RACIAL SETTLEMENT AND METROPOLITAN LAND-USE PATTERNS:
DOES SPRAWL ABET BLACK–WHITE SEGREGATION?1
George Galster2
Department of Geography and Urban Planning
Wayne State University
Jackie Cutsinger
Center for Urban Studies
Wayne State University
Abstract: This paper advances a theory of how metropolitan land-use patterns affect racial
settlement patterns and tests it by measuring the relationship between seven dimensions of landuse patterns and five dimensions of segregation of Blacks and Whites for a representative sample
of 50 large metropolitan areas, using multiple regression analysis. We find substantial, nonlinear
relationships between changes in multiple dimensions of segregation and multiple dimensions of
land use, with most evincing a direct relationship between more compact patterns and segregation once a threshold value is exceeded. The results can be explained holistically by positing that
variations in different dimensions of land-use patterns differentially affect land/housing prices,
inter-group propinquity, interracial commonality of commuting destinations, and spatial mismatch, which in turn appear to affect the ability of a metropolitan area to desegregate. But alterations in certain aspects of land use—density/continuity and job compactness—apparently
spawn a combination of forces that affect desegregation in contrary ways; which force dominates
seemingly depends on how extreme the given land-use pattern has become. These findings hold
implications for those designing land-use policies designed to fight sprawl. [Key words: sprawl,
segregation, land use.]
INTRODUCTION
Few topics in the social sciences have been as comprehensively studied as the metropolitan settlement patterns of various groups categorized by their race or ethnicity. After
generations of qualitative investigations (typified by Glazer and Moynihan, 1963), the
causes of these patterns began to be probed statistically during the 1960s, notably in the
Taeubers’ Negroes in Cities (1965). The sophistication of these analyses advanced with
the advent of multiple regression modeling of cross-metropolitan differences in segregation by Roof and Van Valey (1972), Marshall and Jiobu (1975), and Cloutier (1982), and
subsequently with the use of instrumental variables to control for the endogeneity of
interracial economic disparities and residential choices (Galster, 1987, 1991; Galster and
1
The financial support of the U.S. Geological Survey on this project is gratefully acknowledged. The opinions
expressed herein are those of the authors, and do not necessarily reflect those of USGS or the Board of Governors
of Wayne State University. The authors also wish to thank Jason Booza, Casey Dawkins, Up Lim, John Yinger,
and anonymous referees, who provided extremely beneficial commentary on an earlier draft of this work.
2
Correspondence concerning this article should be addressed to George Galster, Department of Geography and
Urban Planning, Wayne State University, Room 3185 Faculty/Administration Building, Detroit, MI 48202;
telephone: 313-577-9084; fax: 313-577-0022; e-mail: [email protected]
516
Urban Geography, 2007, 28, 6, pp. 516–553.
Copyright © 2007 by V. H. Winston & Son, Inc. All rights reserved.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
517
Keeney, 1988). Investigations during the past decade have probed the determinants of
individuals’ residential choices through the use of such databases as the Public Use
Micro-Sample (e.g., Alba et al., 2000), Panel Study of Income Dynamics (e.g., South and
Crowder, 1997), and Multi-City Study of Urban Inequality (e.g., Ihlanfeldt and Scafidi,
2001, 2002). The focus has been broadened beyond Black–White settlement differences
to include numerous additional racial/ethnic groups (e.g., Denton and Massey, 1988;
Massey and Denton, 1989, 1993; Alba and Logan, 1993; Galster and Santiago, 1995;
Frey and Farley, 1996; Freeman, 2000; Iceland, 2004; Varady, 2005).3
Proceeding on a parallel track has been a rapidly mounting literature on the consequences of particular patterns of metropolitan land use, typically categorized by the
rubric “the costs of sprawl” (Burchell, 1997; Freilich and Peshoff, 1997; Benfield et al.,
1999; McKinney, 2000; Ciscel, 2001; Johnson, 2001; Kahn, 2001; Ewing et al., 2002;
Squires, 2002; Lopez and Hynes, 2003; Nechyba and Walsh, 2004; Sarzynski et al.,
2006). This literature, although far from definitive, has tried to quantify the relationships
between how and where land is developed for residential and nonresidential uses and
what transpires for urban economic, transportation, and ecological systems.4
Unfortunately, the dual literatures on the causes of racial/ethnic segregation and the
consequences of sprawl have developed in isolation from each other (Bullard et al., 2000;
Banerjee and Verma, 2005). In both theoretical and empirical work, the segregation
literature has largely overlooked the role of the spatial development of the metropolitan
built environment, and the sprawl literature has employed none of the important conceptual, theoretical, and empirical developments related to understanding the (often
constrained) residential choices made by different household groups.
There are only two points at which these literatures have intersected in the form of
rigorous, multivariate statistical analyses.5 The first is in the study by Huie and Frisbee
(2000) on the relationship between several sorts of residential densities and five dimensions of Black–White segregation proposed by Massey and Denton (1988). They found
for a 1990 sample of the 58 largest metro areas that: (1) both population per square mile
and the number of residential structures per square mile were strongly negatively correlated with the index of Black centralization; and (2) population per square mile was
positively correlated with the Black isolation index, controlling for region and new housing construction. They conclude that “density is an important part of our understanding of
the processes that are involved in the segregation of individual race/ethnic groups” (Huie
and Frisbee, 2000, p. 521).6
3
Reviews of this voluminous literature have been provided by Galster (1988a), Charles (2003), and Dawkins
(2004). Historical perspectives on segregation trends are presented by Massey and Denton (1993) and Cutler et
al. (1999).
4
Much of this lack of consensus can be traced to lack of precision in operationalizing “sprawl” (Galster et al.,
2001).
5
This claim is not intended to minimize the importance of a substantial body of scholarship in political economy that has argued that the ways public policy shaped suburban development patterns has affected racial settlement patterns (e.g., Jackson, 1985; Massey and Denton, 1993; Rusk, 1993; Orfield, 2002; Drier et al., 2004).
None of this literature has attempted to quantify these relationships through the use of multivariate statistical
models, however.
6
These findings may not be robust regarding alternative measures of segregation. Though it was not the focus
of their study, Cutler et al. (1999) regressed a Black–White dissimilarity index and a Black (footnote continues)
518
GALSTER AND CUTSINGER
The second point of intersection involves three statistical studies of the relationship
between various types of land-use restrictions, metro growth control policies, and racial/
ethnic segregation (measured by dissimilarity indices). Pendall (2000) finds that traditional land-use controls (like multi-family and minimum lot-size zoning) produce a chain
of exclusion in which the suburban supply of affordable rental dwellings is restricted and
segregation thereby abetted. Nelson, Sanchez, and Dawkins (2004) and Nelson, Dawkins,
and Sanchez (2004) observe that metropolitan areas with strong, longstanding growth
containment policies exhibit both lower segregation in 2000 and greater declines in segregation since 1990, but areas with state-mandated local housing elements evince the
opposite patterns. Quigley et al. (2004) found for Los Angeles that suburbs mandating
low-density development are much more likely to attract White than Black, Hispanic, or
Asian households.7
As valuable and provocative as these findings are, these works do not provide an
explicit, unified theoretical framework in which land-use patterns and residential choices
of different racial groups can be understood. Moreover, they do not measure explicitly
any of the multiple aspects of metropolitan land-use patterns besides residential density.
Finally, they typically examine only one dimension of racial segregation (evenness).
The purposes of this article, therefore, are twofold: (1) to offer an exploratory model
of the various ways in which metropolitan land-use patterns (to be operationalized below)
may affect racial settlement patterns; and (2) to test this model by measuring the relationship between seven dimensions of land-use patterns and five dimensions of residential
segregation of non-Hispanic Blacks and non-Hispanic Whites (“Blacks” and “Whites”
hereafter), for a representative sample of 50 large metropolitan areas, using multiple
regression analysis. By doing so, we hope to advance the field in three ways. First, we
will synthesize theory developed primarily by sociologists about the causes of segregation with theory developed primarily by economists about the pricing and allocation of
urban land. Second, just as segregation has been shown to have distinct empirical dimensions, we will demonstrate the same for land-use patterns, employing several analogous
indices as well as others of our own devising. Finally, we will show that the empirical
relationships between 1990–2000 changes in segregation and 1990 levels of various landuse factors are nonlinear and contingent on the initial land-use pattern. These findings
prove to have challenging implications for policymakers wishing to influence metropolitan land-use and racial settlement patterns.
LAND-USE PATTERNS: THE UNDERDEVELOPED
ELEMENT OF RACIAL SEGREGATION RESEARCH
Received theory over the last three decades has reached a consensus that five hypotheses have been consistently advanced as explanations of racial segregation (Clark, 1986;
6
(continued) isolation index on the log of population density for a much larger sample of metropolitan areas
in 1990 and found no strong relationships. However, Dawkins (2005) regressed a gini index of neighborhood
segregation on population density for a sample of 231 metro areas in 2000 and found a positive correlation.
7
Banerjee and Verma (2005) related predominant land-use types to racial-ethnic occupancy patterns in Los
Angeles county municipalities.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
519
Galster, 1988a; Dawkins, 2004). Three, related to interracial differences in ability to pay
for housing, preferences for nonracial attributes of the housing package, and housing
information differences, focus on market demand processes that are not explicitly racial
in their operation. Two others, related to prejudices regarding the racial/ethnic composition of neighborhoods and to racial/ethnic discrimination by private and public actors,
focus on race per se as the driving force of segregation. Our purpose in this overview is
not to argue which hypotheses command the most empirical support,8 but rather to show
that none make recourse to arguments about metropolitan land-use patterns.
Inter-Group Differences in Ability to Pay
Given the acknowledged segregation of homes according to price or rent levels in
most urban neighborhoods (Vandell, 1995), one would expect that households differing
in their economic status (income, wealth, or other measures of purchasing power for
housing) would tend to be segregated, even if their preferences and other characteristics
were identical. Add to this the acknowledged economic disparities among racial groups
on average (e.g., Blank, 2001), and it logically follows that part of racial segregation is
due to these economic status disparities (Bayer et al., 2004).
Inter-Group Differences in Preferences for Nonracial Attributes
The argument here is that dwellings possessing particular structural attributes, environmental amenities, accessibility features, or local public service/tax packages may not
be uniformly distributed across a metropolitan area. If, at the same time, households of
different groups systematically differ in their evaluation of the aforementioned attributes,
then their preferences will lead them to locate in different areas where their most
attractive attribute package is most prevalent (Bayer et al., 2004; Nechyba and Walsh,
2004; Dawkins, 2005).
Inter-Group Differences in Housing Market Information
Once segregation is established it may tend to be self-perpetuating insofar as members
from each group lack current or correct information about housing opportunities in neighborhoods inhabited primarily by other groups (Farley et al., 2000). This view posits that
people acquire much valuable housing market information passively as they conduct their
everyday travels to work, shopping, and friends and family (Clark, 1986). But this information is spatially biased because it may systematically over-represent neighborhoods
where other members of their group reside. When households search for different housing
opportunities, they tend to search in areas with which they are already the most familiar,
thereby perpetuating segregation.
8
An unusually vigorous debate on this point was conducted between Clark (1986, 1988, 1989) and Galster
(1988b, 1989); see also Charles (2003) and Dawkins (2004).
520
GALSTER AND CUTSINGER
Prejudice and Preferences for Neighborhood Racial/Ethnic Composition
Another means by which preferences relate to segregation concerns prejudices regarding neighborhood racial composition. Such preferences clearly have differed among
groups for an extended period. Public opinion polls (Farley et al., 1978, 1994; U.S.
Department of Housing and Urban Development, 1978; Schuman et al., 1985; Darden,
1987; Bobo, 2001) consistently reveal that Blacks and Hispanics generally prefer neighborhoods with roughly equal proportions of Whites and minorities. On the other hand,
Whites generally prefer neighborhoods that are all or predominantly White-occupied,
often because of prejudicial attitudes toward minority neighbors. What is less clear is how
these combinations of preferences are translated into segregation, though a variety of
models have been brought forward (Schelling, 1971; Schnare, 1976; Yinger, 1976; Clark,
2007).
Discrimination by Private and Public Actors
Discriminatory acts by private housing market agents, such as landlords and real estate
agents, can cause segregation if they serve to exclude minority home seekers from nonminority neighborhoods into which they otherwise would be willing and able to move,
and/or if they render occurrences of neighborhood integration more transitory. The
former set of acts includes “exclusion” and “misinformation”; the latter includes “blockbusting” and “steering” (Turner et al., 2002; Galster and Godfrey, 2005). Public-sector
actors may also contribute to racial segregation by their policies and practices. For example, historically there is considerable evidence that many housing authorities allocated
tenants on the basis of the composition of the neighborhood in which the subsidized
development was located (Hirsch, 1983; Popkin et al., 2003). Local political jurisdictions
have often enacted exclusionary zoning and other regulations that, while perhaps not
racially segregationist by intention, may prove to be so in effect (Pendall, 2000; Nelson,
Sanchez, and Dawkins, 2004; Nelson, Dawkins, and Sanchez, 2004).
Bringing Land-Use Patterns into the Analysis
This brief review demonstrates that metropolitan land-use patterns typically do not
play a role in existing theories of racial segregation. The only exception involves the
aforementioned policies and practices of local public authorities. The theories related to
prejudice and interracial differences in incomes and housing preferences can, however,
offer vehicles through which the influence of land-use patterns can be analyzed. Instead
of taking as point of departure a given set of racial prejudices and interracial differences
in ability to pay and housing preferences, we explore theoretically and empirically the
extent to which all these may be a product of metropolitan land-use patterns. As elucidated below, we hypothesize that land-use patterns may influence: (1) prejudices through
the degree to which groups are forced into contact via propinquity; (2) ability to pay
through effects on both incomes and land/housing prices; and (3) where groups will find
their most preferred locations through the spatial distribution of amenities and workplaces.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
521
Fig. 1. Illustration of basic model of land rent and allocation to users in an undeveloped city.
THE BASIC THEORY OF METROPOLITAN URBAN LAND-USE PATTERNS
In this section, we review the basic model developed by urban economists to explain
patterns of metropolitan land uses, which we believe offers a useful foundation for our
analysis. We start with an admittedly oversimplified construct to introduce the model’s
logic, and then employ a more realistic variant. We then extract key lessons from this
model that we subsequently expand upon in the next section as a basis for understanding
potential causal connections between land use and racial segregation.
A Model of Land Pricing and Allocation to Households in an Undeveloped City
The model we draw upon was originally developed by Alonso (1964) and Muth
(1969); we review its basic structure and conclusions below.9 In simplest manifestation,
it posits a stylized, featureless plain on which an employment node has been located.10
Assume that transportation can be obtained in any direction from this node at an out-ofpocket cost of $t per unit of distance d. The issue in question is how perfectly informed
households of two different types, A and B, all of whom work in the node, will locate
themselves in relationship to it.
Alonso-Muth provide an answer by specifying a function that shows how much households from each group would be willing and able to bid for land; households will be
allocated the land for which they register the highest bid. Because no land-use patterns
(other than the single employment node) have yet been established nor any housing built,
9
Alonso and Muth provide different but essentially complementary mathematical treatments. To make the
arguments more accessible, we provide a heuristic argument augmented by graphs.
10
For mathematical simplicity, this employment center is typically modeled as having no area, but simply a
point.
522
GALSTER AND CUTSINGER
the only two things that distinguish locations are their distance d to employment and the
potential size of the parcel that can be occupied once land has been platted. Households’
relative willingness to bid for one location vs. another will thus be shaped in this simple
model only by their transport costs and their preferences for land. They will bid more for
a location closer to the employment node because it will save commuting costs (both outof-pocket and wasted time), but the implied higher price per acre will also lead them to
consume a smaller quantity of land, thereby offsetting somewhat the advantage of greater
proximity. Competition among members of the group will ensure that the difference in
bids between any two parcels exactly offsets these countervailing considerations, resulting in the same level of well-being (utility) for all households in the group. The absolute
ability to bid for any location will depend on that group’s purchasing power.
These land bid functions can be most easily pictured in graphical terms (Fig. 1). The
employment node is located at the origin of a graph showing the relationship between
what households would be willing and able to bid for land at various distances. Because
of simplifying assumptions above, the graph is general for any radius extending from the
node in any direction. Let $a be the cost for undeveloped land in its highest-value, nonurban alternative use (e.g., agriculture, grazing, forestry). Now suppose that members of
group A valued leisure time more (i.e., were more averse to commuting) and/or valued
land less (i.e., were more tolerant of density) than members of group B. This would mean
that the bid differential between any two locations d1–d2 distance apart would be greater
for group A than for B. Graphically, group A’s bid function (A1–A1) would appear more
steeply sloped than B’s (B1–B1; Fig. 1).
It is immediately clear from Figure 1 that group A will be inclined to outbid group B
for locations more proximate to employment. But this alone does not define how high the
level of bidding competition between groups will be pushed before an equilibrium land
price gradient is established and final allocations of particular lot sizes proceed. At this
level of simplification, the intensity with which a group will be willing and able to compete will be directly related to its population and its purchasing power.
To illustrate, let A1–A1 and B1–B1 be the current level of land prices. This implies that
the border between groups will be m (because at distances less than m, A will outbid B,
and vice versa for distances greater than m) and the edge of the city (occupied territory)
will be e (at distances less than e, B will outbid non-urban land users). Does this pattern
of land prices represent an equilibrium? Consider the situation from group B’s perspective. Given their commuting expenses $[t × d] at each location d, and their (common)
incomes IB, they will be willing and able to consume Ld acres of land at each location. If
the sum of these acres across all members of group B exceeds the acreage available from
m to e, bid level B1–B1 clearly is insufficient to be equilibrium, and B’s bids will rise en
masse. Conversely, if the sum of these acres across all members of group B is less than
the acreage available from m to e, bid level B1–B1 is excessive and B’s bids will fall. If
group B’s level of bidding rises they will succeed in obtaining more land both from erstwhile group A occupants as the border shifts closer to the employment node, and from
non-urban users at the urban periphery.11 Of course, group A makes the same calculations
11
Not only will land supplied to group B increase, but the group’s quantity of land demanded per household
will fall as its price per acre rises. Both work to eventually quell the need for further bid increases.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
523
of optimal total land consumption by the group, given current land price level A1–A1,
group A population and A incomes, and adjusts the level of bidding accordingly.
This inter-group competition and adjustment continues until both groups cannot
improve their well-being by adjusting levels of bidding further.12 At this point, equilibrium has been established for a land price gradient, consumption of land at each distance,
size of the urban area, average urban density (given total population and urban area size),
and locations of both household groups, which are rigidly segregated according to distance from the node. In this simple model, segregation is produced to the degree to which
groups differ in their out-of-pocket transportation costs, preferences for land, and preferences for time spent commuting (Mills and Hamilton, 1994).
Extensions of the Model When Applied to a Developed City
The basic principles developed in the model above lead to more interesting implications regarding segregation if we relax some of the simplifying assumptions and allow
household preferences to have more dimensions. For example, assume that an equilibrium allocation of land has occurred at the inception of our stylized city above, and households have subsequently constructed dwellings on their properties. Insofar as dwellings
are often extremely durable compared to the tenure of their occupants, household allocations among parcels in a developed city will need to take existing dwellings into account.
In similar fashion, additional amenities associated with location can be added to the
model, such as environmental quality. So, for example, if we assume that the quality of
built housing13 and/or air quality monotonically increase as one locates farther from the
employment center of our stylized city, the slopes of the bid functions in Figure 1 will
flatten because compensations besides larger lots and lower densities have been added
that are also associated with distance from work. To the extent that one group’s preferences value housing or air quality above the other, their bid function will flatten further.
Another modification can be made by allowing groups to have a preference for certain
types of neighbors. One simple way to model inter-group prejudice here is to let the desirability of a location be influenced by its proximity to households in the other group.14 To
illustrate, assume that all group B members suddenly acquire a prejudice that makes them
averse to living near group A. In the context of Figure 1, this would mean that the group
B bid function must be altered so that it becomes less than it was originally as proximity
to the border m increases; call it function B2–B1. But such a change would create a discontinuous price contour at m, with A members now outbidding B for locations slightly
beyond m. Inasmuch as group A would get more land than they desire to consume, they
can afford to reduce the level of their bidding (to A2–A2); conversely, group B must raise
its bid level to B3–B3 until equilibrium is re-established with a new border m' > m.15 At
this equilibrium, members of group B living at the border m' are compensated for this
12
At this equilibrium, there will be intra-group equality of utility, but inter-group equality of utility is unlikely.
By “quality” we mean the value that households place on the entire bundle of attributes constituting the housing package.
14
Bid models using this approach were first developed by Yinger (1976), Courant and Yinger (1977), Courant
(1978), and Kern (1981).
15
The dynamics of these so-called “border models” were first explored by Bailey (1959).
13
524
GALSTER AND CUTSINGER
disamenity by paying less for housing than other B members surrounded only by others
like themselves (who pay a premium for this amenity). Note how, even if the two groups
had identical transportation cost profiles and preferences, segregation results if one group
expresses an aversion to living near the other and is willing to pay for it.
Finally, the model can easily accommodate housing and land-use restrictions imposed
by the public sector, such as construction and density limitations or nonresidential zoning
regulations. If particular locations have their land-market responses constrained, the ability (and, sometimes, willingness) of different household groups to bid there will be
affected in predictable ways.
Lessons from the Models for the Analysis of Racial Segregation
The foregoing model of how households are allocated to various locations within a
metropolitan area through a market process of competitive bidding provides important
lessons for the analysis of segregation. In the context of the model, residential segregation
results whenever different household groups evince differences in their willingness and/
or ability to bid for the same locations. A group’s willingness to bid is, in turn, related to
its (1) profile of workplace locations and out-of-pocket commuting costs thereto; (2) preferences for land (density) relative to time (which may be devoted to commuting or
leisure); (3) preferences for housing quality or other amenities associated with the parcel;
and (4) prejudices regarding certain types of neighbors. A group’s ability to bid is related
to its real income (and wealth), that is, their purchasing power vis-à-vis land/housing
packages in the city. These lessons form the framework of our analysis of sprawl and
racial segregation because, as will be seen in the next section, metropolitan land-use
patterns likely affect the profile of workplace locations, real incomes, and prejudices
regarding certain types of neighbors.
APPLYING THE THEORY TO UNDERSTANDING THE CONNECTIONS
AMONG MULTIPLE DIMENSIONS OF BOTH RACIAL SEGREGATION
AND METROPOLITAN LAND USE
In this section, we first review briefly the existing literature on the distinct conceptual
and empirical dimensions of spatial patterns of both racial segregation and land use. We
then apply the aforementioned lessons of the land-use theory to comprehend how various
dimensions of land usage might affect the various dimensions of racial segregation.
Five Dimensions of Segregation
White (1986), Massey and Denton (1988), and Massey et al. (1996) have theoretically
elaborated and empirically validated five distinct dimensions of residential segregation:
(1) (Un)Evenness: the degree to which two groups are differentially distributed across
areal units (for consistency with other dimensions’ scaling, we think of this dimension as
unevenness that is positively associated with segregation); (2) Exposure (Isolation): the
degree of potential contact, or the possibility for interaction, between two groups (for
consistency with other dimensions’ scaling, we think of this dimension as isolation—lack
of exposure—which is positively associated with segregation); (3) Concentration: the
RACIAL SETTLEMENT AND LAND-USE PATTERNS
525
relative amount of space occupied by a minority group in the metropolitan area (higher
minority concentration is associated with greater segregation); (4) Centralization: the
degree to which the minority group is located near the center of a metropolitan area
(higher minority centralization is associated with greater segregation); (5) Clustering: the
extent to which areal units inhabited by minority residents adjoin one another (higher
minority clustering is associated with greater segregation). These five dimensions have
become the standard set, which we will employ in our theoretical discussions and empirical modeling below.
Seven Dimensions of Metropolitan Land Use
Unlike the measurement of segregation, the measurement of metropolitan land-use
patterns is in its infancy and little, if any, consensus has developed as to which indices
should be employed (cf. Malpezzi, 1999; Fulton et al., 2001; Glaeser et al., 2001; Kahn,
2001; Malpezzi and Guo, 2001; Ewing et al., 2002; Lopez and Hynes, 2003; Pendall and
Carruthers, 2003; Anthony, 2004). Recent work shows, however, that analogues to the
foregoing dimensions of segregation can fruitfully be employed in the measurement of
land-use patterns.
Galster et al. (2001), Cutsinger et al. (2005), and Wolman et al. (2005) have argued
that at least seven distinct dimensions of metropolitan land use can be identified, and
indices can be operationalized for each to measure job and housing development patterns.
Each of the dimensions is specified such that higher values are associated with attributes
commonly thought of as more compact, less “sprawl-like” development: (1) Density: the
degree to which the housing units or jobs within a metropolitan area are developed in an
intensive manner relative to land area capable of being developed (higher density is less
sprawl-like); (2) Continuity: the degree to which developable land has been developed in
an unbroken fashion throughout a metropolitan area (higher continuity means less “leapfrog” sprawl); (3) (Un)Evenness: the degree to which a given land use (or housing units
or jobs) is distributed unevenly across the square-mile cells comprising a metropolitan
area (greater unevenness is associated with a more compact, less sprawl-like development pattern); (4) Centrality: the degree to which a land use is located nearer the core of
a metropolitan area, relative to the land comprising the metropolis (more centralized
development is less sprawl-like); (5) Proximity: the closeness of housing units, jobs, or
housing unit/job pairs to each other across a metropolitan area, relative to its total land
area (greater proximity is associated with less sprawl-like development); (6) Mixed-Use:
the degree to which housing units and jobs are located in the same square-mile cells comprising a metropolitan area (greater mixing of uses is typically viewed as a less-sprawling
pattern of development); (7) Nuclearity: the degree to which major metropolitan job clusters are disproportionately located in the core, as opposed to dispersed in a multicentric
framework (greater relative clustering of jobs in the core is associated with less sprawllike development).
Potential Causal Connections between Dimensions of Land Use
and Dimensions of Racial Segregation
The foregoing subsections have been a prelude to the main task at hand: articulating
how multiple land-use dimensions may affect dimensions of racial residential segregation.
526
GALSTER AND CUTSINGER
Overall, our analysis suggests that: (1) all dimensions of metropolitan housing and
employment land uses will affect segregation to the degree that they alter real household
incomes through changes in the average price of land/housing packages (“land” hereafter); (2) patterns of housing development will affect segregation to the degree that they
change inter-group prejudices through propinquity and contact; (3) patterns of how
employment nodes are developed will affect segregation to the degree that commuting
cost gradients are changed and interracial income differentials are modified through
adjustments in spatial mismatch; (4) the density of development and the intensity at
which land uses are mixed will affect segregation to the degree that different racial/ethnic
groups differ in their preferences for density and mixture of uses; (5) changes in most
dimensions of housing and employment land uses are likely to generate countervailing
forces influencing segregation. Thus received theory suggests that different aspects
of metropolitan land usage will have different causal impacts, depending on the particular aspect of segregation being considered.16 We therefore conduct a comprehensive
empirical analysis wherein all dimensions of both land use and racial segregation are
investigated.
Land-use patterns, housing prices, and interracial real income differentials. To the
extent that land is developed intensively and compactly for either jobs or housing, the
average price per acre of developed land will be higher, thereby reducing the real incomes
of all residents, everything else being equal. As explained above in the context of the
“income difference” theory of segregation, the fact that the distribution of most minority
groups’ income is shifted to the left of that of Whites, and that most housing is spatially
separated according to affordability, means that some racial segregation will be produced
by interracial differences in ability to pay. But with higher land prices overall this
interracial affordability gap becomes intensified and segregation should, therefore, be
abetted.17
What this implies is that density, continuity, centrality, proximity, and nuclearity
should all be positively related to every dimension of segregation through the land price/
affordability differential argument.18 The evenness of development should have a more
localized impact: in places where the city is more intensively developed, the higher land
prices should lead to a larger affordability gap and hence greater segregation in the vicinity. Although this may yield differences in most dimensions of segregation in the aggregate, there is no a priori reason to expect it will be related to centrality. There is no reason
why the mix of land uses should affect the overall land price gradient unless there is a
consensual preference that a metro’s average degree of land-use mixture represents an
amenity or disamenity of significance.
Housing, propinquity, and prejudice. A metropolitan area with housing developed
at higher densities and in greater proximity is more likely one in which mundane activities of life will bring one in contact with increasingly diverse kinds of people. Indeed,
16
Similar, but less comprehensive, arguments have been advanced by Pendall (2005).
Let the initial mean annual expenditure on housing for White and minority households be PW and PM, respectively, yielding an initial mean interracial affordability gap of PW – PM. Now, if all housing prices rise by M
percent, the new affordability gap will be (1+ M) (PW – PM).
18
The limited empirical evidence on urban form and housing prices is inconsistent; cf. Kahn (2001) and
Wassmer and Baas (2006).
17
RACIAL SETTLEMENT AND LAND-USE PATTERNS
527
propinquity as a determinant of social interaction has been a longstanding sociological
principle (Hawley, 1972). Such interaction, in turn, may provide the basis for eroding
inter-group prejudices, especially to the degree that these contacts are among groups of
similar socioeconomic status (Allport, 1954; Jackman and Crane, 1986; Sigelman et
al., 1996; Emerson et al., 2002; Ihlanfeldt and Scafidi, 2002). However, there is also
theory and evidence to suggest that such interaction may only intensify conflict and
prejudices, especially when minorities are increasing in number and groups of lower
status are mixed (Bobo, 1988; Glaser, 1994; Branton and Jones, 2005). We are therefore unable to predict how more compact development patterns should be associated
with inter-group prejudices and, thereby, levels of residential segregation in all
dimensions.
Jobs, commuting gradients, and spatial mismatch. One factor affecting households’
residential choices is the accessibility of potential residences to employment, which is
defined in terms of the location of their employment. So, to the extent that there are
many different nodes at which significant numbers of jobs are clustered, and that these
nodes differ dramatically in the racial/ethnic composition of their workforces, the probability will rise that workers of different racial/ethnic categories will bid most intensely
for separate locations within the metropolitan area. This implies that a development
pattern with most jobs clustered in or near the city center should have a single, common
peak in the commuting cost gradient (as per Fig. 1) and thus exhibit less racial
segregation.
The distribution of employment may have a second impact on segregation, however,
related to the geographic disjuncture between (typically centrally located) minorityoccupied residential areas and (typically suburban) locations of expanding, well-paying
job opportunities. This spatial mismatch hypothesis, first articulated by Kain (1968), has
been thoroughly investigated and shown to produce interracial differences in employment rates and wages in numerous metropolitan areas (Ihlanfeldt and Sjoquist, 1998).
This means that metropolitan areas having (1) higher shares of jobs more centrally
located, (2) greater numbers of jobs mixed with residential areas, and (3) smaller average distances between jobs and residences would be predicted to have smaller interracial
differences in incomes, and thus lower segregation in all dimensions based on differences in ability to pay.
Differential preferences for land-use amenities. Groups would have no basis for sorting on the basis of different preferences for amenities if such amenities were evenly
distributed across the metropolitan landscape. Two important amenities are generated by
metropolitan development patterns and are likely to vary across space, however. Both
population density and the mixture of residences and jobs at a small geographic scale are
typically disproportionately greater nearer the core (the latter because these areas often
were developed in a time of fewer zoning controls). Intrametropolitan variations in density and land-use mixing will be greater in metros areas with greater spatial unevenness
of residences and jobs. If different racial groups differed substantially in their preferences
for low density and/or homogeneous land-use areas, their bidding would tend to focus on
different parts of the metropolis. If Whites were to have the strongest preferences for
these amenities, this line of argument would imply that metropolitan areas with higher
densities, unevenness, or mixes of land uses will have Whites more segregated; they
528
GALSTER AND CUTSINGER
should especially be located nearer the periphery and concentrated in smaller sectors
where their preferred amenities are available.19
Countervailing forces. The foregoing has articulated different potential causal paths
through which different aspects of metropolitan land use may affect residential segregation. These paths manifest themselves in the following ways: (1) the “propinquity effect”
may change prejudices; (2) the “land/housing price effect” and the “spatial mismatch”
effect may change ability to pay; and (3) the “amenity” effect may change the most
desired locations. Seen holistically, this theoretical discussion reveals a deeper hypothesis: the same dimension of land use may engender countervailing effects on segregation.
This is most readily seen by summarizing the above predictions: (1) density and unevenness directly related to segregation via land price and amenity effects, but (possibly)
inversely related via propinquity effect; (2) continuity directly related to segregation via
the land-price effect, but (possibly) inversely related via the propinquity effect; (3) centrality directly related to segregation via the land-price effect, but inversely related via the
commuting profile effect; (4) proximity directly related to segregation via the land-price
effect, but inversely related via the propinquity and spatial mismatch effects; (5) mixed
use directly related to segregation via amenity effects, but inversely related via the spatial
mismatch effect; (6) nuclearity directly related to segregation via the land-price effect,
but inversely related via commuting profile and spatial mismatch effects.
From theory, we have no way of knowing whether one or the other countervailing
effect predominates over a certain range of the particular land use, or whether they simply
cancel each other out in most circumstances. However, if one or more countervailing
effects are nonlinear in their impacts, it may well be the case that one and then the other
predominates as the range of land-use patterns varies. In our empirical modeling, we will
experiment with nonlinear regression specifications to test for this possibility.
DATA, MEASURES, AND METHODS
Sampling
In this study, we analyzed a sample of 50 metropolitan areas drawn from a pool of the
100 largest metropolitan areas in the United States, based on 1990 population.20 This
sample was regionally stratified and then a proportionate random sample was drawn from
each of the four Census regions. The final sample includes 11 metros from the Northeast
region of the country, 11 metros from the North-Central region, 12 metros from the
Western region, and 16 metros from the Southern region (Table 1).
19
We are unaware of any evidence suggesting interracial differences in amenity preferences (independent of
income), though evidence regarding such differences in components of the dwelling package (Galster, 1979)
makes this a plausible argument.
20
Calculation of the various land-use indices is an extremely resource-intensive process, and support was only
available to work with 50 metros. Moreover, our indices require information from the U.S. Geological Survey’s National Land Cover Data Base; unfortunately, the 2002–2003 NLCDB was not yet available, forcing us
to use 1990 land-use values. Our land-use data are drawn from geographic areas that represent the commutersheds of metropolitan areas. These are explained more fully in Wolman et al. (2005).
529
RACIAL SETTLEMENT AND LAND-USE PATTERNS
TABLE 1. SAMPLE OF 50 METROPOLITAN AREAS
Region
W
W
W
W
W
W
W
W
W
W
W
W
M
M
M
M
M
M
M
M
M
M
M
NE
NE
NE
NE
NE
NE
NE
NE
NE
NE
NE
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
Metro area
Denver, CO
Fresno, CA
Las Vegas, NV
Los Angeles/San Bernardino/Riverside, CA
Phoenix/Mesa, AZ
Portland/Vancouver, OR
Salt Lake City/Ogden, UT
San Diego, CA
San Jose, CA
Seattle/Bellevue/Everett, WA
Stockton/Lodi, CA
Tacoma, WA
Cincinnati, OH
Columbus, OH
Detroit, MI
Fort Wayne, IN
Grand Rapids/Muskegon/Holland, MI
Indianapolis, IN
Milwaukee/Waukesha, WI
Minneapolis/St. Paul, MN
Omaha, NE
St. Louis, MO
Youngstown/Warren, OH
Albany/Schenectady/Troy, NY
Allentown/Bethlehem/Easton, PA
Boston, MA
Buffalo/Niagara Falls, NY
New Haven/Meriden, CT
Philadelphia, PA
Pittsburgh, PA
Providence/Fall River/Warwick, RI
Rochester, NY
Syracuse, NY
Worcester, MA
Atlanta, GA
Baltimore, MD
Baton Rouge, LA
Charlotte, NC
Dallas, TX
El Paso, TX
Houston, TX
Jacksonville, FL
Miami, FL
Mobile, AL
New Orleans, LA
Raleigh/Durham/Chapel Hill, NC
San Antonio, TX
Tulsa, OK
Washington, DC
Wilmington/Newark, DE
MSA code
1990 population
2080
2840
4120
4480
6200
6440
7160
7320
7400
7600
8120
8200
1640
1840
2160
2760
3000
3480
5080
5120
5920
7040
9320
8160
0240
1120
1280
0160
6160
6280
6480
5480
8000
9240
0520
0720
0760
1520
1920
2320
3360
3600
5000
5160
5560
6640
7240
8560
8840
9160
1,622,980
755,580
852,737
8,863,164
2,238,480
1,515,452
1,072,227
2,498,016
1,497,577
2,033,156
480,628
586,203
1,526,092
1,345,450
4,266,654
456,281
937,891
1,380,491
1,432,149
2,538,834
639,580
2,492,525
600,895
742,177
595,081
3,227,707
1,189,288
861,424
4,922,175
2,394,811
1,134,350
530,180
587,884
478,384
2,959,950
2,382,172
528,264
1,162,140
2,676,248
591,610
3,322,025
906,727
1,937,094
476,923
1,285,270
855,545
1,324,749
708,954
4,223,485
513,293
530
GALSTER AND CUTSINGER
Measuring Dimensions of Residential Segregation
Massey and Denton (1988) and Massey et al. (1996), on theoretical and empirical
grounds, have built the case for preferred indices for the five dimensions of segregation,
which are employed in this study. Because these indices are well known and were defined
in the previous citations, they only need listing here: (1) (Un)Evenness: measured with
the Index of Dissimilarity (D) between Black- and White-occupied locations; (2) Exposure (Isolation): measured by the Black Isolation Index (xP*x, or I* hereafter), standardized by the proportion of Blacks in the metropolitan area21; (3) Concentration: measured
by the Delta (DEL) index for Blacks22; (4) Centralization: measured by the index of
Relative Centralization (RCE) of Blacks compared to Whites; (5) Clustering: measured
by the Absolute Clustering (ACL) index for Blacks.23 Higher values of all these measures
are indicative of greater segregation of Blacks. We employ the 1990 and 2000 values of
these indices as calculated by the U.S. Bureau of the Census, based on census tract geography (Iceland et al., 2002). Descriptive statistics and fuller explanations of these indices
are presented in Table 2.
Measuring Dimensions of Metropolitan Land Use
Our measures are based on the conceptual and empirical foundations established in
our prior work (Galster et al., 2001; Cutsinger et al., 2005; Wolman et al., 2005;
Cutsinger and Galster, 2006). Using Geographic Information System (GIS) software, we
first superimposed a virtual grid of square-mile cells over each sampled metropolitan statistical area (MSA). Second, we tabulated numbers of housing units and jobs in each cell,
using data from the 1990 Census of Population block counts and the 1990 Census Transportation Planning Package.24 Third, we specified an “extended urbanized area” (EUA)
as the geography employed for calculating indices; this represents the developed commuter zone of the MSA that is appropriate for measuring land-use patterns.25 Fourth, we
estimated the amount of land in each cell that cannot be developed for physical reasons,
21
Though Massey and Denton (1988) and Massey et al. (1996) recommended the unstandardized P*, we (like
Glaeser and Vigdor, 2003) think the standardized index is the superior segregation measure based on conceptual grounds because it controls for the metro-wide share of the minority. The standardization used was:
100*(1–[b/I])/(1–b), where b is the proportion of Blacks in the metro and I is the unstandardized isolation
index, scaled to a maximum 1.0.
22
Massey et al. (1996) found DEL the superior measure of concentration in their 1990 sample.
23
Massey et al. (1996) found ACL the superior clustering measure in their 1990 sample.
24
The 1990 U.S. Census of Population and Housing provided block-level counts for housing units in each metropolitan area, which were aggregated up to one-square-mile cells overlaid on metros via GIS, our spatial unit
of observation. When gridlines divided census blocks, the number of housing units allocated to each cell was
based on the proportion of the block that fell within a cell. Employment data were obtained from the 1990 Census Transportation Planning Package (CTPP) dataset available from the Bureau of Transportation Statistics.
This dataset included the geographic boundary files for traffic analysis zones (TAZ) and the CTPP Urban Part
II: Place of Work data, which collectively allowed us to allocate the number of jobs to each cell based on the
proportion of each TAZ that fell wholly or partially within a cell.
25
To specify the EUA we exclude from the MSA those square-mile cells that do not have at least (1) 30% or
more of its employees traveling to the urbanized area; and (2) 60 housing units per square mile, thereby avoiding rural areas of the MSA. For conceptual discussion and operational details, see Wolman et al. (2005).
531
RACIAL SETTLEMENT AND LAND-USE PATTERNS
TABLE 2. DESCRIPTIVE STATISTICS FOR BLACK–WHITE SEGREGATION VARIABLES
Dependent
variables
2000
1990
Definitiona
Mean
Std. Dev.
Mean
Std. Dev.
Dissimilarity
(D)
Proportion of Blacks needing to change areal
units to achieve distribution across areal
units equal to Whites’ distribution.
60.71
11.28
64.46
10.63
Standardized
Isolation
(Std. I*)
Average percentage of Blacks in typical
Black’s areal unit, adjusted by the proportion of Blacks in metro area.
75.81
12.45
82.01
11.35
Relative
Centralization (RCE)
Share of Blacks that would have to move
farther from city center to match degree
of centralization of Whites.
33.58
20.41
33.98
22.41
82.21
6.65
84.28
7.17
26.62
15.96
27.25
15.65
Concentration Move across areal units to achieve a uniform
(DEL)
density across the metro area.
Absolute
Clustering
(ACL)
The average number of Blacks in nearby
Census tracts as a proportion of the total
population in nearby tracts.
a
For complete definitions, formulas, and evaluations of these indices, see Massey and Denton (1988),
Massey et al. (1996), and Iceland et al. (2002).
Source: U.S. Census Bureau special report on housing patterns (www.census.gov/hhes/www/housing/
ressseg.html).
employing satellite imagery from the 1992 U.S. Geological Survey’s National Land
Cover Database.26 Finally, we computed 14 land-use indices as described below.
Note at the outset that all our land-use indices are consistently scaled such that higher
values indicate greater compactness and lower degrees of sprawl-like characteristics on
that dimension. The aforementioned seven dimensions of land use we analyze here, typically operationalized with analogous indices for housing and jobs, follow.
Density. We operationalize two comparable density indices, one each for housing
units and jobs within the EUA: (1) Housing Unit Density of Developable Land—the
average number of housing units per square mile of developable land in the EUA; (2) Job
Density of Developable Land—the average number of jobs per square mile of developable land in the EUA.
Continuity. We distinguish two types of continuity: micro-continuity and macrocontinuity. Micro-continuity measures the extent to which developable land within the
EUA has been skipped over; macro-continuity measures the extent to which development
proceeds continuously from the edges of the more intensively developed urbanized area
26
Data from the U.S. Geological Survey’s National Land Cover Database provided information about land coverage types that we used to define for each cell the proportion of land in three categories: developed, developable, and undevelopable.
532
GALSTER AND CUTSINGER
or, instead, exhibits a leapfrog or scattered pattern to the edge of the EUA. Micro-continuity
and macro-continuity are operationalized this way: (1) micro-continuity—percentage of
square-mile cells within the EUA in which 50% or more of the land that is or could
be developed has been developed; (2) macro-continuity—the share of the EUA that is
classified as the Urbanized Area (UA)27 by the U.S. Census Bureau.
Unevenness. Our indices are identical to the Dissimilarity Index, wherein the proportion of all observations of the given land use located in a particular square-mile cell is
contrasted to the proportion of all land in the EUA represented by that cell. We operationalize D indices for both housing and jobs: (1) Housing Unit Unevenness—the percentage
of housing units that would need to move in order to produce an even distribution of
housing units across cells comprising the EUA; (2) Jobs Unevenness—the percentage of
jobs that would need to move in order to produce an even distribution of jobs across cells
comprising the EUA.
Centrality. We define the core of the EUA as the location of city hall and measure the
distance between it and the centroid of each square-mile cell in the EUA, weighted by
the number of homes or jobs in each. We standardize this weighted average distance by
the (unweighted) average distance to city hall from each centroid of the cells comprising
the EUA, so as not to inevitably specify larger EUAs as less centralized.28 Centrality is
operationalized by two indices: (1) Housing Centrality—the ratio of the average distance
to city hall of centroids of the cells comprising the EUA to the average distance to city
hall of a housing unit within the EUA; (2) Job Centrality—the ratio of the average
distance to city hall of centroids of the cells comprising the EUA to the average distance
to city hall of a job within the EUA.
Proximity. Proximity, like centrality, utilizes weighted averages of the distance
between jobs, housing units, or job/housing unit pairs across all square-mile cells in the
EUA so that sparse jobs and housing units on the urban fringe (and therefore less proximate to denser clusters of jobs and housing units near the urban core) do not overly influence estimates. The standardized proximity index adjusts for EUA size in a similar
manner as centrality. We operationalize three proximity indices: (1) Housing Proximity—
the ratio of the average distance among centroids of cells in the EUA to the weighted
average distance among housing units in the EUA; (2) Job Proximity—the ratio of the
average distance among centroids of cells in the EUA to the weighted average distance
among jobs in the EUA; (3) Jobs to Housing Proximity—the ratio of the average distance
among centroids of cells in the EUA to the weighted average distance among jobs and
housing units in the EUA.
Mixed-use. The mixed-use indices are based on exposure (P*) indices and measure
exposure of jobs to housing and vice versa. P* measures the average presence of one
land-use type in the square-mile cells occupied by another type: (1) Mixed-use of Jobs to
Housing—the average number of housing units in the same cell as a job; (2) Mixed-use
of Housing to Jobs—the average number of jobs in the same cell as a housing unit.
27
UAs have at least 1,000 inhabitants per square mile and meet several other criteria.
As per Lopez and Hynes (2003), we believe that sprawl indices should not tautologically vary simply because
a metro is larger in physical scale.
28
RACIAL SETTLEMENT AND LAND-USE PATTERNS
533
Nuclearity. We specify square-mile cells as nuclei, either at the core or sub-centers
outside the core, if they contain 8,000 or more employees, plus any adjacent cells (including
those touching only at their corners) containing 4,000 or more employees. Any two adjacent cells, each of which contains 4,000 or more employees, that were separated from
another nucleus by at least one cell containing less than 4,000 employees, is also considered a nucleus. We operationalize one nuclearity index: Core-dominated Nuclearity—the
ratio of jobs in the core nucleus to jobs in all nuclei (including the core).29 Descriptive
statistics for all land-use indices are displayed in Appendix A.
To reduce the aforementioned 14 indices of metropolitan land-use patterns to a more
manageable number, we conducted a principal components (factor) analysis on indices
calculated for the aforementioned sample of 50 large U.S. metropolitan areas (converted
to EUAs). We found that the analysis produced seven distinct components (Appendix
B).30 Based on the land-use characteristics that are most closely related to (i.e., load most
highly on) each factor, we label these factors: (1) density/continuity, (2) proximity, (3)
job compactness, (4) mixed-use, (5) housing centrality, (6) nuclearity, and (7) housing
unevenness. These seven factors cumulatively explain 94% of the variation in the original
14 indices. Below we provide an intuitive discussion of how each factor measures a
particular spatial aspect of land use.
Both of the density indices and both of the continuity indices load highly on the
density/continuity factor. Higher values of this factor indicate metros that are more intensively developed and have little bypassed developable land. After rotation, this component accounts for 25% of the total variance in the original indices. The proximity factor is
characterized by high loadings from the housing proximity and housing-to-job proximity
indices, and accounts for 15% of the total variance. Higher values of proximity indicate
metros where the distances among homes and among homes and jobs are considerably
less than average distance between centers of the cells comprising the area. The job compactness factor is characterized by high loadings from the job unevenness index, the job
centrality index, and the job proximity index, and accounts for 15% of the total variance.
Higher values of job compactness are indicative of jobs clustered relatively closer to each
other and the core. Both mixed-use indices load highly on the mixed-use factor, which
accounts for 15% of the total variance. Higher values indicate metros having a greater
diversity of residential and nonresidential uses within a square-mile cell. The housing
centrality index is the only measure that loads highly on the housing centrality factor,
which accounts for 8% of the total variance. Higher housing centrality indicates metros
where housing is relatively much closer to the core than the average square-mile unit of
land comprising the area. The sole nuclearity index loads highly onto the nuclearity
factor, which accounts for 8% of the total variance. High nuclearity indicates metros
where the only major employment cluster is in or near the Central Business District.
Lastly, the housing unevenness index loads highly on the housing unevenness factor,
which accounts for 7% of the total variance in the original indices. Higher values of this
29
Core is operationalized as the nucleus containing or adjacent to the cell containing the city hall of the major
municipality defining the metropolitan area.
30
It is merely a coincidence that the number of factors derived is the same as the number of dimensions of land
use that we described and discussed earlier. For details, see Cutsinger et al. (2005).
534
GALSTER AND CUTSINGER
TABLE 3. DESCRIPTIVE STATISTICS FOR LAND-USE VARIABLESa
Explanatory variables
Definition
Mean
Std. Dev.
Density/Continuity
The extent to which housing and jobs are densely
distributed and the degree to which development
occurs in a continuous fashion.
1.75
1.00
Housing Proximity
The extent to which housing units are located nearer
one another, and how near a typical housing unit is to
a typical job, relative to the average distance between
centroids of grids comrpising the EUA.
3.07
1.00
Job Distribution
The degree to which jobs are concentrated in particular
areas, situated nearer the CBD, and located nearer one
another, relative to the average distance between
centroids of grids comprising the EUA.
1.47
1.00
Mixed Use
The extent to which jobs and housing are co-located
within the same square-mile, averaged across grids
comprising the EUA.
2.25
1.00
Housing Centrality
The extent to which housing units are situated nearer
the CBD, relative to the average distance between
centroids of grids comprising the EUA and the CBD.
2.26
1.00
Nuclearity
The degree to which jobs contained in clusters of at
least 8,000 jobs per square mile are located in the core
of the EUA, rather than spread among multiple job
centers.
2.11
1.00
Housing Concentration
The extent to which housing units tend to be clustered in
particular square-mile units comprising the EUA.
2.00
1.00
a
Standard deviations for independent variables are all equal to one because sprawl variables are measured
as z-scores. Means do not equal zero because minimum values from each distribution were added to each
score so that there were no negative values.
Sources: 1990 Census of Population and Housing (U.S. Bureau of the Census, 1991), 1992–1993 National
Land Cover Database (U.S. Geological Survey, 2000), and 1990 Census Transportation Planning Package
(Bureau of Transportation Statistics, 1991).
factor indicate metros with housing concentrated in considerably different degrees across
square-mile cells.
The way in which our 14 land-use indices load onto the seven factors demonstrates
some interesting underlying patterns among the indices that had corresponding measures
for housing and jobs. The factors describing job patterns typically (with the exception of
density) are quite distinct from those describing housing patterns, indicating differences
in how jobs and housing units are dispersed throughout metropolitan space.
We generate variables from these factors by employing the associated factor score
coefficients when all composite variables have been normalized. Because we will be
employing quadratic specifications of each of these seven land-use indices and therefore
desire to eliminate negative values, we rescale by adding the minimum value to each.
Descriptive statistics for these rescaled land-use indices are displayed in Table 3.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
535
Regression Model
Two challenges must be overcome in order to specify a multiple regression model of
segregation and metropolitan land-use patterns: causality (endogeneity) and metropolitan
idiosyncrasies. This paper focuses on uncovering the causal paths from land use to segregation, but it is plausible that causation may also flow in the opposite direction. For example, assume that Whites in a particular, highly segregated metropolitan area show a strong
preference for predominantly White-occupied neighborhoods. However, because of, say,
intensified fair housing enforcement efforts and a robust economy that disproportionately
enriched minority households, many formerly all-White neighborhoods start experiencing a noticeable in-migration of minority households, and segregation declines. Prejudiced Whites may thus be willing to pay an enhanced premium for the few remaining allWhite neighborhoods located far from the mixed ones, thereby signaling developers to
produce exactly those kinds of places, which ultimately become manifested as altered
land-use patterns years later. Thus, past levels of segregation may influence current landuse patterns, although all tend to change slowly over time and influence each other with
a considerable lag.
As for idiosyncrasies, each metropolitan area has a unique history of race relations,
industrial restructuring, and public policies that have either abetted or eroded segregation.
Clearly it is impossible to explicitly measure all these idiosyncratic features or to summarize them with a metro-specific fixed effect, because there would be as many dummy
variables as observations in a typical cross-sectional model.
To meet both of these challenges, we estimate a regression model that relates year2000 cross-metropolitan variation in an index measuring one of i dimensions of segregation (SEG00i) with variation in a matrix of the seven land-use factors measured in 1990
([LANDUSE90]), controlling for the 1990 level of segregation and other changes that
might influence segregation in this metro that are unrelated to the land-use patterns extant
in 1990:
SEG00i = c + αSEG90i + [β][LANDUSE90] +
[θ][(LANDUSE90)2] + [Φ][∆CONTROLS90-00] + ε
(1)
where ε is an error term with usual assumed properties.
Note that although the 2000 level of segregation is modeled as the dependent variable,
because the 1990 segregation level is being controlled [β] can be interpreted as the relationship between 1990 land-use patterns and the change in segregation from 1990 to
2000. Our regression model, therefore, can be interpreted as investigating the following
question:
To what degree is the1990 pattern of metropolitan land use related to 1990–2000
changes in dimensions of Black–White segregation, controlling for decadal
changes in other factors that might affect changes in segregation?
This model meets the challenges of causality and metropolitan idiosyncrasies as follows. Specification of the lag between land-use patterns and segregation measured a
decade later ensures, in conjunction with the lagged value of segregation, that causality is
being measured in the intended way. It must be recognized, however, that the specification
536
GALSTER AND CUTSINGER
in [1] probably understates the longer-term influence of land-use patterns on segregation,
because it controls for SEG90i, which is itself probably shaped by earlier stages of landuse patterns. Metro-specific idiosyncrasies affecting segregation are summarily controlled by employing the lagged value of segregation as an explanatory variable. That is,
each metro’s peculiar historical patterns that have shaped its level of segregation should
already be reflected in SEG90i.
Land-use patterns enter the model in quadratic form in [1]. Following from our prior
theoretical discussion of their potentially countervailing (but nonlinear) influences on the
factors affecting segregation, we would expect [β] and [θ] to have opposite signs,
although theory provides no guidance about which set would have positive signs. Given
this, we employ conservative, two-tailed tests of statistical significance.
As control variables we employ an extensive list of demographic, social, and economic characteristics that have conventionally been employed in previous cross-sectional
models of segregation (Galster, 1987, 1991; Galster and Keeney, 1988; Farley and Frey,
1994; Nelson, Dawkins, and Sanchez, 2004; Nelson, Sanchez, and Dawkins, 2004). Note,
however, that many variables used in prior studies (such as regional dummies) are
rendered superfluous by our specification of the lagged dependent variable. Given the
above interpretation of our model as one explaining changes in segregation, our control
variables are measured as changes from 1990 to 2000. Descriptive statistics for these
variables are presented in Table 4.
In sum, our specification examines cross-sectional variation in Black–White segregation changes from 1990 to 2000 that are associated with coincidental changes in a variety
of economic and demographic changes that play themselves out within varied land-use
contexts extant in 1990.
RESULTS
Inter-Regional Variations in Land-Use Patterns and Changes in Segregation
Table 5 provides a descriptive overview of the interregional variation in our key variables of interest. While most of the interregional differences in our 1990 summary landuse indices are modest, there are some noteworthy patterns. The Midwest region in our
sample, on average, scores lowest on the job distribution, mixed-use, and housing
unevenness factors; the Northeast scores lowest on density/continuity and proximity; the
South scores lowest on housing centrality and nuclearity. Recalling that lower values of
our indices indicate more sprawl-like patterns, the foregoing suggests that it would be
unfair to give any region the blanket label of “most sprawled,” independent of the particular factor being considered. Indeed, these indices cluster in ways that delineate four
metropolitan land-use typologies, each of which manifests a distinctive set of sprawl-like
dimensions (Cutsinger and Galster, 2006).
Changes in Black–White segregation observed from 1990 to 2000 for three indices
are, on average, consistent across all regions in our sample; dissimilarity (D), isolation
(Std. I*), and concentration (DEL) declined in all regions. The West evinces higher-thanaverage declines in all five dimensions of segregation. The only exceptions to desegregation were in the South, where Blacks became slightly more centralized than Whites, and
537
RACIAL SETTLEMENT AND LAND-USE PATTERNS
TABLE 4. DESCRIPTIVE STATISTICS FOR CONTROL VARIABLES
Definitiona
Mean
Std. Dev.
Max Pop
Decades since the metropolitan area reached its maximum
population, figured from 2000.
2.70
2.367
50K Pop
Decades since the metropolitan area reached a population
of 50,000, figured from 2000.
0.70
3.412
Metro area scale
The average distance (in degrees) between centroids of
square-mile cells comprising the EUA.
0.364
0.229
Control variables
Change Minority
Change in the proportion of all minority group members
0.053
0.029
Change Foreign Born
Change in the proportion of the number of foreign born
(naturalized or not a citizen) people
0.032
0.024
Change Rich
Change in the proportion of the number of households
with yearly income over $150,000
0.032
0.018
Change White College Change in the proportion of White people over the age of
25 with a college educationb
0.051
0.016
Change White 25–54
Change in the proportion of White population aged 25 to
54 years
0.006
0.010
Change Pop
Proportional change in the total population
0.210
0.204
Change Prop Pop X
Change in the proportion of the population that is Black
0.011
0.011
a
All changes refer to 1990 to 2000 in the metropolitan area.
Includes non-Hispanic White people with any college education, regardless of whether they earned a
degree.Source: U.S. Census of Population and Housing, 1990 and 2000 (U.S. Bureau of the Census, 1991,
2001; www.AmericanFactfinder.gov).
b
in the Northeast and Midwest, where Blacks clustered more and the same, respectively
(see the bottom panels of Table 5).
Regression Findings
When the model embodied in equation 2 is estimated, a variety of interesting relationships between changes in control variables and changes in segregation emerge.31 Fastergrowing metros exhibited lower dissimilarity and isolation of Blacks. Larger metros and
those that reached their peak populations longer ago were associated with increases in
Black isolation during the 1990s. Just the opposite occurred for areas with growing shares
of minority populations and Whites between the ages of 25 and 54. Metros with greater
expansions of their high-income populations witnessed steeper declines in Black concentration. Finally, growing shares of college-educated Whites in a metro were associated
with less dissimilarity of Black–White locations but greater relative Black centralization.
Here the main focus, however, is on the relationship between land-use patterns in a
metropolitan area and subsequent changes in its Black–White segregation. As predicted
31
The complete set of results for control variables in the regressions are available from the authors.
538
GALSTER AND CUTSINGER
TABLE 5. LAND-USE FACTOR MEANS AND MEAN CHANGES
IN BLACK–WHITE SEGREGATION, BY REGION
Full sample
Northeast
Midwest
South
West
Land-use factors, 1990
Density/Continuity
1.746
1.074
1.919
1.802
2.131
Housing Proximity
3.072
2.721
3.139
3.198
3.164
Job Distribution
1.474
1.675
1.118
1.646
1.388
Mixed Use
2.254
2.941
1.725
1.946
2.520
Housing Centrality
2.263
2.536
2.220
1.975
2.436
Nuclearity
2.111
2.693
2.052
1.601
2.314
Housing Unevenness
1.998
2.429
1.642
1.959
1.983
–3.757
–2.855
–4.264
–2.406
–5.919
–6.201
–3.720
–4.690
–4.780
–11.750
–0.004
–0.008
–0.012
0.007
–0.008
–0.021
–0.021
–0.030
–0.014
–0.021
–0.006
0.025
0.000
–0.029
–0.011
Change in segregation dimensions, 1990–2000
Dissimilarity (D)
Black–White
Standardized Isolation (Std. I*)
Black
Relative Centralization Index (RCE)
Black–White
Concentration (DEL)
Black
Absolute Clustering Index (ACL)
Black
by our theory, our initial trial regressions found that simple, linear models poorly fit relationships between measures of Black–White segregation and land-use measures. Of
course, this could mean that these two sets of variables indeed have no relationship, that
the underlying relationships cancel each other out, or that the relationships are nonlinear
in ways that are poorly captured by the simple, linear specification. In order to explore the
latter possibility we experimented with several sorts of nonlinear specifications. We
found in the case of Black isolation that the quadratic specification worked best. In the
case of Black centralization, concentration, and clustering, a threshold-spline specification (where the slope of the relationship is zero until a critical value of the land-use measure is exceeded) worked best.32 We found no models where land-use measures related
to Black–White dissimilarity in a statistically significant way. In Table 6 we present standardized beta coefficients and associated t-statistics of the relationships that proved
statistically significant.33
32
To be more precise, the land-use measure M spline was defined as follows: M = 0 if M<X; M = (M–X) if
M≥X, where X = threshold.
33
All results come from regressions using the full set of control variables shown in Table 4 and the lagged value
of the dependent variable; full results are available from the authors.
539
RACIAL SETTLEMENT AND LAND-USE PATTERNS
TABLE 6. STATISTICALLY SIGNIFICANT NONLINEAR REGRESSION RESULTS FOR
BLACK–WHITE SEGREGATION AND METROPOLITAN LAND-USE PATTERNSa
Segregation dimensions
Land-use factors
Density/Continuity
Density/Continuity squared
Housing Proximity
Housing Proximity squared
Job Distribution
Job Distribution squared
Mixed Use
Mixed Use squared
Housing Centrality
Housing Centrality squared
Nuclearity
Nuclearity squared
Housing Unevenness
Housing Unevenness
squared
D
Std. I*
RCE
beta
t
beta
t
–0.106b
(–2.928)***
NA
beta
t
beta
t
0.077b
(2.257)**
NA
beta
t
beta
t
DEL
ACL
0.123e
(2.181)**
NA
0.097b
(2.391)**
NA
–0.455
(–2.247)**
1.072
(2.078)**
beta
t
beta
t
beta
t
beta
t
beta
t
beta
t
0.064c
(1.764)*
NA
beta
t
beta
.065d
(1.829)*
NA
t
Adjusted R-square
ANOVA F
Significance
a
.936
24.797
.000
.897
22.927
.000
.960
69.816
.000
.947
52.071
.000
All regressions include controls listed in Table 4 and the 1990 value of the dependent variable.
Spline breaks at 1.5.
c
Spline breaks at 2.11.
d
Spline breaks at 2.
e
Spline breaks at 1.75.
*p < .10, **p < .05, ***p < .01 (two-tailed tests).
b
.918
33.109
.000
540
GALSTER AND CUTSINGER
Fig. 2. Estimated relationship between housing proximity and changes in Black concentration, compared
to baseline. Note: Baseline is metro area with least compact metropolitan form on this dimension.
A few overarching observations are in order at this point. First, only some aspects of
land use seem to be related (in either linear or nonlinear ways) to some aspects of segregation. Second, the dominant relationship observed is that, on several measures, more
sprawl-like land-use patterns are associated with less Black–White segregation (or more
desegregation), although there are two notable exceptions. Third, there appears to be no
single measure of land-use patterns that explains a wide variety of dimensions of segregation. Only density-continuity and housing proximity are strong predictors for even two
dimensions of segregation; three other land-use factors predict only a single dimension of
segregation, and two predict none. Fourth, although the relationships between land-use
patterns and segregation are thus particularized, they are nontrivial in magnitude in these
particular cases, as we shall discuss below. Finally, relative Black centralization is the
dimension of segregation that is most closely related to a wide range of land-use factors.
Now we turn to a discussion of the four dimensions of segregation and their relationships
to land-use patterns.
Consider first the estimated relationship between housing proximity and Black concentration that is portrayed in Figure 2. This figure, and all forthcoming ones that graphically
portray our regression results, shows the decadal changes in the given segregation dimension associated with within-sample variations in the 1990 level of the given land-use factor.34 Increasing values on the horizontal axis are associated with a more compact (less
sprawl-like) metropolitan form; increasing values on the vertical axis are associated with
growing segregation during the 1990s. For extreme sprawl-like settings in which dwellings are widely scattered compared to the metro land area, there appears to be no relationship with Black concentration. Only if the degree of proximity exceeds a threshold of 1.5
34
In all figures, there are sufficient numbers of observations in all ranges of the horizontal axis.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
541
Fig. 3. Estimated relationship between density-continuity and changes in Black clustering, compared to
baseline. Note: Baseline is metro area with least compact metropolitan form on this dimension.
(half the mean; cf. Table 3) does a direct relationship start to emerge. It shows that a metro
with housing proximity two standard deviations above the mean would be predicted to
have a 2.45 times greater increase in Black concentration than a metro with proximity
below the threshold. This difference represents 2.9% of the 1990 mean level of Black concentration and 118% of the observed mean change in this dimension during the 1990s.
The estimated relationship between metropolitan density-continuity and the clustering
of Blacks in contiguous neighborhoods is shown in Figure 3. As in the previous case, the
relationship is characterized by a threshold, but this one appears at the mean value of
density-continuity. Beyond this threshold, the relationship becomes even more strongly
direct than in the previous case, with a metro with housing proximity two standard deviations above the mean predicted to have a 5.6 times greater increase in Black clustering
during the 1990s than a metro with proximity below the threshold. This difference represents 20.6% of the 1990 mean level of Black clustering and 888% of the observed mean
change in this dimension during the 1990s!
Next consider a more nuanced set of findings relating four different land-use factors to
changes in relative Black centralization (Fig. 4). All these relationships also seem best
captured as threshold splines, with modest variations in the parameter specifying the
threshold. Three of the relationships echo those that have been discussed previously: once
housing proximity, housing unevenness, or nuclearity surpass their respective thresholds,
greater degrees of compact land-use patterns are associated with increasing levels of relative Black centralization. As before, the magnitude of these relationships are sizeable:
compared to metros below the threshold, those having values two standard deviations
above their respective means evince higher RCE scores in the range of four to six points,
representing 11.8% to 17.7% of the 1990 mean level and 10 to 15 times the magnitude of
542
GALSTER AND CUTSINGER
Fig. 4. Estimated relationship between various land-use patterns and changes in Black relative centralization, compared to baseline. Note: Baseline is metro area with least compact metropolitan form on this
dimension.
the observed mean change in RCE from 1990–2000. Of special interest, however, is the
distinctly different threshold relationship evinced for density-continuity: once exceeding
the threshold of 1.5, metropolitan areas with higher density and continuity of the residential landscape are associated with less relative Black centralization (i.e., evinced greater
declines in RCE during the 1990s). Compared to metros below the threshold, those having density-continuity values two standard deviations above the mean evince 6.3-point
lower RCE scores, representing 18.5% of the 1990 mean level and almost 16 times the
magnitude of the observed mean change in RCE from 1990–2000.
Finally, we consider the most complex finding of all: the estimated quadratic relationship between job distribution and Black isolation, which evinces a U-shaped plot as portrayed in Figure 5. Figure 5 shows that metros with job distributions ranging between the
mean and values two standard deviations above the mean had substantially greater
declines in Black isolation, everything else being equal. Compared to the metropolitan
area with the least-compact distribution of jobs (at the origin of Fig. 2), one at the minimum of the function (job distribution = 2.75) would be predicted to have a 7.5% greater
decline in the standardized index of Black isolation. This magnitude represents 9.1% of
the 1990 mean and 121% of the actual mean change in isolation observed (cf. Tables 2
and 5). However, metros that had the most compact job distributions had only slightly
different changes in Black isolation than those in the least compact metros. We discuss
possible reasons for these intriguing findings below.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
543
Fig. 5. Estimated relationship between job distribution and changes in Black standardized isolation, compared to baseline. Note: Baseline is metro area with least compact metropolitan form on this dimension.
DISCUSSION
Segregation and the (Sometimes Countervailing) Impacts of Land-Use Patterns
We believe that the totality of the previous results suggest unambiguously the dominance of the land/housing price effect as the primary mechanism through which land-use
patterns influence Black–White residential segregation. To the extent that housing is
developed in a dense, continuous, concentrated, and proximate pattern, the average price
per residential acre of developed land will be higher, thereby magnifying the interracial
affordability gap and intensifying segregation (or equally, limiting desegregation) along
several dimensions.
However, we have also found evidence that suggests the presence of a countervailing
causal mechanism: propinquity that reduces interracial prejudice. A denser residential
settlement pattern proved to be associated with larger reductions in the relative centralization of Blacks’ residential areas, perhaps because Blacks felt more comfortable moving
farther from the core and/or Whites felt more comfortable moving closer to the core. In
this dimension of segregation, the apparent propinquity effects dominate the land-price
effects.
Two additional causal mechanisms must be brought to bear to adequately explain the
quadratic relationship between job compactness and Black isolation. In metropolitan
areas where jobs are widely dispersed (especially if they are in many far-flung subcenters), previous research has found that there will be (1) lower land prices overall
(Kahn, 2001); (2) greater spatial mismatch between jobs and traditionally minorityoccupied neighborhoods (Stoll, 2004); and (3) greater dissimilarity of workplace locations between White and minority employees. By comparison, as jobs are distributed
somewhat more compactly and closer to the core, less mismatch will narrow interracial
income differentials (and housing affordability gaps thereby) and greater interracial commonality of workplace destinations will encourage both Blacks and Whites to bid for the
544
GALSTER AND CUTSINGER
same (more centralized) neighborhoods. Apparently, these forces trump the land/housing
price effect, resulting in a net desegregation situation where Black isolation declines
more. As jobs become extremely concentrated at the core (with no subcenters), however,
the land/housing price effect apparently overwhelms the prior two, yielding roughly the
same net effect on isolation changes as in the case of widely dispersed jobs.
In sum, the evidence clearly shows that the relationship between metropolitan landuse patterns and Black–White segregation can assume a variety of nonlinear forms,
depending on the particular aspect of land use and segregation in question. We would
argue that these findings can only be explained if different dimensions of land-use
patterns generate at least four sorts of impacts (on land/housing prices, inter-group
propinquity, commuting destinations, and job-housing spatial mismatch for Blacks), each
of which affect desegregation in contrary ways. Which force(s) dominate depends on how
extreme a particular land-use pattern has become.
Land-Use Controls, Urban Growth Policy, and Segregation
It follows from our analysis that developing appropriate, effective, and socially beneficial public policy responses to undesirable metropolitan development patterns requires
a careful unpacking of the often (though ambiguously) used concept “sprawl” (Galster et
al., 2001). In establishing the goals of any proposed intervention in a particular metropolitan land market, one must move beyond the vague rhetoric of “fighting sprawl” to targeting the specific dimension(s) of land use that are putatively generating the undesirable
outcome(s).
Moreover, the findings of this research imply that an additional layer of specificity and
sensitivity to unintended outcomes be applied by land-use planners and policymakers. No
longer should it be uncritically assumed that efforts to reduce land-use patterns typically
associated with sprawl will automatically produce conditions more favorable for racial
desegregation. Instead, we suggest that planners and policymakers should consider the
degree to which policy-induced changes in the targeted dimensions of land use will affect
segregation (perhaps in an undesirable fashion).
For example, if one were interested in how a potential land use or growth control
policy might affect trends in Black–White segregation in a given area, the present study
suggests that changes in either the mix of residential and nonresidential uses at the scale
of one square mile or the centrality of housing development are not likely to have any
impact. However, on most other land-use dimensions the policymaker would need to
position her/his metropolitan area before assessing prospective desegregation consequences of a proposed action. For example, if the metro in question were currently at an
extremely low (“sprawl-like”) value of the housing dimension designated for change, a
policy that only modestly nudged that dimension toward a slightly more compact form
would be unlikely to have much impact on segregation, insofar as it would be unlikely to
exceed the threshold. By contrast, metros in the broad middle range of housing development patterns are likely to create substantially more Black–White segregation to the
degree that they successfully move to a less sprawl-like land-use pattern. In the case of
policies affecting job locations, the considerations are also complex. For metros with
highly dispersed job locations, policies that succeeded in transforming the metropolis
toward the middle range of job distribution values would be expected to reap substantial
RACIAL SETTLEMENT AND LAND-USE PATTERNS
545
desegregation payoffs (whether intended or not). However, if the metro in question was
currently in the middle range, it is likely that any further attempts to encourage job subcenters would have the side-effect of stultifying desegregation. All this implies yet a
deeper level of complication: many metros in our sample evince quite varied positions on
the distribution, depending on which land-use factor is in question.
This discussion holds two sobering implications. First, it is possible that some metropolitan areas would inadvertently intensify Black–White segregation were they to more
aggressively fight sprawl in particular ways, given their current land-use patterns.
Second, the anti-sprawl policy instrument being considered may need to avoid altering
unintentionally another dimension of land use besides that being targeted, thereby skirting unintended consequences of exacerbating segregation. This implies that land use and
growth-control policy should be more attuned to treatment by a scalpel rather than a blunt
instrument. That is, policies and programs should be formulated with sensitivity to avoiding potential unintended consequences regarding racial segregation or, more optimistically, to capitalizing on opportunities for reducing both sprawl and segregation
simultaneously, such as instituting requirements of inclusionary zoning for appropriately
sited multi-family housing complexes.
CONCLUSION
In this article, we have investigated theoretically and empirically the relationship
between Black–White segregation and metropolitan land-use patterns for a representative
sample of 50 large U.S. metropolitan areas. We operationalized indices measuring five
dimensions of Black–White segregation and indices measuring seven dimensions of
land-use patterns. To better deduce causality we specified a model of the 1990–2000
change in a given index of segregation as a nonlinear function of each of the seven landuse components in 1990, plus a set of conventional control variables. We found nonlinear
relationships between changes in multiple dimensions of segregation and multiple dimensions of land use, with most demonstrating a direct relationship between more compact
patterns and segregation once a threshold value was exceeded. The magnitude of these
relationships proved substantial.
The patterns of results can be explained holistically by positing that variations in different dimensions of land-use patterns differentially affect land/housing prices, intergroup propinquity, interracial commonality of commuting destinations, and spatial mismatch which, in turn, appear to affect the ability of a metropolitan area to desegregate.
But alterations in certain aspects of land use—density/continuity and job compactness—
apparently spawn a combination of forces that affect desegregation in contrary ways;
which force dominates seemingly depends on how extreme the given land-use pattern
has become. Remarkably, metropolitan areas evincing extremely high or low values on
job compactness demonstrated substantially less desegregation in multiple dimensions
during the 1990s, all else being equal. Thus, although “sprawl” on some dimensions
apparently abets Black–White desegregation, on other dimensions it has the opposite
effect, and for all dimensions the effects are contingent on the initial pattern.
We recognize that our work does not provide a definitive analysis of the topic, but
some caveats are in order. First, we have only investigated changes in segregation during
a single decade, although racial settlement patterns have a long half-life, and thus
546
GALSTER AND CUTSINGER
currently embody the effects of public and private actions occurring decades ago. Second,
we have not analyzed the degree to which the various dimensions of segregation
discussed here are harmful from the perspective of Blacks or the larger society.35 Third,
there may be other aspects of land-use development, such as the connectivity of street
patterns, which may be influential determinants of segregation, yet were unobserved by
us. Finally, we would caution against generalizing findings here to the settlement patterns
of other minority groups.
These caveats notwithstanding, we believe that several conclusions can be drawn from
our work. First, it appears that certain dimensions of land-use patterns have indeed been
an important, though understudied and underappreciated, element in the evolution of
some aspects of Black–White segregation. Second, “sprawl” and “segregation” are terms
that are too ambiguous without further operationalization to guide public policy. And,
third, specific land-use regulations likely have distinct consequences for the development
of future racial settlement patterns, depending on the current context, so planners and
policymakers need to have a heightened sensitivity about the potential for unintended
consequences.
REFERENCES
Alba, R. and Logan, J., 1993, Minority proximity to Whites in suburbs: An individuallevel analysis of segregation. American Journal of Sociology, Vol. 98, 1388–1427.
Alba, R., Logan, J., and Stults, B, 2000, How segregated are middle-class AfricanAmericans? Social Problems, Vol. 47, 543–558.
Allport, G., 1954, The Nature of Prejudice. Reading, MA: Addison-Wesley.
Alonso, W., 1964, Location and Land Use. Cambridge, MA: Harvard University Press.
Anthony, J., 2004, Do state growth management regulations reduce sprawl? Urban
Affairs Review, Vol. 39, 376–397.
Bailey, M., 1959, A note on the economics of residential zoning and urban renewal. Land
Economics, Vol. 35, 288–292.
Banerjee, T. and Verma, N., 2005, Sprawl and segregation: Another side of the Los
Angeles debate. In D. Varady, editor, Desegregating the City. Albany, NY: State
University of New York Press, 200–212.
Bayer, P., McMillan, R., and Rueben, K., 2004, What drives racial segregation? New
evidence using Census microdata. Journal of Urban Economics, Vol. 56, 514–535.
Benfield, F., Raimi, M., and Chen, D., 1999, Once There Were Greenfields. New York,
NY: Natural Resources Defense Council.
Blank, R., 2001, An overview of trends in economic and social well-being, by race. In N.
Smelser, W. J. Wilson, and F. Mitchell, editors, American Becoming, Volume 1.
Washington, DC: National Academies Press, 21–39.
Bobo, L., 1988, Group conflict, prejudice, and the paradox of contemporary racial attitudes. In P. Katz and D. Taylor, editors, Eliminating Racism: Profiles in Controversy.
New York, NY: Plenum Press, 85–116.
35
This is a shortcoming of the segregation consequences literature as a whole. For those who may wish to
explore this further, see Galster (1987, 1991).
RACIAL SETTLEMENT AND LAND-USE PATTERNS
547
Bobo, L., 2001, Racial attitudes and relations at the close of the twentieth century. In N.
Smelser, W. J. Wilson, and F. Mitchell, editors, American Becoming, Volume 1.
Washington, DC: National Academies Press, 262–299.
Branton, R. and Jones, B., 2005, Reexamining racial attitudes: The conditional relationship between diversity and socioeconomic environment. American Journal of Political Science, Vol. 49, 359–372.
Bullard, R., Johnson, G., and Torres, A., 2000, Sprawl City: Race, Politics, and Planning
in Atlanta. Washington, DC: Island Press.
Burchell, R., 1997, Economic and fiscal costs (and benefits) of sprawl. The Urban
Lawyer, Vol. 29, 159–181.
Bureau of Transportation Statistics, 1991, Census Transportation Planning Package.
Washington, DC: U.S. Department of Transportation.
Charles, C. Z., 2003, The dynamics of racial residential segregation. Annual Review of
Sociology, Vol. 29, 167–207.
Ciscel, D., 2001, The economics of urban sprawl: Inefficiency as a core feature of metropolitan growth. Journal of Economic Issues, Vol. 35, 405–413.
Clark, W. A. V., 1986, Residential segregation in American cities: A review and interpretation. Population Research and Policy Review, Vol. 5, 95–127.
Clark, W. A. V., 1988, Understanding residential segregation in American cities: Interpreting the evidence—A reply to Galster. Population Research and Policy Review,
Vol. 7, 113–121.
Clark, W. A. V., 1989, Residential segregation in American cities: Common ground and
differences in interpretation. Population Research and Policy Review, Vol. 8, 193–
197.
Clark, W. A. V., 2007, Race, class and place: Evaluating mobility outcomes for African
Americans. Urban Affairs Review, Vol. 42, 295–314.
Cloutier, N., 1982, Urban residential segregation and Black income. Review of Economics and Statistics, Vol. 64, 282–288.
Courant, P., 1978, Racial prejudice in a search model of the urban housing market.
Journal of Urban Economics, Vol. 5, 329–345.
Courant, P. and Yinger, J., 1977, On models of racial prejudice and urban residential
structure. Journal of Urban Economics, Vol. 4, 272–291.
Cutler, D., Glaeser, E., and Vigdor, J., 1999, The rise and decline of the American ghetto.
Journal of Political Economy, Vol. 107, 455–506.
Cutsinger, J., Galster, G., Wolman, H., Hanson, R., and Towns, D., 2005, Verifying the
multi-dimensional nature of metropolitan land use: Advancing the understanding and
measurement of sprawl. Journal of Urban Affairs, Vol. 27, 235–260.
Cutsinger, J. and Galster, G., 2006, There is no sprawl syndrome: A new typology of
metropolitan land use patterns. Urban Geography, Vol. 23, 228–252.
Darden, J., 1987, Choosing neighbors and neighborhoods. In G. Tobin, editor, Divided
Neighborhoods. Newbury Park, CA: Sage, 15–42.
Dawkins, C., 2004, Recent evidence on the continuing causes of Black–White segregation. Journal of Urban Affairs, Vol. 26, 379–400.
Dawkins, C., 2005, Tiebout choice and residential segregation by race in U.S. metropolitan areas, 1980–2000. Regional Science and Urban Economics, Vol. 35, 734–755.
548
GALSTER AND CUTSINGER
Denton, N. and Massey, D., 1988, Residential segregation of Blacks, Hispanics, and
Asians by socioeconomic status and generation. Social Science Quarterly, Vol. 69,
797–817.
Drier, P., Mollenkopf, J., and Swanstrom, T., 2004, Place Matters, 2nd ed. Lawrence,
KS: University of Kansas Press.
Emerson, M., Kimbro, R., and Yancey, G., 2002, Contact theory extended. Social Science
Quarterly, Vol. 83, 745–761.
Ewing, R., Pendall, R., and Chen, D., 2002, Measuring sprawl and its impact. Washington, DC: Smart Growth America.
Farley, R., Danziger, S., and Holzer, H., 2000, Detroit Divided. New York, NY: Russell
Sage Foundation.
Farley, R. and Frey, W., 1994, Changes in the segregation of Whites from Blacks during
the 1980s: Small steps towards a more integrated society. American Sociological
Review, Vol. 59, 23–45.
Farley, R., Schuman, H., Bianchi, S., Colasanto, D., and Hatchett, S., 1978, Chocolate
city, vanilla suburbs: Will the trend towards racially separate communities continue?
Social Science Research, Vol. 7, 319–344.
Farley, R., Steeh, C., Krysan, M., Jackson, T., and Reeves, K., 1994, Stereotypes and
segregation: Neighborhoods in the Detroit area. American Journal of Sociology, Vol.
100, 750–780.
Freeman, L., 2000, Minority housing segregation: A test of three perspectives. Journal of
Urban Affairs, Vol. 22, 15–35.
Freilich, R. and Peshoff, B., 1997, The social costs of sprawl. The Urban Lawyer, Vol.
29, 183–198.
Frey, W. and Farley, R., 1996, Latino, Asian, and Black segregation in U.S. metropolitan
areas: Are multiethnic metros different? Demography, Vol. 33, 35–50.
Fulton, W., Pendall, R., Nguyen, M., and Harrison, A., 2001, Who sprawls most: How
growth patterns differ across the United States. Washington, DC: Center on Urban
and Metropolitan Policy, Brookings Institution.
Galster, G., 1979, Interracial differences in housing preferences. Regional Science Perspectives, Vol. 9, 1–17.
Galster, G., 1987, Residential segregation and interracial economic disparities: A simultaneous-equations approach. Journal of Urban Economics, Vol. 21, 22–44.
Galster, G., 1988a, Assessing the causes of racial segregation: A methodological critique.
Journal of Urban Affairs, Vol. 10, 395–407.
Galster, G., 1988b, Residential segregation in American cities: A contrary review. Population Research and Policy Review, Vol. 7, 93–112.
Galster, G., 1989, Residential segregation in American cities: A further response to
Clark. Population Research and Policy Review, Vol. 8, 181–192.
Galster, G., 1991, Housing discrimination and urban poverty of African-Americans.
Journal of Housing Research, Vol. 2, 87–122.
Galster, G. and Godfrey, E., 2005, By words and deeds: Racial steering by real estate
agents in the U.S. in 2000. Journal of the American Planning Association, Vol. 71,
251–268.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
549
Galster, G., Hanson, R., Ratcliffe, M., Wolman, H., Coleman, S., and Freihage, J., 2001,
Wrestling sprawl to the ground: Defining and measuring an elusive concept. Housing
Policy Debate, Vol. 12, 681–718.
Galster, G. and Keeney, M., 1988, Race, residence, discrimination, and economic opportunity: Modeling the nexus of urban racial phenomena. Urban Affairs Quarterly, Vol.
24, 87–117.
Galster, G. and Santiago, A., 1995, Puerto Rican segregation in the U.S.: Cause or consequence of economic status. Social Problems, Vol. 42, 361–389.
Glaeser, E., Kahn, M., and Chu, C., 2001, Job Sprawl: Employment Location in U.S.
Metropolitan Areas. Washington, DC: Center on Urban and Metropolitan Policy,
Brookings Institution.
Glaeser, E. and Vigdor, J., 2003, Racial segregation: Promising news. In B. Katz and
R. Lang, editors, Redefining Urban and Suburban America. Washington, DC: Center
on Urban and Metropolitan Policy, Brookings Institution, 215–234.
Glaser, J., 1994, Back to the Black belt: Racial environment and White racial attitudes in
the South. The Journal of Politics, Vol. 56, 21–41.
Glazer, N. and Moynihan, D. P., 1963, Beyond the Melting Pot: The Negroes, Puerto
Ricans, Jews, Italians, and Irish of New York City. Cambridge, MA: MIT Press.
Hawley, A., 1972, Population density and the city. Demography, Vol. 9, 521–529.
Hirsch, A., 1983, Making the Second Ghetto. Cambridge, MA: Cambridge University
Press.
Huie, S. B. and Frisbee, W., 2000, The components of density and the dimensions of
residential segregation. Population Research and Policy Review, Vol. 19, 505–524.
Iceland, J., 2004, Beyond Black and White: Metropolitan residential segregation in multiethnic America. Social Science Research, Vol. 33, 248–271.
Iceland, J., Weinberg, D., and Steinmetz, E., 2002, Racial and Ethnic Residential Segregation in the United States: 1980–2000. Series CENSR-3. Washington, DC: U. S.
Bureau of the Census.
Ihlandfeldt, K. and Scafidi, B., 2001, Black self-segregation as a cause of housing segregation: Evidence from the Multi-City Study of Urban Inequality. Journal of Urban
Economics, Vol. 51, 366–390.
Ihlandfeldt, K. and Scafidi, B., 2002, The neighborhood contact hypothesis: Evidence
from the Multi-City Study of Urban Inequality. Urban Studies, Vol. 39, 619–641.
Ihlanfeldt, K. and Sjoquist, D., 1998, The spatial mismatch hypothesis: A review of
recent studies and their implications for welfare reform. Housing Policy Debate, Vol.
9, 849-892.
Jackman, M. and Crane, M., 1986, Some of my best friends are Black ... Interracial
friendship and Whites’ racial attitudes. Public Opinion Quarterly, Vol. 65, 459–486.
Jackson, K. T., 1985, Crabgrass Frontier: The Suburbanization of the United States.
Oxford, UK, and New York, NY: Oxford University Press.
Johnson, M., 2001, Environmental impacts of urban sprawl: A survey of the literature and
proposed research agenda. Environment and Planning A, Vol. 33, 717–735.
Kahn, M., 2001, Does sprawl reduce the Black/White housing consumption gap? Housing Policy Debate, Vol. 12, 77–86.
Kain, J., 1968, Housing segregation, negro employment, and metropolitan decentralization. Quarterly Journal of Economics, Vol. 82, 175–197.
550
GALSTER AND CUTSINGER
Kern, C., 1981, Racial prejudice and residential segregation: The Yinger model revisited.
Journal of Urban Economics, Vol. 10, 164–172.
Lopez, R. and Hynes, H., 2003, Sprawl in the 1990s: Measurement, distribution and
trends. Urban Affairs Review, Vol. 38, 325–355.
Malpezzi, S., 1999, Estimates of the Measurement and Determinants of Urban Sprawl in
U.S. Metropolitan Areas. Unpublished manuscript, Center for Urban Land Economics
Research, University of Wisconsin, Madison, WI, June 7.
Malpezzi, S. and Guo, W., 2001, Measuring “Sprawl”: Alternative Measures of Urban
Form in U.S. Metropolitan Areas. Unpublished manuscript, Center for Urban Land
Economics Research, University of Wisconsin, Madison, WI, January 15.
Marshall, H. and Jiobu, R., 1975, Residential segregation in U.S. cities: A causal analysis.
Social Forces, Vol. 53, 449–460.
Massey, D. and Denton, N., 1988, Dimensions of segregation. Social Forces, Vol. 67,
281–315.
Massey, D. and Denton, N., 1989, Hypersegregation in U.S. metropolitan areas: Black
and Hispanic segregation along two dimensions. Demography, Vol. 26, 373–393.
Massey, D. and Denton, N., 1993, American Apartheid: Segregation and the Making of
the Underclass. Cambridge, MA: Harvard University Press.
Massey, D., White, M., and Phua, V., 1996, The dimensions of segregation revisited.
Sociological Methods & Research, Vol. 25, 172–206.
McKinney, M. L., 2000, There goes the neighborhood. Forum for Applied Research and
Public Policy, Vol. 15, 23–27.
Mills, E. and Hamilton, B., 1994, Urban Economics, Fifth Edition. New York, NY:
HarperCollins College Publishers.
Muth, R., 1969, Cities and Housing. Chicago, IL: University of Chicago Press.
Nechyba, T. and Walsh, R., 2004, Urban sprawl. Journal of Economic Perspectives, Vol.
18, 177–200.
Nelson, A. C., Dawkins, C., and Sanchez, T., 2004, Urban containment and residential
segregation: A preliminary investigation. Urban Studies, Vol. 41, 423–440.
Nelson, A. C., Sanchez, T., and Dawkins, C., 2004, The effect of urban containment and
mandatory housing elements on racial segregation in U.S. metropolitan areas. Journal
of Urban Affairs, Vol. 26, 339–350.
Orfield, M., 2002, American Metro Politics. Washington, DC: Brookings Institution Press.
Pendall, R., 2000, Local land-use regulation and the chain of exclusion. Journal of the
American Planning Association, Vol. 66, 125–142.
Pendall, R., 2005, Does density exacerbate income segregation? In D.Varady, editor,
Desegregating the City. Albany, NY: State University of New York Press, 175–199.
Pendall, R. and Carruthers, J., 2003, Does density exacerbate income segregation?
Evidence from U.S. metropolitan areas, 1980 to 2000. Housing Policy Debate, Vol.
14, 541–589.
Popkin, S., Galster, G., Temkin, K., Herbig, C., Levy, D., and Richter, E., 2003, Obstacles to desegregating public housing: Lessons learned from implementing eight
consent decrees. Journal of Policy Analysis and Management, Vol. 22, 179–200.
Quigley, J., Raphael, S., and Rosenthal, L., 2004, Local land-use controls and demographic outcomes in a booming economy. Urban Studies, Vol. 41, 389–422.
RACIAL SETTLEMENT AND LAND-USE PATTERNS
551
Roof, W. and Van Valey, T., 1972, Residential segregation and social differentiation in
American urban areas. Social Forces, Vol. 51, 87–91.
Rusk, D., 1993, Cities Without Suburbs. Washington, DC: Woodrow Wilson Center Press.
Sarzynski, A., Wolman, H., Galster, G., and Hanson, R., 2006, Testing the conventional
wisdom about land use and traffic congestion; The more we sprawl, the less we move?
Urban Studies, Vol. 43, 601–626.
Schelling, T., 1971, Dynamic models of segregation. Journal of Mathematical Sociology,
Vol. 1, 143–186.
Schnare, A., 1976, Racial and ethnic price differentials in an urban housing market.
Urban Studies, Vol. 13, 107–120.
Schuman, H., Steeh, C., and Bobo, L., 1985, Racial Attitudes in America. Cambridge,
MA: Harvard University Press.
Sigelman, L., Bledsoe, T., Welch, S., and Combs, M., 1996, Making contact? Black–
White social interaction in an urban setting. American Journal of Sociology, Vol. 101,
1306–1322.
South, S. and Crowder, K., 1997, Residential mobility between cities and suburbs: Race,
suburbanization, and back-to-the-city moves. Demography, Vol. 34, 525–538.
Squires, G., editor, 2002, Urban Sprawl: Causes, Consequences and Policy Responses.
Washington, DC: Urban Institute Press.
Stoll, M., 2004, Sprawl, Race and Job Proximity: Does Urban Sprawl Further Spatially
Isolate Blacks from Jobs? Unpublished paper, UCLA, June 24.
Taeuber, K. and Taeuber, A., 1965, Negroes in Cities: Residential Segregation and
Neighborhood Changes. New York, NY: Anthenum.
Turner, M. A., Ross, S. L., Galster, G., and Yinger, J., 2002, Discrimination in Metropolitan Housing Markets: Results from Phase I of HDS2000. Washington, DC: U.S.
Department of Housing and Urban Development.
U.S. Bureau of the Census, 1991, U.S. Census of Population and Housing, 1990. Summary
of Population and Housing Characteristics. Washington, DC: U.S. Government Printing Office.
U.S. Bureau of the Census, 2001, U.S. Census of Population and Housing. Washington,
DC: U.S. Government Printing Office.
U.S. Department of Housing and Urban Development, 1978, H.U.D. Survey of the Quality
of Community Life. Washington, DC: U. S. Government Printing Office.
U.S. Geological Survey, 2000, National Land Cover Data Set. Sioux Falls, SD: USGS.
Vandell, K., 1995, How ruthless is mortgage default? A review and synthesis of the
evidence. Journal of Housing Research, Vol. 2, 245–264.
Varady, D., editor, 2005, Desegregating the City. Albany, NY: State University of New
York Press.
Wassmer, R. and Baas, M., 2006, Does a more centralized urban from raise housing
prices? Journal of Policy Analysis and Management, Vol. 25, 439–462.
White, M., 1986, Segregation and diversity: Measures in population distribution. Population Index, Vol. 52, 198–221.
Wolman, H., Galster, G., Hanson, R., Ratcliffe, M., Furdell, K., and Sarzynski, A., 2005,
The fundamental challenge in measuring sprawl: Which land should be considered?
Professional Geographer, Vol. 57, 94–105.
Yinger, J., 1976, Racial prejudice and racial residential segregation in an urban model.
Journal of Urban Economics, Vol. 3, 383–406.
552
GALSTER AND CUTSINGER
APPENDIX A. DESCRIPTIVE STATISTICS OF LAND-USE INDICES
Minimum Maximum
Mean
Std.
Deviation
Index
N
Housing Density (Developable land only)
50
364.81
1,906.98
698.035
288.007
Job Density (Developable)
50
257.08
2,320.49
782.279
371.874
Micro-Continuity (Developable)
50
0.13
0.80
0.346
0.126
Macro-Continuity
50
0.19
0.78
0.512
0.147
Housing Unevenness (Developable)
50
0.36
0.66
0.490
0.045
Job Unevenness (Developable)
50
0.51
0.82
0.626
0.072
Standardized Housing Centrality
50
0.79
2.86
1.194
0.313
Standardized Job Centrality
50
0.92
3.51
1.660
0.491
Standardized Housing Unit Proximity
50
1.05
1.97
1.432
0.164
Standardized Job Proximity
50
1.36
4.26
2.070
0.595
Standardized Housing Unit to Job Proximity
50
1.10
2.34
1.634
0.248
Mixed-Use Job to Housing Units
50
366.74
3,160.14
1,724.732
574.472
Mixed-Use Housing Units to Jobs
50
782.26
4,143.29
1,884.693
692.820
Job Nuclearity (Core/Subcenters)
50
0.29
1.00
0.731
0.182
553
RACIAL SETTLEMENT AND LAND-USE PATTERNS
APPENDIX B. ROTATED COMPONENT MATRIX OF LAND-USE INDICESa
Component
Density/
Job
Mixed- Housing
Housing
continuity Proximity distribution
use
centrality Nuclearity unevenness
Housing Density
(Developable)
.813
–.028
.050
.457
.168
–.032
–.127
Job Density
(Developable)
.865
–.020
–.146
.365
.106
–.036
–.065
Micro-Continuity
(Developable)
.892
–.076
–.027
.109
–.115
.025
–.137
Macro-Continuity
.773
.211
–.407
–.016
–.167
–.144
–.007
Housing Unevenness
(Developable)
–.302
.160
–.037
–.074
.314
.149
.852
Job Unevenness
(Developable)
–.638
–.093
.584
–.116
–.251
–.139
.257
.023
.241
.079
.133
.890
–.131
.269
–.126
.225
.853
.162
.213
.150
–.141
.094
.947
.058
.078
.196
–.040
.088
Standardized Job
Proximity
–.168
.504
.816
–.059
–.087
.088
.070
Standardized Housing
to Job Proximity
–.037
.901
.402
–.002
.073
–.030
.056
Mixed-Use Jobs to
Housing
.179
.004
.168
.941
.063
.081
–.034
Mixed-Use Housing
to Jobs
.331
.079
–.108
.902
.064
.018
–.028
Nuclearity (Core/All
Subcenters)
–.039
–.047
.121
.074
–.097
.969
.098
Standardized Housing
Centrality
Standardized Job
Centrality
Standardized Housing
Proximity
a
Extraction method: principal component analysis. Rotation method: varimax with Kaiser normalization.
Rotation converged in nine iterations.