The Spatial Transformation of First-Tier Suburbs

The Spatial Transformation of
First-Tier Suburbs, 1970 to 2000:
The Case of Metropolitan Baltimore
Thomas J. Vicino
Wheaton College
Abstract
The evolution of first-tier suburbs has emerged as an important topic of
scholarly and popular attention in the past decade, yet little is known about
the diversity of neighborhood spatial structure. This article analyzes data on
152 census tracts in 21 first-tier suburban census designated places in metropolitan Baltimore. A total of 49 socioeconomic variables are used to measure
the population, income dynamics, nature of the housing, and structure of the
labor force.
The analysis provides evidence of spatial restructuring in 1970 and 2000.
The racial composition, socioeconomic status, occupation, and nature of
the housing stock differentiate the spatial structure of Baltimore’s first-tier
suburban neighborhoods from one another over time. A typology of five
neighborhoods in 1970 and six in 2000 is derived from a partitional clustering procedure that groups principal components analysis scores. The policy
implications of suburban diversity and decline are discussed.
Keywords: Demographics; Populations; Suburbs
Introduction
Large-scale urban decentralization has transformed the metropolitan
landscape, particularly during the past 50 years (Beauregard 2006). The rim
of urban development is now far from the traditional core. Nearly two out of
every three residents in the metropolitan United States call the suburbs home
(Teaford 2006). There is no doubt that the suburbs consistently attract more
residents and so have become the favored site for development. It is true that
as the cost of transportation climbs, the redevelopment of central cities continues to be an attractive alternative for many residents. Nevertheless, jobs,
investment, and economic growth have become increasingly suburbanized.
HOUSING POLICY DEBATE
VOLUME 19
ISSUE 3
© 2008 METROPOLITAN INSTITUTE AT VIRGINIA TECH. ALL RIGHTS RESERVED.
479
480
Thomas J. Vicino
Metropolitan areas have decentralized from the urban core. In short, the
United States is a metropolitan society dominated by its multifarious suburbs
(Katz and Lang 2005).
The traditional model of this metropolitan landscape posits a declining
central city and growing suburbs (Park, Burgess, and McKenzie 1925). Many
studies have been developed on this simple central city–suburban dichotomy
(Masotti and Hadden 1973). Persistent suburbanization, however, has created a wide range of areas to be subsumed under the category of “suburban,” and they have grown more heterogeneous, rendering the traditional
model obsolete (Orfield 2002). Suburbia increasingly represents a divergent
set of landscapes as suburbs have become sites of immense change (Hanlon,
Vicino, and Short 2006).
Of particular interest are the first‑tier suburbs that flourished during the
1950s and 1960s. Hudnut (2003) defines first-tier suburbs as towns and cities that are located close to central cities and that developed before, during, or
right after World War II. In the 1970s and 1980s, some of these suburbs were
in a state of flux as the early stages of large‑scale social and economic change
such as the deindustrialization of older urbanized areas unfolded (Listokin
and Beaton 1983). Since 1970, various suburban communities have experienced substantial socioeconomic diversification and decline relative to the
outer suburbs, while others did not decline or declined less. In short, experiences varied dramatically (Lucy and Phillips 2006). A full understanding of
the process of first-tier suburban differentiation is, however, lacking.
Typologies provide a way to differentiate not only among but also within
suburbs. As the first-tier suburbs continue to have different social and economic characteristics than the rest of the suburban landscape, it becomes
increasingly important to better understand the diversity of these places and
their characteristics. Moreover, in recent years, the first-tier suburbs have
emerged as a political entity as policy makers and planners have sought to
form regional coalitions to revitalize them (Orfield 2002). Consequently, an
understanding of what differentiates the neighborhoods in the first-tier suburbs is an issue relevant to politics, planning, and public policy alike. Mikelbank fittingly observed that “[t]ypologies serve as a springboard from which
the behavior of complex and diverse phenomena can be more clearly understood. Classification research can help bridge the conceptual gap between
the seemingly unique character of an individual observation and the wellunderstood behavior of groups of similar observations” (2004, 936). Thus,
this article addresses two main research questions:
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
1. Does socioeconomic status persist or change in first‑tier suburbs, and
what are the spatial patterns of socioeconomic status in 1970 and
2000?
2. What are the characteristics of first-tier suburbs in decline, and how were
they different in 1970 and 2000?
This article focuses on Baltimore’s first-tier suburbs. Metropolitan Baltimore serves as a good case study because it is a medium-sized region whose
first-tier suburbs grew for most of the 20th century—until recently. While it
may be difficult to generalize specifically from the case of Baltimore, broad
lessons can be drawn. Puentes and Orfield (2002) note that regions in the
Midwest, which share certain characteristics with Baltimore, have a common set of public problems that are suburban in nature. They argue that
public policy can be harnessed to help these areas. In Baltimore, the first‑tier
suburbs are located close to the central city (within eight miles of the city’s
boundary) and had already experienced substantial development by 1970.
By analyzing the neighborhoods within the first‑tier suburbs, it is possible to
explore the relationship between spatial structure and suburban transformation. This study provides the opportunity to detect changes to and variations
within suburban spatial structure.1
I conduct a principal components analysis (PCA) and a cluster analysis
to examine first‑tier suburban neighborhoods. First, I analyze how patterns
of spatial structure evolved between 1970 and 2000 through a PCA of socioeconomic data in suburban Baltimore. Second, I use a cluster analysis to
classify first‑tier suburban neighborhoods through typologies, paying particular attention to the variety of characteristics of suburban transformation.
I conclude by discussing the policy implications of suburban diversity and
decline.
Theoretical background
As suburbs began to grow during the 1960s, a body of scholarship
emerged to examine the question of whether and how the characteristics
of suburban populations changed over time. This research was collectively
labeled as suburban differentiation because social scientists were interested
in whether the socioeconomic status of suburban populations remained the
same—or persisted—over time. In numerous studies, abundant evidence of
1
No study has yet been done to distinguish among first-tier suburbs over time in the
metropolitan United States.
HOUSING POLICY DEBATE
481
482
Thomas J. Vicino
suburban persistence was presented (Collver and Semyonov 1979; Farley
1964; Guest 1978; Stahura 1979, 1980). The concept of suburban persistence holds that despite population turnover, the socioeconomic characteristics of a suburb persist or remain the same because specific suburbs tend to
attract residents with characteristics similar to those of the current population. According to Farley, “The specialization involved in the origin of a suburb may have implications for its distinctive socioeconomic composition, so
that once a suburb is established, the population that moves into that suburb
tends to resemble the population already living there” (1964, 46).
Another distinct body of work on suburban differentiation also emerged
during this period. Drawing on the work of Park, Burgess, and McKenzie
(1925) from the Chicago School of Sociology, this literature takes the human
ecological perspective, holding that the socioeconomic status of the suburban
population will change over time. Thus, various scholars have attempted to
differentiate suburbs by their function, namely employment and residential
type, to determine the persistence of socioeconomic status (Kramer, Johnson, and Frisbie 1982). Schnore’s classic work (1957, 1959, 1965) differentiates suburbs by residential location and by employment centers, including
manufacturing, commercial, and business activity centers, and found that
suburbs varied by social class. Similarly, Logan (1976) differentiates suburbs
by industrial type and found evidence of socioeconomic stratification. Kish
(1954) differentiates suburbs by social characteristics, including age, race,
and occupation, and found that socioeconomic status changed in different
suburbs. Finally, in another study, Choldin, Hanson, and Bohrer (1980) challenged the suburban persistence model and embraced the human ecological
model. In their analysis of Chicago’s suburban neighborhoods from 1940 to
1970, they found that “suburbs do change in status over time, typically moving gradually downward” (Choldin, Hanson, and Bohrer 1980, 973).
The beginning of the 1990s ushered in a new period of research on the
state of the suburbs by focusing on suburban change, namely diversity and
decline (Orfield 1997). Much has changed since the initial studies on the first
wave of mass suburbanization after World War II. By 1990, suburbs were
the dominant place of residence in the metropolitan United States (Thomas
1998). As more residents populated the suburbs, they grew increasingly
diverse in racial and ethnic terms as well as socioeconomic status (Frey 2003;
Hudnut 2003; Lucy and Phillips 2000). Scholars began to revisit the study
of suburban change by examining patterns of socioeconomic diversity and
decline in first‑tier suburbs to decipher emerging spatial patterns of socioeconomic change (Mikelbank 2004; Orfield 1997).
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
The emergence of this body of research on socioeconomic diversity and
decline in first-tier suburbs presents a new opportunity to revisit an enduring
scholarly inquiry into the study of the suburbs. For decades, social scientists
have studied whether and how suburbs change over time.
Choldin, Hanson, and Bohrer aptly observe that “the question of suburban stability versus change and decline has ramifications in urban sociological
theory as well as public policy” (1980, 973). Indeed, if suburban populations
grow and decline, whether the characteristics of their residents remain stable,
change, or decline is still an enduring question. Our understanding of the
evolution of first-tier suburbs is still quite limited because most studies have
classified both first-tier suburbs and outer suburbs together (Anacker and
Morrow-Jones 2008b). Further, existing studies share the following characteristics: coarse, aggregate geographic scales; broad national samples; and a
focus on the city-suburban dichotomy. As a result, a full understanding of
the process of first-tier suburban differentiation is lacking. Therefore, this
study bridges a gap in the literature by using a finer-grain spatial scale at
the census-tract level, focusing on the evolution of first-tier suburbs for the
Baltimore primary metropolitan statistical area (PMSA) in 1970 and 2000 to
uncover patterns of first-tier suburban differentiation.
Data and methods
In this section, I discuss the types of data that were collected and the
methods that were used for the analysis. First, I present an overview of the
definition of first-tier suburbs and discuss the selection of variables. Next, I
explain the procedures for the PCAs and cluster analyses.
Data
For the purposes of this study, it was first necessary to use census geographic criteria to develop a definition of first‑tier suburbs because urban
data are scale specific to a given geographic area and, in many cases, the
same data collected at different scales can show very different trends (Brenner
2001; Short et al. 1996). After first‑tier suburbs were defined, I then collected quantitative data at the census-tract level to represent the geography
of a neighborhood (Jargowsky 1997). Previous studies on first-tier suburbs
have also used census tracts as an approximate geography for neighborhoods
(Leigh and Lee 2005).
Figure 1 is a map of the first-tier suburbs in metropolitan Baltimore.
Because there are no suburban municipalities there, it was not possible to use
HOUSING POLICY DEBATE
483
housing policy debate
Linthicum
Arbutus
Catonsville
Ferndale
Brooklyn Park
NORTHERN
Lutherville
Pumphrey
Lansdowne
WESTERN
Woodlawn
Lochearn
Pikesville
Figure 1. Map of Baltimore’s First-Tier Suburbs
0
2.5
5
Glen Burnie
10 Miles
City of Baltimore
Edgemere
Dundalk
Essex
Middle River
Overlea
Rosedale
Towson
EASTERN
Parkville
Hampton
484
Thomas J. Vicino
The Spatial Transformation of Baltimore’s First-Tier Suburbs
political boundaries to define the first‑tier suburbs.2 The fact that there were
no incorporated suburbs in Baltimore County or northern Anne Arundel
County meant that an alternative means of identifying first-tier suburbs was
necessary.
I rely on Hanlon and Vicino’s (2007) definition of first-tier suburbs to
identify them in the Baltimore region.3 They use census designated places as
a baseline geography for developing a definition of first‑tier suburbs. These
place boundaries are delineated to collect data on unincorporated areas with
concentrations of population, housing, and commercial sites, and a degree
of local identity (U.S. Bureau of the Census 1994).4 In short, place boundaries are appropriate for identifying suburbs in a metropolitan area that lacks
local political incorporations.5
Two criteria were used to determine whether a suburb would be classified as first tier or not:
1. The first relates to location. Suburbs were classified by default in the first
tier if they shared a boundary with the city of Baltimore.6
2
Maryland has a long tradition of strong county government, and, accordingly, political
power is concentrated at the county rather than the municipal level. Local government boundaries, such as municipal boundaries, are convenient delineations to identify suburbs. Various
scholars have used municipal boundaries for purposes of identification (Bollens 1988; Orfield
1997).
3
Various other scholars have developed competing definitions and terms for first-tier suburbs. While a common definition and term are lacking, similarities abound. These suburbs are
spatially close to the central city and were developed earlier. For example, Bollens (1988) uses
incorporated municipalities and terms them “suburban rings”; Puentes and Warren (2006) use
counties and term them “first suburbs”; Leigh and Lee (2005) use census tracts and term them
“inner-ring suburbs”; and Lucy and Phillips (2000) use census designated places and term them
“old suburbs.”
4
Census designated places are established in cooperation with state and local officials.
5
However, several important caveats are in order. While place-level geography facilitates
the analysis among suburbs within a metropolitan area, it sacrifices the in-depth analysis of
the local, neighborhood scale. Other studies have used alternative geographic scales, and each
has advantages and disadvantages. For example, Puentes and Warren’s (2006) study of 64
urban areas used counties to identity first suburbs. This was advantageous because it allowed
for a broad, national scope to explore suburbs, yet the coarse geographic scale sacrifices the
neighborhood variation within those areas. In another study, Leigh and Lee (2005) used census
tracts to examine the inner-ring suburbs of Philadelphia, providing an opportunity to analyze
the patterns of variation within the neighborhoods of these suburbs, but this study was also
limited in its comparative analysis.
6
The identification of first-tier suburbs with a central-city spatial criterion has political implications. Suburbs sharing a boundary have formed coalitions with other suburbs, as
in metropolitan Cleveland, to confront suburban decline, and in some other cases, such as
Minneapolis–St. Paul, they have even partnered with the central city (Keating and Bier 2008;
Orfield 1997).
HOUSING POLICY DEBATE
485
486
Thomas J. Vicino
2. If suburbs did not meet the first criterion, the second was used. It relates
to the age of development for suburbs that did not share a boundary
with the central city. Suburbs that shared a boundary with a suburb adjacent to the central city were classified as first‑tier suburbs if more than
half of the housing stock was built before 1970.7
The rationale for using the first criterion is that first‑tier suburbs, by
their very nature, are located near the central city and tend to be the oldest
suburban areas of a metropolitan region (Jackson 1985; Sternlieb and Lake
1975).8 These suburbs historically were built close to the urban core and
along infrastructure routes such as streetcar lines (Warner 1978). Baltimore
is no exception because many of the first-tier suburbs bordering the city
were linked to it by rail and streetcar (Harwood 2003). The rationale for the
second criterion was to use the age of the housing stock as a proxy for the
age of a suburb. Hanlon and Vicino (2007) measured the age of the housing
stock as the median age of the unit. They found that with the exception of
Pikesville, most of the housing stock in every first‑tier Baltimore suburb was
built before 1970.9 On average, 70 percent of the housing stock in all 21
first‑tier suburbs was built before that date. In sum, there were 12 places that
bordered the city of Baltimore and 9 others that were adjacent to a first‑tier
suburb that bordered the city.10
Next, it was necessary to determine which census tracts fell within the
boundaries of each first‑tier suburb. Each place contained numerous census tracts or neighborhoods. Table 1 shows the characteristics of Baltimore’s
first-tier suburban neighborhoods. A total of 152 census tracts were identified in the 21 first‑tier suburbs. Using geographic information systems software, I identified all of the census tracts that fell within the boundaries of
each first-tier suburb in the Baltimore PMSA by overlaying the census place
7
1970 was used as a threshold because half of the region’s housing units had already been
built. The suburban housing stock built before 1970 is typified by boring architecture, poor
construction, and a general need for roof replacement (Anacker and Morrow-Jones 2008b).
8
These criteria imply that the housing stock is aging and that property tax revenue may be
declining; it may also suggest a spillover effect of poor central-city residents to nearby first-tier
suburbs (see Anacker and Morrow-Jones 2008a).
9
Pikesville (MD) is the newest first-tier suburb of Baltimore and has the youngest housing
stock.
10
Ground truth analysis verified that these suburbs satisfied the spatial and temporal
criteria. In regions that have recently annexed surrounding suburbs, this definition may be limited. Robert Yin (2002) describes ground truth as a useful method for verifying data collection
and providing context for the environment under study.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Table 1. Characteristics of Baltimore’s First-Tier Suburban Neighborhoods, 2000
Regional
Location
Census
Designated Place
Number of
Census Tracts
Western
Arbutus
Brooklyn Park
Catonsville
Ferndale
Glen Burnie
Lansdowne
Linthicum
Lochearn
Pumphrey
Woodlawn
5
4
12
3
9
4
2
9
2
8
Baltimore
Anne Arundel
Baltimore
Anne Arundel
Anne Arundel
Baltimore
Anne Arundel
Baltimore
Anne Arundel
Baltimore
Northern
Hampton
Lutherville
Parkville
Pikesville
Towson
1
6
9
6
18
Baltimore
Baltimore
Baltimore
Baltimore
Baltimore
Eastern
Dundalk
Edgemere
Essex
Middle River
Overlea
Rosedale
21
6
12
6
3
6
Baltimore
Baltimore
Baltimore
Baltimore
Baltimore
Baltimore
County
Source: U.S. Bureau of the Census (2000).
boundary layer over the census-tract layer in the software. In some cases,
every census tract was an exact spatial match to the census designated place
boundary. In other cases, several census tracts overlapped the boundaries
of two census places. When this occurred, I assigned the census tract to the
census place where most of the land area in the tract fell.11 This technique
of building census places with tracts maintained the identity of each suburb
while capturing the socioeconomic variation within each of them.
The variables I chose provided a way of assessing the socioeconomic
spatial structure over time. The selection was based on well-established
dimensions of urban differentiation. For instance, Shevky and Bell (1955),
early pioneers in the field of urban studies, identified three primary dimensions of urban space in their famous study of Chicago: economic, family,
11
Some 95 percent of census tracts were matches for the census place boundaries.
HOUSING POLICY DEBATE
487
488
Thomas J. Vicino
and ethnic status. Their analysis, as well as subsequent studies, suggested
that the structure of cities could be understood by analyzing data on these
three dimensions (Berry and Rees 1969; Knox 1994; Murdie 1969). Further,
urban scholars still view these three primary dimensions as important elements of urban differentiation in the United States because the same three
have been established in suburban areas (Davies 1984; Hughes 1993; Knox
1994; Perle 1981; Wyly 1999). The variables that I selected all relate to the
primary dimensions that differentiate suburban areas (Short 2007). As such,
they are appropriate for uncovering the spatial structure of Baltimore’s firsttier suburbs.
I collected an array of quantitative data to analyze neighborhood change.
The primary sources were the U.S. Bureau of the Census (2000) and Geolytics’ (2000) Neighborhood Change Database, a commercially available database developed by the Urban Institute and the Rockefeller Foundation.12
Specifically, I compiled data on 49 variables for 152 census tracts in Baltimore’s first‑tier suburbs in 1970 and 2000.13 Then, I organized these data
into six broad categories that were related thematically to population characteristics, family structure, income characteristics, educational attainment,
housing characteristics, and labor market characteristics.14
First, population variables were related to demographic characteristics.
Table 2 reports the list of variables and their descriptions. I included variables that provided relevant information about the population size and racial
composition of each census tract. Previous work has suggested that population trends, such as size and age, as well as the racial and ethnic composition of a community, are important indicators for measuring decline (Orfield
2002). Scholars often used such demographic data as a metric to determine
the socioeconomic condition of suburbs (Lucy and Phillips 2000). Four main
population variables comprised this set of data—size, age, racial and ethnic
composition, and the foreign-born status of the population. The age vari12
Because census tract boundaries change each decade, it was necessary to use this tool to
control for geographic boundary changes. This tool normalized 1970 census tract boundaries
to the 2000 boundaries, thus facilitating the analysis of changes over time.
13
This study is limited to an analysis of suburban change at two points: 1970 and 2000.
One limitation of the study is that the interim years of 1980 and 1990 are excluded.
14
Multicollinearity is not an issue with PCA because inversion of the PCA matrix in not
required. PCA is both a data reduction technique and a technique to detect patterns in the
structure of the data. Thus, the researcher typically loads as many variables as the theory suggests or would explain the phenomenon under investigation. Then, the procedure reduces the
structure of the variables down to a new set of small “principal” components. Some variables
could be multicollinear, but the PCA procedure detects the relationships and reduces them to
new components, so that the relative importance of the variables can be interpreted via the
eigenvalue. Specifically, other studies used a very similar variable selection (Knox 1991; Pratt
and Hanson 1988; Wyly 1999).
housing policy debate
POVRAT
AVHHIN
AVHHINR
FAVINC
FAVINCR
Income
EDUC11
EDUC12
EDUC15
EDUC16
FMC
NEVMAR
FFH
DIVOR
Family
Education
TRCTPOP
YTHPOP
PERS517
PERS64
PERS65P
SHRWHT
SHRBLK
SHRHSP
SHRFOR
Population
High school dropout (11 or fewer years) (%)
High school graduate (12 years) (%)
Some college (13 to 15 years) (%)
4-year college graduate or higher (at least 16 years) (%)
Poverty population (%)
Average household income
Average household income ratio to all suburbs (%)
Average family income
Average family income ratio to all suburbs (%)
Married families with children households (%)
Single, never married households (%)
Female-headed households (%)
Divorced families households (%)
Population size
Youth population (17 and under) (%)
Population aged 18 to 24 (%)
Population aged 18 to 64 (%)
Population aged 65 and over (%)
White, non-Hispanic population (%)
Black, non-Hispanic population (%)
Hispanic (any) population (%)
Foreign-born population (%)
Variable Label
Variable Description
Variable
Category
Table 2. Selection and Communalities of Variables
0.92
0.71
0.89
0.95
0.65
0.96
0.96
0.96
0.96
0.45
0.78
0.69
0.74
0.58
0.87
0.90
0.85
0.89
0.92
0.92
0.67
0.53
0.88
0.92
0.83
0.97
0.89
0.96
0.96
0.96
0.96
0.87
0.93
0.81
0.79
0.63
0.94
0.80
0.84
0.91
0.91
0.91
0.49
0.77
Communality
Extraction Correlation
1970
2000
The Spatial Transformation of Baltimore’s First-Tier Suburbs
HOUSING POLICY DEBATE
489
BDTOT01
BDTOT2
BDTOT3
BDTOT4
BLTYR99
BLTYR89
BLTYR79
BLTYR69
BLTYR59
BLTYR49
BLTYR39
AGGVAL
RNTOCC
OWNOCC
VACHU
Housing
housing policy debate
NA = not applicable.
Labor force
OCC1
Professional and technical occupations (%)
OCC2
Executive, manager, and administrator occupations (%)
OCC3
Sales occupations (%)
OCC4
Administrative support and clerical occupations (%)
OCC5
Production, craft, and repair worker occupations (%)
OCC6
Operator, assembler, and transportation occupations (%)
OCC7
Nonfarm worker occupations (%)
OCC8
Service worker occupations (%)
OCC9
Farm, fishing, and forestry occupations (%)
OCC31
Manufacturing occupations (%)
OCC92
Public administration occupations (%)
UNEMPRT
Unemployed population (%)
Mean communality extraction correlation
0.94
0.82
0.59
0.80
0.89
0.66
0.54
0.83
0.30
0.77
0.79
0.85
0.81
0.89
0.90
0.73
0.81
0.90
0.87
0.70
0.78
0.68
0.70
0.71
0.66
0.80
0.84
0.70
0.86
0.88
0.64
0.77
0.87
0.77
0.83
0.76
0.80
0.90
0.95
0.95
0.66
Communality
Extraction Correlation
1970
2000
One or no bedrooms in the housing unit (%)
0.68
Two bedrooms in the housing unit (%)
0.61
Three bedrooms in the housing unit (%)
0.77
Four bedrooms in the housing unit (%)
0.78
Housing units built in the 1990s (%) NA
Housing units built in the 1980s (%) NA
Housing units built in the 1970s (%) NA
Housing units built in the 1960s (%)
0.88
Housing units built in the 1950s (%)
0.78
Housing units built in the 1940s (%)
0.78
Housing units built in 1939 or earlier (%)
0.89
Average value of the housing units 0.90
Renter-occupied housing units (%)
0.93
Owner-occupied housing units (%)
0.93
Vacant housing units (%)
0.72
Variable Label
Variable Description
Variable
Category
Table 2. Selection and Communalities of Variables continued
490
Thomas J. Vicino
The Spatial Transformation of Baltimore’s First-Tier Suburbs
ables were divided into four cohorts to capture various patterns, including
the population of children (aged 17 and under), the population of young
adults (aged 18 to 24), the population in the workforce (aged 18 to 64), and
the older population (aged 65 and over). The racial and ethnic variables captured non-Hispanic white, non‑Hispanic black, Hispanic, and Asian races.
The other races category was ignored because the prevalence of these groups
in metropolitan Baltimore was negligible.
Second, the family variables dealt with household composition. This category classified households into one of four mutually exclusive types: married families with children, single, never-married households, female‑headed
families, or divorced families.
Third, the income variables related to the economic characteristics of
households and families. The literature suggests that these data are important for identifying trends in poverty and economic segregation, two principal factors for diagnosing the decline of suburban areas (Swanstrom et
al. 2004). I included a series of variables that related to three main income
characteristics:
1. The number and percentage of residents living in poverty (defined by
federal standards)15
2. The average income for households and families
3. The household and family income ratio (to all suburbs)
I computed an income ratio (above 1 or below 1) for each census tract
versus all suburbs in the region. This provided a measure of economic standing.16 Both households and families were analyzed to determine the impact of
the variation of family structure and economic status (Vicino, Hanlon, and
Short 2007).
Fourth, the educational variables added the number of years that a person spent seeking an education; attainment was expressed as an average percentage per category. I classified the population into one of four mutually
exclusive categories: did not graduate from high school (less than 11 years of
15
The federal poverty guidelines, originally developed in 1963 by Mollie Orshansky of
the Social Security Administration, calculated the poverty threshold by multiplying the government’s economy food plan by a factor of three. This was based on the theory that households
generally spend one-third of their net income on food. In 2008, the U.S. Department of Health
and Human Services Poverty Guideline for a household of four is $21,200 in the contiguous
48 states and the District of Columbia (Federal Register 2008).
16
I computed all income data using 1999 dollars to control for inflation. It was necessary
to compute averages instead of medians because census data containing median income figures
were not available at comparable geographic scales over 30 years.
HOUSING POLICY DEBATE
491
492
Thomas J. Vicino
education), high school graduate (exactly 12 years of education), some college (between 13 and 15 years of education), or college graduate (at least 16
years of education).
Fifth, the housing variables related to structural characteristics. There
were four main housing variables based on size, age, value, and tenure. In
terms of size, I calculated the percentage of units that had a particular number of bedrooms, using groups of one or no bedrooms, two bedrooms, three
bedrooms, or four bedrooms as a proxy for the size of a unit. For age, I calculated the percentage of units that were built during a particular decade from
before the 1940s to the 1990s. For value, I calculated the average value in
constant 1999 dollars for each tract. For tenure, I calculated the percentage of
the housing stock that was renter occupied, owner occupied, or vacant. This
information was important for comparing and contrasting the characteristics
of housing in the first-tier suburbs. Recent research has focused on the role
that smaller, older housing played in suburban decline, so I attempted to capture as much information about the nature and quality of the housing stock
as possible (Hudnut 2003). These data were useful in identifying trends.
Last, labor force variables were related to market characteristics. I collected data on the employment status of residents and the occupation of
employed workers. Specifically, I calculated the percentage of the labor force
employed in 1 of 11 different occupations representing the major sectors of
the economy. I used these particular categories because they were the only
ones that were comparable across 30 years at the same geographic scale. In
addition, I computed the percentage of the population that was unemployed
during each decade. This information provided a portrait of the type of work
that residents of the first‑tier suburbs had, as well as the level of labor force
participation. These data measured the types of jobs that residents had, but
not where the jobs were located.
Methods
First, I used the data set to conduct a PCA to understand the structural
relationships among the many different variables. Specifically, PCA is a statistical technique designed to reduce the number of variables in a data set to
several main factors and detect structural relationships among them. PCA
reduces the number of variables and transforms them into a smaller set of
new variables called principal components. The PCA method detects the
relationship between the socioeconomic data and their spatial location, then
identifies patterns in the spatial structure and reports the results as scores for
each suburban neighborhood.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
PCA has long been used in urban studies, and it has been an important
tool in deciphering the spatial organization of urban places (Berry and Horton 1970; Berry and Kasarda 1977; Janson 1980; Perle 1981; Wyly 1999).
During the 1960s and 1970s, the availability of computing power allowed
urbanists to quantitatively test the prevailing social theories of the Chicago
School of the 1920s for the first time (Berry and Kasarda 1977). As a result,
urban research flourished as scholars conducted numerous PCAs on cities
around the world. By the 1980s, the popularity of PCA faded as urban scholars embraced other methods (Knox 1991).
However, over the past decade, scholars have once again turned to PCA
to gain a firmer understanding of the processes shaping suburbia. Some of
the first empirical studies to analyze the decline of suburbs have used PCA.
For example, Lucy and Phillips (2000) analyzed change over a sample of 24
metropolitan areas. Orfield (2002) examined change for the top 25 most
populated regions. Vicino, Hanlon, and Short (2007) investigated urban
restructuring in the East Coast megalopolis, and Wyly (1999) explored suburban change in the Minneapolis metropolitan area. These studies represent several of the first attempts to quantitatively and systematically redefine
changes in the suburban landscape.
In a similar fashion, I conducted a PCA on Baltimore’s first‑tier suburbs
by using census tracts at two points in time: 1970 and 2000. I constructed
a data matrix containing 49 variables for the 152 census tracts (152 x 49)
in Baltimore’s first‑tier suburbs. The matrix contained a total of 7,448 separate numerical values called attributes. The 1970 PCA contained 152 census
tracts, but the 2000 PCA contained only 151.17
The 1970 and 2000 PCAs each generated six principal components.
Table 3 lists them and provides characteristics that I developed based on the
PCA results.18 It was important to first determine how many components to
extract from the data set. Because one of the goals of PCA is to reduce the
number of variables into a smaller, more manageable set, it was necessary to
determine how many components to use for the smaller set. One way was to
17
I deleted a census tract in Southeastern Baltimore County from the 2000 analysis because
no data were associated with it. This area, which was mixed-use residential with industry in
1970, was located in Sparrows Point. By 2000, it contained only industrial sites, so including
it in the analysis was not appropriate.
18
PCA is limited in terms of sample size, labeling group, sensitivity, and interpretation.
Analysis outputs are subject to the researcher’s appropriate explanation of each component,
given each set of limitations.
HOUSING POLICY DEBATE
493
housing policy debate
Professional class households
Fragile family renter households
Poor black households
New suburban family households
Older white households
Young single households
1
2
3
4
5
6
Note: Only PCA loadings ± 0.30 are reported.
1970
Component
Older housing, no-growth households
Young service worker households
Older white households
Black middle-class households
Minority renter households
Poor households
2000
Definition
Table 3. Definitions of PCAs, 1970 and 2000
2.08
2.45
2.82
3.70
5.83
13.45
1970
2.12
2.24
3.46
4.17
7.84
13.58
2000
Eigenvalue
4.74
5.57
6.40
8.40
13.26
30.56
1970
4.25
4.47
6.92
8.34
15.68
27.16
2000
Percentage of Variance
68.93
64.19
58.62
52.22
43.82
30.56
1970
66.83
62.58
58.11
51.19
42.85
27.16
2000
Cumulative Percentage
Extraction Sums of Squared Loadings
494
Thomas J. Vicino
The Spatial Transformation of Baltimore’s First-Tier Suburbs
analyze the eigenvalues for each component on a graph. For both analyses,
there were six components with an eigenvalue of over 2,19 so I extracted
them.
It was then necessary to determine whether or not the six components
explained a large portion of the original data set. Table 3 shows that both
PCAs had substantial explanatory power. The six components extracted for
the 1970 analysis explained over two‑thirds (68.93 percent) of the variation from the original data set. For the 2000 analysis, they explained 66.83
percent. In 1970, component 1 explained about one‑third of the total variation, and component 2 explained 13 percent. The last four components
each explained less than 10 percent of the total variation. The 2000 analysis
results were similar. Thus, the data for 49 variables on 152 cases (census
tracts) could be largely explained by only six components.
Next, it was important to determine whether the variables selected for
this analysis were relevant to the study of suburban transformation. The
PCAs each produced a communality value, ranging from 0 to 1, to ascertain
the relevance of the variables.20 For the 1970 PCA, table 2 shows that the
communality values ranged from 0.45 to 0.96, while for the 2000 PCA, the
range was 0.30 to 0.97. The communality values could be interpreted as
percentages, which were measures of correlation.21 Together, the 49 variables were, on average, 80 percent and 81 percent correlated with the six
extracted components in 1970 and 2000, respectively. For the 1970 PCA,
over two‑thirds of the variables (36 of 49) had values of 0.70 or more. For
the 2000 PCA, more than three-quarters of the variables (41 of 49) had
values of 0.70 or more. This value is a common threshold used in PCA to
demonstrate a strong correlation among variables (Wyly 1999).
The loadings—measures of the degree to which each variable in the set
contributed to the meaning of each new component that the PCA produced—
were then analyzed to interpret the meaning (the characteristics) of each
extracted component. Table 4 displays the loading values that are greater
19
In general, eigenvalues of 2 or higher could be interpreted as having at least twice the
explanatory power of the original set of variables, and so they were a meaningful representation of the larger data set (Kline 1994).
20
Communality is a numerical estimation of the variance and strength of each variable
that is reduced into a set of fewer components. Each communality value indicated how well
the variable correlated with the six extracted components.
21
The communalities measure the percentage of variance in a given variable explained by
all of the components together, and it can be interpreted as the reliability of the new principal
component (Kline 1994).
HOUSING POLICY DEBATE
495
1970
Variable
1970
2000
2. Fragile Family to
Minority Renter
1970
2000
3. Poor Black to
Black Middle Class
1970
2000
4. New Suburban to
Older White
1970
2000
5. Older White to
Young Service
1970
2000
6. Young Single to
Older Housing
TRCTPOP 0.37 –0.31
PERS517 –0.69 0.33
0.45
–0.4
–0.37
0.34 YTHPOP 0.57
0.65 0.61 PERS64 0.39
0.6
0.57 0.53 PERS65P –0.33 –0.58
0.33
0.39 SHRWHT -0.48
–0.62
–0.31 0.63
0.63 SHRBLK 0.4
0.62
0.31 –0.64
–0.63 SHRHSP 0.41 0.33 SHRFOR
0.52 0.55 FMC –0.7 –0.31 –0.32 NEVMAR 0.49
0.33
0.74 0.54 FFH 0.63
0.36 0.49 DIVOR 0.53
0.56 0.41 POVRAT 0.39 0.61
0.39
–0.47 AVHHIN
0.91
–0.89 AVHHINR
0.91
–0.89 FAVINC
0.92
–0.9 FAVINCR
0.92
–0.9 EDUC11 0.78 EDUC12 0.76 EDUC15
0.8
0.4
0.34 EDUC16
0.92
–0.87 BDTOT01 0.67
0.47 0.42 0.5 BDTOT2
–0.52
0.52 0.44 0.41 BDTOT3 –0.64 –0.5 BDTOT4
0.68
–0.52 –0.61 2000
1. Professional Class
to Poor
Table 4. Rotated PCA Loadings for Baltimore’s First-Tier Suburban Neighborhoods, 1970 and 2000
496
Thomas J. Vicino
housing policy debate
1970
Variable
1970
2000
2. Fragile Family to
Minority Renter
1970
2000
3. Poor Black to
Black Middle Class
1970
2000
4. New Suburban to
Older White
1970
2000
5. Older White to
Young Service
1970
2000
6. Young Single to
Older Housing
Note: The method is based on PCA rotation using Varimax with Kaiser normalization in 16 iterations. Only PCA loadings ± 0.30 are reported.
BLTYR99 –0.54
BLTYR89 –0.6
BLTYR79 0.76 BLTYR69 0.4
0.53 BLTYR59 BLTYR49 0.33 0.44 0.45
BLTYR39 0.61 0.37 0.42
AGGVAL
0.9
–0.83 RNTOCC 0.45
0.82
0.54 0.53
0.31 OWNOCC –0.45
–0.82
–0.54 –0.53
–0.31 VACHU 0.4
0.41 0.39
0.31 0.33
OCC1
0.86
–0.85 OCC2
0.91
–0.86 OCC3
0.75 0.31 OCC4 0.66
0.32 OCC5
–0.81
0.67 OCC6
–0.82
0.65 OCC7
–0.6
0.55 OCC8 0.6 0.42 OCC9 0.46 OCC31 0.43 0.44 OCC92 0.34
0.36 UNEMPRT 0.36 0.36 0.34 0.4 2000
1. Professional Class
to Poor
Table 4. Rotated PCA Loadings for Baltimore’s First-Tier Suburban Neighborhoods, 1970 and 2000 continued
The Spatial Transformation of Baltimore’s First-Tier Suburbs
HOUSING POLICY DEBATE
497
498
Thomas J. Vicino
than 0.30 and less than –0.30. Loadings above and below these values did
not significantly contribute to the meaning of the component and are not
shown.
Last, the PCA scores were used to display the spatial patterns in the data
and create an index of positive and negative values that measured the degree
to which each census tract related to the components. The PCA scores were
thus measures of the strength of the representation of the meaning for each
of the six components. A census tract with a high score meant that the component was the defining characteristic of that area; lower scores meant that
the component was not a defining characteristic.
Using the PCA scores, I conducted a partitional cluster analysis to create
a typology of suburban neighborhoods. Recently, numerous urban scholars have used similar grouping techniques to create classifications of suburban places. The approach has been used to identify “suburban immigrant
enclaves” (Logan and Zhang 2004); “at‑risk suburbs” (Orfield 2002); “edgeless cities” (Lang 2003); “healthy suburbs” (Mikelbank 2004); and “middle
America suburbs” (Vicino, Hanlon, and Short 2007), to name just a few. In
this study, I built on the tradition of clustering to create distinct categories
of suburban neighborhoods that depict varying characteristics to determine
the persistence or change in socioeconomic status and the spatial structure of
Baltimore’s first-tier suburbs.
Cluster analysis is a statistical technique that categorizes data into a series
of smaller groups that share similar characteristics. The members of each
group are grouped so that they are strongly associated with each other and
weakly associated with other clusters. I clustered the component scores generated for each census tract from the PCA. Following the analyses of Orfield
(2002) and Short (2007), I grouped the scores by using a k-means clustering technique, a common method22 that was ideal for this study because it
is helpful for large numbers of observations.23 It is necessary to first specify
the number of clusters desired, and scholars have typically defined between
four and seven. Drawing on previous studies, I specified alternative sizes
between three and six clusters for the 1970 and 2000 analyses (Hartigan
1975; Orfield 2002). I compared each specification for ease of interpretation
and selected an appropriate size that facilitated interpretation: five clusters
for the 1970 analysis and six for the 2000 analysis.
22
Other studies have used alternative approaches to cluster analysis. For example, Mikelbank (2004) and Hill, Brennan, and Wolman (1998) use hierarchical clustering with an
agglomeration schedule to have the procedure determine the ideal number of clusters.
23
The 1970 PCA produced 7,448 attributes and the 2000 PCA produced 7,399.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Suburban spatial structure in 1970 and 2000
Baltimore’s first-tier suburban neighborhoods represented a diversity of
places in 1970 and 2000. The analyses showed a significant suburban spatial restructuring. Specifically, on the basis of the results of the PCAs for
1970 and 2000, I have interpreted three primary thematic transformations
of suburban spatial structure: class, race, and age. I will review each theme
in turn.
First, class was a distinguishing feature of suburban spatial structure.
The 1970 PCA detected a strong presence of professional class households. It
was the largest measure of suburban structure in 1970 and explained nearly
a third of the variation in the original data set (table 3). Table 4 shows that
the loadings were the highest for this component. Above all other variables,
they indicated that high socioeconomic status defined household composition. High levels of household and family income, educational attainment,
and professional occupations were the dominant features. By contrast, the
analysis showed that by 2000, poor households emerged. It was also the
strongest measure of suburban structure and explained 27 percent of the
variation in the original data set (table 3). Collectively, the loadings described
the classic characteristics of the underclass (Jargowsky 1997). For instance, in
table 4, female‑headed households and divorced families had high loadings,
and married‑parents families had very low ones (–0.7). This suggests that the
traditional married-parents family was not common in these neighborhoods
(Wilson 1987). Poverty had high loadings, and the income loadings showed
that the population was composed of poor residents. This example captured
the essence of the class characteristics.
Figure 2 shows that the spatial trends for the class transformation from
professional to poor households were especially apparent. In 1970, professional class households were located to the north and west of the city, and
poor households were located to the east and south. By 2000, poor households were still located to the east and south, while nonpoor households were
located to the north. The western first-tier suburban neighborhoods changed
from professional households in 1970 to poor households in 2000. Thus,
the transformation of class in Baltimore’s first-tier suburbs offers evidence
of suburban persistence. There was a clear demarcation between neighborhoods with high and low socioeconomic status in 1970 and 2000. In short,
the spatial segregation of households by class was the dominant feature in
1970 and 2000.
Second, race was another distinguishing feature of suburban spatial
structure. The 1970 PCA captured areas of fragile family renter households.
As table 3 shows, this component accounts for 13 percent of the variation
HOUSING POLICY DEBATE
499
500
Thomas J. Vicino
Figure 2. PCA Scores for Baltimore’s First-Tier Suburban Neighborhoods
Professional
Poor
1970
2000
Fragile Family
Minority Renter
1970
2000
Poor Black
Black Middle Class
1970
2000
PCA Scores
–0.01 to –0.49
–0.50 to –5.08
0 to 0.49
0.50 to 2.68
No Data
housing policy debate
0
3.5
7
14 Miles
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Figure 2. PCA Scores for Baltimore’s First-Tier Suburban Neighborhoods continued
New Suburban Family
Older White
1970
2000
Older White
Young Service Worker
1970
2000
Young Single
Older Housing
1970
2000
PCA Scores
–0.01 to –0.49
–0.50 to –5.08
0 to 0.49
0.50 to 2.68
No Data
0
3.5
7
14 Miles
HOUSING POLICY DEBATE
501
502
Thomas J. Vicino
in the original data set and is strongly related to housing tenure and family composition. Renter-occupied housing units with one bedroom were the
dominant characteristic (table 4). Plus, households were primarily composed
of divorced families, female‑headed families, and singles. By contrast, by
2000, the analysis demonstrated that fragile family renter households were
transformed into minority renter households, accounting for 16 percent of
the variation in the original data set (table 3). Table 4 shows that loadings for
the white population were very low, while they were high for all other races
and ethnicities. For example, black, Hispanic, other races, and foreign-born
populations all loaded at over 0.4, suggesting that there was a strong presence of minority households. In terms of housing, the loadings indicated that
the population rented one‑bedroom units that were primarily built during
the 1970s. The poverty loading value was 0.61, making it the highest value
on the poverty variable out of all six principal components in 2000.
As figure 2 shows, there were small clusters of fragile family renter households throughout first-tier suburban Baltimore in 1970. The largest group
was located directly north of the city in suburban Towson (MD). Small clusters of these neighborhoods were also evident along the western suburban
fringe. In 2000, minority renter households were clustered in the eastern and
northern first-tier suburbs.
The 1970 PCA also revealed the emergence of poor black households.
This component represented over 8 percent of the variation in the original
data set (table 3). Table 4 shows that race was the highest loading value. It
was 0.62 for blacks and –0.62 for whites. The presence of a black population was one of the defining characteristics. Poor black households were also
the only component in the 1970 PCA where the poverty loading value was
significant, at 0.39. By 2000, poor black households had changed into black
middle-class households in first-tier suburban Baltimore. This component
accounted for over 8 percent of the variation in the original data set (table 3).
Table 4 shows that the main variables with high loadings related to population age and size, race, poverty, and occupation. These neighborhoods had
a considerable black population relative to other suburban neighborhoods,
and public sector employment for the residents in these neighborhoods was
the distinguishing feature. This demonstrated the growth of a black middle
class in first‑tier suburban neighborhoods.
Figure 2 shows the spatial distribution of poor black households among
most first‑tier suburbs in 1970. These neighborhoods were concentrated in
the eastern and southwestern neighborhoods of Dundalk, Essex, Middle
River, and Lansdowne. By contrast, the western suburban fringe contained
relatively few, if any, poor black households. There was little evidence that
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
such households were located in the western suburban neighborhoods such
as Woodlawn and Lochearn in 1970. By 2000, a noteworthy transformation had occurred. The black middle-class neighborhoods on the western
and northwestern fringe of suburban Baltimore had grown dramatically. A
contiguous area of 45 neighborhoods was a prominent feature of the landscape. These neighborhoods were exclusively located to the west of the city
in Woodlawn, Lochearn, Pikesville, and Towson.
Third, age was the last distinguishing feature of suburban spatial structure. The 1970 PCA provided evidence of new suburban family households.
The fourth component accounted for over 6 percent of the variation in the
original data set (table 3). Table 4 showed that new suburban family households loaded high on variables relating to the age of the population, housing
age, and housing tenure. Collectively, the loadings indicated that new housing units were built in the 1960s, and young families with children resided in
these neighborhoods. By 2000, these first-tier suburbs had been transformed
into older white households. The 2000 PCA represented 7 percent of the
variation in the original data set (table 3). Table 4 shows that the loadings for
the age and race of the population and housing size variables were significant.
They suggest that the residents of these neighborhoods were mainly elderly
and white, implying that the population was aging without any younger successors. The fourth component provided evidence of patterns of aging among
older white residents in the first-tier suburban neighborhoods.
In spatial terms, figure 2 shows that in 1970, the largest concentration
of new suburban family households was located due north of the city in
Pikesville and east of the city in Middle River and Essex. By 2000, a marked
spatial transformation had occurred in these locations. The older white population was the defining characteristic of the eastern suburbs of Baltimore,
especially Dundalk, Essex, and Middle River. Thus, in 1970, these suburbs
housed new suburban family households, and by 2000 they primarily housed
older white households.
The 1970 PCA also detected older white households and young single
households. These components together accounted for 10 percent of the variation in the original data set (table 3). Table 4 shows that the age and race of
the population and the age of the housing were the only characteristics that
had significant loadings on the fifth component. The loadings for older white
households suggest that an elderly white population resided in older housing units in these neighborhoods in 1970. For young single households, the
age of the population and the family status were the only two variables with
significant loadings, suggesting that there was an emerging trend of young
single households developing in the first-tier suburbs in 1970.
HOUSING POLICY DEBATE
503
504
Thomas J. Vicino
By 2000, the analysis detected that older white households had transformed into younger service-worker households, and young single households
had transformed into older housing. These components explained almost 9
percent of the variation in the original data set (table 3). Table 4 shows that
the loadings for younger service-worker household provide evidence of a
young population, the lack of married parents with children, the presence of
service workers, and unemployment. Together, these variables showed that a
portion of the younger population was employed in the service sector of the
economy. The 2000 PCA identified older housing households, and the loadings collectively suggested that neighborhoods lacked population growth and
had a very old housing stock that dated to before 1950.
In summary, Baltimore’s first-tier suburban neighborhoods were diverse
in both 1970 and 2000. The spatial structure persisted from 1970 to 2000
along the class dimension, and it changed along the dimensions of race and
age. While socioeconomic diversity in the suburbs in 2000 is not surprising, the first-tier suburbs were remarkably diverse in 1970. They housed a
significantly diverse population that included black middle-class residents,
white working-class residents, professionals, and a population that was both
young and old. Overall, these dimensions stratified the spatial structure of
neighborhoods in both time periods, and the PCAs provided evidence of the
transformation of each dimension.
Typology of first-tier suburban neighborhoods
The PCAs provided important baseline information about the patterns
of suburban spatial structure in the first‑tier suburbs in 1970 and 2000 and
established the basis for differentiating among neighborhoods. A cluster
analysis was conducted to better understand the variation of neighborhood
types within the first-tier suburbs. Table 5 summarizes the distribution of the
cases for the cluster analysis in 1970 and 2000. This technique analyzed the
relationships among the PCA scores for each neighborhood and then classified them into smaller, more manageable groups based on a set of common
characteristics, thus facilitating the interpretation of the principal component
scores and allowing for the differentiation of first-tier suburban neighborhoods (Orfield 2002).
First-tier suburban neighborhoods in 1970
The analysis selected for 1970 yielded five clusters. There were 912
individual PCA scores from six principal components for 152 census tracts.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Table 5. Distribution of Cluster Cases in Baltimore’s First-Tier Suburban
Neighborhoods, 1970 and 2000
Definition
1970
Minority presence
Older middle class Blue collar Number of Total Cases
5
17
39
Newer middle class
48
32
181,719
Wealthy 10
7
30,437
Total
152
100
545,296
2000
33
6
100
24
26
48
3
13
37
151
6
18
37
Percentage of
Total Population
27,478
93,568
212,094
Poor Wealthy Blue collar University Black middle class Middle America
Total
11
27
56
Percentage of
Total Cases
Population
16
17
32
2
9
24
100
113,096
76,992
114,373
6,852
45,301
130,990
487,604
22
16
23
2
10
27
100
Table 6 provides a summary of these distinguishing characteristics. In this
section, I will examine the characteristics of the neighborhood clusters in
1970 by comparing four variables: population, economic status, housing,
and labor force.
First, in 1970, the population of all of the first-tier suburban neighborhoods was just over half a million residents. The largest cluster of the
five was blue-collar neighborhoods, which contained 212,094 residents, or
39 percent of the first-tier suburban population. Next, newer middle-class
neighborhoods contained 181,719 residents, or one-third of the population,
followed by older middle-class neighborhoods housing 93,568 residents, or
17 percent of the population. The two smallest clusters, minority presence
and wealthy neighborhoods, each housed approximately 5 percent of the
population. In racial terms, Baltimore’s first-tier suburban neighborhoods
were overwhelmingly white, with the exception of minority presence neighborhoods, which were 25 percent black. Yet this cluster contained only 11
neighborhoods with 27,478 residents in 1970.
HOUSING POLICY DEBATE
505
Population
Economic Status
Housing Labor Force
housing policy debate
Wealthy
n 99% white
n $124,000 in average
n 50% built in the 1960s
n 55% professional
neighborhoods (10)
n 62% aged 18 to 64 household income
n 79% homeownership occupations
n 3,042 average
n 113% more income
n 40% college education
population size than all suburbs
2% poverty
Blue-collar n 98% white
n $50,000 in average
n 40% built in the 1950s
n 48% manufacturing
neighborhoods (56)
n 30% under age 17 household income
n 77% homeownership occupations
n 3,787 average
n Same income level
n 43% high school dropouts
population size as all suburbs
n 5% poverty
Newer middle-class n 97% white
n $58,000 in average
n 87% built in the n 26% manufacturing
neighborhoods (48)
n 65% aged 18 to 64 household income
n 1950s and 1960 occupations
n 3,785 average
n 5% more income
n 65% homeownership
n 44% high school
population size than all suburbs education
n 3% poverty
Minority presence
n 25% black
n $64,000 in average
n 50% built in n 50% service occupations
the neighborhoods (11)
n 75% white household income 1950s and 1960s
n 40% high school education
n 17% aged 18 to 24
n 10% more income than
n 80% homeownership
n 2,498 average all suburbs
population size
n 6% poverty
Older middle-class
n 98% white
n $61,000 in average
n 32% built before 1939
n 20% manufacturing
neighborhoods (27)
n 16% over age 65
household income
n 68% homeownership occupations
n 3,466 average
n 5% more income
n 40% high school
population size than all suburbs education
n 5% poverty
Cluster (N = 152)
Table 6. Typology and Characteristics of Baltimore’s First-Tier Suburban Neighborhoods, 1970
506
Thomas J. Vicino
The Spatial Transformation of Baltimore’s First-Tier Suburbs
In terms of age, the first-tier suburbs primarily housed a younger population, with the exception of older middle-class neighborhoods. Their population had 16 percent of the residents over the age of 65. In 1970, the
population of first-tier suburban Baltimore showed that neighborhoods were
growing with new families in white suburbia, with pockets of older suburbanites, black suburban pioneers, and suburban industrial workers.
Second, the economic status of first-tier suburban neighborhoods in Baltimore was higher than in all other areas in the region in 1970. All five
clusters of first-tier suburban neighborhoods stood out as economically
advantaged compared with the Baltimore PMSA and all other suburbs. For
example, the median household income in the Baltimore PMSA in 1970 was
$38,302, versus $45,546 in all of the suburbs of the Baltimore PMSA.24 By
comparison, in the poorest cluster—blue-collar neighborhoods—the average
household income was $50,000 in 1970. Wealthy neighborhoods, however,
had an average household income of $124,000—113 percent more income
than all other suburbs. The three other neighborhood clusters each had an
average household income of approximately $60,000. Poverty remained at 6
percent or less in all clusters. Three decades ago, the cluster analysis showed
that Baltimore’s first-tier suburban neighborhoods housed the region’s highest-income households relative to other areas.
Third, the housing stock in first-tier suburban Baltimore varied in age
and by tenure status among the neighborhood clusters. In three of them, the
housing stock was relatively new in 1970. In minority presence, newer middle class, and wealthy neighborhoods, at least half of the housing units were
built during the 1950s and 1960s. By contrast, in older middle-class neighborhoods, almost a third of the housing units were built before 1939—the oldest
housing stock in the first-tier suburbs in 1970. In terms of housing tenure,
four out of the five clusters had homeownership rates between 65 percent
and 80 percent. Yet in newer middle-class neighborhoods, approximately
two-thirds of the residents were homeowners, and the rest were renters.
Last, the characteristics of the labor force in first-tier suburban neighborhoods varied substantially in 1970. Wealthy neighborhoods stood out as the
outlier to all other clusters. Over half of all residents worked at professional
occupations, and 40 percent had at least a four-year college degree. These
neighborhoods were clustered to the north of the city. Blue collar neighborhoods exemplified the other end of the labor force spectrum. Approximately
24
This is inflated for 1999 dollars. “All suburbs” is measured by HUD (2000) and is geographically equivalent to the PMSA minus the central city.
HOUSING POLICY DEBATE
507
508
Thomas J. Vicino
half of all residents there held manufacturing jobs, and 43 percent were high
school dropouts. These neighborhoods were clustered to the southeast and
southwest of the city. Thus, the distinctive feature of the labor force in 1970
was the spatial division between the educational and occupational characteristics in first-tier suburban neighborhoods.
In summary, characteristics involving race, economic status, housing,
and the occupations of residents clustered neighborhoods in 1970. Figure 3
is a map of the typology of first‑tier suburban neighborhoods in 1970 and
2000. Race was a prominent spatial feature of first‑tier suburban neighborhoods. With the exception of minority presence neighborhoods, all of the
clusters were white enclaves. Minority residents were the suburban pioneers
of the 1970s. Also, the variation in household income was a distinguishing
feature. Most of the first‑tier suburban neighborhoods had higher income
levels than all other suburbs in the region. Not a single neighborhood in the
first‑tier suburbs was below the suburban average household income level of
$47,721 in 1970. Further, the age of the units in the first‑tier suburban neighborhoods was diverse. Many neighborhoods featured the youngest housing
stock and were the newest locations to feature the latest suburban subdivisions in the region in 1970, while other neighborhoods already had older
stock. The 1970 cluster analysis demonstrated that these factors differentiated the first‑tier suburban neighborhoods.
First-tier suburban neighborhoods in 2000
The cluster analysis for 2000 yielded six clusters. There were 906 individual PCA scores from six principal components for 151 neighborhoods
(see footnote 17). Table 7 provides a summary of the distinguishing characteristics of the clusters. In this section, I will discuss the characteristics of
the neighborhood clusters in 2000 by comparing four variables: population,
economic status, housing, and labor force.
First, the population of the first-tier suburban neighborhoods in 2000
had declined from 1970 to 487,604 residents in the six neighborhood clusters. The largest cluster was middle-America neighborhoods, which housed
130,990 residents, or slightly more than one-quarter of first-tier suburban
Baltimore. Next, both poor and blue-collar neighborhoods each housed 22
percent of the residents, or approximately 114,000 residents each. Wealthy
neighborhoods grew from 1970 to represent 16 percent of the population, or
76,992 residents in the first-tier suburbs. Black middle-class neighborhoods
housed 45,301 residents, or 9 percent. University neighborhoods housed
6,852 residents, or just over 1 percent of the population.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Figure 3. Typology of Baltimore’s First-Tier Suburban Neighborhoods, 1970 and
2000
1970
Typology, 1970
Minority Presence
Older Middle Class
Blue Collar
Newer Middle Class 0
Wealthy
3.5
7
14 Miles
3.5
7
14 Miles
2000
Typology, 2000
Poor
Wealthy
Blue Collar
University
Black Middle Class
Middle America
No Data
0
HOUSING POLICY DEBATE
509
Population
Economic Status
Housing Labor Force
housing policy debate
Middle America
n 86% white
n $59,000 in average
n 64% built before 1970
n 22% manufacturing
neighborhoods (37)
n 11% black household income
n 83% homeownership occupations
n 19% female headed
n 17% less income
n 50% at least some
n 3,540 average than all suburbs college education
population size
n 5% poverty
Black middle-class
n 71% black
n $55,000 in average
n 70% built between
n 40% professional
neighborhoods (13)
n 38% female headed household income the 1950s and the 1960s occupations
n 3,484 average
n 22% less income
n 75% homeownership
n 53% at least some
population size than all suburbs college education
n 6% poverty
University
n 67% white
n $13,000 in average
n 76% built during
n 60% service
neighborhoods (3)
n 25% black household income the 1970s occupations
n 2,284 average
n 76% less income
n 33% homeownership
n 60% at least some
population size than all suburbs college education
n 61% poverty
Wealthy
n 91% white
n $96,000 in average
n 38% built after 1970
n 70% professional
neighborhoods (26)
n 9% foreign born household income
n 77% homeownership occupations
n 13% female headed
n 36% more income
n 60% college
n 2,961 average than all suburbs education
population size
n 4% poverty
Blue-collar n 93% white
n $48,000 in average
n 85% built before 1970
n 30% manufacturing
neighborhoods (48)
n 21% over age 65 household income
n 75% homeownership occupations
n 26% female headed
n 32% less income
n 69% high school
n 3,007 average than all suburbs education or less
population size
n 8% poverty
Poor neighborhoods (24)
n 33% black
n $45,000 in average
n 66% built between
n 60% service
60% white household income the 1950s and the 1970s occupations
n 55% female headed
n 35% less income
n 38% homeownership
n 51% high school
n 4,712 average than all suburbs education or less
population size
n 12% poverty
Cluster (N = 151)
Table 7. Typology and Characteristics of Baltimore’s First-Tier Suburban Neighborhoods, 2000
510
Thomas J. Vicino
The Spatial Transformation of Baltimore’s First-Tier Suburbs
In terms of age, blue-collar neighborhoods stood out as housing onefifth of the residents over 65. As to race, numerous first-tier neighborhoods
grew more diverse. In wealthy and blue-collar neighborhoods, over 90 percent of the population remained white. By contrast, in poor, university, and
black middle-class neighborhoods, the number of blacks increased to nearly
one-third of the population. In 2000, the first-tier suburban population had
grown more diverse, but the first-tier suburban black population continued
to reside primarily in the western neighborhoods.
Second, the economic status of first-tier suburban Baltimore neighborhoods declined in a marked fashion relative to the status of all other areas in
the region in 1970. Four out of six clusters of the first-tier suburban neighborhoods stood out as economically disadvantaged compared with the all of
the other suburbs of the Baltimore PMSA in 2000; only wealthy and middle-America neighborhoods had higher average household incomes that the
suburban average household income of $56,111. Even the wealthy neighborhood cluster lost household income relative to its status in 1970. In addition,
poverty increased throughout the clusters. Poverty stood at 12 percent in
poor neighborhoods and 8 percent in blue-collar neighborhoods. In short,
the economic status of the first-tier suburban neighborhoods declined dramatically from 1970 to 2000.
Third, by 2000, the housing stock had grown older, and tenure was
more diverse in all neighborhood clusters in the first-tier Baltimore suburbs,
compared with all other suburbs in the region. While the first-tier suburban
neighborhoods boasted the qualities and amenities of a newer housing stock
in 1970, the signs of an aging stock were apparent. In all clusters, most of
the units were constructed before 1970, and in many cases, three-quarters
of them were built during the 1950s and 1960s alone. Moreover, housing
tenure remained high in four of the clusters; well over 70 percent of residents
owned a house in wealthy, blue-collar, black middle-class, and middle-America neighborhoods. By contrast, in poor and university neighborhoods, only
one-third of the residents owned a house.
Last, the characteristics of the labor force changed between 1970 and
2000. The decline in manufacturing jobs affected all neighborhoods, especially blue collar and middle America. The disparities between the wealthy
neighborhoods and all other clusters grew ever larger. Some 70 percent of the
residents held professional jobs and 60 percent had at least a four-year college degree. In the other neighborhoods, the rise of service occupations was
evident, and it reached 60 percent in poor neighborhoods.
In summary, the cluster analysis for 2000 demonstrated that there was
substantial variation in the socioeconomic composition of the first‑tier sub-
HOUSING POLICY DEBATE
511
512
Thomas J. Vicino
urban neighborhoods. Overall, several prevailing characteristics defined the
clusters in 2000. Black and white residential populations were segregated in
first‑tier suburban neighborhoods. Without a doubt, there was an increase
in the racial diversity of the first‑tier population. Also, there was considerable income variation in the typology of the six clusters for 2000. Average
household income ranged from $13,000 to $96,000 in 2000, a difference of
$83,000 between the wealthiest and poorest neighborhoods. In addition, the
housing stock was the oldest in the suburbs in 2000. For every cluster, most
of the units were built between the 1950s and the 1970s, but high levels of
homeownership characterized the housing trend. Last, the labor force among
all neighborhoods in the first-tier suburbs was affected by a decline in manufacturing employment and a rise in service sector employment.
Policy implications and conclusions
Today’s suburban landscape is an amalgam of many different types of
neighborhoods in suburbs mixed together in the metropolis. First-tier suburban neighborhoods are diverse—representing a full range of residents,
housing types, socioeconomic status, and jobs. This study demonstrated that
there has been a significant spatial transformation of Baltimore’s first‑tier
suburbs since 1970. The PCAs and cluster analyses illustrated the spatial
transformation of the race of the population, its economic status, and its age,
as well as the labor force. This transformation suggests that suburban decline
had emerged as a strong force throughout Baltimore’s first-tier suburban
neighborhoods by 2000.
Many observed the spatial transformation of those first-tier suburbs
during the 1990s, including key actors in state and local government and
the media (Outen 2005). Baltimore County, which maintains political jurisdiction over most of the region’s first-tier suburbs, developed a strategy to
address the decline of its maturing suburban communities during the mid1990s.25 In 1995, County Executive “Dutch” Ruppersberger created the
countywide Office of Community Conservation, charged with stabilizing
and revitalizing Baltimore County communities. The official mission was to,
“preserve, stabilize, and enhance the human, physical, and economic conditions of the County’s urban communities” (Baltimore County 2000, 144).
Shortly thereafter, the Office launched the Renaissance Development Initia-
25
Baltimore County is large and urbanized, and the only local government is the county;
there are no municipalities.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
tive, which was a countywide revitalization plan for the first-tier suburbs.
For the next decade, Baltimore County planners systematically revitalized
the first-tier suburbs, investing some $1 billion in local revenue.
The Renaissance Development Initiative became a strategy to target an
aging housing stock and struggling commercial strips throughout the first-tier
suburbs. In particular, planners focused on three suburbs—Dundalk, Essex,
and Middle River—to rehabilitate the housing stock and redevelop commercial areas to attract local businesses (Vicino 2008). The policy implication of
suburban neighborhood differentiation is that a finer-grain understanding
of the transformation process helps planners craft redevelopment strategies.
The more information planners, policy makers, and other stakeholders have
about neighborhoods, the more likely it is that informed decisions can be
made about issues such as the allocation of public resources and community
planning.
Baltimore’s first‑tier suburban neighborhoods were highly differentiated
in 1970 and 2000. The analyses showed that every first‑tier suburban neighborhood changed over that 30-year period and that the degree of change
varied. Ultimately, this study provides a lens for understanding the complex
spatial patterns of socioeconomic decline and diversity in first-tier U.S. suburbs. Future research should continue to examine differentiation patterns in
a comparative context.
Author
Thomas J. Vicino is an assistant professor of political science at Wheaton
College in Norton, MA.
The author thanks Katrin B. Anacker, Bernadette F. Hanlon, Donald F.
Norris, John Rennie Short, Todd Swanstrom, and the anonymous reviewers
who provided valuable feedback.
The research described in this article was funded in part by the U.S.
Environmental Protection Agency, under grant CR83105801, and no official
endorsement should be inferred.
References
Anacker, Katrin B., and Hazel A. Morrow-Jones. 2008a. Exploring Public Policies and
Local Tools in Select First Suburbs in Ohio through Expert Interviews. Unpublished paper.
Metropolitan Institute at Virginia Tech.
Anacker, Katrin B., and Hazel A. Morrow-Jones. 2008b. Mature Suburbs, Property Values and Decline in the Midwest? The Case of Cuyahoga County. Housing Policy Debate
19(3):519–552.
HOUSING POLICY DEBATE
513
514
Thomas J. Vicino
Baltimore County. 2000. Master Plan 2010. Towson, MD.
Beauregard, Robert A. 2006. When America Became Suburban. Minneapolis: University
of Minnesota Press.
Berry, Brian J. L., and Frank E. Horton, eds. 1970. Geographic Perspectives on Urban
Systems. Englewood Cliffs, NJ: Prentice-Hall.
Berry, Brian J. L., and John D. Kasarda. 1977. Contemporary Urban Ecology. New York:
Macmillan.
Berry, Brian J. L., and Phillip H. Rees. 1969. The Factorial Ecology of Calcutta. American
Journal of Sociology 74(5):445–91.
Bollens, Scott A. 1988. Municipal Decline and Inequality in American Suburban Rings,
1960–1980. Regional Studies 22(4):277–85.
Brenner, Neil. 2001. The Limits of Scale? Methodological Reflections on Scalar Structuration. Progress in Human Geography 25(4):591–614.
Choldin, Harvey M., Claudine Hanson, and Robert Bohrer. 1980. Suburban Status Instability. American Sociological Review 45(6):972–83.
Collver, Andrew, and Moshe Semyonov. 1979. Suburban Change and Persistence.
American Sociological Review 44(3):480–86.
Davies, Wayne K. D. 1984. Factorial Ecology. Aldershot, England: Gower.
Farley, Reynolds. 1964. Suburban Persistence. American Sociological Review 29:38–47.
Federal Register. 2008. Poverty Guidelines, Research, and Measurement. 73(1):3971–72.
Frey, William H. 2003. Melting Pot Suburbs: A Study of Suburban Diversity. In
Redefining Urban and Suburban America: Evidence from Census 2000. Vol. 1, eds. Bruce
Katz and Robert E. Lang, 155–80. Washington, DC: Brookings Institution Press.
Geolytics. 2000. Neighborhood Change Database. East Brunswick, NJ.
Guest, Avery. 1978. Suburban Social Status: Persistence or Evolution? American
Sociological Review 48(2):251–64.
Hanlon, Bernadette F., and Thomas J. Vicino. 2007. The Fate of Inner Suburbs: Evidence
from Metropolitan Baltimore. Urban Geography 28(3):1–27.
Hanlon, Bernadette F., Thomas J. Vicino, and John Rennie Short. 2006. The New
Metropolitan Reality in the U.S.: Rethinking the Traditional Model. Urban Studies
43(12):2129–43.
Hartigan, John A. 1975. Clustering Algorithms. New York: Wiley.
Harwood, Herbert Jr. 2003. Baltimore’s Streetcars: The Postwar Years. Baltimore: Johns
Hopkins University Press.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Hill, Edward W., John F. Brennan, and Harold L. Wolman. 1998. What Is a Central City
in the United States? Applying a Statistical Technique for Developing Taxonomies. Urban
Studies 35(11):1935–69.
Hudnut, William H. III. 2003. Halfway to Everywhere: A Portrait of America’s First-Tier
Suburbs. Washington, DC: Urban Land Institute.
Hughes, Holly L. 1993. Metropolitan Structure and the Suburban Hierarchy. American
Sociological Review 58(3):417–33.
Jackson, Kenneth. 1985. Crabgrass Frontier: The Suburbanization of the United States.
New York: Oxford University Press.
Janson, Carl-Gunnar. 1980. Factorial Social Ecology: An Attempt at Summary and Evaluation. Annual Review of Sociology 6:433–56.
Jargowsky, Paul A. 1997. Poverty and Place: Ghettos, Barrios, and the American City.
New York: Russell Sage Foundation.
Katz, Bruce, and Robert E. Lang, eds. 2005. Redefining Urban and Suburban America:
Evidence from Census 2000. Vol. 1. Washington, DC: Brookings Institution Press.
Keating, W. Dennis, and Thomas Bier. 2008. Greater Cleveland’s First Suburbs Consortium: Fighting Sprawl and Suburban Decline. Housing Policy Debate 19(3):457–77.
Kish, Leslie. 1954. Differentiation in Metropolitan Areas. American Sociological Review
19(4):388–98.
Kline, Paul. 1994. An Easy Guide to Factor Analysis. London: Routledge.
Knox, Paul. 1991. The Restless Urban Landscape: Economic and Sociocultural Change
and the Transformation of Metropolitan Washington, DC. Annals of the Association of
American Geographers 81(2):181–209.
Knox, Paul. 1994. Urbanization: An Introduction to Urban Geography. Englewood
Cliffs, NJ: Prentice-Hall.
Kramer, Kathryn, Gordon C. Johnson Jr., and W. Parker Frisbie. 1982. Suburban Differentiation: On the Contemporary Utility of the Employment Residence Typology. Journal
of Urban Affairs 4(2):11–24.
Lang, Robert E. 2003. Edgeless Cities: Explaining the Elusive Metropolis. Washington,
DC: Brookings Institution Press.
Leigh, Nancey Green, and Sugie Lee. 2005. Philadelphia’s Space In Between: Inner-Ring
Suburb Evolution. Opolis: An International Journal of Suburban and Metropolitan
Studies 1(1):13–32.
Listokin, David, and W. Patrick Beaton. 1983. Revitalizing the Older Suburb. New Brunswick, NJ: Rutgers University, Center for Urban Policy Research.
HOUSING POLICY DEBATE
515
516
Thomas J. Vicino
Logan, John R. 1976. Industrialization and the Stratification of Cities and Suburban
Regions. American Journal of Sociology 82(2):333–48.
Logan, John R., and Wenquan Zhang. 2004. Identifying Ethnic Neighborhoods with Census Data: Grouping Concentration and Spatial Clustering. In Spatially Integrated Social
Science, ed. Michael F. Goodchild and Donald G. Janelle, 113–26. New York: Oxford
University Press.
Lucy, William H., and David L. Phillips. 2000. Confronting Suburban Decline: Strategic
Planning for Metropolitan Renewal. Washington, DC: Island.
Lucy, William H., and David L. Phillips. 2006. Tomorrow’s Cities, Tomorrow’s Suburbs.
Chicago: APA Planners.
Masotti, Louis H., and Jeffrey K. Hadden, eds. 1973. The Urbanization of the Suburbs.
Vol. 7. Thousand Oaks, CA: Sage Publications.
Mikelbank, Brian A. 2004. A Typology of U.S. Suburban Places. Housing Policy Debate
15(4):935–64.
Murdie, Robert 1969. Factorial Ecology of Metropolitan Toronto, 1951–1961. Research
Paper No. 116. University of Chicago, Department of Geography.
Orfield, Myron. 1997. Metropolitics: A Regional Agenda for Community and Stability.
Washington, DC: Brookings Institution Press.
Orfield, Myron. 2002. American Metropolitics: The New Suburban Reality. Washington,
DC: Brookings Institution Press.
Outen, Donald. 2005. Baltimore County, Maryland: Using the Entire Toolkit for Habitat
Protection. In Nature-Friendly Communities: Habitat Protection and Land Use Planning,
ed. Chris Duerksen and Cara Snyder, 152–73. Washington, DC: Island.
Park, Robert E., Ernest W. Burgess, and Roderick D. McKenzie. 1925. The City. Chicago:
University of Chicago Press.
Perle, Eugene D. 1981. Perspectives on the Changing Ecological Structure of Suburbia.
Urban Geography 2(3):237–54.
Pratt, Geraldine, and Susan Hanson. 1988. Gender, Class, and Space. Environment and
Planning D: Society and Space 6(1):15–35.
Puentes, Robert, and Myron Orfield. 2002. Valuing America’s First Suburbs: A Policy
Agenda for Older Suburbs in the Midwest. Washington, DC: Brookings Institution, Center on Urban and Metropolitan Policy.
Puentes, Robert, and David Warren. 2006. One-Fifth of America: A Comprehensive
Guide to America’s First Suburbs. Survey Series. Washington, DC: Brookings Institution,
Metropolitan Policy Program.
Schnore, Leo. 1957. The Growth of Metropolitan Suburbs. American Sociological Review
22:165–73.
housing policy debate
The Spatial Transformation of Baltimore’s First-Tier Suburbs
Schnore, Leo. 1959. The Timing of Metropolitan Decentralization. Journal of the
American Institute of Planners 25:200–06.
Schnore, Leo. 1965. The Urban Scene. New York: Free.
Shevky, Eshref, and Wendell Bell. 1955. Social Area Analysis. Westport, CT:
Greenwood.
Short, John Rennie. 2007. Liquid City: Megalopolis and the Contemporary Northeast.
Washington, DC: RFF.
Short, John Rennie, Kim Yeung, M. Kuus, and H. Wells. 1996. The Dirty Little Secret of
World Cities Research: Data Problems in Comparative Analysis. International Journal of
Urban and Regional Research 20(4):697–717.
Stahura, John M. 1979. Suburban Status Evolution/Persistence: A Structural Model.
American Sociological Review 44(6):937–47.
Stahura, John M. 1980. Ecological Determinants of the Aging Suburban Populations.
Sociological Quarterly 21(1):107–18.
Sternlieb, George, and Robert Lake. 1975. Aging Suburbs and Black Homeownership.
Annals of the American Academy of Political and Social Science 422(1):105–17.
Swanstrom, Todd, Colleen Casey, Robert Flack, and Peter Dreier. 2004. Pulling Apart:
Economic Segregation among Suburbs and Central Cities in Major Metropolitan Areas.
Living Cities Census Series. Washington, DC: Brookings Institution, Metropolitan Policy
Program.
Teaford, Jon. 2006. The Metropolitan Revolution: The Rise of Post-Urban America. Baltimore: Johns Hopkins University Press.
Thomas, G. Scott. 1998. The United States of Suburbia: How the Suburbs Took Control
of America and What They Plan to Do with It. Amherst, NY: Prometheus.
U.S. Bureau of the Census. 1994. Geographic Areas Reference Manual. Washington,
DC.
U.S. Bureau of the Census. 2000. Census of Population and Housing, Summary File 1.
Washington, DC.
U.S. Department of Housing and Urban Development. 2000. State of the Cities Data
System 1970 to 2000. Washington, DC.
Vicino, Thomas J. 2008. The Quest to Confront Suburban Decline: Political Realities and
Lessons. Urban Affairs Review 43(4):553–81.
Vicino, Thomas J., Bernadette F. Hanlon, and John Rennie Short. 2007. Megalopolis
50 Years On: The Transformation of a City Region. International Journal of Urban and
Regional Research 31(2):344–67.
HOUSING POLICY DEBATE
517
518
Thomas J. Vicino
Warner, Sam Bass. 1978. Streetcar Suburbs: The Process of Growth in Boston, 1870–
1890. Cambridge, MA: Harvard University Press.
Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass,
and Public Policy. Chicago: University of Chicago Press.
Wyly, Elvin. 1999. Continuity and Change in the Restless Urban Landscape. Economic
Geography 74(4):309–38.
Yin, Robert. 2002. Case Study Research: Design and Methods. 3rd ed. Thousands Oaks,
CA: Sage.
housing policy debate