1. Introduction
1.1 Background
1.2 Theoretical Models of Urban Form
1.3 A Brief History of Changes in Urban Form
2. Literature Review
2.1 Introduction
2.2 Factors affecting metropolitan growth patterns
Dispersion vs Concentration
2.1.1 Dispersion
2.1.2 Concentration
2.2.2 Determinants of Metropolitan Spatial Structure
2.2.2.1 Employment Centres
2.2.2.2 Population and Housing
2.3 Empirical Evidence
2.3.1 Non-Metropolitan vs Metropolitan Trends
2.3.2 Intra-Metropolitan Trends
2.3.3 Employment Centres
2.4 Employment Centres in Six Metropolitan Areas
2.4.1 Job Dispersion: Philadelphia and Portland
2.4.2 Large CBDs: New York and Boston
2.4.3 Policentric: San Francisco and Los Angeles
3. Research Methodology
3.1 Research Approach
3.2 Aims and Objectives
3.3 Research Strategy
3.4 Data Collection
5. Empirical Evidence
5.1 Introduction
5.2 Population and Housing
5.3. Employment Centre
5.4 Current Spatial Structure
5.5 Emerging Spatial Structure
5.5 Other factors involved in Spatial Change
6 . Data Analysis
6.1 Data illustration
6.2 Comparison of data and trends
6.3 Existing and Emerging Spatial Structure
6.4 Potential Spatial Development
6.5 Reflex and Summary
7. Conclusion
Urban Spatial Structure Changes/Trends in Employment and Housing in U.S. Metropolitan Areas
Literature Review
This part includes a critical review of the literature on recent spatial structure trends in U.S. metropolitan areas. Suburbanisation occurs throughout the history in U.S. (Bruegmann, 2005). From postwar to 1980s, employment and population growth showed tendency of decentralisation. Central business districts (CBDs) declined while new employment centres/concentrations emerged outside of CBDs. The key point is to find out if these trends have continued. Economic restructuring, public policies, rising congestion, new developments and many other factors may have impact on spatial patterns. These broad trends have continued, however, these trends vary between and within metropolitan areas in the U.S. Three typical patterns of spatial development were discovered. Another emphasis is the intra-metropolitan structure and how it may affect commute and energy consumptions. Research suggests that urban spatial change varies: some metro areas are more polycentric than others; some employments share are retained in CBDs while others are not. Case study of six metropolitan areas is involved to illustrate the changing spatial structure.
Introduction
In recent years, metropolitan areas are featured with decentralised employment and population, declined CBD, the arise of employment centres outside the CBD and suburbanisation. There are literatures on the evolution metropolitan areas (e.g. Berwald, 1982, Jackson, 1985; Chinitz, 1991; Castells; Hall, 1994; Muller, 2004). The causes of changing urban forms include economic restructuring, rising income, technological change, dominance of automobile, public policy (e.g. transportation, housing) and social segmentation.
The interest of urban spatial evolution is focused on the past 20 years. During this span of time, technological advances in information and communication technologies leading to the structural changes in economy are comprehensive. Spatial trends based to the shift to an information economy is a popular argument. Some suggest that as the incomes rise, the demand for amenities and space are relatively elastic. The former will work towards concentration and the latter will work towards dispersion. There are theories that bring up the idea that taxes are significant determinant of urban structure. The implications may also include the transportation costs.
The wide concerns about public policy on reverse or halt these trends are generated by decentralisation and suburbanisation. These concerns include those related to urban sprawl: increased dependence on private vehicles, environmental degradation, inefficient land uses, decline of urban downtown, spatial segmentation, etc. There are numerous public policy efforts put into reversing these trends. Redevelopment programmes in metropolitan centres are actively performed throughout the US. Public transit system have been expanded and working towards higher efficiency. To limit suburban development, some metropolitan areas have used land use controls. The idea of ‘Smart growth’ and ‘transit-oriented development’ have been accepted to solve environment and transportation problems.
It is safe to ask whether longstanding decentralisation trends continue or if new spatial forms are emerging.
The spatial patterns within metropolitan areas may have the greatest impact on commute and energy consumption.
The literature review begin with an overview of factors that affect metropolitan development ninth 1980. These contain economy restructuring, demographics change, policy change and the rise of information economy. These major trends are considered based on theoretical expectations in the discussion of urban economy theories. The empirical literature – employment and population trends at both metropolitan and national level will be discussed in the next section. There will be a case study of six metropolitan areas classified to three typical patterns based on their own empirical work. Finally, the review concludes the implications in regard to future development trends.
2. Factors affecting metropolitan growth patterns
There are two literatures that are directly relevant to this research.
2.1 Dispersion vs Concentration
The urban structure is determined by the relative strength of economies. Cities exist as a result of of being a efficient organisation for economic activities. Traditionally, the influence factors of employment and household location choice are mainly related to urban economics. The choice will then develop into urban structure. Transportation cost is the traditional element that determine the urban form. In recent years, there are much more factors being involved.
Dispersion
The costs of information processing and transmission have reduced to the extent that virtual flows can substitute some physical flows. Reduced communication costs allow firms to have flexibility in location choice. It is possible to locate one centre in downtown and more offices in less costly suburban areas. It also encourage outsourcing and network based service. As the benefits of agglomeration reduce, the costs of agglomeration become obstacles to concentration.
Some argue that cities will be eliminated altogether due to ICT (e.g. Castells, 1989, Cairncross, 1997; Mitchell, 1996). Once the world become highly accessible, there is no value of concentration. As a result, cities will disappear. From a less extreme point of view. dispersion and decentralisation should be expected.
Moreover, dispersion may be prompted by people’s preference for low density living environments. The growth in work mobility and allow workers to have more choice in their household location. Since one of the key factor in company location choice is workforce availability, their residential preferences may lead to employment decentralised. Finally, the quality of life factors involved in preference for suburban locations could contribute to further job decentralisation. The dispersion imply 1) densities in metropolitan areas decline, 2) possible decline in larger metropolitan areas, 3) shift of urban growth to non-metropolitan areas and smaller metropolitan areas.
Concentration
As cities are formed in historic development, they self-reinforcing and self-elimination along the way. Most diverse workforce, most trained experts and large numbers of workers are usually located in large cities and it creates a competitive advantage. Large cities also possess dense transport networks and good access to global transport networks. Since there are greater demand for communication services in large cities, suppliers are able to offer cheaper and better service thanks to scale economies.
There are advantages in labour market pooling. Increased amount of temporary jobs and decreased job stability mean that workers need to seek new opportunities constantly to stay competitive. The employment risk are high and it will be nice to have areas that are highly accessible in jobs. Meanwhile, companies can benefit from a diversified labor supply.
Some suggest that major cities are attractive for being educational and cultural centres as well as destinations for consumption activities. These factors may contribute to a dynamic and vibrant atmosphere that is appealing to highly-educated workers. Such workers may prefer to excitement of urban life over other elements.
There are arguments regarding demographics change will result in more concentration. As baby boomers age, it is likely they will relocate to areas that are more walkable with public transportation access. Another factor is that they no longer need schools in suburban districts for their children, thus motivate them move back to downtown for easier access to amenities.
The concentration imply 1) metropolitan areas continue to grow, 2) metropolitan densities increase, 3) there will be more activities cluster at sub-metropolitan level.
2.2 Determinants of urban spatial structure
This section discuss the theories of urban spatial structure as in the distribution of population and employment. A great number of literature have explained the evolution in metropolitan urban structure in economic terms (Fujita 1989). The existence of employment centre, for example central business district (CBD), is based on scale production and negative economics of transportation and congestion.
The standard urban model assumes there is a single employment centre, and household distribution is based on tradeoff between commute and housing costs (Fujita, 1989). This model predicts that population density will decrease constantly with distance from city centre. Since the housing costs decline as transport costs increase, population density declines with distance and therefore, more housing is consumed. The commuting patterns have also been predicted: the average commute distance equals to the mean distance of gross population to urban centre. Those with stronger preferences for housing will locate further away from the downtown if housing demand varies across households. If these preferences are positively related to income, higher income household would consume more housing units and locate further away while lower income household would locate closer to centre.
The four basic elements that determine the size of city in standard urban model are population, income, agricultural land value and transport costs. An increase in population will result in larger city size but density gradient will not be affected. An increase in income will lead to higher housing demand per household. Therefore the city size will increase and density will decrease. If agricultural land price increases, city size will shrink and density gradient will increase. If transport costs decline, the consumption of housing increases and the city expands thus lead to the decline of density gradient.
In the past few decades, the average income has increased and actual transport costs have declined. Empirical evidence tends to support the theory that population density decline with the distance from city centre. Meanwhile, population density gradient has declined over time (Anas, Arnott and Small, 1998). The spatial extent in metropolitan areas expands as studies on urban sprawl indicates (Glaeser and Kahn, 2003). Higher income households tend to live further away from city centres while lower income households prefer to live closer to city centres and have shorter commutes.
2.2.1 Employment centre
The main critique of standard model is that metropolitan areas are no longer monocentric. Some people believe that metropolitan areas are polycentric while ohters argue that they could be classified as dispersed without any significant employment concentration. If agglomeration economies exist at sub-metropolitan lever, clusters of employment should be observed and can eventually be called employment centres.
What’s the formative factors of multiple employment centres? Traffic congestion, rents,
Over time, to some companies, an existing centre may develop to an extent where negative externalities of locating inside it outweigh the benefits. These forces vary largely by function and industry. These
As globalisation progresses, metropolitans like New York or London may continue to concentrate and more headquarters may seek opportunities to locate close to other headquarters. Meanwhile, concentration in second-tier metropolitan areas may begin to decline as both offices and manufactures relocate.
2.2.2 Population and Housing
As mentioned above,
The standard model predict the lower transport cost and higher income per capita which Mieszkowski and Mills (1993) described as ‘natural evolution’.
The progression of population suburbanisation is based on limited land supply in central cities, abundant supply of cheaper land on the periphery and increasing demand for new housing.
The competitive of suburbanisation is driven by public policy: compare to suburbs, central cities usually have higher taxes, lower quality government services such as public schools, congestion, crime, racial tensions, poor environmental quality (Mieszkowski and Mills, P137). These fiscal social issues in central cities lead to the suburbanisation of high income households. Therefore, suburbs are wealthy while inner cities tend to accommodate more low income population. The fiscal social theory regard suburbanisation as a result of public policy rather than market force. Although suburbanisation can not be attributed to any single policy.
3. Empirical evidence
In this section, the discussion will follow the literature around these questions: 1) what are the trends in share of non-metropolitan and metropolitan population and employment; 2) what are the spatial trends in metropolitan population and employment?
3.1 Non-Metropolitan vs Metropolitan
The population and employment trends will be discussed.
3.1.1 Population
A comprehensive vision is provided: Population continues to urbanise and suburbanise: an increasing share of population reside in metropolitan areas and within these areas, an increasing share reside in suburbs. The definition of suburbs and central cities are based on political boundaries and are quite limited as indicators of spatial trends.
3.1.2 Employment
The table show that between 1969 and 2004, the resident population and employment population have both increased. Increase of Income per capita
On a national scale, these figures could not indicate the disperse of population or employment.
3.2 Intra-Metropolitan Trends
In this section, the population and employment trends within the metropolitan areas are summarised. The following attributes are used to describe the spatial organisation. Density measures the intensity of land use in terms of population, jobs or housing units per land unit. Centralisation indicates the degree of which jobs or population are concentrated around the metropolitan area centre. Concentration measures the degree of which activities are located in small proportion of the metropolitan area. For instance, a metropolitan area may have a high level of job concentration, but can also be highly decentralised. Proximity measures the relative distributions jobs, populations and other activities. Dispersion refers to the degree of spatial organisation. A dispersed job distribution could be one with jobs randomly distributed instead of concentrated or centralised.
3.2.1 Population Distribution
Even though faced with many criticism, population density if the most commonly used empirical measure in urban spatial structure for comparing spatial trends across metropolitan areas or over time. Suburbanisation can emerge with or without leading to
rising density in the suburbs. The change in the economic activities organisation can not be represented by aggregate statistics on a rough suburban/centre classification.
Population density functions have been utilised utilised for cities around the world since the 19th century. One consistent conclusion is that over time, average density tends to get lower and gradients tend to become flatter. Meanwhile transport costs are reduced and per capita income rise. (McDonald, 1989). Kim (2007) used datas of cities with over 25,000 population from 1890 – 2000 to analyse the changes of population over time. The cities’ average population density were increasing up until 1940 before decreasing after. While the common trend is rising and falling in all cities groups, the average is the lowest and remains the lowest for the newer cities, and the average is the highest and remains the highest for the older cities (prior to 1800). The decline in density after 1940 is most obvious in oldest cities which means the average density has decreased. From 1980 to 1990, the trend has attenuated notably.
The estimated density gradient and share of central city population have declined consistently over time. Kim’s research indicates that long time trend of declining population density is attenuating although central city’s share of population continues to decrease.
3.2.2 Case Studies of Population and Employment trends
Case studies of specific metropolitan areas is another method to examine spatial trends. Comparative case studies are limited to widely available data such as US Census data or BEA data. As BEA data is only available at county level, Census is the only data at sub-county level.
CTPP data are based on long-term Census data and provide information about commuting, employment and workers.
Lee, Seo and Webster (2006) examine the commuting pattern, employment trends and specialisation of 12 CMSAs in US to look into the historical changes in metropolitan spatial structure during 1980 to 1990. These metropolitan areas New York, Buffalo, Chicago, Philadelphia, Detroit, Cleveland, Huston, Denver, Los Angeles, San Francisco, Portland and Seattle. According to US Census PUMS data, each CMSA is divided into central city (CC) and not central city (NCC).
In general, the results shows that the differences between fast growth and slow growth cities (expect Houston) together with the different trajectories of central cities.
Horner (2007) uses CTPP data
The result suggest that the larger dispersion of jobs and workers did not lead to increases as large as distribution change might have allowed.
Yang (2008) did a similar case study on Atlanta and Boston using CTPP data for 1980, 1990 and 2000. These two cities are chosen for they are similar in populations and different in urban structures. Together with Tallahassee study, it seems like the workers change their job location to cope with changing spatial patterns.
Finally, Lee (2007) studies the overall pattern of employment and population of six CMSA areas: Boston, New York, Los Angeles and Portland from 1990 to 2000, and Philadelphia and San Francisco from 1980 to 2000.
Table shows that employment decentralisation exist in all six area and they are from different starting points. Portland, with the smallest metropolitan area among the group, is the most centralised followed by New York and Los Angeles. San Francisco Bay may affect the measurement of San Francisco. In general, population changes tend to be smaller with almost no change in Portland or New York. All these metropolitan area have experienced varying degrees of decentralisation and the changes in employment are all greater than in population. These data indicate that within the broad trends of deconcentration and decentralisation, metropolitan areas differ in rate and degree of changes.
3.3 Employment Centres
Whether employment is preponderantly dispersed, or whether concentration of employments or employment centres are significant features of metropolitan areas is a key problem in urban spatial structure.
From economics perspective, an employment centre is a cluster of activity with sufficient magnitude to affect land prices and spatial form. In the case of single centre area, the centre is the zone with highest density of highest land value per unit. In polycentric area, any cluster that can affect land values independently constitutes a centre. In metropolitan areas, there are many clusters of employments with varying topographies.
Garreau (1991) defines ‘edge cities’ as the emerging new centres that are far from CBDs. To be qualified as an edge city, it need to meet five conditions. 1) at least 5 million sqf of rental office space; 2) at least 600,000 sqf of rental retail space 3) more jobs than housing units 4) perceived as a distinct single place 5) was nowhere near to a city 30 years ago.
Tysons Corner, Virginia is described as a prototype of edge city by Garraeu.
Land and Lefurgy (2003) introduce the concept of edgeless city, with isolated buildings spreading across an immense area without a discernible boundary. ‘Edgeless city’ is used to describe a structure at sub-regional level rather than city level. They estimated that there are nearly double the rental office space in edgeless cities than edge city. The emerging spatial structure is interspersed between population and employment without forming any discernible centre.
Employment centres are demonstrated across metropolitans in varying location, size, growth rates and age. The conclusion is that employment centre is an significant aspect in urban spatial structural.
The data in the two tables show that San Francisco and Los Angeles have smaller share of CBD employment compare to Boston, Philadelphia and Boston. Results in Portland are mixed.
The result in these tables also suggest that Philadelphia, Portland, Boston and New York are more monocentric: a significant proportion of all centre employment is located in its major centre. San Francisco and Los Angeles are more polycentric as the CBD owns a smaller share of all centre employment. The difference is depend on how employment centres are defined. Monocentricity is related to with deconcentration. In these more monocentric metropolitan areas, the share of jobs in centres is smaller. On the other side, it seems like employment concentration is associated with both age and size: Philadelphia, New York and Boston may be more monocentric as the core areas were built before 20th century. Portland is more monocentric because its relatively small. The share of all jobs in centres varies between two definitions of centres and across metropolitan area.
Comparing 1990 and 2000, Lee’s analysis is consistent with deconcentration trend: employment growth outside of centre is faster, therefore the share of employment located outside of centres is increasing. The analysis also demonstrates a general trend of fewer centres, except Portland. The total number of centres is expected to decline as more employment will move to non-centre locations from centres.
4. Employment Centres in Six Metropolitan Areas
The majority of metropolitan jobs are located outside of the employment centres in six metropolitan areas.
In recent years, what type of location has gained jobs? There are three important findings from trend analysis.
First, in the 1980s and 1990s, employment continued to dencentralise in metropolitan centre and suburbs.
Second, employment dispersion was more common than subcentering. Employment centres didn’t perform on par with dispersed employment locations, except the case of San Francisco in 1990s. Dynamic employment subcentering was not common as it only occured in San Francisco and Los Angeles. New employment clusters in suburbs almost offset the job loss from older centres in these two areas. In other metropolitan areas, employment growth in subcentres neither compensated for the job share losses in the centre nor kept pace the metropolitan average.
Lastly, the trend analysis of six metropolitan forms show that spatial structures are not evolving in a single direction. The spatial transformation can be classified down to three patterns. Each type of pattern includes a pair of cases.
Los Angeles
Los Angeles is one of the most widely studied areas in metropolitan spatial trends. Some researchers decided Los Angeles region is polycentric (Forestall and Greene 1997, McMillen and Smith 2003, Redfearn 2007) while others claimed that the area is currently ‘beyond polycentric’. The latter suggest that the employment patterns are decentralising and becoming less centralised.
Three Patterns of Spatial Evolution
4.1 Job Dispersion: Philadelphia and Portland
The first type is predominant job dispersion without substantial suburban clustering. Philadelphia and Portland fall into this description where both deconcentration and decentralisation appeared to the greatest extent. In urban core, employment share shrank quickly and subcentres were not strong enough to attract the decentralising jobs. Subsequently, centre employment shares declined.
In Philadelphia, the CBD and its surrounding areas experience significant job losses, as a result, the employment density drop in central location. The main centre that passes a certain density shrank considerably, with its employment share drop to 15.9% in 2000 from 26.2 in 1980. In the 1990s, subcentres also underwent job losses although their employment shares were stable in the 1980s.
4.2 Large CBDs: New York and Boston
A contrary to the first pattern is the case in New York and Boston. Urban cores were much stronger than suburban centres and employment agglomerations remained active throughout the 1990s. The minor job loss in CBDs was mostly compensate by the growth in neighbouring area. Therefore, the share of main centre remained stable. The spatial transformation process in these two metropolitan areas can be summed up as moderate deconcentration and little decentralisation. The overall loss of centre employment share
was less than that in Philadelphia and Portland.
4.3 Policentric: San Francisco and Los Angeles
The last one is quite different from the two previous development patterns. San Francisco and Los Angeles fall into the polycentric classification. Los Angeles experienced absolute shrank of employment agglomeration in regional core while San Francisco experienced it to a lesser extent.
3. Methodology
3.1 Research Approach
The analysis of this dissertation relies mainly on descriptive spatial measures and published statistics.
Having already established the basis of the outlook and history of this dissertation, which is pertinent to which evidence collected will be looked at, now it is necessary to reflect upon how evidence will be collected to support the arguments in the dissertation. There are multiple approaches to social research, the use of historical data analysis via US census is relevant to the dissertation.
3.2 Aims and Objectives
Research attempts to discover the current direction and stage of spatial evolution, the driving forces of spatial changes and the links between urban spatial structure change and development.
What are the prominent features of emerging urban forms?
Are cities becoming more edgy or edgeless?
What are the primary causes of spatial changes?
Are the relationships between metropolitan growth and spatial structure depend on varying metropolitan population size?
What are the consequences of current spatial changes?
3.3 Research Strategy
There are a number of methods to create a well researched picture of spatial structure changes. First, it would be necessary to establish the history of spatial structure in US metropolitan areas and how they operated. This can be done through secondary sources such as “EDGE†OR “EDGELESS†CITIES? URBAN SPATIAL STRUCTURE IN U.S. METROPOLITAN AREAS, 1980 TO 2000∗ (Lee, 2007) and Metropolitan Spatial Trends in Employment and Housing Literature Review (Giuliano, Agarwal, Redfearn).
Next, it is important to analysis the new factors that have influence on spatial structure or traditional factors that affect spatial structure in a way. The method consist doing comparative case studies and examining recent published documents.
Finally, the data needed could be found from both primary and secondary source.
3.4 Data Collection
A good source of primary information can be found in US Census and CTPP, displaying the more up to date data compared to literature review.
To explore and reexamine in some depth the distribution of population and population growth in US metropolitan areas, data from the 1990, 2000 and 2010 Censuses for 35 largest metropolitan areas in the US. The measures include density gradients, density frequency distributions, concentration indices and spatial distributions of growths.


Data needed:
The measures include measures of population distributions in 2010 and changes in those distributions from 1990 to 2010.
Employment Trends, Inside and Outside of Central City
Share of Jobs in Centres, CBD, other Centres in Metropolitan Area
Commute Estimates (km)
Table: Population at Metropolitan, Urbanised, and Principal City Scale in 1990 and 2010, sorted in rank order
Table: Change in Population at Metropolitan, Urbanised, and Principal City Scale in 1990 and 2010, sorted in rank order
Table: Population Density at Metropolitan, Urbanised, and Principal City Scale (1990-2010), sorted in rank order
Table: Change in Population Density at Metropolitan, Urbanised, and Principal City Scale (1990-2010), sorted in rank order
Measurements of population and density for the six metropolitan areas in the US from 1990-2010 Census are presented in tables 1 through 4. Table 1 and 2 display population and density data for 1990 and 2010. Table 4 shows changes in population for metropolitan area, urbanised area and principal city from 1990 to 2010. In each geographic area, data are sorted in descending order by population. In general, metropolitan areas with large populations also have large populations in their respective urbanised areas and principal cities.
Population density and population are highly interconnected at every geographical level. The larger metropolitan areas, urbanised areas and principal cities tend to be more denser.
Growth Distribution and Infill
To have further perceptivity of the growth distribution within metropolitan areas, the paper examined the population growth from 1990 to 2010 in four criteria: 1) the geographical areas that have never been urbanised, 2) the area that urbanised between 2000 and 2010, 3) the areas that urbanised between 1990 and 2000, 4) the areas that were urbanised in 1990. The respective growth occurred in these areas can be seen as measures of urban sprawl and infill.
Table: Change in Population
Infill Development can be define as the metropolitan areas that are urbanised by 1990 and had more growth within the area. Table ? demonstrates the growth distribution in urbanised ares of each metropolitan area. As shown, Portland had the largest margin of infill urban development.
Marginal density equals to the percent change in urbanised land areas divided by percent change in urban population. Portland, Los Angeles and San Francisco have lowest marginal densities which means these metropolitan areas had accommodate the most urban population from 1990 to 2010 with the smallest urbanised area expansions. Some of these measures can reflect the overall trend of population growth. General growth degree in urbanised and non-urbanised area tend to be on par with the metropolitan areas. However, the proportions varied as shown in marginal densities.
Concentration
Measurement of the extent of populations are concentrated within subareas is called concentration. Gini coefficients and Lorenz curves. A high Gini coefficient suggests an uneven distribution which means large numbers of population are concentrated in small areas. It is quite necessary for polycentric urban areas with high density mixed use nodes to have some extents of unevenness in spatial distribution to be fully functioned. An uneven distribution doesn’t necessarily equal to polycentricity, but an even distribution can possibly preclude it.
This paper use the Gini coefficients for the whole metropolitan area and urbanised area. Table ?? demonstrates the Gini coefficients level in 2010 and changes in Gini coefficients level from 1990 to 2010. As the table shows, Portland is highly concentrated at metropolitan level in 2010, with the Gini coefficients larger than 0.9. Part of the reason is that there are large share of rural areas within the metropolitan area. All of the six metropolitan areas have become less concentrated during 1990 and 2010. Philadelphia deconcentrated the most while Portland deconcentrated the least.
Gini coefficients data are also presented in Table ??? at urbanised area level. In 2011, Philadelphia, San Francisco, Boston and New York are most concentrated with San Francisco concentrated the most from 1990 to 2010.
Figure 1
Density Gradients
Table: Change in Density Gradient – Slope, Intercept, Category (1990-2010) sorted in rank order
Legend: A = decentralisation; B = expansion; C = centralisation
Population density gradients measure the extent of population density decline with distance from city centre.
The estimated density gradients’ slopes and intercepts with data from 2010 block group are displayed in Table ??? for the metropolitan areas. It is obvious that New York has the highest estimated intercept and Portland has the steepest estimated density gradient. Los Angeles and San Francisco have relatively flat density gradient.
As shown in Table, three metropolitan areas have expanded since 1990 which means estimated densities are higher from every distance from city centre. In Philadelphia and New York, the density gradients and estimated density in urban core have both dropped. New York is the only metropolitan area with steeping density gradient and rising density in city centre within the study area.
Figure 1: Changes in Employment Shares by Density
Figure 1 displays a summary of employment dispersion trends. All census within the population area are divided into five quintiles by employment density. The densest quintile is split into two deciles. Every density’s groups share of total employment in each year are shown in the bar charts.
As seen, the employment deconcentration is emerging in all six areas. In low density tracts, job shares are increasing while in high density zones, job shares have declined. The largest gain in employment share were in the two lowest density zones. Is is in accordance with the conclusion of Carlino and Chartterjee (2002) that from the 1950s, employments have shifted toward areas with lower density in both intra and inter-metropolitan contexts.
However, there is a significant difference among areas in term of speed and extent of deconcentration. The employment concentration in San Francisco and Los Angeles is notably higher than other metropolitan areas as the majority of jobs are located in the densest quintile. During the studied period, these two polycentric areas also underwent less dispersion.
Growth Patterns of Metropolitan Employment Centers
The definition of CBD and main centre vary in different areas. CBD in New York is strictly defined as the area of midtown south of Central Park and lower Manhattan. It took up 1.2 million jobs in 2000, which is approximately 14% of the job share in the whole metropolitan area. The main centre is an 8-mile-long area that accommodates around 2 millions employments.
Los Angeles’s case is more extreme in comparison. Its CBD only account for 3% of the metropolitan employment. Nevertheless, it has a huge main centre by minimum density method, which is a horizonal corridor for more than 20 miles and accounts for almost a million jobs.
Figure???: Employment Centres in Metropolitan Area, 1990 to 2010. (Find FIGURE)
Despite the presence of large agglomerations, in all six metropolitan areas, the majority of metropolitan jobs are in dispersion outside employment centres.
Lang (2003) suggested that in 13 largest metropolitan areas, edgeless cities account for double office space of edge cities. It is worth mentioning that office sector tends to locate in employment centres rather than areas with priority of other economic activities.
Outside any type of centres, dispersed employment grew to making up between 66 and 88 percent of the metropolitan employment by 2000. With the exception of Los Angeles and San Francisco, under 10 percent of jobs outside the CBDs are clustered in subcentres in other four metropolitan areas. Only these two metropolitan areas are genuinely polycentric as there are more jobs concentrated in subcentres than main centres.
There are a few significant discoveries from the trend analysis. Above all, in all six metropolitan areas, the employment share shrank in core areas, whether defined as main centres or CBDs. Particularly, Los Angeles, New York and Philadelphia went through absolute job losses in their CBDs. The employment shares in CBDs had dropped 2.4% in Los Angeles and 8.7% in Philadelphia in 2010.
Second, it appeared that job dispersion was more common than subcentreing. In almost all cases, jobs had better performance in dispersed location rather than employement centres,
Table: Centre Employment Trends in ???
Three Patterns of Spatial Trends
Portland and Philadelphia
In Portland metropolitan area, more than 240,000 jobs were added since 1990, which is a 33% rise from 1990 to 2010. While most districts within the region experienced the substantial employment growth, most of the growth share were contributed by lower density zones. The dispersed location account for over 90% of the employment growth.
The employment share decreased from 11.3% to 8.2% in CBD. To sum up, Portland metropolitan area become denser and flatter since 1990.
New York and Boston
During the last decade, the spatial structure of Boston remains centralised.
The employment growth rates in CBDs, 4.9% and 8% in 1990 and 2010, are similar or higher than metropolitan average. The employment share in main centre by broad definition, also remained at about 22%. In contrast, there were little job concentrations in suburban Boston in 1990 and they continued to shrink in 2000.
Manhattan is the biggest employment agglomeration in the nation with around two million jobs and kept its predominant position though the 1990s. Although lower Manhattan underwent some extent of job loss in its central area, the lower density parts of the island made up for it. Therefore, the employment share was stable at around 21 percent in the main centre. Compared to Boston, employment centres in suburban areas had a better performance in New York metropolitan area, especially in New Jersey and Long Island. Consequently, the centre employment share only experience a minor loss in New York.
Los Angeles and San Francisco
In these two metropolitan areas, a considerable share of decentralising employment showed signs of reconcentration in suburban centres. Between 1990 and 2000, the share of clustered jobs remained stable.
In Los Angeles metropolitan area, the growth of new clusters in the outer-ring suburbss was only record by nonparametric method.
Seven subcentres disappeared and five merged with others while ten new subcentres emerged. The majority of clustering took place around the border areas of Orange, Riverside and Los Angeles counties, presumably due to the industrial restructuring in the area. This resulted in an approximately three-percentage growth in the employment share of subcentres. However, the densities of these emerging clusters in the outskirts were too low to pass the minimum density test. Therefore, the minimum density method identified a decrease of three percent in employment share from 1990 since only 34 subcentres in 2000 are counted.
However, it seems that minimum density method identifies more subcentres in 2000 while nonparametric method has capture less clustering in the 1990s in San Francisco. Due to the mismatch of data range in different census years, a dramatic drop appeared in subcentres’ employment share between 1990 and 2000.
It can be explained by the fact that data conversion was not applied when using 2000 census tracts boundaries and 2000 data when a huge subcentre is found in Silicon Valley. This subcentre integrates high technology clusters from Mountain View to its east, Milpitas, and accommodating over two hundred and eighty thousand jobs. By GWR method, the total centre employment share is 30.5 percent using 2000 census tracts with combined shares of 6.3 percent in CBDs and 24.2 percent in subcentres. A similar expansion in Silicon Valley and overall rise in clustered employment were also captured by the minimum density method. The two methods deliver the result of a four percentage increase in the employment share of subcentres in San Francisco metropolis between 1990 to 2010.
Number of jobs near the average resident (within 8.8 miles) of the six metropolitan area.
Table: Share of Employment in High-Density ZIP Codes Outside the Urban Core, Selected Metro Areas, 2010
Table: Change in the Geographic Distribution of Jobs, 2007-2010
Table: Employment Located in High-Density ZIP codes, 2010
Table: Employment Clusters, Metropolitan Divisions, and Edge Cities in 6 Metropolitan Statistical Areas, 2010
Table: Static Measures of Metropolitan Area, Colour Coded by Quintile?
Table: Static Measures of Metropolitan Area, Colour Coded by Quintile?