamic suburban
neighborhoods are selected from the cities of Ann Arbor and Birmingham, the two lower density,
and wealthy, suburban neighborhoods are selected from Bloomfield Hills and West Bloomfield,

70

while the two poorer, higher density urban neighborhoods, experiencing disinvestment and
decline, are selected from lower eastside Detroit.
Figure 10: Population densities in the SEMCOG Region
Population
Density, 2010
Persons per Acre
Less than 2.0
2.0 – 4.9
5.0 – 9.9
10.0 – 14.9
15.0 or more

Source: www.semcog.org
* The yellow, orange and red arcs are drawn to show the decreasing levels of population density as visible
on this map.

Figure 10 shows a map of the population densities within the SEMCOG region. The inner
city Detroit neighborhoods were chosen for their built environments characteristics, high
densities, grid street patterns (for high connectivity) and mixed land uses. This area also had a
high concentration of black and lower income residents that either could not afford to move out
of the city with the rest of the population exodus or they were prevented from moving due to
racial discrimination practices. The urban Detroit neighborhoods are the two higher density
mixed land use and connected neighborhoods experiencing disinvestment and decline. Moving
out into the suburbs, Ann Arbor and Birmingham are chosen for their relatively high densities,
mixed land uses and generally grid street patterns. These neighborhoods, while higher density,
maintain higher socio-economic status. Bloomfield Hills and West Bloomfield are the
71

neighborhoods in the last suburban typology grouping. They were chosen because they are
representative of the low density suburbs that upper income groups move to who desire secluded
and homogenous residential enclaves. These jurisdictions are characterized by low densities,
curvilinear street patterns and mostly single land use neighborhoods and districts.
What follows is a description of each of the neighborhoods included in the study, along
with their built environment objective characteristics. For the purposes of the analyses
performed, the six neighborhoods were divided into three typologies, each consisting of two
neighborhoods. The High Density Urban (HDU) typology, characterized by relatively high
connectivity and mixed land uses, includes two neighborhoods from the City of Detroit. The
High Density Suburban (HDS) typology, also characterized by relatively high densities, high
connectivity and mixed land uses, includes a neighborhood from Ann Arbor and a neighborhood
from Birmingham. The last typology is the Low Density Suburban (LDS) typology,
neighborhoods that are characterized by low densities, low connectivity and generally single-use
zoning, and includes a neighborhood from Bloomfield Hills and a neighborhood from West
Bloomfield. For this chapter however, each neighborhood is discussed individually so as to give
a detailed look at each of the six neighborhoods and how these characteristics prompted and
facilitated the groupings into the three typologies.

72

The Inner-City Detroit Neighborhoods
Built Environment: Detroit
The two inner city Detroit neighborhoods depict a built environment that is typical of city
th

th

layout of the late 19 and early 20 centuries. The street network is of a grid pattern, the
density of structures is high, and there are mixed land uses scattered throughout the City. Not
only the density of structures, but the population density is also high for a Michigan context, at
5,135 people per square mile for Detroit as a whole in 2010 (US Census, 2010). The “higher”
density is relative to the built environment of the region, so the Detroit urban neighborhoods
have a higher density than that of the areas surrounding the City. Connectivity, as measured by
the number of 4-way intersections, is also high and that can be expected due to the gridiron street
network.
Vacant homes are a big issue in the City, with 23% of the housing lying vacant in 2010,
and the rest of the dwellings are evenly split between owner and renter occupied housing
(SEMCOG, 2010). Home values are low in the City, with median home values in 2010 being
$80,400 (a decline of 3% in the past decade) and median home rent in 2010 is $747 (an increase
of 16.5% in the last decade) (US Census, 2010). A look at the 2008 land use information shows
41% of the land being used for single family residences (with an additional 2% being used for
multi-family units) and 31% being used for transportation and utility uses. Other land uses have
much smaller shares, and includes some 8% of land in industrial uses, and 6% each in
institutional and park and recreation uses (SEMCOG, 2010).

73

Socio-Economic and Demographic Characteristics: Detroit
The 2010 Census showed that the City lost a quarter of its population within the last
decade. A look at the absolute numbers show that the black population decreased by about
185,000, compared to a decrease of about 44,000 white residents. The black population has also
started to move out of the City. Nevertheless, over 82% of the population in the City of Detroit
was black in 2010, while just under 8% was white, showing a disproportionately high racial
concentration of African Americans in the City (US Census 2010). A look at the educational
attainment, from the 5 year American Community Survey in 2010, shows that the City has a
small percentage—some 12% of the population—that has a bachelor’s degree or higher (with
only 4.6% having a graduate degree or higher), while 34% of the residents have only graduated
high school, and another 23% have not graduated high school (US Census, 2010). This depiction
of a large, poorly educated population is reflected in incomes as well. Median household income
and per capita income decreased between 2000 and 2010 by 26.6% and 21.8% respectively.
Consequently, the number of households and individuals living in poverty increased by 2% (to
30.8% of households) and 6.2% (to 34.5% of individuals) within the past decade (US Census,
2010). These facts depict the hollowing out of the city, and especially by the educated and the
skilled. Middle and upper income groups have moved into the suburban areas, while the lower
income groups, who consist predominantly of black residents, have remained within the city
boundaries.

74

Travel Characteristics: Detroit
A note on travel characteristics that will be discussed in this section is warranted upfront.
The below mentioned characteristics are from the 5 year American Community Survey in 2010.
These numbers reflect the limited travel characteristics of journeys to work only and therefore
are geared to a limited section of the full array of travel behavior that residents engage in on a
daily basis. Similarly, the data itself is from a sample of the city-wide population and therefore
is crude and limiting in nature. All of the particular distinctions in neighborhood travel
characteristics are simply averaged out throughout the jurisdictional boundary. Detailed
information and nuances that are inherent at lower scales of analysis, such as a neighborhood, are
simply not available with this data.
A majority of the residents 16 years and over drove alone to work (73%), with an average
travel time to work of about 26 minutes (US Census 2010). The use of car or van pooling and
public transit decreased in the last decade, from 17% in 2000 to 11.5% in 2010. Only about 8%
of the total working population use public transit to get to work. The City has two forms of
public transit, an automated electric rail system (called the Detroit People Mover), and the much
more extensive bus system run by the Detroit Department of Transportation (DDOT). The
People Mover is an elevated rail line that encircles about 3 miles of the downtown Detroit area.
This system is built to move people around the downtown, rather than to move people between
home and work destinations. The bus system, on the other hand, is quite extensive. The DDOT
has been the principal transit provider in the City since the mid-1970s. It has 48 bus routes
serving Detroit and the surrounding communities (DDOT, 2010). Figure 12 shows the different
routes served by DDOT buses.

75

Neighborhood Specifics: Detroit
The study neighborhoods in this region are adjacent to each other and are located in the
lower eastside of the City (see figure 11). For the first neighborhood, Highway I94 forms the
northern border, Moran Street forms the western border, Kercheval Street is the southern border
and Bewick Street is to the East. Gratiot Avenue is a major street that cuts through this
neighborhood. The second neighborhood is bounded by I94 to the North, Philip Street to the
East, Kercheval Street to the South and Harding Street to the West. Within this second
neighborhood is a considerably large section of land that is industrial, as visible in the map of the
neighborhood. Table 3 compares the population density of the City to those of the census tracts
encompassing both the study neighborhoods for the 2000 and the 2010 Census years.
Figure 11: Detroit Study Neighborhood*

* The above map shows the general placement of the study neighborhoods within the City

76

Figure 12: Detroit Department of Transportation Bus Routes*

Source: Detroit Department of Transportation
* The lines show the prevalence of public transit in the City

The neighborhoods were initially selected using the 2000 Census figures, while the
survey collection was completed towards the end of 2008. Therefore, data for both the 2000 and
2010 Census years are provided. It is fair to assume the actual density figures are closer to the
2010 Census data than to the 2000 data. The 2010 population density of each of the two
neighborhoods was 4,162 and 3,644 people per square mile, showing a decline from the 2000

77

figures. Again, the Detroit 2 neighborhood has a large parcel of land dedicated to a Chrysler
plant and therefore the density figures of this neighborhood are lower than the densities of the
Detroit 1 neighborhood.
Table 3: Population density (per square mile) of Census Tracts Encompassing the Detroit
Study Neighborhoods
Average High Density
Detroit 1
Detroit 2
Urban Typology*
2000
2010
2000
2010
2000
2010
Density
6,896
4,162
5,627
3,644
6,314
3,928
* If the Chrysler plant area is excluded, the average Urban Typology density for 2000 and 2010
would be 6,941 and 4,227 people per square mile respectively.

Objective land use data from neighborhood field surveys show that there is a mix of land
uses-from residential, to commercial, to institutional—scattered throughout the neighborhoods.
Figures 13 and 14 show the land use maps for the neighborhoods. The maps are a visual
snapshot of the mix of land uses within each neighborhood. These maps also depict the relative
lot sizes and compactness of the neighborhoods.
The number of 4-way intersections within each of the neighborhoods is 377 and 294.
Since total length of road miles specifically within the neighborhoods is unavailable, the ratio of
about 94 and 74 (for each of the 2 neighborhoods) 4-way intersections per square mile can be
used to gauge the connectivity measures between neighborhoods. The HDU typology has the
highest number of 4-way intersections among our six study neighborhoods.
Lastly, the prevalence of public transit is shown through the number of bus stops within
each neighborhood. The Detroit 1 and Detroit 2 neighborhoods have 176 and 144 bus stops

78

respectively. These are scattered throughout the neighborhoods depicting an even distribution of
public transit within this area.
Figure 13: Detroit Neighborhood 1 – Land Use Map*

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

79

Figure 14: Detroit Neighborhood 2 – Land Use Map*

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

80

Figure 15(a): Industrial Deterioration in Detroit

15(b): A Residential Street in the Detroit Study Neighborhood

81

The Ann Arbor Neighborhood
Built Environment: Ann Arbor
According to the 2008 SEMCOG data, the City of Ann Arbor has an extensive mix of
land uses. Some 41.6% of the land is allocated to residential uses (out of which 35.3% is for
single family uses). Over 17% of the land is allotted to each government and institutional uses
and to transportation and utility uses. Finally, 8% of the land is in commercial use, while another
2% of the land is in industrial use (SEMCOG, 2010). Population density for the City in 2010
was at about 4,128 persons per square mile, which is considered relatively high for the Michigan
context (US Census, 2010).
There is an almost even split between the owner occupied and renter occupied housing
units within Ann Arbor, with the former group representing 42% and the latter group
representing 52% of the housing. Vacant houses account for only 5% of the housing stock
(SEMCOG, 2010). The robust housing market is also reflected in the median home values. The
median home value in 2010 was $240,400 while median rent was $946, with home values seeing
a slight increase of 2% and rents seeing a slight decrease of 3.4% within the past decade (US
Census, 2010). Figure 16 shows the map of the neighborhood within the City of Ann Arbor.

82

Figure 16: Ann Arbor Study Neighborhood*

* The above map shows the general placement of the study neighborhood within the City
Socio-Economic and Demographic Characteristics: Ann Arbor
The total population of Ann Arbor according to the 2010 Census was 113,934. This
represents a miniscule decline in numbers from the 2000 Census of about 0.1%. Within the last
Census decade (from 2000 to 2010), the percentage of non-Hispanic whites decreased from
72.8% to 70.4%, blacks decreased from 8.7% to 7.6%, and the Asian population saw the largest
change, an increase from 11.9% to 14.3% (Census, 2010). Ann Arbor depicts a relatively
diverse population and this can be attributed, in part, to the presence of the University of
83

Michigan within the City. The presence of the university is also evident in the educational
attainment of the population. An overwhelming majority of 71% have a Bachelor’s Degree or
higher (with 42% having a professional or Graduate degree). In contrast, just 3% of the
population did not graduate high school and 9% of the population were only high school
graduates (US Census-ACS, 2010).
The median household income and per capita income were $52,625 and $30,498
respectively, a decrease of 13% and 12% between the 2000 and 2010 Census. Consequently, the
number of persons and households living in poverty increased from 2000 to 2010, recorded as
20.2% and 17.7% respectively (US Census-ACS, 2010).

Travel Characteristics: Ann Arbor
Keeping in mind the limiting nature of the travel data from the American Community
Survey, the following reflects the travel to work commute patterns of the Ann Arbor population.
The number of people driving alone to work in Ann Arbor decreased by 4.3% between the 2000
and 2010 Census years and was recorded in 2010 at 58.4%. This group represented the majority
of the resident working population. The second largest group was those who walked to work, at
15.5% of the working population, and this group saw a miniscule decline of 0.3% between the
years 2000 and 2010. The numbers of those who used public transportation to get to work
increased by 2.3%, to 8.9% of the working population, the third largest group among Ann
Arbor’s working population (US Census-ACS, 2010). These figures show a good mix of travel
patterns being used by the resident population in Ann Arbor, although the automobile oriented
nature of Ann Arbor residents clearly persists. The closer proximity of work destinations when
84

compared to Detroit was evident in the median time taken to get to work, which was 18.8
minutes (US Census-ACS, 2010).
Ann Arbor’s public transport system includes an extensive bus system run by the Ann
Arbor Transportation Authority. The city does not have a rail transit system, except for the
intercity Amtrak line. The bus system saturates the study neighborhood with about 160 bus stops
within the 2 square mile neighborhood (AATA, 2012). Due to the relatively compact nature of
the city, coupled with the high connectivity and accessibility measures, destinations are closer
together and opportunities for travel via modes other than the automobile are extensive. Walking
and bicycling, as well as using public transit, are modes that are utilized by residents for their
daily travel needs. Overall, however, Ann Arbor is still generally an automobile oriented city.
Figure 17 is a map of The Ride – The transportation system of the Ann Arbor Transportation
Authority.

85

Figure 17: Ann Arbor Transportation Authority Bus Routes*

Source: The Ann Arbor Transportation Authority – The Ride
*The above map shows the prevalence of public transit in the City

86

Neighborhood Specifics: Ann Arbor
The neighborhood chosen for this study is a somewhat “L” shaped area, broadly bounded
by M-14 highway to the north, Maple Road to the west, Scio Church Road to the south and Main
Street and Ferndon Road to the east. This area encompasses the downtown region of the City as
well as some of the University of Michigan campus in the southern part of the neighborhood.
Looking at the population density figures, the City’s overall density of 4,128 people per square
mile is lower than that of the neighborhood, which stood at 5,427 in 2010. The main reason for
this would be that the main core of the City is included within the study neighborhood and this is
where a high concentration of residents are located.
Table 4: Population density (per square mile) of Census Tracts Encompassing the Ann
Arbor Study Neighborhood
Average High Density
Ann Arbor
Suburban Typology*
2000
2010
2000
2010
Density
5,535
5,427
4,696
4,723
* The average refers to the combined figures for both neighborhoods in this typology.

A field survey of the neighborhood revealed an abundant mix of land uses, as evident in
the Ann Arbor neighborhood land use map (Figure 18). The presence of the downtown area and
the University of Michigan campus adds to the land use mix. This neighborhood has, more or
less, a grid street pattern resulting in high connectivity and accessibility measures. The number
of 4-way intersections within this neighborhood is 148. So, comparatively, this neighborhood
has 62 4-way intersections per square mile, a drop from the figures for urban Detroit, but still a
relatively high neighborhood connectivity.

87

Figure 18: Ann Arbor Neighborhood - Land Use Map*

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

With regard to mass transit, the AATA has an extensive bus system as the principle form
of public transit in this neighborhood. This is revealed by the 160 bus stops that are located
within the Ann Arbor neighborhood boundary. The number of bus stops in the Ann Arbor
neighborhood is comparable to the urban Detroit numbers. Figures 19a and 19b show
streetscapes in the Ann Arbor study neighborhood.

88

Figure 19a: Ann Arbor Main Street

Figure 19b: A Residential Street in the Ann Arbor Study Neighborhood

89

The Birmingham Neighborhood
Built Environment: Birmingham
Birmingham is part of the inner ring suburbs of the Detroit region. It covers a small area
of 4.8 square miles, and has a relatively high population density, some 4,188 people per square
mile in 2010 (US Census, 2010). Street patterns are mostly gridiron, depicting a good measure
of connectivity and accessibility. A major north-south street, Woodward Avenue, cuts through
the City, with most non-residential uses concentrated on this main street. There is also a
significant retail and commercial concentration along sections of the east-west streets, including
Maple Road and Fourteen Mile Road. The small spatial extent of the City makes these
commercial and retail strips ‘accessible’ to the residents of Birmingham.
A look at the 2008 land use information from SEMCOG shows that single family
residences make up the majority of the land use, at over 52% of the total land use (with an
additional 1% in multi-family units). A quarter of the land is taken up by transportation and
utility uses, and institutional and park and recreation uses each take up a little over 7% of the
land use. In Birmingham, industrial uses make up a mere 1% of the total land, while commercial
uses make up less than 5% of the land use (SEMCOG, 2010).
Birmingham city’s housing is predominantly owner occupied (66%), with only 9% of the
housing lying vacant (US Census, 2010). Birmingham’s robust housing market is depicted in its
housing values, with the median home value and median rents at $369,200 and $1,191
respectively. Both of these values showed a decline of over 7% in the last decade (US CensusACS, 2010).

90

Figure 20: Birmingham Study Neighborhood Map*

*The above map shows the general placement of the study neighborhood within the City
Socio-Economic and Demographic Characteristics: Birmingham
The City of Birmingham had a population of 20,103 people according to the 2010
Census, an increase of 4.2% over the 2000 Census population figures. This population is
predominantly white (91%), while blacks and Asians make up 3% and 2.5% respectively of the
city’s population (US Census, 2010). Birmingham’s residents have a high level of educational
attainment, with 74% of the population having a bachelor’s degree or higher, while 39% of the
population has a graduate degree or higher. Only 6% of the population has just a high school
degree and 2% of the population has not graduated from high school (US Census-ACS, 2010).
As could be expected, the highly educated make-up of the population is also reflected in high

91

incomes, with the median household and per capita incomes being $101,529 and $69,151
respectively. The City’s predominantly white, highly educated and wealthy residents have few
people and households living in poverty, at 3.8% and 4.8% respectively (US Census-ACS,
2010).

Travel Characteristics: Birmingham
As is evident from the built environment description, the main non-residential land uses
are mostly along the major street, Woodward Avenue, and therefore there would be a very small
percentage of the population who works and lives close enough to this street to be able to walk to
work. There is no local rail public transport system, except for the inter-regional Amtrak station.
There is also an absence of an extensive public transportation system within the City. The
Suburban Mobility Authority for Regional Transportation (SMART) operates and runs a bus
system within the area, but a closer look shows that this bus system reaches only the very few
major roads within the City. Many people are thus relegated to drive to work despite the higher
densities and high accessibility in the City of Birmingham.
The study neighborhood in Birmingham has many non-residential uses clustered along
Woodward Avenue and this affects the travel patterns of the residents within the neighborhood.
Looking at the full array of travel patterns, hence not only limited to work purposes, would
depict a somewhat different travel behavior than what is depicted by the 5 year American
Community Survey data. Notwithstanding, a look at the 2010 Census data on travel work
patterns shows that a majority of the resident, working population drives alone to work (87.5%)
with an average travel time of 22.8 minutes. Over 5% of the working residents worked at home.
92

Those who carpooled made up 3.4% of the working population, those taking public transit to
work made up 0.2% of the working population and those who walked to work made up 2.5% of
the resident working population (US Census-ACS, 2010). These facts depict the lack of an
effective and extensive public transit system in a suburb, even a higher density suburb, where a
large segment of the employment is located outside of the jurisdiction.
Figure 21: SMART Bus Routes Map with Birmingham Inset*

Source: Smart Bus Official Website (smartbus.org)
*The above map shows the “SmartBus” routes, and the inset map shows the prevalence of public
transit in the City.
93

Neighborhood Specifics: Birmingham
The study neighborhood in Birmingham is bounded by Quarton Road to the north,
Buckingham Road to the south, Westwood Road to the west and Eton Road to the east. With
Woodward Avenue also cutting through the case study neighborhood, the street pattern,
combined with the close proximity to this major arterial street, forms a highly connected street
system with a mix of land uses in relatively close proximity. Knowing the 2010 City’s
population density to be at about 4,188 people per square mile, the density within the
neighborhood comes in a bit lower, at 3,828 people per square mile. The presence of the main
street, Woodward Avenue, and other non-residential strips within the neighborhood is the reason
for the slightly lower density, since land uses on and around this avenue are mostly all
commercial, retail and institutional. This is a case where the land use mix contributes to lower
residential densities.
Table 5: Population Density (per square mile) of Census Tracts Encompassing the
Birmingham Study Neighborhood
Average High Density
Birmingham
Suburban Typology*
2000
2010
2000
2010
Density
3,630
3,828
4,696
4,723
* The average refers to the combined figures for both neighborhoods in this typology.

A look at the land use map, Figure 22, shows that most of the mixed land uses within this
neighborhood and clustered along Woodward Avenue. There are some non-residential land uses
scattered through the neighborhood (albeit, at a much lower level than the urban Detroit
94

neighborhoods or even the Ann Arbor neighborhoods). Connectivity, as measured by the
number of 4-way intersections, is also relatively high in this neighborhood. There are 217, 4way intersections within the neighborhood, an average of about 54, 4-way intersections per
square mile. This is lower than the counts for the urban Detroit neighborhoods, but still much
higher than that for Ann Arbor, indicating that this region still has most of its road network in the
gridiron pattern, promoting high connectivity.
Figure 22: Birmingham Neighborhood – Land Use Map*

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

However, the neighborhood abundance in connectivity contrasts to its deficiency in
access to public transit (See Figure 21). The Birmingham neighborhood has only 61 bus stops
located within its boundaries. This number is significantly lower when compared to the number
of bus stops in the Detroit and the Ann Arbor neighborhoods. This lack of alternative travel
95

modes to the car is an important factor in shaping local travel patterns. Figures 23a and 23b
show typical streetscape views of this neighborhood.

Figure 23a: Downtown Birmingham

96

Figure 23b: Typical Birmingham Residential Street

97

The Bloomfield Hills Neighborhood
Built Environment: Bloomfield Hills
Moving to the low density suburbs, the City of Bloomfield Hills lies to the northwest of
Birmingham and has almost the same spatial extent as the City of Birmingham, about 5 square
miles, but it has less than one fifth of the population (3,869) of Birmingham. This results in the
low population density of 774 people per square mile in this city (US Census, 2010). In
Bloomfield Hills we also see a departure from the gridiron street network. Curvilinear streets
dominate the landscape, with fewer 4-way intersections apparent in this city when compared to
Detroit, Ann Arbor, or Birmingham.
Figure 24: Bloomfield Hills Study Neighborhood*

*The above map shows the general placement of the study neighborhood within the City
98

With respect to land uses, the SEMCOG 2008 data shows that over 61% of the land is
used for single family housing, and there are no multi-family units in the City (SEMCOG, 2010).
Two major streets cut through Bloomfield Hills, with Woodward Avenue running north-south
and Long Lake Road running east-west. The commercial uses in Bloomfield Hills make up over
11% of the land area and are concentrated along these two major streets. The City houses the
Cranbrook Institute of Science within the southern portion of its jurisdiction. Cranbrook is a
th

renowned educational institution that houses a Pre-K through 12 grade prep school, an art
museum, a graduate studies school, a science institute, and all within a 319-plus acre campus.
This institution is the main reason that over 13% of the City’s land use is classified as
Government and Institutional. Transportation and utility make up some 12% of the City’s land
use. There is no industrial land in Bloomfield Hills and park and recreation space make up a
miniscule 0.1% of land within this jurisdiction (SEMCOG, 2010). Probably the low densities,
and hence large private lots, give residents enough private open space around their residences to
not require any proximate public lands.
According to the 2010 Census figures, the vast majority of housing in Bloomfield Hills
(81%) is owner occupied, while only 9% is renter occupied and about 10% of the housing stock
in the City is vacant (US Census, 2010). The robust nature of the housing market is reflected in
the house values. The median home value and gross rent is $715,300 and $747 respectively.
About 37% of the owner occupied housing units are valued between $500,000 and $1,000,000,
and another 32% of the housing units are valued at over $1,000,000 (US Census-ACS, 2010).

99

Figure 25: Bloomfield Hills Neighborhood – Land Use Map*

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

Socio-Economic and Demographic Characteristics: Bloomfield Hills
The City of Bloomfield Hills had a population of 3,869 according to the 2010 Census, a
drop of about 1.8% from the previous decade. This City is predominantly white (over 86%),
while blacks and Asians made up 4% and 6.7% of the population respectively (US Census,
2010). Over 69% of the population has a bachelor’s degree or higher (with 37.7% having a
graduate degree or higher). Only 2% of the population did not graduate from high school. Those
graduating high school and those having an associate’s degree made up 15% and 14% of the
population respectively (US Census, 2010). The median household and per capita incomes were

100

$133,370 and $89,538, much higher than that for Birmingham City, a jurisdiction that maintains
a more polarized educational achievement (US Census-ACS, 2010).

Travel Characteristics: Bloomfield Hills
Travel patterns to work for the residents in the City are dominated by driving, with the
average travel time to work in 2010 being about 23 minutes (US Census-ACS, 2010). Over 77%
of the resident working population drives alone to work and another 11.5% carpools. In
Bloomfield Hills, 10% of the population works from home, which in total makes up over 98.5%
of the resident working population. No one takes public transportation and only 1% of the
working population walked to work (US Census-ACS, 2010). What is absolutely clear is the
total dependence on the automobile for travel purposes. Over 10% of the resident population
travels to Detroit for work, while less than 6% of the workers in the City are actually Detroit
residents.
There is no public transportation, rail or bus, within the City. The SMART bus system
just reaches the southern edge of the City (see Figure 25). Lack of connectivity and the
relegation of the relatively few and scattered commercial uses to the two main streets make
walking for daily travel needs extremely difficult. In addition, the lack of mass transit makes
public transportation non-existent.

101

Figure 26: SMART Bus Routes Map with Bloomfield Hills Inset*

*The above map shows the “SmartBus” routes, and the inset map shows the prevalence (rather
the lack of) of public transit in the City.

102

Neighborhood Specifics: Bloomfield Hills
The Bloomfield Hills study neighborhood is bounded by Square Lake Road to the north,
Lone Pine Road to the south, Opdyke Road to the east and Franklin Road to west. The
neighborhood is typical of the City, with the two main streets, Woodward Avenue and Long
Lake Road passing through the eastern and southern section of the neighborhood respectively.
Looking at the population density of the study neighborhood, it is higher than that of the City of
Bloomfield Hills, with the 2010 neighborhood density being 1,141 people per square mile
(compared to 774 people per square mile for the City). This can largely be attributed to the
presence of the Cranbrook Institute in Bloomfield Hills, which is only partially included towards
the southern border of the neighborhood.
Table 6: Population density (per square mile) of Census Tracts Encompassing the
Bloomfield Hills Study Neighborhood
Average Low Density
Bloomfield Hills
Suburban Typology*
2000
2010
2000
2010
Density
1,175
1,141
1,797
1,711
* The average refers to the combined figures for both neighborhoods in this typology.

Connectivity measures are low; the lowest within this built environment typology. The
neighborhood has 33, 4-way intersections, an average of 8 intersections per square mile. This
number is very low compared to the other two neighborhood typologies discussed earlier. The
predominance of curvilinear streets reduces connectivity. The neighborhood is almost
predominantly residential, making accessibility to out-of-home destinations weak. With very
few destinations and other non-residential uses in close proximity, coupled with the curvilinear

103

street network, driving is the mode of travel that the residents are forced to take. Adding to the
predominance of private automobile use is the fact that there are only 15 bus stops within this
neighborhood. As mentioned earlier, the SMART bus system reaches only the edges of the City
and ultimately provides a weak alternative to private vehicle travel. Figures 27a and 27b depict
typical streetscape views of this neighborhood.

Figure 27a: Typical Bloomfield Hills Residential Street

104

Figure 27b: Typical Bloomfield Hills Residential Neighborhood

105

The West Bloomfield Neighborhood
Built Environment: West Bloomfield
The Charter Township of West Bloomfield lies to the west of the City of Bloomfield
Hills. The Township is larger in spatial terms, covering about 31.3 square miles (US Census,
2010). Again, this is a second ring suburb of the Detroit Region and typical of its age, it lacks a
grid street pattern. The streets are generally curvilinear, leading to low connectivity. The
population according to the 2010 Census was 64,690, resulting in a density of about 2,066 people
per square mile (US Census, 2010).
The 2008 land use information from SEMCOG indicates that the Township had 56.4% of
its land in single family residences (with only an additional 2% of land devoted to multi-family
units). There are quite a few lakes in the northern section of the Township, and therefore water
makes up about 14.7% of the land use. Transportation and utility uses make up 13.4% of the
land, while industrial uses make up a mere 0.6% and commercial uses another 3.7% of the land.
Parks and open space make up about 5% of the land, with quite a few golf courses located within
this township. Government and institutional uses make up another 4% of the total land in the
Township, with many churches and schools located within the jurisdiction (SEMCOG, 2010).
Non-residential uses are clustered along the main streets of Orchard Lake Road (north-south
street), Walnut Lake Road, and West Maple Road (both east-west streets).

106

Figure 28: West Bloomfield Study Neighborhood*

*The above map shows the general placement of the study neighborhood within the Township

Housing conditions depict a robust community, with over 77% of the housing units being
owner occupied, 16% being renter occupied, and only 7% lying vacant in 2010 (US Census,
2010). Median home and rent values reflect the socio-economic composition of this jurisdiction,
with 2010 median home value being $291,200 and 2010 median rent being $1,425 (US CensusACS, 2010). Home values are lower than in Bloomfield Hills, however the rents are higher in

107

West Bloomfield. The largest percentage of owner occupied homes, over 30% of the housing,
maintains values between $300,000 and $500,000 (US Census-ACS, 2010).

Socio-Economic and Demographic Characteristics: West Bloomfield
The 2010 Census put the population of the Township at 64,690, showing a minor decline
of about 0.3% people within the last decade. The Township is still largely populated by white
residents, making up 76.5% of the total population (a drop of about 6.7% over the previous
decade). The black population, on the other hand, makes up about 11.3% of the total residents
(showing an increase of 6.2% since the previous decade). The Asian population has been
relatively stable between the two decades, at about 8.4% of the total residents (US Census,
2010). These figures show a steady influx of Black residents into the Township between 2000
and 2010, allowing a growing racial mix in this traditionally white community.
Looking at the educational attainment of the residents in West Bloomfield, some 55% of
the population has a bachelor’s degree or higher (with 25.6% having a graduate degree or
higher). About 6% of the population has not graduated high school and another 15% are only
high school graduates (US Census, 2010). Again, we see here a lower polarization among the
resident population when compared to the City of Bloomfield Hills, and that can be attributed
partly to the fact that we are now looking at the Charter Township of West Bloomfield, an area
much larger than Bloomfield Hills City. The educational attainment data is reflected in the
median incomes as well. The median household and per capita income was $97,004 and $47,201
respectively. In this township, 4% of the population and 4.2% of the households lived in poverty
(US Census-ACS, 2010).
108

Travel Characteristics: West Bloomfield
West Bloomfield Township, like the City of Bloomfield Hills, has no effective public
transportation system. There is no rail transit service and the SMART bus system reaches only
the peripheral areas of the Township (see Figure 29). The built environment characteristics
encourage travel by automobiles more than any other means. Census 2010 figures indicate that
over 88% of the resident working population in the Township drove alone to work, while only
5.8% carpooled. The average travel time to work was about 28 minutes. Another 4.5% of the
townships working population worked from home. Those taking public transportation to work
made up a mere 0.2% and those walking to work consisted of 0.8% of the total resident working
population (US Census-ACS, 2010).
The 2010 5-year American Community Survey data also shows that over 10% of the
resident working population traveled to Detroit for work purposes, whereas over 6% of the
workers in the Township came in from Detroit to work in the Township of West Bloomfield.
Overall, the local travel characteristics show a dependence on automobile travel for a majority of
work travel purposes. Public transit is not available and walking as a means of travel is difficult
due to the lack of connectivity in the street network and a lack of access to proximate work
destinations.

109

Figure 29: SMART Bus Routes Map with West Bloomfield Inset

*The above map shows the “SmartBus” routes, and the inset map shows the prevalence (rather
the lack of) of public transit in the Township.

110

Neighborhood Specifics: West Bloomfield
The study neighborhood within the Township is bordered by Pontiac Road to the north,
Maple Road to the south, Drake Road to the west and Orchard Lake Road to the east. A large
part of the northern section of the neighborhood has open space and that is due to the presence of
a number of golf courses in the area. The 2010 population density for the Township, at 2,066
people per square mile, is a bit lower than that of the neighborhood, which stood at 2,686 people
per square mile. This is largely attributable to the larger amounts of open space within the
Township, which is only partially included within the study neighborhood. The population
density within this neighborhood is higher than that of Bloomfield Hills, however, these
population density figures are much lower than those in the other two neighborhood typologies,
the higher density urban and higher density suburban neighborhoods.
Table 7: Population density (per square mile) of Census Tracts Encompassing the West
Bloomfield Study Neighborhood
Average Low Density
West Bloomfield
Suburban Typology*
2000
2010
2000
2010
Density
2,860
2,686
1,797
1,711
* The average refers to the combined figures for both neighborhoods in this typology.

This neighborhood also depicts spatial structure patterns that are similar to Bloomfield
Hills. A predominance of curvilinear streets reflects low connectivity measures. There are only
thirty, four-way intersections within the neighborhood, an average of 8 per square mile, the same
as that of Bloomfield Hills. These are typical built environment characteristics of low density
suburbs, where gridiron street patterns are almost non-existent. The land use map below (Figure

111

30) also shows the preponderance of residential land uses within this study neighborhood, with
only a handful of “personal service” land uses at the southern edge of the neighborhood. Similar
to Bloomfield Hills, this neighborhood also has 17 bus stops within its boundaries, escalating the
effects of the low connectivity, low accessibility measures that encourage automobile use for
travel. Figures 31a and 31b depict typical streetscape views of this neighborhood.
Figure 30: West Bloomfield Neighborhood – Land Use Map

*The above map shows the prevalence of various land uses within the study neighborhood
(depicted by the different colors)

112

Figure 31a: Typical West Bloomfield Residential Neighborhood

Figure 31b: Typical West Bloomfield Residential Street

113

Commentary
The above description of the study area and neighborhoods show a variation in the socioeconomic, demographic and built environments captured in the six neighborhoods. All the
neighborhoods are approximately 4 square miles in area, and have been chosen so that an
effective analysis can be completed, distinguishing between typical higher density and lower
density urban forms, but also adding a higher density urban built environment characterized by
disinvestment and decline. As mentioned earlier, traditional analyses on travel behavior has
generally focused on robust communities. The inclusion of a neighborhood that is typical of a
high density, high connectivity and high accessibility area, but experiencing severe
disinvestment, has not been the focus of travel behavior studies. The six neighborhoods included
in this research will cover an important gap in the travel behavior research and will add insight
into the travel behavior of residents living in neighborhoods of rapid decline.
The next chapter will discuss the data gathered and analyzed in this dissertation.
Although a general description of the communities and the study neighborhoods have been
provided in this chapter, a more detailed review of the data will be provided in Chapter V. The
chapter will also develop the research objectives, questions, hypotheses and will include the
methodology of the quantitative analysis into travel behavior and environmental burdens
(gasoline consumption and CO emissions).

114

Chapter V - Research Hypotheses, Data, and Methodology
Introduction
This chapter begins by laying out the research hypotheses for this study. In particular,
travel behavior and the resulting environmental burdens (measured by CO emissions and fuel
consumption) are being tested to see if they have any significant associations with the socioeconomic characteristics of residents in study neighborhoods that have differing built
environment characteristics. Following the hypotheses, the next section in this chapter focuses
on the data collected for this study. The data is a combination of primary data, collected through
a mail survey, as well as secondary data from the Michigan Secretary of State and the
Environmental Protection Agency. The last section in the chapter discusses, in detail, the
methodology that was adopted to test the hypotheses and to enrich the understanding of what this
rich dataset has to offer to this line of research. The methodology ranges from Discriminant
Analysis to see the differ impacts that the built environment has on travel behavior and
environmental burdens, to Cluster Analyses to objectively discern the clusters/groups that the
data was lending itself to. As researchers, we have a predetermined idea of what the data might
indicate. A Cluster Analysis gives strength to that thought, in the sense that it adds to the detail
that helps form certain groups of observations. The last set of analyses performed were the
Regressions, which are conducted to test and answer the hypotheses and research question.

115

This dissertation explores the following general hypothesis:
Socio-economic characteristics and built environment conditions have a significant
impact on environmental burdens as related to gasoline consumption and pollutant
emissions
The analyses span six Detroit region neighborhoods. Research at the neighborhood level
would fill a void in current studies where density and socio-economic characteristics have not
been combined to tease out the nuances of travel behavior as related to gasoline consumption and
pollutant emissions. The unique aspect here is the marrying of the built environment, socioeconomic factors and travel patterns at the neighborhood scale, along with the associated fuel
consumption and vehicular emissions. The following research question would aid in
understanding and assessing the main hypothesis. The research question, in turn, is presented
with sub-hypotheses.

Research Question
Do socio-economic characteristics of residents, by neighborhood type, significantly impact
travel patterns, gasoline consumption and vehicular emissions?
Objective
To explore the relationships between socio-economic characteristics and vehicular trips,
along with associated gasoline consumption and pollutant emissions, for different trip purposes
by neighborhood typology. For purposes of this research question, neighborhood types are
categorized into higher density urban (HDU), higher density suburban (HDS), and lower density
116

suburban (LDS) neighborhood typologies, and trip purposes are categorized into work, nonwork, and all trips combined. The categorization of the six neighborhoods into three
‘neighborhood typologies’ will be discussed in the Methodology section of this chapter.

117

Hypotheses
a)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics will travel more often by private
vehicles (measured as a percentage of total trips traveled by car) than
their counterparts with lower socio-economic characteristics.

b)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics travel less frequently by public

Mode of Travel

transit (measured as a percentage of total trips traveled by public transit)
than their counterparts with lower socio-economic characteristics.
c)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics travel less often by walking/biking
(measured as a percentage of total trips traveled by walking/biking) than
their counterparts with lower socio-economic characteristics.

d)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics will travel more miles annually
by private vehicles (thereby consume more fuel and emit more
pollutants) than their counterparts with lower socio-economic

Total Miles

characteristics.

Traveled

118

e)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics will travel fewer miles annually
by public transit than their counterparts with lower socio-economic
characteristics.

f)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics will travel fewer miles annually
by walking/biking than their counterparts with lower socio-economic
characteristics.

g)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics own newer vehicles than their
counterparts with lower socio-economic characteristics.

Vehicle

h)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics own more vehicles per household

Ownership &

than their counterparts with lower socio-economic characteristics.

Characteristics
i)

Within each neighborhood type (HDU, HDS, and LDS), residents with
higher socio-economic characteristics own vehicles with larger engines
than their counterparts with lower socio-economic characteristics.

119

OLS Regression Model:
DEPENDENT VARIABLE = α + β1PERSONAL/HOUSEHOLD INCOME + β2 EDUCATION
+ β3 OCCUPATION + ε
The Dependent Variable would be Travel Mode (Hypotheses a, b, c), Total Miles
Traveled (Hypotheses d, e, f), Vehicle Age (Hypothesis g), Number of Vehicles per Household
(Hypotheses h), and Vehicle Engine Size (Hypotheses i).
The Predictors would be Socio Economic characteristics (Personal and Household
Income, Education, Occupation), and other characteristics affecting the cultural and social
attitudes (age, race, number of dependents).

120

Data
A stratified random mail survey collected data on travel behavior, personal socioeconomic and demographic variables, and automobile ownership data. It was decided to send
out two surveys per household that was selected into the sample. Introductory prompts were sent
out to the sample households in early 2008 introducing the project and alerting the households
that they will be receiving these surveys. Thereafter the surveys were mailed out. Two sets of
reminder prompts were sent out to the households that had not yet responded to the surveys, two
weeks after the surveys were mailed out, and then two weeks thereafter.
In total, there were 1,191 surveys collected. By neighborhood breakdown, there were
128 survey respondents from the Detroit #1 neighborhood, 158 survey respondents from the
Detroit #2 neighborhood, 297 respondents from Ann Arbor, 196 respondents from Birmingham,
211 survey respondents from Bloomfield Hills, and 201 respondents from West Bloomfield. The
mail survey included detailed questions on travel, such as frequency and purpose of trips, travel
mode, and destinations, for a full array of trips, including work, shopping, personal services, and
leisure. Respondents were asked to report travel behavior over a typical week while considering
seasonal distinctions, for instance, reflecting on differences between winter versus summer
travel. The survey response rate was 20%, which is considered good for a mail-out survey
administered to the general population, and particularly for a travel behavior survey (Sommer &
Sommer, 1997; Zimowski et al., 1997). Figure 32, shown below, is a map with the six study
sites.

121

Figure 32: Study Area

The next step was verifying and completing any incomplete information supplied by the
residents on their vehicle ownership and characteristics. This was done by matching the records
from each household to a database of vehicle ownership in Michigan, obtained through the
Michigan Secretary of State. This dataset contained the make, model, year, and body type for
each vehicle registered to each of the addresses in the sample. If any of the information in the
surveys was incomplete, it was matched to the vehicle registration dataset. Only missing
information in any survey was completed through this verification process. So, if a respondent
to a survey said he/she used a 2002 Toyota, and did not provide the model of the vehicle, and if
the vehicle registration dataset had a 2002 Toyota Corolla registered to this address, then the
survey dataset was updated and the model of “Corolla” was added to the dataset. In essence, if
one piece of the puzzle was missing, it was updated. However, if the respondent said they used a

122

different type of vehicle altogether, then the survey response was used (and not the vehicle
registration data).
The most number of incomplete entries were from the Detroit neighborhoods. It was also
in the two Detroit neighborhoods where many of the vehicles that were actually being used were
not registered to respondents’ addresses, and therefore not on the Michigan Secretary of State
vehicle registration database. Lower income residents in these neighborhoods were extensively
using cars that belonged to relatives, or friends, and registered to some other address.
Once the data on vehicles was confirmed and completed, the characteristics of each
vehicle mentioned in the survey was matched to a database by the Environmental Protection
Agency (EPA) wherein emissions and fuel economy was obtained. This database is the Green
Vehicle Guide, maintained by the Office of Transportation and Air Quality, a division of the
EPA. This database is essentially the same as that of a ‘window sticker’ that would get put on a
vehicle if it would be up for sale. All this information is specific to the make, model, year, and
State that the vehicle is registered in. The information is broken down into three parts: vehicle
specifications, environmental information, and other information. Appendix B in the appendices
shows an example of the information provided for a 2005 Lexus ES 330, registered in Michigan.
Fuel economy information, if not available from the EPA database, was retrieved by performing
a similar vehicle specific search on the fueleconomy.gov website, maintained by the US
Department of Energy. Appendix C in the appendices shows an example of the information
provided for a 2005 Lexus ES 330 vehicle registered in Michigan.
Collecting the physical attributes of the neighborhood sites was the next step in the data
collection process. Field trips were made to the study sites to collect information on built

123

environment objective data, such as the land use (residential, commercial, retail, industrial) for
each structure in the study neighborhood, residential building type (single family, apartments,
townhomes, industrial), and an assessment of abandoned properties (these were those that were
visibly abandoned, with boarded up windows and doors). During this period, mid- to late-2008,
the problem with decreasing real estate prices, and increasing foreclosures, abandonment and
vacancies was extensive. During the course of survey collection, over 900 homes to which the
surveys were sent had been vacated. Field trips and returned survey packets and prompts
corroborated that over 75% of the vacated and abandoned houses were from the urban Detroit
neighborhoods.
The study neighborhoods were drawn up in AutoCAD and ArcGIS. The AutoCAD
images helped with the portrayal of the built environment (the urban form), while the ArcGIS
maps were used to calculate connectivity and document the land use mix. Connectivity was
calculated by counting the number of 4-way intersections within the neighborhood boundaries.
Land use mix is calculated as the number of non-residential structures (commercial, retail,
institutional and industrial) within the neighborhood boundaries. Lastly, the data on the location
and number of bus-stops within each neighborhood was also collected. For the urban Detroit and
the Ann Arbor neighborhoods, the coordinate locations for the bus-stops were provided by the
Detroit Department of Transportation (DDOT) and the Ann Arbor Transportation Agency
(AATA) respectively. For the other three suburban neighborhoods of Birmingham, Bloomfield
Hills, and West Bloomfield, the locations were taken from Google maps.

124

Methodology
This section details the data used in the analyses and the different analyses performed to
explore the neighborhood travel behavior. The basic database has demographic, vehicle, and
travel behavior data for respondents in the six neighborhoods, which are categorized into three
different neighborhood typologies—higher density urban (HDU), higher density suburban
(HDS), and lower density suburban (LDS). A list of those variables can be found in the Code
Book (see Appendix D).
One set of data for this analysis involves the built environment objective data, collected
from neighborhood land use and site surveys. Information on land uses (residential, commercial,
retail, and industrial), building types (single family, duplexes, apartments, and factories), and
their concentration within the neighborhoods were collected during the site surveys. This data
was also coupled with connectivity data, the concentration of 4-way intersections within the
neighborhoods, and this built environment objective data all went into defining the three
neighborhood typologies (see table 8 and 9).

Rationale for the Neighborhood Typology Grouping
With respect to basic community characteristics that distinguish these neighborhood
groupings, a number of distinction are worth noting. The two HDU neighborhoods are adjacent
to each other and have the same built environment characteristics. As shown in table 8, the two
Detroit neighborhoods have similar densities, connectivity and accessibility characteristics. It is
also important to note that a big section of land in the Detroit 2 neighborhood is one that was
occupied by the Chrysler Plant. Therefore, density figures, connectivity and accessibility
measures are lower compared to the Detroit 1 neighborhood. An important difference between
125

the Detroit urban neighborhoods and the higher density suburban neighborhoods, of course, is
that the Detroit neighborhoods are lower income and predominately African American, and these
neighborhoods are experiencing severe disinvestment and decline. These two Detroit
neighborhoods, while similar to each other, are very different socio-economically,
demographically, and racially from any other neighborhood grouping, and were grouped together
to form the ‘higher density urban’ (HDU) group. Table 9 shows the built environment
characteristics of the combined neighborhoods.

Table 8: Built Environment characteristics by Neighborhood
Detroit 1 Detroit
Ann
Birmingham Bloomfield
West
2
Arbor
Hills
Bloomfield
Land Use Characteristics (numbers per square mile)
Single Family
Detached Homes
1542.1
1024.7 1575.4
1416.3
424
716.3
Non-residential
structures
77.7
38.1
57.3
54.8
22.6
1.6
Connectivity (numbers per square mile)
4-Way
77.7
49.2
31.1
41.4
5.2
5.2
Intersections
Population Density (numbers per square mile)
Density 2010
4162
3644
5427
3828
1141
2686
Density 2008*
4708.8
4040.6 5448.6
3788.4
1134.2
2720.8
*Estimated density based on decade change values

Table 9: Built Environment characteristics by Neighborhood Type
HDU
HDS
Land Use Characteristics (numbers per square mile)
Single Family Detached Homes
1283.4
1495.9
Non-residential structures
57.9
56.05
Connectivity (numbers per square mile)
4-Way Intersections
63.5
36.3
Population Density (numbers per square mile)
Density 2010
3903
4627.5
Density 2008*
4374.7
4618.5
*Estimated density based on decade change values
126

LDS
570.2
12.1
5.2
1913.5
1927.5

Moving outward, the Ann Arbor neighborhood and the Birmingham neighborhood have
similar built environment characteristics as the HDU neighborhoods, but these are
neighborhoods in more dynamic and robust communities. Although they are not adjacent to
each other, they are in suburban areas with strong urban cores and they have similar density
characteristics as the inner city Detroit neighborhoods. These densities would be considered
relatively high in the Michigan context. They also have similar gridiron street patterns and hence
similar connectivity characteristics. These neighborhoods are also characterized by mixed land
uses, particularly along major roads and the central area of these neighborhoods, their robust and
dynamic urban cores. Ann Arbor has more of a mixed use visible because adjoining and to the
south of the downtown area is the University of Michigan campus and this adds to the diversity
of uses within the neighborhood. Even though the Birmingham neighborhood does not have a
university within its boundary, it has built environment characteristics that are similar to Ann
Arbor, a higher density urban core with a robust land use mix. Both of these neighborhoods are
also generally upper-income and contain a population that is largely white. Given their similar
built environment characteristics, which are relatively unique development patterns within
Michigan, the Ann Arbor and the Birmingham neighborhoods were combined to represent the
‘higher density suburban’ (HDS) group.
The Bloomfield Hills and the West Bloomfield neighborhoods are situated in adjacent
municipalities and have very similar built environment characteristics. These neighborhoods
have low densities, road patterns that are curvilinear and these neighborhoods are almost
completely single use zoned. Socio-economically, these areas are generally wealthy and
predominantly White. Given their built environment characteristics, these two neighborhoods
127

were combined to represent the ‘low density suburban’ (LDS) group. The resultant study area
consists of three neighborhood ‘types’ that have a large enough population base for effective
analyses on travel behavior and environmental burdens to be conducted.

Quantitative Analyses
In this intra neighborhood focus, apart from descriptive statistics, three quantitative
analyses are conducted. A Discriminant Analysis was conducted first on the entire file that
helped distinguish between the three neighborhood types. This analysis aims to predict which
group an observation would fall in, given some characteristic variables. In this study, the
‘groups’ refer to the three neighborhood typologies: HDU, HDS, and LDS. Apart from
prediction, this analysis also aims to enrich the understanding of the data and the characteristics
of the observations in each group. The result is a ‘function’ that is a linear equation which would
help in distinguishing in which group an observation would fall.
After the Discriminant Analysis was conducted, the grand file was separated into three
separate files, one for each neighborhood typology (HDU, HDS, and LDS). On each of the three
files, Cluster Analyses (to see what groupings emerged from the data) and Regression Analyses
(to test the hypotheses) were conducted. A Cluster Analysis also helps in grouping the
observations into groups that share common characteristics. However, a Cluster Analysis and a
Discriminant Analysis are different. The Discriminant Analysis helps distinguish between
known groups (in the case of this study, this analysis would help in distinguishing between
observations in the HDU, HDS and the LDS neighborhoods). A Cluster Analysis, on the other
hand, helps to distinguish between unknown groups. In the case of this study, this analysis is
128

conducted on each neighborhood typology and the observations in each file are not grouped into
any category. The Cluster Analysis would help in categorizing the observations into groups that
share some similar characteristics.
Lastly, Regression Analyses were conducted to see the effect of socio-economic
characteristics on travel behavior and environmental burdens, while holding the built
environment constant. This analysis helps in testing the hypotheses laid out earlier. Regressions
aim to predict or determine causal effects of independent variables on a dependent variable.
Another aim is to also determine associations between independent and dependent variables. In
the case of this study, the Regressions are conducted to mainly determine the associations
between socio-economic variables of residents in a neighborhood typology and their travel
behavior and contributions to environmental burdens. The detailed methodology for each
analysis is provided below.

Discriminant Analysis (DA)
The Discriminant Analysis shows what factors are important in the grouping of residents
in the three neighborhood types: higher density urban, HDU (both neighborhoods in Detroit),
higher density suburban, HDS (these include the neighborhoods in Ann Arbor and Birmingham),
and lower density suburban, LDS (these include neighborhoods in Bloomfield Hills and West
Bloomfield). Starting with a stepwise DA on the variables shown in table 10, the analysis
selected the following variables as significant for the DA: VEHALL, AGE, HINCOME, EDUC,
INC_SUPPORTS, Pct_NonWorkTrips_Walk, and Annual_CO_Emissions.

129

Table 10: Variables used in a Stepwise DA
Demographic Variables
Vehicle Variables
AGE
VEHYOU
PINCOME
VEHALL
HINCOME
VEHYEAR
EDUC
VEHCYLINDER
INC_SUPPORTS
VEHMILES
NUM_CHILD

Travel Variables
Pct_WorkTrips_Walk
Pct_WorkTrips-Drive
Pct_NonWorkTrips_Walk
Pct_NonWorkTrips_Drive
Pct_AllTrips_Transit
Ann_WorkMiles_Walk
Ann_WorkMiles_Transit
Ann_WorkMiles_Drive
Ann_NonWorkMiles_Walk
Ann_NonWorkMiles_Transit
Ann_NonWorkMiles_Drive
Ann_AllMiles_Walk
Ann_AllMiles_Transit
Ann_AllMiles_Drive
Ann_Fuel_Consump
Ann_CO_Emissions

After checking for and transforming the variables for ‘normality’, the DA was rerun with
only the above selected seven variables, transformed as necessary. Only the ‘Age’ and the
“Inc_Supports” variables did not need any transformation, however, log transformations were
done for ‘VehAll’, ‘Educ’, ‘Pct_NonWorkTrips_Walk’ and ‘Ann_CO_Emissions’, and square
root transformation was done for the ‘HIncome’ variable. Appendix D shows the distributions of
the variables used in this analysis (before and after transformations). Many of the variables have
skewed distributions, however, only those used in the analysis have been transformed for
normality as this is a prerequisite for these analysis. Initially, the analysis included 200 cases
from a total of 960 cases. A large number of cases are excluded because at least one variable
was missing. Missing data is an issue with this dataset and all analyses will encounter this
problem. However, the 200 cases included satisfy a general requirement that there should be a
minimum of 5 cases per variable included in the analysis, and since this analysis has 7 variables,
the minimum number of cases should be 35.
130

Another general requirement states that there should preferably be a minimum of 20
cases per variable, with 7 variables in the study, the preferred number of cases would be 140, and
this analysis satisfies that requirement as well. However, it should be cautioned that when these
200 cases get divided into the 3 groups (of neighborhood types), Group 1 (Higher Density
Urban) has 15 cases which satisfies a requirement that the group membership should be greater
than the number of variables, but it does not satisfy the requirement that it would be preferable
that the membership be a minimum of 20 cases. In this case, the results can still be interpreted as
long as the above caution is expressed. The Wilks’ Lambda test shows that both Functions are
significant in the classification of groups in this analysis.
The Box’s M statistic tests the null hypothesis that the dispersion matrices across the
subgroups of the dependent variable are homogenous. With a significance of .027, this result
would reject the null hypothesis at a 99% threshold and accept it at a 95% threshold. If the null
hypothesis is rejected, it means that the variance among the three groups is not homogenous, and
we should use the variance between the groups. In that case, we would run the DA using the
between-group covariance matrix for classification. If we do not reject the null hypothesis, it
means variance between the groups is homogenous and we should look at the variance within
each group. In that case, we use the pooled or within-group covariance matrix for classification.
The within-group covariance matrix was utilized for classification because the between-group
classification did not improve the results. Lastly, the Mahalanobis Distance statistics were used
to calculate and ensure that the cross validated probability of 72.5% classification was
significantly higher than a classification achieved purely ‘by chance’.

131

Cluster Analysis (CA)
In the following analyses, the focus shifts from what separates the three neighborhood
types, to what separates the residents within each neighborhood type, while controlling for the
built environment. Factor and Cluster Analyses have been conducted in this line of research so
as to quantitatively identify different groups of information (Shay and Khattak, 2012). This
analysis is done twice per file, separately for all three files; higher density urban (HDU), higher
density suburban (HDS), and lower density suburban (LDS). The first Cluster Analysis is done
using only the demographic variables to see the clusters that emerge. The second Cluster
Analysis is done using all variables in the file to see the effect of travel behavior and emissions
on the clusters.
To get to the Cluster Analysis, the following steps were performed. A Principal
Components Analysis (PCA) was performed on the variables in the files. The components
generated are orthogonal and multicolinearity issues are contained in this step. The component
scores were then used in a Hierarchical Clustering technique that would lead to the decision on
how many clusters were desirable and appropriate. Those numbers of clusters were then used in
a K-means clustering analysis for the final result. Lastly, a table is presented that gives some
insight into the averages of the main variables in each component so that those ‘Clusters’ may be
understood. Results are summarized in table 11 below:

132

Table 11: Summary of PCA and Cluster Analyses
Higher Density
Higher Density
Low Density Suburban
Urban
Suburban
CA – Demographic & Socio-Economic Variables
PCA
Socio-economic
Employment
Dependants
Components
charac.
Income
Employment
Dependants
Dependants
Vehicle Age/Usage
Race
Vehicle Age/Usage
Race
Employment
# Vehicles owned
# Vehicles owned
# Vehicles Owned
Race
Admin_Service Occup.
Engine Size
# of Clusters
4 (1,29,21,1)
4 (118,83,5,20)
8 (43,20,37,22,60,3,7,1)
<4 cases
2
0
2
See Table #
6
8
10
CA – All Variables
Non-Work Trips
*
Energy Usage
Energy Usage
Mode = Transit
Work Trips
Work Trips
Non-Work Trips
Socio-economic charac.
Employment
# Vehicles owned
Dependants
NonWork trips by
Vehicle Age/Usage
Transit
Race
Dependants
# Vehicles owned
Vehicle Age/Usage
Mode = Walking
Race
Laborer
Employment (for wages) Admin_Service Occup
Gender & Laborer Occup
# of Clusters
4 (1,1,19,10)
8 (52,3,19,64,3,4,3,6)
6 (18,12,2,59,24,7)
<4 cases
2
3
1
See Table #
7
9
11
* Took out Transit mode as there are no travels by this mode in this neighborhood type
PCA
Components

Mode = Transit
Mode = Walking
Destination Work
Energy Usage
Socio-economic
charac.
Vehicle Age/Usage
Race
Dependants
# Vehicles owned

Regression Analyses
To test the hypotheses laid out earlier, this last set of analyses is run separately on the
three neighborhood type files (similar to the Cluster Analyses). The purpose of this analysis is to
capture the relationship of the various characteristics of the residents in a particular

133

neighborhood type to their travel patterns (trip mode and annual miles) and subsequent energy
use. The results are intended to be relational rather than causal as has been stressed in other
studies on the topic (Cervero and Kockelman, 1997). Conducting this analysis separately on the
three neighborhood types allows for the built environment to be controlled and for a focus to be
placed on the ‘other’ factors that impact travel behavior and environmental burdens.
The following steps were performed in this analysis. Starting with the higher density
urban typology, neighborhoods that had the least and the ‘messiest’ data, dimensions from the
Principle Component Analysis were examined on the demographic variables done for the earlier
Cluster analyses. From each dimension, a variable that had one of the highest loadings was used
as my independent variable, thus controlling for multicolinearity issues. The Independent
variables were Num_Children, Dum_Black, Dum_White, Educ, Dum_Employed, HIncome, and
Age. The dependent variables were the percent of trips by each mode (Walking, Transit,
Driving), the annual miles traveled by each mode, for each purpose (Work, Non Work, and All
Trips), and the vehicle characteristics and environmental burdens (specifically gasoline
consumption and carbon monoxide emissions). This analysis was then conducted on the other
two neighborhood typologies, the higher density suburban and lower density suburban
neighborhoods. In all the analyses, data variables were transformed to get them as close to a
normal distribution as possible. As followed for the Discriminant Analysis, the rule that the
skewness and kurtosis should fall within -1.0 and +1.0 was followed. Appendix F shows the
variables before and after transformations as used in the regression analyses.
Even after transformations, there were variables that did not fit the skewness-kurtosis rule
mentioned above. An issue with empirical data is that even though some observations seem like
outliers, they in fact are not and represent the large range of behaviors present in the population.
134

This is largely the case with the dependent variables (specially the percentage traveled by
different modes) in this study. However, other assumptions for OLS regressions were observed
and since the main aim is to get at environmental burdens, the regression models were suitable
for the datasets.

Caution
Caution is raised here as the race variables were coded as dummy variables (Dum_White
coded as 1 for white and 0 for all other races and Dum_Black coded as 1 for Black and 0 for all
other races). The Higher Density Urban file did not have enough white respondents, just as the
Low Density Suburban file did not have enough black respondents. The Higher Density
Suburban file had an adequate number of white and black respondents to run the analysis.
Consequently, the regression models have both these dummy variables included and the ‘other’
race would be the reference. Although the data collected is not ideal for an analysis on ‘race’ as
mentioned above, this variable has been included in order to find out if there was any significant
relationships that emerged from the analyses. Also, some of the neighborhoods did have
considerable variability in ethnic and racial composition, and so to keep it consistent, this
variable was kept in the analyses. Results should be viewed with caution.
Another issue is raised specially for the Low Density Suburban neighborhoods.
Respondents living in these neighborhoods undertook almost no travel by Transit, and therefore
the regression results for Transit trips should be read with caution.

135

Commentary
This chapter has presented the main research question for this study, a set of hypotheses
formulated, and the data and methodology that was used to get at the results of the study. The
quantitative methods discussed above are critical to the study as these have been chosen so that
the main research question and the related hypotheses are sufficiently answered.
Apart from quantitative analyses, a separate line of simple, qualitative analyses is also
conducted so as to uncover more detail and rich information that is critical to this study. This
simple data presentation helps uncover layers of valuable information that put this study and its
results into a niche that has not been explored by existing studies. Although this chapter does not
present qualitative assessment of the data, it is included in the next chapter that discusses the
results of all these analyses in detail.

136

Chapter VI - Results and Discussion
Analysis of Variance
Before the results of the first analysis (the Discriminant Analysis) are presented, the table
below (Table 12) shows the mean values of the different continuous variables used in the various
analyses. Using ANOVA, the significance of the differences of the averages across the three
neighborhood types is presented. The table will show whether the mean of each variable in each
neighborhood type is significantly different from the mean of that variable in the other
neighborhood types.

Vehicle Characteristics
Table 12 shows that most of the variables used in this analysis have averages for the
HDU neighborhoods that significantly differ from those of the suburban neighborhoods. Access
to vehicles increases as you move from the urban neighborhoods to the high density suburban
neighborhoods to the low density suburban neighborhoods. While the residents in the urban
neighborhoods have access to less than one vehicle themselves and just about one vehicle per
household, the suburban neighborhood residents have access to more than one vehicle on
average for themselves and two plus vehicles per household. The average year of the vehicle in
use in the HDU Neighborhoods is about 1999. This means that the average car was about ten
years old in the Detroit neighborhoods. The year of the vehicle is the only variable in this
section that is not significantly different in the three neighborhoods, although it is evident that
the suburban neighborhoods have newer cars.

137

Table 12: Analysis of Variance Table across the three Neighborhood Types
Variable
Description
Higher Density
Urban (1)
N
Mean
(2,3)

Higher Density
Suburban (2)
N
Mean
(1)

(1)

208

0.81

VehAll

# of vehicles available to you
# of vehicles available to all in
household

160

1.00

334

1.95

276

2.22

VehYear

Year of the vehicle

159

1998.72

384

2001.98

323

2003.91

VehCyl

Number of cylinders

149

5.80

369

5.23

303

5.81

VehMiles

Miles on the vehicle

121 78929.30

Age

Age of respondent

227

Educ

Years of Education received
# of children (under 18yrs) living
at home

225

VehYou

Num_Child

(2,3)

(2)

(2,3)

208

(3)

52.10

(2,3)

12.95

(2)

0.87

(3)

2.22

Inc_Supports

# of people your income supports

212

PIncome

Personal Income

209 23995.22

HIncome

Household Income

201 30422.89

Pct_WorkTrips_W

% of Work Trips by Walking

93

Pct_WorkTrips_T

% of Work Trips by Public Transit

93

Pct_WorkTrips_D

% of Work Trips by Driving

93

73.04

Pct_NonWorkTrips_W

220

23.63

Pct_NonWorkTrips_T

% of Non-Work Trips by Walking
% of Non-Work Trips by Public
Transit

220

8.31

Pct_NonWorkTrips_D

% of Non-Work Trips by Driving

220

68.06

(2,3)
(2,3)

138

(3)

13.50

(2,3)

13.46

(3)
(3)

(2,3)
(2,3)

376

1.24

Low Density
Suburban (3)
N
Mean

(1,3)

(1,3)
(1,3)

377 62672.20
391

50.99

389

17.08

377

0.54

377

2.44

(3)
(1)
(1)
(3)

(1,3)

356 62359.55

(1,3)

347 96123.92
265

16.03

265

6.66

265

77.30

389

21.47

389

1.13

(3)

389

(1,3)
(3)
(3)
(1)

(1,3)

77.39

318

1.35

(1,2)

(2)

(1,2)

319

42590.57

317

57.04

324
314

(1,2)
(1)

16.75

0.65
(1,2)

309

2.73

278

73417.27

273

108516.48

196

5.53

196

0.09

196

94.38

323

6.83

323
323

(1,2)
(1,2)
(1,2)
(1,2)
(1,2)
(1,2)
(1)

0.18

(1,2)

92.99

Table 12 (cont’d)
Variable

Description

Pct_AllTrips_W

% of All Trips by Walking

223

22.97

Pct_AllTrips_T

% of All Trips by Public Transit

223

9.84

Pct_AllTrips_D

% of All Trips by Driving

223

67.19

Ann_WorkMiles_W

Annual miles to Work by Walking

93

Ann_WorkMiles_T

Annual miles to Work by Transit

93

Higher Density
Urban (1)
N
Mean
(3)

(2,3)
(2,3)

287.15
(2,3)

299.31

(3)

Higher Density
Suburban (2)
N
Mean
392

20.12

392

2.83

392

(3)

77.05

265
265

(1,3)
(1,3)

130.68
64.18

(1)
(3)

Low Density
Suburban (3)
N
Mean
(1,2)

325

6.92

325

0.20

325

92.88

196
196

(1,2)
(1,2)

88.59
(1)

2.92

(1,2)

1849.67
1926.99
3354.82
Annual miles to Work by Driving
93
265
196
Annual Non-Work miles by
(2,3)
(1)
(1)
224.40
99.89
68.11
Ann_NonWorkMiles_W Walking
220
391
324
Annual Non-Work miles by
(2,3)
(1)
(1)
259.65
9.13
2.28
Ann_NonWorkMiles_T Transit
220
391
324
Annual Non-Work miles by
(3)
(3)
(1,2)
1214.64
855.59
1921.06
Ann_NonWorkMiles_D Driving
220
391
324
Total annual miles for trips by
(3)
(1)
341.14
121.32
Ann_AllMiles_W
Walking
223
392
187.98 325
Total annual miles for trips by
(2,3)
(1)
(1)
380.98
52.49
4.03
Ann_AllMiles_T
Transit
223
392
325
Total annual miles for trips
(3)
(3)
(1,2)
1969.69
2156.09
3938.36
Ann_AllMiles_D
by Driving
223
392
325
Annual Fuel Consumption (in
(3)
(3)
(1,2)
99.64
106.31
212.92
Ann_Fuel_Consump
gallons)
145
355
286
Annual Carbon Monoxide
(3)
(3)
(1,2)
8674.03
8870.33
16300.46
Ann_CO_Emissions
Emissions (gms)
146
356
287
*The numbers in superscript parentheses show which group that mean is significantly different from. For example, the HDU mean for
Ann_WorkMiles_D

(2,3)

VehYou is given as 0.81
. This means that this average is significantly different from that of Group 2 (HDS) and Group 3 (LDS).
All significance is at the 0.05 level.
139

One possibility that should be considered in this discussion is that some suburban
residents have older cars that they maintain as ‘classics’ and that this factor might be responsible
for increasing the average age of the vehicles in these neighborhoods. An average of about
95,000 miles on the vehicles owned and operated by the HDU residents suggests that they have
been around for some time and used considerably. The miles drop drastically as we move to the
suburban neighborhoods, with the HDS vehicles averaging 63,000 miles and the LDS vehicles
averaging 43,000 miles. This can imply that residents in LDS neighborhoods have more vehicle
choice for their different trips and also that their newer vehicles have lower miles on them. The
cylinders on the vehicles tell yet another story though. The HDU and the LDS neighborhood
vehicles have more cylinders from those on the vehicles of the HDS neighborhoods. A couple
things that might help understand this distinction: 1) the HDU residents have older, American
vehicles that typically have larger engines, and 2) the LDS residents have newer, however, more
powerful vehicles that have more cylinders. Many of the respondents from the LDS
neighborhoods live in areas that have unpaved roads or terrain leading to their homes and require
more powerful, all-wheel drive vehicles, particularly during winter months. The HDS residents
have smaller vehicles with fewer cylinders. Even though they have newer vehicles, they
probably do not need the power associated with larger engines and they may be more
environmentally conscious, which in part leads to them living in higher density neighborhoods.

140

Demographic Characteristics
With respect to demographic variables, we see that the residents of the urban (HDU) and
higher density suburban (HDS) neighborhoods are similar in age (about 52 and 51 years old
respectively), while the low density suburban (LDS) residents were comparatively older (about
57 years old). The LDS respondents tended to maintain higher incomes. The move to the lower
density suburbs as you advance in your career and move up the socio-economic ladder can
explain the older average age of the West Bloomfield and Bloomfield Hills neighborhood
respondents. The respondents tended to be slightly wealthier and older in the lower density
suburbs. The lowest mean age among the HDS respondents was likely affected by University of
Michigan in Ann Arbor and the student respondents who live in this neighborhood.
With regard to educational attainment, there are notable differences in averages across
the neighborhoods. The urban respondents have lower education levels (graduated high school),
while the HDS and LDS neighborhoods have higher levels (a Bachelor’s degree). This also
translates into differences in the average income levels by the neighborhood groupings. The
urban respondents have lower levels of average Personal ($24,000) and Household Income
($30,500) compared to the suburban respondents. Even among the suburban respondents, those
in the HDS neighborhoods have a lower Personal ($62,000) and Household income ($96,000)
when compared to respondents living in the LDS neighborhoods (Personal Income at $73,000
and Household Income at $108,000). Overall, there is a stark difference in the earnings of the
Urban and the Suburban residents.

141

With regard to dependents, the numbers of children under the age of 18 years living at
home is the greatest in the urban neighborhoods. The suburban neighborhoods, however, have a
higher number of people in the household that the respondents’ incomes support. This might
have something to do with older children living at home or having older parents being supported
by these families. Having older parents living in the house in the lower density suburbs might
have also contributed to the older average age in these neighborhoods.

Travel Behavior
When we think of work based trips, it is understandable that we will have a limited
distance that people can, and are willing, to walk to get to work. The destination has to be close
enough to encourage people to walk to their workplace, rather than take an alternate mode of
travel. The HDU and the HDS residents undertake about 15% of their work trips by walking.
The LDS residents conduct about 6% of their work trips by walking. This difference is due to
the fact that the LDS neighborhoods are single use, disconnected, and less dense neighborhoods
where very few people will have a work destination that is close enough to walk to. This
distance factor is reflected in the annual miles to work by walking. The HDU residents travel
about 284 miles to work by walking annually, while the LDS residents travel just about 88 miles
annually. Driving as a mode of travel, on the other hand, accounts for between 75% (HDU) to
95% (LDS) of work trips. The HDU and HDS neighborhood residents drive shorter distances
annually to work, while the LDS residents drive twice the amount, indicating that their work
places are further from their residence neighborhoods.

142

While transit is an option for the HDU and the HDS neighborhood respondents, it is not a
practical option for the lower density suburban neighborhood respondents, and hence there is
limited travel by transit in the lower density outer suburbs. This mode of travel is used, however,
by respondents in the HDU and HDS neighborhoods. The HDU residents use transit for almost
the same percentage of work trips as walking and travel almost the same annual distance to work
by transit as they do by walking. The HDS neighborhood residents, on the other hand, walk
more for their work trips than use transit and similarly, they travel longer distances annually by
walking than they do by transit for work based trips.
Since work trips are considered ‘essential’ trips, there is less of an option to use nonautomobile modes of travel for such a structured travel need. The non-work trips, on the other
hand, show that people do exercise the option of using transit and walking to their destinations
when they are not bound by time and location constraints. The percentage of trips by walking
for non-work trips is higher in both the HDU and HDS neighborhoods. The use of transit
actually decreases with non-work based trips, indicating that the residents choose walking over
the use of transit for these trips. It also means that non-work trips are also likely closer to the
respondents’ homes, close enough for walking.
For the LDS neighborhoods, there is not much of a difference between the walking and
the transit options, likely because of the much greater distances to daily destinations from these
neighborhoods. Residents of the LDS neighborhoods have a relatively homogenous travel mode
and use their vehicles for over 90% of all their trips, compared to the urban residents who use
their vehicles for about two-thirds of all their trips. As far as annual distances are concerned, the
residents in the LDS neighborhoods travel double the total distances annually for all trips by car,
compared to residents in the HDU neighborhoods. Consequently the LDS neighborhoods, when
143

compared to the HDU and HDS neighborhoods, consume more than double the fuel annually and
are responsible for emitting more than double the pollutants.
One thing to bear in mind is that the trip purposes are divided into work and non-work
trips. However, many respondents did not enter in details for both trip purposes, which meant
that there were many missing values. When looking at the annual work miles and non-work
miles, the means reflect the presence of the missing values within each trip purpose.
Consequently, the average annual miles (by various modes) to work and non-work destinations
might not add up to the average annual miles for all trips combined. For example, if a
respondent said he traveled 10 miles to work by driving but did not say anything about the travel
to non-work trips, he would be counted in the average for work trips and all trips combined, but
he would not be counted in the non-work trips. Such instances, give rise to different Ns and the
averages (mean values) would be reflective of that and therefore, the average values for work
trips and non-work trips might not add up to the average value for all trips combined (for any
particular mode).
After looking at the averages for the different variables and whether they differ
significantly across the neighborhood types, the next analysis conducted is the Discriminant
Analysis. Although this analysis also aims at distinguishing between the three neighborhood
types, its importance lies in formally indicating what precise variables contribute to the
distinction. We know that the three neighborhoods are different, however, this analysis explores
the specific variables that distinguishes the three neighborhood typologies.

144

Discriminant Analysis (DA)
Overall, the Discriminant Analysis was telling and informative in revealing what
distinguished the residents living in the three neighborhood types. Variables that were included
in this analysis were not only demographic, but also those related to vehicles and travel patterns
(see Table 13). The table below shows the description of the variables selected for this analysis.
Table 13: Variables used in the Discriminant Analysis
Variables
Description
Number of vehicles accessible to all in the household – Log
Ln_VehAll
normal transformed
Age

Age of the respondent

Sqrt_HIncome

Annual Household Income – Square root transformation
Number of years of education received - Log normal
Ln_Educ
transformed
Number of household members that the respondent’s
Inc_Supports
personal income supports
Percent of trips to Non-Work destinations that are
Ln_Pct_NonWorkTrips_Walk undertaken by walking - Log normal transformed
Annual Carbon Monoxide Emissions (in grams) - Log
Ln_Annual_CO_Emissions
normal transformed
The DA analysis resulted in 2 functions (for 3 groups) as shown in table 14 below:
Table 14: Functions at Group Centroids
Function
Neighborhood_Type
1
HDU
-2.969
HDS
.315
LDS
.024

2
-.220
-.245
.791

From the above table we see that Function 1 distinguishes the HDU neighborhoods from
the suburban ones and Function 2 distinguishes the two suburban neighborhoods (the LDS
neighborhoods from the HDS neighborhoods). The HDU neighborhoods are ignored because
145

they are already distinguished from the other neighborhood types by Function 1. Table 15 below
shows the Structure Matrix that identifies the variables that are significantly associated with each
Function. In general, values greater than 0.3 are considered to be significantly associated with
the function.

Table 15: Structure Matrix
Function
Variables
Ln_Educ
Sqrt_HIncome
Age
Ln_NonWorkTrips_Walk
LnAnn_CO_Emissions
Inc_Supports
LnVehAll

1

2
.830
.464
.111
.104
-.193
-.090
.182

.170
.349
.610
-.542
.443
.313
.309

From the above structure matrix, looking at the values for Function 1, we see Education
and Household Income as having the highest contributions to this function that distinguishes the
HDU neighborhoods from the suburban ones. In the column for Function 2, Age, Percent of
Non-Work trips undertaken by Walking, and Annual carbon monoxide emissions have the
largest contributions to this function that effectively distinguishes the two suburban
neighborhoods (the HDS and the LDS neighborhoods). As such, the Functions are shown by the
following equations:
Function 1 = .240 (LnVehAll) + .055 (LnPct_NonWorkTrips_Walk) - .286
(LnAnn_CO_Emissions) - .166 (Age) - .348 (Inc_Supports) + .817 (Ln_Educ) +
.441 (Sqrt_HIncome)
Function 2 = .214 (LnVehAll) - .363 (LnPct_NonWorkTrips_Walk) + .295
(LnAnn_CO_Emissions) + .795 (Age) + .483 (Inc_Supports) + .085 (Ln_Educ) .127 (Sqrt_HIncome)
146

Table 16 below shows the averages of the variables within each group and as a total.
Table 16: Group Statistics
Group
Variables
HDU
LnVehAll
Ln_NonWorkTrips_Walk
LnAnn_CO_Emissions
Age
Inc_Supports
Ln_Educ
Sqrt_HIncome
HDS
LnVehAll
Ln_NonWorkTrips_Walk
LnAnn_CO_Emissions
Age
Inc_Supports
Ln_Educ
Sqrt_HIncome
LDS
LnVehAll
Ln_NonWorkTrips_Walk
LnAnn_CO_Emissions
Age
Inc_Supports
Ln_Educ
Sqrt_HIncome
Total
LnVehAll
Ln_NonWorkTrips_Walk
LnAnn_CO_Emissions
Age
Inc_Supports
Ln_Educ
Sqrt_HIncome

Mean
.3697
3.0953
9.3132
40.9333
3.2000
2.5092
189.0464
.6158
3.3870
8.4939
45.5725
2.7862
2.8600
309.0749
.7277
2.9029
9.1495
53.5851
3.2660
2.8515
327.0032
.6237
3.2513
8.7094
47.1075
2.9300
2.8317
304.2859

Valid N
15
15
15
15
15
15
15
138
138
138
138
138
138
138
47
47
47
47
47
47
47
200
200
200
200
200
200
200

A look at the averages of the variables significant in the first function, distinguishing
urban and suburban neighborhoods, shows that the HDU residents have lower levels of education
and lower household incomes than residents in the suburban neighborhoods. Averages of the

147

variables significant in the second function, distinguishing suburban neighborhoods, reveal that
the LDS neighborhoods had residents that were older, traveled fewer times via walking for nonwork trips and were responsible for higher annual carbon monoxide emissions than the residents
of the HDS neighborhoods. The canonical functions plot is shown below.

Figure 33: Canonical Functions Plot

Discussion
As mentioned earlier, the HDU and the HDS neighborhoods are more like each other in
terms of the built environment than in terms of the socio-economic characteristics of its
residents. This comes out in the results of the Discriminant Analysis. Education and Household
148

Income play a major role in distinguishing the residents of these two neighborhood types. The
LDS and the HDS neighborhoods are more alike in terms of residents’ socio-economic
characteristics, but their built environments are very different. The Discriminant Analysis brings
this aspect of the neighborhood typologies out as well. Travel patterns and emissions make up
the majority of the variables associated with this function. This is understandable since the LDS
neighborhoods are less dense, have less connectivity, and are dominated by single land uses,
when compared to the HDS neighborhoods, limiting the use of non-motorized travel options and
encouraging travel via private vehicles. This analysis, therefore, effectively shows what was
expected: the class dimension that separates the urban neighborhoods from the suburban
neighborhoods and the built environment dimension that separates the two suburban
neighborhoods that have a resident base with similar socio-economic characteristics.
The next section of this chapter presents analyses by neighborhood type, exploring how
variations in socio-economic and demographic characteristics by neighborhood physical form
affect travel, gasoline consumption, and associated emissions. This is done for two reasons: first,
the analyses center on within neighborhood type differences and secondly, this structure will
help present the differences in socio-economic characteristics, travel behavior and environmental
burdens while controlling for the built environment.

149

Higher Density Urban – City of Detroit Neighborhoods
This neighborhood typology typically has lower income, predominantly AfricanAmerican residents. It is a neighborhood typology—neighborhoods experiencing disinvestment
and decline—that is seldom analyzed in travel behavior research. More often than not, these
residents live here because it is where they grew up and call this City home. Low house prices
were the dominant reason why respondents in this neighborhood typology chose to stay here.
However, another important reason was the fact that these homes were their family homes and
they were inherited. These residents are typical of lower eastside Detroit residents. The table
below has some descriptions of the categorical variables used in the analyses:
Table 17: Breakdown of Categorical Variables
Variable
Higher Density
Urban
Number
Percent
RACE
White
18
8.3
Non White
199
91.7
Black
196
90.3
Non-Black
21
9.7
Total
217
100
OCCUPATION
Employed (for wages)
Non-Employed
Total

83
140
223

37.2
62.8
100

The above table shows that there were only 18 respondents from these neighborhoods
that were White, making up a mere 8.3% of the total respondents. The majority of the
respondents were African American, who made up over 90% of the survey respondents. Another
striking feature is the ‘Employment’. The dummy variable for employment was used to identify
everyone that was employed for wages, i.e. a stay at home parent, a student, the unemployed, and
150

the retired were not counted in this category. From table 17, we see that only 37% of the
respondents were ‘employed’ for wages. The average age of the respondents in the urban
neighborhoods was about 52 years. This also suggests that this group of mainly African
American residents did not typically move out of the inner city as they progressed, which is
typical of the urban evolution theory. You would expect to find younger residents in the urban
neighborhoods, and as you move outward into the suburbs, you would expect to find older
residents. This is not the case with the urban typology neighborhoods. We see here a middle
aged, mostly unemployed African American resident base. The table below presents the
summary statistics for the variables in this neighborhood type. All of the variables included in
this table are from the database that contains all the respondents who indicate travel patterns
(whether they have a car or not). In the regression analyses, the last two variables in this table
(Annual fuel consumption and Annual CO Emissions) are used from the database that has only
those respondents that indicated travel patterns by car and had access to a car for their travel.
Therefore the averages in the below table will be smaller than those in the regression and other
tertiary breakdown tables.

151

Table 18: HDU - Descriptive Statistics
N
Minimum
Variable

Maximum

Mean

Std. Dev.

VehYou
208
0
5
.81
.79
VehAll
160
0
4
1.0
.97
VehYear
159
1965
2009
1998.72
6.5
VehCyl
149
2
8
5.80
1.37
VehMiles
121
30
309872
78929.30
59435.63
Age
227
18
94
52.10
16.26
Educ
225
10
19
12.95
2.52
Num_Child
208
0
8
0.87
1.38
Inc_Supports
212
0
8
2.22
1.53
PIncome
209
5000
115000
23995.22
21715.93
HIncome
201
5000
135000
30422.89
27891.67
Pct_WorkTrips_W
93
0.00
100.00
13.50
26.87
Pct_WorkTrips_T
93
0.00
100.00
13.46
28.00
Pct_WorkTrips_D
93
0.00
100.00
73.04
39.48
Pct_OptTrips_W
220
0.00
100.00
23.63
34.03
Pct_OptTrips_T
220
0.00
100.00
8.31
19.78
Pct_OptTrips_D
220
0.00
100.00
68.06
39.36
Pct_AllTrips_W
223
0.00
100.00
22.97
32.87
Pct_AllTrips_T
223
0.00
100.00
9.84
21.43
Pct_AllTrips_D
223
0.00
100.00
67.19
39.33
Ann_WorkMiles_W
93
0.00
10244.00
287.15
1241.47
Ann_WorkMiles_T
93
0.00
10244.00
299.31
1310.86
Ann_WorkMiles_D
93
0.00
25953.20
1849.67
3430.76
Ann_OptMiles_W
220
0.00
13010.40
224.40
977.50
Ann_OptMiles_T
220
0.00
12766.00
259.65
1047.38
Ann_OptMiles_D
220
0.00
11317.80
1214.64
1706.25
Ann_AllMiles_W
223
0.00
15844.40
341.14
1452.12
Ann_AllMiles_T
223
0.00
15917.20
380.98
1625.60
Ann_AllMiles_D
223
0.00
26410.80
1969.69
3209.19
Ann_Fuel_Consump*
145
0.00
1049.53
99.64
163.03
Ann_CO_Emissions*
146
0.00
72825.48
8674.03
13056.73
* Ann_Fuel_Consump and Ann_CO_Emissions show annual total environmental burdens for
travel to end-point destinations.

152

Cluster Analyses: The Higher Density Urban Neighborhood Typology
A Cluster Analysis is conducted first to see what kinds of groups emerged within each
neighborhood typology. As mentioned earlier, two sets of Cluster Analyses were conducted for
each neighborhood type, one with only the demographic variables and one analysis that included
the vehicular and travel behavior variables. For the analysis done with demographic variables on
urban neighborhoods, the results showed that the clusters formed differed mainly on: 1) the
number of children under the age of 18 living at home; 2) the age of the residents; and 3)
whether they were employed (for wages).
Table 19 below shows the results of these analyses. The K-Means 4 cluster analysis
shows 2 out of the 4 clusters as having one member each. Although the analyses identify these
as ‘outliers’, it is reflective of the objective data. In reality, these clusters are realistic and can
hold their value as individual clusters. A look at these ‘outliers’ shows some peculiar
characteristics of the residents in these neighborhoods. The results are presented below.
First the clusters with low membership numbers (<4) are discussed, clusters 1 and 4,
followed by the discussion of the clusters with sufficient membership numbers (>=4), clusters 2
and 3.
o Cluster 1 identified a young (22 years) white respondent, with a higher than average
household income, employed within the Administrative and Service Occupations, and
having access to a high number of vehicles in the household (4). Looking at the averages
presented the table in the beginning of this section, we can see that this resident stands out
from the rest.

153

o Cluster 4 identified a similar aged (21 years) young African American respondent,
employed in the same industry (Administrative and Service), has access to one vehicle,
however, the household income is very low, at $15,000. Both the above groups single out
residents that are unique in some way.
o Cluster 2 presented respondents that are younger than the average aged (41 years), almost
all African American, with a household income that is higher than the average for those
neighborhoods ($47,000). Most of these are employed for wages, have a child under the
age of 18 living at home and have access to a car.
o Cluster 3 brought together a group of mainly African American residents who were older
than the average aged resident (58 years) and consequently had barely any children under
the age of 18 living at home. These respondents had a higher than average household
income (for the neighborhood), even though less than half of them were employed for
wages.

The analysis done with all variables on this neighborhood type reveal differences mainly
in: 1) the annual miles for work related trips done by walking; 2) annual fuel consumption; 3)
carbon monoxide emissions; 4) income; 5) education; and 6) miles on vehicles. In essence,
socio-economic characteristics as well as some travel related characteristics were brought out
(see Table 20). Clusters 1 and 2 came out as having low membership (less than 4 each). Again,
these are anomalies in a way, but they may also stand as individual groups with these particular
characteristics. The results are listed below:

154

o Cluster 1 identified a resident that is African American, has a lower than average household
income, even after being employed as a Professional. This resident is unique because of
his travel behavior. More than 20% of his trips to non-work and all destinations are by
transit and walking. Consequently, he has access to a vehicle that has very low miles on it
(28,000), even after using it for one third of the trips to work.
o Cluster 2 identified a resident that has a very low personal income, being employed for
wages, and conducting about a third of the work trips by driving. He has almost no other
trips by walking or transit and that is probably why this person has an unusually high
number of miles on his vehicle.
o Cluster 3 identified a group of almost all African American residents employed for wages,
with an about average personal income (for the neighborhood). They do not undertake any
trips by transit, only a very few trips by walking, which leaves the bulk of their trips as
being done by driving. Understandably, their vehicles have high mileage on them and they
are responsible for a greater amount of environmental burdens (i.e. they consume more fuel
and emit more Carbon Monoxide).
o Cluster 4 is similar to Cluster 3 in travel patterns, but different in socio-economic
characteristics. They are mostly employed as professionals, have higher personal incomes,
yet have fewer miles on their vehicles, even after depicting a similar travel pattern as the
previous cluster. That is, they maintain high fuel consumption and emit high pollutant
levels.

155

Table 19: Clusters Analysis on Demographic and Socio-Economic Variables – HDU
CL
SES
Dependants
Race
Empl
Veh_Ownership
HIncome
Educ Prof #Child Age
Wh
Bl
Empl Admin Veh-you Veh_all Cylinder
1
45000
14
1
1
22
1
0
1
1
4
4
6
2
46034
14.2 0.31
1.07
40.7 0.03 0.97
0.83
0.31
1.07
1.52
5.7
3
45000 13.05 0.19
0.19 58.19 0.27 0.67
0.43
0.19
1.19
1.33
5.71
4
15000
14
0
0
21
0
1
1
1
1
1
6

Table 20: Clusters Analysis on All Variables – HDU
CL
Transit (Mode)
Walking (Mode)
Work
Energy Use
Non-Work %AllTrps %Non%NonNonMiles-W Trps_D Fuel
CO
Miles
WorkTrps WorkTrips WorkMiles
1
5.6
26.1%
21.4%
21.4%
5.6
0
33.3%
4.5
8.9
2
0
9.1%
0%
0%
2.4
6.3
38.5%
4.7
9.2
3
0
0%
0%
6.5%
1.4
0.5
93%
5.8
10.2
4
0
0.4%
0.5%
6.5%
1.2
1.4
90%
4.7
9.2

Table 20 (cont’d)
CL
SES
Veh Miles
Race
Dependents
Pincome D_Prof Educ V_Miles Wk_Miles_T D_Bl D_Wh Inc_Sup
D_Empl #Child
1
25000
1
12
28000
0
1
0
1
1
0
2
15000
0
14
192029
5.7
1
0
5
1
0
3
39736
0.2 13.2
88712
0
0.8
0.2
2.3
0.9
0.9
4
42000
0.7 15.7
67366
0
1
0
2.4
1
0.2

156

Veh-U
1
1
1.1
1.1

Regression Analyses: The Higher Density Urban (HDU) Neighborhood Typology
After looking at the clusters formed, it is imperative that the hypotheses be tested. The
following Regression Analyses will test the hypotheses. This analysis is undertaken to reveal the
relationships within the neighborhoods. It is not intended to denote causality, nor is it intended
for forecasting purposes; rather it should be looked at as a means to understand the complex
relationships of socio-economic factors that are significant in impacting travel behavior and the
resultant environmental burdens, while controlling for the physical/built environment. The
results are discussed below by trip purpose and the tables, tables 21 and 22, present the results by
mode of travel:


The results indicated that there were no linear relationships between socio-economic
characteristics and trips taken for work purposes.



For the non-work trips (see Tables 21 and 22):
o The percentage of travel undertaken by walking as the mode had a significant and
negative association with age, education and whether the respondents were
employed (for wages) [Hypothesis C]
o The percentage of travel undertaken by public transit had a significant and
negative association with household income [Hypothesis B]
o The percentage of travel undertaken by driving as the mode had a significant and
positive association with household income, education, whether the respondents
were employed (for wages) and the age of the respondents [Hypothesis A]
o The total annual miles for trips undertaken by walking had a significant and
negative association with age, whether the respondents were employed (for
wages) and years of education received [Hypothesis F]
157

o The total annual miles for trips undertaken by public transit had a significant and
negative association with the respondents’ age and whether the respondents were
employed (for wages)[Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with household income and whether the respondents were
employed (for wages) [Hypothesis D]


For all trips (see Tables 21 and 22):
o The percentage of travel undertaken by walking as the mode had a significant and
negative association with age, education and whether the respondents were
employed (for wages) [Hypothesis C]
o The regression results showed that there was no linear relationship between socioeconomic characteristics and all trips by transit [Hypothesis B]
o The percentage of travel undertaken by driving as the mode had a significant and
positive association with the number of dependent children, household income,
employment (for wages) and age of the respondent [Hypothesis A]
o The total annual miles for trips undertaken by walking had a significant and
negative association with years of education received and the age of the
respondent [Hypothesis F]
o The total annual miles for trips undertaken by public transit had a significant and
negative association with the age of the respondent [Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with household income and whether the respondents were
employed (for wages) [Hypothesis D]

158

Table 21: HDU Regression results - Mode
Driving Trips
Variable
Work
Non-Work All

Constant

Work

Transit Trips
NonAll
Work
**

60.072

Work

*

55.727

Walking Trips
Non-Work All

***

47.284 156.787

***

143.949

-21.621

-116.859

-99.676*

74.337

9.388

6.250

8.527**

-3.204

-2.691

-3.631

-6.185

-3.559

-4.895

Dummy_Black

-18.197

7.920

-5.206

10.147

1.518

2.128

8.050

-9.437

3.078

Dummy_White

-15.381

16.135

3.443

0.337

3.487

1.671

15.043

-19.622

-5.114

12.445

27.528*

23.371

-17.428

-2.694

-0.618

4.983

-24.834*

-22.753*

***

***

# Dependant Children

Education
Dummy_Employed
Hshld_Income
Age

R

3.244
0.504

2

Adj. R

19.237

22.187

*

6.249

***

0.613

20.012

*

6.313

***

0.691

-12.481
-0.197

-5.238
**

-3.914

-2.378
**

-3.794

-6.756
-3.047

-0.223

-0.078

-0.108

-0.282

***

-16.949

-2.335
***

-0.535

***

-17.634

-2.519
***

-0.583

.078
2

.193

.181

.063

.080

.049

.049

.144

.158

-.013

.158

.146

-.029

.041

.009

-.045

.108

.122

N
79
172
173
79
172
173
79
172
173
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

159

Table 22: HDU Regression results – Annual Miles
Driving Miles
Variable
Work
Non-Work All

Work

10.079**

6.620

-0.440

0.090

-0.089

-1.172***

-0.168

-0.464

-0.396

0.583

0.536

0.740

0.013

-0.976

-0.887

-0.261

-0.006

0.295

0.918

0.818

-0.495

-1.405

-1.470

0.825

0.277

-1.865

-1.188

-1.715

-1.475

-3.253***

-2.848**

-0.651

1.458*** 2.087***

-1.121

-0.869*

-0.586

-0.137

-0.964**

-0.534

0.895

0.736*** 0.820***

0.071

-0.303

-0.234

0.054

0.100

0.128

-0.034** -0.043***

-0.026

-0.056***

-0.064***

-7.871

-4.067

-4.102

7.429

# Dependant Children

-0.110

0.185

0.429

Dummy_Black

1.585

-0.833

Dummy_White

-0.197
2.164

Dummy_Employed
Hshld_Income

13.899*** 13.147***

2

0.007

0.009

-0.034

.102

2

Adj. R

8.989

-0.064*

Age

R

Work

Walking Miles
Non-Work All

**

Constant

Education

Transit Miles
Non-Work All

.203

.278

.068

.086

.086

.104

.145

.122

.013

.169

.247

-.024

.047

.047

-.016

.109

.084

N
79
172
173
79
172
173
79
172
173
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

160

Table 23: HDU Regression results – Vehicle Characteristics
Vehicle
Veh-All
Veh_Year Veh_Cyl

Constant
# Dependant Children
Dummy_Black
Dummy_White
Education
Dummy_Employed
Hshld_Income
Age

R

-.013
**

-1.323

-0.978
0.411
0.346

*

***

0.297

**

**

4.482

-7.469

-7.067

0.122

0.014

0.388*

0.627**

-0.794

0.974

2.819**

3.351*

-3.838

0.668

2.956**

3.561*

0.077

2.304***

3.355***

-0.177

0.818

*

6.181

1.413
**

1.599

**

**

1.127

0.004

0.278

0.295

-0.006

2

0.047

-0.011

-0.008

-0.005

.304

2

Adj. R

***

-1.605 1965.85

Annual Environmental Burdens
Fuel Consumption
CO Emissions

.234

.047

.341

.285

.262

.165

-.015

.279

.218

N
125
118
114
83
110
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

161



The number of vehicles owned by the household had a significant and positive
association with household income and whether the respondent was employed (for
wages), and a negative association with whether the respondent was black (see Table 23)
[Hypothesis H]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the year of the vehicle used for travel had a significant and positive
association with the household income and the number of years of education received
(see Table 23) [Hypothesis G]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the annual fuel consumption was significantly and positively associated
with the number of dependent children, whether the respondents were black, whether the
respondents were white, years of education and whether the respondents were employed
(for wages) (see Table 23) [Hypothesis D]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the annual carbon monoxide emissions was significantly and positively
associated with the number of dependent children, whether the respondents were black,
whether the respondents were white, years of education and whether the respondents
were employed (for wages) (see Table 23) [Hypothesis D]

The above analyses results show a clear distinction in the socio-economic characteristics
and how they impact travel behavior and environmental burdens. For all trips other than work

162

based trips, socio-economic variables such as household income, employment status (whether
employed for wages), and age of the respondent came out as having a significant impact on
travel behavior. There were also less often, but significant impacts of education and race
(whether the respondent was African American) on travel behavior, vehicle ownership and
vehicle characteristics. With only a third of the respondents being employed, travel to work
regressions did not show significant goodness-of-fit results. Also, given the fact that there was a
small percentage of respondents traveling by modes other than driving to their different
destinations, the “percent trip” variables – those showing mode of travel, had non-normal
distributions even after transformations (and thereby were kept to their original, untransformed
states). The regressions for transit trips, even though they were significant, should be viewed
with caution as the dependent variable as well as the residuals were not typically normally
distributed and that would imply biased results. Even so, the annual distances driven, fuel
consumption and carbon monoxide emission variables were successfully transformed and
regression results were significant.
In order to provide further and more textured insight into travel behavior by socioeconomic characteristics, a breakdown is provided of the urban neighborhoods by income
tertiles, allowing a further exploration into the variation of travel behavior by class.

A Discussion by Income Groupings: The Higher Density Urban (HDU) Neighborhood Typology
Table 24 shows a disaggregation of the respondents of the urban neighborhoods into
three groups: those within the lowest one-third of the household income range, those that fall into

163

the middle one-third of the income range and those that have a household income that is in the
highest one-third of household incomes.
Table 24: HDU Respondents split into three Household Income groups
Household Income
Count
Percentage
$5,000 to $45,000 (Low Income Range)
161
80%
>$45,000 to $90,000 (Middle Income Range)
31
15%
>$90,000 to $135,000 (Upper Income Range)
9
5%
Total
201
100%

Table 24 shows that within the HDU typology, there is a heavy concentration of
respondents in the lowest income grouping. About 80% of the respondents fall in the lowest
income category, while only 5% fall in the highest household income category. This also sheds
more light on the figures in table 11 that show the mean values for various variables. The mean
household income for the respondents in these neighborhoods is $30,423, which falls in the
lowest third of the income range in these neighborhoods. This stark divide is helpful in
understanding the characteristics and behavior of the respondents in the different income groups.
Table 25 (below) shows that the lowest income range, with 80% of the population, has
access to less than one car per household, and those that do have access to a car, have older cars
(10 years old on average). The highest earning, 5% of the respondents, have access to just under
2 cars per household and have newer cars (7 years old on an average). The table on demographic
characteristics (table 24) shows that there is not a difference in the age of the respondents across
the income classes, nor a difference in the number of dependents in the household. There is a
visible difference, however, in the years of education received. The respondents in the lowest
income range have an average of 13 years of education (completed high school), while the ones
in the highest income range have an average of 16 years of education (college graduates).

164

Table 25: HDU Vehicle Ownership

Vehicles for you
Vehicles for all in household
Vehicle Year

Income Range
Low
Middle
Upper
0.8
1.1
1.3
0.8
1.6
1.8
1998
2002
2002

Table 26: HDU Demographic Characteristics

Age
Education
Number of people the Income Supports
Number of children living at home

Income Range
Low
Middle Upper
51.1
50.7
51
12.5
14.5
15.8
2.3
2.5
2.4
1.0
0.6
0.9

Table 27: HDU Mode of Travel Based on Usual Weekly Travel Patterns
Income Range
Low
Middle
Upper
% WorkTrips by Walk
16
12.6
0
% WorkTrips by Transit
16.5
7.7
0.5
% WorkTrips by Drive
67.6
79.7
99.5
% Non-Work Trips by Walk
27.4
12.2
12.2
% Non-Work Trips by Transit
9.2
3.1
6.9
% Non-Work Trips by Drive
63.4
84.7
80.9
% All Trips by Walk
26.8
13.7
6.8
% All Trips by Transit
10.7
4.4
7.1
% All Trips by Drive
62.5
81.9
86.1

Table 27, depicting the mode of travel by the income groupings paints an interesting
picture. For work trips, the respondents in the upper income category rely solely on driving as
the mode of travel, while those in the lowest income range conduct two-thirds of their trips to
work by driving. Those in the lowest income category undertake about 16% of work-trips by
walking and use transit for about another 16% of work-trips.
For non-work trips, the upper income range respondents use driving as the mode for
about 81% of the trips. These respondents walk for 12% of their trips and use transit for 7% of
165

their trips. This remains quite a different travel behavior pattern from the lowest income
category residents. Respondents in the lowest income category use driving for about 63% of the
trips to non-work destinations, while they use transit for 9% and walk for another 28% of their
trips. Again, this is for non-work trips specifically.
The reliance of the upper income residents on the car for all of their work trips might
have something to do with where their jobs are located. Looking at the table on annual average
distances to end-point destinations, we see that these residents travel over four thousand eight
hundred miles on average per year to work destinations in their cars, while the lowest income
range residents travel about one fourth of those miles per year in their vehicles. This difference
is telling as it may suggest that the upper income residents have jobs further from their residence
locations. The longer distances among the upper income grouping imply that many are likely
driving into the Detroit CBD or into the suburbs for their employment. The lowest income range
residents either do not have the educational qualifications or access to vehicles to accept the jobs
in the CBD or outside the city limits. Many of the residents within the lowest income category
are also unemployed. Another important travel pattern to recognize is that all three income
groups travel further distances to work than to non-work destinations, thereby exercising their
choice for different and closer available locations and products in an urban environment
characterized by higher accessibility, density and connectivity.
Lastly, the upper income HDU residents travel over 3.5 times the average annual distance
to end-point locations that the lowest income residents travel for all purposes by car.
Consequently, they have over 2.5 times the annual fuel consumption and the annual carbon
emissions as that of the lowest income range residents. The wealthier respondents have newer

166

cars and that might be a factor in the lower proportional emissions. In any case, the upper
income residents have vehicles and travel patterns that result in greater environmental burdens.

Table 28: HDU Average Annual Miles to End-Point Destinations by Trip Purpose Based on
Usual Weekly Travel Patterns

Annual Miles for WorkTrips by Walk
Annual Miles for WorkTrips by Transit
Annual Miles for WorkTrips by Drive
Annual Miles for Non-Work Trips by Walk
Annual Miles for Non-Work Trips by Transit
Annual Miles for Non-Work Trips by Drive
Annual Miles for All Trips by Walk
Annual Miles for All Trips by Transit
Annual Miles for All Trips by Drive

Income Range
Low
Middle
Upper
157.7
821.0
0.0
238.5
644.1
0.0
1209.3
3172.4
4809.1
199.1
522.6
213.5
214.2
615.7
219.0
1036.6
2590.3
2179.6
259.8
1097.3
213.5
305.9
1066.5
219.0
1501.7
4811.0
5385.6

Table 29: HDU Average Annual Environmental Burdens Based on Usual Weekly Travel
Patterns*
Income Range
Low
Middle
Upper
Annual Fuel Consumption (gallons)
221
484.8
588.6
Annual Carbon Monoxide Emissions (grams)
20140.8 38212.2
51464.8
*The environmental burdens calculations are a result of both end- and return-destination trips.

Discussion
The analyses of the urban neighborhoods reveal a number of important characteristics in
travel patterns within this neighborhood typology. There is a heavy concentration of low income
African American residents in these neighborhoods. However, there are a few residents of
‘other’ races that live here and there are also some residents with higher personal and household
income levels, at least relative to the neighborhood averages. Holding the built environment
167

constant, a number of marked differences in travel behavior of residents are recognized by socioeconomic characteristics. It was thought that these socio-economic characteristics alone would
account for the differences in travel behavior, but as the analysis revealed, there were other
variables that had significant associations with travel. Age is one of those conditions that came
up often in the regressions as significantly affecting travel behavior, even though it is not
included as a socio-economic variable. It is understandable that the older you get, the more you
would travel by a private vehicle rather than public transit or walking. Older residents tend to
choose a more comfortable mode of travel as they are unable to walk longer distances and/or
take public transit.
The main revelation is that even the very few residents with a higher level of socioeconomic characteristics have a remarkably different travel pattern. They have jobs that are
further away, perhaps in the Detroit CBD or outside the city, and have and exercise the option to
travel by their vehicles for most trips. The cluster analysis singled out some of these residents as
‘outliers’, however as mentioned earlier, this is a legitimate class of residents. This is also
precisely the reason why the mode travel variables were skewed but still left as is for the
regressions. Respondents in the lowest income grouping have jobs that are closer and are more
restricted by income to travel more often by other means of transportation than the car for their
work trips. Many of these residents are also simply unemployed. This is also reflected in the
fact that they have access to fewer cars and so they are relegated to travel to nearer destinations.
These residents travel lesser distances by car when they need to travel to non-work destinations,
but these figures are significantly below those of the higher income residents.
Another characteristic that is evident in this neighborhood is that the higher income
residents are responsible for a higher amount of pollutants. Given that they travel further and
168

more often in their vehicles, they consume the expected higher levels of fuel, emit higher levels
of carbon monoxide emissions and consume more fuel. Hence, a class factor emerges in the
selection of vehicle choices, as we see the upper income range residents travel more via their
cars, consume more fuel and are responsible for more emissions than are the lower and middle
income range residents living in the same built environment. At the same time, middle income
respondents in the urban neighborhoods—while consuming less petroleum and emitting fewer
CO emissions than the upper income respondents—consume more fuel and maintain higher CO
emissions than the lowest income respondents. This is supported by both the regression analyses
and the analyses of the income sub-groups.

Higher Density Suburban – Ann Arbor and Birmingham Neighborhoods
In terms of the urban built environment, the spatial structure of the neighborhoods, the
HDS neighborhoods are similar to the HDU neighborhood typology. These are relatively high
density neighborhoods for the Michigan context, their road networks are based on the gridiron
street pattern, and the neighborhoods are characterized by mixed land uses. Looking at the
socio-economic characteristics of residents in the HDS neighborhoods, however, they are
predominantly white, and they have higher incomes and greater education levels than the
residents of the HDU neighborhoods. So in essence, the built environment characteristics of the
HDS neighborhoods are similar to the urban neighborhoods, while the socio-economic
characteristics of these residents are similar to the HDS neighborhoods. Table 30 has some
descriptions of the categorical variables used in the analyses:

169

Table 30: HDS Breakdown of categorical variables
Variable
Higher Density
Suburban
Number
Percent
RACE
White
359
93.7
Non White
24
6.3
Black
7
1.8
Non-Black
376
98.2
Total
383
100
OCCUPATION
Employed (for wages)
Non-Employed
Total

250
142
392

63.8
36.2
100

The above table identifies the low number of African American respondents in these
neighborhoods (7 respondents who make-up 1.8% of the total number of respondents).
However, there is a greater amount of ‘other’ races mixed in these neighborhoods when
compared to the HDU neighborhoods. The HDS neighborhoods had 93.7% white respondents,
1.8% African American respondents and 4.5% of the respondents belonged to ‘Other’ races.
Two thirds of the respondents were employed for wages. Only a small part of these unemployed
respondents can be attributed to the student population of Ann Arbor, since the average age of
the respondents was about 51 years old. A significant portion of the unemployed consist of
retired residents and stay at home parents, since the number of dependents (children living at
home) was close to 1 and the income of the respondents supported 2.4 people on average. The
table below gives the descriptive statistics for the variables assessed for this neighborhood
typology. It should be recognized that these variables are the same as the variables analyzed for
the HDU neighborhoods.

170

Table 31: HDS Descriptive Statistics
N
Minimum
Variable

Maximum

Mean

Std. Dev.

VehYou
376
0
6.0
1.24
0.64
VehAll
334
0
6
1.95
0.86
VehYear
384
1964
2008
2001.98
5.05
VehCyl
369
4
8
5.23
1.26
VehMiles
377
0
315000
62672.20
52165.86
Age
391
19
90
50.99
16.20
Educ
389
12
19
17.08
2.29
Num_Child
377
0
3
0.54
0.90
Inc_Supports
377
0
8.0
2.44
1.35
PIncome
356
5000
155000
62359.55
44186.48
HIncome
347
5000
155000
96123.92
46665.14
Pct_WorkTrips_W
265
0.00
100.00
16.03
31.00
Pct_WorkTrips_T
265
0.00
100.00
6.66
19.34
Pct_WorkTrips_D
265
0.00
100.00
77.30
37.12
Pct_OptTrips_W
389
0.00
100.00
21.47
27.09
Pct_OptTrips_T
389
0.00
74.47
1.13
6.48
Pct_OptTrips_D
389
0.00
100.00
77.39
28.33
Pct_AllTrips_W
392
0.00
100.00
20.12
25.26
Pct_AllTrips_T
392
0.00
71.43
2.83
9.45
Pct_AllTrips_D
392
0.00
100.00
77.05
27.80
Ann_WorkMiles_W
265
0.00
7040.80
130.68
671.76
Ann_WorkMiles_T
265
0.00
4264.00
64.18
358.51
Ann_WorkMiles_D
265
0.00
17602.00
1926.99
2760.72
Ann_OptMiles_W
391
0.00
4581.20
99.89
298.87
Ann_OptMiles_T
391
0.00
1006.20
9.13
61.21
Ann_OptMiles_D
391
0.00
6749.60
855.59
896.22
Ann_AllMiles_W
392
0.00
7065.50
187.98
639.16
Ann_AllMiles_T
392
0.00
4264.00
52.49
304.52
Ann_AllMiles_D
392
0.00
18502.12
2156.09
2593.39
Ann_Fuel_Consump
355
0.00
925.11
106.31
129.34
Ann_CO_Emissions
356
0.00
77708.90
8870.33
10899.63
* Ann_Fuel_Consump and Ann_CO_Emissions show annual total environmental burdens for
travel to end-point destinations.

171

Cluster Analyses: The Higher Density Suburban Neighborhood Typology
The first analysis performed was the Cluster Analysis. As with the urban neighborhood
typology, this analysis was done twice, once using only the demographic variables and the
second time including the travel behavior variables. The Cluster Analysis on the HDS
neighborhoods with demographic variables only revealed groups that were varied in: 1)
household and personal incomes; 2) year of vehicles and the number of miles on the vehicles; 3)
the number of vehicles owned; and 4) race (see Table 32). With all the variables included in the
second Cluster Analysis, the resultant groupings varied a lot more by their travel patterns than
their socio economic patterns. Essentially they differed by: 1) the work and non-work trips
undertaken by walking and driving; 2) the emissions and fuel consumption patterns; and 3) the
use of transit for trips related to work and all trips taken together (see Table 33).
For each of the two Cluster analyses, first the clusters with low membership numbers
(<4) are discussed, followed by the discussion of the clusters with sufficient membership
numbers (>=4).

172

Table 32: Cluster Analysis on Demographic and Socio-economic Variables - HDS
CL Employment
Income
Dependants
Veh_Age
Veh_Ownership
Race
Empl Prof PIncome HIncome Inc_Sup
Child Year Miles
You
All
Wh Bl
1
0.7
0.6
71271
108136
2.4
0.5 2005
33380
1.2
1.9 1.0
0
2
0.8
0.6
55482
92831
3.0
1.0 1998
110069
1.1
1.9 1.0
0
3
0.4
0.4
63000
81000
2.2
0.4 2000
88825
1.2
1.8
0 1.0
4
0.6
0.5
66000
105500
2.8
0.4 2003
49740
2.7
3.6 0.9
0
Table 33: Cluster Analysis on All Variables - HDS
CL
Non-Work Trips
Energy Use
Transit (Mode)
Work
%-Walk %-Drive Non-WorkMile_W
Fuel
CO
%WrkTrps %AllTrps %Trp_W %Trp_D
1
32.9
67.1
3.9
4.9
9.3
1.4
0.4
17.0
81.6
2
55.7
44.2
4.7
4.2
8.7
13.7
6.9
27.4
58.9
3
18
82
1.3
4.6
9.1
3.2
1.5
19.5
77.4
4
13.2
86.8
1.1
4.1
8.5
0.7
0.5
16.9
82.4
5
41.7
58.3
2.7
3.4
7.9
76.7
40.5
0
23.3
6
0
100
0
4.4
8.7
0
0
0
100
7
22.2
77.8
1.3
3.3
7.7
70
40.5
17.8
12.2
8
44.1
45
5
3.6
8.0
30.6
17.7
30.6
38.9

173

Table 33 (cont’d)
CL
SES
1
2
3
4
5
6
7
8

Admin Educ
Veh_All
0
18.4
2.1
0
18
1.7
0.8
14.6
2.1
0
17.9
1.9
1
15.3
2
0
14.3
1.5
0
18
1.7
0.3
18.5
1.8

Non-Work Trips by
Dependents
Veh_Age
Transit
Miles
%Trips
Inc_Supp Child V_Year V_Miles
0
0
2.7
0.7
2002
69876
1.1
0.1
2.3
0.7
1999
108000
0
0
2.5
0.5
2004
52097
0
0
2.9
0.9
2003
57236
0
0
2
0
2001
63667
0
0
1.5
0
2000
63000
0
0
1.7
0
1997
156667
4.6
10.9
3.2
1
2000
77667

Table 33 (cont’d)
CL
Race
Gender/Laborer
Wh Bl
D_Empl Labor Gender
1
1
0
0.8
0
0.6
2
0
1
0.7
0
0.3
3
1
0
1
0
0.2
4
0.9
0
1
0
0.4
5
1
0
1
0
0.7
6
1
0
1
1
0.8
7
0.7
0
0.3
0
0.3
8
1
0
1
0
0.3

174

The results of the analysis on this neighborhood type, while in many ways producing
similar outcomes as in the urban neighborhood grouping, also revealed some differences. These
differences are likely attributable to the built environment characteristics of the higher density
suburban neighborhoods, which have a richer mix of amenities within and in proximity of the
neighborhoods, giving the residents the option of driving, using public transit or walking in
reaching various destinations.
For the K-Means cluster analysis with 4 groups done on the socio-economic variables,
none of the clusters had a deficient number of members. Results are presented below:
o Cluster 1 identified a group of all white respondents who had higher than average personal
and household incomes. Their vehicles were the newest (4 years old on an average) with
the fewest miles on the vehicles.
o Cluster 2 identified a group of all white respondents who had the lowest personal incomes
(average was $55,482), had more dependents, older cars (11 years old on an average) and
higher mileage on the vehicles.
o Cluster 3 identified a group of all African American residents who had below average
household incomes, older cars (9 years old on average) with higher mileage on the cars.
o Cluster 4 identified a majority white group of residents that had high vehicle ownership and
higher than average household incomes.

175

The Cluster Analysis with all variables (demographic and travel behavior) resulted in 8
clusters, with three of them having very low membership numbers.
o Cluster 2 identified a group of 3 members that are African American, two females (who are
employed for wages) and a male. This group is highly educated, has older cars (10 years
old on an average) and therefore higher mileage on their vehicles. What is unique about
this group is the usage of walking as a mode of travel. More than half of their non-work
trips and about 27% of their work trips are done by walking. Consequently, driving is
chosen less frequently as a mode of travel.
o Cluster 5 also had 3 members. This cluster identifies a group of 2 males and 1 female
white respondent who are all employed in the Administrative and Service professions.
Their unique characteristic is that they use transit for two-thirds of their trips for work.
They do not use transit for non-work purposes, rather they walk about 42% of the time for
their non-work trips. Consequently, they consume less fuel and emit fewer pollutants.
o Cluster 7 is a group of mostly females, the majority of who are not employed for wages.
They are extensive users of transit for work purposes and walk for 1/5 of their non-work
trips. They have older cars with very high mileage on the vehicles, however, they consume
less fuel and emit fewer pollutants.
o Cluster 1 is a group of all white respondents, most of them employed for wages and they
are highly educated. This group has over 2 vehicles in the household, and the vehicles are
extensively used for travel (over 80% of work trips and over 67% of non-work trips are by
driving). Consequently, this group is responsible for the highest levels of fuel consumption
and they emit the highest levels of carbon monoxide.

176

o Cluster 3 is a group of white respondents, mostly females who are employed for wages in
the Administrative and Service professions. This group drives most of the times to their
destinations, while they use walking as a mode for about a fifth of their work and non-work
trips. Consequently, they are the second highest fuel consumers and polluters.
o Cluster 4 is a group of white respondents, all of them highly educated and employed for
wages. This group has a higher number of dependents at home (number of people the
income supports and the number of children under the age of 18). They do not use transit
for their trips, they walk to some of their non-work trips, but they mostly drive for the
majority of their trips, the all-purpose trips. They have newer cars with fewer miles on
them and are average polluters and fuel consumers.
o Cluster 6 is a relatively small group of white males that have lower levels of education and
are all employed as laborers. This group of people have no children living at home, and
their unique characteristic is that they drive to all of their trips (work and non-work), yet
they are not the highest polluters and fuel consumers.
o Cluster 8 is a group of white mostly females who are highly educated and have a higher
number of dependents that their income needs to support. This group uses transit for 10%
of their non-work trips and over 30% for their work trips, and they walk for just under 50%
of their non-work trips. All this suggests that they drive less and consequently they are low
consumers of fuel and are low emitters of pollutants.

177

Regressions Analyses: The Higher Density Suburban Neighborhood Typology
The variables in this file were also transformed as necessary to get their distributions as
close to normal as possible. Yet, in this file too, the range of percent traveled by different modes
was skewed and no transformation made those better. This is why they were kept as is for the
analyses. Appendix F shows the before and after transformation statistics. In testing the
hypotheses, the regression analyses showed the following results, discussed below by trip
purpose and presented in tables 34 and 35 by mode of travel:


For work trips (see Tables 34 and 35):
o There was no linear relationship between socio-economic characteristics and
walking to work trips [Hypothesis C]
o The percentage of travel undertaken by public transit had a significant and
negative association with household income [Hypothesis B]
o The percentage of travel undertaken by driving as the mode had a significant and
positive association with household income and whether the respondents were
employed (for wages) [Hypothesis A]
o The analysis on the total annual miles for trips undertaken by walking did not
show any linear relationships [Hypothesis F]
o The total annual miles for trips undertaken by public transit had a significant and
positive association with whether the respondent was black. [Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with whether the respondent was employed (for wages) and a
negative association with the age of the respondent [Hypothesis D]



For the non-work trips (see Tables 34 and 35):
178

o The percentage of travel undertaken by walking as the mode had a significant and
negative association with household income [Hypothesis C]
o The percentage of travel undertaken by public transit had a significant and
negative association with household income [Hypothesis B]
o The percentage of travel undertaken by driving as the mode had a significant and
positive association with household income. [Hypothesis A]
o This analysis showed no linear relationships between annual miles walked to nonwork destinations and socio-economic characteristics [Hypothesis F]
o This analysis showed no linear relationships between annual miles by public
transit to non-work destinations and socio-economic characteristics [Hypothesis
E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with household income and a negative association with
whether the respondent was black [Hypothesis D]


For all trips (see Tables 34 and 35):
o The percentage of travel undertaken by walking as the mode had a significant and
negative association with household income [Hypothesis C]
o The percentage of travel undertaken by public transit had a significant and
negative association with household income [Hypothesis B]
o The percentage of travel undertaken by driving as the mode had a significant and
positive association with household income [Hypothesis A]
o The total annual miles for trips undertaken by walking had a significant and
negative association with the age of the respondent [Hypothesis F]

179

o The total annual miles for trips undertaken by public transit had a significant and
negative association with household income and the age of the respondent
[Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with household income and whether the respondents were
employed (for wages), and a negative association with the age of the respondent
[Hypothesis D]

180

Table 34: HDS Regression results - Mode
Variable

Constant

Driving Trips
Non-Work All

Work

32.564

***

58.614

Transit Trips
Non-Work All

Work

***

50.556

***

39.099

Walking Trips
NonAll
Work

Work

***

4.090 14.517

28.337

**

37.296

**

34.927

# Dependant Children

2.053

-0.800

0.117

-1.995

0.604

-0.582

-0.058

0.196

0.466

Dummy_Black

4.144

-5.991

6.948

-3.362

-0.643

-2.000

-0.782

6.634

-4.948

Dummy_White

9.301

0.639

3.966

-7.127

1.466

-1.540

-2.173

-2.105

-2.426

-0.172

-1.101

-0.992

-0.475

0.025

0.074

0.647

1.077

0.917

2.491

3.463

-2.132

-0.739

0.254

-1.753

-3.717

Education
Dummy_Employed
Hshld_Income
Age

R

**

0.072

***

0.096

***

0.092

-0.050

**

***

-0.019

-0.032

***

-17.507

**

-0.023

***

-0.077

***

-0.060

-0.054

2

0.131

0.177

0.023

0.024

-0.030

0.031

-0.155

-0.147

.080

2

Adj. R

**

19.638

.082

.085

.078

.064

.079

.040

.065

.049

.050

.061

.064

.048

.043

.058

.009

.043

.027

N
224
312
314
224
312
314
224
312
314
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

181

Table 35: HDS Regression results – Annual Miles
Driving Miles
NonAll
Work

Variable

Work

Constant

5.333***

# Dependant Children

0.000

Transit Miles
Non-Work All

Work

**

5.094*** 5.556*** 2.773
-0.146
*

-1.231

-0.122

-0.339

0.190

Work

**

2.339

1.922

Walking Miles
Non-Work All

**

3.190

**

3.604

0.102

-0.170

-0.031

0.076

-0.041

0.198 2.494

0.441

1.339

1.720

-0.158

0.440

**

Dummy_Black

0.861

Dummy_White

-0.581

0.374

0.027

0.027

0.348

0.169

-0.467

0.094

0.005

Education

-0.070

0.011

0.013

-0.021

0.025

0.010

0.033

0.073

0.081

-0.232

-0.047

0.200

-0.554

-0.387

-0.069

Dummy_Employed
Hshld_Income

-0.040

2

Adj. R

0.004
***

Age

R

***

2.908

-0.171
***

0.004

-0.007

***

0.904

***

0.004

*

-0.010

-0.003

***

-0.002

-0.005

***

0.000

-0.011

0.001

-0.013*

-0.006

*

-0.003

**

-0.024

-0.003
***

-0.033

.143
2

.061

.165

.073

.032

.067

.025

.036

.046

.115

.039

.146

.043

.010

.045

-.006

.014

.024

N
224
313
314
224
313
314
224
313
314
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

182

Table 36: HDS Regression results – Vehicle Characteristics
Vehicle
Veh_Year

Veh-All

Constant

***

-2.064

Annual Environmental Burdens
Fuel Consumption
CO Emissions

Veh_Cyl

***

1999.812

4.989***

2.768***

6.807***

# Dependant Children

-0.120

0.235

0.187

-0.047

-0.064

Dummy_Black

-0.310

-0.642

-0.749

0.329

0.489

Dummy_White

-0.121

0.459

-0.227

0.202

0.244

-0.063*

-0.003

0.028

Education

-0.047*

Dummy_Employed

-0.230

Hshld_Income

R

2

Adj. R

***

0.005

-0.007

Age

*

*

*

-0.221

0.276
***

0.021

-0.017

***

***

-0.532

0.754

***

**

0.004

***

0.754

0.002

-3.430E-5

*

0.009

0.002
0.000

.142
2

.161

.144

.124

.087

.119

.139

.121

.101

.064

N
270
306
296
284
285
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

183



The number of vehicles owned by the household had a significant and positive
association with household income and a negative association with the age of the
respondent, the years of education received, and whether the respondent was employed
(for wages) (see Table 36) [Hypothesis H]



For those respondents who have access to and use a private vehicle for any part of their
travel activity, the year of the vehicle used for travel had a significant and positive
association with the household income and a negative association with the years of
education received (see Table 36) [Hypothesis G]



For those respondents who have access to and use a private vehicle for any part of their
travel activity, the engine size of the vehicle used for travel had a significant and positive
association with the household income, and the age of the respondent, and a negative
association with the years of education received, and whether the respondents were
employed (for wages) (see Table 36) [Hypothesis I]



For those respondents who have access to and use a private vehicle for any part of their
travel activity, the annual fuel consumption was significantly and positively associated
with household income and whether the respondents were employed (for wages) (see
Table 36) [Hypothesis D]



For those respondents who have access to and use a private vehicle for any part of their
travel activity, the annual carbon monoxide emissions was significantly and positively
associated with whether the respondents were employed (for wages (see Table 36)
[Hypothesis D]

184

The above regression analyses show a class dimension to some of the travel patterns and
environmental burdens (as was the case with the urban neighborhoods). The difference in these
higher density suburban neighborhoods was that race came up as significant in a few of the
regressions, which is interesting and telling. Although the number of African American
respondents was very low, still the fact that they came out as significant in some travel behavior
was insightful. The results show that African Americans travel further distances annually to
work by transit and they are more likely to travel shorter distances annually to non-work
destinations by car. However, both those regression results must be viewed with caution as the
dependent variable, annual miles traveled, had skewed distributions (even after transformations).
This would imply that the results are biased. The regressions on mode of travel showed that
walking to work trips was not significant. However, the transit trips to all three destination types
(work, non-work, and all) should be viewed with caution as these dependent variables were
highly skewed. As such the associations might not be robust. Similarly, the regressions on
annual miles to non-work trips by transit and to work and non-work trips by walking showed
insignificant associations. The variable for annual miles to non-work trips by transit, even after
transformations, was highly skewed and as such this regression result is understandable.
However, the variables for annual miles by walking to work and non-work trips are only slightly
skewed and therefore the insignificant regression results show more a case of non-significant
linear associations, than the result of skewed variable distributions. In order to get a qualitative
look at the range of travel behavior and environmental burdens, the following section explores
this behavior by class of income.

185

A Discussion by Income Groupings: The Higher Density Suburban Neighborhood Typology
Similar to the analysis of the urban neighborhoods, a breakdown of the respondents of the
higher density suburban neighborhoods into income tertiles (see table 37) will allow for further
texture into the analysis of the socio-economics of travel behavior and resulting impacts on
gasoline consumption and CO emissions.

Table 37: HDS Respondents split into three Household Income Groups
Household Income
Count
Percentage
$5,000 to $50,000 (Low Income Range)
66
19%
>$50,000 to $100,000 (Middle Income Range)
114
33%
>$100,000 to $155,000 (Upper Income Range)
167
48%
Total
347
100%

A visible difference emerges as we look at the income breakdown in the HDS
neighborhoods, see table 35. Almost half of the respondents (48%) fall into the highest one-third
category of household income and only 19% fall into the lowest one-third household income
category. The average household income for all the respondents in this neighborhood type was
$95,920. Tables 38 and 39 below give a more detailed look at the characteristics of these three
groups of respondents within the higher density suburban neighborhoods.

Table 38: HDS Vehicle Ownership
Vehicles for you
Vehicles for all in household
Vehicle Year

Income Range
Low
Middle
Upper
0.9
1.3
1.3
1.4
1.9
2.2
2000
2001
2003

186

Table 39: HDS Demographic Characteristics
Age
Education
Number of people the Income Supports
Number of children living at home

Income Range
Low
Middle
Upper
50.9
49.3
49.7
15.6
17.0
17.8
1.6
2.3
3.0
0.2
0.4
0.8

Tables 38 and 39 show that the upper income group of respondents has newer and more
vehicles for use by the respondents themselves and by everyone else in the household. Although
there is not much of a difference in the ages of the respondents between the three income groups,
the education received (as would be expected) increases as the income of the sub-grouping
increases. Also relevant is that the number of dependents increases with household income. The
upper income range residents have about 3 people that their income supports, and a child under
the age of 18 years living at home, while the lowest income range respondents support less than
2 people on their household income, and most of these households do not have any children
under the age of 18 years living at home.
Table 40: HDS Mode of Travel Based on Usual Weekly Travel Patterns
Income Range
Low
Middle
Upper
% WorkTrips by Walk
28.9
13.2
14.9
% WorkTrips by Transit
21.1
5.8
3.5
% WorkTrips by Drive
50.0
81.0
81.6
% Non-Work Trips by Walk
32.1
20.0
19.0
% Non-Work Trips by Transit
3.7
0.4
0.6
% Non-Work Trips by Drive
64.2
79.6
80.4
% All Trips by Walk
29.3
19.0
18.1
% All Trips by Transit
8.4
2.0
1.6
% All Trips by Drive
62.3
79.1
80.3

187

Table 41: HDS Average Annual Miles to End-Point Destinations by Trip Purpose Based on
Usual Weekly Travel Patterns
Income Range
Low
Middle
Upper
Annual Miles for WorkTrips by Walk
74.9
47.1
228.4
Annual Miles for WorkTrips by Transit
126.0
35.2
77.5
Annual Miles for WorkTrips by Drive
478.9
1963.6
2198.7
Annual Miles for Non-Work Trips by Walk
146.3
101.7
67.8
Annual Miles for Non-Work Trips by Transit
30.5
6.6
3.6
Annual Miles for Non-Work Trips by Drive
789.2
858.1
896.7
Annual Miles for All Trips by Walk
189.4
136.3
233.1
Annual Miles for All Trips by Transit
103.1
32.4
59.9
Annual Miles for All Trips by Drive
1064.9
2300.4
2484.8

Tables 40 and 41 above show the travel patterns of the residents within each income
range. As far the mode of travel is concerned, the middle and upper income range residents
behave in a similar manner while those in the lowest income range have a clear difference in
their mode choice. For work trips, the lower income residents walk twice as many times as the
middle and upper income residents and they use public transit 7 times more than the residents in
the middle and upper income groupings.
The non-work trips do not show such drastic differences, yet variations still do exist. The
lowest income group residents walk for 32% of their non-work trips, while the upper income
residents walk for 19% of their non-work trips. Transit choice is more polarized, with the lowest
income residents using transit for 3.6% of the non-work trips while the upper income residents
use it for less than 1% of their non-work trips.
In examining annual miles traveled to end-point destinations by different modes of
transport, table 39 shows an interesting trend. The middle income residents and the upper
income residents do not travel similar distances necessarily (even though their mode choice was
similar). The upper income residents travel longer distances by walking and public transit than
188

the middle income residents. Another interesting point to recognize is that upper income
residents walk longer annual distances to end-point locations, when compared to even the lower
income grouping.
The lowest income residents walk most and travel longer distances by walking when
compared to the respondents in the middle income groupings. Travel by car shows similar
difference in travel patterns by class, with the upper income residents travelling about 4.5 times
the amount that the lower income residents travel to work annually by car.
Similar to the urban neighborhoods, this analysis shows that upper income residents
travel further from their suburban neighborhoods for work purposes by car. Non-work trips
show a lesser difference. The upper income residents travel only slightly more in distance that
the lower income residents do for non-work trips, and as a result, the total annual distance
traveled for all trip purposes shows the upper income residents travelling about 2.3 times the
distance that the lower income residents do by car.

Table 42: HDS Average Environmental Burdens Based on Usual Weekly Travel Patterns *
Income Range
Low
Middle
Upper
Annual Fuel Consumption (gallons)
115.4
216.4
261.6
Annual Carbon Monoxide Emissions (grams)
9652.6 19058.6
21172.2
*The environmental burdens calculations are a result of both end- and return-destination trips.

The environmental burdens table shows a consistent trend. Since upper income residents
travel about 2.3 times the distance that lower income residents travel, by car specifically, they
consume about 2.3 times the fuel and emit about 2.2 times the pollutants. Similar to the

189

environmental burden patterns among HDU neighborhoods, the wealthier drive more, consume
more gasoline, and are responsible for higher amounts of pollutant emissions, all while
controlling for built environment characteristics.

Discussion
What we see from the analyses on the higher density suburban neighborhoods is a
tempered variation of travel that was evident in the urban neighborhoods. The two
neighborhoods in the HDS grouping do not belong to the same city, nor are they geographically
adjacent to each other, but they are grouped into this category due to their similarities in the built
environment and their similarities in socio-economic characteristics. These are robust
neighborhoods in cities with strong urban cores, at least for a Detroit regional context, a robust
land use mix and a connected street system. They are also dynamic and robust neighborhoods
within wealthy cities. Both sets of the cluster analyses showed groupings that were primarily
based on income, education, employment and travel patterns that consequently affect fuel
consumption and emissions.
The regression analyses also brought out the class dimension associated with
socioeconomic characteristics, travel, and environmental burdens. Although a few results
showed race as a significant factor as well, the analyses led us to accept the hypotheses laid out
earlier, that residents with higher socio-economic characteristics travel more by car, consume
more gasoline and are responsible for higher amounts of pollutant emissions than those with
lower socio-economic characteristics. When controlling for the built environment in these
neighborhoods, the difference visible in travel behavior and environmental burdens all point to
190

socio economic characteristics as an important influence; wealthy people travel more by car,
travel longer distances by car, consume more gasoline and pollute more. Similar to the results in
the urban neighborhood typology, age also gets identified as a significant factor in travel
behavior in the higher density suburban neighborhoods. Although it is not a variable that
indicates socio economic status, it does affect the probability of using alternate travel modes for
various trip purposes. The older population generally prefers traveling by car than walking or
taking public transit.
The HDS neighborhood type also revealed that the middle and upper income range
respondents had very similar patterns relating to mode of travel for the different trip purposes,
but the distance traveled was not that similar. The upper income range respondents walked
further distances annually to all trips than the middle income resident grouping. This is
interesting as you would expect the upper income group to walk less. This might be the result of
the rich amenities in these neighborhoods—or access to rich amenities in the vicinity of these
neighborhoods—and the greater opportunities for walking to various necessary destinations. The
upper income residents might also be more health conscious and be willing to walk more as part
of a healthier lifestyle. While this difference in walking (by the upper and middle income
respondents) is evident, it is very small in comparison to the percentage of trips and miles
walked annually by the lower income range respondents, with the lower income range
respondents walking much more and longer distances for all trip purposes. This is again the
class effect at play. The upper income range respondents drove 2.5 times the distances of the
lower income range respondents for all trips, a substantial variation in travel patterns.
The ‘tempered’ part in the HDS neighborhoods comes from the fact that in the analysis of
the HDU neighborhoods, the respondents in the upper income grouping drove for ten times more
191

trips when compared to the respondents in the lower income grouping. This type of polarized
travel pattern is not evident in the HDS neighborhoods, where the upper income residents
traveled about 2.3 times the distance, consumed about 2.3 times the fuel and were responsible for
about 2.3 times the pollutant emissions. Again, this likely speaks to the availability of richer
amenities—and the greater accessibility to daily necessary destinations—in proximate distance
within the higher density suburban neighborhoods. However, it does reinforce, once again, the
fact that the wealthier drive more and are responsible for higher amounts of environmental
burdens than the lower income residents.

192

Low Density Suburban – Bloomfield Hills and West Bloomfield Neighborhoods
The built environment of the lower density suburban (LDS) neighborhoods is structurally
very different from the urban (HDU) and higher density suburban (HDS) neighborhoods. These
are the lowest density neighborhoods, they are dominated by single land uses, predominantly
residential, and are disconnected, being developed around the curvilinear street system. The
socio economic characteristics of the residents also show a less diverse residential base, in terms
of class. The incomes are high—and as might be expected given the built environment and
incomes, and given the results so far from the analyses—the travel behavior of these respondents
is heavily automobile-dependent. Since transit is not prevalent in these neighborhoods (the
SMART transportation system has a couple routes that come into the peripheries of these
neighborhoods) travel by transit is minimum. Table 43 below shows a breakup of the categorical
variables used in the analyses.
Table 43: LDS Breakdown of Categorical Variables
Variable
Low Density
Suburb
Number
Percent
RACE
White
281
88.6
Non White
36
11.4
Black
11
3.5
Non-Black
306
96.5
Total
317
100
OCCUPATION
Employed (for wages)
Non-Employed
Total

186
134
320

58.1
41.9
100

These neighborhoods have a greater mix of residents than the urban or high density
suburban neighborhoods. Whites make up about 89% of the residents, African American

193

respondents make up 3.5% and respondents of other’ races make up the remaining 7.5% of this
neighborhood typology. The increase in the ‘other’ races reinforces the urban growth theories
that suggest that immigrants that come in as highly qualified take up jobs and residence in the
suburban neighborhoods and do not go through the cycles that typically the white, Hispanic and
African Americans have gone through. A little less than 60% of the respondents were employed
for wages. This might reflect a higher number of stay at home parents and the higher average
age of the respondents (57 years). Table 44 presents the descriptive statistics for the variables
used in the analyses of this neighborhood type. Again, these are the same variables used in the
urban and higher density suburban analyses.

194

Table 44: LDS Descriptive Statistics
N
Minimum
Variable

Maximum

Mean

Std. Dev.

VehYou
318
0
4
1.35
0.64
VehAll
276
0
9
2.22
0.95
VehYear
323
1988
2008
2003.91
3.67
VehCyl
303
4
8
5.81
1.27
VehMiles
319
200
200000
42590.57
39295.87
Age
317
18
92
57.04
15.03
Educ
324
10
19
16.75
2.52
Num_Child
314
0
4
0.65
1.01
Inc_Supports
309
0
10
2.73
1.44
PIncome
278
5000
155000
73417.27
49971.24
HIncome
273
5000
155000
108516.48
44853.69
Pct_WorkTrips_W
196
0.00
100.00
5.53
18.65
Pct_WorkTrips_T
196
0.00
16.67
0.09
1.19
Pct_WorkTrips_D
196
0.00
100.00
94.38
18.67
Pct_OptTrips_W
323
0.00
87.50
6.83
15.39
Pct_OptTrips_T
323
0.00
27.27
0.18
1.85
Pct_OptTrips_D
323
12.50
100.00
92.99
15.45
Pct_AllTrips_W
325
0.00
83.12
6.92
14.84
Pct_AllTrips_T
325
0.00
27.27
0.20
1.89
Pct_AllTrips_D
325
16.88
100.00
92.88
14.92
Ann_WorkMiles_W
196
0.00
7072.00
88.59
553.07
Ann_WorkMiles_T
196
0.00
572.00
2.92
40.86
Ann_WorkMiles_D
196
0.00
69940.00
3354.82
6112.06
Ann_OptMiles_W
324
0.00
2802.80
68.11
283.20
Ann_OptMiles_T
324
0.00
577.20
2.28
32.66
Ann_OptMiles_D
324
0.00
18153.20
1921.06
1972.68
Ann_AllMiles_W
325
0.00
7072.00
121.32
511.75
Ann_AllMiles_T
325
0.00
577.20
4.03
45.41
Ann_AllMiles_D
325
0.00
72867.60
3938.36
5406.37
Ann_Fuel_Consump
286
0.00
4857.84
212.92
353.03
Ann_CO_Emissions
287
0.00
181686.96
16300.46
18430.98
* Ann_Fuel_Consump and Ann_CO_Emissions show annual total environmental burdens for
travel to end-point destinations.

195

Cluster Analyses: The Low Density Suburban Neighborhood Typology
Starting with the Cluster Analyses on demographic variables, the results revealed
groupings based on: 1) race; 2) number of dependents in the household; 3) employment (for
wages) and employment in the professional occupation categories; and 4) the number of vehicles
owned (see Table 45). The second of the last set of analyses, on all variables, revealed groups
that differed on: 1) work and non-work trips undertaken by walking and driving; 2) fuel
consumption and emissions generated; 3) vehicles owned; and 4) the year of and miles on the
vehicles used for travel purposes (see Table 46). For each of the two Cluster analyses, first the
clusters with low membership numbers (<4) are discussed, followed by the discussion of the
clusters with sufficient membership numbers (>=4).

196

Table 45: Cluster Analysis on Demographic and Socio-economic Variables – LDS
CL
Dependants
Employment
Veh_Age
Race
Veh_Ownership
Inc_Sup Child D_Emp D_Prof
Year Miles Wh
Bl
You
All
1
4.4
1.8
0.8
0.7 2006 27901
0.9
0
1.1
2.4
2
2.8
0.5
0
0 2000 79850
0.9
0
1.8
2.4
3
3.1
0.7
1
0.9 2000 94568
1
0
1.3
2.3
4
3
0.4
0.9
0.8 2006 26659
1
0
2.5
3.2
5
2
0.1
0.6
0.5 2006 20583
1
0
1.2
1.7
6
1
0
0.3
0.3 2007
7667
0
0.7
1
1.7
7
5
1.7
0.7
0.6 2004 61957
0
0.9
1
2.9
8
6
1
1
1 2007
800
1
0
3
9
Table 46: Cluster Analysis on All Variables – LDS
CL
Energy Use
Work Trips
Non-Work Trips
Employment
Fuel
CO
%Trps_W
%Trps_D %Trp_W %Trp_D D_Empl D_Prof
1
4.8
9.2
5.1
94.9
1.9
98.1
0.7
0.6
2
4.9
9.3
3.1
96.9
34.5
65.5
0.8
0.8
3
3.7
8.1
100
0
21.4
78.6
0.5
0.5
4
5.6
9.9
2.1
97.7
1.4
98.6
0.9
0.9
5
5.6
10.1
11.9
88
2.1
97.9
1
0.9
6
5.1
9.4
0
100
1.8
98.2
0.7
0.6
Table 46 (Contd.)
CL
Dependants
Veh_Age
Race
Veh_Ownership
Walk (Mode)
Inc_Sup
Child
V_Year V_Miles Wh Bl You All
AllMiles WrkMiles
1
3.9
0.6
2005
36497 0.9
0
2.7
3.3
1.2
0.9
2
3.3
1.1
2004
44392
1
0
1.4
2.2
5.2
0.5
3
3.5
1
2007
17050
1
0
1
1.5
0
0
4
3.1
0.9
2006
24716 0.9
0
1.1
2.1
0.7
0.5
5
3.2
0.8
2000
98638 0.9
0
1.2
2.5
1.8
1.4
6
4
1.3
2005
47557
0 0.7
1
2.7
0.6
0
197

The 8 group K-Means Cluster analysis on the demographic variables yielded two clusters
that had less than 4 members. Results are discussed below:
o Cluster 6 is a group of 3 members, all with a low number of dependents and no children
living at home. The unique part is that this group has 2 African American members and 1
member of another race, and they have the newest cars with the lowest mileage on them.
Two of the three respondents in this group are also not employed for wages.
o Cluster 8 identified one member who is a white respondent that is employed as a
professional, and has a high number of dependents. The uniqueness comes from the fact
that this respondent has a large number of cars accessible to the respondent and the
household (3 for the respondent and 9 for all members of the household). The respondent
also has a new car (2 years old) with very low miles on it.
o Amongst the other clusters, Cluster 1 has an all-white membership of respondents who
have a relatively high number of dependents (4.4) and children living at home (1.8).
They have newer cars (3 year old) with relatively low miles on them.
o Cluster 2 identified a group of almost all white respondents who do not work for wages.
o Cluster 3 identified a group of all white respondents, who are employed as professionals,
and who have vehicles that are older (9 years old) with most miles on them (almost
95,000 miles on average).
o Cluster 4 is composed of white respondents that are mostly employed as professionals.
They have newer cars (3 years old) and a higher number of cars accessible to them (about
2.5 cars on an average).

198

o Cluster 5 has a low number of dependents and almost no children under 18 years living at
home. Relatively few of these respondents are employed for wages, yet they have newer
cars with low miles on them.
o Cluster 7 identified a group of mainly African American respondents, most of who are
employed for wages. This group also has a high number of dependents (5) and children
under the age of 18 living at home (1.7).

The 6 group Cluster analysis on all variables identified only one group with less than 4
members.
o Cluster 3 identified 2 members who are white, have relatively new cars, but the unique
attribute of this group is that they walk to all their work based trips, however, the annual
mileage for walking to work based trips is 0. This might suggest that they have home
based businesses or live in the same structure as their business is in.
o Cluster 1 identified an almost all white group of respondents that have a high number of
dependents, and also a high number of cars for themselves and their household. Most of
their trips for all purposes are by driving, and they have average aged cars that are 4 years
old with relatively low miles on them.
o Cluster 2 identified a group of white respondents who walk to 1/3 of their non-work trips,
the highest average percentage for this neighborhood type.
o Cluster 4 identified a group of white respondents who have high levels of energy use.
They consume the most fuel and are responsible for the highest levels of pollutant
emissions, the carbon monoxide emissions. As would be expected, they drive to most to
their destinations.
199

o Cluster 5 consists of a group of white respondents who have the oldest cars (9 years old)
and have the highest miles on their vehicles. They walk to work about 12% of the time,
and usually drive for non-work trips. Due to their older cars, they are also high energy
users. They are responsible for relatively higher fuel consumption levels and carbon
monoxide emissions.
o Cluster 6 is a group of mostly African American respondents who have a high number of
dependents (4) and children living at home (1.3), the highest numbers for this
neighborhood type. This group drives to almost all of their destinations, be it for work
purposes or for non-work purposes.

Regression Analyses: The Low Density Suburban Neighborhood Typology
In order to test the hypotheses, the next set of analyses are the Regression Analyses. The
Regression Analyses are again broken down into the assessment of ‘work based trips’, ‘nonwork based trips’ and ‘all trips’. The results show the following outcomes:


For work based trips (see tables 47 and 48):
o The analysis on the percentage of travel undertaken by walking as the mode did
not show any linear relationships with socio-economic characteristics [Hypothesis
C]
o The analysis on the percentage of travel undertaken by public transit as the mode
did not show any linear relationships with socio-economic characteristics
[Hypothesis B]

200

o The analysis on the percentage of travel undertaken by driving as the mode did
not turn out significant [Hypothesis A]
o The analysis on the total annual miles for trips undertaken by walking did not
show linear relationships with socio-economic characteristics [Hypothesis F]
o The analysis on the total annual miles for trips undertaken by public transit did
not show linear relationships with socio-economic characteristics [Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with whether the respondent was employed (for wages) and
household income [Hypothesis D]


For the non-work trips (see tables 47 and 48):
o The analysis on the percentage of travel undertaken by walking as the mode did
not show linear relationships with socio-economic characteristics [Hypothesis C]
o The analysis on the percentage of travel undertaken by public transit as the mode
had a negative association with being black and being white and had a positive
association with the age of the respondent [Hypothesis B]. Caution would need to
be raised here as even though the regression is significant on the whole, there
were only a handful of respondents who stated that they travel by public transit
and as such, this result is unrealistic.
o The analysis on the percentage of travel undertaken by driving as the mode did
not show linear relationships with socio-economic characteristics [Hypothesis A]
o The analysis on the total annual miles for trips undertaken by walking did not
show linear relationships with socio-economic characteristics [Hypothesis F]

201

o The analysis on the total annual miles for trips undertaken by public transit had a
negative association with being black and being white and had a positive
association with the age of the respondent [Hypothesis E]. Again, there were
only a few respondents that said they travel by public transit and so this result
should be read with caution.
o The total annual miles for trips undertaken by driving had a significant and
negative association with whether the respondent was employed (for wages) and
the number of children (under 18 years) living at home [Hypothesis D]


For all trips (see tables 47 and 48):
o The analysis on the percentage of travel undertaken by walking as the mode did
not show linear relationships with socio-economic characteristics [Hypothesis C]
o The analysis on the percentage of travel undertaken by public transit as the mode
had a negative association with being black and white, and a positive association
with the age of the respondent [Hypothesis B]. As mentioned earlier, caution
would need to be raised here as even though the regression is significant on the
whole, there were only a handful of respondents who stated that they travel by
public transit and as such, this result needs to be viewed with caution.
o The analysis on the percentage of travel undertaken by driving as the mode
showed no linear relationships with socio-economic characteristics [Hypothesis
A]
o The analysis on the total annual miles for trips undertaken by walking did not
show linear relationships with socio-economic characteristics [Hypothesis F]

202

o The analysis on the total annual miles for trips undertaken by public transit did
not show linear relationships with socio-economic characteristics [Hypothesis E]
o The total annual miles for trips undertaken by driving had a significant and
positive association with household income and whether the respondent was
employed (for wages) and a negative association with the age of the respondent
[Hypothesis D]

203

Table 47: LDS Regression results - Mode
Driving Trips
Variable
Work
Non-Work All

Constant

106.531

Work

Transit Trips
Non-Work All

Work

Walking Trips
Non-Work All

***

87.551***

-0.785

1.184

0.797

-5.746

15.561

11.653

83.255

# Dependant Children

1.123

2.146

2.325

-0.153

0.043

-0.043

-0.970

-2.189

-2.281

Dummy_Black

0.852

5.299

4.673

-0.055

-1.570**

-1.470**

-0.798

-3.729

-3.203

Dummy_White

-6.061

-1.795

-2.886

0.022

-1.785***

-1.676***

6.039

3.581

4.563

Education

-0.778

-0.361

-0.548

0.031

-0.044

-0.028

0.747

0.405

0.576

Dummy_Employed

-0.869

-0.636

0.276

0.042

-0.018

-0.009

0.827

0.654

-0.267

0.025

0.025*

0.022

0.001

0.001

0.001

-0.026

-0.026*

-0.023*

*

*

**

*

Hshld_Income
Age

R

-0.054

2

Adj. R

0.155

0.152

0.001

0.019

0.018

0.053

**

-0.174

**

-0.170

.028
2

.030

.035

.013

.078

.066

.027

.0335

.041

-.016

.002

.007

-.032

.051

.039

-.017

.007

.013

N
163
248
250
163
248
250
163
248
250
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

204

Table 48: LDS Regression results – Annual Miles
Driving Miles
Variable
Work
Non-Work All
***

Constant

3.754*

6.705

# Dependant Children

-0.170

-0.243

**

Work
***

7.078

-0.161

0.055

-1.772

2.095

0.778

-0.058

-0.004

-0.046

0.081

-0.001

-0.025

-0.349

-0.123

0.054

-0.044

-0.377

0.172

-0.021

Dummy_White

-0.020

0.067

-0.035

0.008

0.016

0.049

0.039

0.012

**

***

Hshld_Income
Age

R

**

0.009

-0.341

0.001

0.453

**

0.003

**

-0.016

-0.340

***

*

3.056E-5

0.028

0.009

0.051

0.016

-0.014

0.020

0.539

-0.191

0.274

0.000

0.000

0.000

0.001

-0.002

-0.001

0.000

.056

.159

.013

.079

.028

.135

-.032

*

0.004

1.009

**

-0.008

-0.010

-0.397

***

0.464

.119
2

**

-0.393

-0.031

2

Adj. R

1.314*

Walking Miles
Non-Work All

0.233

1.485

Dummy_Employed

Work

-0.299

Dummy_Black

Education

Transit Miles
Non-Work All

**

-0.028

1.198

*

-0.023

0.004

0.012

.077

.037

.028

.039

.044

.051

.010

-.016

.011

.016

N
163
249
250
163
249
250
163
249
250
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

205

Table 49: LDS Regression results – Vehicle Characteristics
Vehicle
Veh-All
Veh_Year Veh_Cyl
***

Constant

1.069 2002.450

# Dependant Children

0.067

Dummy_Black
Dummy_White
Education
Dummy_Employed
Hshld_Income
Age

R

8.803

***

5.588***

4.253***

0.297

0.013

-0.019

-0.038

0.236

0.167

-0.130

0.143

-0.200

0.121

-1.423

-0.463

-0.554**

-0.455

-0.003

-0.127

-0.058

0.013

0.017

0.118
***

0.004

-0.148
***

0.011

***

0.562

-0.157
**

***

0.003

0.004

**

-0.010

0.590
0.003

***
***
**

-0.014

-0.004

2

0.027

0.013

.109

2

Adj. R

Annual Environmental Burdens
Fuel Consumption
CO Emissions

.056

.048

.255

.216

.079

.025

.014

.230

.190

N
215
246
232
219
220
Note: * p < 0.1; ** p < 0.05; *** p < 0.01; Values reported are coefficients; Significant coefficients are in bold. Shaded columns
represent the regressions that did not result in a good model fit.

206



The number of vehicles owned by the household had a significant and positive
association with household income (see table 49) [Hypothesis H]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the analysis on the year of the vehicle used for travel did not show linear
relationships with socio-economic characteristics (see table 49) [Hypothesis G]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the analysis on the engine size of the vehicle used for travel did not show
linear relationships with socio-economic characteristics (see table 49) [Hypothesis I]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the annual fuel consumption was significantly and positively associated
with household income and whether the respondents were employed (for wages) and
negatively associated with the age of the respondent and whether the respondent was
white (see table 49) [Hypothesis D]



For those respondents that have access to and use a private vehicle for any part of their
travel activity, the annual carbon monoxide emissions was significantly and positively
associated with whether the respondents were employed (for wages), the household
income, and negatively associated with the age of the respondent and the number of
children under the age of 18 years living at home (see table 49) [Hypothesis D]

The regression analyses show that many of the results on travel mode and annual miles
for different trip purposes came out as insignificant. The results suggest that employment for
wages had a positive association with the annual miles of car travel, as did the household
207

income. This can be understood as the socio economic characteristics of the residents in this
neighborhood would not significantly shape the variation in travel patterns of the residents using
other modes of travel. Given the existing literature on urban form and travel patterns, this can be
expected due to the extreme built environment characteristics of these neighborhoods—low
density, single use zoned, and disconnected neighborhoods—which are built as highly isolated
residential pods, generally forcing everyone to drive in these neighborhoods, a travel behavior
outcome that likely generated the insignificant results. The highly skewed data on mode travel is
also reflective of this. The transit mode travel regressions that did turn out significant should be
ideally considered insignificant because there were only a handful of respondents (at the most)
who used transit for any part of their travel needs. The same would be the case for some of the
other significant regressions such as the annual miles to non-work and all destinations combined
by driving. Even though these results were significant, the dependent variables mentioned above
have a skewed distribution and therefore the results would be biased. On the other hand, the
environmental burdens did turn out significant and those had a class dimension. So even though
there was not a significant variation in the travel patterns that socio-economic characteristics
were facilitating, there was a significant impact of the socio-economic characteristics on the
environmental burdens (i.e. fuel consumption and pollutant emissions) and this would be the
case primarily due to the type of vehicle that the respondents in different income categories
owned. A more nuanced assessment of this variation can be teased out in the following
qualitative analysis.

208

A Discussion by Income Groupings: The Low Density Suburban Neighborhood Typology
In order to further understand the different characteristics and behavior of respondents,
the respondents were divided into three income groups (low, medium, and high). This income
based structure for the lower density suburban neighborhood typology is as follows.
Table 50: LDS Respondents split into three Household Income groups
Household Income
Count
Percentage
$5,000 to $50,000
38
14%
>$50,000 to $100,000
74
27%
>$100,000 to $155,000
161
59%
Total
273
100%

The concentration of the wealth in the lower density suburban neighborhoods becomes
evident in table 50. Over half of the respondents (59%) fall into the highest one-third of the
household income range, while only 14% belong to the lowest one-third of the household income
range. The average household income for the entire neighborhood type was $108,164. This was
the highest average income among the three neighborhood groupings (HDU, HDS, and LDS).

Table 51: LDS Vehicle Characteristics

Vehicles for you
Vehicles for all in household
Vehicle Year

Income Range
Low
Middle
Upper
1.2
1.4
1.4
1.7
2.0
2.4
2003
2004
2004

209

Table 52: LDS Demographic Characteristics
Age
Education
Number of people the Income Supports
Number of children living at home

Income Range
Low
Middle
Upper
66.9
57.5
52.9
15.1
16.6
17.3
1.7
2.4
3.2
0.2
0.6
0.9

The range in vehicle characteristics among the three income groupings, see table 51,
shows an expected trend. The higher the income, the greater the number of cars in the household
and the newer the cars. The range of vehicle characteristics (such as the number of vehicles and
age of the vehicle) between the low income group and the upper income group is not much
though, so most of the people in this neighborhood type, regardless of income, have similar
vehicle characteristics. The demographic table, on the other hand shows a significant range in
the age of the respondent. The low income groups are much older than the upper income groups.
This could be expected as the older respondents would start entering retirement and their
incomes would be lower. The education and number of dependents in the house also depicts an
expected trend. The older group consists largely of empty nesters and their income needs to
support fewer people.
Table 53: LDS Mode of Travel Based on Usual Weekly Travel Patterns

% WorkTrips by Walk
% WorkTrips by Transit
% WorkTrips by Drive
% Non-Work Trips by Walk
% Non-Work Trips by Transit
% Non-Work Trips by Drive
% All Trips by Walk
% All Trips by Transit
% All Trips by Drive

Income Range
Low
Middle
Upper
5.8
6.7
5.9
0
0
0.1
94.2
93.3
94
9.6
6.8
6.1
0
0.4
0.1
90.4
92.8
93.8
9.0
6.8
6.5
0
0.4
0.2
91.0
92.8
93.4

210

Table 54: LDS Average Annual Miles to End-Point Destinations by Trip Purpose Based on
Usual Weekly Travel Patterns
Income Range
Low
Middle
Upper
Annual Miles for WorkTrips by Walk
0.0
44.0
133.4
Annual Miles for WorkTrips by Transit
0.0
0.0
5.0
Annual Miles for WorkTrips by Drive
1295.3
2386.9
4158.7
Annual Miles for Non-Work Trips by Walk
21.4
119.0
46.4
Annual Miles for Non-Work Trips by Transit
0
8.1
0.5
Annual Miles for Non-Work Trips by Drive
1796.2
1935.9
2012.9
Annual Miles for All Trips by Walk
21.4
147.3
140.9
Annual Miles for All Trips by Transit
0.0
8.0
4.0
Annual Miles for All Trips by Drive
2156.0
3533.5
4957.6

The mode of travel table, table 53, shows a more or less homogenous travel behavior by
the respondents of all three income groups. As mentioned before, public transit is not used on a
regular basis as it is not widely available in these neighborhoods. Walking is used sparingly,
only for non-work trips. Respondents in the lowest income grouping walk for about 9% of their
non-work trips, while respondents in the upper income grouping walk for about 6% of their nonwork trips. Most of the travel in the lower density suburban neighborhood typology, regardless
of income, is by car.
The annual miles table, table 54, shows greater diversity. The public transit mileage is
quite similar and very low. However, the annual miles traveled by driving, shows a clear
distinction by class. Respondents in the upper income grouping travel farther distances by car
for all trip purposes. The greatest variance is for work trips, where the upper income group
travels over 3.2 times the total distance that the lower income grouping travels. For non-work
trips, the upper income group travels a distance that is 1.1 times greater than the total distance
that the lower income grouping travels. For all trips combined, the upper income group traveled
2.3 times the total annual distance to end-point locations in comparison to the lower income
211

grouping. The middle income group has values that fall in the middle of the range that the upper
and lower income groups have.

Table 55: LDS Average Environmental Burdens Based on Usual Weekly Travel Patterns
Income Range
Low
Middle
Upper
Annual Fuel Consumption
194.2
339.0
580.6
Annual Carbon Monoxide Emissions
16306.0
29506.6
41698.6
*The environmental burdens calculations are a result of both end- and return-destination trips.

The table on environmental burdens shows, once again, a familiar and expected outcome.
The upper income sub-group consumes about 3 times the annual fuel (in gallons) in comparison
to the lower income sub-group. The pollutant emissions show similar trends, with the
respondents in the upper income grouping emitting about 2.6 times the annual carbon monoxide
levels (in grams) when compared to respondents in the lower income grouping, while the middle
income group has values that fall in between the upper and lower income sub-groups. Apart
from the variance in the travel behavior, the higher pollutant emissions among the wealthier
could be attributed to the upper income group having more powerful engines and larger vehicles
than the lower income group. These vehicle characteristics would lead to the higher amounts of
fuel required by the vehicles and higher amounts of emissions generated by these vehicles.

212

Discussion
The results of the analyses on the LDS neighborhood type showed clusters that mainly
differed on the number of vehicles, age of the vehicles, miles on the vehicles, annual miles to all
trips by driving, the number of dependents, and age. These clusters indicated that the groups
were formed more on the different socio-economic characteristics than on the travel patterns.
This facet was expected. We know these neighborhoods—due to the characteristics of their built
environment—encourage travel by private vehicle and it was no surprise to see this scale of
automobile dependence. The regressions, even though most on travel behavior were
insignificant, are telling of this facet. These are low density, largely single use zoned, and
disconnected neighborhoods, where everyone drives.
The most of the regressions on the mode of travel and the annual miles traveled for
different purposes were insignificant. The annual miles traveled for all purposes by driving
resulted in household income, employment (for wages) and the age of the respondent as the
characteristics that significantly affected this type of travel. So what this insignificance tells us is
that after holding the built environment constant, there was no significant relationship with the
various socio-economic characteristics and variation in travel behavior. Two things are worth
noting at this point. First, there is not that much of a range in the socio economic characteristics
themselves, and second, there was not much of a difference in travel behavior. Knowing that the
latter is true for these neighborhoods, the regressions show that travel patterns for the residents in
these neighborhoods is heavily dependent on using a private vehicle. The insignificant mode
choice regressions imply this. The annual miles regressions, however, did imply that socio
economic characteristics—like income and employment—have a positive effect, and the age of
the resident has a negative effect on the annual miles for all trips that will be traveled by driving.
213

The breakdown by income sub-groups supports the regression results. It is evident that
the upper income grouping of residents drove more than three times the annual distance to work
than the lower income grouping of residents. For non-work purposes, the lower income range
residents walked the most percentage of trips, compared to the other two income groups, but they
did so for the least total miles. The middle income group walked the greatest distance for these
trip purposes. Combining all the trips, the upper income grouping drove three times the distance
annually when compared to the distances that the lower income grouping drove, and that fact
was translated into much greater environmental burdens. Respondents in the upper income
grouping were responsible for about three times the fuel consumption and 2.6 times the carbon
monoxide pollutant emissions than the respondents in the lower income range grouping.

Commentary
The purpose of this research, and the specific focus of the intra-neighborhood analysis,
was to control for the built environment and tease out the relationships between socioeconomic
characteristics and travel behavior, along with the associated environmental burdens, the annual
gasoline consumption and the annual carbon monoxide emissions. Across all three
neighborhoods, the main finding was that class had a significant relationship with travel patterns
and environmental burdens, both gasoline consumption and vehicle emissions. Even while
living in the same neighborhood, people with different socio economic characteristics behaved
differently. The wealthier traveled more times and longer distances with their private vehicles
than did the lower income residents and consequently, they were responsible for higher

214

environmental burdens. A summary of the travel behavior and environmental burdens by
income group for the three neighborhood typologies is presented in table 56.
The Cluster Analyses on each neighborhood type showed that the suburban
neighborhoods, both HDS and LDS, had larger variations (resulting in clusters) in travel patterns
than in the socio-economic characteristics of the residents. The variations in socio-economic
characteristics were more prevalent in the urban neighborhoods (HDU) and the higher density
suburban (HDS) neighborhoods, with the HDS neighborhoods showing a greater variation in
socio-economic characteristics than the HDU neighborhoods. Variations in travel patterns were
greater for the HDS neighborhoods than the LDS neighborhoods.

215

Table 56: Summary of Travel Behavior by Income Group and Neighborhood Typology
Higher Density Urban
Low
Middle
Higher
Income
Income
Income
All
Mode Choice (Average per Respondent)
% Work Trips by Walk

16

12.6

0

13.5

% Work Trips by Transit

16.5

7.7

0.5

13.5

% Work Trips by Drive

67.6

79.7

99.5

73.0

% Non-Work Trips by Walk

27.4

12.2

12.2

23.6

% Non-Work Trips by Transit

9.2

3.1

6.9

8.3

% Non-Work Trips by Drive

63.4

84.7

80.9

68.1

% All Trips by Walk

26.8

13.7

6.8

23.0

% All Trips by Transit

10.7

4.4

7.1

9.8

% All Trips by Drive

62.5

81.9

86.1

67.2

Average Annual Distances (Average Miles per Respondent to end-point destinations)
Work Trips by Walk

157.7

821.0

0.0

287.2

Work Trips by Transit

238.5

644.1

0.0

299.3

Work Trips by Drive

1209.3

3172.4

4809.1

1849.7

Non-Work Trips by Walk

199.1

522.6

213.5

224.4

Non-Work Trips by Transit

214.2

615.7

219.0

259.7

Non-Work Trips by Drive

1036.6

2590.3

2179.6

1214.6

All Trips by Walk

259.8

1097.3

213.5

41.1

All Trips by Transit

305.9

1066.5

219.0

381.0

All Trips by Drive

1501.7

4811.0

5385.6

1969.7

221

485

589

99.6

20141

38212

51465

8674.0

Avg. Fleet Fuel Economy (miles per gallon)

21.7

20.5

20.4

21.3

Avg. Fleet CO Emissions (grams per mile)

4.3

4.4

4.4

4.3

Fuel Consumption (gallons)
CO Emissions (grams)

216

Table 56 (Contd.)
Low
Income

Higher Density Suburban
Middle
Higher
Income
Income

All

Mode Choice (Average per Respondent)
% Work Trips by Walk

28.9

13.2

14.9

16.0

% Work Trips by Transit

21.2

5.8

3.5

6.7

% Work Trips by Drive

50.0

81.1

81.6

77.3

% Non-Work Trips by Walk

32.1

20.0

19.0

21.5

% Non-Work Trips by Transit

3.7

0.4

0.6

1.1

% Non-Work Trips by Drive

64.2

79.6

80.4

77.4

% All Trips by Walk

29.3

19.0

18.1

20.1

% All Trips by Transit

8.4

2.0

1.6

2.8

% All Trips by Drive

62.3

79.1

80.3

77.1

Average Annual Distances (Average Miles per Respondent to end-point destinations)
Work Trips by Walk

74.9

47.1

228.4

130.7

Work Trips by Transit

126.0

35.2

77.5

64.2

Work Trips by Drive

478.9

1963.6

2198.7

1927.0

Non-Work Trips by Walk

146.3

101.7

67.8

99.9

Non-Work Trips by Transit

30.5

6.6

3.6

9.1

Non-Work Trips by Drive

789.2

858.1

896.7

855.6

All Trips by Walk

189.4

136.3

233.1

188.0

All Trips by Transit

103.1

32.4

59.9

52.5

All Trips by Drive

1064.9

2300.4

2484.8

2156.1

Average Annual Environmental Burdens (Average per Respondent)
Fuel Consumption (gallons)

115

216

262

106.3

9653

19059

21172

8870.3

Avg. Fleet Fuel Economy (miles per gallon)

23

21.8

20.7

21.6

Avg. Fleet CO Emissions (grams per mile)

4.0

4.2

4.1

4.1

CO Emissions (grams)

217

Table 56 (Contd.)
Higher Density Suburban
Middle
Higher
Income
Income

Low
Income

All

Mode Choice (Average per Respondent)
% Work Trips by Walk

5.8

94.4

6.8

6.1

6.8

0.4

0.1

0.2

92.8

93.8

93.0

6.8

6.5

6.9

0.4

0.2

0.2

91.0

% All Trips by Drive

94

0

% All Trips by Transit

93.3

9.0

% All Trips by Walk

0.1

90.4

% Non-Work Trips by Drive

0.1

0

% Non-Work Trips by Transit

0

9.6

% Non-Work Trips by Walk

5.5

94.2

% Work Trips by Drive

5.9

0

% Work Trips by Transit

6.7

92.8

93.4

92.9

Average Annual Distances (Average Miles per Respondent to end-point destinations)
Work Trips by Walk

0.0

44.0

133.4

88.6

Work Trips by Transit

0.0

0.0

5.0

2.9

1295.3

2386.9

4158.7

3354.8

21.4

119.0

46.4

68.1

0

8.1

0.5

2.3

1796.2

1935.9

2012.9

1921.1

21.4

147.3

140.9

121.3

0.0

8.0

4.0

4.0

2156.0

3533.5

4957.6

3938.4

Work Trips by Drive
Non-Work Trips by Walk
Non-Work Trips by Transit
Non-Work Trips by Drive
All Trips by Walk
All Trips by Transit
All Trips by Drive

Average Annual Environmental Burdens (Average per Respondent)
Fuel Consumption (gallons)

194

Avg. Fleet CO Emissions (grams per mile)

4.0
218

212.9

29507

41699

16300.5

21.0

Avg. Fleet Fuel Economy (miles per gallon)

581

16306

CO Emissions (grams)

339

20.7

19.2

20.1

4.3

4.2

4.2

The regression analyses showed that after holding the built environment constant, socioeconomic characteristics had significant relationships with travel behavior and environmental
burdens, especially when it came to non-work trips and all trips combined. Age, as a sociodemographic variable came out as significant in many of the regressions affecting travel by
walking and some by public transit. It is reasonable to expect that age is a characteristic that
would shape mode of travel. Older residents prefer to travel by car than to travel by public
transit or walking to their various destinations.
Household income, education, and employment status (whether the respondent was
employed for wages) consistently came out as significant for travel associated with public transit
and driving. In general, the more educated the respondents are, those with higher household
incomes and the greater the chance that the respondent was employed for wages, the greater the
propensity to drive to destinations compared to those with lower values for those similar socioeconomic characteristics. This is precisely what was hypothesized: that those with higher socioeconomic characteristics drive more often and longer distances. These residents are also,
consequently, responsible for higher levels of environmental burdens. The regressions within
each neighborhood built environment typology allow the analysis to focus on the socio-economic
composition of the population and their impacts on travel. One important dimension to travel
captured in this analysis was that the variation in travel behavior by residents in the same
neighborhood was significant for mainly those neighborhood typologies that had higher
densities, mixed land uses, and higher connectivity (the HDU and the HDS neighborhoods). In
the LDS neighborhoods, the built environment was developed to facilitate automobile travel and
therefore only the regressions on travel behavior for all trips combined and the environmental

219

burdens came out as significant. This too is an important recognition in the outcomes. If the
neighborhood typology itself has no options for alternative travel modes and destinations, then it
forces everyone living in such environments to use their automobiles for any and all travel
purposes (except walking for leisure). Another important issue that needed recognition was that
many of the mode travel variables were skewed and so the regression results would be biased.
Therefore it is important to understand that the analyses results would be sensitive to this. At the
same time, since the regressions are not used for prediction, rather just to indicate relationships,
it is safe to use the results, albeit with caution. The main point however, is that the
environmental burdens variables and regressions were significant and can be stated with
confidence.
The breakdown of the neighborhood types into income subgroups, based on tertiles,
provided added texture to the study of the socio-economic impacts of travel and the resulting
environmental burdens. These impacts are discussed by travel purpose below:

Work-based Trips
Looking at mode of travel, in the urban (HDU) and the higher density suburban (HDS)
neighborhoods, the upper income range residents used their vehicles about 1.5 times more than
the lower income residents did for work purposes. In the lower density suburban (LDS)
neighborhoods, there was no difference between the income groups’ usage of vehicles as their
preferred mode of travel for work purposes. Due to the characteristics in urban form, virtually
everyone needed to drive to work.

220

The difference in the distance traveled between the higher and lower income groups in all
neighborhood typologies was significant. In the HDU neighborhoods, on average, the higher
income grouping residents traveled about three times the annual distance to work than the lower
income grouping residents. In the HDS neighborhoods, the higher income range residents
traveled over 3.7 times the average annual distance to work when compared to the lower income
range residents. In the LDS neighborhoods, the higher income range residents traveled over 3.3
times the average total annual distance to work when compared to the lower income range
residents. In the case of all three neighborhood typologies, the higher income groupings had jobs
that were further away from their home when compared to the lower income groupings.
In case of the HDU neighborhoods, the higher end jobs are located in the suburbs and the
higher income residents—at least in the context of Detroit income averages—from the Detroit
neighborhoods are required/need to travel great distances to get to their place of employment
(Darden and Kamel, 2000; Darden and Thomas, 2013). At the same time, data from the survey
shows that these higher income residents will not necessarily sell their current homes in the inner
city and move closer to their jobs. This, in fact, can be expected for all neighborhoods, since
people switch jobs on shorter time intervals than they are necessarily willing to—or can afford
to—move from their neighborhood. On a regional scale, in the case of metropolitan areas, travel
to work are the high pressure commutes where people have to travel longer distances to get to
work opportunities, in large part because people will change jobs but will unlikely relocate every
time they get a new job.
For the HDS neighborhoods, the burden of travel for work purposes is the lowest. So
while the study shows in general that the upper income residents travel further for work purposes
than the lower income residents, it also shows that the upper income groupings in the HDU and
221

LDS neighborhoods travel 1.7 and 1.9 times the annual distance that the upper income grouping
travels to work in the HDS neighborhoods. This is reflective of both the built environment
characteristics and the fact that the higher paying jobs have moved out into, or in close
proximity, to these HDS areas, where along with employment opportunities, the quality of life is
also high and desired by many upper income residents. Many Ann Arbor residents, for instance,
have jobs in relative close proximity due to the rich and robust characteristics of this activity
node. In contrast, in the Detroit lower eastside neighborhoods, despite the high densities, mixed
land uses, and high connectivity, there are simply not these types of employment opportunities.
In part, the extreme disinvestment and decline within Detroit has shaped these outcomes.
For the LDS neighborhoods, it is recognized from the built environment objective data
that these are large homes on large lots in isolated residential neighborhood pods. From the
Census data and from the neighborhood surveys, it is also evident that these are expensive homes
in upper income neighborhoods. The residents of these neighborhoods are generally higher
income earners that are willing to drive long distances to work, as evident in time and distance,
in order to attain the particular quality of life desired, living in isolated and secluded
neighborhoods with large homes on large lots.

Non-Work Trips
For non-work trip purposes, we start seeing the residents make use of other travel modes,
particularly walking. As is expected, the lower income range residents walk and take transit
(where available) more than the middle and upper income range residents. The HDU
neighborhood residents face a situation where they have only a few good choices when it comes
222

to shopping and service availability. Shopping for food is a necessity, however, the residents in
the urban neighborhoods have to travel further to get to supermarkets that are located away from
the city, and particularly in the surrounding suburbs. As these neighborhoods experience
disinvestment and decline, basic daily destinations are unavailable within these neighborhoods—
whether national supermarket chains or personal services—and the residents have to travel
further distances due to a lack of opportunities and choices. Hence, these neighborhoods
maintain lower accessibility levels to many necessary daily destinations, despite the higher
densities, the high connectivity, and the mixed land uses.
In the HDS neighborhoods though, for travel that is less constrained by time and location
(non-work trips), people tend to vary their mode of travel and make greater use of walking and
public transit. This lends support, again, to the notion that within these robust communities,
there are ample choices for destinations of different purposes and this gives the residents the
choice to use alternative forms of travel to the car. The increase in walking as a mode of
transportation for non-work purposes is the least in the LDS neighborhoods and this was
expected since these neighborhoods rarely have any destinations close by that would be
accessible by walking (unless residents go for leisurely walks in their neighborhood).
Across all regressions in the three neighborhood types, socio economic characteristics
had a significant relationship with the different modes of travel (and especially walking) and the
miles traveled. All three income groups in all neighborhoods walked more times to non-work
destinations than to work destinations, however, the lower income groups walked to all
destinations the most.

223

As for transit, the lack of an extensive transit system in the suburbs leads to a lower use
of this mode of transportation. The prevalence of transit use is highest in the HDU
neighborhoods, and the use of transit was higher for work purposes by the lower and middle
income group residents than for non-work trip purposes. Only the higher income range residents
increased their usage of transit from work trips to non-work trips, that too only marginally. Even
so, the lower income grouping in the HDU neighborhoods used transit the most. This is similar
pattern as in the HDS neighborhoods. The lower income grouping residents are the ones who
avail of this mode of travel the most. For the LDS neighborhoods, transit is not very prevalent as
a mode of travel. Public transit use is minimal and in part because of the built environment of
these neighborhoods, which makes mass transit provision in these urban built forms costly.

Environmental Burdens
The environmental burdens show similar trends as the driving patterns. Since the
wealthier subgroup of residents in all three neighborhood types drive farther distances, they are
responsible for the highest consumption levels of fuel and the highest emissions of carbon
monoxide on an annual basis (grams per respondent). The vehicle characteristics of the
wealthier residents play a role in this too. Larger and more powerful engines, larger sized
vehicles, and driving conditions all affect fuel consumption and emissions. The wealthier
residents are more likely to own powerful, faster, and larger vehicles, such as SUVs and
minivans. These vehicle characteristics combined with the driving patterns are the variables that
account for such high levels of fuel consumption and CO emissions.

224

In the HDU neighborhoods, the higher income subgroup drives about 3.5 times the
distance (annual average per respondent to end-point destinations) that the lower income group
does. Accordingly, the higher income grouping consumes 2.7 times the annual fuel (annual
average gallons per respondent) and emits 2.6 times the carbon monoxide (annual average grams
per respondent) than the lower income grouping in urban Detroit.
As far as the average fuel economy of the fleet of vehicles that each income group in
urban Detroit owns, it becomes evident that the average fuel economy decreases as one moves
from the lower income grouping to the upper income grouping. The average fleet fuel economy
per respondent varies from 21.7 miles per gallon (low income group) to 20.5 miles per gallon for
the middle income group and 20.4 miles per gallon for the upper income group. Similarly, the
average fleet carbon monoxide emissions per respondent also increase from the lower income to
the upper income groupings. The average carbon monoxide emissions for the lower income fleet
was 4.3 grams per mile, for the middle income fleet it was 4.4 grams per mile, and for the upper
income fleet it was 4.4 grams per mile.
In the HDS neighborhoods, when comparing the upper and lower income groupings, the
upper income subgroup drives 2.3 times the distance for all trip purposes combined (annual
average per respondent to end-point destinations), consumes 2.3 times more fuel (annual average
gallons per respondent), and is responsible for about 2.2 times more emissions (annual average
grams per respondent). The average fuel economy for the fleet shows the same trend (with the
higher income groups being responsible for higher levels of environmental burdens), as fuel
consumption varies from 23 miles per gallon for the lower income group, to 21.8 miles per
gallon for the middle income group fleet and 20.7 miles per gallon for the upper income group
fleet. The average carbon monoxide emissions range from 4.0 grams per mile for the lower
225

income group fleet, to 4.2 grams per mile for the middle income fleet and 4.1 grams per mile for
the upper income group fleet. These HDS neighborhoods shows similar trends for driving
distance, fuel consumption and emissions as they generally depict that the wealthier consume
more fuel and emit more pollutants.
For the last neighborhood typology, the LDS neighborhoods, similar trends continue, as
the upper income group drives about 2.3 times the distance for all trip purposes combined
(annual average per respondent), consumes 3 times the fuel (annual average gallons per
respondent) and emits 2.6 times the pollutants (annual average grams per respondent) when
compared to the lower income group. The proportionate increases in fuel consumption and
emissions are also shaped by the vehicle characteristics. On average, the lower income subgroup
fleet fuel efficiency is about 21 miles per gallon, while it is 20.7 miles per gallon for the middle
income subgroup fleet and 19.2 miles per gallon for the upper income subgroup fleet. The
average carbon monoxide emissions range from 4.0 grams per mile for the low income fleet, to
4.3 grams per mile for the middle income fleet and 4.2 grams per mile for the upper income fleet.
These figures all point to the fact that as ones income increases, the vehicles that they own tend
to be larger, more powerful and have higher energy requirements in travel. Within this
neighborhood, we see once again, that there is a steady decline in fuel efficiency as the income
range increases from the lowest to the highest income subgroups. Again, this positive trend
between increasing environmental burdens and incomes is captured despite the fact that lower
income groups have older vehicles that due to their lower technical sophistication will tend to
generate higher MPGs and CO emissions.
All these results reiterate and stress the point that the wealthier drive farther distances,
consume more fuel and are responsible for higher levels of emissions than lower income
226

households, even when they live within the same built environment. This is a unique
contribution of this analysis, since we are using performance characteristics of actual cars
owned, with specific weight, engine, and power traits, best case emission levels pertaining to the
specific make, model and year of the vehicle, and assigning them to a full array of driving
destinations, including both work trips, non-work trips and all trips combined. This detail of
analysis has not been provided in the existing literature and it is the unique and critical
contribution of this work.

227

Chapter VII - Conclusion
Research on transportation issues has been conducted on a wide range of topics, from
streetscape design to travel demand to transportation related greenhouse gas emissions and
climate change. The latter issue has received considerable attention as global climate change and
its vast consequences have been discussed with increasing fervor in recent years. With
transportation, as an industry, being the fastest growing contributor of greenhouse gasses, studies
relating to the planning and functioning of the urban built environment have focused on the
transportation aspect for a deeper understanding of this industry’s contribution to increasing
greenhouse gasses.
Transportation topics have been researched at various geographical scales depending on
the specific area of research. Greenhouse gas emissions and the transportation industry have
typically been researched at larger geographic scales, such as national and state levels of scale.
Travel demand and behavior studies typically focus on lower geographic scales, such as cities
and travel analysis zones (TAZ). Road network and performance studies have been researched
at the lowest scale, such as road and street network segments.
Travel behavior studies have typically looked at built environment characteristics of the
area (usually robust neighborhoods or cities) and socio economic characteristics of residents to
uncover how these affect travel demand and VMT. Greenhouse gas emission studies have
focused on driving patterns, built environment characteristics and road networks to uncover how
these variables affect emission levels. The gap in the literature, therefore, is the effects of the
socio-economic characteristics of residents, including in neighborhoods experiencing
disinvestment and decline, on travel behavior and the consequent environmental burdens, and

228

specifically gasoline consumption and carbon monoxide (CO) emissions. This focus would help
understand the travel behavior of residents with differing socio-economic characteristics and
identify the set of characteristics that are responsible for higher levels of environmental burdens.
Just as it is important to know the built environment characteristics that contribute to higher
levels of emissions, so it is important to know the set of characteristics of people who contribute
to higher levels of emissions (and fuel consumption).
Another important dimension of this research was the opportunity to gather primary data
through a mail survey. Getting travel behavior data at the neighborhood scale is critical to
transportation studies, and yet highly underrepresented, because this type of data collection is
very time consuming and costly. This was an important reason for this particular study design: it
would be completed with rich data, at a low scale, and fill a critical gap in the research.
Detroit, Michigan is a city in the US that has experienced unprecedented growth in the
th

early 20 century on the basis of its manufacturing industry. Automobile manufacturing took
the City to unprecedented heights, however, as business cycles go, this industry also experienced
extensive economic downturns that crippled the City. With the economic engine suffering, other
issues--such as suburbanization, segregation, crime, education, and poor policies--came to the
forefront and wrapped the City in a self-perpetuating downward spiral. A study done in this
unique City and its suburbs would add value in itself, given the conditions that neighborhoods in
the City were experiencing, severe disinvestment and decline. Exploring travel behavior in
communities experiencing disinvestment and decline is considered an under-represented aspect
of urban geography and community planning.

229

The main research aim was to see if the built environment and socio-economic
characteristics impact travel behavior and more specifically, environmental burdens (measured
as vehicular fuel consumption and carbon monoxide (CO) emissions). The finer details involved
analyzing how variations in socio-economic and demographic characteristics of residents living
within the same built environment impacted travel and the consequent environmental burdens.
This formed the crux of this study: to get an understanding of those resident characteristics that
contributed to higher or lower environmental burdens, while also including within the study
neighborhoods experiencing disinvestment and decline as well as robust and dynamic
neighborhoods.
There were six neighborhoods involved in the study. These neighborhoods were grouped
in twos (based on similar built environment characteristics) and were defined as three specific
neighborhood typologies: the higher density urban (HDU) neighborhoods, the higher density
suburban (HDS) neighborhoods, and the lower density suburban (LDS) neighborhoods. The two
urban Detroit neighborhoods formed the HDU typology, Ann Arbor and Birmingham formed the
HDS topology and Bloomfield Hills and West Bloomfield formed the LDS typology. The built
environment characteristics that defined these groupings were density of structures, street
network patterns, and the amount of and placements of mixed land uses.
A Discriminant Analysis was performed on the three typologies to see what the main
factors were that distinguished the three neighborhood types. The results of the discriminant
analysis stressed that the HDU neighborhood was different from the other two suburban
neighborhoods based mainly on household income and educational attainment. In essence, the
residents of the urban Detroit neighborhoods had lower educational attainment and lower
household income compared to those of the suburban neighborhood residents.
230

The suburban neighborhoods differed from each other mainly based on the age of the
residents, the share of non-work trips that are undertaken by walking, and the annual CO
emissions. This result is telling as two of the three distinguishing factors between the two
suburban neighborhood typologies relate to travel. Since the HDS neighborhoods are more
compact, have higher accessibility and connectivity, more people can undertake their non-work
trips by walking, and consequently, the residents in these neighborhoods are responsible for
fewer annual environmental burdens (gasoline consumption and CO emissions). So a class
dimension separates the Urban from the Suburban typologies, while a behavior dimension, a
consequence of the different physical forms, separates the HDS from the LDS typologies.
The next analysis that was performed was a K-Means Cluster analysis on each of the
three typologies. This analysis was conducted to see what groups the data themselves were
forming. Two sets of Cluster Analyses were conducted on each of the three neighborhood
typologies: the first Analysis was conducted on only the demographic variables, while the second
Analysis was conducted on demographic as well as travel behavior variables. Since the
typologies were already selected based on their built environment characteristics, the Cluster
Analysis was conducted on the other variables (excluding the built environment characteristics)
to see the types of groups that emerged within each typology.
The results of the Cluster Analyses on demographic variables only indicated that for the
HDU typology, the groupings were mainly based on the number of dependents, the age of the
residents, and whether they were employed for wages. The analysis done on the HDS typology
resulted in groups that were mainly based on income, number of vehicles owned, age and usage
of the primary vehicle used for travel, and the race of the respondent. The difference in the basis
of the groups formed within the HDU and HDS typology shows that the prominent ranges in the
231

HDU typology were related to the class of the residents (income and education). The higher
income among residents in the HDS typology resulted in groups based more on demographic
factors. The results of the Cluster Analysis on the LDS typology indicated groups based on the
number of dependents in the household, employment (for wages in general and in professional
categories in particular) and the number of vehicles owned. The LDS groups show a smaller
range of characteristics (like the HDU typology) that the groups were based on. The main
difference being that employment (in professional categories) was also one of the main grouping
factors.
The results of the Cluster Analyses on demographic as well as travel behavior variables
indicated that for the HDU typology, groups were mainly based on travel behavior variables such
as annual miles to work by walking, annual environmental burdens (fuel consumption and CO
emissions), miles on vehicles, income and education. The analysis on the HDS typology
revealed groups based on trips undertaken by different modes of travel (walking, transit and
driving), and annual environmental burdens (fuel consumption and CO emissions).
The difference between the bases of clusters formed between the Higher Density Urban
and Suburban typologies indicates that the HDS typology has a rich mix of amenities in
proximity of the residential neighborhoods so as to give the residents the option of using the
various modes of travel to get to various destinations. Fewer groups based on travel behavior for
the HDU typology suggests that even though the typology has characteristics of high
connectivity and access, there is not a good variety of travel options to realistically enable
different modes of travel for the different trip purposes.

232

The results of the Cluster Analysis on the LDS typology indicated that the groups were
based on trips undertaken by walking and driving, annual environmental burdens (fuel
consumption and CO emissions), number of vehicles owned and the age and usage of the
primary vehicle. These groupings are very similar to those that the HDS analysis resulted in,
with an emphasis on mode of travel and environmental burdens. The difference being that the
LDS typology had more grouping variables that related to the vehicles owned and used.
In summary the Cluster Analyses indicated that the HDS typology had a rich variation in
the characteristics of the residents, and more specifically, a good variation in travel behavior.
This is important because we can expect that residents living in the LDS typology neighborhoods
use their vehicles for most of their travel purposes, however, residents living in the HDU
typology neighborhoods confront similar built environment characteristics as residents living in
the HDS typology neighborhoods and yet do not have this type of variation in their travel
behavior. This distinction in the richness in travel behavior variation might be due to the fact
that these HDU neighborhoods are experiencing disinvestment and decline and so there are not
that many destinations and safe travel choices for the residents in these urban neighborhoods.
The last set of quantitative analyses conducted was the OLS Regressions and these were
performed in order to test the hypotheses. The hypotheses essentially stated that within each
neighborhood typology, the residents with higher socio-economic characteristics (higher
incomes, higher levels of educational attainment, probability of being employed for wages,
number of dependents, race and age of the respondent) would use their vehicles more than
walking or using public transit for any and all of their travel purposes. Consequently, these
residents would be responsible for higher levels of environmental burdens. In contrast, those
residents with lower socio-economic characteristics would use other modes of transportation
233

more often and use their vehicles for fewer miles and consequently be responsible for lower
levels of environmental burdens. The hypotheses also stated that residents with higher socioeconomic characteristics would own more vehicles per household, own newer vehicles and own
vehicles with stronger engines compared to those with lower socio-economic characteristics
(within the same neighborhood typology).
The regression results, in general, indicated an acceptance of the above hypotheses.
Across all three neighborhoods, the main finding was that class had a significant relationship
with travel patterns and environmental burdens. Even while living in the same neighborhood,
people with different socio-economic characteristics behaved differently. The wealthier traveled
more times and longer distances with their private vehicles compared to lower-income groups,
and consequently, they were responsible for higher environmental burdens—higher gasoline
consumption and CO emissions—despite the fact that the poorer households had older vehicles
that are less fuel-efficient.
The regression analyses showed that after holding the built environment constant, socioeconomic characteristics had significant relationships with travel behavior and environmental
burdens, especially when it came to non-work trips and all trips combined. Age, as a sociodemographic variable, came out as significant in most of the regressions affecting travel. Older
residents prefer to travel by car, over walking and transit, to reach their various destinations.
Household income, education, and employment status (whether the respondent was employed for
wages) consistently came out as significant for travel associated with public transit and driving.
In general, the more educated, those with higher household incomes and respondents employed
for wages, were likely to drive to destinations compared to those with lower values for those

234

similar socio-economic characteristics. These residents were also, consequently, responsible for
higher levels of environmental burdens, greater gasoline consumption and higher CO emissions.
One important aspect was that the variation in travel behavior by residents was
significant for mainly those neighborhood typologies that had higher densities, mixed land uses,
and higher connectivity (i.e. the HDU and the HDS neighborhoods). When it came to the LDS
neighborhoods, the built environment was developed to facilitate automobile travel and therefore
only the regressions on travel behavior for all trips combined and environmental burdens came
out as significant. This is also an important aspect of this research. If the neighborhood
typology itself has no options for alternative travel modes and destinations, then it forces
everyone living in such environments to use their automobiles for any and all travel purposes. In
the low density, single land use, and disconnected suburbs, the only way to move between daily
destinations is by car.
To give the Regression analyses and overall results more meaning, each of the three
typologies was broken down into three classes by the income of residents: categorizing them into
Lower Income, Middle Income, and Higher Income groups. Travel behavior and environmental
burdens were then categorized according to the income profile of the residents. What this did
was provide a deeper, yet a more nuanced breakdown, of how residents with varying incomes,
living in the same neighborhood, had differing travel behaviors and contributions to
environmental burdens.
Across all three typologies, there was a constant recognition that the wealthier drive
farther distances, consume more fuel and are responsible for higher levels of emissions than
lower income households, even when they live within the same built environment. This is a

235

unique contribution of this analysis, since we are using performance characteristics of actual cars
owned, with specific weight, engine, and power traits, best case emission levels pertaining to the
specific make, model and year of the vehicle, and assigning them to a full array of actual driving
destinations by households, including work trips and non-work trips.
Another important contribution of this work is bringing the issue of ‘class’ to the
forefront of travel behavior analysis. This issue has been dealt with in the literature by looking at
the travel behavior (not environmental burdens as a result of the travel behavior) of entire
neighborhoods with similar class structures. This study shows that even the relatively wealthier
population living in declining neighborhoods contributes more to environmental burdens than the
poorer residents in the same neighborhoods. In addition to the specific case of including
neighborhoods experiencing extreme decline, there is a variation in class structure within all
neighborhood typologies. Simply, the wealthier, whether they live in robust communities or in
neighborhoods experiencing disinvestment and decline, travel more often and further distances
by car and are responsible for higher levels of environmental burdens.
These results reveal that, even in neighborhoods that would generally be looked at as
maintaining a similar class characteristic, the more subtle class variations within the
neighborhood generate significant differences in travel behavior and resulting environmental
burdens. This is a novel contribution to the existing study of travel behavior, and particularly in
the context of analyzing gasoline consumption and CO emissions. This has never previously
been examined in travel behavior research that has focused on environmental burdens, including
gasoline consumption and/or pollutant emissions.

236

An important recognition should be made with regard to the impact of neighborhood
disinvestment and decline on neighborhood access to daily destinations. Travel behavior studies
have shown that lower income residents travel further distances (especially for work based trips).
They also tend to live in inner city areas so that they can avail of alternative travel modes to get
to destinations, as they are less likely to own vehicles (Grengs, 2012; Pendall, 2000; Squires and
Kubrin, 2005). This study also shows that even though the lower income population lives in
urban Detroit areas characterized by high-densities, high-connectivity and mixed land uses, the
extensive disinvestment within these neighborhoods have increased travel distances to a variety
of destinations, including amenities that would be considered necessary daily destinations. In
fact, for Detroit urban residents—despite the higher-density, greater land-use mix and higherconnectivity, which are neighborhood characteristics that should improve access—distances to
reach shopping, personal services, and leisure destinations are similar to those of residents living
in isolated, disconnected, residential neighborhoods in the low-density Detroit suburbs.
For example, the average distance to national/regional supermarkets is higher for the
Detroit urban residents (5.5 miles) than for the low-density suburban residents (3.1 miles),
indicating that these stores are locating away from the inner city and closer to the wealthier, lowdensity suburban neighborhoods. In fact, for all store types combined, the urban Detroit residents
travel about the same average distances for retail/grocery shopping as do the low-density
suburban residents (3.3 miles and 3.4 miles respectively). Similarly, for personal services, the
Detroit residents and the low-density suburban residents travel about the same average distances
(3.4 miles and 3.5 miles respectively). There has been such extensive disinvestment in the highdensity Detroit urban neighborhoods that they face comparable, and sometimes even longer

237

distances, to daily destinations when compared to residents living in the low-density, isolated
Detroit suburbs.
The lower income eastside Detroit residents face a greater burden, in terms of travel
distances, to reach basic daily necessities not because of the urban built environment, but
because of the lack of availability of adequate destination choices in proximity to their homes.
What the extreme disinvestment and decline in Detroit has accomplished is to transform, what
are traditionally viewed as high-accessibility urban neighborhoods, into low-density suburbs, at
least in terms of accessibility characteristics. However, despite these similarities in distances to
destinations between the lower-income Detroit urban respondents and the upper-income, lowdensity Detroit suburban respondents, the higher income suburban respondents—on an annual
basis—travel about twice the distances by car, consume about twice the gasoline (gallons per
year), and emit about two times the CO levels associated with automobile travel. It is thus at
multiple dimensions that we can conclude that environmental burdens associated with travel are
shaped by class, with higher income earners disproportionately contributing to negative
anthropogenic environmental impacts.
This study is important on a number of dimensions. First and most importantly, it
contributes to a gap in the research on the social impacts on travel and environmental burdens.
Secondly, it combines qualitative and quantitative analyses to help understand the various
nuances of the effects of class on travel and environmental burdens. Both these types of analyses
are distinct and have critical contributions to the existing research in geography, urban planning,
urban design and economics, and as such, are important in their own respects. Thirdly, this study
is important to multiple target agencies, officials and researchers, such as transportation planners,

238

researchers on sustainability issues, environmental planning and urban planning professionals
and policymakers.
There are a number of policy implications of this research and its outcomes. One
addresses the current policy debate in Michigan on how the revenue generation for road repair
should be collected from the Michigan population. It is a highly controversial issue that will go
to the ballots in November of 2013, if approved by the Senate. Governor Snyder has proposed to
the raise the Michigan gas tax in order to raise $1.2 billion annually for road improvements. The
increase in the gas tax (currently from 19 cents per gallon for gasoline and 15 cents per gallon for
diesel to about 33 cents for both fuels) would put the responsibility of maintaining Michigan
roads on those who use them. Opposing lawmakers, on the other hand, have suggested an
alternative way to raise funds -- cancel the sales tax on gasoline and increase the gas tax by the
same amount as the sales tax (so that people do not pay more at the pump). They have also
suggested an increase in the overall state sales tax (of about 1-2 percent) to make up for the loss
in sales tax revenue from fuel sales (this would also lead to an overall increase in tax revenue
that would go to education and revenue sharing) (Eggert, 2013; Oosting, 2013).
This research has shown that the lower the income among the population, the more that
they will tend to walk and the more that they will tend to drive fewer miles in comparison to
their wealthier counterparts. A flat sales tax across the Michigan population would remove the
onus from those who use the roads most, and spread the cost of maintaining roads and highways
to all Michigan residents, which would disproportionately burden the Michigan lower-income
population who do not use cars as much for their travel needs. In the context of the considerable
diversity in automobile use by income, and given that road use by automobile travel has clear
cost implications linked to those populations driving, policy would need to be introduced that
239

would ensure a costing structure based as closely as possible on marginal costs (Vojnovic, 2000).
This would require very clear relationships to be established between car use, road use (along
with its wear-and-tear), and revenue collection for road and highway repairs. Without the tax on
road repair linked to gasoline consumption, both inefficiencies and inequities would be generated
within Michigan’s public revenue and service delivery structure. This would be realized with
Michigan lower income populations paying for services delivered, maintained and consumed--in
this case, roads/highways and their maintenance--disproportionately by the Michigan wealthy
(Vojnovic 1999, 2000a, 2000b).
The findings from this study also provide policy insight into the existing structure of the
Gas Guzzler Tax. The Gas Guzzler tax was originally implemented in 1978 and is a tax that is
placed on manufacturers for sales of vehicles that do not meet the minimum 22.5 miles per
gallon fuel efficiency (EPA, 2012). An important aspect of this tax is that it is not levied on
SUVs, minivans and pickup trucks. These vehicles are a major part of the US passenger vehicle
fleet and they are known to be larger vehicles and typically consume higher levels of fuel and
emit higher levels of pollutants. These vehicles, as my research has shown, also tend to be driven
by the wealthier population, and also over longer distances.
These vehicles were not included in the Energy Tax Act because in 1978 they formed a
very small proportion of the US vehicle fleet. However, the use of these vehicles is much more
extensive now and the Gas Guzzler tax should be reflective of this. If the U.S. is to seriously
address greenhouse gas emissions, the federal government needs to expand the Energy Tax Act
to include the most serious gasoline consumers and pollutant emitters. There is no reason to
exempt SUVs, minivans and pickup trucks from national efforts to reduce Greenhouse Gas

240

Emissions, and particularly given that these vehicles are associated with the most significant
environmental impacts and are driven disproportionately by the wealthy.
This study brings together multiple fields and can lead to many more collaborative
studies and research on social characteristics, environmental and climate change issues,
transportation planning and the urban built environment. By using actual data on the full-array
of travel behavior, along with automobile ownership characteristics at the household level, a
unique perspective is provided on fuel consumption and vehicle emissions. This environmental
and travel data has the added advantage of being tied to socio-economic and demographic
household characteristics. This research provides important insight to policymakers that can be
used to shape both local public finances associated with road provision and maintenance, and
also environmental policy tied to the growing global interest in Greenhouse Gas Emissions.
What becomes evident along both lines of policy is the political interest in exempting the
wealthy from being responsible for their disproportionate scales of resource consumption and
environmental burdens.

Future Research Possibilities
Future research on this topic could explore in more detail the variables that impact travel
behavior for work trips. This study had several regressions that turned out insignificant for
socio-economic characteristics and work based trips. The reason for this could be that residents
have less of a choice in mode of travel and the distance they travel to work that can be impacted
by socio-economic characteristics. Lower income and minority populations have more limited
job options and have less of a choice on where their place of employment would be. The higher
241

income residents do have and exercise greater choice in terms of where they live and work. A
further study into work based trips and environmental burdens would be useful. In addition, the
regression analyses resulted in low adjusted R-squares. This indicates that the models explained
a small part of the variation in travel behavior and environmental burdens. This would need to
be further investigated and other factors that determine travel behavior would need to be
included in future models to assess the explanatory power of the models. Other socio-economic
factors, such as family structure and gender, and other built environment factors, such as
connectivity, accessibility and density, could be included in models to determine if these help in
explaining a greater proportion of the variation in travel behavior and environmental burdens.
In addition, a deeper analysis of the contributions of race to travel behavior and
environmental burdens would also be important. Although this study did include variables of
race within the analyses, there were just not that much of variation among the population to
account for a full analysis. Within the Detroit neighborhoods, there were too few whites and
within the LDS neighborhoods, there were too few blacks. The ‘other’ race faced the same
issues as well, a result of the well-known racial polarization of these areas. A study that would
include a larger variation and number of residents of different races would be useful in
determining the contributions of this important variable to the environmental burdens discourse.
Apart from this study making critical contributions to existing research, it has tremendous
potential for future research as well. The detailed dataset that this study comes from has the
potential to extend the environmental burdens analysis to different trip destinations, different
geographic scales and social characteristics. Future potential areas of research include analyzing
environmental burdens at an individual/household level of scale emphasizing built environment
impacts on environmental burdens. This will be done by examining the built environment within
242

a half mile buffer zone around each household and examine the impacts of the personal built
environment factors on travel behavior and environmental burdens.
Another potential research opportunity would include a merger of gender and family
structure characteristics on environmental burdens. This would allow an analysis into the
differing environmental burdens between men and women. Yet another opportunity would be an
analysis of environmental impacts broken down into a greater variety of trip destinations than
just the work, non-work and all destinations included in this study. In short, these data sets and
the study of environmental burdens have the potential for research that can add to the existing
understanding of environmental impacts, and especially given the sorely lacking and
underrepresented dimension of this research direction at the household level and the
neighborhood scale.

243

APPENDICES

244

APPENDIX A - Data Validation
Table 57: Data Validation - Detroit

Demographic
Characteristic
Race
% White
% Black
% Other
Income
Mean Household
Per Capita

Detroit
2010 Census Tracts
(encompassing the
neighborhood)

Detroit City

Study Survey

8%
82%
10%

5%
93%
2%

8%
90%
2%

$39,327
$15,261

$31,863
$12,834

$28,102
$21,809

Table 58: Data Validation – Ann Arbor

Demographic
Characteristic
Race
% White
% Black
% Other
Income
Mean Household
Per Capita

Ann Arbor
2010 Census Tracts
(encompassing the
neighborhood)

Ann Arbor City

Study Survey

70%
8%
22%

84%
5%
11%

94%
2%
4%

$77,239
$32,734

$77,975
$34,937

$88,160
$54,926

Table 59: Data Validation – Birmingham

Demographic
Characteristic
Race
% White
% Black
% Other
Income
Mean Household
Per Capita

Birmingham
2010 Census Tracts
(encompassing the
neighborhood)

Birmingham City

Study Survey

91%
3%
6%

93%
3%
4%

94%
1%
5%

$151,887
$67,580

$166,467
$72,079

$113,896
$73,869

245

Table 60: Data Validation – Bloomfield Hills

Demographic
Characteristic
Race
% White
% Black
% Other

Bloomfield Hills City

Bloomfield Hills
2010 Census Tracts
(encompassing the
neighborhood)

Study Survey

86%
4%
10%

92%
1%
7%

$211,077
$79,184

Income
Mean Household
Per Capita

85%
5%
10%
$209,541
$81,551

$109,278
$73,933

Table 61: Data Validation – West Bloomfield

Demographic
Characteristic
Race
% White
% Black
% Other
Income
Mean Household
Per Capita

West Bloomfield
2010 Census Tracts
(encompassing the
neighborhood)

City

Study Survey

77%
11%
12%

73%
14%
13%

84%
5%
11%

$127,804
$47,269

$122,077
$43,488

$106,758
$69,118

246

Appendix B – Example of EPA Window Sticker
Figure 34: Example of an EPA Window Sticker

247

Figure 34 (cont’d)

248

Appendix C – Example of Fuel Economy from the Department of Energy
Figure 35: Example of a Fuel Economy Sheet

249

Appendix D – Code Book
Table 62: Code Book
CODE
Demographic Variables
Age

NAME

DESCRIPTION

SOURCE

Age

Age of respondent in years
Mean values of Personal
Income Ranges on the survey
Mean values of Household
Income Ranges on the survey
Years of formal education
received
Number of people the
household income supports
Number of children under the
age of 18 years living at home

Survey

PIncome

Personal Income

HIncome
Educ

Household Income
Educational
Attainment

Inc_Supports

Income Supports

Num_Child

Number of Children

Computed from survey
Computed from survey
Computed from survey
Survey
Survey

Vehicle Variables
VehYou

Vehicles for you

VehAll
VehYear

Vehicles for all
Vehicle Year

VehCylinder
VehMiles

Cylinders
Mileage

Number of vehicles for use by
respondent primarily
Number of vehicles for use by
all in the household
Year of the Vehicle Model
Number of cylinders in Vehicle
Engine
Number of miles on vehicle

Percent Work Trips
Walk
Percent Work Trips
Drive

Percent of all trips to work
undertaken by walking
Percent of all trips to work
undertaken by driving

Survey
Survey
Survey/MI SOS
Survey/MI SOS
Survey

Travel Variables
Pct_WorkTrips_Walk
Pct_WorkTrips-Drive

250

Computed from survey
Computed from survey

Table 62 (cont’d)

Pct_NonWorkTrips_Walk

Percent Non-Work
Trips Walk

Pct_NonWorkTrips_Drive

Percent Non-Work
Trips Drive

Pct_AllTrips_Transit
Ann_WorkMiles_Walk

Percent All Trips
Transit
Annual Work Miles
Walk

Ann_WorkMiles_Drive

Annual Work Miles
Transit
Annual Work Miles
Drive

Ann_NonWorkMiles_Walk

Annual Non-Work
Miles Walk

Ann_WorkMiles_Transit

Annual Non-Work
Ann_NonWorkMiles_Transit Miles Transit
Ann_NonWorkMiles_Drive

Annual Non-Work
Miles Drive

Ann_AllMiles_Walk

Annual All Miles
Walk

Ann_AllMiles_Transit

Annual All Miles
Transit

Percent of all trips to non-work
destinations undertaken by
walking
Percent of all trips to non-work
destinations undertaken by
driving
Percent of all trips to all
destinations undertaken by
public transit
Total annual miles for all trips
to work undertaken by walking
Total annual miles for all trips
to work undertaken by Public
Transit
Total annual miles for all trips
to work undertaken by Driving
Total annual miles for all trips
to non-work destinations
undertaken by walking
Total annual miles for all trips
to non-work destinations
undertaken by Public Transit
Total annual miles for all trips
to non-work destinations
undertaken by Driving
Total annual miles for all trips
to all destinations undertaken
by walking
Total annual miles for all trips
to all destinations undertaken
by Public Transit
251

Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey
Computed from survey

Table 62 (cont’d)
Ann_AllMiles_Drive

Annual All Miles
Drive

Total annual miles for all trips
to all destinations undertaken
by Driving

Ann_Fuel_Consump

Annual Fuel
Consumption

Total annual gallons of fuel
consumption

Ann_CO_Emissions

Annual Carbon
Monoxide
Emissions

Total annual grams of Carbon
Monoxide emissions

252

Computed from survey
Computed from survey and EPA’s Office of
Transportation and Air Quality (OTAQ) site
on Green Vehicle Guide
(http://www.epa.gov/greenvehicles/Index.do)
Computed from survey and EPA’s Office of
Transportation and Air Quality (OTAQ) site
on Green Vehicle Guide
(http://www.epa.gov/greenvehicles/Index.do)

Appendix E – Distributions of Variables (for Discriminant Analysis)

The Discriminant Analysis was conducted on one single file that had all three neighborhood
types coded from 1 to 3 (1=HDU, 2=HDS, and 3=LDS)
The stepwise DA (Discriminant Analysis) chose the following variables.
Table 63: Results of Stepwise DA
Variable
N
Minimum Maximum
VehAll
774
0
9
Age
939
18
94
HIncome
825
5000
155000
Educ
942
10
19
Inc_Supports
899
0
10
Pct_NonWork_
Trips_Walk
936
0
100
Ann_CO_Emissions
793
0 439966.80

Mean
1.85
53.27
83981.82
15.96
2.49

Skewness
0.70
0.04
0.03
-0.47
0.87

Kurtosis
3.44
-0.61
-1.38
-1.10
0.70

16.84
12401.97

1.65
10.79

1.79
175.80

For normally distributed variables, the skewness and kurtosis numbers should fall between -1.0
and +1.0. After trying different transformations (log normal, square root, and inverse), the
following transformed variables were used for the DA.
Table 64: Transformed Variables used in DA
Variable
N
Minimum Maximum
LnVehAll
705
0.00
2.20
Age
939
18.00
94.00
Sqrt_HIncome
825
70.71
393.70
Ln_Educ
942
2.30
2.94
Inc_Supports
899
0
10.00
Ln_Pct_NonWork_
Trips_Walk
390
0.21
4.61
Ln_Ann_CO_Emissions
743
3.61
12.99

253

Mean
0.62
53.27
270.61
2.75
2.49

Skewness
-0.12
0.04
-0.44
-0.69
0.87

Kurtosis
-0.45
-0.61
-0.97
-0.73
0.70

3.42
8.82

-0.66
-0.41

0.09
0.45

Appendix F – Distributions of Variables (for Regression Analyses)
The Regression Analyses were conducted on three separate files corresponding to the three neighborhood typologies (HDU, HDS, and
LDS). The following tables represent the distributions for the variables that were measured on a scale/ratio basis used in the
regressions.
Table 65: Original Distribution of Variables - HDU Neighborhoods
Variable
N
Minimum Maximum
Mean
Variance
Age
227
18
94
52.10
264.42
HIncome
201
5000
135000 30422.89
7.78
Educ
225
10
19
12.95
1.46
Num_Child
208
0
8
0.87
1.91
Pct_WTrips_W
93
0
100
13.50
721.73
Pct_WTrips_T
93
0
100
13.46
784.10
Pct_WTrips_D
93
0
100
73.04
1558.28
Pct_OptTrips_W
220
0
100
23.63
1157.80
Pct_OptTrips_T
220
0
100
8.31
391.09
Pct_OptTrips_D
220
0
100
68.06
1549.43
Pct_AllTrips_W
223
0
100
22.97
1080.52
Pct_AllTrips_T
223
0
100
9.84
459.12
Pct_AllTrips_D
223
0
100
67.19
1546.49
Ann_WMiles_W
93
0
10244
287.15 1541254.9
Ann_WMiles_T
93
0
10244
299.31 1718341.1
Ann_WMiles_D
93
0
25953
1849.67 11770138.1
Ann_OptMiles_W
220
0
13010
224.40
955502.8
Ann_OptMiles_T
220
0
12766
259.65 1097011.4
Ann_OptMiles_D
220
0
11318
1214.64 2911302.1
Ann_AllMiles_W
223
0
15844
341.14 2108656.2
Ann_AllMiles_T
223
0
15917
380.98
2642589
Ann_AllMiles_D
223
0
26411
1969.69 10298916.2
Ann_Fuel_Consumption
105
1.49
1049.53
137.60
26578.3
Ann_CO_Emissions
105
131.04
72825.48 11931.21
1.705E8
254

Skewness
-0.058
1.458
1.093
1.769
2.01
1.99
-0.99
1.19
3.03
-0.75
1.28
2.72
-0.74
6.48
6.17
4.41
10.83
8.77
2.70
8.26
8.04
3.62
2.97
2.15

Kurtosis
-0.508
2.066
0.470
3.322
3.23
2.77
-0.71
0.00
9.71
-1.08
0.29
7.56
-1.11
47.54
41.56
26.65
136.32
96.29
9.71
77.06
72.21
18.79
10.88
5.28

The above variables were transformed to the following to get as close to a normal distribution as possible:
Table 66: Transformed Distribution of Variables - HDU Neighborhoods
Variable
N
Minimum Maximum Mean Variance Skewness
Age
227
18
94
52.10
264.42
-0.06
Ln_Hshld_Income
201
8.52
11.81
9.88
1.01
-0.11
Ln_Educ
225
2.30
2.94
2.54
0.03
0.74
Sqrt_Num_Child
208
0.00
2.83
0.55
0.57
0.89
Pct_WTrips_W
93
0
100
13.50
721.73
2.01
Pct_WTrips_T
93
0
100
13.46
784.10
1.99
Pct_WTrips_D
93
0
100
73.04
1558.28
-0.99
Pct_OptTrips_W
220
0
100
23.63
1157.80
1.19
Pct_OptTrips_T
220
0
100
8.31
391.09
3.03
Pct_OptTrips_D
220
0
100
68.06
1549.43
-0.75
Pct_AllTrips_W
223
0
100
22.97
1080.52
1.28
Pct_AllTrips_T
223
0
100
9.84
459.12
2.72
Pct_AllTrips_D
223
0
100
67.19
1546.49
-0.74
LnAnn_WMiles_W
93
0
9.23
0.98
6.00
2.26
LnAnn_WMiles_T
93
0
9.23
0.95
5.96
2.31
LnAnn_WMiles_D
93
0
10.16
4.75
13.24
-0.41
LnAnn_OptMiles_W
220
0
9.47
2.12
7.45
0.72
LnAnn_OptMiles_T
220
0
9.45
1.33
7.07
1.62
LnAnn_OptMiles_D
220
0
10.46
5.42
8.17
-1.00
LnAnn_AllMiles_W
223
0
9.67
2.27
8.15
0.69
LnAnn_AllMiles_T
223
0
9.68
1.51
7.99
1.47
LnAnn_AllMiles_D
223
0
10.18
5.68
8.91
-0.99
LnAnn_Fuel_Consumption
105
0.91
6.96
4.24
1.63
-0.28
LnAnn_CO_Emissions
105
4.88
11.20
8.69
1.76
-0.42

255

Kurtosis
-0.51
-1.21
-0.10
-0.72
3.23
2.77
-0.71
0.00
9.71
-1.08
0.29
7.56
-1.11
3.49
3.71
-1.61
-1.10
0.88
-0.29
-1.07
0.44
-0.29
-0.52
-0.30

Table 67: Original Distribution of Variables - HDS Neighborhoods
Variable
N
Minimum Maximum
Mean
Age
391
19
90
50.99
HIncome
347
5000
155000 96123.92
Educ
389
12
19
17.08
Num_Child
377
0
3
0.54
Pct_WTrips_W
265
0
100
16.03
Pct_WTrips_T
265
0
100
6.66
Pct_WTrips_D
265
0
100
77.30
Pct_OptTrips_W
389
0
100
21.47
Pct_OptTrips_T
389
0
100
1.13
Pct_OptTrips_D
389
0
100
77.39
Pct_AllTrips_W
392
0
100
20.12
Pct_AllTrips_T
392
0
100
2.83
Pct_AllTrips_D
392
0
100
77.05
Ann_WMiles_W
265
0
7041
130.68
Ann_WMiles_T
265
0
4264
64.18
Ann_WMiles_D
265
0
17602
1926.99
Ann_OptMiles_W
391
0
4581
99.89
Ann_OptMiles_T
391
0
1006
9.13
Ann_OptMiles_D
391
0
6750
855.59
Ann_AllMiles_W
392
0
7066
187.98
Ann_AllMiles_T
392
0
4264
52.49
Ann_AllMiles_D
392
0
18502
2156.09
Ann_Fuel_Consumption
347
0.44
925.11
108.76
Ann_CO_Emissions
347
37.13
77708.90
9098.97

256

Variance
262.37
2.178E9
5.23
0.81
961.31
374.04
1377.85
733.74
41.96
802.75
637.88
89.36
772.98
451257.6
128528.9
7621591.4
89325.6
3747.2
803207.2
408523.9
92734.3
6725655.6
16728.9
1.188E8

Skewness
0.29
-0.185
-0.91
1.44
1.84
3.16
-1.26
1.19
8.22
-1.19
1.28
4.23
-1.13
8.22
9.72
2.22
9.86
12.33
2.65
7.48
11.00
2.42
2.36
2.46

Kurtosis
-0.65
-1.19
-0.19
0.75
2.02
9.72
-0.09
0.53
78.04
0.48
0.81
20.56
0.24
71.23
103.75
6.29
133.99
186.02
10.53
64.28
137.71
7.85
7.38
8.06

The above variables were transformed to the following to get as close to a normal distribution as possible:
Table 68: Transformed Distribution of Variables - HDS Neighborhoods
Variable
N
Minimum Maximum
Mean
Age
391
19
90
50.99
Sqrt_HIncome
347
70.71
393.70
298.01
Educ
389
12
19
17.08
Sqrt_Num_Child
377
0
1.73
0.40
Pct_WTrips_W
265
0
100
16.03
Pct_WTrips_T
265
0
100
6.66
Pct_WTrips_D
265
0
100
77.30
Pct_OptTrips_W
389
0
100
21.47
Pct_OptTrips_T
389
0
100
1.13
Pct_OptTrips_D
389
0
100
77.39
Pct_AllTrips_W
392
0
100
20.12
Pct_AllTrips_T
392
0
100
2.83
Pct_AllTrips_D
392
00
100
77.05
LnAnn_WMiles_W
265
0
8.86
1.26
LnAnn_WMiles_T
265
0
8.36
0.69
LnAnn_WMiles_D
265
0
9.78
5.61
LnAnn_OptMiles_W
391
0
8.43
2.12
LnAnn_OptMiles_T
391
0
6.91
0.26
LnAnn_OptMiles_D
391
0
8.82
6.10
LnAnn_AllMiles_W
392
0
8.86
2.45
LnAnn_AllMiles_T
392
0
8.36
0.65
LnAnn_AllMiles_D
392
0
9.83
6.93
LnAnn_Fuel_Consumption
347
0.37
6.83
4.10
LnAnn_CO_Emissions
347
3.64
11.26
8.49

257

Variance
262.37
7337.7
5.23
0.38
961.31
374.04
1377.85
733.74
41.96
802.75
637.88
89.36
772.98
5.52
3.52
9.31
6.13
1.20
2.58
7.08
3.21
2.41
1.33
1.45

Skewness
0.29
-0.70
-0.91
1.03
1.84
3.16
-1.26
1.19
8.22
-1.19
1.28
4.23
-1.13
1.47
2.47
-0.95
0.49
4.09
-2.30
0.40
2.56
-1.85
-0.21
-0.42

Kurtosis
-0.65
-0.27
-0.19
-0.70
2.02
9.72
-0.09
0.53
78.04
0.48
0.81
20.56
0.24
0.59
4.60
-0.45
-1.43
15.60
6.26
-1.41
5.05
6.42
0.01
0.65

Table 69: Original Distribution of Variables - LDS Neighborhoods
Variable
N
Minimum Maximum
Mean
Variance
Age
317
18
92
57.04
226.0
HIncome
273
5000
155000 108516.48
2.012E9
Educ
324
10
19
16.75
6.35
Num_Child
314
0
4
0.65
1.02
Pct_WTrips_W
196
0
100
5.53
348.0
Pct_WTrips_T
196
0
16.67
0.09
1.41
Pct_WTrips_D
196
0
100
94.38
348.47
Pct_OptTrips_W
323
0
87.50
6.83
236.78
Pct_OptTrips_T
323
0
27.27
0.18
3.43
Pct_OptTrips_D
323
12.50
100
92.99
238.82
Pct_AllTrips_W
325
0
83.12
6.92
220.36
Pct_AllTrips_T
325
0
27.27
0.20
3.57
Pct_AllTrips_D
325
16.88
100
92.88
222.49
Ann_WMiles_W
196
0
7072
88.59
305884.6
Ann_WMiles_T
196
0
572
2.92
1669.3
Ann_WMiles_D
196
0
69940
3354.82 37357268.9
Ann_OptMiles_W
324
0
2803
68.11
80199.5
Ann_OptMiles_T
324
0
577
2.28
1066.8
Ann_OptMiles_D
324
0
18153
1921.06 3891470.3
Ann_AllMiles_W
325
0
7072
121.32
261892.6
Ann_AllMiles_T
325
0
577
4.03
2062.2
Ann_AllMiles_D
325
0
72868
3938.36 29228784.5
Ann_Fuel_Consump
284
4
4858
214.42
124629.4
Ann_CO_Emissions
284
349
181687 16428.26
3.397E8

258

Skewness
-0.12
-0.50
-0.85
1.30
3.94
14.00
-3.92
2.76
12.15
-2.71
2.75
11.52
-2.69
10.81
14.00
8.05
6.79
17.07
3.41
9.13
12.35
7.88
9.27
4.55

Kurtosis
-0.41
-1.01
-0.20
0.41
15.86
196.00
15.74
7.79
159.82
7.50
7.97
146.34
7.63
132.44
196.00
80.17
54.17
299.95
19.01
109.68
153.75
89.60
111.50
32.85

The above variables were transformed to the following to get as close to a normal distribution as possible:
Table 70: Transformed Distribution of Variables - LDS Neighborhoods
Variable
N
Minimum Maximum
Mean
Age
317
18
92
57.04
Sqrt_HIncome
273
70.71
393.70
320.05
Educ
324
10
19
16.75
Sqrt_Num_Child
314
0
2
0.46
Pct_WTrips_W
196
0
100
5.53
Pct_WTrips_T
196
0
16.67
0.09
Pct_WTrips_D
196
0
100
94.38
Pct_OptTrips_W
323
0
87.50
6.83
Pct_OptTrips_T
323
0
27.27
0.18
Pct_OptTrips_D
323
12.50
100
92.99
Pct_AllTrips_W
325
0
83.12
6.92
Pct_AllTrips_T
325
0
27.27
0.20
Pct_AllTrips_D
325
16.88
100
92.88
LnAnn_WMiles_W
196
0
8.86
0.52
LnAnn_WMiles_T
196
0
6.35
0.03
LnAnn_WMiles_D
196
0
11.16
6.68
LnAnn_OptMiles_W
324
0
7.94
0.98
LnAnn_OptMiles_T
324
0
6.36
0.05
LnAnn_OptMiles_D
324
0
9.81
7.12
LnAnn_AllMiles_W
325
0
8.86
1.23
LnAnn_AllMiles_T
325
0
6.36
0.06
LnAnn_AllMiles_D
325
0
11.20
7.76
LnAnn_Fuel_Consumption
284
1.68
8.49
4.85
LnAnn_CO_Emissions
284
5.86
12.11
9.24

259

Variance
-0.12
6107.5
-0.85
0.44
348.0
1.41
348.47
236.78
3.43
238.82
220.36
3.57
222.49
3.13
0.21
8.01
4.19
0.24
1.22
5.3
0.36
1.43
1.09
1.06

Skewness
-0.12
-0.94
-0.85
0.86
3.94
14.00
-3.92
2.76
12.15
-2.71
2.75
11.52
-2.69
3.25
14.00
-1.62
1.85
10.84
-1.87
1.59
9.32
-1.85
-0.28
-0.40

Kurtosis
-0.41
0.11
-0.20
-1.02
15.86
196.00
15.74
7.79
159.82
7.50
7.97
146.34
7.63
9.10
196.00
1.39
1.98
120.38
9.84
1.00
87.80
9.76
0.13
-0.15

BIBLIOGRAPHY

260

BIBLIOGRAPHY

Adams, C. F., Fleeter, H. B., Kim, Y., Freeman, M., & Cho, I. (1996). Flight from Blight and
Metropolitan Suburbanization Revisited. Urban Affairs Review, 31(4), 529-543.
Adams, J. and VanDrasek, B. (1993) Minneapolis-St Paul: People, Place, and Public Life.
Minneapolis: University of Minnesota Press.
Ahn, K. and Rakha, H. (2008). The effects of route choice decisions on vehicle energy
consumption and emissions. Transportation Research Part D: Transport and Environment,
13(3), 151-167.
Ananat, E. (2011). The Wrong Side(s) of the Tracks: The Causal Effects of Racial Segregation
on Urban Poverty and Inequality. American Economic Journal: Applied Economics, 3(2),
34-66.
Bachman, W., Sarasua, W., Hallmark, S., and Guensler, R. (2000). Modeling regional mobile
source emissions in a geographic information system framework. Transportation Research
Part C: Emerging Technologies, 8(1-6), 205-229.
Bartholomew, K. and Ewing, R. (2008). Land Use-Transportation Scenarios and Future Vehicle
Travel and Land Consumption: A Meta Analysis. Journal of the American Planning
Association, 75(1), 13-27.
Baum-Snow, N. (2007). Suburbanization and transportation in the monocentric model. Journal
of Urban Economics, 62(3), 405-423.
Baum-Snow, N. (2007b) Did Highways Cause Suburbanization? Quarterly Journal of
Economics, 122(2), 775-805.
Baum-Snow, N. and Lutz, B. (2011). School Desegregation, School Choice, and Changes in
Residential Location Patterns by Race. American Economic Review, 101(7), 3019-3046.
Bento, A. M., Cropper, M. L., Mobarak, A. M., & Vinha, K. (2005). The Effects of Urban
Spatial Structure on Travel Demand in the United States. The Review of Economics and
Statistics, 87(3), 466-478.
Bhat, C. and Guo, J. (2007). A comprehensive analysis of built environment characteristics on
household residential choice and auto ownership levels. Transportation Research Part B:
Methodological, 41, 506-526
Binelli, Mark. (2012). Detroit City is the place to be: the afterlife of an American metropolis.
Metropolitan Books: New York, NY.

261

Bledsoe, T., Welch, S., Sigelman, L., and Combs, M. (1995). Residential Context and Racial
Solidarity among African Americans. American Journal of Political Science, 39(2), 434458.
Blight Busters Detroit. (2013). Official Website. Available at http://blightbustersdetroit.com/.
Accessed on July 15, 2013.
Blumenberg, E. (2004). En-gendering Effective Planning: Spatial Mismatch, Low-Income
Women, and Transportation Policy. Journal of the American Planning Association, 70(3),
269-281.
Blumenberg, E. and Manville, M. (2004). Beyond the Spatial Mismatch: Welfare Recipients and
Transportation Policy. Journal of Planning Literature, 19(2), 182-205.
Board of County Auditors. (1926). Manual: County of Wayne Michigan. Retrieved on April 29,
2013 at http://archive.org/stream/manualcountyofwa00wayn#page/n0/mode/1up
Boriboonsomsin, K. and Barth, M. (2009). Impacts of Road Grade on Fuel Consumption and
Carbon Dioxide Emissions Evidenced by use of Advanced Navigation Systems.
Transportation Research Record, 2139, 21-30.
Boustan, L. (2012). School Desegregation and Urban Change: Evidence from City Boundaries.
American Economic Journal: Applied Economics, 4(1), 85-108.
Brown, J., Morris, E., and Taylor, B. (2009). Planning for Cars in Cities: Planners, Engineers,
th
and Freeways in the 20 Century. Journal of the American Planning Association, 75(2),
161-177.
Brown, M., Southworth, F., and Sarzynski, A. (2008). Shrinking the Carbon Footprint of
Metropolitan America. Metropolitan Policy Program. Brookings Institution: Washington,
DC.
Brownstone, D., & Golob, T. (2009). The impact of residential density on vehicle usage and
energy consumption. Journal of Urban Economics, 65, 91-98.
Bureau of Transportation Statistics. (2007). National Transportation Statistics. Bureau of
Transportation: Washington, DC
Cao, X., Mokhtarian, P., and Handy, S. (2006). Neighborhood Design and Vehicle Type Choice:
evidence from Northern California. Transportation Research Part D, 11, 133-145.
Cao, X., Mokhtarian, P., and Handy, S. (2007). Cross-sectional and quasi-panel explorations of
the connection between the built environment and auto ownership. Environment and
Planning A, 39, 830-847.
Card, D., & Krueger, A. (1996). School Resources and Student Outcomes: An Overview of the
Literature and New Evidence from North and South Carolina. Journal of Economic
Perspectives, 10(4), 31-50.
262

Carroll-Larson, J. and Caplan, A. (2009). Estimating the Effectiveness of a Vehicle Miles
Traveled Tax in Reducing Particulate Matter Emissions. Journal of Environmental Planning
and Management, 52(3), 315-344.
CBS Detroit. (2013). Meijer Readying for Official Grand Opening in Detroit. Online at
http://detroit.cbslocal.com/2013/07/11/meijer-readying-for-official-grand-opening-in-detroit/.

Accessed on July 11, 2013.
Center for Disease Control –CDC (2010). Adult Obesity Facts. Online at
http://www.cdc.gov/obesity/data/adult.html Accessed May 15, 2012.
Cervero, R. and Murakami, J. (2010). Effects of Built Environments on Vehicle Miles Traveled:
Evidence from 370 US Urbanized Areas. Environment and Planning A, 42(2), 400-418.
Cervero, R., & Radisch, C. (1996). Pedestrian Versus Automobile Oriented Neighborhoods.
Transportation Policy, 3, 127-141.
Chen, C., Gong, H., and Paaswell, R. (2008). Role of the built environment on mode choice
decisions: additional evidence on the impact of density. Transportation, 35, 285-299.
Cho, D. and Marr, K. (2009). “U.S. Expands Aid to Auto Industry”. The Washington Post.
Published January 17, 2009. Available online at http://www.washingtonpost.com/wpdyn/content/article/2009/01/16/AR2009011602874_2.html%3Fhpid%3Dtopnews. Accessed
on April 25, 2013.
Clark, W. (2007). Race, Class, and Place: Evaluating Mobility Outcomes for African Americans.
Urban Affairs Review, 42(3), 295-314.
Clark, W. and Blue, S. (2004). Race, Class and Segregation Patterns in U.S. Immigrant Gateway
Cities. Urban Affairs Review, 39(6), 667-688.
Cohn, S., & Fosselt, M. (1998). The other reason job suburbanization hurts blacks: The
relationship between the Location and Racial Composition of employment in Detroit and
Atlanta, 1980. Urban Affairs Review, 34(1), 94-125.
Counts, Jeff. (2011). Detroit and Ann Arbor: A great destination. Woodstock, VT: The
Countryman Press.
Crane, R., & Crepeau, R. (1998). Does Neighborhood Design Influence Travel? A Behavioral
Analysis of Travel Diary and GIS Data. Transportation Research Part D, 3(4), 225-238.
Dahl, Bill. (2001). Motown: The golden years. Iola, WS: Krause Publications.
Darden and Kamel (2000)
Darden, J. T. & Kamel, S. M. (2000). Black Residential Segregation in the City and Suburbs of
Detroit: Does Socioeconomic Status Matter? Journal of Urban Affairs, 22(1), 1-13.

263

Darden, J., Hill, R., Thomas, J., and Thomas, R. (1987). Detroit: Race and Uneven Development.
Philadelphia: Temple University Press.
Darden, Joe and Thomas, Richard. (2013). Detroit: Race Riots, Racial Conflicts and Efforts to
Bridge the Racial Divide. East Lansing, MI: Michigan State University Press.
Darden, Joe, Stokes, Curtis and Thomas, Richard. (2007). The State of Black Michigan, 19672007. East Lansing, MI: Michigan State University Press.
Darden, Joe. (2009). Race Matters in Metro Detroit, in Schaetzl, R., Darden, J. and Brandt,x,
(Eds) Michigan Geography and Geology. XXXXXX
Data Driven Detroit (2012). State of the Detroit Child. Report prepared for the Skillman
Foundation. Detroit, MI: Skillman Foundation.
Davey, M. and Walsh, M. W. (2013). “Detroit Creditors and Unions Prepare for Bankruptcy
Fight”. New York Times, published July 19, 2013. Available online at
http://www.nytimes.com/2013/07/20/us/breadth-of-bankruptcy-fight-detroit-faces-becomingclear.html?ref=us&_r=0. Accessed on July 19, 2013.
Detroit Historical Society. (2013). Encyclopedia of Detroit: Arsenal of Democracy. Available
online at http://detroithistorical.org/learn/encyclopedia-of-detroit/arsenal-democracy.
Accessed on July 7, 2013.
Detroit Public Schools. (2011). DPS reaches 62 percent graduation rate, the highest since state
began new cohort methodology in 2007. Retrieved on 29 April, 2013 at
http://detroitk12.org/content/2011/02/22/dps-reaches-62-percent-graduation-rate-the-highestsince-state-began-new-cohort-methodology-in-2007/

DeVise, P. (1976). The Suburbanization of Jobs and Minority Employment. Economic
Geography, 52(4), 348-362.
Drummond, William J. (2010) Statehouse Versus Greenhouse. Journal of the American Planning
Association 76(4), 413-433.
Dunphy, R., & Fisher, K. (1996). Transportation, Congestion, and Density: New Insights.
Transportation Research Record, 1551, 89-96.
Duvall, R., Norris, G., Burke, J., Olson, D., Vedantham, R. and Williams, R. (2012).
Determining Spatial Variability in PM2.5 Source Impacts Across Detroit, MI. Atmospheric
Environment, 47(1), 491-498.
Eggert, D. (2013). “Gas tax increase proposed to help pay for Michigan road repairs”. ABC
Action News. Available online at http://www.wxyz.com/dpp/news/sales-tax-could-increaseto-help-pay-for-michigan-road-repairs. Accessed on August 27, 2013.

264

Environmental Protection Agency (EPA). (2012). Gas Guzzler Tax – Program Overview. EPA –
Office of Transportation and Air Quality. Available online at
http://www.epa.gov/fueleconomy/guzzler/420f12068.pdf. Accessed on August 27, 2013.
Ewing, R. and Cervero, R. (2001). Travel and the Built Environment. Transportation Research
Record, 1780, 87-114.
Ewing, R. and Cervero, R. (2010). Travel and the Built Environment. Journal of the American
Planning Association, 76(3), 265-294.
Ewing, R., Pendall, R., Chen, D. (2002). Measuring Sprawl and its Impact. Washington DC:
Smart Growth America.
Ewing, R., Pendall, R., Chen, D. (2003a). Measuring Sprawl and its transportation impacts.
Transportation Research Record, 1831(X), 175-183.
Ewing, R., Schieber, R., Zegeer, C. (2003b). Urban Sprawl as a Risk Factor in Motor Vehicle
Occupant and Pedestrian Facilities. American Journal of Public Health, 93(X), 1541-1545.
Farley, R., Couper, M., and Krysan, M. (2007). Race and Revitalization in the Rust Belt: A
Motor City Story. A Report on findings from the University of Michigan’s 2004 Detroit Area
Study. Report by the Population Studies Center, University of Michigan: Ann Arbor, MI.
Farley, Reynolds, Danzinger, Sheldon, and Holzer, Harry. (2000) Detroit Divided. New York:
Russell Sage Books.
Fischer, M. (2008). Shifting Geographies: Examining the Role of Suburbanization in Blacks’
Declining Segregation. Urban Affairs Review, 43(4), 475-496.
Fitzgerald, J. (2010) Emerald Cities: Urban Sustainability and Economic Development. New
York: Oxford University Press.
Frey, W. (2000). Melting Pot Suburbs: A Census 2000 study of suburban diversity. Washington,
DC: Center on Urban and Metropolitan Policy, Brookings Institution.
Gallagher, J. (2013). A Fresh Start: Whole Foods opens in Detroit with hoopla, eager crowd (and
some shoppers). Detroit Free Press. Available online at
http://www.freep.com/article/20130605/BUSINESS06/306050069/whole-foods-openingcustomers-wait. Accessed on June 10, 2013.
Gallagher, John. (2010). Reimagining Detroit: Opportunities for Redefining an American City.
Detroit, MI: Wayne State University Press.
Galster, G., Hanson, R., Ratcliffe, M. R., Wolman, H., Coleman, S., & Freihage, J. (2001).
Wrestling Sprawl to the Ground: Defining and Measuring an Elusive Concept. Housing
Policy Debate, 12(4), 681-717.
Galster, George. (2012). Driving Detroit. Philadelphia, PA: University of Pennsylvania Press.
265

Gas Guzzler Tax, 26 USC § 4064
Glaeser, E., Kahn, M., & Rappaport, J. (2008). Why do the poor live in cities? The Role of
public transportation. Journal of Urban Economics, 63(1), 1-24.
Glaeser, E., Kahn, M., and Rappaport, J. (2008). Why do the poor live in cities? The Role of
public transportation. Journal of Urban Economics, 63(1), 1-24.
Goetzke, F. and Weinberger, R. (2012) Separating contextual from endogenous effects in
automobile ownership models. Environment and Planning A, 44, 1032-1046.
Greene, J. and Winters, M. (2006). Leaving Boys Behind: Public High School Graduation Rates.
Civic Report 48. New York: Manhattan Institute for Policy Research.
Grengs, J. (2010). Job Accessibility and Modal Mismatch in Detroit. Journal of Transport
Geography, 18(1), 42-54.
Harrison, B. (1974). Urban Economic Development. Washington D.C.: The Urban Institute.
Hitchens, P. (2011). “From Motown to Ghost Town: How the once mighty Detroit is heading
down a long, slow road to ruin”. Daily Mail. Available online at
http://www.dailymail.co.uk/news/article-2012971/From-Motown-Ghost-town-How-mightyDetroit-heading-long-slow-road-ruin.html. Accessed on May 18, 2013.
Holtzclaw, J., Clear, R., Dittmar, H., Goldstein, D., & Haas, P. (2002). Location Efficiency:
Neighborhood and Socio-Economic Characteristics Determine Auto Ownership and Use Studies in Chicago, Los Angeles and San Francisco. Transportation Planning and
Technology, 25, 1-27.
Huffstutter, P. (2009). Investors see farms as way to grow Detroit. Published in the LA Times
on December 27, 2009. Available online at http://articles.latimes.com/2009/dec/27/nation/lana-detroit-farms27-2009dec27. Accessed on April 30, 2013.
ICPSR. (n.d.) Detroit Area Studies Series. In Inter-University Consortium for Political and
Social Research. Retrieved on July 8, 2012, from
http://www.icpsr.umich.edu/icpsrweb/ICPSR/series/00151.
Independent Lens. (2013). Welcome to Detroit: America’s Fastest Shrinking City. Film
produced for PBS Video. Aired on May 27, 2013. Available online at
http://video.pbs.org/video/2354657351/.

Jackson, K. T. (1987) Crabgrass Frontier: the suburbanization of the United States. Oxford
University Press.
Jacobs, J. (1961). The Death and Life of Great American Cities. Penguin Books:
Harmondsworth, UK

266

Jordan, S., Ross, J. P., & Usowski, K. G. (1998). U.S. Suburbanization in the 1980s. Regional
Science and Urban Economics, 28(5), 611-627.
Kahn, M. (2006). The Quality of Life in Sprawled Versus Compact Cities. ECMT Regional
Round Tables, 137. (OECD, Paris).
Kain, J. (1967) Postwar metropolitan development: housing preferences and auto ownership.
American Economic Review, 57, 223-234.
Kain, J. (1968) Housing segregation, Negro employment, and metropolitan decentralization. The
Quarterly Journal of Economics, 82, 175-197.
Kerner Commission. (1968). Report of the National Advisory Commission on Civil Disorders.
US Government Printing Office: Washington DC.
Kirby, D. and LeSage, J. (2009). Changes in commuting to work times over the 1990 to 2000
period. Regional Science and Urban Economics, 39(4), 460-471.
Kockelman, K. M. (1996). Travel Behavior as a Function of Accessibility, Land Use Mixing, and
Land Use Balance: Evidence from the San Francisco Bay Area. UC Berkeley, Berkeley.
Konduri, K., Ye, X., Sana, B., and Pendyala, R. (2011). Joint Model of Vehicle Type Choice
and Tour Length. Transportation Research Record, 2255, 28-37.
Kopecky, K. and Suen, R. (2010). A Quantitative Analysis of Suburbanization and the Diffusion
of the Automobile. International Economic Review, 51(4), 1003-1037.
Krizek, K. J. (2003). Residential Relocation and Changes in Urban Travel: Does NeighborhoodScale Urban Form Matter? Journal of the American Planning Association, 69, 125-158.
Kromer, M. A., Bandivadekar, A. & Evans, C. (2010). Long-Term Greenhouse Gas Emission
and Petroleum Reduction Goals: Evolutionary Pathways for the Light-Duty Vehicle Sector.
Energy Policy 35, 387-397.
Lankao, P. R., Tribbia, J. L. & Nychka, D. (2009). Testing Theories to Explore the Drivers of
Cities' Atmospheric Emissions. Ambio 38(4), 236-244.
Leichenko, R. (2001) Growth and Change in U.S. Cities and Suburbs, Growth and Change, 32,
326-354.
Lewis, R. (2002). The Changing Fortunes of American Central-City Manufacturing, 1870-1950.
Journal of Urban History, 28(5), 573-598.
Lin, J., & Long, L. (2008). What Neighborhood Are You In? Empirical Findings of
Relationships Between Household Travel and Neighborhood Characteristics. Transportation
35, 739-758.

267

Lindhjem, C., Pollack, A., DenBleyker, A. and Shaw, S. (2012). Effects of Improved Spatial and
Temporal Modeling of On-Road Vehicle Emissions. Journal of the Air & Waste
Management Association, 62(4), 471-484.
Lindsey, M., Schofer, J., Durango-Cohen, P. and Gray, K. (2011). The Effect of Residential
Location on Vehicle Miles of Travel, Energy Consumption and Greenhouse Gas Emissions:
Chicago Case Study. Transportation Research Part D: Transport and Environment, 16(1), 19.
Liu, C. and Shen, Q. (2011). An Empirical Analysis of the Influence of Urban Form on
Household Travel and Energy Consumption. Computers Environment and Urban Systems,
35(5), 347-357.
Lutsey, N. & Sperling, D. (2009). Greenhouse Gas Mitigation Supply Curve for the United
States for Transport Versus Other Sectors. Transportation research Part D, 14, 222-229.
Mach, A. (2013). “‘We’ve lost respect for life’: Detroit records deadliest year in decades”. NBS
News. Online at http://usnews.nbcnews.com/_news/2013/01/04/16352612-weve-lostrespect-for-life-detroit-records-deadliest-year-in-decades?lite. Accessed on May 18, 2013.
Manski, C. (2000). Economic Analysis of Social Interactions. Journal of Economic Perspective,
14, 115-136.
Mashayekh, Y., Jaramillo, P., Chester, M., Hendrickson, C., and Weber, C. (2011). Costs of
Automobile Air Emissions in US Metropolitan Areas. Transportation Research Record,
2233, 120-127.
Maynard, N. (2011). “Detroit: A Boom Town Goes Bust”. PBS Newshour. Available Online at
http://www.pbs.org/newshour/rundown/2011/03/-this-entry-is-cross-posted.html. Accessed
on April 25, 2013.
McClam, E. (2013). “Ex-Detroit mayor Kilpatrick convicted of range of corruption charges”.
NBC News, published March 11, 2013. Available online at
http://usnews.nbcnews.com/_news/2013/03/11/17269402-ex-detroit-mayor-kilpatrickconvicted-of-range-of-corruption-charges?lite. Accessed on April 25, 2013.
McCollum, D. & Yang, C. (2009). Achieving Deep Reductions in Us Transport Greenhouse Gas
Emissions: Scenario Analysis and Policy Implications, Energy Policy 37, 5580-5596.
McGrath, D. T. (2005). More evidence on the spatial scale of cities. Journal of Urban
Economics, 58(1), 1-10.
McGuckin, and Srinivasan, (2003). Journey to Work Trends in the United States and its Major
Metropolitan Areas 1960-2000. US Department of Transportation Federal Highway
Administration, Office of Planning. Publication No. FHWA-EP-03-058.
Meier, August, and Rudwick, Elliott. (1981). Black Detroit and the Rise of the UAW. Oxford:
Oxford University Press.
268

Meyer, M. (2010) Greenhouse Gas and Climate Change Assessment. Journal of the American
Planning Association 76(4), 402-412.
Michigan Department of Transportation (MDOT). 2013. “One small step for public transit, one
giant leap for Detroit (Part 2)”. MDOT Website. Available online at
http://www.michigan.gov/mdot/0,4616,7-151-63670_63671_63933-293895--,00.html.
Accessed on April 25, 2013.
Mieszkowski, P. & Mills, E. (1993) The causes of Metropolitan Suburbanization. Journal of
Economic Perspectives, 7(3), 135-147.
Mirel, J. (1999). The Rise and Fall of an Urban School System: Detroit 1907-81. Second
Edition. The University of Michigan Press, Ann Arbor, MI.
Moore, A., Staley, S., and Poole, R. (2010). The role of VMT reduction in meeting climate
policy goals. Transportation Research Part A: Policy and Practice, 44(8), 565-574.
Morrow, R., Gallagher, K. S., Collantes, G. & Lee, H. (2010). Analysis of Policies to Reduce Oil
Consumption and Greenhouse-Gas Emissions from the Us Transportation Sector. Energy
Policy 38, 1305-1320.
National Park Service. National Register of Historic Places. US Department of the Interior.
Online at http://www.nps.gov/nr/. Accessed on 4/1/2013.
National Research Council and the Committee on Health, Environmental, and Other External
Costs and Benefits of Energy Production and Consumption. (2010). Hidden Cost of Energy:
Unpriced Consequences of Energy Production and Use. National Academies Press:
Washington, DC.
Newman, P. and Kenworthy, J. (1999). Sustainability and Cities: Overcoming Automobile
Dependence. Washington D.C.: Island Press.
Newman, P., & Kenworthy, J. (1989). Cities and Automobile Dependence: An International
Sourcebook: Gower Publishing.
Oosting, J. (2013). Snyder calls for hike in gas tax, registration fees: ‘If you use the roads more,
you should pay more’. MLive, February 08, 2013. Available online at
http://www.mlive.com/politics/index.ssf/2013/02/snyder_backs_hike_in_gas_tax_r.html.
Accessed on August 27, 2013.
Pendall R, 2000,“Local land use regulation and the chain of exclusion” Journal of the American
Planning Association 66(2), 125-142.
Plotkin, S. (1999). Technologies and Policies for Controlling Greenhouse Gas Emissions from
the U.S. Automobile and Light Truck Fleet. Transportation Quarterly 53(2), 7-30.
Potoglou, D. (2008). Vehicle-type choice and neighborhood characteristics: An empirical study
of Hamilton, Canada. Transportation Research Part D, 13(3), 177-186.
269

Pucher, J., & Renne, J. (2003). Socioeconomics of Urban Travel: Evidence from the 2001
NHTS. Transportation Quarterly, 57(3), 49-77.
Renolds, F., Danziger, S., and Holzer, H. (2000). Detroit Divided: A Volume in the Multi-City
Study of Urban Inequality. Russell Sage Foundation: New York.
Reschovsky, C. (2004). Journey to Work: 2000. United States Census Bureau: Washington DC.
Rhodes, J. and Jeffries, J. (2010). “Motor City Panthers” in Jeffries, J. (Ed) On the Ground: The
Black Panther Party in Communities Across America. University Press of Mississippi, Pp
125-185.
Rodriguez, D. A., Khattak, A. J., & Evenson, K. R. (2006). Can New Urbanism Encourage
Physical Activity? Comparing a New Urbanist Neighborhood with Conventional Suburbs.
Journal of the American Planning Association, 72(1), 43-54.
Rossetti, M. and Eversole, B. (1993). Journey to Work Trends in the United States and its Major
Metropolitan Areas, 1960-1990. United States Department of Transportation, Federal
Highway Administration: Washington DC.
Ryan, B. (2006) Morphological change through residential redevelopment: Detroit, 1951-2000.
Urban Morphology, 10(1), Pp. 5-22.
Ryan, B. (2012) Design after Decline: How America rebuilds Shrinking Cities. Philadelphia, PA:
University of Pennsylvania Press.
Ryan, B. (2012). “People Want These Houses: The Suburbanization of Detroit”, in Ryan, B.
(Ed.) Design After Decline: How America Rebuilds Shrinking Cities. Philadelphia, PA:
University of Pennsylvania Press, Pg 84-127.
Schiller, P., Bruun, E., and Kenworthy, J. (2010). An Introduction to Sustainable Transportation:
Policy, Planning and Implementation. Earthscan: London
Schimek, P. (1996). Household Motor Vehicle Ownership and Use: How Much does Residential
Density Matter? Transportation Research Record, 1552, 120-125.
Schmalensee, R. (1998). World Carbon Dioxide Emissions: 1950-2050. The Review of
Economics and Statistics, 80(1), 15-27.
Schwartz, Y. and Gold, J. (2012). “Detroit may let abandoned buildings burn; film documents
firefighters’ tough times”. NBC News. Available online at
http://usnews.nbcnews.com/_news/2012/04/24/11376348-detroit-may-let-abandonedbuildings-burn-film-documents-firefighters-tough-times?lite. Accessed on May 18, 2013.
Seelye, Katharine. (2011). Detroit Census Confirms a Desertion like no other. New York Times,
Published March 23, 2011, Pg A1.

270

Shay, E. and Khattak, A. (2012). Household Travel Decision Chains: Residential Environment,
Automobile Ownership, Trips and Mode Choice. International Journal of Sustainable
Transportation, 6(2), 88-110.
Shelley v. Kraemer. 334 U.S. 1. Supreme Court of the U.S. 3 May 1948.
Shoup, D. (2005). The High Cost of Free Parking. American Planning Association: Chicago, IL
Shrank, D. and Lomax, T. (2009) The 2009 Urban Mobility Report. Texas Transportation
Institute, Texas A&M: Austin.
Sperling, D. & Gordon, D. (2008). Advanced Passenger Transport Technologies. Annual Review
of Environment and Resources 33, 63-84.
Squires G, & Kubrin C, 2005,“Privileged places: race, uneven development and the geography
of opportunity in urban America” Urban Studies 42(1), 47-68.
Stone, B. (2008). Urban Sprawl and Air Quality in Large Us Cities. Journal of Environment
Management 86, 688-698.
Stone, B. J., Mednick, A. C., Holloway, T., & Spak, S. N. (2007). Is Compact Growth Good for
Air Quality? Journal of the American Planning Association, 73(4), 404-418.
Su, Q. (2010). Travel Demand in the US Urban Areas: A System Dynamic Panel Data Approach.
Transportation Research Part A: Policy and Practice, 44(2), 110-117.
Sugrue, Thomas (2010). From Motor City to Motor Metropolis: How the Automobile Industry
Reshaped Urban America. Online at:
http://www.autolife.umd.umich.edu/Race/R_Overview/R_Overview2.htm#popsugrue Accessed
on March 14, 2013.
Sugrue, Thomas. (1996). The Origins of the Urban Crisis: Race and Inequality in Post-war
Detroit. Princeton, NJ: Princeton University Press.
Sultana, S. (2002). Job/Housing Imbalance and Commuting Time in the Atlanta Metropolitan
Area: Exploration of Causes of Longer Commuting Time. Urban Geography, 23(8), 728749.
Taylor, B. D., & Mauch, M. (1998). Gender, Race and Travel Behavior: An Analysis of
Household-serving Travel and Commuting in the San Francisco Bay Area. Paper presented
at the Federal Highway Administration's National Conference on Women's Travel Issues.
The Heidelberg Project. (2013). Official Website. Available online at http://heidelberg.org/.
Accessed on July 10, 2013.
Thomas, J. (1997). Redevelopment and Race: Planning a Finer City in Postwar Detroit.
Baltimore, MD: John Hopkins University Press.

271

Thomas, June. (1997). Redevelopment and Race: Planning a Finer City in Postwar Detroit.
Baltimore: John Hopkins University Press.
Thomas, Richard (1992). Life for Us Is What We Make It: Building Black Community in Detroit,
1915-1945. Bloomington: Indiana University Press.
Ulfarsson, G. F., & Carruthers, J. I. (2006). The cycle of fragmentation and sprawl: a conceptual
framework and empirical model. Environment and Planning B, 33, 767-788.
Vojnovic, I. 1999. “The environmental costs of modernism: an assessment of Canadian cities.”
Cities: The International Journal of Urban Policy and Planning. Vol. 16, No. 5. Pp. 301313.
Vojnovic, I. (2000a). “Municipal restructuring, regional planning, and fiscal accountability:
designing municipal tax zones in two Maritime regions.” Canadian Journal of Regional
Science. Vol. 23. No.1. Pp. 49-72.
Vojnovic, I. (2000b). Shaping Metropolitan Toronto: A study of linear infrastructure subsidies,
1954-66. Environment and Planning B, 27(2), 197-230.
Vojnovic, I. (2006). Building Communities to Promote Physical Activity: A Multi-Scale
Geographical Analysis. Geografisca Annaler, 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.
Vojnovic, I., Jacson-Elmoore, C., Holtrop, J., and Bruch, S. (2006). The Renewed Interest in
Urban Form and Public Health: Promoting Increased Physical Activity in Michigan. Cities,
23(1), 1-17.
Vojnovic, I., Lee, J., Kotval-K, Z., Podagrosi, A., Varnakovida, P., LeDoux, T., and Messina, J.
(forthcoming). 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.
Vojnovic, I., Smith S., Kotval, Z., and Lee, J. (2007). “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. Pp. 93-129
Voyles, Kenneth and Rodigue, Mary (2012) The enduring legacy of the Detroit Athletic Club:
driving the Motor City. Charleston, SC: History Press.
Walker, R., & Lewis, R. D. (2001). Beyond the Crabgrass Frontier: industry and the spread of
North American cities, 1850-1950. Journal of Historical Geography, 27(1), 3-19.
Warner, S. (2004). Streetcar Suburbs: The Process of Growth in Boston, 1870 – 1900.
Cambridge, MA: Harvard University Press.
272

Weber, J., & Kwan, M.-P. (2003). Evaluating The Effects of Geographic Contexts on Individual
Accessibility: A Multilevel Approach. Urban Geograpy, 24(8), 647-671.
Weinberger, R. and Goetzke, F. (2010). Unpacking preference: How previous xperience affects
auto ownership. Urban Studies, 47, 2111-2128.
Weinberger, R., Seaman, M., and Johnson, C. (2009). Residential off-street parking: Car
ownership, VMT and related carbon impacts. Transportation Research Record, 2118, 24-30.
Weingroff, R. (1996). Federal-Aid Highway Act of 1956: Creating the Interstate System. Public
Roads, 60(1), 1-14.
Wheeler, S. (2008). State and Municipal Climate Change Plans. Journal of the American
Planning Association, 74(4), 481-496.
Whitt, A. and Yago, G. (1985). Corporate Strategies and the Decline of Transit in U.S. Cities.
Urban Affairs Review, 21(1), 37-65.
Williams, Jeremy. (2009). Detroit: The Black Bottom Community. Charleston, SC: Arcadia
Publishing.
Winsborough, H. H. (1963). An Ecological Approach to the Theory of Suburbanization. The
American Journal of Sociology, 68(5), 565-570.
Woosley, M. (2008). “Top Suburbs to live well”. Forbes.com Published on March 26, 2008.
Available online at http://www.forbes.com/2008/03/25/suburbs-quality-lifestyle-forbeslifecx_mw_0326realestate.html. Accessed on April 25, 2013.
Xie, F. and Levinson, D. (2010). How Streetcars shaped Suburbanization: a Granger causality
analysis of land use and transit in the Twin cities. Journal of Economic Geography, 10(3),
453-470.
Yang, C., McCollum, D., McCarthy, R. and Leighty, W. (2009). Meeting an 80% Reduction in
Greenhouse Gas Emissions from Transportation by 2050: A Case Study in California.
Transportation research Part D, 14, 147-156.
York, R. (2003). Cross-National Variation in the Size of Passenger Car Fleets: A Study in
Environmentally Significant Consumption. Population and Environment, 25(2), 119-140.
Zhang, H. (2011). School Desegregation and White Flight Revisited: A Spatial Analysis from a
Metropolitan Perspective. Urban Geography, 32(8), 1208-1226.
Zhao, Y. and Kockelman, K. (2000). “Household Vehicle Ownership by vehicle type:
st
Application of a Multivariate negative binomial model”. In: Proceedings of the 81 TRB
Annual Meeting, Washington, DC.
Zolnik, E. (2011). The effects of sprawl on private-vehicle commuting distances and times.
Environment and Planning B: Planning and Design, 38(6), 1071-1084.
273