Citation
Dynamics of the amenity city

Material Information

Title:
Dynamics of the amenity city an agent-based simulation of neighborhood location decisions
Creator:
Kim, Yuseung
Publication Date:
Language:
English
Physical Description:
xiv, 165 leaves : illustrations ; 28 cm

Subjects

Subjects / Keywords:
Multiagent systems ( lcsh )
Amenity migration ( lcsh )
Neighborhoods ( lcsh )
Residential mobility ( lcsh )
Amenity migration ( fast )
Multiagent systems ( fast )
Neighborhoods ( fast )
Residential mobility ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 149-165).
General Note:
College of Architecture and Planning
Statement of Responsibility:
by Yuseung Kim.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
677842537 ( OCLC )
ocn677842537
Classification:
LD1193.A735 2010d K55 ( lcc )

Downloads

This item has the following downloads:


Full Text
DYNAMICS OF THE AMENITY CITY: AN AGENT-BASED SIMULATION OF
NEIGHBORHOOD LOCATION DECISIONS
by
Yuseung Kim
B.S.,Yonsei University, 1999
M.R.P., Cornell University, 2001
A thesis submitted to the
University of Colorado Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Design and Planning
2010


2010 by Yuseung Kim
All rights reserved.


This thesis for the Doctor of Philosophy
degree by
Yuseung Kim
has been approved
by
O'-
oLfUL'
Z8
2-o \ o
Date


Kim, Yuseung (Ph.D., Design and Planning)
Dynamics of the Amenity City: An Agent-Based Simulation of Neighborhood Location
Decisions
Thesis directed by Professor Brian Muller
ABSTRACT
The purpose of this study is to identify urban amenity-based residential location decision
factors by interviewing experts in the fields and to test the factors explanatory power by
applying the Agent-Based Modeling (ABM) technique. For the simulation model, it was
assumed that household agents decide their residential location based on the physical
environments and non-physical or social environments. Physical environments include the
relative locations of urban amenities as well as the relative locations of traditional location
decision factors such as transportation networks to potential housing locations. Non-
physical or social environments include both positive and negative social interactions
between households and their neighbors.
Interview results confirm the emergence of urban amenities as important location decision
factors in local housing markets in the study areas in the Colorado Front Range region. The
ABM simulation scenarios with the traditional job-accessibility variables, however,
outperform the simulation scenarios with urban amenity variables in all three test areas:
Boulder, Denver, and Louisville. The study results show that during the study period
(1990-2000), job-accessibility-based location decision factors were more powerful than
urban amenity factors in the residential location decision processes. Considering the
increased discussions about the role of urban amenities in attracting human capital and the
cultural trends towards entertainment and consumption-oriented lifestyles, a different
outcome with higher explanatory power for the urban amenity variables is expected for the
simulation outcomes with more recent data.


Testing the explanatory power of the existing amenity-based urban development and
location decision theories such as Consumer City, Creative Center, and the City as an
Entertainment Machine discussions result in heterogeneous outcomes in each study area.
Simulation outcomes show that each study area has a unique set of urban amenities that are
effective in attracting people, as well as a unique set of comparative importance among
them. It is argued that a careful analysis of the local conditions and the area-specific
location decision factors are required before the application of the amenity-based
development theory to a local place. Finally, using the best performing amenity-based
model in each study area, future demographic distribution patterns in the study areas are
predicted with possible future simulation scenarios.
This abstract accurately represents the content of the
publication.
Signei
es thesis. I recommend its
Brian Muller


DEDICATION PAGE
I dedicate this dissertation to my parents and my wife for their love and endless support
while I was completing this dissertation.


ACKNOWLEDGMENT
My special thanks to my advisor, Brian Muller, for his guidance and support to my
research. I also wish to thank the members of my committee for teaching and sharing their
valuable knowledge. Dr. Thomas Clark gave me insightful comments at every step of my
research progress. Dr. Kevin Krizek assisted me with his constructive suggestions on my
research, while arranging financial support for me as a program chair. Graham Billingsley,
FAICP, shared his experience of more than 30 years of planning practices. Dr. Subhrendu
Gangopadhyay helped me to refine my model with his expertise in the modeling and
simulation study. This dissertation would not have been possible without them.
I also would like to express my appreciation to my colleagues in the Land Use Futures Lab
and all the PhD students at the University of Colorado for their friendship, encouragement,
and inspiration.


TABLE OF CONTENTS
Figures.........................................................................xi
Tables........................................................................xiii
Chapter
1. Introduction...............................................................1
1.1 Background.................................................................1
1.2 Research Tool and Research Questions.......................................3
1.3 Outline of the Study.......................................................4
2. Theory of Urban Residential Location.......................................5
2.1 Phase 1(1920s ~): Market-Based Approaches..................................5
2.1.1 Bid-Rent Approach........................................................5
2.1.2 Human Ecology Approach...................................................7
2.2 Phase II (1950s ~): Non-Market-Based Approaches............................8
2.2.1 Political Economy Approach...............................................8
2.2.2 Residential Mobility Approach............................................9
2.3 Phase III (2000s ~): Amenity-Based Approach................................11
2.3.1 Residential Location Decision Factors...................................11
2.3.2 Consumption-based Neighborhood Development...............................15
2.3.3 Urban Amenity-Based Location Decision Theory..............................16
2.3.3.1 Consumer Cities.........................................................17
2.3.3.2 The City as an Entertainment Machine....................................18
2.3.3.3 Superstar Cities........................................................19
2.4 Demographics in Urban Residential Market..................................20
2.4.1 Major Instigators of Urban Resurgence...................................20
2.4.1.1 Young Professionals.....................................................21
2.4.1.2 The Seniors.............................................................24
2.4.2 Social Networks Among Demographic Groups................................28
2.4.2.1 Social Interactions and Conflicts.......................................28
2.4.2.2 Neighborhood Effects....................................................30
2.5 The Role of the Local Governments.........................................33
viii


3. Theory and Applications of Urban Modeling...................................36
3.1 Historical Development.......................................................37
3.1.1 Emergence of Urban Models..................................................37
3.1.2 Large Scale Model Critiques................................................39
3.1.3 Land Use-Transportation Models.............................................40
3.1.4 Planning Support Systems...................................................41
3.2 Evolution of Urban Modeling: CA/ABM........................................45
3.2.1 Cellular Automata (CA).....................................................45
3.2.2 Agent-Based Model (ABM)....................................................46
3.2.3 CA/ABM Approach............................................................47
3.3 CA/ABM applications........................................................48
3.3.1 Residential Location Modeling..............................................49
3.3.2 Segregation Modeling.......................................................53
4. Research Methods.............................................................56
4.1 Study Areas................................................................56
4.2 Data.......................................................................58
4.2.1 Expert Interview...........................................................58
4.2.2 Data in Digital Format.....................................................60
4.3 Variable Selection.........................................................63
4.4 Agent-Based Model..........................................................64
4.4.1 Construction of Agents.....................................................64
4.4.1.1 Agent Categorization......................................................64
4.4.1.2 Agent Behavior............................................................66
4.4.2 Construction of an Agent-Based Model......................................69
4.4.2.1 Simulation Model..........................................................69
4.4.2.2 Validation of the Model...................................................71
4.4.3 Future Simulation..........................................................76
4.4.3.1 Amenity District Policy Simulation........................................76
4.4.3.2 Location Preference Simulation............................................76
4.4.3.3 Different Size of Neighborhood Simulation.................................77
5. Results......................................................................78
5.1 Interview Results............................................................78
5.2 Explanatory Model Simulation Results.......................................81
5.2.1 Pilot Simulation...........................................................81
5.2.2 Real Simulation............................................................82
IX


5.2.2.1 Boulder....................................................................82
5.22.2 Denver.................................................................86
5.2.2.3 Louisville.................................................................90
5.2.2.4 Model Evaluation...........................................................94
5.2.3 Addition of Neighborhood Variable..........................................97
5.2.3.1 Boulder.................................................................97
5.2.3.2 Denver.................................................................99
5.2.3.3 Louisville................................................................101
5.2.4 Amenity-Based Theory Test.................................................103
5.2.4.1 Boulder................................................................103
5.2.4.2 Denver....................................................................106
5.2.4.3 Louisville................................................................108
5.3 Prognostic Model Simulation Results.......................................110
5.3.1 Pilot Simulation..........................................................110
5.3.2 Future Simulation in the Study Areas......................................112
6. Conclusion..................................................................117
6.1 Research Findings...........................................................117
6.1.1 Interview.................................................................117
6.1.2 Explanatory Model.........................................................118
6.1.3 Prognostic Model..........................................................120
6.2 Significance of Research....................................................121
6.3 Limitations of Research.....................................................123
6.4 Future Research Agenda......................................................124
Endnotes..........................................................................125
Appendix
A. Survey Instrument.........................................................131
B. Simulation Outcomes with Neighborhood Variable............................134
C. Sensitivity Test Results..................................................143
Bibliography......................................................................149
x


FIGURES
Figure
4.1 Study Areas, Front Range Region, Colorado.....................................56
4.2 Hierarchical Diagram for the AHP Survey.......................................59
4.3 Distributions of Urban Amenities in the Study Areas, 1990, 2000...............61
4.4 Calculation of the Amenity Score..............................................67
4.5 Comparison of the Total Score Surfaces among Three Agent Types................69
4.6 Computational Approach........................................................71
4.7 Conceptual Diagram of Model Run and Model Validation Process..................73
5.1 Pilot Simulation Results......................................................81
5.2 Similarity Index Change, Pilot Simulation.....................................82
5.3 Real Agent Distribution, Boulder, CO, 1900 and 2000...........................83
5.4 Simulation Results, Boulder, CO...............................................84
5.5 Similarity Index Change, Boulder, CO, 1990-2000...............................85
5.6 Distribution of Urban Amenities by Agent Type, Boulder, CO, 1990..............86
5.7 Real Agent Distribution, Denver, CO, 1900 and 2000............................87
5.8 Simulation Results, Denver, CO................................................88
5.9 Similarity Index Change, Denver, CO, 1990-2000................................89
5.10 Distribution of Urban Amenities by Agent Type, Denver, CO, 1990...............90
5.11 Real Agent Distribution, Louisville, CO, 1900 and 2000........................91
5.12 Simulation Results, Louisville,CO.............................................92
5.13 Similarity Index Change, Louisville, CO, 1990-2000............................93
5.14 Distribution of Urban Amenities by Agent Type, Louisville, CO, 1990...........94
5.15 Box Plotted Aggregated Discrepancy Index......................................95
5.16 Similarity Index Change with 3X3 Neighborhood, Boulder, CO, 1990-2000.........98
5.17 Similarity Index Change with 3X3 Neighborhood, Denver, CO, 1990-2000.........100
5.18 Similarity Index Change with 3X3 Neighborhood, Louisville, CO, 1990-2000 ....102
5.19 Box Plot, Discrepancy Index for Amenity-Based Theory Test, Boulder, CO......104
XI


5.20 Amenity-Based Theory Test Results, Boulder, CO, 1990 and 2000.............105
5.21 Box Plot, Discrepancy Index for Amenity-Based Theory Test, Denver, CO.....106
5.22 Amenity-Based Theory Test Results, Denver, CO, 1990 and 2000..............107
5.23 Box Plot, Discrepancy Index for Amenity-Based Theory Test, Louisville, CO.108
5.24 Amenity-Based Theory Test Results, Louisville, CO, 1990 and 2000..........109
5.25 Pilot Future Simulation Results: Clustering vs. Even Distribution of Amenities... 110
5.26 Pilot Future Simulation Results: Changing vs. Permanent Location Preference.... 111
5.27 Pilot Future Simulation Results: Varying Neighborhood Sizes...............111
5.28 Future Simulation Results: Changing vs. Permanent Preference..............113
5.29 Similarity Index Change: Boulder, Denver, Louisville, 2000-2020...........114
5.30 Future Simulation Results: Varying Neighborhood Sizes, Boulder, CO........115
5.31 Future Simulation Results: Varying Neighborhood Sizes, Denver, CO.........115
5.32 Future Simulation Results: Varying Neighborhood Sizes, Louisville, CO.....116
A. l AHP Survey Instrument.....................................................132
B. l Simulation Results with 3X3 Neighborhood, Boulder, CO.....................134
B.2 Simulation Results with 5X5 Neighborhood, Boulder, CO.....................135
B.3 Simulation Results with 7X7 Neighborhood, Boulder, CO.....................136
B.4 Simulation Results with 3X3 Neighborhood, Denver, CO......................137
B.5 Simulation Results with 5X5 Neighborhood, Denver, CO......................138
B.6 Simulation Results with 7X7 Neighborhood, Denver, CO......................139
B.7 Simulation Results with 3X3 Neighborhood, Louisville, CO..................140
B.8 Simulation Results with 5X5 Neighborhood, Louisville, CO..................141
B. 9 Simulation Results with 7X7 Neighborhood, Louisville, CO..................142
C. l Box Plotted Sensitivity Test Results, Boulder, CO.........................144
C.2 Box Plotted Sensitivity Test Results, Denver, CO..........................146
C.3 Box Plotted Sensitivity Test Results, Louisville, CO......................148
xii


TABLES
Table
2.1 Location variables in select hedonic price models.............................13
3.1 Comparison of the urban models,...............................................43
3.2 Comparison of select agent-based residential location models..................52
4.1 Population change, 1990-2000, study areas.....................................57
4.2 GIS data layers...............................................................60
4.3 Quarterly Census of Employment and Wages data structure.......................60
4.4 Urban amenity changes in the study areas, 1990-2000...........................62
4.5 Urban amenity variables from the existing literature and pilot interviews.....63
4.6 Agent categorization..........................................................65
4.7 Controllable model parameters.................................................68
4.8 Simulation scenarios..........................................................74
4.9 Weight sensitivity test.......................................................75
4.10 Urban amenity-based theory test...............................................75
5.1 AHP survey results, Boulder, CO...............................................79
5.2 AHP survey results, Denver, CO................................................80
5.3 AHP survey results, Louisville, CO............................................80
5.4 Aggregated discrepancy index..................................................96
5.5 ANOVA analysis results for comparing the discrepancy indexes..................96
5.6 Aggregated discrepancy index with neighborhood variable, Boulder, CO..........98
5.7 Aggregated discrepancy index with neighborhood variable, Denver, CO..........100
5.8 Aggregated discrepancy index with neighborhood variable, Louisville, CO......102
5.9 Aggregated discrepancy index for amenity-based theory test, Boulder, CO......104
5.10 Aggregated discrepancy index for amenity-based theory test, Denver, CO.......106
5.11 Aggregated discrepancy index for amenity-based theory test, Louisville, CO...108
6.1 Average variances between weights by agent types.............................118
6.2 Comparison of aggregated discrepancy indexes.................................120
xiii


C.l Sensitivity test results, Boulder, CO................................................143
C.2 Sensitivity test results, Denver, CO.................................................145
C.3 Sensitivity test results, Louisville, CO.............................................147
XIV


1.
INTRODUCTION
1.1 Background
With the geographic redistribution of population and resurgence of the central cities in
metropolitan areas in the US (in spite of the decentralization of the employment centers in
recent years) the discussions of human capital and amenity-based location decision theories
are gaining momentum.' There are an increasing number of human capitalists who argue
the importance of people as the motor force behind urban development (Florida, 2002;
Lloyd and Clark, 2001; Glaeser et al., 2001; Glaeser, 1998; Lucas, 1988; Romer, 1994).
Those human capitalists are in part indebted to the insight of Jane Jacobs. Jacobs (1961)
noted several decades ago the ability of cities to attract people and thus spur economic
growth. Lucas (1988) argued the productivity effects that come from the clustering of
human capital as the critical factor in regional prosperity. Glaeser (1998) found empirical
evidence that human capital is one of the central factors in regional growth. According to
Glaeser, such clustering of human capital is the ultimate cause of regional agglomeration
of firms-not merely to tap the advantages from linked networks of customers and
suppliers. Florida (2002) correlated the proportion of particular types of workers, the
creative class of artists, engineers, software writers, teachers, and the degree of regional
development and found a strong relationship between them. He defined spatial
concentrations of this creative class as creative centers and explained:
The creative centers are not thriving for such traditional economic reasons as access
to natural resources or transportation routes. Nor are they thriving because their local
governments have given away the store through tax breaks and other incentives to
lure business. They are succeeding largely because creative people want to live there.
The companies then follow the peopleor, in many cases, are started by them.
Creative centers provide the integrated eco-system or habitat where all forms of
creativityartistic and cultural, technological and economiccan take root and
flourish.
1


Human capitalists in the urban study area commonly agree on the importance of urban
amenities and the role of social networks in dense urban areas in attracting people,
especially high human capital with high educational and income levels.
The willingness to pay for urban amenities relative to the amenities of suburbs may
have increased. For example, it may be that the cities have always had more
museums, but the value place on proximity to museums rises with income and
education (Glaeser and Gottlieb, 2006).
Workers in the elite sectors of the postindustrial city make quality of life demand,
and in their consumption practices can experience their own urban location as if
tourists, emphasizing aesthetic concerns. These practices impact considerations about
the proper nature of amenities to provide in contemporary cities (Lloyd and Clark,
2001).
Creative people are not moving to these places for traditional reasons. The physical
attractions that most cities focus on buildingsports stadiums, freeways, urban malls
and tourism and entertainment districts that resemble theme parks-are irrelevant,
insufficient or actually unattractive to many urban dwellers. What they look for in
communities may be high-quality amenities and experiences, an openness to diversity
of all kinds, and above all else the opportunity to validate their identities as creative
people (Florida, 2002).
This view is a radical departure from the traditional interpretation of the residential
location decision process because traditional studies on the housing search were focused on
the producer motivation of the location decision process: they are based on the trade-off
relationship between job accessibility and rent rate. Furthermore, those trade-off
relationship-based housing location studies had either not incorporated space or did so
using rudimentary Euclidean measures in residential location process.
It is quite probable, however, that cities reason for being changed (i.e. from city as a
growth machine (Molotch, 1976) to city as an entertainment machine (T. N. Clark,
2004)) as well as their residents reason for living within them changed (i.e. from make a
living to enjoy a life) over the past several decades. Also, the amenity-based residential
location theories may have comparable explanatory power to the traditional location
theories at least for certain demographic groups.
The study on the role and the significance of urban amenities in the housing location
decision process becomes more important when we consider that there are increasing
2


numbers of political leaders in North America who are convinced by human capital
discussions." They are exercising their planning efforts to attract high human capital to
their localities. These policies include cash contributions, financing aid, tax abatements,
zoning and building code variances, new or improved infrastructure, preparation of an
entertainment-oriented district, and the creation of dedicated taxes for physical
improvements.
However, without clear understanding of the role and the significance of urban amenities
in the residential location decision process, those kinds of local monetary and efforts
spending may end up as unfortunate urban policies. Again, we need clear understanding of
the residential location decision process.
1.2 Research Tool and Research Questions
In this research, the significances of various urban amenity variables will be tested using
the agent-based modeling (ABM) technique. Agent-based models (ABMs) are computer
representations of systems that are comprised of multiple, interacting actors (Brown,
2005). Agent behavioral rules for this study will be developed by interviewing experts in
the local housing markets in the Colorado Front Range region. The developed model will
be applied to three Colorado cities with varying city sizes and with different distribution
patterns of urban amenities: Boulder, Denver, and Louisville. By simulating the individual
actions of heterogeneous households with diverse locational preferences, and measuring
the resulting system behavior and outcomes over time, the model will provide a useful tool
for understanding urban phenomena and the processes and dynamics behind them.
I will use the agent-based residential location decision model in two ways. The first one is
the explanatory way: to use the model as a means to understand the urban housing location
decision process using the historic data. My explanatory research question is What is the
role of urban amenities in the residential location decision process? The sub-question
related to this question is asking the explanatory power of the various existing amenity-
3


based location decision theories: Which amenity-based urban development and location
decision theory successfully describes demographics and residential distribution trends?
The second way of using the agent-based residential location decision model is the
prognostic way: to use the model to extrapolate the trends, and to evaluate scenarios, and to
predict the future states. My prognostic research questions is How changes in location
decision factors such as urban policy change or individual preference change will impact
the future distribution of population and the formation of the neighborhood?
Finally, I will test the usefulness of the agent-based approach in urban modeling by
constructing an agent-based residential location decision model and applying the model for
the future simulation. The question is Can an agent-based model be a useful
computational model for urban simulation?
1.3 Outline of the Study
Chapter 1 introduces the research topic and provides the relevant context for it. This
chapter includes background of the problem, research questions, and outline of the study.
Chapter 2 is a description of the existing studies on urban residential location theories and
their deficiencies. Chapter 3 is a description of the existing studies on urban modeling and
their deficiencies. Chapter 4 is a description of my research design plan. This chapter
includes explanations on data, data collection methods including my interview plan, data
analysis method, assumptions, agent typology, and simulation model schema. Chapter 5 of
the study is an analysis of the interview and survey results as well as an application of the
survey results to the simulation model. The spatial distribution of heterogeneous agents is
analyzed. The main objective of chapter 6 is to summarize the study and highlight its
primary contribution to the theory of urban residential location. I underscore the findings
from this study, including the significance of urban amenities in residential location
decision processes that will be analyzed from the simulation model.
4


2. THEORY OF URBAN RESIDENTIAL LOCATION:
A REVIEW OF LITERATURE
In this chapter, I review the discussions on existing theories on urban residential location
decision process and evaluate their explanatory powers. I follow the chronological order of
the development of the theories from market-based (1920s~) to non-market-based (1950s~)
to more recent amenity-based (2000s~) location decision theories. Additionally, I review
the studies on the demographic structures in urban housing markets and the role of the
local governments on the residential location decision. Most articles and research reviewed
in this chapter are from academic journals or independent volumes by academic authors.
However, due to the novelty of the topic, there are several issues that were not studied yet
in academia. In those cases, I review newspapers and professional journals in real estate
and land development fields.
2.1 Phase I (1920s ~): Market-Based Approaches
2.1.1 Bid-Rent Approach
The market-based approach has its origins in the sociological observations of the Chicago
School in the 1920s (Maclennan, 1982). The market approachs main concept of bid rent
functions provides the foundation for most microeconomic models of urban spatial
structure (Anas, 1982). Ricardo (1921) is credited with the idea of bid rent function, which
explains that the price of agricultural land is determined by its fertility. In von Tinmens
location model, accessibility replaces fertility as the determinant of land rent. Ever since
von Thiinen gave his version of the monocentric city-region, geographers, urban
economists, and planners have been working on theories of city structure that can both
explain and predict the way in which cities are formed and have evolved (Huu Phe &
5


Wakely, 2000). Alonso (1964) expanded upon the von Thiinen model to the model of
household location decisions. When applied to the housing location model, housing cost
has a trade-off relationship with the commuting cost: housing costs should rise as the
distance to the employment center falls since households would be willing to pay more in
order to save time getting to work. This is basically an equilibrium model of location
choice. When all households are satisfied with their location choices, locational
equilibrium occurs, that is, no household wants to change its location. If there is a change
in conditions that temporarily disrupts the equilibrium, all households are expected to
change residence without cost and instantaneously by repeating the bid-auctioning process.
Extensions of the market-based approach include: estimation of the housing-price and
land-rent function (J. R. Jackson, 1979; Mills, 1969); estimation of the relationship
between various housing characteristics (including location and different components of
housing) and the price of housing (Kain & Quigley, 1970; Straszheim, 1975; King, 1975);
estimation of population and employment density functions (Mills, 1972; Muth, 1969); a
model of urban land use that uses linear and nonlinear programming to allocate land to
alternative activities (Mills, 1976); and the discussion of the several renditions of the
traditional monocentric model (Wheaton, 1977). The access/space trade-off theorys
success in replicating empirical regularities in Western cities has made the theory a
preferred analytical tool (Bassett & Short, 1980), and it is now described as the dominant
paradigm of urban economic research (Maclennan, 1982) and a normative theory (Fujita,
1989).
As U.S cities evolve during the last half of the twentieth century, however, the explanatory
power of this traditional model become weaker. Central city resurgence is a good example
of the phenomenon that the market approach fails to explain. The trade-off theory suggests
that the rich have a natural propensity to live in large parcels in the suburbs, where the land
is cheap and the environment is good, because they can afford the transportation costs. The
poor live in the inner city because they cannot pay for high transport costs. Despite a heavy
degree of simplification, this was more or less accepted until the 1970s and 1980s, when
the rich moved into derelict areas in the inner city, renovated them and stayed, in a
6


widespread phenomenon later called gentrification (Hamnett & Williams, 1980). This
migration is difficult to explain without the risk of contradicting the fundamental
assumptions of the access/space trade-off theory.
2.1.2 Human Ecology Approach
Based on bid-rent functions, Burgess (Park & Burgess, 1925) explained distribution of
social groups within urban areas. His model depicts urban land use in concentric rings: The
CBD is in the middle of the model, and the city expanded in rings with different land uses.
He categorizes urban land in six different land uses: CBS as a center, factory zone, the
transition zone of mixed residential and commercial uses, low-class residential zone (inner
city), better-quality middle-class homes (outer suburbs), and commuter zone.
Hoyt (1939) modified the concentric model of Burgess, suggesting that zones expand
outward from the city center along railroads, highways, and other transportation arteries.
Hoyt theorized that cities tend to grow in wedge-shaped patterns, or sectors, emanating
from the CBD and center on major transportation routes. Higher levels of access meant
higher land values; thus, many commercial functions would remain in the CBD, but
manufacturing functions would develop in a wedge surrounding transportation routes.
Residential functions would grow in wedge-shaped patterns with a sector of low-income
housing bordering manufacturing/industrial sectors (traffic, noise, and pollution makes
these areas the least desirable), while sectors of middle- and high-income households were
located furthest away from these functions.
Harris and Ullman (1945) proposed multiple nuclei model in their article The Nature of
Cities. According to their model, a city contains more than one center around which
activities revolve. Some activities are attracted to particular nodes while others try to avoid
them. For example, a university node may attract well-educated residents, pizzerias, and
bookstores, whereas an airport may attract hotels and warehouses. Incompatible land use
7


activities will avoid clustering in the same area, explaining why heavy industry and high-
income housing rarely exist in the same neighborhood.
Even though these early attempts to explain the arrangement of urban land uses indicate a
broader principle of urban organization, they are based on early twentieth century urban
development and transportation patterns. Changes in urban environments such as
advancement in transportation and information technology, changing local urban politics,
globalizing economy, and diversified location preferences by households as well as firms
made zoning of the cities difficult.
2.2 Phase II (1950s ~): Non-Market-Based Approaches
2.2.1 Political Economy Approach
The political economy approach analyzes the residential distribution and city structure in
terms of social groups inhabiting urban areas, with some groups taking advantage or
control over others. Research in this group focuses on differentiation and disparities
between social groups with relation to the accessibility to urban goods and services, on
how division of urban space reflect different structures of consumption, and on the material
and symbolic values that different groups attach to these different consumption landscapes
(Soja, 2000; Topalov, 1989; Dear & Flusty, 1998; Harvey, 1989; Scott & Soja, 1996).
The political economy group critiques the market-based theories as ahistorical (Huu Phe &
Wakely, 2000). Researchers in this group suggest that residential location patterns are less
strongly influenced by market competition, land and house prices and more strongly
influenced or even manipulated by capital through monopolistic rent (Harvey, 1973; Smith,
1987). They argue that space differentiation leads to increasing inequity in access to other
urban services.
Smiths (1987, 1979) rent gap theory is an example of this groups theoretical
explanations to the urban phenomenon of gentrification. He claimed that capital makes
8


investment decisions in decayed urban areas because of the potential gains from substantial
increases in land rent and land value. Population turnover is a consequence of the process.
This effect is due to the gap that has been created between the current rent and the expected
rent after rehabilitation, i.e., between capitalized rent and potential rent (Diappi & Bolchi,
2008). Smith argued that the rent gap is but one illustration of the uneven spatial
development characteristic of market societies, and thus revitalization is a back-to-the-city
movement by capital not people (Smith, 1979). His model is criticized, however, as a
partial attempt to explain gentrification. The model ignores public sectors and the demand
side by not identifying the role of local governments and private actors in neighborhood
revitalization (Hamnett, 1984, 1991; Ley, 1986).
More recently, Fainstein (2001), focusing on London and New York, described the
contemporary urban development patterns as a confluence of an explosion of speculative
building for profit and a surging demand for space within a context of local incentives to
growth. She explained that, in addition to the traditional market factors of demand and
supply, individual developers and government policy play an active role in the dynamics of
the real estate market. Her work shows the importance of the way in which the
development industry is organized within the local economy, and the institutional relations
of the development sector emerge as a key factor in urban regeneration process.
Critics suggest that the political economy argument about monopolistic rent has never
really been supported by empirical data (Kivell, 1993). Another criticism is that this
approach ignores powerful impetuses for residential location, such as life-cycles, personal
preferences, and taste (Huu Phe & Wakely, 2000).
2.2.2 Residential Mobility Approach
Another non-market approach to residential location, originating in geography and
sociology, is based on residential mobility. This approach includes works that focus on the
household characteristics and on dissatisfaction, or push factors, inducing mobility
9


(Waddell, 2000). Most of the research in this group apply Rossis (1955) lifecycle
hypothesis. Lifecycle theory explains that people adjust their housing consumption to fit
changing household needs with their progression through the cycle of life: for example,
change in household size, age of household members, and marriage status. This is basically
a disequilibrium model of residential mobility, which posits that residential moves occur
when households fall out of equilibrium in their housing consumption (Clark & Moore,
1978; Quigley & Weinberg, 1977).
Coupe and Morgan (1981) extended the lifecycle approach to suggest that changes in
household and personal characteristics are not the only factors that should be considered in
household relocation studies. They noted that housing choices may be affected by
residential history and market factors or forces that are external to the household. Building
further on this concept, Clark and Onakas (1983) was a unique study that attempted to
consider an amalgamation of factors driving residential relocation and mobility processes.
They characterized residential mobility as a combination of an adjustment move (adjusting
to the market), an induced move (changes in household composition and lifecycle), and a
forced move (loss of housing unit or job). Researchers in this group, however, generally
focus on household decisions to move but not on the destinations of the migration.
The life-cycle hypothesis, however, is also not free from criticism. Critics of the life-cycle
model point to the poor articulation of life-cycle concepts in mobility research (W.A.V.
Clark & Dieleman, 1996; Pickles & Davies, 1985; Quigley & Weinberg, 1977) and the
normative nature of the model. There is some indication that the use of head of household
and household variables as proxies for life-cycle stages misrepresents the relationship
between life-cycle change and mobility (Wenning, 1995). Also, research suggests that the
effects of family structure on mobility decisions, at least in terms of the movement from
renting to owning, may be declining (Gyourko & Linneman, 1996).
10


2.3 Phase III (2000s ~): Amenity-Based Approach
2.3.1 Residential Location Decision Factors
A large body of research has accumulated over the last 30 years on location choice by
households (Dieleman, 2001; Luger, 1996; Timmermans et. al., 1994; Michelson, 1977,
1987; Boehm, 1982; De Jong, et al., 1981; Bell, 1968; Foote, et al., 1960). According to
the traditional utility maximization theory, households make residential location decisions
based on the accessibility to their workplaces in order to minimize commute costs.
Frequently, however, households have other priority factors besides job accessibility when
they make residential choices.
M. J. Kim and Morrow-Jones (2005) found, using a survey data from recent home buyers,
that the traditional housing location variable, accessibility to job, was relatively
unimportant compared to other variables, including the following: housing characteristics
(floor plan, quality of construction, and cost), community characteristics (safety of the
neighborhood and good investment or resale value), and school quality (school reputation
and quality of schools academic programs). Zondag and Pieters (2005) also argued that
the role of job accessibility is significant but small compared with the effect of
demographic factors, neighborhood amenities, and dwelling attributes in explaining
residential location choices.
Existing research focused on environmental factors and neighborhood composition in
residential location decision process. Factors in the environmental category include the
presence of environmental amenities, clean air, scenic views, and recreation opportunities
including access to parks and open space preserved natural habitats (Rouwendal & Meijer,
2001; Wales, 1978; Gawande et al., 2001; Homsten & Fredman, 2000; Tyrvainen, 2001;
Tyrvainen & Vaananen, 1998; Colwell et al., 2002; Greenberg & Lewis, 2000). Factors in
neighborhood composition category include housing type and open space, the presence of
nearby retail and service facilities, racial differences, and neighborhood preference (Parkes
et al., 2002; Bowes & Ihlanfeldt, 2001; T. Kim et al., 2005).
11


Another market-based approach, hedonic price model, has been used to estimate the price
of housing or land, based on a list of variables that can be used in the residential location
decision study. In this model, the price of housing or land is comprised not only of
characteristics relating to the structure itself-such as type of house, size, number or rooms
or the existence of central heatingbut also of characteristics determined by location. The
latter includes the classic element of urban economic modes, accessibility. There are
additional location-determined characteristics, such as the character of neighboring houses
and households, localized traffic effects and the quality of the micro environment, and
local public goods such as schools (Cheshire & Sheppard, 1995). Table 2.1 provides
tabulated location-specific variables from existing literature. Commonly applied location-
specific variables include park, distance to CBD, neighborhood characteristics like median
income, and transportation accessibility.
12


Table 2.1 Location variables in select hedonic price models
Cheshire & Sheppard (1995) Tse & Love (2000) J. Wu et al. (2004) Anderson & West (2006) Kong et al. (2007) Cavailhes et al. (2009)
Natural Amenities / Disamenities Open Space Park River Small Park Special Park Park Scenery Forest Green Space Agricultural Land
Lake Golf Course Green Space Steepness
Wetland Lake Bush
Slope River Water
Elevation Flooding
Urban Amenities / Disamenities Shopping Center Sports Facilities Cemetery View CBD Commercial Zone Industrial Zone Traffic Condition CBD Cemetery CBD Plaza Town Hall
Neighborhood Characteristics Main Occupation Housing Density Population Density Land Use Population
Ethnic Group Income Income Income
New Construction Zoning Crime Population under 18 Population over 65 Zoning
School School District Education Environment
Transportation Bus Route Public Transportation Major Road Road/ Railroad
13


Another group of researchers have focused on the location decision factors from housing
producers perspective: developers location decision factors. Early research on developer
behavior was conducted by a University of North Carolina group in the 1960s and early
1970s (Chapin & Weiss, 1962; Donnely et al., 1964; Kaiser, 1966; Weiss et al., 1966;
Kaiser, 1968; Kenney, 1972). Developers locational decision factors from their research
include accessibility to jobs and urban services, physical and socio-economic
characteristics of properties, and zoning protection. However, their research interests
covered the wider issue of the residential development process including the spatial
distribution of subdivisions, the distribution of site characteristics, the categorization of
developer types, and the target price range for the development but less on developers
location decision factors.
During the 1980s there were several approaches to identify developers decision factors for
development location. Baerwald (1981) described developers location decision making
process as a three-staged scanning process by spatial scale and listed site selection criteria
for each stage. His criteria for the first stage includes general development cost variables
and marketability data for a whole metropolitan region. For the second stage, criteria
include accessibility, site characteristics, and governmental policy variables for scanning
relatively large areas. For the third stage, criteria include detailed site characteristics for the
examination of specific sites. Hepner (1983) identified the factors involved in the process
of land use conversion and arranged the factors in a hierarchy according to importance
elicited from the developer survey. His research found prestige, urban amenity, and local
government services as important factors in developers location decisions. More recently,
Smersh et al. (2003) identified spatial factors in developer choices by tracking new housing
units in each section (square mile) in Alachua County, Florida by the year built over a
twenty-year period. Significant variables from their finding included the distance from a
central business district, distance from highway interchanges, housing cost, and the level of
regulation.
14


2.3.2 Consumption-based Neighborhood Development
Among researchers in the political economy school, urban revitalization, redevelopment, or
gentrification is viewed as a consumption practice through which new middle classes seek
to distinguish themselves from the old middle class (Ilkucan & Sandikci, 2005). They
perceive urban redevelopment and gentrification process as a spatial manifestation of the
values of the new middle class. The new middle class, an offspring of post-Fordism, is
a presumed to urge to stay away from the tastes and values associated with old middle and
working classes (Featherstone, 1991; Lury, 1996). New middle class consumption
practices include shopping and attendance in cultural and social amenities, which are
expressive of lifestyle. Boorstin (1973) first introduced the term consumption
community to refer to informal groups expressing shared needs, values, or lifestyles
through distinctive consumption patterns.
Cova (1997) argued that this consumption community still has a linking value, similar
to the sense of community that emerged in the traditional neighborhood, and this shared
consciousness is reflected in consumption practices of the new residents, creating a sense
of what are appropriate consumption practices and what are not. Sack (1988) argued that
consumption molded people's consciousness of place, helped them to construct real places,
connected the realms of nature, social relations and meaning, and revealed how
geographical settings are constitutive of contemporary tensions and paradoxes.
Following this argument, many local governments in the United States exercise
commercial development policies based on the hypothesis that if they prepare new
commercial developments, they will attract new workers to the neighborhoods. This
hypothesis goes further: The new workers will require services and facilities that the city
must provide; the city in turn will look to the developers who are profiting from the
commercial development to share the costs of those services and facilities. However, the
existing research on this hypothetical relationship is limited, and the causal relationship
between the attraction of people and commercial development is still unclear.
15


Analyzing three American cities, Boston, San Francisco, and Santa Monica, which have
adopted downtown planning policies that link large-scale commercial development with
housing, transit, and employment, Keating (1986) concluded that these policies are likely
to have a significant but marginal effect on social problems aggravated by downtown
growth. He also argued that these policies should be tied to comprehensive downtown
plans. Andrew and Merrian (1988) emphasized the importance of the economic feasibility
and the legal foundations for the commercial development-driven urban policies.
2.3.3 Urban Amenity-Based Location Decision Theory
Studies on the effect of urban amenities in attracting people to central locations are very
limited, especially in Northern Europe and North America, where the omission is
encouraged by an ascetic Protestantism, labeling non-work as quasi-sinful (T. N. Clark,
2004). However, there is an expanding discussion on the role of amenities in an urban
context as a pull factor of urban migration.
The early stages of urban influx were analyzed as phenomena associated with the
countercultural lifestyle, including avant-garde artists (P. Jackson, 1985) and gay
communities (Castells, 1983). Kasarda et al. (1997) explained the movement as higher-
income households search for the locations with the lowest tax burden and the greatest
bundle of location amenities. They speculated that non-urban residents who have moved to
central cities are attracted by amenities such as a rich diversity of lifestyles, significant
architectural resources, diverse commercial opportunities, and other entertainment options
such as ethnic restaurants and cultural activities. Suchman (2005) argued that quality of
place has become a paramount concern when employers and people shop for a place to be.
He listed urban amenities such as theaters, museums, sports facilities, restaurants, bars,
parks, and civic spaces, and areas of historic or architectural interest in addition to urban
areas inherent appeal as the symbolic, physical, economic, and cultural centers of
metropolitan spaces. By analyzing the 1970-2000 US Census data, Chen and Rosenthal
16


(2008) found that improving consumer amenities typically attracts retirees, especially
highly educated people.
Urban amenities and urban dwellers are considered to have a reciprocal impact on each
other. Waldfogel (2003) found empirical support in the markets for broadcast radio,
newspapers, and restaurants. For example, when there is a larger local consumer base for a
certain format of radio station, caliber of newspaper, or style of restaurant, more of them
exist in a city. By revealed preference, that greater variety increases city dwellers welfare
because the more options there are for residents that share a particular set of tastes, the
more they consume.
Next, three widely known discussions around the issue of urban amenity will be
introduced.
2.3.3.1 Consumer Cities
Researchers interested in central city economies of post-industrial cities and urban
revitalization projects are increasingly focused on consumption-based industriesincluding
cultural facilities, specialized boutiques and eateries, and retailing and leisure complexes
whose success depends on harnessing the lifestyles of various groups with disposable
income to particular kinds of consumption practices (Hannigan, 1998; Zukin, 1998). David
Brooks in his book Bobos in Paradise (2000) described the changing definition of social
class using the concept of consumption, Karl Marx argued that classes are defined by
their means of production but it could be true that, in the information age at least, classes
define themselves by their means of consumption.
Glaeser et al. (2001) argued that nowadays thriving cities are consumer cities ones that
attract highly educated households through appealing cultural amenities, such as museums,
restaurants and the opera. They demonstrated that between 1977 and 1995, live
performance venues and restaurants are correlated with the future population growth. By
contrast, bowling alleys are correlated with population decline. They argued that high
17


amenity cities have grown faster than low amenity cities. Urban rents have gone up faster
than urban wages, suggesting that the demand for living in cities has risen for reasons
beyond rising wages (Glaeser et al., 2001).
Glaesers four critical urban amenities include the following: (1) the presence of a rich
variety of services and consumer goods: restaurants, theaters, and an attractive mix of
social partners that are hard to transport and are therefore local goods; (2) aesthetics and
physical setting: architectural beauty, and mild weather, which is the single most important
determinant of population or housing price growth at the county level; (3) good public
services including good schools and less crime; and (4) speed including the range of
services (and jobs) available in metropolitan areas as a function of the ease with which
individuals can move around. Finally, he emphasized the role of cities and dense urban
areas as places facilitating humans to interact socially. He provided evidence for the United
States, suggesting that the resurgence of big cities in 1990s is due, in part, to the increased
demand for social interactions and to the reduction in crime, which had made it difficult for
urban residents to enjoy the social amenities.
2.3.3.2 The City as an Entertainment Machine
T. N. Clark (2004) proposed that contemporary cities are entertainment machine[s]. He
argues that workers in the elite sectors of the postindustrial city make the quality of life
demand, and in their consumption practices can experience their own urban location as if
tourists, emphasizing aesthetic concerns. In their argument, these practices impact
considerations about the proper nature of amenities to provide in contemporary cities, and
the city becomes an entertainment machine, leveraging culture to enhance economic well-
being.
Using data from 3,111 US counties, he measured the impact of amenities on population
growth. He used two types of amenities: natural amenities, including six components such
as moderate temperature and water while constructed amenities included the opera, juice
18


bars, museum, and Starbucks. He found that sub-populated groups behave differently.
College graduates are more numerous where there are fewer natural but more constructed
amenities. Seniors are the opposite: their numbers increase with more natural amenities,
but less with constructed amenities. Residents filing high tech patents live in locations with
both natural and constructed amenities.
2.3.3.3 Superstar Cities
Gyourko et al. (2006) identified a handful of metropolitan areas experiencing the
concentration of high-income people and consequential housing price growth that
significantly exceeded the national average, leading to a widening gap across locations in
average house prices.
In their discussion of Superstar Cities, they showed that high-demand cities have income
distributions that are shifted to the right: low-income families can live there only if they
have a very strong preference for the city while high-income families can live there even if
they only modestly prefer it. As the national high-income population grows, the greater
number of high-income families outbid relatively low income families (as well as some
high-income families) who are unwilling or unable to pay a higher premium to live in their
preferred location. They found that such superstar locations experience supra-normal house
price growth and a shift of their income distributions to the right as they experience inflows
of high-income households and outflows of their lowest-income residents. They explained
this phenomenon by an inelastic supply of land in some attractive locations combined with
an increasing number of high-income households nationally. They did not specify the list
of amenities that superstar cities provide, but their findings imply that there must be
something unique and attractive about superstar cities, otherwise potential residents would
turn to cheaper locations and superstar cities would not be able to sustain excess price
growth.
19


2.4 Demographics in Urban Residential Market
2.4.1 Major Instigators of Urban Resurgence
It is widely observed that urban resurgence, especially from young professionals and
entrepreneurs, is related to the emergence of strong creative and technological industries
from the 1970s (Clay, 1979; C. Hamnett, 1991; Lang, 1982; Smith, 1979). During the
economic boom of the late 1990s, high-tech companies led the charge downtown in
order to accommodate new economy workers who appreciate the aesthetics of downtown
work and life. The move-to-the-downtown is a response to the new economys insistence
that ideasand the people who generate themare a companys most valuable commodity
(Suchman, 2002). More evidence by the US Department of Housing and Urban
Development shows that cities have become centers for high-tech job growth (U.S. Dept,
of Housing and Urban Development, 2000). High-tech jobs make up almost 10 percent of
all jobs in central cities according to the report, which is nearly identical to the percentage
found in the suburbs. Furthermore, high-tech job growth in cities increased by 26.7%
between 1992 and 1997, more than three times their overall increase.
With the changing economic structures of the cities, structures of families are also
changing. Evidence shows that household structure is changing on a national level. In
1940, less than eight percent of all households consisted of people living alone; today,
singles make up 25% of American households. Between 1940 and 2000, the number of
unmarried people living together as couples increased by 72%, to 5.47 million. By 2020,
married couples with children are projected to account for only one in five households. The
traditional nuclear family, a working father, stay-at-home mother, and two or more
children, constitutes less than one quarter of all households today. As population and
household types become increasingly diverse and as they demand housing appropriate for
their lifestyles, a broader range of housing types is needed, and urban dwellings become a
popular option among these diverse household types (Lurz, 1999).
20


Existing literature commonly identifies increasing numbers of the seniors and young
professionals as two of the biggest demographic groups who instigate the urban
resurgence. These two groups share a number of characteristics that make them a good
match for downtown living. First, they are not concerned with school quality, something
that often deters families from living in central cities. Second, they often seek low-
maintenance housing that does not require extensive yard work and home repairs. Third,
both groups tend to have the time, money, and inclination to partake in urban amenities
(Sohmer, 1999).
2.4.1.1 Young Professionals
These upper-middle class groups of people are in between their twenties and early forties.
This populationpeople who are delaying marriage or putting off having childrenis
growing. According to the U.S. Census Bureau, approximately 67% of American
households are currently childless (with no children under 18). By 2010, projections show
this figure jumping to 72 percent. At the same time, they are the most mobile people in the
American population. Over the five-year period from 1995 to 2000, some 6.6 million 25 to
34 year-olds moved from one metropolitan area to another (Cortright, 2006).
There are a number of researchers who identify young professionals as generators of urban
resurgence and focus on their motivations for migrating to cities. Examining the inner-city
neighborhoods in Washington, DC, Gale (1979) found that the earliest new settlers in an
inner-city neighborhood are likely to be single males searching for their first homes. The
lack of children tends to make this initial pioneer class of in-migrations oblivious to the
risks associated with the deteriorated buildings and higher crime rates common to many
depressed inner-city neighborhoods. As more and more migrants move into and invest in
the neighborhood, it becomes more stable, and new classes of risk-prone and, eventually,
risk-averse residents begin to occupy renovated dwellings.
21


Using migration data from the 1982 American Housing Survey by the US Census Bureau,
Spain (1989) found that single, childless householders are more likely to choose the city
over the suburbs. Using a mail questionnaire of home buyers in Cincinnati, Varady (1990)
found that college-educated, childless households desiring employment accessibility and
cosmopolitan amenities would tend to locate in the city and that race and income were
important determinants of a households location decision. Kern (1984) described the
characteristics of urban renovators. The typical renovator is wealthy, young, highly
educated, and either single or married with less than two children. Such people are
attracted to central-city locations because by (1) patronize cultural establishments in the
central city, (2) have a relatively high commuting cost, and (3) have relatively low
demands for housing and land.
Analyzing the US Census data, Cortright (2006) found that about a third of the 50 largest
metropolitan areas saw increases in their 25 to 34 year-old population between 1990 and
2000. In contrast, several metropolitan areas saw declines in their 25 to 34 year-old
population of more than 20%. So the data shows that the residential location decisions of
the young people are disproportionately favoring certain metropolitan areas. More
specifically, the growth in the number of college-educated young adults is fueling
prosperity in places like Austin, Charlotte, Atlanta, Portland, Raleigh-Durham, and
Phoenix. The concentration of young people in fewer cities makes those cities even more
attractive places for talented people, creating a powerful gravitational pull for other young
people and forming a positive feedback loop.
Even within metropolitan areas, place is playing an increasingly important role. During the
1990s, the preference of young adults for close-in neighborhoods (within 3 miles of the
regions center) increased sharply. In 1990, 25 to 34 year-olds were about ten percent more
likely than other residents in the metropolitan area to live in the close-in neighborhoods.
By 2000, these young adults were more than 30 percent more likely than other
metropolitan residents to live in these close-in neighborhoods. Strikingly, the relative
attractiveness of central neighborhoods to young adults increased significantly and in every
one of the top 50 metropolitan areas in the 1990s. In 1990, in the aggregate, 25 to 34 year-
22


olds were about 12% more likely than other Americans to live in a close-in neighborhood;
by 2000, they were 33% more likely to live in these close-in neighborhoods (Cortright,
2006).
Traditionally, employment opportunities, family factors and housing are the most
frequently cited reasons for moving for all generations (Schachter, 2004). However, many
young people, particularly the well-educated, seem to be putting a higher priority on
quality of life factors. While economic growth is still an important determinant of
migration, an analysis of movement patterns of young adults showed that well-educated
persons were actually more likely to move to a place with slower job growth than the place
they left almost 60% of the time (Kodrzycki, 2001). This evidence buttresses the
conclusions of Florida (2002), who argued that talented workers are increasingly drawn to
amenities, and also that of Glaeser, who noted that the decisive economic advantage of
cities increasingly derives from the kinds of public and private consumption opportunities
they provide (Glaeser et al., 2001).
Leys study (1986) provided evidence that a density of restaurants and art galleries shows a
strong correlation with urban regenerations, establishing a foundation for broader
arguments made by Florida (2002) and T. N. Clark (2004) that the growth of old, dense
cities has been significantly driven by the preferences of young professionals for diversity
and proximity to amenity-heavy locales. In addition, the Myers and Gearin (2001) study
confirmed the contribution of an already logical decision-rule assigned for college
students, i.e., they prefer to live close to the campus (which is their place of work).
European studies also emphasize the importance of urban amenities. Cheshire (1995)
observed the demographic migration patterns in Europe and found that high quality cities
(for example, cathedral and university towns) have frequently gained population, primarily
amongst young high-income households, attracted by the amenities of the cities.
Overall, the evidence indicates that people in the younger generation have locational
preferences for the core downtown areas where they can enjoy a high quality of life with a
dense and diverse set of urban amenities in addition to the employment and family factors.
23


2.4.1.2 The Seniors
Urban resurgence generator groups in the senior population include empty-nesters, retirees
and healthy, affluent, and socially connected senior people. There are several studies on the
migration patterns by these active senior groups. Valerio (1997) confronted common
stereotypical views about seniors and argued that rather than being poor, sickly, penny-
pinching individuals on the fringe of society, the average senior person is a relatively
healthy, affluent consumer of luxury goods whose in-migration promises net positive
returns to a community and net losses to the community he or she leaves. Ezell (2006),
coined the term ruppies, retired urban professionals: A ruppy volunteers at the theater as
an usher, or serves on a task force for helping the homeless, or works at a marathon
handing out water among other nonpaid voluntary jobs. John Mcllwain at the Urban Land
Institute argued that people now heading into their 60s, as opposed to older retirees, are
more used to, and attracted by, the urban lifestyle and living among people of different
ethnic backgrounds, incomes, and ages (Greene, 2006). Using data from the Puget Sound
Transportation panel, Krizek and Waddell (2002) identified nine classifications of lifestyle
in an effort to address the interaction of daily activity participation and travel patterns with
longer-term household choices of vehicle ownership, residential location, and employment
location. Two lifestyle groups, retirees and transit users, have preferences on less auto-
dependent lifestyles according to their analysis. They estimated an increase in demand for
urban residences with less home maintenance and improved rates of transit and walking by
a growing retiree group comprised of aging of baby boomers.
The demographic trend predicts a huge increase of population in these senior groups in the
near future. From 1990 to 2000, the percentage of the nations households between 55 and
64 slipped slightly from 13.5% to 13.2%, but from 2000 to 2010, the percentage is
projected to jump to 17.4% (Masnick & Di, 2000). According to U.S. Census Bureau
projections, a substantial increase in the number of older people will occur during the 2010
to 2030 period, after the first baby boomers turn 65 in 2011. After 2010 the leading edge of
the boomers will pass age 65, and growth among the senior population will substantially
exceed that of younger adults, an unprecedented social and economic development. This is
24


best seen in the ratio of those aged 65 and older as compared to working-age adults (aged
25 to 64). After decades of relative stability, this ratio will surge 30% in the 2010s and a
further 29% in the 2020s (Myers, 2007), altering the balance to which we have long been
accustomed (Myers & Ryu, 2008).
It is also predicted that a number of empty nesters will continue to grow as baby boomers
age. After their children leave home, empty nesters often change their lifestyle in a way
that favors downtown livingthey relocate to condominiums or townhouses and spend
more of their disposable income on leisure activities. This change in lifestyle may in fact
translate into living in a downtown apartment and patronizing downtown restaurants and
cultural facilities (museums and concert halls). If even a modest portion of empty-nester
households trades suburban homes for urban ones, it is estimated that the market for
downtown housing will boom.
There is a greater number of existing studies on location choice among the seniors focused
on the regional pattern of migration and the place characteristics of the destination
communities when compared to the number of studies on young peoples behaviors. Like
the migration pattern of the younger generations, later-life migration increases the
geographic concentration of the older population. Florida has been the most important
receiving state for the past three decades, followed by Arizona and California. Other major
destinations include Texas, Colorado, the coastal areas of the Southeast and the Pacific
Northwest, the Ozarks (Arkansas and Missouri), and the lake regions of the North Central
states. Although retirement migration from the North to the Sunbelt may have subsided
somewhat during the 1980s, the major regional patterns established during the past few
decades have continued to the present day (Fuguitt, 1993; Golant, 1990; Graff & Wiseman,
1990; Lin, 1999; Longino & Fox, 1995).
Walters (2000) identified three types of migration of seniors: (1) amenity migration: a
search for attractive climate and leisure amenities; (2) assistance migration: a search for
residential and economic dependence to other family members; and (3) migration in
response to severe disability: a relocation that tends to result in institutionalization or other
shared living arrangements. Among the three types of migration, he argues that only
25


amenity migration has a distinctive spatial pattern; the other two types do not. Comparable
to Walters research, De Jong et al. (1995) concluded that the decision to live with family
members is based primarily on factors other than disability. Specifically, they find that
neither initial disability nor increasing disability influences the odds of choosing kinship as
a reason for moving.
Choi (1996) used data from the Longitudinal Study of Aging (LSOA) to assess the motives
of non-institutionalized migrants. Choi found that the primary reasons for moving are the
desire for kinship (20%); financial problems (18%); the poor health of the respondent
(17%); the desire for attractive amenities (13%); and the poor health, death, or
institutionalization of a spouse (11%). (The low proportion citing amenity reasons is likely
to have resulted from the exclusion of retirees younger than age 70.)
Investigating migration to and within the Northeast, Shin (1990) found that seniors in
positive-shift migration streams, those who move from counties of low senior net
migration to counties of high senior net migration, often match the profile of amenity
migrants: They tend to be high-income, married, and to live in high-cost housing.
Conversely, seniors in negative-shift migration streams often exhibit the characteristics
expected of assistance migrants: They tend to be low-income, widowed, and living with
their adult children.
Surveying nearly 600 retired migrants in North Carolina, Haas (1993) and Haas and Serow
(1997) reported that certain origin characteristics are consistently mentioned as
unattractive. These include unpleasant climate (especially among higher-income migrants),
problems of urban areas (especially among younger and rural-destination migrants), high
property taxes, high cost of living (especially among lower-income migrants), and few or
no family residing in area (especially among female migrants). The destination
characteristics most often mentioned as attractive include scenic beauty, four mild
seasons, recreational opportunities, cultural amenities, and warm year-round climate
(especially among older migrants).
Using a logit model, Valerio (1997) identified seven place characteristics as significant
factors in potential senior migration decisions. In order of importance, these characters
26


were a localitys (1) rate of population change in a previous decade (proxy for friends and
family in the area); (2) number of cold days annually; (3) monthly median rent; (4) percent
of population resident in urban areas; (5) miles of coastline and square miles of inland
water; (6) acres of national parks and number of state parks; and (7) residential property
tax. Interestingly, monthly median rent was positively correlated with the senior migration
decision. Valerio treats this as a proxy of urban amenity level.
Existing studies on the issue of senior migration commonly list the existence of family
members and friends, financial issues including living costs, level of natural and urban
amenities (with more emphasis on the natural amenities), and health issues as location
decision factors of this group of people.
By its very nature, there is yet no empirical study on the migration patterns of baby
boomerswhose first cohort will turn 65 in 2011-as a senior group. However, it is not
difficult to estimate that their set of residential preferences is very different from the elder
generations preferences when considering their financial affluence, environmental
concerns, and demand for diverse lifestyles. As Florida (2008) described:
Where boomers flock, bargains disappear, and the neighborhood butcher shop is
replaced by a pan-Asian fusion restaurant and a hardware store gives way to a high-
end remodeling center.
Historically, baby boomers as a collective group have a strong impact on the
neighborhood. During the 1970s and 1980s, their passage into their early twenties spawned
the first stage of inner-city revitalization and gentrification. Their passage into the family
formation and settlement years of their thirties and forties fueled dramatic single-family
home construction and suburban growth. Now, they are entering a new stage of their life,
and predicting their locational behavior becomes an important issue for many scholars as
well as urban planners.
Among a few existing studies on baby boomers residential location behavior, Myers and
Gearins (2001) research presented their preferences for more densely configured housings
in more central locations. By analyzing demographic dynamics and projections, they
estimated that home buyers aged 45 and older who prefer denser, more compact housing
27


alternatives will account for 31% of total homeowner growth during the 2000-10 period,
double the same segments market share in the 1990s.
2.4.2 Social Networks Among Demographic Groups
2.4.2.1 Social Interactions and Conflicts
Mainly due to small size and the resulting high possibility of daily encounters between
members of a residential community, social interactions in a community may generate
externalities that can be significant in a housing location decision process.
There have been several studies on the impact of social interactions on residential mobility.
Brueckner et al. (1996) argued that endogenous and exogenous amenities (which are forms
of interactions) have a central effect on the distribution of rich and poor households
between different locations. Kan (2007), using Panel Study of Income Dynamics (PSID)
data, found that a households possession of local social capital has a negative effect on its
residential mobility, and this negative effect of local social capital may be stronger on
long-distance mobility than on short-distance mobility. By analyzing the existing literature
on the relationship between social network and migration decision, he identified several
potential channels through which the possible social networks in a destination facilitates
migration: (1) members of ones social networks are a source of material aid (e.g.,
accommodation); (2) emotional support (e.g., encouragement), which is important for new
immigrants, can be derived from ones social networks; and (3) one may obtain important
information (e.g., living environment, job opportunities, etc.) from ones social networks.
The general consensus among existing studies is that same ethnic or income group
interaction strengthens the social network and decreases migration decision, but inter-
group interactions weaken the social network and increase the mobility. However, while
there are many existing studies on the social conflicts between existing residents and new
residents in renovated or redeveloped urban areas, academic research on the relationship
28


between new resident groups (i.e., old and young) is scarce mainly because of its novelty.
Instead, an article by Casselman in the Wall Street Journal from May 11, 2007 reported the
instances of generational discordance:
One time I went up there and the twenty-somethings had the whole place
monopolized, she recalls, and I thought, well, not today. Ms. Lammel says she and
some of her cohorts have a strategy for reclaiming the space, at least temporarily:
They're planning a covered-dish pool party. Anyone is welcome, she says in her
pleasant Southern drawl. But well see who shows up.
...and many of the young buyers want their neighbors to be more like them. Ricky
Florita, a 29-year-old mortgage banker in Nashville, says he avoided buying in
Viridian in part because he heard it was attracting an older crowd. Instead, he
signed a contract for a $160,000 condo in Icon, another project by Bristol
Development Group, that will feature a media lounge and a pool plaza with grilling
cabanas when it opens next year. I really think it's going to be a singles scene, Mr.
Florita says. Every time you were in the sales center, you saw really attractive
women buying these condos.
There are also numerous testimonies from developers about the unexpected emergence of
senior people within their development projects that were originally intended to target
younger generations:
My buyers are bit the young urban professionals who started the loft conversion
business in New York and Chicago. What I get are affluent empty-nesters who are
tired of constantly driving into the city for entertainment.
- Lewis Kostiner, Developer (from Lurz, 1999)
Our target market was young military families. That's why we worked very hard to
keep our price point low. We hit the target, but we've also attracted a lot of empty-
nester retirees from out of state, especially in the late stages of the project, as the town
really came into its own. The new town hall was under construction when we started.
The school has now been completely remodeled. There's a new, historic-styled post
office and a new fire station. It's really amazing when you consider that our first two
houses were sold while looking at old, rundown trailers across the street. The retirees
like the idea of being in town, where they can walk most of the places they want to
go.
- Robert Turner, Developer (from Lurz, 1999)
A year before the building is set to open, just 30% of the unitsstudios start at
$885,000have sold and those that have sold haven't necessarily gone to the intended
demographic. Early buyers have been quite a mixed group with a wide range of ages.
It's not as young as I thought.
- Andre Balazs, Developer (from Casselman, 2007)
29


but it's not so easy to control demographics in the open market. Some of the
buildings are drawing unexpected buyers: people old enough to be the parents of the
kids down the hall. And that's leading to territorial conflicts, social snubs-even
planned boardroom coups.
- Lee Schaefer, Developer (from Casselman, 2007)
One of the possible reasons for this new entry of the senior people into a real estate
development that had originally been targeted to younger generations is that these groups
are the two main consumer groups of urban lifestyles and amenities. However, it is also
plausible to assume that developers did not envisage this emerging senior market.
Consequentially, there has been a limited supply of urban residential products targeting this
affluent group. Another plausible explanation is that the seniors prefer to be located near
younger generations. As Florida (2008) described, for many empty-nesters and retirees, a
key factor in their location choiceand in almost everything they dois proximity to their
children and grandchildren. While children may return home after college or when their
parents become ill, an increasing trend is for parents, especially those with means, to
follow their kids. Whatever the reasons are for this generational spatial overlapping, it is
anticipated that instances of generational conflict between these two demographic groups
will become more common as long as they are competing for the limited amount of space
downtown.
2.4.2.2 Neighborhood Effects
Researchers have consistently argued that neighborhoods influence a households
residential location decision. Schellings segregation model (1971) was the first attempt to
measure the impact of the individual neighborhood preferences on the spatial pattern of
demographic distribution. However, Schellings conceptual model has mostly been
explored in hypothetical settings, and there was a limited attempt to apply his model to a
practical application. More recently and more empirically, Cutler and Glaeser (1997)
tracked racial segregation in US metropolitan areas from the late 19th century to 1990s.
Based on the dissimilarity index, segregation increased continuously until the 1970s but
30


has since declined. The decline is primarily attributable to a movement of black households
outwards from city centers into suburbs that were formerly white-dominated. By contrast,
Abramson et al. (1994) found that, between 1970 and 1990, income segregation increased
in US metropolitan areas. The dissimilarity index rose by 11% for poor households and the
isolation index increased by nine percent, van Ham and Feijten (2008) explored the
influence of the neighborhood characteristics (percentage of rented dwellings, low-income
households, and ethnic minorities in the neighborhood) on the different categories of
residents wish to leave their neighborhood. Their main result showed that, with an
increasing percentage of people from an ethnic minority in the neighborhood, more people
have the wish to leave the neighborhood. However, this is to a lesser extent the case for
members of ethnic minorities themselves.
At the same time, diverse measures of segregation have been developed: evenness
(dissimilarity), exposure (isolation), concentration (the amount of physical space occupied
by the minority group), clustering (the extent to which minority neighborhoods are
contiguous), and centralization (proximity to the city centre) (see Cutler et al., 1999, for a
more detailed discussion).
Age segregation of residential neighborhoods received some research attention in the
1970s and 1980s (Chevan, 1982; Cowgill, 1978; Pampel & Choldin, 1978; Ward et al.,
1985) but has largely been ignored since then. A possible reason for this peak may have
been the events of the late 1960s and early 1970s: a period with much discussion of
generation gaps and slogans such as Never trust anyone over 30. It was also a time of
increased sensitivity towards isms that lead to separation and exclusion, starting with the
awareness of racism and a call for civil rights (Hagestad & Uhlenberg, 2005). It is
puzzling, however, that the literature on age segregation never combined an interest in the
potential conflict between the young and old. Indeed, publications related to the separation
of young people have a different perspective than those related to older people. The
literature on age segregation of children and youth takes a social problems perspective,
emphasizing the costs of separation. However, the literature on segregation of older people
tends to emphasize the benefits of separation, particularly residential. In the first case, one
31


finds discussions of juvenile delinquency, troubled families, and children with behavior
problems. In the second case, security, simplified service delivery, and easy access to peers
are stressed.
The studies find moderate levels of segregation between the older and younger
populations. They also report an aging of the suburbs, which might have increased the
potential for young people to interact with older individuals (Fitzpatrick & Logan, 1985).
But none of this research was able to demonstrate that residential age segregation made
much difference in the lives of old and young. Since 1990, a great deal of attention has
been given to the effects of neighborhood on the well-being of adolescents (e.g., Sampson
et al., 2002). The concept of social capital, which emphasizes the critical role of social ties,
is suggested as one of the primary mechanisms linking neighborhoods to individual
outcomes. However, neighborhood age composition has not been one of the central
characteristics included in these studies. Thus, little is known about current levels of age
segregation in neighborhoods, or the implications of residential segregation for cross-age
perceptions and interactions.
Although there have been several attempts to measure the neighborhood effects on a
households residential location decision, relatively little is known about the nature of the
neighborhood effects. There are several possible explanations. First, the neighborhood
effects that may result in a households migration decision can be generated from a series
of interactions with neighbors of different cultures. Second, they also can be originated
from the existing prejudice against different demographic groups and their cultures. More
plausibly, they possibly came from both: negative interactions that reinforce prejudice.
If negative interactions are the main reasons for migration decisions, tolerance will be one
of the key factors in controlling mobility: a community with highly tolerant people will
have low mobility rate and represent a high level of diversity. Whereas, if the prejudice
among people is the main reason, staying time in a community will be one of the key
factors: accumulation of positive interactions with different cultures (or people in different
cultures) among neighbors will result in a low mobility rate and represent a high level of
diversity. As Allport (1954) pointed out in his classic study, a key weapon against
32


stereotypes and prejudices is intergroup contact, which allows individuals the opportunity
to challenge homogenized categories and see beyond stigmatized characteristics to other
relevant qualities of persons in a pivotal category.
2.5 The Role of the Local Governments
Existing research on the effects of the role of the local policies on the residential location
decision shows ambiguous results. Friedman (1981) employed maximum-likelihood
estimation of a multinomial logit to examine the effects of local public services and other
community attributes on residential location decisions by families. The model was
estimated for six subsamples differing by household size, income, and age of head of
household. For most of the subsamples that were examined, the conclusion was that local
public services and other community characteristics play only minor roles in determining
residential location choice.
Uyar and Brown (2005) employed McFadden's (1978) discrete choice model to test for the
significance of dwelling-specific local taxes and public services on household location
decisions within a single taxing jurisdiction. Their findings indicated that such variables
are significant determinants of location decisions even within a single taxing jurisdiction,
and should not be assumed away. Bayoh et al. (2006) employed a hybrid conditional logit
choice model using data on the characteristics and destination of homeowners who
engaged in intra-metropolitan moves among 17 school districts within the Columbus, Ohio.
The model was used to test the relative influence of local fiscal and public goods versus
household-level characteristics in determining household location choices across central
city and suburban school districts. Results provided evidence of both a "natural evolution"
of households to the suburbs, due to job location, residential filtering, and household
income and lifecycle effects, and flight from blight, due to lower school quality, higher
crime levels, and lower average income levels in the city. In comparing the magnitudes of
these variables, they found that school quality exerted the strongest influence: a one-
33


percent increase in the school quality of the city district increased the probability of
choosing a city residence by 3.7%. In contrast, the effects of household income and other
individual characteristics were relatively modest. The findings provided support for a
flight from blight suburbanization process that was dominated by differences in
neighborhood quality between the city and suburbs.
Public efforts to revitalize declining central cities have a relatively long history. The U.S.
Federal Urban Renewal (then called Urban Redevelopment) Program was launched under
Title I of the Housing Act of 1947. Although the program was intended to [provide] more
and better housing through the spot removal of residential slums (Keyes, 1969), the next
two decades witnessed the replacement of people and their homes with comprehensively
planned structures (mostly office buildings, commercial complexes, and luxury housing)
and, all too often, empty tracts (Y. Zhang & Fang, 2004). Moreover, urban renewal
programs were criticized for simply pushing slum dwellers (mostly African-Americans and
other minorities) to other parts of the city or to the suburbs, thereby exacerbating the ills
the program sought to solve (Gans, 1968).
More recently, however, local municipalities have adapted innovative inner city
revitalization programs to achieve several goals: raising property values and taxes,
increasing sales tax revenues, stemming crime, creating construction and permanent jobs,
improving civic image, and attracting tourists. Their programs include cash contributions,
financing aid, tax abatements, zoning and building code variances, new or improved
infrastructure, implementation of growth boundaries, and the creation of dedicated taxes
for physical improvements. Additionally, some cities use neighborhood conservation
programs to support existing communities and to maintain their desirability. Some cities
partner with nonprofit community development corporations to kick-start housing
production in lagging markets. Some business improvement districts (BIDs) actively
promote and facilitate housing redevelopment and new construction (Haughey, 2002).
The impact of governmental programs is difficult to generalize because successful cases
are location-specific. While some planning policies and regulations are generating the
resurgence of the downtown (i.e., Seattle and Portlands state-mandated growth controls
34


which promoted development inside the urban growth boundaries), the resurgence of other
downtown areas in cities like New York and Chicago are mainly market-driven.
Consequently, the studies on the effectiveness of the various governmental programs are
scarce.
Mainly due to the absence of research on the effectiveness of the governmental programs
to generate inner city revitalization, many local governments invest their budget into the
symbolic mega-construction projects in downtowns hoping these projects will attract
people into the areas. However, there are increasing numbers of discussions that the
function of cities and the characteristics of citizens have changed. Cities reason for being
and their residents reason for living within themmay be the new lifestyle that only dense
and diverse urban environment can provide. As Florida (2002, 2005a, 2005b, 2008) argued
in a series of his creative class books, the physical attractions that most cities focus on
buildingsports stadiums, freeways, urban malls and tourism and entertainment districts
that resemble theme parksare probably be irrelevant, insufficient or actually unattractive
to many urban dwellers. Instead, what they look for in communities may be high-quality
amenities and experiences, an openness to diversity of all kinds, and above all else the
opportunity to validate their identities (Florida, 2008). More research on the role of local
government that controls the level of urban amenity by regulating their types, amounts, and
distributions are required.
35


3. THEORY AND APPLICATIONS OF URBAN MODELING
An urban model is an abstracted representation of a real-urban system. In general,
models act as a vehicle to enable experimentation with theory in a predictive sense and to
enhance understanding which may be prior to predictions of situations as yet unrealized,
for example, in the future (Batty, 2009). A more conventional definition of urban models is
the following:
Representations of functions and processes which generate urban structure in terms of
land use, population, employment and transportation, usually embodies in computer
programs that enable location theories to be tested against data and predictions of
future locational patterns to be generated
- International Encyclopedia of Human Geography (Thrift & Kitchin (eds.), 2009)
In constructing urban models (act of urban modeling), the assumed real urban is broken
down into a form amenable to analysis, by representing with only those variables that truly
affect the behavior of the urban system and by clarifying the relationships between those
variables (Taha, 2003).
Simulation is an act of running constructed models. According to Decker (1993), An
(urban) simulation ... can serve as an accessible surrogate for the citys complex systems,
extensive spatial structure, or environmental influences. The ability to clearly see and
understand alternative development scenarios has always been a goal of planning
(Simpson, 2001). The complex nature of cities, however, with dynamically changing
parameters and large numbers of independent/dependent actors make it difficult to estimate
the results of planning policies, and there have been requests to adapt a simulation
approach to urban system for a long period of time.
In this chapter, a review of the development of computer-based modeling and simulation in
urban planning field is provided. Specific interests are paid to the applications of agent-
based approaches in residential location decision models.
36


3.1 Historical Development
3.1.1 Emergence of Urban Models
In terms of cities, the kinds of urban theory that are basic to the development of computer
models are those that are traditionally called location theories: Theories that propose
mechanisms that enable industries, services and households to locate in space with
economic constraints of income and profitability. In turn, these economies are conditioned
primarily by distance between land-uses associated with these activities, which depend
upon a range of market conditions essentially underpinned by trade. Thus distance and
movement are central to such theory (Batty, 2009).
It has been over 50 years since computer models were first developed in the urban domain
(B. Harris, 1965). The first generation of urban models treated the urban system as a static
entity whose land uses and activities were to be simulated as a cross-section in time and
whose dynamics were largely regarded as self-equilibrating (Batty, 2009). Early urban
modelers include Isard (1960) and Alonso (1964), who presented models based on spatial
interaction ideas from social physics and macro-economic models such as input-output
analysis. These urban models generally dealt with land use and land price, but many urban
components such as population, job, services, and transportation were beyond these
models framework (Benenson & Torrens, 2004).
It was Lowry (1964) who first attempted to relate such urban components into the model
(Lowry, 1964). In his model, the urban system was structured in a simple way, with three
activity sectors: (1) a basic sector, including industrial, business, and administrative
activities, whose clients are mostly non-local; (2) a retail sector, dealing with the local
population; and (3) a household sector. The basic assumption of the model was based on
the gravity model, which is a simple concept: If the distance between two points equals d,
then any interaction between the objects located at these points is inversely proportional to
some power of d. Even though the implementation of the model to the City of Pittsburgh
37


was quite successful, Lowrys model was static, which means that cities in this model can
be seen as largely unchanging; this assumption is unrealistic in many ways.
The first dynamic urban model was developed by Forrester (1969). He saw positive
feedbacks as the main source of complex and counterintuitive behavior of natural systems
in general and urban flavors in particular. He proposed that one positive feedback
dominates the system for a time and then this dominance is shifted to another feedback,
responsible for the other part of the system. The behavior of the system changes so much
with such a shift that the two regimes seems unrelated. At the same time, while one of the
feedbacks dominates, the system maintains resistance to the other, thus marginal, changes.
Forresters work made some meaningful contribution to the field of urban modeling and
simulation such as use of computer as the main tool for investigating the behavior of
complex urban system; emphasis of model structure and relationship between seemingly
independent urban phenomena; incorporate as few variables as possible into a model to
reduce the volume of possibility. At the same time, his works, however, generated a storm
of criticism. The major criticism of Forresters work is on the complete ignorance of
geographic theory and non-spatial aspect of his model (K. Chen, 1972). Explicit
incorporation of geographic knowledge into the models was soon developed, but the
negative attitude remained and could, perhaps, be one of the reasons behind the general
retreat of geographic and urban science from quantitative methods and modeling during the
1970s and 1980s (Benenson & Torrens, 2004).
Limited numbers of urban modeling early research were developed within the integrated
regional framework that merges the models of Lowry and Forrester, and they followed the
tradition of mathematical ecology and economy (Day, 1982; Dendrinos & Mullally, 1985).
Other studies include the studies of competition between two social groups for space (W.
Zhang, 1989) or between economic sectors (Dendrinos & Sonis, 1990; W. Zhang, 1993;
W. Zhang, 1994), and explanation of emerging urban hierarchy (Rosser, 1994).
One meaningful approach of that time by a group of modelers at the National Bureau of
Economic Research was The Detroit Prototype of the NBER Urban Simulation Model,
which focused on the housing location decision based on access to employment, a
38


disaggregated equation for different household types and different housing types and
spatially located zones (Ingram et al., 1972). Even though this model was not free from the
criticism, modeling approaches of this model were based on individual, market processes,
and developer decision provided theoretical and methodological background for the arrival
of agent-based models.
3.1.2 Large Scale Model Critiques
Traditional urban models, developed in the style of the spatial interaction model, were
pioneered at a time in which the field of urban simulation was radically different from
current situation. Computing power of that time was relatively less accessible and
sophisticated than it is today and detailed data sets were not widely available (Torrens,
2001). Those traditional urban models were criticized heavily (Lee, 1973; 1994; Sayer,
1979), and one of the most famous criticisms on urban comprehensive modeling came
from Lee in 1973. In his Requiem for Large-Scale Models Lee challenged all three
widely known urban models at that time: the Forrester model, PLUM (an evolutionary of
the Lowry model), and NBER and listed seven sins of large-scale models.
His seven sins of the large-scale models are: (i) hypercomprehensiveness, the models
attempt to design too of a complex system in a single shot and attempt to serve too many
purposes at the same time; (ii) grossness, coarse level of aggregated data to be used by
most policy makers; (iii) hungriness, the models tremendous requirement of data; (iv)
wrongheadedness, the use of too many variables that even the model builder might not
perceive and the resulting deviation between claimed model behavior and the equations or
statements that actually govern model behavior; (v) complicatedness, the difficulty of
understanding black box model behavior mainly due to multicollinearity and
misspecification of variables; (vi) mechanicalness, intrinsic procedural and numerical
errors in computing; and (vii) expensiveness, the high price of data collection and
parameter estimation.
39


Even though the development of micro-computation technology and Geographic
Information System (GIS) negated some of his criticisms and enabled a cost-effective
development of urban models, Lees criticisms of the large-scale model are still valid and
highly relevant today, particularly when contrasted with newer models currently being
developed in academic contexts: their centralized approach, a poor treatment of dynamics,
weak attention to detail, shortcomings in usability, reduced flexibility, and lack of realism
(Torrens, 2001). In the next section, recent development of modeling and simulation
techniques that offer the potential for improving the usability of traditional models will be
explored.
3.1.3 Land Use-Transportation Models
Since the early efforts to construct operational urban models were mostly based on social
physical paradigm (the gravity model), there have been several attempts to embrace
theories from a wide range of urban sciences. Transportation engineers, urban economists,
social scientists, and geographers tried to incorporate land use-transportation interactions
(trip and location decisions codetermine each other), economic forces of city growth (cities
are systems of market in which households try to match their space needs and location
preferences with their budget restrictions), and social theories of urban development (the
spatial development of cities is the result of individual or collective appropriation of space)
into the urban models.
Consequently, many recently developed urban models are built around the theme of land-
use transportation feedback cycle. They include the following: the California Urban
Futures (CUF) model developed at the University of California at Berkeley (Landis & M.
Zhang, 1998); DELTA, the land use/economic modeling package by Simmonds
Consultancy (Simmonds, 1999, 2001); the Integrated Land Use, Transportation,
Environment (ILUTE) modeling system developed at several Canadian universities (Miller
& Salvini, 2001); the integrated modeling package developed by Echenique and Partners
40


(MEPLAN) (Williams 1994; Hunt & Simmonds, 1993); the land use transportation model
developed in the Oregon Transportation and Land Use Model Integration Program
(TLUMIP) (ODOT, 2001); the transportation and land use model (TRANUS) developed
by de la Barra (1989); and UrbanSimthe microsimulation model of location choice of
households and firms at the University of California Berkeley (Waddell, 1998, 2001).
Most of the models introduced here simulate beyond just land use and human
transportation interactions: They forecast housing stock, nonresidential building stock, and
goods transportation as well. All of the models here assume that the transportation is
always in equilibrium, i.e., that travel flows reflect travel times and costs on the network.
However, some models (CUF, ILUTE, and UrbanSim) assume that the land use system is
dynamic (or quasi-dynamic as they work with discrete time periods) in that they explicitly
model the adjustment processes over time. These models are based on the assumption that
some adjustment processes are faster than others and that the differences in speed are so
large that urban systems are normally in disequilibrium (Wagener, 2004). Finally, some
models (ILUTE, TLUMIP, and UrbanSim) apply microsimulation of land use model based
on individual household or firm level behavior.
3.1.4 Planning Support Systems
With the development of information science with data and Electronic Data Processing
(EDP) in the 1960s, information and Management Information Systems (MIS) in the
1970s, and knowledge and Decision Support Systems (DSS) in the 1980s, Planning
Support System (PSS) from the 1990s became a popular topic in urban planning field
(Klosterman, 2001). There have been several meaningful modeling efforts in developing
computer-assisted planning tools. These efforts include the following: METROPILUS, a
land use model that has roots in the early model of Lowry developed by Putman (1995);
INDEX, a GIS-based PSS that estimates the potential impact of community land use and
design decisions developed by Criterion Planners (Allen, 2000); What if?, a scenario-
41


based, policy-oriented PSS developed by Klosterman (2007); and CommunityVIZ, a GIS-
based PSS for community planning and design applications developed by the Orton Family
Foundation (Kwartler, 1998). Some urban models built around the concept of land-use
transportation interactions are also included in this PSS category: CUF (vector-based
model), CUF II (development of CUF but has a raster data structure), and CURBA
(focuses on the interactions between land use change and habitat loss), TRANUS and
UrbanSim.
While there is a rapid and increasing number of the development of PSS tools, there are a
limited number of successfully applied PSSs in real world planning practices. They are, to
practitioners in general, too rigid, too theoretical, too costly, un-dynamic, unrealistic, and
unreliable. More efforts on the development of urban modeling tool that help planners in
anticipatory decision making situations, i.e., planning situations, are required.
42


Table 3.1 Comparison of the urban models*
CUF II CURBRA INDEX
Model Type Land use change Urban growth GIS, urban impact
Thematic Scope Urban simulation Urban growth, environmental and ecological quality Land use, transportation, housing, employment, natural environment
Spatial Resolution One-hectare One-hectare User defined
Temporal Resolution Custom Custom User defined
User Non-technical planning participants Land use planners, policy makers, environmentalists Non-technical planning participants
Operational Methods Logit, regression Logit, regression Causal inference, correlation, linear programming, network analysis, time-series
Urban Land Use Categories Single, multi family, Commercial, Industrial, Development No category User defined
Non-Urban Land Use Categories Agricultural, Forest, Wetland, Water, Preservation, Park Agricultural, Forest, Wetland, Water, Preservation, Park User defined
Independent Variables on Land Use Patterns Transportation, zoning, master plan, impact fee, sewer and water fee Transportation, zoning, master plan Zoning, master plan
43


Table 3.1 (Cont.)
What if? UrbanSim MEPLAN
Model Type GIS GIS, urban economic/land use market, logit, hedonic Travel demand, urban economic/land use market, hedonic
Thematic Scope Land use evaluation and change analysis Land use, transportation, economics, environmental impacts Spatial economic-based I/O
Spatial Resolution User defined User defined User defined
Temporal Resolution User defined User defined User defined
User Non-technical planning participants Land use and transportation planners, community participants Planners, transportation engineers, economists
Operational Methods GIS Expert system, logit, regression, Monte Carlo Logit, network analysis
Urban Land Use Categories User defined User defined User defined
Non-Urban Land Use Categories User defined User defined User defined
Independent Variables on Land Use Patterns Transportation, zoning, master plan Transportation, zoning, master plan, fiscal policy Transportation, zoning, master plan, fiscal policy, impact fee, tax, sewer and water fee, subsidy, road toll, parking fee, registration fee
* Rearranged from EPA (2000), Klosterman (2001), and Sietchiping (2004)
44


3.2 Evolution of Urban Modeling: CA/ABM
3.2.1 Cellular Automata (CA)
Recent research efforts toward a new urban model that are detailed but flexible,
conceptually understandable, cost-effective, dynamic, realistic and reliable include the
adaption of cellular automata (CA) and agent-based modeling (ABM) approaches. At the
most rudimentary level, a cellular automaton is an array or lattice of regular spaces or cells.
At any given time, a particular cell is in one of a finite number of allowed states, and that
state will change according to the states of neighboring cells in the lattice according to a
uniformly applied set of transition rules. Cells alter their states iteratively and
synchronously through the repeated application of these rules. A CA is thus composed of
four principle elements: a lattice, a set of allowed states, neighborhoods defined by the
lattice, and transition rules. In addition a fifth, temporal component can be considered
(Torrens & O'Sullivan, 2001). From the International Encyclopedia of Human Geography
(2009), CA model in urban simulation context is defined as:
A class of spatially disaggregate models, often pictured as being formed on a 2-
dimensional lattice of cells, where each cell represents a land use and where
embodying processes of change in the cellular state are determined in the local
neighbourhood of any and every cell. Such models can be seen as simplifications of
agent-based models where the focus is on emergent spatial patterns through time.
CA models have many advantages for modeling urban phenomena, including their
decentralized approach, the link they provide to complexity theory, the connection of form
with function and pattern with process, the relative ease with which model results can be
visualized, their flexibility, their dynamic approach, and also their affinities with
geographic information systems and remotely sensed data (Torrens, 2000). Perhaps the
most significant of their qualities, however, is their relative simplicity. By mimicking how
macro-scale urban structures may emerge from the myriad interactions of simple elements,
CA offers a framework for the exploration of complex adaptive systems. However, with
CA, this principal innovative feature is also one of its greatest weaknesses. CA models are
45


constrained by their simplicity, and their ability to represent real-world phenomena is often
diluted by their abstract characteristics (Torrens & O'Sullivan, 2001).
3.2.2 Agent-Based Model (ABM)
The target of urban modeling and simulation, some real world phenomenon which the
researchers are interested in, is always a dynamic entity, changing over time and reacting
to its environment. With the complexity and dynamic features of urban phenomenon,
especially when the relationship between the target and variables of the model is nonlinear,
traditional analytic reasoning using logic or by using mathematics can be very difficult or
impossible. In these cases, disaggregated and dynamic simulation is often the only way
(Gilbert & Troitzsch, 1999).
Early efforts to construct urban models built around representing the actions and behavior
of individual agents located in space include Chapin & Weiss (1968) and Ingram et al.,
(1972) but it was since mid-1990s, an agent-based simulation approach began to emerge
(Benenson et al., 2002), and this was mainly because of the micro-computer capacity
development.
Agent-based models (ABMs) are computer representations of systems that are comprised
of multiple, interacting actors (Brown, 2005). In an urban planning context, agents include
households, developers, enterprises, and/or planners. By simulating the individual actions
of diverse actors, and measuring the resulting system behavior and outcomes over time,
ABMs provide useful tools for studying urban phenomena that operate at multiple scales
and organizational levels and their effects.
What distinguishes ABM from CA is that, whereas CA usually has a fixed interaction
topology (i. e., which neighbors a cell interacts with is fixed by the cellular geometry),
interactions in the ABM can be dynamically changed as the model runs, because they are
defined at the agent level, rather than in terms of the partitioning of space (Brown, 2005).
46


Several characteristics define ABM: autonomous agents, communication and interaction
between agents, or cooperation and competition between agents, and the impact of agent
decision on the environment (Parker et al., 2003). Finally, ABMs for urban research are
nearly always spatially explicit, which means that the agents and/or their actions are
referenced to particular locations on the urban area. For this reason, many ABMs have
either direct or indirect interaction with GIS (Brown, 2005).
3.2.3 CA/ABM Approach
There have been several attempts to overcome the limits of CA models in which the
immobility of the cells is a prerequisite and to apply the concept of individualism of
ABMs. Benenson (1999) and Portugali (2000) proposed the Entity-Based (EB) modeling
approach. The EB model considers the city as consisting of two interacting layers. The first
represents immobile urban components and is described by a CA-type model, whose
elementary units are infrastructure elements that can be treated as innately homogeneous:
land parcels, houses, street segments. The second layer of the EB model represents the
instantaneous spatial distribution of mobile urban decision makers, the dynamics of which
is described by means of an agent-based (AB) model. The benefit of separating between
the infrastructure and population layers is that this approach allows researchers to account
for the different rates of growth of the two layers (i.e., fast growing population and slow
development of infrastructure) (Benenson et al., 2002).
Torrens and Benenson (2005) extended the CA concept to incorporate the ABM concept,
and proposed Geographic Automata Systems (GAS). In this framework, geographic
phenomena as a whole are considered as the outcomes of the collective dynamics of
multiple animate and inanimate geographic automata.
In CA framework, an automation, A, belonging to a CA lattice is expressed as follows
(Torrens & Benenson, 2005):
A ~ (S, T, N) (3.1)
47


where S denotes a set of states and N denotes automata neighboring A, and defines the set
of cells for drawing input information which is necessary for the application of transition
rules T.
Whereas, in the GAS framework, a geographic automation system, G, consists of seven
components:
G~(K;S,Ts;L,Ml;N,Rn) (3.2)
where K denotes a set of types of automata featured in the GAS, L denotes the geo-
referencing conventions that dictate the location of automata in the system, ML denotes the
movement rules for automata, governing changes in their location in time, and RN denotes
the rules that govern changes of automata relations to the other automata. The minimalistic
approach of GAS in which most urban phenomena are abstracted in a simple simulation
framework allows for a degree of standardization between automata models and other
systems, including GIS (Torrens & Benenson, 2005).
3.3 CA/ABM applications
CA/ABM has been applied to various areas in urban planning field, such as land-use/cover
change (Laine & Busemeyer, 2004; D. Parker et al., 2001; Waddell, 2001), urban growth
(Fernandez et al., 2005; Jantz et al., 2004), urban ecology (Gimblett, 2002; Becu et al.,
2003; Brown et al., 2004), pedestrian movement and evacuation (Blue & Adler, 2001;
Haklay et al., 2001; Kerridge et al., 2001; Zacharias et al., 2005), transportation (Kii &
Doi, 2005; Teodorovic, 2003), housing and industrial location (Arentze & Timmermans,
2007; Irwin & Bockstael, 2002; Mulbrandon, 2007; Torrens, 2001), and segregation
(Benenson et al., 2002; Laurie & Jaggi, 2003; J. Zhang, 2004; O'Sullivan & Macgill, 2005;
Chen et al., 2005; Bruch & Mare, 2006; Benard & Wilier, 2007). The next section focuses
specifically on the application of CA-ABM to the study of residential location.
48


3.3.1 Residential Location Modeling
A majority of existing residential location studies adapting the ABM approach focused on
exurban population migration and location decision behavior (Brown et al., 2002;
Fernandez et al., 2005; Rand et al., 2002; Yin, 2004; Yin & Muller, 2007; Muller et al.,
2008). Less attention was paid to urban population behavior mainly because of the
extensive impacts of suburban migration and land development at the urban-rural fringe.
Also, more heterogeneous characteristics of urban housing seekers when compared to
exurban housing seekers made the urban residential location modeling process difficult.
Research on urban migration and household location decision paid specific attention to the
dynamics of urban real estate market. J. Jackson et al. (2008) explored the process of
gentrification and land value change in Boston, Massachusetts using four classes of agents
(professionals, college students, non-professional, and the seniors) and set of simple
decision rules, which are based on the social class and economic ability of the agent. In
their model, agent decision is influenced by memory of favorite location and by
communication between friends of similar class. However, an agents decision to move in
their model is only driven by an economic imperative: an agent decides to look for a new
location when it can no longer afford its current place, but no other household mobility
factors are considered.
Using a virtual monocentric urban real estate market, Filatova et al. (2008) adapted agents
with heterogeneous preferences for location. Their model output was analyzed using a
series of macro-scale economic and landscape pattern measures, including land rent
gradients estimated using simple regression. They demonstrated that heterogeneity in
preference for proximity alone is sufficient to generate urban expansion and that
information on agent heterogeneity is needed to fully explain land rent variation over
space.
Nara and Torrens (2005) used four types of fixed agents (market, sub-area, property, and
fixed land) and one type of mobile agent (resident) in simulating gentrification process in
Salt Lake City, Utah. Location decision and housing choice of resident agent is based on
49


four housing characteristics (property value, housing type, house size, and accessibility)
and two neighborhood characteristics (economic status and ethnicity). Their model
demonstrated the gentrification process when gentrifiers (new residents with higher
economic status) and gentrifiable properties (new properties with higher value) introduced
to the study area.
In ABM mobility and location decision studies, the concept of lifecycle is widely adapted
(Waddell, 1998, 2000). Torrens (2001) used three lifecycle stages: young, middle, and
senior in constructing a residential location model. Households in different lifecycle stages
have different preferences for housing types: young and middle agents without children
prefer apartment and middle agents with children and senior agents prefer houses. Because
the model was developed as a prototype residential location model based on simple
unrealistic assumption, the practical usability is limited. Fontaine and Rounsevell (2009)
presented a framework to model future residential demand for housing in a polycentric
region based on the household life cycle stages. The model was calibrated for the region of
East Anglia in the United Kingdom using a semi-quantitative procedure. Their results
demonstrated non-uniform, spatial patterns of urban sprawl with some locations
experiencing greater urban development pressure than others. The proximity to the
principal city was identified as the main demand factor for residential housing. By
explicitly modeling agent behavior and interactions, they represented the response and
adaptation strategies of a population to changing circumstances.
Devisch et al. (2009) tested the concept of the bounded rationality. They presented an
agent-based model developed to simulate residential choice behavior in a non-stationary
housing market. The model was built around the assumption that agents at different
lifecycle stages have incomplete and imperfect knowledge and thus have to base their
decisions on beliefs. They illustrated how the agents deal with the uncertainty inherent in
these beliefs, both at the level of a single agent, deciding among a set of successive actions,
and at the level of a group of agents, negotiating over the price of a house. Torrens (2007)
applied the GAS (geographic automata systems) approach to explore residential mobility
and location decision process of households. In his model, residential mobility is treated as
50


a two-stage process. First, households decide whether to initiate a relocation event or not,
looking both to internal and to external stressors in formulating this decision. If they decide
to engage in a housing search, they do so based on preferences that relate to their internal
characteristics and those of the larger community and submarket in which the search is
focused. The applicability of the methodology used in the model is demonstrated through
the development of a rich model of residential mobility, in which individual households
interact with other households and real-estate infrastructure, dynamically in space and
time, to form synthetic communities and artificial property submarkets.
Li and Liu (2007) used multicriteria evaluation techniques to determine some of the
parameters for the agent-based residential development model in Guangzhou, China.
Empirical data from GIS were used to define the agent's properties, and a sensitivity
analysis was also carried out to assess the influences of parameters on simulation
outcomes. B. M. Wu et al. (2008) presented a dynamic simulation model that projected the
future population of the city of Leeds as a basis for policy analysis and scenario planning.
They argued that microsimulation modeling is not entirely effective in the representation of
student populations. They suggest that agent-based modeling and microsimulation are
powerful as complementary technologies for individual-based modeling.
51


Table 3.2 Comparison of select agent-based residential location models
J. Jackson et al (2008) Filatova et al (2008) Fontaine & Rounsevell (2009) Nara & Torrens (2005) B. M. Wu et al (2008)
Thematic Scope Gentrification Land market Urbanization Gentrification Spatial population distribution
Location Boston, MA Virtual East Anglia, UK Salt Lake City, UT Leeds, UK
Agents Professionals Buyer Households at different lifecycle stages Resident First year undergraduates
College students Seller Other undergraduates
Non- professionals Master students
The seniors Doctoral students
Accessibility Variables College campus Commercial district CBD Environmental amenities Road network Train station Downtown Highway University
Key service area Mall
Market town Grocery
Large city
Coastline
Rivers and
waterbody
Green
recreational
area
Flooding zone
Coastal cliff
erosion zone
Other
Variables
Affordability Rent
Neighborhood
Recommendation
Location Memory
Property value Same cohort
Property size
Housing type
Neighborhood
economic
status
Neighborhood
ethnic profile
52


3.3.2 Segregation Modeling
Residential segregation based on ethnicity or economic status or religious affiliation is one
of the consistently explored areas by ABM modelers from various fields including
sociology, politics, economics, and urban planning. One of the earliest and best-known
models of residential segregation was developed by Schelling (1971, 1978). In the
Schelling model, agents inhabit a checkerboard (sixty-four squares in eight rows and
eight columns). The agents are divided into two classes that are meant to represent any
binary social division that could affect the distribution of agents in space (e.g., men and
women, blacks and whites, French-speaking and English-speaking). Agents are initially
distributed randomly across the lattice. From that point on, agents observe their
neighborhood (3X3), and change locations if the number of agents of the other type
exceeds some threshold. Schelling showed that populations could reach high levels of
segregation even when agents are willing to remain in neighborhoods in which up to two-
thirds of their neighbors are members of the other group. His model succinctly represents
what most agent-based models try to find out: how peoples interactions could lead to
results that were neither intended nor expected.
Following Schellings seminal paper, a number of researchers extended the standard
Schelling model to investigate a variety of questions. Some of recent variations of the
Schelling models are introduced here.
Laurie and Jaggi (2003) modeled the role of vision in the dynamics of Schelling model.
As outcomes of non-linear function of vision (a distance agents can authentically see
their neighborhoods) and neighborhood preference threshold on the segregation index, they
identified three regimes of segregation: an unstable regime, where societies invariably
segregate; a stable regime, where integrated societies remain stable; and an intermediate
regime where a complex behavior is observed. J. Zhang (2004) demonstrated that
segregation is a stochastically stable state that tends to emerge and persist in the long run
regardless of initial state. Using a stochastic evolutionary game theoretic model, he showed
53


that slight asymmetry in residential preferences between two groups would produce
endogenous segregation.
OSullivan and Macgill (2005) investigated the impact of neighborhood scale on the
dynamics of segregation using two neighborhood types: continuous or local neighborhood
(immediately neighboring households in a lattice of residential locations) and bounded
neighborhood (regional neighborhood containing their residential locations). He found that
the time taken for the model to settle on a stable arrangement increases with neighborhood
size, the model to settle on a stable arrangement increases with neighborhood size, and that
larger neighborhoods may lead to situations where many households are unhappy but
unable to relocate to locations preferable to their location. K. Chen et al. (2005)
incorporated household preferences over location characteristics other than racial
composition preferences in their model. Extended household preference variables in their
model include housing price and neighborhood density, and the model allowed income
heterogeneity across races and among households of same race. Their preliminary findings
indicate that patterns of segregation can emerge even when individuals are wholly
indifferent to neighborhood racial composition, due to competing preferences over
neighborhood density. Bruch and Mare (2006) tested the sensitivity of residential choice
functions by testing three types of neighborhood threshold functions: Schelling-type
discrete choice function, stepwise function, and continuous functions with various slopes.
They found that the same average level of tolerance but different response functions give
rise to different neighborhood formation patterns.
Bernard and Wilier (2007) extended the Schelling model to incorporate the wealth and
status of agents and the desirability and affordability of residences. They analyzed the
effects of the degree of the status-wealth correlation and the extent to which the wealth of
residents shapes the affordability of residences on levels of status and wealth segregation.
Their findings are that the greater the correlation between status and wealth, the more the
agents tend to segregate, either due to choice (for the wealthy and high status) or exclusion
(for the poor and low status). Yin (2009) added economic factors (i.e. housing price) in
addition to the neighborhood racial composition factors. Using Buffalo, New York data,
54


she showed that segregation in the city could generate from the interaction of racial and
economic factors.
Non-Schelling model-based segregation research using ABM was developed by Benenson
et al. (2002). Benenson applied an Entity-based (EB) approach to simulate the dynamics of
ethnic distribution in the Yaffo area of the city of Tel Aviv, Israel, during the period 1955-
95. In their model, each householder is considered as a separate entity with its own
cultural, religious, or ethnic properties and whose residential behavior is defined by the
properties of the surrounding infrastructure entities (that is, the physical environment) and
other householders (the social environment). They defined two dissonances as dwelling
and residential dissonance: difference between the properties of an agent and the properties
of the neighbors and the dwelling types. They assumed that with the increase in combined
dissonances, the probability of leaving a residence increases, and the probability of
occupying a vacant residence decreases. The main difference of their model from the
Schelling model is that they defined the neighborhood based on the adjacency of Voronoi
polygons constructed around the centroid of the buildings instead of using lattice of cells.
Jayaprekash et al. (2009) presented a model of the interaction of segregation and
suburbanization in determining residential location. The model incorporated differential
income between two classes of agents, a simplified market mechanism for the purchase of
housing, and a simple geographic structure of one central city and four symmetrically
arranged suburbs. Agents derived utility from neighborhood racial composition, the size of
their lot, private amenities that are specific to neighborhoods and public amenities that
stretch across municipalities. They found that the public-amenities term leads to a positive-
feedback loop in which migration to suburbs increases the public amenities in those
municipalities while lowering amenities in the central city, thus sparking further migration.
When the minority agents were uniformly less affluent than the majority agents, this
dynamic produced discontinuity in segregation as measured by centralization.
55


4. RESEARCH METHODS
4.1 Study Areas
Study sites are selected in the historic downtown districts in three Colorado cities with
varying city sizes and with different distribution patterns of urban amenities: Boulder,
Denver, and Louisville. These three cities have a high level of urban and natural amenities,
and they are commonly ranked highly on the best place-type national surveys1".
The degree of the population growth in three study areas during the 1990-2000 period
shows less than the Colorado average level, since three study areas are located within
densely developed downtown areas (Table 4.1).
56


Table 4,1 Population change, 1990-2000, study areas
1990 2000 Change Change %
Boulder, CO YP 11,726 8,194 -3,532 -30.1%
MA 1,868 2,701 833 44.6%
TS 1,760 1,300 -460 -26.1%
Other 2,695 6,407 3,712 137.7%
Total 18,049 18,602 553 3.1%
HH 9,112 9,258 146 1.6%
Housing 9,519 9,508 -11 -0.1%
Louisville, CO YP 4,602 4,601 -1 0.0%
MA 956 2,411 1,455 152.2%
TS 437 606 169 38.7%
Other 2,856 3,608 752 26.3%
Total 8,851 11,226 2,375 26.8%
HH 3,299 4,131 832 25.2%
Housing 3,404 4,199 795 23.4%
Denver, CO YP 18,026 20,378 2,352 13.0%
MA 5,329 7,543 2,214 41.5%
TS 3,426 2,629 -797 -23.3%
Other 3,575 3,399 -176 -4.9%
Total 30,356 33,949 3,593 11.8%
HH 19,644 22,113 2,469 12.6%
Housing 23,553 23,479 -74 -0.3%
Colorado YP 1,418,386 1,707,088 288,702 20.4%
MA 588,224 953,432 365,208 62.1%
TS 329,443 416,073 86,630 26.3%
Other 958,341 1,224,668 266,327 27.8%
Total 3,294,394 4,301,261 1,006,867 30.6%
HH 1,282,489 1,658,238 375,749 29.3%
Housing 1,477,349 1,808,037 330,688 22.4%
U.S. YP 99,775,147 104,004,252 4,229,105 4.2%
MA 46,371,009 61,952,636 15,581,627 33.6%
TS 31,241,831 34,991,753 3,749,922 12.0%
Other 71,321,886 80,473,265 9,151,379 12.8%
Total 248,709,873 281,421,906 32,712,033 13.2%
HH 91,947,410 105,480,101 13,532,691 14.7%
Housing 102,263,678 115,904,641 13,640,963 13.3%
YP: Yong Professionals (age between 20-45), MA: Middle Ages (age between 45-65),
TS: The Seniors (age over 65), Other: Others (age between 0-20), HH: Households
57


4.2 Data
The data for this study were collected primarily from two sources: interviews with
individuals involved in urban residential markets as well as data in digital format such as
GIS and historic urban amenity location data from local governments and census data from
the U.S. Census Bureau.
4.2.1 Expert Interview
First, I conducted interviews with experts of the local housing markets in the three study
area between summer 2009 and spring 2010. Pilot interviews with three realtors, two
developers, and one planner were conducted for the preparation and refinement of the
survey instruments. Interviews with 15 realtors in the Denver-Boulder metropolitan area
were conducted after the pilot interviews. The interview was not tightly structured at the
beginning stage of the each interview, thus allowing the interviewees to raise issues
concerning urban residential location patterns they felt were important. Interviewees were
guided in the following ways:
- After I greet the interviewee and introduce myself, I provided a brief overview of the
purpose of the research project, which is to find out the mechanisms by which different
demographic groups make their residential location decisions.
- I asked interviewees to provide a brief overview of their firm or their personal
experiences in the field and to provide any materials that would assist me in
understanding their work and projects.
- I prompted the interviewees to talk about the locational preferences of their customers
and neighborhood preferences between their customers.
- As issues and topic areas emerged, I asked the interviewees to address any related areas if
they had not covered them in their unstructured conversations.
- Finally, I asked the interviewees to fill out an AHP survey instrument that were designed
to calculate different weights on various location decision factors by different
demographic groups.
58


The Analytical Hierarchy Process (AHP) (Satty, 1977) is a survey approach which allows
researchers to rank and evaluate preferences.1V Given a choice between two features,
subjects are asked to rate the relative importance of one over the other on a scale of one to
nine. The output of the AHP is a prioritized ranking or weighting of each decision
alternative (Atthirawong & McCarthy, 2002). For the AHP survey, subjects were given a
list of urban amenities and asked to compare a set of pairwise urban amenities for different
demographic groups. The quantitative output is intended to establish the relative
importance of the location decision factors. Fifteen variables in five categories from the
literature review and pilot interviews were used. To cross-validate the survey results, the
weights for location decision variables from the AHP survey were compared with the
verbally described comparative importance of the variables from the interviews.
Football Field, Ball Park, Theme Park
Recreational Amenity Recreation Center

Neighborhood Park, Trail, Bike Trail
Coffee Shop, Cafe
"The Third Places" Restaurant

Bar, Pub, Night Club
Orchestra, Opera, Musical
Cultural Amenity 1 1 J 1 1
Museum, Gallery
Library, Bookstore
Grocery Stores, Pharmacy
Shopping - Shopping Center
Small Retail Store
Public Transportation
Transportation < Highway Ramp

Major Arterial
Figure 4.2 Hierarchical Diagram for the AHP Survey
59


4.2.2 Data in Digital Format
To construct a computer-based urban residential location simulation model, various GIS
layers including road, zoning, and parcel data for the study areas were collected and
prepared for the spatial simulation. US Census data was used to analyze population
distribution patterns for different demographic groups at the census block level, which is
the smallest available unit of analysis.
Table 4.2 GIS data layers
Layer Source
Zoning Boulder and Denver Planning Department
Street Network U.S. Census Bureau TIGER/Line
Parcel Boulder and Denver Planning Department
Census Block Boundary U.S. Census Bureau TIGER/Line
Census Block Group Boundary U.S. Census Bureau TIGER/Line
Quarterly Census of Employment and Wages (QCEW) data was used to geocode the
historic locations of various types of urban amenities. Figure 4.3 and Table 4.4 show the
types and distribution of urban amenities in the study areas in the years 1990 and 2000.
Table 4.3 Quarterly Census of Employment and Wages data structure (N=l,363,213)
Table Element Descriptions
Year Year data collected
LGLNM Name of the reporting unit for legal purposes, both for single and multi-unit employers. Referred to as the Corporate Name in many systems
ADDRESS Identifies the physical location address
SIC Standard Industrial Classification code to identify the primary economic activity of the reporting unit
NAICS North American Industry Classification System codes are uniform industrial codes used by the United States, Canada, and Mexico to identify the primary activity of an establishment
60





J;%r4P£~> 'l^fef^FP ^
o ;3hzp&kt iktup 1t ^ _ J^Sq % <^p- t
Htfer!?fe f "r
y^fTrtfif
.^fz *fpWF
,?-<"£* * 1 _**4r _
arnoi
-SStef
SJ

1
*. .
|&5,Fpyj|S|
*-; irtf 1% ii(f
aliSEri
&naaRenl
O 9p Center
O ftetic tronst
(4) Denver, 2000
o. o r
-C^'t-jrj-
nA*r.:
iig
-i: V" ~ ^ =-2 r~ F -
z ~ Prj~*f7 Z 2 -f _
*^vF -t t,1" jPF; r4~
~e^~L -r r*t j -f as^jPM
L Ts^i
.^Ft^.r-rk :-.-^fJ-t-: rrr y -ft-t--4 ,tt -i .. _ __
H.z;Tt^a5'-^.n.n.r.;^x^r:c44:rt!Tr-t:j^T -r sj- }: atevrrF*
- *W j rrr !3 4: ^
oIV ~ _r-J JT < *1 o1o
-o* f z F- r_ -u J- tH
p-4|g
3a.
tf ^
'pRip
*
-* S
.--. 0 ..,- ... fry-....
a 9 Grocery See Center !
O Urn lAiMum Srou RtM
9 Ctte CVero O 9>i Center J
O Sal Part Put O tVeee Trowa j
ReaawM rt^wayRinv I
(6) Louisville, 2000
Figure 4.3 Distributions of Urban Amenities in the Study Areas, 1990, 2000
61


Table 4,4 Urban amenity changes in the study areas, 1990-2000
City ii g m n i ,:r Amenities 1990 2000 Increase Increase Rate
Boulder Football Field, Ball Park, Theme Park 3 6 3 100.0%
Recreation Center 11 35 24 218.2%
Neighborhood Park, Trail, Bike Trail 43 43 0 0.0%
Coffee Shop, Cafe 18 23 5 27.8%
Restaurant 75 190 115 153.3%
Bar, Pub, Night Club 5 11 6 120.0%
Orchestra, Opera, Musical 4 19 15 375.0%
Museum, Gallery 1 2 1 100.0%
Library, Bookstore 10 12 2 20.0%
Grocery Stores, Pharmacy 13 27 14 107.7%
Shopping Centers 1 1 0 0.0%
Small Retail Stores 60 109 49 81.7%
Boulder Total 244 478 234 95.9%
Denver Football Field, Ball Park, Theme Park 1 6 5 500.0%
Recreation Center 11 12 1 9.1%
Neighborhood Park, Trail, Bike Trail 21 21 0 0.0%
Coffee Shop, Cafe 16 16 0 0.0%
Restaurant 98 193 95 96.9%
Bar, Pub, Night Club 12 28 16 133.3%
Orchestra, Opera, Musical 6 11 5 83.3%
Museum, Gallery 7 7 0 0.0%
Library, Bookstore 3 9 6 200.0%
Grocery Stores, Pharmacy 10 27 17 170.0%
Shopping Centers 2 2 0 0.0%
Small Retail Stores 64 114 50 78.1%
Denver Total 251 446 195 77.7%
Louisville Football Field, Ball Park, Theme Park 1 1 0 0.0%
Recreation Center 3 5 2 66.7%
Neighborhood Park, Trail, Bike Trail 36 36 0 0.0%
Coffee Shop, Cafe 1 2 1 100.0%
Restaurant 6 30 24 400.0%
Bar, Pub, Night Club 1 3 2 200.0%
Orchestra, Opera, Musical 1 2 1 100.0%
Museum, Gallery 2 2 0 0.0%
Library, Bookstore 1 3 2 200.0%
Grocery Stores, Pharmacy 1 9 8 800.0%
Shopping Centers 0 0 0 NA
Small Retail Stores 4 4 0 0.0%
Louisville Total 57 97 40 70.2%
62


4.3 Variable Selection
Urban amenity variables for the simulation model were collected from the existing
literature and pilot interviews with the key informants in urban residential housing market
in the study areas. Accessibility to cultural amenities and shopping amenities are
commonly agreed as important location decision factors in the literature, while some pay
more attention to the entertainment amenities and others focus on small-scale social places
like coffee shops or cafes.
Table 4.5 Urban amenity variables from the existing literature and pilot interviews
Consumer City (Glaeser et al., 2006, 2001) The City as an Entertainment Machine (Lloyd et al., 2001) Creative Center (Florida, 2008, 2005a, 2002) Pilot Interview
Baseball field Baseball field Football field, Ball park, Theme park
Health club Recreation center
Bicycle path, Park Trail Neighborhood park, Trail, Bike trail
Cafe, Juice bar Cafe Coffee shop, Cafe
Restaurant Restaurant Restaurant
Bar or Tavern Bar, Night club Bar, Pub, Night club
Classical Music, Concert Opera Music venue Orchestra, Opera, Musical
Museum, Art gallery Museum Gallery Museum, Gallery
Bookstore, Library Library, Bookstore
Whole food store Grocery store, Pharmacy
Shopping center
Shop Boutique Shop Small retail store
Theater Theater
University
63


4.4 Agent-Based Model
4.4.1 Construction of Agents
4.4.1.1 Agent Categorization
The simulation model has three agent types: young professionals, middle age, and seniors.
These agent groups cover almost entire demand groups in an urban residential market since
only the age group below 20 is not included in the agent categorization. The traditional
form of a household, parents with children, was not categorized as an independent agent
type in the model mainly because of the unattractiveness of downtown as their housing
location and their relative small proportions in downtown housing market. The issue of
poorly performing public schools remains an unsolved problem for almost all American
cities and is a real deterrent to families who cannot afford to send their children to private
school. Creative programs are being explored to address the problem, but to date, solutions
have been elusive (Haughey, 2002).
In this ABM simulation, the model uses a household as an agent instead of an individual.
Lin (1997) discusses the advantages of a household-level approach to migration modeling.
Because migration decisions are often made by households rather than individuals,
household-level analyses are sometimes more consistent with rational expectations. For
example, a migrating individual may willingly undergo a net loss of opportunities or
amenities so that other household members experience a net gain. Using 1990 census data,
Lin presents several empirical models that show how individuals and households can differ
in their destination choices. These differences are generally modest.
From an empirical point of view, using an age-based agent categorization method in a
housing location decision study has a great advantage as opposed to the more complicated
and possibly difflcult-to-fmd alternatives.v Existing studies indicate that little statistical
explanatory power is lost in studies where age is the only or the most convenient variable
to use in describing the point reached in life (Morrow-Jones & Wenning, 2005). Variables
64


for which age can be a proxy include the following: income, status, wealth, need for more
or less space, need for low maintenance, accessible residences, and more help in daily tasks
or long-term care.
Table 4.6 Agent categorization
Agent Age Properties
38 out of 50 largest metropolitan areas saw increases in their 20- to 45-year-old population between 1990 and 2000. (average increase rate: 13.08%, national increase rate: 4.24%)
Young Professionals 20-45 Urban renovators: wealthy, young, highly educated, highly mobile and either single or married with less than two children, patronize cultural establishments in the central city, have a relatively high commuting cost, have relatively low demands for housing and land.
From 2000 to 2010, the percentage of the nations households between 55 and 64 is projected to jump to 17.4 percent.
Middle Ages 45-65 home buyers aged 45 and older who prefer denser, more compact housing alternatives will account for 31 percent of total homeowner growth during the 2000-10 period, double the same segments market share in the 1990s (Myers and Gearin, 2001).
According to the U.S. Census Bureau projections, a substantial increase in the number of older people will occur during the 2010 to 2030 period, after the first baby boomers turn 65 in 2011.
The Seniors Over 65 Growth among the senior population will substantially exceed that of younger adults, an unprecedented social and economic development.
It is predicted that a number of empty nesters will continue to grow as baby boomers age. After their children leave home, empty nesters often change their lifestyle in a way that favors downtown living.
65


4.4.1.2 Agent Behavior
Agent behavior rules are structured based on the literature review and interview results.
Each agent selects residential locations based on its preference score for each cell that
consists of the urban amenity score, the transportation network score and the neighborhood
score.
Total Score*; = a/ Amenity Accessibility Score*, + a2 Amenity Density Score*, +
bi Transportation Accessibility Score*, + b2 Transportation Density Score*, +
ci Neighborhood Score*, (4.1)
where,
i represents agent types.
- j represents each cell.
a/, a2, b*, b2, c/ are the weights for each score.
Exponential decay function was applied when calculating amenity and transportation
accessibility scores to reflect the decaying characteristics of the utility (benefits) related to
the distance between amenities and an origin. A quarter mile radius was applied as a
walking distance boundary in estimating the decay coefficient.
Accessibility Score*, = Ln (L (e~d Dx) wxi) (4.2)
where,
d represents the decay coefficient for the distance variable
D represents the Euclidian distance to the nearest amenity or transportation facility
W represents the AHP weight from the interview
x represents amenity and transportation facility types
66


Then, accessibility scores were normalized within the range of [0, 1] so that these factors
are comparable with other scores in the decision making. The scores were normalized by
using following function:
, Score Min
Score =----------
Max Min
(4.3)
Ball park
Recreation center
Neighborhood park, Trail
Coffee shop, Cafe
Restaurant
Bar, Pub, Night club
Orchestra, Opera, Musical
Museum, Gallery
Library, Bookstore
Grocery, Pharmacy
Shopping Center, Mall
Small Retails
Public Transportation
Highway Ramp
Major Arterial
X Weighty
X Weight y
X Weighty
X Weighty
X Weighty
X Weighty
X Weighty
X Weight y
X Weight y
X Weight y
X Weighty
X Weighty
X Weighty
X Weighty
X Weight y
Amenity Score Surface Agent;
ar
Figure 4.4 Calculation of the Amenity Score
67


Density Scores for urban amenities and transportation variables were calculated by
dividing the number of total amenities or transportation facilities in a quarter-mile radius
neighborhood by the maximum of the number of total amenities or transportation facilities
in a quarter-mile radius neighborhood. Neighborhood scores were calculated by dividing
the number of same type agents in a neighborhood by the total number of cells in a
neighborhood. The size of the neighborhood in calculating neighborhood score is not fixed
but controlled to measure the conceptual size of the neighborhood in housing location
decision process. Finally, total scores are calculated dynamically at each time step. Figure
4.5 shows overlaid total score surface for three agent types at one time step.
Table 4.7 Controllable model parameters
Category Attribute
Urban Amenities Recreational Amenities: ball park, rec center, small park "The Third Places*": cafe, restaurant, bar Cultural Amenities: opera, museum, library Shopping: grocery, shopping center, small retail
Transportation Networks Public transportation, highway ramp, major arterial
Neighborhoods 3X3, 5X5, 7X7 neighborhood size
Weights a, weights on amenity score b, weights on transportation score c, weights on neighborhood score
Mobility In-migration rate Out-migration rate Move-within rate
Oldenburg (1989)
68


Figure 4.5 Comparison of the Total Score Surfaces among Three Agent Types
4.4.2 Construction of an Agent-Based Model
4.4.2.1 Simulation Model
In this study, an ABM model was constructed on a JAVA-based platform. Agents are
populated in a 100 X 60 cell lattice with a cell size of 10,000 square feet based on 1990
Census data at the Block level. The focus of this ABM simulation is the spatial distribution
of agents in the urban area. At each time step (representing one year), a number of agents
in each group are generated based on historical data, and they start searching for their
housing locations. At the same time step, a number of agents in each group who were
previously located in the urban area decide to re-locate in (1) different neighborhoods in
69


the study area or (2) different locations outside the study area. At each scanning stage,
agents select a location based on a selective combination of individual preferences,
including neighborhood composition; transportation network: and the level of amenities,
which includes amount, type, and distribution.
The model dynamically generates several outputs during and after the completion of the
simulation: (1) overall spatial distribution of agents, (2) number of agents in each group in
each neighborhood, and (3) similarity index representing agent distribution patterns. The
similarity index, is calculated as follows:
Similarity =
n
P =
STi
N
(4.4)
where,
n represents the total number of agents
Pj represents the percentage of the same type agents in agent i's neighborhood
STi represents the number of same type agent in agent i's neighborhood
N represents the total number of cells in a neighborhood
70


Creating Urban Infrastructure

Populating Ur ban Amenities
Agents Creation
Search for Home
r
Moving In
I
Moving Out
i
CinQ
''Simulation Result.
Figure 4.6 Computational Approach
GIS Data
QCEW Data
Census Data
Lit Review
Interview
AHP
4.4.2.2 Validation of the Model
Validation concerns how well model outcomes represent real system behavior. Therefore,
validation involves comparing model outputs with real-world observations or the product
of another model or theory assumed to adequately characterize reality (Parker et al., 2003).
Initially, 1990 historic distribution of urban amenities were geocoded to the GIS
environment using QCEW data. Then, three heterogeneous demographic groups were
populated to the model, and the simulation was executed. The resulting distribution of
different demographic groups was compared with the 2000 US Census data at the Block
level for model evaluation.
71


Since model outcomes and the Census data have different spatial units, model outcomes
had be converted to comparable units. In this study, the proportion comparison method was
used. Cell-based model outcomes were grouped based on the Census Block boundary, and
for the Census data and model outcomes together, the proportions of each agent group in a
Census Block boundary were calculated. The differences in the proportions of each agent
group between the Census data and model outcomes were averaged to generate a model
discrepancy index.
Aggregated Discrepancy Index =
( 3
1=1
7=1
n
(4.5)
where,
Rcj is the ratio of agent type i in the reality
Rsj is the ratio of agent type i in the simulation result
n is the number of Census Blocks
72


Agent distribution, 1990
(Real, unknown)
mci
Agent distribution, 1990
at Census Block level
(Real, known)
Agent distribution, 1990
(Presumed)
Model Run
Agent distribution, 2000
(Simulated)
Agent distribution, 2000
at Census Block level
Agent distribution, 2000
(Real, unknown)
Figure 4.7 Conceptual Diagram of Model Run and Model Validation Process
Several simulation scenarios were developed for validation. First, to test the significance of
urban amenity variables and transportation variables, each set of variables as well as
combined sets of variables were applied in a model run and the model results were
compared with 2000 Census data. Additionally, three different sizes of neighborhoods
(3X3, 5X5, and 7X7 cell neighborhood size) were applied to the simulation model to test
the conceptual boundary of the neighborhood when people make housing location
decisions (Table 4.8).
73


Second, to measure the comparative importance between urban amenities and
transportation variables and neighborhood preference, the weights assigned to each set of
variables were controlled (Table 4.9).
Third, to test the explanatory powers of the different amenity-based urban residential
location theories, variables selected from Consumer City and Creative Center literature
were applied to the model. The full model, which includes all urban amenity variables was
also included for comparison purposes. Finally, the best model showing the best
performance was selected for each study area, and additionally transportation and
neighborhood variables were tested with the best model (Table 4.10).
Table 4.8 Simulation scenarios
Amenity and Transportation Variables
Neighborhood
Variables
1-1 Amenity accessibility 3X3 5X5 7X7
1-2 Amenity accessibility + Amenity density 3X3 5X5 7X7
1-3 Transportation accessibility 3X3 5X5 7X7
1-4 Transportation accessibility + Transportation Density 3X3 5X5 7X7
1-5 Amenity accessibility + Amenity density + Transportation accessibility 3X3 5X5 7X7
1-6 Amenity accessibility + Amenity density + Transportation accessibility + Transportation Density 3X3 5X5 7X7
74


Table 4,9 Weight sensitivity test
Variable Weight Variable Weight Variable Weight
II-1 Neighborhood 0 Transportation 0 Amenity 0
II-2 Neighborhood 0 Transportation 0 Amenity 1
II-3 Neighborhood 0 Transportation 0 Amenity 2
II-4 Neighborhood 0 Transportation 1 Amenity 0
II-5 Neighborhood 0 Transportation 1 Amenity 1
II-6 Neighborhood 0 Transportation 1 Amenity 2
II-7 Neighborhood 0 Transportation 2 Amenity 0
II-8 Neighborhood 0 Transportation 2 Amenity 1
II-9 Neighborhood 0 Transportation 2 Amenity 2
11-10 Neighborhood 1 Transportation 0 Amenity 0
11-11 Neighborhood 1 Transportation 0 Amenity 1
11-12 Neighborhood 1 Transportation 0 Amenity 2
11-13 Neighborhood 1 Transportation 1 Amenity 0
11-14 Neighborhood 1 Transportation 1 Amenity 1
11-15 Neighborhood 1 Transportation 1 Amenity 2
11-16 Neighborhood 1 Transportation 2 Amenity 0
11-17 Neighborhood 1 Transportation 2 Amenity 1
11-18 Neighborhood 1 Transportation 2 Amenity 2
11-19 Neighborhood 2 Transportation 0 Amenity 0
11-20 Neighborhood 2 Transportation 0 Amenity 1
11-21 Neighborhood 2 Transportation 0 Amenity 2
11-22 Neighborhood 2 Transportation 1 Amenity 0
11-23 Neighborhood 2 Transportation 1 Amenity 1
11-24 Neighborhood 2 Transportation 1 Amenity 2
11-25 Neighborhood 2 Transportation 2 Amenity 0
11-26 Neighborhood 2 Transportation 2 Amenity 1
11-27 Neighborhood 2 Transportation 2 Amenity 2
Table 4.10 Urban amenity-based theory test
Test Model
III-l Consumer City model
III-2 Creative Center model
III-3 Full model
III-4 Best model + Transportation
III-5 Best model + Transportation + Neighborhood
75


4.4.3 Future Simulation
4.4.3.1 Amenity District Policy Simulation
Many local governments in North America exercise a number of planning policies to
regulate the distribution of urban amenities in their boundaries. The impact of these public
policy interventions was tested using scenarios designed to depict various types, amounts,
and distributions of urban amenities. Scenarios were developed to reflect real world
situations in different planning regimes. Scenarios for testing include: even distribution of
urban amenities and clustering of urban amenities. Generated spatial distribution patterns
of agents from each scenario are analyzed and the results inform a discussion on policy
implications.
4.4.3.2 Location Preference Simulation
The second simulation scenario sets are not related to public policy but are related to
individual lifestyles. The current model assumes that agents change their location
preferences over time. For example, each demographic group has different location
preferences, and each individual agents location preference is governed by his or her
group. In the current model, an agents demographic group categorization is changed as
time step progresses (i.e. from middle age to the senior groups). However, it is possible
that one agents locational preference is not changing over time (i.e. considering the wealth
and health status of baby boomers, it is probable to assume that they will preserve their
active and entertainment-oriented lifestyle for long periods of time). Therefore, two
different scenarios, one with changing locational preferences and another with non-
changing locational preferences, were prepared and tested.
76


4.4.3.3 Different Size of Neighborhood Simulation
To test how agents neighborhood concept and size interact with neighborhood preferences
and lead to complex clustering behavior, the neighborhood size was controlled. The
simulations were executed with four different neighborhood sizes: no neighborhood, 3X3,
5X5, and 7X7 cell size neighborhood. Considering the current trend of urban resurgence
and the formation of urban neighborhood, it is plausible to assume that the size of
conceptual neighborhood will change in the foreseeable future. Heterogeneous distribution
patterns of the agent distribution are expected with varying sizes of the neighborhood.
77


5. RESULTS
In this chapter, I report the simulation results from the explanatory as well as prognostic
uses of the model. Both explanatory and prognostics models are pilot-tested against virtual
urban spaces and then applied with the real data. In the explanatory use, 29 combinations
of the location decision variable were tested in each study area. In the prognostic use, three
sets of future scenarios were tested.
5.1 Interview Results
Tables 5.1-5.3 show the AHP survey results from the study areas. The weights in the tables
represent the comparative importance of each location decision variable by agent types in
each study area. This comparative importance becomes the weight for the location decision
variables in the simulation model.
In the Boulder and Denver study areas, survey results show similar patterns that bar, pub,
nightclub as the most important location decision factor for young professional group and
restaurant was most important for the middle age group. In Louisville, neighborhood
park, trail, bike trail was ranked first for the young professional as well as for the middle
age group. For the senior group, the survey results from the three study areas show
heterogeneous outcomes: cafe, coffee shop ranked first in Boulder, opera, orchestra,
musical ranked first in Denver, and grocery store ranked first in Louisville.
In the Boulder study area, fitness-related amenities like neighborhood park, trail, bike
trail and recreation center ranked highly for the young professional group. Grocery
store and bar, pub, nightclub ranked highly for middle age group, and grocery store
and library, bookstore ranked highly for the senior group. Overall, the third place
category variables ranked highly for all three agent groups.
78


In the Denver study area, neighborhood park, trail, bike trail and cafe, coffee shop
ranked highly for both young professional and middle age groups. Library, bookstore
and cafe, coffee shop ranked highly for the senior group. Overall, the third place and
recreational amenity category variables ranked highly for young professional and middle
age groups, and cultural amenity category variable ranked highly for the senior group.
In the Louisville study area, recreation center ranked highly for both young professional
and middle age groups. Also, cafe, coffee shop ranked highly for young professionals
and grocery store ranked highly for middle ages group. Neighborhood park, trail, bike
trail and small retail store ranked highly for the senior group. Overall, similar to the
results from Denver, recreational amenity category variables ranked highly for young
professional and middle age groups. For the senior group in Louisville, variables in the
shopping category ranked highly in the survey results.
Table 5.1 AHP survey results, Boulder, CO
Variable Young Professionals Middle Ages Seniors
Weight Rank Weight Rank Weight Rank
Ball park 0.01823 15 0.01556 14 0.00959 14
Rec center 0.13831 3 0.10087 6 0.07532 5
Park, Trail 0.14086 2 0.13203 4 0.06654 6
Cafe 0.11483 5 0.11010 5 0.13936 1
Restaurant 0.04094 9 0.17091 1 0.06322 7
Bar 0.17472 1 0.14725 3 0.03286 10
Opera 0.02567 11 0.05760 9 0.06075 8
Museum 0.02198 12 0.02198 13 0.02114 11
Library 0.09899 7 0.08574 8 0.11974 3
Grocery 0.12591 4 0.16382 2 0.12076 2
Center 0.07795 8 0.01048 15 0.00496 15
Small 0.02666 10 0.03063 12 0.01804 12
Public 0.11217 6 0.09624 7 0.10960 4
Ramp 0.01916 13 0.05408 10 0.01402 13
Arterial 0.01916 13 0.03407 11 0.05131 9
79


Table 5.2 AHP survey results, Denver, CO
Variable Young Professionals Middle Ages Seniors
Weight Rank Weight Rank Weight Rank
Ball park 0.09337 4 0.08972 6 0.03363 9
Rec center 0.06914 5 0.05654 9 0.01702 15
Park, Trail 0.14101 2 0.17290 2 0.06711 5
Cafe 0.09516 3 0.12235 3 0.13663 3
Restaurant 0.04394 8 0.17591 1 0.06522 6
Bar 0.33740 1 0.12063 4 0.01934 12
Opera 0.01563 15 0.06203 8 0.19549 1
Museum 0.02447 11 0.07810 7 0.13184 4
Library 0.04134 9 0.09930 5 0.16470 2
Grocery 0.05115 7 0.04483 10 0.05448 8
Center 0.05411 6 0.02497 13 0.01803 13
Small 0.01743 13 0.01953 15 0.02410 11
Public 0.03122 10 0.02516 12 0.06189 7
Ramp 0.01814 12 0.03842 11 0.01753 14
Arterial 0.01630 14 0.02413 14 0.02744 10
Table 5.3 AHP survey results, Louisville, CO
Variable Young Professionals Middle Ages Seniors
Weight Rank Weight Rank Weight Rank
Ball park 0.06696 6 0.09327 5 0.06422 8
Rec center 0.14094 2 0.15474 2 0.07978 5
Park, Trail 0.26510 1 0.17341 1 0.13785 2
Cafe 0.10390 3 0.06006 8 0.06686 7
Restaurant 0.09988 4 0.08945 6 0.02294 12
Bar 0.07919 5 0.07625 7 0.01665 14
Opera 0.02009 12 0.02231 14 0.02259 13
Museum 0.01974 13 0.05181 11 0.01477 15
Library 0.05997 9 0.05676 9 0.02535 11
Grocery 0.06158 7 0.13246 3 0.14472 1
Center 0.06007 8 0.11690 4 0.07934 6
Small 0.02336 11 0.05316 10 0.13561 3
Public 0.01839 14 0.02135 15 0.08264 4
Ramp 0.04244 10 0.04234 12 0.04293 9
Arterial 0.01702 15 0.02662 13 0.03341 10
80


5.2 Explanatory Model Simulation Results
5.2.1 Pilot Simulation
Before testing the model using the real data, a simulation model was tested against a virtual
urban space for the model verification purpose: to test the model to see whether it behaves
as expected. This process is referred to as testing inner validity of the model (Brown,
2005). At the initial stage, agents in each demographic group were randomly distributed.
Population distribution was simulated over 30 years. For modeling simplicity, the number
of inflow and outflow of the population was fixed, and it was assumed that there was no
change in total population during the simulation period. In the 100 by 60 grid space, there
were 135 young professional, 135 middle age, and 90 senior agents at any time of the
simulation. At each stage of iteration, 360 randomly selected agents leave this space, and
40 randomly selected agents relocated their housing location within this virtual urban
space. Simulation results show the increasing similarity index over time: from 40%
initially up to 77% after the 30 iterations. The simulation model follows the logic of the
expected behaviors through computer programming.
Year 20 Year 25 Year 30
Figure 5.1 Pilot Simulation Results
81


80
30 rr-ir-r-rr-r
12345678 9101112131415161718192021222324252627282930
Figure 5.2 Similarity Index Change, Pilot Simulation
5.2.2 Real Simulation
5.2.2.1 Boulder
After the pilot simulation, residential location decision process and resulting population
distribution change was simulated using the real data over a ten-year period from 1990 to
2000. At the initial stage, agents in each demographic group are distributed based on 1990
Census data at the Block level: There are 6,662 young professional agents, 1,062 middle
age agents, and 1,000 senior agents in the Boulder study area. Two same type agents
occupy one cell in the urban space for the simulation. Model parameters were decided
based on the historic demographic change data. The total number of agents increased 1.6%
over the 10 year period. Annually, 201 young professional agents, 133 middle age agents,
and 35 senior agents moved into this simulation environment. Also, 355 randomly selected
agents left this space, and 89 randomly selected agents relocated their housing location
within this space annually.
82


(1) Dominant Agent by Census Block, 1990
(3) Cell-based Agent Distribution, 1990 (4) Cell-based Agent Distribution, 2000
Figure 5.3 Real Agent Distribution, Boulder, CO, 1900 and 2000
Six scenarios with different variable sets were tested and the simulation results were
compared with 2000 Census data (Figure 5.4). Most model outcomes show a certain degree
of agreement with actual demographic distribution from the Census data: clustering of
middle ages in the Western part of the study area and the concentration of young
professionals in the East. Considering the fact that about 52% of the agents relocated
during the simulation period, it is estimated that the models showed significant levels of
accuracy.
83


1-5. Amenity (Accessibility + Density) + 1-6. Amenity (Accessibility + Density) +
Transportation (Accessibility) Transportation (Accessibility + Density)
Figure 5.4 Simulation Results, Boulder, CO
Figure 5.5 shows the change of the similarity index during the ten-year simulation period
in the Boulder study area. During the simulation period, similarity indexes for most
scenarios showed similar patterns: They decreased during the first seven- to eight-year
period, and they started to rebound after that time period. However, two simulation
scenarios (1-4 and 1-6) with the transportation network variables did not show the
rebounding patterns and continually decreased during the simulation period. Several
84


explanations are possible: (1) In the Boulder study area, heterogeneous sets of urban
amenities preferred by different agent groups that are spatially mixed and concentrated in
small size location (see Figure 5.6), did not generate the clustering of the same agent type
neighborhood, (2) the variances between weights for the same variables by agent types
were too small to generate clustering of the same type agent neighborhoods. The average
AHP weight variance between agent types was 0.0011423. (3) After several years, most
preferred areas by all agent types were filled and, in less-preferred areas, clustering by the
same-type agents happened. (4) Variances for the transportation variables, however, were
too small to generate rebound in the scenarios 1-4 and 1-6 (Average variance between agent
type for urban amenity variables were 0.0013606 and for transportation variables were
0.0002690).
--1-1
1-2
-0-1-4
-B-l-5
A I-6
Figure 5.5 Similarity Index Change, Boulder, CO, 1990-2000
85


Y : Top 3 Young Professional Amenities: Bar, Park, Rec Center
M : Top 3 Middle Age Amenities: Restaurant, Grocery, Bar
S : Top 3 Senior Amenities: Cafe, Grocery, Library
: Other Amenities
w .m.
Figure 5.6 Distribution of Urban Amenities by Agent Type, Boulder, CO, 1990
5.2.2.2 Denver
At the initial stage of the simulation in the Denver study area, there were 13,220 young
professional agents, 3,910 middle age agents, and 2,515 senior agents based on the Census
data. Denver has a higher population and household density than Boulder, and five same-
type agents occupy one cell in the urban space for the simulation. Also, the Denver study
area shows a higher rate of population growth when compared with the Boulder study area.
The total number of agents increased 12.6% over the ten-year period. Annually, 748 young
professional agents, 331 middle age agents, and 52 senior agents moved into this
simulation environment. Also, 885 randomly selected agents left this space, and 221
86


Full Text

PAGE 1

DYNAMIC S O F TH E AMENIT Y CITY : A N AGENT-BASE D SIMULATIO N O F NEIGHBORHOO D LOCATIO N DECISION S b y Yuseun g Ki m B.S.,Yonse i University 199 9 M.R.P. Cornel l University 200 1 A thesi s submitte d t o th e Universit y o f Colorad o Denve r i n partia l fulfillmen t o f th e requirement s fo r th e degre e o f Docto r o f Philosophy Desig n an d Plannin g 201 0

PAGE 2

201 0 b y YuseungKi m Al l right s reserved

PAGE 3

Thi s thesi s fo r th e Docto r o f Philosoph y degre e b y Yuseun g Ki m ha s bee n approve d b y Thoma s Clar k J~an* Z B 2-o\0 Dat e

PAGE 4

Kim Yuseun g (Ph.D. Desig n an d Planning ) Dynamic s o f th e Amenit y City : A n Agent-Base d Simulatio n o f Neighborhoo d Locatio n Decision s Thesi s directe d b y Professo r Bria n Mulle r ABSTRAC T Th e purpos e o f thi s stud y i s t o identif y urba n amenity-base d residentia l locatio n decisio n factor s b y interviewin g expert s i n th e field s an d t o tes t th e factors explanator y powe r b y applyin g th e Agent-Base d Modelin g (ABM ) technique Fo r th e simulatio n model i t wa s assume d tha t househol d agent s decid e thei r residentia l locatio n base d o n th e physica l environment s an d non-physica l o r socia l environments Physica l environment s includ e th e relativ e location s o f urba n amenitie s a s wel l a s th e relativ e location s o f traditiona l locatio n decisio n factor s suc h a s transportatio n network s t o potentia l housin g locations Non physica l o r socia l environment s includ e bot h positiv e an d negativ e socia l interaction s betwee n household s an d thei r neighbors Intervie w result s confir m th e emergenc e o f urba n amenitie s a s importan t locatio n decisio n factor s i n loca l housin g market s i n th e stud y area s i n th e Colorad o Fron t Rang e region Th e AB M simulatio n scenario s wit h th e traditiona l job-accessibilit y variables however outperfor m th e simulatio n scenario s wit h urba n amenit y variable s i n al l thre e tes t areas : Boulder Denver an d Louisville Th e stud y result s sho w tha t durin g th e stud y perio d (1990-2000) job-accessibility-base d locatio n decisio n factor s wer e mor e powerfu l tha n urba n amenit y factor s i n th e residentia l locatio n decisio n processes Considerin g th e increase d discussion s abou t th e rol e o f urba n amenitie s i n attractin g huma n capita l an d th e cultura l trend s toward s entertainmen t an d consumption-oriente d lifestyles a differen t outcom e wit h highe r explanator y powe r fo r th e urba n amenit y variable s i s expecte d fo r th e simulatio n outcome s wit h mor e recen t data

PAGE 5

Testin g th e explanator y powe r o f th e existin g amenity-base d urba n developmen t an d locatio n decisio n theorie s suc h a s Consume r City Creativ e Center an d th e Cit y a s a n Entertainmen t Machin e discussion s resul t i n heterogeneou s outcome s i n eac h stud y area Simulatio n outcome s sho w tha t eac h stud y are a ha s a uniqu e se t o f urba n amenitie s tha t ar e effectiv e i n attractin g people a s wel l a s a uniqu e se t o f comparativ e importanc e amon g them I t i s argue d tha t a carefu l analysi s o f th e loca l condition s an d th e area-specifi c locatio n decisio n factor s ar e require d befor e th e applicatio n o f th e amenity-base d developmen t theor y t o a loca l place Finally usin g th e bes t performin g amenity-base d mode l i n eac h stud y area futur e demographi c distributio n pattern s i n th e stud y area s ar e predicte d wit h possibl e futur e simulatio n scenarios Thi s abstrac t accuratel y represent s th e conten t o f th e candidate' s thesis I recommen d it s publication Signe i fJ\" Bria n Mulle r

PAGE 6

DEDICATIO N PAG E I dedicat e thi s dissertatio n t o m y parent s an d m y wif e fo r thei r lov e an d endles s suppor t whil e I wa s completin g thi s dissertation

PAGE 7

ACKNOWLEDGMEN T M y specia l thank s t o m y advisor Bria n Muller fo r hi s guidanc e an d suppor t t o m y research I als o wis h t o than k th e member s o f m y committe e fo r teachin g an d sharin g thei r valuabl e knowledge Dr Thoma s Clar k gav e m e insightfu l comment s a t ever y ste p o f m y researc h progress Dr Kevi n Krize k assiste d m e wit h hi s constructiv e suggestion s o n m y research whil e arrangin g financia l suppor t fo r m e a s a program chair Graha m Billingsley FAICP share d hi s experienc e o f mor e tha n 3 0 year s o f plannin g practices Dr Subhrend u Gangopadhya y helpe d m e t o refin e m y mode l wit h hi s expertis e i n th e modelin g an d simulatio n study Thi s dissertatio n woul d no t hav e bee n possibl e withou t them I als o woul d lik e t o expres s m y appreciatio n t o m y colleague s i n th e Lan d Us e Future s La b an d al l th e Ph D student s a t th e Universit y o f Colorad o fo r thei r friendship encouragement an d inspiration

PAGE 8

TABL E O F CONTENT S Figure s x i Table s xii i Chapte r 1 Introductio n 1 1. 1 Backgroun d 1 1. 2 Researc h Too l an d Researc h Question s 3 1. 3 Outlin e o f th e Stud y 4 2 Theor y o f Urba n Residentia l Locatio n 5 2. 1 Phas e I (1920 s ~) : Market-Base d Approache s 5 2.1. 1 Bid-Ren t Approac h 5 2.1. 2 Huma n Ecolog y Approac h 7 2. 2 Phas e I I (1950 s ~) : Non-Market-Base d Approache s 8 2.2. 1 Politica l Econom y Approac h 8 2.2. 2 Residentia l Mobilit y Approac h 9 2. 3 Phas e II I (2000 s ~) : Amenity-Base d Approac h 1 1 2.3. 1 Residentia l Locatio n Decisio n Factor s 1 1 2.3. 2 Consumption-base d Neighborhoo d Developmen t 1 5 2.3. 3 Urba n Amenity-Base d Locatio n Decisio n Theor y 1 6 2.3.3. 1 Consume r Citie s 1 7 2.3.3. 2 Th e Cit y a s a n Entertainmen t Machin e 1 8 2.3.3. 3 Supersta r Citie s 1 9 2. 4 Demographic s i n Urba n Residentia l Marke t 2 0 2.4. 1 Majo r Instigator s o f Urba n Resurgenc e 2 0 2.4.1. 1 Youn g Professional s 2 1 2.4.1. 2 Th e Senior s 2 4 2.4. 2 Socia l Network s Amon g Demographi c Group s 2 8 2.4.2. 1 Socia l Interaction s an d Conflict s 2 8 2.4.2. 2 Neighborhoo d Effect s 3 0 2. 5 Th e Rol e o f th e Loca l Government s 3 3 VI M

PAGE 9

3 Theor y an d Application s o f Urba n Modelin g 3 6 3. 1 Historica l Developmen t 3 7 3.1. 1 Emergenc e o f Urba n Model s 3 7 3.1. 2 Larg e Scal e Mode l Critique s 3 9 3.1. 3 Lan d Use-Transportatio n Model s 4 0 3.1. 4 Plannin g Suppor t System s 4 1 3. 2 Evolutio n o f Urba n Modeling : CA/AB M 4 5 3.2. 1 Cellula r Automat a (CA ) 4 5 3.2. 2 Agent-Base d Mode l (ABM ) 4 6 3.2. 3 CA/AB M Approac h 4 7 3. 3 CA/AB M application s 4 8 3.3. 1 Residentia l Locatio n Modelin g 4 9 3.3. 2 Segregatio n Modelin g 5 3 4 Researc h Method s 5 6 4. 1 Stud y Area s 5 6 4. 2 Dat a 5 8 4.2. 1 Exper t Intervie w 5 8 4.2. 2 Dat a i n Digita l Forma t 6 0 4. 3 Variabl e Selectio n 6 3 4. 4 Agent-Base d Mode l 6 4 4.4. 1 Constructio n o f Agent s 6 4 4.4.1. 1 Agen t Categorizatio n 6 4 4.4.1. 2 Agen t Behavio r 6 6 4.4. 2 Constructio n o f a n Agent-Base d Mode l 6 9 4.4.2. 1 Simulatio n Mode l 6 9 4.4.2. 2 Validatio n o f th e Mode l 7 1 4.4. 3 Futur e Simulatio n 7 6 4.4.3. 1 Amenit y Distric t Polic y Simulatio n 7 6 4.4.3. 2 Locatio n Preferenc e Simulatio n 7 6 4.4.3. 3 Differen t Siz e o f Neighborhoo d Simulatio n 7 7 5 Result s 7 8 5. 1 Intervie w Result s 7 8 5. 2 Explanator y Mode l Simulatio n Result s 8 1 5.2. 1 Pilo t Simulatio n 8 1 5.2. 2 Rea l Simulatio n 8 2 I X

PAGE 10

5.2.2. 1 Boulde r 8 2 5.2.2. 2 Denve r 8 6 5.2.2. 3 Louisvill e 9 0 5.2.2. 4 Mode l Evaluatio n 9 4 5.2. 3 Additio n o f Neighborhoo d Variabl e 9 7 5.2.3. 1 Boulde r 9 7 5.2.3. 2 Denve r 9 9 5.2.3. 3 Louisvill e 10 1 5.2. 4 Amenity-Base d Theor y Tes t 10 3 5.2.4. 1 Boulde r 10 3 5.2.4. 2 Denve r 10 6 5.2.4. 3 Louisvill e 10 8 5. 3 Prognosti c Mode l Simulatio n Result s 11 0 5.3. 1 Pilo t Simulatio n 11 0 5.3. 2 Futur e Simulatio n i n th e Stud y Area s 11 2 6 Conclusio n 11 7 6. 1 Researc h Finding s 11 7 6.1. 1 Intervie w 11 7 6.1. 2 Explanator y Mode l 11 8 6.1. 3 Prognosti c Mode l 12 0 6. 2 Significanc e o f Researc h 12 1 6. 3 Limitation s o f Researc h 12 3 6. 4 Futur e Researc h Agend a 12 4 Endnote s 12 5 Appendi x A Surve y Instrumen t 13 1 B Simulatio n Outcome s wit h Neighborhoo d Variabl e 13 4 C Sensitivit y Tes t Result s 14 3 Bibliograph y 14 9 x

PAGE 11

FIGURE S Figur e 4. 1 Stud y Areas Fron t Rang e Region Colorad o 5 6 4. 2 Hierarchica l Diagra m fo r th e AH P Surve y 5 9 4. 3 Distribution s o f Urba n Amenitie s i n th e Stud y Areas 1990 200 0 6 1 4. 4 Calculatio n o f th e Amenit y Scor e 6 7 4. 5 Compariso n o f th e Tota l Scor e Surface s amon g Thre e Agen t Type s 6 9 4. 6 Computationa l Approac h 7 1 4. 7 Conceptua l Diagra m o f Mode l Ru n an d Mode l Validatio n Proces s 7 3 5. 1 Pilo t Simulatio n Result s 8 1 5. 2 Similarit y Inde x Change Pilo t Simulatio n 8 2 5. 3 Rea l Agen t Distribution Boulder CO 190 0 an d 200 0 8 3 5. 4 Simulatio n Results Boulder C O 8 4 5. 5 Similarit y Inde x Change Boulder CO 1990-200 0 8 5 5. 6 Distributio n o f Urba n Amenitie s b y Agen t Type Boulder CO 199 0 8 6 5. 7 Rea l Agen t Distribution Denver CO 190 0 an d 200 0 8 7 5. 8 Simulatio n Results Denver C O 8 8 5. 9 Similarit y Inde x Change Denver CO 1990-200 0 8 9 5.1 0 Distributio n o f Urba n Amenitie s b y Agen t Type Denver CO 199 0 9 0 5.1 1 Rea l Agen t Distribution Louisville CO 190 0 an d 200 0 9 1 5.1 2 Simulatio n Results Louisville,C O 9 2 5.1 3 Similarit y Inde x Change Louisville CO 1990-200 0 9 3 5.1 4 Distributio n o f Urba n Amenitie s b y Agen t Type Louisville CO 199 0 9 4 5.1 5 Bo x Plotte d Aggregate d Discrepanc y Inde x 9 5 5.1 6 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Boulder CO 1990-200 0 9 8 5.1 7 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Denver CO 1990-200 0 10 0 5.1 8 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Louisville CO 1990-200 0 ....10 2 5.1 9 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Boulder C O 10 4 X I

PAGE 12

5.2 0 Amenity-Base d Theor y Tes t Results Boulder CO 199 0 an d 200 0 10 5 5.2 1 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Denver C O 10 6 5.2 2 Amenity-Base d Theor y Tes t Results Denver CO 199 0 an d 200 0 10 7 5.2 3 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Louisville C O 10 8 5.2 4 Amenity-Base d Theor y Tes t Results Louisville CO 199 0 an d 200 0 10 9 5.2 5 Pilo t Futur e Simulatio n Results : Clusterin g vs Eve n Distributio n o f Amenities.. 11 0 5.2 6 Pilo t Futur e Simulatio n Results : Changin g vs Permanen t Locatio n Preference... 11 1 5.2 7 Pilo t Futur e Simulatio n Results : Varyin g Neighborhoo d Size s Il l 5.2 8 Futur e Simulatio n Results : Changin g vs Permanen t Preferenc e 11 3 5.2 9 Similarit y Inde x Change : Boulder Denver Louisville 2000-202 0 11 4 5.3 0 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Boulder C O 11 5 5.3 1 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Denver C O 11 5 5.3 2 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Louisville C O 11 6 A. l AH P Surve y Instrumen t 13 2 B. l Simulatio n Result s wit h 3X 3 Neighborhood Boulder C O 13 4 B. 2 Simulatio n Result s wit h 5X 5 Neighborhood Boulder C O 13 5 B. 3 Simulatio n Result s wit h 7X 7 Neighborhood Boulder C O 13 6 B. 4 Simulatio n Result s wit h 3X 3 Neighborhood Denver C O 13 7 B. 5 Simulatio n Result s wit h 5X 5 Neighborhood Denver C O 13 8 B. 6 Simulatio n Result s wit h 7X 7 Neighborhood Denver C O 13 9 B. 7 Simulatio n Result s wit h 3X 3 Neighborhood Louisville C O 14 0 B. 8 Simulatio n Result s wit h 5X 5 Neighborhood Louisville C O 14 1 B. 9 Simulatio n Result s wit h 7X 7 Neighborhood Louisville C O 14 2 C. l Bo x Plotte d Sensitivit y Tes t Results Boulder C O 14 4 C. 2 Bo x Plotte d Sensitivit y Tes t Results Denver C O 14 6 C. 3 Bo x Plotte d Sensitivit y Tes t Results Louisville C O 14 8 XI I

PAGE 13

TABLE S Tabl e 2. 1 Locatio n variable s i n selec t hedoni c pric e model s 1 3 3. 1 Compariso n o f th e urba n model s 4 3 3. 2 Compariso n o f selec t agent-base d residentia l locatio n model s 5 2 4. 1 Populatio n change 1990-2000 stud y area s 5 7 4. 2 GI S dat a layer s 6 0 4. 3 Quarterl y Censu s o f Employmen t an d Wage s dat a structur e 6 0 4. 4 Urba n amenit y change s i n th e stud y areas 1990-200 0 6 2 4. 5 Urba n amenit y variable s fro m th e existin g literatur e an d pilo t interview s 6 3 4. 6 Agen t categorizatio n 6 5 4. 7 Controllabl e mode l parameter s 6 8 4. 8 Simulatio n scenario s 7 4 4. 9 Weigh t sensitivit y tes t 7 5 4.1 0 Urba n amenity-base d theor y tes t 7 5 5. 1 AH P surve y results Boulder C O 7 9 5. 2 AH P surve y results Denver C O 8 0 5. 3 AH P surve y results Louisville C O 8 0 5. 4 Aggregate d discrepanc y inde x 9 6 5. 5 ANOV A analysi s result s fo r comparin g th e discrepanc y indexe s 9 6 5. 6 Aggregate d discrepanc y inde x wit h neighborhoo d variable Boulder C O 9 8 5. 7 Aggregate d discrepanc y inde x wit h neighborhoo d variable Denver C O 10 0 5. 8 Aggregate d discrepanc y inde x wit h neighborhoo d variable Louisville C O 10 2 5. 9 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Boulder C O 10 4 5.1 0 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Denver C O 10 6 5.1 1 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Louisville C O 10 8 6. 1 Averag e variance s betwee n weight s b y agen t type s 11 8 6. 2 Compariso n o f aggregate d discrepanc y indexe s 12 0 XII I

PAGE 14

C. l Sensitivit y tes t results Boulder C O 14 3 C. 2 Sensitivit y tes t results Denver C O 14 5 C. 3 Sensitivit y tes t results Louisville C O 14 7 XI V

PAGE 15

1 INTRODUCTIO N 1. 1 Backgroun d Wit h th e geographi c redistributio n o f populatio n an d resurgenc e o f th e centra l citie s i n metropolita n area s i n th e U S (i n spit e o f th e decentralizatio n o f th e employmen t center s i n recen t years ) th e discussion s o f huma n capita l an d amenity-base d locatio n decisio n theorie s ar e gainin g momentum. Ther e ar e a n increasin g numbe r o f huma n capitalist s wh o argu e th e importanc e o f peopl e a s th e moto r forc e behin d urba n developmen t (Florida 2002 ; Lloy d an d Clark 2001 ; Glaese r e t al. 2001 ; Glaeser 1998 ; Lucas 1988 ; Romer 1994) Thos e huma n capitalist s ar e i n par t indebte d t o th e insigh t o f Jan e Jacobs Jacob s (1961 ) note d severa l decade s ag o th e abilit y o f citie s t o attrac t peopl e an d thu s spu r economi c growth Luca s (1988 ) argue d th e productivit y effect s tha t com e fro m th e clusterin g o f huma n capita l a s th e critica l facto r i n regiona l prosperity Glaese r (1998 ) foun d empirica l evidenc e tha t huma n capita l i s on e o f th e centra l factor s i n regiona l growth Accordin g t o Glaeser suc h clusterin g o f huma n capita l i s th e ultimat e caus e o f regiona l agglomeratio n o f firms--no t merel y t o ta p th e advantage s fro m linke d network s o f customer s an d suppliers Florid a (2002 ) correlate d th e proportio n o f particula r type s o f workers th e "creativ e class o f artists engineers softwar e writers teachers an d th e degre e o f regiona l developmen t an d foun d a stron g relationshi p betwee n them H e define d spatia l concentration s o f thi s creativ e clas s a s "creativ e centers an d explained : Th e creativ e center s ar e no t thrivin g fo r suc h traditiona l economi c reason s a s acces s t o natura l resource s o r transportatio n routes No r ar e the y thrivin g becaus e thei r loca l government s hav e give n awa y th e stor e throug h ta x break s an d othe r incentive s t o lur e business The y ar e succeedin g largel y becaus e creativ e peopl e wan t t o liv e there Th e companie s the n follo w th e people—or i n man y cases ar e starte d b y them Creativ e center s provid e th e integrate d eco-syste m o r habita t wher e al l form s o f creativity—artisti c an d cultural technologica l an d economic—ca n tak e roo t an d flourish 1

PAGE 16

Huma n capitalist s i n th e urba n stud y are a commonl y agre e o n th e importanc e o f urba n amenitie s an d th e rol e o f socia l network s i n dens e urba n area s i n attractin g people especially hig h huma n capita l wit h hig h educationa l an d incom e levels Th e willingnes s t o pa y fo r urba n amenitie s relativ e t o th e amenitie s o f suburb s ma y hav e increased Fo r example i t ma y b e tha t th e citie s hav e alway s ha d mor e museums bu t th e valu e plac e o n proximit y t o museum s rise s wit h incom e an d educatio n (Glaese r an d Gottlieb 2006) Worker s i n th e elit e sector s o f th e postindustria l cit y mak e "qualit y o f life demand an d i n thei r consumptio n practice s ca n experienc e thei r ow n urba n locatio n a s i f tourists emphasizin g aestheti c concerns Thes e practice s impac t consideration s abou t th e prope r natur e o f amenitie s t o provid e i n contemporar y citie s (Lloy d an d Clark 2001) Creativ e peopl e ar e no t movin g t o thes e place s fo r traditiona l reasons Th e physica l attraction s tha t mos t citie s focu s o n building—sport s stadiums freeways, urba n mall s an d touris m an d entertainmen t district s tha t resembl e them e parks-ar e irrelevant insufficien t o r actuall y unattractiv e t o man y urba n dwellers Wha t the y loo k fo r i n communitie s ma y b e high-qualit y amenitie s an d experiences a n opennes s t o diversit y o f al l kinds an d abov e al l els e th e opportunit y t o validat e thei r identitie s a s creativ e peopl e (Florida 2002) Thi s vie w i s a radica l departur e fro m th e traditiona l interpretatio n o f th e residentia l locatio n decisio n proces s becaus e traditiona l studie s o n th e housin g searc h wer e focuse d o n th e produce r motivatio n o f th e locatio n decisio n process : the y ar e base d o n th e trade-of f relationshi p betwee n jo b accessibilit y an d ren t rate Furthermore thos e trade-of f relationship-base d housin g locatio n studie s ha d eithe r no t incorporate d spac e o r di d s o usin g rudimentar y Euclidea n measure s i n residentia l locatio n process I t i s quit e probable however tha t cities reaso n fo r bein g change d (i.e fro m "cit y a s a growt h machine (Molotch 1976 ) t o "cit y a s a n entertainmen t machine (T N Clark 2004) ) a s wel l a s thei r residents reaso n fo r livin g withi n the m change d (i.e fro m "mak e a living t o "enjo y a life" ) ove r th e pas t severa l decades Also th e amenity-base d residentia l locatio n theorie s ma y hav e comparabl e explanator y powe r t o th e traditiona l locatio n theorie s a t leas t fo r certai n demographi c groups Th e stud y o n th e rol e an d th e significanc e o f urba n amenitie s i n th e housin g locatio n decisio n proces s become s mor e importan t whe n w e conside r tha t ther e ar e increasin g 2

PAGE 17

number s o f politica l leader s i n Nort h Americ a wh o ar e convince d b y huma n capita l discussions. The y ar e exercisin g thei r plannin g effort s t o attrac t hig h huma n capita l t o thei r localities Thes e policie s includ e cas h contributions financin g aid ta x abatements zonin g an d buildin g cod e variances ne w o r improve d infrastructure preparatio n o f a n entertainment-oriente d district an d th e creatio n o f dedicate d taxe s fo r physica l improvements However withou t clea r understandin g o f th e rol e an d th e significanc e o f urba n amenitie s i n th e residentia l locatio n decisio n process thos e kind s o f loca l monetar y an d effort s spendin g ma y en d u p a s unfortunat e urba n policies Again w e nee d clea r understandin g o f th e residentia l locatio n decisio n process 1. 2 Researc h Too l an d Researc h Question s I n thi s research th e significance s o f variou s urba n amenit y variable s wil l b e teste d usin g th e agent-base d modelin g (ABM ) technique Agent-base d model s (ABMs ) ar e compute r representation s o f system s tha t ar e comprise d o f multiple interactin g actor s (Brown 2005) Agen t behaviora l rule s fo r thi s stud y wil l b e develope d b y interviewin g expert s i n th e loca l housin g market s i n th e Colorad o Fron t Rang e region Th e develope d mode l wil l b e applie d t o thre e Colorad o citie s wit h varyin g cit y size s an d wit h differen t distributio n pattern s o f urba n amenities : Boulder Denver an d Louisville B y simulatin g th e individua l action s o f heterogeneou s household s wit h divers e locationa l preferences an d measurin g th e resultin g syste m behavio r an d outcome s ove r time th e mode l wil l provid e a usefu l too l fo r understandin g urba n phenomen a an d th e processe s an d dynamic s behin d them I wil l us e th e agent-base d residentia l locatio n decisio n mode l i n tw o ways Th e firs t on e i s th e explanator y way : t o us e th e mode l a s a mean s t o understan d th e urba n housin g locatio n decisio n proces s usin g th e histori c data M y explanator y researc h questio n i s "Wha t i s th e rol e o f urba n amenitie s i n th e residentia l locatio n decisio n process? Th e sub-questio n relate d t o thi s questio n i s askin g th e explanator y powe r o f th e variou s existin g amenity 3

PAGE 18

base d locatio n decisio n theories : "Whic h amenity-base d urba n developmen t an d locatio n decisio n theor y successfull y describe s demographic s an d residentia l distributio n trends? Th e secon d wa y o f usin g th e agent-base d residentia l locatio n decisio n mode l i s th e prognosti c way : t o us e th e mode l t o extrapolat e th e trends an d t o evaluat e scenarios an d t o predic t th e futur e states M y prognosti c researc h question s i s "Ho w change s i n locatio n decisio n factor s suc h a s urba n polic y chang e o r individua l preferenc e chang e wil l impac t th e futur e distributio n o f populatio n an d th e formatio n o f th e neighborhood? Finally I wil l tes t th e usefulnes s o f th e agent-base d approac h i n urba n modelin g b y constructin g a n agent-base d residentia l locatio n decisio n mode l an d applyin g th e mode l fo r th e futur e simulation Th e questio n i s "Ca n a n agent-base d mode l b e a usefu l computationa l mode l fo r urba n simulation? 1. 3 Outlin e o f th e Stud y Chapte r 1 introduce s th e researc h topi c an d provide s th e relevan t contex t fo r it Thi s chapte r include s backgroun d o f th e problem researc h questions an d outline o f th e study Chapte r 2 i s a descriptio n o f th e existin g studie s o n urba n residentia l locatio n theorie s an d thei r deficiencies Chapte r 3 i s a descriptio n o f th e existin g studie s o n urba n modelin g an d thei r deficiencies Chapte r 4 i s a descriptio n o f m y researc h desig n plan Thi s chapte r include s explanation s o n data dat a collectio n method s includin g m y intervie w plan dat a analysi s method assumptions agen t typology an d simulatio n mode l schema Chapte r 5 o f th e stud y i s a n analysi s o f th e intervie w an d surve y result s a s wel l a s a n applicatio n o f th e surve y result s t o th e simulatio n model Th e spatia l distributio n o f heterogeneou s agent s i s analyzed Th e mai n objectiv e o f chapte r 6 i s t o summariz e th e stud y an d highligh t it s primar y contributio n t o th e theor y o f urba n residentia l location I underscor e th e finding s fro m thi s study includin g th e significanc e o f urba n amenitie s i n residentia l locatio n decisio n processe s tha t wil l b e analyze d fro m th e simulatio n model 4

PAGE 19

2 THEOR Y O F URBA N RESIDENTIA L LOCATION : A REVIE W O F LITERATUR E I n thi s chapter I revie w th e discussion s o n existin g theorie s o n urba n residentia l locatio n decisio n proces s an d evaluat e thei r explanator y powers I follo w th e chronologica l orde r o f th e developmen t o f th e theorie s fro m market-base d (1920s) t o non-market-base d (1950s) t o mor e recen t amenity-base d (2000s~ ) locatio n decisio n theories Additionally I revie w th e studie s o n th e demographi c structure s i n urba n housin g market s an d th e rol e o f th e loca l government s o n th e residentia l locatio n decision Mos t article s an d researc h reviewe d i n thi s chapte r ar e from academi c journal s o r independen t volume s b y academi c authors However du e t o th e novelt y o f th e topic ther e ar e severa l issue s tha t wer e no t studie d ye t i n academia I n thos e cases I revie w newspaper s an d professiona l journal s i n rea l estat e an d lan d developmen t fields 2. 1 Phas e I (1920 s ~) : Market-Base d Approache s 2.1. 1 Bid-Ren t Approac h Th e market-base d approac h ha s it s origin s i n th e sociologica l observation s o f th e Chicag o Schoo l i n th e 1920 s (Maclennan 1982) Th e marke t approach' s mai n concep t o f "bi d rent function s provide s th e foundatio n fo r mos t microeconomi c model s o f urba n spatia l structur e (Anas 1982) Ricard o (1921 ) i s credite d wit h th e ide a o f bi d ren t function whic h explain s tha t th e pric e o f agricultura l lan d i s determine d b y it s fertility I n vo n TMnen' s locatio n model accessibilit y replace s fertilit y a s th e determinan t o f lan d rent Eve r sinc e vo n Thiine n gav e hi s versio n o f th e monocentri c city-region geographers urba n economists an d planner s hav e bee n workin g o n theorie s o f cit y structur e tha t ca n bot h explai n an d predic t th e wa y i n whic h citie s ar e forme d an d hav e evolve d (Hu u Ph e & 5

PAGE 20

Wakely 2000) Alons o (1964 ) expande d upo n th e vo n Thiine n mode l t o th e mode l o f househol d locatio n decisions Whe n applie d t o th e housin g locatio n model housin g cos t ha s a trade-of f relationshi p wit h th e commutin g cost : housin g cost s shoul d ris e a s th e distanc e t o th e employmen t cente r fall s sinc e household s woul d b e willin g t o pa y mor e i n orde r t o sav e tim e gettin g t o work Thi s i s basicall y a n equilibriu m mode l o f locatio n choice Whe n al l household s ar e satisfie d wit h thei r locatio n choices locationa l equilibriu m occurs tha t is n o househol d want s t o chang e it s location I f ther e i s a chang e i n condition s tha t temporaril y disrupt s th e equilibrium al l household s ar e expecte d t o chang e residenc e withou t cos t an d instantaneousl y b y repeatin g th e bid-auctionin g process Extension s o f th e market-base d approac h include : estimatio n o f th e housing-pric e an d land-ren t functio n (J R Jackson 1979 ; Mills 1969) ; estimatio n o f th e relationshi p betwee n variou s housin g characteristic s (includin g locatio n an d differen t component s o f housing ) an d th e pric e o f housin g (Kai n & Quigley 1970 ; Straszheim 1975 ; King 1975) ; estimatio n o f populatio n an d employmen t densit y function s (Mills 1972 ; Muth 1969) ; a mode l o f urba n lan d us e tha t use s linea r an d nonlinea r programmin g t o allocat e lan d t o alternativ e activitie s (Mills 1976) ; an d th e discussio n o f th e severa l rendition s o f th e traditiona l monocentri c mode l (Wheaton 1977) Th e access/spac e trade-of f theory' s succes s i n replicatin g empirica l regularitie s i n Wester n citie s ha s mad e th e theor y a preferre d analytica l too l (Basset t & Short 1980) an d i t i s no w describe d a s "th e dominan t paradig m o f urba n economi c research (Maclennan 1982 ) an d a normativ e theor y (Fujita 1989) A s U. S citie s evolv e durin g th e las t hal f o f th e twentiet h century however th e explanator y powe r o f thi s traditiona l mode l becom e weaker Centra l cit y resurgenc e i s a goo d exampl e o f th e phenomeno n tha t th e marke t approac h fail s t o explain Th e trade-of f theor y suggest s tha t th e rich hav e a natura l propensit y t o liv e i n larg e parcel s i n th e suburbs wher e th e lan d i s chea p an d th e environmen t i s good becaus e the y ca n affor d th e transportatio n costs Th e poo r liv e i n th e inne r cit y becaus e the y canno t pa y fo r hig h transpor t costs Despit e a heav y degre e o f simplification thi s wa s mor e o r les s accepte d unti l th e 1970 s an d 1980s whe n th e ric h move d int o derelic t area s i n th e inne r city renovate d the m an d stayed i n a 6

PAGE 21

widesprea d phenomeno n late r calle d gentrificatio n (Hamnet t & Williams 1980) Thi s migratio n i s difficul t t o explai n withou t th e ris k o f contradictin g th e fundamenta l assumption s o f th e access/spac e trade-of f theory 2.1. 2 Huma n Ecolog y Approac h Base d o n bid-ren t functions Burges s (Par k & Burgess 1925 ) explaine d distributio n o f socia l group s withi n urba n areas Hi s mode l depict s urba n lan d us e i n concentri c rings : Th e CB D i s i n th e middl e o f th e model an d th e cit y expande d i n ring s wit h differen t lan d uses H e categorize s urba n lan d i n si x differen t lan d uses : CB S a s a center factor y zone th e transitio n zon e o f mixe d residentia l an d commercia l uses low-clas s residentia l zon e (inne r city) better-qualit y middle-clas s home s (oute r suburbs) an d commute r zone Hoy t (1939 ) modifie d th e concentri c mode l o f Burgess suggestin g tha t zone s expan d outwar d fro m th e cit y cente r alon g railroads highways an d othe r transportatio n arteries Hoy t theorize d tha t citie s ten d t o gro w i n wedge-shape d patterns o r sectors emanatin g fro m th e CB D an d cente r o n majo r transportatio n routes Highe r level s o f acces s mean t highe r lan d values ; thus man y commercia l function s woul d remai n i n th e CBD bu t manufacturin g function s woul d develo p i n a wedg e surroundin g transportatio n routes Residentia l function s woul d gro w i n wedge-shape d pattern s wit h a secto r o f low-incom e housin g borderin g manufacturing/industria l sector s (traffic noise an d pollutio n make s thes e area s th e leas t desirable) whil e sector s o f middle an d high-incom e household s wer e locate d furthes t awa y from thes e functions Harri s an d Ullma n (1945 ) propose d "multipl e nucle i model i n thei r articl e 'Th e Natur e o f Cities. Accordin g t o thei r model a cit y contain s mor e tha n on e cente r aroun d whic h activitie s revolve Som e activitie s ar e attracte d t o particula r node s whil e other s tr y t o avoi d them Fo r example a universit y nod e ma y attrac t well-educate d residents pizzerias an d bookstores wherea s a n airpor t ma y attrac t hotel s an d warehouses Incompatibl e lan d us e 7

PAGE 22

activitie s wil l avoi d clusterin g i n th e sam e area explainin g wh y heav y industr y an d high incom e housin g rarel y exis t i n th e sam e neighborhood Eve n thoug h thes e earl y attempt s t o explai n th e arrangemen t o f urba n lan d use s indicat e a broade r principl e o f urba n organization the y ar e base d o n earl y twentiet h centur y urba n developmen t an d transportatio n patterns Change s i n urba n environment s suc h a s advancemen t i n transportatio n an d informatio n technology changin g loca l urba n politics globalizin g economy an d diversifie d locatio n preference s b y household s a s wel l a s firm s mad e "zoning o f th e citie s difficult 2. 2 Phas e I I (1950 s ~) : Non-Market-Base d Approache s 2.2. 1 Politica l Econom y Approac h Th e politica l econom y approac h analyze s th e residentia l distributio n an d cit y structur e i n term s o f socia l group s inhabitin g urba n areas wit h som e group s takin g advantag e o r contro l ove r others Researc h i n thi s grou p focuse s o n differentiatio n an d disparitie s betwee n socia l group s wit h relatio n t o th e accessibilit y t o urba n good s an d services o n ho w divisio n o f urba n spac e reflec t differen t structure s o f consumption an d o n th e materia l an d symboli c value s tha t differen t group s attac h t o thes e differen t consumptio n landscape s (Soja 2000 ; Topalov 1989 ; Dea r & Flusty 1998 ; Harvey 1989 ; Scot t & Soja 1996) Th e politica l econom y grou p critique s th e market-base d theorie s a s ahistorica l (Hu u Ph e & Wakely 2000) Researcher s i n thi s grou p sugges t tha t residentia l locatio n pattern s ar e les s strongl y influence d b y marke t competition lan d an d hous e price s an d mor e strongl y influence d o r eve n manipulate d b y capita l throug h monopolisti c ren t (Harvey 1973 ; Smith 1987) The y argu e tha t spac e differentiatio n lead s t o increasin g inequit y i n acces s t o othe r urba n services Smith' s (1987 1979 ) "ren t ga p theory i s a n exampl e o f thi s group' s theoretica l explanation s t o th e urba n phenomeno n o f gentrification H e claime d tha t capita l make s 8

PAGE 23

investmen t decision s i n decaye d urba n area s becaus e o f th e potentia l gain s fro m substantia l increase s i n lan d ren t an d lan d value Populatio n turnove r i s a consequenc e o f th e process Thi s effec t i s du e t o th e ga p tha t ha s bee n create d betwee n th e curren t ren t an d th e expecte d ren t afte r rehabilitation i.e. betwee n capitalize d ren t an d potentia l ren t (Diappi & Bolchi 2008) Smit h argue d tha t th e ren t ga p i s bu t on e illustratio n o f th e uneve n spatia l developmen t characteristi c o f marke t societies an d thu s revitalizatio n i s a back-to-the-cit y movemen t b y capita l no t people (Smith 1979) Hi s mode l i s criticized however a s a partia l attemp t t o explai n gentrification Th e mode l ignore s publi c sector s an d th e deman d sid e b y no t identifyin g th e rol e o f loca l government s an d privat e actor s i n neighborhoo d revitalizatio n (Hamnett 1984 1991 ; Ley 1986) Mor e recently Fainstei n (2001) focusin g o n Londo n an d Ne w York describe d th e contemporar y urba n developmen t pattern s a s a confluenc e o f a n explosio n o f speculativ e buildin g fo r profi t an d a surgin g deman d fo r spac e withi n a contex t o f loca l incentive s t o growth Sh e explaine d that i n addition t o th e traditiona l marke t factor s o f deman d an d supply individua l developer s an d governmen t polic y pla y a n activ e rol e i n th e dynamic s o f th e rea l estat e market He r wor k show s th e importanc e o f th e wa y i n whic h th e developmen t industr y i s organize d withi n th e loca l economy an d th e institutiona l relation s o f th e developmen t secto r emerg e a s a ke y facto r i n urba n regeneratio n process Critic s sugges t tha t th e politica l econom y argumen t abou t monopolisti c ren t ha s neve r reall y bee n supporte d b y empirica l dat a (Kivell 1993) Anothe r criticis m i s tha t thi s approac h ignore s powerfu l impetuse s fo r residentia l location suc h a s life-cycles persona l preferences an d tast e (Hu u Ph e & Wakely 2000) 2.2. 2 Residentia l Mobilit y Approac h Anothe r non-marke t approac h t o residentia l location originatin g i n geograph y an d sociology i s base d o n residentia l mobility Thi s approac h include s work s tha t focu s o n th e househol d characteristic s an d o n dissatisfaction o r pus h factors inducin g mobilit y 9

PAGE 24

(Waddell 2000) Mos t o f th e researc h i n thi s grou p appl y Rossi' s (1955 ) lifecycl e hypothesis Lifecycl e theor y explain s tha t peopl e adjus t thei r housin g consumptio n t o fit changin g househol d need s wit h thei r progressio n throug h th e cycl e o f life : fo r example chang e i n househol d size ag e o f househol d members an d marriag e status Thi s i s basicall y a disequilibriu m mode l o f residentia l mobility whic h posit s tha t residentia l move s occu r whe n household s fal l ou t o f equilibriu m i n thei r housin g consumptio n (Clar k & Moore 1978 ; Quigle y & Weinberg 1977) Coup e an d Morga n (1981 ) extende d th e lifecycl e approac h t o sugges t tha t change s i n househol d an d persona l characteristic s ar e no t th e onl y factor s tha t shoul d b e considere d i n househol d relocatio n studies The y note d tha t housin g choice s ma y b e affecte d b y residentia l histor y an d marke t factor s o r force s tha t ar e externa l t o th e household Buildin g furthe r o n thi s concept Clar k an d Onaka' s (1983 ) wa s a uniqu e stud y tha t attempte d t o conside r a n amalgamatio n o f factor s drivin g residentia l relocatio n an d mobilit y processes The y characterize d residentia l mobilit y a s a combinatio n o f a n adjustmen t mov e (adjustin g t o th e market) a n induce d mov e (change s i n househol d compositio n an d lifecycle) an d a force d mov e (los s o f housin g uni t o r job) Researcher s i n thi s group however generall y focu s o n househol d decisions t o mov e bu t no t o n th e destination s o f th e migration Th e life-cycl e hypothesis however i s als o no t fre e fro m criticism Critic s o f th e life-cycl e mode l poin t t o th e poo r articulatio n o f life-cycl e concept s i n mobilit y researc h (W.A.V Clar k & Dieleman 1996 ; Pickle s & Davies 1985 ; Quigle y & Weinberg 1977 ) an d th e normativ e natur e o f th e model Ther e i s som e indicatio n tha t th e us e o f hea d o f househol d an d househol d variable s a s proxie s fo r life-cycl e stage s misrepresent s th e relationshi p betwee n life-cycl e chang e an d mobilit y (Wenning 1995) Also researc h suggest s tha t th e effect s o f famil y structur e o n mobilit y decisions a t leas t i n term s o f th e movemen t fro m rentin g t o owning ma y b e declinin g (Gyourk o & Linneman 1996) 1 0

PAGE 25

2. 3 Phas e II I (2000 s ~) : Amenity-Base d Approac h 2.3. 1 Residentia l Locatio n Decisio n Factor s A larg e bod y o f researc h ha s accumulate d ove r th e las t 3 0 year s o n locatio n choic e b y household s (Dieleman 2001 ; Luger 1996 ; Timmerman s et al. 1994 ; Michelson 1977 1987 ; Boehm 1982 ; D e Jong e t al. 1981 ; Bell 1968 ; Foote e t al. 1960) Accordin g t o th e traditiona l utilit y maximizatio n theory household s mak e residentia l locatio n decision s base d o n th e accessibilit y t o thei r workplace s i n orde r t o minimiz e commut e costs Frequently however household s hav e othe r priorit y factor s beside s jo b accessibilit y whe n the y mak e residentia l choices M J Ki m an d Morrow-Jone s (2005 ) found usin g a surve y dat a fro m recen t hom e buyers tha t th e traditiona l housin g locatio n variable accessibilit y t o job wa s relativel y unimportan t compare d t o othe r variables includin g th e following : housin g characteristic s (floo r plan qualit y o f construction an d cost) communit y characteristic s (safet y o f th e neighborhoo d an d goo d investmen t o r resal e value) an d schoo l qualit y (schoo l reputatio n an d qualit y o f schools academi c programs) Zonda g an d Pieter s (2005 ) als o argue d tha t th e rol e o f jo b accessibilit y i s significan t bu t smal l compare d wit h th e effec t o f demographi c factors neighborhoo d amenities an d dwellin g attribute s i n explainin g residentia l locatio n choices Existin g researc h focuse d o n environmenta l factor s an d neighborhoo d compositio n i n residentia l locatio n decisio n process Factor s i n th e environmenta l categor y includ e th e presenc e o f environmenta l amenities clea n air sceni c views an d recreatio n opportunitie s includin g acces s t o park s an d ope n spac e preserve d natura l habitat s (Rouwenda l & Meijer 2001 ; Wales 1978 ; Gawand e e t al. 2001 ; Hornste n & Fredman 2000 ; Tyrvainen 2001 ; Tyrvaine n & Vaananen 1998 ; Colwel l e t al. 2002 ; Greenber g & Lewis 2000) Factor s i n neighborhoo d compositio n categor y includ e housin g typ e an d ope n space th e presenc e o f nearb y retai l an d servic e facilities racia l differences an d neighborhoo d preferenc e (Parke s e t al. 2002 ; Bowe s & Ihlanfeldt 2001 ; T Ki m e t al. 2005) 1 1

PAGE 26

Anothe r market-base d approach hedoni c pric e model ha s bee n use d t o estimat e th e pric e o f housin g o r land base d o n a lis t o f variable s tha t ca n b e use d i n th e residentia l locatio n decisio n study I n thi s model th e pric e o f housin g o r lan d i s comprise d no t onl y o f characteristic s relatin g t o th e structur e itself—suc h a s typ e o f house size numbe r o r room s o r th e existenc e o f centra l heating—bu t als o o f characteristic s determine d b y location Th e latte r include s th e classi c elemen t o f urba n economi c modes accessibility Ther e ar e additiona l location-determine d characteristics suc h a s th e characte r o f neighborin g house s an d households localize d traffi c effect s an d th e qualit y o f th e micr o environment an d loca l publi c good s suc h a s school s (Cheshir e & Sheppard 1995) Tabl e 2. 1 provide s tabulate d location-specifi c variable s fro m existin g literature Commonl y applie d location specifi c variable s includ e park distanc e t o CBD neighborhoo d characteristic s lik e media n income an d transportatio n accessibility 1 2

PAGE 27

Tabl e 2. 1 Locatio n variable s i n selec t hedoni c pric e model s Cheshir e & Sheppar d (1995 ) Tse & Lov e (2000 ) J W u e t al (2004 ) Anderso n &Wes t (2006 ) Kon g e t al (2007 ) Cavailhe s etal (2009 ) Natura l Amenitie s / Disamenitie s Urba n Amenitie s / Disamenitie s Neighborhoo d Characteristic s Ope n Spac e Mai n Occupatio n Ethni c Grou p Ne w Constructio n Shoppin g Cente r Sport s Facilitie s Cemeter y Vie w Par k Rive r Lak e Wetlan d Slop e Elevatio n CB D Commercia l Zon e Industria l Zon e Traffi c Conditio n Housin g Densit y Incom e Smal l Par k Specia l Par k Gol f Cours e Lak e Rive r CB D Cemeter y Populatio n Densit y Incom e Crim e Par k Scener y Fores t Gree n Spac e CB D Plaz a Lan d Us e Gree n Spac e Agricultura l Lan d Steepnes s Bus h Wate r Floodin g Tow n Hal l Populatio n Incom e Zonin g Zonin g Populatio n unde r 1 8 Populatio n ove r 6 5 Schoo l Schoo l Distric t Educatio n Environmen t Transportatio n Bu s Rout e Publi c Transportatio n Majo r Roa d Road / Railroa d 1 3

PAGE 28

Anothe r grou p o f researcher s hav e focuse d o n th e locatio n decisio n factor s from housin g producers perspective : developers locatio n decisio n factors Early researc h o n develope r behavio r wa s conducte d b y a Universit y o f Nort h Carolin a grou p i n th e 1960 s an d earl y 1970 s (Chapi n & Weiss 1962 ; Donnel y e t al. 1964 ; Kaiser 1966 ; Weis s e t al 1966 ; Kaiser 1968 ; Kenney 1972) Developers locationa l decisio n factor s fro m thei r researc h includ e accessibilit y t o job s an d urba n services physica l an d socio-economi c characteristic s o f properties an d zonin g protection However thei r researc h interest s covere d th e wide r issu e o f th e residentia l developmen t proces s includin g th e spatia l distributio n o f subdivisions th e distributio n o f sit e characteristics th e categorizatio n o f develope r types an d th e targe t pric e rang e fo r th e developmen t bu t les s o n developers locatio n decisio n factors Durin g th e 1980 s ther e wer e severa l approache s t o identif y developers decisio n factor s fo r developmen t location Baerwal d (1981 ) describe d developers locatio n decisio n makin g proces s a s a three-stage d scannin g proces s b y spatia l scal e an d liste d sit e selectio n criteri a fo r eac h stage Hi s criteri a fo r th e firs t stag e include s genera l developmen t cos t variable s an d marketabilit y dat a fo r a whol e metropolitan region Fo r th e secon d stage criteri a includ e accessibility sit e characteristics an d governmenta l polic y variable s fo r scannin g relativel y larg e areas Fo r th e thir d stage criteri a includ e detaile d sit e characteristic s fo r th e examinatio n o f specifi c sites Hepne r (1983 ) identifie d th e factor s involve d i n th e proces s o f lan d us e conversio n an d arrange d th e factor s i n a hierarch y accordin g t o importanc e elicite d from th e developer survey Hi s researc h foun d prestige urba n amenity an d loca l governmen t service s a s importan t factor s i n developers locatio n decisions Mor e recently Smers h e t al (2003 ) identifie d spatia l factor s i n developer choice s b y trackin g ne w housin g unit s i n eac h sectio n (squar e mile ) i n Alachu a County Florid a b y th e yea r buil t ove r a twenty-yea r period Significan t variable s fro m thei r findin g include d th e distanc e fro m a centra l busines s district distanc e fro m highwa y interchanges housin g cost an d th e leve l o f regulation 1 4

PAGE 29

2.3. 2 Consumption-base d Neighborhoo d Developmen t Amon g researcher s i n th e politica l econom y school urba n revitalization redevelopment o r gentrificatio n i s viewe d a s a consumptio n practic e throug h whic h ne w middl e classe s see k t o distinguis h themselve s fro m th e ol d middle clas s (Ilkuca n & Sandikci 2005) The y perceiv e urba n redevelopmen t an d gentrificatio n proces s a s a spatia l manifestatio n o f th e value s o f th e "ne w middl e class. Th e "ne w middl e class, a n offsprin g o f post-Fordism i s a presume d t o urg e t o sta y awa y fro m th e taste s an d value s associate d wit h ol d middl e an d workin g classe s (Featherstone 1991 ; Lury, 1996) "Ne w middle class consumptio n practice s includ e shoppin g an d attendanc e i n cultura l an d socia l amenities whic h ar e expressiv e o f lifestyle Boorsti n (1973 ) firs t introduce d th e ter m "consumptio n community t o refe r t o informa l group s expressin g share d needs values o r lifestyle s throug h distinctiv e consumptio n patterns Cov a (1997 ) argue d tha t thi s "consumptio n community stil l ha s a "linkin g value, simila r t o th e sens e o f communit y tha t emerge d i n th e traditiona l neighborhood an d thi s share d consciousnes s i s reflecte d i n consumptio n practice s o f th e ne w residents creatin g a sens e o f wha t ar e appropriat e consumptio n practice s an d wha t ar e not Sac k (1988 ) argue d tha t consumptio n molde d people' s consciousnes s o f place helpe d the m t o construc t rea l places connecte d th e realm s o f nature socia l relation s an d meaning an d reveale d ho w geographica l setting s ar e constitutiv e o f contemporar y tension s an d paradoxes Followin g thi s argument man y loca l government s i n th e Unite d State s exercis e commercia l developmen t policie s base d o n th e hypothesi s tha t i f the y prepar e ne w commercia l developments the y wil l attrac t ne w worker s t o th e neighborhoods Thi s hypothesi s goe s further : Th e ne w worker s wil l requir e service s an d facilitie s tha t th e cit y mus t provide ; th e cit y i n tur n wil l loo k t o th e developer s wh o ar e profitin g fro m th e commercia l developmen t t o shar e th e cost s o f thos e service s an d facilities However th e existin g researc h o n thi s hypothetica l relationshi p i s limited an d th e causa l relationshi p betwee n th e attractio n o f peopl e an d commercia l developmen t i s stil l unclear 1 5

PAGE 30

Analyzin g thre e America n cities Boston Sa n Francisco an d Sant a Monica whic h hav e adopte d downtow n plannin g policie s tha t lin k large-scal e commercia l developmen t wit h housing transit an d employment Keatin g (1986 ) conclude d tha t thes e policie s ar e likel y t o hav e a significan t bu t margina l effec t o n socia l problem s aggravate d b y downtow n growth H e als o argue d tha t thes e policie s shoul d b e tie d t o comprehensiv e downtow n plans Andre w an d Merria n (1988 ) emphasize d th e importanc e o f th e economi c feasibilit y an d th e lega l foundation s fo r th e commercia l development-drive n urba n policies 2.3. 3 Urban Amenity-Base d Locatio n Decisio n Theor y Studie s o n th e effec t o f urba n amenitie s i n attractin g peopl e t o centra l location s ar e ver y limited especially i n Norther n Europ e an d Nort h America wher e th e omissio n i s encourage d b y a n asceti c Protestantism labelin g "non-work a s quasi-sinfu l (T N Clark 2004) However ther e i s a n expandin g discussio n o n th e rol e o f amenitie s i n a n urba n contex t a s a pul l facto r o f urba n migration Th e earl y stage s o f urba n influ x wer e analyze d a s phenomen a associate d wit h th e countercultura l lifestyle includin g avant-gard e artist s (P Jackson 1985 ) an d ga y communitie s (Castells 1983) Kasard a e t al (1997 ) explaine d th e movemen t a s higher incom e households searc h fo r th e location s wit h th e lowes t ta x burde n an d th e greates t bundl e o f locatio n amenities The y speculate d tha t non-urba n resident s wh o hav e move d t o centra l citie s ar e attracte d b y amenitie s suc h a s a ric h diversit y o f lifestyles significan t architectura l resources divers e commercia l opportunities an d othe r entertainmen t option s suc h a s ethni c restaurant s an d cultura l activities Suchma n (2005 ) argue d tha t qualit y o f plac e ha s becom e a paramoun t concer n whe n employer s an d peopl e sho p fo r a plac e t o be H e liste d urba n amenitie s suc h a s theaters museums sport s facilities restaurants bars parks an d civi c spaces an d area s o f histori c o r architectura l interes t i n addition t o urba n areas inheren t appea l a s th e symbolic physical economic an d cultura l center s o f metropolita n spaces B y analyzin g th e 1970-200 0 U S Censu s data Che n an d Rosentha l 1 6

PAGE 31

(2008 ) foun d tha t improvin g consume r amenitie s typicall y attract s retirees especiall y highl y educate d people Urba n amenitie s an d urba n dweller s ar e considere d t o hav e a reciproca l impac t o n eac h other Waldfoge l (2003 ) foun d empirica l suppor t i n th e market s fo r broadcas t radio newspapers an d restaurants Fo r example whe n ther e i s a large r loca l consume r bas e fo r a certai n forma t o f radi o station caliber o f newspaper o r styl e o f restaurant, mor e o f the m exis t i n a city B y reveale d preference tha t greate r variety increase s cit y dwellers welfar e becaus e th e mor e option s ther e ar e fo r resident s tha t shar e a particula r se t o f tastes th e mor e the y consume Next thre e widel y know n discussion s aroun d th e issu e o f urba n amenit y wil l b e introduced 2.3.3. 1 Consume r Citie s Researcher s intereste d i n centra l cit y economie s o f post-industria l citie s an d urba n revitalizatio n project s ar e increasingl y focuse d o n consumption-base d industries—includin g cultura l facilities specialize d boutique s an d eateries an d retailin g an d leisur e complexes — whos e succes s depend s o n harnessin g th e lifestyle s o f variou s group s wit h disposabl e incom e t o particula r kind s o f consumptio n practice s (Hannigan 1998 ; Zukin 1998) Davi d Brook s i n hi s boo k Bobos in Paradise (2000 ) describe d th e changin g definitio n o f socia l clas s usin g th e concep t o f "consumption, "Kar l Mar x argue d tha t classe s ar e define d b y thei r mean s o f productio n bu t i t coul d b e tru e that i n th e informatio n ag e a t least classe s defin e themselve s b y thei r mean s o f consumption. Glaese r e t al (2001 ) argue d tha t nowaday s thrivin g citie s ar e "consume r cities one s tha t attrac t highl y educate d household s throug h appealin g cultura l amenities suc h a s museums restaurant s an d th e opera The y demonstrate d tha t betwee n 197 7 an d 1995 liv e performanc e venue s an d restaurant s ar e correlate d wit h th e futur e populatio n growth B y contrast bowlin g alley s ar e correlate d wit h populatio n decline The y argue d tha t hig h 1 7

PAGE 32

amenit y citie s hav e grow n faste r tha n lo w amenit y cities Urba n rent s hav e gon e u p faste r tha n urba n wages suggestin g tha t th e deman d fo r livin g i n citie s ha s rise n fo r reason s beyon d risin g wage s (Glaese r e t al. 2001) Glaeser' s fou r critica l urba n amenitie s includ e th e following : (1 ) th e presenc e o f a rich variet y o f service s an d consume r goods : restaurants theaters an d a n attractiv e mi x o f social partner s tha t ar e har d t o transpor t an d ar e therefor e loca l goods ; (2 ) aesthetic s an d physica l setting : architectura l beauty an d mil d weather whic h i s th e singl e mos t importan t determinan t o f populatio n o r housin g pric e growt h a t th e count y level ; (3 ) goo d publi c service s includin g goo d school s an d les s crime ; an d (4 ) spee d includin g th e rang e o f service s (an d jobs ) availabl e i n metropolita n area s a s a functio n o f th e eas e wit h whic h individual s ca n mov e around Finally h e emphasize d th e rol e o f citie s an d dens e urba n area s a s place s facilitatin g human s t o interac t socially H e provide d evidenc e fo r th e Unite d States suggestin g tha t th e resurgenc e o f bi g citie s i n 1990 s i s due i n part t o th e increase d deman d fo r socia l interaction s an d t o th e reductio n i n crime whic h ha d mad e i t difficul t fo r urba n resident s t o enjo y th e socia l amenities 2.3.3. 2 Th e Cit y a s a n Entertainmen t Machin e T N Clar k (2004 ) propose d tha t contemporar y citie s ar e "entertainmen t machine[s]. H e argue s tha t worker s i n th e elit e sector s o f th e postindustria l cit y mak e th e "qualit y o f life demand an d i n thei r consumptio n practice s ca n experienc e thei r ow n urba n locatio n a s i f tourists emphasizin g aestheti c concerns I n thei r argument thes e practice s impac t consideration s abou t th e prope r natur e o f amenitie s t o provid e i n contemporar y cities an d th e cit y become s a n entertainmen t machine leveragin g cultur e t o enhanc e economi c well being Usin g dat a fro m 3,11 1 U S counties h e measure d th e impac t o f amenitie s o n populatio n growth H e use d tw o type s o f amenities : natura l amenities includin g si x component s suc h a s moderat e temperatur e an d wate r whil e constructe d amenitie s included th e opera juic e 1 8

PAGE 33

bars museum an d Starbucks H e foun d tha t sub-populate d group s behav e differently Colleg e graduate s ar e mor e numerou s wher e ther e ar e fewe r natura l bu t mor e constructe d amenities Senior s ar e th e opposite : thei r number s increas e wit h mor e natura l amenities bu t les s wit h constructe d amenities Resident s filin g hig h tec h patent s liv e i n location s wit h bot h natura l an d constructe d amenities 2.3.3. 3 Supersta r Citie s Gyourk o e t al (2006 ) identifie d a handfu l o f metropolita n area s experiencin g th e concentratio n o f high-incom e peopl e an d consequentia l housin g pric e growt h tha t significantl y exceede d th e nationa l average leadin g t o a widenin g ga p acros s location s i n averag e hous e prices I n thei r discussio n o f "Supersta r Cities, the y showe d tha t high-demand citie s hav e incom e distribution s tha t ar e shifte d t o th e right : low-incom e familie s ca n liv e ther e onl y i f the y hav e a ver y stron g preferenc e fo r th e cit y whil e high-incom e familie s ca n liv e ther e eve n i f the y onl y modestl y prefe r it A s th e nationa l high-incom e populatio n grows th e greate r numbe r o f high-incom e familie s outbi d relativel y lo w incom e familie s (a s well a s som e high-incom e families ) wh o ar e unwillin g o r unabl e t o pa y a highe r premiu m t o liv e i n thei r preferre d location The y foun d tha t suc h supersta r location s experienc e supra-norma l hous e pric e growt h an d a shif t o f thei r incom e distribution s t o th e righ t a s the y experienc e inflow s o f high-incom e household s an d outflow s o f thei r lowest-incom e residents The y explaine d thi s phenomeno n b y a n inelasti c suppl y o f lan d i n som e attractiv e location s combine d wit h a n increasin g numbe r o f high-incom e household s nationally The y di d no t specif y th e lis t o f amenitie s tha t supersta r citie s provide bu t thei r finding s impl y tha t ther e mus t b e somethin g uniqu e an d attractiv e abou t supersta r cities otherwis e potentia l resident s woul d tur n t o cheape r location s an d supersta r citie s woul d no t b e abl e t o sustai n exces s pric e growth 1 9

PAGE 34

2. 4 Demographic s i n Urba n Residentia l Marke t 2.4. 1 Majo r Instigator s o f Urba n Resurgenc e I t i s widel y observe d tha t urba n resurgence especiall y fro m youn g professional s an d entrepreneurs i s relate d t o th e emergenc e o f stron g creativ e an d technologica l industrie s fro m th e 1970 s (Clay 1979 ; C Hamnett 1991 ; Lang 1982 ; Smith 1979) Durin g th e economi c boo m o f th e lat e 1990s high-tec h companie s "le d th e charg e downtown i n orde r t o accommodat e "ne w econom y workers wh o appreciat e th e aesthetic s o f downtow n wor k an d life Th e move-to-the-downtow n i s a respons e t o th e ne w economy' s insistenc e tha t ideas—an d th e peopl e wh o generat e them—ar e a company' s mos t valuabl e commodit y (Suchman 2002) Mor e evidenc e b y th e U S Departmen t o f Housin g an d Urba n Developmen t show s tha t citie s hav e becom e center s fo r high-tec h jo b growt h (U.S Dept o f Housin g an d Urba n Development 2000) High-tech job s mak e u p almos t 1 0 percen t o f al l job s i n centra l citie s accordin g t o th e report whic h i s nearl y identica l t o th e percentag e foun d i n th e suburbs Furthermore high-tec h jo b growt h i n citie s increase d b y 26.7 % betwee n 199 2 an d 1997 mor e tha n thre e time s thei r overal l increase Wit h th e changin g economi c structure s o f th e cities structure s o f familie s ar e als o changing Evidenc e show s tha t househol d structur e i s changin g o n a nationa l level I n 1940 les s tha n eigh t percen t o f al l household s consiste d o f peopl e livin g alone ; today single s mak e u p 25 % o f America n households Betwee n 194 0 an d 2000 th e numbe r o f unmarrie d peopl e livin g togethe r a s couple s increase d b y 72% t o 5.4 7 million B y 2020 marrie d couple s wit h childre n ar e projecte d t o accoun t fo r onl y on e i n fiv e households Th e traditiona l nuclea r family a workin g father stay-at-hom e mother an d tw o o r mor e children constitute s les s tha n on e quarte r o f al l household s today A s populatio n an d househol d type s becom e increasingl y divers e an d a s the y deman d housin g appropriat e fo r thei r lifestyles a broade r rang e o f housin g type s i s needed an d urba n dwelling s becom e a popula r optio n amon g thes e divers e househol d type s (Lurz 1999) 2 0

PAGE 35

Existin g literatur e commonl y identifie s increasin g number s o f th e senior s an d youn g professional s a s tw o o f th e bigges t demographi c group s wh o instigat e th e urba n resurgence Thes e tw o group s shar e a numbe r o f characteristic s tha t mak e the m a goo d matc h fo r downtow n living First the y ar e no t concerne d wit h schoo l quality somethin g tha t ofte n deter s familie s fro m livin g i n centra l cities Second the y ofte n see k low maintenanc e housin g tha t doe s no t requir e extensiv e yar d wor k an d hom e repairs Third bot h group s ten d t o hav e th e time money an d inclinatio n t o partak e i n urba n amenitie s (Sohmer 1999) 2.4.1. 1 Youn g Professional s Thes e upper-middl e clas s group s o f peopl e ar e i n betwee n thei r twentie s an d earl y forties Thi s population—peopl e wh o ar e delayin g marriag e o r puttin g of f havin g children—i s growing Accordin g t o th e U.S Censu s Bureau approximatel y 67 % o f America n household s ar e currentl y childles s (wit h n o childre n unde r 18) B y 2010 projection s sho w thi s figure jumpin g t o 7 2 percent A t th e sam e time the y ar e th e mos t mobile peopl e i n th e America n population Ove r th e five-yea r perio d fro m 199 5 t o 2000 som e 6. 6 millio n 2 5 t o 3 4 year-old s move d fro m on e metropolita n are a t o anothe r (Cortright 2006) Ther e ar e a numbe r o f researcher s wh o identif y youn g professional s a s generator s o f urba n resurgenc e an d focu s o n thei r motivation s fo r migratin g t o cities Examinin g th e inner-cit y neighborhood s i n Washington DC Gal e (1979 ) foun d tha t th e earlies t ne w settler s i n a n inner-cit y neighborhoo d ar e likel y t o b e singl e male s searchin g fo r thei r firs t homes Th e lac k o f childre n tend s t o mak e thi s initia l pionee r clas s o f in-migration s obliviou s t o th e risk s associate d wit h th e deteriorate d building s an d highe r crim e rate s commo n t o man y depresse d inner-cit y neighborhoods A s mor e an d mor e migrant s mov e int o an d inves t i n th e neighborhood i t become s mor e stable an d ne w classe s o f "risk-prone and eventually "risk-averse resident s begi n t o occup y renovate d dwellings 2 1

PAGE 36

Usin g migratio n dat a fro m th e 198 2 America n Housin g Surve y b y th e U S Censu s Bureau Spai n (1989 ) foun d tha t single childles s householder s ar e mor e likel y t o choos e th e cit y ove r th e suburbs Usin g a mai l questionnair e o f hom e buyer s i n Cincinnati Varad y (1990 ) foun d tha t college-educated childles s household s desirin g employmen t accessibilit y an d cosmopolita n amenitie s woul d ten d t o locat e i n th e cit y an d tha t rac e an d incom e wer e importan t determinant s o f a household' s locatio n decision Ker n (1984 ) describe d th e characteristic s o f urba n renovators Th e typica l renovato r i s wealthy young highl y educated an d eithe r singl e o r marrie d wit h les s tha n tw o children Suc h peopl e ar e attracte d t o central-cit y location s becaus e b y (1 ) patroniz e cultura l establishment s i n th e centra l city (2 ) hav e a relativel y hig h commutin g cost an d (3 ) hav e relativel y lo w demand s fo r housin g an d land Analyzin g th e U S Censu s data Cortrigh t (2006 ) foun d tha t abou t a thir d o f th e 5 0 larges t metropolita n area s sa w increase s i n thei r 2 5 t o 3 4 year-ol d populatio n betwee n 199 0 an d 2000 I n contrast severa l metropolitan area s sa w decline s i n thei r 2 5 t o 3 4 year-ol d populatio n o f mor e tha n 20% S o th e dat a show s tha t th e residentia l locatio n decision s o f th e youn g peopl e ar e disproportionatel y favorin g certai n metropolita n areas Mor e specifically th e growt h i n th e numbe r o f college-educate d youn g adult s i s fuelin g prosperit y i n place s lik e Austin Charlotte Atlanta, Portland Raleigh-Durham an d Phoenix Th e concentratio n o f youn g peopl e i n fewe r citie s make s thos e citie s eve n mor e attractiv e place s fo r talente d people creatin g a powerfu l gravitationa l pul l fo r othe r youn g peopl e an d formin g a positiv e feedbac k loop Eve n withi n metropolitan areas plac e i s playin g a n increasingl y importan t role Durin g th e 1990s th e preferenc e o f youn g adult s fo r close-i n neighborhood s (withi n 3 mile s o f th e region' s center ) increase d sharply I n 1990 2 5 t o 3 4 year-old s wer e abou t te n percen t mor e likel y tha n othe r resident s i n th e metropolita n are a t o liv e i n th e close-i n neighborhoods B y 2000 thes e youn g adult s wer e mor e tha n 3 0 percen t mor e likel y tha n othe r metropolita n resident s t o liv e i n thes e close-i n neighborhoods Strikingly th e relativ e attractivenes s o f centra l neighborhood s t o youn g adult s increase d significantl y an d i n ever y on e o f th e to p 5 0 metropolita n area s i n th e 1990s I n 1990 i n th e aggregate 2 5 t o 3 4 year 2 2

PAGE 37

old s wer e abou t 12 % mor e likel y tha n othe r American s t o liv e i n a close-i n neighborhood ; b y 2000 the y wer e 33 % mor e likel y t o liv e i n thes e close-i n neighborhood s (Cortright 2006) Traditionally employmen t opportunities famil y factor s an d housin g ar e th e mos t frequentl y cite d reason s fo r movin g fo r al l generation s (Schachter 2004) However man y youn g people particularl y th e well-educated see m t o b e puttin g a highe r priorit y o n qualit y o f lif e factors Whil e economi c growt h i s stil l a n importan t determinan t o f migration a n analysi s o f movemen t pattern s o f youn g adult s showe d tha t well-educate d person s wer e actuall y mor e likel y t o mov e t o a plac e wit h slowe r jo b growt h tha n th e plac e the y lef t almos t 60 % o f th e tim e (Kodrzycki 2001) Thi s evidenc e buttresse s th e conclusion s o f Florid a (2002) wh o argue d tha t talente d worker s ar e increasingl y draw n t o amenities an d als o tha t o f Glaeser wh o note d tha t th e decisiv e economi c advantag e o f citie s increasingl y derive s fro m th e kind s o f publi c an d privat e consumptio n opportunitie s the y provid e (Glaese r e t al. 2001) Ley' s stud y (1986 ) provide d evidenc e tha t a densit y o f restaurant s an d ar t gallerie s show s a stron g correlatio n wit h urba n regenerations establishin g a foundatio n fo r broade r argument s mad e b y Florid a (2002 ) an d T N Clar k (2004 ) tha t th e growt h o f old dens e citie s ha s bee n significantl y drive n b y th e preference s o f youn g professional s fo r diversit y an d proximit y t o amenity-heav y locales I n addition th e Myer s an d Geari n (2001 ) stud y confirme d th e contributio n o f a n alread y logica l decision-rul e assigne d fo r colleg e students i.e. the y prefe r t o liv e clos e t o th e campu s (which i s thei r plac e o f "work") Europea n studie s als o emphasiz e th e importanc e o f urba n amenities Cheshir e (1995 ) observe d th e demographi c migratio n pattern s i n Europ e an d foun d tha t hig h qualit y citie s (fo r example cathedra l an d universit y towns ) hav e frequentl y gaine d population, primaril y amongs t youn g high-incom e households attracte d b y th e amenitie s o f th e cities Overall th e evidenc e indicate s tha t peopl e i n th e younge r generatio n hav e locationa l preference s fo r th e cor e downtow n area s wher e the y ca n enjo y a hig h qualit y o f lif e wit h a dens e an d divers e se t o f urba n amenitie s i n addition t o th e employmen t an d famil y factors 2 3

PAGE 38

2.4.1. 2 Th e Senior s Urba n resurgenc e generato r group s i n th e senio r populatio n includ e empty-nesters retiree s an d healthy affluent an d sociall y connecte d senio r people Ther e ar e severa l studie s o n th e migratio n pattern s b y thes e activ e senio r groups Valeri o (1997 ) confronte d commo n stereotypica l view s abou t senior s an d argue d tha t rathe r tha n bein g poor sickly penny pinchin g individual s o n th e fring e o f society th e averag e senio r perso n i s a relativel y healthy affluen t consume r o f luxur y good s whos e in-migratio n promise s ne t positiv e return s t o a communit y an d ne t losse s t o th e communit y h e o r sh e leaves Ezel l (2006) coine d th e ter m "nippies, retire d urba n professionals : A rupp y volunteer s a t th e theate r a s a n usher o r serve s o n a tas k forc e fo r helpin g th e homeless o r work s a t a maratho n handin g ou t wate r amon g othe r nonpai d voluntar y jobs Joh n Mcllwai n a t th e Urba n Lan d Institut e argue d tha t peopl e no w headin g int o thei r 60s, a s oppose d t o olde r retirees ar e mor e use d to an d attracte d by th e urba n lifestyl e an d livin g amon g peopl e o f differen t ethni c backgrounds incomes an d age s (Greene 2006) Usin g dat a fro m th e Puge t Soun d Transportatio n panel Krize k an d Waddel l (2002 ) identifie d nin e classification s o f lifestyl e i n a n effor t t o addres s th e interactio n o f dail y activit y participatio n an d trave l pattern s wit h longer-ter m househol d choice s o f vehicl e ownership residentia l location an d employmen t location Tw o lifestyl e groups retiree s an d transi t users hav e preference s o n les s auto dependen t lifestyle s accordin g t o thei r analysis The y estimate d a n increas e i n deman d fo r urba n residence s wit h les s hom e maintenanc e an d improve d rate s o f transi t an d walkin g b y a growin g retire e grou p comprise d o f agin g o f bab y boomers Th e demographi c tren d predict s a hug e increas e o f populatio n i n thes e senio r group s i n th e nea r future Fro m 199 0 t o 2000 th e percentag e o f th e nation' s household s betwee n 5 5 an d 6 4 slippe d slightl y fro m 13.5 % t o 13.2% bu t fro m 200 0 t o 2010 th e percentag e i s projecte d t o jum p t o 17.4 % (Masnic k & Di 2000) Accordin g t o U.S Censu s Burea u projections a substantia l increas e i n th e numbe r o f olde r peopl e wil l occu r durin g th e 201 0 t o 203 0 period afte r th e firs t bab y boomer s tur n 6 5 i n 2011 Afte r 201 0 th e leadin g edg e o f th e boomer s wil l pas s ag e 65 an d growt h amon g th e senio r populatio n wil l substantiall y excee d tha t o f younge r adults a n unprecedente d socia l an d economi c development Thi s i s 2 4

PAGE 39

bes t see n i n th e rati o o f thos e age d 6 5 an d olde r a s compare d t o working-ag e adult s (age d 2 5 t o 64) Afte r decade s o f relativ e stability thi s rati o wil l surg e 30 % i n th e 2010 s an d a furthe r 29 % i n th e 2020 s (Myers 2007) alterin g th e balanc e t o whic h w e hav e lon g bee n accustome d (Myer s & Ryu 2008) I t i s als o predicte d tha t a numbe r o f empt y nester s wil l continu e t o gro w a s bab y boomer s age Afte r thei r childre n leav e home empt y nester s ofte n chang e thei r lifestyl e i n a wa y tha t favor s downtow n living—the y relocat e t o condominium s o r townhouse s an d spen d mor e o f thei r disposabl e incom e o n leisur e activities Thi s chang e i n lifestyl e ma y i n fac t translat e int o livin g i n a downtow n apartmen t an d patronizin g downtow n restaurant s an d cultura l facilitie s (museum s an d concer t halls) I f eve n a modes t portio n o f empty-neste r household s trade s suburba n home s fo r urba n ones i t i s estimate d tha t th e marke t fo r downtow n housin g wil l boom Ther e i s a greate r numbe r o f existin g studie s o n locatio n choic e amon g th e senior s focuse d o n th e regiona l patter n o f migratio n an d th e plac e characteristic s o f th e destinatio n communitie s whe n compare d t o th e numbe r o f studie s o n youn g people' s behaviors Lik e th e migratio n patter n o f th e younge r generations later-lif e migration increase s th e geographi c concentratio n o f th e olde r population Florid a ha s bee n th e mos t importan t receivin g stat e fo r th e pas t thre e decades followe d b y Arizon a an d California Othe r majo r destination s includ e Texas Colorado th e coasta l area s o f th e Southeas t an d th e Pacifi c Northwest th e Ozark s (Arkansa s an d Missouri) an d th e lak e region s o f th e Nort h Centra l states Althoug h retiremen t migratio n fro m th e Nort h t o th e Sunbel t ma y hav e subside d somewha t durin g th e 1980s th e majo r regiona l pattern s establishe d durin g th e pas t fe w decade s hav e continue d t o th e presen t da y (Fuguitt, 1993 ; Golant 1990 ; Graf f & Wiseman 1990 ; Lin 1999 ; Longin o & Fox 1995) Walter s (2000 ) identifie d thre e type s o f migration o f seniors : (1 ) amenit y migration : a searc h fo r attractiv e climat e an d leisur e amenities ; (2 ) assistanc e migration : a searc h fo r residentia l an d economi c dependenc e t o othe r famil y members ; an d (3 ) migratio n i n respons e t o sever e disability : a relocatio n tha t tend s t o resul t i n institutionalizatio n o r othe r share d livin g arrangements Amon g th e thre e type s o f migration h e argue s tha t onl y 2 5

PAGE 40

amenit y migratio n ha s a distinctiv e spatia l pattern ; th e othe r tw o type s d o not Comparabl e t o Walters research D e Jon g e t al (1995 ) conclude d tha t th e decisio n t o liv e wit h famil y member s i s base d primaril y o n factor s othe r tha n disability Specifically the y fin d tha t neithe r initia l disabilit y no r increasin g disabilit y influence s th e odd s o f choosin g kinshi p a s a reaso n fo r moving Cho i (1996 ) use d dat a fro m th e Longitudina l Stud y o f Agin g (LSOA ) t o asses s th e motive s o f non-institutionalize d migrants Cho i foun d tha t th e primar y reason s fo r movin g ar e th e desir e fo r kinshi p (20%) ; financia l problem s (18%) ; th e poo r healt h o f th e responden t (17%) ; th e desir e fo r attractiv e amenitie s (13%) ; an d th e poo r health death o r institutionalizatio n o f a spous e (11%) (Th e lo w proportio n citin g amenit y reason s i s likel y t o hav e resulte d from th e exclusio n o f retiree s younge r tha n ag e 70. ) Investigatin g migration t o an d withi n th e Northeast Shi n (1990 ) foun d tha t senior s i n positive-shif t migration streams thos e wh o mov e fro m countie s o f lo w senio r ne t migratio n t o countie s o f hig h senio r ne t migration ofte n matc h th e profil e o f amenit y migrants : The y ten d t o b e high-income married an d t o liv e i n high-cos t housing Conversely senior s i n negative-shif t migration stream s ofte n exhibi t th e characteristic s expecte d o f assistanc e migrants : The y ten d t o b e low-income widowed an d livin g wit h thei r adul t children Surveyin g nearl y 60 0 retire d migrant s i n Nort h Carolina Haa s (1993 ) an d Haa s an d Sero w (1997 ) reporte d tha t certai n origi n characteristic s ar e consistentl y mentione d a s unattractive Thes e includ e unpleasan t climat e (especiall y amon g higher-incom e migrants) "problem s o f urba n areas (especiall y amon g younge r an d rural-destinatio n migrants) hig h propert y taxes hig h cos t o f livin g (especiall y amon g lower-incom e migrants) an d "fe w o r n o famil y residin g i n area (especiall y amon g femal e migrants) Th e destinatio n characteristic s mos t ofte n mentione d a s attractiv e includ e sceni c beauty "fou r mil d seasons, recreationa l opportunities cultura l amenities an d war m year-roun d climat e (especiall y amon g olde r migrants) Usin g a logi t model Valeri o (1997 ) identifie d seve n plac e characteristic s a s significan t factor s i n potentia l senio r migratio n decisions I n orde r o f importance thes e character s 2 6

PAGE 41

wer e a locality' s (1 ) rat e o f populatio n chang e i n a previou s decad e (prox y fo r friend s an d famil y i n th e area) ; (2 ) numbe r o f col d day s annually ; (3 ) monthl y media n rent ; (4 ) percen t o f populatio n residen t i n urba n areas ; (5 ) mile s o f coastlin e an d squar e mile s o f inlan d water ; (6 ) acre s o f nationa l park s an d numbe r o f stat e parks ; an d (7 ) residentia l propert y tax Interestingly monthl y media n ren t wa s positivel y correlate d wit h th e senio r migratio n decision Valeri o treat s thi s a s a prox y o f urba n amenit y level Existin g studie s o n th e issu e o f senio r migration commonl y lis t th e existenc e o f famil y member s an d friends financia l issue s includin g livin g costs leve l o f natura l an d urba n amenitie s (wit h mor e emphasi s o n th e natura l amenities) an d healt h issue s a s locatio n decisio n factor s o f thi s grou p o f people B y it s ver y nature ther e i s ye t n o empirica l stud y o n th e migratio n pattern s o f bab y boomers—whos e firs t cohor t wil l tur n 6 5 i n 2011-a s a senio r group However i t i s no t difficul t t o estimat e tha t thei r se t o f residentia l preference s i s ver y differen t fro m th e elde r generation' s preference s whe n considerin g thei r financia l affluence environmenta l concerns an d deman d fo r divers e lifestyles A s Florid a (2008 ) described : Wher e boomer s flock bargain s disappear an d th e neighborhoo d butche r sho p i s replace d b y a pan-Asia n fusio n restauran t an d a hardwar e stor e give s wa y t o a high en d remodelin g center Historically bab y boomer s a s a collectiv e grou p hav e a stron g impac t o n th e neighborhood Durin g th e 1970 s an d 1980s thei r passag e int o thei r earl y twentie s spawne d th e firs t stag e o f inner-cit y revitalizatio n an d gentrification Thei r passag e int o th e famil y formatio n an d settlemen t year s o f thei r thirtie s an d fortie s fuele d dramati c single-famil y hom e constructio n an d suburba n growth Now the y ar e enterin g a ne w stag e o f thei r life an d predicting thei r locationa l behavio r become s a n importan t issu e fo r man y scholar s a s wel l a s urba n planners Amon g a fe w existin g studie s o n bab y boomers residentia l locatio n behavior Myer s an d Gearin' s (2001 ) researc h presente d thei r preference s fo r mor e densel y configure d housing s i n mor e centra l locations B y analyzin g demographi c dynamic s an d projections the y estimate d tha t hom e buyer s age d 4 5 an d olde r wh o prefe r denser mor e compac t housin g 2 7

PAGE 42

alternative s wil l accoun t fo r 31 % o f tota l homeowne r growt h durin g th e 2000-1 0 period doubl e th e sam e segment' s marke t shar e i n th e 1990s 2.4. 2 Socia l Network s Amon g Demographi c Group s 2.4.2. 1 Socia l Interaction s an d Conflict s Mainl y du e t o smal l siz e an d th e resultin g hig h possibilit y o f dail y encounter s betwee n member s o f a residentia l community socia l interaction s i n a communit y ma y generat e externalitie s tha t ca n b e significan t i n a housin g locatio n decisio n process Ther e hav e bee n severa l studie s o n th e impac t o f socia l interaction s o n residentia l mobility Brueckne r e t al (1996 ) argue d tha t endogenou s an d exogenou s amenitie s (whic h ar e form s o f interactions ) hav e a centra l effec t o n th e distributio n o f rich an d poo r household s betwee n differen t locations Ka n (2007) usin g Pane l Stud y o f Incom e Dynamic s (PSID ) data foun d tha t a household' s possessio n o f loca l socia l capita l ha s a negativ e effec t o n it s residentia l mobility an d thi s negativ e effec t o f loca l socia l capita l ma y b e stronge r o n long-distanc e mobilit y tha n o n short-distanc e mobility B y analyzin g th e existin g literatur e o n th e relationshi p betwee n socia l networ k an d migration decision h e identifie d severa l potentia l channel s throug h whic h th e possibl e socia l network s i n a destinatio n facilitate s migration : (1 ) member s o f one' s socia l network s ar e a sourc e o f materia l ai d (e.g. accommodation) ; (2 ) emotiona l suppor t (e.g. encouragement) whic h i s importan t fo r ne w immigrants ca n b e derive d fro m one' s social networks ; an d (3 ) on e ma y obtai n importan t informatio n (e.g. livin g environment jo b opportunities etc. ) fro m one' s socia l networks Th e genera l consensu s amon g existin g studie s i s tha t sam e ethni c o r incom e grou p interactio n strengthen s th e socia l networ k an d decrease s migration decision bu t inter grou p interaction s weake n th e socia l networ k an d increas e th e mobility However whil e ther e ar e man y existin g studie s o n th e socia l conflict s betwee n existin g resident s an d ne w resident s i n renovate d o r redevelope d urba n areas academi c researc h o n th e relationshi p 2 8

PAGE 43

betwee n ne w residen t group s (i.e. ol d an d young ) i s scarc e mainl y becaus e o f it s novelty Instead a n articl e b y Casselma n i n th e Wall Street Journal from Ma y 11 200 7 reporte d th e instance s o f generationa l discordance : "On e tim e I wen t u p ther e an d th e twenty-something s ha d th e whol e plac e monopolized, sh e recalls "an d I thought well no t today. Ms Lamme l say s sh e an d som e o f he r cohort s hav e a strateg y fo r reclaimin g th e space a t leas t temporarily : They'r e plannin g a covered-dis h poo l party "Anyon e i s welcome, sh e say s i n he r pleasan t Souther n drawl "Bu t we'l l se e wh o show s up. ...an d man y o f th e youn g buyer s wan t thei r neighbor s t o b e mor e lik e them Rick y Florita a 29-year-ol d mortgag e banke r i n Nashville say s h e avoide d buyin g i n Viridia n i n par t becaus e h e hear d i t wa s attractin g "a n olde r crowd. Instead h e signe d a contrac t fo r a $160,00 0 cond o i n Icon anothe r projec t b y Bristo l Developmen t Group tha t wil l featur e a medi a loung e an d a poo l plaz a wit h "grillin g cabanas whe n i t open s nex t year I reall y thin k it' s goin g t o b e a single s scene, Mr Florit a says "Ever y tim e yo u wer e i n th e sale s center yo u sa w reall y attractiv e wome n buyin g thes e condos. Ther e ar e als o numerou s testimonie s fro m developer s abou t th e unexpecte d emergenc e o f senio r peopl e withi n thei r developmen t project s tha t wer e originall y intende d t o targe t younge r generations : "M y buyer s ar e bi t th e youn g urba n professional s wh o starte d th e lof t conversio n busines s i n Ne w Yor k an d Chicago Wha t I ge t ar e affluen t empty-nester s wh o ar e tire d o f constantl y drivin g int o th e cit y fo r entertainment. Lewi s Kostiner Develope r (fro m Lurz 1999 ) "Ou r targe t marke t wa s youn g militar y families That' s wh y w e worke d ver y har d t o kee p ou r pric e poin t low W e hi t th e target bu t we'v e als o attracte d a lo t o f empty neste r retiree s fro m ou t o f state especiall y i n th e lat e stage s o f th e project a s th e tow n reall y cam e int o it s own Th e ne w tow n hal l wa s unde r constructio n whe n w e started Th e schoo l ha s no w bee n completel y remodeled There' s a new historic-style d pos t offic e an d a ne w fir e station It' s reall y amazin g whe n yo u conside r tha t ou r firs t tw o house s wer e sol d whil e lookin g a t old rundow n trailer s acros s di e street Th e retiree s lik e th e ide a o f bein g i n town wher e the y ca n wal k mos t o f th e place s the y wan t t o go. Rober t Turner Develope r (fro m Lurz 1999 ) A yea r befor e di e buildin g i s se t t o open jus t 30 % o f th e units—studio s star t a t $885,000~hav e sol d an d thos e ma t hav e sol d haven' t necessaril y gon e t o th e intende d demographic Early buyer s hav e bee n quit e a mixe d grou p wit h a wid e rang e o f ages It' s no t a s youn g a s I thought. Andr e Balazs Develope r (fro m Casselman 2007 ) 2 9

PAGE 44

"bu t it' s no t s o eas y t o contro l demographic s i n th e ope n market Som e o f th e building s ar e drawin g unexpecte d buyers : peopl e ol d enoug h t o b e th e parent s o f th e kid s dow n th e hall An d that' s leadin g t o territoria l conflicts socia l snubs-eve n planne d boardroo m coups. Le e Schaefer Develope r (fro m Casselman 2007 ) On e o f th e possibl e reason s fo r thi s ne w entr y o f th e senio r peopl e int o a rea l estat e developmen t tha t ha d originall y bee n targete d t o younge r generation s i s tha t thes e group s ar e th e tw o mai n consume r group s o f urba n lifestyle s an d amenities However i t i s als o plausibl e t o assum e tha t developer s di d no t envisag e thi s emergin g senio r market Consequentially ther e ha s bee n a limite d suppl y o f urba n residentia l product s targetin g thi s affluen t group Anothe r plausibl e explanatio n i s tha t th e senior s prefe r t o b e locate d nea r younge r generations A s Florid a (2008 ) described fo r man y empty-nester s an d retirees a ke y facto r i n thei r locatio n choice—an d i n almos t everythin g the y do~i s proximit y t o thei r childre n an d grandchildren Whil e childre n ma y retur n hom e afte r colleg e o r whe n thei r parent s becom e ill a n increasin g tren d i s fo r parents especially thos e wit h means t o follo w thei r kids Whateve r th e reason s ar e fo r thi s generationa l spatia l overlapping i t i s anticipate d tha t instance s o f generationa l conflic t betwee n thes e tw o demographi c group s wil l becom e mor e commo n a s lon g a s the y ar e competin g fo r th e limite d amoun t o f spac e downtown 2.4.2. 2 Neighborhoo d Effect s Researcher s hav e consistentl y argue d tha t neighborhood s influenc e a household' s residentia l locatio n decision Schelling' s segregatio n mode l (1971 ) wa s th e firs t attemp t t o measur e th e impac t o f th e individua l neighborhoo d preference s o n th e spatia l patter n o f demographi c distribution However Schelling' s conceptua l mode l ha s mostl y bee n explore d i n hypothetica l settings an d ther e wa s a limite d attemp t t o appl y hi s mode l t o a practica l application Mor e recentl y an d mor e empirically Cutle r an d Glaese r (1997 ) tracke d racia l segregatio n i n U S metropolita n area s fro m th e lat e 19t h centur y t o 1990s Base d o n th e dissimilarit y index segregatio n increase d continuousl y unti l th e 1970 s bu t 3 0

PAGE 45

ha s sinc e declined Th e declin e i s primaril y attributabl e t o a movemen t o f blac k household s outward s from cit y center s int o suburb s tha t wer e formerl y white-dominated B y contrast Abramso n e t al (1994 ) foun d that betwee n 197 0 an d 1990 incom e segregatio n increase d i n U S metropolita n areas Th e dissimilarit y inde x ros e b y 11 % fo r poo r household s an d th e isolatio n inde x increase d b y nin e percent va n Ha m an d Feijte n (2008 ) explore d th e influenc e o f th e neighborhoo d characteristic s (percentag e o f rente d dwellings low-incom e households an d ethni c minoritie s i n th e neighborhood ) o n th e differen t categorie s o f residents wis h t o leav e thei r neighborhood Thei r mai n resul t showe d that wit h a n increasin g percentag e o f peopl e fro m a n ethni c minorit y i n th e neighborhood mor e peopl e hav e th e wis h t o leav e th e neighborhood However thi s i s t o a lesse r exten t th e cas e fo r member s o f ethni c minoritie s themselves A t th e sam e time divers e measure s o f segregatio n hav e bee n developed : evennes s (dissimilarity) exposur e (isolation) concentratio n (th e amoun t o f physica l spac e occupie d b y th e minorit y group) clusterin g (th e exten t t o whic h minorit y neighborhood s ar e contiguous) an d centralizatio n (proximit y t o th e cit y centre ) (se e Cutle r e t al. 1999 fo r a mor e detaile d discussion) Ag e segregatio n o f residentia l neighborhood s receive d som e researc h attentio n i n th e 1970 s an d 1980 s (Chevan 1982 ; Cowgill 1978 ; Pampe l & Choldin 1978 ; War d e t al. 1985 ) bu t ha s largely bee n ignore d sinc e then A possibl e reaso n fo r thi s pea k ma y hav e bee n th e event s o f th e lat e 1960 s an d earl y 1970s : a perio d wit h muc h discussio n o f "generatio n gaps an d slogan s suc h a s "Neve r trus t anyon e ove r 30. I t wa s als o a tim e o f increase d sensitivit y toward s "isms tha t lea d t o separatio n an d exclusion startin g wit h th e awarenes s o f racis m an d a cal l fo r civi l rights (Hagesta d & Uhlenberg 2005) I t i s puzzling however tha t th e literatur e o n ag e segregatio n neve r combine d a n interes t i n th e potentia l conflic t betwee n th e youn g an d old Indeed publication s relate d t o th e separatio n o f youn g peopl e hav e a differen t perspectiv e tha n thos e relate d t o olde r people Th e literatur e o n ag e segregatio n o f childre n an d yout h take s a "socia l problems perspective emphasizin g th e cost s o f separation However th e literatur e o n segregatio n o f olde r peopl e tend s t o emphasiz e th e benefit s o f separation particularl y residential I n th e firs t case on e 3 1

PAGE 46

finds discussion s o f juvenil e delinquency troubled families an d childre n wit h behavio r problems I n th e secon d case security, simplifie d servic e delivery an d eas y acces s t o peer s ar e stressed Th e studie s fin d moderat e level s o f segregatio n betwee n th e olde r an d younge r populations The y als o repor t a n agin g o f th e suburbs whic h migh t hav e increase d th e potentia l fo r youn g peopl e t o interac t wit h olde r individual s (Fitzpatric k & Logan 1985) Bu t non e o f thi s researc h wa s abl e t o demonstrat e tha t residentia l ag e segregatio n mad e muc h differenc e i n th e live s o f ol d an d young Sinc e 1990 a grea t dea l o f attentio n ha s bee n give n t o th e effect s o f neighborhoo d o n th e well-bein g o f adolescent s (e.g. Sampso n e t al. 2002) Th e concep t o f socia l capital whic h emphasize s th e critica l rol e o f socia l ties i s suggeste d a s on e o f th e primar y mechanism s linkin g neighborhood s t o individua l outcomes However neighborhoo d ag e compositio n ha s no t bee n on e o f th e centra l characteristic s included i n thes e studies Thus littl e i s know n abou t curren t level s o f ag e segregatio n i n neighborhoods o r th e implication s o f residentia l segregatio n fo r cross-ag e perception s an d interactions Althoug h ther e hav e bee n severa l attempt s t o measur e th e neighborhoo d effect s o n a household' s residentia l locatio n decision relativel y littl e i s know n abou t th e natur e o f th e neighborhoo d effects Ther e ar e severa l possibl e explanations First th e neighborhoo d effect s tha t ma y resul t i n a household' s migratio n decisio n ca n b e generate d fro m a serie s o f interaction s wit h neighbor s o f differen t cultures Second the y als o ca n b e originated fro m th e existin g prejudic e agains t differen t demographi c group s an d thei r cultures Mor e plausibly the y possibl y cam e fro m both : negativ e interaction s tha t reinforc e prejudice I f negativ e interaction s ar e th e mai n reason s fo r migratio n decisions toleranc e wil l b e on e o f th e ke y factor s i n controllin g mobility : a communit y wit h highl y toleran t peopl e wil l hav e lo w mobilit y rat e an d represen t a hig h leve l o f diversity Whereas i f th e prejudic e amon g peopl e i s th e mai n reason stayin g tim e i n a communit y wil l b e on e o f th e ke y factors : accumulatio n o f positiv e interaction s wit h differen t culture s (o r peopl e i n differen t cultures ) amon g neighbor s wil l resul t i n a lo w mobilit y rat e an d represen t a hig h leve l o f diversity A s Allpor t (1954 ) pointe d ou t i n hi s classi c study a ke y weapo n agains t 3 2

PAGE 47

stereotype s an d prejudice s i s intergrou p contact whic h allow s individual s th e opportunit y t o challeng e homogenize d categorie s an d se e beyon d stigmatize d characteristic s t o othe r relevan t qualitie s o f person s i n a pivota l category 2. 5 Th e Rol e o f th e Loca l Government s Existin g researc h o n th e effect s o f th e rol e o f th e loca l policie s o n th e residentia l locatio n decisio n show s ambiguou s results Friedma n (1981 ) employe d maximum-likelihoo d estimatio n o f a multinomia l logi t t o examin e th e effect s o f loca l publi c service s an d othe r communit y attribute s o n residentia l locatio n decisions b y families Th e mode l wa s estimate d fo r si x subsample s differin g b y househol d size income an d ag e o f hea d o f household Fo r mos t o f th e subsample s tha t wer e examined th e conclusio n wa s tha t loca l publi c service s an d othe r communit y characteristic s pla y onl y mino r role s i n determinin g residentia l locatio n choice Uya r an d Brow n (2005 ) employe d McFadden' s (1978 ) discret e choic e mode l t o tes t fo r th e significanc e o f dwelling-specifi c loca l taxe s an d publi c service s o n househol d locatio n decision s withi n a singl e taxin g jurisdiction Thei r finding s indicate d tha t suc h variable s ar e significan t determinant s o f locatio n decisions eve n withi n a singl e taxin g jurisdiction an d shoul d no t b e assume d away Bayo h e t al (2006 ) employe d a hybri d conditiona l logi t choic e mode l usin g dat a o n th e characteristic s an d destinatio n o f homeowner s wh o engage d i n intra-metropolita n move s amon g 1 7 schoo l district s withi n th e Columbus Ohio Th e mode l wa s use d t o tes t th e relativ e influenc e o f loca l fisca l an d publi c good s versu s household-leve l characteristic s i n determinin g househol d locatio n choice s acros s centra l cit y an d suburba n schoo l districts Result s provide d evidenc e o f bot h a "natura l evolution o f household s t o th e suburbs du e t o jo b location residentia l filtering an d househol d incom e an d lifecycl e effects an d "fligh t fro m blight, du e t o lowe r schoo l quality highe r crim e levels an d lowe r averag e incom e level s i n th e city I n comparin g th e magnitude s o f thes e variables the y foun d tha t schoo l qualit y exerte d th e stronges t influence : a one 3 3

PAGE 48

percen t increas e i n th e schoo l qualit y o f th e cit y distric t increase d th e probabilit y o f choosin g a cit y residenc e b y 3.7% I n contrast th e effect s o f househol d incom e an d othe r individua l characteristic s wer e relativel y modest Th e finding s provide d suppor t fo r a "fligh t fro m blight suburbanizatio n proces s tha t wa s dominate d b y difference s i n neighborhoo d qualit y betwee n th e cit y an d suburbs Publi c effort s t o revitaliz e declinin g centra l citie s hav e a relativel y lon g history Th e U.S Federa l Urba n Renewa l (the n calle d Urba n Redevelopment ) Progra m wa s launche d unde r Titl e I o f th e Housin g Ac t o f 1947 Althoug h th e progra m wa s intende d t o "[provide ] mor e an d bette r housin g throug h th e spo t remova l o f residentia l slums (Keyes 1969) th e nex t tw o decade s witnesse d th e replacemen t o f peopl e an d thei r home s wit h comprehensivel y planne d structure s (mostl y offic e buildings commercia l complexes an d luxur y housing ) and al l to o often empt y tract s (Y Zhan g & Fang 2004) Moreover urba n renewa l program s wer e criticize d fo r simpl y pushin g slu m dweller s (mostl y African-American s an d othe r minorities ) t o othe r part s o f th e cit y o r t o th e suburbs thereb y exacerbatin g th e ill s th e progra m sough t t o solv e (Gans 1968) Mor e recently however loca l municipalitie s hav e adapte d innovativ e inne r cit y revitalizatio n program s t o achiev e severa l goals : raisin g propert y value s an d taxes increasin g sale s ta x revenues stemmin g crime creatin g constructio n an d permanen t jobs improvin g civi c image an d attractin g tourists Thei r program s includ e cas h contributions financin g aid ta x abatements zonin g an d buildin g cod e variances ne w o r improve d infrastructure implementatio n o f growt h boundaries an d th e creatio n o f dedicate d taxe s fo r physica l improvements Additionally som e citie s us e neighborhoo d conservation program s t o suppor t existin g communitie s an d t o maintai n thei r desirability Som e citie s partne r wit h nonprofi t communit y developmen t corporation s t o kick-star t housin g productio n i n laggin g markets Som e busines s improvemen t district s (BEDs ) activel y promot e an d facilitat e housin g redevelopmen t an d ne w constructio n (Haughey 2002) Th e impac t o f governmenta l program s i s difficul t t o generalize becaus e successfu l case s ar e location-specific Whil e som e plannin g policie s an d regulation s ar e generatin g th e resurgenc e o f th e downtow n (i.e. Seattl e an d Portland' s state-mandate d growt h control s 3 4

PAGE 49

whic h promote d developmen t insid e th e urba n growt h boundaries) th e resurgenc e o f othe r downtow n area s i n citie s lik e Ne w Yor k an d Chicag o ar e mainl y market-driven Consequently th e studie s o n th e effectivenes s o f th e variou s governmenta l program s ar e scarce Mainl y du e t o th e absenc e o f researc h o n th e effectivenes s o f th e governmenta l program s t o generat e inne r cit y revitalization man y loca l government s inves t thei r budge t int o th e symboli c mega-constructio n project s i n downtown s hopin g thes e project s wil l attrac t peopl e int o th e areas However ther e ar e increasin g number s o f discussion s tha t th e functio n o f citie s an d th e characteristic s o f citizen s hav e changed Cities reaso n fo r being — an d thei r residents reaso n fo r livin g withi n them—ma y b e th e ne w lifestyl e tha t onl y dens e an d divers e urba n environmen t ca n provide A s Florid a (2002 2005a 2005b 2008 ) argue d i n a serie s o f hi s "creativ e class books th e physica l attraction s tha t mos t citie s focu s o n building—sport s stadiums freeways urba n mall s an d touris m an d entertainmen t district s tha t resembl e them e parks—ar e probabl y b e irrelevant insufficien t o r actuall y unattractiv e t o man y urba n dwellers Instead wha t the y loo k fo r i n communitie s ma y b e high-qualit y amenitie s an d experiences a n opennes s t o diversit y o f al l kinds an d abov e al l els e th e opportunit y t o validat e thei r identitie s (Florida 2008) Mor e researc h o n th e rol e o f loca l governmen t tha t control s th e leve l o f urba n amenit y b y regulatin g thei r types amounts an d distribution s ar e required 3 5

PAGE 50

3 THEOR Y AN D APPLICATION S O F URBA N MODELIN G A n urba n mode l i s a n abstracte d representatio n o f a "real-urban system I n general model s ac t a s a vehicl e t o enabl e experimentatio n wit h theor y i n a predictiv e sens e an d t o enhanc e understandin g whic h ma y b e prio r t o prediction s o f situation s a s ye t unrealized fo r example i n th e futur e (Batty 2009) A mor e conventiona l definitio n o f urba n model s i s th e following : Representation s o f function s an d processe s whic h generat e urba n structur e i n term s o f lan d use population employmen t an d transportation usuall y embodie s i n compute r program s tha t enabl e locatio n theorie s t o b e teste d agains t dat a an d prediction s o f futur e locationa l pattern s t o b e generate d Internationa l Encyclopedi a o f Huma n Geograph y (Thrif t & Kitchi n (eds.) 2009 ) I n constructin g urba n model s (ac t o f urba n modeling) th e assume d "rea l urban i s broke n dow n int o a for m amenabl e t o analysis b y representin g wit h onl y thos e variable s tha t trul y affec t th e behavio r o f th e urba n syste m an d b y clarifyin g th e relationship s betwee n thos e variable s (Taha 2003) Simulatio n i s a n ac t o f runnin g constructe d models Accordin g t o Decke r (1993) "A n (urban ) simulatio n .. ca n serv e a s a n accessibl e surrogat e fo r th e city' s comple x systems extensiv e spatia l structure o r environmenta l influences. Th e abilit y t o clearl y se e an d understan d alternativ e developmen t scenario s ha s alway s bee n a goa l o f plannin g (Simpson 2001) Th e comple x natur e o f cities however wit h dynamicall y changin g parameter s an d larg e number s o f independent/dependen t actor s mak e i t difficul t t o estimat e th e result s o f plannin g policies an d ther e hav e bee n request s t o adap t a simulatio n approac h t o urba n syste m fo r a lon g perio d o f time I n thi s chapter a revie w o f th e developmen t o f computer-base d modelin g an d simulatio n i n urba n plannin g fiel d i s provided Specifi c interest s ar e pai d t o th e application s o f agent base d approache s i n residentia l locatio n decisio n models 3 6

PAGE 51

3. 1 Historica l Developmen t 3.1. 1 Emergenc e o f Urban Model s I n term s o f cities th e kind s o f urba n theor y tha t ar e basi c t o th e developmen t o f compute r model s ar e thos e tha t ar e traditionall y calle d locatio n theories : Theorie s tha t propos e mechanism s tha t enabl e industries service s an d household s t o locat e i n spac e wit h economi c constraints o f incom e an d profitability I n turn thes e economie s ar e conditione d primaril y b y distanc e betwee n land-use s associate d wit h thes e activities whic h depen d upo n a rang e o f marke t condition s essentiall y underpinne d b y trade Thu s distanc e an d movemen t ar e centra l t o suc h theor y (Batty 2009) I t ha s bee n ove r 5 0 year s sinc e compute r model s wer e firs t develope d i n th e urba n domai n (B Harris 1965) Th e firs t generatio n o f urba n model s treate d th e urba n syste m a s a stati c entit y whos e lan d use s an d activitie s wer e t o b e simulate d a s a cross-sectio n i n tim e an d whos e dynamic s wer e largel y regarde d a s self-equilibratin g (Batty 2009) Earl y urba n modeler s includ e Isar d (1960 ) an d Alons o (1964) wh o presente d model s base d o n spatia l interactio n idea s fro m socia l physic s an d macro-economi c model s suc h a s input-outpu t analysis Thes e urba n model s generall y deal t wit h lan d us e an d lan d price bu t man y urba n component s suc h a s population, job services an d transportatio n wer e beyon d thes e models framewor k (Benenso n & Torrens 2004) I t wa s Lowr y (1964 ) wh o firs t attempte d t o relat e suc h urba n component s int o th e mode l (Lowry 1964) I n hi s model th e urba n syste m wa s structure d i n a simpl e way wit h thre e activit y sectors : (1 ) a basi c sector includin g industrial business an d administrativ e activities whos e client s ar e mostl y non-local ; (2 ) a retai l sector dealin g wit h th e loca l population ; an d (3 ) a househol d sector Th e basi c assumptio n o f th e mode l wa s base d o n th e gravit y model whic h i s a simpl e concept : I f th e distanc e betwee n tw o point s equal s d the n an y interactio n betwee n th e object s locate d a t thes e point s i s inversel y proportiona l t o som e powe r o f d Eve n thoug h th e implementatio n o f th e mode l t o th e Cit y o f Pittsburg h 3 7

PAGE 52

wa s quit e successful Lowry' s mode l wa s static whic h mean s tha t citie s i n thi s mode l ca n b e see n a s largel y unchanging ; thi s assumptio n i s unrealisti c i n man y ways Th e first dynami c urba n mode l wa s develope d b y Forreste r (1969) H e sa w positiv e feedback s a s th e mai n sourc e o f comple x an d counterintuitiv e behavio r o f natura l system s i n genera l an d urba n flavor s i n particular H e propose d tha t on e positiv e feedbac k dominate s th e syste m fo r a tim e an d the n thi s dominanc e i s shifte d t o anothe r feedback responsibl e fo r th e othe r par t o f th e system Th e behavio r o f th e syste m change s s o muc h wit h suc h a shif t tha t th e tw o regime s seem s unrelated A t th e sam e time whil e on e o f th e feedback s dominates th e syste m maintain s resistanc e t o th e other thu s marginal changes Forrester' s wor k mad e som e meaningfu l contributio n t o th e fiel d o f urba n modelin g an d simulatio n suc h a s us e o f compute r a s th e mai n too l fo r investigatin g th e behavio r o f comple x urba n system ; emphasi s o f mode l structur e an d relationshi p betwee n seemingl y independen t urba n phenomena ; incorporat e a s fe w variable s a s possibl e int o a mode l t o reduc e th e volum e o f possibility A t th e sam e time hi s works however generate d a stor m o f criticism Th e majo r criticis m o f Forrester' s wor k i s o n th e complet e ignoranc e o f geographi c theor y an d non-spatia l aspec t o f hi s mode l (K Chen 1972) Explici t incorporatio n o f geographi c knowledg e int o th e model s wa s soo n developed bu t th e negativ e attitud e remaine d an d could perhaps b e on e o f th e reason s behin d th e genera l retrea t o f geographi c an d urba n scienc e fro m quantitativ e method s an d modelin g durin g th e 1970 s an d 1980 s (Benenso n & Torrens 2004) Limite d number s o f urba n modelin g earl y researc h wer e develope d withi n th e integrate d regiona l framewor k tha t merge s th e model s o f Lowr y an d Forrester an d the y followe d th e traditio n o f mathematica l ecolog y an d econom y (Day 1982 ; Dendrino s & Mullally 1985) Othe r studie s includ e th e studie s o f competitio n betwee n tw o socia l group s fo r spac e (W Zhang 1989 ) o r betwee n economi c sector s (Dendrino s & Sonis 1990 ; W Zhang 1993 ; W Zhang 1994) an d explanatio n o f emergin g urba n hierarch y (Rosser 1994) On e meaningfu l approac h o f tha t tim e b y a grou p o f modeler s a t th e Nationa l Burea u o f Economi c Researc h wa s Th e Detroi t Prototyp e o f th e NBE R Urba n Simulatio n Model whic h focuse d o n th e housin g locatio n decisio n base d o n acces s t o employment a 3 8

PAGE 53

disaggregate d equatio n fo r differen t househol d type s an d differen t housin g type s an d spatiall y locate d zone s (Ingra m e t al. 1972) Eve n thoug h thi s mode l wa s no t fre e fro m th e criticism modelin g approache s o f thi s mode l wer e base d o n individual marke t processes an d develope r decisio n provide d theoretica l an d methodologica l backgroun d fo r th e arriva l o f agent-base d models 3.1. 2 Larg e Scal e Mode l Critique s Traditiona l urba n models develope d i n th e styl e o f th e spatia l interactio n model wer e pioneere d a t a tim e i n whic h th e field o f urba n simulatio n wa s radicall y differen t fro m curren t situation Computin g powe r o f tha t tim e wa s relativel y les s accessibl e an d sophisticate d tha n i t i s toda y an d detaile d dat a set s wer e no t widel y availabl e (Torrens 2001) Thos e traditiona l urba n model s wer e criticize d heavil y (Lee 1973 ; 1994 ; Sayer 1979) an d on e o f th e mos t famou s criticism s o n urba n comprehensiv e modelin g cam e fro m Le e i n 1973 I n hi s "Requie m fo r Large-Scal e Models Le e challenge d al l thre e widel y know n urba n model s a t tha t time : th e Forreste r model PLU M (a n evolutionar y o f th e Lowr y model) an d NBE R an d liste d seve n sin s o f large-scal e models Hi s seve n sin s o f th e large-scal e model s are : (i ) hypercomprehensiveness th e model s attemp t t o desig n to o o f a comple x syste m i n a singl e sho t an d attemp t t o serv e to o man y purpose s a t th e sam e time ; (ii ) grossness coars e leve l o f aggregate d dat a t o b e use d b y mos t polic y makers ; (iii ) hungriness th e model' s tremendou s requiremen t o f data ; (iv ) wrongheadedness th e us e o f to o man y variable s tha t eve n th e mode l builde r migh t no t perceiv e an d th e resultin g deviatio n betwee n claime d mode l behavio r an d th e equation s o r statement s tha t actuall y gover n mode l behavior ; (v ) complicatedness th e difficult y o f understandin g "blac k box mode l behavio r mainl y du e t o multicollinearit y an d misspecificatio n o f variables ; (vi ) mechanicalness intrinsi c procedura l an d numerica l error s i n computing ; an d (vii ) expensiveness th e hig h pric e o f dat a collectio n an d paramete r estimation 3 9

PAGE 54

Eve n thoug h th e developmen t o f micro-computatio n technolog y an d Geographi c Informatio n Syste m (GIS ) negate d som e o f hi s criticism s an d enable d a cost-effectiv e developmen t o f urba n models Lee' s criticism s o f th e large-scal e mode l ar e stil l vali d an d highl y relevan t today particularl y whe n contraste d wit h newe r model s currentl y bein g develope d i n academi c contexts : thei r centralize d approach a poo r treatmen t o f dynamics wea k attentio n t o detail shortcoming s i n usability reduce d flexibility an d lac k o f realis m (Torrens 2001) I n th e nex t section recen t developmen t o f modelin g an d simulatio n technique s tha t offe r th e potentia l fo r improvin g th e usabilit y o f traditiona l model s wil l b e explored 3.1. 3 Lan d Use-Transportatio n Model s Sinc e th e earl y effort s t o construc t operationa l urba n model s wer e mostl y base d o n socia l physica l paradig m (th e gravit y model) ther e hav e bee n severa l attempt s t o embrac e theorie s fro m a wid e rang e o f urba n sciences Transportatio n engineers urba n economists socia l scientists an d geographer s trie d t o incorporat e lan d use-transportatio n interaction s (tri p an d locatio n decision s codetermin e eac h other) economi c force s o f cit y growt h (citie s ar e system s o f marke t i n whic h household s tr y t o matc h thei r spac e need s an d locatio n preference s wit h thei r budge t restrictions) an d socia l theorie s o f urba n developmen t (th e spatia l developmen t o f citie s i s th e resul t o f individua l o r collectiv e appropriatio n o f space ) int o th e urba n models Consequently man y recentl y develope d urba n model s ar e buil t aroun d th e them e o f "land us e transportatio n feedbac k cycle. The y includ e th e following : th e Californi a Urba n Future s (CUF ) mode l develope d a t th e Universit y o f Californi a a t Berkele y (Landi s & M Zhang 1998) ; DELTA th e lan d use/economi c modelin g packag e b y Simmond s Consultanc y (Simmonds 1999 2001) ; th e Integrate d Lan d Use Transportation Environmen t (ILUTE ) modelin g syste m develope d a t severa l Canadia n universitie s (Mille r & Salvini 2001) ; th e integrate d modelin g packag e develope d b y Echeniqu e an d Partner s 4 0

PAGE 55

(MEPLAN ) (William s 1994 ; Hun t & Simmonds 1993) ; th e lan d us e transportatio n mode l develope d i n th e Orego n Transportatio n an d Lan d Us e Mode l Integratio n Progra m (TLUMIP ) (ODOT 2001) ; th e transportatio n an d lan d us e mode l (TRANUS ) develope d b y d e la Barr a (1989) ; an d UrbanSim~th e microsimulatio n mode l o f locatio n choic e o f household s an d firm s a t th e Universit y o f Californi a Berkele y (Waddell 1998 2001) Mos t o f th e model s introduce d her e simulat e beyon d jus t lan d us e an d huma n transportatio n interactions : The y forecas t housin g stock nonresidentia l buildin g stock an d good s transportatio n a s well Al l o f th e model s her e assum e tha t th e transportatio n i s alway s i n equilibrium i.e. tha t trave l flow s reflec t trave l time s an d cost s o n th e network However som e model s (CUF ILUTE an d UrbanSim ) assum e tha t th e lan d us e syste m i s dynami c (o r quasi-dynami c a s the y wor k wit h discret e tim e periods ) i n tha t the y explicitl y mode l th e adjustmen t processe s ove r time Thes e model s ar e base d o n th e assumptio n tha t som e adjustmen t processe s ar e faste r tha n other s an d tha t th e difference s i n spee d ar e s o larg e tha t urba n system s ar e normall y i n disequilibriu m (Wagener 2004) Finally som e model s (ILUTE TLUMIP an d UrbanSim ) appl y microsimulatio n o f lan d us e mode l base d o n individua l househol d o r fir m leve l behavior 3.1. 4 Plannin g Suppor t System s Wit h th e developmen t o f informatio n scienc e wit h dat a an d Electroni c Dat a Processin g (EDP ) i n th e 1960s informatio n an d Managemen t Informatio n System s (MIS ) i n th e 1970s an d knowledg e an d Decisio n Suppor t System s (DSS ) i n th e 1980s Plannin g Suppor t Syste m (PSS ) fro m th e 1990 s becam e a popula r topi c i n urba n plannin g fiel d (Klosterman 2001) Ther e hav e bee n severa l meaningfu l modelin g effort s i n developin g computer-assiste d plannin g tools Thes e effort s includ e th e following : METROPILUS a lan d us e mode l tha t ha s root s i n th e earl y mode l o f Lowr y develope d b y Putma n (1995) ; INDEX a GIS-base d PS S tha t estimates th e potentia l impac t o f communit y lan d us e an d desig n decision s develope d b y Criterio n Planner s (Allen 2000) ; Wha t if? a scenario 4 1

PAGE 56

based policy-oriente d PS S develope d b y Klosterma n (2007) ; an d CommunityVIZ a GIS base d PS S fo r communit y plannin g an d desig n application s develope d b y th e Orto n Famil y Foundatio n (Kwartler 1998) Som e urba n model s buil t aroun d th e concep t o f land-us e transportatio n interaction s ar e als o included i n thi s PS S category : CU F (vector-base d model) CU F I I (developmen t o f CU F bu t ha s a raste r dat a structure) an d CURB A (focuse s o n th e interaction s betwee n lan d us e chang e an d habita t loss) TRANU S an d UrbanSim Whil e ther e i s a rapi d an d increasin g numbe r o f th e developmen t o f PS S tools ther e ar e a limite d numbe r o f successfull y applie d PSS s i n rea l worl d plannin g practices The y are t o practitioner s i n general to o rigid to o theoretical to o costly un-dynamic unrealistic an d unreliable Mor e effort s o n th e developmen t o f urba n modelin g too l tha t hel p planner s i n anticipator y decisio n makin g situations i.e. plannin g situations ar e required 4 2

PAGE 57

Tabl e 3. 1 Compariso n o f th e urba n models' 1 CUFI I CURBR A INDE X Mode l Typ e Lan d us e chang e Urba n growt h GIS urba n impac t Themati c Scop e Urba n simulatio n Urba n growth environmenta l an d ecologica l qualit y Lan d use transportation housing employment natura l environmen t Spatia l Resolutio n One-hectar e One-hectar e Use r define d Tempora l Resolutio n Custo m Custo m Use r define d Use r Non-technica l plannin g Lan d us e planners participant s polic y makers environmentalist s Non-technica l plannin g participant s Operationa l Method s Logit regressio n Logit regressio n Causa l inference correlation linea r programming networ k analysis time-serie s Urba n Lan d Us e Single mult i family N o categor y Categorie s Commercial Industrial Developmen t Use r define d Non-Urba n Lan d Us e Categorie s Agricultural Forest Wetland Water Preservation Par k Agricultural Forest Wetland Water Preservation Par k Use r define d Independen t Variable s o n Lan d Us e Pattern s Transportation zoning Transportation zoning maste r plan impac t fee maste r pla n sewe r an d wate r fe e Zoning maste r pla n 4 3

PAGE 58

Table3.1(Cont. ) mmmm^mmm Wha t if ? UrbanSi m MEPLA N Mode l Typ e GI S GIS urba n economic/lan d us e market logit hedoni c Trave l demand urba n economic/lan d us e market hedoni c Themati c Scop e Lan d us e evaluatio n an d Lan d use transportation Spatia l economic-base d chang e analysi s economics I/ O environmenta l impact s Spatia l Resolutio n Use r define d Use r define d Use r define d Tempora l Resolutio n Use r define d Use r define d Use r define d Use r Non-technica l plannin g Lan d us e an d Planners transportatio n participant s transportatio n planners engineers economist s communit y participant s Operationa l Method s GI S Exper t system logit Logit networ k analysi s regression Mont e Carl o Urba n Lan d Us e Categorie s Use r define d Use r define d Use r define d Non-Urba n Lan d Us e Categorie s Use r define d Use r define d Use r define d Independen t Variable s o n Lan d Us e Pattern s Transportation zoning maste r pla n Transportation zoning Transportation zoning maste r plan fiscal polic y maste r plan fisca l policy impac t fee tax sewe r an d wate r fee subsidy roa d toll parkin g fee registratio n fe e Rearrange d from EP A (2000) Klosterma n (2001) an d Sietchipin g (2004 ) 4 4

PAGE 59

3. 2 Evolutio n o f Urban Modeling : CA/AB M 3.2. 1 Cellula r Automat a (CA ) Recen t researc h effort s towar d a ne w urba n mode l tha t ar e detaile d bu t flexible conceptuall y understandable cost-effective dynamic realisti c an d reliabl e includ e th e adaptio n o f cellula r automat a (CA ) an d agent-base d modelin g (ABM ) approaches A t th e mos t rudimentar y level a cellula r automato n i s a n arra y o r lattic e o f regula r space s o r cells A t an y give n time a particula r cel l i s i n on e o f a finit e numbe r o f allowe d states an d tha t stat e wil l chang e accordin g t o th e state s o f neighborin g cell s i n th e lattic e accordin g t o a uniforml y applie d se t o f transitio n rules Cell s alte r thei r state s iterativel y an d synchronousl y throug h th e repeate d applicatio n o f thes e rules A C A i s thu s compose d o f fou r principl e elements : a lattice a se t o f allowe d states neighborhood s define d b y th e lattice an d transitio n rules I n addition a fifth tempora l componen t ca n b e considere d (Torren s & O'Sullivan 2001) Fro m th e Internationa l Encyclopedi a o f Huma n Geograph y (2009) C A mode l i n urba n simulatio n contex t i s define d as : A clas s o f spatiall y disaggregat e models ofte n picture d a s bein g forme d o n a 2 dimensiona l lattic e o f cells wher e eac h cel l represent s a lan d us e an d wher e embodyin g processe s o f chang e i n th e cellula r stat e ar e determine d i n th e loca l neighbourhoo d o f an y an d ever y cell Suc h model s ca n b e see n a s simplification s o f agent-base d model s wher e th e focu s i s o n emergen t spatia l pattern s throug h time C A model s hav e man y advantage s fo r modelin g urba n phenomena includin g thei r decentralize d approach th e lin k the y provid e t o complexit y theory th e connectio n o f for m wit h functio n an d patter n wit h process th e relativ e eas e wit h whic h mode l result s ca n b e visualized thei r flexibility thei r dynami c approach an d als o thei r affinitie s wit h geographi c informatio n system s an d remotel y sense d dat a (Torrens 2000) Perhap s th e mos t significan t o f thei r qualities however i s thei r relativ e simplicity B y mimickin g ho w macro-scal e urba n structure s ma y emerg e fro m th e myria d interaction s o f simpl e elements C A offer s a framewor k fo r th e exploratio n o f comple x adaptiv e systems However wit h CA thi s principa l innovativ e featur e i s als o on e o f it s greates t weaknesses C A model s ar e 4 5

PAGE 60

constraine d b y thei r simplicity an d thei r abilit y t o represen t real-worl d phenomen a i s ofte n dilute d b y thei r abstrac t characteristic s (Torren s & O'Sullivan 2001) 3.2. 2 Agent-Base d Mode l (ABM ) Th e targe t o f urba n modelin g an d simulation som e "rea l world phenomeno n whic h th e researcher s ar e intereste d in i s alway s a dynami c entity changin g ove r tim e an d reactin g t o it s environment. Wit h th e complexit y an d dynami c feature s o f urba n phenomenon especiall y whe n th e relationshi p betwee n th e targe t an d variable s o f th e mode l i s nonlinear traditiona l analyti c reasonin g usin g logi c o r b y usin g mathematic s ca n b e ver y difficul t o r impossible I n thes e cases disaggregate d an d dynami c simulatio n i s ofte n th e onl y wa y (Gilber t & Troitzsch 1999) Earl y effort s t o construc t urba n model s buil t aroun d representin g th e action s an d behavio r o f individua l agent s locate d i n spac e includ e Chapi n & Weis s (1968 ) an d Ingra m e t al. (1972 ) bu t i t wa s sinc e mid-1990s a n agent-base d simulatio n approac h bega n t o emerg e (Benenso n e t al. 2002) an d thi s wa s mainl y becaus e o f th e micro-compute r capacit y development Agent-base d model s (ABMs ) ar e compute r representation s o f system s tha t ar e comprise d o f multiple interactin g actor s (Brown 2005) I n a n urba n plannin g context agent s includ e households developers enterprises and/o r planners B y simulatin g th e individua l action s o f divers e actors an d measurin g th e resultin g syste m behavio r an d outcome s ove r time ABM s provid e usefu l tool s fo r studyin g urba n phenomen a tha t operat e a t multipl e scale s an d organizationa l level s an d thei r effects Wha t distinguishe s AB M fro m C A i s that wherea s C A usuall y ha s a fixe d interactio n topolog y (i e. whic h neighbor s a cel l interact s wit h i s fixe d b y th e cellula r geometry) interaction s i n th e AB M ca n b e dynamicall y change d a s th e mode l runs becaus e the y ar e define d a t th e agen t level rathe r tha n i n term s o f th e partitionin g o f spac e (Brown 2005) 4 6

PAGE 61

Severa l characteristic s defin e ABM : autonomou s agents communicatio n an d interactio n betwee n agents o r cooperatio n an d competitio n betwee n agents an d th e impac t o f agen t decisio n o n th e environmen t (Parke r e t al. 2003) Finally ABM s fo r urba n researc h ar e nearl y alway s spatiall y explicit whic h mean s tha t th e agent s and/o r thei r action s ar e reference d t o particula r location s o n th e urba n area Fo r thi s reason man y ABM s hav e eithe r direc t o r indirec t interactio n wit h GI S (Brown 2005) 3.2. 3 CA/AB M Approac h Ther e hav e bee n severa l attempt s t o overcom e th e limit s o f C A model s i n whic h th e immobilit y o f th e cell s i s a prerequisit e an d t o appl y th e concep t o f individualis m o f ABMs Benenso n (1999 ) an d Portugal i (2000 ) propose d th e Entity-Base d (EB ) modelin g approach Th e E B mode l consider s th e cit y a s consistin g o f tw o interactin g layers Th e firs t represent s immobil e urba n component s an d i s describe d b y a CA-typ e model whos e elementar y unit s ar e infrastructur e element s tha t ca n b e treate d a s innatel y homogeneous : lan d parcels houses stree t segments Th e secon d laye r o f th e E B mode l represent s th e instantaneou s spatia l distributio n o f mobil e urba n decisio n makers th e dynamic s o f whic h i s describe d b y mean s o f a n agent-base d (AB ) model Th e benefi t o f separatin g betwee n th e infrastructur e an d populatio n layer s i s tha t thi s approac h allow s researcher s t o accoun t fo r th e differen t rate s o f growt h o f th e tw o layer s (i.e. fas t growin g populatio n an d slo w developmen t o f infrastructure ) (Benenso n e t al. 2002) Torren s an d Benenso n (2005 ) extende d th e C A concep t t o incorporat e th e AB M concept an d propose d Geographi c Automat a System s (GAS) I n thi s framework geographi c phenomen a a s a whol e ar e considere d a s th e outcome s o f th e collectiv e dynamic s o f multipl e animat e an d inanimat e geographi c automata I n C A framework a n automation A belongin g t o a C A lattic e i s expresse d a s follow s (Torren s & Benenson 2005) : A~(S,T,N ) (3.1 ) 4 7

PAGE 62

wher e S denote s a se t o f state s an d N denote s automat a neighborin g A an d define s th e se t o f cell s fo r drawin g inpu t informatio n whic h i s necessar y fo r th e applicatio n o f transitio n rule s T Whereas i n th e GA S framework a geographi c automatio n system G consist s o f seve n components : G~(K;S,T S ;L,M L ;N,R N ) (3.2 ) wher e K denote s a se t o f type s o f automat a feature d i n th e GAS L denote s th e geo referencin g convention s tha t dictat e th e locatio n o f automat a i n th e system M L denote s th e movemen t rule s fo r automata governin g change s i n thei r locatio n i n time an d R N denote s th e rule s tha t gover n change s o f automat a relation s t o th e othe r automata Th e minimalisti c approac h o f GA S i n whic h mos t urba n phenomen a ar e abstracte d i n a simpl e simulatio n framewor k allow s fo r a degre e o f standardizatio n betwee n automat a model s an d othe r systems includin g GI S (Torren s & Benenson 2005) 3. 3 CA/AB M application s CA/AB M ha s bee n applie d t o variou s area s i n urba n plannin g field suc h a s land-use/cove r chang e (Lain e & Busemeyer 2004 ; D Parke r e t al. 2001 ; Waddell 2001) urba n growt h (Fernande z e t al. 2005 ; Jant z e t al. 2004) urba n ecolog y (Gimblett 2002 ; Bec u e t al. 2003 ; Brow n e t al. 2004) pedestria n movemen t an d evacuatio n (Blu e & Adler 2001 ; Hakla y e t al. 2001 ; Kerridg e e t al. 2001 ; Zacharia s e t al. 2005) transportatio n (Ki i & Doi 2005 ; Teodorovic 2003) housin g an d industria l locatio n (Arentz e & Timmermans 2007 ; Irwi n & Bockstael 2002 ; Mulbrandon 2007 ; Torrens 2001) an d segregatio n (Benenso n e t al. 2002 ; Lauri e & Jaggi 2003 ; J Zhang 2004 ; O'Sulliva n & Macgill 2005 ; Che n e t al. 2005 ; Bruc h & Mare 2006 ; Benar d & Wilier 2007) Th e nex t sectio n focuse s specificall y o n th e applicatio n o f C A-AB M t o th e stud y o f residentia l location 4 8

PAGE 63

3.3. 1 Residentia l Locatio n Modelin g A majorit y o f existin g residentia l locatio n studie s adaptin g th e AB M approac h focuse d o n exurba n populatio n migration an d locatio n decisio n behavio r (Brow n e t al. 2002 ; Fernande z e t al. 2005 ; Ran d e t al. 2002 ; Yin 2004 ; Yi n & Muller 2007 ; Mulle r e t al. 2008) Les s attentio n wa s pai d t o urba n populatio n behavio r mainl y becaus e o f th e extensiv e impact s o f suburba n migratio n an d lan d developmen t a t th e urban-rura l fringe Also mor e heterogeneou s characteristic s o f urba n housin g seeker s whe n compare d t o exurba n housin g seeker s mad e th e urba n residentia l locatio n modelin g proces s difficult Researc h o n urba n migration an d househol d locatio n decisio n pai d specifi c attentio n t o th e dynamic s o f urba n rea l estat e market J Jackso n e t al (2008 ) explore d th e proces s o f gentrificatio n an d lan d valu e chang e i n Boston Massachusett s usin g fou r classe s o f agent s (professionals colleg e students non-professional an d th e seniors ) an d se t o f simpl e decisio n rules whic h ar e base d o n th e socia l clas s an d economi c abilit y o f th e agent I n thei r model agen t decisio n i s influence d b y memor y o f favorit e locatio n an d b y communicatio n betwee n friend s o f simila r class However a n agent' s decisio n t o mov e i n thei r mode l i s onl y drive n b y a n economi c imperative : a n agen t decide s t o loo k fo r a ne w locatio n whe n i t ca n n o longe r affor d it s curren t place bu t n o othe r househol d mobilit y factor s ar e considered Usin g a virtua l monocentri c urba n rea l estat e market Filatov a e t al (2008 ) adapte d agent s wit h heterogeneou s preference s fo r location Thei r mode l outpu t wa s analyze d usin g a serie s o f macro-scal e economi c an d landscap e patter n measures includin g lan d ren t gradient s estimate d usin g simpl e regression The y demonstrate d tha t heterogeneit y i n preferenc e fo r proximit y alon e i s sufficien t t o generat e urba n expansio n an d tha t informatio n o n agen t heterogeneit y i s neede d t o full y explai n lan d ren t variatio n ove r space Nar a an d Torren s (2005 ) use d fou r type s o f fixe d agent s (market sub-area property an d fixe d land ) an d on e typ e o f mobil e agen t (resident ) i n simulatin g gentrificatio n proces s i n Sal t Lak e City Utah Locatio n decisio n an d housin g choic e o f residen t agen t i s base d o n 4 9

PAGE 64

fou r housin g characteristic s (propert y value housin g type hous e size an d accessibility) an d tw o neighborhoo d characteristic s (economi c statu s an d ethnicity) Thei r mode l demonstrate d th e gentrificatio n proces s whe n gentrifier s (ne w resident s wit h highe r economi c status ) an d gentrifiabl e propertie s (ne w propertie s wit h highe r value ) introduce d t o th e stud y area I n AB M mobilit y an d locatio n decisio n studies th e concep t o f lifecycl e i s widel y adapte d (Waddell 1998 2000) Torren s (2001 ) use d thre e lifecycl e stages : young middle an d senio r i n constructin g a residentia l locatio n model Household s i n differen t lifecycl e stage s hav e differen t preference s fo r housin g types : youn g an d middl e agent s withou t childre n prefe r apartmen t an d middl e agent s wit h childre n an d senio r agent s prefe r houses Becaus e th e mode l wa s develope d a s a prototyp e residentia l locatio n mode l base d o n simpl e unrealisti c assumption th e practica l usabilit y i s limited Fontain e an d Rounsevel l (2009 ) presente d a framewor k t o mode l futur e residentia l deman d fo r housin g i n a polycentri c regio n base d o n th e househol d lif e cycl e stages Th e mode l wa s calibrate d fo r th e regio n o f Eas t Angli a i n th e Unite d Kingdo m usin g a semi-quantitativ e procedure Thei r result s demonstrate d non-uniform spatia l pattern s o f urba n spraw l wit h som e location s experiencin g greate r urba n developmen t pressur e tha n others Th e proximit y t o th e principa l cit y wa s identifie d a s th e mai n deman d facto r fo r residentia l housing B y explicitl y modelin g agen t behavio r an d interactions the y represente d th e respons e an d adaptatio n strategie s o f a populatio n t o changin g circumstances Devisc h e t al (2009 ) teste d th e concep t o f th e bounde d rationality The y presente d a n agent-base d mode l develope d t o simulat e residentia l choic e behavio r i n a non-stationar y housin g market Th e mode l wa s buil t aroun d th e assumptio n tha t agent s a t differen t lifecycl e stage s hav e incomplet e an d imperfec t knowledg e an d thu s hav e t o bas e thei r decision s o n beliefs The y illustrate d ho w th e agent s dea l wit h th e uncertaint y inheren t i n thes e beliefs bot h a t th e leve l o f a singl e agent decidin g amon g a se t o f successiv e actions an d a t th e leve l o f a grou p o f agents negotiatin g ove r th e pric e o f a house Torren s (2007 ) applie d th e GA S (geographi c automat a systems ) approac h t o explor e residentia l mobilit y an d locatio n decisio n proces s o f households I n hi s model residentia l mobilit y i s treate d a s 5 0

PAGE 65

a two-stag e process First household s decid e whethe r t o initiate a relocatio n even t o r not lookin g bot h t o interna l an d t o externa l stressor s i n formulatin g thi s decision I f the y decid e t o engag e i n a housin g search the y d o s o base d o n preference s tha t relat e t o thei r interna l characteristic s an d thos e o f th e large r communit y an d submarke t i n whic h th e searc h i s focused Th e applicabilit y o f th e methodolog y use d i n th e mode l i s demonstrate d throug h th e developmen t o f a ric h mode l o f residentia l mobility i n whic h individua l household s interac t wit h othe r household s an d real-estat e infrastructure dynamicall y i n spac e an d time t o for m syntheti c communitie s an d artificia l property submarkets L i an d Li u (2007 ) use d multicriteri a evaluatio n technique s t o determin e som e o f th e parameter s fo r th e agent-base d residentia l developmen t mode l i n Guangzhou China Empirica l dat a fro m GI S wer e use d t o defin e th e agent' s properties an d a sensitivit y analysi s wa s als o carrie d ou t t o asses s th e influence s o f parameter s o n simulatio n outcomes B M W u e t al (2008 ) presente d a dynami c simulatio n mode l tha t projecte d th e futur e populatio n o f th e cit y o f Leed s a s a basi s fo r polic y analysi s an d scenari o planning The y argue d tha t microsimulatio n modelin g i s no t entirel y effectiv e i n th e representatio n o f studen t populations The y sugges t tha t agent-base d modelin g an d microsimulatio n ar e powerfu l a s complementar y technologie s fo r individual-base d modeling 5 1

PAGE 66

Tabl e 3. 2 Compariso n o f selec t agent-base d residentia l locatio n model s Themati c Scop e Locatio n Agent s Accessibilit y Variable s Othe r Variable s J Jackso n e t a l (2008 ) Gentrificatio n Boston M A Professional s Colleg e student s Non professional s Th e senior s Colleg e campu s Commercia l distric t Affordabilit y Neighborhoo d Recommendatio n Locatio n Memor y Filatov a e t a l (2008 ) Lan d marke t Virtua l Buye r Selle r CB D Environmenta l amenitie s Ren t Fontain e & Rounsevel l (2009 ) Urbanizatio n Eas t Anglia U K Household s a t differen t lifecycl e stage s Roa d networ k Trai n statio n Ke y servic e are a Marke t tow n Larg e cit y Coastlin e River s an d waterbod y Gree n recreationa l are a Floodin g zon e Coasta l clif f erosio n zon e Nar a & Torren s (2005 ) Gentrificatio n Sal t Lak e City U T Residen t Downtow n Highwa y Mal l Grocer y Propert y valu e Propert y siz e Housin g typ e Neighborhoo d economi c statu s Neighborhoo d ethni c profil e B M W u e t a l (2008 ) Spatia l populatio n distributio n Leeds U K Firs t yea r undergraduate s Othe r undergraduate s Maste r student s Doctora l student s Universit y Sam e cohor t 5 2

PAGE 67

3.3. 2 Segregatio n Modelin g Residentia l segregatio n base d o n ethnicit y o r economi c statu s o r religiou s affiliatio n i s on e o f th e consistentl y explore d area s b y AB M modeler s fro m variou s fields includin g sociology politics economics an d urba n planning On e o f th e earlies t an d best-know n model s o f residentia l segregatio n wa s develope d b y Schellin g (1971 1978) I n th e "Schellin g model, agent s inhabi t a checkerboar d (sixty-fou r square s i n eigh t row s an d eigh t columns) Th e agent s ar e divide d int o tw o classe s tha t ar e mean t t o represen t an y binar y socia l divisio n tha t coul d affec t th e distributio n o f agent s i n spac e (e.g. me n an d women black s an d whites French-speakin g an d English-speaking) Agent s ar e initiall y distribute d randoml y acros s th e lattice Fro m tha t poin t on agent s observ e thei r neighborhoo d (3X3) an d chang e location s i f th e numbe r o f agent s o f th e othe r typ e exceed s som e threshold Schellin g showe d tha t population s coul d reac h hig h level s o f segregatio n eve n whe n agent s ar e willin g t o remai n i n neighborhood s i n whic h u p t o two third s o f thei r neighbor s ar e member s o f th e othe r group Hi s mode l succinctl y represent s wha t mos t agent-base d model s tr y t o fin d out : ho w people' s interaction s coul d lea d t o result s tha t wer e neithe r intende d no r expected Followin g Schelling' s semina l paper a numbe r o f researcher s extende d th e standar d Schellin g mode l t o investigat e a variety o f questions Som e o f recen t variation s o f th e Schellin g model s ar e introduce d here Lauri e an d Jagg i (2003 ) modele d th e rol e o f "vision i n th e dynamic s o f Schellin g model A s outcome s o f non-linea r functio n o f "vision ( a distanc e agent s ca n authenticall y se e thei r neighborhoods ) an d neighborhoo d preferenc e threshol d o n th e segregatio n index the y identifie d thre e regime s o f segregation: a n unstabl e regime wher e societie s invariabl y segregate ; a stabl e regime wher e integrate d societie s remai n stable ; an d a n intermediat e regim e wher e a comple x behavio r i s observed J Zhan g (2004 ) demonstrate d tha t segregatio n i s a stochasticall y stabl e stat e tha t tend s t o emerg e an d persis t i n th e lon g ru n regardles s o f initia l state Usin g a stochasti c evolutionar y gam e theoreti c model h e showe d 5 3

PAGE 68

tha t sligh t asymmetr y i n residentia l preference s betwee n tw o group s woul d produc e endogenou s segregation O'Sulliva n an d Macgil l (2005 ) investigate d th e impac t o f neighborhoo d scal e o n th e dynamic s o f segregatio n usin g tw o neighborhoo d types : continuou s o r loca l neighborhoo d (immediatel y neighborin g household s i n a lattic e o f residentia l locations ) an d bounde d neighborhoo d (regiona l neighborhoo d containin g thei r residentia l locations) H e foun d tha t th e tim e take n fo r th e mode l t o settl e o n a stabl e arrangemen t increase s wit h neighborhoo d size th e mode l t o settl e o n a stabl e arrangemen t increase s wit h neighborhoo d size an d tha t large r neighborhood s ma y lea d t o situation s wher e man y household s ar e unhapp y bu t unabl e t o relocat e t o location s preferabl e t o thei r location K Che n e t al (2005 ) incorporate d househol d preference s ove r locatio n characteristic s othe r tha n racia l compositio n preference s i n thei r model Extende d househol d preferenc e variable s i n thei r mode l includ e housin g pric e an d neighborhoo d density an d th e mode l allowe d incom e heterogeneit y acros s race s an d amon g household s o f sam e race Thei r preliminar y finding s indicat e tha t pattern s o f segregatio n ca n emerg e eve n whe n individual s ar e wholl y indifferen t t o neighborhoo d racia l composition du e t o competin g preference s ove r neighborhoo d density Bruc h an d Mar e (2006 ) teste d th e sensitivit y o f residentia l choic e function s b y testin g thre e type s o f neighborhoo d threshol d functions : Schelling-typ e discret e choic e function stepwis e function an d continuou s function s wit h variou s slopes The y foun d tha t th e sam e averag e leve l o f toleranc e bu t differen t respons e function s giv e ris e t o differen t neighborhoo d formatio n patterns Bernar d an d Wilie r (2007 ) extende d th e Schellin g mode l t o incorporat e th e wealt h an d statu s o f agent s an d th e desirabilit y an d affordabilit y o f residences The y analyze d th e effect s o f th e degre e o f th e status-wealt h correlatio n an d th e exten t t o whic h th e wealt h o f resident s shape s th e affordabilit y o f residence s o n level s o f statu s an d wealt h segregation Thei r finding s ar e tha t th e greate r th e correlatio n betwee n statu s an d wealth th e mor e th e agent s ten d t o segregate eithe r du e t o choic e (fo r th e wealth y an d hig h status ) o r exclusio n (fo r th e poo r an d lo w status) Yi n (2009 ) adde d economi c factor s (i.e housin g price ) i n additio n t o th e neighborhoo d racia l compositio n factors Usin g Buffalo Ne w Yor k data 5 4

PAGE 69

sh e showe d tha t segregatio n i n th e cit y coul d generat e from th e interactio n o f racia l an d economi c factors Non-Schellin g model-base d segregatio n researc h usin g AB M wa s develope d b y Benenso n e t al (2002) Benenso n applie d a n Entity-base d (EB ) approac h t o simulate th e dynamic s o f ethni c distributio n i n th e Yaff o are a o f th e cit y o f Te l Aviv Israel durin g th e perio d 1955 95 I n thei r model eac h householde r i s considere d a s a separat e entit y wit h it s ow n cultural religious o r ethni c propertie s an d whos e residentia l behavio r i s define d b y th e propertie s o f th e surroundin g infrastructur e entitie s (tha t is th e physica l environment ) an d othe r householder s (th e socia l environment) The y define d tw o dissonance s a s dwellin g an d residentia l dissonance : differenc e betwee n th e propertie s o f a n agen t an d th e propertie s o f th e neighbor s an d th e dwellin g types The y assume d tha t wit h th e increas e i n combine d dissonances th e probabilit y o f leavin g a residenc e increases an d th e probabilit y o f occupyin g a vacan t residenc e decreases Th e mai n differenc e o f thei r mode l fro m th e Schellin g mode l i s tha t the y define d th e neighborhoo d base d o n th e adjacenc y o f Vorono i polygon s constructe d aroun d th e centroi d o f th e building s instea d o f usin g lattic e o f cells Jayaprekas h e t al (2009 ) presente d a mode l o f th e interactio n o f segregatio n an d suburbanizatio n i n determinin g residentia l location Th e mode l incorporate d differentia l incom e betwee n tw o classe s o f agents a simplifie d marke t mechanis m fo r th e purchas e o f housing an d a simpl e geographi c structur e o f on e centra l cit y an d fou r symmetrically arrange d suburbs Agent s derive d utilit y fro m neighborhoo d racia l composition th e siz e o f thei r lot privat e amenitie s tha t ar e specifi c t o neighborhood s an d publi c amenitie s tha t stretc h acros s municipalities The y foun d tha t th e public-amenitie s ter m lead s t o a positive feedbac k loo p i n whic h migratio n t o suburb s increase s th e publi c amenitie s i n thos e municipalitie s whil e lowerin g amenitie s i n th e centra l city thu s sparkin g furthe r migration Whe n th e minorit y agent s wer e uniforml y les s affluen t tha n th e majorit y agents thi s dynami c produce d discontinuit y i n segregatio n a s measure d b y centralization 5 5

PAGE 70

4 RESEARC H METHOD S 4. 1 Stud y Area s Stud y site s ar e selecte d i n th e histori c downtow n district s i n thre e Colorad o citie s wit h varyin g cit y size s an d wit h differen t distributio n pattern s o f urba n amenities : Boulder Denver an d Louisville Thes e thre e citie s hav e a hig h leve l o f urba n an d natura l amenities an d the y ar e commonl y ranke d highl y o n th e bes t place-typ e nationa l surveys'" Th e degre e o f th e populatio n growt h i n thre e stud y area s durin g th e 1990-200 0 perio d show s les s tha n th e Colorad o averag e level sinc e thre e stud y area s ar e locate d withi n densel y develope d downtow n area s (Tabl e 4.1) Figur e 4. 1 Stud y Areas Fron t Rang e Region Colorad o 5 6

PAGE 71

Tabl e 4. 1 Populatio n change 1990-2000 stud y area s 199 0 200 0 Chang e Chang e % Boulder C O Y P M A T S Othe r Tota l H H Housin g Louisville C O Y P M A T S Othe r Tota l H H Housin g Denver C O Y P M A T S Othe r Tota l H H Housin g Colorad o Y P M A T S Othe r Tota l H H Housin g U.S Y P M A T S Othe r Tota l H H Housin g 11,72 6 1,86 8 1,76 0 2,69 5 18,04 9 9,11 2 9,51 9 4,60 2 95 6 43 7 2,85 6 8,85 1 3,29 9 3,40 4 18,02 6 5,32 9 3,42 6 3,57 5 30,35 6 19,64 4 23,55 3 1,418,38 6 588,22 4 329,44 3 958,34 1 3,294,39 4 1,282,48 9 1,477,34 9 99,775,14 7 46,371,00 9 31,241,83 1 71,321,88 6 248,709,87 3 91,947,41 0 102,263,67 8 8,19 4 2,70 1 1,30 0 6,40 7 18,60 2 9,25 8 9,50 8 4,60 1 2,41 1 60 6 3,60 8 11,22 6 4,13 1 4,19 9 20,37 8 7,54 3 2,62 9 3,39 9 33,94 9 22,11 3 23,47 9 1,707,08 8 953,43 2 416,07 3 1,224,66 8 4,301,26 1 1,658,23 8 1,808,03 7 104,004,25 2 61,952,63 6 34,991,75 3 80,473,26 5 281,421,90 6 105,480,10 1 115,904,64 1 -3,53 2 83 3 -46 0 3,71 2 55 3 14 6 -1 1 1 1,45 5 16 9 75 2 2,37 5 83 2 79 5 2,35 2 2,21 4 -79 7 -17 6 3,59 3 2,46 9 -7 4 288,70 2 365,20 8 86,63 0 266,32 7 1,006,86 7 375,74 9 330,68 8 4,229,10 5 15,581,62 7 3,749,92 2 9,151,37 9 32,712,03 3 13,532,69 1 13,640,96 3 -30.1 % 44.6 % -26.1 % 137.7 % 3.1 % 1.6 % -0.1 % 0.0 % 152.2 % 38.7 % 26.3 % 26.8 % 25.2 % 23.4 % 13.0 % 41.5 % -23.3 % -4.9 % 11.8 % 12.6 % -0.3 % 20.4 % 62.1 % 26.3 % 27.8 % 30.6 % 29.3 % 22.4 % 4.2 % 33.6 % 12.0 % 12.8 % 13.2 % 14.7 % 13.3 % YP : Yon g Professional s (ag e betwee n 20-45) MA : Middl e Age s (ag e betwee n 45-65) TS : Th e Senior s (ag e ove r 65) Other : Other s (ag e betwee n 0-20) HH : Household s 5 7

PAGE 72

4. 2 Dat a Th e dat a fo r thi s stud y wer e collecte d primaril y fro m tw o sources : interview s wit h individual s involve d i n urba n residentia l market s a s wel l a s dat a i n digita l forma t suc h a s GI S an d histori c urba n amenit y locatio n dat a fro m loca l government s an d censu s dat a from th e U.S Censu s Bureau 4.2. 1 Exper t Intervie w First I conducte d interview s wit h expert s o f th e loca l housin g market s i n th e thre e stud y are a betwee n summe r 200 9 an d sprin g 2010 Pilo t interview s wit h thre e realtors tw o developers an d on e planne r wer e conducte d fo r th e preparatio n an d refinemen t o f th e surve y instruments Interview s wit h 1 5 realtor s i n th e Denver-Boulde r metropolitan are a wer e conducte d afte r th e pilo t interviews Th e intervie w wa s no t tightl y structure d a t th e beginnin g stag e o f th e eac h interview thu s allowin g th e interviewee s t o rais e issue s concernin g urba n residentia l locatio n pattern s the y fel t wer e important Interviewee s wer e guide d i n th e followin g ways : Afte r I gree t th e interviewe e an d introduc e myself I provide d a brie f overvie w o f th e purpos e o f th e researc h project whic h i s t o fin d ou t th e mechanism s b y whic h differen t demographi c group s mak e thei r residentia l locatio n decisions I aske d interviewee s t o provid e a brie f overvie w o f thei r fir m o r thei r persona l experience s i n th e fiel d an d t o provid e an y material s tha t woul d assis t m e i n understandin g thei r wor k an d projects I prompte d th e interviewee s t o tal k abou t th e locationa l preference s o f thei r customer s an d neighborhoo d preference s betwee n thei r customers A s issue s an d topi c area s emerged I aske d th e interviewee s t o addres s an y relate d area s i f the y ha d no t covere d the m i n thei r unstructure d conversations Finally I aske d th e interviewee s t o fil l ou t a n AH P surve y instrumen t tha t wer e designe d t o calculat e differen t weight s o n variou s locatio n decisio n factor s b y differen t demographi c groups 5 8

PAGE 73

Th e Analytica l Hierarch y Proces s (AHP ) (Satty 1977 ) i s a surve y approac h whic h allow s researcher s t o ran k an d evaluat e preferences. 1 V Give n a choic e betwee n tw o features subject s ar e aske d t o rat e th e relativ e importanc e o f on e ove r th e othe r o n a scal e o f on e t o nine Th e outpu t o f th e AH P i s a prioritize d rankin g o r weightin g o f eac h decisio n alternativ e (Atthirawon g & McCarthy 2002) Fo r th e AH P survey subject s wer e give n a lis t o f urba n amenitie s an d aske d t o compar e a se t o f pairwis e urba n amenitie s fo r differen t demographi c groups Th e quantitativ e outpu t i s intended t o establis h th e relativ e importanc e o f th e locatio n decisio n factors Fiftee n variable s i n fiv e categorie s fro m th e literatur e revie w an d pilo t interview s wer e used T o cross-validat e th e surve y results th e weight s fo r locatio n decisio n variable s fro m th e AH P surve y wer e compare d wit h th e verball y describe d comparativ e importanc e o f th e variable s fro m th e interviews Recreationa l Amenit y Footbal l Field Bal l Park Them e Par k Recreatio n Cente r Neighborhoo d Park Trail Bik e Trai l "Th e Thir d Places Coffe e Shop Caf e Restauran t Bar Pub Nigh t Clu b Cultura l Amenit y Orchestra Opera Musica l Museum Galler y Library Bookstor e Shoppin g Grocer y Stores Pharmac y Shoppin g Cente r Smal l Retai l Stor e Transportatio n 1 Publi c Transportatio n Highwa y Ram p Majo r Arteria l Figur e 4. 2 Hierarchica l Diagra m fo r th e AH P Surve y 5 9

PAGE 74

4.2. 2 Dat a i n Digita l Forma t T o construc t a computer-base d urba n residentia l locatio n simulatio n model variou s GI S layer s includin g road zoning, an d parce l dat a fo r th e stud y area s wer e collecte d an d prepare d fo r th e spatia l simulation U S Censu s dat a wa s use d t o analyz e populatio n distributio n pattern s fo r differen t demographi c group s a t th e censu s bloc k level whic h i s th e smalles t availabl e uni t o f analysis Tabl e 4. 2 GI S dat a layer s Laye r Zonin g Stree t Networ k Parce l Censu s Bloc k Boundar y Censu s Bloc k Grou p Boundar y Sourc e Boulde r an d Denve r Plannin g Departmen t U.S Censu s Burea u TIGER/Lin e Boulde r an d Denve r Plannin g Departmen t U.S Censu s Burea u TIGER/Lin e U.S Censu s Burea u TIGER/Lin e Quarterl y Censu s o f Employmen t an d Wage s (QCEW ) dat a wa s use d t o geocod e th e histori c location s o f variou s type s o f urba n amenities Figur e 4. 3 an d Tabl e 4. 4 sho w th e type s an d distributio n o f urba n amenitie s i n th e stud y area s i n th e year s 199 0 an d 2000 Tabl e 4. 3 Quarterl y Censu s o f Employmen t an d Wage s dat a structur e (N = 1,363,213 ) Tabl e Elemen t Yea r LGLN M ADDRES S SI C NAIC S Description s Yea r dat a collecte d Nam e o f th e reportin g uni t fo r lega l purposes bot h fo r singl e an d multi-uni t employers Referre d t o a s th e Corporat e Nam e i n man y system s Identifie s th e physica l locatio n addres s Standar d Industria l Classificatio n cod e t o identif y th e primar y economi c activit y o f th e reportin g uni t Nort h America n Industr y Classificatio n Syste m code s ar e unifor m industria l code s use d b y th e Unite d States Canada an d Mexic o t o identif y th e primar y activit y o f a n establishmen t 6 0

PAGE 75

. 1T^ = # X^ -f t ^ J h [3 T % L*i%a>'^Q 3 L^--TO~tf"ir^ 7 ***^ — ^ M ~'LtL-J"U~ t o O &icpCwit w J (1 ) Boulder 199 0 •*' JI 5 8 J& %&• VqUl j -.f t t ~ B i ^ r --T& ? S3I T b-6" ^ n" 1 1 Otf t r ^ • • > *tT1 c o I F 1 S ? £ V (2 ) Boulder 200 0 *t 4 • > — i t a i— i j J ( ^ JTT i_ j RMira U • (3 ) Denver 199 0 \ 5 4 • BI T (4 ) Denver 200 0 A Bir^^Ji t'^r f i t ^tsO "*1T1?^™'„ 1 ~' •fr y ~"~ • • B w O Oscv y • OccCwM t O UBur y Unman • ^nMBc U • Ctl i • Cn n O 9iCf n O BrtP M • t M O MIU B ifii> M (5 ) Louisville 199 0 tr ^ 0 +J ^ • — *~* • *fu o • i a a n ^ 9 ..,,..• : "* ^ Is ~"" ~ -4 l~~~ ^ 1 3 1^-^— 3 ^ • ^^^^^^^^ • • ~, ~ — ^O O LBnr v • hkiteu m • Srn u RAM I > • Cit e • Ops n O *at > Corn s [ 0 BnlP w • Pu l O Out*;rrw | • • • r+^myBnTi J | (6 ) Louisville 200 0 Figur e 4. 3 Distribution s o f Urba n Amenitie s i n th e Stud y Areas 1990 200 0 6 1

PAGE 76

Tabl e 4. 4 Urba n amenit y change s i n th e stud y areas 1990-200 0 Cit y Amenitie s 199 0 200 0 Increas e Increas e Rat e Boulde r Footbal l Field Bal l Park Them e Par k Recreatio n Cente r Neighborhoo d Park Trail Bik e Trai l Coffe e Shop Caf e Restauran t Bar Pub Nigh t Clu b Orchestra Opera Musica l Museum Galler y Library Bookstor e Grocer y Stores Pharmac y Shoppin g Center s Smal l Retai l Store s Boulde r Tota l Denve r Footbal l Field Bal l Park Them e Par k Recreatio n Cente r Neighborhoo d Park Trail Bik e Trai l Coffe e Shop Caf e Restauran t Bar Pub Nigh t Clu b Orchestra Opera Musica l Museum Galler y Library Bookstor e Grocer y Stores Pharmac y Shoppin g Center s Smal l Retai l Store s Denve r Tota l Louisville Footbal l Field Bal l Park Them e Par k Recreatio n Cente r Neighborhoo d Park Trail Bik e Trai l Coffe e Shop Caf e Restauran t Bar Pub Nigh t Clu b Orchestra Opera Musica l Museum Galler y Library Bookstor e Grocer y Stores Pharmac y Shoppin g Center s Smal l Retai l Store s Louisvill e Tota l 3 1 1 4 3 1 8 7 5 5 4 1 1 0 1 3 1 6 0 24 4 1 1 1 2 1 1 6 9 8 1 2 6 7 3 1 0 2 6 4 25 1 1 3 3 6 1 6 1 1 2 1 1 0 4 5 7 6 3 5 4 3 2 3 19 0 1 1 1 9 2 1 2 2 7 1 10 9 47 8 6 1 2 2 1 1 6 19 3 2 8 1 1 7 9 2 7 2 11 4 44 6 1 5 3 6 2 3 0 3 2 2 3 9 0 4 9 7 3 2 4 0 5 11 5 6 1 5 1 2 1 4 0 4 9 23 4 5 1 0 0 9 5 1 6 5 0 6 1 7 0 5 0 19 5 0 2 0 1 2 4 2 1 0 2 8 0 0 4 0 100.0 % 218.2 % 0.0 % 27.8 % 153.3 % 120.0 % 375.0 % 100.0 % 20.0 % 107.7 % 0.0 % 81.7 % 95.9 % 500.0 % 9.1 % 0.0 % 0.0 % 96.9 % 133.3 % 83.3 % 0.0 % 200.0 % 170.0 % 0.0 % 78.1 % 77.7 % 0.0 % 66.7 % 0.0 % 100.0 % 400.0 % 200.0 % 100.0 % 0.0 % 200.0 % 800.0 % N A 0.0 % 70.2 % 6 2

PAGE 77

4. 3 Variabl e Selectio n Urba n amenit y variable s fo r th e simulatio n mode l wer e collecte d from th e existin g literatur e an d pilo t interview s wit h th e ke y informant s i n urba n residentia l housin g marke t i n th e stud y areas Accessibilit y t o cultura l amenitie s an d shoppin g amenitie s ar e commonl y agree d a s importan t locatio n decisio n factor s i n th e literature whil e som e pa y mor e attentio n t o th e entertainmen t amenitie s an d other s focu s o n small-scal e socia l place s lik e coffe e shop s o r cafes Tabl e 4. 5 Urba n amenit y variable s fro m th e existin g literatur e an d pilo t interview s Consume r Cit y Th e Cit y a s a n Creativ e Cente r (Glaese r e t al. 2006 Entertainmen t Machin e (Florida 2008,2005a Pilo t Intervie w 2001 ) (Lloy d e t al. 2001 ) 2002 ) Footbal l field, Bal l park Them e par k Recreatio n cente r Neighborhoo d park Trail Bik e trai l Coffe e shop Caf e Restauran t Bar Pub Nigh t clu b Orchestra Opera Musica l Museum Galler y Library Bookstor e Grocer y store Pharmac y Shoppin g cente r Smal l retai l stor e Basebal l field Restauran t Ba r o r Taver n Classica l Music Concer t Museum Ar t galler y Basebal l field Healt h clu b Bicycl e path Par k Cafe Juic e ba r Restauran t Bar Nigh t clu b Oper a Museu m Bookstore Librar y Whol e foo d stor e Trai l Caf e Musi c venu e Galler y Sho p Theate r Boutiqu e Sho p Theate r Universit y 6 3

PAGE 78

4. 4 Agent-Base d Mode l 4.4. 1 Constructio n o f Agent s 4.4.1. 1 Agen t Categorizatio n Th e simulatio n mode l ha s thre e agen t types : youn g professionals middl e age an d seniors Thes e agen t group s cove r almos t entir e deman d group s i n a n urba n residentia l marke t sinc e onl y th e ag e grou p belo w 2 0 i s no t include d i n th e agen t categorization Th e traditiona l for m o f a household parent s wit h children wa s no t categorize d a s a n independen t agen t typ e i n th e mode l mainl y becaus e o f th e unattractivenes s o f downtow n a s thei r housin g locatio n an d thei r relativ e smal l proportion s i n downtow n housin g market Th e issu e o f poorl y performin g publi c school s remain s a n unsolve d proble m fo r almos t al l America n citie s an d i s a rea l deterren t t o familie s wh o canno t affor d t o sen d thei r childre n t o privat e school Creativ e program s ar e bein g explore d t o addres s th e problem bu t t o date solution s hav e bee n elusiv e (Haughey 2002) I n thi s AB M simulation th e mode l use s a househol d a s a n agen t instea d o f a n individual Li n (1997 ) discusse s th e advantage s o f a household-leve l approac h t o migration modeling Becaus e migratio n decisions ar e ofte n mad e b y household s rathe r tha n individuals household-leve l analyse s ar e sometime s mor e consisten t wit h rationa l expectations Fo r example a migratin g individua l ma y willingl y underg o a ne t los s o f opportunitie s o r amenitie s s o tha t othe r househol d member s experienc e a ne t gain Usin g 199 0 censu s data Li n present s severa l empirica l model s tha t sho w ho w individual s an d household s ca n diffe r i n thei r destinatio n choices Thes e difference s ar e generall y modest Fro m a n empirica l poin t o f view usin g a n age-base d agen t categorizatio n metho d i n a housin g locatio n decisio n stud y ha s a grea t advantag e a s oppose d t o th e mor e complicate d an d possibl y difficult-to-fin d alternatives / Existin g studie s indicat e tha t littl e statistica l explanator y powe r i s los t i n studie s wher e ag e i s th e onl y o r th e mos t convenien t variabl e t o us e i n describin g th e poin t reache d i n lif e (Morrow-Jone s & Wenning 2005) Variable s 6 4

PAGE 79

fo r whic h ag e ca n b e a prox y includ e th e following : income status wealth nee d fo r mor e o r les s space nee d fo r lo w maintenance accessibl e residences an d mor e hel p i n dail y task s o r long-ter m care Tabl e 4. 6 Agen t categorizatio n Agen t Ag e Propertie s 3 8 ou t o f 5 0 larges t metropolitan area s sa w increase s i n thei r 20 t o 45-year-ol d populatio n betwee n 199 0 an d 2000 (averag e increas e rate : 13.08% nationa l increas e rate : 4.24% ) Youn g Professional s Urba n renovators : 20-4 5 wealthy young highl y educated highl y mobile an d eithe r singl e o r marrie d wit h les s tha n tw o children patroniz e cultura l establishment s i n th e centra l city hav e a relativel y hig h commutin g cost hav e relativel y lo w demand s fo r housin g an d land Fro m 200 0 t o 2010 th e percentag e o f th e nation' s household s betwee n 5 5 an d 6 4 i s projecte d t o jum p t o 17. 4 percent Middl e Age s hom e buyer s age d 4 5 an d olde r wh o prefe r denser mor e 45-6 5 compac t housin g alternative s wil l accoun t fo r 3 1 percen t o f tota l homeowne r growt h durin g th e 2000-1 0 period doubl e th e sam e segment' s marke t shar e i n th e 1990 s (Myer s an d Gearin 2001) Accordin g t o th e U.S Censu s Burea u projections a substantia l increas e i n th e numbe r o f olde r peopl e wil l occu r durin g th e 201 0 t o 203 0 period afte r th e firs t bab y boomer s tur n 6 5 i n 2011 Th e Senior s Ove r 6 5 Growt h amon g th e senio r populatio n wil l substantiall y excee d tha t o f younge r adults a n unprecedente d socia l an d economi c development I t i s predicte d tha t a numbe r o f empt y nester s wil l continu e t o gro w a s bab y boomer s age Afte r thei r childre n leav e home empt y nester s ofte n chang e thei r lifestyl e i n a wa y tha t favor s downtow n living 6 5

PAGE 80

4.4.1. 2 Agen t Behavio r Agen t behavio r rule s ar e structure d base d o n th e literatur e revie w an d intervie w results Eac h agen t select s residentia l location s base d o n it s preferenc e scor e fo r eac h cel l tha t consist s o f th e urba n amenit y score th e transportatio n networ k scor e an d th e neighborhoo d score Tota l Score, y = a / • Amenit y Accessibilit y Score/, + a 2 • Amenit y Densit y Score,; + b / • Transportatio n Accessibilit y Score/ + b 2 • Transportatio n Densit y Score/ + c / • Neighborhoo d Score/ (4.1 ) where / represent s agen t types j represent s eac h cell a/ & 2 hi, b 2 C / ar e th e weight s fo r eac h score Exponentia l deca y functio n wa s applie d whe n calculatin g amenit y an d transportatio n accessibilit y score s t o reflec t th e decayin g characteristic s o f th e utilit y (benefits ) relate d t o th e distanc e betwee n amenitie s an d a n origin A quarte r mil e radiu s wa s applie d a s a walkin g distanc e boundar y i n estimatin g th e deca y coefficient Accessibilit y Score/ = L n ( 2 ( e d Dx ) • w„) (4.2 ) where d represent s th e deca y coefficien t fo r th e distanc e variabl e D represent s th e Euclidia n distanc e t o th e neares t amenit y o r transportatio n facilit y W represent s th e AH P weigh t fro m th e intervie w x represent s amenit y an d transportatio n facilit y type s 6 6

PAGE 81

Then accessibilit y score s wer e normalize d withi n th e rang e o f [0 1 ] s o tha t thes e factor s ar e comparabl e wit h othe r score s i n th e decisio n making Th e score s wer e normalize d b y usin g followin g function : Score' = Scor e Mi n Ma x Mi n (4.3 ) Bal l par k Recreatio n cente r Neighborhoo d park Trai l Coffe e shop Caf e Restauran t Bar Pub Nigh t clu b Orchestra Opera Musica l Museum Galler y Library Bookstor e Grocery Pharmac y Shoppin g Center Mal l Smal l Retail s Publi c Transportatio n Highwa y Ram p Majo r Arteria l Amenit y Scor e Surfac e Agent • Figur e 4. 4 Calculatio n o f th e Amenit y Scor e 6 7

PAGE 82

Densit y Score s fo r urba n amenitie s an d transportatio n variable s wer e calculate d b y dividin g th e numbe r o f tota l amenitie s o r transportatio n facilitie s i n a quarter-mil e radiu s neighborhoo d b y th e maximu m o f th e numbe r o f tota l amenitie s o r transportatio n facilitie s i n a quarter-mil e radiu s neighborhood Neighborhoo d score s wer e calculate d b y dividin g th e numbe r o f sam e typ e agent s i n a neighborhoo d b y th e tota l numbe r o f cell s i n a neighborhood Th e siz e o f th e neighborhoo d i n calculatin g neighborhoo d scor e i s no t fixe d bu t controlled t o measur e th e conceptua l siz e o f th e neighborhoo d i n housin g locatio n decisio n process Finally tota l score s ar e calculate d dynamicall y a t eac h tim e step Figur e 4. 5 show s overlai d tota l scor e surfac e fo r thre e agen t type s a t on e tim e step Tabl e 4. 7 Controllabl e mode l parameter s Categor y Attribut e Urba n Amenitie s Recreationa l Amenities : bal l park re c center smal l par k "Th e Thir d Places*" : cafe restaurant ba r Cultura l Amenities : opera museum librar y Shopping : grocery shoppin g center smal l retai l Transportatio n Network s Neighborhood s Weight s Mobilit y Publi c transportation highwa y ramp majo r arteria l 3X3 5X5 7X 7 neighborhoo d siz e a weight s o n amenit y scor e b weight s o n transportatio n scor e c weight s o n neighborhoo d scor e In-migratio n rat e Out-migratio n rat e Move-withi n rat e "Oldenbur g (1989 ) 6 8

PAGE 83

Figur e 4. 5 Compariso n o f th e Tota l Scor e Surface s amon g Thre e Agen t Type s 4.4. 2 Constructio n o f a n Agent-Base d Mode l 4.4.2. 1 Simulatio n Mode l I n thi s study a n AB M mode l wa s constructe d o n a JAVA-base d platform Agent s ar e populate d i n a 10 0 X 6 0 cel l lattic e wit h a cel l siz e o f 10,00 0 squar e fee t base d o n 199 0 Censu s dat a a t th e Bloc k level Th e focu s o f thi s AB M simulatio n i s th e spatia l distributio n o f agent s i n th e urba n area A t eac h tim e ste p (representin g on e year) a numbe r o f agent s i n eac h grou p ar e generate d base d o n historica l data an d the y star t searchin g fo r thei r housin g locations A t th e sam e tim e step a numbe r o f agent s i n eac h grou p wh o wer e previousl y locate d i n th e urba n are a decid e t o re-locat e i n (1 ) differen t neighborhood s i n 6 9

PAGE 84

th e stud y are a o r (2 ) differen t location s outsid e th e stud y area A t eac h scannin g stage agent s selec t a locatio n base d o n a selectiv e combinatio n o f individua l preferences includin g neighborhoo d composition ; transportatio n network : an d th e leve l o f amenities whic h include s amount type an d distribution Th e mode l dynamicall y generate s severa l output s durin g an d afte r th e completio n o f th e simulation : (1 ) overal l spatia l distributio n o f agents (2 ) numbe r o f agent s i n eac h grou p i n eac h neighborhood an d (3 ) similarit y inde x representin g agen t distributio n patterns Th e similarit y index i s calculate d a s follows : 1 Similarity = — V p l ,, (4.4 ) P= N where n represents the total number of agents Pi represents the percentage of the same type agents in agent i's neighborhood STi represents the number of same type agent in agent i's neighborhood N represents the total number of cells in a neighborhood 7 0

PAGE 85

Creatin g Urba n Infrastructur e Populatin g Urba n Amenitie s Agent s Creatio n Searc h fo r Hom e I GI S Dat a QCE W Dat a Censu s Dat a Li t Revie w Intervie w AH P JAV A 1 I n n A GI S Outpu t Figur e 4. 6 Computationa l Approac h 4.4.2. 2 Validatio n o f th e Mode l Validatio n concern s ho w well mode l outcome s represen t rea l syste m behavior Therefore validatio n involve s comparin g mode l output s wit h real-worl d observation s o r th e produc t o f anothe r mode l o r theor y assume d t o adequatel y characteriz e realit y (Parke r e t al 2003) Initially 199 0 histori c distributio n o f urba n amenitie s wer e geocode d t o th e GI S environmen t usin g QCE W data Then thre e heterogeneou s demographi c group s wer e populate d t o th e model an d th e simulatio n wa s executed Th e resultin g distributio n o f differen t demographi c group s wa s compare d wit h th e 200 0 U S Censu s dat a a t th e Bloc k leve l fo r mode l evaluation 7 1

PAGE 86

Sinc e mode l outcome s an d th e Censu s dat a hav e differen t spatia l units mode l outcome s ha d b e converte d t o comparabl e units I n thi s study th e proportio n compariso n metho d wa s used Cell-base d mode l outcome s wer e groupe d base d o n th e Censu s Bloc k boundary an d fo r th e Censu s dat a an d mode l outcome s together th e proportion s o f eac h agen t grou p i n a Censu s Bloc k boundar y wer e calculated Th e difference s i n th e proportion s o f eac h agen t grou p betwee n th e Censu s dat a an d mode l outcome s wer e average d t o generat e a mode l discrepanc y index Aggregated Discrepancy Index = 2M f 3 ^\Ra-Rsi\ 1= 1 V (4.5 ) where Rci i s th e rati o o f agen t typ e i i n th e realit y Rsi i s th e rati o o f agen t typ e / i n th e simulatio n resul t n i s th e numbe r o f Censu s Block s 7 2

PAGE 87

Agen t distribution 199 0 (Real unknown ) Figur e 4. 7 Conceptua l Diagra m o f Mode l Ru n an d Mode l Validatio n Proces s Severa l simulatio n scenario s wer e develope d fo r validation First t o tes t th e significanc e o f urba n amenit y variable s an d transportatio n variables eac h se t o f variable s a s well a s combine d set s o f variable s wer e applie d i n a mode l ru n an d th e mode l result s wer e compare d wit h 200 0 Censu s data Additionally thre e differen t size s o f neighborhood s (3X3 5X5 an d 7X 7 cel l neighborhoo d size ) wer e applie d t o th e simulatio n mode l t o tes t th e conceptua l boundar y o f th e neighborhoo d whe n peopl e mak e housin g locatio n decision s (Tabl e 4.8) 7 3

PAGE 88

Second t o measur e th e comparativ e importanc e betwee n urba n amenitie s an d transportatio n variable s an d neighborhoo d preference th e weight s assigne d t o eac h se t o f variable s wer e controlled (Tabl e 4.9) Third t o tes t th e explanator y power s o f th e differen t amenity-base d urba n residentia l locatio n theories variable s selecte d fro m Consume r Cit y an d Creativ e Cente r literatur e wer e applie d t o th e model Th e ful l model whic h include s al l urba n amenit y variable s wa s als o include d fo r compariso n purposes Finally th e bes t mode l showin g th e bes t performanc e wa s selecte d fo r eac h stud y area an d additionall y transportatio n an d neighborhoo d variable s wer e teste d wit h th e bes t mode l (Tabl e 4.10) Tabl e 4. 8 Simulatio n scenario s Amenit y an d Transportatio n Variable s ^~ 11 Amenit y accessibilit y 3X 3 5X 5 7X 7 12 Amenit y accessibilit y + Amenit y densit y 3X 3 5X 5 7X 7 13 Transportatio n accessibilit y 3X 3 5X 5 7X 7 14 Transportatio n accessibilit y + Transportatio n Densit y 3X 3 5X 5 7X 7 T ^ Amenit y accessibilit y + Amenit y densit y v
PAGE 89

Tabl e 4. S II1 II2 II3 II4 II5 II6 II7 II8 II9 11-1 0 11-1 1 11-1 2 11-1 3 11-1 4 11-1 5 11-1 6 11-1 7 11-1 8 11-1 9 11-2 0 11-2 1 11-2 2 11-2 3 11-2 4 11-2 5 11-2 6 11-2 7 ) Weigh t sensitivit y Variabl e Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d Neighborhoo d tes t Weigh t 0 0 0 0 0 0 0 0 0 2 2 2 2 2 2 2 2 2 Variabl e Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transporta t Transportat ] Transportat Transportat Transporta t Transporta t Transportat Transportat Transportat Transportat Transportat Transportat Transportat Transportat Transportat io n io n io n io n io n io n io n io n o n o n o n o n o n o n o n o n o n o n o n o n o n o n o n o n o n o n o n Weigh t 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 Variabl e Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Amenit y Weigh t 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 Tabl e 4.1 0 Urba n amenity-base d theor y tes t Tes t Mode l IIIl III2 III3 III4 III5 Consume r Cit y mode l Creativ e Cente r mode l Ful l mode l Bes t mode l Bes t mode l + + Transportatio n Transportatio n + Neighborhoo d 7 5

PAGE 90

4.4. 3 Futur e Simulatio n 4.4.3. 1 Amenit y Distric t Polic y Simulatio n Man y loca l government s i n Nort h Americ a exercis e a numbe r o f plannin g policie s t o regulat e th e distributio n o f urba n amenitie s i n thei r boundaries Th e impac t o f thes e publi c polic y intervention s wa s teste d usin g scenario s designe d t o depic t variou s types amounts an d distribution s o f urba n amenities Scenario s wer e develope d t o reflec t rea l worl d situation s i n differen t plannin g regimes Scenario s fo r testin g include : eve n distributio n o f urba n amenitie s an d clusterin g o f urba n amenities Generate d spatia l distributio n pattern s o f agent s fro m eac h scenari o ar e analyze d an d th e result s infor m a discussio n o n polic y implications 4.4.3. 2 Locatio n Preferenc e Simulatio n Th e secon d simulatio n scenari o set s ar e no t relate d t o publi c polic y bu t ar e relate d t o individua l lifestyles Th e curren t mode l assume s tha t agent s chang e thei r locatio n preference s ove r time Fo r example eac h demographi c grou p ha s differen t locatio n preferences an d eac h individua l agent' s locatio n preferenc e i s governe d b y hi s o r he r group I n th e curren t model a n agent' s demographi c grou p categorizatio n i s change d a s tim e ste p progresse s (i.e fro m middl e ag e t o th e senio r groups) However i t i s possibl e tha t on e agent' s locationa l preferenc e i s no t changin g ove r tim e (i.e considerin g th e wealt h an d healt h statu s o f bab y boomers i t i s probabl e t o assum e tha t the y wil l preserv e thei r activ e an d entertainment-oriente d lifestyl e fo r lon g period s o f time) Therefore tw o differen t scenarios on e wit h changin g locationa l preference s an d anothe r wit h non changin g locationa l preferences wer e prepare d an d tested 7 6

PAGE 91

4.4.3. 3 Differen t Siz e o f Neighborhoo d Simulatio n T o tes t ho w agents neighborhoo d concep t an d siz e interac t wit h neighborhoo d preference s an d lea d t o comple x clusterin g behavior th e neighborhoo d siz e wa s controlled Th e simulation s wer e execute d wit h fou r differen t neighborhoo d sizes : n o neighborhood 3X3 5X5 an d 7X 7 cel l siz e neighborhood Considerin g th e curren t tren d o f urba n resurgenc e an d th e formatio n o f urba n neighborhood i t i s plausibl e t o assum e tha t th e siz e o f conceptua l neighborhoo d wil l chang e i n th e foreseeabl e future Heterogeneou s distributio n pattern s o f th e agen t distributio n ar e expecte d wit h varyin g size s o f th e neighborhood 7 7

PAGE 92

5 RESULT S I n thi s chapter I repor t th e simulatio n result s fro m th e explanator y a s wel l a s prognosti c use s o f th e model Bot h explanator y an d prognostic s model s ar e pilot-teste d agains t virtua l urba n space s an d the n applie d wit h th e rea l data I n th e explanator y use 2 9 combination s o f th e locatio n decisio n variabl e wer e teste d i n eac h stud y area I n th e prognosti c use thre e set s o f futur e scenario s wer e tested 5. 1 Intervie w Result s Table s 5.1-5. 3 sho w th e AH P surve y result s fro m th e stud y areas Th e weight s i n th e table s represen t th e comparativ e importanc e o f eac h locatio n decisio n variabl e b y agen t type s i n eac h stud y area Thi s comparativ e importanc e become s th e weigh t fo r th e locatio n decisio n variable s i n th e simulatio n model I n th e Boulde r an d Denve r stud y areas surve y result s sho w simila r pattern s tha t "bar pub nightclub a s th e mos t importan t locatio n decisio n facto r fo r youn g professiona l grou p an d "restaurant wa s mos t importan t fo r th e middl e ag e group I n Louisville "neighborhoo d park trail bik e trail wa s ranke d firs t fo r th e youn g professiona l a s wel l a s fo r th e middl e ag e group Fo r th e senio r group th e surve y result s fro m th e thre e stud y area s sho w heterogeneou s outcomes : "cafe coffe e shop ranke d firs t i n Boulder "opera orchestra musical ranke d firs t i n Denver an d "grocer y store ranke d firs t i n Louisville I n th e Boulde r stud y area fitness-relate d amenitie s lik e "neighborhoo d park trail bik e trail an d "recreatio n center ranke d highl y fo r th e youn g professiona l group "Grocer y store an d "bar pub nightclub ranke d highl y fo r middle ag e group an d "grocer y store an d "library bookstore ranke d highl y fo r th e senio r group Overall "th e thir d place categor y variable s ranke d highl y fo r al l thre e agen t groups 7 8

PAGE 93

I n th e Denve r stud y area "neighborhoo d park trail bik e trail an d "cafe coffe e shop ranke d highl y fo r bot h youn g professiona l an d middle ag e groups "Library bookstore an d "cafe coffe e shop ranke d highl y fo r th e senio r group Overall "th e thir d place an d "recreationa l amenity categor y variable s ranke d highl y fo r youn g professiona l an d middl e ag e groups an d "cultura l amenity categor y variabl e ranke d highl y fo r th e senio r group I n th e Louisvill e stud y area "recreatio n center ranke d highl y fo r bot h youn g professiona l an d middle ag e groups Also "cafe coffe e shop ranke d highl y fo r youn g professional s an d "grocer y store ranke d highl y fo r middle age s group "Neighborhoo d park trail bik e trail an d "smal l retai l store ranke d highl y fo r th e senio r group Overall simila r t o th e result s fro m Denver "recreationa l amenity categor y variable s ranke d highl y fo r youn g professiona l an d middl e ag e groups Fo r th e senio r grou p i n Louisville variable s i n th e "shopping categor y ranke d highl y i n th e surve y results Tabl e 5. 1 AH P surve y results Boulder C Q Bal l par k Re c cente r Park Trai l Caf e Restauran t Ba r Oper a Museu m Librar y Grocer y Cente r Smal l Publi c Ram p Arteria l Youn g Professional s Weigh t 0.0182 3 0.1383 1 0.1408 6 0.1148 3 0.0409 4 0.1747 2 0.0256 7 0.0219 8 0.0989 9 0.1259 1 0.0779 5 0.0266 6 0.1121 7 0.0191 6 0.0191 6 Ran k 1 5 3 2 5 9 1 1 1 1 2 7 4 8 1 0 6 1 3 1 3 Middl e Age s Weigh t 0.0155 6 0.1008 7 0.1320 3 0.1101 0 0.1709 1 0.1472 5 0.0576 0 0.0219 8 0.0857 4 0.1638 2 0.0104 8 0.0306 3 0.0962 4 0.0540 8 0.0340 7 Ran k 1 4 6 4 5 1 3 9 1 3 8 2 1 5 1 2 7 1 0 1 1 Senior s Weigh t 0.0095 9 0.0753 2 0.0665 4 0.1393 6 0.0632 2 0.0328 6 0.0607 5 0.0211 4 0.1197 4 0.1207 6 0.0049 6 0.0180 4 0.1096 0 0.0140 2 0.0513 1 Ran k 1 4 5 6 1 7 1 0 8 1 1 3 2 1 5 1 2 4 1 3 9 7 9

PAGE 94

Tabl e 5. 2 AH P Variabl e Bal l par k Re c cente r Park Trai l Caf e Restauran t Ba r Oper a Museu m Librar y Grocer y Cente r Smal l Publi c Ram p Arteria l Tabl e 5. 3 AH P Variabl e Bal l par k Re c cente r Park Trai l Caf e Restauran t Ba r Oper a Museu m Librar y Grocer y Cente r Smal l Publi c Ram p Arteria l surve y results Denver C O Youn g Professional s Weigh t Ran k 0.0933 7 0.0691 4 0.1410 1 0.0951 6 0.0439 4 0.3374 0 0.0156 3 0.0244 7 0.0413 4 0.0511 5 0.0541 1 0.0174 3 0.0312 2 0.0181 4 0.0163 0 4 5 2 3 8 1 1 5 1 1 9 7 6 1 3 1 0 1 2 1 4 Middl e Age s Weigh t Ran k 0.0897 2 0.0565 4 0.1729 0 0.1223 5 0.1759 1 0.1206 3 0.0620 3 0.0781 0 0.0993 0 0.0448 3 0.0249 7 0.0195 3 0.0251 6 0.0384 2 0.0241 3 surve y results Louisville C O Youn g Professional s Weigh t Ran k 0.0669 6 0.1409 4 0.2651 0 0.1039 0 0.0998 8 0.0791 9 0.0200 9 0.0197 4 0.0599 7 0.0615 8 0.0600 7 0.0233 6 0.0183 9 0.0424 4 0.0170 2 6 2 1 3 4 5 1 2 1 3 9 7 8 1 1 1 4 1 0 1 5 6 9 2 3 1 4 8 7 5 1 0 1 3 1 5 1 2 1 1 1 4 Middl e Age s Weigh t Ran k 0.0932 7 0.1547 4 0.1734 1 0.0600 6 0.0894 5 0.0762 5 0.0223 1 0.0518 1 0.0567 6 0.1324 6 0.1169 0 0.0531 6 0.0213 5 0.0423 4 0.0266 2 5 2 1 8 6 7 1 4 1 1 9 3 4 1 0 1 5 1 2 1 3 Senior s Weigh t 0.0336 3 0.0170 2 0.0671 1 0.1366 3 0.0652 2 0.0193 4 0.1954 9 0.1318 4 0.1647 0 0.0544 8 0.0180 3 0.0241 0 0.0618 9 0.0175 3 0.0274 4 Senior s Weigh t 0.0642 2 0.0797 8 0.1378 5 0.0668 6 0.0229 4 0.0166 5 0.0225 9 0.0147 7 0.0253 5 0.1447 2 0.0793 4 0.1356 1 0.0826 4 0.0429 3 0.0334 1 Ran k 9 1 5 5 3 6 1 2 1 4 2 8 1 3 1 1 7 1 4 1 0 Ran k 8 5 2 7 1 2 1 4 1 3 1 5 1 1 1 6 3 4 9 1 0 8 0

PAGE 95

5. 2 Explanatory Mode l Simulatio n Result s 5.2. 1 Pilo t Simulatio n Befor e testin g th e mode l usin g th e rea l data a simulatio n mode l wa s teste d agains t a virtua l urba n spac e fo r th e mode l verificatio n purpose : t o tes t th e mode l t o se e whethe r i t behave s a s expected Thi s proces s i s referre d t o a s testin g "inne r validity o f th e mode l (Brown 2005) A t th e initia l stage agent s i n eac h demographi c grou p wer e randoml y distributed Populatio n distributio n wa s simulate d ove r 3 0 years Fo r modelin g simplicity th e numbe r o f inflo w an d outflo w o f th e populatio n wa s fixed, an d i t wa s assumed tha t ther e wa s n o chang e i n tota l populatio n durin g th e simulatio n period I n th e 10 0 b y 6 0 gri d space ther e wer e 13 5 youn g professional 13 5 middle age an d 9 0 senio r agent s a t an y tim e o f th e simulation A t eac h stag e o f iteration 36 0 randoml y selecte d agent s leav e thi s space an d 4 0 randoml y selecte d agent s relocate d thei r housin g locatio n withi n thi s virtua l urba n space Simulatio n result s sho w th e increasin g similarit y inde x ove r time : fro m 40 % initiall y u p t o 77 % afte r th e 3 0 iterations Th e simulatio n mode l follow s th e logi c o f th e expecte d behavior s throug h compute r programming Yea r 2 0 Yea r 2 5 Yea r 3 0 Figur e 5. 1 Pilo t Simulatio n Result s 8 1

PAGE 96

8 0 7 5 7 0 6 5 6 0 5 5 5 0 4 5 4 0 3 5 3 0 i 1 1 r—-,-— , rr —-T—T— r—i—r— i r—r— ; 1 r—!~i ; r—r— , : 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293 0 Figur e 5. 2 Similarit y Inde x Change Pilo t Simulatio n 5.2. 2 Rea l Simulatio n 5.2.2. 1 Boulde r Afte r th e pilo t simulation residentia l locatio n decisio n proces s an d resultin g populatio n distributio n chang e wa s simulate d usin g th e rea l dat a ove r a ten-yea r perio d fro m 199 0 t o 2000 A t th e initia l stage agent s i n eac h demographi c grou p ar e distribute d base d o n 199 0 Censu s dat a a t th e Bloc k level : Ther e ar e 6,66 2 youn g professiona l agents 1,06 2 middl e ag e agents an d 1,00 0 senio r agent s i n th e Boulde r stud y area Tw o sam e typ e agent s occup y on e cel l i n th e urba n spac e fo r th e simulation Mode l parameter s wer e decide d base d o n th e histori c demographi c chang e data Th e tota l numbe r o f agent s increase d 1.6 % ove r th e 1 0 yea r period Annually 20 1 youn g professiona l agents 13 3 middl e ag e agents an d 3 5 senio r agent s move d int o thi s simulatio n environment. Also 35 5 randoml y selecte d agent s lef t thi s space an d 8 9 randoml y selecte d agent s relocate d thei r housin g locatio n withi n thi s spac e annually 8 2

PAGE 97

(1 ) Dominan t Agen t b y Censu s Block 199 0 (2 ) Dominan t Agen t b y Censu s Block 200 0 (3 ) Cell-base d Agen t Distribution 199 0 (4 ) Cell-base d Agen t Distribution 200 0 Figur e 5. 3 Rea l Agen t Distribution Boulder CO 190 0 an d 200 0 Si x scenario s wit h differen t variabl e set s wer e teste d an d th e simulatio n result s wer e compare d wit h 200 0 Censu s dat a (Figur e 5.4) Mos t mode l outcome s sho w a certai n degre e o f agreement wit h actua l demographi c distributio n from th e Censu s data : clusterin g o f middle age s i n th e Wester n par t o f th e stud y are a an d th e concentratio n o f youn g professional s i n th e East Considerin g th e fac t tha t abou t 52 % o f th e agent s relocate d durin g th e simulatio n period i t i s estimate d tha t th e model s showe d significan t level s o f accuracy 8 3

PAGE 98

I-1 Amenit y (Accessibility) \ ':•*• •: • ,;V_,,. : • -Wi j .3. • 1-2 Amenit y (Accessibilit y + Density ) 1-3 Transportatio n (Accessibility ) 1-4 Transportatio n (Accessibilit y + Densit y • •• r •_•_•_ • • H • • • • • 1J1 • • • P -'-H*4 • • ... .. .•"•? K • 1 • V • -.-•> : — • -.. •; • • •• -••'•• J t, :•.'• •"••ttf c 1-5 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility) Figur e 5. 4 Simulatio n Results Boulder C O 1-6 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e 5. 5 show s th e chang e o f th e similarit y inde x durin g th e ten-yea r simulatio n perio d i n th e Boulde r stud y area Durin g th e simulatio n period similarit y indexe s fo r mos t scenario s showe d simila r patterns : The y decrease d durin g th e firs t seven t o eight-yea r period an d the y starte d t o reboun d afte r tha t tim e period However tw o simulatio n scenario s (14 an d 1-6 ) wit h th e transportatio n networ k variable s di d no t sho w th e reboundin g pattern s an d continuall y decrease d durin g th e simulatio n period Severa l 8 4

PAGE 99

explanation s ar e possible : (1 ) I n th e Boulde r stud y area heterogeneou s set s o f urba n amenitie s preferre d b y differen t agen t group s tha t ar e spatiall y mixe d an d concentrate d i n smal l siz e locatio n (se e Figur e 5.6) di d no t generat e th e clusterin g o f th e sam e agen t typ e neighborhood (2 ) th e variance s betwee n weight s fo r th e sam e variable s b y agen t type s wer e to o smal l t o generat e clusterin g o f th e sam e typ e agen t neighborhoods Th e averag e AH P weigh t varianc e betwee n agen t type s wa s 0.0011423 (3 ) Afte r severa l years mos t preferre d area s b y al l agen t type s wer e fille d and i n less-preferre d areas clusterin g b y th e same-typ e agent s happened (4 ) Variance s fo r th e transportatio n variables however wer e to o smal l t o generat e reboun d i n th e scenario s 14 an d 16 (Averag e varianc e betwee n agen t typ e fo r urba n amenit y variable s wer e 0.001360 6 an d fo r transportatio n variable s wer e 0.0002690) Figur e 5. 5 Similarit y Inde x Change Boulder CO 1990-200 0 8 5

PAGE 100

Y Y Y Y O D Y Y Y 0 o o :Y MM S M Y 6 o t > o o o \ M Y w i v o S M "• T • • M M o Y o 0 o o Y Y Y O o Y O o o D O o o Y Y Y S M S o 8 Y O n O O o o o o o o O O 0 O 0 • 1 W s "M M ^ & Y Y : To p 3 Youn g Professiona l Amenities : Bar Park Re c Cente r M : To p 3 Middl e Ag e Amenities : Restaurant Grocery Ba r S : To p 3 Senio r Amenities : Cafe Grocery Librar y : Othe r Amenitie s Figur e 5. 6 Distributio n o f Urba n Amenitie s b y Agen t Type Boulder CO 199 0 5.2.2. 2 Denve r A t th e initia l stag e o f th e simulatio n i n th e Denve r stud y area ther e wer e 13,22 0 youn g professiona l agents 3,91 0 middl e ag e agents an d 2,51 5 senio r agent s base d o n th e Censu s data Denve r ha s a highe r populatio n an d househol d densit y tha n Boulder an d five same typ e agent s occup y on e cel l i n th e urba n spac e fo r th e simulation Also th e Denve r stud y are a show s a highe r rat e o f populatio n growt h whe n compare d wit h th e Boulde r stud y area Th e tota l numbe r o f agent s increase d 12.6 % ove r th e ten-yea r period Annually 74 8 youn g professiona l agents 33 1 middl e ag e agents an d 5 2 senio r agent s move d int o thi s simulatio n environment Also 88 5 randoml y selecte d agent s lef t thi s space an d 22 1 8 6

PAGE 101

randoml y selecte d agent s relocate d thei r housin g locatio n withi n thi s spac e annually I n th e Denve r stud y area 61 % o f agent s relocate d durin g th e simulatio n period A compariso n betwee n actua l 199 0 an d 200 0 demographi c distributio n show s th e concentratio n o f middl e ag e populatio n i n th e southeaster n par t o f th e stud y are a durin g th e ten-yea r period (1 ) Dominan t Agen t b y Censu s Block 199 0 (2 ) Dominan t Agen t b y Censu s Block 200 0 (3 ) Agen t Distribution 199 0 (4 ) Agen t Distribution 200 0 Figur e 5. 7 Rea l Agen t Distribution Denver CO 190 0 an d 200 0 Unlik e Boulder mos t mode l outcome s fai l t o sho w a hig h degre e o f agreemen t wit h actua l demographi c distributio n fro m th e Censu s dat a (Figur e 5.8) : Mode l outcome s coul d no t predic t th e clusterin g o f middl e age s i n th e southeaster n par t o f th e stud y area ; instea d mode l outcome s sho w th e clusterin g o f middle age s i n th e norther n o r wester n par t o f th e stud y area 8 7

PAGE 102

1-1 Amenit y (Accessibili t 1-2. Amenit y (Accessibilit y + Densit ; 1-3 Transportatio n (Accessibility ) 1-4 Transportatio n (Accessibilit y + Density ) 1-5 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) Figur e 5. 8 Simulatio n Results Denver C O 1-6 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e 5. 9 show s th e tren d o f th e similarit y inde x chang e durin g th e ten-yea r simulatio n perio d i n Denve r stud y area Unlik e th e Boulde r stud y area th e similarit y indexe s fo r th e differen t scenario s d o no t sho w homogeneou s patterns : Som e sho w increasin g pattern s overal l (1-1 1-2,1-5) som e sho w n o chang e overal l (1-3 1-6) an d th e on e scenari o show s a n overal l decreasin g patter n (1-4) Severa l explanation s ar e possible (1 ) Sinc e th e simulatio n scenario s wit h urba n amenitie s onl y (11 an d 1-2 ) sho w th e increasin g pattern s

PAGE 103

overall i t i s presume d tha t differen t set s o f urba n amenitie s preferre d b y differen t agen t group s whic h ar e spatiall y separate d generate d thos e outcome s (Figur e 5.10) (2 ) I t i s als o probabl e tha t relativel y highe r degre e o f th e weigh t variance s betwee n agen t typ e i n th e stud y are a fo r th e urba n amenit y variable s strengthe n thes e outcome s (Averag e varianc e fo r urba n amenit y variabl e i s 0.0043737 ) (3 ) Scenario s wit h transportatio n variable s (1-3,I 4 an d 1-6 ) sho w overal l decreasin g o r n o changin g patterns an d i t i s presume d tha t th e smal l weigh t varianc e betwee n th e agen t type s fo r th e transportatio n variabl e generat e thes e outcome s (Averag e varianc e fo r th e transportatio n variabl e i s 0.00018723) Figur e 5. 9 Similarit y Inde x Change Denver CO 1990-200 0 8 9

PAGE 104

M,>YM S % K V MJ W s > M a MMV-YMS YM ^ • JM 3 M rn s M YM S M M A H M V M Y M Y M ..M M YM S Y M Y M Y M ftl I Y M Y M e Y M Y M • M JM VM S Y "YMS Y M Y M Y M Y M e o Y M e o • •• • W Y : To p 3 Youn g Professiona l Amenities ; Bar Park Caf e M : To p 3 Middl e Ag e Amenities : Restaurant Park Caf e S : To p 3 Senio r Amenities : Music Library Caf e • : Othe r Amenitie s M M YM M 'YM # M M ^ M ~ M M Figur e 5.1 0 Distributio n o f Urba n Amenitie s b y Agen t Type Denver CO 199 0 5.2.2. 3 Louisvill e A t th e initia l stag e i n th e Louisvill e stud y area ; ther e wer e 2,53 2 youn g professiona l agents 52 7 middl e ag e agents an d 24 0 senio r agents Louisvill e ha s th e lowes t populatio n an d househol d densit y amon g th e thre e stud y areas an d on e agen t occupie s on e cel l i n th e urba n spac e fo r th e simulation Louisville however show s th e highes t rat e o f populatio n growt h amon g thre e stud y areas an d th e tota l numbe r o f th e agent s increase d 25.2 % ove r th e ten-yea r perio d betwee n 199 0 an d 2000 Annually 12 3 youn g professiona l agents 10 4 middl e ag e agents an d 2 1 senio r agent s move d int o thi s simulatio n environment. Also 16 5 randoml y selecte d agent s lef t thi s space an d 4 1 randoml y selecte d agent s relocate d thei r housin g locatio n withi n thi s space I n Louisvill e stud y area 70 % o f agent s relocate d durin g 9 0

PAGE 105

th e simulatio n period A compariso n betwee n 199 0 an d 200 0 demographi c distributio n show s th e clusterin g o f middl e age s i n th e northwester n par t o f th e stud y are a an d th e clusterin g o f th e senior s i n th e northeaster n part (1 ) Dominan t Agen t b y Censu s Block 199 0 (2 ) Dominan t Agen t b y Censu s Block 200 0 (3 ) Agen t Distribution 199 0 (4 ) Agen t Distribution 200 0 Figur e 5.1 1 Rea l Agen t Distribution Louisville CO 190 0 an d 200 0 Mos t mode l outcome s sho w a degre e o f agreemen t wit h actua l demographi c distributio n fro m th e Censu s dat a (Figur e 5.12) : clusterin g o f th e senio r populatio n i n th e northeaster n par t o f th e stud y area However th e mode l outcome s d o no t predic t th e clusterin g o f middl e age s i n th e northwester n par t o f th e stud y area 9 1

PAGE 106

1-1 Amenit y (Accessibilit y 1-2 Amenit y (Accessibilit y + Density ) 1-3 Transportatio n (Accessibility ) 1-4 Transportatio n (Accessibilit y + Density ) 1-5 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility) 1-6 Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e 5.1 2 Simulatio n Results Louisville C O Figur e 5.1 3 show s differen t pattern s o f th e similarit y inde x chang e i n th e Louisvill e stud y are a t o th e othe r tw o stud y areas Mos t simulatio n scenario s sho w droppin g pattern s afte r showin g a n increas e o f similarit y inde x fo r th e firs t tw o t o fiv e year s wit h a n exceptio n o f 15 scenario Possibl e explanation s are : (1 ) Differen t set s o f urba n amenitie s preferre d b y differen t agen t group s whic h wer e spatiall y separated an d thi s amenit y distributio n patter n generate d clusterin g o f th e agent s (Figur e 5.14) (2 ) Afte r severa l year s o f movin g an d 9 2

PAGE 107

relations mos t preferre d area s wer e fille d wit h agents and i n less-desirabl e areas wher e th e amenitie s ar e evenl y distributed agent s ar e mixe d an d generate d lo w similarit y indexes (3 ) Transportatio n variables onl y scenario s (13 an d 1-4 ) sho w rapidl y droppin g similarit y indexes Again lo w varianc e betwee n agen t type s fo r th e transportatio n variable s explain s thes e outcome s (Averag e varianc e fo r urba n amenit y variabl e i s 0.001392 2 an d fo r transportatio n variabl e i s 0.0004612 ) Figur e 5.1 3 Similarit y Inde x Change Louisville CO 1990-200 0 9 3

PAGE 108

(CM S m YM S YM S YM S YM S YM S YM S YM S Y M MS S S Y oY M YM S YM S YM S YM S YM S YM S YM S YM S Y : To p 3 Youn g Professiona l Amenities : Park Re c Center Caf e M : To p 3 Middl e Ag e Amenities : Park Re c Center Grocer y S : To p 3 Senio r Amenities : Grocery Park Smal l Retai l : Othe r Amenitie s YM S YM S M S Figur e 5.1 4 Distributio n o f Urba n Amenitie s b y Agen t Type Louisville CO 199 0 5.2.2. 4 Mode l Evaluatio n T o evaluat e th e mode l performance aggregate d discrepanc y indexe s wer e calculate d fro m eac h scenari o outcom e i n eac h stud y area Th e simulatio n wa s execute d 1 0 time s fo r eac h scenario an d th e averag e value s o f th e aggregate d discrepanc y indexe s ar e reporte d (Tabl e 5.4) Figur e 5.1 5 show s box-plotte d inde x value s fro m multipl e mode l runs A one-wa y analysi s o f varianc e (ANOVA ) tes t wa s execute d t o compar e th e mean s o f th e discrepanc y indexe s fro m si x simulatio n scenario s i n eac h stud y area F-tes t result s i n th e Tabl e 5. 5 sho w evidenc e agains t th e nul l hypothesi s tha t al l mean s ar e th e same ANOV A tes t confirm s th e heterogeneit y o f th e simulatio n outcome s b y differen t scenarios 9 4

PAGE 109

Overall al l th e simulatio n outcome s fro m Boulde r showe d th e lowes t aggregate d discrepanc y indexe s an d thu s th e bes t mode l performance O n average Boulde r model s hav e 0.144 4 discrepanc y inde x whic h mean s th e averag e differenc e o f agen t proportion s betwee n realit y an d th e mode l outcome s i s 14.4% Fo r al l thre e stud y areas th e transportatio n mode l (transportatio n accessibilit y plu s transportatio n densit y model 1-4 ) showe d th e bes t performance Th e urba n amenit y mode l (amenit y accessibilit y plu s amenit y densit y model 1-2 ) showe d lowe r performanc e tha n th e transportatio n mode l i n al l thre e stud y areas Counter-intuitively th e combine d mode l (amenit y accessibilit y + amenit y densit y + transportatio n accessibilit y + transportatio n density 1-6 ) showe d lowe r performanc e tha n th e urba n amenit y mode l o r th e transportatio n model Possibl e explanatio n i s tha t th e distributio n o f urba n amenitie s an d th e distributio n o f th e transportatio n facilitie s ar e spatiall y separated an d th e gravit y force s fro m tw o heterogeneou s set s o f variable s wer e neutralize d i n th e combine d model Also i t i s possibl e t o presum e that i n th e rea l world ther e ar e tw o type s o f peopl e i n eac h agen t group : urba n amenit y seeker s wh o emphasiz e th e proximit y t o urba n amenitie s an d dail y commuter s wh o emphasiz e th e proximit y t o th e transportatio n networks I n th e combine d model i t i s possibl e tha t bot h type s o f agent s en d u p i n location s somewher e betwee n urba n amenitie s an d transportatio n networks O 220 0 O 210 0 O 200 0 O.190 0 0.180 0 O 1 70 0 O 160 0 O.150 0 O 140 0 0.130 0 ^ ^ M 12 13 14 15 Boulde r Denve r Louisvill e 16 Figur e 5.1 5 Bo x Plotte d Aggregate d Discrepanc y Inde x 9 5

PAGE 110

Tabl e 5. 4 Aggregate d discrepanc y inde x (average d valu e afte r 1 0 iterations ) Variable s Boulde r Denve r Louisvill e Discrepanc y betwee n 199 0 an d 200 0 Censu s Dat a 0.144 1 0.109 1 0.186 9 11 Amenit y Accessibilit y 0.150 4 0.174 0 0.179 6 12 Amenit y Accessibilit y + Amenit y Densit y 0.143 6 0.162 5 0.165 3 13 Transportatio n Accessibilit y 0.142 6 0.169 9 0.163 9 14 Transportatio n Accessibilit y + Transportatio n Densit y 0.137 8 0.159 6 0.154 3 1 5 Amenit y Accessibilit y + Amenit y Densit y + 0 146 4 0 167 4 0 175 3 Transportatio n Accessibilit y 16 Amenit y Accessibilit y + Amenit y Densit y + Q ^ Q ^ Q ^ Transportatio n Accessibilit y + Transportatio n Densit y Averag e Tabl e 5. 5 ANOV A analysi s result s fo r < Boulde r Denve r Louisvill e Sourc e Betwee n Group s Withi n Group s Tota l Betwee n Group s Withi n Group s Tota l Betwee n Group s Withi n Group s Tota l D F 5 5 4 5 9 5 5 4 5 9 5 5 4 5 9 :omparin g th e S S 0.00088 7 0.00058 4 0.00147 1 0.00132 5 0.00048 4 0.00180 9 0.00400 3 0.00108 8 0.00509 1 0.144 4 discrepanc y i M S 0.00017 7 1.08E-0 5 0.00026 5 8.95E-0 6 0.00080 1 2.01E-0 5 0.166 7 ndexe s F 16.3836 7 29.5994 4 39.7344 1 0.167 9 P-valu e 8.21E-1 0 2.45E-1 4 6.43E-1 7 9 6

PAGE 111

5.2. 3 Additio n o f Neighborhoo d Variabl e 5.2.3. 1 Boulde r Thre e differen t size s o f neighborhood s (3X3 5X5 an d 7X 7 cel l neighborhoo d size ) wer e applie d t o th e simulatio n mode l (Se e Appendi x B fo r th e graphics) an d th e aggregate d discrepanc y inde x wa s calculated O n average simulatio n outcome s wit h neighborhoo d variable s fo r al l si x modelin g scenario s d o no t sho w a n improve d mode l performance Also th e mode l performanc e decrease d wit h th e increase d siz e o f neighborhood I t i s assume d tha t neighborhoo d compositio n wa s no t a significan t facto r whe n householder s decide d upo n a residentia l locatio n i n Boulder Interestingly th e bes t performanc e modelin g scenari o withou t a neighborhoo d variable whic h wa s th e transportatio n accessibilit y plu s transportatio n densit y mode l (1-4) di d no t outperfor m othe r scenario s wit h neighborhoo d variable s i n al l thre e neighborhoo d applications Instead th e 14 scenari o performe d lowes t i n al l neighborhoo d siz e tests I n a 3X 3 neighborhoo d case th e amenit y accessibilit y plu s amenit y densit y plu s transportatio n accessibilit y mode l (1-5 ) performe d best I n a 5X 5 case th e transportatio n accessibilit y mode l (1-3 ) performe d best I n a 7X 7 case amenit y accessibilit y plu s amenit y densit y mode l (1-2 ) performe d best Figur e 5.1 6 show s th e chang e o f th e similarit y inde x durin g th e ten-yea r simulatio n perio d i n th e Boulde r stud y are a wit h 3X 3 neighborhood Al l scenario s wit h differen t size s o f th e neighborhoo d showe d a linea r increas e o f th e similarit y inde x ove r th e simulatio n period 9 7

PAGE 112

Tabl e 5. 6 Aggregate d discrepanc y inde x wit h neighborhoo d variable Boulder C O Variable s 3X 3 5X 5 7X 7 11 Amenit y Accessibilit y 12 Amenit y Accessibilit y + Amenit y Densit y 13 Transportatio n Accessibilit y 0.156 1 0.157 5 0.162 2 0.156 4 0.160 3 0.153 8 0.155 6 0.155 3 0.159 1 14 Transportatio n Accessibilit y + Transportatio n Densit y 0.157 4 0.158 2 0.162 6 Amenit y Accessibilit y + Amenit y Densit y + 15 16 Transportatio n Accessibilit y Amenit y Accessibilit y + Amenit y Densit y + Transportatio n Accessibilit y + Transportatio n Densit y 0.154 1 0.158 1 0.154 8 0.157 1 0.155 5 0.157 0 Averag e 0.156 1 0.157 5 0.158 2 Figur e 5.1 6 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Boulder CO 1990-200 0 9 8

PAGE 113

5.2.3. 2 Denve r O n average simulatio n outcome s wit h a 3X 3 neighborhoo d variabl e outperforme d th e withou t neighborhoo d variabl e cas e an d othe r tw o neighborhoo d case s wit h th e bigge r neighborhoo d sizes Th e mode l performanc e decrease d wit h th e increase d siz e o f th e neighborhood Th e transportatio n accessibilit y plu s transportatio n densit y mode l (1-4 ) showe d th e bes t performanc e i n al l neighborhoo d siz e applications I n contras t t o th e cas e o f th e Boulde r stud y area i t i s assume d tha t hom e seeker s conside r neighborhoo d compositio n whe n the y decide d upo n a residentia l locatio n i n th e Denve r stud y area However th e conceptua l boundar y o f th e neighborhoo d i s presume d t o b e small o r householder s onl y sca n th e compositio n o f th e immediat e neighborhoo d fo r potentia l housin g locations Figur e 5.1 7 show s th e chang e o f th e similarit y inde x durin g th e ten-yea r simulatio n perio d i n th e Denve r stud y are a wit h th e 3X 3 neighborhood Al l scenario s wit h a neighborhoo d variabl e sho w a linea r increas e o f th e similarit y inde x ove r th e simulatio n period However th e slope s o f th e line s fo r al l scenario s decrease d slightl y ove r time 9 9

PAGE 114

Tabl e 5. 7 Aggregate d discrepanc y inde x wit h neighborhoo d variable Denver C O Variable s 3X 3 5X 5 7X 7 11 Amenit y Accessibilit y 12 Amenit y Accessibilit y + Amenit y Densit y 13 Transportatio n Accessibilit y 0.166 9 0.173 4 0.176 7 0.167 1 0.174 3 0.172 5 0.164 5 0.170 6 0.170 4 14 Transportatio n Accessibilit y + Transportatio n Densit y 0.162 9 0.168 1 0.165 6 Amenit y Accessibilit y + Amenit y Densit y + 15 16 Transportatio n Accessibilit y Amenit y Accessibilit y + Amenit y Densit y + Transportatio n Accessibilit y + Transportatio n Densit y 0.165 0 0.172 7 0.172 5 0.165 8 0.169 5 0.172 7 Averag e 0.165 4 0.171 4 0.171 7 Figur e 5.1 7 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Denver CO 1990-200 0 10 0

PAGE 115

5.2.3. 3 Louisvill e O n average simulatio n outcome s wit h neighborhoo d variable s fo r al l si x modelin g scenario s di d no t sho w improve d mode l performance Th e mode l performanc e decrease d substantiall y wit h th e increase d siz e o f neighborhood : Th e averag e aggregate d discrepanc y inde x fo r no-neighborhoo d applicatio n wa s 0.167 9 an d th e sam e inde x fo r 3X3 5X5 an d 7X 7 neighborhoo d applications wer e 0.1956 0.2059 an d 0.2068 respectively I t i s assumed tha t neighborhoo d compositio n wa s no t a significan t facto r whe n householder s decide d upo n a residentia l locatio n i n Louisville Interestingly lik e th e cas e o f th e Boulde r stud y area th e bes t performanc e modelin g scenari o withou t th e neighborhoo d variable whic h wa s th e transportatio n accessibilit y plu s transportatio n densit y mode l (1-4) di d no t outperfor m othe r scenario s wit h a neighborhoo d variabl e i n al l thre e neighborhoo d applications Th e 14 scenario however di d no t sho w th e lowes t performanc e i n an y neighborhoo d size s tests I n a 3X 3 neighborhoo d case th e ful l mode l (1-6 ) performe d best I n a 5X 5 case th e amenit y accessibilit y plu s amenit y densit y mode l (1-2 ) performe d best I n a 7X 7 case lik e a 3X 3 model th e ful l mode l performe d best Figur e 5.1 8 show s th e chang e o f th e similarit y inde x durin g th e ten-yea r simulatio n perio d i n th e Louisvill e stud y are a wit h a 3X 3 neighborhood Al l scenario s wit h a neighborhoo d variabl e sho w a n increas e o f th e similarit y indexe s ove r th e simulatio n period lik e th e Boulde r an d Denve r stud y areas Th e slope s o f th e line s fo r al l scenarios however wer e rapidl y decreasin g ove r time an d afte r eigh t t o nin e year s o f th e simulation th e indexe s stoppe d increasing A possibl e explanatio n i s tha t th e proces s o f assimilatio n durin g th e simulatio n reache d th e maximu m statu s afte r eigh t t o nin e year s (Se e graphic s i n Appendi x B) 10 1

PAGE 116

Tabl e 5. 8 Aggregate d discrepanc y inde x wit h neighborhoo d variable Louisville C O Variable s 3X 3 5X 5 7X 7 11 Amenit y Accessibilit y 12 Amenit y Accessibilit y + Amenit y Densit y 13 Transportatio n Accessibilit y 0.199 5 0.216 4 0.209 1 0.195 2 0.193 1 0.201 6 0.200 2 0.209 3 0.220 6 14 Transportatio n Accessibilit y + Transportatio n Densit y 0.199 2 0.207 4 0.212 7 15 Amenit y Accessibilit y + Amenit y Densit y + Transportatio n Accessibilit y T Amenit y Accessibilit y + Amenit y Densit y + Transportatio n Accessibilit y + Transportatio n Densit y Averag e 0.191 4 0.208 6 0.199 2 0.187 9 0.200 5 0.197 5 0.195 6 0.205 9 0.206 8 Figur e 5.1 8 Similarit y Inde x Chang e wit h 3X 3 Neighborhood Louisville CO 1990-200 0 10 2

PAGE 117

5.2. 4 Amenity-Base d Theor y Tes t T o evaluat e th e explanator y powe r o f th e existin g amenity-base d urba n developmen t an d locatio n decisio n theories thre e set s o f amenit y variables representin g differen t amenity base d theories wer e prepare d an d teste d agains t th e histori c demographi c distributio n pattern s o f th e thre e stud y areas First urba n amenit y variable s fro m Glaeser' s Consume r Cit y discussio n includ e th e following : bal l park restaurant, bar musi c venue an d store Second Florida' s Creativ e Cente r discussio n include s th e following : trail cafe musi c venue gallery an d store Finally th e ful l mode l tha t include s al l variable s fro m th e theorie s an d th e pilo t intervie w wa s simulated Lloyd' s Th e Cit y a s a n Entertainmen t Machin e discussio n include s ever y variabl e excep t th e larg e shoppin g center Afte r identifyin g th e best-performin g amenit y theory-base d model transportatio n variable s an d 3X 3 neighborhoo d variable s wer e als o adde d t o th e bes t model 5.2.4. 1 Boulde r I n th e Boulde r stud y area th e Creativ e Cente r mode l wa s identifie d a s th e bes t mode l describin g th e agen t distributio n patter n betwee n 199 0 an d 200 0 wit h th e lowes t discrepanc y inde x o f 0.1409 Th e additio n o f transportatio n variable s increase d th e explanator y powe r o f th e mode l wit h th e inde x o f 0.1396 bu t addin g th e neighborhoo d variabl e di d no t improv e th e mode l performance 10 3

PAGE 118

Tabl e 5. 9 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Boulder C O Discrepanc y Inde x IIIl Consume r Cit y III2 Creativ e Cente r III3 Ful l Mode l 0.148 5 0.140 9 0.143 6 III4 Creativ e Cente r III5 Creativ e Cente r + Transportatio n 0.139 6 + Transportatio n + Neighborhoo d 0.155 1 Figur e 5.1 9 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Boulder C O 10 4

PAGE 119

(1 ) Consume r Cit y Mode l (2 ) Creativ e Cente r Mode l .!•. • ". < • • -. >•• --ft: R (3 ) Ful l Mode l (4 ) Creativ e Cente r + Transportatio n Mode l (5 ) Creativ e Cente r + Transportatio n + 3X 3 Neighborhoo d Mode l Figur e 5.2 0 Amenity-Base d Theor y Tes t Results Boulder CO 199 0 an d 200 0 10 5

PAGE 120

5.2.4. 2 Denve r I n th e Denve r stud y area th e Consume r Cit y mode l wa s identifie d a s th e bes t mode l describin g th e agen t distributio n patter n betwee n 199 0 an d 200 0 wit h th e discrepanc y inde x o f 0.1605 Th e addition o f transportatio n variable s an d th e neighborhoo d variable however di d no t improv e th e mode l performance Tabl e 5.1 0 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Denver C O Discrepanc y Inde x ra-i III2 III3 III4 III5 Consume r Cit y Creativ e Cente r Ful l Mode l Consume r Cit y Consume r Cit y + Transportatio n + Transportatio n + Neighborhoo d 0.160 5 0.171 3 0.162 5 0.164 5 0.164 8 Figur e 5.2 1 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Denver C O 10 6

PAGE 121

(1 ) Consume r Cit y Mode l (2 ) Creativ e Cente r Mode l (3 ) Ful l Mode l (4 ) Consume r Cit y + Transportatio n Mode l (5 ) Consume r Cit y + Transportatio n + 3X 3 Neighborhoo d Mode l Figur e 5.2 2 Amenity-Base d Theor y Tes t Results Denver CO 199 0 an d 200 0 10 7

PAGE 122

5.2.4. 3 Louisvill e I n th e Louisvill e stud y area th e ful l mode l wit h al l variable s wa s identifie d a s th e bes t mode l describin g th e agen t distributio n patter n betwee n 199 0 an d 200 0 wit h th e discrepanc y inde x o f 0.1653 Lik e th e Denve r stud y area addition o f th e transportatio n variable s an d th e neighborhoo d variabl e di d no t improv e th e mode l performance Tabl e 5.1 1 Aggregate d discrepanc y inde x fo r amenity-base d theor y test Louisville C O Discrepanc y Inde x —••——pi—wfr^^wiiiii i • I \mm*m*mi I.I I u n i n 1 1 mum i i m i mi l iwtmm**m*mmm*mm\ > n in—•*—— — I N H i I • i n i ii j 111.H 1 III1 Consume r Cit y 0.170 7 III2 Creativ e Cente r 0.166 9 III3 Ful l Mode l 0.165 3 III4 Ful l Mode l + Transportatio n 0.169 4 III5 Ful l Mode l + Transportatio n + Neighborhoo d 0.187 9 0.20 0 ] 1 0.19 5 0.19 0 0.18 5 0.18 0 0.17 5 | 0.17 0 I 1 I 0.16 5 l 1 I l 0.16 0 1 • 0.15 5 0.15 0 I 1 1 1 — I 1 II M III2 III3 llt4 IU5 Figur e 5.2 3 Bo x Plot Discrepanc y Inde x fo r Amenity-Base d Theor y Test Louisville C O 10 8

PAGE 123

(1 ) Consume r Cit y Mode l (2 ) Creativ e Cente r Mode l (3 ) Ful l Mode l (4 ) Ful l Mode l + Transportatio n Mode l (5 ) Ful l Mode l + Transportatio n + 3X 3 Neighborhoo d Mode l Figur e 5.2 4 Amenity-Base d Theor y Tes t Results Louisville CO 199 0 an d 200 0 10 9

PAGE 124

5. 3 Prognosti c Mode l Simulatio n Result s 5.3. 1 Pilo t Simulatio n Severa l futur e simulatio n scenari o set s wer e teste d usin g th e virtua l urba n space First tw o differen t distributio n pattern s o f urba n amenitie s wer e prepared : (1 ) clustere d urba n amenitie s an d (2 ) eve n distribute d urba n amenitie s (Figur e 5.25) Th e simulatio n scenari o wit h clustere d urba n amenitie s generate d a larg e siz e o f th e same-typ e agen t neighborhood s an d th e eve n distributio n scenari o generate d a smalle r siz e o f th e neighborhoods Second tw o differen t scenarios on e wit h changin g locationa l preference s an d anothe r wit h non-changin g locationa l preferences wer e simulate d a t th e pilo t stud y area Figur e 5.2 6 show s differen t agen t distributio n pattern s b y th e scenarios The y generate d differen t agen t distributio n patterns bu t th e degree s o f th e concentratio n b y th e sam e typ e agen t wer e similar 11 0

PAGE 125

. I J • ZZZm :::: ; H I Ill l I I 1 1 )iiiiinTl i r iiiiiiiiiiniiii i niiiiiiiiinn n 'm i min i iinilllll l i i A 111 1 i i 111111 1 •- • 201 0 201 5 iiimiiiiiiiiiiiiii m iiiiiiiiiiiiiiiiiiiiiiiiin i mrmTinviTfl T iiiiiini m mrnm nmq p iiimiiiiiiimiiiimimiiimiwiiiiiimm i iiiiiiiiiiiiiiiiiiiiiiiiniiiiiiiimiiiiiiTTTTTT i iiimiiimiimimimiiiiiiiimiiuiiiim m 0 1OZ>3O5O6O7DeoSO10 O (1 ) Changin g Locatio n Preferenc e (2 ) Permanen t Locatio n Preferenc e Figur e 5.2 6 Pilo t Futur e Simulatio n Results : Changin g vs Permanen t Locatio n Preferenc e Finally simulatio n scenario s wit h differen t neighborhoo d size s wer e prepare d an d tested Th e large r conceptua l boundar y o f th e agent s generate d th e bigge r size s o f th e sam e typ e o f agen t neighborhoods J E 3X 3 VS VS 5X 5 7X 7 Figur e 5.2 7 Pilo t Futur e Simulatio n Results : Varyin g Neighborhoo d Size s 11 1

PAGE 126

5.3. 2 Futur e Simulatio n i n th e Stud y Area s Tw o set s o f th e futur e simulatio n scenario s wer e applie d t o th e thre e stud y areas : (1 ) changin g an d permanen t locatio n preference s an d (2 ) varyin g neighborhoo d sizes Th e bes t performanc e urba n amenity-base d model s wer e applie d fo r th e futur e simulatio n durin g 200 0 t o 202 0 periods Fo r th e Boulde r stud y area th e Creativ e Cente r plu s transportatio n variable s model ; fo r th e Denve r stud y area th e Consume r Cit y model ; an d fo r th e Louisvill e stud y area th e ful l mode l wa s simulated Location s o f urba n amenitie s wer e obtained fro m 200 0 QCE W data Tota l populatio n chang e i s assume d t o b e followin g th e histori c trend s i n eac h stud y area Figur e 5.2 8 show s th e result s fro m th e futur e simulatio n base d o n changin g locatio n preferenc e an d permanen t locatio n preferenc e i n thre e stud y areas Thre e graph s i n Figur e 5.2 9 show s th e chang e o f th e similarit y inde x durin g th e futur e simulatio n i n th e stud y areas I n al l thre e stud y areas th e futur e simulatio n scenari o outcome s wit h changin g locatio n preference generate d highe r similarit y indexes Thi s ca n b e interprete d a s that i f peopl e chang e thei r locatio n preferenc e a s the y age thi s behavio r wil l increas e th e possibilit y o f generatin g a homogeneou s neighborhood O n th e othe r hand i f peopl e preserv e thei r locatio n preferenc e an d hav e th e sam e preferenc e ove r thei r lif e course thi s wil l decreas e th e possibilit y o f generatin g a homogeneou s neighborhood Considerin g th e wealt h an d th e healt h statu s o f bab y boomers thi s i s a probabl e scenari o sinc e the y ar e expecte d t o preserv e thei r activ e an d entertainment-oriente d lif e styl e fo r a lon g perio d o f time Figure s 5.3 0 an d 5.3 1 sho w th e result s fro m th e futur e simulatio n base d o n varyin g neighborhoo d sizes Simila r t o th e outcome s fro m th e pilo t futur e simulation th e large r conceptua l boundar y o f th e agent s generate d th e bigge r size s o f th e homogeneou s neighborhood s i n al l stud y areas For m th e simulatio n result s betwee n 1990-2000 i t wa s assume d tha t peopl e i n th e stud y area s hav e relativel y smal l conceptua l neighborhoo d size s whe n the y mak e housin g locatio n decisions Th e futur e simulatio n outcome s show however tha t a smal l chang e i n people' s conceptua l siz e o f a neighborhoo d ma y generat e a 11 2

PAGE 127

bi g chang e i n th e demographi c distributio n o f th e neighborhoo d an d ma y generat e segregatio n betwee n neighborhoods (1 ) Changin g Preference Boulde r (2 ) Permanen t Preference Boulde r (3 ) Changin g Preference Denve r (4 ) Permanen t Preference Denve r (5 ) Changin g Preference Louisvill e (6 ) Permanen t Preference Louisvill e Figur e 5.2 8 Futur e Simulatio n Results : Changin g vs Permanen t Preferenc e 11 3

PAGE 128

200 1 200 2 200 3 200 200 S 200 S 200 7 200 8 20O S 201 0 201 1 20* 2 201 3 201 4 201 5 201 8 201 7 201 8 201 9 2O2 0 (1 ) Boulde r - — Changin g -•-Permanen t 2O0 1 200 2 200 3 200 4 2O0 S 200 8 20O 7 200 8 200 9 201 0 201 1 201 2 201 3 201 4 201 S 201 8 201 7 201 8 201 6 202 0 (2 ) Denve r (3 ) Louisvill e Figur e 5.2 9 Similarit y Inde x Change : Boulder Denver Louisville 2000-202 0 11 4

PAGE 129

^1 ) Creativ e Center N o Neighborhoo d (2 ) Creativ e Cente r + 3X 3 Neighborhoo d (3 ) Creativ e Cente r + 5X 5 Neighborhoo d (4 ) Creativ e Cente r + 7X 7 Neighborhoo d Figur e 5.3 0 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Boulder C O (1 ) Consume r City N o Neighborhoo d (2 ) Consume r Cit y + 3X 3 Neighborhoo d (3 ) Consume r Cit y + 5X 5 Neighborhoo d (4 ) Consume r Cit y + 7X 7 Neighborhoo d Figur e 5.3 1 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Denver C O 11 5

PAGE 130

(1 ) Ful l Model N o Neighborhoo d (2 ) Ful l Mode l + 3X 3 Neighborhoo d ^-tj^xv* (3 ) Ful l Mode l + 5X 5 Neighborhoo d (4 ) Ful l Mode l + 7X 7 Neighborhoo d Figur e 5.3 2 Futur e Simulatio n Results : Varyin g Neighborhoo d Sizes Louisville C O 11 6

PAGE 131

6 CONCLUSIO N Give n th e increasin g deman d fo r analytica l framework s tha t enabl e u s t o explai n th e contemporar y pattern s o f th e redistributio n o f th e populatio n aroun d citie s an d th e emergenc e o f amenity-base d urba n developmen t an d residentia l locatio n theories I becam e intereste d i n learnin g mor e abou t th e natur e an d functio n o f changin g residentia l locatio n decisio n factor s b y households Th e purpos e o f thi s stud y wa s t o identif y amenity-base d locatio n decisio n variable s an d t o tes t thei r explanator y powe r usin g th e agent-base d simulatio n method Afte r validatin g th e simulatio n model I teste d futur e scenario s a s well 6. 1 Researc h Finding s 6.1. 1 Intervie w Th e comparativ e importanc e o f th e locatio n decisio n variable s b y ag e group s i n eac h stud y are a wa s identifie d fro m interview s wit h realtor s i n th e stud y areas I n th e Boulde r stud y area th e "Thir d Place categor y variable s includin g cafe restaurant an d ba r wer e identifie d a s th e mos t importan t locatio n decisio n factor s fo r al l ag e groups I n th e Denve r stud y area th e 'Thir d Place categor y variable s a s wel l a s "Recreationa l Amenity categor y variable s suc h a s bal l park recreation center neighborhoo d park an d trai l wer e identifie d a s th e mos t importan t factor s fo r youn g professional s an d middl e ag e groups Fo r th e senio r group "Cultura l Amenity categor y variable s includin g musi c hall museum an d librar y wer e identifie d a s th e mos t importan t factors I n th e Louisvill e stud y area "Recreatio n Amenity categor y variable s wer e mos t importan t t o youn g professional s an d middl e ag e groups an d "Shopping categor y variable s includin g grocery shoppin g center an d smal l retai l wer e mos t importan t t o th e senior s group 11 7

PAGE 132

Stud y area s sho w differen t level s o f th e variance s betwee n weight s fo r th e sam e variable s b y differen t agen t types Th e Denve r stud y are a ha s th e highes t leve l o f variance s betwee n th e weight s overall I n th e Boulde r an d Louisvill e areas th e averag e variance s betwee n weight s wer e abou t on e thir d o f tha t o f th e Denve r stud y area I n al l thre e stud y areas th e averag e variance s fo r th e transportatio n variable s wer e smalle r tha n th e averag e variance s fo r th e urba n amenit y variables Thes e difference s i n th e leve l o f variance s betwee n th e stud y area s generate d heterogeneou s demographi c distributio n outcome s b y th e stud y areas Also th e differen t level s o f variance s betwee n urba n amenit y variable s an d transportatio n variable s als o generate d heterogeneou s outcome s b y differen t modelin g scenario s i n eac h stud y area Intervie w result s confirme d th e significanc e o f urba n amenitie s a s importan t residentia l locatio n decisio n factor s i n loca l housin g markets I t wa s als o foun d tha t urba n amenitie s hav e a differen t comparativ e importanc e i n differen t housin g markets Tabl e 6. 1 Averag e variance s betwee n weight s b y agen t type s Betwee n urba n amenit y variabl e weight s Betwee n transportatio n variabl e weight s Betwee n al l variabl e weight s Boulde r 0.00136 1 0.00026 9 0.00114 2 Denve r 0.00437 4 0.00018 7 0.00353 6 Louisvill e 0.00139 2 0.00046 1 0.00120 6 6.1. 2 Explanator y Mode l I n thi s study i t wa s assume d tha t peopl e decid e thei r housin g locatio n base d o n th e physica l environment s an d non-physica l o r socia l environments Physica l environment s includ e relativ e location s o f urba n amenitie s a s well a s th e relativ e location s o f traditiona l locatio n decisio n factor s lik e transportatio n network s t o potentia l housin g locations Non physica l o r socia l environment s includ e bot h positiv e an d negativ e socia l interaction s betwee n household s an d thei r neighbors 11 8

PAGE 133

I n th e simulatio n model th e interaction s betwee n th e agent s an d th e physica l environment s wer e modele d usin g th e weight s fro m AH P interview s wit h realtor s i n th e region Th e socia l interaction s i n th e mode l wer e teste d b y controllin g th e siz e o f th e conceptua l neighborhoo d durin g th e relocatio n process Severa l differen t combination s o f th e variable s wer e prepare d t o tes t differen t simulatio n scenario s an d th e simulatio n outcome s wer e compare d wit h th e actua l demographi c distributio n pattern s i n eac h stud y area Th e simulatio n scenario s wit h transportatio n variable s outperforme d th e scenario s wit h urba n amenit y variable s i n al l thre e stud y areas I t i s assumed tha t durin g th e simulatio n perio d (1990-2000) traditiona l locatio n decisio n factor s wer e stil l mor e significan t tha n urba n amenit y factor s durin g th e housin g locatio n decisio n processe s i n th e stud y areas Considerin g th e increase d discussion s abou t th e rol e o f urba n amenitie s i n attractin g huma n capita l an d th e cultura l trend s toward s entertainmen t an d consumption-oriente d lifestyles a differen t outcom e wit h highe r explanator y powe r fo r th e urba n amenit y variable s i s expecte d fo r th e simulatio n result s wit h mor e recen t data Inclusio n o f neighborhoo d factor s di d no t improv e th e mode l performanc e wit h th e exceptio n o f th e Denve r stud y are a wit h th e 3X 3 neighborhoo d size. I t i s assume d tha t peopl e prefe r th e age-mixe d communit y o r a t leas t the y d o no t conside r th e ag e structur e o f th e neighborhoo d whe n the y mad e thei r housin g locatio n decisions I t i s als o assumed peopl e browse d onl y nearb y neighborhood s eve n i f the y decide d t o conside r thei r potentia l neighborhood' s demographi c profiles Testin g th e explanator y powe r o f th e existin g amenity-base d urba n developmen t an d locatio n decisio n theorie s resulte d i n heterogeneou s outcome s i n eac h stud y area Th e Creativ e Cente r mode l wit h th e transportatio n variabl e performe d bes t i n th e Boulde r stud y area an d th e Consume r Cit y mode l performe d bes t i n th e Denve r stud y area an d th e ful l mode l wit h al l variable s performe d bes t i n th e Louisvill e stud y area Th e simulatio n result s fro m th e amenity-base d theorie s di d no t outperfor m th e simulatio n result s fro m th e transportatio n variables-onl y model s i n eac h stud y area Th e difference s betwee n them however wer e ver y small 11 9

PAGE 134

Th e researc h findings wit h distinc t best-performin g theorie s fo r differen t stud y area s war n th e potentia l risk s o f indifferen t application s o f a singl e "famous amenity-base d urba n developmen t theor y t o differen t loca l areas Simulatio n outcome s showe d tha t eac h stud y are a ha s a uniqu e se t o f urba n amenitie s tha t ar e effectiv e i n attractin g people a s well a s a uniqu e se t o f comparativ e importanc e amon g them A carefu l analysi s o f th e loca l condition s an d th e area-specifi c locatio n decisio n factor s ar e require d befor e th e generalizatio n o f th e amenity-base d developmen t theory Tabl e 6. 2 Compariso n o f aggregate d discrepanc y indexe s Transportatio n mode l Urba n amenit y mode l Amenit y theory-base d mode l Boulde r 0.137 8 0.143 6 0.139 6 Denve r 0.159 6 0.162 5 0.160 5 Louisvill e 0.154 3 0.165 3 0.165 3 6.1. 3 Prognosti c Mode l Usin g th e best-performin g amenit y theory-base d mode l i n eac h stud y area severa l futur e simulatio n scenario s wer e tested First whe n comparin g th e simulatio n outcome s b y differen t spatia l distributio n pattern s o f urba n amenities th e simulatio n scenari o wit h clustere d urba n amenitie s generate d a larg e siz e o f th e same-typ e agen t neighborhoods an d th e eve n distributio n scenari o generate d a smalle r siz e o f eac h neighborhood Thi s simulatio n outcom e showe d th e potentia l rol e o f urba n policy Eve n thoug h loca l government s canno t directl y contro l individua l housin g locations th e researc h result s showe d tha t the y ca n possibl y contro l th e demographi c distribution s i n thei r municipa l boundarie s b y regulatin g th e spatia l distributio n o f amenities Second whe n comparin g th e simulatio n outcome s fro m changin g locatio n preferenc e scenari o an d permanen t locatio n preferenc e scenario th e outcome s fro m th e changin g locatio n preferenc e scenari o represente d a highe r degre e o f clusterin g b y th e sam e typ e 12 0

PAGE 135

agents I t i s estimate d tha t th e incremen t o f th e health y an d wealth y senior s i n urba n area s woul d generat e a mor e age-mixe d communit y i n thes e areas Third th e simulatio n outcome s showe d increasin g size s o f th e homogeneou s cluster s o f th e sam e typ e agent s an d th e highe r leve l o f segregatio n betwee n th e cluster s wit h th e incremen t o f th e conceptua l neighborhoo d sizes Th e degre e o f clusterin g wa s muc h highe r i n th e neighborhoo d siz e tes t tha n i n th e locatio n preferenc e chang e test Th e simulatio n outcome s showe d tha t th e chang e i n people' s conceptua l neighborhoo d siz e hav e a mor e significan t impac t o n th e demographi c distributio n pattern s tha n th e chang e i n th e comparativ e importanc e o f urba n amenitie s i n thei r housin g locatio n decisio n process 6. 2 Significanc e o f Researc h I n th e literatur e revie w sectio n o f thi s study I suggeste d tha t conventiona l theoretica l framework s coul d no t full y explai n th e pattern s o f populatio n an d househol d distributio n o f contemporar y cities Althoug h th e theor y o f urba n residentia l locatio n ha s bee n a topi c o f interes t an d stud y i n academi a fo r severa l decades th e changin g natur e o f ou r urba n environmen t require s ne w element s t o thi s theory An y theor y develope d fo r thi s purpos e need s t o addres s th e uniqu e feature s o f urba n housin g markets First fro m a suppl y side th e theor y need s t o identif y th e pul l factor s whic h attrac t peopl e t o th e previousl y less-attractiv e locations Second fro m a deman d side th e theor y need s t o identif y th e instigator s o f ne w urba n trend s an d specif y thei r locationa l preferences Third th e theor y need s t o identif y th e powe r o f th e socia l interaction s betwee n member s o f th e cities Fourth th e theor y need s t o includ e heterogeneou s nature s o f household s i n differen t demographi c groups Fifth th e theor y need s t o b e abl e t o handl e th e possibilit y o f non-linea r behavio r an d multipl e equlibri a i n a loca l housin g market I n thi s study a n attemp t wa s mad e t o identif y th e element s o f thi s ne w theory First peopl e workin g i n th e urba n residentia l market realtor s wer e interviewe d t o discer n ne w housin g 12 1

PAGE 136

locatio n factor s includin g variou s form s o f urba n amenitie s an d thei r role s i n attractin g differen t demographi c groups Second tw o demographi c groups youn g professional s an d seniors wer e identifie d fro m th e intervie w an d b y reviewin g existin g studie s a s th e proponent s o f th e recen t redistributio n o f populatio n aroun d urba n areas Third th e rol e o f socia l interaction s i n a loca l housin g marke t wa s teste d b y controllin g th e siz e o f th e conceptua l neighborhoo d b y agents Fourth b y usin g agent-base d simulatio n approach th e heterogeneou s natur e o f differen t demographi c group s wa s modeled B y micro-controllin g heterogeneou s agent' s locationa l preferences I wa s abl e t o observ e changin g aggregate d outcome s a t th e macr o level Fifth simulatio n outcome s showe d non-linea r behavior s whe n agent s generate d th e sam e typ e agen t clusters I t i s difficul t t o expec t t o observ e thes e kind s o f behavior s fro m th e traditiona l modelin g approac h base d o n a homogeneou s populatio n wit h perfec t marke t information Additionally b y constructin g an d runnin g a mode l o f residentia l location s b y differen t demographi c groups i t wa s possibl e no t onl y t o tes t th e theor y o f spatia l pattern s o f residentia l developmen t bu t als o t o analyz e th e spatia l clusterings o f th e population s i n differen t socio-economi c groups A s Galste r (2002 ) argues housin g policie s i n th e U.S ha s increasingl y bee n see n a s a mean s o f achievin g th e wide r targe t o f reducin g socia l stratificatio n an d socia l exclusion Th e mode l result s wil l provid e planners polic y makers an d urba n scholar s a mor e profoun d understandin g o f urba n residentia l developmen t patterns an d wil l provid e a foundatio n fo r a n urba n theor y tha t i s potentiall y mor e policy relevant I n thi s study I hav e adapte d agent-base d modelin g approac h a s a researc h tool Also I hav e adapte d amenity-base d an d neighborhood-base d locatio n decisio n approache s i n constructin g agen t behaviora l rules Thi s researc h adde d ne w element s t o al l thre e existin g approache s adapte d i n th e study First t o th e agent-base d approach I'v e identifie d ne w locatio n decisio n variable s an d applie d t o th e model Mos t o f th e existin g agent-base d inne r cit y residentia l locatio n decisio n studie s applie d th e traditiona l locatio n decisio n factor s suc h a s distanc e t o CBD transportatio n networks an d som e environmenta l amenities Second t o amenity-base d locatio n decisio n approach I'v e teste d explanator y 12 2

PAGE 137

powe r o f th e existin g amenity-base d theorie s an d trie d t o evaluat e th e existin g theorie s usin g th e empirica l data Thi s kin d o f comparativ e stud y ha s bee n rarel y attempted Finally t o th e neighborhood-base d locatio n decisio n approach I'v e use d dynami c characteristic s o f th e people I use d age-base d classificatio n metho d i n m y model Mos t existin g neighborhood-base d locatio n decisio n studie d use d non-dynami c characteristic s o f peopl e suc h a s race economic wealth o r educationa l statu s a s locatio n decisio n factors 6. 3 Limitation s o f Researc h Ther e ar e severa l limitation s t o thi s study mos t o f whic h ar e limitation s o f th e dat a I use d fo r constructin g an d validatin g th e simulatio n model On e limitatio n wa s i n th e numbe r o f interviewee s i n eac h stud y area Fiv e realtor s wer e interviewe d i n eac h stud y area an d th e househol d behavior s i n th e housin g market s wer e filtere d throug h thei r responses I trie d t o minimiz e thi s limitatio n b y selectin g bes t performin g realtor s i n th e stud y area s wit h th e highes t sale s volume s an d year s o f experience : Mos t interviewee s ha d mor e the n 1 5 year s wort h o f experienc e i n thei r respectiv e markets Anothe r are a o f concer n i s th e limit s o f micro-leve l residentia l locatio n data Sinc e i t wa s impossibl e t o obtai n individua l householde r demographi c an d housin g locatio n dat a fo r th e entir e stud y areas th e househol d agent s wer e populate d base d o n th e aggregate d rati o a t eac h Censu s Bloc k a t th e initia l stag e o f th e simulation Also th e simulatio n outcome s wer e aggregate d an d compare d wit h th e Censu s Bloc k leve l data Th e discrepanc y inde x use d i n thi s stud y wa s a comparativ e indicato r o f th e mode l performanc e wit h othe r simulatio n scenario s an d th e inde x shoul d no t b e interprete d a s a n absolut e mode l performanc e evaluator I n thi s study si x simulatio n scenario s wer e use d initiall y wit h differen t set s o f locatio n decisio n variables an d the n neighborhoo d variable s wer e adde d t o th e model Also set s o f variable s fro m existin g theorie s wer e tested I t i s stil l possibl e tha t ther e ar e unteste d combination s tha t ma y resul t i n th e highe r mode l performances I t was however 12 3

PAGE 138

impossibl e t o tes t ever y possibl e combinatio n sinc e th e inpu t o f a variabl e whic h wa s independentl y significan t di d no t guarante e th e improvemen t o f th e mode l performance. v l A fina l are a t o b e note d pertain s t o th e natur e o f th e compute r simulatio n approac h i n urba n study An y urba n simulatio n mode l require s a certai n degre e o f simplificatio n o f th e rea l worl d phenomena A t th e curren t leve l o f abstraction th e simulatio n outcome s i n thi s stud y appea r sufficien t i n scop e fo r th e research' s purpose I t is however stil l possibl e tha t ther e i s a n unidentifie d optima l degre e o f abstractio n fo r thi s researc h purpose Decidin g a n appropriat e leve l o f abstractio n wil l remai n a n ongoin g challeng e t o modelers 6. 4 Futur e Researc h Agend a Th e model s wer e simulate d fo r th e ten-yea r perio d fro m 199 0 t o 200 0 an d th e mode l outcome s wer e compare d wit h 200 0 Censu s data A simulatio n fo r a longe r perio d o f tim e an d comparin g tha t wit h mor e recen t Censu s dat a woul d resul t i n increase d mode l validity Also applicatio n o f thi s mode l t o othe r stud y area s wit h variou s types amounts an d distribution s o f urba n amenitie s wil l increas e th e reliabilit y o f th e model I n thi s study th e adaptatio n o f a n agent-base d modelin g approac h showe d stron g potentia l a s a polic y testin g too l capturin g importan t informatio n abou t th e urba n housin g market T o maximiz e thi s potential mor e attempt s ar e neede d t o communicat e wit h al l thos e who m suc h modelin g wil l infor m an d t o appl y ABM s t o th e plannin g practices Achievin g mor e validit y an d increasin g th e reliabilit y o f th e mode l woul d b e prerequisites Th e transformatio n o f urba n areas i n term s o f demography business an d cultur e i s creatin g no t onl y ne w opportunitie s bu t als o ne w challenge s fo r planners polic y makers an d developers a s wel l a s scholar s wh o ar e studyin g cities A conceptually soun d locatio n theor y an d th e developmen t o f th e simulatio n mode l wit h th e theor y wil l provid e al l professional s wit h a mor e profoun d understandin g o f cities 12 4

PAGE 139

ENDNOTE S I Sixtee n o f th e 2 0 larges t citie s i n th e Unite d State s gaine d populatio n fro m 199 0 t o 200 0 (Suchman 2002) O n average homeownership educatio n levels an d racia l an d ethni c diversit y i n downtow n area s o f 4 4 majo r U.S citie s increase d betwee n 197 0 an d 200 0 (Birch 2005) Accordin g t o Glaese r e t al (2001) betwee n 196 0 an d 1990 th e rat e o f growt h o f commute s wher e th e househol d live s i n th e cit y increase d whil e th e growt h rat e o f commute s originatin g i n th e subur b fell Withi n cities th e high-incom e populatio n wa s movin g close r t o th e centra l downtow n area Luc y an d Phillip s (2006 ) argu e tha t th e cours e o f America n metropolita n histor y ma y b e reversin g a s centra l citie s ar e revivin g an d man y suburbs especially thos e wit h middle-age d housin g buil t betwee n 194 5 an d 1970 ar e sinkin g int o obsolescenc e an d decay Usin g censu s 200 0 data the y foun d tha t mos t centra l citie s increase d i n populatio n durin g th e 1970s 1980s an d 1990s The y als o foun d tha t pe r capit a incom e o f th e centra l cities White s o n averag e wer e highe r tha n th e Whites incom e i n suburb s o f metropolita n area s i n 2000 u p slightl y fro m 1990 Relativ e media n incom e declin e i n larg e citie s wa s attributabl e t o mor e minoritie s bein g present i n mos t instances rathe r tha n avoidanc e o f citie s b y middle an d upper-incom e Whites Usin g mor e recen t data the y foun d 1 5 o f 2 0 larg e centra l citie s increase d i n relativ e pe r capit a incom e (fou r decrease d an d on e remaine d th e same) an d 1 1 o f 2 0 increase d (seve n decrease d an d tw o remaine d th e same ) i n relativ e media n famil y incom e fro m 200 0 t o 2003 B y analyzin g residentia l developmen t pattern s fro m 199 6 t o 200 0 amon g 5 0 o f th e larges t M S A s i n th e U.S. Steinacke r (2003 ) foun d tha t citie s ar e constructin g a t leas t thei r fai r shar e o f ne w housing give n thei r geographi c size an d conclude d tha t th e amoun t o f residentia l developmen t i n citie s ha s bee n muc h greate r tha n th e medi a coverag e o f suburba n spraw l suggests I I Nort h America n municipalitie s recentl y adaptin g urba n policie s tha t ai m t o attrac t hig h huma n capita l includ e bu t no t limite d t o Tallahassee E l Paso Charlotte Duluth Tacoma Lon g Beach Virgini a Beach Toronto Montreal an d Ottawa 12 5

PAGE 140

" Boulde r selecte d a s on e o f th e "Bes t Place s t o Liv e Th e 3 0 Cooles t Neighborhood s i n America Men' s Journal 200 9 ranke d # 1 i n "America' s To p 2 5 Town s t o Liv e Well Forbes.com 200 9 ranke d # 1 i n "To p Te n Bes t Midsiz e Metropolita n Areas Bizjournals.com 200 9 ranke d a s a favorit e eco-friendl y smal l tow n i n 'Th e West' s 2 0 Bes t Smal l Towns Sunse t Magazine 200 9 ranke d # 1 i n "America' s Brainies t Metropolita n Areas America n Cit y Busines s Journals 200 5 ranke d # 1 fo r single s i n "America n Communit y Survey ranke d #2 1 fo r professional s ranke d #1 3 fo r famil y wit h childre n ranke d # 3 fo r empt y nester s ranke d #4 2 fo r retiree s U S Censu s Bureau 200 5 Denve r ranke d # 1 i n "America' s Mos t Popula r Bi g Cities Pe w Researc h Center 200 9 ranke d # 2 i n "Mos t Popula r America n Cit y t o Liv e In Harri s Interactiv e Poll 200 9 ranke d # 1 i n "Bes t Citie s fo r Singles ranke d # 5 i n "Bes t Citie s fo r Culture ranke d #1 0 i n "Bes t Citie s fo r Nightlife Forbe s Magazine 200 5 Louisvill e ranke d # 1 i n "Bes t Plac e t o Live (# 1 2009 # 3 2007 # 5 2005 ) CNN/MONE Y an d MONE Y Magazin e ranke d # 1 i n "Bes t Plac e t o Rais e a Family Ber t Sperlin g & Pete r Sander Bes t Place s t o Rais e You r Family Frommers 200 6 12 6

PAGE 141

1 V Th e Analytica l Hierarch y Proces s (Satty 1977 ) i s a decisio n approac h designe d t o ai d i n th e solutio n o f comple x multipl e criteri a problems i n a numbe r o f applicatio n domains Th e decision-make r judge s th e importanc e o f eac h criterio n i n pair-wis e comparison s an d th e outcom e o f AH P i s a prioritize d rankin g o r weightin g o f eac h decisio n alternativ e (Atthirawon g & McCarthy 2002) Th e firs t ste p o f AH P i s t o establis h a structura l hierarchy I n thi s step th e proble m i s decompose d int o a hierarch y o f goal criteria sub-criteri a an d alternatives Hierarch y indicate s a relationshi p betwee n element s o f on e leve l wit h thos e o f th e leve l immediatel y belo w (Bhusha n & Rai 2004) Th e secon d ste p i s t o establis h comparativ e judgment A se t o f compariso n matrice s o f al l element s i n a leve l o f th e hierarch y wit h respec t t o a n elemen t o f th e immediatel y highe r leve l ar e constructe d s o a s t o prioritiz e an d conver t individua l comparativ e judgment s int o rati o scal e measurements Th e preference s ar e quantifie d b y usin g a nine-poin t scal e (Atthirawon g & McCarthy 2002) Scal e o f preferenc e betwee n tw o element s (adapte d fro m Golde n e t al. 1989 ) Numerica l value s 1 3 5 7 9 2,4,6,8 Reciprocal s Definitio n Equall y importan t o r preferre d Slightl y mor e importan t o r preferre d Strongl y mor e importan t o r preferre d Ver y strongl y mor e importan t o r preferre d Extremel y mor e importan t o r preferre d Intermediat e value s t o reflec t compromis e Use d t o reflec t dominanc e o f th e secon d alternativ e a s compare d wit h th e firs t 12 7

PAGE 142

I n th e thir d step th e pair-wis e comparison s o f variou s criteri a generate d a t th e secon d ste p ar e organize d int o a squar e matrix Th e diagona l element s o f th e matri x ar e 1 Th e criterio n i n th e rth ro w i s bette r tha n criterio n i n they't h colum n i f th e valu e o f elemen t (/,_/ ) i s mor e tha n 1 ; otherwis e th e criterio n i n they't h colum n i s bette r tha n tha t i n th e rth row Th e (/' 0 elemen t o f th e matri x i s th e reciproca l o f th e (/,/ ) element A = an a 2 i a„\ a i2 a a. a„i. ttu Q, in a mi (1 ) I n th e fina l step th e principa l eigenvalu e an d th e correspondin g normalize d righ t eigenvecto r o f th e compariso n matri x giv e th e relativ e importanc e o f th e variou s criteri a bein g compared Th e element s o f th e normalize d eigenvecto r ar e terme d weight s wit h respec t t o th e criteri a o r sub-criteri a an d ratin g wit h respec t t o th e alternatives AW = On an... a 2 a 22.. a at a ni... a MI am a mi w2 (2 ) Satt y (1977 ) use s th e maxima l eigenvalu e (A™^ ) t o fin d th e genera l W i n Eq (2 ) (A-k m IW=Q (3 ) A se t o f linea r equation s fo r W\, W 2 ,..., W n ca n b e obtaine d fro m Eq (3) th e exac t value s o f W\, W 2 ,..., W n i s compute d b y th e normalize d conditio n a s follows : Wi+ W 2 +...+ W n = 1 (4 ) 12 8

PAGE 143

Consistenc y Tes t Th e \ ma x valu e i s a n importan t validatin g paramete r i n AHP I t i s use d a s a referenc e inde x t o scree n informatio n b y calculatin g th e consistenc y rati o C R (Saaty 2000 ) o f th e estimate d vecto r i n orde r t o validat e whethe r th e pair-wis e compariso n matri x provide s a completel y consisten t evaluation Th e consistenc y rati o i s calculate d vi a th e followin g steps : 1 ) Calculat e th e eigenvecto r o r th e relativ e weight s an d A ma x fo r eac h matri x o f orde r n 2 ) Comput e th e consistenc y inde x fo r eac h matri x o f orde r n b y th e formulae : Cl = (X max -n)/(n-\) (5 ) 3 ) Th e consistenc y rati o i s the n calculate d usin g th e formulae : C R = CI/R I (6 ) wher e R J i s a know n rando m consistenc y inde x obtained fro m a larg e numbe r o f simulatio n runs an d i t varie s dependin g upo n th e orde r o f matrix Th e acceptabl e C R rang e varie s accordin g t o th e siz e o f matri x i. e 0.0 5 fo r a 3 X 3 matrix 0.0 8 fo r a 4 X 4 matri x an d 0. 1 fo r al l large r matrices n> = 5 (Chen g & Li 2001) I f th e valu e o f C R i s equa l to o r les s tha n tha t value i t implie s tha t th e evaluatio n withi n th e matri x i s acceptabl e o r indicate s a goo d leve l o f consistenc y i n th e comparativ e judgment s represente d i n tha t matrix I n contrast i f C R i s mor e tha n th e acceptabl e value inconsistenc y o f judgment s withi n tha t matri x ha s occurre d an d th e evaluatio n proces s shoul d therefor e b e reviewed reconsidered an d improved A n acceptabl e consistenc y propert y help s t o ensur e decision make r reliabilit y i n determinin g th e prioritie s o f a se t o f criteri a (Atthirawon g & McCarthy 2002) 12 9

PAGE 144

v Othe r possibl e categorizatio n method s include : (1 ) Occupation-based : (i ) downtow n offic e an d servic e industr y workers (ii ) students faculty an d staf f o f downtow n universitie s an d educationa l institutions (iii ) employee s o f othe r institutions suc h a s hospitals locate d i n th e city (iv ) municipa l employees especiall y i n citie s tha t requir e everyon e wh o work s fo r th e cit y t o liv e withi n th e cit y limits an d (v ) retirees (2 ) Motivation-based : (i ) curren t an d forme r resident s o f downtow n neighborhoods an d (ii ) "Urba n Pioneers, peopl e who regardles s o f thei r demographi c profile simpl y lik e cit y lif e an d ar e willin g t o b e amon g th e firs t ne w resident s i n a n emergin g are a o f th e city (3 ) Immigratio n status-based : (i ) immigrant s an d (ii ) non-immigrants (4 ) Ethnicity-based : (i ) white (ii ) Africa n American (iii ) America n Indian (iv ) Asian (v ) Latin o American an d (vi ) multiracia l American V 1 Ther e ar e tota l o f fifteen locatio n variable s i n th e mode l an d th e sam e variable s ca n b e use d twic e fo r th e calculatio n o f th e accessibilit y scor e an d fo r th e densit y score Additionall y ther e ar e thre e neighborhoo d variable s wit h varyin g sizes Th e numbe r o f th e possibl e combination s i s a s follow : 4 X GoC + 30 C 2 + 30C 3 + + 30C2 8 + 3oC 2 9 + 30C30 ) = 4,294,967,29 2 13 0

PAGE 145

APPENDI X A SURVE Y INSTRUMEN T Ope n Question s fo r Unstructure d Intervie w You r background experience specialty an d focu s areas ? Wha t ar e th e genera l housin g marke t trend s i n you r area ? Wh o ar e attracte d t o th e downtown ? (percentage ) Wha t attrac t peopl e t o th e downtow n (urba n amenities) ? Wha t ar e othe r factor s i n housin g locatio n decisio n (job climate famil y members) ? Downtow n housin g market : Ar e ther e differen t housin g market s fo r differen t demographi c groups ? Wha t ar e th e difference s betwee n the m (financia l power preferences) ? I s ther e an y cas e o f generationa l conflic t betwee n differen t ag e group s i n th e area ? Wha t ar e th e strength s o f socia l networ k inside/betwee n demographi c group s whe n the y mak e housin g locatio n decisions ? B y ag e group ? Wha t ar e th e turnove r rate s i n you r market ? B y ag e group ? Wha t ar e thei r neighborhoo d preferences ? B y ag e group ? An y area s i f w e hav e no t covere d durin g th e conversations 13 1

PAGE 146

INSTRUCTIONS : Pleas e respon d b y placin g a n 'Y (fo r youn g professional' s preference ) an d = B (fo r babybooraers ) andS (fo r Seniors ) o n th e appropriat e locatio n t o indicat e whic h facto r i s mor e importan t an d ho w importan t i s i t relativ e t o th e othe r facto r fo r eac h o f th e followin g pairwis e comparisons Youn g Professionals : 2 0 4 5 Babyboomers : 4 5 Seniors : ove r 6 5 Conrearisox c Betwee n -6 5 Catezorie s Ex3ee VyOT g Sttoa s Mcdsni e Eqsa S Madsaat o SSoa s Vaayoeq a Ezssoa a Criterio n A 0 0 0 0 Recreationa l Amenitie s Recreationa l Amenitie s Recreationa l Amenitie s Recreationa l Amenitie s "Th e Thir d Places "Th e Thir d Places Th e Thir d Places Cultura l Amenitie s Cultura l Amenitie s Shoppin g 0 0 0 O • s 0 7 igur e A. 1 AH P Surve y Instrumen t Criterio n B "Th e Thir d Places Cultura l Amenitie s Shoppin g Transportatio n Cultura l Amenitie s Shoppin g Transportatio n Shoppin g Transportatio n Transportatio n : 13 2

PAGE 147

Recreationa l Amejatte ; Footbal l Said Ballpark Footbal l fijld. Ballpsris Toercejai k Recreatio n cesse r 3 ) 0 S Recreatio n cente r Neighborhoo d pari Trail BOS S trai l Neighborhoo d park Trail SSetrail "Tk e Thir d Places Coffe e sbop Caf e Coffe e shop Caf e Restauran t • Restauran t Bar Pub Nightclu b Bar Pub Nightclu b Cultura l Ameutte ; Orchesca Opera Musica l Orchestra Opera Musica l Museum Gallei y Museura Galler y Library 3ookstoi e Library Bookstor e Shoornn a Grocei y sore Pharmac y Grocer y sare Pbarrcac y Shoppin g cexe r Shoppin g cente r Smal l retailer s Smal l retailer s Transportatto s Publi c transportatio n Pubh c transportatio n KigSwa y latu p Highwa y ram p Majo r arteria l Majo r arteria l Figur e A 1 (Cont. ) 13 3

PAGE 148

APPENDI X B SIMULATIO N OUTCOME S WIT H NEIGHBORHOO D VARIABL E B. l Boulde r 3X 3 Neighborhoo d . • jv •• • (1 ) Amenit y (Accessibility) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) Transportatio n (Accessibilit y + Density ) Figur e B. l Simulatio n Result s wit h 3X 3 Neighborhood Boulder C O 13 4

PAGE 149

5X 5 Neighborhoo d t V • l&T ".. H • .-•••. • -jpj •.! • (1 ) Amenit y (Accessibility ) (2 ) Amenit y (Accessibilit y + Density ) •_ S • • v • L ^x (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 2 Simulatio n Result s wit h 5X 5 Neighborhood Boulder C O 13 5

PAGE 150

7X 7 Neighborhoo d • • • • :, 1 • i \ ites ; (1 ) Amenit y (Accessibility) (2 ) Amenit y (Accessibilit y + Density ) •••'. : L li" : • .-W r • r in: t (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) •='•••£* * • H ""•.... • • (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 3 Simulatio n Result s wit h 7X 7 Neighborhood Boulder C O 13 6

PAGE 151

B. 2 Denve r 3X 3 Neighborhoo d (1 ) Amenit y (Accessibility ) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility) (4 ) Transportatio n (Accessibilit y + Density ) M M J-' m "-7*,. .n?-• •• C .... . i / *• • T . :•* £ (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 4 Simulatio n Result s wit h 3X 3 Neighborhood Denver C O 13 7

PAGE 152

5X 5 Neighborhoo d (1 ) Amenit y (Accessibility) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 5 Simulatio n Result s wit h 5X 5 Neighborhood Denver C O 13 8

PAGE 153

7X 7 Neighborhoo d •'iTl l •!• " ".••. j-" p • \ :'. : ••'i t ^ *••. .. • r . \ n • i • • „ • a a I (1 ) Amenit y (Accessibility ) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility) • • • M^l l (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 6 Simulatio n Result s wit h 7X 7 Neighborhood Denver C O 13 9

PAGE 154

B. 3 Louisvill e 3X 3 Neighborhoo d (1 ) Amenit y (Accessibility) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 7 Simulatio n Result s wit h 3X 3 Neighborhood Louisville C O 14 0

PAGE 155

5X 5 Neighborhoo d (1 ) Amenit y (Accessibility ) (2 ) Amenit y (Accessibilit y + Density ) (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility ) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 8 Simulatio n Result s wit h 5X 5 Neighborhood Louisville C O 14 1

PAGE 156

7X 7 Neighborhoo d (1 ) Amenit y (Accessibility ) (2 ) Amenit y (Accessibilit y + Density ) r rT (3 ) Transportatio n (Accessibility ) (4 ) Transportatio n (Accessibilit y + Density ) (5 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibility) (6 ) Amenit y (Accessibilit y + Density ) + Transportatio n (Accessibilit y + Density ) Figur e B. 9 Simulatio n Result s wit h 7X 7 Neighborhood Louisville C O 14 2

PAGE 157

APPENDI X C SENSITIVIT Y TES T RESULT S C. l Boulde r Table d II1 II2 II3 II4 II5 II6 II7 II8 II9 11-1 0 11-1 1 11-1 2 11-1 3 11-1 4 11-1 5 11-1 6 11-1 7 11-1 8 11-1 9 11-2 0 11-2 1 11-2 2 11-2 3 H-2 4 11-2 5 11-2 6 11-2 7 Sensitivit y tes t results Boulder C O Weigh t o n Neighborhoo d 0 0 0 0 0 0 0 0 0 2 2 2 2 2 2 2 2 2 Weigh t o n Transportatio n 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 Weigh t o n Amenit y 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 Discrepanc y Inde x 0.197 0 0.143 6 0.143 6 0.137 8 0.145 5 0.146 7 0.137 8 0.142 4 0.145 5 0.154 5 0.156 4 0.153 9 0.157 4 0.157 1 0.153 2 0.155 4 0.154 9 0.153 3 0.154 5 0.156 8 0.156 4 0.154 3 0.156 8 0.155 7 0.157 4 0.155 9 0.157 1 14 3

PAGE 158

0.2 0 0.1 9 0.1 8 0.1 7 0.1 6 0.1 5 0.1 4 1 0 1 0 0 r,0 0 f l 0 0 V o n r S i 0 + | 4 1 9 r o * 1 1 t* i r S 1 i 0 [ J 111 II2 II3 II4 II5 II6 II7 II8 II9 IH O 11-1 1 11-1 2 11-1 3 11-1 4 11-1 5 11-1 6 11-1 7 11-1 8 11-1 9 II-2 0 11-2 1 II-2 2 II-2 3 II-2 4 II-2 5 II-2 6 II-2 7 Figur e C. l Bo x Plotte d Sensitivit y Tes t Results Boulder C O 14 4

PAGE 159

C. 2 Denve r Tabl e C. 2 II1 II2 II3 II4 II5 II6 II7 II8 II9 11-1 0 11-1 1 11-1 2 11-1 3 11-1 4 11-1 5 11-1 6 11-1 7 11-1 8 11-1 9 11-2 0 11-2 1 11-2 2 11-2 3 11-2 4 Sensitivit y tes t results Denver C O Weigh t o n Neighborhoo d 0 0 0 0 0 0 0 0 0 2 2 2 2 2 2 Weigh t o n Transportatio n 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 Weigh t o n Amenit y 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 Discrepanc y Inde x 0.177 7 0.162 5 0.162 5 0.159 6 0.166 8 0.166 3 0.159 6 0.161 8 0.166 8 0.172 2 0.167 1 0.167 0 0.162 9 0.165 8 0.166 2 0.162 8 0.163 7 0.165 3 0.172 2 0.168 7 0.167 1 0.164 7 0.165 4 0.166 3 11-2 5 11-2 6 11-2 7 2 2 2 2 2 2 0 1 2 0.162 9 0.166 2 0.165 8

PAGE 160

0.1 9 • 0.18 0.1 7 0.1 6 • I ? H 1 4 I J i hv M T 1 JJJ^M M T • T | I H 112 113 114 115 116 117 118 119 11-1 0 11-1 1 11-1 2 11-1 3 IH 4 IH 5 II-1 6 11-1 7 IM 8 11-1 9 II-2 0 11-2 1 II-2 2 II-2 3 II-2 4 II-2 5 II-2 6 11-2 7 Figur e C. 2 Bo x Plotte d Sensitivit y Tes t Results Denver C O 14 6

PAGE 161

C. 3 Louisvill e Tabl e C 3 Sensitivit y tes t results Louisville C O Weigh t o n Weigh t o n Weigh t o n Discrepanc y Neighborhoo d Transportatio n Amenit y Inde x IIl 0 0 0 0.215 9 II2 0 0 1 0.165 3 II3 0 0 2 0.165 3 II4 0 1 0 0.154 3 II5 0 1 1 0.169 4 II6 0 1 2 0.175 7 II7 0 2 0 0.154 3 II8 0 2 1 0.165 9 II9 0 2 2 0.169 4 11-1 0 11-1 1 11-1 2 11-1 3 11-1 4 11-1 5 1 11-1 6 1 11-1 7 11-1 8 1 I 0 I 0 [ 0 1 1 [ 1 I 1 1 2 2 L 2 0 1 2 0 1 2 0 1 2 0.209 2 0.195 2 0.192 1 0.199 2 0.187 9 0.188 2 0.194 8 0.189 2 0.188 6 11-1 9 2 0 0 0.209 2 11-2 0 2 0 1 0.197 1 11-2 1 2 0 2 0.195 2 11-2 2 2 1 0 0.203 8 11-2 3 2 1 1 0.197 6 11-2 4 2 1 2 0.190 6 11-2 5 2 2 0 0.199 2 11-2 6 2 2 1 0.197 7 11-2 7 2 2 2 0.187 9 14 7

PAGE 162

Figur e C. 3 Bo x Plotte d Sensitivit y Tes t Results Louisville C O 14 8

PAGE 163

BIBLIOGRAPH Y Abramson A J. Tobin M. & VanderGoot M (1994) The changing geography of metropolitan opportunity: The segregation of the poor in U.S. metropolitan areas: 1970 to 1990. Washingto n D.C. : Aspe n Institute ; Urba n Institute Allen E (2000) Environmental characteristics of smart growth neighborhoods. Washingto n D.C. : Nationa l Resource s Defens e Council Allport G W (1954) The nature of prejudice. Cambridge MA : Addison-Wesley Alonso W (1964) Location and land use: Toward a general theory of land rent. Cambridge MA : Harvar d Universit y Press Anas A (1982) Residential location markets and urban transportation: Economic theory, econometrics, and policy analysis with discrete choice models. Ne w York : Academi c Press Anderson S T. & West S E (2006) Ope n space residentia l propert y values an d spatia l context Regional Science and Urban Economics, 36(6), 773-789 Andrew C I. & Merriam D H (1988) Defensibl e linkage Journal of the American Planning Association, 54(2), 199-209 Arentze T. & Timmermans H (2007) A multi-agen t activity-base d mode l o f facilit y locatio n choic e an d use DisP, 170, 33-44 Atthirawong W. & McCarthy B (2002) A n applicatio n o f th e analytica l hierarch y proces s t o internationa l locatio n decision-making Proceedings of the 7th Cambridge Research Symposium on International Manufacturing. Cambridge MA : Centr e fo r Internationa l Manufacturing Cambridg e University Baerwal d T J (1981) Th e sit e selectio n proces s o f suburba n residentia l builders Urban Geography, 22, 339-357 Bassett K. & Short J R (1980) Housing and residential structure: Alternative approaches. London ; Boston : Routledg e & K Paul Batty M (2009) Urba n modelling I n R Kitchi n & N Thrif t (Eds.) International encyclopedia of human geography. Amsterdam Netherlands : Elsevier Bayoh I. Irwin E.G & Haab T (2006) Determinant s o f residentia l locatio n choice : Ho w importan t ar e loca l publi c good s i n attractin g homeowner s t o centra l cit y locations? Journal of Regional Science, 46, 97-120 14 9

PAGE 164

Becu N. Perez P. Walker A. Barreteau O. & Page C L (2003) Agen t base d simulatio n o f a smal l catchmen t wate r managemen t i n norther n Thailand : Descriptio n o f th e CATCHSCAP E model Ecological Modelling, 1 70(2-3), 319-331 Bell W (1968) Th e city th e suburb an d a theor y o f socia l choice I n S Gree r e t al (Eds.) The New Urbanization (pp 132-168) Ne w York : St Martin' s Press Benard S. & Wilier R (2007) A wealt h an d status-base d mode l o f residentia l segregation The Journal of Mathematical Sociology, 31(2), 149-174 Benenson I (1999) Modelin g populatio n dynamic s i n th e city : Fro m a regiona l t o a multi agen t approach. Discrete Dynamics in Nature and Society, 5(2-3) 149-170 Benenson I. & Torrens P M (2004) Geosimulation : Automata-based modelling of urban phenomena. Hoboken NJ : Joh n Wile y & Sons Benenson I. Omer I., & Hatna E (2002) Entity-base d modelin g o f urba n residentia l dynamics : Th e cas e o f Yaffo Te l Aviv Environment and Planning B, 29(4), 491 512 Bhushan N. & Rai K (2004) Strategic Decision Making: Applying the Analytic Hierarchy Process. London : Springer-Verlag Birch E L (2005) Who lives downtown. Washingto n D.C. : Th e Brooking s Institution Blue V J. & Adler J L (2001) Cellula r automat a microsimulatio n fo r modelin g bi directiona l pedestria n walkways Transportation Research Part B, 35(3), 293-312 Boehm T P (1982) A hierarchica l mode l o f housin g choice Urban Studies, 19, 17-31 Boorstin D J (1973) The Americans: The democratic experience. Ne w York : Rando m House Bowes D. & Ihlanfeldt K (2001) Identifyin g th e impact s o f rai l transi t station s o n residentia l property values Journal of Urban Economics, 50, 1-25 Brooks D (2000) Bobos in paradise: The new upper class and how they got there. Ne w York : Simo n & Schuster Brown D G (2005) Agent-base d models I n H Gis t (eds. ) The earth's changing land: An encyclopedia of land-use and land-cover change (pp 7-13) Westport CT : Greenwoo d Publishin g Group Brown D G. Page S E. Riolo R. & Rand W (2002) Modelin g th e effect s o f greenbelt s a t th e urban-rura l fringe Proceeding s o f th e Internationa l Modelin g an d Softwar e Societ y (pp 190-195) Lugano Switzerland Brown D G. Page S E. Riolo R. & Rand W (2004) Agent-base d an d analytica l modelin g t o evaluat e th e effectivenes s o f greenbelts Environmental Modelling & Software, 79(12) 1097-1109 Bruch E E. & Mare R D (2006) Neighborhoo d choic e an d neighborhoo d change American Journal of Sociology, 112(3), 667-709 15 0

PAGE 165

Brueckner J. Thisse J. & Zenou Y (1996) Why is central Paris rich and downtown Detroit poor? An amenity-based theory. COR E Discussio n Paper s 1996065 Universit e catholiqu e d e Louvai n Cente r fo r Operation s Researc h an d Economic s (CORE) Casselman B (2007 Ma y 11) Th e hom e front : Anima l hous e meet s th e empt y nest ; condo s buil t fo r hipster s dra w folk s ove r 50 ; showdow n a t th e pool The Wall Street Journal, p Wl Castells M (1983) The city and the grassroots: A cross-cultural theory of urban social movements. Berkeley : Universit y o f Californi a Press Cavailhes J. Brossard T. Cavailhes J. Foltete J.C. Hilal M. Joly D e t al (2009) GIS-base d hedoni c pricin g o f landscape Environmental and Resource Economics, 44(4), 571-590 Chapin F S. & Weiss S F (1962) Factors influencing land development. Chape l Hill NC : Universit y o f Nort h Carolina Institut e fo r Researc h i n Socia l Science Chapin F S. & Weiss S F (1968) A probabilisti c mode l fo r residentia l growth Transportation Research, 2(4) 375-390 Chen K (1972) Urban dynamics: Extensions and reflections. Sa n Francisco CA : Sa n Francisc o Press Chen K. Irwin E G. Jayaprakash C & Warren K (2005) Th e emergenc e o f racia l segregatio n i n a n agent-base d mode l o f residentia l location : Th e rol e o f competin g preferences Computational and Mathematical Organization Theory, 77(4) 333-338 Chen Y. & Rosenthal S S (2008) Loca l amenitie s an d life-cycl e migration : D o peopl e mov e fo r job s o r fun? Journal of Urban Economics, 64(3), 519-537 Cheng E W L. Li H (2001) Analyti c hierarch y process : A n approac h t o determin e measure s fo r busines s performance Measuring Business Excellence, 5(3) 30-37 Cheshire P (1995) A ne w phas e o f urba n developmen t i n Wester n Europe ? Th e evidenc e fo r th e 1980s Urban Studies, 32(1), 1045-1063 Cheshire P. & Sheppard S (1995) O n th e price s o f lan d an d th e valu e o f amenities Economica, 62(246), 247-267 Chevan A (1982) Age housin g choice an d neighborhoo d ag e structure The American Journal of Sociology, 87(5), 1133-1149 Choi N G (1996) Olde r person s wh o move : Reason s an d healt h consequences Journal of Applied Gerontology, J 5(3), 325-344 Clark T N (2004) Th e cit y a s a n entertainmen t machine Research in Urban Policy, 9, Oxford UK : Elsevier 15 1

PAGE 166

Clark W A V. & Dieleman F M (1996) Households and housing: Choice and outcomes in the housing market. Ne w Brunswick NJ : Cente r fo r Urba n Polic y Research Clark W A V. & Onaka J L (1983) Lif e cycl e an d housin g adjustmen t a s explanation s o f residentia l mobility Urban Studies, 20(1), 47-57 Clark W A V. & Moore E G (1978) Population mobility and residential change. Evanston IL : Dept o f Geography Northwester n University Clay P L (1979) Neighborhood renewal: Middle-class resettlement and incumbent upgrading in American neighborhoods. Lexington MA : Lexingto n Books Colwell P. Dehring C & Turnbull G (2002) Recreatio n deman d an d residentia l location Journal of Urban Economics, 51, 418-428 Cortright J (2006) The young and restless in a knowledge economy. CEO s fo r Cities Retrieve d Ja n 01 2010 fro m th e Cente r fo r Houston' s Futur e website : http://www.centerforhoustonsfuture.org/cmsFiles/Files/The%20Young%20and%20 R estless%20in%20a%20Knowledge%20Economv.pd f Coupe R T. & Morgan B S (1981) Toward s a fulle r understandin g o f residentia l mobility : A cas e stud y i n Northampton England Environment and Planning A, 13, 201-215 Cova B (1997) Communit y an d consumption : Toward s a definitio n o f th e linkin g valu e o f product s o r services European Journal of Marketing, 31, 297-316 Cowgill (1978) Residentia l segregatio n b y ag e i n America n metropolitan areas Journal of Gerontology Kirkwood, 33(3), 446-453 Cutler D M. & Glaeser E (1997) Ar e ghetto s goo d o r bad? Quarterly Journal of Economics, 112(3), 827-872 Cutler D. Glaeser E. & Vigdor J (1999) Th e ris e an d declin e o f th e America n ghetto Journal of Political Economy, 107(3), 455-506 Day R H (1982) Emergenc e o f chao s fro m neoclassica l growth Geographical Analysis, 13(A), 315-327 D e Jong G F. Wilmoth J M. Angel J L. & Cornwell G T (1995) Motive s an d th e geographi c mobilit y o f ver y ol d Americans The Journals of Gerontology, 505(6) S395-S404 D e Jong G F. & Fawcett J.T (1981) Motivation s fo r migration : A n assessmen t an d a value-expectanc y researc h mode l i n migration decisio n making : multidisciplinar y approache s t o microleve l studie s I n G F D e Jon g & R W Gardne r (Eds.) i n Developed and developing countries. Ne w York : Pergamo n Press d e l a Barra T B (1989) Integrated land use and transport modeling. Cambridge : Cambridg e Universit y Press 15 2

PAGE 167

Dear M. & Flusty S (1998) Postmoder n urbanism Annals of the Association of American Geographers, 88(1), 50-72 Decker J (1993) Simulatio n methodologie s fo r observin g large-scal e urba n structures Landscape and Urban Planning, 26(1-4) 231-250 Dendrinos D S. & Mullally H (1985) Urban evolution: Studies in the mathematical ecology of cities. Oxfor d Oxfordshir e ; Ne w York : Oxfor d Universit y Press Dendrinos D S. & Sonis M (1990) Chaos and socio-spatial dynamics. Ne w York : Springer-Verlag Devisch O T J. Timmermans H J P. Arentze T A. & Borgers A W J (2009) A n agent-base d mode l o f residentia l choic e dynamic s i n nonstationar y housin g markets Environment and Planning A, 41, 1997-2013 Diappi L. & Bolchi P (2008) Smith' s ren t ga p theor y an d loca l rea l estat e dynamics : A multi-agen t model Computers, Environment and Urban Systems, 32(1) 6-18 Dieleman F M (2001 ) Modelin g residentia l mobility : A Revie w o f Recen t Trend s i n Research Journal of Housing and the Built Environment, 16, 249-265 Donnelly T G. Chapin F.S. & Weiss S F (1964) A Probabilistic Model for Residential Growth. Chape l Hill NC : Universit y o f Nort h Carolina Institut e fo r Researc h i n Socia l Science Environmenta l Protectio n Agency (2000) Projecting land use change: A summary of models for assessing the effects of community growth and change on land use pattern. Washingto n D.C. : Author Ezell K (2006) Retire downtown: The lifestyle destination for active retirees and empty nesters. Kansa s Cit y MO : Andrew s McMeel Fainstein S (2001) The city builders: Property development in New York and London, 1980-2000. (2 n d ed.) Lawrence KA : Universit y o f Kansa s Press Featherstone M (1991) Consumer Culture and Postmodernism, London : Sage Fernandez L E. Brown D G. Marans R W. & Nassauer J I (2005) Characterizin g locatio n preference s i n a n exurba n population : implication s fo r agent-base d modeling Environment and Planning B, 32, 799-820 Filatova T. Parker D C & Vee n va n der A (2008) Introducin g preference heterogeneit y int o a monocentri c urba n model : A n agent-base d lan d marke t model Secon d Worl d Congress o n Socia l Simulation Fairfax VA : WCSS Fitzpatrick K M. & Logan J R (1985) Th e agin g o f th e suburbs 1960-1980 American Sociological Review, 50(1) 106-117 Florida R L (2002) The rise of the creative class: And how it's transforming work, leisure, community and everyday life. Ne w York : Basi c Books Florida R L (2005a) Cities and the creative class. Ne w York : Routledge 15 3

PAGE 168

Florida R L (2005b) The flight of the creative class: The new global competition for talent. Ne w York : HarperBusiness Florida R L (2008) Who's your city? : How the creative economy is making where to live the most important decision of your life. Ne w York : Basi c Books Fontaine C Rounsevell M (2009) A n agent-base d approac h t o mode l futur e residentia l pressur e o n a regiona l landscape Landscape Ecology, 24, 1237-1254 Foote N N. Abu-Lughod J. Foley M.M & Winnick L (1960) Housing Choice and Housing Constraints, Ne w York : McGraw-Hil l Boo k Company Forrester J W (1969) Urban dynamics. Cambridge MA : M.I.T Press Friedman J (1981) A Conditiona l logi t mode l o f th e rol e o f loca l publi c service s i n residentia l choice Urban Studies, 18(3), 347-358 Fuguitt G (1993) Th e changin g concentratio n o f th e olde r nonmetropolita n population Journal of Gerontology, 48(6), S278-S288 Fujita M (1989) Urban economic theory: Land use and city size. Cambridg e England ; Ne w York : Cambridg e Universit y Press Gale D E (1979) Middl e clas s resettlemen t i n olde r urba n neighborhoods : Th e evidenc e an d th e implications Journal of the American Planning Association, 45(3), 293-304 Galster G (2000) Identifyin g neighborhoo d thresholds : A n empirica l exploration Housing Policy Debate, 11(3), 701-732 Gans H J (1968) People and plans; essays on urban problems and solutions. Ne w York : Basi c Books Gawande K. Berrens R & Bohara A (2001 ) A consumption-base d theor y o f th e environmenta l Kuznet s curve Ecological Economics, 37, 101-112 Gilbert G N. & Troitzsch K G (1999) Simulation for the social scientist. Buckingham ; Philadelphia PA : Ope n Universit y Press Gimblett H R (2002) Integrating geographic information systems and agent-based modeling techniques for simulating social and ecological processes. Oxford ; Ne w York : Oxfor d Universit y Press Glaeser E L (1998) Ar e citie s dying ? Journal of Economic Perspectives, 12(2), 139-160 Glaeser E L & Gottlieb J D (2006) Urba n resurgenc e an d th e consume r city Urban Studies, 43(S), 1275-1299 Glaeser E L. Kolko J. & Saiz A (2001) Consume r city Journal of Economic Geography, 7(1) 27-50 Golant S M (1990) Post-198 0 regiona l migration pattern s o f th e U.S elderly population Journal of Gerontology, 45(4), S135-S140 15 4

PAGE 169

Golden B L. Wasil E A. & Harker P T (1989) The analytic hierarchy process: Applications and studies. Berlin ; Heidelberg ; Ne w York : Springer Graff T O. & Wiseman R F (1990) Changin g patter n o f retiremen t countie s sinc e 1965 Geographical Review, 80(3), 239-251 Greenberg M. & Lewis M (2000) Brownfield s redevelopment preference s an d publi c involvement : A cas e stud y o f a n ethnicall y mixe d neighborhood Urban Studies, 37, 2501-2514 Greene K (2006 Octobe r 2) Forge t gol f courses beache s & mountains The Wall Street Journal, Rl Gyourko J. & Linneman P (1996) The changing influence of education, income, family structure, and race on homeownership by age over time. Philadelphia PA : Rea l Estat e Cente r Wharto n Schoo l o f th e Universit y o f Pennsylvania Gyourko J. Mayer C & Sinai T (2006) Superstar cities. Cambridge MA : Nationa l Burea u o f Economi c Research Inc Haas W (1993) Amenit y retiremen t migratio n process : A mode l an d preliminar y evidence Gerontologist, 33(2), 212-220 Haas W. & Serow W (1997) Retiremen t migratio n decisio n making : Lif e cours e mobility sequencin g o f events socia l tie s an d alternatives Journal of the Community Development Society, 25(1) 116-130 Hagestad G O. & Uhlenberg P (2005) Th e socia l separatio n o f ol d an d young : A roo t o f ageism Journal of Social Issues, 61(2), 343-360 Haklay M. O'Sullivan D. Thurstain-Goodwin M. & Schelhorn T (2001) "S o g o downtown" : Simulatin g pedestria n movemen t i n tow n centres Environment and Planning B, 25,343-359 Hamnett C (1984) Gentrificatio n an d residentia l locatio n theory : A revie w an d assessment Geography and the Urban Environment: Progress in Research and Applications, 6, 283-319 Hamnett C (1991) Th e blin d me n an d th e elephant : Th e explanatio n o f gentrification Transactions of the Institute of British Geographers, 16(2), 173-189 Hamnett C & Williams P R (1980) Socia l chang e i n London : A stud y o f gentrification Urban Affairs Review, 15(4), 469-487 Hannigan J (1998) Fantasy city: Pleasure and profit in the postmodern metropolis. London ; Ne w York : Routledge Harris B (1965) Ne w tool s fo r planning Journal of the American Planning Association, 31(2), 90-95 Harris C & Ullman E (1945) Th e natur e o f cities The ANNALS of the American Academy of Political and Social Science, 242( 1) 7-17 15 5

PAGE 170

Harvey D (1973) Social justice and the city. Baltimore : John s Hopkin s Universit y Press Harvey D (1989) The condition ofpostmodernity: An enquiry into the origins of cultural change. Oxford UK ; Cambridge MA : B Blackwell Haughey R M (2002) Fittin g in Urban Land, Jan, 50-56 Hepne r G (1983 ) A n analysi s o f residentia l developer locatio n factor s i n a fas t growt h urba n region Urban Geography, 4, 355-363 Hornsten L. & Fredman P (2000) O n th e distanc e t o recreationa l forest s i n Sweden Landscape and Urban Planning, 51, 1-10 Hoyt H (1939) The structure and growth of residential neighbourhoods in American cities. Washingto n D.C. : Federa l Housin g Administration Hunt J D. & Simmonds D C (1993) Theor y an d applicatio n o f a n integrate d land-us e an d transpor t modelin g framework Environment and Planning B, 20, 221-244 Hu u Phe H. & Wakely P (2000) Status qualit y an d th e othe r trade-off : Toward s a ne w theor y o f urba n residentia l location Urban Studies, 37(1) 7-35 Ilkucan A. & Sandikci O (2005) Gentrificatio n an d consumption : A n explorator y study I n G Meno n & A R Ra o (Eds.) Advances in Consumer Research (pp 474-479 ) Duluth MN : Associatio n fo r Consume r Research Ingram G K. Kain J F. & Ginn J R (1972) The Detroit prototype of the NBER urban simulation model. Ne w York : Nationa l Burea u o f Economi c Research ; distribute d b y Columbi a Universit y Press Irwin E G. & Bockstael N E (2002) Interactin g agents spatia l externalitie s an d th e evolutio n o f residentia l lan d us e patterns Journal of Economic Geography, 2, 31-54 Isard W (1960) Methods of regional analysis: An introduction to regional science. Cambridge : Publishe d jointl y b y th e Technolog y Pres s o f th e Massachusett s Institut e o f Technolog y an d Wiley Ne w York Jackson J R (1979) Intraurba n variatio n i n th e pric e o f housing Journal of Urban Economics, 6(4), 464-479 Jackson J. Forest B. & Sengupta R (2008) Agent-base d simulatio n o f urba n residentia l dynamic s an d lan d ren t chang e i n a gentrifyin g are a o f Boston Transactions in GIS, 72(4) 475-491 Jackson P (1985) Neighborhoo d chang e i n Ne w York : Th e lof t conversio n process Tijdschrift Voor Economische En Sociale Geografie, 76(3) 202-215 Jacobs J (1961) The death and life of great American cities. Ne w York : Rando m House Jantz C A. Goetz S J. & Shelley M K (2004) Usin g th e SLEUT H urba n growt h mode l t o simulat e th e impact s o f futur e polic y scenario s o n urba n lan d us e i n th e Baltimore-Washingto n metropolita n area Environment and Planning B, 30, 251 271 15 6

PAGE 171

Jayaprakash C Warren K. Irwin E. & Chen K (2009) Th e interactio n o f segregatio n an d suburbanizatio n i n a n agent-base d mode l o f residentia l location Environment and Planning B, 36(6) 989-1007 Kain J F. & Quigley J M (1970) Measurin g th e valu e o f housin g quality Journal of the American Statistical Association, 55(330) 532-548 Kaiser E J (1968) Locationa l decisio n factor s i n a produce r mode l o f residentia l development Land Economics, 44(3), 351-36 2 Kaiser E J (1966) Towards a model of residential developer location behavior. Unpublishe d doctora l thesis Universit y o f Nort h Carolina Chape l Hill Kan K (2007) Residentia l mobilit y an d socia l capital Journal of Urban Economics, 61(3), 436-457 Kasarda J. Appold S J. Sweeney S H. & Sieff E (1997) Central-cit y an d suburba n migratio n patterns : I s a turnaroun d o n th e horizon ? Housing Policy Debate, 8, 307 358 Keating W (1986) Linkin g downtow n developmen t t o broade r communit y goals : A n analysi s o f linkag e polic y i n thre e cities Journal of the American Planning Association, 52(2), 133-141 Kenney K B (1972) The residential land developer and his land purchase decision. Unpublishe d doctora l thesis Universit y o f Nort h Carolina Chape l Hill Kern C (1984) Uppe r incom e residentia l reviva l i n th e city : Som e lesson s fro m th e 1960 s an d 1970 s fo r th e 1980s Research in Urban Economics, 4, 79-96 Kerridge J. Hine J. & Wigan M (2001) Agent-base d modellin g o f pedestria n movements : Th e question s tha t nee d t o b e aske d an d answered Environment and Planning B, 28(3), 327-34 1 Keyes L C (1969) The rehabilitation planning game: A study in the diversity of neighborhood. Cambridge MA : MI T Press Kii M. & Doi K (2005) Multiagen t land-us e an d transpor t mode l fo r th e polic y evaluatio n o f a compac t city Environment and Planning B, 32, 485-504 Kim M J. & Morrow-Jones H A (2005) Curren t determinant s o f residentia l locatio n decisions I n D M Levinso n & K J Krize k (eds.) Access to Destinations (pp 149 170) Oxford : Elsevier Kim T. Horner M W. & Marans R W (2005) Lif e cycl e an d environmenta l factor s i n selectin g residentia l an d jo b locations Housing Studies, 20(3), 457-473 King A G (1975 ) Th e deman d fo r housing : Integratin g th e role s o f journe y t o work neighborhoo d quality an d prices I n N E Terlecky j & Nationa l Burea u o f Economi c Researc h (Eds.) Household production and consumption. Ne w York : Nationa l Burea u o f Economi c Research : distribute d b y Columbi a Universit y Press 15 7

PAGE 172

Kivell P (1993) Land and the city: Patterns and processes of urban change. London ; Ne w York : Routledge Klosterman R E (2001) Plannin g suppor t systems : a ne w perspectiv e o n computer-aide d planning I n R K Brai l & R E Klosterman (Eds.) Planning support systems: Integrating GIS, models, and visualization tools. Redland CA : ESR I Press Klosterman R E (2007) Deliberatin g abou t th e future I n L D Hopkin s & M A Zapata (Eds.) Engaging the future: Forecasts, scenarios, plans, and projects. Hollis NH Purita n Press Kodrzycki Y K (2001) Migratio n o f recen t colleg e graduates : Evidenc e fro m th e nationa l longitudina l surve y o f youth New England Economic Review, 13-34 Kong F H. Yin H W. Nakagoshi N (2007) Usin g GI S an d landscap e metric s i n th e hedoni c pric e modelin g o f th e amenit y valu e o f urba n gree n space : a cas e stud y i n Jina n City China Landscape and Urban Planning, 79, 240-252 Krizek K. & Waddell P (2002) Analysi s o f lifestyl e choices : Neighborhoo d type trave l patterns an d activit y participation Transportation Research Record, 1807(\), 119 128 Krugman P R (1996) The self-organizing economy. Cambridge MA : Blackwel l Publishers Kwartler M (1998) Regulatin g th e goo d yo u can' t thin k of Urban Design International, 3(1-2), 13-21 Laine T. & Busemeyer J (2004) Comparin g agent-base d learnin g model s o f land-us e decisio n making Proceeding of the Sixth International Conference on Cognitive Modeling, (pp 142-147) Mahwah NJ : Lawrenc e Earlbaum Landis J D. & Zhang M (1998) Th e secon d generatio n o f th e Californi a urba n futur e model par t 1 an d par t 2 Environment and Planning B, 25, 657-666 795-824 Lang M H (1982) Gentrification amid urban decline: Strategies for America's older cities. Cambridge MA : Ballinge r Pub Co Laurie A J. & Jaggi N K (2003) Rol e o f 'vision i n neighbourhoo d racia l segregation : A varian t o f th e Schellin g segregatio n model Urban Studies, 40{\3), 2687-2704 Lee D B (1973) Requie m fo r large-scal e models Journal of the American Institute of Planners, 39(3) 163-178 Lee D B (1994) Retrospectiv e o n large-scal e urba n models American Planning Association. Journal of the American Planning Association, 60{\), 35-40 Ley D (1986) Alternativ e explanation s fo r inner-cit y gentrification : A Canadia n assessment Annals of the Association of American Geographers, 76(4), 521-535 15 8

PAGE 173

Li X. & Liu X P (2007) Definin g agents behavior s t o simulat e comple x residentia l developmen t usin g multicriteri a evaluation Journal of Environmental Management, 85(4), 1063-1075 Lin G (1997) Elderl y migration : househol d vs individua l approach. Papers in Regional Science, 76(3), 285-300 Lin G (1999) Assessin g change s i n interstat e migration pattern s o f th e Unite d State s elderl y population 1965-1990 InternationalJournal of Population Geography, 5(6) 411-424 Lloyd R. & Clark T N (2001) Th e cit y a s entertainmen t machine I n K F Gatha m (Eds.) Research in urban sociology, Vol 6, Critical perspectives on urban redevelopment, (pp 357 378) Oxford : JAI/Elsevier Longino C F. & Fox R (1995) Retirement migration in America. Houston : Vacation Lowry I S (1964) A model of metropolis. Sant a Monica CA : Ran d Corporatio n & Pittsburg h Regiona l Plannin g Association Lucas R J (1988) O n th e mechanic s o f economi c development Journal of Monetary Economics, 22(\), 3-42 Lucy W H. & Phillips D L (2006) Tomorrow's cities, tomorrow's suburbs. Chicago : America n Plannin g Association Luger M I (1996) Quality-of-lif e difference s an d urba n an d regiona l outcomes : A review Housing Policy Debate, 1, 749-771 Lury, C (1996) Consumer culture, Ne w Brunswick NJ : Rutger s Universit y Press Lurz B (1999 July) Interestin g infill Professional Builder, 50-56 Maclennan D (1982) Housing economics: An applied approach. London ; Ne w York : Longman Masnick G S. & Di Z X (2000) Updating and extending the joint center household projections using new census bureau population projections, Cambridge MA : Join t Cente r fo r Housin g Studie s Harvar d University McFadden D (1978) Modelin g th e choic e o f residentia l location I n A Karquis t e t al (Eds.) Spatial Interaction Theory and Planning Models. Amsterdam : North Holland Michelson W (1977) Environmental choice, human behavior, and residential satisfaction, Ne w York : Oxfor d Universit y Press Miller E J. & Salvini P A (2001 ) Th e integrate d lan d use transportation environmen t (ILUTE ) microsimulatio n modelin g system : Descriptio n an d curren t status I n D A Hensche r (Eds.) The leading edge in travel behaviour research, Selected Papers from the 9th International Association for Travel Behaviour Research Conference. Gol d Coast Queensland Australia 15 9

PAGE 174

Mills E S (1969) Th e valu e o f urba n land I n H S Perloff & Resource s fo r th e future (Eds.) The quality of the urban environment; essays on new resources in an urban age. Washington : Resource s fo r th e Future ; distribute d b y th e John s Hopkin s Press Baltimore Mills E S (1976) Plannin g an d marke t processe s i n urba n model I n W S Vickrey & R E Grieson (Eds.) Public and urban economics: Essays in honor of William S. Vickrey. Lexington Mass. : Lexingto n Books Mills E S. & Resource s fo r th e Future (1972) Studies in the structure of the urban economy. Baltimore : Publishe d fo r Resource s fo r th e Futur e b y John s Hopkin s Press Molotch H (1976) Th e cit y a s a growt h machine : Towar d a politica l econom y o f place American Journal of Sociology, 82(2), 309-332 Morrow-Jones H A. & Wenning M V (2005) Th e housin g ladder, th e housin g life cycl e an d th e housin g life-course : upwar d an d downwar d movemen t amon g repea t home-buyer s i n a U.S metropolita n housin g market Urban Studies, 42, 1739-1754 Mulbrandon M C (2007) An agent-based model to examine housing price, household location choice, and commuting times in Knox County, Tennessee. Unpublishe d master s o f scienc e thesis Universit y o f Maryland Colleg e Park Muller B. Yin L. Kim Y. & Alexandrescu F (2008) Th e dynamic s o f lan d developmen t i n resor t communities : a multiagen t simulatio n o f growt h regime s an d housin g choice Environment and Planning A, 40(1), 1728-1743 Muth R F (1969) Cities and housing; the spatial pattern of urban residential land use. Chicago : Universit y o f Chicag o Press Myers D. & Gearin E (2001) Curren t preference s an d futur e deman d fo r dense r residentia l environments Housing Policy Debate, 12(A), 633-659 Myers D (2007) Immigrants and boomer : Forging a new social contract for the future of America. Ne w York : Russel l Sag e Foundation Myers D. & Ryu S (2008) Agin g bab y boomer s an d th e generationa l housin g bubble : Foresigh t an d mitigatio n o f a n epi c transition Journal of the American Planning Association, 74(1), 17-33 Nara A. & Torrens P (2005) Inner-city gentrification simulation using hybrid models of cellular automata and multi-agent systems. Pape r presente d a t th e Geocomputatio n 200 5 conference Universit y o f Michigan Oldenbur g R (1989) The Great Good Place. Ne w York : Parago n House Orego n Departmen t o f Transportation-ODOT (2001) Transportation and Land Use Model Integration Program: Overview of the First Generation Models. Retrieve d Januar y 3 2010 fro m th e Orego n Stat e Governmen t website : http://www.oregon.gov/ODOT/TD/TP/docs/TMR/GE N 1 /0 1 FedRpt.pd f 16 0

PAGE 175

O'Sullivan D. & Macgill J R (2005) Impact s o f neighbourhoo d scal e o n th e dynamic s o f segregation Pape r presente d a t SIR C 200 5 Th e 17t h Annua l Colloquiu m o f th e Spatia l Informatio n Researc h Centre Dunedin Ne w Zealand Pampel F C & Choldin H M (1978) Urba n locatio n an d segregatio n o f th e aged : A block-leve l analysis Social Forces, 56(4), 1121-1139 Park R E & Burgess E W (1925) The city. Chicago : Universit y o f Chicag o Press Parker D. Manson S M. Janssen M A. Hoffmann M J. & Deadman P (2003) Multi-agen t system s fo r th e simulatio n o f land-us e an d land-cove r change : A review Annals of the Association of American Geographers, 93(2), 314-337 Parker D. Berger T. & Manson S M (2001) Agent-base d model s o f land-us e an d land cove r change Repor t an d Revie w o f a n Internationa l Workshop Bloomington IN : LUC C Focu s 1 Publicatio n 6 Parkes A. Kearns A & Atkinson R (2002) Wha t make s peopl e dissatisfie d wit h thei r neighborhoods? Urban Studies, 39, 2413-2438 Pickles A. & Davies R (1985) Th e longitudina l analysi s o f housin g careers Journal of Regional Science, 25( 1) 85-101 Portugali J (2000) Self-organization and the city. Berlin ; Ne w York : Springer Putman S H (1995) The EMPAL and DRAM location and landuse models: An overview. Pape r prepare d fo r FHWA TMI P lan d us e modelin g conference Dallas TX Quigley J M. & Weinberg D H (1977) Intra urba n residentia l mobility : A revie w an d synthesis International Regional Science Review, 2(1) 41-66 Rand W. Zellner M. Page S E. Riolo R L. Brown D G. & Fernandez L E (2002) Th e comple x interactio n o f agent s an d environments : A n exampl e i n urba n sprawl Proceedings of AGENT 2002, Chicago IL Ricardo D (1921) Th e first si x chapter s o f the principles of political economy and taxation, 1817. Ne w York ; London : Macmilla n an d co Romer P (1994) Th e origin s o f endogenou s growth Journal of Economic Perspectives, 8(1). 3-22 Rosse r Jr. J B (1994) Dynamic s o f emergen t urba n hierarchy Chaos, Solitons & Fractals, 4(A), 553-561 Rossi P H (1955 ) Why families move: a study in the social psychology of urban residential mobility. Glencoe IL : Fre e Press Rouwendal J & Meijer E (2001) Preference s fo r housing jobs an d commuting : a mixe d logi t analysis Journal of Regional Science, 41, 475-505 Saaty T L (1977) A scalin g metho d fo r prioritie s i n hierarchica l structures Journal of Mathematical Psychology, 15, 234-281 16 1

PAGE 176

Sack R D (1988) Th e consumer' s world : Plac e a s context Annals of the Association of American Geographers, 78, 642-65 Sampson, R J. Morenoff J D. & Gannon-Rowley T (2002) Assessin g neighborhoo d effects : Socia l processe s an d ne w direction s i n research Annual Review of Sociology, 28(1), 443-478 Sayer R A (1979) Understandin g urba n model s versu s understandin g cities Environment and Planning A, 11, 853-862 Schachter J (2004) Geographical mobility, 2002-2003. Washingto n D.C. : U.S Dept o f Commerc e Economic s an d Statistic s Administratio n Burea u o f th e Census Schelling T C (1971) Dynami c model s o f segregation Journal of Mathematical Sociology, 1, 143-186 Schelling T C (1978) Micromotives and macrobehavior. Ne w York : Norton Scott A J. & Soja E W (1996) The city: Los Angeles and urban theory at the end of the twentieth century. Berkeley CA : Universit y o f Californi a Press Shin S (1990) An analysis of person-environment interaction in elderly migration streams to and within the northeast, 1975-1980. Unpublishe d doctora l thesis Universit y o f Connecticut Sietchiping R (2004) A geographic information systems and cellular automata-based model of informal settlement growth. Unpublishe d doctora l Dissertation Th e Universit y o f Melbourne Simmonds D C (1999) Th e desig n o f th e DELT A land-us e modelin g package Environment and Planning B, 26, 665-84 Simmonds D C (2001 ) Th e objective s an d th e desig n o f a ne w land-us e modelin g package : DELTA I n G P Clarke & M Madden (Eds.) Regional science in business (pp 159-88) Berlin ; Heidelberg : Springer Simpson D M (2001) Virtua l realit y an d urba n simulatio n i n planning : A literatur e revie w an d topica l bibliography Journal of Planning Literature, 15(3), 359-376 Smersh G T. Smith M T an d Schwartz A L Jr. (2003) Factor s Affectin g Residentia l Propert y Developmen t Patterns Journal of Real Estate Research, 25(1) 61-75 Smith N (1979) Towar d a theor y o f gentrification : A bac k t o th e cit y movemen t b y capital no t people Journal of the American Planning Association, 45(4), 538-548 Smith N (1987) Gentrificatio n an d th e ren t gap Annals of the Association of American Geographers, 77(3) 462-465 Sohmer R (1999) Downtow n housin g a s a n urba n redevelopmen t tool : Hyp e o r hope Housing Policy Debate, 10(2), 477-505 Soja E W (2000) Postmetropolis : Critical studies of cities and regions. Oxford ; Maiden MA : Blackwel l Publishers 16 2

PAGE 177

Spain D (1989) Wh y highe r incom e household s mov e t o centra l cities Journal of Urban Affairs, 77(3) 283-299 Steinacker A (2003) Infil l developmen t an d affordabl e housing : Pattern s fro m 199 6 t o 2000 Urban Affairs Review, 38(4), 492-509 Straszheim M R (1975) An econometric analysis of the urban housing market. Ne w York : Nationa l Burea u o f Economi c Research : Distribute d b y Columbi a Universit y Press Suchman D R. & Urba n Lan d Institute (2002) Developing successful infill housing. Washington D.C. : Urba n Lan d Institute Taha H A (2003) Operations research: An introduction (7 t h ed.) Uppe r Saddl e River NJ : Prentic e Hall Teodorovic D (2003) Transpor t modelin g b y multi-agen t systems : A swar m intelligenc e approach Transportation Planning and Technology, 26(4) 289-312 Thrift N J. & Kitchin R. ed (2009) International encyclopedia of human geography. Amsterdam Netherlands : Elsevier Timmermans H. Molin E & va n Noortwijk L (1994 ) Housin g choic e processes : State d versu s reveale d modelin g Approaches Journal of Housing and Built Environment, 9, 215-227 Topalov C (1989) A histor y o f urba n research : Th e French experienc e sinc e 1965 International Journal of Urban and Regional Research, 13(4), 625-651 Torrens P M (2000) How cellular models of urban systems work (WP-28) London : Centr e fo r Advance d Spatia l Analysis Universit y Colleg e London Torrens P M (2001) Can geocomputation save urban simulation? Throw some agents into the mixture, simmer and wait... (WP-32) Centr e fo r Advance d Spatia l Analysis Universit y Colleg e London Torrens P M (2007) A geographi c automat a mode l o f residentia l mobility Environment and Planning B, 34, 200-222 Torrens P M. & Benenson I (2005) Geographi c automat a systems International Journal of Geographical Information Science, 19(4), 385-412 Torrens P M. & O'Sullivan D (2001) Cellula r automat a an d urba n simulation : Wher e d o w e g o fro m here ? Environment and Planning B, 28, 163-168 Tse R Y C & Love P E D (2000) Measurin g residentia l propert y value s i n Hon g Kong Property Management, 18(5), 366-374 Tyrvainen L. & Vaananen H (1998) Th e economi c valu e o f urba n fores t amenities : A n applicatio n o f th e contingen t valuatio n method Landscape and Urban Planning, 43, 105-118 16 3

PAGE 178

Tyrvainen L (2001) Economi c valuatio n o f urba n fores t benefit s i n Finland Journal of Environmental Management, 62, 75-92 U.S Dept o f Housin g an d Urba n Development (2000) The state of the cities, 2000. Washington D.C. : U.S Dept o f Housin g an d Urba n Development Uyar B. & Brown K H (2005) Impac t o f loca l publi c service s an d taxe s o n dwellin g choic e withi n a singl e taxin g jurisdiction : A discret e choic e model Journal of Real Estate Research, 27(4) 427-443 Valerio C (1997) Elderly Americans: Where they choose to retire. Ne w York : Garlan d Pub va n Ham M. & Feijten P (2008) Wh o want s t o leav e th e neighbourhood ? Th e effec t o f bein g differen t fro m th e neighbourhoo d populatio n o n wishe s t o move Environment and Planning A, 40, 1151-1170 Varady D P (1990) Influence s o n th e city-suburba n choic e A stud y o f Cincinnat i homebuyers Journal of the American Planning Association, 56(1), 22-40 Waddell P (1998) A n urba n simulatio n mode l fo r integrate d polic y analysi s an d planning : residentia l locatio n an d housin g marke t component s o f UrbanSim 8th World Conference on Transport Research, Antwerp Belgium Waddell P (2000) A behaviora l simulatio n mode l fo r metropolita n polic y analysi s an d planning : Residentia l locatio n an d housin g marke t component s o f UrbanSim Environment and Planning B: Planning and Design, 27, 247-263 Waddell P (2001) Towar d a behaviora l integratio n o f lan d us e an d transportatio n modeling 9th International Association for Travel Behavior Research Conference. Waldfogel J (2003) Preference externalities: An empirical study of who benefits whom in differentiated product markets. Cambridge MA : Nationa l Burea u o f Economi c Research Inc Wales T (1978) Labo r suppl y an d commutin g time Journal of Econometrics, 8, 2 15-226. Walters W H (2000) Type s an d pattern s o f later-lif e migration Geogrqfiska Annaler: Series B, Human Geography, 82(3), 129-147 Ward R A. LaGory M. & Sherman S R (1985) Neighborhoo d an d networ k ag e concentration : Doe s ag e homogeneit y matte r fo r olde r people ? Social Psychology Quarterly, 48(2), 138-149 Wegener M (2004 ) Overvie w o f land-us e transpor t models I n D A Hensche r & K Button (Eds.) Handbook in Transport 5. Kidlington : Pegamon/Elsevie r Science Weiss S F. Smith J E. Kaiser E J an d Kenney K G (1966 ) Residential developer decisions. Chape l Hill NC : Universit y o f Nort h Carolina Institut e fo r Researc h i n Socia l Science 16 4

PAGE 179

Wenning M V (1995) Rethinking the link: Residential mobility, housing and the life cycle. Unpublishe d doctora l thesis Th e Ohi o Stat e University Columbus OH Wheaton W C (1977) Incom e an d urba n residence : A n analysi s o f consume r deman d fo r location American Economic Review, 67(4), 620-631 Williams I W (1994) A mode l o f Londo n an d th e Sout h East Environment and Planning B, 21. 535-553 Wu B M. Birkin M H & Rees P H (2008) A spatia l microsimulatio n mode l wit h studen t agents Journal of Computers, Environment and Urban Systems, 32, 440 453 Wu J. Adams R M & Plantinga A J (2004) Amenitie s i n a n urba n equilibriu m model : residentia l developmen t i n Portland Oregon Land Economics, 80(1), 19-32 Yin L (2004) Exurban residential development and the attraction of natural amenities. Unpublishe d doctora l dissertation Universit y o f Colorad o Denver Yin L (2009) Th e dynamic s o f residentia l segregatio n i n Buffalo : A n Agent-base d Simulation Urban Studies, 46(\3), 2749-2770 Yin L. & Muller B (2007) Residentia l locatio n an d th e biophysica l environment : exurba n developmen t agent s i n a heterogeneou s landscape Environment and Planning B, 34, 279-295 Zacharias J. Bernhardt T. & d e Montigny L (2005) Computer-simulate d pedestria n behavio r i n shoppin g environment Journal of Urban Planning and Development, 73/(3) 195-200 Zhang J (2004) Residentia l segregatio n i n a n all-integrationis t world Journal of Economic Behavior & Organization, 54(A), 533-550 Zhang W (1989) Coexistenc e an d separatio n o f th e tw o residentia l group s a n interactiona l spatia l dynami c approach Geographical Analysis, 21, 91-102 Zhang W (1993) Locatio n choic e an d lan d us e i n a n isolate d state Annals of Regional Science, 27(1), 23-39. Zhang W (1994) Dynamic s o f interactin g spatia l economies Chaos, Solitions & Fractals, 4(4), 595-604 Zhang Y. & Fang K (2004) I s histor y repeatin g itself? : Fro m urba n renewa l i n th e unite d state s t o inner-cit y redevelopmen t i n china Journal of Planning Education and Research, 23(3), 286-298 Zondag B. & Pieters M (2005) Influenc e o f accessibilit y o n residentia l locatio n choice Transportation Research Record, 1902, 63-70 Zukin S (1998) Urba n lifestyles : Diversit y an d standardisatio n i n space s o f consumption Urban Studies, 35(5-6), 825-839 16 5