An Introduction to Regional Economics
Edgar M. Hoover and Frank Giarratani
The Spatial Structure of Urban Areas


In this chapter we are concerned more specifically with spatial relations within the individual urban or metropolitan area.1 Such an area includes a principal city with an intensively developed core or downtown area (the central business district, or CBD) and a surrounding fringe of suburbs and satellites linked to the principal city by trade, commutation, and other socioeconomic interaction.

It would be hard to think of any significant question or proposition of urban economics not involving space, distance, or location as a fundamental concern, since the essence of a city lies in the close proximity of diverse activities and persons. So urban economics is just part of the broader field of spatial or regional economics. But it is such a large part that it is often studied as a distinct entity. Our intention in this chapter is not to survey urban economics as a field of study but to expose some essential aspects of the spatial structure of urban economies. In this way, it will be possible to affirm some principles of spatial economics that are particularly relevant and useful in understanding the development of cities and their problems.


The land-use analysis that was developed in Chapter 6 allowed us to identify a number of characteristics associated with urban activities which implied a willingness to bid high rents for the more central locations. The movement of people and the importance of direct face-to-face contact contribute to the advantages of central locations for these activities. The applicable transfer rates are high, and linkages among nonresidential activity units, among households, and among residential units and nonresidential units are substantial. Further, the realization of agglomeration economies most often requires close and frequent contact. These economies enhance the attractiveness of central locations and tend to bring units together, not merely in the same city but in the same district of a city.

7.2.1 Independent Locations

While these access and agglomerative factors are quite important in explaining the outcome of the bidding process by which land is allocated among competing urban uses, it is necessary to recognize that some kinds of locations within an urban area can be regarded as independently determined. In fact, there are two distinct bases for exogenous determination of locations in an urban area. For some activities, certain topographical or other natural site features are essential; this means that the lie of the land narrows the choice to one or a very small number of locations. Ports for water traffic illustrate this, and there are some urban areas where the topography limits airport sites almost as drastically. In the distant past, considerations of defense played a major part in locating the heart of the city and the city itself. Localized recreational facilities such as beaches also illustrate this kind of factor, and in a few urban areas extractive industries (mainly mining) occur and are, of course, limited to certain special sites.

There is a further type of exogenously determined location where the independent influence arises not from site features as much as from the fact that the activity requires contact with the outside world. Not just water ports but all kinds of terminal and interarea transport activities come under this head. Since there are great economies of scale in interregional transport and in terminal handling of goods, the urban area’s gateways to and from the outside world constitute a set of focal points, whose locations within the area help to determine—rather than just being determined by—the other activities of the area. This does not mean, of course, that such terminal locations are absolutely and permanently unresponsive to the changing patterns of other activities in the area served. Such terminals are from time to time shifted so as to improve local accessibility or to make way for more insistent claimants for the space. But the terminal locations do play an active role in shaping the pattern and are to be viewed as part of the basic framework around which other activities are fitted.

7.2.2 The Center

There is also a strong element of exogenous determination in the location of the point of "maximum overall accessibility" within the urban area. If we think of this, for example, as the place where all the people of the area could assemble with the least total man-miles of travel, it is the "median center of population" and would seem to depend simply upon the location of the various types of residence. But travel is cheaper and faster along developed routes, and the cost and layout of these routes are affected by scale (traffic volume) and topography. Thus, evaluated in terms of travel cost and time, the focal maximum-access point can be regarded as a rather stable datum, even though the extent and importance of its access advantage over other points can change radically. In major American urban areas, despite great overall growth, far-reaching change, and redistribution of activities, this focal point has usually shifted only a relatively short distance over periods measured in decades and generations; and the earlier central foci are well within what we currently recognize as the central business district.

This concept of a single "most central" focal point in an urban area is significant and useful in developing simplified bases for understanding the overall pattern. Obviously, it has its limitations, some of which will be discussed now and others later. First, there are really a variety of distinguishable central points of this sort, depending on what kinds of people or things we are imagining to be assembled with a minimum of total expense or effort. The employed workers of the area are not distributed in quite the same pattern as the total population, the shopping population, the school-attending population, the office workers, the industrial blue-collar workers, the theater-going or the library-using population; there might be a different optimum location from the standpoint of access to each of these types of people. Where goods rather than people are moving (for example, in the case of wholesale activity or production serving such local needs as daily newspapers or bread), the transport conditions are different, and this may again mean a different optimum-access point. Second, we have to recognize that, in varying degrees, the concept of one single point serving as the origin or destination for all flows of a specified type is unrealistic, and defensible only as a convenient fiction. Thus if we identify some central point as having best access to the homes of the entire clerical office force of an urban area, this does not imply that all offices should logically be concentrated there. What it does imply is that, solely from the standpoint of commuting access for the clerical workers and ignoring claims of alternative uses of space, it would make sense for the density of clerical employment to peak at that point.

7.2.3 Neighborhood Externalities

In gaining perspective on the role of access and agglomerative factors in urban location decisions, it is also necessary to recognize that proximity can have unfavorable as well as favorable effects. "Neighborhood character’ ‘—in terms of cleanliness, smells, noise, traffic congestion, public safety, variety interest, and general appearance—is important in attracting some kinds of use and repelling others. Prestige types of residence or business are, of course, particularly sensitive to this kind of advantage, which is often more important than any access consideration as such. High-income householders may be willing to lengthen their work journey greatly for the sake of neighborhood amenity or agreeable surroundings.

The usual effect of this type of consideration is to make neighborhoods more homogeneous within themselves and more unlike other neighborhoods: a tendency toward areal specialization by uses, or "segregation" in the broad sense.2 With few exceptions, a given type of activity finds advantage in being in a neighborhood devoted to reasonably similar kinds of uses, and disadvantage in being in violent contrast to the neighborhood pattern. Zoning controls and planned street layouts play a part in reinforcing this tendency.

7.2.4 Scale Economies and Urban Land Use

Many of the points just raised imply that the broad continuous zones of economic activity suggested by von Thünen’s simplified model of patterns of land use in Chapter 6 would be substantially modified in an intraurban setting. When that model is applied to such extensive activities as agricultural or residential land users, it is not really necessary to consider the size of the individual location unit in terms of output or occupied land area, since such zones contain a large number of adjacent units. Accordingly, in that instance, we look for explanations of rent-paying ability and location in terms of inputs, costs, outputs, and rents on a per-acre basis. We could appropriately consider costs as affected by intensity of land use rather than by the size of the producing unit, the firm, or the cluster.

Consider, however, an activity such as university education, which on the basis of its production characteristics can best be located, say, 5 miles from the center of the metropolitan area supplying the bulk of the students. A more central location would mean excessive land costs, while a less central one would mean poor access to the homes of the commuting students and perhaps also to various other urban activities with which contact is desired. If we brashly apply the basic von Thünen model, we get the answer that a university should occupy a ring-shaped zone with a 5-mile radius. If the amount of space needed were, say, 300 acres, the ring-shaped campus would be about 80 feet wide and more than 31 miles long. Since such a layout would preclude both having a sizable stadium and getting to classes on time, it is clearly unacceptable. In the interest of its own internal logistics, the university would prefer a blob to a doughnut. Two different institutions in the same city might find some external-economy advantage in being close to one another in a single "university district," but if they are intensely competing for commuter students, they might prefer to locate on opposite sides of town.

The point here is that a university campus is a location unit subject to considerable economies of scale, so that there will be only a few unit locations, perhaps only one, in any given urban area; at the same time it is sufficiently space using to need an off-center or even suburban location. The same principle applies to any activity with these characteristics. As a result, the concentric ring pattern appears within urban areas only with respect to certain broad classes of activities such as residence. For other noncentral uses, the pattern can range from scattered fragments of a ring to a single off-center concentration. Still further complication is introduced by the fact that each such concentration can become a focal point for a neighborhood constellation of associated land uses. Any sizable urban area contains a number of such subcenters in addition to the principal downtown center.


7.3.1 Bases of Simplification

A number of factors relevant to intraurban location decisions have been catalogued above, but basically there are three kinds of considerations that determine the relative desirability of locations for individual location units, such as households or business establishments. These are (1) environmental characteristics, (2) access, and (3) cost. They reflect the fact that the users of a site are concerned with it in three distinct ways. They occupy it, as residents or producers, and are thus concerned with its "site and neighborhood," or immediate environmental qualities. They, as well as goods and services, move between this site and others and are therefore concerned with its convenience of access to other places. Finally, they have to pay for its use and are therefore concerned with its cost.

It should be evident now that in reducing the complex factor of access within an urban area to the simple form of proximity to a single focal point—as was done in Chapter 6—some important aspects of urban economic activity are put aside. Within a city, it is as if all intraurban journeys were to or from downtown and all shipments of goods also passed through downtown.3Additionally, such an analysis eliminates all differentiation of sites with respect to topography, amenity, and environmental advantage. These two simplifications also imply ignoring the manifold types of external-economy effects and environmental attractions and repulsions that have been discussed. In effect, each type of activity is thought of as being independently attracted (by access considerations) toward the urban center. The only interdependence among the locations of the various activities arises, then, from the fact that they are bidding against one another for space.

Nevertheless, as a starting point for understanding urban spatial structure, monocentric models can be very useful. While they abstract from some important features of the urban environment, they expose others that are fundamental in understanding urban spatial patterns.

7.3.2 The Density Gradient

Perhaps the most elementary aspect of an urban pattern that is illuminated by monocentric models such as those discussed in Chapter 6 is the way in which intensity of land use varies with distance from the center. Implicit here is the concept of a city as a multitude of space-occupying location units seeking close contact. If these location units are affected by more or less the same kind of access attraction (as, for example, households are affected by the desire to shorten the journey to work) and have some leeway in the amount of space they occupy, we should expect their density (intensity of space use) to be at a peak at the center (the optimum total-access point) and to fall off in all directions with increasing distance from the center. Such a tendency can be described by a density gradient, where density is a negative function of radial distance.

In this simple scheme, the decline of density with distance depends (1) on the rate at which the area’s noncentral activity units (households, in a purely residential journey-to-work model) are willing to trade off spaciousness of home sites against a quicker or cheaper journey to the center (reflecting lower land rents with increased distance from the CBD), and (2) on the time and money cost of transport. Obviously, a variety of circumstances—such as better transport in some directions than in others and variations in site quality—can complicate this neat symmetrical picture in the real world.

As Colin Clark has demonstrated, the gradient of population density with respect to radial distance, in a wide selection of large modern cities, has a consistent shape, identifiable as an exponential function. The exponential shape of the density gradient is predicted by virtually all monocentric models of urban land use,4and consistent with this prediction, Clark found that residential density does tend to fall by a uniform percentage with each unit increase in distance from the center. Density gradients can therefore be specified by two parameters: D0, the peak density at the center, and b, a slope factor, in the following formula:

Dx=D0 e - bx

where x represents radial distance and e is 2.718 ..., the base of natural logarithms.5Several actual gradients are shown in Figure 7-1.

Since this particular conformation describes residential density, it fits in only those parts of the urban area that are primarily residential. The "peaking" of residential densities actually resembles a volcano more than a sharp conical mountain peak. There is a crater of lower density in the innermost zone, where nonresidential activities predominate. The D0 parameter in the gradient formula is thus fictional, representing an extrapolation to what the gross residential density would theoretically be at the center if nonresidential uses did not preempt the most central locations. Alternatively, it is possible (though more difficult in terms of data availability) to construct the gradient on the basis of net residential density.6

More recent analysis by Muth, Berry, Alonso, Mills, and others has confirmed the prevalence of this exponential form of residential density gradient, and has developed and begun to test some useful explanatory hypotheses about its determinants.7

In brief, it appears that:

1. Larger cities have, in addition to higher central densities, lower slope coefficients (i.e., flatter slope).

2. The observed decline of population per acre with increased distance from the city center is actually a combination of at least three different gradients. As we go outward from the center, the number of housing units per acre falls, and so does the proportion of people living outside of households (for example, in institutions, hotels, and rooming houses); but the declining density effect of those two variables is partially offset by rising household size.

3. The central density is largely determined by conditions (such as transport, communication, production technology, income levels, and occupation structure) during the period when the city became established. Once set, the basic form of the city (particularly in the central area where investment in structures is heaviest) is subject to considerable inertia. At any given time, then, the age of a city (definable in terms of the date at which it attained some specified minimum size, such as 50,000) is highly correlated with its central density. The familiar dichotomy between newer American "auto-oriented" cities such as Phoenix and older "pre-auto" cities recognizes this effect.

Perhaps the most intensive statistical analysis of urban residential density gradients is to be found in Richard Muth’s work. After a series of statistical tests of the relation of distance to gross residential density figures in 46 U.S. cities in 1950 (based on samples of 25 Census tracts in each city), he concluded that "the negative exponential function in distance from the CBD [Central Business District] alone fits population density data for American cities in 1950 rather well."8 This held true despite the fact that there were numerous deviations from regularity and that the exponential formula as a regression equation accounted for only about half of the observed intracity variation of density (among tracts within any city). The fitted density gradients varied widely in their slopes, with b ranging from 0.18 to 1.2. These b values correspond to density declines of 17 percent and 70 percent respectively for each mile of distance.

Muth then looked for factors to explain why some cities had steeper density gradients than others. He found that flatter gradients (that is, lower values for b) were significantly associated with each of the following characteristics of the urban area:

• High automobile ownership

• High income level

• High proportion of nonwhite to total population

• Large size (population) of urban area

• Low degree of concentration of the metropolitan area’s manufacturing employment in the central city

• Low quality (in terms of condition and plumbing facilities) of housing in the central city

Finally, he found by further analysis that "the distribution of population between the central city and its suburbs and the land used by the urbanized area are largely governed by the same forces influencing the population distribution within the central city." Two main qualifications to this general statement appeared. First, an influx of lower-income persons into the central city is apparently associated with a greater degree of suburbanization of population, whereas within the central city the effect is in the opposite direction (a steeper density gradient). This as Muth suggests makes sense, considering that the central city is a separate fiscal unit and the presence of a larger low income group tends to make the tax burden heavier for the upper income groups and for business firms, whose incentive to escape to other jurisdictions is thereby increased.9

Various empirical investigations have brought to light similar fairly consistent density gradients for certain nonresidential types of land as well. Otis Dudley Duncan presents a gradient of manufacturing employees per thousand square feet of land occupied (that is, net manufacturing employment density) for Chicago in 1951, showing a reasonably good fit to the exponential formula, with a slope substantially flatter than that of the typical residential density gradient.10 Daytime population likewise shows the same kind of gradient. In this case, the slope is much steeper and the central density much higher than for residential population. Finally, it appears that the gradient of land values in urban areas also follows the same general exponential form.

Analysis of the behavioral factors underlying these gradient patterns poses many complications. If all households could be assumed alike in preferences and place of work, the form of the gradients of residential densities and rents could be read as representing the individual household’s trade-off between more space and quicker access. But it is not so simple. We know that this tradeoff is affected by income level. Higher-income families tend to live farther out than lower-income families, particularly if allowance is made for presence or absence of young children. This means that the observed overall residential density and land value gradients represent in part the gradation of trade-offs: The analysis of residential distributions involves an additional dimension. Similarly for "manufacturing employment density"—in the case of the employment density gradient referred to earlier, a breakdown of manufacturing into twenty-five industry groups disclosed that they displayed very different degrees of centrality, associated with employment density. A still finer breakdown would, of course, show the same kind of differentiation within an industry group.

7.3.3 Land-Use Zones: The Burgess Model

It is clear that to go beyond such elementary explanations, some explicit attention must be paid to the heterogeneity of both residential and non-residential land uses in a more complicated conceptual scheme. Early attempts in this direction relied on highly descriptive characterizations of urban areas.11

The Burgess zonal hypothesis is a schematic model developed along these lines in the 1920s. 12 Its kinship with von Thünen’s much older zonal model of rural land uses around an urban focal point and modern analyses of urban land use (see Chapter 6) is obvious. Activities are grouped on the basis of concentration in successive distance zones from the center outward, in this order:

1. Central business district activities: department stores and smart shops, office buildings, clubs, banks, hotels, theaters, museums, organization headquarters

2. Wholesaling

3. Slum dwellings (in a zone of blight invaded from the center by business and light manufacturing)

4. Middle-income industrial workers’ residences

5. Upper income single-family residences

6. Upper income suburban commuters’ residences

This research is an important example of inductive generalization applied to regional analysis. Burgess moved from his descriptive exercise to put forward a simplified dynamic model. The Burgess hypothesis was that these land-use zones preserve their sequence, but as the city grows each zone must spread and move outward, encroaching on the next one and creating zones of transition and land-use succession. He emphasized the transitional problem created in the third (blighted) zone.

In the Burgess model, we have an elementary classification of urban land uses by locational types that is still useful as a starting point. Downtown uses, light manufacturing, wholesaling, and three or four levels of residence characterized by income level are singled out as significantly different and important location types. Finally, heavy industry is not in the Burgess model at all, which makes sense in the light of the location factors discussed earlier. Heavy industry requires large level sites with good transport to and from the outside world, and access to the urban "center of gravity" is of little relevance since most of the inputs (except labor) and outputs are nonlocal.

One of the most important generalizations introduced by the Burgess model concerns residential locational preferences. In his scheme, the richer people are, the farther they live from the city center. As mentioned in Chapter 6, this pattern is characteristic of cities in the United States even at the present time. However, the analysis of residential location behavior developed in that chapter (see Section 6.6) made it clear that such a pattern is not universally relevant. Rather, personal preferences and characteristics of individual economies, such as the nature of transfer costs in the daily commute to work, can account for the location patterns of heterogeneous income groups. Nevertheless, the concept of land-use succession and the transition of neighborhoods from one income group to another have figured prominently in shaping the spatial patterns of metropolitan areas.


Some approaches to the explanation of urban spatial patterns have stressed tendencies toward differentiation according to direction, rather than according to distance from the center. The sector theory is associated historically with Homer Hoyt and has been stated as follows: "growth along a particular axis of transportation usually consists of similar types of land use. The entire city is considered as a circle and the various areas as sectors radiating out from the center of that circle; similar types of land use originate near the center of the circle and migrate outward toward the periphery."13 Hoyt’s formulation was mainly concerned with residential land use and assigned a dominant role to the forces determining the direction of expansion of the highest-class residential district.

In terms of the existing pattern at any given time in an urban area, it is easy to explain sectoral differentiation on the basis of such factors as (1) topographical and other "natural" variation, (2) the presence of a limited number of important radial transport routes, and (3) the previously discussed incentives toward a greater concentration of any one activity than a symmetrical concentric ring layout would afford. But the Hoyt hypothesis is couched primarily in dynamic terms, as an explanation of persistent sectoral differences in the character of development. And in that context, it introduces two further useful concepts.

One of these concepts is that of succession of uses of a given site or neighborhood area. Except at the outer fringe of urban settlement, each type of land use as it expands is taking over from an earlier urban use; by and large, the growth process involves (as described earlier in the context of the simple monocentric model) an outward encroachment of each type of activity into the next zone out. Some such transitions are cheaper or easier than others, and the extension tends to be in the direction of easiest transition. Thus obsolete mansions are conveniently converted into funeral homes; row houses and apartments are easily converted, subdivided, and downgraded into low-income tenements; and obsolete factory space is easily used for wholesaling and storage. The "filtering" theory of succession of uses in the urban housing market implies gradual and continuous, rather than abrupt, change in residential neighborhood character.

The other useful concept might be called minimum displacement. The growth process uproots all kinds of housing and business activities in the zones of transition, forcing them to seek new locations. Copious empirical evidence bears out the reasonable presumption that when these moves are made by householders or by small neighborhood-serving businesses, there is a strong preference for remaining as close as possible to the old location. This cohesion or inertia, which is quite rational in the light of both economic and social considerations, tends to perpetuate a sectoral differentiation and to cause a particular activity to move gradually outward along the line of least resistance, rather than into another sector.


Although a city or metropolis generally has one identifiable main center, there are subordinate centers as well. Spatially, an urban area is multinuclear, and some models of urban spatial structure particularly stress the development of subcenters. Recent trends have entailed the rapid sprawl and coalescence of originally discrete cities and towns into larger metropolitan and megalopolitan complexes, bringing this multinuclear aspect into prominence as a basic characteristic of the urban pattern. Even a small individual city usually contains a number of important business centers or other focal points outside the central business district.

Any consumer-serving activity that can attain its economies of scale and agglomeration without having to serve the entire urban area from a single center will increase its proximity to consumers by branching out into shopping centers, each serving a part of the whole area.14 Each shopping center is in turn a concentration of employment activity, a focal access point for work, shopping, and recreational trips. The basic concentric patterns of access advantage, centripetal movement of people, and centrifugal movement of goods and services are replicated in each part of the urban area, albeit for a more limited range of activities than those represented downtown. Local peaks of the gradients of residential density, land values, intensity of land use, and access potential appear around each of these subcentral points, like hillocks on the shoulders of a mountain.

While part of the subcenter phenomenon can be explained, as above, on the basis of its efficiency in providing consumer-serving activities, other forces are in effect. This is evident as soon as we recognize that among the types of activity that usually do agglomerate in one place within an urban area, there are many for which the central business district simply is not an economic location. These activities are highly concentrated but typically off-center.

For some activities, the basic reason is inherent in their production functions—they do not use space intensively enough to afford downtown land, but at the same time their internal-access requirements call for a more compact zone of occupation than a ring would provide. This case was examined in section 7.2.4, with a university campus as the example. Off-center cluster is the typical pattern for research centers, cultural centers, concentrations of automobile salesrooms, and to an increasing extent, wholesale produce markets and other wholesaling activities with strong external economies of cluster but substantial space requirements.

There is an interesting exception to this principle of "blob rather than doughnut." The building of fast suburban beltways around major cities has made it more feasible for some activities (for example, electronics and other light industries) to assume an extended distribution along at least a sizable arc—that is, part of a doughnut.

Second, the tendency to concentration at the expense of symmetry is found in specific types of residential land use as well, reflecting among other things the preference for neighborhood homogeneity that acts like an agglomerative force for any particular class of residence (such as high-income single-family houses) even where low densities are involved.

A still further basis for off-center concentration appears in situations where the activity serves a market that is itself lopsidedly distributed in relation to the overall area. For example, if residential areas occupied by higher income and educational groups are predominantly to the northwest of the city center, trade and service activities catering especially to those groups will find the point of maximum market access potential somewhere northwest of the city center. This pattern also applies for those activities that mainly serve markets outside their own urban area (such as export activities). Access considerations for such activities dictate location close to intercity transport terminals or major highways.

Finally, special topographical or other site features may make a particular off-center location optimal even though it does not have the best access. The availability of a large level tract amid generally hilly topography may well be the decisive factor for such uses as airports or major industrial developments.

Thus a typology of urban subcenters might include:

1. Retail shopping subcenters each serving a surrounding residential area

2. Subcenters based primarily on nodal advantages of transport—for example, at junctions of major traffic arteries or transit routes

3. Subcenters based essentially on a single large-scale unit, such as a major industrial plant or sports stadium

4. Subcenters that were formerly separate towns, now engulfed by the spreading metropolitan area

5. Subcenters based on transport terminals connecting to the outside world— for example, near airports

6. Subcenters based on special natural advantages of site

Any particular subcenter may, of course, qualify under more than one heading.


We have discussed the location of activities within cities in terms of four simple schematic models: the density gradient, Burgess’s concentric land-use zones, sectoral differentiation, and systems of subcenters. Each of these throws into relief some recognizable features of urban patterns, though none provides by itself a really good likeness.

These simple analytical constructs are not to be regarded as rival, mutually exclusive theories of urban form. They are, in fact, mutually consistent and complementary, and each has something to contribute to our understanding of the whole pattern. Subcenters merely represent a replication of the basic concepts involved in the density gradient and concentric zone models; namely, an ordered sequence of land uses of different intensities and types around a common focal point. In the view that emphasizes sectoral differentiation, there is still the idea of an outward spread from a center and a recognition of the agglomerative tendencies of particular types of land use. Shifts associated with urban growth and change can be, as we shall see in the next section, analyzed in terms of all four of the basic constructs set forth in this chapter.

It should be noted also that even the simplified economic models of urban spatial form developed by theorists and econometricians usually superimpose substantial refinements and elaborations on the basic density-gradient, zonal, sectoral, or subcenter framework used. For example, some monocentric models of residential density, based on the density gradient concept, have introduced a commuting cost variable that depends not merely on distance to the city center but also on the development density of the territory traversed, which is presumed to affect congestion and therefore speed of travel.

In real cities, spatial patterns are much more complex than in any model (if they were not, models would be unnecessary!) and may appear largely haphazard at first sight. To explain them, we have to analyze in depth the "natural" differentiation of sites and the neighborhood linkages between activities to which the sector and subcenter theories merely allude. We have to take into account the network and nodal structure of urban transport, which makes variation in access advantage less simple and continuous than smooth gradients and nice round concentric zones would suggest. Specifically in the case of retailing areas, we have to recognize the pattern of ribbon development wherein commercial areas sometimes extend for miles along a single major street in response to the attractions of access to a moving stream of customers rather than to a fixed residential or employment concentration. We must also recognize the locational effects of public decision making as embodied in zoning, housing finance, property taxation, and placement of public facilities.

Most importantly, an understanding of the spatial layout of a city requires some idea of the processes of change. Present locations and neighborhoods embody to a large extent decisions made in the past, when conditions were different. The pattern is always behind the times and involved in a never-ending process of adjustment. Accordingly, we now turn to the subject of changes in the spatial structure of urban areas.


Most of the urban problems that concern us today can be traced to underlying changes in land use, location, or locational advantage that make life or business survival more difficult for some group or groups. The regional economist rightly stresses the spatial origins and implications of such problems—where his peculiar talents are most likely to be relevant. The present section is concerned with the principal kinds of change that have been occurring and seem likely to occur in the spatial patterns of urban areas.

7.7.1 General Effects of Urban Growth

Several simplified models of urban form have been presented, primarily as static descriptions or rationalizations of spatial structure. Let us now put some of these models to work and see what they may be able to suggest regarding dynamic shifts in patterns. First of all, we shall ask them what may be expected to happen simply as the result of urban growth. The locational effects of rising levels of affluence and new technologies of production and transport will subsequently be examined in terms of specific types of urban activities.

One appropriate way to see the structural implications of pure size is to make cross-sectional comparisons among urban areas of different size classes in the same country at the same time. What differences, then, are associated with larger city size as such? Some of the most obvious ones can be rationalized in terms of the basic density-gradient model. Increased total size has both intensive and extensive impacts. The central densities or other measures of peak central intensity rise, while at the same time development pushes farther out. Residential densities in any given zone increase, except that the central nonresidential crater expands. Increases in density are greatest, in percentage terms, at the outer fringe of urban development.

We also envisage (as impacts of growth per se) the successive pushing out and widening of the various more or less concentric zones of activity already discussed in the context of the original Burgess model. An increase in the length of all types of journeys and hauls of goods is likewise to be expected.

But as such journeys and shipments become lengthier and more expensive with expansion of the area, there are adjustments to combat or partially offset the increase in travel time and other transfer costs. Subcenters for various individual activities or groups of activities play a growing role in a larger urban area because the total market in the area, for more kinds of goods and services, becomes big enough to support two or more separate production or service centers at an efficient scale rather than just one. Further, the larger size of the area, with its expanded and more variegated manpower, services, materials, and markets also provides the basis for an increasing number of subcenters of nonresidential activity that are not simply oriented to the neighborhood consumer market but may serve the whole area and outside markets as well.

It would appear, then, that growth as such helps to account for the flattening of density gradients that has characteristically shown up as a trend in our American cities—though there are other important reasons as well.

The picture of changing patterns in an urban area that is simply getting more populous, without major changes in technology or income level, is this. Development proceeds both vertically (more intensive use of space) and horizontally (use of more space). Each specialized zone of activities widens and moves outward, encroaching on its outer neighbor and giving way to its inner neighbor. New types of activities arise in the central area. The variety of types of activity and occupancy increases. Off-center foci of activities increase in number, size, diversity, and importance. The gradients of residential density and land value become higher but flatter. The average length of journeys and the total amount of travel and internal goods transfer increase—but not as much as they would if all nonresidential activity remained as highly concentrated at the center as it was originally. The pattern of transport flow becomes more complex, with more criss-crossing and more nonradial traffic. Traffic studies show that the larger the urban area, the smaller is the fraction of its internal travel that enters the central business district.

With the increased variety of activities, occupations, and life styles represented in a larger area, and the proliferation of more and more orders and types of subcenters, it is clear that an urban area’s growth is associated with a more elaborately differentiated pattern of land uses: more spatial division of labor and more specialization of functions. This increased macroscale heterogeneity fosters, somewhat paradoxically at first sight, increased homogeneity within individual neighborhoods and other subareas, or segregation in the broad sense of the term. We have considered earlier the various pressures for microscale homogeneity within urban areas; and these pressures can operate to a greater degree in the framework of a larger and more varied community complex. One manifestation of this tendency is the magnitude of the problem of de facto racial segregation of schools (that is, reflecting neighborhood composition) in larger cities. Another is the problem (again, most evident in the larger cities) of accommodating intensely cohesive specialized business concentrations such as the Manhattan garment district and urban wholesale produce markets, which are highly resistant to piecemeal moving or adjustment. A third problem, likewise more evident in the largest metropolitan areas, is political and economic conflict between the main central city and the surrounding suburbs, which resist merger or basic coordination with the central city or with one another.

Thus it appears that many of the most pressing problems of larger urban areas today—ranging from traffic congestion to racial discord, city-suburb conflict, and the fiscal crises of central cities—can be traced in some part to sheer size and growth. They are implicit in even the simplest models of urban structure. More broadly still, it is clear that larger agglomerations must raise challenging problems of divergence of private costs and benefits from social ones (and local from overall), in view of the intensified proximity impacts: scarcity of space, pollution of water and air, environmental nuisances, and generally increased interdependence of interests. Such problems are part of the price to be paid for the economic and social advantages of greater diversity of contact and opportunity that constitute the very reason for the city’s existence. In Chapter 13 we shall turn again to these issues and focus more explicitly on some spatial aspects of urban problems.

This hypothetical and mainly deductive picture of trends of change in a single growing area conforms closely, as would be expected, with what we observe empirically in a cross-sectional comparison of urban areas of different sizes in one country at one time. Moreover, we recognize in this picture many familiar features corresponding to observed historical and current trends; and we can infer that simple growth plays a part in accounting for them, and can be expected to exert a similar influence in the future.

7.7.2 Changes in Density Gradients for Major Types of Urban Activity

Observed trends in density-gradient parameters are not fully explicable in terms of the effects of growth per se but reflect also the influences of other factors. Available data indicate definitely that urban density gradients have been getting flatter for many decades at least, and that their central-density parameters have characteristically declined in the present century, at least in the urban areas of more developed countries.15 Similar trends have been found in the density gradients of employment in manufacturing, wholesale trade, and retail trade in a sample of six U.S. metropolitan areas (see Figure 7-2). The lines in this figure are not density gradients; they measure the slopes of the gradients at successive dates. For each activity at each date, Edwin Mills fitted to the historical data a density-gradient formula of the exponential type described earlier in section 7.3.2, in which density of the activity declines by a fixed percentage with each unit increase of distance from the city center. Where the line for a given activity slants downward, as occurs consistently in the figure, this shows a flattening of the density gradient during that time interval.

It appears from Figure 7-2 that trade and service activities (in these cities at least) were suburbanizing faster than residential population, and at increasing rates, for at least three or four decades prior to 1963; and that manufacturing employment tended to suburbanize at a somewhat slower pace between 1920 and 1948 but quite rapidly thereafter.

The flattening of the urban residential density gradient has been shown to extend back to 1880 at least for a smaller sample of four metropolitan areas.16

For the discussion that follows, it is convenient to consider urban activities under four major types with distinctive locational characteristics: commodity-exporting, administrative and informational, residential, and consumer-serving. For each of these we shall identify and try to explain the dominant trends of locational change.

7.7.3 Location of Commodity-Exporting Activities

Commodity-exporting activities are primarily manufacturing industries; though a few urban areas (see Table 9-3) export significant amounts of crops or minerals, and some wholesaling involves exports of goods to a wider area than the city and suburbs. We have just noted some evidence of the suburbanization of both manufacturing and wholesaling.

An important instance of the outward shift of wholesaling is the transfer (in 1969) of the Paris produce market, which actually serves much of the rest of France as well, from Les Halles in central Paris to new quarters at suburban Rungis. Produce markets in many American cities (such as Boston and New York) have been similarly relocated, and wholesale establishments of other types as well are increasingly represented in suburban industrial zones.

In manufacturing at least, this suburbanization trend goes back even further than Figure 7-2 shows. One of the earliest systematic investigations dates it from 1889:

Between 1879 and 1889, manufacturing activity was growing more rapidly in most large metropolitan cities than in the surrounding districts... Since 1889, manufacturing activity has grown more rapidly in the suburban sections surrounding great manufacturing cities than in the manufacturing cities themselves. 17

Improvements in Census data made possible Daniel Creamer’s more detailed analyses for the period since 1899, which are summed up in Table 7-1. Because the data are not presented in precisely comparable terms by all censuses, and because the picture of location shifts is affected by changes in the classification of specific areas as they grow, three different time series are shown in this summary table. It is clear from each series, however, that location types C and F (suburban areas around important industrial cities) have shown faster industrial growth than those cities themselves (location types A and D respectively).18 Suburbanization becomes increasingly apparent in the more recent period; by the 1960s, the popularity of outlying locations for new and expanded manufacturing plants was so obvious as hardly to require documentation. How can this tendency be explained?

More Extensive Plant Layouts. One important reason for this trend emerges from changes in manufacturing technology, relating particularly to the ways in which energy and goods in process are moved about within the plant. Comparing an old factory with a modern factory, one is immediately struck by the high, compact, almost cubical shape of the old, and the low, sprawling shape of the new. The old type dates back to the days when motive power was supplied by steam engines transmitted by belts and shafting, calling for the closest possible proximity of the individual power-using units of equipment. Early in the twentieth century, there was a nearly universal shift to electric power, transmitted to individual motors on each piece of equipment. Since additional cable costs relatively little, much more extensive layouts become possible. This in turn contributed to the adoption of conveyors and assembly-line layouts, in which machines bring the goods to the successive stages of processing or fabricating equipment.

Such considerations did not apply in heavy processing industries requiring tall structures and moving materials through pipes in liquid, gaseous, or powder form (such as oil refineries, primary chemical plants, smelters, cement plants, flour mills, distilleries, or breweries). Nor did they apply to small-scale light industries that could effectively operate in rented upstairs space in loft buildings and were, in general, strongly dependent on external economies of cluster. But for nearly all other types of manufacturing, the attractions of a horizontal layout became large. With this increased desire for more spacious sites, the enticements of the cheaper land of the suburbs were naturally strong.

The desire for more space has had other bases as well, such as a growing tendency to anticipate expansion needs, increased emphasis on amenity and visibility, the need to provide parking space, and a fear of being hemmed in by surrounding development.

Impressive evidence of the increased appetite for space emerged from a comprehensive economic study of the Pittsburgh region in the early 1960s. Relevant findings were:

[Plants relocating within Allegheny County, 1957-1959]

In the eleven cases in which the area of site and of buildings at both old and new locations were specified, the average site area per plant had increased from 4.6 to 19.6 acres, or 300 percent, and the average building area per plant had increased from 90,000 to 122,000 square feet, or 36 percent. [This sample consists primarily of rather large manufacturing plants.] The much greater expansion of site area than of building area indicates a desire for more open space for storage, loading, and parking, and for subsequent expansion. The site area per employee was at least doubled in each of these eleven relocations, and was increased by a factor of more than 20 in two cases.

[Plants that had not relocated but reported need for more space]

The average estimate of additional space [site area] required was 153 percent, but the average estimate of increased employment associated with those requirements was only 38 percent. These figures imply a desire to increase the amount of space per employee by 83 percent.

[Respondents, primarily occupying rented space in multitenant buildings, who reported need for more floor space]

Although only fourteen of the respondents reporting inadequate floor space gave the requested information on amount of additional space needed and additional employment expected, in all but one of those cases the percentage increase in floor space was at least as great as the increase in employment. On the average, 138 percent more floor space was called for, with an associated increase in employment of only 44 percent. These figures imply a desire to increase the amount of floor space per employee by 65 percent.19

Changes in Transport Technology. Another change contributing to the suburbanization of commodity-exporting activities comes from transport technology, and specifically from the improvement of motor vehicles and highways that enabled a good part of the inputs of such activities and a still larger part of their outputs to be shipped by truck. This change became important in the 1920s. Earlier, manufacturing establishments relied heavily on the horse and wagon for the intraurban movement of commodities, while the interregional shipment of a large portion of materials inputs as well as their outputs was effected by rail.

In an insightful analysis of the effect of transport on urban spatial patterns, Leon Moses and Harold F. Williamson, Jr., point to changes in the relative cost of interregional versus intraregional transfer of commodities as an important factor encouraging decentralization.20 The efficiency of rail transport depends to a large extent on scale economies associated with freight handling and large-lot shipments. During the early stages of urban development, it was often the case that no more than one central terminal could be maintained economically in a given city. By clustering about the central terminal, manufacturers could benefit by receiving shipments directly from rail sidings. Also, the clustering minimized the distances involved for shipments among local establishments—an important consideration given the inefficiency of the horse-drawn wagon.

The influence of truck transport came in two stages. Moses and Williamson point out that early in this century (1900—1920) the truck replaced the horse and wagon for intraurban shipments but that manufacturers were still tied to rail transport for shipments to and from the city. In this early phase, locational ties to the urban "core" were weakened; as the cost of intraurban transfer was reduced, suburbanization was encouraged. However, the full impact of truck transport was not realized until much later. As the interstate highway system became more fully developed (after World War II), suburban export terminals became common, and the second phase of decentralization came into full swing.21 It has most strongly affected wholesaling and the lighter types of manufacturing that ship high-value outputs in small consignments; but even steel mills and other heavy industries have come to ship substantial parts of their output over the roads.

However, an interesting reversal occurred in the 1960s in the method of transporting new automobiles from factories and assembly plants. Statistics compiled by the Automobile Manufacturers’ Association show that in 1959, 90 percent of such traffic was by road and only 8 percent by rail. The railroads then devised equipment and tariffs that made it more economical to ship by rail for medium and long distances, and by about 1970 the rails were carrying the majority of new cars shipped. Barge shipments of new automobiles, which had been nearly 8 percent of the total in 1949, had become insignificant by the mid-1960s.

The use of highway transport greatly widens the choice of locations since the road network is many times finer than the rail network and offers an almost unlimited choice of stopping places. For direct shipment in whole truckloads, there is no need to be near any transport terminal, and many piggyback loading yards have been conveniently placed for suburban access. An outlying plant location speeds the receipt and delivery of goods by obviating slow and expensive trucking through congested city streets.

Access to Labor Supply. A third factor contributing to industrial sub-urbanization is labor supply. Moses and Williamson have argued that in the earliest years of this century, the intraurban movement of people was much more efficient than the intraurban movement of goods and services. Trolleys and commuter railroads freed workers from residing in close proximity to the downtown, even while many manufacturing establishments were still tied to locations at or near central freight terminals.

During this early period, the urban center was truly a hub of economic activity, where streetcars and railroads brought workers and commodities together on a daily basis. However, as automobile ownership became a common characteristic of urban life, the locational consequences of a decentralized labor force became more apparent. The urbanization of population and its motorization have made it feasible to attract an adequate work force to locations outside of any major population center, and business location decisions now reflect labor’s local mobility. As noted in the last section of Chapter 10, some suburban locations are better than downtown in terms of access to the supply of high-income professional personnel. Locations in beltway zones can provide quick access to labor from a sizable arc of the metropolitan circumference.

This does not exhaust the list of reasons for the increased attractiveness of suburban locations for exporting activities. Business firms have become increasingly influenced by amenity, prestige, and public relations. A suburban location with attractively landscaped grounds, exposed to the view of thousands of daily travelers on a busy expressway, has an advertising value not to be underrated.

Finally, it is important to note that, in the aggregate, all of the forces motivating suburbanization acquire further importance from the changing composition of productive and distributive activities. Higher income levels and the proliferation of products, brands, and successive stages of processing mean an increasing proportion of the lighter types of activity— those involving relatively little weight loss or orientation to transported inputs and relatively high sensitivity to quick market access, environmental amenity, and local public relations.

7.7.4 Location of Administrative and Other Information-Processing Activities

A rapidly increasing proportion of activities produce intangible outputs that are delivered through personal contact or communications media, with little or no shipment of any actual goods.22 Since new information obsolesces rapidly (yesterday’s newspaper is trash, and last week’s memo may serve only to clutter the files) and since human time is expensive, market-access advantage for such activities is measured primarily in terms of time.

Technological advance has greatly speeded long-distance communication and personal travel, though in our time there has been relatively slight improvement for the short haul. The locational impact is clearly visible in the rapidly growing operations of administration, data processing, and research. Individual business corporations have been increasingly consolidating such operations at headquarters and reducing the relative importance of field offices. The unchecked trend toward business amalgamation, which in the 1960s involved a striking increase in "conglomerates," or multi-industry corporations, has played a part in this trend; for the acquiring firm customarily adds to its headquarters staff and drastically cuts the headquarters staff of the acquired firm even when the latter retains its name and the status of a division of the larger complex. New York and other headquarters cities have been frantically erecting new downtown skyscrapers since World War II to keep pace with an apparently insatiable demand for office space.

Within urban areas, headquarters offices have been rather tightly concentrated within the central business district. This concentration can be ascribed to the multifarious daily interfirm contacts required (and also, to some extent, to the prestige value of new skyscrapers and downtown addresses, and the stake that some large corporations and related financial institutions have in downtown property values).

At the same time, the suburbs hold strong attractions for office and informational activities that are least subject to the access needs and external economies of downtown cluster. As the "head office" activities of large firms have grown, they have at the same time tended to split into downtown and suburban (or even nonmetropolitan) categories. Routine data processing and other clerical work can fairly easily be shifted out of expensive downtown office space, leaving the "top brass" behind. The major concern in a split is access to adequate clerical manpower and womanpower in the suburbs. For research laboratories, the advantage of the suburbs is much more positive, and this is reflected in their customary location. Suburbanizing factors include need for ample space; proximity to the preferred residential areas of professional workers and technicians; access to universities and scientific institutions; absence of undue noise, distraction, air pollution, vibration, and the like; and a degree of isolation from inquisitive competitors and from the distracting demands of the production divisions of the same firm for solutions to their day-to-day production problems.23

Table 7-2 provides some data applicable to the two major categories of employment just discussed, though it covers the headquarters offices and research facilities of manufacturing firms only. We note in this table the very rapid growth of both activities in the period covered and the strong concentration of central-office employment within central cities, and of research employment in more peripheral locations.

It would appear from the data in Table 7-2 that downtown and other central-city areas are rapidly losing their hold on central-office employment. The period covered, however, was only nine years; and a more recent and more intensive study of the location trends in such employment foresaw much less drastic decentralization. This study was conducted by the Regional Plan Association of New York and was primarily directed at assessing the position and prospects of office work in the New York region, but it reached the following important conclusions at the national level:

Large metropolitan areas are and seem likely to remain [the nation’s] dominant office centers....
…The central business districts of the nation’s largest 21 metropolitan areas have been, on the whole, holding their own in the past decade (roughly the 1960s]; while population decentralized, offices did not…
Because office jobs are suited to city centers, they offer the nation a chance to harness private enterprise to renew older cities and keep them attractive to all income and ethnic groups.24

Table 7-2 shows the research and development employment of manufacturing firms increasingly concentrated in the suburbs and satellite communities of metropolitan areas. It is likely that the bulk of such jobs shown as located in cities were actually in establishments well outside the central business district.

Two other major categories of research facility are those of government agencies and commercial research firms. The locational considerations are quite similar to those already cited for research laboratories of manufacturing firms, except that there may be no separate downtown headquarters office. For example, a 1966 report on research laboratories in the Washington, D.C., area (see map, Figure 7-3) observed:

The picking up of the research business coincided happily with the opening of the 65-mile, six-lane Beltway, which rings the District of Columbia about 10 miles from the center. Just as the small companies were beginning to outgrow their original quarters, the Beltway opened up to give swift access from anywhere in the Maryland-Virginia metropolitan area to the 14 largest federal labs. The highway.., runs through some wooded areas that are ideal for development as industrial parks, and are shielded by some of the nation’s toughest residential zoning laws. Smokeless, tidy R&D [research and development] is about the only industry that home-conscious residents will tolerate in Maryland’s Montgomery and Prince George counties, and Virginia’s Arlington and Fairfax counties.
More than a dozen companies immediately set up shop near the Beltway, including four in the publicized "new town" of Reston, Va. . . . Local boosters predict a research boom on the Beltway rivaling that on Boston’s Route 128.25

This prediction has proved to be quite accurate.

7.7.5 Residential Location

Urban populations have become richer, more leisured, and more widely mobile in terms of their day-to-day journeys within urban areas. These changes have been associated with more dispersed residential location patterns. Analysis of the residential density gradients, as noted earlier in this chapter, discloses that such gradients are flatter in cities where income levels and car ownership are higher, and that the rich characteristically live farther out than the poor, particularly if they have children.

Tables 7-3 and 7-4 provide some relevant evidence of the major trends. When we simply compare the central cities of metropolitan areas with the remainders of those areas and with the nonmetropolitan United States, it is clear that the bulk of American population growth between 1950 and 1980 took place in metropolitan suburbs. Nonmetropolitan areas, which had grown much more slowly than metropolitan areas during most of this period, also had substantial population increases in the 1970s.26 The suburbanization trend is also characteristic of the nation’s black population. Although in the 1950s and 1960s the black population increased more rapidly in central cities than in suburbs, the 1970s were a period of suburbanization for blacks. Table 7-3 shows also that sometime between 1950 and 1960, the black population became more metropolitan than the nation as a whole, and that blacks were more than proportionately represented in central-city populations as early as 1950 and have become increasingly more so, even in the face of the suburbanization of blacks in the 1970s.

Table 7-4 is taken from one of the reports of the New York Metropolitan Region Study of the late 1950s and refers to a broad belt of municipalities within the New York metropolitan area intermediate between New York City itself and the outer ring of suburban or exurban territory. In this table, individual communities are classified by income level, and various characteristics are shown for each income class. Higher family income appears strongly associated with smaller communities, lower residential density, prevalence of single-family dwellings, rapid population growth, and distance from Manhattan.

It appears, then, that (1) urban population in the aggregate has been rapidly suburbanizing, (2) higher-income people have shown the strongest preferences for suburban location (see Section 7.3.3.), and (3) blacks remain highly concentrated in central cities, even though they have joined in the suburbanization movement in recent years.

Since part of the explanation of the overall suburbanization of population lies in rising levels of income and leisure, and since the wealthier can more easily afford spacious sites and modernity, we are not surprised to see the upper-income groups leading the outward trek and continuing to live farther from the center than those with lower incomes. So far, relatively few upper-income people, mainly those without children, have moved into close-in areas despite the access advantages and amenities now available.

The migration patterns reflected in Table 7-3 imply financial incentives that may have encouraged the suburbanization trend. The large influx of low-income blacks to central cities in the 1950s, coupled with the mobility of higher-income whites, left many older cities with serious fiscal problems. Higher-income individuals could avoid much of the tax burden associated with the rapid in-migration of low-income individuals by moving to the suburbs (see section 7.3.2). Thus urban fiscal distress is seen by some as being caused by suburbanization, while at the same time that suburbanization may well have been one consequence of the fiscal pressures exerted by in-migrants to the urban areas.27

One of the most important factors promoting suburbanization is government subsidy to home owners. The federal government has had an explicit policy of encouraging home ownership since 1934, when the Federal Housing Administration (FHA) was created. Prior to that time, lending institutions typically would extend loans for only about 50 percent of the market value of a home, and the term of a mortgage was usually less than 10 years.28 Mortgage loans that are insured by the FHA against risk of default have much more favorable provisions from the homeowner’s perspective. The terms of such loans run to 30 years, and much lower down payments are required.

Tax policies also have made it easier to purchase a home. The federal government presently allows homeowners to deduct the full value of interest payments and property taxes from their taxable income. Given the progressive nature of the federal income tax, this means proportionately larger savings for higher-income households. A less obvious—but nevertheless important—subsidy is involved also in the failure to tax homeowners for the" value" of their dwellings. A person who owns rental property must pay tax on rental income; thus rent is a measure of the occupancy value of the rental unit. Such a value may be imputed to owner-occupied dwellings as the amount that the owner would have to pay in order to obtain comparable housing in the rental market. The failure to tax this imputed income biases investment away from rental units and toward owner-occupied units.

The combination of these subsidies has made home ownership more affordable and attractive.29 Since single-family homes are extensive land users (as compared to multifamily dwellings), the bulk of housing development of this type naturally takes place where the price of land is low— namely in the suburbs.

The aging of housing and neighborhoods also plays an important role in shifting residence patterns. According to the filter-down theory, housing deteriorates with the sheer passage of time. Thus if new housing is bought mainly by the well-to-do, housing units will in the course of time be handed down to occupants lower and lower on the income scale. Each stratum of urban society except the top will have access to housing relinquished by the stratum above.30

There is substantial correlation between the age and the condition of structures. Moreover, housing (and the same applies to nonresidential structures) can become less useful with the passage of time, independent of any physical deterioration. Preferences change. The design of a house that was well adapted for a typical well-to-do family of 1890 or 1920 may not correspond to what a similar family prefers in the 1980s in the context of newer alternatives. Neighborhood land-use layouts in terms of lot sizes, front and back yards, block sizes, street widths, and the like are likewise vulnerable to obsolescence and loss of favor in the face of changing conditions and tastes. Finally there is the factor of prestige attached to newness per se, whether it refers to the family car, the family dwelling, or the neighborhood.

Nevertheless, there are a fair number of instances in which old neighborhoods and old housing are visibly involved in a filter-up process. Small-scale remodeling and larger-scale conversion can play a significant role in housing market adjustment. Indeed, while this has meant that some neighborhoods in older cities have been judiciously refurbished (as exemplified by Georgetown in Washington, D.C., and Beacon Hill in Boston), it has also been an important mechanism by which entire suburbs may change to accommodate higher-income residents as a city grows.

The net residential density of new developments on the suburban fringe responds to cyclical ups and downs in land prices, construction costs, and housing demand, as well as to shifts in the relative demands of various income groups for such housing. For example, in the New York metropolitan region during the 1950s, average lot size in new suburban subdivisions was growing by roughly 4 percent per annum.31 But a number of considerations suggest that this rapid growth in size at the margin has not been maintained since, and that the density in new fringe settlement may even have risen.

One such consideration is the rise in land prices, construction costs, and interest rates, which have made spacious and spaciously sited dwellings more and more costly. Another important factor is the ability of different income groups to bid for new suburban housing. The Burgess hypothesis described the twentieth-century American norm in terms of the rich living farther from city centers than the poor. One rationalization for this pattern was the fact that only upper-income people could afford to indulge a preference for new housing, and that such people were also more generally provided with automobiles (thus being more independent of public transportation) and with the leisure time to enjoy suburban living.

But these perquisites have become less exclusive. Car ownership has extended gradually to all but the rock-bottom income group; lower-income people quite often work shorter hours than people with higher incomes and greater responsibility; and finally, various types of mobile and modular homes have appeared, making new detached private housing at last available to people over a wider range of the income distribution. The point here is that these developments have brought medium-income and lower-income housing at rather high densities to fringe areas. Some single-family areas, of course, are at any given time shifting to multifamily development by conversion of large old houses and erection of new apartment buildings; and densities at that stage jump to a much higher level. But we observe also that the peak densities associated with inner-city slums seem to be generally lower now than in former times.32

The foregoing discussion discloses a number of the factors that help to explain the observed trend toward flattening of residential density gradients. One of the important explanations is, of course, the increasing decentralization of nearly all types of employment. Admittedly, there is a degree of circularity in explaining residential suburbanization on the basis of an increasingly suburban pattern of employment opportunities and at the same time explaining the suburbanization of business activities on the basis of access to an increasingly suburban consumer market and labor force. On each side of the relationship, however, other decentralizing factors have been noted. The market and commuting linkages between residence and jobs merely reinforce the suburbanization incentives to which each is subject.

7.7.6 Location of Consumer-Serving Activities

Consumer-serving activities (of which retail trade is the largest category) have been subject to some interesting locational shifts, different from those of any of the other categories of activity so far discussed.

The important location factors for consumer-serving activities within an urban area are (1) access to population, (2) economies of scale and agglomeration, and (3) space requirements affecting the activity’s ability to bid for expensive land.

Market access means somewhat different things for different types of retailing or consumer servicing. Convenience goods such as cigarettes, magazines, or candy bars are bought mainly "on the run" by people bound on some other errand to which these purchases are incidental, and the market is a moving stream of possible customers. Similarly, filling stations are located along major streets with moderate to heavy traffic density.33 Daytime population, rather than simply residential population, is the relevant measure of market for a wide range of goods or services that are bought both by housewives and by employees during lunch hour. Access to residential or nighttime population is the most relevant for stores and shopping centers to which most journeys are made from homes (such as food supermarkets). Finally, some kinds of specialty shops and services—for example, large bookstores, antique shops, and luxury boutiques—cater mainly to certain groups of the population, who may be concentrated in particular neighborhoods.

We have already seen that both residential and daytime population distributions have been suburbanizing as a result of growth per se plus changes in income, transport, and job location. It is no surprise, then, that consumer-serving activity in general has shown a similar outward trend. Downtown department stores, for example, have not flourished. Many have disappeared or merged, and many of the rest have established suburban branches or even shifted outright to a suburban location. Restaurant, theater, and hotel-motel businesses have reacted similarly.

Two factors other than market access, however, have also affected retail and consumer-service location trends. One of these is agglomeration economies. The degree to which such economies can be realized is limited (as Adam Smith said long ago) by the extent of the market, or the number of customers who can be attracted to any one location. The motorized shopper can not only travel farther but can also buy in larger quantities at one time (for example, a whole week’s food shopping at the supermarket), which makes a long journey more worthwhile. At the same time, some kinds of stores have been able to realize new scale economies by labor-saving store layouts and mechanized sales, and to exploit still further the advantages of goods variety that depend on the sales volume of a store or a cluster of competing stores. Consequently, the broad pattern of decentralization within urban areas has been associated with increasing agglomeration in large stores and large shopping centers.

Finally, the location patterns of consumer-serving activities (except the sidewalk-convenience type) have been significantly affected by the larger space requirements imposed by the need for parking space. This consideration reinforces the trend to the suburbs, and perhaps also reinforces the tendency to cluster in large shopping centers, where the pooling of parking space can lead to its more efficient utilization.


Location within urban areas is especially affected by need for movement of people and direct personal contact, with time consequently playing the major role in transfer costs and access advantage. Complex linkages among units and activities, and competition for space, are also important location factors in the urban context. The monocentric land-use models developed in Chapter 6, which emphasized these factors, abstract from other characteristics of urban spatial structure.

Various highly simplified models of urban spatial form are helpful in analyzing the operation of the more basic location factors. Simplest of all is the concept of land-use intensity rising to a peak at the city center, as predicted by monocentric models of urban land use. Residential density and certain other variables do characteristically decline with distance from the center at a rate expressed by a density-gradient slope; and inter-city differences in the slope reflect such characteristics as city size, availability and cost of transport, income, and the age of the city.

An early, descriptive analysis of urban patterns pictured the city in terms analogous to the land-use models discussed in Chapter 6, with a sequence of concentric zones devoted to different broad types of activity. Land-use succession within urban areas can be characterized as changes in the size or spacing of such zones.

Real urban land-use patterns depart drastically from the concentric ring scheme for many reasons. Direction from the center can be as important as distance because of topography, major transfer routes, and intracity cluster economies and other forces promoting neighborhood homogeneity and specialization. Further, any urban area of substantial size has, in addition to its main center, a number of subcenters.

Increased population in an urban area, independently of any changes in income or technology, helps to explain many of the major observed trends in urban form and travel patterns, such as more extensive and intensive use of space, flattening of density and rent gradients, longer intracity journeys and shipments, and a diminished role of the central business district relative to subcenters and suburbs.

Commodity-exporting activities in cities (primarily manufacturing and wholesaling) have been decentralized for many decades as the result of a number of factors, including requirements for more spacious layouts and sites, use of highway transport for both goods and workers, a more decentralized and mobile labor force, and in some cases the attraction of suburban amenities.

Administrative and other information-handling activities have been locationally affected by revolutionary improvements in communications and data processing, as well as by a strong attraction toward the suburbs for the upper-income workers involved. Suburban locations predominate for research facilities and have attracted much office employment as well, though the central business districts of large cities have kept some of their old preeminence as locations for corporate headquarters.

Residential location patterns in urban areas have been decentralizing (as measured by flatter population density gradients) since the latter part of the nineteenth century at latest. This trend appears associated with larger city size, more income and leisure, and more widespread automobile ownership. The increasing concentration of black metropolitan populations in the central cities seems to have come to a halt by 1970. In the decade to follow, there were substantial increases in the black populations of metropolitan suburbs. In the shift to suburban homes, especially by higher-income families, government policies that subsidize home ownership and preference for newer housing have been important factors.

Consumer-serving activities such as retail trade have in general followed population shifts, but at the same time they have clustered increasingly in subcenters because of scale and other agglomeration economies and the enhanced mobility and affluence of their customers.



Monocentric urban models

Minimum displacement

Density gradient


Gross and net residential density

Ribbon development

Central density

Commodity-exporting activities

Burgess zonal hypothesis

Information-processing activities

Land-use succession

Filter-down and filter-up of housing

Sector theory



Harland Bartholomew, Land Uses in American Cities (Cambridge, Mass.: Harvard University Press, 1955).

J. V. Henderson, Economic Theory and the Cities (New York: Academic Press, 1977).

Edgar M. Hoover and Raymond Vernon, Anatomy of a Metropolis (Cambridge, Mass.: Harvard University Press, 1960).

Edwin S. Mills, Urban Economics, 2nd ed. (Glenview, Ill.: Scott, Foresman, 1980).

Leon Moses and Harold Williamson, "The Location of Economic Activity in Cities," American Economic Review, 57 (May 1967), 211-222.

Raymond Vernon, Metropolis 1985 (Cambridge, Mass.: Harvard University Press, 1960).

William C. Wheaton, "Monocentric Models of Urban Land Use: Contributions and Criticisms," in Peter Mieszkowski and Mahlon Straszheim (eds.), Current Issues in Urban Economics (Baltimore, Md.: Johns Hopkins University Press, 1979), pp. 107-129.


1. Some of the content of this chapter is adapted from Edgar M. Hoover, "The Evolving Form and Organization of the Metropolis," in Harvey S. Perloff and Lowdon Wingo, Jr. (eds.), Issues in Urban Economics (Baltimore: Johns Hopkins University Press, 1968), pp. 237-284.

2. See Richard F. Muth, Urban Economic Problems (New York: Harper & Row, 1975), pp. 87-92, for a simple theoretical statement of this point.

3. A considerable (and rising) proportion of journeys from homes are to destinations other than downtown; and for most nonresidential activities as well, the markets and input sources can be at many points other than the city center. In recognition of this, the "access-potential" approach to interaction over distance, previously discussed in Chapter 2, has been used for the empirical analysis of travel patterns within urban areas. The method is of some value in describing and predicting transportation demands, residential development patterns, and locational choice for consumer-oriented activities (retail trade and services). See, for example, T. R. Lakshmanan and Walter G. Hansen, "A Retail Market Potential Model," Journal of the American Institute of Planners, 31, Special Issue on Urban Development Models (May 1965), 134-143. In this study of shopping centers in the Baltimore metropolitan area, it was found that the actual sales at the various centers (or, in some cases, the number of shopping trips to those centers, estimated from transportation survey data) corresponded well to what would be predicted on the basis of an index of access to the homes of consumers (weighted by their total retail expenditures).

4. With respect to residential density, for example, it can be shown that population density will assume a negative exponential form similar to that of land rents when the price elasticity of demand for housing space is unity. See Edwin S. Mills, Urban Economics (Glenview, Ill.: Scott, Foresman, 1972), p. 84.

5. Colin Clark, "Urban Population Densities," Journal of the Royal Statistical Society, Series A, 114 (1951), 490-496.

The percentage rate of decline in density per unit of distance is 100(e-b— 1). This same form of density gradient can alternatively be expressed in terms of logarithms of the densities, as follows; In Dx =ln D0bx. The logarithm of density is thus linearly related to distance, and the gradient can be plotted as a straight line on a chart if a logarithmic (ratio) scale is used for density.

6. Bruce Newling has proposed a more sophisticated formula that does provide for a "central crater" and lends itself to a dynamic model of urban growth in which the zone of peak density moves outward over time. See Bruce F. Newling, "The Spatial Variation of Urban Population Densities," Geographical Review, 59, 2 (April 1969), 242-252.

Net residential density means population per acre of land actually in residential use. It has been shown that in the Chicago area, the fit of the gradient formula is better for net than for gross residential density. See Carol Kramer, "Population Density Patterns," CATS (Chicago Area Transportation Study) Research News, 2 (1958), 3-10; and Chicago Area Transportation Study, Final Reports, vols. 1-2 (1959-1960).

7. See particularly Richard F. Muth, Cities and Housing: The Spatial Pattern of Urban Residential Land Use (Chicago: University of Chicago Press, 1969); William Alonso, Location and Land Use (Cambridge, Mass.: Harvard University Press, 1964); Brian J. L. Berry, J. W. Simmons, and R. J. Tennant, "Urban Population Densities: Structure and Change," Geographical Review, 53, 3 (July 1963), 389-405; and Brian J. L. Berry, "Research Frontiers in Urban Geography," in Philip M. Hauser and Leo F. Schnore (eds.), The Study of Urbanization (New York: Wiley, 1965), pp. 403-430.

8. Muth Cities and Housing, p. 184.

9. Ibid., p. 183.

10. Otis Dudley Duncan. Population Distribution and Community Structure," Gold Spring Harbor Symposia on Quantitative Biology, 22 (1957), 357-371.

11. Recent efforts in developing models of urban spatial structure are deductive in nature and rely on systems of equations characterizing the behavior of various economic sectors. The models may he normative, in that they solve for the spatial allocation of people, goods, housing, and land about the central business district so as to maximize a social welfare function, or positive, in that they solve for a competitive equilibrium. See Edwin S. Mills and James Mackinnon, "Notes on the New Urban Economics," Bell Journal of Economics, 4, 2 (Autumn 1973), 593-601; and J. V. Henderson, Economic Theory and the Cities (New York: Academic Press, 1977).

Closely related to these efforts are attempts to simulate urban economies by using computers to seek numerical solutions to equation systems that characterize housing demand and supply in relation to locations of work places. See Gregory K. Ingram, "Simulation and Econometric Approaches to Modeling Urban Areas," in Peter Mieszkowski and Mahlon Straszheim (eds.), Current Issues in Urban Economics (Baltimore: Johns Hopkins University Press, 1979), pp. 130-164.

12. For a brief account of the Burgess model and models involving subcenters and sectors (discussed later in this chapter), and for references to the original sources, see Chauncy D. Harris and Edward L. Ullman, "The Nature of Cities," in Harold M. Mayer and Clyde F. Kuhn (eds.), Readings in Urban Geography (Chicago: University of Chicago Press, 1959), pp. 282-286.

13. Homer Hoyt, The Structure and Growth of Residential Neighborhoods in American Cities, U.S. Federal Housing Administration (Washington, D.C.: Government Printing Office, 1939). The quotation is from Harris and Ullman, "The Nature of Cities," p. 283. See also, in the same volume, Homer Hoyt, "The Pattern of Movement of Residential Rental Neighborhoods," pp. 499-510. A handy collection of most of his writings over the period 1916-1966 is Homer Hoyt, According to Hoyt (Washington, D.C.: Homer Hoyt Associates, 1968).

14. In Chapter 8 we shall find that this is but part of a more general hierarchical structure of urban activity explained by the central-place model. See also Brian J. L. Berry, "Research Frontiers in Urban Geography," in Hauser and Schnore, (eds.), Study of Urbanization, pp. 407-408. Berry’s article, in bibliographical notes appended on pp. 424-430, cites literature on both interurban and intraurban applications of central-place analysis.

15. See Berry, Simmons, and Tennant, "Urban Population Densities," p. 399, for relevant evidence concerning the residential density gradient for the Chicago urban area for all decennial years, 1860-1950. Clark, "Urban Population Densities," traces the steady flattening of the London density gradient from 1801 to 1941, with central density also showing signs of a decline in more recent decades.

16. Mills, Urban Economics, pp. 100-101.

17. Glenn E. McLaughlin, Growth of American Manufacturing Areas, Monograph No. 7 (Pittsburgh: University of Pittsburgh, Bureau of Business Research, 1938), p. 186. His conclusions were based on U.S. Census data for the 13 largest Census Industrial Areas (composed of whole counties and groups of counties) and their central cities.

18. Industrial Areas (location categories A+B+C in Table 7-1) and also selected other important industrial counties (category F) were identified by the Census of Manufactures in 1929 and replaced by the Standard Metropolitan Areas concept in 1947. The Industrial Area was a unit based on concentration of at least 40,000 manufacturing wage earners in an important industrial city, its county, and adjacent important industrial counties. In 1929 there were 34 Industrial Areas; applying the same criteria in later years, Creamer had 49 by 1963. This is obviously a more exclusive category than the more recent Standard Metropolitan Statistical Area (SMSA), of which there were 323 by 1980. In the period 1929-1963, the number of B cities (see Table 7-1) ranged from 12 to 23; the number of D cities and F counties, from 41 to 61; and the number of F counties, from 47 to 94.

19. Ira S. Lowry, Portrait of a Region, vol. 2 of the Economic Study of the Pittsburgh Region conducted by the Pittsburgh Regional Planning Association (Pittsburgh: University of Pittsburgh Press, 1963), p. 73. The three passages quoted here are from a section prepared by Edgar M. Hoover. Italics in original.

20. Leon Moses and Harold F. Williamson, ‘"The Location of Economic Activity in Cities," American Economics Review, 57, 2 (May 1967), 211-222.

21. For an interesting theoretical analysis of the effect of a suburban export terminal on urban spatial structure, see Michelle J. White, "Firm Suburbanization and Urban Subcenters," Journal of Urban Economics, 3, 3 (July 1976), 323-343.

22. See the discussion at the beginning of Chapter 3 regarding the transfer of commodities and information. For a penetrating study of the processes involved in office activity and their locational significance, see J. B. Goddard, "Office Communications and Office Location: A Review of Current Research," Regional Studies, 5, 4 (December 1971), 263-280.

23. This last consideration may seem far-fetched, but it was repeatedly stressed by responsible corporate officials in personal interviews associated with the Pittsburgh Regional Planning Association’s Economic Study of the Pittsburgh region in the early 1960s.

24. John P. Keith, president of the Regional Plan Association, in the foreword to Regina Belz Armstrong, The Office Industry: Patterns of Growth and Location, a Report of the Regional Plan Association (New York: Regional Plan Association, 1972), p. vii.

25. Research Labs Swarm to Capital," Business Week (23 April 1966), 145.

26. Much more will be said about this turnaround in nonmetropolitan population growth in Chapter 8.

27. See David F. Bradford and Harry H. Kelejian, "An Econometric Model of the Flight to the Suburbs," Journal of Political Economy, 81, 3 (May/June 1973), 566-589.

28. See John C. Weicher, "Urban Housing Policy," in Peter Mieszkowski and Mahlon Straszheim (eds.), Current Issues in Urban Economics (Baltimore: Johns Hopkins University Press, 1979), p. 472.

29. The percentage of homes that are owner-occupied has risen from 45.7 percent in 1940 to 68.7 percent in 1980. See U.S. Bureau of the Census, Statistical Abstract of the United States: 1961, 82nd ed. (Washington, D.C.: U.S. Government Printing Office, 1961), Table 1066, p. 764; and Statistical Abstract of the United States 1982-1983, 103rd ed. (Washington, D.C.: U.S. Government Printing Office, 1982), Table 1352, p. 752.

30. This filter-down process is indeed familiar in the used-car market; but it does not fit as well when applied to housing. Housing deterioration is by no means so closely related to age as is deterioration of automobiles. Instead, condition depends primarily on maintenance, the structure itself being almost indefinitely lasting if adequately maintained. See Ira S. Lowry, "Filtering and Housing Standards: A Conceptual Analysis," Land Economics, 36 (November 1960), 362-370. For a survey of more recent efforts to relate the depreciation and deterioration of dwelling units to residential succession see John M. Quigley, "What Have We Learned about Housing Markets?" in Peter Mieszkowski and Mahlon Straszheim (eds.), Current Issues in Urban Economics (Baltimore: Johns Hopkins University Press, 1979), pp. 417-420.

31. Edgar M. Hoover and Raymond Vernon, Anatomy of a Metropolis (Cambridge, Mass.: Harvard University Press, 1959), Table 49, p. 220. These figures are brought down to 1960 in Regional Plan Association, Spread City (New York: Regional Plan Association, 1962).

32. In Manhattan, for example, the highest density area in 1900 had 400,000 people per square mile, compared with a maximum of 165,000 in any area in 1850; but the maximum was down to 260,030 in 1940 and 221,000 in 1957. Density in the 1957 peak density area had fallen to 171,000 by 1968. The situation was similar in Brooklyn, where a peak of 147,000 was passed in 1930, and in Jersey City, with a 1920 peak of 75,500. In Brooklyn, however, the somewhat newer slum areas of Bedford-Stuyvesant and Brownsville both increased slightly in density between 1960 and 1968. During the same period, densities in Central and East Harlem fell about 20 percent. See Hoover and Vernon, Anatomy of a Metropolis, Table 50, p. 224, and more recent estimates supplied by the Regional Plan Association of New York. The areas involved are wards, assembly districts, and New York City health areas.

33. Retailing location factors have been researched in great detail, and rating systems devised to pinpoint especially desirable sites. For example, it has been determined that filling stations along a commuting artery generally do better business if located on the right-hand side of the road for homecoming commuters—partly, perhaps, because commuters are in less of a hurry on the homebound trip than they are in the morning rush hour.

Back to Contents

Back to Chapter Six

Next Chapter