Professional Documents
Culture Documents
Barkley IdentifyingSpreadversus 1996
Barkley IdentifyingSpreadversus 1996
Functions Approach
Author(s): David L. Barkley, Mark S. Henry and Shuming Bao
Source: Land Economics , Aug., 1996, Vol. 72, No. 3 (Aug., 1996), pp. 336-357
Published by: University of Wisconsin Press
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
University of Wisconsin Press is collaborating with JSTOR to digitize, preserve and extend
access to Land Economics
cation of economic
ity in theactivity
periphery, Gaile terms (Mack
the result- a
Schaeffer 1993). ing impact as spread. A decline in the abso-
Of interest in this lute
study is the
level of economic activitynet effect
in the periph-
of population growth in an with
ery in conjunction economic
core expansion is area
metropolitan core evidence
on population change
of a backwash effect. Economic
nonmetropolitan theory
areas proximate
suggests that the net effect of tothe a
more distant from the core.
spread-backwash processGaile (1980)
varies according to
proposes that economic development
the characteristics in t
of the core and surround-
core impacts the surrounding region
ing area (e.g., size and growth ratethroug
of the
a complex set of dynamic spatial
core, industrial structure of theprocesse
core, exist-
These processes (summarized in Table
ing spatial distribution of development, dis-
include intraregional flows
tribution of: private
of transportation and communica-ca
tal; private and public
tion networks, expenditures
and the spatial distribution of f
goods and services;sociopolitical information power). and tech
nology; residents and commuters; and poli
cal influence and public investments. Su
flows generally provide III. DELINEATION
both OF beneficial
REGIONAL a
detrimental impacts on ECONOMIC
the AREAS
peripheral r
gion with the net result depending on t
relative magnitude of It longthe has positive
been recognized among
and re-
neg
tive forces. If the cumulative gional scientists that the economic
process develop
result
in an increase in the absolute level of activ- ment prospects of rural areas generally can
TABLE 1
INTRA-REGIONAL FLOWS STIMULATED BY GROWTH OF THE METROPOLITAN CORE AREA
Flows of People
Rural labor commutes to the urban area for employment (spread)
Urban families relocate residences to rural areas because of lower real estate costs and perceived higher
of life (spread)
Rural residences migrate to the urban areas for better access to employment and urban lifestyle (back
Flows of Firms/Employment
Firms in the mature or declining stage of product life cycle locate in rural areas to take advantage of lo
and land costs (spread)
Firms in the innovative or growing stage of product life cycle locate in urban areas to benefit from agg
economies, markets, and specialized labor (backwash)
to FEAs so that commuter flows are maxi- density patterns within the overall decreas-
mized within an FEA.2 ing rate. Neither the often observed "crest"
The eight FEAs identified (see Figure 1) or "crater" in density at the city center nor
for the study area include four that are the leveling off of residential density at the
relatively large, each containing at least 11 boundary of the metropolis can be repre-
counties (Augusta, GA; Charlotte, NC; sented by this function. Recent literature is
Columbia, SC; and Greenville, SC). The also skeptical of the appropriateness of
three FEAs bordering the Atlantic coast monocentric models as descriptors of re-
(Charleston, SC; Myrtle Beach, SC; and Sa- gional form since such models do not repre-
vannah, GA) all have six or fewer counties. sent areas with multiple metropolitan cen-
The small number of counties in these FEAs ters, growth corridors, edge cities, and
results from their inability to expand east oflower-ordered central places (Fishman 1990;
their core counties. The remaining FEA is aWaddell and Shukla 1993).
five-county region surrounding the small Anderson (1982) suggests that density
MSA of Florence, SC.3 functions with irregularities (crater at
center, undulations at non-core centers, lev-
IV. DENSITY FUNCTIONS AND eling off at fringe) may be estimated accu-
SPATIAL PROFILES rately with the following cubic spline speci-
fication:
Specification Alternatives
(ChaPdone PEA)
(Grenvfa PEA)
Plraae.
(Poec PEA)
C <Y
(Colmablemat A9 C S del
PEA am emoty
I (a=a city
(CCbmdnton PEA
(Savanah FEA
(Savwmh IFEA
FIGURE 1
MULTIPLE COUNTY FUNCTIONAL ECONOMIC AREAS: SOUTH CAROLINA REGION
In D, In D
t+1 x
t+1
_ : distance ;
+1
j t?' (x) tj t?'
1 W
Do > Do D. aDo
Y1+1, %, Y y,*+ < y'
b1+1 > b' b?1+ > b,
C. Core Growth-Hinterland Stagnation D. Backwash
In D, In D,
1+1 t+1
x x
b' tf' b'
1+1 t 1
Do > Do o > Do
b'+ > yt' by+ > by
bPt: b' 11b ' b1+1< b,
FIGURE 2
DENSITY FUNCTIONS AND SPATIAL TRANSFORMATIONS
V. ESTIMATED POPULATION
sity. The density functions are representa-
DENSITY FUNCTIONS
tive of those estimated with a linear spline
regression with one "knot" at the boundary
of the urbanized area.5 Cubic Spline Regressions
"Spread through growth" describes the
spatial transformation where residential
A spline function is a system of piece-wise
density in both the urban core and polynomial
periph- approximations. The function it-
selfgrows.
eral rural areas increase as the city is continuous and its derivatives also are
continuous.
That is, significant growth spills over to theSuits, Mason, and Chan (1978,
hinterlands but not at the expense of132) propose that "spline functions are a
growth
device
in the core. In terms of changes for approximating the shape of a
in the
density function, "spread throughcurvilinear
growth" stochastic function without the
may be depicted by an increase in necessity
city cen-of pre-specifying the mathematical
ter density (Do), little change inform of the function." Cubic spline func-
density
tions(b)
gradient (y), and an inflection point are at
especially useful in estimating pop-
greater distance from nodal centerulation
(referdensity
to functions for areas with
Figure 2A). craters near the core's center, undulations
where
"Spread through decentralization" subcenters are located, and a leveling
refers
to metropolitan growth occurring off of density in the rural hinterlands.7
primarily
away from the city center along with The cubic spline estimation begins by
signif-
icant spillovers to rural areas (Figure dividing2B).
the x axis (in this case the distance
Population spillovers evident primarily axis) intoat
segments where the dividing points
the metro fringe are referred to as are"metro-
referred to as "knots." Three segments
politan decentralization." Significant of equal length (two interior knots) are ad-
spillovers to hinterland areas are termed vised unless there are important reasons to
"regional deconcentration." The spread do otherwise (Suits, Mason, and Chan 1978).
through decentralization structural change In this study, however, we follow the proce-
may be represented by a stable or declining
central city density (D"+I < D'), a shal-
lower density gradient (y'+l < y'), and a 5Throughout this paper, the term "nodal center"
refers to the identified central business district in the
more distant boundary (b'+l > b').
FEA's metropolitan city. "Urban core" and "urban
"Core growth with periphery and hinter- field" refer to the urbanized area of the FEA's
land stagnation" refers to the development metropolitan core. The "fringe" is the rural area near
scenario where rural areas in the FEA re- est the urban core while "hinterland" is the area more
distant from the core.
ceive limited spillover of residential housing
activity as a result of growth in the FEA's 6The spatial structural changes reflected by 1980
and 1990 density functions capture only the spread and
core. In this case, the center density and
backwash effects closely associated with population
density gradient would increase but littlechange. Ideally, information on fluctuations in per
change would be evident in the density
capita income, employment opportunities, and quality
function's inflection point (Figure 2C). of life would be included to more completely assess the
impacts of metro growth on hinterland regions. How-
Finally, Gaile (1980) and Krugman (1991)ever, population growth or decline is viewed as an
note that nodal growth may lead to declin-important barometer of an area's economic well-being.
ing population densities in the surrounding
Population changes reflect changes in the area's tax
periphery and hinterlands if the core's
base, property values, retail markets, and employment
growth generates expenditure, capital, andopportunities. In addition, for the study area used in
this analysis, population density and per capita income
migration flows that are unfavorable to eco-
changes were significantly correlated. Tracts with rela-
nomic development in the hinterlands.
tively rapid population density growth also tended to be
These backwash effects may be represented those with more rapid per capita income growth.
by the following changes in the metro den- 7Density functions for the eight FEAs also were
estimated using the negative exponential, quadratic
sity function: an increase in center density,
gamma, and lognormal functional forms. The cubic-
a steeper gradient, and a reduction in fringe
spline estimations consistently provided the "best fit"
or hinterland population densities.6 and are the only results reported here.
and slope values for the 1980 and areas. 1990 Interestingly enough, the frequently
density functions for the eight FEAs indi-
noted density crater at the nodal center was
cate a general trend of metropolitan evident decen- only for the most populous
tralization. In all the FEAs except Myrtle metropolitan areas (Charlotte, Columbia,
Beach, nodal center density values declined and Greenville). Also, the exception to
(lower intercept values) and the density thegra-
above generalization is again the
dients became flatter (smaller negative co-
tourism/retirement area of Myrtle Beach.
efficients on distance) between 1980We andsuspect that the "irregularities" in this
1990. Thus, residential population near function
theresult from the situation that dis-
FEA core declined relative to population tance in
is not always a good measure for
census tracts further from the metro center. proximity to the nodal center because nu-
In the Myrtle Beach FEA, however, the
merous wetlands' areas prevent direct ac-
intercept increased while the density cess from one location to another.
gradi-
ent became only marginally "flatter." As noted above, the results of the cubic
Growth in the Myrtle Beach area appears spline toregression estimations are consistent
be spread throughout tracts both near with an tooverall trend of regional population
and distant from the FEA nodal center." dispersal through decentralization. The
Our general finding of spread through graphs in Figure 3 demonstrate, however,
metropolitan decentralization is consistent that the magnitude and spatial dimensions
with the- conclusions of earlier studies of of this decentralization vary by economic
changing urban residential and business lo- region, and these trends do not always fit
cation patterns (e.g., Anderson 1985; Parr neatly into the four categories previously
1987; Waddell and Shukla 1993). Yet signif- described. The identified changes in density
icant coefficients for the quadratic and cubic patterns are grouped into three "hybrid"
distance terms and the interval control categories:
dummy variables indicate undulations in the
density gradient and significant changes in 1. Fringe Growth through Decentraliza-
the density function itself over the range of
the FEA. Because of these additional con-
tion-Hinterland Stagnation. Four
FEAs with medium-sized nodal cities
siderations, a clearer appreciation of
(Augusta, Columbia, Greenville, and
changes in population patterns throughout
the entire FEA is provided by graphical Savannah) exhibited declining popula-
tion densities near the center and in-
representations of the estimated density
functions (Figure 3). The reader will note creasing densities in areas approxi-
that the changes in the FEAs' density func- mately 10 to 30 miles from the center.
tions (as shown in Figure 3) generally are
not dramatic for the period 1980 to 1990.
This absence of major shifts in density func- 11 The Chow test was applied to determine if the
tions probably is attributable to the study's
1980 and 1990 population density functions for each
FEA differed significantly. No significant structural
relatively short time frame. Thus, our find-
changes in density functions were detected for any of
ings should be interpreted as preliminary the eight FEAs. However, the Chow test does not
evidence of spread or backwash within these
identify density changes for specific segments within
economic regions. Additional evidence may the overall density functions. Thus we recast the mod-
be provided through analysis of spatial
els in terms of changes in population density from 1980
to 1990 as a function of distance from the core and
structural change over a more extended time
estimated these models using cubic-spline regressions.
period. For all the estimated equations except the Florence
The density functions in Figure 3 gener- FEA, the coefficients on the distance variables and
ally have the anticipated shape--density de- internal control dummy variables, and the overall F-
clines rapidly with distance from the core, statistics were significantly different from zero. Thus
for seven of the eight FEAs, changes in population
"bumps" exist in the gradient at the loca-
density from 1980 to 1990 were significantly different
tion of lower-ordered central places, and from zero. The results of these regressions are avail-
population density levels off in the rural able from the authors upon request.
Tract Tract
SAugusta
dWAg tF domily
6000 3500
2500
4000
2000
3000
1500 - -.. . . . . . . . . . . . . .
2000 .
1000 . . . . ..- --
1000500
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Dikence(Mlbs) from core tect Dietanoe(lbs) from com trct
Tract TOWt
o ion Savann
- -- 1980 Densy Function- - - 1980 Densty Function
1990 Denity Function 1990 Density Function
........--------------------- 0000
o4000 - - - - - - - - - - - - - - - - - - - - 4000 ------ ----
Mr2M ----------------------------------- _- - - - -
0 0
0 10 20 30 40 so so 0 10 20 30 40 so so
Beyond 30 miles, population densities Two FEAs with rapid growth (Char-
remained essentially unchanged from lotte, Charleston) experienced small
1980 to 1990. Metro area population declines in population densities near
growth and dispersal had little or no their nodal centers, but increasing res-
impact on lower-ordered central placesidential densities in tracts approxi-
and rural areas in the hinterland re- mately 5 to 35 miles from their cen-
gions of these FEAs. ters. Tracts more distant from the
2. Fringe Growth through Decentraliza- metro areas exhibited declining popu-
tion-Hinterland Decline (Backwash). lation densities. Growth at the urban
Tract motr
000 100- - - - - - - - - - - - - - - - - - - 2
low - - - - - - - - - - - - - -- - - -I 1000 - - - - - - - - - - - - - - - - - - - -
00
S10 20 0 40 50 0 10 20 30 40 50
Dian(ce(ns) from core tract Disanc(cils) from com tract
Tract Trct
oputio
dfrwiyFlorence FEA"ly
pop Myrtle Beach FEA
15oo- - - - - - ------------ a
5oo -- - --
0 0
0 10 20 30 40 O 0 0 10 20 30 40 s0 o 0
Distanoe(nis) fromcore tract Distanom(nmis) from core tract
FIGURE 3
(CONTINUED)
2000 - - - - - - - - - - - - - - - - 2000
0 0
0 10 20 30 40 50 eo 0 10 20 30 40 50 s0
Ostance(miles) Irom core tract omeno(mm) from cor
1- 1 00 ---------------.----
0 0
0 10 20 30 40 so 6s 0 10 20 30 40 60 60
FIGURE 4
ESTIMATED DENSITY FUNCTIONS, INTERSTATE VS. NON-INTERSTATE HIGHWAYS PRO
1980 AND 1990
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Distance(miles) from core tact Disltance(miles) from core tract
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Distance(miles) from core tract Distanoe(miles) from core tract
FIGURE 4
(CONTINUED)
FEA hinterland) benefited little from metro FEAs is consistent with earlier studies of
growth. Hinterland population density was the influence of growth centers on sur-
stagnant in four of the FEAs and declined rounding rural areas. We do not propose,
in two others. Little regional population however, that our findings are necessarily
deconcentration was evident among the representative of urban-to-rural spread ef-
eight economic areas. fects in all economic regions. With the ex-
The lack of extensive urban-to-rural ception of Charlotte, the metro areas in our
study are relatively small. Urban disec-
spread effects in our eight Southeastern
1980 1990
Variable Coefficient t-value Coefficient t-value
APPENDIX TABLE 1
(CONTINUED)
Charleston FEA
1980 1990
Charlotte FEA
1980 1990
Columbia FEA
1980 1990
Florence FEA
1980 1990
Variable Coefficient t-value Coefficient t-value
APPENDIX TABLE 1
(CONTINUED)
Greenville FEA
1980 1990
Myrtle B
1980 1990
Savannah FEA
1980 1990
aThe numbe
is a dummy
following kn
bN= the num
APPENDIX TABLE 2
CUBIC-SPLINE REGRESSION RESULTS FOR REGIONAL POPULATION DENSITY
FUNcrIONS, TRACTS WITH INTERSTATE HIGHWAYS VERSUS NO INTERSTATE,
CHARLOTTE, CHARLESTON, COLUMBIA, AND GREENVILLE, 1980 AND 1990
1980 1990
Charlotte: Non-
1980 1990
Charleston: I
1980 1990
Charleston
1980 1990
APPENDIX TABLE 2
(CONTINUED)
1980 1990
Columbia: N
1980 1990
Greenville: I
1980 1990
Greenville: N
1980 1990
aRefer to Append
bThe number of t
tracts in the FEA because both sets included the same tracts at the nodal center.
This content downloaded from
103.200.32.12 on Thu, 16 Nov 2023 06:36:23 +00:00
All use subject to https://about.jstor.org/terms
356 Land Economics August 1996
McDonald, J., and H. W. Bowman. 1976. metropolitan "Some Areas, ed. D. L. Barkley. Boulder,
Tests of Alternative Urban PopulationCO: Westview Press.
Den-
sity Functions." Journal of Urban Economics Schmitt, B. Forthcoming. "Advantages Compara-
3:242-52.
tifs, Dynamique de Population et Dynamique
McGranahan, D. A., and J Salsgiver. 1993. "Re- D'Empoli Des Espaces Ruraux." Revue
cent Population Change in Adjacent Non- d'Economie Regionale et Urbaine.
metro Counties." Rural Development Perspec- Scott, A. J. 1983. "Location and Linkage Sys-
tives 8 (3):2-7. tems: A Survey and Reassessment." The An-
Mieszkowski, P., and E. S. Mills. 1993. "The nals of Regional Science 17 (1):1-39.
Causes of Metropolitan Suburbanization."Skaburskis, A. 1989. "Inversions in Urban Den-
Journal of Economic Perspectives 7 (3):135-47. sity Gradients: A Brief Look at the Vancou-
Miller, J. P. 1993. "Small and Midsize Enterprise ver Metropolitan Area's Density Profile." Ur-
Development: Prospects for Nonmetropolitan ban Studies 26:397-401.
Areas." In Economic Adaptation: AlternativesSmith, S. M. 1993. "Service Industries in the
for Nonmetropolitan Areas, ed. D. L. Barkley. Rural Economy: The Role and Potential Con-
Boulder, CO: Westview Press.
tribution." In Economic Adaptation: Alterna-
Mills, E. S. 1972. Studies in the Structure of the
tives for Nonmetropolitan Areas, ed. D. L.
Urban Economy. Washington, DC: Resources
for the Future.
Barkley. Boulder, CO: Westview Press.
Suarez-Villa, L. 1989. The Evolution of Regional
Muth, R. F. 1969. Cities and Housing: The Spatial
Economics: Entrepreneurship and Macroeco-
Pattern of Urban ResidentialLand Use. Chicago:
nomic Change. New York: Praeger.
University of Chicago Press.
. 1961. "The Spatial Structure of the ?. 1988. "Metropolitan Evolution, Sectoral
Housing Market." Papers and Proceedings of Economic Charge, and the City Size Distribu-
tion." Urban Studies 25:1-20.
the Regional Science Association 7:207-20.
O'hUallachian, B., and M. Satterwaite. 1992. Suits, D., A. Mason, and L. Chan. 1978. "Spline
"Sectoral Growth Patterns at the Metropoli- Functions Fitted by Standard Regression
tan Level: An Evaluation of Economic Devel- Methods." The Review of Economics and
Statistics 60:132-39.
opment Incentives." Journal of Urban Eco-
nomics 26 (1):19-32. Thurston, L., and A. M. J. Yezer. 1994. "Causal-
Parr, J. B. 1987. "The Development of Spatial ity in the Suburbanization of Population and
Structure and Regional Economic Growth." Employment." Journal of Urban Economics
35:105-18.
Land Economics 63 (May):113-27.
. 1985. "A Population Density Approach Waddell, P., and V. Shukla. 1993. "Employment
to Regional Spatial Structure." Urban Studies Dynamics, Spatial Restructuring, and the
22 (4):289-303. Business Cycle." Geographical Analysis 25
Rowley, T. R., and S. L. Porterfield. 1993. "Re- (1):35-52.
moving Rural Development Barriers Through White, K. J. 1978. "A General Program for
Telecommunications: Illusion or Reality?" In Econometric Methods-SHAZAM." Econo-
Economic Adaptation: Alternatives for Non- metrica 46 (1):239-40.