Agglomeration In Urban Growth

This article was downloaded by:[University of Sydney]
On: 5 September 2007
Access Details: [subscription number 778083596]
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Urban Studies
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713449163

The Agglomeration Process in Urban Growth
Joel Bergsman a; Peter Greenston a; Robert Healy a
a
The Urban Institute, Washington, D.C.

Online Publication Date: 01 October 1972
To cite this Article: Bergsman, Joel, Greenston, Peter and Healy, Robert (1972)
'The Agglomeration Process in Urban Growth', Urban Studies, 9:3, 263 - 288
To link to this article: DOI: 10.1080/00420987220080431
URL: http://dx.doi.org/10.1080/00420987220080431

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,
re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly
forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents will be
complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be
independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or
arising out of the use of this material.
© Taylor and Francis 2007

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

JOEL BERGSMAN, PETER GREENSTON AND
ROBERT HEALY

The tendency of people and economic activities
to concentrate in urban areas in the United
States is persistent and increasing . Despite pollution, congestion, high taxes and high land costs
the metropolitan area has remained the preferred
site for the fastest growing economic activities,
particularly services . Traditional locational factors such as transportation and power have
become more equally available among cities of
various sizes, but the metropolis has retained its
attraction, capitalising on its role as a rich source
of information and professional talent (Hoover
and Vernon, 1962) .
Urban growth is an agglomeration process .
Business firms find it profitable to cluster together spatially with firms in their own and other
industries . The large metropolitan area contains not only a large number of different industries, but also a large final product market, a
labour pool, good communications, and a
variety of specialised services . On the other
hand, for some activities large cities may offer
more diseconomies than economies . Thus we
view urban growth as continual agglomeration
and deglomeration of economic activities,
responding to external economies and diseconomies created by previous location decisions of
firms and individuals (Lichtenberg, 1960 ;
Chinitz, 1961) .
In this paper we report results of the first step
in a continuing research effort . The objective of

the entire project is to analyse the forces that
determine changes in size and economic structure
of metropolitan areas in the United States . We
hope to quantify the partial effects of various
forces, using multiple regression analysis applied
to all urban economic activities in over 200
metropolitan areas . In this paper we describe
sets of `industrial clusters' . These clusters,
which are groups of economic activities that have
similar locational patterns, may be regarded as
the building blocks of metropolitan economies .
We discuss alternative methods of defining
industrial clusters, describe many clusters, and
speculate on some of the forces which seem to
hold the clusters together . In further research,
we will analyse these forces more rigorously .
We also present a mutually exclusive and collectively exhaustive grouping of 186 activities on
the 2- and 3-digit level (comprising all urban
activities) into clusters .
Concepts and Methodology
Agglomerating forces
Agglomerating forces, although fundamental
to the urban growth process, are not well understood . A really compt pensive theoretical
statement has yet to be achieved, and empirical
studies are far from satisfactory . In the literature, agglomeration economies are divided into
`localisation' and `urbanisation' economies

The authors are with The Urban Institute, Washington, D .C.
263

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

264

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

(Hoover, 1937) . Localisation economies result
from the proximity of several firms engaged in
the same activity . They help explain such
specialised concentrations as jewellery in Providence, fur goods in New York, and aircraft in
Los Angeles. The study of localisation economies tells us little about why an activity gets
started in a place, but can explain much about
why some activities are concentrated in only a
few of the many places that apparently offer
similar advantages .
Urbanisation economies create two other types
of clustering . The first is the clustering of firms
in industrial complexes such as oil refining/
petrochemicals, food processing/container manufacturing, and metal working/machinery . The
linkages which hold these firms together are
often input-output flows (Isard et al., 1959) . At
times these flows are direct, with one firm serving
as the supplier or customer of another . In
other cases, two firms are linked because of their
common relationship with a third .
The second type of urbanisation economies
causes both individual firms and industrial
complexes to locate in or near large cities . For
example, over 90% of national employment in
such disparate activities as paints and varnishes,
periodicals, security and commodity brokerage,
millinery, and aircraft manufacturing is located
within metropolitan areas . Style apparel and
advertising are both concentrated in New York.
But probably they do not attract each other ;
rather each is there because of some other
characteristics of the city, such as its primacy
in communications media and its role as a
style setter for the nation as a whole (Meier,
1970) .
The attributes of metropolitan areas are more
attractive for some types of activities than for
others . Along with external economies, the
metropolis presents many disadvantages, many
of them a direct result of the clustering process
itself. Thus textile products follow two location
patterns : manufacturing of standardised goods
such as bedding and underwear is located in low
wage, non-metropolitan locations in the South
and Middle West while designing of high

quality fabrics or high fashion goods is concentrated in the largest metropolitan areas .
The various types of agglomeration economies
are not easy to distinguish in practice . Los
Angeles has higher than average employment in
aircraft, in part because aircraft is an urbanoriented industry, in part because there are many
industries in Los Angeles that trade with aircraft,
and in part because of the historical influence of
localisation economies . In general, localisation
economies tend to heighten the influence of urbanisation economies, rather than cancel them out .
Thus, an activity may be found not evenly spread
among the large cities, but highly concentrated
in the few cities where the activity got started .
Basic methodology and data
There already is a large body of research on
the localisation, clustering and urbanisation of
firms, which points the way for our own work .
Geographers have pointed out patterns of localisation (Nelson, 1955 ; Alexandersson, 1956 ;
Murphy, 1966) . Economists have speculated
about agglomeration effects, but have tended to
concentrate their work at the state or regional
level (Perloff et al., 1960 ; Fuchs, 1962) . Other
economists have tried to measure agglomeration
economies (Rocca, 1970 ; Gordon, 1971 ; Shefer,
1971), while others have formulated planning
models that include agglomeration effects (Isard
et al., 1959) . Richter (1969) and Streit (1969)
have studied clustering in manufacturing, without considering interaction with service and
other supporting industries . Most recently,
Stanback and Knight (1970) have described how
employment shares in highly aggregated sectors
vary with city size, and how some of these sectors
exhibit common locational patterns .
The work reported in this paper differs from
most earlier work in two important ways : better
data, and absence of a priori restrictions on
clustering behaviour.
The Standard Industrial Classification (SIC)
groups activities without regard to locational
behaviour . Any SIC group, therefore, may contain two or more activities that have different



Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH
locational patterns . Trying to make sense out of
the location of SIC 37, Transportation Equipment, which includes automobiles, aircraft, ships
and railroad equipment, is not a promising task .
Even SIC 372, Aircraft, includes aircraft
assembly (concentrated on the west coast) and
aircraft engines and parts (spread over the
`manufacturing belt' in the north-eastern and
north-central U.S .) . We therefore start with
highly disaggregated data . The data file used
for this paper describes the industrial structure
in 1963, using employment figures from the
Census of Manufactures, Census of Business,
County Business Patterns, Census of Governments and other sources. This file covers 144
manufacturing activities (3-digit SIC detail) and42
non-manufacturing activities (mostly 2-digit SIC
detail) . Geographically, the data cover all of the
203 SMSAs defined in 1963 in the conterminous
United States .' In future work we plan to move
to an even more detailed basic data set .
We use the structure of employment as the
measure of the structure of economic activity
because it is the only measure available at sufficiently high levels of disaggregation .
One serious data problem at the 3- and 4-digit
level, for SMSAs, is the Census Bureau's disclosure rule . This rule prevents publication of
employment data for any industry-place combination where the number of firms is small . A
small number of firms may account for thousands of jobs, particularly in manufacturing.
Fortunately, the Census Bureau releases figures
on number of plants by county and industry for
each of 7 size classes . We have combined information about the number of plants by size class
with data for mean number of employees per
plant (for each industry nationally) for the same
size classes and estimated the data which the
Census Bureau did not release . These estimates,

265

made on the 4-digit level, were then aggregated
to the 2-digit level and compared with published
2-digit data. Any difference between actuals and
estimates was then distributed across the 4-digit
estimates .
This estimation procedure is our solution to
the dilemma that other researchers have solved
by either aggregating activities or places . We
thus do not have to settle either for highly
aggregated industrial data (e .g . 2-digit SIC)
which inherently contain a lot of locational
`noise', or for highly aggregated geographic units
(e .g . states or multi-state regions), which are
very poor representations of local economic units .
SMSAs, while not perfect descriptions of the
coverage of local economies, are far closer than
states or multi-state regions .
The second major distinguishing characteristic
of our approach is that we describe first, and ask
questions later. We do not approach the data
with hypotheses about the nature of agglomerations . Although appealing hypotheses do exist
in the literature, we prefer a general search of the
data for whatever kinds of clustering patterns
may exist, to be followed by more rigorous
analysis of causes. This approach does not
limit the types of clusters that can be discerned,
as would be the case if, for example, we searched
only for locational patterns following inputoutput linkages.
Definition of clusters
Our basic conception of an industrial cluster
is a group of activities that tend to locate in the
same places. The most direct and simplest way
to embody this idea in a measurable criterion is
the correlation coefficient between the number
of people employed in each of two activities, the
units of observation being the SMSAs . 2 This

t For New York and Chicago, the Standard Consolidated Areas are used (aggregatioti,(bf the central
SMSA with one or more contiguous SMSAs) . For several New England cities, where SMSAs
are defined on a township rather than a county basis, we have used Metropolitan State Economic
Areas which are based on counties . Throughout this paper, we sometimes use the word 'city'
to describe metropolitan areas.
2 There are several other measures of industrial association, the best known of which is P . Sargant
Florence's 'coefficient of geographic association' (1943) . The superiority of the simple correlalation coefficient has been pointed out by McCarty (1956) and Richter (1969) .

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

2 66

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

remains our basic measure of association, but
additional measures and analyses are needed, for
four reasons .
First, using pair-wise correlations to describe
agglomerations is unwieldy, and imposes an
arbitrarily simple structure on the data . Firms
in a dozen or more different activities may have
a common location pattern, which can be interpreted only when we can view them as a group or
cluster . Moreover, pair-wise correlations among
186 activities produces a matrix of over 17,000
individual correlation coefficients, which is
hardly convenient for hypothesis testing . A
technique is needed to define industrial clusters
containing any number of activities, and to
separate those cities where a particular cluster is
important from those where it is not .
Factor analysis provides the necessary technique . Factor analysis can be viewed as a
method that reduces the dimensionality of a given
n-dimensional space, with minimum loss of information. Each of our 186 different activities
can be seen as a variable ; together these 186
variables define a 186-dimensional space . Factor analysis is an efficient computational tool for
creating a smaller number of new variables which
(given their number) reproduce as much of the
information (i.e . variance) present in the original
186 dimensions as is possible . The new variables
created by the factor analysis (called `factors')
can be arranged ('rotated') so that each factor is
highly correlated with a few of the original variables, and not highly correlated with the others .
Each factor thus identifies a group of the original
variables which are more highly correlated
among themselves than with the other variables
-in our case, an industrial cluster . An index
of the importance of each observation (SMSA)
for each factor can also be computed . These
indices, called `factor scores', measure the extent
to which each locational pattern is present in
each SMSA . (See Rummel, 1970, for a comprehensive explication of the techniques of factor
analysis .)
The second reason why correlations of total
employment are insufficient is that a few large
cities have a large share of the total national

employment in several industries, and therefore
have individualistic economic structures . These
extreme observations dominate analyses of the
industries involved . Another wa of saying this
is that the distributions of many of the variables
are severely skewed . This means that factor
analysis based on correlations of total employment reveals, among other things, the ways in
which some large cities are unique . Such analysis is less useful, however, in revealing clusters
that appear mostly in smaller cities . We have
therefore done additional factor analyses based
on correlations between per capita employment
in pairs of activities . These analyses reveal more
details about clustering in smaller cities .
Another way to hold the effects of city size
constant is, first, to estimate, for each activity,
employment as a linear function of population .
The differences between observed employment
and that predicted by these functions then
measure unusually high or unusually low levels
of each activity in each city. As a second step,
factor analysis based on these differences reveals
very sharply the singular economic structures of
the larger cities .
A third problem with using correlations of
total employment is that total employment in
most activities is highly correlated with the size
of the city (as measured by population) . Therefore, many of the variables measuring total
employment are highly correlated with each
other, and this obscures clustering behaviour not
associated with city size . Again, factor analysis
based on per capita employment solves this
problem .
The fourth problem relates not to the definition of the variables to be correlated, but to the
assumed relation between the variables . Simple
correlation coefficients are most appropriate to
linear relationships. Actual agglomeration
effects, however, may be either linear (e .g . if
caused by input-output flows) or non-linear
(e.g. if activities cluster in order to be near
information sources, or clients, then there may
be important threshold effects). To investigate
this, we experimented with analyses based on
rank-order correlation coefficients, as well as

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

with more general techniques such as the Automatic Interaction Detector (Sonquist and Morgan, 1964), which does not require any prior
specification of functional form. It turned out
that, because of the non-normality of the raw
data, the analyses based on linear correlations
between total and per capita employment
respectively had already revealed virtually all of
the clustering that the non-linear methods
showed. However, the rank-order analysis was
of some use in describing higher-order aggregations of the clusters revealed by the simple correlation coefficients .
To sum up, the results reported in this paper
are based on three factor analyses . Individualistic structures of larger cities were described by
analysis of residuals of regressions of employment on population. This analysis also revealed
a number of more common clusters . Additional
clusters, especially those appearing in smaller
cities, were taken from factor analysis based on
per capita employment . Factor analysis based on
rank-order correlations between per capita
employment described aggregations of the more
detailed clusters, and also showed how outlying
observations affected the other analyses .
Results : The Nature of Agglomerations
Our results show that clustering is an important characteristic of the locations of economic
activities in the present-day urban United States .
The patterns of these agglomerations, however,
are complex, and the clusters are not so sharply
defined that all the details of any precise description are obvious . We present our results,
therefore, as one possible way to describe a complex aspect of the world. The basic characteristics we describe are clear and unequivocal, but
the details are open to different interpretations .
The first results to be described are the individualistic economic structures of each of eight
cities . Next, we describe a few of the several

267

dozen clusters found more commonly. A
number of activities that each locate more or less
independently of all others are listed later . In
the next section we present a mutually exclusive
and collectively exhaustive aggregation of the
186 activities, which can be thought of as a reaggregation of the SIC into groups with relatively
high locational homogeneity .
Singular economic structures of eight cities
Six of the largest metropolitan areas in the
United States, and two medium-sized areas, have
highly individualistic economic structures . These
cities are New York, Chicago, Los Angeles,
Boston, Washington, D.C ., Philadelphia, Milwaukee and Akron . More precisely, while the
economies of other cities are composed of
groups of activities found throughout urban
America, these eight cities contain groups that
are not found, to any significant degree, in other
cities .
Table I shows the details of New York's
structure . Our results (using 1963 data) correspond closely to the description in the New
York Metropolitan Study of the late 1950s
(Hoover and Vernon, 1959 ; Lichtenberg, 1960),
and so are not very surprising . New York is
shown to be not only a national centre for nonmanufacturing activities such as finance, advertising, transportation services, legal services, real
estate, publishing, wholesale trade, insurance,
etc., but also a centre for manufacturing of nonstandardised goods such as certain clothing,
toys, sport equipment, pens, cosmetics, notions,
luggage, jewellery, etc. Central office employment is shown by `administrative and auxiliarymanufacturing' ; the same category for nonmanufacturing is also strongly present in New
York, appearing just below the arbitrary cut-off
we imposed for inclusion in the tables . Our
results also confirm the clear dominance of New
York in these lines, as measured by the high
factor loadings shown in Table 1 . 3

3 The square of the factor loading is equal to the locational variance explained by the factor, as a
ratio to the total variance . Thus, for the particular formulation used here, the `New York' factor
explains from 50 to 95 % of the locational variance of the residuals in 25 of the activities just
mentioned and from 10 to 50 % of roughly 25 more .

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

268

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY
Table 1
SINGULAR STRUCTURE : NEW YORK
(A) Most Important Activities
SIC
62
237
317
234
238
233
731
47
235
81
239
60
236
394
65
273
272
395
278
284
67
396
49
73,

excl.
731
274
279
84
283
50
44
226
319
72
399
316
64
48
63
225
384
364
391
86
232
333
264
387
289
76,

excl .

Name

Loading

Security and commodity brokers and exchanges
Fur goods
Handbags and personal leather goods
Women's and children's undergarments
Miscellaneous apparel and accessories
Women's and misses' outerwear
Advertising
Transportation services
Hats, caps and millinery
Legal services
Miscellaneous fabricated textile products
Banking
Children's outerwear
Toys and sporting goods
Real estate
Books
Periodicals
Pens, pencils, office and art supplies
Blankbooks and bookbinding
Soap, cleaners, and toilet goods
Investment companies
Costume jewellery and notions
Electric, gas, and sanitary services

0. 983
0972
0. 966
0. 952
0946
0 . 946
0. 945
0 .936
0927
0 .924
0 .912
0 .909
0 .893
0 .883
0 .878
0877

Miscellaneous business services, except advertising

0727

Miscellaneous publishing
Printing trade services
Museums, botanical, zoological gardens
Drugs
Wholesale trade
Water transportation
Textile finishing, except wool
Leather goods, N .E .C .
Personal services
Miscellaneous manufactures II
Luggage
Insurance agents, brokers and service
Communication
Insurance carriers
Knitting mills
Medical instruments and supplies
Electric lighting and wiring equipment
Jewellery, silverware and plated ware
Non-profit membership organisations
Mens' and boys' furnishings
Primary non-ferrous metals
Miscellaneous converted paper products
Watches, clocks and watchcases
Miscellaneous chemical products

0 .718
0 . 701
0. 697
0681
0 . 678
0. 676
0661
0 . 659
0658
0. 653
0648

Consumer-oriented repair shops and services

0 . 365

Administrative and auxiliary-manufacturing

0. 356

0869
0856
0 . 814
0 . 813
0 . 760
0731
0 .727

0629
0. 596
0591
0589
0. 579
0. 558
0. 550
0. 517
0. 471
0. 430
0 . 403
0. 391
0. 367

769

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

269

Table 1-continued
(B) Least Important Activities
SIC
Name
42
Trucking and warehousing
202
Dairy products
61
Non-bank credit agencies
327
Concrete, gypsum and plaster products
359
Miscellaneous machinery, except electrical
332
Iron and steel foundries
331
Steel
354
Metalworking machinery
299
Miscellaneous petroleum and coal products
339
Miscellaneous primary metal products
344
Fabricated structural metal products
203
Canned, cured and frozen foods
345
Screw machine products, bolts, etc.
356
General industrial machinery
382
Mechanical measuring and control devices
349
Miscellaneous fabricated metal products
346
Metal stampings
371
Motor vehicles and equipment
07
08
Agricultural services, forestry and fisheries
09
325
Structural clay products
379
Miscellaneous transportation equipment
40
Railroad transportation
374
Railroad equipment
19
Ordnance
Automobile repair and rental
75
322
Glass and glassware, pressed or blown
Grain mill products
204

Also interesting is New York's singular lack
of certain activities . Many metalworking and
metal products activities are relatively underrepresented, as are a variety of other activities.
Remember that `New York' includes counties in
New Jersey, Connecticut, and New York state
outside of the city proper, where high land prices
and other barriers to such activities are not so
important as in New York City itself.
The Chicago area, by contrast to New York,
is specialised in many metalworking activities
and other `heavy manufacturing' . (See Table 2 .)
The unique Chicago factor does not show a
relatively high amount of service, trade, or other
market-centre activities . This conclusion is
somewhat modified by Chicago's rank of tenth
of the 203 cities in employment per capita in a
more widespread market-centre factor (discussed
below) . We conclude that Chicago is one of
many important market centres, but does not
approach New York's unique concentration of
market centre functions .

Loading
-0.746
-0. 617
-0. 607
-0568
-0. 517
-0 . 505
-0 .490
-0. 466
-0 . 453
-0 .450
-0 .436
-0424
-0 .415
-0 .400
-0394
-0 . 393
-0 . 386
-0 . 385
-0 . 383
--0 . 38!
-0 .366
-0 . 365
-0-362
-0-359
-0. 353
-0. 345
-0-344

Los Angeles' singular economic structure is
characterised by motion pictures, ordnance, and
aircraft, as is well known, and also by a number
of other activities shown in Table 3 . The appearance of SIC 61, which includes consumer
finance companies, may be an example of the
oft-observed tendency of Los Angelenos to lead
the nation in innovative behaviour-in this case,
personal indebtedness. The high score of automobile repairing confirms another aspect of Los
Angeles' popular image, as does the low loading
on SIC 41, which includes bus services, etc .
Boston, the fourth highly individualistic city,
is characterised by footwear manufacturing,
medical, .and educational services, and electronic
components (Table 4) . As with most cities with
singular economic structures ; ,the results confirm
and quantify existing impressions about the city .
Philadelphia has a singular concentration of
certain textile and clothing activities, petroleum
products and plastics, and a few other activities
(Table 5). Washington, D .C., not surprisingly,




Downloaded By: [University of Sydney] At: 04:45 5 September 2007

270

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

Table 2
SINGULAR STRUCTURE : CHICAGO

(A) Most Important Activities
SIC
Name
Radio and television receivers
365
393
Musical instruments
334
Secondary non-ferrous metals
275
Commercial printing
348
Miscellaneous fabricated wire products
307
Miscellaneous plastic products
349
Miscellaneous metal products
374
Railroad equipment
207
Confectionery
295
Paving and roofing materials
341
Metal cans
285
Paints and allied products
40
Railroad transportation
Miscellaneous manufactures II
399
281
Industrial inorganic and organic chemicals
Partitions and fixtures
254
Metal stampings
346
265
Paperboard containers and boxes
363
Household appliances
345
Screw machine products, bolts, etc .
249
Miscellaneous wood products
274
Miscellaneous publishing
347
Miscellaneous metal services
209
Miscellaneous food preparations
353
Construction and related machinery
Local and interurban passenger transit
41
259
Miscellaneous furniture and fixtures
Grain mill products
204
42
Trucking and warehousing
Medical instruments and supplies
384
279
Printing trade services
364
Electrical lighting and wiring equipment
336
Non-ferrous foundries
387
Watches, clocks, and watchcases
52-59
Retail trade
Administrative and auxiliary-non-manufacturing
264
Miscellaneous converted paper products
84
Museums, botanical, zoological gardens
205
Bakery products
331
Steel
278
Blankbooks and bookbinding
398
Miscellaneous manufactures I
339
Miscellaneous primary metal products
276
Manifold business forms
358
Service industry machines
Special industry machinery
355
(B) Least Important Activities
Automobile repair and rental
75
76,
excl.
Consumer-oriented repair shops and services
769
89
Miscellaneous services
07 )
08
Agricultural services, forestry and fisheries
09 )}
67
Investment companies

Loading
0. 881
0. 860
0. 854
0816
0 . 806
0. 789
0 . 770
0. 736
0 . 724
0-722
0-718
0. 717
0. 704
0. 685
0. 671
0. 667
0 . 626
0-608
0. 591
0. 590
0. 582
0. 576
0. 570
0. 539
0. 537
0. 519
0 . 515
0-515
0. 491
0. 480
0 .466
0. 459
0442
0. 433
0. 423
0-421
0. 418
0. 411
0. 407
0 . 400
0-397
0 . 384
0-376
0-374
0 . 370
0. 363

-0-480
-0 . 387
-0 . 366
-0. 364
-0. 336

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

SINGuLAR

Table 3
Sraucruity Los

ANGELES

(A) Most Important Activities

sic

Name
Motion pictures
Nonclassifiable establishments
Ordnance
Miscellaneous transportation equipment
Automobile repair and rental
Household furniture
Aircraft
Miscellaneous services

78
99
19
379
75
251
372
89
76,
excl .
769
61
48
359
73
c. 1 .
731
52-59
769
15
16
17
249
79
07
08
09
347

1

Loading
0. 914
0. 859
0.789
0.779
0 .688
0 .668
0 .633
0 .609

Consumer-oriented repair shops and services

0.603

Non-bank credit agencies
Communication
Miscellaneous non-electrical machinery

0. 591
0. 565
0. 548

Miscellaneous business services, except advertising

0. 515

Retail trade
Miscellaneous repair services

0. 514
0.494

Construction

0.460

Miscellaneous wood products
Amusement and recreation services

0.400
0.353

Agricultural services, forestry and fisheries

0.345

Miscellaneous metal services

0 .344

(B) Least Important Activities
Administrative and auxiliary-manufacturing
49
Electric, gas, and sanitary services
41
Local and interurban passenger transit
279
Printing trade services
67
Investment companies
86
Non-profit membership organisations
331
Steel
Table 4

-0 .567
_0.406
-0 .398
-0 .376
-0.354
-0.348
-0.344

SINGULAR STiLucruRE : BOSTON

Most Important Activities

sic
313
314
93
311
302
82
80
229
367
381
223
271
277
306
41
355
207
328
1 8

Name
Footwear cut stock .
Footwear, except rubber
Local government
Leather tanning and finishing
Rubber footwear
Educational services
Medical services
Miscellaneous textile goods
Electronic components
Scientific instruments
Weaving and finishing mills, wool
Newspapers
Greeting card publishing
Fabricated rubber products, N.E.C.
Local and interurban passenger transit
Special industry machinery
Confectionery
Cut stone and stone products

Loading
0-905
0. 889
0. 871
0 .850
0 .781
0 .735
0 .685
0.685
0 .511
0.504
0-456
0.408
0.401
0. 401
0. 401
0.377
0. 354
0-339





Downloaded By: [University of Sydney] At: 04:45 5 September 2007

272

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY
Table 5
SINGULAR STRUCTURE : PHILADELPHIA

(A) Most Important Activities
SIC

231
299
225
227
312
382
202
282
232
228
205
361
66

Name

Loading

Men's and boys' suits and coats
Miscellaneous petroleum and coal products
Knitting mills
Carpet mills
Industrial leather belting
Mechanical measuring and control devices
Dairy products
Plastic materials and synthetics
Men's and boys' furnishings
Yarn and thread mills
Bakery products
Electric test and distributing equipment
Combined real estate, insurance, loan, law offices

0. 710
0 . 610
0. 507
0. 502
0.458
0 .447
0-431
0 .407
0 .405
0 . 392
0 . 380
0 .341
0 .339

(B) Least Important Activities

209

Miscellaneous food products

-0. 380

Table 6
SINGULAR STRUCTURE : WASHINGTON, D.C .

(A) Most Important Activities

91
86
15
16
17
82
271

Federal government
Non-profit membership organisations

0 .847
0 .676

Construction

0 . 398

Educational services
Newspapers

0 .371
0-359

(B) Least Important Activities

371
354

Motor vehicles and equipment
Administrative and auxiliary-manufacturing
Metalworking machinery

is outstanding in federal government activities,
non-profit membership organisations (this includes churches, civic and trade associations,
unions, etc.), educational services, and newspapers (Table 6) . Concentrations of -mployment in newspapers, relative to city size, do not
appear in New York or other large cities, but
rather in a number of state capitals and in
Washington.
We now come to the two medium-sized cities
that exhibit singular economic structures. The
results are shown in Tables 7 and 8 . Milwaukee

-0 .382
-0-356
-0.355

is notable for a number of metal products ;
Akron for tyres and rubber products .
Explanations of why these eight cities have
such individualistic economic structures must
await further research . We can, however,
mention some hypotheses . The singular structures are the results of concentrations of large
parts of several entire national industries in just
one city . These concentrations, in turn, may
occur only in industries that have two characteristics : first, they have something like a national
market (i .e . low transport costs relative to value

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

273

Table 7
SINGULAR STRUCTURE : MILWAUKEE

Most Important Activities
SIC

375
362
351
208
339
353
315
332

Name

Loading

Motorcycles and bicycles
Electrical industrial apparatus
Engines and turbines
Beverages
Miscellaneous primary metal products
Construction and related machinery
Leather gloves and mittens
Iron and steel foundries

0816
0 . 779
0686
0 . 579
0493
0 .479
0 .419
0419

Table 8
SINGULAR STRUCTURE : AKRON

Most Important Activities
SIC

301
303
306
344

Name

Loading

Tyres and inner tubes
Reclaimed rubber
Fabricated rubber products, N .E .C.
Fabricated structural metal products

of product), and second they have either great
scale economies internal to the plant or great
localisation economies. Localisation economies
may be much more important than internal scale
economies . (Localisation economies are those
resulting from concentrations of many similar
firms .) For such activities, inertia may be crucial
-they tend to grow fastest where they already
are, and thus may remain concentrated in a few
places, or even only one place, long after the
reasons they got started in those places have
disappeared .
More widespread clusters
There are several dozen pairs or groups of
industries which are found together in more than
one city, or, as is usually the case, across the whole
system of cities . The most important of these, in
terms of the number of industries involved and
the amount of variance in total employment
explained, is a group of what we might call
, market-centre' activities. (See Table 9 .) These
industries, almost exclusively non-manufacturing, include business services, legal services, miscellaneous services (which includes data processing), real estate, wholesale trade, banking,

0 . 925
0. 911
0. 703
0 . 335

and several others . The market centre industries
are most important in cities that are either national or regional capitals (Atlanta, San Francisco,
New York) or that service large and prosperous non-urban hinterlands (Charlotte, North
Carolina ; Des Moines, Iowa). Of the twenty
largest SMSAs, only Buffalo (New York),
Detroit and Pittsburgh fail to rank in the top
third among all 203 SMSAs in specialisation in
these activities.
The market centre cluster appeared in every
factor analysis we performed, regardless of
whether the employment measure was per capita,
total, or the residual unexplained by a linear
function of population . The cluster is revealed
essentially unchanged using rank-order rather
than the more usual product-moment correlation
coefficients .
Identification of other clusters, however,
depends on the measure of association used .
Take, for example, the classic industrial complex
of petroleum and chemicals . Cities with high
per capita employment in these industries .
among them Houston, Beaumont (Texas) and
New Orleans, also have high per capita employment in water transportation and shipbuilding .
When residuals from total employment are used .



Downloaded By: [University of Sydney] At: 04:45 5 September 2007

274

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY
Table 9

MARKET CENTER CLUSTER

sic
73,
excl.
731
81
89
65
50
60
64
731
63
72
15
16
17
86
61
41
75
52-59
275
62
279
45
271
82
78
76,
excl .
769

Name

Loading

Miscellaneous business services, except advertising

0. 780

Legal services
Miscellaneous services
Real estate
Wholesale trade
Banking
Insurance agents, brokers and service
Advertising
Insurance carriers
Personal services

0 . 752
0 .745
0. 733
0.699
0 . 686
0 . 651
0 . 639
0603
0. 578

Construction

0 .561

Non-profit membership organisations
Non-bank credit agencies
Local and interurban passenger transit
Automobile repair and rental
Retail trade
Commercial printing
Security, commodity brokers and services
Administrative and auxiliary-non-manufacturing
Printing trade services
Air transportation
Newspapers
Educational services
Motion pictures

0 . 537
0 .530
0 -526
0 .509
0 .505
0. 412
0. 409
0 .396
0.395
0. 374
0. 355
0. 355
0. 349

Consumer-oriented repair shops and services

0 . 340

however, this larger grouping is not revealed .
Petroleum and chemicals are shown to be associated strongly with each other, as are water
transportation and shipbuilding . The four
industries, however, are not associated in a
single group when total employment residuals
are used to measure location .
This phenomenon is very common in our
analysis . In general it represents the consequences of imposing a structure, no matter how
sophisticated, on a complex reality . By using
different measures and several factor analyses,
we describe this reality more fully than we would
by using a single criterion . If an industry must
reach some minimum size before it produces
sufficient external economies to attract (and
cluster with) another industry, we may find a
pattern in total employment which is not found
at all in per capita employment. Similarly,

analysis of per capita data reveals patterns of
clustering that exist only in smaller cities, and
that get swamped by the variations among large
cities when we look at the distribution of total
employment.
We now describe several other clusters that
are revealed by most or even all of our variations
in the way we defined employment, in the measure of association and in the number of factors
rotated . Most are composed of industries that
either produce similar products or that have
very strong input-output ties .
One of these clusters is composed of cigarettes,
tobacco stemming and redrying, and chewing
and smoking tobacco . These industries are
heavily concentrated in some of the cities of the
principal tobacco-growing region-Richmond
(Virginia), Durham and Winston-Salem (North
Carolina) and Louisville and Lexington (Ken-

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

tucky) . There is some tendency for knitting
mills, miscellaneous_ textile products and plastics
to cluster with the tobacco industries, but the
association is probably accidental, as the
industries neither trade with one another nor
appear to have the same locational needs .
Another cluster apparently oriented to natural
resources is made up of forest products industries
including logging, sawmills, veneer and plywood,
and miscellaneous wood products . These industries are important in terms of total employment in Eugene and Portland (Oregon) and
Seattle (Washington) . In addition, they provide
an important share of the economies of several
small Southern cities, such as Pine Bluff (Arkansas) and Asheville (North Carolina).
Footwear, footwear cut stock (soles and uppers
for shoes) and wool weaving mills compose
another cluster . Worcester and Brockton
(Massachusetts), Binghampton (New York) and
Lewiston (Maine) have substantial employment
in these industries, and despite their modest size
this cluster appears in the analysis based on total
employment residuals as well as in that based on
per capita employment . The dominant city on
this cluster, however, is Boston, which contains
about 20 % of the nationwide urban employment
in these industries .
These industries are located together for
historical reasons, reinforced perhaps by the
availability of low-wage labour . The historical
attraction of the textile industry and the shoe
industry to New England's water power and,
later, to its immigrants is no new discovery . The
various factor analyses reveal some other patterns that are not so obvious. For example,
rubber footwear is rather strongly represented
in Boston and Providence but in almost none of
the smaller towns where employment in other
branches of the shoe industry is important . We
might speculate that it was only in the largest of
the traditional shoemaking cities that entrepreneurs could shift to making a type of shoe
that required entirely different materials and
manufacturing techniques.
Another interesting pattern that emerges from
the factor analysis is the tendency of electronic
us 9r

1 8 r

275

components manufacturing to associate somewhat with the shoe/textile cluster . This is true
not only in Boston but in Manchester (New
Hampshire), Providence, Brockton, Lewiston
and elsewhere in New England . Again, we
might speculate that components manufacturers
have been attracted to the cheap, factory-trained
labour that a declining shoe industry has been
releasing .
Another cluster that appears quite consistently
is composed of hotels and recreation services .
Also associated with this cluster, although not to
the same degree in every factor analysis, are
retail trade, personal services and air transportation, along with a group of activities generally
associated with rapid population growthconstruction, concrete and real estate . Needless
to say, this cluster is most important in resort
cities, such as Miami, Las Vegas, Reno, Atlantic
City (New Jersey) and Phoenix (Arizona) .
Many of our 186 industries are not clear and
unequivocal members of one and only one
cluster. This should not be surprising . Most
industries are subject to a variety of locational
forces . Some textile firms seek out low-wage
labour in the South ; others are bound by inertia
to New England locations . The metal can industry locates partly in places where the food
canning industry is important, partly in certain
smaller cities that are centres of metalworking
and machine building, and disproportionately
(18% of nationwide urban employment) in
Chicago. This variety of locational patterns
makes it impossible to describe completely the
industry's locational behaviour by placing it in
one well-defined industrial cluster .
A cluster of metalworking and machine building industries, composed of screw machine
products (nuts and bolts), cutlery, metalworking
machinery, general industrial machinery, industrial leather belting, mechanical measuring
instruments, metal cans, and special industry
machinery, is a close approximation to the metalworking complex which is often mentioned as a
possible industrial base for growth centres .
Employment in these industries makes up an
important part of the economies of Rockford



Downloaded By: [University of Sydney] At: 04:45 5 September 2007

2 76

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

(Illinois), Hartford (Connecticut), Lorain (Ohio)
and Worcester (Massachusetts), but the bulk of
total employment is in large cities such as
Chicago, Detroit and New York .
The industries in this cluster show a marked
concentration in the manufacturing belt .' Not
a single city outside this region had any major
degree of specialisation in these activities .
Interestingly, the cities where these metalworking industries are most important show no
specialisation in either steel production (a major
supplier) or motor vehicles (a major customer).
This finding tends to confirm the conclusion of
an Italian study that :
'the presence of a new iron and steel centre
does not automatically give rise to new industries but only to isolated units such as
cement manufacture from slag, etc. As for
the attraction of an iron and steel works to
mechanical engineering industries, we have
seen these tend basically to be set up in areas
where the intermediate and auxiliary industries
indispensable to them already exist . Modem
conditions of efficiency and competitiveness,
as may be relied upon by industries in the
vicinity of the necessaryy auxiliary and subsidiary units, largely offset the additional cost
of transport of iron and steel products . . .'
(EEC, 1956) .
Many of the specialised services that the other
machine building activities require--the making
of jigs, moulds and dies, and rebuilding of
machine tools-are included in the 'metalworking machinery' industry . 'Other services, how-

ever, such as machine shops and electroplating,
are in industries that show very little association
with this cluster.
Although steel and motor vehicles do not tend
to locate in cities where machine building is
important, the concentration of machine building employment in the manufacturing belt
indicates that interactions may cross SMSA
boundaries. In later work, we plan to introduce
explicitly the influence of activities in other,
nearby, SMSAs .
Wages in the industries composing the metalworking cluster average about 13% above the
mean for manufacturing as a whole . The impact
of these higher wages is illustrated by the fact
that manufacturing wages in the ten cities scoring highest on this factor are 15% higher than
the national average.
If the metalworking/machine building cluster
is an example of how a favourable industry mix
can raise local wages, a `low wage/apparel'
cluster shows the opposite kind of orientation .
The cluster, as identified by the per capita factor
analysis, is shown in Table 10 . Women's and
children's undergarments and men's and boys'
furnishings also load high on this factor . These
activities have wages more than 30 % below the
national average for all manufacturing . Characterised by low skill and a high labour input,
they are most important in the economies of
Wilkes-Barre, Scranton, and York (Pennsylvania), Lima (Ohio) and Tampa (Florida). In
total employment, however, these industries (with
the exception of cigars and men's furnishings)
are overwhelmingly concentrated in New York .

Table 10
IoW
SIC
212
233
395
236
238
4

WAOEIAPPAREL CLUSTER

Name
Cigars
Women's and misses' outerwear
Pens, pencils, office and art supplies
Children's outerwear
Miscellaneous apparel

Loading
0. 832
0. 745
0-666
0-444
0-334

The manufacturing belt is a group of contiguous counties characterised by high value added in
manufacturing. The belt covers most of the states of Massachusetts, Connecticut, Rhode Island,
New York, New Jersey, Pennsylvania, Maryland, Ohio, Michigan, Indiana, and Illinois and parts
of Maine, New Hampshire, Delaware, West Virginia, Iowa and Wisconsin (Pred, 1965) .

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

Two of the industries in the 'low wage/
apparel' cluster, children's outerwear and
women's undergarments, are also associated
with another group of low or low-to-average
wage industries shown in Table 11 . Cities most
specialised in the industries in this cluster include
Lancaster (Pennsylvania), Springfield (Massachusetts), Laredo (Texas) and Little Rock
(Arkansas).
Children's outerwear and women's undergarments also are associated with a factor which
describes the attraction of some branches of the
textile and apparel industries to the low wage,
non-union labour of the small and medium
sized SMSAs in the South, as shown in Table 12 .
The greatest degree of specialisation in these
industries occurs in Greenville (South Carolina),
Asheville and Greensboro (North Carolina) and
Augusta (Georgia) .
A number of SMSAs are engaged in activities
linked in some way to agriculture. Canning,
sugar refining, services to agriculture and the
manufacture of metal cans are associated, with
the greatest specialisation occurring in Stockton
(California), West Palm Beach (Florida) and
Savannah (Georgia).

An association was found among three members of the electrical goods industry : electric test
and distributing equipment, communication
equipment and electronic components. Cities
specialising in these activities include Pittsfield
(Massachusetts), Springfield (Illinois), Lynchburg (Virginia) and Anaheim (California) .
Specialisation in these industries shows less
correlation with office machines and computers,
and very little correlation with specialisation in
other electrical products, such as radio and
television receivers.
The richness of information revealed by the
various factor analyses, and some of the difficulties of interpretation, can be seen by examining the patterns of employment in the steel
industry. Steel employment is most highly
concentrated in Pittsburgh and Chicago, each of
which has over 15% of national employment in
the industry . The cities most specialised in
making steel, however, are mainly smaller
places, such as Steubenville and Youngstown
(Ohio) and Gadsden (Alabama) . This specialisation is often accompanied by a specialisation in
the manufacture of pottery and ceramic products, which include such steel-using items as

Table 11
LABOUR INTENSIVE CLUSTER

SIC

398
235
387
329
334
236
367
234

Name

Miscellaneous manufactures I
Hats, caps and millinery
Watches, clocks and watchcases
Miscellaneous non-metallic mineral products
Secondary non-ferrous metals
Children's outerwear
Electronic components
Women's and children's undergarments

Loading
0 . 846
0. 713
0 .609
0.534
0. 524
0. 509
0.442
0. 435

Table 12
SOUTHERN TEXTILE CLUSTER

SIC

222
226
221
227
236
355
234
232

277

Name

Weaving mills, synthetics
Textile finishing, excluding wool
Weaving mills, cotton
Carpet miffs
Children's outerwear
Special industry machinery
Women's and children's undergarments
Men's and boys' furnishings

Loading
0 .900
0 . 800
0 . 707
0 . 599
0451
0. 393
0. 339
0. 337



Downloaded By: [University of Sydney] At: 04:45 5 September 2007

278

JOEL BERGSMAN, PETER GREENSTON AND ROBERT MEALY

bathtubs and toilets . To a much lesser extent
steel cities also specialise in structural clay products, which include the firebricks used to line
blast furnaces and steel furnaces . Steel is also
associated with the glass and glass bottle
industries, particularly in Pittsburgh, which has
the highest total employment in each of these
industries . At least part of this clustering of
steel and glass may be due to a common attraction to low-cost fuel sources . There is little
tendency for steel to be located in the same cities
as many major steel-using activities, such as
machinery, motor vehicles, shipbuilding, or
metal fabricating . Naturally, the large industrial
cities such as Chicago and Detroit have substantial employment in steel as well as in the
steel-using industries . Among the specialised
machine building cities, however, there are as
many with little or no steelmaking (Rockford,
Illinois ; Flint, Michigan) as there are with a
larger than average share (Lorain, Ohio ;
Worcester, Massachusetts) . Conversely, some
of the steel-making cities, such as Provo (Utah)
and Steubenville (Ohio) have developed very
little employment in industries which use steel .
A handful of activities, including aircraft and
ordnance, have location patterns almost completely independent of those of any other industries . In the case of aircraft, we might
speculate that internal scale economies and
localisation economies combined with wartime
dispersion policy and the relative ease of shipment of the inputs and outputs to produce the
concentrations of employment in such disparate
places as Los Angeles, Seattle and Wichita
(Kansas) . For ordnance, political rather than
market forces may have determined many of the
locations .
The preceding discussion relates to clustering
on a level of aggregation where all urban economic activity is divided among two or three dozen
clusters plus perhaps a dozen independent industries. Clustering can, of course, be analysed
and described on any level of aggregation . As
an example, we will describe briefly one higher
level, where there are about ten larger clusters
plus about a dozen independent activities .

On this more aggregate level, the most important cluster consists of manufacturing activities ; metal products, machinery and equipment
and precision instruments are the outstanding
groups . The second most important cluster is
the market-centre group, and the third is a
textile-clothing agglomeration . Other clusters,
which are less important than the first three, are
a group of food products excluding food canning,
petroleum/chemicals/port activities, wood products, food canning, tobacco products, and
paper products . Although these groups can be
described in terms similar to one- and two-digit
SIC groups, they are only rough approximations
of the SIC . Even at this level it is possible to
construct a classification in which each group is
far more locationally homogeneous than is the
SIC.
Results : Mutually Exclusive Clusters
Industrial clusters, which the factor analyses
help us to define, are to be used in two different
ways in subsequent research. One use is to
analyse clustering as such : What are the nature
of industrial agglomerations, and what forces
link the agglomerating activities? The other use
is as a description of the whole urban economic
structure, for analysis of that structure. We will
test hypotheses about what characterises a city
that has a given concentration of certain activities, and the clusters should serve as dependent
variables in that analysis .
These two uses of the clusters do not have the
same requirements . The principal conflict is
whether the clusters should be mutually exclusive groups of activities . The regression
models mentioned above would be much more
useful if the dependent variables could be `added
up' to form a total description of a city's economic structure . This requires that each activity
be put into one, and only one, cluster . However,
some activities have more-or-less equally strong
links to more than one cluster . For analysing
agglomerative behaviour, such activities should
be included in all relevant clusters ; for the forthcoming regression analysis, we have decided to

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

classify them as separate clusters, which will of
course be somewhat correlated with the clusters
with which they are linked.
To form mutually exclusive clusters for later
use in the regression analysis, we face two
additional important choices . The first relates
to what we want the regression analysis to
explain . The factor analyses indicated that
clustering behaviour measured with total employment (which emphasises what is taking
place in the larger cities) differs somewhat from
clustering measured with per capita employment
(which emphasises clustering in smaller cities) .
Thus, if we wish to explain the distribution
among cities of total employment in an
industry, we will use somewhat different clusters
than we will if we wish to explain variations
in the degree of specialisation of cities in that
industry .
The second choice we must make relates to
the degree of `tightness' of the clusters and the
number of clusters formed . The analyst must
choose between . on the one hand, a large
number of small clusters in which the locational
patterns of activities are very similar, or on the
other hand, a smaller number of larger clusters,
which include activities with somewhat less
similar locational patterns. For ease of analysis
and presentation, we want to aggregate the 186
activities to a fairy small number, but the more
we aggregate the more noise we introduce into
the clusters .' The factor analyses do not present
us with a single, unambiguous set of clusters, but
show us the tradeoffs we might make between
the number of clusters and their internal tightness.
Actually assigning industries to the appropriate
cluster involves some further choices . Most
S

6

279

activities are unambiguously associated with one
and only one cluster, by very high factor loadings on one factor and (necessarily) low loadings
on the rest . Other activities are associated most
strongly with a factor with which no other activity is closely identified . Some of these are real
'loners'-they have only very low factor loadings
on factors with which other activities are strongly
associated . Others are only partly loners-they
are most closely identified with their own factor,
but are also identified with another factor that
is in turn strongly identified with other activities .
In this last case we had to decide whether to
treat these industries as separate 'independent
locaters' or place them in a larger, looser, cluster
with several other activities .
The list that follows describes one possible
aggregation of our 186 industries, in this case
based on the analysis of per capita employment. 6
This aggregation resulted in 42 clusters, six
industries with strong links to more than one
cluster, and 21 independent locaters .
The
correlation of each industry with its cluster
provides a measure of each cluster's 'tightness' .
The fact that we formed 69 clusters rather than
a larger or smaller number was somewhat
arbitrary, and represents one judgement as to
an appropriate compromise between the number
of clusters and their internal consistency .
In the list in Table 13, two numbers appear to
the left of the name of each activity . The first,
in parentheses, is the simple correlation of the
activity with the cluster as a whole (both
measured by per capita employment) . The
second is the SIC number of the activity .
The names of the clusters are for short identification purposes only, and in some cases may
not describe the cluster very accurately .

Whatever level of aggregation we choose, the activities in each of our clusters will be far more
similar in their locational behaviour than the activities at a comparable level of aggregation in an
SIC sector.
In making the mutually exclusive clusters, we first left each activity in the factor on ) ch it had the
highest loading. Three modifications were then made : (1) An activity with a loading greater than
0. 33 on more than one factor was defined to be a separate cluster . This eliminated activities from
a particular cluster if any other factor accounted for as much as 10% (0. 332) of its variance. (2)
Certain activities which loaded highest on a factor where no other activity loaded highest were
assigned to the cluster where they had their second-highest loading, providing that the secondhighest loading was at least 0. 33. (3) Two clusters containing disparate types of activities with
similar locational patterns were each split into two .







Downloaded By: [University of Sydney] At: 04:45 5 September 2007

"SO

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY
Table 13
MUTUALLY EXCLUSIVE CLUSTERS

Cluster 1 : Metal Products I
312
Industrial leather belting (transmission belting, leather mechanical packings)
(0 . 47)
(0 . 82)
342
Cutlery, hand tools and hardware (edge tools, locks, razor blades)
345
Screw machine products, bolts, etc . (rivets, screws, washers)
(0 . 86)
(0-87)
354
Metalworking machinery (machine tools, dies)
General industrial machinery (pumps, ball bearings)
(0 . 77)
356
Highest per capita employment in : Rockford, III. ; Hartford, Conn . ; Worcester, Mass. ; Trenton, N.J . ; Lorain, Ohio .

Cluster 2 : Metal Products 11
(0 .49)
259
Miscellaneous furniture and fixtures (venetian blinds, restaurant furniture)
332
Iron and steel foundries
(0-84)
(0 .46)
348
Miscellaneous fabricated wire products (barbed wire, coathangers)
351
Engines and-turbines
(0-83)
(0. 76)
359
Miscellaneous non-electrical machinery (piston rings, machine shops)
Toys and sporting goods
(0-56)
394
Highest per capita employment in : Muskegon, Mich. ; Tyler, Texas ; Saginaw, Mich. ; Lynchburg, Va . ; Tuscaloosa, Ala . ;
Peoria, 111 .

Cluster 3 : Metal Products 111
(0 . 63)
254
Partitions and fixtures (shelving, wood and metal display cases)
(0 . 73)
325
Structural clay products (bricks, ceramic tile)
339
Miscellaneous primary metal products (ferrous and non-ferrous forgings, metal powder)
(0-79)
Highest per capita employment in : Canton, Ohio ; Terre Haute, Ind . ; Worcester, Mass. ; Augusta, Me. ; Lansing, Mich . ;
Muncie, Ind .

Cluster 4 : Precision Instruments I
336
Non-ferrous foundries
(0. 51)
(0 . 88)
369
Miscellaneous electrical equipment (batteries, X-ray equipment)
382
Mechanical measuring and control devices (gas meters, thermostats)
(0 , 85)
Optical instruments and lenses (microscopes, spectrometers)
(0 . 55)
383
Highest per capita employment in : Ann Arbor, Mich. ; Bay City, Mich . ; Rockford, Ill. ; Muncie, Ind . ; Rochester, N .Y .

Cluster 5 : Precision Instruments II
(0.39)
384
Medical instruments and supplies (dressings, hearing aids)
(0. 81)
385
Ophthalmic goods (eyeglass frames and lenses)
(0. 98)
386
Photographic equipment and supplies
Highest per capita employment in : Rochester, N .Y. ; Binghampton, N.Y. ; Worcester, Mass . ; Boston, Mass . ; Erie, Pa .

Cluster 6: Electrical and Electronic
(0 . 52)
357
Office machines and computers
(0 . 57)
361
Electric test and distributing equipment (transformers, switchgear)
(0 . 85)
366
Communication equipment (telephone equipment, radio-TV transmitters)
(0 . 57)
367
Electronic components (semiconductors, antennas)
Highest per capita employment in : Pittsfield, Mass . ; Binghampton, N.Y. ; Anaheim, Ca. ; Lynchburg, Va . ; Manchester,
N .H .













Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

281

Table 13-continued
Cluster 7 : Miscellaneous Intermediates I
329
(0. 99)
Miscellaneous non-metallic mineral products (asbestos products, abrasives, mineral wool insulation)
(042)
334
Secondary non-ferrous metals
Highest per capita employment in : Albany, N .Y. ; Lancaster, Pa . ; Worcester, Mass . ; Trenton, N .J . ; Buffalo, N .Y .

Cluster 8 : Miscellaneous Intermediates II
(0 . 59)
326
Pottery and related products (china, toilets, porcelain electrical supplies)
(0. 99)
331
Steel
Highest per capita employment in : Steubenville, Ohio ; Youngstown, Ohio ; Gadsden, Ala . ; Provo, Utah ; Pueblo, Colo. ;
Pittsburgh, Pa.

Cluster 9: Machinery
(0 . 33)
347
Miscellaneous metal services (plating, coating, polishing)
(0 .42)
362
Electrical industrial apparatus (electric motors, welding apparatus)
Motor vehicles and equipment (including buses, trucks and truck trailers)
(0 .98)
371
Highest per capita employment in : Flint, Mich . ; Muncie, Ind . ; Lima, Ohio ; Ann Arbor, Mich . ; Fort Wayne, Ind .

Cluster 10 : Chemicals
(0. 99)
282
Plastic materials and synthetics (synthetic rubber and fibres)
(0. 47)
286
Gum and wood chemicals (charcoal, turpentine)
Highest per capita employment in : Asheville, N.C . ; Pensacola, Fla. ; Richmond, Va. ; Chattanooga, Tenn. ; Beaumont,
Tex.

Cluster 11 : Petroleum/Chemicals
(0. 88)
281
Industrial organic and inorganic chemicals
(0. 85)
291
Petroleum refining
(0 . 30)
299
Miscellaneous petroleum and coal products (lubricating oils, fuel briquettes)
Highest per capita employment in : Beaumont, Tex. ; Lake Charles, La . ; Galveston, Tex. ; Charleston, W.Va. ; Baton
Rouge, La. ; Wilmington, Del.

Cluster 12: Rubber Products
Tyres and inner tubes
(0. 94)
301
Reclaimed rubber
(0. 73)
303
(0. 63)
306
Fabricated rubber products, N .E.C . (rubber rafts, rubberised fabrics)
Highest per capita employment in : Akron, Ohio ; Gadsden, Ala. ; Tuscaloosa, Ala. ; Jackson, Mich. ; Des Moines, Iowa.

Cluster 13 : Glass and Glass Products
(0. 59)
321
Flat glass
(0. 83)
322
Pressed and blown glass (bottles)
(0. 47)
323
Products of purchased glass (aquariums, glass doors)
Highest per capita employment in : Toledo, Ohio ; Huntington-Ashland, W.Va. ; Wheeling, W.Va. ; Nashville, Tenn. ;
Waco, Tex .






Downloaded By: [University of Sydney] At: 04:45 5 September 2007

282

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

Table 13-continued
Cluster 14: Appliances and Furniture
(0 . 87)
251
Household furniture
(0-87)
363
Household appliances
Highest per capita employment in : Fort Smith, Ark. ; Evansville, Ind. ; Grand Rapids, Mich . ; Umisville, Ky . ; Lynchburg, Va . ; Greensboro, N.C .

Cluster 15 : Quarries
(0 . 96)
324
Hydraulic cement
(0. 58)
328
Cut stone and stone products
Highest per capita employment in : Allentown, Pa. ; Knoxville, Tenn . ; Bakersfield, Ca. ; San Bernard ino, Ca. ; Bay City,
Mich.

Cluster 16 : Grain Products
(0 . 96)
204
Grain mill products (flour, cereals, animal feeds)
(0 . 80)
209
Miscellaneous food preparations (coffee, macaroni, ice)
Highest per capita employment in : Decatur, Ill . ; Cedar Rapids, Iowa ; St Joseph, Mo . ; Terre Haute, Ind . ; Peoria, Ill.

Cluster 17: Food Canning
071
(0 . 52)
08
Agricultural services, forestry and fisheries
09
203
Canned, cured and frozen foods
(0 .94)
206
Sugar (cane and beet sugar, molasses)
(0 .47)
Highest per capita employment in : Brownsville, Tex. ; Stockton, Ca. ; San Jose, Ca . ; Rochester, N .Y. ; Savannah, Ga.

Cluster 18: Lumber
241
Logging
(0 . 96)
242
Sawmills
(0. 96)
Millwork, veneer, plywood
(0. 97)
243
(0. 52)
249
Miscellaneous wood products (poles, cork products)
Highest per capita employment in : Eugene, Ore. ; Tacoma, Wash . ; Dubuque, Iowa ; Pine Bluff, Ark. ; Portland, Ore .

Cluster 19: Wood Products
244
Wooden containers
(0. 86)
(0 . 88)
266
Building paper and board mills (tarpaper, wallboard)
Highest per capita employment in : Macon, Ga . ; Mobile, Ala . ; Dubuque, Iowa ; Lewiston, Me . : Savannah, Ga .

Cluster 20: Textile Products
(0 . 87)
221
Weaving mills, cotton
(0. 84)
222
Weaving mills, synthetics
(0 . 77)
226
Textile finishing, except wool (dyeing)
(0 . 54)
227
Carpet mills
(0 . 49)
232
Men's and boys' furnishings (shirts, neckties, work clothing)
Highest per capita employment in : Greenville, S .C. ; Augusta, Me. ; Lewiston, Me . ; Greensboro, N .C. ; Fall River, Mass.

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

283

Table 13-continued
Cluster 21 : Shoes, etc .
(0. 78)
223
Weaving and finishing mills, wool
(0 . 89)
313
Footwear cut stock (soles, uppers, buckles)
(0. 99)
314
Footwear, except rubber
Highest per capita employment in : Lewiston, Me. ; Manchester, N.H . ; Brockton, Mass . ; Binghampton, N.Y. : Portland,
Me.

Cluster 22: Paper Products
(0 .81)
262
Paper mills, except building paper
(0 . 60)
263
Paperboard mills
(0 . 91)
Miscellaneous converted paper products (envelopes, bags, wallpapers)
264
Highest per capita employment in : Green Bay, Wisc . ; Kalamazoo, Mich. ; Pine Bluff, Ark . ; Monroe, La . ; Hamilton,
Ohio .

Cluster 23: Tobacco/Textile
211
Cigarettes
(0-82)
Chewing and smoking tobacco
(0. 41)
213
(0. 77)
Tobacco stemming and redrying
214
Knitting mills
(0. 78)
225
(0. 37)
239
Miscellaneous fabricated textile products (curtains, textile bags, canvas products)
Highest per capita employment in : Durham, N.C. ; Winston-Salem, N .C . ; Reading, Pa . ; Richmond, Va . ; Greensboro,
N .C.

Cluster 24 : Labour Intensive I
(0 .70)
234
Women's and children's undergarments
Hats, caps and millinery
(0 .60)
235
(0 . 70)
236
Children's outerwear (dresses, coats, suits)
(0 . 54)
349
Miscellaneous fabricated metal products (barrels, safes, valves)
(0 . 55)
387
Watches, clocks and watchcases
(0 .77)
398
Miscellaneous manufactures I (brooms, linoleum, matches)
Highest per capita employment in : Lancaster, Pa . ; .Scranton, Pa. ; Johnstown, Pa. ; Springfield, Mass . ; Greenville, S.C .

Cluster 25: Labour Intensive II
(0 . 72)
224
Narrow fabric mills (fabric tapes, webbing)
(0. 59)
229
Miscellaneous textile goods pace goods, tyre cord)
(0. 49)
238
Miscellaneous apparel (robes, raincoats, belts)
(0. 66)
302
Rubber footwear
(0. 64)
364
Electric lighting and wiring equipment (light bulbs, switches)
(0. 87)
391
Jewellery, silverware and plated ware
(0. 92)
396
Costume jewellery and notions
Highest per capita employment in : Providence, R .I. ; Fall River, Mass . ; New Haven, Conn . ; Bridgeport, Conn . ; Boston,
Mass.

Cluster 26: Style Goods
(0 .46)
237
Fur goods
(0 .99)
317
Handbags and personal leather goods (wallets, watch straps)
Highest per capita employment in : Utica, N .Y. ; New York, N.Y. ; Fall River, Mass. ; Springfield, Mass . ; Springfield .
Ohio.

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

284

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

Table 13-r0 ztinued
Cluster 27: Printin,-

(0. 58)
274
Miscellaneous publishing (maps, catalogues)
(070)
276
'danifold business forms (sales books)
(0-70)
27S
Blankbooks and bookbinding (checkbooks, looseleaf binders)
(0. 69)
279
Printing trade services (typesetting, photoengraving)
Highest per capita employment in : Springfield, Mass . ; York, Pa . ; Dayton, Ohio ; Terre Haute, Ind . ; Topeka, Kans . :
Chicago, Ill . ; Dallas, Tex . : New York, N.Y.
Cluster 28: Low Wage I

(0-99)
228
Yarn and thread mills
(0-34)
261
Pulp mills
(0-40)
319
Leather goods, N .E .C . (holsters, leashes)
Highest per capita employment in : Albany, Ga. ; Chattanooga, Tenn. ; Macon, Ga . ; Columbus, Ga . ; Providence, R .I . :
Scranton, Pa .
Cluster 29 : Low Wage II

(0-63)
212
Cigars
231
Men's and boys' suits and coats
(058)
(0-96)
233
Women's and misses' outerwear (dresses, coats)
(0 . 58)
395
Pens, pencils, office and art supplies
Highest per capita employment in : Wilkes-Barre, Pa . ; Scranton, Pa. ; Fall River, Mass . ; Allentown, Pa . ; York, Pa . ;
Atlantic City, N .J . ; New York, N.Y.
Cluster 30 : Miscellaneous I

(060)
2 53
Public building furniture (church pews, bleachers, school furniture)
(0-93)
283
Drugs
(0-63)
393
Musical instruments
Highest per capita employment in : Kalamazoo, Mich . ; Indianapolis, Ind . ; New London, Conn . ; Richmond, Va . ;
Grand Rapids, Mich .

Cluster 31 : Miscellaneous I/

(0-67)
307
Miscellaneous plastic products (plastic hose, plastic bowls)
(0-78)
358
Service industry machines (vending, refrigeration equipment, commercial laundry equipment)
(061)
365
Radio and television receivers
Highest per capita employment in : Syracuse, N .Y . ; Evansville, Ind . ; Trenton, N .J . ; Scranton, Pa. ; York, Pa .
Cluster 32 : Miscellaneous III

(0. 56)
(0.99)

315
352
381

Leather gloves and mittens
Farm machinery
Scientific instruments
(0-51)
Highest per capita employment in : Davenport, owa ; Dubuque, Iowa ; Waterloo, Iowa ; Racine, Wise. ; Des Moines,
Iowa .

Cluster 33 : Miscellaneous IV

(0 . 84)
273
Books (publishing and printing)
(0 .69)
284
Soap, cleaners, toilet goods (perfume, polishes)
(0-56)
311
Leather tanning and finishing
(0-74)
343
Plumbing and heating apparatus
Highest per capita employment in : Racine, Wise . ; Lorain-Elyria, Ohio ; Grand Rapids, Mich . ; Louisville, Ky . ; Binghampton, N .Y.





Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

285

Table 13-continued
Cluster 34: Miscellaneous V
(0-90)
287
Agricultural chemicals (fertiliser, insecticides)
(0 . 64)
295
Paving and roofing materials (asphalt paving mixtures)
Highest per capita employment in : Savannah, Ga . ; Charleston, S.C . ; Albany, Ga. ; Tampa, Fla . ; Norfolk, Va .

Cluster 35: Miscellaneous V/
(0. 88)
277
Greeting card publishing
(0. 76)
316
Luggage
Highest per capita employment in : Manchester, N .H . ; Kansas City, Kans
; Denver, Colo. ; Topeka, Kans . ;
.
.-Mo
Cincinnati, Ohio.

Cluster 36: Market Centre Business Services
(0. 49)
275
Commercial printing
(0. 56)
48
Communication (telephone company, radio-TV broadcasting)
(0. 69)
60
Banking
(0. 43)
62
Security and commodity brokers and exchanges
(0. 80)
63
Insurance carriers
731
Advertising
(0.53)
73,
Miscellaneous business services except advertising (employment agencies, testing laboratories,
(0. 69) excl .
731
management consultants)
(0. 63)
81
Legal services
(0 .67)
86
Non-profit membership organisations (churches, labour unions, trade associations)
(0. 63)
89
Miscellaneous services (data processing, non-profit research)
Highest per capita employment in : Des Moines, Iowa ; Hartford, Conn . ; New York, N.Y . ; Lincoln, Nebr. ; Springfield,
Ill .

Cluster 37: Market Centre, Consumer Services
15
16
Construction
(0 . 70)
17
205
Bakery products
(0 . 32)
271
(0 . 50)
Newspapers
(0-18)
285
Paints and allied products (varnish, putty)
Local and interurban passenger transit (taxis, bus lines)
(0 . 54)
41
(0-83) 52-59
Retail trade
(0 . 49)
61
Non-bank credit agencies (savings and loans, finance companies)
(0-155)
64
Insurance agents, brokers and service
(0 . 63)
65
Real estate
(021)
66
Combined real estate, insurance, loan, law offices
(0-68)
72
Personal services (laundries . barber shops, funeral parlours)
(0 . 65)
75
Automobile repair and rental
(0 . 30)
78
Motion pictures (movie production, movie theatres)
(0 . 39)
80
Medical service (physicians, dentists, hospitals)
(0 . 32)
82
Educational services (schools, colleges, libraries)
(024)
99
Non-classifiable establishments
Highest per capita employment in : Las Vegas, Nev. ; Reno, Nev . ; Charlotte, N.C . ; Atlantic City, N.J . ; Sacramento, Ca . ;
Des Moines, Iowa.







JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

286
Table 13-continued
Cluster 38: Ports
(0. 85)
373
(0. 85)
44
(0. 46)
47
Highest per capita

Ship and boat building
Water transportation
Transportation services (freight forwarding, stockyards, travel agents)
employment in : Galveston, Tex . ; Beaumont, Tex . ; New Orleans, La. ; New London, Conn . ; Mobile,
Ala . ; Newport News, Va .

Cluster 39 : ResortslRapid Growth
(0 . 60)
327
Concrete, gypsum and plaster products
701
Hotels
(0-86)
(0 . 94)
79
Amusement and recreation services (racetracks, golf courses)
Highest per capita employment in : Reno, Nev . ; Las Vegas, Nev . ; Atlantic City, N .J. ; Miami, Fla. ; Fort Lauderdale,
Fla . ; Santa Barbara, Ca .

Cluster 40: Transport and Repair Services
(0. 84)
42
Trucking and warehousing
(0-69)
45
Air transportation
76,
excl .
Consumer-oriented repair shops and services (reupholstery, radio repair)
(0. 30)
769
(0-45)
Miscellaneous repair shops (welding repair, armature rewinding)
769
Highest per capita employment in : Winston-Salem, N .C. ; Miami, Fla. ; Charlotte, N.C . ; Atlanta, Ga . ; Odessa, Tex . ;
Dallas, Tex.

Cluster 41 : Miscellaneous VII
(0 . 39)
202
Dairy products
265
Paperboard containers and boxes
(0. 38)
(0 . 43)
272
Periodicals
50
Wholesale trade
(0-96)
Highest per capita employment in : Bridgeport, Conn . ; Springfield, Mo . ; Charlotte, N .C. ; Des Moines, Iowa ; Atlanta,
Ga. ; Dallas, Tex.

Cluster 42 : Miscellaneous VIII
(0-96)
67
Investment companies (holding companies, mutual funds)
84
Museums, botanical, zoological gardens
(0-68)
Highest per capita employment in : New London, Conn. ; Minneapolis, Minn . ; Las Vegas, Nev. ; Tulsa, Okla. ; Des
.
Moines, Iowa.

Industries with Strong Links to More than One Cluster
289
Miscellaneous chemical products (explosives, printing ink)
341
Metal cans
Fabricated structural metal products (metal doors, boiler shops)
344
346
Metal stampings (auto body parts, bottle caps, kitchen utensils)
Special industry machinery (food and textile machinery)
355
Pipelines
46





Downloaded By: [University of Sydney] At: 04:45 5 September 2007

THE AGGLOMERATION PROCESS IN URBAN GROWTH

287

Table 13-continued
Independent Industries : -Few Locations
19
Ordnance
252
Office furniture (wood and metal)
333
Primary non-ferrous metals
Non-ferrous rolling and drawing (copper wire, aluminium foil)
335
372
Aircraft
374
Railroad equipment
375
Motorcycles and bicycles
379
Miscellaneous transportation equipment (mobile homes)
399
Miscellaneous manufactures U (signs and advertising displays, tobacco pipes, umbrellas)
Administrative and auxiliary-manufacturing

Independent Industries: Many Locations
10-14
Mining
201
Meat products
207
Confectionery (candy, cocoa, chewing gum)
208
Beverages (soft drinks, distilled liquor, beer, wine)
353
Construction and related machinery (earthmovers, cranes, mining and oil field machinery, elevators)
40
Railroad transportation
Electric, gas and sanitary services
49
91
Federal government
State government
92
93
Local government
Administrative and auxiliary-non-manufacturing

Conclusion
We have seen that highly disaggregated data,
and units of observation (SMSAs) which
roughly approximate local market areas, permit
us to discern important similarities in the
locational patterns of groups of industries.
Often this `clustering' occurs among industries
in widely separated sectors of the Standard
Industrial Classification .
The structure of this clustering is complex .
First, all clustering is not mutually exclusive.
Although many activities are clearly associated
with one and only one cluster, many others are
not. Therefore, any mutually exclusive description does some violence to the more complex
reality.
Moreover, there do not seem to be any
`natural' levels of aggregation . We have not
exhaustively tested all possible levels of aggregation, but we have tested many different methodologies and arrived at clustering at several different
levels, and all of these levels seem to make sense .
We conclude that the analyst may choose the
t 5

level of aggregation most appropriate to his
purpose.
The clustering of activities is an important
determinant of the economic growth of cities .
In our continuing research we are further analysing this process.
REFERENCES
(1956) . The Industrial Structure
ofAmerican Cities. Lincoln : University of Nebraska

ALEXANDERSSON, G.

Press .
CHINITZ, B. (1961). Contrasts in Agglomeration : New

York and Pittsburgh. American Economic Review,
Papers and Proceedings . May, pp . 279-289.
EUROPEAN ECONOMIC COMMUNITY (1966) . Studr
for the Promotion of an Industrial Development Pole
in Southern Italy. Economic and Financial Affairs
Series, No . 5 . Brussels : European Economic
Community.
FLORENCE, P. S ., et al. (1943). Industrial Location and
National Resources. Washington D .C. : National
Resources Planning Board.
Fucus, V. (1962). Changes In the Location of Manufacturing in the United States Since 1929. New
Haven : Yale University Press.
GORDON, P. (1971). Urban Agglomeration Economies .
Presented to the Regional Science Association, 14
April (mimeo).

Downloaded By: [University of Sydney] At: 04:45 5 September 2007

288

JOEL BERGSMAN, PETER GREENSTON AND ROBERT HEALY

HOOVER, E . M . (1937) . Location Theory and the Shoe
and Leather Industries . Cambridge, Mass . : Harvard
University Press.
HOOVER, E . M . and VERNON, R . (1962). Anatomy of
a Metropolis. New York : Anchor Books .
ISARD, W ., SCHOOLER, E. and VIETORISz, T. (1959) .
Industrial Complex Analysis and Regional Development . New York : John Wiley and Sons .
LICHTENBEAG, R. M . (1960). One-Tenth of a Nation .
Cambridge, Mass. : Harvard University Press .
MCCARTY, H ., HOOK, J . and KNOS, D . (1956). The
Measurement of Association in Industrial Geography .
Iowa City : State University of Iowa .
MEIER, R . L . (1970). A Communications Theory of
Urban Growth . Cambridge, Mass . : M .I .T . Press .
MURPHY, R. (1966). The American City : An Urban
Geography . New York : McGraw-Hill .
NELSON, H . J . (1955). A Service Classification of American Cities. Economic Geography, Vol . 3 1, pp . 189-210 .
PERLOFF, H . S. et al . (1960) . Regions, Resources and
Economic Growth. Lincoln : University of Nebraska
Press .
PRED, A. (1965) . The Concentration of High ValueAdded Manufacturing. Economic Geography, Vol .
41, No. 2, pp. 108-132.

RICHTER, C . E . (1969) . The Impact of Industrial
Linkages on Geographic Association. Journal of
Regional Science, Vol. 9, No. 1, pp. 19-28 .
RoccA, C . A . (1970) . Productivity in Brazilian Manufacturing . Appendix 2 of Bergsman, Joel, Brazil:
Industrialization and Trade Policies . Cambridge,
Eng. : Oxford University Press .
RUMMEL, R . J . (1970) . Applied Factor Analysis.
Evanston : Northwestern University Press .
SHEFER, D. (1971). An International Comparison of
Production Parameters-The United States vs . Israel .
Haifa, Israel : Center for Urban and Regional
Studies, Technion. August .
SONQUIST, J . A . and MORGAN, J. N . (1964), The
Detection of Interaction Effects : A Report on a
Computer Program for the Selection of Optimal
Combinations of Explanatory Variables. Ann Arbor :
University of Michigan Institute for Social
Research .
STANBACK, T. M . and KNIGHT, R. V . (1970) . The
Metropolitan Economy . New York : Columbia
University Press .
STREIT, M . E. (1969) . Spatial Associations and Economic Linkages Between Industries. Journal of
Regional Science, Vol . 9. No . 2, pp. 177-188 .

Acknowledgements. This is a revised version of a paper presented at the Resources for the FutureUniversity of Glasgow Conference on Economic Research Relevant to the Formulation of National
Urban Development Strategies, Glasgow, Scotland, August 30-September 3 1971 . The research was
supported by funds from the U .S . Department of Housing and Urban Development . The views in
this paper are those of the authors and do not necessarily reflect those of The Urban Institute or its
sponsors.
The authors gratefully acknowledge the advice and comments of William Alonso, Brian Berry,
Benjamin Chinitz, and Thomas Vietorisz, and the assistance of Andreas Andrikopoulos and WinStanley Luke.