Professional Documents
Culture Documents
Agglomeration economies are economies external to the firm, and plant, but
internal to a particular location. In this paper the notion of agglomeration economies
is formally investigated. The implications of the concept for urban structure and
industrial organization are then traced. Under specified considerations, weak econo-
mies of agglomeration, as defined in this paper, are shown to be necessary for the
existence of an urban area.
1. INTRODUCTION
An urban agglomeration is a geographic concentration of economic
activity. The generally accepted economic basis for this agglomeration is
some form of scale economy, a notion that spatial proximity of activities
makes resources more efficient than if such activities are spatially dispersed.
The concept of economies of scale has been examined intensively from both
a theoretical and an empirical viewpoint, but not with completely satisfac-
tory results and not in a spatial context. (See Gold [3].) The agglomerative
economies concept, on the other hand, has not received the same attention.
There are basically three approaches to the scale concept underlying
agglomerative economies; economies internal to a firm at a given location,
which we shall label the company town; economies external to a firm at a
given location, but internal to an industry at that location, called localiza-
tion economies; and finally, economies external to both the firm and the
specific industry at a particular location, called urbanization economies.
Peat, Marwick, and Partners [13] and Henderson [5] discuss the latter two
‘We thank Jack Meyer, Ron Braeutigam, John Ledyard, Mike Marrese, Bruce Petersen, and
a referee for helpful comments.
91
0094-1190/84 $3.00
Copyright 0 19X4 hy Academic Press, Inc
All rights of reproduction in any form reserved.
92 GOLDSTEIN AND GRONBERG
concepts in detail. All three approaches imply that the benefit that firms
derive from operating in an urban environment is based on a scale concept,
One early example of this link between the functional structure of
industry and its geographic structure is provided by Stigler [17]. “Localiza-
tion is one method of increasing the economic size of an industry and
achieving the gains of specialization. The auxiliary and complementary
industries that must operate in intimate cooperation can seldom do so
effectively at a distance” [17, p. 1891. Stigler conjectures that geographic
dispersion is a luxury enjoyed by large industries, specifically those large
enough so that even small units of production can benefit from specializa-
tion. This reasoning suggestsone important cause of spatial agglomeration
which we pursue below, the incentive to reduce the costs of coordinating
closely linked activities.
An example of urbanization economies is specialized services in large
urban areas which do not exist in smaller areas. In a rural area, a
manufacturing firm which operates a fleet of trucks must have on hand its
own mechanics, or use local “general” mechanics. In a large urban area, the
firm can draw upon firms which specialize in maintenance of large trucks.
(See Isard [6].) W e will provide an alternative way of viewing agglomerative
economies, based on the theory of multiproduct firms as developed by
Panzar and Willig. We will argue that urban areas can be viewed as
“ vehicles” for spatial integration, in the same way that vertically integrated
firms gain efficiency from engaging in multioutput production. It is not
simply the scale of activity in the area that is important in our approach,
but the improvement in productive efficiency from placing related activities
at a common location.
We can distinguish our concept from the scale concepts by considering
the empirical implications of each approach. For localization economies, we
measure the scale factor using total employment or output in a given
industry, in a given urban area; for urbanization economies, we use total
population or employment of the entire city. Agglomerative economies
based on a multioutput concept cannot be measured with such aggregate
measuressince it is the presenceof a specific set of interrelated activities in a
given location that is the source of the efficiency gain. One possible measure
of agglomerative economies would then be the reduction in unit costs that
members of this set derive when they are in spatial proximity.
Our concept can also answer a question raised by Henderson [5] regard-
ing systemsof cities. “If resources are more efficient in one size city than in
another, then radically different size cities should not survive in a market-
oriented economy. Yet radically different size cities have always coexisted in
market economies” [5, p. 11.If one major source of agglomerative economies
is productive efficiency from locating specific sets of economic activities in
spatial proximity, then we can have a variety of city sizes and a variation in
ECONOMIES OF AGGLOMERATION 93
the set of activities that takes place across cities. How large a city is and
what activities it contains depend on the particular set of interrelated
activities that operates within its boundaries.
Before we consider supporting empirical work, we note two other reasons
for agglomeration which do not fit neatly into the scheme outlined above.
Mills [9] argues that the most important aspect of agglomerative economies
“is statistical in nature, and is an application of the law of large numbers.
Sales of outputs and purchases of inputs fluctuate in many firms and
industries for random, seasonal, cyclical, and secular reasons. To the extent
that fluctuations are imperfectly correlated among employers, an urban area
with many employers can provide more nearly full employment of its labor
force than can an urban area with few employers” [9, p. 111.
Finally, there may be false price signals at small scales that lead more
firms to locate in a specific location than would do so in the absenceof this
“market failure.” This situation is analogous to congestion on an express-
way. The price paid for use of the urban infrastructure is not marginal social
cost, but the lower price, marginal private cost. The result is urban ag-
glomeration.
The approach to be presented formally includes the various scale con-
cepts, in addition to the multiproduct notion. The market failure idea could
be considered a special case of the multiproduct concept, in which private
costs of locating in an urban area do not reflect social costs. Mills’
suggestion, however, requires a dynamic context and is not dealt with in this
model.
2. EMPIRICAL SUPPORT FOR THE CONCEPT OF
AGGLOMERATIVE ECONOMIES
Lichtenberg’s study of the New York Metropolitan Region in the late
1950s is the classic case study of localization economies. He identified 87
industries “as dominated by external-economy factors of location and
which are relatively concentrated in the New York Metropolitan Region”
[6, p. 791. Most of these industries were related to the production of
clothing, in which proximity to buyers and suppliers was important.
Henderson [5] uses 1972 data for the United States and 1970 data for
southern Brazil to examine the nature and extent of scale economies in
manufacturing. He seeks to identify these scale economies as either urban-
ization or localization ones. Henderson’s empirical work is developed using
a constant returns to scale firm production function with a Hicks-neutral
external shift factor (which is a function of scale and technology measures
specific to an industry in an area). Henderson finds that in general, external
economies of scale are due to localization factors rather than urbanization
ones. Nothing, however, is revealed about the possible strength of a “multi-
product” basis for agglomeration when a Hicks-neutral shift factor is
94 GOLDSTEIN AND GRONBERG
‘Moomaw [lo], reexamines two recent studies linking productivity and city size using
Hicks-neutral shift factors. He found that large American cities have substantial productivity
advantages which are much more important in their nonmanufacturing sectors. Harris and
Hopkins [4], who do examine interindustry effects, find that the basic chemical industry and the
iron and steel industry had the most influence on change in output in other industries. Basic
chemicals influenced the location of 11 other industries, six through transport costs, one as
major buyer and five because of the agglomeration effects of being a major supplier. [Iron and
Steel] influenced the location of 10 other industries. Other important influencing industries
are Agriculture, Paper, Rubber and Plastics, and Other Nonferrous Metals” [4, p. 921.
ECONOMIES OF AGGLOMERATION 95
3Teece [18] points out that scope economies indicate that joint production must occur, but
that “joint” production can be obtained using contracts as well as managerial hierarchies.
Whether the one is more efficient than the other is basically an empirical question which turns
on transactions costs. Stigler [17] makes a similar point.
41n a related paper, Chipman [2] developed the concept of “parametric external economies
of scale,” a concept which can be related to agglomerative economies. He distinguished the
entrepreneur’s notion of his production function, the subjective production function, from the
actual or objective production function. Each entrepreneur believes “that his firm is operating
under constant returns to scale, and any departures from this assumed output-factor
relationship are interpreted by him as brought about by a perturbation in his unit-homoge-
neous production function even if such departures are caused in part by changes in his own
level of output. Such shifts are, in turn, assumed to be governed by the level of output in the
industry” [2, p. 3491.Chipman argues that this last assumption could be relaxed to make such
shifts a function of employment in other industries. This would cover agglomerative economies.
Our approach will not rely on a distinction between a subjective and an objective production
function. We will restrict firms by making assumptions about economies of scope, about
operating costs about which the firm is fully infmed. Using a Chipman-type model, we would
find that the market would not necessarily produce spatial agglomeration since the firms would
not perceive the effect of such spatial agglomerations on their production function.
ECONOMIES OF AGGLOMERATION 97
(1)
In Table 1 we summarize the notation to be used. For example, C( y,‘, w)
represents the cost function of a multiproduct firm which locates all
production activities at site r. This firm benefits from both economies of
scope and economies of agglomeration. At the other extreme,
i i c(q;,,w), cq-;;=yp
r=l i=l r
represents the total cost of producing the products in the subset S when
those products are produced at more than one location by more than one
firm. There are both diseconomies of scope and diseconomies of agglomera-
tion. Finally, in order to concentrate on interindustry effects, we assume w
to be uniform over space. There are no initial differences in the relative
attractiveness of various sites. The only factor that can make one site more
or less attractive than another to a particular firm is the presence or absence
of other firms.
98 GOLDSTEIN AND GRONBERG
TABLE 1
Cost of Production under Various Assumptions
Economies of
agglomeration
Diseconomies of
agglomeration
wherel<lrzandzqG,=y,,.
r
(14
To rule out an urban area with only multiproduct firms, we must assume
diseconomies of scope in the Panzar-Willig sense,that is, that
in the subset s, if full advantage is taken of spatial economies of scale, but not spatial cost
complementarities.
We can say that economies of agglomeration are due to spatial cost complementarities and
not spatial scale economies if (1) holds and in addition if
‘0 < c. e < 1.
100 GOLDSTEIN AND GRONBERG
Following Panzar and Willig, if there did not exist at least one subset of N
containing two or more products, which was invulnerable to such fragmen-
tation, no multiproduct urban areas could exist in competitive equilibrium.
A similar argument holds for the case of diseconomies of scope.
We have now established the consequencesof the presence or absenceof
economies of agglomeration for the spatial arrangement of firms in competi-
tive equilibrium.‘j The source of these economies, input sharing, can now be
examined.
6We have said nothing about the existence of either spatial or nonspatial monopoly. The
approach developed can shed light on this issue. What is necessary is a more precise
specification of the relationship between the spatial structure of demand and the spatial
structure of costs.
ECONOMIES OF AGGLOMERATION 101
tion will increase (or not change) and therefore the cost of producing all
goods together will increase (or not change). Therefore, superadditivity (or
not subadditivity) of the cost of producing the shared resource is incompati-
ble with economies of agglomeration.
8. SUMMARY
By taking account of both the spatial production and organizational costs
of operating a firm, we have developed a foundation for the economics of
agglomeration. We have, moreover, indicated that multiproduct urban areas
do not necessarily involve multiproduct firms. Clearly vertical and horizon-
tal integration in the traditional industrial organization, that is, where the
104 GOLDSTEIN AND GRONBERG
focus is on the firm, are strongly linked to the location strategy of the firm,
that is, where the focus is the urban area.
REFERENCES
1. W. .I. Baumol and R. D. Willig, Fixed costs, sunk costs, entry barriers and sustainability of
monopoly, Quart. J. Icon., 95, 405-431 (1981).
2. J. S. Chipman, External economies of scale and competitive equilibrium, Quart. J. Econ.,
84,341-385(1970).
3. B. Gold, On size, scale and returns: A survey, J. Econ. Lir., 19, 5-33 (1981).
4. C. C. Harris, Jr. and Frank E. Hopkins, “Locational Analysis,” Heath, Lexington, Mass.
(1972).
5. .I. Vernon Henderson, “Efficiency of Resource Usage and City Size,” Brown University
Working Paper No. 82-14, August, 1982.
6. E. M. Hoover, “An Introduction to Regional Economics,” Second ed., New York, Knopf,
1975.
7. W. Isard, “Introduction to Regional Science,” Prentice-Hall, Englewood Cliffs, N.J.
(1975).
8. T. C. Koopmans and M. J. Beckmarm, Assignment problems and the location of economic
activities, Econometrica, 25, 53-76 (1957).
9. E. S. Mills, “Urban Economics,” Second ed., Scott, Foresman and Co., Glenview, Ill.
(1980).
10. R. Moomaw, Productivity and city size: A critique of the evidence, Quart. J. Econ., 95,
675-688(1981).
11. J. C. Panzar and R. D. Willig, “Economies of Scale and Economies of Scope in
Multi-Output Production,” Bell Laboratories Economics Discussion Paper #33 (1975).
12. J. C. Panzar and R. D. Willig, Economies of scope, Pap. Proc. Amer. Econ. Assoc., 71,
268-272.
13. Peat, Marwick and Partners, “Economics of Agglomeration,” Report to the City of
Toronto Planning Board (1975).
14. F. Perroux, Note on the concept of growth poles, in “Regional Economics,” (D. L. McKee
et al., Eds.), pp. 93-104, The Free Press,New York (1970).
15. F. M. Scherer, “Industrial Market Structure and Economic Performance,” Second ed.,
Rand McNally, Chicago (1980).
16. R. W. Schmenner, “The Manufacturing Location Decision: Evidence from Cincinnati and
New England,” Economic Development Research Report (1978).
17. G. J. Stigler, The division of labor is limited by the extent of the market, J. Pol. Econ.,
185-193 (1951).
18. D. Teece, Economies of scope and the scope of the enterprise, J. Econ. Behavior
Organization, 1, 223-245 (1980).
19. 0. E. Williamson, “Corporate Control and Business Behavior,” Prentice-Hall, Englewood
Cliffs, N.J. (1970).
20. 0. E. Williamson, The modem corporation: Origins, evaluation, attributes, J. Econ. Lit.,
19,1537-1568 (1981).