Urban Affairs Review http://uar.sagepub.

com/

Are Fragmentation and Sprawl Interlinked?: North American Evidence

Eran Razin and Mark Rosentraub
Urban Affairs Review 2000 35: 821
DOI: 10.1177/10780870022184697

The online version of this article can be found at:

http://uar.sagepub.com/content/35/6/821

Published by:

http://www.sagepublications.com

On behalf of:

The Urban Politics Section, American Political Science Association

Additional services and information for Urban Affairs Review can be found at:

Email Alerts: http://uar.sagepub.com/cgi/alerts

Subscriptions: http://uar.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://uar.sagepub.com/content/35/6/821.refs.html

>> Version of Record - Jul 1, 2000

What is This?

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
Razin, Rosentraub / RESEARCH
URBAN
NOTE
AFFAIRS REVIEW / July 2000

RESEARCH NOTES

ARE FRAGMENTATION AND

SPRAWL INTERLINKED?
North American Evidence
ERAN RAZIN
Hebrew University
MARK ROSENTRAUB
Indiana University

The association between municipal fragmentation and suburban sprawl is examined, based on a
cross-sectional analysis of all U.S. and Canadian metropolitan areas with more than 500,000 res-
idents in the 1990s. Results reveal that this association is rather weak but significant and is sus-
tained even when the less fragmented and more compact Canadian metropolitan areas are
excluded from the analysis. The impact of residential sprawl on fragmentation is significant, but
fragmentation does not predict sprawl. Low levels of fragmentation do not guarantee compact
development, but lack of excessive fragmentation might be a precondition for compact develop-
ment in North America.

SPRAWL AND FRAGMENTATION:

THE ASSUMED LINKAGE

There is some irony in the similarity of issues confronting urban areas in

the 1890s and the 1990s. Then, as now, the growth spiraling away from center
cities led to sprawl and new governments (Sayre and Kaufman 1960; Peiser
1984; Real Estate Research Corporation 1974; Bourne 1992). Critics then, as
now, described sprawl as uncoordinated low-density development frequently
“leapfrogging” over green space and open land. To reduce sprawl and
enhance the quality of life, reformers in both time periods proposed consoli-
dating fragmented governance systems comprising a myriad of jurisdictions
with control over local land-use policies (Peirce 1993).

URBAN AFFAIRS REVIEW, Vol. 35, No. 6, July 2000 821-836

© 2000 Sage Publications, Inc.

821

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
822 URBAN AFFAIRS REVIEW / July 2000

Others are less critical of sprawl, noting that low-scale development has
not emerged as a drag on the U.S. economy (Gordon and Richardson 1997).
In addition, some areas with numerous governments have attempted to retard
sprawl (e.g., Oregon and New Jersey). The level of intergovernmental service
delivery arrangements that exist today (Thompson 1997) also suggests that
fragmented governments are neither necessarily wasteful nor unable to coop-
erate (Parks and Oakerson 1989).
It is somewhat surprising that little evidence exists indicating whether a
“Tiebout-style” framework for urban governance produces sprawl-like con-
ditions. Some accept ending fragmentation as axiomatic to ensure an envi-
ronmentally responsible future. If governmental fragmentation is unrelated
to sprawl—people may simply prefer smaller governing units or they may
prefer low-density developments independent of governing structures—then
those addressing any perceived negative consequences from sprawl will have
to look at policy responses other than reducing the number of governmental
units. If fragmentation does indeed lead to sprawl, and if it can be determined
that sprawl is indeed negative, changing governance systems may be the right
step to change land-use patterns.
There seems to be acceptance of the idea that the existence of numerous
local governments encourages sprawl by discouraging uncoordinated plan-
ning (Downs 1994; Norris 1998) and encouraging competition over taxable
property (Razin 1998). The acceptance of the association between sprawl
and fragmentation of powers was probably made most clear when Downs
(1998) defined suburban sprawl as including a fragmentation of land-use
powers and large disparities in fiscal capacity between cities. Our study
examines whether fragmentation and sprawl are indeed associated in North
America and whether arguments for a causal relationship in a particular
direction can be supported.

DATA AND METHODOLOGY

A cross-sectional examination assessed the association between fragmen-

tation and sprawl in all metropolitan areas with at least 500,000 inhabitants in
North America. Each metropolitan area was ranked on several measures of
fragmentation and sprawl, based on data from the U.S. Bureau of the Census
and Statistics Canada. An examination of correlation coefficients between
these measures preceded a regression analysis of the factors that explained
fragmentation and sprawl.
Because most of the largest urban concentrations in the United States are
defined by the U.S. Bureau of the Census as two-tiered metropolitan areas

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 Razin, Rosentraub / RESEARCH NOTE 823

(composed of several primary metropolitan statistical areas [PMSAs] that

form one consolidated metropolitan statistical area [CMSA]), a choice had to
be made to include one or the other unit in multivariate analyses. Neither
seemed to be optimal; PMSAs are sometimes narrowly defined and include
only central or suburban portions of a metropolitan region. Several PMSAs,
such as Nassau-Suffolk and Monmouth-Ocean, are completely suburban.
CMSAs are sometimes too broadly defined for an examination of fragmenta-
tion and sprawl. The Washington-Baltimore CMSA includes two physically
separated metropolitan areas. Moreover, delineation of PMSAs and CMSAs
is based on county boundaries; hence it can include large tracts of rural land
beyond actual metropolitan boundaries. The Riverside–San Bernardino
PMSA in California has the Colorado River as its eastern boundary. In that
PMSA, population density measures would be meaningless.
The most appropriate unit for comparing metropolitan population densi-
ties in the United States is the urbanized area (U.S. Bureau of the Census
1993a). Thus, an aggregation of PMSAs according to the criterion of maxi-
mum similarity to the urbanized area definition was used for fragmentation
measures. In some metropolitan areas, this meant using CMSA-level data for
the metropolitan area and in others using PMSAs or a combination of
PMSAs. For metropolitan areas in New England, data for NECMAs (New
England county metropolitan areas) were available; as a result, urbanized
areas with more than 500,000 inhabitants (such as Boston and Worcester)
were combined to fit most closely the NECMA definition.
The study included five measures of municipal fragmentation: (1) total
number of local governments per 10,000 residents (including special-pur-
pose governments), (2) general-purpose governments per 10,000 residents,
(3) the existence of multipurpose metropolitan-level government, (4) the pro-
portion of population in the largest city, and (5) the proportion of population
in cities with at least 100,000 residents. Measures for U.S. metropolitan areas
were based on data from the 1992 census of governments, and measures for
the Canadian regions were based on the 1996 census. Data for the total num-
ber of local governments per 10,000 residents were not available in Canada.
A composite fragmentation measure was calculated based on two measures:
general-purpose governments per capita and the proportion of population in
cities with more than 100,000 residents. An additional composite measure
gave extra weight to metropolitan areas that had a metropolitan tier of local
governments.
The fragmentation data reported in selected U.S. census publications have
some anomalies. In Indianapolis, for example, the number of general-pur-
pose governments reported by the U.S. Bureau of the Census is greater than
the actual number because the creation of a consolidated city left some

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
824 URBAN AFFAIRS REVIEW / July 2000

governmental units with minimal responsibilities. In Canada, several gen-

eral-purpose governments are small Indian reservations. However, we
refrained from any adjustments of census data, as these would have been par-
tial and subjective.
Measures of sprawl are difficult to operationalize. Ewing (1997)
downplayed the role of density and viewed poor accessibility and lack of
functional open space as indicators of sprawl. That perspective does not
define sprawl directly but measures its perceived negative implications.
Low-density development driven by market forces may come out quite com-
pact according to these measures. Moreover, it can be argued that density is
indeed the first, although not only, criterion of sprawl. Downs (1998) found
no significant relationship between most traits of sprawl and measures of
urban decline. He thus assumed that the relationship might be inherent in all
forms of U.S. suburban growth, not just finely defined suburban sprawl.
Our study used measures of density to measure sprawl: (1) the percentage
of dwellings in single-unit detached houses, (2) population per square kilo-
meter, and (3) housing units per square kilometer. Measures of density per
square kilometer were calculated in Canada for urbanized cores of Canadian
metropolitan areas, which were assumed to correspond most closely to
urbanized areas in the United States. However, it should be acknowledged
that the definitions are not identical. Also, some municipalities in Canadian
metropolitan areas are split into land defined as urbanized core and land
defined as urban or rural fringe. In these cases, we used rough estimations of
population and land area included in the urbanized core. All three concepts
served to define a composite measure of residential sprawl.
Measuring sprawl for nonresidential land uses requires surveys and data
beyond the scope of this study. On the basis of available data, we could only
calculate an indirect measure: a ratio of the proportion of a region’s jobs
located in central cities, defined according to criteria used by the U.S. Bureau
of the Census, divided by the proportion of a region’s population in the cen-
tral cities. This ratio has two major limitations: (1) It depends on the U.S.
Bureau of the Census’s definition of central cities, which is not the most
appropriate for our objectives, and (2) it depends on the proportion of popula-
tion residing in central cities. When central cities are large and include a high
proportion of a metropolitan area’s total population, the ratio cannot be very
high.
The regression models on each of the fragmentation and sprawl measures
referred to the following control variables:

1. Country: Canada or the United States. Metropolitan areas in Canada are ex-
pected to be less fragmented and more compact (Goldberg and Mercer 1986).

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 Razin, Rosentraub / RESEARCH NOTE 825

2. Population of the metropolitan area. Larger metropolitan areas are expected to

be denser and less fragmented in per capita terms.
3. Age of the metropolitan area. Metropolitan areas that were already large in
1940, thus less influenced by the automobile, are expected to be relatively
dense.
4. Housing costs. High housing costs indicate high land values and encourage
high-density development.
5. Annexation procedures. Metropolitan areas are expected to be more frag-
mented where annexation requires the consent of all sides involved.
6. Proportion of poor population. Racial tensions and the desire of the middle
and upper classes to isolate themselves from poor minorities could be a major
factor instigating fragmentation.
7. Region. Generalizations based on the experiences of older northeastern and
midwestern metropolitan areas in the United States are not necessarily rele-
vant to those in the South and West (Rothenberg Pack 1998).

Our cross-sectional analysis was limited in its ability to detect causality.

Levels of municipal fragmentation in particular metropolitan areas change
over time, and land-use patterns may reflect municipal structures of the past
rather than present ones. Thus, an analysis of measures of change could have
been viewed as more appropriate for the study—for example, studying the re-
lationship between fragmentation levels at the beginning of each decade with
the change in population densities in the same decade (influenced by popula-
tion growth and expansion of the urbanized area). However, such an approach
also entails problems of data collection and definitions. As a first approach to
the issue, a cross-sectional study that detects relationships as they exist at a
particular moment seemed more appropriate.

RANKING NORTH AMERICAN METROPOLITAN

AREAS BY FRAGMENTATION AND SPRAWL

The least fragmented metropolitan area in North America is Honolulu (see

Table 1). The complete lack of municipal fragmentation in Honolulu reflects
the unique centralization of power in Hawaii. Canadian metropolitan areas
are also characterized by low levels of fragmentation and have particularly
high proportions of population in cities of more than 100,000 and in the larg-
est city in the region. Small metropolitan areas that lack a strong, dominant
central city are most fragmented. Boston, Atlanta, and Washington are three
large metropolitan areas with a particularly small proportion of the popula-
tion in their central city but are less fragmented according to other measures.
The specific traits of local government in different states and provinces
also influence fragmentation. It has been noted that Pittsburgh is the most

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 826

TABLE 1: Least and Most Fragmented Metropolitan Areas in North America by Selected Measures of Municipal Fragmentation,
1992 (United States) and 1996 (Canada)

Fragmentation
Measure Total Local General-Purpose Proportion of Proportion of
Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014

(Metropolitan Fragmentation Governments per Governments per Population in Population in

Metropolitan Area Government)a Measureb 10,000 Residents 10,000 Residents the Largest City Cities of 100,000+

Least fragmented metropolitan areas

1. Honolulu, HI 100.0 100.0 0.05 0.01 100.0 100.0
2. Toronto, ON 100.0 92.6 .NA 0.08 15.3 87.6
3. Calgary, AB 95.0 95.0 .NA 0.11 93.5 93.5
4. Hamilton, ON 94.4 84.4 .NA 0.14 51.6 73.6
5. Winnipeg, MB 93.5 93.5 .NA 0.16 92.7 92.7
6. El Paso, TX-NM 92.0 92.0 0.56 0.11 87.6 87.6
7. Vancouver, BC 90.3 80.3 .NA 0.22 28.1 68.2
8. Norfolk–Virginia Beach, VA 89.4 89.4 0.20 0.11 27.9 82.6
9. Ottawa-Hull, ON-PQ 86.8 76.8 .NA 0.29 32.0 63.6
10. Phoenix, AZ 84.4 84.4 0.85 0.15 42.2 73.8
Most fragmented metropolitan areas
89. St. Louis, MO-IL 34.8 34.8 3.06 1.22 13.8 13.8
90. Allentown-Bethlehem, PA-NJ 31.9 31.9 3.45 1.45 16.6 16.6
91. Louisville, KY-IN 31.0 31.0 2.84 1.77 26.3 26.3
92. Ann Arbor, MI 29.6 29.6 2.53 1.68 20.5 20.5
93. Albany-Schenectady-Troy, NY 29.4 29.4 3.38 1.46 11.8 11.8
94. Syracuse, NY 28.5 28.5 3.06 1.76 20.9 20.9
95. Pittsburgh, PA 25.7 25.7 3.64 1.74 14.7 14.7
96. Youngstown, OH 22.1 22.1 2.71 1.53 14.6 0.0
 97. Harrisburg, PA 9.6 9.6 4.79 2.21 8.3 0.0
98. Scranton, PA 0.0 0.0 5.09 2.74 12.3 0.0

SOURCE: U.S. Bureau of the Census (1994); County and City Extra (1998); Statistics Canada (1997).
NOTE: The rankings include all metropolitan areas of 500,000 residents or more: metropolitan statistical areas (MSAs) in the United States and census metro-
Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014

politan areas (CMAs) in Canada. In the case of consolidated MSAs divided into primary MSAs (PSMAs) in the United States, PMSAs were grouped in a man-
ner that resembles most closely the urbanized area definitions. NA = not available.
a. Based on two variables: general-purpose governments per 10,000 residents and proportion of population in cities of 100,000+. Each variable was trans-
formed into a 0 to 100 range, 100 being the value for the least fragmented metropolitan area. The composite measure is an average of these two transformed
variables. Ten extra points (up to a maximal value of 100.0 for the measure) were given to metropolitan areas that have multipurpose metropolitan government.
b. A similar fragmentation measure that does not include the 10 extra points for metropolitan government.
827
828 URBAN AFFAIRS REVIEW / July 2000

fragmented large metropolitan area in North America. However, our data

indicate that fragmentation characterizes all metropolitan areas in Pennsyl-
vania. Metropolitan areas in Ohio, New York, and Michigan also tend to be
fragmented, and those in Missouri and Kansas are fragmented mainly in
terms of local governments per capita. Low levels of fragmentation charac-
terize Hawaii, California, and the Southwest. The more fragmented metro-
politan areas in Canada were in the province of Quebec (Montreal and Que-
bec), but even those were not as fragmented as the more fragmented U.S.
metropolitan areas.
Canadian metropolitan areas have low levels of residential sprawl (see
Table 2). New York–North New Jersey, Miami, and Los Angeles (including
Orange County) are the more compact American metropolitan areas. Although
many cite Los Angeles as the perfect example of sprawl (Ewing 1997; Leo
1998), its urbanized area has one of the highest residential densities in the
United States as a result of high land prices (Gordon and Richardson 1998).
Large metropolitan areas, characterized by high land prices, are indeed
prominent in the list of metropolitan areas with the lowest levels of residen-
tial sprawl. Metropolitan areas with the highest levels of residential sprawl
were mid-size and smaller ones, mainly in the Midwest, Southeast, and cen-
tral plains (Table 2), where land is cheaper than in cities with less sprawl.
Metropolitan areas with the highest concentrations of employment in cen-
tral cities (data not shown) tended to be small and characterized by a low pro-
portion of population in their central cities. Honolulu and Ottawa-Hull were
remarkable in the concentration of employment in their relatively larger cen-
tral cities. Metropolitan areas with the lowest concentration of employment
in their central cities tended to be very large ones (Philadelphia, Chicago,
New York, Los Angeles) and those with declining cores such as Detroit. In
New York, Los Angeles, and Chicago, residents were more concentrated in
central cities than were employment opportunities. Jobs were more concen-
trated than population in central cities in all Canadian metropolitan areas.

THE CORRELATION BETWEEN

FRAGMENTATION AND SPRAWL

The rankings of both fragmentation and sprawl do hint at a possible asso-

ciation. The most striking feature is the position of Canadian metropolitan
areas as being both less fragmented and with less sprawl. However, the asso-
ciation depends substantially on the measures used. Among fragmentation
measures, the number of general-purpose governments per capita was highly
correlated with the total number of local governments per capita (R = 0.81).

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 Razin, Rosentraub / RESEARCH NOTE 829

The proportion of cities with a population in excess of 100,000 was highly

correlated with the proportion of population in the largest city (0.83). Our
composite fragmentation measures average two weakly correlated measures:
general-purpose governments per capita and the proportion of population
residing in large cities.
The Pearson correlation coefficients between the composite fragmenta-
tion and residential sprawl measures were significant but not very robust
(0.38 and 0.33, the latter one referring to the fragmentation measure that does
not take into account general-purpose metropolitan governments). To what
extent do these correlations merely reflect the tendency of Canadian metro-
politan areas to be less fragmented and more compact than U.S. ones? The
correlation coefficients between the composite fragmentation and residential
sprawl measures were indeed lower when only the U.S. metropolitan areas
were included in the calculations (0.27 between fragmentation, including
metropolitan government, and residential sprawl and 0.26 between fragmen-
tation, excluding metropolitan government, and residential sprawl). How-
ever, the coefficients were positive and significant at the 0.01 level.
An examination of the position of individual metropolitan areas with
regards to the composite fragmentation and residential sprawl measures (see
Figure 1) reveals that a substantial number of metropolitan areas are charac-
terized by substantial sprawl despite low levels of fragmentation. Such met-
ropolitan areas include Calgary, El Paso, Edmonton, Norfolk, Jacksonville,
Albuquerque, Mobile, Oklahoma City, the Dallas/Fort Worth area, and Hous-
ton. These metropolitan areas are usually characterized by large supplies of
relatively inexpensive land that has allowed sprawl, even when the local gov-
ernment has not been particularly fragmented. The opposite, however, is
much less true. There are only a few metropolitan areas that are more frag-
mented than average and less sprawling than average, and there is no metro-
politan area that is remarkably compact despite being fragmented (see Fig-
ure 1). Thus it may be argued that in a municipally fragmented environment,
controlling sprawl is unlikely, or perhaps more compact residential develop-
ment reduces the prospects for extreme fragmentation.

THE FACTORS INFLUENCING

FRAGMENTATION AND SPRAWL

Age of the metropolitan area was found to be the most prominent variable
predicting fragmentation (see Table 3). Older metropolitan areas were more
fragmented than younger ones. As expected, the Canada-U.S. distinction was
another major explanatory variable: Canadian metropolitan areas were less

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 830

TABLE 2: Metropolitan Areas in North America with Lowest and Highest Residential Sprawl by Measures of Residential Sprawl,
1990s

Residential % of Dwellings in Population per Housing Units per

Metropolitan Area Sprawl Measurea Single-Unit Detached Housesb Square Kilometerc Square Kilometerc
Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014

Lowest residential sprawl

1. Montreal, PQ 91.81 29.8 2,060.6 843.4
2. Toronto, ON 88.93 44.4 2,469.0 872.0
3. New York–North New Jersey, NY-NJ 85.41 36.9 2,088.2 813.7
4. Miami, FL 84.22 40.2 2,096.2 836.8
5. Los Angeles, CA 78.35 48.6 2,239.7 799.4
6. Fort Lauderdale, FL 70.81 37.3 1,461.3 734.2
7. Honolulu, HI 64.18 44.9 1,761.1 616.3
8. Ottawa-Hull, ON-PQ 61.88 44.1 1,572.4 621.7
9. San Francisco–Oakland, CA 60.79 48.3 1,603.2 653.1
10. Chicago-Gary, IL-IN 60.57 49.2 1,655.1 644.7
Highest residential sprawl
89. Youngstown, OH 13.40 73.9 834.4 345.2
90. Kansas City, MO-KS 12.45 66.6 646.4 278.2
91. Little Rock, AR 11.62 65.3 591.9 259.9
92. Greenville, SC 11.28 67.7 647.0 271.1
93. Tulsa, OK 10.86 67.5 602.1 274.5
94. Nashville, TN 10.60 59.9 457.8 201.1
95. Birmingham, AL 10.45 66.7 602.3 253.1
96. Knoxville, TN 7.42 68.1 537.4 235.8
97. Mobile, AL 5.19 68.4 507.4 206.1
98. Oklahoma City, OK 5.17 68.0 468.2 211.4

SOURCE: U.S. Bureau of the Census (1993a, 1993b); Statistics Canada (1993, 1997).
 a. Each of the three sprawl variables was transformed into a 0 to 100 range, 100 being the value for least residential sprawl. The composite measure is an aver-
age of these three transformed variables.
b. Data refer to metropolitan statistical areas or the combination of primary metropolitan statistical areas (PMSAs) that conform most closely to the urbanized
areas in the United States and to census metropolitan areas (CMAs) in Canada.
c. Calculated for urbanized areas in the United States and for urbanized cores of CMAs in Canada (in the case of CMAs that are consolidated with secondary
Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014

CAs, only the primary CMA [PCMA] was included in the calculation).
831
832 URBAN AFFAIRS REVIEW / July 2000

Figure 1: Metropolitan Areas in North America by Municipal Fragmentation and Resi-

dential Sprawl, 1990s

fragmented. Annexation method was also a major variable: A requirement

for majority approval in areas to be annexed caused metropolitan areas to be
more fragmented. Large population and high proportion of poor population
were associated with lower fragmentation in U.S.-only models. Perhaps frag-
mentation pressures are associated more with middle-class population;
hence a high proportion of lower-income population produces a less frag-
mented pattern.
The major sprawl variable influencing fragmentation was population den-
sity. Higher densities were predictors of lower fragmentation, and this influ-
ence was significant when taking into account the impact of the other explan-
atory variables (see Table 3). Excluding the Canadian metropolitan areas
from the models reduced the explanatory power of population density, but it
remained significant at the 0.02 level.
The major variable influencing the composite residential sprawl measure
was the Canada-U.S. distinction followed closely by median rent of housing
units (a land price measure) and population size. Canadian metropolitan

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 TABLE 3: Factors Influencing Municipal Fragmentation and Residential Sprawl Among North American Metropolitan Areas: Ordi-
nary Least Squares Regression Analyses, 1990s

Dependent Variables
Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014

Fragmentation
Fragmentation Measure (Metropolitan Residential Sprawl
Measure (Metropolitan Government), Residential Sprawl Measure,
a
Government) United States Onlya Measureb United States Onlyb
Independent Variables B Significance B Significance B Significance B Significance

Constant 25.004 0.33 –1.586 0.95 –6.256 0.42 –27.227 0.03

c
Canada–United States 17.293 0.00 28.181 0.00
% housing units built 1939 or earlier –0.979 0.00 –0.901 0.00 0.233 0.15 0.274 0.08
Population 0.000 0.00 0.000 0.00
Ln (population) 2.613 0.19 3.960 0.05
Population per square kilometer 0.013 0.00 0.011 0.02
Central metro area in a larger metro region 11.408 0.04 10.304 0.05
General-purpose governments per capita –4.947 0.14 –2.832 0.38
Median rent of housing units 0.070 0.00 0.092 0.00
d
Majority approval for annexation required –7.483 0.00 –8.316 0.00
e
% population below poverty level 0.610 0.02 0.658 0.04
Adjusted R2 0.60 0.54 0.57 0.56
n 98 89 98 89

a. High values represent low fragmentation.

b. High values represent lower residential sprawl.
c. Canada = 1, United States = 0.
d. Majority approval in area to be annexed required in 1990 = 1; majority approval not required = 0.
833

e. Included only in U.S.-only regressions.

834 URBAN AFFAIRS REVIEW / July 2000

areas, metropolitan areas with high land values, and large metropolitan areas
were less sprawling (see Table 3). Metropolitan areas that were practically
the central parts of broader metropolitan regions were also more compact.
Poverty and age were also positively associated with more compact metro-
politan areas.
Fragmentation did not have any significant impact on the composite mea-
sure of residential sprawl (see Table 3) or on any specific measure of sprawl.
The number of general-purpose governments per capita, which was the frag-
mentation variable most closely correlated with residential sprawl, was not
significant in the models. Regression models of aggregate variables on a sam-
ple of less than 100 might be sensitive to the exact definition of the model.
However, no combination of independent variables on any of the dependent
residential sprawl variables could produce a significant impact of fragmenta-
tion on residential sprawl.
A strong explanatory power of the proportion of population in cities of
100,000 or more residents on the concentration ratio of employment in cen-
tral cities is largely definitional. Nevertheless, a regression model (not
shown) revealed that high employment concentrations in central cities char-
acterized younger metropolitan areas, small metropolitan areas, and Cana-
dian metropolitan areas.

FRAGMENTATION AND SPRAWL ARE LINKED, BUT

ONE IS NOT AN INHERENT ATTRIBUTE OF THE OTHER

There is a relatively weak association between fragmented government

structures and sprawl. The positive correlation is partly a consequence of the
inclusion of Canadian metropolitan areas in the analysis, which are generally
less fragmented and less sprawling than U.S. ones. However, the correlation
between fragmentation and sprawl remains significant, although weak, when
only U.S. metropolitan areas are examined. The regression models illustrate
that although the impact of residential sprawl on fragmentation is significant,
fragmentation does not predict residential sprawl. Variations in residential
sprawl are largely explained by the basic differences between national politi-
cal/land market/planning systems (the distinction between Canada and the
United States) and by land values.
Low levels of fragmentation do not guarantee compact development. In
contrast, the combination of high levels of fragmentation and low levels of
sprawl is rare. Compact forms of development do not coexist with most frag-
mented municipal patterns, perhaps because fragmentation precludes the

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
 Razin, Rosentraub / RESEARCH NOTE 835

control of sprawl or because compact residential development reduces pros-

pects for fragmentation.
Canadian metropolitan areas were indeed characterized by less municipal
fragmentation and less sprawl. Older metropolitan areas were more frag-
mented but more compact. However, age primarily promotes fragmentation,
and its impact on compaction is secondary, probably because all metropoli-
tan areas in North America grew vastly in the age of the automobile.
The above results stress that given the weak association of fragmentation
and sprawl, assessments of the pros and cons of sprawl cannot always assume
fragmentation as an integral aspect of sprawl or as a major factor responsible
for sprawl. Nevertheless, lack of excessive fragmentation seems to be a pre-
condition for compact development in North America, although it is not a
sufficient condition. Noting that sprawl does indeed increase fragmentation
does not imply that fragmented areas cannot work together to correct out-
comes that are seen as negative. However, assessments of sprawl and frag-
mentation should address these phenomena separately.

REFERENCES

Bourne, L. S. 1992. Self-fulfilling prophecies? Decentralization, inner city decline, and the qual-
ity of urban life. Journal of the American Planning Association 58:509-13.
County and city extra: Annual metro, city and county data book. 1998. Lanham, MD: Bernan.
Downs, A. 1994. New visions for metropolitan America. Washington, DC: Brookings Institution.
. 1998. The big picture, how America’s cities are growing. The Brookings Review 16 (3):
8-12.
Ewing, R. 1997. Is Los Angeles-style sprawl desirable? Journal of the American Planning Asso-
ciation 63:107-26.
Goldberg, M. A., and J. Mercer. 1986. The myth of the North American city. Vancouver: Univ. of
British Columbia Press.
Gordon, P., and H. W. Richardson. 1997. Are compact cities a desirable planning goal? Journal
of the American Planning Association 63:95-106.
. 1998. Prove it: The costs and benefits of sprawl. The Brookings Review 16 (3): 23-26.
Leo, C. 1998. Regional growth management regime: The case of Portland, Oregon. Journal of
Urban Affairs 20:363-94.
Norris, D. F. 1998. Local government reform in the US—and why it differs so greatly from Brit-
ain. In Local government reorganization: The review and its aftermath, edited by S. Leach,
113-30. London: Frank Cass.
Parks, R., and R. Oakerson. 1989. Metropolitan organization and governance: A local public
economy approach. Urban Affairs Quarterly 25:18-29.
Peirce, N., with C. Johnson and J. S. Hall. 1993. Citistates: How urban America can prosper in a
competitive world. Washington, DC: Seven Locks Press.
Peiser, R. B. 1984. Does it pay to plan suburban growth? Journal of the American Planning Asso-
ciation 50:419-33.

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014
836 URBAN AFFAIRS REVIEW / July 2000

Razin, E. 1998. Policies to control urban sprawl: planning regulations or changes in the “rules of
the game.” Urban Studies 35:321-40.
Real Estate Research Corporation. 1974. The costs of urban sprawl: Detailed cost analysis.
Washington, DC: Government Printing Office.
Rothenberg Pack, J. 1998. Metropolitan areas, regional differences. The Brookings Review 16
(3): 26-31.
Sayre, W. S., and H. Kaufman. 1960. Governing New York city: Politics in the metropolis. New
York: Norton.
Statistics Canada. 1993. 1991 census: Profile of census metropolitan areas and census agglom-
erations. Ottawa: Author.
. 1997. 1996 census: Census metropolitan areas, census agglomeration and census tract
reference maps. Ottawa: Author.
Thompson, L. 1997. Citizen attitudes about service delivery modes. Journal of Urban Affairs
19:291-302.
U.S. Bureau of the Census. 1993a. 1990 census of population and housing, supplementary
reports: Urbanized areas of the United States and Puerto Rico. Washington, DC: Govern-
ment Printing Office.
. 1993b. 1990 census of population and housing, supplementary reports: Metropolitan
areas as defined by the Office of Management and Budget, June 30, 1993. Washington, DC:
Government Printing Office.
. 1994. 1992 census of governments: Vol. 1. Government organization. Washington, DC:
Government Printing Office.

Eran Razin is an associate professor of geography at the Hebrew University of Jerusa-

lem. His recent publications focus on the impact of local government organization on
development and disparities (Environment and Planning C, Urban Studies, Space and
Polity) and on immigrant entrepreneurs in different urban milieus (International Migra-
tion Review, Urban Affairs Review, Urban Geography, TESG).

Mark Rosentraub is a professor and associate dean at Indiana University’s School of

Public and Environmental Affairs. His recent work has appeared in Public Finance Re-
view, Public Administration Review, and the Journal of Urban Affairs. The second and
revised edition of his Major League Losers: The Real Costs of Sports and Who’s Paying
for It was published in 1999 by Basic Books.

Downloaded from uar.sagepub.com at Nat. Taichung Univ. of Sci. & Tech. on April 25, 2014