Content uploaded by John I Carruthers
Author content
All content in this area was uploaded by John I Carruthers on Feb 03, 2014
Content may be subject to copyright.
1 Introduction
Over the last several decades there has been a sustained interest in measuring the
relative costs of alternative forms of development in US metropolitan areas (Burchell,
1998; Frank, 1989; RERC, 1974). Throughout, a major emphasis has been on the
question of whether or not urban sprawl
ö
low-density, discontinuous, suburban-style
development, often characterized as the result of rapid, unplanned, and/or uncoordi-
nated growth (Nelson et al, 1995)
ö
undermines the cost-effective provision of urban
services. This issue is important because, unlike many other criticisms of sprawl, it
provides a practical point of departure for debates over the role that governments
should play in regulating the outcome of urban growth. In particular, the high service
costs allegedly incurred through far-flung development patterns serve as a key source
of leverage for urban planners and others advocating the use of growth management
and `smart growth' programs to promote more compact urban areas (for example, see
Ewing, 1997). But despite claims that land-use regulation is necessary to maintain
efficient service provision, the supporting evidence remains thin and inconclusive.
How does the character of urban development affect the cost of services, and what
does this imply for land-use planning and growth management efforts administered in
the name of economic efficiency?
In this paper we respond to these questions with an exploratory analysis of the
influence that alternative development patterns have on twelve measures of public
expenditure: total direct, capital facilities, roadways, other transportation, sewerage,
trash collection, housing and community development, policy protection, fire protec-
tion, parks, education, and libraries. Our objectives are threefold. First, we provide a
Urban sprawl and the cost of public services
John I Carruthers
Mundy Associates LLC, 1825 Queen Anne Avenue North, Seattle, WA 98109, USA;
e-mail: carruthers@mundyassoc.com
GudmundurFUlfarsson
Daniel J Evans School of Public Affairs, Box 353055, University of Washington, Seattle,
WA 98195-3055, USA; e-mail: gfu@u.washington.edu
Received 6 August 2001; in revised form 28 October 2002
Environment and Planning B: Planning and Design 2003, volume 30, pages 503 ^ 522
Abstract. One of the principle criticisms of urban sprawl is that it undermines the cost-effective
provision of public services. In this paper the authors examine whether or not this is true through
an exploratory analysis of the influence that alternative development patterns have on twelve meas-
ures of public expenditure: total direct, capital facilities, roadways, other transportation, sewerage,
trash collection, housing and community development, police protection, fire protection, parks,
education, and libraries. The objectives of the analysis are threefold. First, the authors, through a
background discussion, provide a brief overview of previous research on the relationship between
urban development patterns and the cost of public services. Second, through empirical analysis, they
examine how the character of urban development affects per capita public outlays in a cross-section
of 283 metropolitan counties during the 1982^ 92 time period. A separate equation is estimated for
each measure of expenditure, providing substantive evidence on how density, the spatial extent of
urbanized land area, property value, and political fragmentation affect the cost of services. Finally, the
authors use the results of the empirical analysis to develop a set of policy recommendations and
directions for future research.
DOI:10.1068/b12847
brief overview of previous research on the relationship between urban form and the
cost of public services. Second, we examine how the character of urban development
affects per capita public outlays in a cross-section of 283 metropolitan counties during the
1982 ^ 92 time period. A separate equation is estimated for each measure of expenditure,
providing substantive evidence on how density, the spatial extent of urbanized land area,
property value, and political fragmentation affect the cost of services. Third, we use the
results of the empirical analysis to develop a set of policy recommendations and
directions for future research. We suggest that growth management programs may be
justified from the standpoint of public finance and that development impact fees have
significant potential for mitigating the fiscal effects of urban sprawl. Future research
should focus on evaluating the cost of services by means of alternative measures of urban
form, on determining whether or not the quality of service provision is affected by the
physical character of development, and on evaluating the relative costs and benefits of
political fragmentation, after taking into account its influence on urban sprawl.
The paper is organized into four sections. In section 2 we review previous research
on the relationship between urban development and the cost of services in US metro-
politan areas. In the section 3 we present the empirical analysis, including the research
hypotheses, modeling framework, and estimation results. In section 4 we provide a
discussion of the modeling results, focusing on policy recommendations and directions
for future research. Finally, in section 5, we conclude the paper with a summary of the
research findings.
2 Background
Widespread interest in evaluating the causes and consequences of alternative develop-
ment patterns emerged in the 1960s, just following the first major postwar boom of
suburban development in the United States. At the time, urban sprawl was a relatively
new phenomenon, so much of the early research focused on defining its key character-
istics and its relationship to newly evolving land markets (Bahl, 1968; Clawson, 1962).
Although generally inconclusive about the costs and benefits of sprawl, these and other
studies collectively characterized it as being composed of low-density, scattered, strip,
and leapfrog development patterns and as being eminently associated with land specu-
lation, suburbanization, and political fragmentation (Burchell, 1998; Downs, 1999).
Since its initial rise, sprawl has come to represent the dominant mode of growth in
most US metropolitan areas and, as a result, it continues to generate extensive debate
over its desirability as a pattern of land use. On the one hand, proponents defend urban
sprawl as a fulfillment of consumer preferences whereas, on the other hand, detractors
fault it for contributing to numerous social and economic problems (Ewing, 1997;
Gordon and Richardson, 1997).
Despite its intensity, this debate has been hampered by a failure on both sides to
distinguish sprawl from general suburbanization and by a lack of criteria for establish-
ing what constitutes an ideal urban form in the first place. Suburbanization often occurs
at high densities
ö
as the experiences of Las Vegas, Los Angeles, Phoenix, and many
other western cities have shown
ö
even though it is still considered sprawl. Meanwhile,
in other instances, low-density suburbanization has produced many communities that
present none of the problems, such as environmental degradation, socioeconomic
segregation, and traffic congestion, that sprawl is commonly faulted for. Because of
this dichotomy, density emerges as only part of the picture and can sometimes provide a
misleading image of urban form. Placed in a broader context, the problem of sprawl in
Los Angeles, for example, may stem more from the city's tremendous land area and
extreme separation of land uses than density alone (Burchell et al, 1998). Even so,
504 J I Carruthers, G F Ulfarsson
density remains the most common measure used to describe urban form because of its
intuitive appeal and the difficulty of obtaining data on alternative measures.
On a more pragmatic level, there remains the question of what constitutes an ideal
urban form. Planning is a normative profession, responsible for shaping cities into what
they `ought' to be, but there are very few rigorous criteria for justifying one outcome
over another (Talen and Ellis, 2002). But urban form matters in meaningful ways
to people who live, work, and/or otherwise spend time in cities, so policies that aim to
shape it should be guided by well-founded theory and have a clear set of objectives
(Carruthers, 2002a). The physical outcome of urban development directly affects the
livability, property values, transportation alternatives, and many other aspects of
the urban environment and therefore is central to the planning process. This is expressed
in one of the fundamental theories of urban form (Lynch, 1981), which suggests that
cities may be evaluated on the basis of five dimensions (vitality, sense, fit, assess, and
control) and two metacriteria (efficiency and justice). Together, these social use values
describe how well a city serves the needs of its populace and promotes its quality of life.
Ultimately, they suggest that an ideal urban form is one that is dynamic and responsive
to the needs of its residents
ö
in short, one that produces net benefits for the public at
large and that may continually be adapted to minimize negative externalities.
Among the most tangible points of departure for evaluating urban form is urban
planners' well-known contention that sprawl undermines the cost-effective provision of
public services (Altshuler and Gome
¨z-Iba
¨n
¬ez, 1993; Kaiser et al, 1995). In particular, it
is argued that, for many services, the cost per unit of development rises as densities
decrease (Kelly, 1993; Knaap and Nelson, 1992; Nelson et al, 1995; Porter, 1997). That is,
low-density, spatially expansive development patterns lead to greater costs because of
the large investments required to extend roadways and other types of infrastructure that
transmit water, sewage, electricity, and other services long distances to reach relatively
fewer numbers of people (Carruthers, 2002a). Urban sprawl may also undermine
economies of scale for other services, including police protection and public education,
by lowering the density of individual consumers. That public goods and services are
priced according to their average as opposed to their marginal cost adds to the problem,
as land developers have little motivation to help maintain a cost-effective urban form.
The location of new development continues to be determined by land speculation and
potential for profit instead of its impact on aggregate public welfare. As an outcome,
growth commonly enjoys significant subsidies, as the costs it imposes end up being
financed through collective property tax revenues (Bruekner, 2000; Lee, 1981).
The logic behind this reasoning is straightforward but the supporting evidence
remains thin, and little is known about the actual relationship between urban form
and the cost of services
ö
if any exists at all. As a practical matter, site planners and
engineers have investigated how alternative development patterns affect the cost of
delivering physical infrastructure, including roads, schools, sewers, and other public
facilities. Although many of these studies find that low-density developments are more
expensive to support, they have produced few generalizable conclusions because of
their site-specific focus and an overall lack of standardized measures of expenditure
(for a thorough review, see Frank, 1989; for a recent analysis of this type, see Speir and
Stephenson, 2002). The most well known of these was undertaken by the Real Estate
Research Corporation (RERC, 1974), publishing its findings under the title The Costs of
Sprawl. RERC attempted to use an `internally consistent' set of estimates for the direct
costs of alternative development patterns to illustrate that urban sprawl was approx-
imately twice as expensive to serve as `high-density planned' development patterns. But
despite its extensive impact, the RERC study has also received significant criticism over
the years for its methodology and failure to control for the influence of factors other
Urban sprawl and the cost of public services 505
than density that affect the cost of service provision (Altshuler and Gome
¨z-Iba
¨n
¬ez, 1993;
Ladd, 1998).
More recently, a series of regression-based analyses conducted by Ladd and Yinger
(1991) and Ladd (1992; 1994) suggests that greater densities are associated with higher,
not lower, public service expenditures. Drawing on cross-sectional data and controlling
for other determinants of spending, Ladd and Yinger found that the cost of services
rises with density, contradicting the findings of earlier site-based analyses. Specifically,
using a `piecewise' regression procedure, Ladd (1992) illustrated that the relationship
may be U-shaped, first declining as density increases but then increasing sharply,
leading to average costs that exceed the minimum by as much as 43% in very dense
counties. The implication is that urban services are subject to economies and disecono-
mies of scale
ö
a finding that is explained in terms of the `harshness' of high-density
areas or, in other words, the increased traffic congestion, crime rates, and other
conditions associated with urban environments (Ladd, 1998). If this is the case, urban
sprawl may not be as costly as planners claim, undermining the rationale for policies
aimed at shaping compact development patterns.
Despite the high quality and methodological rigor of these analyses, other evidence
suggests that further research drawing on cross-sectional data is needed before the
relationship between urban form and service expenditures is fully understood. For
example, in an analysis of the 159 counties forming the 25 largest metropolitan areas
in the United States, Pendall (1999) finds that public indebtedness is associated with
urban sprawl. Although the direction of causation examined is the opposite of that
examined here, the implication is that low-density development patterns require greater
public expenditures to support them than do high-density development patterns. More-
over, in a cross-section of 283 metropolitan counties we have found that density has
a negative influence on the cost of infrastructure, including roadways and sewers
(Carruthers, 2000b; Carruthers and Ulfarsson, 2002). Unlike Ladd's (1992; 1994) expen-
diture model, in which indicator variables were used to partition the dataset categorically
by density, we assumed a linear overall relationship and focused specifically on the
interconnected influences of several characteristics of urban development. The results
suggest that per capita spending on infrastructure declines at greater densities but
increases with the spatial extent of urbanized land area and property values. It may
therefore be the `spread' of a metropolitan area and that relative strength of its property
tax base, rather than its `bulk', that leads to greater per capita service expenditure.
Finally, in addition to these and other characteristics of the built environment,
urban sprawl has also been described in terms of the political structure of metropolitan
regions (Burchell, 1998; Downs, 1994; 1999). In particular, new local governments
and special districts are often formed in order to increase and/or maintain the
quality of service provision in newly urbanizing areas (Foster, 1997; Lewis, 1996).
This process is fundamental to the perpetuation of sprawl because new incorporations
and service districts literally enable suburban development to proceed at the urban
fringe (Carruthers, 2003; Carruthers and Ulfarsson, 2002). In turn, the `fiscal zoning'
and growth-control strategies often employed in these communities work to lower
densities, virtually ensuring that metropolitan areas become more spread out over
time. Even so, the thinking among many suburban residents is that the formation of
small general and special purpose governments helps to secure the highest possible
quality of public services for the lowest possible price. In this way, the process of political
fragmentation compounds questions regarding the relationship between the character of
urban development and service expenditures by simultaneously promoting the physical
dimensions of urban sprawl and seeking to achieve greater cost-effectiveness in service
delivery (Fischel, 1985).
506 J I Carruthers, G F Ulfarsson
In theoretical terms, the role that fragmentation plays in reducing the cost of public
services may be understood through the Tiebout model of metropolitan governance.
Specifically, the model suggests that, assuming a mobile population capable of `voting
with their feet', highly fragmented metropolitan areas should exhibit relatively lower
per capita service expenditures as communities minimize their operating costs in order
to attract and retain residents (Tiebout, 1956). This expectation has generally been
reinforced by empirical research finding that the greater the number of jurisdictions,
the lower their overall expenditure, but little is known about how spending on different
types of services is affected (Dowding et al, 1994). The question is an important one
because it has direct implications for the role of planning and growth management in
many metropolitan regions, especially if the physical characteristics of development
also make a difference.
In sum, the relationship between urban form and public service expenditure
remains ambiguous and controversial. Early research developed a strong foundation
for characterizing sprawl but remains inconclusive about the desirability of sprawl as a
form of land use. Meanwhile, normative theory suggests that an ideal urban form is
one that maximizes social use value by creating net benefits to the public at large and that
may be adapted to minimize negative externalities. Within this context, public service
expenditures represent a tangible point of departure for evaluating the impact of urban
sprawl on aggregate public welfare. Site-based analyses have attempted to address the
issue from a practical standpoint, estimating the infrastructure costs associated with
alternative development patterns, but these have provided only limited insight because
of their overemphasis on density as singular determinant of public spending. Regression-
based analyses have produced conflicting evidence, partly because of methodological
differences but also because of differences in the way the character of urban development
is measured (an issue discussed in section 3.1). Adding to the complexity, the political
structure of fragmented metropolitan regions may also affect service provision by
promoting greater cost-effectiveness, even as it works to create low-density development
patterns. Meanwhile, the alleged costs of sprawl continue to serve as leverage for growth
management and `smart growth' programs aimed at shaping compact, high-density urban
areas in both academic and applied planning forums. Although it is impossible for any
single analysis to account for all the relative costs and benefits of alternative development
patterns, there is a clear need for more detailed testing of how the various physical and
political dimensions of metropolitan areas affect public service expenditure.
3 Empirical analysis
In this section we present an empirical analysis of the relationship between alternative
development patterns and expenditure on public services in a cross-section of 283
metropolitan counties, observed over the 1982 ^92 decade. The process is divided into
three steps. In the first we provide a framework for measuring the physical and political
characteristics of urban development (section 3.1). In the second we describe the data
and specify the empirical model (section 3.2). In the third we deliver the estimation
results for twelve different forms of expenditure: total direct, capital facilities, roadways,
other transportation, sewerage, trash collection, housing and community development,
police protection, fire protection, parks, education, and libraries (section 3.3).
3.1 Measuring the characteristics of urban development
Two key factors have detracted from analysts' ability to estimate accurately the costs of
alternative development patterns in the past: the interconnected influences of different
physical characteristics of urban development, and the difficulty of obtaining appro-
priate measurements. First, most research has been narrowly focused on the question
Urban sprawl and the cost of public services 507
of how density, as a singular measure of urban development, influences public service
expenditure. This approach is problematic because although density may help to create
economies of scale for certain urban services it does not unilaterally describe the
character of urban areas. For example, many services are also subject to economies
of geographic scope, which depend on the spatial extent of the area they provide for
ö
especially where facilities are immobile and the cost of service delivery varies from
location to location (Knaap and Nelson, 1992). Moreover, measurement of the influ-
ence of density in isolation may yield misleading results because dense urban areas also
have high land values and therefore generate greater property taxes (Ewing, 1997). In this
way, density may `pay for itself', obscuring the actual costs that it generates.Where this is
the case, density is likely to be positively correlated with the cost of service delivery,
because of the greater spending through property tax revenues, not because of the
physical form of the development itself.
Second, because of data limitations, analysts have until recently been limited
to measuring the density of specific sites or, in the case of cross-sectional analyses, to
using county land area as the spatial unit. As mentioned in section 2, the site-based
approach has limitations because its findings do not necessarily apply beyond a
localized area. The use of counties as the spatial unit of analysis is even more problem-
atic because their large size obscures actual urban density. Counties are also ineffective
spatial units for measuring changes in density through time
ö
because the spatial
unit remains fixed, any amount of population growth, by definition, leads to greater
density. When measured this way, changes in county density over time more accurately
represent changes in population than in the character of development occurring within.
In this analysis we address these two shortcomings by accounting simultaneously
for the influence of density, the spatial extent of urbanized land area, and property
value. Measurement of these dimensions is made possible by the National Resources
Inventory (NRI), of the US Department of Agriculture (USDA), which records the
number of acres of urbanized land and other major land-use categories at the county
level every five years (USDA, 2002). Using these data: density is measured as the
number of jobs and people per acre of urbanized land; the spatial extent of urbanized
land area in a county is given by the total number of developed acres; and property
value is expressed as the total locally assessed property value per acre of urbanized
land. Employment plus population is used to calculate density, because the amount of
developed land depends both on residential land use and on nonresidential land use.
Also, in the case of property value, it is assumed that assessed land value corresponds
primarily to urban development because the analysis focuses on metropolitan counties
and it is impossible to separate the data according to land use. Together, the three
measures provide a more realistic profile of the character of urban development within
a county than is possible when the land area of the county itself is used. The approach
has the added benefit of allowing the spatial unit to vary over time.
In addition to physical characteristics, in our empirical analysis we control for
the effects of political fragmentation. Urban service expenditure is closely linked to the
underlying political landscape
ö
especially where land-use authority is distributed
among numerous jurisdictions. In order to capture this effect, political structure is
measured in terms of per capita municipalities and per capita special districts, with
higher values corresponding to greater fragmentation. For example, a perfectly frag-
mented metropolitan area, with each person being his or her own mayor, would have a
measure of 1 on an index of per capita municipalities. Note that measurement of
political fragmentation in this way `double counts' people, because total population is
the denominator both in per capita municipalities and per capita special districts. This
is appropriate because it is hypothesized that the ratio of each type of government to
508 J I Carruthers, G F Ulfarsson
the total population affects public service expenditure. In addition to these measures,
because we make use of county-level data (described in section 3.2), an indicator
variable is used to distinguish between central city and suburban counties; the two
types of counties often have very different socioeconomic characteristics, so it is
reasonable to expect that their spending patterns may vary systematically.
Finally, it is important to note the limitations and strengths of the analytical
framework just described. At issue here is the scale at which the analysis is conducted.
On the one hand, these measures
ö
density, the spatial extent, or spread, of urbanized
land, property value, and political fragmentation
ö
are limited in the sense that they
cannot capture the place-to-place variation in urban form that occurs within large
metropolitan areas. It is quite possible, for example, to have `traditional' urban neigh-
borhoods imbedded in an overall pattern of urban sprawl. At a finer scale, urban form
may be described in terms of centralization, concentration, connectivity, grain (which
describes land-use mix), and numerous other measures that deal more directly with
patterns of land use and that are capable of drawing out localized variation (Alberti,
1999; Galster et al, 2001). On the other hand, the measures incorporated in this
analysis capture a great deal of variation in the overall character of metropolitan areas
and provide a detailed basis for testing hypotheses regarding how alternative develop-
ment patterns affect public service expenditures within an interregional framework.
Although none of these variables measures sprawl directly, given previous interpreta-
tions in the planning literature, lower density, larger urbanized land area, greater
fragmentation, and suburban county classification may generally be viewed as lying
at that end of the spectrum.
3.2 Empirical model
In order to investigate the relationship between public service expenditures and the
physical and political dimensions of sprawl, the variables described in section 3.1 are
imbedded within an equation containing additional variables measuring sources of
revenue:
efB,P,R,u,(1)
where expenditure, e, on a given public service is a function of: the characteristics of
the built environment, including density, urbanized land area, and property value (B);
political characteristics, including per capita municipal governments, per capita special
districts, and the indicator variable marking counties that contain a central city (P);
revenue, including local tax and intergovernmental sources (R); and u, a vector of
unobserved effects. The two revenue variables are defined as the total value of locally
assessed taxes
ö
because communities rely on different combinations of property and
sales taxes and tax rates
ö
and state plus federal aid, respectively. These variables were
originally tested in disaggregated form, including separate variables for property tax,
state aid, and federal aid, but the detail added little to the results of the analysis:
property tax alone was not an adequate explanatory variable, and state and federal
aid were often collinear with one another.
In table 1 (see over) we provide a definition of each independent variable and
summarize the effects that each is expected to have on the dependent variables. One
variable that is conspicuously absent from the list is income, which plays a major role
in shaping residents' preferences for public services (Ladd, 1992; 1994). Income was
tested in the model but competed with property value because of multicollinearity
between the two variables
ö
when income was added, property value would become
insignificant and/or reverse its sign, and the density variable was also negatively
affected. For this reason, income was discarded as an explanatory variable.
Urban sprawl and the cost of public services 509
The functional relationship identified in equation (1) is specified as an econometric
model with variables collected for 283 counties located in fourteen states at three points
in time: 1982, 1987, and 1992. The dataset includes all metropolitan counties in Arizona,
California, Colorado, Florida, Georgia, Idaho, Nevada, New Mexico, North Carolina,
Oregon, Tennessee, Texas, Utah, and Washington (1998 Census definition). These states
are similar in the sense that each ranks among the top twenty most rapidly growing
states in the country, but they also capture significant geographic diversity. The wide-
spread growth in the states makes them especially good locations for examining the
relationship between urban development patterns and service expenditure because their
cites are evolving rapidly, producing changes that may be traced through the long-
itudinal structure of the dataset. In figure 1 we show all counties included in the
analysis (shaded dark gray), and in table 2 (see over) we list descriptive statistics and
data sources for each dependent and independent variable used in the analysis.
Two accommodations are made in order to make the best use of the dataset within an
econometric framework. First, because we employ cross-sectional and time-series data
the model is specified by using a fixed-effects estimation method, adding constant terms
for the years 1987 and 1992, and thirteen of the fourteen states
ö
one from each group
(1982 and Texas) is omitted in order to avoid perfect multicollinearity with the overall
intercept. This controls for correlation across locations, because observations from all
locations at a particular time are likely to share unobservable effects, and for correlation
through time, because observations from a particular location are likely to share unob-
servable effects. The fixed-effects approach also helps us to minimize any omitted variable
bias that may affect the parameter estimates. Ideally, location-specific fixed effects would
be added for each county in the dataset, but this would require the addition of too many
additional constant terms (282 instead of 13). The state-specific fixed effects represent a
good compromise because they capture everything that sets a given state apart from the
rest through time. Second, because the data are based on countywide aggregations, and
because the counties are of different sizes, it is likely that the observations are hetero-
scedastic. Within an ordinary least squares (OLS) regression model, heteroscedasticity
causes the estimates of the coefficients to become inefficient; although they remain
unbiased and consistent, the usual estimate of the variance ^ covariance matrix becomes
Tabl e 1. Expected influence of the independent variables on public expenditure.
Variable Expected Variable definition
effect
Built environment
density ÿAverage number of people plus jobs per acre
of urbanized land
urbanized land Total number of acres of urbanized land
property value Average locally assessed property value per
acre of urbanized land (US$1000 per acre)
Political characteristics
per capita municipal ÿNumber of municipal governments
governments (thousands) headquartered in county, divided by population
per capita special ÿNumber of special districts headquartered in
districts (thousands) county, divided by population
central city indicator 1 if the county contains a central city; 0 if not
Revenue
per capita local tax revenue Total value of locally assessed tax dollars
within county, per person
per capita intergovernmental Total value of state plus federal aid (US$)
revenue received by general purpose governments within
county, per person
510 J I Carruthers, G F Ulfarsson
biased, thereby making it difficult to make statistical inferences about the coefficients. In
order to avoid this problem, the model is estimated using White's heteroscedasticity-
consistent estimator for the variance ^covariance matrix. For a discussion of fixed effects
and White's heteroscedasticity-consistent estimator, see Kennedy (1998).
As little is known about the exact nature (shape) of the relationship between urban
development patterns and service expenditure, extensive sensitivity testing was con-
ducted in order to achieve the best possible fit between the dependent and independent
variables. As Ladd (1992) found, we also found that neither linear nor log ^ linear
forms are appropriate; instead, a semilog form was adopted by taking the log of the
dependent variable only. This allows the function itself to be nonlinear but still to
preserve the linear-in-parameters assumption necessary to estimate an equation properly
by using OLS. Because of the semilog form, the estimated coefficients were interpreted
as percentages
ö
that is, a unit change in the independent variable produces a percent-
age change in the dependent variable (Kennedy, 1998). The result is an econometric
specification of equation (1), with the following functional form:
ln e
it
al
j
t
t
bx
it
e
it
, (2)
where iranges over all counties; tranges over the three time periods (1982, 1987, and
1992); jranges over the fourteen states; arepresents the overall constant; lrepresents
the locational fixed effects; trepresents the temporal fixed effects; brepresents a vector
of estimable coefficients; xrepresents the vector of independent variables given by
equation (1); and erepresents the stochastic error term. As noted above, two of the
fixed effects are restricted in each equation (t
1982
0, and l
Texas
0) in order to
avoid perfect multicollinearity with the overall intercept (a).
0 270 540 810 1080 km
N
Figure 1. Counties included in the empirical analysis.
Urban sprawl and the cost of public services 511
512 J I Carruthers, G F Ulfarsson N:/pdf-prep/
Tabl e 2 . Descriptive statistic and data sources for the dependent and independent variables.
Mean Median Standard Data sources
deviation
Dependent variables USCG, various years c; REIS, various years
per capita total expenditure (US $) 1 463.31 1 348.74 984.48
per capita spending (US$) on
capital facilities 188.56 144.56 258.66
roadways 58.63 56.31 31.90
other transportation 16.46 3.99 38.53
sewerage 34.08 24.70 36.70
trash collection 19.61 16.79 17.12
housing 22.19 16.60 23.61
police protection 59.93 55.82 27.65
fire protection 29.03 26.98 21.12
parks 26.22 19.62 24.27
education 547.05 542.40 173.17
libraries 7.63 6.11 6.51
Built environment
density (number of jobs and people per acre) 4.80 3.90 4.41 USDA, 2002; REIS, various years
urbanized land (acres) 68 677.64 45 900.00 96 784.27 USDA, 2002; REIS, various years
property value (US $ thousands per acre) 67 258.69 42 003.99 96 784.27 USDA, 2002; USCG, various years b
Political characteristics
per capita municipal governments (thousands), tÿ5 0.07 0.04 0.09 USCG, various years a; REIS, various years
per capita special districts (thousands), tÿ5 0.12 0.08 0.13 USCG, various years a; REIS, various years
Revenue
per capita local tax revenue (US$) 494.99 454.00 213.01 USCG, various years b; REIS, various years
per capita intergovernmental revenue (US$) 408.88 376.23 195.21 USCG, various years b; REIS, various years
Note: tÿ5, previous period.
3.3 Estimation re sults
By using data from the Compendium of Governmental Finances of the US Census of
Government (USCG, various years c), we estimated twelve separate models for per
capita spending on public services: total direct, capital facilities, roadways, other trans-
portation, sewerage, trash collection, housing and community development, police
protection, fire protection, parks, education, and libraries. In table 3 we provide a
description of each dependent variable, as defined by the census survey form used to
collect the data; the first two variables
ö
per capita total direct expenditure and per
capita spending on capital facilities
ö
are aggregate measures, extending over all indi-
vidual types of services. The results for each model are presented in table 4 (see over),
showing the OLS estimates and t-statistics for all independent variables. The number of
included observations varies slightly across equations because observations where the
dependent variable was equal to zero were dropped. Because of the exploratory nature
of the analysis, greater emphasis is placed on the hypothesis tests (the t-statistics) than
the coefficients; although the coefficients are useful for judging the relative magnitude
of the influence of a significant variable, they should not be interpreted literally.
The OLS estimates provide strong support for the hypothesis that public service
expenditure is closely linked to the physical and political structure of metropolitan
areas. First, the parameter estimates for density are negative and significant in several
of the models, suggesting that it creates economies of scale for: public spending on the
whole (total direct expenditure), capital facilities, roadways, police protection, and
education. For each of these services, the per capita cost decreases as densities increase,
Tabl e 3. Description of dependent variables (source: USCG, 2000).
Variable Description
Per capita total direct
expenditure
Sum of direct expenditure, including salaries and wages
Per capita spending
on capital facilities
Sum of capital outlays, including new construction, the purchase
of equipment, and outlays on land and existing structures
Per capita spending
on roadways
Expenditure on the construction and maintenance of municipal
streets, sidewalks, bridges and toll facilities, and street lighting, on
snow removal, and on highway engineering, control, and safety
Per capita spending
on other transportation
Expenditure on municipal airports, parking facilities, and sea and
inland port facilities and subsidies to private transit facilities
Per capita spending
on sewerage
Expenditure for the construction, maintenance, and operation
of sanitary and storm sewer systems and sewage disposal plants
Per capita spending
on trash collection
Expenditure on street cleaning and the collection and disposal
of garbage
Per capita spending on
housing and community
development
Expenditure on urban renewal, slum clearance, and housing
projects
Per capita spending
on police protection
Expenditure on municipal police agencies, including coroners,
medical examiners, vehicular inspection activities, and traffic
control and safety activities
Per capita spending
on fire protection
Expenditure incurred for firefighting and fire prevention,
including contributions to volunteer fire units
Per capita spending
on parks
Expenditure on parks and recreation, including playgrounds, golf
courses, swimming pools, museums, marinas, community music,
drama, celebrations, zoos, and other cultural activities
Per capita spending
on education
Expenditure on local schools
Per capital spending
on libraries
Expenditure on municipal and nongovernmental libraries
Urban sprawl and the cost of public services 513
Tabl e 4 . Ordinary least squares estimates for expenditure equations: aggregate expenditure (total direct and on capital facilities) and expenditure on roadway,
other transportation, sewerage, trash collection, housing and community development (`housing'), police protection, fire protection, parks and recreation,
education and libraries.
Total direct Capital facilities Roadways Other transportation
coefficient tcoefficient tcoefficient tcoefficient t
Constant 6.07* 82.27 3.69* 27.75 3.29* 32.43 ÿ1.82* ÿ5.48
Built environment
density ÿ0.03* ÿ2.52 ÿ0.03* ÿ1.86 ÿ0.06* ÿ7.09 0.10* 3.41
urbanized land 2.8410
ÿ7
* 1.55 6.0310
ÿ7
* 1.82 3.1210
ÿ7
* 1.30 2.6610
ÿ6
* 3.62
property value 8.6310
ÿ7
* 1.80 1.2310
ÿ6
* 1.61 1.6310
ÿ6
* 3.83 ÿ3.8010
ÿ6
*ÿ3.32
Political characteristics
per capita municipal governments ÿ0.47* ÿ3.27 ÿ0.51* ÿ1.54 0.62* 3.72 ÿ0.90 ÿ0.82
per capita special districts ÿ0.21 ÿ0.85 ÿ0.43 ÿ1.28 ÿ0.06 ÿ0.31 ÿ0.80* ÿ1.40
central city indicator 0.12* 4.78 0.09* 1.83 0.03 0.64 0.91* 6.56
Revenue
per capita local tax revenue 1.2710
ÿ3
* 6.90 1.6710
ÿ3
* 5.70 1.0710
ÿ3
* 4.27 3.9710
ÿ3
* 8.48
per capita intergovernment revenue 1.3010
ÿ3
* 9.26 1.0210
ÿ3
* 5.00 4.5810
ÿ4
* 3.40 2.2110
ÿ3
* 4.00
Temporal effects
1987 ÿ0.02 ÿ0.65 0.08* 1.54 0.07* 1.55 ÿ0.11 ÿ0.70
1992 ÿ0.08* ÿ1.99 0.14* 2.00 ÿ0.04 ÿ0.95 ÿ0.39* ÿ2.10
Locational effects
Arizona ÿ0.04 ÿ0.94 ÿ0.37* 3.62 0.59* 9.02 ÿ1.00* ÿ2.38
California ÿ0.21* ÿ3.36 ÿ0.43* ÿ3.68 0.13* 1.63 ÿ0.46* ÿ1.58
Colorado 0.09 1.12 0.27* 2.10 0.48* 5.60 ÿ0.41* ÿ1.51
Florida 0.10* 2.70 0.27* 3.47 0.23* 4.87 0.75* 4.13
Georgia 0.07* 2.18 0.33* 3.40 ÿ0.02 ÿ0.42 ÿ0.43* ÿ1.71
Idaho 0.01 0.12 0.29* 1.92 0.38* 4.36 1.54* 5.45
North Carolina 3.8710
ÿ3
0.07 ÿ0.35* ÿ3.58 ÿ1.09* ÿ10.31 ÿ0.09 ÿ0.38
New Mexico 0.01 0.08 0.27* 2.05 0.40* 3.27 0.49* 1.51
Nevada 0.10* 1.93 0.56* 2.61 0.64* 5.61 1.48* 2.57
Oregon ÿ0.01 ÿ0.18 ÿ0.20* ÿ1.96 0.15* 1.90 ÿ0.36 ÿ0.99
Tennessee 0.42* 7.55 0.09 0.96 0.44* 8.35 0.34 1.08
Utah 0.25* 3.79 0.38* 2.52 0.11* 1.45 0.40 0.88
Washington 0.14 1.27 0.31* 2.94 0.39* 4.33 1.33* 3.98
Number of observations 849 849 849 684
Adjusted R
2
0.62 0.40 0.61 0.42
514 J I Carruthers, G F Ulfarsson
Urban sprawl and the cost of public services 515
Tabl e 4 (continued).
Sewerage Trash collection `Housing' Police protection
coefficient tcoefficient tcoefficient tcoefficient t
Constant 1.76* 9.40 1.28* 7.07 1.31* 5.16 2.93* 35.11
Built environment
density 0.03 1.12 ÿ0.02 ÿ0.71 0.01 0.31 ÿ0.02* ÿ1.87
urbanized land 5.4210
ÿ7
1.08 7.3010
ÿ7
* 1.89 6.8610
ÿ7
* 1.88 3.2010
ÿ7
* 2.25
property value ÿ1.4610
ÿ6
*ÿ1.34 2.6510
ÿ7
0.24 5.9810
ÿ7
0.64 6.3210
ÿ7
* 1.37
Political characteristics
per capita municipal governments ÿ1.82* ÿ3.11 0.21 0.45 ÿ2.54* ÿ3.21 ÿ0.63* ÿ1.95
per capita special districts ÿ1.24* ÿ2.80 ÿ1.39* ÿ2.69 ÿ1.97* ÿ3.34 ÿ0.46* ÿ1.83
central city indicator 0.18* 2.04 0.39* 4.48 0.52* 5.33 0.12* 4.82
Revenue
per capita local tax revenue 2.2810
ÿ3
* 7.82 1.3610
ÿ3
* 5.66 8.2810
ÿ4
* 2.51 1.4710
ÿ3
* 9.43
per capita intergovernment revenue 1.3810
ÿ3
* 4.42 1.8010
ÿ3
* 5.51 2.0510
ÿ3
* 4.89 7.3310
ÿ4
* 4.97
Temporal effects
1987 ÿ0.07 ÿ0.62 ÿ0.17* ÿ1.97 ÿ0.23* ÿ2.30 1.9110
ÿ3
0.07
1992 ÿ0.12 ÿ1.10 ÿ0.02 ÿ0.26 ÿ0.29* ÿ2.37 ÿ1.5310
ÿ3
ÿ0.03
Locational effects
Arizona ÿ0.17 ÿ0.74 ÿ0.11 ÿ0.83 ÿ0.16 ÿ0.79 0.37* 7.67
California 0.08 0.45 ÿ1.40* ÿ5.53 ÿ0.16 ÿ0.71 0.23* 3.59
Colorado 0.17 1.00 ÿ1.62* ÿ4.04 0.43* 1.99 0.16* 2.37
Florida ÿ0.67* ÿ3.50 0.56* 6.03 ÿ0.12 ÿ0.82 0.38* 10.80
Georgia ÿ0.36* ÿ2.57 0.13* 1.44 0.20 1.29 0.10* 2.66
Idaho 0.81* 3.99 0.93* 5.67 1.25* 3.26 0.49* 5.87
North Carolina ÿ0.61* ÿ3.35 0.37* 4.46 0.30* 1.94 0.07* 1.62
New Mexico ÿ0.27 ÿ0.73 ÿ0.04 ÿ0.20 ÿ0.43* ÿ1.62 0.47* 6.56
Nevada ÿ0.48 ÿ1.16 ÿ2.82* ÿ7.40 0.40* 1.80 0.88* 7.40
Oregon 0.50* 3.26 ÿ1.78* ÿ5.90 0.92* 4.48 0.04 0.80
Tennessee 0.24* 1.61 0.65* 5.41 0.85* 4.69 0.21* 4.18
Utah 0.49* 2.21 0.17 0.83 0.23 0.70 0.28* 1.56
Washington 0.59* 3.89 ÿ0.17 ÿ0.92 0.10 0.54 0.18* 2.97
Number of observations 753 839 849 849
Adjusted R
2
0.34 0.41 0.32 0.70
516 J I Carruthers, G F Ulfarsson
Tabl e 4 (continued).
Fire protection Parks and recreation Education Libraries
coefficient tcoefficient tcoefficient tcoefficient t
Constant 1.58* 8.73 1.13 6.08 5.64* 74.03 0.33* 2.09
Built environment
density ÿ0.02 ÿ0.78 ÿ4.8610
ÿ3
ÿ0.28 ÿ0.04* ÿ4.48 ÿ0.01 ÿ0.49
urbanized land 3.9410
ÿ7
1.16 6.1110
ÿ7
* 1.50 ÿ1.8910
ÿ7
*ÿ1.36 57210
ÿ7
* 1.68
property value 7.2610
ÿ7
0.65 4.2410
ÿ7
0.48 9.6510
ÿ7
* 2.52 32210
ÿ9
5.0010
ÿ3
Political characteristics
per capita municipal governments ÿ2.64* ÿ6.52 ÿ2.10* ÿ3.64 0.03 0.18 ÿ1.61* ÿ4.15
per capita special districts ÿ0.57 ÿ1.24 ÿ1.26* ÿ3.17 ÿ0.34* ÿ1.32 ÿ1.22* ÿ3.43
central city indicator 0.32* 5.96 0.39* 6.48 ÿ0.03* ÿ1.38 0.18* 2.61
Revenue
per capita local tax revenue 2.0710
ÿ3
* 5.71 2.5910
ÿ3
* 6.39 7.9310
ÿ4
* 5.78 21810
ÿ3
* 7.03
per capita intergovernment revenue 1.1510
ÿ3
* 3.87 1.0210
ÿ3
* 3.72 1.2510
ÿ3
* 6.99 6.9710
ÿ4
* 3.17
Temporal effects
1987 ÿ0.07 ÿ0.91 ÿ0.14* ÿ1.86 0.02 0.92 ÿ0.02 ÿ0.29
1992 ÿ0.15* ÿ1.55 ÿ0.28* ÿ2.88 ÿ0.06 ÿ1.19 ÿ0.08 ÿ0.95
Locational effects
Arizona 0.31* 2.29 0.52* 4.23 ÿ0.08* ÿ1.58 0.55* 3.68
California 0.24* 1.58 0.34* 2.05 ÿ0.32* ÿ4.59 0.52* 3.97
Colorado 0.19 1.28 0.90* 5.77 ÿ0.03 ÿ0.53 0.59* 4.27
Florida 0.36* 4.96 0.60* 6.28 ÿ0.10* ÿ3.73 0.31* 3.18
Georgia ÿ0.01 ÿ0.11 0.02 0.18 ÿ0.27* ÿ6.14 ÿ0.57* ÿ4.93
Idaho 1.03* 7.89 0.92* 5.81 ÿ0.08 ÿ0.93 1.13* 8.05
North Carolina 0.14 1.21 0.23* 1.65 ÿ0.12* ÿ4.01 0.46* 3.87
New Mexico 0.39* 2.17 0.81* 3.53 ÿ0.17* ÿ2.63 0.77* 3.82
Nevada 0.95* 4.60 1.59* 8.41 ÿ0.24* ÿ3.11 0.86* 5.52
Oregon 0.74* 6.33 0.32* 2.63 0.12* 1.99 0.64* 5.13
Tennessee 0.46* 3.79 0.26* 1.64 ÿ0.17* ÿ3.34 ÿ0.20* ÿ1.66
Utah 0.85* 3.81 1.52* 5.96 0.14* 2.13 1.22* ÿ7.82
Washington 0.71* 5.91 0.99* 6.21 ÿ0.08* ÿ1.37 1.19* 7.38
Number of observations 846 843 849 827
Adjusted R
2
0.53 0.58 0.55 0.54
* Significant at p<0:10.
Note: For units of measurement, see table 2.
with the greatest savings realized in areas with very high densities. An individual police
officer patrolling a square mile in a dense urban area may provide protection to many
more people than his or her counterpart in a suburban area. Likewise, fewer roads are
needed in high-density areas, and school systems may be operated more efficiently
ö
fewer (though larger) schools and less bussing of pupils are needed, for example. Among
the rest of the models, density is insignificant and/or negative except in two logical
instances: other transportation and sewerage.The positive coefficient in the equation for
other transportation makes sense given the increased need for parking garages, public
transit, and other facilities in high-density areas. Similarly, the positive correlation
between density and sewerage is likely attributable to the use of private septic systems
and lack of stormwater systems in low-density areas. Assuming a one-tailed hypothesis
test in the opposite direction to that specified
ö
positive instead of negative
ö
the density
coefficient in the sewerage equation is significant at an 85% confidence level. Overall,
the models provide good evidence that density works to increase the cost-effectiveness
of public service expenditure.
Second, the spatial extent of urbanized land is positive and significant in most of
the models, indicating that the spread of a metropolitan area plays an important role in
determining public service expenditure. As explained in the background discussion,
urban sprawl requires roadways and sewer systems to be extended over long distances
to reach relatively fewer people. Trash collection and street cleaning activities must
cover larger areas and, similarly, police and fire protection are spread thin, requiring
more patrols and, potentially, more station houses to achieve a given level of service. In
the case of parks and libraries, a greater number of facilities must be built in order for
people throughout the metropolitan area to enjoy equal access. In one instance
ö
education
ö
the urbanized land coefficient is significant and negative, but this effect
is more likely to be a result of the overall size of the urban area than its spatial extent.
The coefficient is positive and insignificant in the housing and community development
equation, indicating that the spatial extent of urban development has little effect on
spending on these services.
Third, property value is significant in five of the twelve equations and positively
correlated with per capita spending for all services except for other transportation and
sewerage. These findings illustrate the balancing effect that property value has in
helping dense urban areas support themselves
ö
and also how an examination of the
influence of density in isolation may provide misleading results. As property values are
generally the greatest in high-density areas, their contributions to public revenue
through property taxes enable density to support itself. In the case of other transporta-
tion and sewerage, the negative coefficient is logical, because parking garages, sewerage
treatment plants, and other locally undesirable facilities are less likely to be built in
areas with high property values.
Fourth, the three political characteristics are significant in most of the equations,
highlighting the role that political fragmentation plays in influencing patterns of public
spending. Specifically, all of the coefficients for per capita municipal governments and
per capita special districts carry negative signs, except in the roadways model, where
municipal fragmentation has a positive influence, and in the trash collection and
education models, where the respective variables are insignificant. These findings
suggest that the formation of small general and special purpose governments may
work to lower per capita spending, although it remains unclear just how this occurs.
Although it is possible that greater efficiencies are achieved through competition
among jurisdictions, some analysts have suggested that fragmentation may create fewer
opportunities for budget maximization and/or may reduce communities' willingness to
provide certain types of collective services (Dowding et al, 1994). In either case, further
Urban sprawl and the cost of public services 517
research is needed in order to develop a better understanding of how fragmentation
affects public service expenditures. Meanwhile, the central city indicator captures signifi-
cant differences between the spending patterns of central city and suburban counties. In
all cases except for education, the parameter estimates indicate that more money is
spent on public services in central cities. This finding is realistic, as central cities
commonly house facilities such as parks and museums that are used by the metropolitan
area at large and are often where infrastructure systems converge. The negative sign on
the education coefficient is interesting because it reinforces the notion that higher
quality school systems are located in suburban areas.
Finally, the remaining control variables
ö
revenue and the temporal and locational
fixed effects
ö
fulfill their expected role within the equations. Being perhaps the most
important determinants of public spending, local tax revenue and intergovernmental
revenue are significant and positive in all equations. The temporal fixed effects are only
occasionally significant, indicating that little time-specific correlation exists among
locations at the times of observation. The locational fixed effects, in contrast, are
mostly significant, revealing important state-to-state differences in per capita spending
patterns. Unfortunately, because the fixed effects capture an amalgamation of unob-
served effects, they have no straightforward interpretation; instead, they highlight the
need for further research aimed at uncovering state-level variables that affect local
governments' spending patterns.
4 Discussion
The results of the empirical analysis (summarized in table 4) illustrate the numerous
ways in which the characteristics of urban development affect public service expendi-
tures. Collectively, they point to two overarching conclusions: (1) the physical pattern
of development has a multidimensional effect, with density, urbanized land area, and
property value all influencing the per capita value spent on service provision; and (2)
one way or another, the political structure of metropolitan areas makes a difference,
with greater fragmentation being associated with lower expenditure. Although the first
of these findings is a well-known argument that is widely accepted among the planning
community (Kaiser et al, 1995) there was little in the way of supporting evidence prior
to this study. What follows are several policy-relevant insights and directions for future
research.
By far the most salient finding of the analysis is that the per capita cost of most
services declines with density (after controlling for property value) and rises with the
spatial extent of urbanized land area. This reinforces planners' claim that urban sprawl
undermines cost-effective service provision, and lends support to growth management
and `smart growth' programs aimed at increasing the density and contiguity of metro-
politan areas
ö
at least from the standpoint of public finance. In particular, the models
show that there are savings to be gained in numerous areas, especially where both the
density and the spread of the metropolitan area matter for the cost of service delivery.
One important exception is sewerage, but further investigation is needed to determine
whether the positive correlation is attributable to the increased cost or increased use of
sanitary and stormwater sewage systems in high-density areas. In other words, the
coefficient may reflect the greater reliance on septic tanks and above-ground storm-
water drainage in low-density areas. The positive influence of the urbanized land area
variable (though not quite significant within acceptable tolerances) suggests that this
may be the case because it indicates that sewerage systems are more expensive when
spread over greater areas. Although this evidence does not unilaterally justify growth
management, it indicates that communities may wish to carefully evaluate whether or
not greater efficiencies could be achieved through their urban form.
518 J I Carruthers, G F Ulfarsson
Empirical research on the effectiveness of state-based growth management programs
suggests that they may help to reduce public expenditures through their influence on
urban form. Specifically, programs that require local governments to produce plans
that are consistent with state-defined goals and objectives and that incorporate urban
growth boundaries (such as in Oregon) have been found to increase urban densities,
which in turn affect the cost of public services. Programs that do not require consis-
tency among jurisdictions' planning activities (such as in Georgia) and/or that rely on
concurrency (such as in Florida) may inadvertently contribute to sprawl, thereby
raising the cost of services (for an analysis demonstrating these results, see Carruthers,
2002b).
As an extension, the strong link between urban form and service expenditures
reinforces the rationale for `market-based' approaches to growth management, such
as the use of development impact fees. As described in section 2, one of the principal
complaints of urban sprawl is that it often ends up being financed by the public-at-
large through average cost pricing mechanisms. Impact fees alter this situation by
shifting some or all of the costs of growth to the private sector, forcing developers to
consider more seriously the costs of alternative development patterns (Altshuler and
Gome
¨z-Iba
¨n
¬ez, 1993). As these costs are eventually passed on to homebuyers
ö
making
new housing more expensive
ö
low-density development patterns may continue to be
accommodated as long as market demand is sufficient to uphold the increase in price.
Ultimately, the effect on the physical pattern of development rests on the elasticity of
demand for low-density growth. Although it is probably unrealistic to assess impact
fees for the ongoing costs of service provision, evidence suggests that it may be
relatively easy to shift the costs of physical infrastructure to the private sector (Speir
and Stephenson, 2002). It may therefore be worthwhile to compel new development to
finance the roads, sewerage, schools, and other infrastructure that it requires. For
example, in the average county in the dataset, capital facilities account for about 13%
of total direct expenditure
ö
a substantial proportion of their overall budgets. Density,
urbanized land area, and property value are all highly significant in the capital facilities
model, providing good evidence in favor of assessing impact fees at least for physical
infrastructure. Even if growth continues to proceed at low densities, the increased price
of housing and other development will strengthen the tax base, raising the amount of
revenue available to support the ongoing costs of operation.
The results of this analysis point to several directions for future research. First,
there is a need for additional work to incorporate alternative measures of urban form
of the sort mentioned in section 3.1. For example, Galster et al (2001) have recently
defined seven distinct dimensions of urban land-use patterns beyond density: centrality,
clustering, concentration, contiguity, nuclearity, mixed use, and proximity. Each of
these has been developed for and tested in thirteen metropolitan areas. Similarly,
Alberti (1999) has emphasized the need to look beyond density and to include measures
of centralization, connectivity, and grain in studies of urban form, especially with
respect to its impact on the environment. Although the development of these types of
measures for multiple metropolitan areas presents a considerable challenge, they hold
much promise for offering further insight into the relationship between urban form and
the cost of public services.
Second, given the potential savings to be gained through more compact urban
development patterns, a major question that remains is whether or not the quality of
service is affected. In this paper we have dealt with intermediate outputs but not the
final outputs eventually consumed by the public. Future research should focus on
evaluating how the character of urban development influences people's enjoyment of
public services
ö
congestion, for example, may overshadow the benefits of reduced cost
Urban sprawl and the cost of public services 519
if it significantly lowers the accessibility of a given service. However, the increased
property values of high-density areas may yield sufficient revenue to maintain a high
enough level of service provision to offset the effects of congestion and/or to provide
specialized forms of services that are unavailable in other areas. These issues are
important because, ultimately, citizen support for growth management programs and
for other policies aimed at shaping more compact development patterns is likely to rest
heavily on how the outcome affects their quality of life.
Finally, the finding that fragmentation is associated with lower per capita spending
suggests that there is a trade-off to be made between the physical and political
structure of metropolitan areas. In particular, a number of studies have shown evi-
dence that fragmentation contributes to urban sprawl in a physical sense by lowering
densities and/or promoting growth at the urban fringe (Carruthers, 2002b; 2003;
Carruthers and Ulfarsson, 2002; Lewis, 1996; Pendall, 1999; Shen, 1996). So, even if
the lower costs are attributable to interjurisdictional competition, as the Tiebout model
suggests, they may not offset the effects caused by the physical pattern of development.
Likewise, if the correlation reflects the limitations of smaller tax bases, the creation
of new municipalities and special districts may not be an advantageous approach to
dealing with public services
ö
no matter what the effect of the physical pattern of
development. In any case, further applied research aimed at uncovering the nature
of the relationship between fragmentation and service expenditures and at evaluating
the relative costs and benefits of alternative political structures is needed before any
substantive conclusions can be made.
5 Summary and conclusions
Over the last several decades there has been a sustained interest in evaluating the
relative costs of alternative forms of development in US metropolitan areas. In this
paper we examined this issue through an analysis of the relationship between the
physical and political structure of metropolitan areas and twelve separate measures
of public expenditure: total direct, capital facilities, roadways, other transportation,
sewerage, trash collection, housing and community development, police protection,
fire protection, parks, education, and libraries. Our primary contribution has been to
provide empirical evidence of the widely held
ö
but largely unfounded
ö
belief among
planners that urban sprawl raises the cost of providing public services. In this way, we
have contributed to the sprawl ^ antisprawl debate in favor of more compact cities;
although US metropolitan areas will continue to suburbanize, the results presented
here suggest that they may maintain a more cost-effective urban form by doing so at
higher densities and by consuming less land. Although public service expenditures
represent just one aspect of urban performance, minimising the cost of such services
to residents produces net benefits to the public at large, as long as the quality of those
services remains unaffected. Talen and Ellis (2002) recently called for research to
develop well-validated criteria for identifying desirable outcomes of urban develop-
ment. The findings of this analysis represent substantive evidence that, at least from
the standpoint of public finance, a more compact urban form is a desirable planning
goal.
Acknowledgements. Special thanks to Peter Johnson for his invaluable research assistance, and to
the two anonymous reviewers for their insightful comments and suggestions. An earlier version of
this paper was presented at the meetings of the Pacific Regional Science Conference Organization
in Portland, Oregon, in July 2001.
520 J I Carruthers, G F Ulfarsson
References
Alberti M, 1999, ``Urban p atterns and environm ental pe rformance: what do we know?'' Journal
of Planning Education and Research 19 151 ^ 16 3
Altshuler A, Gome
¨z-Iba
¨n
¬ez J, 1993 Regulation for Revenue: The Political Economy of Land Use
Exactions (Lincoln Institute of Land Policy, Cambridge, MA)
Bahl R, 1968, ``A land speculation model: the role of the property tax as a constraint to urban
sprawl'' Journal of Regional Science 8199 ^ 208
Brueckner J, 2000, ``Urban sprawl: diagnosis and remedies'' International Regional Science
Review 23 16 0 ^ 171
Burchell R, 1998 Transportation Cooperative Research Program Report 39: The Costs of Sprawl
ö
Revisited (National Academy Press, Washington, DC)
Carruthers J, 2002a, ``Evaluating the effectiveness of regulatory growth management programs:
an analytical framework'' Journal of Planning Education and Research 21 406 ^ 420
Carruthers J, 2002b,``The impacts of state growth management programs: a comparative
analysis'' Urban Studies 39 1959 ^ 1982
Carruthers J, 2003,``Growth at the fringe: the impact of political fragmentation in United States
metropolitan areas'' Papers in Regional Science 84 forthcoming
Carruthers J, Ulfarsson G, 2002, ``Fragmentation and sprawl: evidence from interregional
analysis'' Growth and Change 33 312 ^ 340
Clawson M, 1962, ``Urban sprawl and speculation in suburban land'' Land Economics 38 9 9 ^ 111
Dowding K P J, Biggs S, 1994, ``Tiebout: a survey of the empirical literature'' Urban Studies 31
767^798
Downs A, 1994 New Visions for Metropolitan America (Lincoln Institute of Land Policy,
Cambridge, MA)
Downs A, 1999,``Some realities about sprawl and decline'' Housing Policy Debate 10 955 ^ 974
Ewing R, 1997, ``Is Los Angeles-style sprawl desirable? Journal of the American Planning
Association 63 107 ^ 12 6
Fischel W, 1985 The Economics of Zoning Laws: A Property Rights Approach to American Land
Use Controls (Johns Hopkins University Press, Baltimore, MD)
Foster K, 1997 The Political Economy of Special Purpose Government (Georgetown University
Press, Washington, DC)
Frank J, 1989 The Costs of Alternative Development Patterns: A Review of the Literature (The
Urban Land I nstitute,Washington, DC)
Galster G, Hanson R, Ratcliffe M,Wolman H, Coleman S, Freihage J, 2001,``Wrestling sprawl
to the ground: defining and measuring an elusive concept'' Housing Policy Debate 12 681^717
Gordon P, Richardson H, 1997, ``Are compact cities a desirable planning goal?'' Journal of the
American Planning Association 63 95 ^ 106
Kaiser E, Godschalk D, Chapin S, 1995 Urban Land Use Planning (University of Illinois Press,
Urbana, IL)
Kelly E, 1993 Managing Community Growth: Policies, Techniques, and Impacts (Praeger,Westport)
Kennedy P, 1998 A Guide to Econometrics (MIT Press, Cambridge, MA)
Knaap G, Nelson A, 1992 The Regulated Landscape: Lessons on State Land Use Planning From
Oregon (Lincoln Institute of Land Policy, Cambridge, MA)
Ladd H, 1992, ``Population growth, density, and the costs of providing publi c servi ces'' Urban
Studies 29 273 ^ 295
Ladd H, 1994, ``Fiscal impacts of local population growth: a conceptual and empirical analysis''
Regional Science and Urban Economics 24 661 ^ 686
Ladd H, 1998, ``Land use regulation as a fiscal tool'', in Local Government Tax and Land Use
Policies in the United States: Understanding the Links Ed. H Ladd (Lincoln Institute of Land
Policy, Ca mbridge, MA) pp 55 ^ 74
Ladd H, Yinger J, 1991 America's Ailing Cities: Fiscal Health and the Design of Urban Policy
(Johns Hopkins University Press, Baltimore, MD)
Lee D,1981, ``Land use planning as a response to market failure'', in The Land Use Policy Debate in
the United States Ed. J de Neufville (Plenum Press, New York) pp 149 ^ 164
Lewis P,1996 Shaping Suburbia: How Political Institutions Organize Urban Development (University
of Pittsburgh Press, Pittsburgh, PA)
Lynch K, 1981 Theory of Good City Form (MIT Press, Cambridge, MA)
Nelson A, Duncan J, Mullen C, Bishop K, 1995 Growth Management: Principles and Practices
(Planners Press, American Planning Association, Chicago, IL)
Urban sprawl and the cost of public services 521
Pendall R, 1999, ``Do l and-use controls cause sprawl?'' Environment and Planning B: Planning
and Design 26 555 ^ 571
Porter D, 1997 Managing Growth in America's Communities (Island Pr ess, Wash ington, DC)
REIS, various years Regional Economic Information System Geospatial and Statistical Data
Center, University of Virginia, http://fisher.lib.virginia.edu/reis/
RERC, 1974 The Costs of Sprawl: Environmental and Economic Costs of Alternative Residential
Development Patterns at the Urban Fringe Real Estate Research Corporation (US
Environmental Protection Age ncy, Washington, DC)
Shen Q, 1996, ``Spatial impacts of locally enacted growth controls: the San Francisco Bay Region
in the 1980s'' Environment and Planning B: Planning and Design 23 61 ^ 91
Speir C, Stephenson K, 2002, ``Does sprawl cost us all? Isolating the effects of housing patterns
on public water and sewer costs'' Journal of the American Planning Association 68 56 ^ 70
Talen E, Ellis C, 2002, ``Beyond relativism: reclaiming the search for good city form'' Journal of
Planning Education and Research 22 36 ^ 49
Tiebout C, 1956, ``A pu re theory of local expenditures'' Journal of Political Economy 64 416 ^ 424
USCG, various years a Gover nment Organization, Vol ume 1, Number 1 US Census of Governments
(US G overnment Printing Office, Washington, DC)
USCG, various years b Assessed Valuations for Local Gene ral PropertyTaxation,Volume 2, Number 1
US Census of Governments (US Government Printing Office, Washington, DC)
USCG, various years c Compendium of Government Finances, Volume 4, Number 5 US Census of
Governments (US Government Printing Office, Washington, DC)
USCG, 2000,``2000 Annual Survey of Local Government Finances'', form F-21, US Census of
Governments http://www.census.gov/govs/forms/2001/f21.pdf
USDA, 2002,``1997 National Resources Inventory (NRI)'', US Department of Agriculture;
machine-readable data, available on CD-ROM, National Cartography and Geospatial Center,
PO Box 6567, Fort Worth, TX 76115-0567
ß 2003 a Pion publication printed in Great Britain
522 J I Carruthers, G F Ulfarsson
