-1
Has
.9.

arumm.

.v'l - a,
:.>uwr£ww.unﬁd.
. ll
Lnﬂﬁnﬂﬂfnmﬂ

s.

_
. 4
awaieﬂfss

9': IS!
.5. a...

t 1.1....
w! Aurel.
‘

 

u .
99...}. a. .u
.vhizv-KLED ‘ it.
ad. ér...ix.....tV 112.:
55....

 

I
.li?
31.5.

i I $.92
.1: ﬁfitlﬁllt.
£21.55!

_._1

 

 

|u

 

yarn LIBRARY
" Mich ;, "tate
Uriviar§jix____l

 

 

 

This is to certify that the
dissertation entitled

DO LOCAL ROADS MATTER? LINKING LOCAL ROADS
SPENDING TO DECENTRALIZATION IN THE DETROIT
METROPOLITAN AREA

presented by

ANNALIE L. CAMPOS

has been accepted towards fulﬁllment
of the requirements for the

Ph. D. degree in Geggraphy

 

 

 

Major Professor’s Signature
5M7 25’9”; 2 0 o 2

 

Date

MSU is an Afﬁnnative Action/Equal Opportunity Employer

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc8Pres/ClRC/Date0ue.indd

DO LOCAL ROADS MATTER? LINKING LOCAL ROADS SPENDING TO
DECENTRALIZATION IN THE DETROIT METROPOLITAN AREA

By

Annalie L. Campos

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Geography

2009

ABSTRACT

DO LOCAL ROADS MATTER? LINKING LOCAL ROADS SPENDING TO
DECENTRALIZATION IN THE DETROIT METROPOLITAN AREA
By

Annalie L. Campos

The purpose of this dissertation research is to determine the causal relationship
between local roads spending (i.e., state-allocated road spending and locally-raised road
spending) and the decentralization of people in the context of the Detroit Metropolitan
Area between 1980 and 1990. This research adopts a human ecology methodological
framework and develops a conceptual model that broadly integrates local roads spending
with proposed explanations of decentralization. The conceptual model is developed
within social geographic inquiry and is informed by three perspectives (a) ecological and
spatial assimilation, (b) neighborhood preference, and (0) place stratiﬁcation in
understanding the decentralization process. This research integrates each of these three
proposed theoretical perspectives into the human ecological framework to conceptualize
how local roads spending affected decentralization during that decade.

A standard multiple regression and spatial regression approaches were
implemented to estimate the effect of local road spending on decentralization. Three
models were estimated using three different measures of decentralization as the
dependent variable in these models, including: 1) percent change in general population
density; 2) percent change in the relative concentration of non-whites; and 3) percent

change in population density of non-whites. The ﬁrst dependent variable measured the

pattern of decentralization in the general population and the latter two dependent
variables measured the non-whites’ pattern of decentralization.

The dependent variables were each regressed against a lagged value of state-
allocated road spending at the municipality level controlling for local road stock, distance
from the central business district (i.e., downtown Detroit), household median income,
percent college educated non-whites, percent of the non-white population, and proportion
of white to total homeowners.

The result from the ﬁrst model (i.e., percent change in general population density)
indicated that state-allocated road spending had a signiﬁcant effect on decentralization (,8
= 0.56, p < 0.10). The result from the second model (i.e. the percent change in the relative
concentration of non-whites) indicated that locally-raised road spending had a signiﬁcant
effect on decentralization (,B = 1.76, p <0.001). On the other hand, the results ﬁom the
third model (i.e., percent change in population density of non-whites) did not indicate
signiﬁcance of either state-allocated or locally-raised local road spending on non-whites’
decentralization. Other variables that were important in all three models of
decentralization were percent non-whites and distance from the central business district.
These ﬁndings show that both state-allocated and locally-raised spending were important
in decentralization in the Detroit Metropolitan Area but the populations impacted varied.
While future research Should continue to investigate the effects of both spending on racial
differences in decentralization, the study recommends an emphasis on the state-allocated

road spending mechanism and its policy implications.

To my husband Danny, and daughters Katrina, Micaela, and Leanna, my parents,

and in memory of my grandmother, Romana

iv

ACKNOWLEDGEMENTS

This dissertation is an outcome of collective efforts and I would like to recognize
and thank many individuals here. First and foremost, I thank God who gave me the hope
and strength in overcoming many hurdles with completing a dissertation. I am also
greatly blessed for having a strong army of friends from the Fellowship of Christian
Internationals and my family who never wavered in their support and encouragement.

My committee had provided guidance and support in all dimensions of
dissertation development and writing. I feel very fortunate to have had Dr. Ashton
Shortridge for his objective comments and prompt attention to my dissertation needs. His
patience and kindness throughout my graduate program are very much appreciated. I am
thankful to Dr. Sue Grady for her concerns for my academic and personal well-being. Her
guidance and support improved the content and structure of my dissertation. I would also
like to thank Dr. John Schweitzer who inspired me to learn statistics and pursue research
that matters and enjoyable. I am grateful for his, and Liz’s constant optimism and
encouragement. Last but not least, I would like to thank Dr. Soji Adelaja for his
leadership in my committee, support, and guidance. I admire his wisdom and spirit of
innovation. I appreciate his generosity, both time and ideas, and attention to quality
writing and research.

I am indebted to my professors in geography as well as many others who have
contributed to my personal and professional growth. 1 am grateful to Dr. Joe Darden for
his guidance and encouragement. I beneﬁted from our discussions, which improved my

dissertation signiﬁcantly. 1 will be always thankful to Dr. Bruce Pi gozzi who has an open

door policy and willingness to listen and interact with students. I am grateful to Dr. Jay
Harman as well for his support and words of wisdom. I am thankful for Drs. Catherine
Yansa and Cynthia Simmons whose encouragement and ﬁiendship meant a lot to me as a
woman geographer. I appreciate Dr. Antoinette Winklerprins for her support during my
entire graduate program. Her advice and assistance in obtaining research and travel grants
were invaluable. I am also thankful to Dr. Groop for his insights and advice. My deep
appreciation goes to the staff in the geography department especially to Sharon Ruggles,
Judy Reginek, and Wilson Ndovie for their friendship and assistance at different phases
of my graduate program. My deep appreciation is for LuAnn Gloden at the Land Policy
Institute for her assistance, encouragement, patience, and genuine friendship.
I would not have been able to complete my graduate program without ﬁnancial support. I
am truly thankful to the Land Policy Institute, Urban Affairs Programs, Families and
Communities Together (FACT), the Julian Samora Research Institute, and the Ofﬁce of
the University Outreach and Engagement. The research experience I gained from
working in different projects in these organizations has been invaluable. The Geography
Department and the College of Social Science have also provided assistantship, research
and travel support, and a dissertation completion fellowship. I am truly grateful for these
and many other kinds of support that the Department of Geography provided especially a
work space, copying, and an ofﬁce telephone, which added to my sense of security
during those long nights.

Many staff from the Michigan Department of Transportation provided assistance
with data and interpretation of results. Ed Tricee, Mary C'umberworth, Richard Turcotte

are very much appreciated. I also beneﬁted from discussions with David Bertram of the

vi

Michigan Township Association, and Dr. Jim McConell who also provided me with
comprehensive materials about the study region.

Many friends and colleagues are truly like family. I thank Connie and Bob for being our
wonderful parents in East Lansing, They have stood by us and our children as we juggled
work, school, and parenting. I am grateful to Donna and Bob McKelvey, Rex and Vangie
Alocilja and our ﬁiends at the Bible Study group. They have encouraged, inspired,
laughed, and cried with me. I also appreciate Yohannes and Ahadu for their meaningful
friendship, support, and encouragement. Dr. Dozier Thornton deserves a special
recognition for his belief in what I can accomplish, ﬁnancial and academic support, his
wisdom, and friendship.

My geography pals, past and present, have made graduate school fun and
memorable. I especially thank Steve Aldrich, Ivan Ramirez, Eric Sandberg, Sarah
Hession, and Rita Pereira who were always supportive and encouraging. They all made
graduate school something to look back with a big smile.

My deepest gratitude is to my husband Danny for the compromises he made so I
can pursue and complete my program. I also thank my children for their understanding
and cooperation. Lastly, I remain thankful in everything, for the joys and challenges that

came along with this journey.

vii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. xi

LIST OF FIGURES .......................................................................................................... xii

CHAPTER ONE: INTRODUCTION ................................................................................. 1

Background to the Study ............................................................................................ 1

An Overview of Causes of Decentralization in US Metropolitan Areas ............. 4

The Research Context and Questions ........................................................................ 6

Roads, Government Spending, and Decentralization ........................................ 10

Highway Development and Decentralization .................................................... 12

Local Roads and Decentralization ..................................................................... l4

Purpose of the Study ................................................................................................ 15

The Research Framework ........................................................................................ 16

Signiﬁcance of the Study ......................................................................................... 18

CHAPTER TWO: RELEVANT LITERATURE ............................................................. 20

Decentralization in US Metropolitan Areas ............................................................. 21

Deﬁnition of Decentralization ........................................................................... 22

Theoretical Background ........................................................................................... 25

The Spatial Assimilation Perspective ................................................................ 27

Socioeconomic Status and Population Decentralization ........................... 29

The Place Stratiﬁcation Perspective .................................................................. 33

Social Barriers to Population Decentralization ......................................... 33

Discrimination in Housing ........................................................................ 37

The Neighborhood Preference Perspective ........................................................ 40

Preference for Homogeneous versus Integrated Neighborhood ............... 40

The Role of Government Policy in Population Decentralization ............................ 42

Government Policy on Roads and Decentralization .......................................... 42

Potential Links between Local Roads and Decentralization .............................. 45

Summary .................................................................................................................. 49

CHAPTER THREE: THE CONCEPTUAL MODEL ...................................................... 51

The Human Ecology Framework ............................................................................. 52

Sociological Interpretations of Human Ecology ................................................ 53

The Classic Ecological Models ................................................................. 54

Human Ecology within Geographic Inquiry ...................................................... 58
Understanding Patterns of Decentralization using a Human Ecology

Framework ............................................................................................................... 61

The Conceptual Model ............................................................................................. 62

Hypotheses of the Study .......................................................................................... 69

CHAPTER FOUR: THE EMPIRICAL MODEL ............................................................. 71

Model Speciﬁcation ................................................................................................. 72

viii

Speciﬁc Aims ........................................................................................................... 76

CHAPTER FIVE: DATA AND METHODS ................................................................... 78
The Study Area ........................................................................................................ 78
General Considerations for Selecting the Case Study Area ............................... 81
Regional Population Trends ...................................................................... 82

Land Use Change and Pattern of Uneven Development .......................... 84

Income Disparity ....................................................................................... 85

Racial Segregation .................................................................................... 86

Job Distribution ......................................................................................... 87

The Study Period and Communities .................................................................. 88
The Unit of Analysis and Observation ............................................................... 89
The Dependent Variable .................................................................................... 91
Population Decentralization ...................................................................... 91

The Independent Variables ................................................................................ 95
Local Roads Spending and Stock ............................................................. 95

Income and Education ............................................................................. 100
Proportion of White Homeowners .......................................................... 102

Percent Non-white Population ................................................................ 103

Distance from the Central Business District ........................................... 104

Data and Data Selection ......................................................................................... 105
Data Collection and Processing ....................................................................... 105
The Database .................................................................................................... 106
Empirical Model Estimation .................................................................................. 108
Ordinary Least Squares .................................................................................... 109
Variable Transformation ......................................................................... 1 10

Spatial Lag Model ............................................................................................ 11 1
Summary of Data and Estimation Procedure ......................................................... 113
CHAPTER SIX: RESULTS AND DISCUSSION ......................................................... 114
Descriptive Characteristics of the Study Region ................................................... 115
Population Density ........................................................................................... 115
Patterns of Distribution of Non-whites ............................................................ 120
Socioeconomic Characteristics and Homeownership ...................................... 126
Local Roads Spending and Local Roads Stock ............................................... 128
The Local Road Revenue, State-Allocated ............................................. 128

Patterns of MTF Allocation in the Detroit Metropolitan Area ............... 129

The Locally-raised Road Revenue, Local Contributions ........................ 130
Analytical Results .................................................................................................. 140
Model Estimation ............................................................................................. 140
Patterns of Decentralization: Total Population ....................................... 143
Ordinary Least Squares Regression Results, Full Speciﬁcation ....... 143

Spatial Regression Results, Full Speciﬁcation ................................. 145

Ordinary Least Squares Regression Results, Reduced Form ............ 149

Spatial Regression Results, Reduced Form ...................................... 150

Summary of Findings: Decentralization of Total Population ................. 153

ix

Patterns of Decentralization, Non-white Population .............................. 153
Relative Concentration: Ordinary Least Squares and Spatial

Regression Results, Full Speciﬁcation ............................................. 154

Relative Concentration: Ordinary Least Squares and Spatial
Regression Results, Reduced Form .................................................. 157

Non-white Population Density: Ordinary Least Squares and
Spatial Regression Results, Full Speciﬁcation ................................. 159

Non-white Population Density: Ordinary Least Squares and
Spatial Regression Results, Reduced Form ...................................... 161
Summary of Findings: Patterns of Non-White Decentralization ............ 164
Discussion .............................................................................................................. 165
Effects of Local Roads Spending ..................................................................... 165
Effects of Socioeconomic Factors ................................................................... 166
Effects of Racial Composition ......................................................................... 167
Effects of Spatial or Locational Attributes ...................................................... 171
CHAPTER SEVEN: CONCLUSIONS .......................................................................... 174
Summary of Key Findings ..................................................................................... 175
Conclusions and Implications ................................................................................ 177
Conclusions ...................................................................................................... 1 83
Contributions .................................................................................................... l 87
Theoritical Contributions ........................................................................ 187
Methodological Contributions ................................................................ 190
Policy and Broader Impacts .................................................................... 190
Limitations and Future Work ................................................................................. 192
Framework ....................................................................................................... 192
Data and Methods ............................................................................................ 195
APPENDICES ................................................................................................................ 200

Appendix A. List of Study Communities (cities and villages) in the Detroit

Metropolitan Area .................................................................................................. 201
Appendix B. Dependent and Independent Variables: Untransforrned ......... j .......... 203
Appendix C. Bivariate Correlations ....................................................................... 205
REFERENCES ............................................................................................................... 207

Table 5.1.
Table 5.2.
Table 6.1.

Table 6.2.

Table 6.3.

Table 6.4.

Table 6.5.

Table 6.6.

Table 6.7.

Table 7.1.

LIST OF TABLES

Changes in the Michigan Transportation Fund Distribution Formula ............. 98
Variables and their Sources ........................................................................... 107
Descriptive Characteristics of the Study Community .................................... 138
Coefﬁcients of Standard and Spatial Regression (Full Model)

Percent Change in Population Density: Detroit MSA, 1980-1990 ................ 148
Coefﬁcients of Standard and Spatial Regression (Reduced Model),

Percent Change in Population Density: Detroit MSA, 1980-1990 ................ 152
Coefficients of Standard and Spatial Regression (Full Model)

Percent Change in Non-white Concentration, Detroit MSA, 1980-1990 ...... 156

Coefﬁcients of the Standard and Spatial Regression (Reduced Model),
Percent Change in Non-white Concentration, Detroit MSA, 1980-1990 ...... 158

Coefﬁcients of the Standard and Spatial Regression (Full Model),
Percent Change in Non-white Population Density Detroit MSA,
1980-1990 ...................................................................................................... 160

Coefﬁcients of the Standard and Spatial Regression (Reduced Model),
Percent Change in Non-white Population Density
Detroit MSA, 1980-1990 ............................................................................... 163

Comparison of Signiﬁcant Explanations of Decentralization
Between the Total and Non-white Population ............................................... 177

xi

LIST OF FIGURES
(Images in this dissertation are presented in color)

Figure 2.1. Percent of Population Living in Central Cities and Suburbs in US.

Metropolitan Areas, 1910-2000 ....................................................................... 22
Figure 3.1. A Conceptualized Relationship between Local Roads and
Population Decentralization ............................................................................ 64
Figure 5.1. The Study Area ................................................................................................ 80
Figure 5.2. Detroit Metropolitan Area Population Trend, 1900-2008 ............................... 83
Figure 6.1. Percent Population Density Change, Detroit MSA
1980-1990 ..................................................................................................... 117
Figure 6.2. Population Change by Travel Time: 1972-2002 .......................................... 119
Figure 6.3. Relative Concentration of Non-whites, Detroit MSA, 1980 ........................ 122
Figure 6.4. Relative Concentration of Non-whites, Detroit MSA, 1990 ........................ 123
Figure 6.5. Percent Change in Relative Concentration of Non-whites,
Detroit MSA, 1980-1990 .............................................................................. 125
Figure 6.6. Household Median Income by distance from Detroit CBD, 1980 ................ 127

Figure 6.7. MTF Allocation: Proportion of the City of Detroit to the
Total Metropolitan Region, 1952-2002 ......................................................... 131

Figure 6.8. Per Person State-allocated Local Road Revenue ($), Detroit MSA, 1977....132

Figure 6.9. Per Person State-allocated Local Road Revenue ($), Detroit MSA, 1987....133

Figure 6.10. Per Person Locally-raised Road Revenue ($), Detroit MSA, 1977 ............. 134
Figure 6.11. Per Person Locally-raised Road Revenue ($), Detroit MSA, 1987 ............. 135
Figure 6.12. Local Road Density, Detroit MSA, 1977 .................................................... 136
Figure 6.13. Local Road Density, Detroit MSA, 1987 .................................................... 137
Figul‘e 6.14. Percent Change in Population Density Model Residuals ........................... 144

xii

CHAPTER ONE: INTRODUCTION
Background to the Study

Metropolitan areas, which are deﬁned as “geographic areas with large population
nucleus, together with adjacent communities that exhibit a high degree of social and
economic integration with that nucleus” (U. S. Bureau of the Census, 2000), are dynamic
human-dominated landscapes and their transformation remain an enduring subject in
settlement formation, urbanization, and decentralization processes. In the US, the
transformation of metropolitan areas has been characterized by profound shifts in

demographic trends and distinct changes in the spatial distribution of people. These shifts

have proceeded in two distinct spatial patterns.

During the ﬁrst half of the twentieth century, U. S. metropolitan areas
developed into highly centralized regions primarily due to the industrialization boom.
Ci ti es in the North and Midwest regions, in particular, became the dominant destination
pl aces for people from international and domestic non-metropolitan origins, as well as for
many Southern blacks who were attracted to better life and economic opportunities that
the North could offer, and away from realities of racial persecution endemic in the South

(Teaford, 1993). As a result, major cities such as New York, Chicago, and Detroit

experi enced unprecedented population growth.
By the 19508, many of these cities experienced their highest centralization

point, thus the transformation of these cities into social, cultural, political, and economic
pO‘Nel‘houses. Over time, however, the dominance of central cities in these regions waned

althollgh cities like Chicago and New York have experienced some revitalization while

cities like Detroit remained ﬁscally, socially, and ecologically challenged and immensely
distressed.

In the second half of the twentieth century, the explosive growth in suburban
areas and/or the simultaneous and persistent out-migration of central city residents for
suburban locations ushered a drastic transformation of metropolitan areas particularly in
the North and Midwestern regions (Hobbs & Stoops, 2002). Hobbs and Stoops (2002)
indicate drastic shifts in population distribution when they found that in 1950, the average
population density in a given central city was 7,517 persons per square mile. Half a
century later, in 2000, a central city had an average population density of 2,716 persons
per square mile (Hobbs & Stoops, 2002; p. 38). In contrast, communities outside of
central cities during the same time period experienced population gains, indicating an
average population density increase from 175 to 208 persons per square mile. This
pattern of reversal in population growth in central cities and suburban areas has been a
well observed phenomenon in many Northeast and Midwestern cities in metropolitan
areas in the US.

The reversal in population trends has become increasingly understood in such
terms as “decentralization” (Mieszkowski & Smith, 1991; Glaeser & Khan, 2001),
“suburbanization” (Mieszkowski & Mills, 1993; Jordan et al., 1998), “counter
urbanization” (Berry, 1980), dispersion (Boume, 1989), and the more general and catch-
all phrase that many urban scholars label as “urban sprawl” (Squires, 2002; Duany et al.,
2000; Breugrnann, 2005; Lopez & Hynes, 2003).

This dissertation research uses the term decentralization to describe the geography

of urban transformation characterized by signiﬁcant population growth in suburban areas,

while the central city experiences stagnation, slow growth, and in many instances, a
substantial decline in population and economic activities (Hobbs & Stoops, 2002; Berube
& Fonnan, 2002). Within metropolitan areas, decentralization has attracted much
attention because of the fundamental change in the way people live, work, and play.
Decentralization has prompted a widespread concern among urban scholars, interest
groups, policymakers and their constituents and has stimulated extensive research on
causation.

At the metropolitan level, decentralization presents two primary problems that are
inherently geographical. On one hand, decentralization results in central city decline and
negative social and economic consequences. This means that for central cities, the
decentralization of people is associated with capital and job ﬂight, and ﬁscal challenges
due to a diminished competitiveness of cities for attracting and retaining residents and
businesses. Over time, the decline in central cities and its effects are disproportionately
spread across different population groups, with African Americans and other ethnic
minorities experiencing the brunt of poor public service provision, congestion, isolation,
and a limited accessibility to jobs and other opportunities (Glaeser et al., 2004;

J argowsky, 2002). On the other hand, decentralization affects suburban communities
particularly since it imposes ecological challenges associated with the unplanned pattern
of development in outlying areas (Squires, 2002; Power, 2001; Downs, 1999).

Accordingly, beyond the central city, decentralization manifests in the following
problems: 1) a signiﬁcant loss of farmland and open space due to land development in

support of housing and infrastructure needs (Lopez et al., 1988); 2) landscape

fTi’lgl'nentation due to uncoordinated land use planning (Wilson et al., 2003), and; 3)

environmental pollution due to excessive intrametropolitan driving (Handy, 2006;
Johnson, 2001 ), among other negative consequences. Hence, residents in ﬁinge and
outlying communities are also adversely affected but in forms and intensities that are
different from central city residents.

As the current pattern of decentralization threatens the ﬁscal, social, economic,
and physical health of communities, the urgency to re-evaluate and recast public policies
has intensiﬁed as has the quest for explanations that underlie it. There is widespread
recognition that until the underlying factors of decentralization are well understood,
policy responses will remain off target, if not signiﬁcantly ineffective for reversing the
decentralization of people, the sprawling pattern of grth in metropolitan areas, and
negative consequences that are associated with these processes.

This study will contribute to the understanding of some of the causes of
decentralization that are currently not well understood. Speciﬁcally, the study will add to
an understanding of the effects of government policy on the geography of opportunities
for quality of life in general, and patterns of decentralization of people in particular.
Furthermore, the study will examine the inﬂuence of local roads spending on the pattern

of decentralization in metropolitan areas.

An Overview of Causes of Decentralization in U. S. Metropolitan Areas

Previous geographic research provides an understanding of decentralization of
p€0ple in metropolitan areas as a process shaped by different factors and circumstances
Operating at different geographic scales over time. These factors are inextricably linked I

and are not easily disentangled. At the macro-scale, the out-migration of people and jobs

from central cities reﬂect: 1) demographic transition including decreasing mortality
followed by decreasing birth rates; 2) advances in communication and technology;

3) deindustrialization manifested in shifts from manufacturing to service based industry;
4) government policies on housing and highway development; 5) the convergence of
military and aviation research along with advances in the air conditioning and
refrigeration technologies in the South and West that further stimulated the reversal of
migration flow from the “frostbelt” to “sunbelt” regions; and 6) the emergence of global
cities that facilitated increasing diversity in cities and exchange of goods and services
while these cities maintain command and global control over ﬁnancial markets (Sassen,
2002; Bruegmann, 2005; Duany et al., 2000; Ebner, 1985; Fisher & Mitchelson, 1981;
Berry, 1980; Squires, 2002). These macro 1eve1 causes reinforced metropolitan and local
circumstances prompting changes in the spatial distribution of people and the overall
metropolitan landscape.

At the metropolitan scale, causes of decentralization range from individual wealth
or ﬁnancial status and other socio-economic factors that allow increasing mobility for
some people (Clark, 2007, 2009; Freeman, 2008; Clark & Ware, 1997; Mieszkowski &
Mills, 1993), to the variation of local community amenities that entice people to relocate
in such communities (Bayoh et al., 2006; Koles & Muench, 2002; Clark et al., 2002), to
race relations (Alba & Logan, 1993; Charles, 2003; Darden etal., 1987; Fischer, 2008;
Iceland & Wilkes, 2006; J argowsky, 2002; Massey & Denton, 1993; Zubrinsky & Bobo,
1996), to changing taste and preferences for rural lifestyle (Belden et al., 1998), and to

fCl’atures of the natural and physical environment that either restrict or accommodate

exPansive development patterns (Burchﬁeld et al., 2005). Another body of knowledge

holds that the reality of local political fragmentation reinforced by uncoordinated local
land use and growth control mechanisms has a signiﬁcant contribution in shaping the
decentralization in metropolitan areas (Byun et al., 2005; Ulfarsson & Carruthers, 2006).
Overall, these mutually reinforcing metropolitan level and broader processes have
signiﬁcant effects in the way business, residential location choices, and recreation are

realized.

The Research Context and Questions

While evidence suggests that the decentralization of people in US. metropolitan
areas has been occurring since the ﬁrst quarter of the century but proceeded more rapidly
after the Second World War (Otterstrom, 2003; Hobbs & Stoops, 2002; Mieszkowski &
Mills, 1993; Ebner, 1985), the pattern of decentralization of people by race has been
distinctively uneven across groups. Minorities, deﬁned as “all people who are Hispanic,
and are races other than white including those who indicated White alone in Census
2000” (US. Bureau of the Census, 2000), have less participation in the decentralization
process and the precise causes are far from settled (Clark, 2009, 2007; Fischer, 2008;
Logan, 2003; Darden, 1990).

In the last few decades, however, signiﬁcant increases in suburbanization rates for
blacks or African Americans and corresponding decreases in levels of their segregation
from the majority whites have been observed in many metropolitan areas (Clark 2009;
Fischer, 2008; Freeman, 2008). While it appears that there is an improvement or progress

in the quest for residential integration particularly for blacks in many metropolitan areas

in the country, interpretation of this trend must be pursued with caution.

A closer scrutiny of the trend suggests that the increasing suburbanization of
Aﬁican Americans does not readily equate to spatial assimilation and residential
integration with majority whites. Blacks continue to experience the highest level of
segregation from whites relative to their Asian, Hispanic, and other ethnic group
counterparts (Alba & Logan, 1993; Chung & Brown, 2007; Fischer, 2008; Johnston et al.,
2007). Essentially, there has been a sustained separation between the minority population
groups from whites, and some cities like Chicago, Cleveland, Detroit, Milwaukee,
Newark, and Philadelphia remain “hypersegregated”, which is a condition that
characterizes higher levels of spatial separation of blacks from whites based on all ﬁve
dimensions of segregation that include unevenness, isolation, clustering, concentration,
and centralization (Massey & Denton, 1993; Wilkes & Iceland, 2004). This incidence of
“hypersegregation” reﬂects a group’s lack of propensity and capability for social mobility
and spatial mobility, which may explain the differential level of suburbanization across
racial groups (Massey & Denton, 1993).

The different patterns of decentralization and the sociospatial separation by race
are signiﬁcant social issues that continue to stimulate discussions among policymakers
and urban scholars from many disciplines. These discussions reﬂect deep concerns for
the sustained differential locational outcomes by race in metropolitan areas. Where there
is a high disparity of levels of suburbanization and “hypersegregation” between racial
groups, the segregated group is more likely to experience isolation (Wilson, 1987;
Massey & Denton, 1993), a decline in access to employment and other opportunities

(Darden et al., 1987), concentrated poverty (Wilson, 1987), higher incidence of teenage

Pregnancy (Hogan & Kitagawa, 1985), poor school performance (Charles et al., 2004)

and poor maternal and infant health outcomes (Grady, 2006). Because of these negative
consequences, there has been an enduring interest in identifying causes of
decentralization. More speciﬁcally, there has been an enduring interest on the causes of
differential rate of suburbanization of non-whites and their locational outcomes.

From a social geographer’s point of view, the differential rate of decentralization
across racial groups have been viewed on the basis of the following perspectives: 1)
ecological or spatial assimilation; 2) neighborhood preference; and 3) place stratiﬁcation.
These perspectives provide relevant theoretical foundations for much of the analyses
regarding differential patterns of suburbanization and locational outcomes of minorities.

According to the ecological perspective, the differential rate of decentralization
and locational outcome is a result of competition, invasion, succession, and adaptation
processes (Park et al., 1925; Alba & Logan, 1993; Darden, 1985, 2000; Lawrence, 2003).
In empirical tests, the ecological perspective identiﬁes socio-economic status measured in
income, education, and occupation as primary determinants of patterns of
decentralization by race. Further, results of these tests have shown that income,
education, and occupation are signiﬁcant explanations for the likelihood that an

. individual or a group will assimilate with the majority population group, the whites, as

the spatial assimilation theory asserts (Park et al., 1925; Massey & Denton, 1985).
Therefore, individuals and population groups with better socio-economic characteristics
are in a better position for relocating in suburban communities, which are places with
better residential and locational endowments (Clark, 2009; Clark & Ware, 1997).

Alternatively, the neighborhood preference perspective proposes that the

differential rate of decentralization by race is an outcome of sheer individual choice for

certain demographic composition and amenities in the “destination community” (Clark,
2009; Bouma-Doff, 2007; Freeman, 2000; Farley et al., 1997). The neighborhood
preference perspective ties in with the place stratiﬁcation perspective, which asserts that
causes of differential decentralization by race are due to structural forces that induce the
relocation of certain population groups in suburban communities while others, mostly
minorities or people of color, are left behind in central cities. The main idea for the place
stratiﬁcation reasoning is that prejudice and discrimination cause people to distance
themselves from certain population groups, particularly blacks or Aﬁican Americans and
other ethnic minorities through the process of “white ﬂight” and other discriminatory
behavior (Gotham, 2002; Crowder, 2000; Alba & Logan, 1993; Charles, 2003).

These three theoretical perspectives have been used extensively to analyze
residential patterns and decentralization processes. However, prior applications and
empirical studies of these tend to overlook impacts of public policies on patterns of
decentralization. Speciﬁcally, the effects of policy on transportation development at the
local level have had inadequate attention. It is possible that the lack of focus of policy
impacts regarding local transportation may have obscured effects of the pattern of
distribution of funds for the development of local infrastructures and amenities that serve
local needs. Essentially, what is not captured in previous studies using one of these three
perspectives is the potential effect of patterns of distribution of public funds across
metropolitan landscapes and how these may have exacerbated differential levels of

decentralization across population groups.

To address the gap, a human ecology methodological framework is adopted in

this dissertation research to understand the relationship between local roads and patterns

of decentralization. The human ecology framework enables analyzing the relationship
between roads and decentralization within a broad and holistic theoretical base.
Speciﬁcally, this dissertation research focuses on local or non-highway roads and their
possible inﬂuence in the decentralization of people by race. The Detroit Metropolitan
Area in Michigan is the case study area.

The research addresses three questions: 1) Do local roads matter in the
decentralization of people? 2) Does the pattern of distribution of local roads investment
facilitate decentralization? and 3) Does the pattern of distribution of local roads spending
explain patterns of decentralization of non-whites in metropolitan areas?

This research attempts to understand the relationship between local roads and
patterns of decentralization, and the possible contribution of govermnent policy to this

relationship.

Roads, Government Spending, and Decentralization
Roads and road building had become vital components in the evolution and
transformation of human settlements even before the introduction of a formal road system
and management strategies by the Roman empire in the pre-modern civilization (Bund,
1920). This transfonnative capacity of roads in a landscape is tied to the fact that roads
are basic infrastructure for small villages and large and complex conurbations alike.
Roads provide economic, social, and military or security beneﬁts, and shape the internal
structure of a place or a region.
In the twentieth century, roads and road building in the United States have be-
COme of paramount signiﬁcance beyond these basic beneﬁts. Roads and road building

have been linked to the marked transformation of the social and physical landscapes

10

across the country. For instance, the social transformation in the U. S. particularly the
birth of the civil rights movement has been linked to the events on the road and Rosa
Parks (Hanson and Giuliano, 2004; p.3). In the same vein, roads and road building
facilitated the spatial transformation of the metropolitan landscape, which manifests in
the way people and jobs are distributed. Conversely, people have engaged in the process
of landscape transformation through road building, which shapes the internal structure of
any given place or region.

A signiﬁcant part of many discussions about the metropolitan landscape
transformation has focused on the effects of government policies on roads and road
building. This is not surprising given the fact that government policies are often invoked
as major causes of decentralization in US. metropolitan areas (Hartgen, 2003; Voss &
Chi, 2006; Baum-Snow, 2007). Indeed, research in the last few decades indicates the
signiﬁcance of the interstate highway development in the uneven growth across
metropolitan areas, making some communities more attractive for residential and
industrial relocation than others (Handy, 2005; Downs, 1999; Garrison et al., 1959; Rii,
1983). Thus, the signiﬁcant impact of roads and road building on decentralization, and
the policies that underlie these cannot be ignored. Roads and road-building have been
subjects of intense discussions since these are linked to many positive as well as negative
social, economic, and ecological outcomes.

First, roads form a blueprint for local and regional transportation system, the

network of waterlines, and other linear infrastructures. The capacity of these basic
infrastructures determines how much growth can be accommodated in a given time

peri Od. Second, roads enable local housing development, and social and economic

ll

activities to function. The quality of road services and related infrastructure can either
serve as “pull” or “push” factors for residential relocation, which in turn affects local
growth and land use change. Third, roads and highways enable infra—regional exchange
with broader markets. The quality of roads services affects the nature of inter-
metropolitan and intra-metropolitan mobility and accessibility in a given locale or region,
which in turn affects industry location decisions that can trigger shifts in the distribution
of people and jobs, and land use change (Hanson & Giuliano, 2004; Rodrigue, 2006).
Thus, roads and road building are linked to the physical transformation and
decentralization in metropolitan areas. These have also contributed to social
transformation as the Rosa Parks experience has demonstrated. Essentially, roads and
road building do not just have a paramount inﬂuence over the form and character of
human settlements, but they shape the social landscape as well (Hanson & Giuliano,
2004; de Boer, 1986). Roads are intricately woven into the social fabric of community
life. Their production and management entail complex processes, decisions, and policies.
Oﬁentimes, these processes, decisions, and policies have unintended consequences that
may provide beneﬁts to some individuals and population groups but costs and injustice to

others.

High way Development and Decentralization
As research shows, strong evidence for the hi ghway-decentralization connection
exists. The highway development is recognized as one of the major facilitators of
decentralization in metropolitan areas (Hartgen, 2003; Voss & Chi, 2006; Baum-Snow,

2007). Highways facilitated increases in spatial interactions between regions, linked

12

settlements across the country (Berry & Garrison, 1958; Berry & Horton, 1970), and the
dispersion of economic (retail) activities (Brueckner, 2005). Additionally, highways
generated spillover effects inducing the displacement of industries in one location and
shifting them to another (Chandra & Thompson, 2000). For the most part, these
displacements meant the deconcentration of jobs and people in central cities (Handy,
2005; Downs, 1999; Davis, 1996; Garrison et al., 1959; Rii, 1983), subsequently
transforming the form, structure, and function of cities.

Despite the strength of the evidence supporting the highways-decentralization
connection, there remains much to be understood about roads and the possible facilitating
effect that these may have on population decentralization in metropolitan areas. Some
have cast doubt on the role of highways on decentralization (Gramlich, 1994), the
direction of the causality (Kain, 1970; Zhang, 2001), and the possibility that factors other
than the highway development have contributed to the decentralization in metropolitan
areas.

The ambiguity of the hi ghways-decentralization relationship raises questions on
whether highways have directly contributed to decentralization or whether these merely
serve communities that would have otherwise developed due to factors other than
highway development. Such ambiguity requires a more comprehensive analysis of
decentralization that accounts for other factors with possible synergistic and

complementary effects. This dissertation research will contribute to this debate.

13

Local Roads and Decentralization

The inﬂuence of local roads or non-highway roads on the decentralization of
people is plausible and deserves further investigation. This is particularly so since similar
to the functions that highways provide, local roads are relevant in the socio-spatial
structure and development of local communities. Local roads enable connectivity
between major roads or highways and local suburban communities, provide base
infrastructure, and serve as an amenity that affects quality of life (Black, 2003; Hanson &
Giuliano, 2004). Local roads also affect health and is linked to obesity (Saelens et al.,
2003), and sense of safety (Darcin & Darcin, 2006; Marans & Sheehan, 2003). As such,
local roads and their development have been given greater consideration than the
development or delivery of other public goods and services like local parks, libraries, and
police protection, among others. It is more likely that communities may exist without
parks, police protection and all other types of goods and amenities but may not equally do
without local roads that provide the basic infrastructure. Thus, the economic and socio-
spatial relevance of local roads present strong rationales for examining the effects of local
roads on the decentralization of people in a metropolitan area.

While some urban scholars provide insights into the relationship between local
roads and decentralization, only few have undertaken studies on the direct link between
them. And these studies do not examine the impacts of local roads on racial differences in
decentralization. There is less information on how local roads may directly spur the
uneven decentralization of people in general and by racial groups in particular.

Studies of local roads with reference to decentralization have primarily focused

on costs (Burchell et al., 2002; Carruthers & Ulfarsson, 2003; Holcombe & Williams,

14

2008; Ladd, 1998) and efﬁciency and equity consideration of road investments
(Mikelbank & Jackson, 2000; Siggerud, 2004), travel behavior (Cervero & Kockelman,
1997; Cervero 2003), health and quality of life (Vojnovic et al., 2006; Saelens et al.,
2003), sense of community and other neighborhood social engagements (Toit et al.,
2007), and economic and job creation (Guild, 2000; Kemmerling & Stephan, 2002).
These studies indirectly link local roads to decentralization and the dearth of existing
studies that directly link them provides mixed ﬁndings. Furthermore, none of these
studies examined the impact of local roads on the differential rate of suburbanization
across population by race. For these reasons, there is a substantial room for improvement
in the existing literature on the relationship between local roads and the decentralization

of people in metropolitan areas.

Purpose of the Study

The purpose of this dissertation research is to determine the causal relationship
between local roads spending and the decentralization process between 1980 and 1990, a
decade of continued population decline in central cities particularly in the Northeast and
Midwest regions in the US. (Hobbs & Stoops, 2002; Otterstrom, 2003; Berube &
Forman, 2002) and is therefore, a relevant time frame for analyzing patterns of
decentralization and its causes. This research speciﬁcally examines the effects of local
road stock and road spending on decentralization and the differential rate of
decentralization for minorities in the Detroit Metropolitan region. The research
determines the relationship between local roads and patterns of decentralization with

proposed environment, population, and behavioral explanatory factors deﬁned within the

15

human ecology conceptual model of decentralization. Ultimately, the study seeks to
contribute to the on-going discussion of how government policies, through local roads
spending, contribute to the decentralization process in the Detroit Metropolitan region
and how that decentralization shaped the spatial pattern of sprawl between 1980 and
1990. The ﬁndings from this research shed insights into the role that government policies
play in sustaining the disparity and spatial inequities across metropolitan areas.

Speciﬁcally, the dissertation: l) develops a theoretical model of human ecology
that integrates population, behavior, and built-environment input variables into the
analysis of the causal relationship between local roads and decentralization; 2)
determines the Spatial pattern of the state’s spending on local road development at the
community level; 3) characterizes the spatial pattern of population density change and
concentration of non-whites within these communities; and 4) determines the causal link
between local roads and decentralization, by relating local roads and road stock to
population density change and concentration of non-whites as measures of

decentralization using the Detroit Metropolitan region as the case study region.

The Research Framework
For this dissertation research, the human ecology framework enables an
understanding of decentralization as an outcome of dynamic interactions and
relationships among population characteristics, behavior, and environmental factors. The
framework conceptually situates an analysis of decentralization within the context of
dynamic relationships among factors such as socioeconomic characteristics of the

population using income and education variables, behavior using preferences for

16

residential location based on demographic characteristics and racial discrimination
variables, and features of the built environment such as roads stock, road spending, and
locational factors such as distance from the central business district (CBD).

The relationship between government spending on local roads and
decentralization is understood in terms of effects of local road spending due to
improvements in intra—metropolitan accessibility. The relationship between
socioeconomic characteristics and decentralization is viewed as expression of people’s
capacity for social mobility, which translates into spatial mobility. The relationship
between behavior and decentralization is understood in terms of individual and collective
preference for local communities with particular demographic characteristics. As
relationships and interactions change among these factors, the built environment and land
use pattern change, and subsequently, the spatial distribution of people across the
metropolitan landscape.

Within the context of the human ecology framework, and with the goal of
making contributions to the body of knowledge in the decentralization process, the
dissertation focuses on local roads and government spending. More speciﬁcally, it seeks
to determine how variation of these is linked to the changes in population density and
concentration of non-whites at the municipality level across the study region, while
accounting for income, education, home ownership, demographic composition, and
locational factors.

These proposed explanatory factors reﬂect three theoretical perspectives of
residential patterns including spatial assimilation, neighborhood preference, and place

stratiﬁcation perspectives that are integrated into the human ecology framework of

17

decentralization adopted in this dissertation research. In sum, the human ecology adopted
in this dissertation research orients an analysis of decentralization that recognizes the
complex, multi dimensional, and dynamic nature of the decentralization phenomenon.
The human ecology offers an inclusive framework for studying the possible effects of

government policies on local roads ﬁom multiple but complementary perspectives.

Significance of the Study

This dissertation research contributes to the theoretical, methodological, and
policy discussions on the decentralization process in metropolitan areas. First, the study’s
focus on local roads extends previous conceptualization of decentralization as one that is
explained, to a large extent, by the accessibility beneﬁts derived through the major
highways development. It addresses questions about the role of government policy at the
local level on the decentralization of people and the racial differences in pattern of
decentralization.

The goal of the research is to determine the effects of local roads which may shed
insights into the sustained segregation of minorities from the white majority population
group. Second, using the human ecology as a methodological approach allows an
integrated analysis of the relationship between local roads and the decentralization of
people following a broad theoretical base that integrates government policies with

pOpulation, behavior and environmental input variables. Third, empirical results have
policy implications for the role of the government and possible local spending oversight
on decentralization and its negative consequences. Findings would provide a strong

justi ﬁcation for an objective policy change regarding the distribution of local roads funds.

18

The rest of this dissertation is organized as follows. Chapter Two provides a
review of prior studies of the decentralization phenomenon and factors and processes
shaping it. It also presents the theoretical perspective adopted in this study. Chapter Three
provides the conceptual framework followed by Chapter Four, which presents the
empirical model of the relationship between local roads and the decentralization process.
Chapter Five presents the data and methods. Chapter Six presents the results of the
statistical analysis followed by discussion of results. Chapter Seven provides the

summary and conclusions.

19

CHAPTER TWO: RELEVANT LITERATURE

Recognizing that the population decentralization in metropolitan areas is linked to
many positive and negative consequences, urban scholars and policymakers have
increasingly sought for concrete explanations that facilitate such transformation.
Research has been extensive and ﬁndings based on analyses of decentralization from
different vantage points suggest that causes of decentralization reﬂect a plethora of
social, economic, demographic, political, and ecological variables. The goal of this
chapter is to provide some of the major theoretical and empirical ﬁndings on
decentralization and analyzes possible links between local roads and decentralization in
metropolitan areas.

This chapter is organized into four sections. The ﬁrst section provides a brief
description of the shifts in population trend in US metropolitan areas during the twentieth
century. The idea is to present these shifts in population trend to establish the relevance
for studying decentralization and its causes. This section also articulates common terms
and concepts that are used to characterize the pattern of metropolitan growth. The second
section discusses theoretical explanations of decentralization informed by the residential
mobility and segregation literatures. This is followed by a review of the empirical
evidence of causes of decentralization. The third section presents plausible links between
decentralization and local roads. The chapter ends with a summary of causes of

decentralization and identiﬁes themes and variables for which an empirical model of the
relationship between decentralization and local roads can be developed and tested in the

context of the Detroit Metropolitan Area.

20

Decentralization in US. Metropolitan Areas

Throughout the twentieth century, U.S. metropolitan areas experienced a steady
grth though this pattern of growth proceeded in two distinct spatial patterns. During
the ﬁrst half of the twentieth century, cities and suburban communities experienced
population growth, with cities growing at an unprecedented rate and concentration of
economic activities. In the second half of the twentieth century, a reversal of the pattern
of growth occurred with suburban areas gaining rapid population growth while central
cities were experiencing a simultaneous stagnation, slow growth, or a signiﬁcant
population decline in some regions particularly in the North and Midwest regions in the
US. (Hobbs & Stoops, 2002; Kim, 2007).

As Figure 2.1 shows, U.S. metropolitan areas experienced a steady population
increase between 1910 and 2000 with central cities indicating their highest centralization
point in the 19503. It shows that on average, 32.8 percent of the total metropolitan
population lived in central cities in the 19503. By the 19903, many metropolitan areas in
the US were decentralized with old industrial cities in the Northeast and Midwest regions
experiencing a “hollowing out” particularly in neighborhoods closest to a central business
district, while few cities experienced renaissance and revitalization of different
proportions (Berube & Forman, 2002). At the turn of the twenty ﬁrst century, 30.3
percent of the total metropolitan population lived in central cities while 50 percent lived
in suburban areas. This translates into an aggregate population loss for central cities by

2.5 percent (Hobbs & Stoops, 2002), despite the revival of some major cities in the
Northeast and Midwest regions. These trends and spatial distribution of people in

metropolitan areas during the second half of the twentieth century stimulated extensive

21

research on underlying causes and consequences. These have also elicited a number of

terms or descriptors for characterizing the pattern of growth.

 

 

 

 

 

 

 

 

 

 

100‘ . ” ,
iCJSuburb

-° ICentralcity
a ..
=3 80 F__ __ T7
E —-w
8 T a N ‘9.
i a E i i 3
— m
8
I-
“II
o
o\°

 

 

 

           

 

 

I 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000

I Year

Figure 2.1. Percent of Population Living in Central Cities and Suburbs in
US. Metropolitan Areas, 1910-2000

Source: Hobbs and Stoops, 2002. Demographic Trends in the Twentieth Century,
US. Bureau of the Census Special Report, CENR No. 4.
Definition of Decentralization
The transformation in US. metropolitan areas particularly during the second half
of the twentieth century has been described in terms such as decentralization,
suburbanization, dispersion, counter urbanization, deconcentration, and urban Sprawl.
Though these terms may have been deﬁned and adopted in different ways, they
emphasize a pattern of diffusion of people, jobs, and other phenomena in an uneven and

unprecedented way. Speciﬁcally, these terms characterize an uneven pattern of urban

22

development with a speciﬁc reference to population growth and development in suburban
areas while the central city experiences stagnation, slow growth, and in most cases,
decline.

Hoyt (1941) conveys his description of the US. urban transformation using the
term “decentralization.” He notes that decentralization proceeds in two reinforcing ways.
The city breaks up into small towns, and industries and people increasingly relocate in
the peripheries. Boume (1989), on the other hand, uses the term “dispersion” to
characterize a decentralized form of urban development. Closely related to Hoyt’s and
Bourne’s description, Berry (1980) uses the term “counter urbanization” to emphasize
population deconcentration in urbanized areas and the rise of smaller, low-density, and
homogeneous communities as determining characteristics of the pattern of urban
development. Berry’s deﬁnition calls our attention to a depiction of urban development
that not only discusses the population shifts and urban-suburban interactions but also
decomposes the pattern of development that results from these interactions by concrete
measures of spatial distribution and demographic composition.

Alternatively, many urban scholars have used the term “suburbanization” to
characterize a process of increasing relocation of people and economic activities from the
core to the peripheries. In essence, suburbanization is a term that has been used
interchangeably with decentralization, dispersion, deconcentration, and urban sprawl
(Fisher & Mitchelson 1981; Mieszkowski and Mills 1993; Jordan et a1. 1998; Carlino,
2000).

Over the last few decades, scholarly and popular articles have increasingly

described the pattern of development in US. metropolitan areas using the term “urban

23

sprawl” (Bruegmann, 2005; Squires, 2002; Lopez & Hynes, 2003). Urban sprawl is a
“catch all” phrase that characterizes the range of concepts given above. While many
urban scholars and policymakers are in agreement of the many negative consequences
associated with the urban sprawl phenomenon, a general consensus about its deﬁnition
and measurement is far from settled.

For example, Lopez & Hynes (2003) identify nine attributes that are often used to
characterize sprawl. These include: 1) low density development; 2) separation of land
uses; 3) leap-frog development; 4) strip retail development; 5) automobile-dependent
development; 6) development at the periphery; 7) employment decentralization; 8) loss of
peri-urban, rural agriculture, open space, and; 9) fragmented governmental responsibility
and oversight. These authors, however, argue that a measure of urban sprawl based on
population density captures the core feature of the phenomenon. They developed a sprawl
measure based on the concentration and density dimensions of urban sprawl.

Galster et a1. (2001) emphasize land use dynamics as a core underlying dimension
of urban sprawl. They identify eight distinct dimensions of land use pattern and measure
of urban sprawl including: 1) density; 2) continuity; 3) concentration; 4) clustering; 5)
centrality; 6) nuclearity; 7) mixed uses, and 8) proximity. Similarly, Ewing et a1. (2003)
measure urban sprawl based on four land use dynamics including: 1) density, 2) land use
mix, 3) degree of centering, and 4) street accessibility. These different dimensions
accentuate urban sprawl as an outcome of land use dynamic, and stress highlight urban
sprawl as a land use issue.

Given above, the different dimensions and measures clearly elucidate the

complexity of the sprawl phenomenon and its underlying causes. As the review of Lopez

24

& Hynes show, urban sprawl is a pattern of development that reﬂects aspects beyond
land use dynamics to include spatial interactions through shifts in the ﬂow and diffusion
of people and jobs between central city and suburban areas. Essentially, urban sprawl
represents social, economic, spatial, and ecological dimensions of human and
environment relations.

For the purposes of this dissertation research, the term decentralization is used to
describe the uneven pattern of development in metropolitan areas. Speciﬁcally,
decentralization is conceived of as a process involving population gains for communities
outside of a central city between two time periods. Furthermore, the research treats
decentralization as a pattern of suburbanization of people by racial groups. Thus, patterns
of decentralization of the general and non-white population in metropolitan areas are
analyzed. To operationalize decentralization, three indicators are used. These include: 1)
percent change in population density, 2) percent change in relative concentration of non-
white population; and 3) percent change in non-white population density. Furthermore,
the terms decentralization and suburbanization is interchangeably used for the rest of the

dissertation.

Theoretical Background
Decentralization has characterized many metropolitan areas in the US. in the past
several decades and is understood as a signiﬁcant out-migration of people from central
cities to outlying areas. One prominent subject is the differential participation of various

population groups in the decentralization process. Research has established that there is a

25

sustained difference in levels of suburbanization between majority whites and non-whites
(Fischer, 2003; Iceland et al., 2002).

Over the past several decades, non-whites have experienced increasing levels of
decentralization (Charles, 2003; Fischer, 2003). However, their increased pattern of
suburbanization does not necessarily equate to spatial assimilation with the majority
white population (Fischer, 2008; Iceland & Wilkes, 2004; Charles, 2003). Blacks, in
particular, have been suburbanizing but they continue to experience high levels of
segregation from whites, leading some scholars to assert that a “reghettoization” instead
of integration has been taking place in suburban areas (Alba & Logan, 1993).
Accordingly, many US cities and metropolitan areas experience high levels of
segregation, or for a good number of cities, continue to be “hypersegregated” as is the
case for Detroit, Chicago, Cleveland. Milwaukee, Newark, and Philadelphia (Massey &
Denton. 1993; Wilkes & Iceland, 2004).

The high level of segregation between blacks and whites is not unparalleled.
Blacks are the most segregated individuals compared to others in different racial and
ethnic groups. According to the US. Bureau of the Census report in 2000, 58 percent of
the Asian population lived in suburbs in 2000, up by 5 percent from 53 percent in 1990;
49 percent of Latino population, up by 3 percent from 46 percent in 1990; 39 percent of
Blacks or African-American population up by 5 percent from 34 percent in 1990. As can
be observed, Blacks or Aﬁican-American suburbanites remain signiﬁcantly lower in
number compared to Asian and Latino groups despite a comparatively similar rate of

suburbanization across these population groups. Compared to whites, which was 71

26

percent in 2000, black suburbanization rate trails to white suburbanization rate by a
signiﬁcant 32 percent points (Logan, 2003).

What explains the differential levels of suburbanization across racial groups?
What might government policies on local community development contribute to the
differential suburbanization of people by race? The causes of decentralization in general
and differential suburbanization by racial groups in particular, reﬂect complex processes
that interact at different geographic scales across time.

The existing literature in social geography indicates three major theoretical
perspectives underlying the differential pattern of suburbanization by race, including: 1)
the ecological or place assimilation; 2) neighborhood preference; and 3) place
stratiﬁcation. These perspectives inﬂuenced much of the theorization of racial differences
in decentralization and the sustained segregation in metropolitan areas. These

perspectives are described in the next section.

The Spatial Assimilation Perspective
The spatial assimilation perspective asserts that individuals and population groups

demonstrate differential mobility patterns and propensity to assimilate with the majority
population group based on socioeconomic and cultural competence (Massey & Denton,
1985; Massey, 1985). Central to the spatial assimilation perspective is the notion that
individuals and population groups will attain a greater life’s chances once adapted and
assimilated with the majority population group. With its ecological underpinnings from
the Chicago School of Sociology (Park et al., 1925; Massey, 1985), spatial assimilation

perspective emphasizes “natural” processes involving competition, invasion, and

27

succession resulting in a spatial distribution of people based on “ability to pay” and
acculturation or adaptive characteristics (Burgess, 1928). These characteristics —
measured in income, education, length of stay, and language proﬁciency - determine
social mobility which in turn, facilitates spatial mobility (Park et al., 1925; Massey &

Denton, 1985).

The literature identiﬁes three ways that spatial assimilation could proceed
including: 1) structural assimilation; 2) cultural assimilation; and, 3) segmented
assimilation. Structural assimilation reﬂects a pattern of assimilation between minorities
and the majority population group based on socioeconomic factors. It asserts that income,
education, and employment facilitate suburbanization and increase the propensity for

residential integration.

Cultural assimilation, on the other hand, emphasizes the role of cultural
adjustment — language proﬁciency, length of stay — in the quest for assimilation of
minority population groups, particularly foreign-bom minorities, to the majority whites
(Massey & Denton, 1985). Factors that reﬂect these two assimilation tendencies shape

the third one, segmented assimilation.

Models that analyze segmented assimilation reﬂect socio economic and
cultural/contextual factors and how these offer “divergent destinies” or assimilation paths
(Zhou, 1997; Portes and Rombaut, 1996). Central to segmented assimilation is the notion
that context, which encompasses physical, social, cultural characteristics of the host
community that immigrants come in ﬁrst contact, determine their assimilation path. The

immigrant could become: 1) an upwardly mobile and integrated with the host

28

community; 2) trapped in a culture of poverty; and 3) adapted or assimilated with the

host’s culture but retains ethnic cultural traits (Zhou, 1997).

The importance of spatial mobility in spatial assimilation is such that spatial
mobility increases non-whites’ prOpensity for participation in the opportunity structure
such as access to jobs, quality housing and education, and other quality of life factors. As
assimilation theorists contend, spatial mobility is an important step toward racial
integration and assimilation with the broader society (Massey & Denton, 1985, 1993;
Massey, 1985). Therefore, a better participation of non-whites in the pattern of
decentralization would enhance greater opportunities for assimilation and integration

between groups.

Since the focus of this study is about population decentralization in metropolitan
areas, with central city population groups that are traditionally exposed to factors
constraining social and spatial mobility (i.e., blacks or African Americans and
Hispanics), the study reviews empirical evidence for structural factors and less attention
has been placed for reviewing empirical evidence for cultural and segmented assimilation
factors of spatial mobility although all three assimilation tendencies could operate

simultaneously at the same time.

Socioeconomic Status and Population Decentralization

Socioeconomic status affects people’s propensity to assimilate with the majority
population group (i.e. whites of Caucasian descent). The consensus among proponents of
the spatial assimilation perspective, more speciﬁcally structural assimilation, is that

measures of socioeconomic status such as income, education, and employment are

29

predictors of an individual or group propensity for spatial mobility. A number of studies
by William Clark and others underscore the relevance socioeconomic status in residential

location patterns.

Clark and his associates provide evidence for the relevance of socioeconomic
factors in the segregation of different racial groups from whites. In a study that uses data
from the Multi—City Study of Urban Inequality, Clark (2007, 2009) found that
socioeconomic matters in preferences for residential location among whites, blacks,
Asians, and Hispanics in four major cities — Los Angeles, Boston, Atlanta, and Detroit.
He found that income and education are consistent in explaining propensity for these
racial groups to live in diverse residential places. For blacks, increases in income and
education indicate increases in their propensity to live with whites in majority white
neighborhoods. However, there is still a strong indication that preference for same race
residential location persists across these racial groups. In particular, whites still indicate
resistance to living in neighborhoods with high percentage of black residents, while
blacks still indicate a preference for no less than 50% black and 50% other race
combination. As these ﬁndings show, socioeconomic factors —- income and education —
are signiﬁcant factors in explaining residential patterns in decentralization in

metropolitan areas.

To analyze the segregation of blacks from whites in Southern California, Clark
and Ware (1997) use dissimilarity and relative exposure indices of segregation across
census tracts in Los Angeles, Orange, Riverside, San Bemardino, and Ventura for ﬁve
categories of income and seven categories of education. By comparing dissimilarity and

exposure indices for blacks and whites within the same income and education categories,

30

they ﬁnd that decreases in levels of segregation of blacks corresponded with increases in
levels of socioeconomic status. They concluded that socioeconomic status does explain

segregation.

Iceland and Wilkes (2006) provide a strong support for the relevance of
socioeconomic status in residential patterns. Their study evaluated the relevance of
spatial assimilation and place stratiﬁcation perspectives in patterns of segregation for
African-Americans, Asians, and Hispanics groups from whites in all US metropolitan
areas for the period 1990 and 2000. They used the dissimilarity index to measure
segregation between each of these racial groups and whites. Consistent with previous
ﬁndings, Iceland and Wilkes ﬁnd that the spatial assimilation perspective is well
supported among Hispanics and Asians only. Blacks were found to have the highest
levels of segregation for all time periods relative to Hispanics and Asians. Their ﬁndings
indicate that high levels of socioeconomic status (measured in mean scores of education,
income, occupation, and poverty status) are associated with lower measures of

segregation speciﬁcally for Hispanics and Asians population groups, but not for blacks.

However, Iceland and Wilkes (2006) ﬁnd that the change in the socioeconomic
status among blacks — increases in the number of blacks that moved out of the "in
poverty" and "less than high school education" groups — affects segregation patterns. This
ﬁnding shows that as blacks move up in the socioeconomic structure, they are more
likely to experience less isolation from the majority whites. Essentially, blacks with
higher socioeconomic status have a higher probability for participating in the

suburbanization process and being assimilated with whites.

31

Support for the assimilation perspective is further stressed by Freeman (2008)
when he re-examined the relevance of socioeconomic status on locational outcome of
blacks in US. metropolitan areas at the individual level for 1970 to 2000. He found that
household income and education explained the propensity of blacks to assimilate with
whites. Accordingly, more high income and educated blacks live with whites than poorer
and less educated blacks. The strength of this ﬁnding, however, was the same for all time
periods, which suggests that while socioeconomic status matters, the strength of this
relationship did not improve. This ﬁnding raises questions about other possible causes
that may have contributed to the signiﬁcant but “static” role of socioeconomic status in

the pattern of mobility and residential outcome of blacks in metropolitan areas.

Despite a consistent support for the spatial assimilation perspective in levels of
suburbanization and pattern of segregation across metropolitan areas, the lack of
improvement (i.e. decreases) in segregation across time periods and the sustained high
levels of segregation of blacks from whites for all time periods in past studies, challenge
the notion of the differential levels of suburbanization and segregation as a pure “class-
based” phenomenon. The fact that the level of segregation for blacks with respect to other
racial groups is consistently high regardless of geography contributes to this ambiguity
(Clark, 1992; 2009). Blacks from central cities are as highly segregated as their black
suburbanite counterparts relative to other racial groups (Darden et al., 1987; Darden,
2000; Clark, 2009). For these reasons, the role of race in residential pattern and the
decentralization of people by race cannot be ignored. There is simply an overwhelming
evidence to suggest that race is a pervasive explanation in residential patterns in US

metropolitan areas.

32

The Place Stratiﬁcation Perspective

The place stratiﬁcation perspective asserts that social barriers impede spatial
mobility for many minorities (Massey & Denton, 1993; Charles, 2003). Central to place
stratiﬁcation perspective is the notion that society is hierarchically organized with
different population groups ranked in such a way that the more advantaged groups locate
in places with better living arrangements while disadvantaged groups remain in
undesirable places (Charles, 2003). The main argument for place stratiﬁcation holds that
some population groups are not integrated with majority whites because of factors other

than socioeconomic status.

Scholarly works indicate that racial prejudice and discrimination against the
disadvantaged population groups, particularly minorities and people of color. Racial
prejudice and discrimination have been proposed as mechanisms for constraining social

mobility, which in turn impede spatial mobility.

Social Barriers to Population Decentralization

Previous research has shown that there are many factors that are simultaneously at
work in determining and sustaining the differential levels of suburbanization and
separation between population groups (Massey & Denton, 1993). Generally, research
motivated by place stratiﬁcation perspective views the separation between racial groups
as an expression of individual’s or group’s aversion for other groups on the basis of race
(Darden, 2004; Jackson, 1994; Harris, 1999, 2001). In residential spaces, the
geographical separation between racial groups tends to reﬂect: 1) individual avoidance

due to prejudice; and 2) discrimination in the housing market (Squires, 2003; Williams et

33

al., 2005; Massey, 2005). These have become the core mechanisms for sustaining the
pervasive separation across racial groups in residential spaces (Charles 2003; Darden,

2003).

The most recent ﬁnding that suggests a possible presence of factors that reﬂect
the place stratiﬁcation perspective in residential pattern comes from the work of Fischer
(2008). Fischer ﬁnds that the average decrease in suburban segregation of blacks from all
other racial groups — including whites, Hispanics, Asians and those categorized as “all
others” — in 248 metropolitan areas across the US. corresponded with areas that
experienced rapid increases in the black population. Fischer alludes to the notion that
suburban residential landscapes are reconﬁguring consistent with “white tolerance

thresholds for contact with blacks, or other factors related to place stratiﬁcation” (p. 492).

Different tolerance levels for integation contribute to residential segregation and
are a hindrance toward suburbanization for minorities and their residential integration
with majority whites (Clark, 1992). In his study in Los Angeles, Clark found that
individuals are more willing to move into neighborhoods with a race combination that is
similar to their own. He found those who indicated “no preference” mostly moved into
their own-race neighborhood. Whites prefer same race the strongest while blacks prefer
the same race the weakest. Also, whites would prefer to move into neighborhoods with
low percentage blacks, while blacks prefer to move into a 50 percent black and 50
percent other races (Charles, 2003; Farley, 1995). This behavior suggests that while there
is some acceptance for the idea of an integrated neighborhood, there remains an
overwhelming unwillingness for actually sharing residential spaces (Clark, 2009; Farley,

1995). This evidence shows differences in tolerance levels for people of different race

34

and culture. Such tolerance level may reﬂect motivations for establishing a well-deﬁned

racial line among different groups.

Massey & Denton published the American Apartheid (Massey & Denton, 1993),
one of the most important works that illustrates the effects of prejudice and
discrimination in residential spaces. By studying the pattern of segregation across major
cities in the US, these authors show an extreme isolation of blacks from education and
employment opportunities combined with “white ﬂight” perpetuates the “black
underclass.” These authors indicate that prejudice and discrimination impede blacks from
participation in the opportunity structure in suburban areas particularly in jobs, education,

and housing market.

A signiﬁcant amount of support for place stratiﬁcation factors in explaining
different levels of suburbanization for minorities or people of color, particularly blacks,
can also be gleaned from segregation studies that consistently found high levels of pattern
of segregation between blacks and whites relative to other racial groups. Iceland and
Wilkes (2006) demonstrate that scale and lack of within group variation make a
difference in explaining high level of segregation between blacks and whites at the
metropolitan level for the years 1990 and 2000. Contrary to Asian and Hispanic
population groups, the residential pattern for blacks does not support the spatial
assimilation reasoning. Blacks’ residential segregation does not improve (decrease) with
increasing socioeconomic status at the metropolitan level. The authors regard this ﬁnding
as a support to place stratiﬁcation reasoning. However, a further analysis of within-black
socioeconomic status and its effect on level of segregation yields an interesting result.

The authors show that blacks with higher socioeconomic status tend to live in less

35

segregated areas than their poor black counterparts, implying that socioeconomic status
still does matter in explaining the residential pattern but with exceptions.

Johnston et al. (2007) revisits the work of Massey and Denton (1988) on
segregation measures and applied the 5 dimensional segregation indices at the census
tract level between whites and three racial groups, namely: blacks, Hispanics, and Asians.
Consistent with prior ﬁndings, Johnston and his associates ﬁnd the segregation of blacks
the highest. Similarly, Logan et a1. (1996) ﬁnd that socioeconomic factors —
homeownership, income, education, and nativity/language — explain residential pattern
for Hispanic and Asians but not for Blacks. Both studies Show that there is a greater

exposure of Asians and Hispanics to whites in residential areas but lesser for blacks.

Darden and his associates undertook a series of studies to test the relevance of
ecological theory on suburbanization and segregation across different ethnic groups
(Darden, 1977, 1985, 1986, Darden et al. 1987; Darden & Kamel, 2000). Their consistent
ﬁnding shows that blacks lag in levels of suburbanization, and empirically shows that

race essentially distorts the ecological theory.

For example, Darden (1990) indicates that professional status enables whites,
Asians, Hispanic, and American Indian in the North Central Region to experience higher
suburbanization rates. This did not appear to be the case for blacks with professional
status. Darden contends that participation in suburbanization across these population
groups “exists along a continuum deﬁned more by race than socioeconomic and
professional status (p. 20). Suburbanization rates for these population groups based on
professional status were found to proceed in the following order: whites, American

Indian, Asians, Hispanics, and blacks.

36

Similar tests were undertaken by Darden in Kansas (1986) and Chicago (1987).
He examined the residential segregation between blacks and whites, while controlling for
socioeconomic status —— occupation, income, and education. The analysis of both yields a
strong support for the signiﬁcance of race in the segregation of blacks from whites. The
black-non Hispanic white segregation was consistently high despite the Similarity in level
of socioeconomic status between blacks and non-Hispanic whites communities that were

studied.

In Detroit, patterns of segregation between blacks and whites were high in central
city but a much higher pattern was observed in the suburbs (Darden et al., 1987; p. 87).
This ﬁnding leads many scholars to believe that the residential pattern in Detroit ﬁts well
with the process of “reghettoization,” a description for a process whereby “re—

segregation” rather than dispersal of blacks in suburban areas is occurring (Clay, 1979).

As these studies show, race has been shown to be a pervasive factor in the
sustained separation across different racial groups. The different tolerance levels,
individual prejudice, and a consistent ﬁnding of the high levels of segregation between
Blacks and whites relative to other racial groups demonstrate the strength of the Place
Stratiﬁcation perspective for understanding the differential levels of suburbanization and

pattern of segregation in metropolitan areas.

Discrimination in Housing

Empirical evidence also shows that segregation is an outcome of institutional
forms of discrimination that are prevalent in the housing market. These discriminatory
practices occur in the real estate, lending, and insurance industries. There is also a strong

evidence for the inﬂuence of government agencies and programs at different levels, and

37

local jurisdictions. However, the manner by which these discriminatory practices are
carried out are not easily detected, ranging from overt to covert practices that have
become sophisticated and undetected over the years (Williams et al., 2005; Massey,

2005).

Institutional forms of discrimination in the housing market have been carried out
by real estate and ﬁnancing industries through the practice of imposing restrictive
covenants, red lining (Massey & Denton, 1993; Yinger, 1995), block busting (Sugrue,
1996), racial steering (Yinger, 1995), and by denying home mortgage loans even when
applicants (mostly blacks) are qualiﬁed to receive approval for ﬁnancing (Yinger, 1995;
Alba and Logan, 1993). Excluding black realtors from participating in white dominated
realtor’s organization was practiced (Darden, 1977, 2001), as well as discrimination in
the property insurance industry (Squires, 2003). Blacks were also subjected to the “color
tax”, a phenomenon that a resident pays for being black in America (Darden, 1973).
Blacks were paying higher rent than whites for housing of the same quality or equal rents

for lower-quality housing.

The pervasiveness of discriminatory practices resulted in the uneven
homeownership rates across racial groups in the US. (Haan, 2007). It led to many
policies and programs that attempt to mitigate discrimination in the housing market. The
impact of these policies, however, remains vague. According to Williams et al. (2005),
“discrimination in the housing market has become increasingly difﬁcult to detect and
decreases in incidence of abuses may not reﬂect the reality.” Indeed, the Housing and

Urban Development has reported that an estimated two million or more abuses in housing

38

discrimination in the US occur. These and many more incidents of abuses at different

stages involved in buying or renting a home are unreported (Farley & Squires, 2005).

Because housing discrimination has become increasingly undetected, Williams et
al. (2005) suggest that percent homeownership would be a relevant measure for detecting
discrimination in housing. The percent homeownership at a disaggregate level across
different racial groups provide a sense of the distribution of mortgage lending activity. It
is also a relevant indicator for successful mortgage lending that can possibly provide an
indirect way for measuring discrimination in lieu of the “paired audit” approach, which is
one of the more sophisticated and resource intensive form of uncovering discriminatory

behavior in the housing market (De Rango, 2001; Galster, 1992).

Overall, the place stratiﬁcation perspective provides an alternative interpretation
for sustained separation across racial groups. The differential levels of suburbanization
reﬂect individual prejudice, and differential levels of tolerance facilitated by institutional
forms of discrimination in the housing market. These have signiﬁcant effects for how
preference for certain neighborhoods is shaped. Indeed, research has shown that the
neighborhood preference perspective ties in with place stratiﬁcation and intersects with
features of spatial assimilation perspective. Therefore, all three perspectives are mutually
reinforcing theoretical explanations for the differential levels of suburbanization and

pattern of segregation.

39

The Neighborhood Preference Perspective

The neighborhood preference perspective asserts that residential segregation is a
function of a group’s or individual’s preference for a particular residential location
(Schelling 1971; Clark, 1991; Clark & Fossett, 2008). Central to this perspective is the
notion that individuals and groups have a propensity to reside in communities with a
demographic combination that they perceive desirable. It highlights the notion that
residents locate in particular communities as a manifestation of “in-group” attraction
among residents (Chung & Brown, 2007; p. 314). Hence, the spatial distribution of
people by race can be understood as an individual choice or voluntary decision to reside
in a particular placed based on the demographic composition of the destination

community (Clark, 1991; Farley, et. al., 1979).

Preference for Homogeneous versus Integrated Neighborhood

Studies on preferences for residential neighborhood based on racial composition
have provided a consistent ﬁnding for all racial groups. Generally, the ﬁnding shows that
different racial and ethnic groups tend to exhibit an “own race” preference for residential
location (Clark & Fossett, 2008; Clark, 2009).

For example, ﬁndings from the Multi-City Study of Urban Inequality (MCSUI),
which was a multi method survey to understand issues related to poverty and urban
inequality, provides compelling evidence for the relevance of neighborhood preference
perspective in patterns of decentralization. Analysis for Detroit, Boston, Atlanta, and Los
Angeles reveal similar ﬁnding with regards to preference for neighborhood composition

among blacks and whites. Blacks prefer an integrated neighborhood with a combination

40

of 50 percent blacks and 50 percent other race (Clark, 1991, 2009; Farley et al., 1997).
Blacks’ preference for integrated neighborhood increases with decreasing education and
income. Accordingly, nearly half of blacks with low income also preferred an integrated
neighborhood. On the other hand, middle class blacks prefer neighborhoods with the
greater presence of white residents (Clark, 1991, 2009; Farley et al., 1997).

For whites, they tend to have a stronger preference for low percentage of blacks in
their neighborhoods. The overall ﬁnding for all racial groups in the study reveals that
whites, blacks, Asians, and Hispanics, all have a strong “own-race” preference in their
choice of residential neighborhood. Except for whites, however, all racial groups
indicated preference for integrated neighborhoods.

The analysis of Farley et a1. (1997) with regards to the Asian racial group
deserves further examination. Farley and his associates ﬁnd that the tolerance of whites
increases with Asians and other fair skinned minorities especially those of whites of
Hispanic origin. This behavior has been described as one consistent with the concept of
“desirability continuum,” in which members of white ethnic groups are viewed as less
inferior than Orientals, and the latter were seen as more desirable than blacks (Lieberson,

1980).

Results from studies on residential preferences reviewed above indicate individual
preferences for certain demographic characteristics as underlying explanation for
residential patterns. It is clear that preferences for residential spaces are inﬂuenced by
factors other than an individual’s “ability to pay” status (Clark 1992, 2009). It appears
that perceived hostility in the destination community is a relevant factor affecting an

individual’s preference for certain residential locations.

41

The Role of Government Policy in Population Decentralization

The review of the different theoretical perspectives and empirical evidence for
racial differences in decentralization and segregation areas reveals two things. First, the
existing literature of residential pattern and segregation are not clear about the direct
effects of government policies on sustained pattern of separation in residential spaces.
There is a lack of study that demonstrates a possible direct effect of local policies in the
separation of people by race. And, a study that directly links government policies,
particularly local roads spending policies on decentralization is not adequate.

Second, theoretical perspectives that motivate studies about underlying processes
of suburbanization and segregation are clearly not mutually exclusive (Chung & Brown,
2007). Most of the existing studies analyze residential pattern using one or two
theoretical frameworks and neglect to recognize that understanding complex processes
like decentralization and segregation require a multiple perspective approach.

Thus, this study is an attempt to understand the effects of local roads on racial
differences in decentralization. It does so by using a human ecological methodological
framework, which accommodates an analysis of the relationship between roads and

decentralization using a broad theoretical base and multiple perspective approach.

Government Policy on Roads and Decentralization
Roads and road building are basic infrastructures that shape human settlements.
Indeed, roads serve as a blueprint to any land use organization and formation and
facilitate spatial interactions (Black, 2003). Roads serve as a major conduit for growth in

one location and decline in another (Boamet, 1998). The extent that roads, both highways

42

and local roads facilitate accessibility and reduce transportation cost, roads can be viewed
as a signiﬁcant facilitator of decentralization in metropolitan areas.

Indeed, the pattern of decentralization in metropolitan areas has been tied to
government policies on road development. Empirical evidence shows that roads,
speciﬁcally highways, have facilitated the dispersion of people and jobs across
metropolitan areas and regions throughout the country. Baum-Snow (2007) ﬁnds that a
constructed highway that passes through a central city impacts the population distribution
of the total metropolitan area. He estimates that a constructed highway in a metropolitan
area that passes through a central city reduces that city’s population by about eighteen
percent (18%). Additionally, his ﬁnding shows that the city would have grown by eight
percent (8%) if the highway system was not constructed.

In another study about highway impacts, V058 and Chi (2006) ﬁnd that the
expansion of all major highways in the state of Wisconsin increased the population of
- surrounding communities. They estimate that between 1970 and 2000, metropolitan civil
divisions (MCDS) that are located ten to twenty miles away from an expanded highway
experienced population growth. A similar ﬁnding related to the decentralizing effect of
interstate highways comes from Chandra and Thompson (2000), who ﬁnd that retail
activities across regions are redistributed due to the construction of a highway system.
The constructed highway displaced some industries in one county and shifted them to
another (Chandra and Thompson, 2000).

The study of Haugwhout (1999) reinforces and sheds additional light to the link
between infrastructure investment (including transportation) and pattern of employment

distribution. Haughwout (1999) used a spatial equilibrium model to specify the marginal

43

effects of infrastructure investment and found that state infrastructure investments tend to
facilitate spatial distribution of growth in favor of the suburban or less dense areas within
the state. His ﬁndings indicate that infrastructure investment affects the relative
attractiveness of places and redirects the spatial allocation of economic activities. He also
stresses that central cities are underprovided when it comes to public investment and tend
to cope with maintaining extensive infrastructure stock while the tax base or local ﬁscal
support declines.

A similar analysis of the impacts of infrastructure investment on economic
development was undertaken by Berechman et a1. (2006). This study provides evidence
for highway development impacts at different spatial and temporal scales. Berechman
and others attempt to verify previous ﬁndings on the differential effects of transportation
investment on economic development given spatial/ geographical and time lags. A
production function type of the basic model was estimated, with output as the dependent
variable, and labor, private capital, public capital (transportation capital) and
unemployment as independent variables. A lagged model in which 1 to 5 years time lags
were introduced was estimated to test for possible delays in output response to additional
transportation investment. Another model tested for spillover effects on neighboring
areas via the Moran’s 1 spatial autocorrelation test. The results from an analysis of data
on 48 states across the U. S. for the periods 1990 and 2000 indicate a signiﬁcant impact
of transportation investment on production output at different geographical scales
including state, county, and municipality. Results further indicate output elasticities of
0.37, 0.34, and 0.01 for state, county, and municipality levels, respectively. Finally, the

“spillover” effects of highway investment were found to have mostly affected

44

municipalities. This means that impacts of highway investment onto neighboring areas
where highways are located tend to increase at a decreasing level of analysis (i.e. from
state to municipality).

Clearly, these studies provide a strong evidence for the spillover and dispersive
effects of interstate highways and have shown that highways promote beneﬁts and
advantages in some places, but costs and disadvantages in others. Also, these studies
provide an evidence for the transformative capacity of highway development in
urbanized areas from mono-centric to complex polycentric urban environments in just a

few decades.

Potential Links between Local Roads and Decentralization

While highways and major throughways were found to facilitate the
decentralization in metropolitan areas, the role of local roads particularly in non-whites’
participation in the decentralization process is not quite well understood. Two
explanations are appropriate at this point. First, aside from the scant literature on the
relationship between local roads and decentralization, ﬁndings about the relationship tend
to be mixed. A second reason, which is tied to the ﬁrst, is that most studies that address
local roads do not focus on decentralization but other outcomes such as costs of
investment on infrastructure and urban form, and local roads and travel or commuting
behavior. These studies often neglect an analysis of local roads as if these do not matter
in household relocation decision. Further, these studies do not shed any information about
the possible facilitative and dispersive impacts of local roads investment. Essentially,

without a direct analysis of the relationship between these processes, it is not clear

45

whether local roads or government spending on local roads, indeed, spurs
decentralization.

Hartgen (2003) ﬁnds a weak relationship between them roads and growth in
North Carolina. He shows that only about 15 to 25 percent of the variation in grth is
explained by road improvements such as urban and rural widening. He concludes that
prior population density level, not road investment, appear to have more signiﬁcance in
explaining the pattern of growth in the twelve commuting regions that comprise one
thousand ﬁve hundred ﬁfty one or 1551 census tracts in North Carolina during the 19903.
Vojnovic (2000) on the other hand, shows a positive association between municipal
expenditure on roads and suburban sprawl in the Metropolitan Area of Toronto. Both
studies, however, did not directly explain how the relationship is linked to non-whites’
pattern of decentralization. The discrepant ﬁndings from these studies may be related to
place (i.e., Toronto versus North Carolina), scale of analysis, and time period covered in
the study, and show the complexity of this research question.

The paucity of studies and a lack of consistent results on the relationship between
local roads and the decentralization of people in metropolitan areas necessitate further
research. Most existing studies on local roads focus on costs, and urban form and travel,
and overlook addressing the possible local roads-population decentralization relationship.

For example, in determining the relationship between urban form and different
types of government expenditures, Burchell et a1. (2002) examined the relative costs of
infrastructure, land, services, and housing, and compared these costs within the context of
sprawl (sprawl being characterized by low-density development) and compact or high

density pattern of growth at the county level in different regions in the US. Speciﬁcally

46

for local roads, these authors ﬁnd that between the projected years 2000 and 2025, a
projected cost saving to the US government is 188,305 or 9.2 percent of lane miles and
$110 Billion or 11.8 percent of road costs under the compact and high density urban
form.

Unfortunately, Holcombe and Williams (2008) do not support their ﬁnding. These
authors examined population density as it relates with per capita municipal government
expenditure including infrastructure, services, and total expenditure in 487 large and
small cities across the US and show that higher population density does not lower per
capita government expenditures. On the other hand, their test on the relationship between
population density and infrastructure expenditures yields the expected result, indicating
that higher population density lowers infrastructure cost in small cities. Unfortunately,
only highways and sewer expenditures were included in their study and local roads were
ignored in the measure of infrastructure.

Despite the lack of emphasis on local roads in the study by Holcombe and
Williams, other ﬁndings in their study show the beneﬁts that could be gained from higher
density development particularly for small cities. Residents in small cities with higher
density are more likely to achieve economies of scale from infrastructure services, which
translates into lower cost on infrastructure services for individual residents.

The inconsistent ﬁndings about urban form and the cost of infrastructure
development to the government resonate in other areas of study such as in commuting
and travel behavior (Cervero & Kockelman, 1997; Cervero, 2003). Most studies within
this domain test for the relevance of some planning design principles (i.e. neotraditional

and smart growth) on travel behavior and further evaluate whether or not development

47

pattern that reﬂects any of these design principles are effective in reducing private
motorized commuting and curbing the pattern of urban sprawl. Thus, local roads are
implicitly considered in these studies but the direct link to population decentralization is
not analyzed.

For example, Cervero and Kockelman (1997) ﬁnd that impacts of urban form
based on different physical design elements (density, diversity, pedestrian-oriented
design) alter travel behavior in terms of a reduction in non-work trips, and increases in
non- motorized travel. The study ﬁnds that neighborhoods with gridiron street designs
and restricted commercial parking are signiﬁcant variables in reducing household level
non-work related vehicle miles traveled (VMT) and the use of single occupant vehicles.
However, their ﬁnding is contradictory to what Crane and Crepeau (1998) found in their
study on the relationship between street design and travel behavior. The latter ﬁnd no
signiﬁcant link between street network pattern and short and long-term non-work travel
behavior.

In another study, Sultana and Weber (2007) ﬁnd that travelers from communities
characterized by pattern of urban sprawl (i.e. low-density) tend to experience longer
commute time and mileage. Their ﬁndings also show that socioeconomic characteristics
of the travelers are equally important factors in commuting behavior. The ﬁnding implies
that in explaining commuting behavior, socioeconomic characteristics and the pattern of
development that the commuters reside are closely interwoven and either attribute or
factor could explain commuting behavior alone.

As these studies show, there is a lack of an analysis of the direct relationship

between local roads and decentralization. The existing studies reviewed do not examine

48

whether a particular urban form (i.e. low-density, high-density, and pedestrian-oriented)
given its corresponding type of road network (i.e. gridiron) is given consideration in
individual or household relocation decisions. Moreover, these studies do not provide
insights into possible mechanisms by which local roads or the type of road network in a
given community facilitates the decentralization of people by race. So far, existing
studies on local roads and decentralization or sprawl, presented above are informative but
ﬁndings are still inconclusive. The paucity of studies that directly assess local roads
impacts on decentralization contributes to such gap. This study attempts to contribute in
ﬁlling this gap.
Summary

This chapter provides a brief description of the shifts in population trend in US.
metropolitan areas during the twentieth century. It emphasizes the post World War 11
population decentralization and assesses existing literatures related to its underlying
causes. A review of the relevant theoretical and empirical evidence reveals the lack of
information on the direct relationship between local roads spending and decentralization.
However, informed by three theoretical perspectives in patterns of residential mobility
and segregation, the review highlights education, income, demographic composition that
deﬁnes the homogeneity or heterogeneity of communities, incidence of housing
discrimination, and locational or spatial attributes as determinants of patterns of
decentralization.

The next chapter discusses a conceptual framework that reﬂects a multi
perspective approach in understanding the decentralization of people in metropolitan

areas. Using a human ecology methodological framework, the role of government policy

49

is integrated into a broad based human ecology conceptual model that recognizes
explanatory factors shaping decentralization from spatial assimilation, place stratiﬁcation,
and neighborhood preference perspectives. These theoretical perspectives were used to
identify the proposed explanatory factors of patterns of decentralization across the white

and non-white population groups.

50

CHAPTER THREE: THE CONCEPTUAL MODEL

This chapter builds upon the review of the relevant information related to the
decentralization and residential patterns in metropolitan areas presented in Chapter Two.
It draws from the different theoretical perspectives of decentralization and segregation —
spatial assimilation, place stratiﬁcation, neighborhood preference perspectives — as
theoretical foundations for the development of a conceptual model of the relationship
between local roads and decentralization of people in metropolitan areas. What follows
describes a human ecology conceptual model that characterizes the causal relationship
between local roads and decentralization.

Using human ecology, the conceptual model developed for this study situates the
role of government policy within other proposed explanatory factors based on the review
of the three existing interrelated theoretical perspectives of decentralization and
segregation. This conceptual model reﬂects a broad and holistic theoretical base that
recognizes the essence of population, behavior, and environment factors shaping pattems
of mobility and spatial distribution of people across space. Further, the model helps orient
an analysis of patterns of decentralization that is inclusive and recognizes the complex,
multidimensional, and dynamic features of the decentralization phenomenon.

The chapter begins with an overview of human ecology followed by the
description of the conceptual relationship between local roads and the decentralization of

people in metropolitan areas. The last section lists the hypotheses of the study.

51

The Human Ecology Framework

The subject of human ecology is recognized to have existed long before the
concept was formally institutionalized in the 19203 (Borden, 2008). It is a subject with a
rich intellectual tradition beginning when the term “ecology” was introduced by Haeckel
in 1866 (Lawrence, 2003) and has become a foundational concept for many ecological
studies of human systems particularly in the social science disciplines. Gross (2004)
alerts us to the fact that human ecology was conceptualized and applied in different ways
as different disciplines (i.e. geographers, sociologists, and ecologists) discovered or
rediscovered the relevance of the human ecology framework in analyzing
population/human and environment relations. This means that human ecology has been
deﬁned and conceived of in different ways, mostly reﬂecting disciplinary orientation and
perspective (Borden, 2008; Gross, 2004; Lawrence 2003). It has been deﬁned and
conceived of as an approach, framework, method, study, program and has a wide range of
application across the social and natural sciences (Gross, 2004).

This dissertation research uses the conceptualization of human ecology as
understood by sociologists and human geographers particularly among social
geographers, which primarily conceptualizes human ecology as a study of human
environment relations treats human ecology as an approach for analyzing human systems
that recognizes complex, multi dimensional, and dynamic nature of such systems. As the
subsequent paragraphs would show, human ecology as a framework has had a strong
presence in both sociology and geography disciplines. The subject of human ecology has

stimulated extensive discussions with scholars from both disciplines forging areas of

52

agreements and disagreements as they attempt to erect disciplinary boundary and identify

“catchment areas” (Gross, 2004).

Sociological Interpretations of Human Ecology

The sociological literature traces the origin of human ecology to the works of
Park, Burgess, McKenzie and others at the Chicago School of Sociology during the ﬁrst
half of the twentieth century. In its original interpretation, the sociological perspective of
human ecology advanced by Park and others adopts biological concepts in plant and
animal ecologies, and emphasizes processes of competition, invasion, and succession as
core processes underlying human systems (Park et. al., 1925). Park deﬁned human
ecology as a “study of the spatial and temporal relations of human beings as affected by
the selective, distributive, and accommodative forces of the environment” (Park, 1925; p.
63). This interpretation of human ecology essentially treats human ecology as a theory for
studying human populations and their ability to adapt to the changing environmental
conditions and the patterns of land use organization that result from the adaptation
process. Essentially, human ecology emphasizes evolutionary processes underlying the
formation and growth of cities--“from simple societies to complex conurbations”
(McKenzie, 1967; p.73).

The pioneering authors of human ecology were not only interested in describing
the pattern of physical and social structure of the study area (i.e. the City of Chicago) and
their changes but also the effects or consequences of these changes on social behavior
and institutions. Further, these authors were keen about understanding how spatial

relationships are linked to urban residents’ experiences with life in the cities including

53

psychological, moral, and spiritual aspects of their experiences. As such Park and his
associates focused on the analysis of segregation, culture, vice, mental illness, divorce
and family issues, and a host of sociological issues using neighborhoods in the City of
Chicago as study areas.

In articulating the human ecology framework, Park emphasizes a model of urban
change characterized by four co-evolutionary interacting variables or components, which
is similar to a human ecological model adopted by human geographers (Meade, 1977;
Meade & Earickson, 2000). The components include: 1) population; 2) artifact
(technological culture); 3) customs and beliefs (or the non-material culture); and natural
resources (Park, 1936; p. 15). Park’s human ecological framework highlights the essence
of the population-environment relations in shaping social and spatial structure of cities
recognizing that technology (i.e. transportation technology) and non-material
culture/behavior (social organization) are equally signiﬁcant components aside ﬁom
population and the enviromnent.

The recognition of the technological culture in the human ecology framework is
noteworthy. The advances in technology, particularly transportation technology has
played a signiﬁcant part in urban growth and the sociospatial transformation of cities and

metropolitan areas, as the next section would reveal.

The Classic Ecological Models
Three major ecological models, which include the Burgess Concentric Zone,
Hoyt’s Sectoral model, and Harris and Ullman Multiple-nuclei model are recognized as

classical human ecological models for explaining change in sociospatial structure and

54

growth in cities. The Burgess Concentric Zone model (1925) presents a dynamic theory
for the social and land use organization of the City of Chicago. In this dynamic model,
competition, invasion, succession, and adaptation processes, which are core to the human
ecological perspective of urban spatial change, are processes underlying residential and
neighborhood change. The model, presents the notion of the “survival of the ﬁttest”
advanced in the study of animal and plant ecologies, wherein individuals and social
groups are conceived of as competing for survival. In urban systems, population groups
compete for the best and most desirable locations in a given locale, such as in a
neighborhood, city or metropolitan region based on economic factors. Burgess theorizes
that the outcome of competition is a land use organization shaped in concentric rings
reﬂecting competitive advantage for certain activities and social groups with respect to a
central business district (Burgess 1925).

Burgess describes the city pattern as consisting of ﬁve zones: Zone 1 is the
central business district (CBD) where intensiﬁcation of land use is prominent and
commercial, ﬁnancial, administrative, recreational, and other institutional services are
located; Zone 2 is the zone of transition, which is occupied by poor and old residential
establishments, and where spillover of businesses and light manufacturing establishments
are located; Zone 3 is the working class zone; Zone 4 is the upper class residential area;
and, Zone 5 is the commuter zone or suburban area, which is mostly devoted to new
developments. As described above, these different zones reﬂect a land use organization
that describes the pattern of city growth through mechanisms of change involving

competition, invasion, and succession processes described above. As a result of these

55

processes, activities of each zone tend to extend to the next, thus shaping neighborhood
change and land use organization.

The effect of competition and evolution on urban growth has been viewed by
Hoyt (1939) from a different dimension. The sectoral model of Hoyt (1939) is an
extension to the Burgess’s concentric zone model. It proposes that expansion of urban
areas follow a sectoral and wedge-like pattern rather than concentric rings, and grants
importance to transportation routes such as highways in the growth and expansion of
cities. It incorporates both distance and direction from a central business district in the
analysis of the pattern of urban expansion and growth, expressed in residential patterns
based on class structure (Yeates, 1965) and in the present land use organization,
commercial development along highway transport routes (Boamet, 1998; Chandra &
Thompson, 2000; Handy, 2005).

Criticisms about the relevance of traditional concentric and sectoral models of
urban transformation during the postwar years stimulated discussions and subsequent
formulation of alternative theories such as the Multiple-nuclei model (Harris & Ullman,
1945) of urban change. One major criticism to the concentric and sectoral models holds
that these models are no longer relevant in describing urban form during postwar years
even though the CBD is still a major center of social and economic relations in a given
locale. The multiple nuclei model of Harris and Ullman (1945) provides a notion that as
city grows and expands, similar and complementary activities tend to group together,
thus, creating the formation of new sub centers. Over time, these sub-centers act as nodes
from which a certain land use organization and pattern develops. This notion has been

empirically tested and ﬁndings have shown that the multiple nuclei model provides a

56

better description of the urban transformation process that is more in sync with
contemporary metropolitan form and structure.

As urban landscapes evolved into a more complex systems and new perspectives
have dominated in the analysis of such systems, the three classical human ecological
models of Burgess, Hoyt, and Harris and Ullman have continued to yield useful results
for understanding urban growth and land organization (Shearmur & Charron, 2004).
They provide important insights for how individuals, households, and jobs are spatially
distributed and the processes that propagate the distribution. Further, these models reveal
the social and implicitly, the power structure, in as much as the_spatial structure of a
metropolitan landscape. Overtime, however, sociologists have advanced alternative
perspectives that recognize forces of urban growth and sociospatial processes beyond the
ideas of competition, dominance, and succession as structures of the classical human
ecological perspective.

Some alternative models of human ecology reject the biological orientation of
classical human ecology and tend to dismiss conﬂict as a driving mechanism for
sociospatial and growth patterns in metropolitan landscapes (Hawley, 1950). Others
represent a more holistic and synergistic model that addresses questions of sociospatial
change and other processes of human systems from an integrated social and environment
perspectives.

A number of human ecology models recognize the importance of population,
environment, and technology but also stresses organizational dynamics as a critical
interacting dimension shaping socio-spatial and other processes of human systems

(Duncan et al., 1959; Duncan, 1964; Dietz & Rosa, 1994; Dunlap et al., 1994).

57

Essentially, this theoretical approach highlights the interrelations between social system
and natural processes. For example, the population, organization, environment, and
technology or POET (Duncan, 1964) model represents a holistic approach that
emphasizes these four interacting and synergistic components of POET as dynamic

causal forces effecting changes described above.

Human Ecology within Geographic Inquiry

While human ecology is a well researched topic in the sociology discipline, the
literature on human ecology in geographic inquiry also reﬂects an extensive engagement
of geographers in the theorization and application of the human ecology subject. Indeed,
similar to sociologists, the idea of human ecology as a holistic approach has been a well-
recognized interpretation of human ecology in geographic inquiry. However, although
geographers recognize human ecology as an integrative study of human-environment
relations (Barrows, 1923; Gross, 2004), the early conceptualization of human ecology in
geography has had strong environmental deterministic orientation.

Goode (1904), for example, deﬁnes human ecology as “the geographical
conditions of culture” (p. 584) and asserts that human ecology “passes beyond geography
(p. 584).” Huntington also casts human ecology towards a more physical and
environmental deterministic interpretation as exempliﬁed in his work on different
climatic regions on civilization (Gross, 2004, p. 589). In the 19203, a shiﬁ in
interpretation from a primarily environmental deterministic to a more holistic orientation
has been made partly as a corrective effort. Barrows (1923), indicates “geography as

human ecology” and deﬁned human ecology as “the objective of geographic inquiry and

58

geographers’ inquiry must be in the pursuit of the mutual relations between man and his
natural, the physical and the biological aspects, of the environment” (p. 3). Thus, it
appears that a paradigm shift within the geography discipline with respect to the
conceptualization and application of human ecology occurred. As the extensive literature
shows, many geographers were engaged in many ecological studies and human ecology
certainly had a niche within this body of studies.

According to Zimmerer (1996), geographers adopt and apply ﬁve ecological
approaches in the study of interrelations among organisms and the environment. These
include: 1) human ecology-including natural hazards approach; 2) cultural historical
ecology-from the Sauerian school; 3) systems ecology-a subset of human ecology; 4)
adaptive dynamics ecology-sometimes referred as cultural ecology, and 5) political
ecology. These different genres of geographical ecology indicate diverse ways that
geographers frame their analysis, with each approach tied to particular methods, subject
of interest, and epistemological perspectives.

ln social geography, human ecology continues to be prominent particularly within
the medical geography sub discipline. Accordingly, “the social model of human ecology
is said to survive in geography in the form of human ecologies of disease” (Del Casino &
Marston, 2006). This human ecology model approaches vulnerability and diffusion of
disease from a multi dimensional perspective and asserts that relationships between
population (biological selves), habitat (natural, social, and built environments), behavior
(beliefs, social organization) and technology shape disease ecology (Meade and
Earickson, 2000). These components of human ecology of disease, in some ways,

parallels with the human ecology of Park (1936), which comprises four co-evolutionary

59

and multi-dimensional components described previously.

Given the holistic, integrative and multidimensional orientation that human
ecology is conceptualized in the geography and sociology disciplines as described above,
this study adopts a human ecology framework that recognizes the relevance of these
(environment, population, and behavior) components in understanding patterns of
decentralization. Akin to the human ecology framework proposed by Park (1936), and
Meade (1977) and Meade & Earickson (2000), the human ecology framework adopted in
this study recognizes the essence of the environment, population, and behavioral (EPB)
forces shaping patterns of decentralization. Essentially, the pattern of decentralization in
a metropolitan area is conceived of as a pattern shaped by the interacting environment,
population and behavioral factors in a given region. Thus, the environment, population,
and behavior model of human ecology is a relevant model for analyzing the
transformation of the urban landscape, particularly in explaining racial differences in
decentralization in metropolitan areas. The human ecology framework offers a highly
inclusive approach despite inherent differences in applications between disciplines and
articulation of some of the key components of the framework.

For example, Catton (1994) asserts that sociologists and geographers differ in
their interpretation of ecological concepts. Speciﬁcally, geographers and sociologists
have different interpretations and conceptualizations of environments. For geographers,
environment means the physical and biotic aspects and they emphasize the integration of
the physical and social elements in the analysis of human behavior, processes, and
phenomena. For sociologists, speciﬁcally the ecological school of Park et al. (1925), the

environment meant the social and cultural surrounding a person or group, not necessarily

60

the land water, air and biotic factors of the environment (Catton, 1994; p. 84). Catton
basis this distinction on the projects that sociologists pursued, which include extensive
work on urban spatial structure and mapping of crime, ethnicity, and other economic,
social, and demographic variables. However, Catton clearly misses the original
interpretation of human ecology model, which speciﬁcally identiﬁes the physical/natural
environment as one of the four interacting components of the classical ecological model,
and to reiterate include: population, artifact (or the technological culture), customs and
beliefs (or the non-material culture), and natural resources (Park, 1936; p. 15).
Regardless of differences in interpretation of key components of the human
ecology framework, one common view for scholars from both geography and sociology
disciplines is that human ecology is a holistic and inclusive approach, and is a relevant
framework for understanding complex processes of human systems such as urban growth

and decentralization in metropolitan areas.

Understanding Patterns of Decentralization using a Human Ecology Framework
The pattern of decentralization in metropolitan areas is a result of dynamic,
complex, and interdependent interactions and relationships between people and the
environment. Because people structure and interact with their environment and
environment facilitates in the formation of human behavior, an understanding of how
they interact and results in decentralization necessitates viewing decentralization from an
integrated perspective. Viewed in this vein, human ecology described in the previous
section, is a relevant framework for structuring an analysis of the relationship between

local roads (spending and stock) and the decentralization of people in metropolitan areas

61

while accounting for population (socio-economic characteristics), behavior (social
structure), and the environmental (local roads, locational) factors.

Therefore, this dissertation adopts human ecology as a methodological
framework. The framework structures an analysis of the relationship between local roads
and patterns of decentralization as a phenomenon resulting from the dynamic human and
environment relations. Further, it enables an analysis that integrates three theoretical
perspectives of residential mobility and locational outcome using the segregation and
residential mobility literatures within the social geography sub discipline. These
perspectives include: ecological or spatial assimilation, neighborhood preference and
place stratiﬁcation perspectives. As reviewed in Chapter Two, these perspectives and
their empirical application provide relevant background for the development of the
conceptual and empirical models of this dissertation research. The review of these
perspectives helped identify the proposed causal factors of patterns of decentralization in

metropolitan areas.

The Conceptual Model
Following the theoretical perspectives for decentralization and residential patterns
described in Chapter Two, the ecological or spatial assimilation, place stratiﬁcation, and
neighborhood preference theoretical perspectives are weaved into the role of local roads
and government policies on local roads development. The integrated perspective enables
a broad-based conceptual human ecology framework of Environment, Population, and
Behavior (EPB) that is relevant for explaining the relationship between local roads and

patterns of decentralization by race. The model of decentralization by race weaves

62

concepts of local community attractiveness via roads stock, roads spending, and distance
from the CBD as amenity and locational input factors, behavioral concepts via
neighborhood preference and racial discrimination variables, into the ecological concepts
of competition and spatial assimilation via socio-economic status as spatial mobility
factors.

Figure 3.1 describes an Environment, Population, and Behavior (EPB) human
ecology conceptual model that reﬂects three critical and interrelated components that
underlie patterns of decentralization by race in metropolitan areas. These model
components reﬂect the underlying theoretical explanations of population decentralization
that recognize the relevance of local roads as an environmental amenity variable as well
as locational factors measured by distance from CBD (i.e., distance from downtown
Detroit) and other existing population (i.e. socio-economic factors) and behavioral (i.e.
neighborhood preference and racial discrimination) explanations of the decentralization

phenomenon.

63

 

Population Component
Socioeconomic Status (Income, Education)

      
   

 

V

   

Patterns of Decentralization

 

 

by Race
Built Environment Component Behavior Component
Amenity and locational factor Attitude/Preference
(Local roads spending, location (White homeownership, proportion of non-
preference—distance from CBD) whites)

 

 

 

Figure 3.1. A Conceptualized Relationship between Local Roads and Population
Decentralization

Based on the theoretical explanation of the relationship between local roads
spending and differential levels of decentralization of people by race, the conceptual

model can be speciﬁed as:
D,-, = f (ROADit, POP), , SEC” , 83); where, (3.1)
D), is an indicator of decentralization (Population density change, percent change in

relative concentration or representation of non-whites using the Location Quotient

approach, and non-white population density change) at the local level i in a given time t;

ROADi, is local roads stock and spending and represent environmental amenity due to

government behavior effects on decentralization; POP” is the population characteristics

or spatial assimilative factors of the community via the socioeconomic characteristics;

64

SE C), represents neighborhood preference and place stratiﬁcation factors via

demographic composition and representation of white homeownership; a,- is the random

error or the remaining variability in a given community i; and t denotes a time period in
year.

As Equation 3.1 shows, understanding the process of population decentralization
is a complex matter and requires a multi perspective approach that recognizes population,
behavior, and environmental factors as speciﬁed in the human ecological framework
adopted in this study. Each component or dimension of the model is explained further in
subsequent paragraphs. The conceptual model posits that the decentralization of people in
metropolitan areas in general, and by race in particular, is shaped by past condition of the
interrelated causal factors that reﬂect the multi-dimensional human ecology conceptual
model adopted in this study. This means that lag value of the speciﬁed causal factors —
population via socioeconomic characteristics, behavior via neighborhood preference and
place stratiﬁcation factors, and environment via local roads and spending, and locational
factors —— explain the decentralization of people in metropolitan areas. Essentially, the
EPB conceptual human ecology model highlights that patterns of decentralization are
shaped by environmental factors, the competitive and assimilative capacity of a
population group, and behavioral factors. The expectation is that different levels of
participation by different racial groups should reﬂect these three conceptual components
and their identiﬁed proposed explanatory factors. The model and the different

components are described below.

65

Model Component 1
The Built Environment (Local roads spending, Local roads, Locational factors)

The environment component of the human ecology model adopted in this study
reﬂects the relevance of the built environment as a vital consideration for identifying
selecting residential or business location. The choice of residential location is affected by
many spatial and locational factors including the internal structure of the built
environment, which is shaped by roads network. As described in the introductory chapter,
local roads provide basic infrastructures for local communities and their development and
maintenance to optimal standard are added amenities that affect intra and inter-
community accessibility, property value, and quality of life particularly for the sense of
safety that these evoke upon community residents and strangers alike.

In communities where local roads investment is substandard, the quality of roads
services is more likely compromised. Accordingly, low quality of road services affects
the relative attractiveness of a given community for residential or job/industries
relocation. Given the sensitivity of roads quality and services on investment patterns,
local roads spending can be a critical component in community formation and growth.
Local roads spending should affect the relative attraction and retention of people and
industries.

It follows that variability of local roads spending across communities in a
metropolitan landscape should affect the structure of the built-environment, patterns of
distribution of people, the density and intensiﬁcation of land use, and the culture and
lifestyle (behavior) that are promoted and upheld by the design and form of roads

network established. Thus, local roads and the role of government spending are relevant

66

factors that are expected to contribute to the pattern of growth and decentralization of
people in a given metropolitan area.

Another factor that affects patterns of decentralization is locational attributes,
speciﬁcally the relative location of communities with respect to a central business district
(CBD). The relative location of communities, whether a central city, inner ring suburban,
or exurban community, may provide an indication of relative attractiveness of such
communities for residential and industrial relocation. Since suburban communities have
become the destination communities for many individuals and businesses, it is highly
likely that the location of community with reference to a CBD is an important
determinant of pattern of grth and decentralization in metropolitan areas.

These two factors local roads (stock and spending) and locational attributes are
environmental conditions that individuals and businesses consider in relocation decisions.
Patterns of growth and decentralization in metropolitan areas should reﬂect these two
factors.

Model Component 2- Population, Socioeconomic status (Spatial Assimilation
Perspective)

As presented in Chapter Two, patterns of decentralization by race reﬂect
variations in individual’s or group’s capacity for spatial mobility as the ecological or
spatial assimilation perspective asserts. This means that higher levels of socioeconomic
status, measured in income, education, or occupation should increase the capacity of
individuals and population groups for moving into suburban neighborhoods where
relatively, better amenities and quality of life are often available. For minorities or non-

whites, higher socio-economic status should enable them social mobility, subsequently,

67

spatial mobility. This means that minorities would have a greater Opportunity for
participating in decentralization and eventually, assimilation with the majority white
population group.

Model Component 3--Behavi0r

Social/Racial preferences (Place Stratiﬁcation and Neighborhood Preference
Perspectives)

While spatial mobility is contingent upon socio-economic status as the ecological
and spatial assimilation perspective asserts, certain behavioral conditions may facilitate
or impede patterns of mobility of individuals and groups particularly those individuals
and population groups who may perceive hostility from their potential neighbors in a
residential context. Therefore, the participation of minorities in the patterns of
decentralization should reﬂect the variation of these behavioral causal factors across
communities in a given metropolitan region.

As the neighborhood preference asserts, the demographic composition of a
residential location. The demographic composition affects the relative attractiveness of a
community for residential location. Therefore, the relative attractiveness of a given
community should reﬂect preferences for either homogeneous or diverse communities.
On the other hand, structural forces may constrain individuals or groups from
participating in the decentralization process.

As presented in Chapter Two, discrimination and prejudice in destination
communities and other informal and institutional forms of discrimination are behaviors
that impede minorities from relocating in communities where these behaviors are
perceived to exist. Therefore, preference for certain demographic characteristics and

structural forces should affect the pattern of decentralization in metropolitan areas. These

68

factors signify the relevance of behavior in the patterns of decentralization. More
speciﬁcally, they provide alternative but complementary explanations to the spatial
assimilation, and environmental or amenity-based explanations of patterns of
decentralization.

The conceptual model described above theoretically speciﬁes the effects of the
population, behavior, and built environment on decentralization of people in metropolitan
areas. It speciﬁcally proposes that patterns of decentralization by race is explained by
environmental factors or local roads (spending and stock), socio-economic factors or
income and education, demographic composition or the percent non-whites, racial
discrimination or proportion of white homeowners, and locational factor or diStance from
a central business district. The conceptual model is tested in the context of the Detroit

Metropolitan Area using cities and villages as units of observation.

Hypotheses of the Study
This research will test the following hypotheses:
0 municipalities with higher roads stock and road spending are expected to
facilitate increases in population and that increase in population will

contribute to the decentralization of people;

0 Communities with higher local roads spending and road stock, higher
representation of non-white population, lower representation of white
homeowners, higher income and education, and located increasingly away
from a central business district are more likely to increase the

representation of non-white population, and that increase in population

69

contributes to increasing level of participation of non-whites in the

decentralization of a metropolitan area.

Whether the two hypotheses presented are supported in the context of the Detroit
MSA is the subject in the remaining chapters. The next chapter presents the empirical
model of the relationship between local roads and decentralization. Informed by the
literature on segregation and residential patterns, the empirical model of decentralization
is speciﬁed, and is integrated into the human ecological framework adopted in this

dissertation study.

70

CHAPTER FOUR: THE EMPIRICAL MODEL

The Enviromnent, Population, and Behavior (EPB) human ecology conceptual
model developed in this dissertation research and described in Chapter Three provides a
rational for developing an empirical model of the relationship between local roads and
decentralization. The empirical model highlights decentralization by race as a process
shaped by improvements in the urban built enviromnent through local roads spending,
and distance from the CBD as a locational factor affecting preference for either central
city or suburban relocation. The empirical model also emphasizes socioeconomic status
as population characteristic that determines spatial assimilation; demographic
composition as determinants of an individual or a group’s preference for certain
neighborhoods; and representation of whites’ homeownership as proxy for attitude of
discrimination against certain population groups. The two variables, demographic
composition and representation of white homeowners are behavioral factors affecting
patterns of decentralization of people by race as the neighborhood preference and place
stratiﬁcation perspectives assert.

Consistent with the conceptual model, the empirical estimation speciﬁcally
involves testing for the effects of local roads (roads stock, road spending) on the
differential levels of decentralization across population groups measured in non-white
population density change and relative concentration of non-whites using the Location
Quotient (LQ) approach. The model speciﬁes that decentralization is a function of road
spending and road stock, income, education, proportion of white homeowners, percent
non-white population, and distance variables. The empirical model is developed using

relevant studies of decentralization within the literatures in segregation and residential

71

llli
de.

10C

pattern presented in Chapter Two. The empirical model speciﬁes causal factors of
decentralization and estimates the model in the context of Detroit MSA. The empirical
estimation is based on aggregate measures using the municipality (cities and villages) as
units of analysis. l

Chapter Four is organized into two sections. The ﬁrst section describes the

empirical model. The second lists speciﬁc aims of the dissertation study.

Model Specification

The process of central city out-migration as discussed above is shaped by
complex interdependent factors involving environmental, population, and behavioral
factors. The built environment and its attractiveness for household and industrial location
contribute to local community growth. Therefore, government spending on local
inﬁastructure should facilitate the attractiveness of a given community for household
relocation.

The importance of local roads in a community’s attractiveness for residential
relocation goes beyond spatial beneﬁts (Yago, 1983). Local roads impact residents’
quality of life in terms of their everyday experiences with trafﬁc and congestion,
pedestrian safety, pollution and noise (Marans & Sheehan 2003). Moreover, local roads
enable connectivity and walkability. On the other hand, local roads are used to separate
people into certain neighborhoods and increase the social stratiﬁcation by race, class, and
income (Grannis, 1998). Therefore, local roads can be seen as facilitators of
decentralization but at the same time deterrents to residential integration. It follows that

local roads and the quality of the service these provide are desirable amenities that shape

72

mobility and accessibility patterns in a given region with few exceptions. The empirical
evidence that this study hopes to generate should shed insights into the impacts of roads
on the decentralization patterns in metropolitan areas.

Aside from the inﬂuence of roads in the development of the built environment,
population characteristics (socioeconomic factors) determine capacity for spatial mobility
of individuals and households and are very critical in explaining patterns of
decentralization. Socioeconomic factors such as income and education determine social
mobility, hence, spatial mobility as the spatial assimilation perspective asserts. On the
other hand, evidence suggests that race distorts the ecological theory and the stark spatial
and social separation between majority whites and people of color in residential spaces
exists regardless of socioeconomic status (Alba & Logan, 1993; Darden, 1989; F arley,
1979; Madden, 2003). Accordingly, blacks with higher level socioeconomic status are
still segregated in suburban areas in as much as those blacks who are living in central
cities (Darden, 2003; Darden & Kamel, 2000).

Therefore, the empirical evidence shows that while ecological theory provides a
consistent explanation for the spatial distribution of people across metropolitan
landscapes for some groups, it cannot fully explain alone the differential levels of
decentralization of people and their locational pattern.

Preference for certain locations on the basis of demographic characteristics, and
racial discrimination and prejudice explain the different levels of decentralization by race.
To determine the explanatory power of these variables on decentralization, researchers
have often used proxy variables in lieu of more direct measures that are not easily

available unless a sophisticated approach such as paired-testing is employed (National

73

Research Council, 2002). The percent non-white and proportion of white homeowners
are used as proxy variables of the neighborhood preference and place stratiﬁcation
explanation of decentralization. These variables are approximation of preference for
either homogeneous or diverse residential location, and proportion of white homeowners
as measure of racial discrimination.

It follows that the spatial distribution of communities with better local roads stock
and investment may reveal the location and direction of growth in a metropolitan area.
One can therefore explain, via regression analysis, the causal relationship between the
differential levels of decentralization and local road spending and road stock (road
density) while accounting for income and education as ecological or assimilation
variables, percent non-whites as the neighborhood preference variable, and percent white
homeownership measured as proportion of white homeowners for the housing
discrimination as a place stratiﬁcation variable. A distance variable is included in the
model to account for the effects of location as a determinant of preference for either
central city or suburban location. More speciﬁcally, it measures the separation effect of
distance on the pattern of decentralization of population groups based on race. A fully

speciﬁed model of decentralization is presented below.

ADecentralization it =f(Road spending and density”. 1);, Income“. 1); ,Education(,.1),- ,
%Non-whites(,.1),~, Proportion of white home owners(,.1),- ,

Distance, 8;); where, (4.1)

ADecentralizationit uses indicators including percent change in total population density,

percent change in non-white concentration, and percent change in non-white density for

74

community i at a given time period t, between 1980 and 1990 which are years that many
central cities in the Northeast and Midwest have had drastic decline in population. These

are indicators of local community attractiveness or absorption capacity of a given
community; Road spending and road density“. 1);, are amenity and policy effects
variables including state allocated road ﬁmding, locally-raised road funds, and local roads

density; Income“. 1)i and Education“. 1);, are variables for socioeconomic status that

affect spatial mobility; % Non-whites". 1); is a variable for neighborhood composition
reﬂecting preference for either homogeneous or diverse community; proportion of White

home owners(,.1),- is a proxy variable for housing discrimination; and Distance is a

variable for locational factors that determine preference for either central city or suburban
relocation; e,- is random errors or remaining variability, and t-I represents previous or

lagged values of variables in the model.

Using the standard multiple regression and spatial regression approaches, the
empirical model estimates decentralization using three measures of decentralization
described previously, as a function of lag values of local roads spending (state-allocated
and locally-raised) and local road density while accounting for socio-economic factors
income, education, percent white homeowners, percent non-whites, and distance from the
CBD. The multivariate model speciﬁcation provides an estimate of the effects of local
roads on decentralization in the context of the Detroit Metropolitan Area.

The lag speciﬁcation of the model accounts for the effects of past conditions or
values of explanatory variables on each of the three measures of decentralization. The lag

speciﬁcation is used recognizing that the effect of these explanatory variables on pattern

75

of decentralization takes some time. The empirical model is tested in the context of the

Detroit Metropolitan region to accomplish the following speciﬁc aims of the study.

Specific Aims

Having speciﬁed the empirical model, the study aims to:

0 develop a conceptual model to explain the relationship between local
roads and decentralization, and test this model in the context of the Detroit
Metropolitan region using a human ecology methodological framework,
accounting for income and education as ecological and spatial assimilation
variables, percent non—white as neighborhood preference variable,
proportion of white homeowners for place stratiﬁcation variable, and
distance between a municipality to the Detroit CBD as a spatial or

locational preference variable.

0 characterize the pattern of decentralization in the study region using three
measures of decentralization, which include: 1) percent change in
population density of the general population; 2) the percent change in the
relative concentration of non-whites; and 3) percent change in non-white

population density,
0 characterize the pattern of local road spending across the region; and

0 determine links between local roads and patterns of decentralization for
the general and non-white population while accounting for other proposed

predictors including household median income, percent non-white with

76

college education, proportion of white homeowners, demographic

composition, and distance from the CBD.

The next chapter describes the Detroit Metropolitan Area as the case study area. It
also describes the data and methods employed in the research, and the analytical

approach.

77

CHAPTER FIVE: DATA AND METHODS

The primary goal of this dissertation research is to understand the possible
inﬂuence of local roads on the decentralization process in metropolitan areas within a
social geographic inquiry. As described in Chapter Three, the research conceptualizes the
relationship between local roads and decentralization using a human ecology
methodological framework that integrates environment, population, and behavioral
attributes as proposed explanation for racial differences in decentralization. To
implement the empirical model described in Chapter Four, this chapter introduces the
data, methods, and estimation procedures.

The chapter is divided into three sections. The ﬁrst section describes the study
area and the general considerations for choosing the area, the study period, and the unit of
analysis. The second section describes the variables, data used and their sources, data
processing, and organization procedure. The third section describes the estimation
approach adopted for determining the relationship between local roads and

decentralization. A brief summary of the data and analytical approach end the chapter.

The Study Area
Since the latter part of the twentieth century, metropolitan areas in the US. have
become home to more than 50 percent of the entire population in the U. S. By 2000,
metropolitan areas absorbed about 80 percent of the entire U. S. population of 281 .4
million residents, and have become increasingly characterized by distinct demographic

distribution (Hobbs & Stoops, 2002; p. 33).

78

All across the US, 75 percent of metropolitan residents are whites, while about
25 percent are non-whites (Hobbs & Stoops, 2002; p. 77). Indeed, non-whites, who are
represented by blacks and all other individuals from other races, including foreign-born
residents tend to have a strong preference for metropolitan locations (US. Bureau of the
Census, 2000; Kent et al., 2001). Their presence and concentration in metropolitan areas,
more speciﬁcally in central cities, makes a metropolitan area a relevant geographic
context for understanding the different patterns of decentralization of the general and
non-white population groups. The possible causal relationship between local roads and
population decentralization is examined in the context of the Detroit Metropolitan Area in
Michigan as the case study area.

Also known as the Detroit Metropolitan region or the Detroit Tri-County Area,
the Detroit Metropolitan Area comprises the Wayne, Oakland, and Macomb counties
with 131 communities located within this region. These communities include 75 cities, 15
villages, and 41 townships with Northville and Richmond straddled between two counties

(Figure 5.1).

79

Sour

 

 

“-

 

Washtenaw
Legend

— Interstate Highway

D County

   
 
 

 

 

 

0 25 5 londﬂes
F+4—+4—+4—+4

 

 

 

 

Figure 5.1 The Study Area

Source: Developed by the author from accessible maps in Michigan Geographic
Data Library, Center for Geographic Information.

80

General Considerations for Selecting the Case Study Area

While there are many regions that are relevant for determining the causal
relationship between local roads and the decentralization of people by race, the Detroit
Metropolitan region offers a particularly more compelling context for unraveling
different causes of decentralization. The Detroit Metropolitan region represents the most
decentralized region in the state of Michigan with an unprecedented rate of
suburbanization of the white population. Also, the region is characterized by uneven
development, a high incidence of concentrated poverty in the central city, and a
residential pattern that starkly separates people by race and class. Aside from the
population dynamics, land consumption in the region far outpaces population grth by
an average ratio of 13 to 1 between the years 1960 and 1990. This means that the land
consumption in the region was 13 times faster than the region’s population grth
between those years (Michigan Land Use Leadership Council, 2003).

The distinct demographic distribution, unprecedented land consumption, and a
sustained uneven development, incidence of concentrated poverty makes the region a
highly relevant context for understanding complex causes of decentralization. These
circumstances in the Detroit Metropolitan region require an understanding of the role of
government policies, particularly policies on local road development that are more likely
to have contributed to the decentralization process.

As the most urbanized region in the state, the Detroit Metropolitan region has an
extensive road infrastructure. A total of 17,949 miles of primary and local roads, trunk

lines, and freeways are developed and maintained by three separate county road

81

dc

m

Re

leg]

the

a .

tlp
berm:

commissions in Oakland, Wayne, and Macomb counties.l This roads stock represents 65
percent of the total roads in southeast Michigan and show the intensity of the built
enviromnent in the region.

In light of the pattern of road development and the signiﬁcant population shifts
that have been occurring in the region, the Detroit Metropolitan region offers an excellent
opportunity for unraveling the inﬂuence of local road spending and decentralization. The
analysis of the relationship between these processes reveals additional insights into the
deep inequality in the allocation of road funds across communities and its possible link to
the decentralization process. A more detailed description of the circumstances in Detroit

Metropolitan region is presented in the next paragraphs.

Regional Population Trends

The Detroit Metropolitan region is the most urbanized and populous region in
southeast Michigan with a land area of 1,967 square miles and 183 square miles of water
(SEMCOG, 2002). Population shifts in the Detroit Metropolitan region reﬂect the
dynamic growth and development in the region over time as Figure 5.2 shows. The
region experienced an unprecedented population grth and development particularly in
the ﬁrst three quarters of the twentieth century. In the 703 and 803, however, the region
experienced population decline at the rate of 3.8 between 1970 and 1980, and 3.3 percent
between 1980 and 1990. But by the 19903, the region gained population by 3.3 percent,

which was equivalent to what it had lost in the previous decade (Hobbs & Stoops, 2002).

 

' This information is available at the Southeast Michigan Council of Government website
at http://www.semcog.org/Data/Apps/comprof/transportation.cfm?cpid=8999 last accessed on
August 15, 2009. The southeast Michigan region, which comprises Wayne, Macomb, Oakland,
Washtenaw, Monroe, Livingston, and St. Clair counties, has 28,000 miles in road stock.

82

Current estimates indicate that the Detroit Metropolitan region 0.78 percent of its
population between 2000 and 2008. Wayne County lost an estimated 4.2 percent of the
entire metro population (SEMCOG, 2008). Indeed, as of July 2008, the Southeast
Michigan Council of Government estimated that some 86,020 persons have left Wayne
County for other locations since 2000 (SEMCOG, 2008).

While the population trend in the Detroit Metropolitan region as a whole may
suggest a grim condition, the intraregional dynamics presents a different context. While
Wayne County experienced a signiﬁcant population loss since 2000, Oakland and
Macomb counties experienced population gains at the rate of 0.9 and 5.6 percent,
respectively. It follows that the population loss in the Detroit Metropolitan region is
largely attributed to the losses in communities in Wayne County, particularly the city of
Detroit, which has experienced a steady population decline since the 19503 (SEMCOG,

2008).

4,500,000
4,000,000 1
5 3,500,000 1
:3 3,000,000
32,500,000 1
3 2,000,000 -
5 1,500,000 ~
.3 1,000,000 -
500,000
1890 1900 1910 1920 1930 1940 1950 1960 1970 1900 1990 2000 2010

Year

 

 

Figure 5.2 Detroit Metropolitan Area Population Trend, 1900-2008
Source: Southeast Michigan Council of Governments (2002, 2008)

83

ml

ice

He,

up

dt‘q‘

llm

The City of Detroit. The City of Detroit is located within Wayne County and is
the largest city in the region and the State of Michigan with 951,270 residents as of 2000
(US. Bureau of the Census, 2000). As one of the aging industrial cities in the Midwest,
the city of Detroit has been experiencing a signiﬁcant population and job losses. As of the
2008 population estimate, the city has 860,521, which was less than half of the
population in the 19503 when the city experienced its highest centralization point of 1.85
million people (SEMCOG, 2008). The signiﬁcant population 1033 in the city resulted in
high rental and home vacancy rates, making the city of Detroit America’s second most
abandoned city (Greenburg, 2009).

The high rate of housing vacancy and abandonment in Detroit is one of many
manifestations but also a cause of other more complex and reinforcing outcomes of
decentralization. Land use change and uneven development, economic and residential
segregation, joblessness, and the concentration of poverty have all become major
distinguishing features for the city of Detroit. Essentially, the steady decline in the city
created many economic, social, and environmental woes that are disproportionately

experienced by different segments of the metro population.

Land Use Change and Pattern of Uneven Development

A signiﬁcant outcome of the unprecedented population growth in the city of
Detroit during the ﬁrst half of the twentieth century was the centrality of the pattern of
development in the region. Detroit attracted people from across the country and around
the world due to available jobs in the car manufacturing and related industries (Darden et

al., 1987; Teaford, 1993). During the second half of the twentieth century, however, the

84

C011

Cor

squa
{our
1610
dam:
30051
the in
0000‘
160110
1993).
P311130.

need 0

111mm,

.

concentration of housing production and jobs shifted toward the peripheries.
Consequently, the metropolitan landscape signiﬁcantly changed.

In the 19903, an estimated 33 percent of the region’s land area of about 1,967
square miles was already urbanized (Limoges, 2001). This urbanized area was then about
four times the size of the urbanized area in the 19503. This signiﬁcant change in land use
reﬂect primarily, the conversion of tracts of land into urban uses in response to increasing
demand for housing, infrastructure, and other institutional support services.
Subsequently, the increasing suburban development, signiﬁcant industrial shift toward
the fringes occurred, followed by the retail industries (Darden et al., 1987). Thus, the
Detroit Metropolitan region was transformed from a highly centralized to a less
centralized urban region and has since been characterized by “edge cities” (Garreau,
1992). People and capital ﬂow towards the ﬁinges of the city became the dominant
pattern of mobility for people and industries in the region, leaving the central city in dire

need of economic and physical revitalization.

Income Disparity

The Detroit Metropolitan region is characterized by a deep disparity in income
between counties and cormnunities. For example, the per capita income for Wayne
County in 2000 was $20,058, which was below state level per capita income of $22,168.
Oakland and Macomb counties, on the other hand, were both above the state level with
$32,534 and $24,446, respectively. The median household income of Bloomﬁeld Hills
City in Oakland County was $105,000 and was ranked 17th in the country. Birmingham

City in Oakland County was $37,061. On the other hand, Ecorse and Dearborn cities in

85

d1

10:1

the

Wayne County had a per capita income of $9,781 and $21,488, respectively (US. Bureau
of the Census, 2000).

Refer to Figure 6.6, which illustrates the trend of income distribution in the region
by communities and with respect to distance from the CBD of Detroit. This income
disparity across communities in the region reﬂects the spatial distribution of people by

race in the region.

Racial Segregation

Residential segregation in the Detroit Metropolitan region is one of the highest in
the county with blacks or African Americans, experiencing the highest level of
segregation from whites. This means that blacks are segregated the highest from whites
compared to Asians, Hispanics, and other racial groups (Darden et al., 1987; Rusk, 2003).

The pattern of high levels of segregation between blacks and whites in the Detroit
region has been consistent. Between 1940 and 1970, the white-black racial segregation in
the Detroit Metropolitan region was high, with index of dissimilarity that ranged from
83.7 to 88.9 (Darden et. al., 1987; p. 87). A decrease in segregation was noted in the
19803 (85.8), down by 3.1 percent from the previous decade. Similarly, Rusk (2003)
indicated a decline in segregation in the region when he found a dissimilarity index of 81
in 2000, down by 2 percent points from the previous decade (1990). These decreases in
levels of segregation, however, do not readily translate into improvements in the quality
of life especially for blacks. These population groups continue to experience high levels

of isolation not just in the residential space but in educational and other spaces of

opportunity.

86

School segregation is markedly high in the Detroit Metropolitan region and is one
of the highest in the country (Frankenberg et al., 2003; Rusk, 2003; Orﬁeld, 2002). The
index of dissimilarity was 82 percent in 1997 (Orﬁeld, 2002; p. 52) and 89 percent as of
2000 (Rusk, 2003; p. 40). These indices simply show that the demographic composition
of schools closely resembles the residential patterning in the region. They reveal not just
the inequality in residential and educational opportunities but also disparities in access to
and participation in jobs and other opportunities. In essence, residential and school
segregation are inextricably linked with patterns of job and workforce distribution in the

region.

Job Distribution

Job distribution in the Detroit Metropolitan region is highly decentralized and is
one of the extremely decentralized employment centers in the country (Glaeser et al.,
2001). Across 100 largest metropolitan areas in the country, an average 22 percent of the
jobs are located within 3 miles of the CBD while 35 percent are located beyond 10 miles
of the CBD. In Detroit, only 5 percent of the jobs are located within 3 miles while 78
percent are located beyond 10 miles of the CBD (Glaeser et al., 2001; p. 5). Further, only
22 percent of jobs are located within 10 miles, which is far below the national average of
65 percent (Katz et al., 2003). Indeed, while Detroit was still the leading employment
center in the Detroit Metropolitan region in 2000, its share of the region’s employment
has dropped from 38 percent in 19703 to 13 percent in 2000 (SEMCOG, 2002).

In light of these circumstances in Detroit Metropolitan region, the region has

become associated not for its production of the Model Ts that put the city on the map and

87

the
cm

100

hr.
10 b1
ng
lie n
incre

lOm

H

the minds of people around the country and the world, but for its production of edge
cities (Garreau, 1992). In Detroit, the centrality and function of cities in the monocentric,
modern period became increasingly superseded by multi-centric urban form as new
developments — shopping centers, ofﬁce complexes, schools and other institutions — in
the suburbanized areas created new patterns of how people live, work, and play in the
metropolitan region.

Given the situation in the Detroit Metropolitan region, there is a critical
opportunity for studying mechanisms and processes underlying income disparity,
segregation, and patterns of decentralization. This research aims to contribute to an
understanding of these processes by examining the effects of local road spending on
decentralization and how government policies may have sustained these effects over
time. To understand these effects, the dissertation research carefully identiﬁes the
geographic and temporal context, unit of analysis, variables and measurement, data,

methods and analytical approach. These are presented in subsequent sections.

The Study Period and Communities
The analysis of the relationship between local roads and decentralization focuses
between 1980 and 1990, which were years that decentralization in major cities continued
to be prominent, particularly in the Midwest (Berube & Forman, 2002; SEMCOG, 2002).
Figure 5.2 shows the population trend in the Detroit Metropolitan region and highlights
the marked decline in the region’s population between 1970 and 1990. The steady
increase in the region’s population prior the 703 is noteworthy. It conﬁrms the region’s

dominance in attracting people, and as a center of the car manufacturing industry in the

88

country. Whether the change in population trends was affected by government policies
on local roads raises questions about the unintended consequences of government
decisions. By focusing on the decade of distinct population decline in the region, the
study presents a critical opportunity for determining how government policy contributed
to the pattern of decentralization. The analysis focuses on 88 cities and villages, out of
131 communities that comprise the region. Forty one (41) townships were excluded from
the analysis because townships do not have road jurisdiction in Michigan and are
therefore, excluded in the distribution of government road funds. Appendix A lists the

cities and villages in the study region.

The Unit of A nalysis and Observation

The relationship between local roads and population decentralization is examined
using the municipalities (cities and villages) as units of analysis. The study then makes
generalizations on the relationship between local roads and decentralization using
information that best characterize and measure the pattern of decentralization and its
proposed causes at the municipality level. The municipality is the ﬁnest geographical unit
that local roads stock and spending variables are documented, particularly in the State of
Michigan (Michigan Department of Transportation Series Report No. 139 and 162).
Therefore, the analysis of the impacts of roads on the decentralization process is limited

at this level of aggregation.

89

Variables and Measurement
To reiterate, the empirical model of the relationship between local roads and
decentralization in metropolitan areas and other hypothesized predictors of

decentralization within a social geographic inquiry speciﬁes:

ADecentralizationit =f(Road spending and density(,_1),°, Income(t.1),°, Education (t. 1);,
%Non-white3(,_1),', Proportion of white home owners(t_1),-,

Distance, 8;); where, (5.1)
ADecentralizationit uses indicators including percent change in total population density,

percent change in non-white concentration, and percent change in non-white density for
community i at a given time period t, between 1980 and 1990. These are indicators of

local community attractiveness or absorption capacity of a given community; Road

spending and road density(,.1),-, are amenity and policy effects variables including state
allocated road funding, locally-raised road funds, and local roads density; Income(t_1),-
and Education(,_1),-, are variables for socioeconomic status that affect spatial mobility; %
Non-whites”. 1),- is a variable for neighborhood composition reﬂecting preference for

either homogeneous or diverse community; proportion of white home owners(,_1); is a
proxy variable for housing discrimination; and Distance is a variable for locational
factors that determine preference for either central city-suburban relocation; e,- is random

errors or remaining variability, and t-I represents previous or lagged values of variables
in the model.
The selection of variables in the empirical model was guided by the existing

literature on decentralization within the residential mobility and segregation literature

90

domains. The model is tested to determine the decentralization of the population in
general, and the differential levels of suburbanization between races in particular using
population density and concentration of non-white population as dependent variables.
Each dependent variable is regressed on local roads stock and spending, income,
education, percent white homeownership, percent non-white, and distance between a

central city and community as independent variables in the proposed empirical model.

The Dependent Variable

Population Decentralization

As presented in Chapter One, the decentralization in metropolitan areas is oﬁen
described as a pattern of urban transformation characterized by an uneven population
grth across the metropolitan landscape. A decentralized region exhibits a stagnant,
slow growth, or a signiﬁcant population decline in the central city resulting in a
decreased population density on one hand, and an increase in population and population
density in communities outside of the central city on the other.

Although many indicators of decentralization have been identiﬁed (Galster et al.,
2001; Ewing, 1997; Lopez & Hynes, 2003), previous studies have consistently indicated
the relevance of population density used either singly (Lopez & Hynes, 2003) or in
combination of many other dimensions (Galster et al., 2001; Ewing, 1997). A
conventional approach for assessing patterns of decentralization is by estimating a
population density change (an increase or decrease) or density gradient within a
metropolitan landscape (Jordan et al., 1998; Mieszkowski and Mills, 1993; Carlino &

Mills, 1987). A decrease in population density in a central city accompanied by an

91

increase in population density outside the conﬁnes of the city demonstrates a
decentralized pattern of growth in a metropolitan area.

Population Density Change. A simple equation of population density change in a
given community (6. g. city or village) i between 1980 and 1990 is presented in Equation
5.2. The equation speciﬁes population gains, when the rate of change is positive and a
population 1033 if otherwise. Further, it assumes that the land area is the same between
1980 and 1990. Equation 5.2 also suggests the attractiveness of community i for
residential location. A positive value indicates an increase in population for community i
and may suggest that community i is a “destination community” in a given metropolitan
region. On the other hand, a negative value indicates a population 1033 for community i,
which may indicate the unattractiveness of community i for residential location.

Population density change of community i takes the form:

 

P - P -
APOPDeniJ980—l990 = 01" - & Where (52)
Am" 1990 Am“ 1980

APopDeni, 1980,1990 is the change in population density at community i between 1980 and
1990 or the change in number of individuals per unit area or space; Popi is population

count or number of individuals in community i; and Areai is the total area (e. g. in square

miles) of community i.

Differential Levels of Decentralization by Race. Equation 5.2 provides a general
measure of the population dynamics in the region. The population dynamics is further
understood by isolating levels of decentralization of people by race in metropolitan areas.

It does so by speciﬁcally measuring the relative concentration or representation of the

92

non-white population group across communities in the region using the Location
Quotient (LQ) approach.

The Location Quotient (LQ) is an approach in Economic Base Analysis (EBA)
that has been increasingly adopted in describing patterns of demographic and socio-
economic processes in metropolitan landscapes, including residential and segregation
patterns (Chung & Brown, 2007; Brown & Chung, 2008; Ray & Bergeron, 2007) and in
crime incidence analysis (Zhang & Peterson, 2007; Guhathakurta & Mushkatel, 2000).
The LQ measures the concentration of the phenomenon of interest such as the
homogeneity or diversity of a population group in a particular locale (i.e. a block, census
tract, village, and city) relative to a larger reference areal unit. The value of LQ ranges
from less to more than 1 (0 - 00). A value of the LQ that is equal to 1 indicates that the
concentration or representation of a particular population group is the same to that of the
larger reference unit; a value of less than 1 indicates an under representation of the
population group; and a value of greater than 1 indicates an over representation or
concentration of the population group relative to the reference or larger areal unit.

Unlike the traditional global indicator of segregation such as the dissimilarity
index (DI), the LQ characterizes the local pattern of concentration of a particular
population group (e. g. non-whites) with respect to the general population of that group in
a larger areal unit (e. g. a metropolitan area). The variation in LQ across the landscape
illustrates the distribution of low to high concentration of non-whites in the region. When
mapped, the distribution characterizes a description of the relative representation of non-

whites and may represent destination communities for the non-white population.

93

The change in LQ between two time periods indicates a relative change of the
level of concentration of non—white population group in a given community between
these periods and may suggest communities that have become “destination communities”
for the non-white population group. Values derived from these measures of
decentralization are mapped to determine the spatial pattern of grth and decline, the
racial fabric across communities in the study region, as well as the change of the racial

fabric across time. In this study, the Location Quotient change for community i takes the

 

 

 

 

form:
Emu-mi /_ " Popnon-W,i /
Pop Total,i . _ POP TotaLi //
/ P op non - W. Metro * 1// Pop non-w, Metro
. PapTotal, Metro 1990 ' P0P Total,Metro 1980 ’
ALQumran-1990 = P
‘ 0P non-w,i
[ l)0!) Total,i ) ’
" P0P non-w, Metro J
3 Po
PTota|,Metro 1980
where (5.3)

ALwawgo is the Location Quotient change of community i between 1980 and 1990;
Popmmw, ,-is the total number of individuals from the non-white population group in
community i; PomeM is the total number of individuals in community i including the
non-white and white (of Caucasian origin) population group; Popnomw, Mm, is the number

of individuals from the non-white population group of the study region, and POPTota|,Me"-o

is the total number of individuals in the entire study region as the reference unit.

94

1
01

Thus, Equation 5.2 is used to describe the dynamics and intensity of the
distribution of the people in a given community. Equation 5.3 further captures the
population dynamics by characterizing the concentration and representation of the non-
white population goup for each municipality. Additionally, a third measure of the pattern
of decentralization of non-whites, percent non-white population density is used. This
measure enables the pattern of decentralization of the general and non-white population
to be comparable.

The Independent Variables
Local Roads Spending and Stock

As the conceptual model indicates, government spending on local roads is
expected to facilitate decentralization in metropolitan areas. Improvements in the built
environment due to local roads spending facilitates better intra-community accessibility,
increases in property value, and beneﬁts that have implications for quality of life. Higher
local road spending rather than lower spending in communities, therefore, will more
likely improve the quality of local road stock and services.

One must acknowledge that state-allocated local roads revenue alone may not
explain fully the mobility dynamics across communities and subsequently, patterns of
decentralization. Road building is a complicated process and different sources of revenue
may develop one or different road projects at a time. And, variation of road funds
allocation could result in variation of local roads services, which have implications to the
attractiveness of a given community for business or household relocation. Therefore, the
variation of roads funds allocation could be tied to the variation of local roads services

and would have implications for grth and land use change.

95

.11:
101
101'

101

1201

111

11g

and

Two variables are used in the model as independent variables for testing the
inﬂuence of local roads on the decentralization of people in metropolitan areas. These
closely connected variables are: l) the per capita local roads spending allocated by the
state government or the Michigan Transportation Fund dollars and the locally-raised road
funds, and 2) the local road stock in miles. The per capita local roads spending in a given
year is calculated by dividing the total road dollar allocation for the community by the
total number of people, resulting in a per person dollar road funds/spending variable. A
road density is measured by dividing total local road miles of the combined major and
local street by land area in square miles for each city and village in the study area. Both
per capita dollar spending and road density are used in the model.

To understand local roads allocation and spending in Michigan, a discussion that
focuses on the sources of revenue for local roads development across the study

communities is necessary. This is discussed in the next subsection.

Sources of Revenue for Local Roads Development and Maintenance. In
Michigan, as in other states, the allocation of the Federal and State transportation revenue
for transportation development at the State and local levels is a complicated matter
(Cadot et al., 2006). In terms of roadways expenditure, Federal monies are earmarked for
highways and local roadways development with guidelines on how funds are to be used
and what government bodies or entities are eligible to receive them. At the local level, the
operating funds for the development and maintenance of roads come primarily from the
Michigan Transportation Fund and locally-raised revenues. Other sources may come

from matching funds and borrowings through bonds within the Act 51 guidelines.

96

The Michigan Transportation Fund. In terms of the State resources, only the
Michigan Transportation Fund and the Transportation Economic Development Fund
(Categories A and F) are accessible to local communities (SEMCOG, 2002). The
Michigan Transportation Fund (MTF) formerly known as the Motor Vehicle Highway
Fund is the primary revenue for the development of roads and transportation networks in
the state of Michigan. The MTF is generated through the motor fuel taxation, vehicle
registrations, and sales of auto parts. MTF revenues are distributed following an
“external” (distribution to the MDOT, counties, and cities and villages) and “internal”
(distribution/use within cities and villages) formula as prescribed in Act 51 of 1951 as
amended under Michigan Compiled Laws MCL 247.651 Section 10. It was established
through the Public Act 51 of 1951, which governs state appropriations for interstate and
local highways and public transportation programs (Hamilton, 2003). The act was
introduced through the House Bill number 42 on January 17, 1951 and was enacted on
July 1, 1951 (Journal of the Senate, 1951).

The motivation for the passage of Act 51 of 1951 can be traced back to the
country’s interest in highway transportation to aid in defense and national development
(Pohl & Brown, 1997). It grew out of the Good Roads Federation Study in 1947, which at
that time, indicated “a need of $1 .5 billion dollars over a 15-year period to bring the
entire highway system of Michigan -— State, county, and city roads to safe and tolerable
standards” (Pohl & Brown, 1997, p. 9).

Technically, the formula-based state spending on transportation in Michigan
addressed the issues associated with the disproportionate or inequitable allocation scheme

during the years prior to the enactment of Act 51. Prior to Act 51, the allocation scheme

97

was a matter of “political haggling” and funds for road development generally accrued
favorably to rural interests (Warner, 1962). The enactment of Act 51 was a signiﬁcant
improvement to the road ﬁnancing strategy and represented a major development toward
an objective based formula free from political interference. Warner (1962) identiﬁed the
signiﬁcant contributions of Act 51 as: 1) The establishment of a new special fund, the
Motor Vehicle Highway Fund (that was to become the Michigan Transportation Fund in
later years), for isolating highway user revenues from general state revenues; 2) the
introduction of road and street classiﬁcation as a means for allocating user revenues; and
3) the promotion of cities to equal eligibility with the counties for user revenue grants.
Indeed, the Act 51 policy provided a strong reform to the distribution mechanism
underlying road allocation in the state.

Since its creation in 1951 by the state legislature, the external formula has
undergone a few changes (Welke et. al., 2000). Table 5.1 below shows the changes and
the years in which they were made. The external distribution formula of 1985 is still
applied to this date (Welke et al., 2000).

Table 5.1. Changes in the Michigan Transportation Fund Distribution Formula

 

 

 

 

 

 

 

Year MDOT (%) County (%) Cities/villages (%)
1952 44.0 37.0 19.0
1957 47.0 35.0 18.0
1972 44.5 35.7 19.8
1978 46.7 34.3 19.0
1982 41.9 37.4 20.7
1985 39.1 39.1 21.8

 

 

 

 

 

 

Source: Transportation Funding for the Twenty First Century, Welke et al., 2000

98

The “internal” formula guides the MTF distribution to cities and villages across
the state of Michigan. Population and road stock (in miles) are used to determine the
amount of dollars allocated for each place or community (SEMCOG, 2002). The amount
allocated, therefore, can be affected by the increase or decrease of population and
changes in road stock. It is also affected by broader factors such as the general MTF
annual revenues through gas tax, vehicle registration, and sales of motor vehicle parts.

Locally-raised Revenues. The second most important source of revenue for local
roads development and maintenance come from the local govermnent sources through
special tax levies, general fund appropriations, general obligation funds, and special
assessment bonds. These revenues are used to match funds from other sources for local
roads construction (Hamilton, 2003). Cities and villages could also avail of federal
allocations in connection with designated high priority and demonstration projects
speciﬁed by the US. Secretary of Transportation. However, the Michigan Transportation
Fund is the major source of revenue for road development and maintenance at the local
level. It is also used as payment for debt service for any borrowings that a city or village
incurs provided it is in accordance with the Act 51 guidelines.

One might question why road spending variable is included as one exogenous
variable in the analysis of decentralization when in fact the MTF variable itself is a result
of a formula that includes population and road stock. One might similarly ask why
population and road stock could not have been used as explanatory variable of
decentralization instead of state-allocated (MTF) road spending. Two pieces of
information must be well-understood about the state-allocated (MTF) local road

spending.

99

00‘
11;

p11

;)
H

is

lll1

11111
101

like

First, the objective of the study is to examine the role of public policy on
decentralization, which the state-allocated local road revenue (MTF) proxies. In
examining such role, it is the policy instrument or mechanism that is directly linked to the
process of interest (i.e. decentralization) and only indirectly accounts for road stock and
population not vice versa. A direct test of the state-allocated local road spending and
decentralization relationship is a direct test of the effects of policy, which one would not
have observed using the population count and road stock information. Second, it is also
important to recognize that policy makes the formula but they cannot change the
variables that are linked through the formula (i.e. population and road stock). Given
above, it is therefore relevant to use the state-allocated (MTF) local roads revenue not
population and local roads as a proposed predictor of decentralization.

One must also recognize that the state-allocated local roads and locally-raised
revenues are the major sources of funding for local roads development. The impact of
these on the decentralization of people in the study region is analyzed controlling for

other identiﬁed predictors, which are presented in the following subsection.

Income and Education

Although a range of factors explain decentralization in metropolitan areas, the
empirical evidence indicates income and education as predictors of the spatial mobility of
people in metropolitan areas (Iceland & Wilkes, 2006; Timberlake & Iceland, 2007). The
improvement in income and education among minorities, and adjustment to the “host”
culture in the case of ethnic immigrants, facilitates suburbanization, and in turn, the

likelihood of assimilation to the “major” population group (Massey, 1985). Further,

100

’T")

Ci

income and educational attainment undergird the choices, preference, and the groups’
ability to access better residential amenities and economic opportunities. These
socioeconomic factors facilitate the transition of the urban spatial and social landscape by
processes of competition, invasion, succession, and enable adaptive capacity of minority
population groups based on these “ability to pay” attributes (Park et al.,1925; Burgess,
1925; Hoyt, 1933).

For this study, the income and education variables are drawn from the Census CD
1980 and 1990, which is a data source developed by the Geolytics Incorporated based on
the Long Form census data ﬁle (Geolytics Incorporated, CD1980 & CD1990).
Unfortunately data gaps still exist despite a well-organized and sophisticated resource
such as the Census CD. These gaps are addressed in this dissertation research by using
data from the National Historical Geographic Information System statistical and mapping
data developed by the Population Center of the University of Minnesota.

Two variables that assess capacity for spatial mobility of individuals and
population groups are used. The income variable and education variables use household
median income and proportion of non—white college educated individuals who are 25
years and over, and with at least four years of college education as determinants for the
capacity of individuals and population groups for spatial mobility. Consistent with
previous studies, these variables are identiﬁed as determinants of locational outcome and

residential pattern in the metropolitan region as the spatial assimilation theory predicts.

101

Proportion of White Homeowners

While income and education are relevant factors that explain the pattern of
decentralization, these explanatory factors have become increasingly challenged in light
of the persistent trends of differential levels of suburbanization across these groups. For
example, trends of suburbanization for blacks or African Americans have dramatically
improved in the last few decades but a pattern of high segregation between blacks and
whites still persists. Such evidence challenges the “pure class” based explanation of
decentralization. The income and education reasoning purported by spatial assimilation
theorists simply cannot hold as the only explanatory factors for the differential levels of
suburbanization across different population groups. An alternative explanation
emphasizes racial discrimination and prejudice as important factors (Massey & Denton,
1993; Darden et al., 1987).

The percent white homeownership has been used as a proxy for the discrimination
and prejudice variables (Williams, et al., 2005). Percent white homeownership indirectly
measures the incidence of housing discrimination and prejudice, which have become
increasingly difﬁcult to detect and measure since the passage of Fair Housing Act in 1968
(Massey & Denton, 1993; p. 59). Essentially, the FHA promoted anti-discrimination in
local government and private sectors particularly in the real estate, and mortgage lending
and insurance industries (Y inger, 1995). However, the extent that the law has
successfully reduced discrimination and prejudice in the residential landscape has been
contentious.

Empirical evidence indicates a decline in segregation and increases in

suburbanization for minorities and ethnic groups, but the pattern of high segregation

102

particularly for blacks and Hispanics persists. This reality mars any little improvement
(decline in segregation) that has been achieved so far and increases the claims for the
validity of discrimination and prejudice as dominant factors explaining patterns of
decentralization and segregation in metropolitan areas.

For this dissertation research, the proportion of white homeowners to the
proportion of white population within each of the study communities is used as a proxy
variable for the possible incidence of housing discrimination and prejudice. The variable
represents a crude local approximation for housing discrimination as a factor explaining
the low level of non-white decentralization. The white home ownership index ranges
from a value 0 to 00, with 0 equal to no representation to high representation of white

home owners in relation to the total white residents in a community.

Percent Non-white Population

An alternative explanation to the decentralization process in metropolitan areas
highlights the relevance of neighborhood preference as an underlying cause. Generally,
the empirical evidence indicates that different racial/ethnic groups tend to exhibit an
“own race” preference for residential location (Clark & Fossett, 2008; Clark, 2009).
Blacks, in particular express strong preference for integrated neighborhoods while whites
express strong aversion to integrated and highly diverse neighborhoods. The empirical
evidence indicates that whites are willing to move into a neighborhood with no more than
30 percent minorities. Blacks on the other hand, are willing to move into a community

when there is a presence of 50 percent blacks and 50 percent from other races or ethnic

103

groups. These preferences shape locational outcome and the general pattern of
suburbanization by different population groups.

For this study, the percent non-white is a proxy variable for the neighborhood
preference concept as an explanation to the differential levels of decentralization across
population groups. The percent non-white measures the likelihood of other non-whites to
relocate in a community based on the racial composition. It is plausible to predict that
communities with a higher concentration of non-whites would attract other non-white

residents than those with lower concentration of non-whites.

Distance from the Central Business District

Aside from the proposed determinants of decentralization described above, the
distance variable enables an analysis of the changes in the distribution of people and
intensity of land use across the metropolitan landscape from a monocentric framework.
The distance from the CBD measures locational attributes and the extent of spatial
separation between communities (Mieszkowski & Mills, 1993), which are proxy
variables for the signiﬁcance of geographical and spatial considerations that affect
locational decisions (Belden et al., 1998; Burchﬁeld et al., 2005).

The distance effect on the decentralization process in the study region is measured
in terms of areal distance or a Euclidean measure of separation between the Detroit CBD
and each study community. The distance variable is derived by measuring the centroid to
centroid distance using the Hawth’s Tools ArcGIS extension in ArcMapTM 9.1 (Beyer,

2004). This distance variable is used in the regression model.

104

Data and Data Selection
To estimate the empirical model, data preparation including data selection,
collection, processing, and integration procedures were undertaken. Three sets of data
were needed to develop and test the causal relationship between local roads and the
decentralization process. These include data on 1) demographic and socioeconomic
characteristics, 2) the roads spending and road stock, and 3) GIS shapeﬁles for describing

patterns of decentralization, and estimating spatial regression models.

Data Collection and Processing

Data were retrieved from existing sources. The demographic and socioeconomic
data were extracted from the Census CD 1980 and 1990 (GeoLytics Incorporated). Other
data were also extracted ﬁ'om the National Historical Geographic Information Systems or
NHGIS (2004) developed by the Population Center at the University of Minnesota. Both
GeoLytics Incorporated and the NHGIS developed software and data products using data
from the census of population and housing. Both organizations used the long and short
forms of the decennial census.

Data on roads spending and road stock in miles were retrieved from the Michigan
Department of Transportation, Financial Services Division and the MDOT Library. These
departments provided MDOT Report Series No. 139 & 162. Speciﬁcally, data by city and
village for the total receipts of the Michigan Transportation Fund and locally raised
revenues, and local road stock were retrieved and integrated with other data relevant for
the analysis. Data processing and integration were performed using a MS Excel 2003

spreadsheet and ArcMapTM version 9.1.

105

301

1‘0

The description and presentation of these information involved mapping using
existing shapeﬁles obtained from the ofﬁcial website of the Michigan Geographic Data
Library, Center for Geographic Information or CGI (http://wwwmcgi.state.mi.us/mgdl/,

2009).

The Database

The ﬁnal database contained the following ﬁelds: Name of municipality (NAME),
percent population density change (%_DENSCH), percent change in non-white
concentration (%_NWLQCH), per person state allocated local road revenue from
Michigan Transportation Funds (MTFPP), per person locally-raised revenue
(LOCALPP), proportion of white homeowners (%_WHOME), percent non-white
(%_NONW), household median income (MEDINC), percent of non-whites, 25 years old
and over with at least four years college education (%_CONW), local road density
(LOCRDENS), and the distance measure between the Detroit CBD and each of
community in the study region (DISTDET).

Since some explanatory variables use monetary value as a measure, dollar values
for these variables are converted into 2008 dollar amount to account for inﬂation using
conversion factors on consumer price index (CPI) that are listed in the Bureau of Labor
Statistics website. The conversion of dollar values into current values enhances
comparability when analyzing objects and entities that changes values over time such as
money. A summary of the variables used in the analysis and their data sources is listed in

Table 5.2.

106

Table 5.2. Variables and their Sources

 

 

 

 

 

 

 

Variable Description of variable Source
%_DEN SCH Percent change in population Computed by the author
densityl980-l 990 based on data from the
Census CD 1980-1990
GeoLytics, Incorporated
%_NWLQCH Percent change in non-white Computed by the author
concentration, 1980-1990 using population count from
Census CD 1980 and 1990
MTFPP77 Per capital state road revenue, in S, Michigan Dept of
1977 Transportation, Report #162
and #139
LOCALPP77 Per capita locally-raised road Michigan Dept of
revenue, in $, 1977 Transportation, Report #162
LOCRDENS77 Local road density=total local roads Computed by the author
in miles/area, 1977 using data from MDOT
report #162

 

FAEDINcso

Household median income, 1980

National Historical GIS,
Population Center, U of
Minnesota

 

%_CONWSO

Percent of non-whites, 25+ years 4
years college education, 1980

Census CD 1980-1990
GeoLytics, Incorporated

 

I %_WHOME80

Percent white homeowners/ Percent
white population, 1980

Census CD 1980-1990
GeoLﬂics, Incorporated

 

 

%_NONWSO Percent non-whites (Blacks, Native Computed by the author
Americans, Aleut, Alaskan, Asians based on data ﬁ'om Census
and Paciﬁc Islander, Others), 1980 CD 1980-1990 GeoLytics,

Incorporated
DIS TDET Distance from Detroit CBD, in Computed by the author

 

miles

 

using Hawth’s Tools
ArcGIS Extension

 

107

 

are g
e115

m

100.
Re;
001
Ob;

be‘

Empirical Model Estimation

In quantitative analysis, correlations and regression are statistical techniques that
are generally adopted to determine whether a relationship between two or more processes
exists. Bivariate correlation statistics measure the degree of association between two

variables, while regression analysis takes one step further by exploring the degree of ﬁt

of this relationship (Kachigan, 1986).
Multivariate regression analysis is used in determining the relationship between

local roads and decentralization process in the context of the Detroit Metropolitan region.
Regression is a statistical technique that is most often adopted in modeling spatial and
non spatial processes and is used in both simple and multivariate estimation. The overall
objectives of regression analysis are: l) to determine whether or not a relationship exists
between two variables (in simple/univariate regression); 2) to describe the nature of the
relationship, should one exist, in the form of mathematical equation; 3) to assess the
degree of accuracy of description or prediction achieved by the regression equation, and
4) in the case of multiple regression, to assess the relative importance of the various
Predictor variables in their contribution to variation in the criterion or dependent variable
(Kachi gan,1986; p. 239).

The Ordinary Least Squares technique is used in the estimation of the empirical
model 0f the causal relationship between local roads and the decentralization of people
b 3’ race Within the context of the Detroit Metropolitan region. The dependent variable in
the model is a rate of change between two time periods (1980 and 1990). The

i I-ldel’endent variables are lagged variables or the initial condition and are therefore

Stlmated coefﬁcrents of the extent that the rnrtral condition affects the trajectory or

108

future direction of growth and subsequently patterns of decentralization. The use of
lagged independent variables is necessary given that a cross section data may not lend a
straightforward interpretation of cause and effect. However, to the extent that initial
condition precedes growth and decentralization outcomes, one must consider the
“priorness” of the initial condition. Thus, the causal relationship between local roads
spending and decentralization is better understood by recognizing the initial condition of

explanatory variables and the extent by which it affects future grth and

decentralization.

Ordinary Least Squares
The Ordinary Least Squares technique is based on the following assumptions: 1)
explanatory or predictor variables are independent; 2) The expected or mean value of the
residuals is zero; 3) residuals are homoskedastic and independent from one another; and
4) residuals are normally distributed (Bailey & Gattrell, 1995). Essentially, OLS assumes
a linear relationship between the predictor and criterion variables and seeks to estimate

the parameter/s that provides the best ﬁtting linear relationship. The OLS equation takes

the fom1:

It
Y=a+ Zﬂixi+£;where (5.4)

3:]

Y is the dependent or criterion variable, a is the Y intercept (the value of Y when x = O),

B i is the coefﬁcients of the independent variables, X. is independent or predictor

V ' -
arrables, and s 13 the error term.

109

To understand how each variable adequately explains the decentralization process
in metropolitan areas, the dissertation adopts a stepwise regression via forward selection
approach. This approach helps “single out” the simplest model that describes the
decentralization process or other processes of interest. Essentially, the approach enables a

process of identifying signiﬁcant independent variables in the model, one variable at a

time (Kachigan, 1986).

Variable Transformation

Prior to modeling, univariate tests for normality are performed on the model
variables to account for normality assumption of the Ordinary Least Squares (OLS). The
histograms for all variables in the model are plotted and presented in Appendix B.
Natural log transformation was performed for a few variables but since these did not

yield meaningful coefﬁcients particularly since the dependent variable is measured in
proportion (Bailey & Gatrell, 1995; p. 276-277), a re-estimation without the log
transformation was made and results were retained for further analysis.
Assumption of independence of errors is further tested by employing the Moran’s
I Statisti c on the model residuals (Moran, 1950). These tests have implications for the
goodness of ﬁt of the model (Bailey & Gattrell, 1995).
Moran’s I is a weighted correlation coefﬁcient that identiﬁes deviations from
Spatial randomness. If a pattern of distribution of the values of variable or certain spatial
process 00 curs, the process is said to indicate spatial autocorrelation (Anselin, 2005). A
Spatially autocorrelated process indicates a biased and imprecise OLS estimator. It is

b ‘ . .
1 ased beczause events that are clustered or concentrated Will have a greater impact on the

110

estimate (Kachigan, 1986). With clustering of events, fewer number of independent

observations than assumed affects the precision of the estimate. Moran’s I takes the form:

"2 chj(yc ‘TXJ’,’ —;)

 

I- c=lj=l
n _ 2 ;where, (5,5)
2(yc —y) X Z we]
C=1 c¢j

n is the number of observations or communities, yc is the value of a variable at location c,

yj is the value of a variable at location j, y is the mean value ofvariable y, wcj- is the

weight coefﬁcient for the lag separation between c and 1' based on a proximity weight
matrix W.

Moran’s 1 takes the value of -l.0 to 1.0. Values with positive coefﬁcients suggest
that the surrounding values of an observation i at a speciﬁc location c increase as the
value of i increases. Conversely, negative coefﬁcients suggest that the surrounding values

0f observation i at location c decrease as value of i increases. A zero coefﬁcient suggests
that spatial dependence is not observed in the process under study. This diagnostics test
on mod 61 residuals provides information on the ﬁt of the OLS model and allows

al temati ve modeling technique to be explored.

Spatial Lag Model
Since area data tends to exhibit violations in assumption of independence in errors
%d ObserVation (Anselin, 2005), the spatial lag model as an alternative to the standard

LS regrtitssron technique 18 used. The altematrve model 13 chosen to capture second

111

order effects or the spatial autocorrelation of model residuals that a standard OLS
regression do not account for, thereby making inﬂated and biased estimates (Bailey &
Gatrell, 1995; p. 287).

The spatial lag form introduces a spatial autocorrelation coefﬁcient rho, p, which
indicates a diffusion or “spillover” effect of the process. It captures the inﬂuence of
neighboring values of dependent variables on the dependent variable or process of

interest. To operationalize a spatial lag model, a spatial weight matrix W is developed
using alternative proximity measurements. A k nearest neighbor proximity matrix is

adopted in this research and the model is estimated by maximum likelihood estimation

(Anselin, 2005).

From Bailey & Gatrell (1995, p. 288), the spatial lag k model is:
Y = Xﬂ+pWY+ a; where, (5.6)

Y, the random variable of the process of interest, is a function of a trend component X )3,-
a second order component pWY, and s is vector of independent random errors. And,

13 is a vector of parameters to be estimated, p is a spatial autoregressive coefﬁcient

indicating the inﬂuence of nearby values of Y, W is the proximity matrix.

For the empirical model estimation of the relationship between local roads and
patterns of decentralization, the R statistical software (R, 2004) and GeoDA (Anselin,
2 005) are used. The R analysis used three libraries including: foreign (DebRoy et a1,
4/ 1 90005), maptools (Bivand et al, 3/4/2005), and spdep (Bivand et a1, 4/19/2005). The

S tatistical Package for Social Science (SPSS) version 17 is also used for initial data

112

processing and management. Data integration and mapping were performed using

ArcGIS 9.1.

Summary of Data and Estimation Procedure
The empirical estimation of the relationship uses the Ordinary Least Squares
(OLS) regression technique. Subsequently, a stepwise regression is performed to address
possible multicollinearity issue among independent variables and simplify the model.
Further, Moran’s 1 statistic is determined to account for the possible violation of
independence in the model residuals. Finally, an alternative model, spatial lag, is
speciﬁed to account for the structure or second order effects of the process.

The next chapter presents the descriptive statistics for the study area. It also
presents the results of the statistical model estimation of the relationship between local
roads and the differential levels of decentralization of people by race in the Detroit

‘Metropolitan region. A discussion of results of model estimation, and a brief summary of

the ﬁndings are also included.

113

CHAPTER SIX: RESULTS AND DISCUSSION
To reiterate, the objective of this study is to understand the relationship between
local roads spending and the decentralization of people in metropolitan areas. Also, the
speciﬁc goals of the study, listed in Chapter 4, are to 1) develop the conceptual model of
the causal relationship between local roads and decentralization using three measures of
decentralization; 2) describe the pattern of population density change and the
concentration or representation of non-white population; 3) describe the road spending

variables and their distribution; and 4) test the empirical model using data from a speciﬁc

case study of the Detroit Metropolitan region.

In this dissertation research, decentralization is understood using three speciﬁc
indicators: 1) percent density change of the general or total population; 2) percent change
in non-white concentration or representation; and 3) percent change in non-white
population density of each community across the study region. The effects of local roads
spending on decentralization are estimated while accounting for other proposed predictor
variabl es: income and education, the demographic composition or percent non-white, the

Proportion of white homeowners to the proportion of white population, and distance

between the Detroit CBD.
This chapter presents the empirical estimates, statistical results of the hypotheses

tests, and discussions of ﬁndings. The ﬁrst section of the chapter presents the descriptive

Statistics - This is followed by the analytical results of the hypotheses testing. A discussion

0 f key results ends the chapter.

114

Descriptive Characteristics of the Study Region
The Detroit Metropolitan region comprises a total of 131 townships, villages, and
cities, with two communities straddled between two counties (Southeast Michigan
Council of Governments, 2002). Eighty eight (88) of these communities, all of which are
incorporated cities and villages, are identiﬁed as study communities. These communities
vary in land area, and land use intensities (See Figure 1.1, Chapter 1). The average
population in a given community was 38,156 in 1980 and 35,306 in 1990. The land area

across the study communities is on average, 9.31 square miles and ranges from 0.25 to

l 42.92 square miles.

Population Density
The population density across the study region ranged from 131 to 10,172 persons
per square mile in 1980, and 179 to 7,422 in 1990. The population density in a given
comm unity in the region averaged 3,528 in 1980 and 3,047 in 1990. The decrease in the
average population density in the study region between 1980 and 1990 reﬂects population
redistri bution as Opposed to shifts in land area since it is assumed that land area for each
Community remained static between 1980 and 1990.

Indeed, unlike in the earlier decades, very few city or village incorporations took
place between 19803 and 19903 in the region (SEMCOG, 2002). Within the study region,
only three city incorporations are noted in the 19803 including Lake Angelus (1984),
ALIbum H ills (1983), and Rochester Hills (1984). Thus, the average percent change in
130plllaliiorl density across the study region, which is 23 percent, reﬂects the percent

I)qulaﬁon growth in a given community. It shows the dynamic nature of the growth

115

process such that while a community may have increased in population density by as high .
as 454 percent between 1980 and 1990, some others decreased by 95.31 percent.

An examination of the spatial distribution of the changes in percent population density
across the study region reveals the location of communities that experienced a signiﬁcant
change in population density. Figure 6.1 shows that the central city (Detroit),
communities immediately surrounding the central city, as well as Pontiac experienced a
stark decline in population density. Many outlying and small communities in the region,
on the other hand, experienced an increase in population density, thus, providing
evidence of the decentralization process occurring in the region during the study period.
As it shows, communities north of the city of Detroit, a few in the northwest side and a
small cluster in the south of the city experienced population density increase between the
1980 and 1990 by as much as 454%. On the other hand, the city of Detroit and many of
the surrounding communities declined by as low as 95.31 percent while a couple of

communities indicated no change between 1980 and 1990.

116

     

I: .9531 - 68.48
- -68.47 - -27.03
- -27.02 - 5.22
- 5.23 - 40.97
- 40.98 - 454.64
I: Township

0 2.5 5 10 Miles
1—1—1—1—1—1—1—1-1

Figure 6.1. Percent Population Density Change,
Detroit MSA, 1980-1990

117

A more interesting snapshot of population redistribution in the region can be
gleaned from Figure 6.2. Figure 6.2 shows the combined population for each group of
communities according to their distance from the central city of Detroit. The travel time
estimates are adopted from the Southeast Michigan Council of Governments (SEMCOG,
2002), which provided an estimated travel time between Detroit CBD and individual
communities in the region during peak or rush hours. Communities located within 1 to 15
minutes from the central business district of the city of Detroit decreased in population
since the 19703. Communities located within 16 to30 minutes away ﬁ'om the central
business district of Detroit, or the inner ring suburbs, also experienced a similar trend
though some slight increase in the population is observed since the late 19803. The
outlying communities, those located beyond 30 minutes away from the central business
district of Detroit experienced increases in population. What is important to note is that
communities that are located within 30 to 45 minutes away from the city of Detroit were
fast growing relative to other communities in the region. How do racial composition,
socio-economic characteristics, and local roads stock and spending reﬂect the pattern of
population distribution in the region?

The next section provides descriptive statistics of each of the factors proposed to
explain the pattern of decentralization in metropolitan areas. The proposed factors are
derived from a review of segregation and residential patterns literature presented in

Chapter Two.

118

 

03 r»

I\ N N [s 1 1 1

or 00 N 60 .

‘- 93 93 g 82 g N 60+ Travel time from CBD
Year " ‘- N (minutes)

Figure 6.2. Population Change by Travel Time: 1972-2002

Source: Computed by the author using population estimates from the US Census Bureau,
Census of Governments ﬁles, 1972-2002

119

Patterns of Decentralization of Non-whites

An interesting aspect of the variation in community level population density is its
possible connection to the spatial distribution of the non-white population in the region.
The pairwise correlation statistics (Appendix C) indicates that these two processes are
positively associated at a statistically signiﬁcant level of less than 0.05 though the
strength of this relationship is not strong (Pearson r = 0.232, p < 0.05).

The non-white population, which comprises the total population count of blacks
or Aﬁican Americans, Hispanics, Native Americans, Eskimo, Aleut, Asians and Paciﬁc
Islander, and others, are situated unevenly across the study region. Figure 6.3 and 6.4
show that while there is a representation of the non-white population across the different
communities of the study region, they are mainly overrepresented or highly concentrated
in a few communities. Communities with location quotient greater than 1.0 indicate a
relative concentration of non-whites as can be gleaned from Figure 6.3 and 6.4. These
maps identify speciﬁc communities that have highest relative concentration of the non-
white population in the region and reveal the link between concentration of non-whites
and community population 1033.

Figure 6.3 shows that in 1980, cities such as Highland Park, Detroit, Plymouth,
Inkster, Ecorse, Pontiac, New Haven, River Rouge, Utica, South Lyon, and Wixom were
home to a large proportion of non-whites (LQ > 1.0) relative to the Detroit Metropolitan
region as the reference area. Seven of these ten cities have the highest concentration of
non-white population including Highland Park (LQ=3.52), Detroit (LQ=2.63), Plymouth
(LQ=2.52), 1nkster(LQ=2.45), Ecorse (LQ=1.64), Pontiac (LQ=1.55), New Haven

(LQ=1.40), and River Rouge (LQ=1.37).

120

A similar distribution of non-whites’ relative concentration in the 1990 can be
gleaned from Figure 6.4. Communities that indicated a high concentration of non-whites
in 1980 remained highly concentrated in 1990. Oak Park became highly concentrated by
1990 as well. Altogether, Highland Park, Detroit, New Haven, Inkster, Pontiac, Ecorse,
Oak Park, River Rouge and Southﬁeld had high concentration of non-whites among the
study communities in the region. Except for Southﬁeld, these cities had a Location
Quotient that is greater than 1.00 (LQ > 1.0).

While Figures 6.3 and 6.4 provide a static description of the presence and
concentration of the non-white population in the study region, a map of the percent
change in the concentration of the non-white population between 1980 and 1990 offers
another dimension for analyzing the dynamic character of the distribution of people by
race within a decade. Figure 6.5 shows that although many communities in the study
region had a considerable presence of non-whites, non-whites are not overrepresented in
these communities relative to the region as a whole. These communities had location
quotients less than 1.0 or a LQ of 0.01 to 0.99. However, it can be argued that these
communities may have increasingly become the “destination” communities for non-

whites.

121

     

[:l 0.00
- 0.01—0.41
- 0.42 - 0.99
- 1-00

- 1.01 _ 3.52
1:] Townshlp

0 2.5 5 10 Miles
l—l—l—I—l—I—l—l—l

Figure 6.3. Relative Concentration of N on-whites,
Detroit MSA, 1980

122

 

Figure 6.4. Relative Concentration of Non-whites,
Detroit MSA, 1990

123

Speciﬁcally, communities in the periphery or the ﬁrst ring suburban areas
immediately surrounding the city of Detroit, and some distant villages toward the north
experienced gains in non-white residents by over 200 percent within a decade. The top
gainers or retainers of non-white population between 1980 and 1990 include Farmington,
Garden, Warren, Royal Oak, Melvindale, Farmington Hills, Ortonville, Novi, Dearbom,
and Sterling Heights with percentage change that ranges from 551 to 1,235 percent
(Figure 6.5). On the other hand, ﬁve communities were least gainers or retainers of non-
white population including Highland Park, River Rouge, Gibraltar, Hamtramck, Fraser,
Detroit, and Pontiac. These seven communities indicated a gain of non-white population
that ranges from 1 to 18 percent.

However, the gains in these seven communities and the overall redistribution and
concentration of non-white population need be interpreted with caution. It is possible that
the increase in non-white population could be due to “white ﬂight” and not necessarily an
inﬂux of non-whites to these communities. It is also possible that both processes, whites
ﬂeeing and blacks moving in, to be simultaneously occurring. Nonetheless, the shifts in
rate of concentration reveal the degree of change in the socio-spatial separation between

whites and non~whites.

124

I MAC OMB
OAKLAND
I P

w ‘

   

               
 

I

I“ "'

   

   

  

 
      
   

. l... N
"I l A
I 'l‘.

0
1‘36 1:] —98.32 - -31.24
- -31.23 - 69.63

- 69.64 - 214.05
- 214.06 - 551.46
- 551.47 - 1235.18
[:1 Township

1
r

0 2.5 5 10 Miles
l-+—+—l--|—l-H—|

I“ﬁgure 6.5. Percent Change in the Relative Concentration of Non-whites,
Detroit MSA, 1980-1990

125

Socioeconomic Characteristics and Homeownership
As described above, the population redistribution in the Detroit Metropolitan
region ﬁts well with the process of decentralization. The Detroit Metropolitan region
experienced signiﬁcant population redistribution and a distinct racial composition.
Measures of socioeconomic characteristics and homeownership reﬂect the population and
racial fabric. The household median income, educational attainment, and homeownership
reveal the disparity across the study communities.

The average household median income across the study communities was
$61,923 . 79 in 1980 and $68,170.69 in 1990. The income disparity across the region is
clearly revealed in the range, which is $149,797.38 in 1980 and $228,520 in 1990. Figure
6.6 illustrates a scatter plot of the distribution of communities by household median
income with respect to distance of each study community from the Detroit CBD. The plot
generally indicates a decreasing trend of household median income with communities
that are located farther ﬁom the CBD. However, the plot also indicates that most of the

economically endowed communities are located within 10 and 25 miles of the Detroit

CBD.

126

 

200,000 -

A 180,000 - o
3’? 160,000-
l40,000 ‘ 0
120,000 ‘ O
. [100,000 - 0°

0
80,000- 00 ° o°°0

O O
991W
’ o 0
40,000- 0 o O % 0% 0° 80
0

Household Medran Income

 

 

 

 

20,000 - °
0 u . u r .
0 10 20 30 4O 50
Distance from Detroit CBD (miles)

 

Figure 6.6. Household Median Income by Distance from Detroit CBD (1980)

The disparity across the study communities is not just revealed by household
median income but by levels of educational attainment and homeownership as well. On
average, there is a lower proportion of non-whites with college education across the study
communities. On average, a given community has less than 1 percent or 0.55 percent of
non-white population who had at least 25 years old and completed 4 years of college
education in 19803. A majority of educated residents in a given community were still
mostly whites. Similarly, the proportion of white homeowners to the proportion of white
population in a given community is also very high. On average, almost all of white
residents in a given community in 1980 owned their home (average estimated
homeownership index = 0.97). In 1990, this proportion was even higher, 1.03, indicating
that some communities have had over representation of whites who own a home relative

to the community’s white population.

127

The high representation of white home owners in a given community could
suggest greater incidence of discrimination in housing. In the Detroit Metropolitan
region, this white homeownership index ranges from 0.67 to 1.86 in 1980 and 0.95 to
1.54 in 1990. Clearly, a high proportion of whites own their home relative to their non-

white counterparts.

Local Roads Spending and Local Roads Stock
The variation in population size, land use intensity, and socio-economic and
demographic characteristics in the study region reﬂects variation in the amount of funds
that each community receives for local roads development and maintenance from the
state government and other sources including locally-raised or contributions coming from

the local government. Hence, the local roads stock and spending reveals such variation.

The Local Road Revenue, State Allocated

The Michigan Transportation Fund allocation for local roads development
remains the primary source of revenue for local roads development and maintenance in
the state of Michigan. For cities and villages, the average per person allocation amounted
to $59.24 in 1977. The dollar amount that each community receives reﬂects the
community’s population size and road mileage in a given time period. The state
allocation across these communities ranges from $37.42 to $138.01 and $5.03 to $89.70
per person in 1977 and 1987, respectively. The decline in average allocation from 1977 to

1987 is noteworthy. It suggests many factors affecting the allocation of revenues

128

including population shifts as well as the total amount of revenue that is generated by the

state as a whole prior distribution (Welke et al., 2000).

Patterns of M TF Allocation in the Detroit Metropolitan Area

All cities and villages in the Detroit MSA have consistently received the state
allocated local roads revenue through the Michigan Transportation Fund (MTF) of Public
Act 51 of 1951. However, disparity in allocation has been observed as can be gleaned
from Figure 6.7. This ﬁgure shows that the total share of the city of Detroit has declined
over time. The proportion of the MTF allocation for the city of Detroit in relation to total
allocation of the metropolitan region ranged from 56% in 1953 to 37 % in 2002,
indicating a decreasing trend of the city’s share over time. This is not surprising,
considering the pattern of population distribution in the region described above.

Figure 6.8 and 6.9 present spatial distribution of state-allocated roads funds across
the study communities in 1977 and 1987. The values are expressed in dollars per person
and show the variation of allocation across the communities. These two maps show a
slightly different distribution of local road funds. In particular, some communities in
Oakland and Macomb counties have had relatively higher per capita state-allocated local
road funds than communities in Wayne county in 1977 (Figure 6.8). This pattern changed
by 1987 (Figure 6.9) when all counties including Wayne, Oakland, and Macomb have
had some communities that experienced a relatively higher per capita roads allocation.

The reduction in per capita allocation from 1977 to 1987 is noteworthy. The

change in per capita allocation for communities ﬁom a maximum of $138.01 in 1977 to

129

$89.97 is maybe explained by several factors including a reduction in overall MTF

revenue which affects local road allocation.

The Locally-raised Road Revenue, Local Contributions

To reiterate, the locally-raised road revenue presents another major source of
funding for local roads development and maintenance for Michigan communities.
Locally-raised road revenue generation for many communities has declined. Speciﬁcally
between 1977 and 1987, communities experienced an average decrease in their ability to
generate and spend for their local roads development and maintenance, which is similar
to the pattern of state spending on local roads. The trend indicates that for local
communities, their capacity to fund local roads development tended to decline over time
as well.

The average per person road revenue that a given community generated amounted
to $49.40 and $16.41 in 1977 and 1987, respectively (Table 6.1). The locally-raised road
dollars ranged ﬁom $0 to $355.00 in 1977 and $0 to 290.53 in 1987. The zero value
indicates that some communities either did not report any local contributions or these
communities may have fully relied from other sources for development and maintenance
of their road stock. Figure 6.10 and 6.11 present variation of per capita locally-raised
road revenues that individual communities generated for 1977 and 1987. These maps
show the changes between the speciﬁed time periods described above.

In terms of local roads stock, the local road stock across the study communities
did not signiﬁcantly change in a decade. Between 1977 and 1987, there was an average

increase of 1.97 miles per square mile across the study communities. The average local

130

road density was 21.29 in 1977 and 23.21 in 1987. Figure 6.12 and 6.13 show the
individual local road density by year and the slight change that can be gleaned from them.
They show that a slight increase in road density in 1987 in some communities in Oakland

and Macomb counties is apparent.

 

100
090
050 -
070
060 7
0.50 :

DetroitIMSA

 

040
030 7
020 -
010 -

000 . - . - - : -
1950 1960 1970 1980 1990 2000 2010

Year

 

 

 

Figure 6.7. MTF Allocation: Proportion of the City of Detroit to the Total
Metropolitan Region, 1952-2002

Source: Computed by the author based on MDOT Reports #139

131

     

- 0.01- 5534
- 55.95- 67.65
- 67.66- 96.81
- 96.82- 138.01
I: Township

0 2.5 5 10 Miles
|-—1—+—1+1—1—1—1

Figure 6.8. Per Person State-allocated Local Roads Revenue (5),
Detroit MSA, 1977

I32

     

- 0.01- 47.58
- 4759— 55.25
- 55.26- 69.58
- 69.59 - 89.97
C] Township

0 2.5 5 10 Miles
l—H—l—l—l—l—l—I

Figure 6.9. Per Person State-allocated Local Roads Revenue (S),
Detroit MSA, 1987

133

     

- 0.01 - 59.00
- 59.01 — 103.00
- 103.01 - 188.00
- 188.01 - 355.00
:1 Township

0 2.5 5 10 Miles
H—l—H—l—I—l—l

Figure 6.10. Per Person Locally-raised Road Revenue (S),
Detroit MSA, 1977

134

 

[:1 0.00
- 0.01 - 22.93
- 22.94 - 38.18
- 38.19 - 65.90
- 65.91 — 290.53
:1 Township

0 2.5 5 10 Miles
l—i—i—H—i—i—i-l

Figure 6.11. Per Person Locally-raised Road Revenue (S),
Detroit MSA, 1987

135

   

 

- 4.34 - 8.20
- 8.21 - 11.64
- 11.65 - 16.14
- 16.15-21.29
l:l Township

0 2.5 5 10 Miles
1—1—1—1—1—4—1—1—1

Figure 6.12. Local Road Density, Detroit MSA, 1977

136

   

     

- 4.74 - 8.58
- 8.59 - 11.66
- 11.67- 15.77
- 15.78 -23.21
:1 Township

0 2.5 5 10 Miles
1—1—1—1—1—1—1—1—1

Figure 6.13. Local Road Density, Detroit MSA, 1987

137

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

New 82 8. a 88 2 R2 8 a 88 888 .82 58.88 832 83
3:. 3.2 8.98 o 2 S2 .888 8 m 28 88 Baa-£83
8.8 4.8 me o 8 R2 .8588 88 m 98 88 Baa-383
2.2 8.8 8.2 88 .8 $2 .888 88 m 2.2 .83 88 2%
8.2 8.8 5.2: 8.8 8 R2 .888 88 m 2.2 .88 88 2.8
88 82 42 88 8 82 883 9 8888; an; .8 8888
8.0 8.0 of So 8 82 883 a 8888; 2:? 8 8:885
8.: 8.8 8.8 o 8 82 8:388 6553-8: 83 8
8.2 w: 8.8 o 8 82 8:288 via-8: 88 8
43 $8 8.2 o 8 82 so E +8 88:8 a» 4 58 as; 8:8 .8
8.5.8 8.8 18 8.838 888.2 8 82 .m 582: 888 2838:
8.48. 2 N 28.8. e 8.83: 888.8 3 82 .m 6885 8:88 2888:
8.8m 228 2.83 87 mm 82-82 3 888885 223-8: .888 .8
86 one mm.m o mm 82 .04 gouge-8:8 ones-eoz
3.0 «no mom :8 8 82 On senescence 6:882
8.8 3.8 8.8.4 2.8- mm 82-82 .888 588 8888 888.2
83 O8. 53 :08- mm 82 -82 888 £88 88388
83 Sow 3..» a: 8 82 base 8:288
Sam 83 N22 5 mm 82 588 88388
$8.2 _ 08.8 88.80; 2 m 8 82 8:288
832 82.8 828.2 82 mm 82 8888
Sun— .Ew 502 £58?wa SEE: Z o—Dwﬂdxr
moumcﬂm 339880

 

 

 

bias—530 bosom 25 we murmtoaoauaanv gag—mnemon— ~6 Sash.

138

139

 

 

 

 

 

 

 

 

8 8885 2 28>
8.: :8 8.22 8.: 8 2:: 88m 5 82
8.: 8.2 2.8 N2 8 828 88:8 58282 889 so: 8:85
8.: 8.2 8.8 2% 5 $2 8 8 8 88: 588: 88. 83

 

 

 

 

Adz—SUV motor—33.35 955839 ﬁe via-n.

Analytical Results

To determine the relationship between local roads and population
decentralization, an empirical estimation of three fully speciﬁed models of three
indicators of pattern of decentralization for the general and non-white population is
performed via the standard Ordinary Least Squares (OLS) approach. A stepwise
regression is then performed to identify a less comprehensive and underlying sub-model
(if it exists) of the process. To determine the ﬁt of the model and possible violation of the
OLS assumptions, the model residuals are examined using the Moran’s I statistic.
Subsequently, considering the likelihood of spatial autocorrelation of the residuals, an
alternative spatial lag model speciﬁcation is estimated to account for possible violation of
the assumption of independence of errors in OLS and to improve the predictive power of

the model.

Model Estimation

Two hypotheses are tested to determine the causal relationship between local
roads and decentralization. The three measures of the pattern of decentralization, as
dependent variables, are: 1) percent change in population density as a measure of pattern
of decentralization of the general or total population; 2) percent change in relative
concentration of non-white as a measure of the pattern of decentralization of non-white
population; and 3) percent change in population density of non-whites as a second
measure of pattern of decentralization of the non-white population. Each measure is

regressed against eight proposed predictor variables.

140

Recall that from Chapter Four, the empirical model of population decentralization
is developed using three theoretical perspectives of segregation and residential patterns
literature — spatial assimilation, neighborhood preference, and place stratiﬁcation — and
were integrated into the human ecological framework of decentralization. To reiterate,

the model for estimation takes the form:

ADecentralization it =f(Road spending and density(,.1),-, Income(,_1)i, Education“. 1);,
%Non-whites(,.1)i, Proportion of White home owners(,_1),-,
Distance, 8;); where, (6.1)

As explained above, there are three dependent variables used as indicators of
decentralization. The road spending and road density variables include state allocated
road revenue, locally-raised revenues for local roads development, and local roads
density as measures of policy effects on decentralization. The income and education
variables reﬂect ﬁnancial status that measures capacity for spatial mobility. The percent
non-whites measures neighborhood composition, which reﬂects individual or group’s
preference for either homogeneous or diverse community. The proportion of white
homeowners is an index of relative concentration of white homeowners in a community
as proxy for housing discrimination. Finally, distance is a measure of proximity to a
central business district (CBD), speciﬁcally to the city of Detroit. It reﬂects locational
preference and impacts of spatial separation between communities and the CBD on
patterns of population decentralization. The error term, 8,, is random error or remaining

variability, and t-I represents previous or lagged values of variables in the model.

141

The ten year lag reﬂects the expected length of tenure for all homeowners across
the US. That is, ten years is the median number of years between a purchase and resell
of a given house (National Association of Realtors, 2008). The lagged variable in the
model essentially captures the impact of previous conditions of the conceptualized
predictors on the future condition of the process of interest, such as population density
change as a measure of the pattern of decentralization in a given region.

The model described above posits that the decentralization of people in
metropolitan areas is a function of government policy on local roads development
(environmental component). However, it also accounts for other proposed predictors
reﬂecting impacts of socio-economic or ﬁnancial status (population component based on
ecological and spatial assimilation perspective), demographic composition (behavior
component based on neighborhood preference perspective), and social exclusion or
discrimination (behavior component based on spatial stratiﬁcation perspective) as
conceptualized predictors for the pattern of decentralization of people.

For the rest of the chapter, Equation 6.2 is used to explain decentralization using
the three indicators presented above. See Table 5.1 for a summary of variable description

and their sources.

ADecentralization1930_1990i = ’60 + p(°/o __ DENSCHi) + ﬂ] MTFPP 77" + ﬂzLOCALPP 77;

+ ﬁ3L0CRDENS 77 ,. + ,64MEDINC80; + ﬂ5°/o_C0NW 80,-
+ o6%_ NONW 80,- + ,67%_WH0ME 80,- + ,BsblSTDETi + a,
(6.2)

The following section provides results of the empirical estimation of each of the

three models of decentralization. For each model, a full and reduced (by stepwise

142

regression) form are estimated using the standard Ordinary Least Squares and Spatial
regression techniques. Discussions of results of each model are presented in a separate

section of the chapter.

Pattern of Decentralization: Total Population
Decentralization of Total Population: Ordinary Least Squares Regression Results, Full
Specification

Fully speciﬁed and reduced model were estimated to determine the relationship
between local roads and pattern of decentralization of the total population while
controlling for other proposed predictor variables. The model intends to capture what
explains decentralization in general, which is the pattern of decentralization of combined
whites and non-whites population groups.

Results of a fully speciﬁed model (percent change in population density)
estimation using the Ordinary Least Squares (OLS) approach indicates an adjusted R
square value of 0.40. Hence, 40 percent of the variation on the percent change in
population density is explained by eight (8) conceptualized predictors. The Moran’s 1
statistic is 0.01 and is not signiﬁcant, which indicates that spatial autocorrelation of the

model residual is not an issue in the model. A residual map is presented in Figure 6.14.

143

 

- 402.54 - -S9.65
- .59.“ - -9.12
- -9.12 - -0.01
- -0.02 - nus
- 118.24 - 360.53
C] Township

   

0 2.5 5 10 Miles
I-H—l—i—l—i—l—I

Figure 6.14. Percent Change in Population Density Model Residuals

144

Four of the eight conceptualized predictor variables in the model are statistically
signiﬁcant. The estimated coefﬁcient of government policy effects or state—allocated local
road spending on decentralization is 0.86 and is statistically signiﬁcant at less than 5
percent level (p< 0.05). The coefﬁcient of socio-economic effects or household median
income is 0.001 and is statistically signiﬁcant at less than 1 percent level (p<0.01). The
coefﬁcient for demographic composition, which is percent non-white is -0.97 and is
statistically signiﬁcant at less than 5 percent level (p< 0.05). The coefﬁcient for the
spatial or locational attributes and spatial separation measure, which is distance from the
central business district is 4.51 and is statistically signiﬁcant at less than 0.] percent
(p<0.001).

The coefﬁcients indicate that communities with higher per capita state-allocated
road spending, household median income, and with increasing distance from the CBD
experienced an increase in population density and this increase reﬂects a pattern of
decentralization of the total population. However, the stronger presence of non-whites in
a given community has a negative effect on the rate of change of population density. It
appears that a stronger presence of non-white residents in a given community induces

repulsion ﬁ'om rather than attraction to the community.

Decentralization of Total Population: Spatial Regression Results, Full Speciﬁcation

Data tends to result in a violation of normality, constant variance, and
independence of errors (Bailey & Gatrell, 1995; Kachigan, 1986). For area data, data that
are aggregated into different area] units such as census zones, municipalities, voting

districts, the OLS assumptions do not hold well and are often violated. This is particularly

145

so since “areal data give rise to many spatial relationships” (Bailey & Gatrell, 1995; p.
20) and these spatial relationships can take in many forms with corresponding relevant
measurement or description.

For example, if one wants to assess the degree of segregation between and among
particular population groups, there are several ways for describing their spatial
relationship. Clustering, exposure, isolation, and concentration are possible measures that
each can be described in a number of ways (Massey & Denton, 1985). If the interest is
spatial distribution of shoppers across a region, travel cost and travel time maybe used to
describe the spatial relationship. Some spatial relationships are described in days traveled
by feet in societies or communities that lack modern transportation. Hence, different
spatial relationships can be gleaned from aerial data and relevant measures can be used to
describe these relationships.

Due to an inherent presence of spatial relationship in areal data, it is highly likely
that spatial autocorrelation in the model residual will be observed. Hence, a spatial lag
model is estimated to capture spatial autocorrelation of errors, and improve prediction.
For a spatial lag model, a proximity matrix using 3 and 5 nearest neighbors is developed
for this study.

Results of the spatial lag model estimation conﬁrmed the notion of a presence of
spatial autocorrelation in the model residuals. The spatial autocorrelation coefﬁcient rho
(p) is high and statistically signiﬁcant at less than 0.1% (p = 0.43; p<0.001) for the spatial
lag model using the 3 nearest neighbors, and p = 0.38 (p<0.001 ) for the 5 nearest
neighbors. In addition, the effect of the presence of non-white population with college

education is signiﬁcant. The coefﬁcients are -6.32 (p<0.05) using the 3 nearest neighbors,

146

and -5.13 (p<0.10) using the 5 nearest neighbors proximity matrix. Hence, an inverse
relationship seems to exist between the presence of non-whites and the pattern of
decentralization of the general or total population.

Results from the spatial lag model estimation conﬁrmed a presence of the spatial
autocorrelation in the model residuals. These also suggest that the spatial lag model
performed better than the OLS as can be observed from the tests for multivariate
normality including the log likelihood, Schwarz and AIC criteria (Anselin, 2005; p. 175).
Both the spatial lag models using 3 and 5 nearest neighbors have a higher log likelihood
values (3nn, -481.583; 5nn, -483.94) than that of the OLS (-487.14) estimation. The AIC
(3nn, 983.06; 5nn, 987.89) and Schwarz (3nn, 1007.83; 5nn, 1012.66) criteria for spatial
lag models are also slightly lower than the OLS values (AIC=992.28, Schwarz=1014.58).
Comparatively speaking, the spatial lag model performed better than the OLS model
estimates and is able to capture the secondary effects of the process. Except for the effect
of the college educated non-white residents and the presence of non-white population in a
given community, the coefﬁcients of the signiﬁcant predictors (state-allocated road
spending, household median income, percent non-whites, and distance from the central
business district) are consistently higher and statistically signiﬁcant. Table 6.2
summarizes these results.

These results show positive effects of state-allocated local road spending on
decentralization. As the coefﬁcients show, local road spending facilitates population
growth in suburban communities and this growth contributes to the decentralization in

metropolitan areas.

147

Save a 48.0%N t... .8? a I. gov m .. :32 858$an

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

taco Ian-v.0 22
* *Nﬁo Ed 2366px ~ MSW-52
8.22 Ea z: 3.82 $.32 8:25 53%
an? gas a. as ES? 8er: MS
$.23 ”.23 8.93 380 02
MS mg 88% a
one 93 03:3 m .a<
So as as as 23% m
Imam m: I... 5. 28m m: I... 5. Ea a :23 use Susan
5.8- 8.8 39.- 9.3- 8.3 3%. $2 £28388; 223 .x.
33 as *8- .03- 3o *3. 82 8338 ages s
$.2- 25 Km. ...Nme- 5m KN- oma 08:8 a E v a; +8 £33-82 x
188d ago tags 180.0 825 1285 82 0585 E38 2288:
08. 9: Bo. So. a: 5o- :2 £23 32 83
8o 26 3.0 8o 26 :3 Re 2522 22 822-283
wood :3 EMS «So :3 ES E: 8:22 Re 2%
* z 3.x- on: 5.2:. “9.8. 3: *Nnoe- %Szou

Echo—$000 5km E20580 EEoEooU “ohm E06580 Spawn-5

MS Edam eavesm m3 Exam $25
30 Begum So “panama
cam-V conﬁaz “882 m 39 eoﬁwaz “882 m

c8723—

.<m2 £950: £323: :o=a_=nom 5 $530 Eon-Sm .2232 :5: .538..qu 15.3w can FEE—Sm .«o manﬂoEocU .N6 2...;

148

Decentralization of Total Population: Ordinary Least Squares Regression Results,
Reduced F orrn

A stepwise regression is performed to simplify the model of pattern of
decentralization of the general or total population using the same convention adopted in
the fully speciﬁed model estimation. Results from the OLS procedure yield a ﬁnal model
with four signiﬁcant predictor variables including: 1) the state-allocated local road
revenue; 2) household median income; 3) demographic composition, which is percent
non-white ; and 4) distance between a given community and the Detroit central business
district or CBD. The OLS and spatial regression model estimation results are presented in
Table 6.3.

Results of the OLS stepwise estimation procedure in the restricted or reduced
form indicates an adjusted R square of 0.42, which means that 42 percent of the variation
in percent change of population density is explained by four conceptualized predictors.
The Moran’s I statistic is 0.02 and is not signiﬁcant, which indicates that spatial
autocorrelation of the model residual is not an issue.

The coefﬁcients indicate state-allocated road revenue (0.74, p<0.05), household
median income (0.0011, p<0.01), percent non white (-0.99, p<0.05), distance from the
CBD (4.93, p<0.001) are signiﬁcant predictors of the pattern of decentralization of the
total population. The coefﬁcients show that higher state road allocation for communities,
socioeconomic or ﬁnancial status, and locational factors are signiﬁcant in explaining
decentralization. However, greater presence of non-whites negatively affects the
decentralization process such that a community with higher proportion of non-white

population induces repulsion from rather than attraction to a community.

149

Decentralization of Total Population: Spatial Regression Results, Reduced F orrn

An alternative spatial lag model is also estimated with the reduced form of the
model to conﬁrm the presence of autocorrelation of errors in the process. The Moran’s 1
statistic using the 3 nearest neighbors indicates a value of 0.02 and is not statistically
signiﬁcant. However, with 5 nearest neighbors, Moran’s I increased to 0.10 and is
statistically signiﬁcant at the 5 percent level. Therefore, an alternative model in the form
of spatial lag is performed.

The results of the spatial lag model estimation conﬁrm that spatial
autocorrelation in the model residuals exists. The spatial autocorrelation coefﬁcient rho,
p, indicate a positive and statistically signiﬁcant value at less than 0.1 % (3 nearest
neighbors, p = 0.36, p<0.l%; 5 nearest neighbors, p = 0.32, p<0.1%).

In comparing the results of both OLS and spatial lag model, it is clear that the
spatial lag model slightly performs better than the OLS estimation and the spatial
autocorrelation is also addressed. The tests for multivariate normality including the log
likelihood, Schwarz and AIC criteria (Anselin, 2005; p. 175) show improvements of the
spatial lag estimates to the OLS estimates. The values from these tests indicate that the
spatial lag estimates ﬁt the observed process better than the OLS estimates. Both the
spatial lag models using 3 and 5 nearest neighbors have higher log likelihood values
(3nn, -483.76; 5nn, —485.56) than that of the OLS (-487.97) estimation. The AIC (3nn,
979.53.06; Snn, 983.11) and Schwarz (3nn, 994.39; 5nn, 997.98) criteria for spatial lag
models are also slightly lower than the OLS values, thus, indicating that spatial lag model

performs better than the OLS model.

150

The slight shift in the strength of the effects of local roads policy on
decentralization is noteworthy. The reduced OLS estimate is 0.74 and is statistically
signiﬁcant. The spatial lag model using the 3 nearest neighbors’ proximity matrix is 0.48
and is not statistically signiﬁcant. On the other hand, the spatial lag model estimate using
the 5 nearest neighbors indicates a coefﬁcient of 0.56 and is statistically signiﬁcant at the
10 percent level. The discrepancy in strength and signiﬁcance suggests the sensitivity of
the estimates to the deﬁnition and speciﬁcation of the proximity matrix. Table 6.3

summarizes these results.

151

8.on a 48.on r; .5? a t .8? m .. :32 880885

 

 

 

 

 

 

 

 

 

 

 

 

 

*de ***©m.o cam
.58 8.8 282QO _ was:
8.8 ~28 8.8 ~38 8:35 888
8.34. 8.58- 8.me 8.2:,- 88:% MS
_ 88 888 8.88 8.23 o?
8.2 8.8 888 a
as 98 088 m .3
8.8 38 $8 $8 08% a
3.3m 88 $8.4 ram: 88 :84 8:5 a 889 :58 8:85
.88- as 8.8- .88- 38 .88- 82 8:288 383-8: 8
:88.o 88.0 to 88 88.8 88.8 $888 82 0585 882: 2838:
«88 88 1.8 8.0 43 *4; RS 0822 82 3%
38.0:- 28 38.3;- ILQS- 2.8 time:- ezﬁmzou
8205000 Sum E20580 E06580 “ohm Eamon-HBO
ms 888 88% m3 88% 38% 888>
So @8888 So @8858
A53 mconawﬁz 80802 m A58 $3,362 8232 m

 

 

gag-cue .<w—Z

39:2— "335: notwinom E 99.20 2.3.5.— 5252 coo—€95 mesa—wax :5an a...“ Beta-3w he .8:oner .m6 033.

152

Summary of Findings: Decentralization of Total Population

The results of the empirical model estimation of the pattern of decentralization of
the total population show that generally, pattern of decentralization is explained by 1)
government policy effects, speciﬁcally state-allocated local road revenue, 2)
socioeconomic status, speciﬁcally household median income 3) diversity or homogeneity
of the population, speciﬁcally percent non-whites in a community, and 4) spatial and
locational attributes, speciﬁcally the distance from the CBD variable. The effect of
government policy on local roads development is positive and signiﬁcant, which supports
the ﬁrst hypothesis of the study. The ﬁrst hypothesis states: “communities with higher
road stock and road spending are expected to facilitate increases in population and that
increase in population will contribute to the decentralization of people. "

Indeed, as the results show, the state’s road spending in Michigan does facilitate
the pattern of decentralization of the total population. Whether or not government policy

on local roads development exhibits the same effect on the pattern of non-white

decentralization is presented in the second model estimation.

Pattern of Decentralization, Non-white Population

The relationship between local roads and decentralization of the non-white
population is determined by estimating two models, the relative percent change of
concentration and percent change of non-white density as dependent variables. The
empirical model estimation follows all the estimation procedures adopted for estimating

patterns of decentralization of the general population. The model attempts to capture the

153

effect of local roads while controlling for other proposed predictors on the pattern of

decentralization of non-white population.

Decentralization of Non-white Population (Relative Concentration): Ordinary Least
Squares and Spatial Regression Result, Full Specification

Results from model estimation of the fully speciﬁed and reduced model of non-
white decentralization (percent change of relative concentration of non-whites) yield a
relatively consistent estimates and statistical signiﬁcance for both OLS and spatial lag
models. The test for spatial autocorrelation, Moran’s I is not signiﬁcant. Also, the ﬁnding
of no autocorrelation is conﬁrmed with the spatial lag model estimation. The spatial
autocorrelation coefﬁcient rho (p) is not signiﬁcant for both the full and reduced model
speciﬁcations. Further, the multivariate normality tests log likelihood, and AIC and
Schwarz criteria indicate consistent results, which show that spatial lag models with
different proximity matrices, did not perform better than the OLS. Based on these tests, it
appears that spatial autocorrelation is not signiﬁcant and does not present an issue to the
OLS model estimation. Therefore, estimates from the OLS model remains a reasonable
estimate of the pattern of decentralization for non-whites.

Results of the fully speciﬁed model indicate an adjusted R square of 20 percent
which means that 20 percent of the variation is explained by the eight proposed
predictors. Three out of eight (8) proposed predictors indicate statistical signiﬁcance in
the model, including: 1) Locally-raised road ﬁrnds; 2) % non-whites; and 3) distance

from CBD.

154

Speciﬁcally, the signiﬁcance of locally-raised revenues in road development on
decentralization is noteworthy. The coefﬁcient for the locally-raised roadrevenue is 1.6
(p<0.01), and is statistically signiﬁcant at the 1 percent level. Similarly, spatial or
locational effects explain the pattern of decentralization of non-whites but the effect is
negative. The coefﬁcient is -10.48 and is statistically signiﬁcant at the 5 percent level
(p<0.05). The result suggests that non-whites decentralize but their pattern of
decentralization is constrained to communities located closer to the CBD. On the other
hand, the greater presence of non-whites affects the pattern of decentralization of other
non-whites. The coefﬁcient is -7.55 and is statistically signiﬁcant at 1 percent (p<0.01).

Table 6.4 summarizes these results.

155

Sodve. .12.. ._o.ov m1. .modv n. _.. ”—96— pogoﬁcwﬁ

 

 

 

 

 

 

 

 

 

 

 

 

 

Ed 26 2E
mod cod inﬂamed _ M582
”v.53 mash: cm. 52 mag-BS 55:0 ganom
mmgc- mg 5. @320- mw.m$- 33:85 me:
539 oowmﬁ 93$ 36mm” 02
Mt}:- wN-d 288% m
88 8.8 08% a .a<
omd wmd omd wad 085cm x
*ocd- m; :22:- smnd- N; *mvdT 8?: 5 €th 50¢ 8585
5.3 862 Nudm Mama 863 mud»- owoﬁ 8580:3880: 3E3 .x.
€1.02.- mod 1.3.5- .3136. mod *Imm.\-- cam: noun—30m SEB-coc .x.
mo. _ T 8.2 we; T mm._ T 8.2 we; T owe“ owuzoo E 8» v ,8» +3 833-52 R.
ﬂood- m Sod ﬂood- m Sod- m Sod ﬂood- omoﬁ 2:85 5838 20:33:
om; $6 :2 cm; 2.0 34 :2 98% use Boo-H
1%; who 1. $4 1%; who .1. 3; RS 85>»: 38 3385.83
EN 2: omd EN 3; man 2.3 2832 38 38m
modem 3&3 oodwm 3.3-m omwﬁ 393 872-5200
0.0305000 Sum EBB-coo 0 022050500 Sum 8205000
93 38m 38% e3 888m 888m 83%;
mac cacgm qu @8855
3:3 £82302 5382 m Q20 88,362 8232 m

 

 

33-23— .<m2
=9.qu "aetaeunooae0 333-.52 E own—.30 Boo-Sm .2352 :55 neﬁmouwom .3»an tan 33:35 me 85.9580 .96 033.

156

Decentralization of Non-white Population (Relative Concentration): Ordinary Least
Squares and Spatial Regression Result, Reduced Form

The reduced and ﬁnal model using the OLS and spatial regression approach
indicate results consistent with the fully speciﬁed model. The adjusted R square value is
22.0, which is up by 2 percent compared to the fully speciﬁed model. This indicates that
22 percent of the variance in the model is explained by three predictor variables. The
reduced model essentially indicates that the decentralization of non-whites is affected by
the local govemment’s efforts to develop local roads infrastructure. The coefﬁcient is
1.76 and is statistically signiﬁcant at the 0.1 percent level (p<0.01). Similarly, the
locational factor has signiﬁcance in the pattern of decentralization of non-whites but the
effect is such that, the pattern of decentralization for non-whites is constrained to
communities closer to the CBD. The coefﬁcient is negative, -8.44, and is statistically
signiﬁcant at the 0.1 percent level (p<0.001). Also, the greater presence of the non-white
population in a community has a negative effect on the pattern of non-whites’
decentralization. The coefﬁcient is -7.46 and is statistically signiﬁcant at the 0.1 percent
level (p<0.001).

The results indicate that the pattern of decentralization of the non-white
population is explained by locally-raised road funds, greater presence of non-white
population, and locational or spatial attributes or distance from the CBD factors. Table

6.5 summarizes these results.

157

5°.on 1* .de A .1, .modv Q n. ”_0>0_ 00:02.:cwa

 

 

 

 

 

 

 

 

 

 

 

a; 2.0 a 22
8.0 $6 massage a was:
_ 209 3&2 a. $2 822 8:25 Magnum
£3 s- ”2%. 2.20- $.80- 82:85 03
N332 WEE ﬁgs 2.32 02
0.3 o? 225a a
mg mg 03:3 m g
Rd 3o Rd m3 23s a
toms- a; 3.0. 33- com 3.”- 8:5 a :23 53 8529
ENE- 32 E01- ENE- 22 30:- 32 8:238 223-8: .x.
332 one 1:0: 1:3 coo 50: Re 855. 82 828-3303
$398.6. 3.? 1&2: 58.0% 8% 53.3% Pzﬁmzoo
«CumoEvOU gnu-Em “2205000 320E000 hobm aﬁumoﬂwooU
03 30an 05:80 03 33% 032% saunas
So 3055 So “saga
953 222902 30502 W 25.9 28:052 “852 m

 

 

 

 

33-33 £92 .255 "5:22.850

81.3.5.2 3 $530 2.00.5“— .20—52 too—.6015 :Emnohwom .52me 1:0 32:55 05 no 3535000 .m6 012;.

158

Decentralization of Non-white Population (Non-white Population Density): Ordinary
Least Squares and Spatial Regression Results, Full Speciﬁcation

To further analyze the effects of local roads and other predictor variables on the
participation of non-whites in the decentralization process, an alternative model is
estimated using a density measure for non-white population as the dependent variable.
The fully speciﬁed model indicates that 24 percent of the variation is explained by the
eight proposed predictor variables. Presence of autocorrelation in the model residuals is
examined using the Moran’s 1 statistic coefﬁcient. The coefﬁcient is 0.009 and is not
statistically signiﬁcant (1 = 0.009, p = 0.67). A spatial lag model is further estimated but
results indicate that the lag model did not improve the prediction of the OLS model. The
spatial autocorrelation coefﬁcient p is positive but not statistically signiﬁcant (p = 0.06).
Therefore, the model estimate from OLS is retained for analysis.

Overall, results indicate that only the locational or geographical factor, distance
from the central business district is statistically signiﬁcant at less than 0.1 percent. The
estimate indicates that for every mile that a community is located farther away ﬁ'om the
Detroit CBD, a corresponding decline of 4.8 percent in the non-white population density
is expected. Similarly, the model indicates that a greater presence of non-whites also
affect the overall non-White’s decentralization pattern. Thus, the strength of the locational
or geographical factor and demographic composition are important considerations in
understanding the pattern decentralization non-whites.

Table 6.6 summarizes the results of the full OLS model estimation of the percent

change of non-white density and the proposed predictor variables.

159

Evan .53!N I... .8? a .1. gov a .. ”:32 85055

 

 

 

 

 

 

 

 

 

 

 

 

:.o- 80 a
mod mod mam-528% d @552
added" 0.003 0.33 0.33 :otoEU ~33ng
H0.va- 00:043- mmdwv- $.me 085:3: we:
8.33 3...“? 2423 omxcdd 02
hmé 50+ oamuﬁm u
go Rd 05% m .6...
Ed de de de 0.833 M
.3156- oo; 3.3%.? *Inwd- do; 31%.? 3:8 E :obuQ an 8:8me
84mm 2.0m mwam mm.mm 3.0m mw.mm dwﬁ mEmSFsooEon BBB X.
MWmud. vvd avnd- «mud- vvd «Ed- dwe :ouﬂsmom BBB-co: .x.
m: _N.m 0m; d0._ KM 0m._ owi owe—Bo S. we» v we» +mN 6:53-52 o\°
voodd- mdddd good- good. mdddd voddd- of“: 0885 .8608 238:0:
Rd- mm." wnd- Nwd- mm; mud- :2 .3800 38 :80:
:d Ed 2d mod 9d d H d :3 35>»: 38 BEE->280:
mmd- omd Gd- ad- dmd “Nd- E-a 3:960 008 88m
**0m.mm_ mfmv Imvéﬁ imﬁvﬂ adv 3&ng PZ<PmZOU

056580 5cm EBB-coco 620580 Sum E20580
an 3an esteem wﬁ Exam wanasm 832:3

qu Banzai m-HO 0.305%
3:3 megnwmo Z $882 w A508 £08362 3832 m

 

 

 

dag-dwa <m2 =9.qu—

"bﬁno: nets—.53— 85»?qu E omega-U 0:99.; .9232 :55 noﬁmohwom Guam 0:: Ewanﬁw 2: .3 356580 0.0 «3:.

160

Decentralization of Non-white Population (Non-white Population Density): Ordinary
Least Squares and Spatial Regression Results, Reduced Form

A reduced form of the model of the pattern of non-white decentralization is
estimated using the percent change non-white density as the dependent variable. Results
indicate that the reduced form performed slightly better than the fully speciﬁed model
(Table 6.7). The adjusted R square is 0.26 percent indicating that 26 percent of the
variation on the dependent variable is explained by the eight proposed predictor
variables. The test for the presence of autocorrelation in the model residual, Moran’s 1,
indicates a value of 0.004 and .003 for the OLS model using two proximity matrices.
Both are not statistically signiﬁcant. The test for any improvement of the OLS model
estimate using the spatial lag model indicates nearly consistent coefﬁcients. The
multivariate normality tests AIC, Schwarz and log likelihood were slightly different
between OLS and spatial lag models but no signiﬁcant difference exist between the
model. The autocorrelation coefﬁcients rho (p) for both models using the 3 and 5 nearest
neighbors are negative but are not statistically signiﬁcant.

The results of the estimation are consistent with ﬁndings of the fully speciﬁed
model. They conﬁrm the effects of locational factor (distance from the CBD) and
demographic composition as signiﬁcant predictors of the pattern of decentralization of
non—whites. The estimates show that the effects of distance from the CBD is negative
indicating that non-whites decentralize but they do relocate in communities closer to the
CBD. Results also indicate that neighborhood composition is a signiﬁcant factor
explaining non-whites decentralization pattern. The model indicates that a greater
presence of non-whites in a community appears to deter locational decision. The

coefﬁcient suggests that a 1 percent change in the non-white population corresponds to a

161

0.74 decline in the rate of change of population density at a statistical signiﬁcance of 10
percent. Consistent with the other measure of decentralization, the model estimates
indicate that locational factors and demographic composition have signiﬁcant effects on

the pattern of decentralization of non—whites.

162

2v}N ,Sodvm t; .8? a r. gov a * :32 85%:me

 

 

 

 

 

 

 

 

 

 

 

 

Ed- 83. a
83 256 233$ ~ €822
$.23 $.33 8.23 9.93 8:35 ~52th
3331 8.31 3.2? 8.3? 82:3: woo
as? 8.03 8.»; 868 02
3.2 3.2 games“, a
03 8d 23a m .3
Rd ”No mmo ”do 05% m
1%.? Rd :53- 53¢- Rd :33- 8:8 a maﬁa Soc 8:929
“who- as «Ed- «Ed- 93 “Ed- one 8:238 233-5: R.
3:32 5.2 1.3.5 :Lvm: $2 $3.5 b72528

E20580 “ohm E20580 E20580 “ohm E20580
93 Exam @5qu wﬁ Esra, seesaw 85E;

30 32% So “52%
3:0 £92302 3282 m 3:0 £85902 5832 m

 

 

 

93—453 .(m—Z =9.qu innuen—

=ot£=nom 833-52 5 own-EU 239.?— 5232 .5253: :33"...qu 33.25 was FEE—Sm 2: he 856580 .56 93.;

163

Summary of Findings: Patterns of Non-White Decentralization

Results of the model estimation indicate three consistent factors that explain the
pattern of non-white decentralization. Local contributions for local road development, the
demographic composition as measure of homogeneity or heterogeneity of communities,
and locational factor or distance from the CBD, were signiﬁcant predictors. These results
provide some support to the second hypothesis, which states “municipalities with higher
local roads spending, higher representation of non-white population, lower
representation of white homeownership, higher income and education, and increasing
distance from a central business district are more likely to facilitate increasing
representation of non-white population and that increase in population contributes to
increasing level of participation of non-whites in the decentralization process. "

Results of the model estimation indicate support to the hypothesized relationship
between local roads spending and decentralization of non-whites albeit in a different
context. As the results show, only the local government contributions for local road
development, not the state’s allocation, are signiﬁcant. Also, contrary to expectation,
non-whites participate in the decentralization process but their pattern of mobility is
constrained to communities closer to the CBD. Furthermore, greater representation or
concentration of non-whites in a given community affects decentralization of non-whites
in a negative direction. Hence, the demographic composition of a community, locational,
and locally-raised road investments affect the relative community attractiveness for non-

white household relocation, which in turn affect the general pattern of decentralization of

non-whites.

164

Discussion

The results presented above help explain the pattern of decentralization in the
context of Detroit Metropolitan region and show the signiﬁcance of key explanatory
factors in the process. Findings highlight the inﬂuence of local roads spending,
socioeconomic or ﬁnancial status, demographic composition, and locational or spatial
attributes in patterns of decentralization in metropolitan areas. However, ﬁndings also
indicate that different population groups respond to these factors in different ways. In the
remainder of this chapter, these ﬁndings are reiterated within the context of the stated

hypotheses and research questions.

Effects of Local Roads Spending

The result of model estimation suggests a direct and signiﬁcant inﬂuence of local
roads spending on the pattern of decentralization particularly for the general population in
metropolitan areas. This result also provides a considerable basis for contending that
growth in a given community is tied to initial level of expenditure on physical
infrastructures such as local roads. Those communities with higher average local roads
allocation gained additional residents a decade later, indicating the direct effects of local
roads spending on the relative attractiveness of such communities for residential location.

This ﬁnding is consistent with the notion that government spending on
infrastructure (including transportation) inﬂuenced the decentralization of people in
metropolitan areas. It speciﬁcally conﬁrms that local road spending results in differential

levels of suburbanization across racial groups. Therefore, the evidence presented in this

165

study is consistent with previous ﬁndings on highways and major throughways effects on
decentralization, which indicates that government spending facilitates decentralization.

The extent to which state’s road spending allocation affects the pattern of mobility
of non-whites is not clear. The coefﬁcient of the state’s allocation for local roads
development is positive but is not statistically signiﬁcant. The lack in signiﬁcance of state
allocated spending on non-whites may suggest that non-whites are more tied to urban
jobs, may rely more on urban amenities, have limited ability to move out of cities, and
encounter resistance or constraints to moving into suburbs and nearby rural areas.

This study shows that state-allocated revenue for local roads development rewards
high population and high road stock communities but shifts over time to reward
communities that are growing and increasingly located away from the central business
district or CBD. This shift presents a disadvantage to communities with higher
representation of non-whites, which are mostly located in closer proximity to the CBD.
Because many of these communities experience population decline, which means a
decline in state-allocated dollars, they must rely on their taxing capabilities to
compensate for such loss to address local roads and road infrastructure needs. Hence, the
pattern of decentralization of non-whites appear to be sensitive to the location, taxing
capacity of the community reﬂecting tax base, size of community, and other measures of

ﬁscal capacity of a given community.

Effects of Socioeconomic Factors
This study ﬁnds that household median income is a positive and signiﬁcant

explanation for the pattern of decentralization of the general population. In the context of

166

the Detroit Metropolitan region, communities with higher income in the earlier decade
have attracted additional residents of similar economic background. This ﬁnding is
consistent with the ecological and assimilation perspective, which asserts that the uneven
mobility patterns across different individuals and population groups reﬂect their
socioeconomic background, and for foreign born individuals, cultural competence to
assimilate with majority whites (Massey & Denton, 1985; Massey, 1985). The ﬁndings of
this study that income is not signiﬁcant in the non-white model of decentralization
suggest that factors other than socioeconomic status facilitate non-white mobility. It
shows that higher income does not to translate into greater mobility for minorities.

There are at least three possible reasons for the lack of evidence supporting the
ecological and spatial assimilation theories in non-whites’ pattern of decentralization.
First, the spatial stratiﬁcation reasoning could be a dominant process in the context of the
Detroit Metropolitan region exempliﬁed by the persistent separation of other racial
groups from whites (particularly blacks) in residential, employment, educational contexts.
Second, preference for communities with certain demographic composition and other
contextual factors could be the dominant explanation as well. And third, spatial or
locational considerations may have stronger effects on decentralization for both the total

and non-white population.

Effects of Racial Composition
The ﬁnding in this study shows that the pattern of decentralization of both the
general and non-white population is affected by the community’s demographic

composition. It shows that the behavior of both general and non-white population groups

167

reﬂects avoidance from moving into communities with greater presence of non-whites.
This ﬁnding is consistent with the literature, which provides evidence to suggest that race
is a dominant explanation for the residential patterns in US metropolitan areas.

As the ﬁnding in this study shows, the pattern of black suburbanization and
blacks’ persistent separation from majority whites in the case of the Detroit Metropolitan
region is a microcosm of a broader condition. This study also conﬁrms prior ﬁndings that
blacks may have increasingly participated in the decentralization process, but their
residential pattern reveals persistent separation from majority white population. The
persistent separation between blacks and whites has strengthened notions of
discrimination, prejudice, and other range of factors including a lack of interest for
integration among non-whites themselves as explanations.

Indeed, blacks have consistently experienced the highest level of segregation from
whites relative to other racial groups despite an increased level of black suburbanization
by as much as 50 percent or more in some areas, at the regional (Darden, 1977; Farley,
1986; Clark, 1991) and national (Farley & Frey, 1994) levels. Blacks lag behind Asians,
and other racial groups in terms of socio—economic achievements, and assimilative
capacity with the white majority population group (Clark, 1992, 2009; Clark & Ware,
1997; Iceland & Wilkes, 2006).

In essence, the persistent resistance to integration among whites to non-whites is a
more valid explanation for not ﬁnding that income is not statistically signiﬁcant
explanation for the pattern of decentralization of non-white population in this study.
Given the history of entrenched segregation in the Detroit Metropolitan region (Sugrue,

2005; Darden et al., 1987), the ecological theory would be supported in the pattern of

168

decentralization of the total population. With non-whites, particularly blacks, race, not
socio-economic status is expected to explain the pattern of decentralization particularly
since race distorts ecological theory (Darden et al., 1987; Darden and Kamel, 2000).
Therefore, a class-based analysis of the segregation of blacks and whites in the region is
not a sufﬁcient dimension for understanding the persistent segregation of blacks from the
majority whites (Darden & Kamel, 2000).

The evidence in this study therefore provides support for the ecological and
assimilation theory in explaining the pattern of decentralization of the general population
in the Detroit Metropolitan region, but not for the non-white population. It appears that in
the context of Detroit, the spatial stratiﬁcation reasoning may support the pattern of non-
white decentralization.

However, it is also possible that attitude and preference for a particular
demographic composition of the community underlies the pattern of decentralization of
non-whites in the Detroit Metropolitan region. This notion, which ties to the
discrimination reasoning, brings to a second possible explanation for the lack of evidence
supporting the ecological and assimilation theory in the pattern of decentralization among
non—whites in this study.

Findings in this study show that the demographic composition (percent non-
whites) in a community and the pattern of decentralization for total and non-white
population are inversely related. Communities with a higher presence of non-whites in
the previous decade experienced a decrease in both the rate of growth and concentration
of non-whites. This result suggests that both groups, the general and non-whites, avoid

communities with greater presence of non-whites.

I69

For the general population, the inverse effect of the demographic or racial
composition on the pattern of decentralization is consistent with the literature and
possible support for the “own race” preference theory (Charles 2002; Clark & Fossett,
2008; Clark, 2009). Whites indicate the strongest preference for own race consistent with
neighborhood preference perspective and also in some ways place stratiﬁcation
perspective, which emphasizes the low tolerance thresholds of whites for contacts with
blacks (Fischer, 2008). Blacks, Hispanics and Asians indicate preference for integrated or
50-50 mix neighborhoods (Clark, 1992). Blacks in particular, indicate preference for
neighborhoods with 50 percent blacks and 50 percent other races (Clark, 1992).

For non-whites in the Detroit Metropolitan region, ﬁnding a negative relationship
between demographic composition and pattern of decentralization is not consistent with
expectation. The result indicates that higher representation of non-whites in a given
community a decade earlier corresponds with a decreasing percent concentration of non-
whites for that community. This ﬁnding suggests that the total or aggregate and non-
white population groups demonstrate avoidance from communities with higher
proportion of non-whites. Total population density and non-white concentration
decreased in communities with higher proportion of non-whites a decade earlier.

This ﬁnding is not surprising given that different racial groups that indicate
preference for integrated and mixed neighborhoods, may not in reality translate into
behavior (Clark, 1992). Clark (1992) showed that despite the stated preferences for mixed
neighborhoods among Anglos, Asians, Hispanics, and blacks in his study of over 2000
households in California, a very low proportion of Hispanics (0.06), and blacks (0.07)

actually moved into mixed neighborhoods. Asians and Anglos did not actually move into

170

these neighborhoods. Those who indicated “no preference” actually chose to live in their
own race neighborhoods (Clark, 1992). In essence, evidence of what people say they
prefer and how they behave are two different processes. The reality seems to indicate that
demographic composition or the homogeneity or diversity of the community, reﬂects
residential location choice.

The evidence in the Detroit Metropolitan region may reﬂect this inconsistency in
motivation and behavior. It is also possible that non-whites prefer integrated
neighborhoods with a racial combination other than their stated 50 percent blacks and 50
percent combination of other races. It is also possible that contextual factors not
demographic or racial composition of destination communities underlie the pattern of
non—white decentralization in metropolitan areas. Communities with better amenities tend
to be homogeneous communities with higher property values. This suggests that a
combination of class and contextual factors, coupled with job availability in the suburban
areas are synergistically affecting the pattern of decentralization among non-whites rather

than purely demographic or racial composition reasoning.

Effects of Spatial or Locational Attributes
Results conﬁrm previous ﬁndings that the decentralization of people reﬂects
changes in population density with distance from the CBD. Indeed, such ﬁnding is
consistent with the pattern of decentralization of the general population. On the other
hand, non-whites relocate to communities closer to the central city but with lower
proportion of the non-white residents. It appears that non-whites have a preference for

communities closer to the central city, which has some presence of other non-whites, and

171

access to better job opportunities and locational and institutional amenities relative to the
central city. This behavior suggests that non-whites are being selective in choosing their
destination communities. Essentially, the pattern of non-white decentralization can be
explained by non-whites’ preference for proximity to jobs and amenities near the CBD
combined with a sense of security and belonging in the (low) presence of other non-
whites.

In the case of Detroit, the pattern of non-white decentralization is consistent with
the above observation. In Detroit, more jobs are located outside than inside of the city’s
boundary. Only 5 percent of jobs are located within 3 miles, 22 percent within 10 miles,
and majority or about 78 percent are located beyond 10 miles of the city of Detroit ( et
al., 2001, p. 5). Thus, for non-whites in the study region, the concentration of fewer other
non-whites in communities closer to the central city and presence of job opportunities in
these communities may have strong inﬂuence in their pattern of decentralization.

The ﬁnding in this study that shows the general and non-whites’ preference for
communities with lower presence of other non-whites is noteworthy. It supports the
notion that attitude does not necessarily equate into a behavior. Consistent with the
literature, the ﬁnding shows that even when whites and Asians may indicate openly a
preference for integrated neighborhoods, their residential patterns does not match with
such stated preference. In reality, they tend to choose to live in homogeneous rather than
diverse communities (Clark, 1992).

The mismatch between behavior and stated preference for integrated residential
neighborhoods may come from sheer “avoidance” by some population groups to live with

non-whites. It is also highly possible that racial avoidance to integrated communities is

172

not an expression of pure prejudice and discrimination against non-whites. Instead the
avoidance is an indirect expression for preference for high amenity, better locational
attributes, and quality life (Harris, 2001). Since most homogeneous communities tend to
have better amenities than in diverse or integrated communities, people in general (both
whites and non-whites) would tend to choose these locations with lower presence of non-
whites.

The range of explanations for patterns of decentralization of people by race is
wide and complex. Ultimately, as ﬁndings in this dissertation show, a combination of
processes, not a single process explain the divergent patterns of residential choice and
patterns of locational outcomes for different population groups. This study suggests that
local roads funding allocation policy, socioeconomic or ﬁnancial status, demographic or
racial composition and spatial attributes explain the pattern of decentralization of the

general and non-white population in the Detroit Metropolitan region.

173

CHAPTER SEVEN: CONCLUSIONS

This study examines pattern of decentralization in metropolitan areas and
investigates causal factors within the social geographic inquiry. Informed by three
theoretical perspectives of segregation and residential mobility, the study adopts a human
ecology framework that recognizes the effects of environment, population, and behavior
(Meade, 1977; Meade and Earickson, 2000) on the complex and multidimensional
decentralization phenomenon. The human ecology framework adopted in this study helps
orient an analysis of patterns of decentralization that integrates local roads spending with
the conventional theoretical explanations of patterns of decentralization within the
segregation and residential mobility literature including: 1) spatial assimilation, 2)
neighborhood preference, and 3) place stratiﬁcation. The effect of local road spending on
decentralization controlling for other explanatory factors identiﬁed through these three
theoretical perspectives is examined using standard multiple regression and spatial
regression techniques. Essentially, the study examines the relationship between local
roads spending and patterns of decentralization controlling for household median income,
percent college educated non-whites, percent non-white population, proportion of white
home owners, and locational factor such as local road stock and distance from the central
business district. This relationship is understood in the context of the Detroit
Metropolitan Area as the case study area.

This chapter draws this dissertation to a conclusion. The ﬁrst section provides a
summary of key ﬁndings. The second section provides conclusions and implications of

the study. The last section identiﬁes limitations and recommendations for future studies.

174

Summary of Key Findings

Results from this study show that there is a distinct pattern of decentralization for
the general and non-white population groups. The overall ﬁndings indicate that patterns
of decentralization can be explained by 1) local roads spending; 2) socio- economic
factor 3) demographic composition, and 4) locational attributes. Each population group,
the general and non-whites, responds to these factors in different but similar in some
ways.

The pattern of decentralization for the general population is explained by the
state-allocated funding for local roads development, which comes from the Michigan
Transportation Fund of Act 51 of 1951. On the other hand, non-whites’ pattern of
decentralization is explained by locally-raised road funding. This ﬁnding shows that the
state’s road funding policy favors the general population but not the non-white
population. This means that higher state-allocated local road revenue in the initial year is
associated with positive percent change in population density of the general population
but not for the non-white population between 1980 and 1990.

With reference to effects of socioeconomic status on patterns of decentralization,
this study highlights the signiﬁcance of socioeconomic status, speciﬁcally income, in the
pattern of decentralization of the general population. The conclusion that income is a
facilitator of pattern of decentralization of the general population conﬁrms previous
ﬁndings and is consistent with ecological and spatial assimilation theory. On the other
hand, income is not signiﬁcant in explaining the pattern of decentralization of non-
whites. Further, as this study shows, education and income variables are not signiﬁcant

predictors of non-whites’ pattern of decentralization in metropolitan areas.

175

The effects of demographic factors, speciﬁcally percent non-white residents on
the decentralization process for both general and non-white population groups are
signiﬁcant and of the same direction. Communities with greater presence of other non-
whites experienced slower growth between 1980 and 1990. This ﬁnding suggests that for
both whites and non-whites, the presence of other non-whites affects their locational
decisions. Both population groups tend to relocate in communities with a lower presence
of non-whites. This behavior is consistent with the concept of “avoidance” that the place
stratiﬁcation theory asserts (Massey & Denton, 1993; Darden, 2004; Jackson, 1994). This
behavior also lends some support to the neighborhood preference theory, and shows that
demographic composition affects the attractiveness of a community for residential
location (Clark, 1992, 2009).

For both groups, decentralization is affected by locational factors, speciﬁcally
distance from the CBD, but in opposite ways. For non-whites, results seem to suggest
that their preferred destination communities are located at or near the CBD. On the other
hand, the general population, which in this study combines whites and non—whites
population, indicates a tendency to choose communities that are located increasingly
away from the CBD. As this ﬁnding indicates, both groups have proclivity for residential
spaces based on locational attributes, which may reﬂect preferences for amenities and
related socio-economic and environmental contexts.

Table 7.1 presents a summary of key ﬁndings of the study. The corresponding
signs denote signiﬁcance and direction of the effects of each proposed explanatory
factors on patterns of decentralization by racial groups. Only the statistically signiﬁcant

explanatory factors are presented in the summary.

176

Table 7.1. Comparison of Signiﬁcant Explanations of Decentralization Between the
Total and Non-white Population

 

 

 

 

 

 

 

 

 

Explanation of Decentralization Total/Aggregate Non-white
Population Population
Local roads spending (Environment)
State allocated \/ (+) Not signiﬁcant
Locally-raised Not signiﬁcant \l (+)
Socioeconomic status (Population)
Household Median Income \/ (+) Not signiﬁcant
Demographic composition (Behavior)
% Non-whites \l (-) \1 (-)
Locational attributes (Environment)
Distance from CBD \l (+) \l (-)

 

\l signiﬁcant; (+) positive direction; (- ) negative direction

Conclusions and Implications

Results of the empirical model estimation provide some answers to the questions
that motivated this study, which are: 1) Do local roads matter in the decentralization of
people? 2) Does the pattern of distribution of local roads investment facilitate
decentralization? and 3) Does the pattern of distribution of local roads funds explain the
pattern of decentralization of non-whites in metropolitan areas?

Given these questions, the study attempts to examine the role of local roads in the
differential patterns of decentralization in metropolitan areas. It does so using the Detroit
Metropolitan region as the case study area between 1980 and 1990, which are years of
persistent high levels of out-migration of people from central cities in many metropolitan
regions in the Midwest and northeast including the Detroit Metropolitan region.

The study focuses on the relationship between local roads spending and the
pattern of decentralization of non-whites using the municipality as the unit of analysis.

Essentially, this research seeks to achieve four speciﬁc objectives.

177

The ﬁrst objective of the dissertation is to develop a conceptual model to explain
the relationship between local roads and decentralization and test this model in the
context of the Detroit Metropolitan region. To accomplish this objective, the research
adopts a human ecology framework and uses this framework to develop a Environment,
Population, and Behavior (EPB) conceptual model that broadly integrates local roads
spending into existing proposed explanations of decentralization.

Recall in Chapter Two and Three that these existing proposed explanations of
decentralization are drawn from the segregation and residential mobility literatures,
speciﬁcally from three theoretical perspectives of decentralization including: (a)
ecological and spatial assimilation, (b) neighborhood preference, and (c) place
stratiﬁcation perspectives. These proposed explanatory factors from the three theoretical
perspectives and government policy on local roads reﬂect the foundational elements of
the environment, population, and behavior (EPB) human ecology model adopted in this
study. The human ecology conceptual model in this study, therefore, helps orient an
analysis of decentralization that recognizes the multidimensional, complex, and
interrelated processes that shape the decentralization phenomenon.

As this study has shown and in the context of the Detroit Metropolitan region,
there is support for the effects of govermnent spending policy on local roads development
controlling for factors that reﬂect the ecological and spatial assimilation perspective,
neighborhood preference perspective, and locational factors.

The second objective of the dissertation is to characterize the pattern of
decentralization in the study region. To accomplish this objective, three indicators of

decentralization were used. The ﬁrst indicator, percent change in population density,

178

measures the pattern of decentralization of the general population. The second and third
indicators, the percent change in the relative cOncentration of non-whites and non-white
population density, measure patterns of non-whites decentralization. The non-white
population includes African American blacks, Hispanics, Asians and Paciﬁc Islanders,
Native Americans, Alaskans, Aleuts, and people who identiﬁed in the “other” race
category.

Descriptive statistics, simple graphs, and mapping procedures are employed to
illustrate the population distribution and patterns of decentralization in the Detroit
Metropolitan region. Findings indicate that the central city and most of its adjacent
communities experienced a decline in population between 1980 and 1990 while more
distant suburban communities gained new residents. Also, these distant communities did
not only gain population but also maintain lower population density levels. This study
highlights the tendency of distant communities to attract and absorb both residents and
businesses and show that the pattern of growth in the region is such that suburban
communities are not only low density in character, but are also fast growing relative to
the central city and inner ring suburban communities.

On the other hand, the pattern of decentralization for non-whites is constrained to
communities located increasingly closer to the CBD. As this study shows, non-whites are
increasingly represented in communities within the ﬁinges of the central city of Detroit.

The third objective is to characterize the pattern of local road spending across the
study region. Two sources of local roads revenue are examined: a) the state-allocated,
which is a revenue source through the Michigan Transportation Fund and governed by

Act 51 of 1951, and b) locally-raised revenues, which are revenues for local roads

179

development from different sources including the local govemment’s general funds,
special assessments, bonds, and revenues from other “borrowings”

As Chapter Six indicates, revenue sources for local roads development that
include the state-allocated and locally-raised I) declined between the 19703 and 19803, 2)
were unevenly allocated and expended across communities, and 3) possibly reﬂect the
minimal average change in road production between the 19705 and 19803.

The average per person estimate for state-allocated revenue in a given community
amounted to $59.24 in 1977 and $50.08 in 1987. Similarly, the average per person
estimate for locally-raised roads revenue amounted to $49.40 in 1977 and $16.41 in 1987.
As observed, per capita state-allocation and locally-raised revenues declined over time,
which could be a possible explanation for the minimal change in the production of local
roads stock. Figure 6.8 to 6.11 show the differential distribution and allocation of revenue
sources for local roads development in the study region while Figures 6.12 and 6.13
highlight the minimal increase in production of local roads, which tend to have occurred
in some suburban communities in Detroit.

On average, a given community in the study region had a local road density of
21.29 miles per area (in square mile) in 1977. By 1987, the average road density was
23.21 miles, which indicates a positive change of 1.82 miles. Described above, this
minimal increase in average road density reﬂects production of roads in suburban areas
rather than in central city and inner ring suburban areas in the region. It may also reﬂect
the constraints imposed by ﬁscal challenges due to declining revenue sources. However,

it may also reﬂect the fact that mature communities have more established infrastructure

180

stock. New construction of roads could have been less in mature communities than in
new developing communities, which tend to be located in suburban areas.

The distribution of the state-allocated road funds clearly reﬂects the diminishing
share of more established communities like the city of Detroit. This is not surprising
since the formula that guides road funding allocation in Michigan accounts for population
and road mileage. Therefore, communities with declining population like the city of
Detroit, which has a considerable road stock would be at a disadvantage compared to
communities that are gaining population and building base infrastructures at the same
time. The former would experience a decrease in road funding in subsequent allocation
periods while the latter will experience gains.

It follows that the formula-based road funding allocation, which is touted to be
grounded on equity considerations (Warner, 1962; Welke et al., 2000), rewards suburban
communities and penalizes more mature communities that tend to have declining
population and aging, more established road infrastructure. Because the funding
allocation clearly responds to changes in population size, suburban communities which
tend to attract people and businesses are more likely to increase state-allocated road funds
over time. The additional local road funds can help develop existing infrastructure or
develop new ones to support growth and maintain an optimum level of road service
quality, which in turn promote the relative attractiveness of a given community that in
turn could spur more growth and land use changes.

Thus, there is a simultaneous and reinforcing effect of government spending on
local roads and suburban community growth. A higher initial state-allocated road funds

could facilitate improvements in roads service and attract growth in as much as growth in

181

community can generate higher state-allocated revenues and other funding sources,
including those coming from the local, state, and federal sources.

As the ﬁnding shows, the state’s spending on local roads facilitates the pattern of
decentralization of the general population, which behavior reﬂects a pattern of mobility
tied to a preference for communities that are located increasingly away from the Detroit
CBD. This means that the formula-based road allocation rewards suburban communities
as a result of population changes but also by other factors not directly accounted for in
the formula. The formula has internal biases, which means that it implicitly accounts for
class, politics, and location not mere population count and roads infrastructure stock
(Adelaja et al., 2009). Essentially, the allocation formula reinforces patterns of
decentralization in metropolitan areas by the way it rewards suburbanites, high income
households, and Republican dominated communities. This means that the formula, with
its tendency to reward suburban communities that have largely white residents, fails to
accomplish what it intends to do. That is, “the formula is inherently limited in providing a
mechanism for an equitable road funding allocation across communities in metropolitan
areas in Michigan” (Adelaja et al., 2009).

The fourth objective of the dissertation is to test the empirical model and
determine the links between local roads and patterns of decentralization of the general or
total and non-white population while accounting for other proposed predictors such as
household median income, percent non-whites of 25 years and up with 4 years college
education, proportion of white homeowners, local road took and distance from the CBD.
These predictors reﬂect three theoretical perspectives on residential and decentralization

patterns in metropolitan areas including: spatial assimilation, neighborhood preference,

182

and place stratiﬁcation. Findings from the empirical models yield three critical

conclusions.

Conclusions

First, local roads spending from the state government facilitates the pattern of
decentralization of the general population, but fails to explain the concentration and
decentralization of the non-white population. On the other hand, the locally-raised road
revenue facilitates decentralization of the non-white population but fails to explain the
pattern of decentralization of the general population. Second, both the general and non-
white population groups suburbanize, but they exhibit avoidance from communities with
higher representation of non-whites. Third, distance from the CBD matters in the pattern
of suburbanization for both general and non-white population groups. The general
population suburbanizes with greater distance from the CBD. On the other hand, non-
whites suburbanize but tend to choose communities in closer proximity to the CBD.

An important distinction between the general and non-white population is the
signiﬁcance of socioeconomic status, more speciﬁcally income, in their pattern of
decentralization. The result shows that for the general population, increasing income
facilitates spatial mobility and explains decentralization. On the other hand, the spatial
mobility of non-whites is not explained by income or any measure of socioeconomic
status. Therefore, ﬁndings in this study support the assimilation theory in the context of
the pattern of decentralization of the general population. For non-whites, ﬁndings from
this study tend to support the neighborhood preference theory. Also, patterns of

decentralization for both general and non-white population groups did not indicate

183

support for the place stratiﬁcation perspective in the context of decentralization process
in the study region.

Finding a positive and signiﬁcant relationship between local roads spending,
speciﬁcally state-allocated road funds, and pattern of the decentralization of the general
population raises questions about the behavior that the general population actually
represents. In the context of the Detroit Metropolitan area, it is highly conceivable that
the general population is in reality the behavior of the white population group despite the
fact that the general population is the total population comprising white and non-white
residents combined. This conjecture is formed on the following basis.

First, there is an overwhelming evidence to suggest that patterns of
decentralization in US. metropolitan areas are cases of white suburbanization. As in
many metropolitan areas, Detroit experienced a high incidence of “white ﬂight,” which is
a concept that is speciﬁcally tied to the behavior of whites ﬂeeing to the suburban
communities to avoid crime, poor social services and infrastructures, and increasingly
diverse central cities. Accordingly, this study shows that “white ﬂight” has indeed
characterized patterns of decentralization in the region when as the results indicate, non-
whites pattern of decentralization is constrained to communities at or closer to the CBD.
One can therefore conceive of the behavior of the general population as one reﬂecting a
behavior of the majority white population, particularly in the case of the Detroit
Metropolitan region. However, this interpretation remains speculative since this study did
not formally investigate the relationship between local roads spending and white

decentralization.

184

Second, income explains decentralization as the ecological and spatial
assimilation theory asserts. Therefore, spatial mobility and the opportunity to live in
suburban areas are tied to one’s ability to pay and access to opportunities that afford a
continuous income stream. Since whites have continued to fare relatively best in income
and occupational structures across racial groups, they tend to have the highest
participation in the decentralization process and are more likely to live in communities
and neighborhoods with amenities and better access to job opportunities. One may
therefore interpret the ﬁnding in this study that reveals a positive and signiﬁcant effect of
household median income on the pattern of decentralization of the general population as
simply a process reﬂecting the behavior of the predominantly white population.
However, as mentioned previously, no formal analysis on whites’ decentralization was
undertaken. Therefore this interpretation is subject to future empirical investigation.

Overall, the signiﬁcance of income, preference for suburban living as indicated by
locational factors such as distance from the CBD, and preference for lower presence of
non-white population on pattern of decentralization of the general population is
consistent with previous ﬁndings about some of the causes of decentralization of the
majority white population (Bruegmann, 2005; Squires, 2002; Duany et al., 2000). Hence,
even though the general population in this study consists of the total or aggregate
population, which includes all whites and non-whites combined, the pattern of
decentralization of the general population may simply tantamount to whites’ pattern of
decentralization. With Detroit Metropolitan Area as the case study area, the

overwhelming central city-suburban polarity by race substantiates the interpretation of

185

pattern of whites’ decentralization implicit in the behavior and pattern of general
population described above.

This dissertation research, as a whole, contributes to ongoing discussions and
research about the role of public policies on transportation in metropolitan areas. The
research conﬁrms the notion of synergy and “complementarity” between the state’s local
transportation spending and highway spending in explaining patterns of decentralization.
The study shows that increases in state’s spending on local roads facilitate sprawl, hence
the ﬁrst hypothesis of the study is supported.

The state-allocation facilitates the decentralization of the general population but at
the expense of declining communities, which based on ﬁndings in this dissertation tend to
be the preferred destination communities for non-whites. It follows that the state-
allocated local road dollars facilitate the increasing diffusion of people from the central
city to an increasingly distant communities in the region. On the other hand, the locally-
raised road revenue facilitates the relative concentration of non-whites in communities
that are in close proximity to the CBD. It appears that non-whites’ decisions to relocate
are tied to geographical and racial contexts of the destination communities. Therefore, the
second hypothesis of the study is partially supported. In the context of Detroit, locally-
raised revenues, demographic composition, and locational or spatial attributes facilitate
the pattern of non-white concentration and density in a metropolitan region.

The representation of non-whites in the pattern of decentralization is therefore
tied to a community’s ability to raise funds for local roads development and maintenance,
its demographic composition, and locational characteristics. Speciﬁcally the presence of

other non-whites and the location of community with reference to distance from the

186

CBD, not state policies on local roads development, affect the participation of non-whites
in patterns of decentralization in metropolitan areas. However, as ﬁndings indicate,

locally-raised road funds do explain the pattern of non-white decentralization.

Contributions
The ﬁndings in this study offer evidence for the positive and signiﬁcant effects of

local roads spending on the patterns of decentralization in metropolitan areas. As such,
this study has shown that the state govemment’s policy on local roads development in
Michigan promotes decentralization particularly for the general population, which in the
context of Detroit may suggest whites’ pattern of decentralization. There are intellectual
merits and signiﬁcant policy and broader impacts that can be drawn from the ﬁndings in
this study. In the following paragraphs, the theoretical, methodological, and policy

implications of the study’s ﬁndings are described.

Theoretical Contributions

This dissertation study provides contributions to the theoretical knowledge base,
particularly in: a) segregation and residential mobility; and 2) decentralization and urban
sprawl. The theoretical contribution that this study makes to social geography,
particularly in the segregation and residential mobility literature rests in its emphasis of
an analysis of the decentralization phenomenon from a multi perspective but inclusive
and integrated framework. Essentially, the human ecology conceptual framework
developed in this study helps orient an analysis of decentralization that is holistic and

recognizes the multidimensionality, complexity, and dynamic nature of the

187

decentralization phenomenon. The multi perspective approach sets the current study apart
from most social geographic inquiry, particularly on decentralization, residential
mobility, and segregation, which have traditionally employed more focused approaches,
and analyses of decentralization from a singular rather than holistic and multiple
perspectives (see review in Chung & Brown, 2007).

The tendency to adopt a less holistic framework in decentralization, segregation,
and residential mobility processes can be gleaned from previous works. The empirical
evidence provide an understanding of racial differences in patterns of decentralization
from mostly economic and sociological points of view and tend to ignore the direct
relevance of government policies (particularly on local roads) and other tangible factors
that could affect decentralization. For example, ﬁndings from many empirical tests
provide support for spatial assimilation, neighborhood preference, and place stratiﬁcation
on racial differences in decentralization but lacking is a support for possible direct,
simultaneous, and reinforcing effects of government spending on local roads and
proposed explanatory factors.

By adopting an environment, population, behavior (EPB) human ecology
framework in this study, and by integrating effects of government spending and
locational attributes into conventional explanations of decentralization and segregation,
the study shows the relevance of a holistic, inclusive, and integrative approach in the
study of decentralization. The human ecology conceptual framework developed and
adopted in this study, therefore, presents a creative approach in the study of

decentralization.

188

More speciﬁcally, the contribution of this study rests in demonstrating that human
ecology still survives and can be practiced in other forms of social geographic inquiry
such as segregation and differential patterns of decentralization, not just in medical
geography as Del Casino and Marston (2006) have claimed. This study has shown that
environmental, population, and behavioral attributes are foundational elements of a
holistic and inclusive framework that is relevant for understanding a complex and
dynamic urban processes such as the decentralization phenomenon.

This study also contributes to the knowledge base on decentralization and urban
sprawl by demonstrating that local roads and the underlying pattern of govermnent
spending explain population decentralization in metropolitan areas. This ﬁnding
highlights the plausible complementary and synergistic effect of local roads spending on
highway spending in light of the trend of people and job mobility that has characterized
many metropolitan areas including Detroit. This pattern of mobility has coincided with
the implementation of some roads spending policies such as the Act 51 of 1951 that
governs policies on roads and road building in Michigan, which gives a reason to believe
that indeed, government policies on roads development reinforces and facilitates patterns
of decentralization.

As this study has shown, there is some evidence to support the claim that
infrastructure spending in general, including local roads spending, affects growth. The
beneﬁts of such growth accrue disproportionately to certain segments of the population

and geographies, mostly favoring the white majority and suburban communities.

I89

Methodological Contributions

While this study did not introduce new methods of data collection and analytical
approach to the study of decentralization and segregation, and did not employ a rigorous
spatial approach to the study of patterns of decentralization, the emphasis on spatial
regression framework signiﬁes a recognition for the relevance of space in social
geographic analysis of complex urban processes such as decentralization and segregation.
The emphasis on spatial effects in previous studies of decentralization and segregation
was less emphasized.

In previous segregation studies, a proclivity to description of patterns (Massey &
Denton, 1988; Darden, 1985, 1987; Darden & Kamel, 2000; Chung & Brown, 2007)
rather than prediction of causes occupied many urban scholars. When predictive approach
was employed, the role of space in predictive empirical modeling of segregation and
decentralization was often ignored and spatial regression framework was hardly
employed.

As this study has shown, and increasingly in some contemporary works on
decentralization and segregation, the additional emphasis on spatial regression framework
offers a more robust analytical approach that addresses some gaps in previous predictive

modeling approaches that primarily rely on standard multiple regression techniques.

Policy and Broader Impacts
Beyond the theoretical and methodological implications, ﬁndings in this study
have policy implications and potential contributions to the broader society. Given the

positive and signiﬁcant effect of local roads on patterns of decentralization and its

190

implications for sustaining deep separation across population groups by race and class,
the role of government action on the pattern of decentralization in metropolitan areas
deserves considerable attention. Speciﬁcally, there is a critical need for a policy review of
the mechanisms that underlie the state’s allocation of local roads revenue.

As presented in Chapter Five, the mechanism for allocating local roads revenues
from the state government still governs the current distribution. It is expected, therefore,
that the effects of this mechanism will sustain the uneven growth in metropolitan areas
and hasten land use change as has been the experience in Michigan metropolitan areas in
the past several decades. There is a need to closely examine and adopt changes to the
allocation formula that explicitly accounts for other factors not directly recognized in the
current formula. Unless a considerable effort to change the current mechanism of
distribution of local road funds, patterns of differential levels of decentralization for the
general and non-white population would persist.

Findings in this study are expected to contribute to future discussions related to
any proposed changes of the allocation formula. Essentially, the evidence of the positive
relationship between local roads spending and decentralization presented in this study
should motivate state policy makers, local communities, and local leaders to reassess the
current road spending allocation mechanism and propose changes to it. It is expected that
with a changed allocation formula, communities and their residents could usher a more
equitable sharing of state-allocated local road revenue, which in turn could facilitate
better opportunities for developing local road stock particularly in declining and more

ﬁscally challenged communities in the study region.

191

Limitations and Future Work

This dissertation research provides evidence for the role of government in
facilitating the sustained separation between population groups. The results from this
study conﬁrm the unintended consequences of government policies on patterns of growth
in metropolitan areas. Speciﬁcally, the dissertation shows that local roads spending
matters in shaping the patterns of decentralization by race in metropolitan areas.

While intellectual, methodological, and policy implications can be drawn from
this study, there are also a number of limitations that should be recognized. These
limitations on theoretical orientation, data and methods, and approach serve as caveats for

interpreting the results with caution.

Framework

While the human ecological framework enables a holistic and more inclusive
analysis of a given phenomenon, the analysis pursued in this study did not explore the
hierarchy of possible causes of patterns of decentralization. Essentially, this study focuses
on the cross-sectional variation of patterns of decentralization and their causes but does
not account for hierarchical aspects of such causes, and no causal factors beyond the local
and regional level are investigated. The current framework adopted lacks recognition of
the reality that causes of patterns of decentralization are complex and interrelated. These
causes operate at different levels and geographies and at different times.

In the Detroit Metropolitan region, for example, deindustrialization, race
relations, government policies, and global crises are some of interacting and reinforcing

causes of patterns of decentralization in the region (Sugrue, 1996; Darden et al., 1987).

192

The deindustrialization process that occurred through the late twentieth century is just
one of the roots that can be traced to the decentralization and inevitable decline in the city
of Detroit. It resulted in massive shutdown of car manufacturing factories in the city, and
the creation of new and relocation of manufacturing centers in suburban communities that
triggered an unprecedented job loss and devastation in the central city. The job loss and
devastation in cities like Detroit has been reinforced with the existence of strong militant
union in manufacturing cars that hastened deindustrialization. Deindustrialization,
essentially, changed the urban geography of Detroit. Consequently, people dispersed and
redistributed across the metropolitan landscape as the working population followed the
new job opportunities available outside of the city boundary.

Racial animosity in the city has also changed Detroit. The intense competition for
jobs and other resources exacerbated the smoldering racial divide in the city that led to
deadly riots, most notably the riot in 1967 that left 43 people dead and over 2,000
buildings burned (Sugrue, 1996).

Beyond the conﬁnes of the city, broader causal factors that affected the car
industry in Detroit and throughout the U. S. have had devastating effects on Detroit’s
population. In particular, during the 19703 and 19803, the energy crisis, rising fuel cost,
and international market competition with Japan and Germany were signiﬁcant factors
explaining the changes in Detroit’s metropolitan landscape. Further, as this study shows,
the state’s spending on local roads along with highways development by the federal
government have had a decentralizing effect on Detroit. The conﬂuence of these causal

factors and numerous others at the local, regional, and global levels has created a

193

landscape in Detroit characterized by high level of decentralization and separation of
people by race and class.

Given the complex, multi level causes of decentralization in metropolitan areas
like Detroit, future studies may employ theoretical frameworks that lend an analysis of
decentralization that not only recognizes cross-sectional variation but also causes and
effects of decentralization through space and time. The alternative frameworks may also
incorporate and emphasize the role of power relations shaping decentralization, which is
relevant for analyses of phenomena and processes involving government policies that
generally address public goods production and provision. These alternative ﬁameworks
will be useful in expanding the scope and orientation of the current study.

There could be other alternative frameworks and analytical approaches that can be
employed for extending this study. One such approach is to focus on focus on mediating
and moderating factors of the relationship between local roads and patterns of
decentralization by race. Another is to employ simultaneous equation modeling as an
alternative approach for determining causal relationships between phenomena that are
jointly dependent. The range of possible approaches and frameworks for future studies is
open. However, the utility of ﬁndings in this study cannot be undermined. This study
serves as benchmark for future studies of the relationship between local roads and
decentralization for which a meaningful comparison between past and future ﬁndings can

be made.

194

Data and Methods

The fact that no further disaggregation was made to isolate the different patterns
of decentralization and mobility tendencies of the different racial groups in the study
region presents a limitation to this study. By categorizing all non-white racial groups into
one group, ﬁndings may not represent the patterns of decentralization for each separate
population groups, particularly since each population group may demonstrate different
mobility patterns. This means that within the non-white population group, which
represents all individuals in the community other than whites of Caucasian descent, each
group could have different abilities to pay and different opportunities for spatial mobility.

Also, each population group tends to exhibit “own race” preference in selecting a
residential location. Therefore, the aggregation of all racial groups into one category may
have obscured the potential response effects of local roads spending on racial patterns of
decentralization by race.

Another limitation of the study is the fact that other policy-induced or supported
amenity variables were not considered in the analysis. This means that the study lacks
considerations for other basic amenities such as school, sewer, police protection, and
other basic services, which have been found to explain decentralization and urban sprawl.
In particular, impacts of school spending on patterns of decentralization is a highly
relevant topic and may shed insights into the role of government spending policies in
patterns of growth and suburban sprawl.

The relevance of school spending on decentralization is noteworthy. Schools in
suburban communities continue to attract household relocation and are considered

amenities that make some communities preferred destinations, particularly for families

195

with school-aged children. If schools are amenities that attract families to relocate in
suburban communities, we should expect to see more non-whites with school-aged
children relocating in suburban communities. Yet, the participation of non-whites in
decentralization and their presence in suburban schools have remained low and less than
impressive compared to whites. Therefore, other factors that represent government
spending must be pursued and incorporated for a more comprehensive analysis of
impacts of government spending on decentralization.

The challenge that researchers could face is not in the different amenity variables
to include but on the availability of information that would lend an analysis that meets
consistency in scale and time of observation. As this study has uncovered, information on
school spending, local roads spending, police protection, and other amenity variables are
collected and stored in different ways, each corresponding to particular geographies, and
institutional needs and requirements. For example, school districts and local governments
have different administrative and/or political boundaries. Therefore, the available
information reﬂects these boundary differences.

Another limitation of this study rests in the unavailability of data that can be used
as measurable attributes for select proposed causal factors of decentralization. For
example, the study used proportion of white home owners as a proxy variable for racial
discrimination. This poses some issues since whites’ home ownership as a variable for
racial discrimination may suggest different processes and may lend different
interpretations other than racial discrimination. It can represent a measure of a
community’s socio-economic status and quality of life since home ownership may

indicate a particular level of income, community stability, levels of sense of community,

196

among others. Nonetheless, the application and use of home ownership as proxy for
racial discrimination in previous studies justiﬁes its use in the present study in lieu of
more direct information about racial discrimination that are often obtained by
sophisticated and expensive approaches such as paired testing approach (National
Research Council, 2002).

Related to this limitation is the difﬁculty associated with a lack of available
information that establishes a direct link between typology of roads (i.e. local versus
major streets) that speciﬁcally induces decentralization, and types of roads spending (i.e.
state versus locally-raised) and subsequently patterns of decentralization. The ideal
scenario for the study was to establish a direct link between speciﬁc roads spending
sources (i.e. local, state, and federal) and types of roads that are directly sprawl inducing
(i.e. newly constructed roads that extend or support new developments). However,, it was
not possible to determine whether a particular local roads revenue directly supported the
development of a particular roads system.. Nonetheless, the data that were available,
which were aggregate expenditure on total local roads by different levels of government
were useful in determining the effects of different spending sources on patterns of
decentralization. The most available information about roads spending was able to
provide some evidence of the relationship between local roads and patterns of
decentralization. The evidence shows that local roads spending is associated with
decentralization and sustained separation between whites and non-whites in the context
of the Detroit Metropolitan region.

Given these data limitations, future studies may beneﬁt from undertaking a more

qualitative approach to an analysis of the relationship between local roads spending and

197

patterns of decentralization. Efforts to seek direct a information from local communities
on their sources of revenue for local roads development and how these are expended
should provide a more in-depth analysis of the effects of roads and road spending on
decentralization.

Another glaring limitation of the study is the limited time period covered in the
analysis. The present study looked at a decade worth of information and may not be
adequate for generalizing ﬁndings beyond the period covered. Findings may not be a
representative for all time periods. Also, this study has a speciﬁc focus on Detroit, which
similar to the limited time covered in this study, raises questions about generalizability of
ﬁndings. The study is essentially, a micro study of patterns of decentralization by race in
space and time. Therefore, consideration for longitudinal data analysis and replication of
the study in different regions in the state of Michigan and in other states in the U. S.
should beneﬁt our understanding of the role roads and government policies on roads play
in decentralization. Future studies can beneﬁt from panel data analysis that could account
for the effects of past multi-year local road spending on current patterns of
decentralization.

This study has sought to unravel relationships between two complex processes
from a multiple yet integrated perspectives. The study of the relationship between local
roads spending and patterns of decentralization elucidates the unintended consequences
of government policy on different segments of the population. As this study has shown,
state-allocated roads spending shapes the separation of people by race and class in

metropolitan areas. What is more critical, however, is not the patterns of decentralization

198

that are shaped by disproportionate allocation of local roads revenue but the lack of
initiatives toward changing the mechanism underlying the pattern of allocation.

The tasks for changing the mechanism for local road spending allocation would
be daunting, complex, and requires collective action. Who, where, and how these tasks
should commence remain to be sorted out. This study, however, may offer a small

beginning to a very big and signiﬁcant task.

199

APPENDICES

200

Appendix A. List of Study Communities (cities and villages) in the Detroit

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Metropolitan Area
Name of city County GEO_NAME
Allen Park Wayne Allen Park city
Armada Macomb Armada
Auburn Hills Oakland Auburn Hills city
Belleville Wayne Belleville city
Berkley Oakland Berkley city
Beverly Hills Oakland Beverly Hills
Bingham Farms Oakland Bingham Farms
Birmingham Oakland Birmingham city
Bloomfield Hills Oakland Bloomfield Hills city
Center Line Macomb Center Line citL
Clarkston Oakland Clarkston City
C lawson Oakland Clawson city
Dearbom Wayne Dearbom city
Dearbom Heights Wayne Dearbom Hetghts city
Detroit Wayne Detroit city
Eastpointe Macomb East Detroit city
Ecorse Wayne Ecorse city
Farminiton Oakland Farmington city
F armington Hills Oakland Farmington Hills city
F emdale Oakland F emdale city
Flat Rock Wayne Flat Rock city
Franklin Oakland Franklin
Fraser Macomb Fraser city
Garden City Wayne Garden City
Gibraltar Wayne Gibraltar city
Grosse Pointe Wayne Grosse Pointe city
Grosse Pointe Farms Wayne Grosse Pointe Farms
Grosse Pointe Park Wayne Grosse Pointe Park
Grosse Pointe Shores Wayne Grosse Pointe Shores city
Grosse Pointe Woods Wayne Grosse Pointe Woods
Hamtramck Wayne Hamtramck city
Harper Woods Wayne Harper Woods city
Hazel Park Oakland Hazel Park city
Highland Park Wayne Highland Park city
Holly Oakland Holly
Huntington Woods Oakland Huntington Woods city
Inkster Wayne Inkster city
Keego Harbor Oakland Keego Harbor city
Lake Angelus Oakland Lake Angelus city
Lake Orion Oakland Lake Orion
Lathrup Village Oakland Lathrup Village city
Leonard Oakland Leonard
Lincoln Park Wayne Lincoln Park city

 

201

 

Appendix A. Cont’d.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Livonia Wayne Livonia city
Madison Heights Oakland Madison Heights city
Melvindale Wayne Melvindale city
Memphis Macomb Memphis

Milford Oakland Milford

Mt Clemens Macomb Mount Clemens city
New Baltimore Macomb New Baltimore city
New Haven Macomb New Haven
Northville Oakland Northville city
Novi Oakland Novi city

Oak Park Oakland Oak Park city
Orchard Lake Oakland Orchard Lake Village
Ortonville Oakland Ortonville

Oxford Oakland Oxford

Pleasant Ridge Oakland Pleasant Ridge city
Plymouth Wayne Plymouth city
Pontiac Oakland Pontiac city
Richmond Macomb Richmond city
River Rouge Wayne River Rouge city
Riverview Wayne Riverview city
Rochester Oakland Rochester city
Rochester Hills Oakland Rochester Hills city
Rockwood Wayne Rockwood city
Romeo Macomb Romeo

Romulus Wayne Romulus city
Roseville Macomb Roseville city
Royal Oak Oakland Royal Oak city
South Lyon Oakland South Lyon city
Southﬁeld Oakland Southﬁeld citL
Southgate Wayne Southgate city

St Clair Shores Macomb St. Clair Shores city
Sterling Heights Macomb Sterling Heights city
Sylvan Lake Oakland Sylvan Lake city
Taylor Wayne Taylor city

Trenton Wayne Trenton city

Troy Oakland Troy city

Utica Macomb Utica city

Walled Lake Oakland Walled Lake city
Warren Macomb Warren city

Wayne Wayne Wayne city
Westland Wayne Westland city
Wixom Oakland Wixom city
Wolverine Lake Oakland Wolverine Lake
Woodhaven Wayne Woodhaven city
Wyandotte Wayne Wyandotte city

 

202

 

Appendix B. Dependent and Independent Variables: Untransformed

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

70 . 40 '
60 d 35 "
a so. a 30-
: E
Q Q 25 q
5 40 " 5
8- 8- 20 'l
a 30 ' a 15 .
20 'l 10 .
10 ‘ 5 ‘
0 - 0 .
%_DENSCH ' %_NWLQCH
35 20 ‘
30 'i >1 16 II
5‘ 25 . 0
5 5 12 ,
g 20 " ﬂ
{3;- 15 . E 8 -
k‘ 10 .
4 d
5 II
0 ~ 0 '
NWDENSCH LOCRDENS77
40 - 45 '
35 . 40 ‘
5" 30 I >3 35 '
= 25 - 3 3° '
0 Q)
5 20 - 5 25 '
g- 15 8' 20 '
ta 1 E 15 -
l0 ‘ 10 ..
5 u 5 u
o l 0 -
LOCALPP77 MEDINC80

 

 

 

 

203

 

 

 

Appendix B. Dependent and Independent Variables: Untransformed (Cont’d.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

60 - 90 .
50 . so .
70 .
>> >a
: 40- a 60 .
O 0 50 .
5 30 . 5
g g 40 -
g. 20 . Ln 30 .
10 - :3
0 'l 0 q
MTFPP77 %_wrromrzso
70 ‘ 25 .
60 ' 20 .
e 50- 2‘
5 40 . g 15 i
’ a
g 30 ' a 10 i
‘" 20 . 5
10 4
o 0 +
%_NWHITE80 DISTDET
so -
40 j
>z
U
5 30 -
5
g 20 -
[a
10 -
0 i ..... Ia
%_CONW80

 

 

 

204

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 2 E M: M: 2 3 2 +8 $2.8
g. 85. ”3. NE. SN. 3%. a s a» $323.18: ox.
_ «8.- v8.- .mmm. me. ”2.. NS; t 8328 .x. E
mm an an an mm mm 2 23— 2:85
Se. Sm. :o. o2. mm». a 562: 228.8:
_ 25. So.- 29. at. «8.- t 802592 GO
2- ns an on on 2 m6
moo. coo. So. Non. Q 3 can 38.. .82 .80..-
_ :N _ n- red..- :emm.- 35. t 555963 E
3- wwo. an . my. W m 02.26: .32 woman
_ 3m. MM ”- van. a #:80— :8qu com
: . Ema-.53 5
on as as 2 m 2:55..
25. 3%. Q memo. 88m .538 com
_ :34. :8. t Ema-:2 5
mm mm Z owe—£0
m8. Q bicep coca—smog .X.
_ mm. L Ema-51.x. 5
mm Z E3830 cosmos
Q 833-8: E omega .X.
_ t 5332...; 5
8: € 5 5 3v 5 as 5 € 3 032:9
nae-5:239 83333 U hazwmﬁvt

205

:3 VA 1. .85 Va ..

”.96— oogommcwmm 2 3205000 cows—8.50 new-80m

 

 

 

 

 

 

 

S a «w 3 2 E M: 2 8o. 2 2 a8

5. 2.. ms. to. So. 2 _. 8o. 3 o««. a :28 5% 8:55

_ m8; «0 _. R _ .- m2.- ..m«c.- a _. ..S«. :53. «8.- t 5955 8:
a «a 2 3 E 3 an 3 a z -

8o. So. «3. «.3. am. a G. o _ o. 25. a amtmwmrmcnﬂxﬁ”
_ :89- SN. «.8.- So.. $o.- EV.- .EN,. :08.- t o

«w «a «w M: R 8a. «m «a 2 maﬁa

N8. cm. a». new. R m8. _«o. a £283er Ox.

_ .m.«~.- NM _. on... so. So. SN. .MNN. t omeOza .x. a:

 

 

 

 

 

 

 

 

 

 

.93.:er netﬂohenv 83.32M— U «£5223.

206

REFERENCES

207

REFERENCES

Adelaja, A., Campos, A., & Hailu, Y. (2009). Equity and politics in road-spending
allocation in Michigan: Implications for metropolitan affairs. In R. J elier & G.
Sands (Eds.), Sustaining Michigan, Metropolitan policies and strategies. East
Lansing: Michigan State University Press.

Alba, R. D., & Logan, J. R. (1993). Minority proximity to whites in suburbs: An
individual level analysis of segregation. American Journal of Sociology, 98(6),
1388-1427.

Anselin, L. (2005). Exploring spatial data with GeoDA: A Workbook: Center of
Spatially Integrated Social Science.

Bailey, T. & Gattrell, A. (1995). Interactive Spatial Data Analysis. New York: Pearson
Education Limited.

Barrows, H. H. (1923). Geography as human ecology. Annals of the Association of
American Geographers, 13(1), 1-14.

Baum-Snow, N. (2007). Did highways cause suburbanization? Quarterly Journal of
Economics, 122(2), 775.

Bayoh, I., Irwin, E. G., & Haab, T. (2006). Determinants of residential location choice:
How important are local public goods in attracting homeowners to central city
locations? Journal of Regional Science, 46(1), 97-120.

Belden, Russonello, & Stewart. (1998). Choices between asphalt and nature: American
discuss sprawl. Washington DC: Belden Russonello & Stewart Research and
Communications. '

Berechman, J ., Ozmen, D., & Ozbay, K. (2006). Empirical analysis of transportation
investment and economic development at state, county and municipality levels.
Transportation, 33(6), 537-551.

Berry, B. J. L. & Garrison, W. (1958). The functional bases of the central place
hierarchy. Economic Geography, 34, 145-154.

Berry, B. J. L. & Horton, F. E. (1970). Geographic perspectives on urban systems.
Englewood Cliffs, NJ: Prentice Hall.

Berry, B. J. L. (1980). Urbanization and counter urbanization in the United States. Annals
of the American Academy of Political and Social Science, 45] (Changing cities: A
challenge to planning), 13-20.

Berube, A., & Forman, B. (2002). Living on the edge: Decentralization within cities in
the 1990s. Washington DC: The Brookings Institution, Center on Urban and
Metropolitan Policy.

208

Beyer, H. L. (2004). Hawth's Analysis Tools for ArcGIS.

Bivand, R., Anselin, N., Vemat, A., Carvalho, M., Dray, S., Halversma, R., et al. (2005).
The spdep Package. Version 0.3-12. (Version 4/19/2005).

Bivand, R., & Lewin-Koh, N. (2005). The maptools Package. Version 0.4-13.

Black, W. (2003). Transportation: A geographical analysis. New York: The Guilford
Press.

Boamet, M. G. (1998). Spillovers and the locational effects of public infrastructure.
Journal of Regional Science, 38(3), 381-400.

Bobo, L.D. & Zubrinsky C. 1996. Attitudes on residential integration: perceived status
differences, mere in-group preference or racial prejudice? Social Forces. 74, 883-
909.

Borden, R. (2008). A brief history of SHE: Reﬂections on the founding and ﬁrst twenty
ﬁve years of the Society for Human Ecology. Human Ecology Review, 15(1), 95-
108.

Bouma-Doff, W. L. 2007. Involuntary isolation: ethnic preferences and residential
segregation. Journal of Urban Aﬂairs.29 (3), 289-309.

Boume, L. (1989). Are new urban forms emerging? Empirical tests for Canadian urban
areas"

Brown, L. A. & Chung, S. Y. 2008. Market-Led Pluralism: Rethinking Our
Understanding of Racial/Ethnic Spatial Patterning in US. Cities. Annals of the
Association of American Geographers. 98 (1), 180-212.

Brueckner, J. K. (2005). Transport subsidies, system choice, and urban sprawl. Regional
Science and Urban Economics, 35(6), 715-733.

Bruegmann, R. (2005). Sprawl: A compact history. Chicago: University of Chicago
Press.

Bund, J .W.W. (1920). The evolution of the road. The Journal of the Royal Society for the
Promotion of Health, 41, 309-314.

Bureau of Labor Statistics, available at http://www.bls.gov, last accessed on February 25,
2009.

Burchell, R. W., Downs, A., & Seskin, S. (2002). Title Costs of sprawl -2000 (No. TCRP
Report #74): Transportation Research Board Publication.

Burchﬁeld, M., Overman, H. G., Puga, D., & Turner, M. A. (2005). Causes of sprawl: A
portrait from space. The Quarterly Journal of Economics, 121(2), 587-633.

209

Burgess, E. (1925). The growth of a city: An introduction to a research project. In R.
Park, E Burgess, & R. McKenzie (Eds), The City. Chicago: University of
Chicago Press.

Burgess, E. W. (1928). Residential segregation in American cities. Annals of the
American Academy of Political and Social Silence, 140, 105-115.

Byun, P., Waldorf, B. S., & Esparza, A. X. (2005). A revisionist model of
suburbanization and sprawl: The role of political fragmentation, growth control,
and spillovers. Journal of Planning Education and Research, 24, 252-264.

Cadot, 0., Roller, L. H. & Stephan, A. (2006). Contribution to productivity or pork
barrel? The two faces of infrastructure investment. Journal of Public Economics,
90(6-7), 1133-1153.

Carlino, E. G., & Mills, E. S. (1987). The determinants of county growth. Journal of
Regional Science, 27(1), 39-54.

Carlino, G. (2000). From centralization to deconcentration: People and jobs spread out.
Federal Reserve Bank of Philadelphia Business Review (N ovember/December),
15-27.

Carruthers, J. I. (2003). Growth at the fringe: The inﬂuence of political fragmentation in
United States metropolitan areas. Papers in Regional Science, 82(4), 475-499.

Carruthers, J. I., & Ulfarsson, G. F. (2003). Urban sprawl and the cost of public services.
Environment and Planning B-Planning & Design, 30(4), 503-522.

Catton, W. (1987). The world’s most polymorphic species.” Bioscience, 37(6):4l3-4l9.

Catton, W. R. (1994). Foundations of human ecology. Sociological Perspectives, 37(1),
75-95.

Cervero, R. (2003). Road expansion, urban growth, and induced travel - A path analysis.
Journal of the American Planning Association, 69(2), 145-163.

Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: Density, diversity,
and design. Transportation Research -D, 2(3), I99-219.

Chandra, A., & Thompson, E. (2000). Does public infrastructure affect economic
activity? Evidence from the rural interstate highway system. Regional Science
and Urban Economics, 3 0(4), 457-490.

Charles, C.Z. (2002). Comfort zones: immigration, assimilation, and the neighborhood
racial-composition preferences of Latinos and Asians. Paper presented at Annual
meeting of the American Association, Chicago.

Charles, C. Z. (2003). The dynamics of racial residential segregation. Annual Review of

210

Sociology, 29, 167-207.

Charles, C. 2., Dinwiddie, G., & Massey, D. (2004). The continuing consequences of
segregation: Family stress and college academic performance. Social Science
Quarterly, 85(5), 1353-1373.

Chung, S., & Brown, L. A. (2007). Racial/ethnic residential sorting in spatial context:
Testing the explanatory frameworks. Urban Geography, 28(4), 312-339.

Clark, T., Lloyd, R., Wong, K., & Jain, P. (2002). Amenities drive growth. Journal of
Urban Affairs, 24(5), 493-515.

Clark, W. (1991). Residential preferences and neighborhood racial segregation: a test of
the Schelling segregation model. Demography, 28, 1-19.

Clark, W., & Ware, J. (1997). Trends in residential segregation by socioeconomic status
in California. Urban Affairs Review, 32, 825-843.

Clark, W. A. (1992). Residential preferences and residential choices in a multi-ethnic
context. Demography, 29, 451-466.

Clark, W. A. V. (2007). Race, Class, and Place: Evaluating Mobility Outcomes for
African Americans. Urban Affairs Review, 42(3), 295-314.

Clark, T., & Fossett, M. (2008). Understanding the social context of the Schelling
segregation model. Proceedings of the National Academy of Sciences, 105(11),
4109-4114.

Clark, W. A. V. (2009). Changing Residential Preferences across Income, Education, and
Age: Findings from the Multi-City Study of Urban Inequality. Urban Affairs
Review, 44(3), 334-355.

Clay, P. L. (1979). The Process of Black Suburbanization. Urban Affairs Review, 14(4),
405-424.

County Roads Association of Michigan, available at
http://www.micountyroads.org/page.cfm/6/ last accessed on March 10, 2009.

Crane, R., & Crepeau, R. (1998). Does neighborhood design inﬂuence travel? A
behavioral analysis of travel diary and GIS data. Transportation Research -D,
3(4), 225-238.

Crowder, K. (2000). The racial context of white mobility: An individual level assessment
of the white ﬂight hypothesis. Social Science Research, 29(2), 223-257.

Darcin, M., & Darcin, S. (2006). Relationship between quality of life and child trafﬁc
fatalities. 39(4), 826-832.

211

Darden, J. (1973). Afro-American in Pittsburgh: the Residential Segregation of a People:
Lexington: D. C. Heath.

Darden, J. T. (l 977). Blacks in the suburbs: Their numbers are rising but patterns of
segregation persist. Vital Issues (December 1977).

Darden, J. T. (1985). The residential segregation of Native Americans in cities and
suburbs of Michigan. The East Lakes Geographer, 20, 84-88.

Darden, J. T. (1986). The signiﬁcance of race and class in residential segregation.
Journal of Urban Affairs, 8, 49-55.

Darden, J. T. (1987). Socioeconomic status and racial residential segregation: Blacks and
Hispanics in Chicago. International Journal of Comparative Sociology, 28(1 &
2), 1-13.

Darden, J. T. (1989). Blacks and other racial minorities: The signiﬁcance of color in
inequality. Urban Geography, 10(6), 562-577.

Darden, J. T., Kamel, S. M. (2000). Black residential segregation in suburban Detroit:
Empirical testing of the ecological theory. Review of Black Political Economy,
27(3), 103-123.

Darden, J. T., Kamel, S. M. (2000). Black residential segregation in the city and suburbs
of Detroit: Does socioeconomic status matter? Journal of Urban A ffairs, 22(1), 1.

Darden, J. T. (2001). Detroit Divided. Association of American Geographers. Annals of
the Association of American Geographers, 91(3), 582.

Darden, J. T. (2003). Residential segregation: The causes and social and economic
consequences. In C. Stokes & T. Melendez (Eds), Racial Liberalism and the

Politics of Urban America (pp. 321-344). East Lansing, MI: Michigan State
University Press.

Darden, J. T. (2004). The significance of white supremacy in the Canadian metropolis of
Toronto. Lewiston, N.Y.: Edwin Mellen Press.

Darden, J. T., Hill, R., Thomas, R., & Thomas, J. (1987). Detroit, race and uneven
development. Philadelphia: Temple University Press.

Davis, B. J. J. (1996). Spatial patterns and distributions of the locations of racial groups
and employment opportunities in Michigan, 1960-1990. Unpublished 1379715,
Michigan State University, United States -- Michigan.

de Boer, E. (1986). Transport Sociology: social aspects of transport planning. Oxford:
Pergamon.

212

DebRoy, S. R., Bivand, D., Murcoch, N., Lewin-Koh, F. Warmerdam, T. Lumley, D., et
a1. (2005). The foreign Package. Version 0.8-7.

Del Casino, V., & Marston, S. (2006). Social geography in the United States: everywhere
and nowhere. Social and Cultural Geography, 7(6), 995-1009.

DeRango, K. (2001). Black-White Segregation, Discrimination, and Home Ownership.
Upjohn Institute Staff Working Paper 01-71 (WP01-7l), W.E. Upjohn Institute
for Employment Research.

Dietz, T.,& Rosa, E. (1994). Rethinking the environmental impacts of population, afﬂuence,
and technology.” Human Ecology Review, 1, 277-300.

Downs, A. (1999). Some realities about sprawl and decline. Housing Policy Debate, 10,
955-974.

Duany, A., Plater-Zyberk, E., & Speck, J. (2000). Suburban nation: The rise of sprawl
and the decline of American dream. New York: North Point Press.

Duncan, O. (1959). Human ecology and population studies. In P. Hauser & O. Duncan
(Eds), The study of population: an inventory and appraisal (pp. 678-716).
Chicago: University of Chicago Press.

Duncan, 0., & Duncan, B. (1955). Methodological analysis of segregation indexes.
American Sociological Review, 20, 210-217.

Duncan, 0., Schnore, L., & Rossi, P. (1959). Cultural, Behavioral, and Ecological
Perspectives in the Study of Social Organization. The American Journal of Sociology,
65(2), 132-153.

Duncan, O. (1964). "From social system to ecosystem." Sociological Inquiry, 31, 140-149.

Dunlap, R., Loren L., & Rosa, E. (1994). Understanding environmental problems: A
sociological perspective.” In Beat Bﬁrgenmeier (Ed), Economy, Environment, and
Technology. Armonk, NY: ME. Sharpe.

Ebner, M. H. (1985). Re-reading suburban America: Urban population deconcentration,
1810-1980. American Quarterly, 37(3), 368-381.

Ewing, R. (1997). Is Los Angeles style sprawl desirable? Journal of American Planning
Association, 63, 107-126.

Ewing, R., Pendall, R., & Chen, D. (2003). Measuring sprawl and its transportation
impacts. In Travel Demand and Land Use 2003 (pp. 175-183).

Farley, J .E. (1986). Segregated city, segregated suburbs: to what extent are they products
of black-white socioeconomic differentials? Urban Geography, 7(2), 164-171.

Farley, J. E. (1995). Race Still Matters: The Minimal Role of Income and Housing Cost

213

as Causes of Housing Segregation in St. Louis, 1990. Urban Affairs Review,
31(2), 244-254.

Farley, R. (1997). Racial Trends and Differences in the United States 30 Years After the
Civil Rights Decade. Social Science Research, 26 (3): 235-262.

Farley, J. and Squires, G. (2005). Fences and Neighbors: Segregation in 21 st-Century
Segregation in let-Century America. Contexts 4, 33-39.

Farley, R., Fielding, E., and Krysan, M. (1997). The residential preferences of blacks and
whites: A four-metropolis analysis. Housing Policy Debate, 8(4), 763-800.

Farley, R., and W. H. Frey. 1994. Changes in the segregation of Whites from Blacks
during the 19803: Small steps toward a more integrated society. American
Sociological Review 59:23—45.

Farley, R., Schuman, H., Bianchi, S., Colasanto, D. & Hatchett, S. (1978). Chocolate city,
vanilla suburbs: Will the trend toward racially separate communities continue?
Social Science Research. 7, 319-344.

Farley, R., Bianchi, S., & Colasanto, D. (1979). Barriers to the racial integration of
neighborhoods: The Detroit case. The Annals of the American Academy, 441 , 97-
113.

Fischer, M. J. (2003). The Relative Importance of Income and Race in Determining
Residential Outcomes in US. Urban Areas, 1970-2000. Urban Affairs Review,
38(5), 669-696.

Fischer, M. J. (2008). Shifting Geographies: Examining the Role of Suburbanization in
Blacks' Declining Segregation. Urban Aﬂairs Review, 43(4), 475-496.

Fisher, J. S., & Mitchelson, R. L. (1981). Forces of change in the American settlement
pattern. Geographical Review, 71(3), 293-310.

Frankenberg, E., Lee, C., & Orﬁeld, G. (2003). A Multiracial Society with Segregated
Schools: Are We Losing the Dream?

Freeman, L. (2008). Is Class Becoming a More Important Determinant of Neighborhood
Attainment for Aﬁ'ican-Americans? Urban Affairs Review, 44(1), 3-26.

Galster, G. (1992). Research on Discrimination in Housing and Mortgage Markets:
Assessment and Future Directions," Housing Policy Debate, 3(2), 639-684.

Galster, G., Hanson, R., Ratcliffe, M. R., Wolman, H., Coleman, 8., & Freihage, J.
(2001). Wrestling sprawl to the ground: Deﬁning and measuring an elusive
concept. Housing Policy Debate, 12(4), 681-717.

Garreau, J. (1992). Edge cities: Life on the new frontier: Anchor Books.

214

Garrison, W. L., Berry, B. J. L., Marble, D. F., Nystuen, J. D. & Morrill, R. L. M. (1959).
Studies of highway development and geographic change. Seattle: University of
Washington Press in cooperation with the Bureau of Public Roads of the
Department of Commerce and the Washington State Highway Commission.

GeoLytics, I. (1970-2000). Neighborhood Census Database (NCDB). East Brunswick,
NJ.

Glaeser, E. L., Khan, M., & Chu, C. (2001). Job sprawl: Employment location in US
metropolitan areas. Washington, D. C.: The Brookings Institution, Center on
Urban and Metropolitan Policy.

Glaeser, E. L., & Kohlhase, J. E. (2004). Cities, regions and the decline of transport costs.
Papers in Regional Science, 83(1), 197-228.

Goode, P. (1904). The human response to the physical environment. Journal of
Geography, 3, 333-343.

Gotham, K. F. (2002). Beyond invasion and succession: School segregation, real estate
blockbusting, and the political economy of neighborhood racial transition. City &
Community, 1(1), 83-111.

Grady, S. (2006). Racial disparities in low birth weight and the contribution of residential
segregation: a multilevel analysis. Social Science and Medicine, 63(12), 1013-
3029.

Gramlich, E. M. (1994). Infrastructure investment - a Review-Essay. Journal of
Economic Literature, 32(3), 1176-1196.

Grannis, R. (1998). The importance of trivial streets: Residential streets and residential
segregation.

Greenburg, Z. (2009). America's emptiest cities (Publication. Retrieved February 16,
2009, from Forbescom: http://www.forbes.com/2009/02/l 2/cities-ten-top-
lifestyle-real-estate_02 1 2_cities.html

Gross, M. (2004). Human geography and ecological sociology: The unfolding of a human
ecology, 1890 tol930- and beyond. Social Science History, 28(4), 575-605

Guhathakurta, S., & Mushkatel, A. H. (2000). Does Locational Choice Matter? A
Comparison of Different Subsidized Housing Programs in Phoenix, Arizona.
Urban Aﬂairs Review, 35(4), 520-540.

Guild, R. L. (2000). Inﬁ'astructure investment and interregional development. Theory,

evidence and implications for planning. Public Works Management and Policy,
4(4), 274-285.

215

Hamilton, W. (2003). Act 51 Primer: A Guide to 1951 Public Act 51 and Michigan
Transportation Funding. In Agency (Ed).

Handy, S. (2005). Smart growth and the transportation - Land use connection: What does
the research tell us? International Regional Science Review, 28(2), 146-167.

Handy, S. (2006). The road less driven. Journal of the American Planning Association,
72(3), 274.

Hanson, S. & Giuliano, G. (Eds). (2004). The geography of urban transportation (Third
ed.). New York London: The Guilford Press.

Harris, D. R. (1999). "Property Values Drop When Blacks Move in, Because...": Racial
and Socioeconomic Determinants of Neighborhood Desirability. American
Sociological Review, 64(3), 461 -479.

Harris, D. R. (2001). Why are whites and blacks averse to black neighbors?*. Social
Science Research, 30(1), 100-116.

Hartgen, D. (2003). Highways and sprawl in North Carolina. Raleigh, North Carolina

Haughwout, A. F. (1999). State infrastructure and the geography of employment. Growth
and Change, 30(4), 549-566.

Hawley, A. (1950). Human ecology: A theory of community structure. New York: Ronald
Press.

Hobbs, F., & Stoops, N. (2002). Demographic trends in the 20th century. Washington
DO: U. S. Government Printing Ofﬁce.

Hogan, D. P., & Kitagawa, E. M. (1985). The impact of social status, family structure,
and neighborhood on the fertility of black adolescents. American Journal of
Sociology 19, 825-855

Holcombe, R. G., & Williams, D., W. (2008). The impact of population density on
municipal government expenditures. Public Finance Review, 36(23), 359-373.

Hoyt, H. (1933). One hundred years of land value in Chicago. University of Chicago
Press.

Hoyt, H. (1939) The Structure and Growth of Residential Neighborhoods in American
Cities, Federal Housing Administration, Washington, DC.

Hoyt, H. (1941). Forces of urban centralization and decentralization. The American
Journal of Sociology, 46(6), 843-852.

Iceland, J ., Wienberg, D., & Steinmetz, E. (2002). Racial and ethnic residential
segregation in the United States: 1980-2000. US. Census Bureau, Census Special

216

Report, CENSR-3. Washington, DC: US. Government Printing Ofﬁce.

Iceland, J., & Wilkes, R. (2006). Does socioeconomic status matter? Race, class, and
residential segregation. Social Problems, 53(2), 248-273.

Jackson, T. D. (1994). Prejudice as a determinant of the residential preferences of whites
and Blacks. Unpublished Type of Work: Dissertation, University of Michigan.

J argowsky, P. (2002). Sprawl, Concentration of Poverty, and Urban Inequality. In G. D.
Squires (Ed.), Urban sprawl: Causes, consequences, and policy responses.
Washington, DC: The Urban Institute Press.

Johnson, M. P. (2001). Environmental impacts of urban sprawl: A survey of the
literature and proposed agenda. Environment and Planning A, 33, 717-735.

Johnston, R., Poulsen, M., & Forrest, J. (2007). Ethnic and Racial Segregation in US.
Metropolitan Areas, 1980-2000: The Dimensions of Segregation Revisited. Urban
Aﬂairs Review, 42(4), 479-504.

Jordan, 3., Ross, J. P., & Usowski, K. G. (1998). U. S. suburbanization in the 1980s
Regional Science and Urban Economics, 28, 611-627.

Journal of the Senate, Senate, Regular Session, 8 Congressional Record (1951).

Harris, D. R. (2001). Why are whites and blacks averse to black neighbors? Social
Science Research, 30(1), 100-116.

Kachigan, S. (1986). Statistical analysis: An interdisciplinary introduction to univariate
and multivariate methods. New York: Radius Press.

Katz, 3., Berube, A., Prince, R., & Smith, H. (2003). Detroit in focus: A profile from
census 2000. The Brookings Institution Center on Urban Metropolitan Policy.
Washington, DC.

Kain, J. F. (1970). Postwar changes in land use in the American city. In D. P. Moynihan
(Ed.), Toward a national urban policy. New York: Basic Books.

Kemmerling, A., & Stephan, A. (2002). The contribution of local public infrastructure to
private productivity and its political economy: Evidence from a panel of large
German cities. Public Choice, 113, 403-424.

Kent, M. M., Pollard, K. M., Haaga, J ., & Mather, M. (2001). First glimpses from the
2000 US. census. Population Bulletin, 57(2), 1-39.

Kim, S. (2007). Changes in the nature of urban spatial structure in the United States,
1890-2000. Journal of Regional Science, 47(2), 273-287.

Koles, M., & Muench, D. (2002). Why people are moving to suburbia (and beyond):

217

Examining the pull factors in the Fox Valley. Journal of Extension, 40(6).

Ladd, H. (1998). Evidence of discrimination in mortgage lending. The Journal of
Economic Perspectives, 12(2), 41-62.

Lawrence, R. A. (2003). Human ecology and its applications. Landscape and Urban
Planning, 65, 31-40.

Lieberson, Stanley. 1980. A Piece of the Pie: Blacks and White Immigrants Since
1880. Berkeley: University of California Press.

Limoges, E. (2001). Detroit and 300 years of metropolitan growth. Detroit: Southeast
Michigan Council of Governments.

Logan, J. R. (2003). “America’s Newcomers.” Lewis Mumford Center Working Paper
Series.

Lopez, R., & Hynes, P. (2003). Sprawl in the 19908: Measurement, distribution, and
trends. Urban Affairs Review, 38(3), 325-355.

Lopez, R. A., Adelaja, A. O., & Andrews, M. S. (1988). The Effects of Suburbanization
on Agriculture. American Journal of Agricultural Economics, 70(2), 346-358.

Madden, J. F. (2003). The changing spatial concentration of income and poverty among
suburbs of large US Metropolitan areas. Urban Studies, 40(3), 481-503.

Marans, R. W. & Sheehan, M. O. (2003). Understanding environmental quality through
quality of life studies: The 2001 DAS and its use of subjective and objective
indicators. What will it take to halt sprawl? Landscape and Urban Planning,
65(1-2), 73-84.

Massey, D. (1985). Ethnic residential segregation: A theoretical synthesis and empirical
review. Sociology and Social Research, 69, 315-350.

Massey, D., & Denton, N. (1985). Spatial assimilation as a socioeconomic outcome.
American Sociological Review, 50, 94-106.

Massey, D., & Denton, N. (1988). The dimensions of residential segregation. Social
Forces, 67, 281-315.

Massey, D., & Denton, N. (1989). Hypersegregation in US metropolitan areas: Black and
Hispanic segregation along ﬁve dimensions. Demography, 26, 378.

Massey, D., & Denton, N. (1993). American Apartheid: Segregation and the making of
the underclass. Cambridge: Harvard University Press.

Massey, D. S. (2005). Racial Discrimination in Housing: A Moving Target. Social
Problems, 52(2), 148-151.

218

Meade, M. (1977). Medical geography as human ecology: The dimension of population
movement. Geographical Review 67(4), 383.

Meade, M. & Earickson, R. (2000). Medical Geography. (Ed). New York, London.
Guilford Press.

Michigan Geographic Data Library available at http://www.mcgi.state.mi.us/mgdl/; last
accessed on January 28, 2009.

Michigan Land Use Leadership Council (2003). Michigan 's Land, Michigan 's Future:
Final report of the Michigan 's Leadership Council. Lansing: The State
Government of Michigan.

Mieszkowski, P., & Mills, E. S. (1993). The Causes of Metropolitan Suburbanization.
Journal of Economic Perspectives, 7(3), 135-147.

Mieszkowski, P., & Smith, B. (1991). Analyzing decentralization: The case of Houston.
Regional Science and Urban Economics 21, 183-1 99.

Mikelbank, B. A., & Jackson, R. W. (2000). The role of space in public capital research.
International Regional Science Review, 23(3), 235-258.

Moran, P. (1950). Notes on continuous stochastic phenomena. Biometrika, 3 7, 17-23.

Michigan Land Use Leadership Council (2003). Michigan ’s Land, Michigan 's Future:
Final report of the Michigan's Leadership Council. Lansing: The State
Government of Michigan.

Orﬁeld, M. (2002). American metropolitics: The new suburban reality. Washington, DC:
Brookings Institution

National Association of Realtors (2008). Proﬁle of homebuyers and sellers. Available at
http://www.realtor.org/wps/wcm/connect/, last accessed on April 15, 2009)

National Research Council (2002) Measuring Housing Discrimination
in a National Study: Report of a Workshop. Committee on National Statistics.
Angela Williams Foster, Faith Mitchell, Stephen E. Fienberg, (Eds). Division of
Behavioral and Social Sciences and Education. Washington, DC: National
Academy Press.

Otterstrom, S. (2003). Population concentration in United states city-systems from 1790
to 2000: Historical trends and current phases. T ijdschrift voor Economishe en
Sociale Geograﬁe, 94(4), 477-495.

Park, R., Burgess, E. W., & McKenzie, R. (1925). The City. Chicago, IL: University of
Chicago Press.

Park, RE. (1936). Human ecology. American Journal of Sociology, 42, 1-15.

219

Pohl, D. G., & Brown, N. E. (1997). The history of roads. Paper presented to the
Association of Southern Michigan Roads Commission.

Portes, A. & Rumbaut, R. (1996). Immigrant America: A Portrait, 2"d Edition, Berkeley,
CA: University of California Press.

Power, A. (2001). Social Exclusion and Urban Sprawl: Is the Rescue of Cities Possible?
Regional Studies, 35(8), 731 - 742.

Ray, 3., & Bergeron, J. (2007). Geographies of ethnocultural diversity in a second-tier
city: Moving beyond where people sleep [Electronic Version]. Our diverse cities,
4, 4 from http://canada.metropolisnet/pdfs/NDC%200ntario%20FRA4)df.

Rii, H. U. (1983). Genesis, early growth, and impact of the transportation system on
Detroit, 1805-1900 Unpublished 8315500, Michigan State University, United
States - Michigan.

Rodrigue, J .-P., Comtois, C., & Slack, B. (2006). The Geography of transport systems.
London: Routledge.Rusk, D. (2003). Cities without suburbs: A Census 2000
update (Third ed). Washington DC: Woodrow Wilson Center Press.

Sassen, S. (2002). Global Networks/Linked Cities. New York and London: Routledge.

Saelens, B. 15., IF, S. & Frank, L. D. (2003). Enviromnental correlates of walking and
cycling: ﬁndings from the transportation, urban design, and planning literatures.
Annals of Behavioral Medicine, 25(2), 8091.

Schelling, T. (1971). Dynamic Models of Segregation. Journal of Mathematical
Sociology. 1, 143-186.

SEMCOG. (2002). Historical population and employment by minor civil division,
Southeast Michigan. Detroit: The Southeast Michigan Council of Governments.

SEMCOG. (2003). Land use change in Southeast Michigan: Causes and consequences.
Detroit: Southeast Michigan Council of Governments.

SEMCOG. (2008). Population and households in Southeast Michigan, 2000-2008.
Detroit: Southeast Michigan Council of Governments.

Si ggerud, K. (2004). Surface transportation: Many factors affect investment decisions
(No. GAO-04-744). Washington, DC.

Statistical Package for Social Science soﬁware (SPSS version 14.0)

Squires, G. D. (2002). Urban sprawl: causes, consequences, and policy responses.
Washington, D. C.: The Urban Institute Press.

220

Squires, G. D. (2003). Racial proﬁling, insurance style: Insurance redlining and the
uneven development of metropolitan areas. Journal of Urban Aﬂairs, 25(4), 391—
410.

Sugrue, T. (1996). The Origins of the Urban Crisis: Race and inequality in postwar
Detroit. Princeton: Princeton University Press.

Sugrue, T. (2005). The Origins of the Urban Crisis: Race and inequality in postwar
Detroit (Princeton Studies in American Politics). Princeton: Princeton University
Press.

Sultana, S., & Weber, J. (2007). J oumey-to-work patterns in the age of sprawl: Evidence
from two midsize southern metropolitan areas. The Professional Geographer,
59(2), 193-208.

Teaford, J. C. (1993). Cities of the heartland: The rise and fall of the industrial Midwest.
Bloomington and Indianapolis: Indiana University Press.

Timberlake, J ., & Iceland, J. (2007). Change in Racial and Ethnic Residential Inequality
in American Cities, 1970-2000

Toit, d L., Cerin, E., Leslie, E., & Owen, N. (2007). Does walking in the neighborhood
enhance local sociability? Urban Studies, 44(9), 1677-1695.

US Census Bureau, 2000 available at http://quickfacts.census.gov/qfd/states/26/ last
accessed on February 15, 2009.

US. Bureau of the Census, Census of Population and Housing. Washington, DC: &
Census.

Ulfarsson, G. F., & Carruthers, J. I. (2006). The cycle of fragmentation and sprawl: a
conceptual framework and empirical model. Environment and Planning B-
Planning & Design, 33(5), 767-788.

Vojnovic, I. (2000). Shaping Metropolitan Toronto: a study of linear infrastructure
subsidies, 1954-66. Environment and Planning B—Planning & Design, 27(2), 197-
230.

Vojnovic, I., J ackson-Elmoore, C., Holtrop, J ., & Bruch, S. (2006). The renewed interest
in urban form and public health: Promoting increased physical activity in
Michigan. Cities, 23(1), 1-17.

Voss, P. R., & Chi, G. O. (2006). Highways and population change. Rural Sociology,
71(1), 33-58.

Wilkes, R., & Iceland, J. (2004). Hypersegregation in the twenty-ﬁrst century.
Demography, 41(1), 23-36.

221

Williams, R., Nesiba, R., & McConnell, E. D. (2005). The Changing Face of Inequality
in Home Mortgage Lending. Social Problems, 52(2), 181-208.

Warner, S. L. (1962). The distribution of highway user taxes in Michigan. Unpublished
MA, Michigan State University, East Lansing.

Welke, R., Hart, S., Johnson, R., Hoffman, P., Tatter, J ., Kelly, T., et al. (2000).
Transportation funding for the 21st century. Lansing, MI: The State Government
of Michigan.

Wilson, W. 1987. The Truly Disadvantaged: The Inner City, The Underclass, and Public
Policy. Chicago: University of Chicago Press

Wilson, E. H., Hurd, J ., Civco, D., Prisloe, M. & Arnold, C. (2003). Development of a
geospatial model to quantify, describe, and map urban growth. Remote Sensing of
the Environment, 86, 275-285.

Yago, G. (1983). The Sociology of Transportation. Annual Review of Sociology, 9, 171-
190.

Yinger, J. 1995. Closed Doors, Opportunities Lost: The Continuing Costs of Housing
Discrimination. New York: Russell Sage Foundation. Zhou, M. (1997).
Segmented Assimilation: Issues, controversies, and recent research on the new
second generation. Special Issue: Immigrant adaptation and native-bom responses
in the making of Americans. International Migration Review, 31 (4), 975-1008.

Zhang, T. (2001). Community features and urban sprawl: The case of the Chicago
metropolitan region. Land Use Policy, 18, 221-232.

Zhang, H. & Peterson, MP. (2007). A spatial analysis of neighborhood crime in Omaha,
Nebraska using alternative measures of crime rates. Internet Journal of
Criminology (www.intemetjoumalofcriminologv.com).

Zimmerer, KS. (1996). Ecology as Cornerstone and Chimera in Human Geography. In
C. Earle, K. Mathewson & M. Kenner (Eds), Concepts in Human Geography,
Rowan and Littleﬁeld Publishers: Lanham, 161-188.

Zubrinsky, C. L., & Bobo, L. (1996). Prismatic metropolis: Race and residential
segregation in the City of the Angels. Social Science Research, 25, 335-374.

222