Factors Causing Urban Sprawl

There are at least two methods to count change of population and employment altogether, as opposed to only examining the directions of change, i.e., for the two “Same level of compactness/sprawl with change in land use,” in the above table. One type of methods is to count the floor space for residential and commercial in use. The most ideal data to calculate if space is more used or less to judge if central city is used us more or less.

However, the data are mostly unavailable. The other type of method is to count the land area for residential and commercial in use. The variable is not ideal since the real floor space used up cannot be correctly counted. The data may not be counted correctly also when land use is mixed in the buildings. In addition, mostly the data are unavailable.

The way of thinking about dynamics of growth, redistribution, and impact of sub-area characteristics in terms of population dimension alone. The growth of population

constitutes what and the amount has to be distributed in space in a metropolitan area.

Housing stock formulates the level of attraction in terms of quantity on the one hand.

Unmet housing demand in this sub-area may cause the force of spill-over, or in the central city case, urban sprawl. So the above involves three variables representing different roles:

population growth represents the amount to be distributed; housing stock represent the force of pulling, and; unmet housing demand cause the force of pushing. For example, for central city, the overall population growth of metropolitan, and then blown up by shrinking

household size, formulates the amount needed to be distributed, which cause change of urban form, but not necessarily compactness or “sprawl;” (t-t1—t0).

3. Factors Causing Urban Sprawl

Factors causing sprawl may involve two phenomena: change of population (growth, mostly, and decrease), and (re)distribution of activities (both residential and employment). Growth definitely leads to distribution of new population and employment, affected by pushing, pull, and accessibility characteristics of places in a metropolitan area, and accessibility ability of population and employment. Redistribution of existing population and employment is also affected by characteristics of places in a metropolitan area. In a fast growing metropolitan area, the impact of distribution of growing residential and employment spatial needs on urban form could be larger than that of redistribution of existing residential and employment needs.

Growth will definitely affect degree of compactness/sprawl since the urban form will change due to more space used by human. However, growth will not necessarily lead to compact or sprawling form; in contrast, new urban form depends on policies digesting their new spatial demand. Hence growth may reinforce the impact of exiting compactness or

“sprawl” tendency possibly shaped by such policies housing policies, transportation policies;

for example, newly resulted unmet (increasing) housing demand in central city may reinforce sprawling trends.

[鍵入文字]

Table 2 Four Population-Household-Employment-Based Archetypal Central Cities: Taiwan

No △△△△POP

Type of central city, by Pop., HH, and Employment

Policy Implications or

Further Analysis Needed Jobs/Housing Balance Overall Overall Overall Overall △△△△POP Overall Compactness/SprawlOverall Compactness/SprawlOverall Compactness/SprawlOverall Compactness/Sprawl Policy ImplicationsPolicy ImplicationsPolicy ImplicationsPolicy Implications

1 + + +
Growing central city (CC);

Toward compact CC.

To serve spatial needs from centralizing population and employment to lead to more compact urban form, and maintain residential livability.

Residential activities replacing employment activities.

What could have been missed:

Wrongly diagnosed as more compact CC since being observed from population alone.

(Towards jobs/housing balance? If toward balanced, then the cc is heading toward good urban form; if not, it’s not.)

Toward balance

Away from balance-More economically robust CC

N/A N/A
Acceptable

Planning-oriented policy intervention to induce

increasing housing demand but with diminishing population.

Employment activities replacing residential activities, yet with more residential demand.

What could have been missed:

Wrongly-Diagnosed as less compact CC since being observed from population alone, and yet with more residential demand.

Increasing housing supply can serve housing needs, and hence leads to more compact CC.

Toward balance

Away from balance-More economically robust CC

N/A N/A
Acceptable

Market-oriented policy meet CC-bound migration of population.

6 -- + --
Weakening CC, but with

increasing housing demand.

What could have been missed:

Decentralizing CC, yet the residential need in CC was not identified if not observing from HH.

Increasing housing supply can serve housing needs, and hence leads to compact CC.

N/A N/A N/A
Market-oriented policy compact CC since being observed from population alone and yet with more employment demand.,

Toward balance

Away from balance-More economically robust CC

N/A N/A
Acceptable

Planning-oriented policy intervention to induce

(Diminishing metro or

decentralizing urban form? Needs to analyze the overall growth rate of the metro and

compactness/sprawl index.)

N/A
+ Decentralizing
+ Decentralizing CC but still of compactness in general

-- Sprawling

Planning-oriented policy intervention to induce CC-bound migration of population and employment?

-- Diminishing

Growth -- the growth of population constitutes what and the amount have to be distributed in space in a metropolitan area. Housing stock formulates the level of attraction in terms of

quantity on the one hand. Unmet housing demand in this sub-area may cause the force of spill-over (push out forces), or in the central city case, urban sprawl.

So the above involves three variables representing different roles: population growth represents the amount to be distributed; housing stock represent the force of pulling, and; unmet housing demand cause the force of pushing. For example, for central city, (1) the overall population growth of metropolitan, and then blown up by shrinking household size, formulates the amount needed to be distributed, which cause change of urban form, but not necessarily compactness or “sprawl;” (t-t1—t0).

The outward pushing forces between time 1 and time 2 (i.e., expected POP t2 – Real POP t2)(the pink part), lead to outward population movement/change at the outskirt at the ending year of the period (t2), or the change of degree of urban “sprawl” between t1 and t2 (e.g., pop-based sprawl t2- pop-based sprawl t1).

Unmet demand for space for population and employment cause push out effect. Unmet demand is a collective result of housing demand and supply, in terms of both quantity and quality.

Characteristics in terms of both quality and quantity. For instance, the simplest one, highway system differs from general road systems in terms of quality, and also the length of roads per se differs in terms of quantity. Another example of quantity difference alone is the unmet housing demand. If not, then the good thing is that both compactness and sprawl can be gauged. In this case, a sub-area in a metropolitan area can have both push-outward and pull-inward factors and the overall condition. For central city, push effect means “sprawl” forces, and pull effect means“compactness”effect. For suburban areas, push effect may“compactness”

force or further“sprawl”effect.

If a direction is defined as forces causing compactness and sprawl, then forces can be defined, for explaining sprawl, as pushing outward forces of central city, pulling-outward forces of suburban areas, on the one hand. And on the other hand, pulling-inward forces of central cities, and pushing outward of suburban areas. The analysis unit seems to be the metropolitan areas since the sprawl and compactness is measured at the metropolitan area level, sprawling forces can also be measured at the metropolitan level, such as pushing-outward forces and pulling-outward forces, and the compactness effects

4. Methods

7

This research employs a time-series-based framework to analyze the dynamics of urban

“sprawl” factors and their impacts on “sprawl.” The time series framework incorporates time points, years 1966, 1980, 1990, and 2000, for 36 metropolitan areas in Taiwan. Due to both time-series and cross-sectional data format, panel data analysis is employed, which is conducted in Stata 10.

Population and employment should be both taken into account to gauge compactness/sprawl since they are two major activities that use up space in a metropolitan area. Hence, USI is based not only on population (e.g., Pop-Moran’s I in Table 1), but also employment (e.g.,

Pop-Emp-Moran’s I in Table 1), which altogether can reveal if urban “sprawl” measured

according to population spatial distribution pattern is not really happening if employment spatial distribution is counted; for example, population-based decentralized central city may be

substituted by employment space, which does not weaken central city but possibly the other way around, and may not be treated as urban “sprawl” but a result of “population bid out by

commercials.”

To count population and employment together for measuring compactness/sprawl, an assumption is adopted that one percentage point loss/emigration of population in one sub-area in a metropolitan area can be made up by the gain/immigration of one percentage point of

employment, and vice versa. This assumption can lead to weightings of employment and population, which allow counting them altogether, and the following index is developed (Table 1):

Mix Index of Population and Employment

= (Population + Employment * W) population-based, or

= (Population / W + Employment) employment-based

Where W = Metropolitan Population / Metropolitan Employment

A panel regression model is built to probe factors causing the spatial distribution of population in a metropolitan developing in a sprawling fashion, based on the experience of intermediate and large 23 metropolitan areas in Taiwan for 1966, 1980, 1990, and 2000. This panel model, with change in the degree of urban “sprawl” as independent variable, composing three time-point time-series data, allows it to examine the extent to which factors, for example at the beginning of a time point (e.g., 1966 of the 1966-1980) affects the change of urban form during this time point. A variety of the 23 metropolitan areas expands the variance of degrees of change of urban “sprawl” and the characteristics of the metropolitan areas. The selection of change of urban “sprawl” in a dynamic sense, as opposed to degree of urban “sprawl” in a stable sense, makes it possible to uncover what makes a metropolitan more compact or sprawling (more in the notes). The findings of this model serve to examine the research hypothesis.

Following are framework and variables of the panel regression model of the dynamic urban

“sprawl” model. Out of the 36 metropolitan areas in Taiwan (Yang, 2001), only 26 intermediate to large metropolitan areas are selected. Secondary data are collected for four time points, 19661, 1980, 1990, and 20002, ten to 15 years apart, which is probably not too short to observe changes in a metropolitan areas in terms of land use, transportation infrastructure,

socio-economic characteristics, and urban form. These four time points formulate three time time points of observation, i.e., 1966~1980, 1980~1990, and 1990~2000, making up 78 (3*26) observations in total.

Change of global Moran’s I, an urban sprawl index (USI for short hereafter), is chosen as dependent variable to measure degree of urban “sprawl,” and independent variables are classified according their impacts on urban “sprawl” (Equation 1). Global Moran’s I is able to gauge the degree to which high-density sub-areas in a metropolitan area are clustered; it, hence, is able to distinguish compactness from sprawl—the more compact the urban form, the higher the value of Moran’s I3 (Tsai, 2005). Hence positive value of change of Moran’s I indicates a metropolitan area is becoming more compact, and vice versa. Moran’s I are calculated from the city-based data for each metropolitan areas with Spatial Autocorrelation Tool of ArcGIS 9.2. The

weighting in calculating the Moran coefficient is the inverse distance between the centroids of two cells, which more sensitive and accurate in characterizing metropolitan forms than contiguity criteria (i.e., 0 for discontinuous cells, and 1 for continuous cells).

Theoretically potential factors leading to more sprawling urban form are classified into pushing-outward, pulling-outward, and accessibility categories (Equation 1, Table 1), a gravity-model like theory. First of all, pushing-outward factors can further be classified into pushing outward from central city and pushing outward in general. The former are those pushing population out of central cities, primarily unmet housing demand for central-city living in terms of quantity, quality, and price of housing, and quality of community theoretically; data dated back more than four decades are only available for quantity-based for this empirical study.

As a consequence, variables are incorporated that measure the housing demand not served by new housing in central city between previous time point (i.e., (t-1)) and current time point (i.e., t0), not by existing empty housing stock at previous time point, or by both a whole (Table 1).

Hence, pushing-outward factors are mostly related to under-served housing demand. The pushing outward in general category are those factors that measures increasing spatial needs, which if not served vertically (higher floor area rate), is likely to be served horizontally, that is moving outward. The variables include population growth rate, change of household size, which altogether contribute to household growth rate.

Secondly, by the same token, pulling-outward factors are those pulling population out of

1 1966 is selected instead of 1970 because some most significant variables for this research are only available then.

2 Some data are only available one year before or after.

3 Sprawling indexes measuring the degree that activities are evenly distributed, such as Shannon’s entropy and Sprawl Index, are not selected since they are unable to measure spatial relationship (Tsai, 2005).

9

central cities to outskirts, majorly attraction of outskirt characteristics, including quality of communities, and quality, quantity, and price of housing. Due to the limited availability of the forty-year-old data, merely new housing added during the time point in question, and existing unoccupied housing at the beginning of the time point are incorporated. In contrast to under-served housing demand of pushing-outward factors, the incorporated pulling-outward factors are supply of housing.

Finally, accessibility variables are those that facilitate the pushing/pulling outward factors, and it can be further broken down by public policy related, and socioeconomic characteristics: the former include highway length, number of highway entrance/exits, intercity roadway length, and the latter include wage, automobile ownership, and moped ownership (Table 1).

Other than the above three types of factors causing urban sprawl, there are variables that might also affects the change of urban form over time that needs to be controlled for. These control variables may include characteristics of a metropolitan area, such as population at the beginning time point of the observation time point (i.e., t-1), percentage change of population size (t-1~t0), percentage change of household size (t-1~t0), and population density (t-1) (Table 1).

The factors affecting change of urban form in terms of urban sprawl/compactness can be of dynamic or stable fashion. In this dynamic model measuring the change of degree of urban sprawl/compactness during an observation time point (i.e., △USIi, t-(t-1) in equation 1), variables shown above can be both the change of certain characteristics during a time point (i.e., dynamic);

for example, incremental number of highway interchange during a time point may lead to worsening degree of urban sprawl. On the other hand, and the characteristic at certain time point (i.e., stable) may also cause more sprawling urban form during a time point; for example, a metropolitan of high population density (at the beginning of the time point) may have higher floor area cap, and hence given others equal, might still develop in a more compact shape; or pressure from high-density living might cause emigration from high-density to low-density sub-areas, and hence leads to less compact form. In short, the above sprawl-causing factors are classified according to the mechanism that variables affect urban form; they can be further broken by time-series (dynamic) and difference between metropolitan areas (cross-sectional) for the panel regression model.

USIi, t0~(t-1)

＝α

_i,＋

β

₁X_push ^＋

β

₂X_pull ^＋

β

₃Xaccessibility ^＋

β

₄X_control ^＋

ε

_{i, t0}

(1)

Where

△USIi, t0-(t-1)is change of urban sprawl index (USI) of metropolitan area i from previous time point (t-1) to current time point (t0), where i = metropolitan area, and t = time

α

i,is the unknown intercept for each entity X_{push (t-1)} is the pushing outward factors,

X_{pull (t-1)} is the pulling outward factors

Xaccessibility(t-1)is factors of accessibility connecting between inner cities and outskirt,

β

is the coefficients for X, and

ε

is is the error term

It is noteworthy that the independent variables come in two kinds of format: the stable and dynamic; the stable statistics shows the status at the beginning or mid-term of an observation time point, and dynamic statistics gauges the change of characteristic of a dynamic statistic. In this change of urban sprawl model, this stable arrangement of independent variables lead to

cross-sectional like analysis, i.e., different characteristic of metropolitan areas, regardless which time point they occurs, leads to more compact or less sprawling development in the future. On the other hand, the dynamic arrangement measures whether the change of status leads to a more compact/less sprawling urban form in the future, which is more like time-series analysis.

11

Table 3 Variables adapted in developing panel models of urban “sprawl,”

by pushing outward, pulling outward, and accessibility categories

Variable Alternative Variable(s)

Urban Sprawl Index (USI):

Change of Sprawl Index (SI), population based (△SI(t-1)~t0) Pushing outward factors:

Overall pushing outward:

Household growth rate between previous time point (i.e., t-1) and current time point (i.e., t0) (△HH%(t-1)~t0)

Population growth rate between previous time point and current time point (△Pop%(t-1)~t0)

Percentage change in household size (△HH-Size% (t-1)~t0) Pushing outward from central cities:

Ratio of household growth not served by new housing in central cities between previous time point and current time to point to metropolitan households [=(predicted household growth in central cities(t-1)~t0 – new housing in central city(t-1)~t0)/ total metropolitan households t0]

(△HHcc-Not-Served-By-New-Housingcc (t-1)~t0)

Ratio of Household growth not served by new and existing unoccupied housing in central cities between previous time point and current time pointto metropolitan households

[=(predicted household growth in central cities(t-1)~t0 –new and existing empty housing in central cities) t0 / total metropolitan households t0]

(Ratio-△HHcc-Not-Served-By-All-Housingcc(t-1)~t0) Ratio of Migration from central cities to outskirts between

previous time point and current time point to metropolitan households [(predicted number of households in central cities

t0 – real number of household in central cities t0)/ total metropolitan households t0]

(Ratio-Migrationcc (t-1)~t0) Pulling outward factors:

Unoccupied housing stock in the outskirt in previous time point (Outskirt-Housing-Stock(t-1)) (absolute quantity)

(Quantity-based, as opposed to price-, and quality-based)

Percentage unoccupied housing stock in the outskirt in previous time point

(=Outskirt-Housing-Stock(t-1)/Metropolitan-Housing-Stock(t-1)) (%Outskirt-Housing-Stock(t-1)) (relative quantity) New housing built from previous time point to current time point

in the outskirt (Outskirt-New-Housing(t-1)~t0) (absolute quantity)

Percentage new housing built from previous time point to current time point in the outskirt

(=Outskirt- New-Housing(t-1)~t0/

Metropolitan-New-Housing(t-1)~t0) (relative quantity) (%Outskirt-New-Housing(t-1)~t0)

Ratio of all available housing in the outskirts to metropolitan households

(New housing (t-1)~t0 plus unoccupied housing stock in outskirts (t-1))/

total metropolitan households t0) (Ratio-Outskirt-All-Housing(t-1)~t0) Accessibility variables:

Public policy related:

Highway length at mid-time-period (i.e., t-0.5) (KM) (Highway-KM(t-0.5))

Highway density at mid-time-period (KM/KM²) (Highway-Density(t-0.5))

Number of highway exits at mid-time-period (Highway-Exit(t-0.5))

Density of highway exits at mid-time-period (Highway-ExitDensity(t-0.5))

Socioeconomic characteristics

Household automobile ownership at mid-time-period (Automobile-Ownership(t-0.5))

Household moped ownership at mid-time-period (Moped-Ownership(t-0.5))

Control variables

Population at previous time point (Pop (t-1))

Household density at previous time point (HH-Density (t-1))

Growth rate of the whole metro is picked to calculate newly developed demand for

housing/space in central city. Submitting the population change between t1 and t2 in central city (the green part) from the population growth above (population t1 * growth rate t1-t2)(the purple part) the magnitude of pushing-outward forces are calculated (the pink part) (Figure 2).

Expected POP growth = can be assumed as the average growth rate of the

Figure 2 The Way of Calculating Pushing-outward Forces

在文檔中 土地混合使用對住宅價格的影響---台北市實證研究 (頁 26-35)