Embed Size (px)
Citation preview
University of Nebraska at Omaha University of Nebraska at Omaha
DigitalCommons@UNO DigitalCommons@UNO
Student Work
7-1-2002
The Impact of Apartment Complexes on Property Values of The Impact of Apartment Complexes on Property Values of
Single-Family Dwellings West of lnterstate-680 in Omaha, Single-Family Dwellings West of lnterstate-680 in Omaha,
Nebraska Nebraska
Lesli M. Rawlings University of Nebraska at Omaha
Follow this and additional works at: https://digitalcommons.unomaha.edu/studentwork
Recommended Citation Recommended Citation Rawlings, Lesli M., "The Impact of Apartment Complexes on Property Values of Single-Family Dwellings West of lnterstate-680 in Omaha, Nebraska" (2002). Student Work. 1145. https://digitalcommons.unomaha.edu/studentwork/1145
This Thesis is brought to you for free and open access by DigitalCommons@UNO. It has been accepted for inclusion in Student Work by an authorized administrator of DigitalCommons@UNO. For more information, please contact [email protected].
The Impact of Apartment Complexes on Property Values of
Single-Family Dwellings West of Interstate-680 in Omaha, Nebraska
A Thesis
Presented to the
Department o f Geography and Geology
and the
Faculty of the Graduate College
University of Nebraska
In Partial Fulfillment
of the Requirements for the
Master of Arts Degree in Geography
University of Nebraska at Omaha
by
Lesli M. Rawlings
July 2002
UMI Number: EP73385
All rights reserved
INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMTDlss«ta.fon Rubl shm§
UMI EP73385
Published by ProQuest LLC (2015). Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code
ProQuest*—
ProQuest LLC.789 East Eisenhower Parkway
P.O. Box 1346 Ann Arbor, Ml 48106-1346
THESIS ACCEPTANCE
Acceptance for the faculty of the Graduate College, University of Nebraska, in partial fulfillment of the
Requirements for the degree Master of Arts, University of Nebraska at Omaha.
Committee
' 'N \ \ ^ ’s \ ' \Chairperson
a «• c~\
The Impact of Apartment Complexes on the Property Values of
Single-Family Dwellings West of Interstate 680 in Omaha, Nebraska
ABSTRACT
Lesli M. Rawlings, M.A.
University of Nebraska, 2002
Advisor: Dr. Charles R. Gildersleeve
The literature concerning land use exhaustively describes why homeowners
dislike apartment complexes, but fails to analyze the problem quantitatively. The
objectives of this thesis were to determine if the selling price of single-family dwellings
increased with increasing distance from the apartment complex and to determine if the
selling price o f single-family dwellings decreased with increasing structural density of
apartment complexes in Omaha, Nebraska. Fifty apartments built in the year 2000 or
before, and 1,665 single-family dwellings, which sold in 1999 and 2000 within 914.4
meters of an apartment complex, were geocoded by address. Data needed to test the
hypotheses were sales price, structural characteristics of the dwellings, straight-line
distance, and the number o f apartments influencing each dwelling. Quantitative methods
employed to test the hypotheses included a full and reduced attribute multiple regression
model, factor analysis, and regression analysis using factor scores. The results indicated
that sales price did increase with increasing distance from an apartment complex only
with the utilization of factor analysis and regression using factor scores. However, the
increased number o f apartment complexes proved to decrease the sales price of single
family dwellings when implementing all of the multivariate analyses.
ACKNOWLEDGEMENTS
The purpose of a Master’s thesis is to illustrate a student’s capability to complete
an original research project, which contributes to existing knowledge. For the most part,
completing a Master’s thesis is not effortless or unproblematic, but rather it is a thought-
provoking process, which requires a student to utilize knowledge acquired throughout her
academic career in addition to learning and implementing new methods of analysis. And
although I encountered typical problems associated with research such as sleepless
nights, writer’s block, paper jams, corrupted data files, meticulously examining databases
for errors, and receiving a white hair during the process; I feel that my pursuit to
complete a Master of Arts degree in Geography was a rewarding experience. I have
expanded my knowledge of geography in addition to meeting fellow geographers whom I
consider friends.
I would like to thank several people for their valuable assistance, especially my
thesis committee. I believe that my thesis advisor Dr. Charles Gildersleeve and
committee members Dr. Karen Falconer Al-Hindi, Dr. Michael Peterson, and Dr. Louis
Pol form the ultimate “Thesis Committee Dream Team.” I thank Dr. Gildersleeve for his
unwavering leadership as a thesis advisor who continually provided me with
encouragement. Dr. Gildersleeve contributed with ideas concerning urban geography,
land use studies, and his experience as a former Omaha Planning Board member. Dr.
Karen Falconer Al-Hindi provided invaluable assistance with this thesis, which included
introducing me to the geographers’ perspective of New Urbanism (neo-traditionalism), a
pictorial explanation of a reverse distance decay model, and suggesting an additional
viewpoint relating to the significance of my thesis. Dr. Peterson gave constructive ideas
with reference to measuring distance in Arc View and map development. I am grateful
for Dr. Pol who offered statistical assistance on various occasions and contributed to the
literature review with the article concerning the impact Section 8 housing on the value of
single-family dwellings. I would also like to thank Dr. Rutherford Platt, from the
University of Massachusetts, who granted me permission to use his zoning hierarchy
diagram, which I adapted for this thesis.
Former UNOmaha Geography students also played a crucial role with this thesis.
I would like to thank Shawn Simon, who assisted me with the data acquisition, answering
my detailed questions regarding structural characteristics o f housing, and recalling
information from his extensive mental map of Omaha. Since the beginning of my
graduate studies, Laura Banker has been my GIS mentor. She has provided me with
insight regarding the creation of buffers in Arc View 3.2 and advising the implementation
o f using Avenue Scripts to measure straight-line distance.
I give a special thanks to the Graduate Studies Department for awarding me the
$1,000 Thesis Scholarship for the Spring 2002 semester. This scholarship paid for
various expenses pertinent toward the completion of this thesis.
Last but not least, I would like to thank my parents Sonja and James Rawlings.
They have continuously supported my academic pursuits and without them I would not
have excelled as a graduate student. I hope I will be able to make similar sacrifices if I
have children.
CONTENTS
THESIS ACCEPTANCE................................................................................ ii
ABSTRACT.............................................................................................................................. iii
ACKNOWLEDGMENTS....................................................................................................... iv
CONTENTS......................................................................................................................... vi
LIST OF ILLUSTRATIONS.............................................................................................. viii
LIST OF TABLES...................................................................................................................ix
Chapter
1. INTRODUCTION................................................................................................... 1
Nature of the Problem and Justification..................................................4
Research Objectives..................................................................................... 6
Hypotheses and Rationale.......................................................................... 7
Significance of Research..............................................................................9
Definition of General Terms.................................................................... 10
Study Area.................................................................................................... 12
Chapter Summary.......................................................................................15
2. LITERATURE REVIEW................................................................................... 16
Undesired Residential Properties.............................................................16
The Market Value of New Urbanism...................................................... 21
Price Distance Gradients with the Impact of Externalities................ 23
GIS Applications in Real Estate...............................................................25
Summary of Literature..............................................................................26
3. METHODOLOGY 29
Data................................................................................................................ 29
Data Acquisition........................................................................................... 31
GIS Procedures. ........................................................................................ 32
Database Organization and Data Evaluation..........................................35
Data Analysis.................................................................................................36
Summary of Methodology.......................................................................... 39
4. DISCUSSION OF RESULTS........................................................................... 41
Descriptive Statistics.....................................................................................41
Pearson Correlation......................................................................................43
Multiple Regression......................................................................................45
Factor Analysis.................................................... 48
Regression Utilizing Factor Scores.............................................................54
Summary of Results......................................................................................56
5. CONCLUSION....................................................................................................58
APPENDIX................................................................................................................ 61
A. ESRI Avenue Script.......................................................................................... 61
B. Scatter Plots: Sales Price Versus Housing Characteristics........................65
C. Scatter Plots: Sales Price Versus Hypotheses Variables............................69
D. Pearson Correlation Matrices........................................................................ 75
BIBLIOGRAPHY...................................................................................................80
LIST OF ILLUSTRATIONS
Figure Page
1. Hierarchy of Zoning..............................................................................................................2
2. Eagle Run Apartments Adjacent to Single-family Dwellings..........................................5
3. Conceptual Model of Distance.............................................................................................8
4. Conceptual Model of the Number of Apartments.............................................................. 8
5. Study Area...........................................................................................................................12
6. Typical Zoning Surrounding Apartments and Single-family Dwellings........................ 13
7. Heaviest Concentration of Apartments and Zoning.........................................................14
8. Methodological Design...................................................................................................... 30
9. ArcView 3.2 Interface........................................................................................................ 34
10. Component Plot in Rotated Space................................................................................... 53
LIST OF TABLES
Table Page
1. List of Variables..................................................................................................................31
2. Descriptive Statistics.......................................................................................................... 42
3. Descriptive Statistics for the Apartment Complexes...................................................... 43
4. Pearson Correlation Matrix............................................................................................... 44
5. Regression Results for Model 1........................................................................................ 46
6. Regression Results for Model 2 ........................................................................................47
7. Collinearity Statistics for Models 1 and 2 ........................................................................ 48
8. Communalities for Factor Analysis..................................................................................49
9. Total Variance Explained..................................................................................................50
10. Unrotated Component Matrix......................................................................................... 52
11. Rotated Component Matrix............................................................................................. 52
12. Component Score Coefficient Matrix............................................................................ 54
13. Regression Coefficients of Factor Scores for Model 3 .................................................55
14. Collinearity Statistics for Model 3 ..................................................................................56
1
Chapter 1
INTRODUCTION
Court decisions, zoning laws, politicians, the federal government, and not-for-
profit groups have all ingrained the idea that the American dream is to own a single
family detached dwelling. During the 1930s, the federal government promoted
homeownership when it entered the privatized housing industry by creating the Federal
Housing Administration and the Federal National Mortgage Association (Fannie Mae),
which provide mortgage insurance (Levy 2000). In addition to these federal programs,
favorable tax treatment is given to homeowners. For instance, homeowners may deduct
their mortgage interest and property taxes from their taxable income, while renters cannot
(Levy 2000). Consequently, the federal government has contributed to the significant
increase of home ownership rates from 1945 to 1980 through federal housing programs
and tax policy (Hartshorn 1992 and Levy 2000). However by1980, the construction of
multifamily dwellings was almost equivalent to the construction of single-family
dwellings in the suburbs because o f the elevated housing costs and the densification of
the suburbs (Hartshorn 1992). In 1997, Omaha’s home ownership rate was 59 percent,
which represented more than a 2 percent decrease from 1980 (Omaha Master Plan 1997).
Housing is the largest single type of land use in the United States. Housing has
also been explained to represent a person’s socioeconomic status, power, and identity
(Adams 1984). The social factors of housing mentioned above should not be
underestimated because it has been suggested that apartments adjacent to single-family
neighborhoods challenge even threaten the so-called American dream (National
2
Apartment Association (NAA) 1997). Such occurs because the social fabric o f the
American society considers home ownership higher on the social ladder than renting
(Adams 1987). In order to preserve this hierarchy, zoning ordinances protect single
family dwellings. Thus, social inequalities exist where particular housing types are
placed.
Zoning plays an important role because it dictates what specific types of land use
can be developed in a particular area. Exclusionary zoning has protected detached single
family residences from other land uses such as apartments, mobile home parks,
businesses, and industries. Density factors, supposed aesthetic harm, and the lowering of
property values are the most significant reasons given for zoning to exclude apartments
(Keating 1995). As a result, single-family dwellings are at the apex of the zoning
hierarchy as show in Figure 1 (Adams 1987 and Platt 1996).
Single-familyDwelling
Apartm entComplex
Commercial
HeavyIndustry
Fig. 1. A hierarchy of zoning classifications adapted by permission, from Rutherford H. Platt, Land Use and Society: Geography, Law, and Public Policy (Washington D.C.: Island Press, 1996), 236.
3
Euclidean Zoning in effect prohibits the development of apartment complexes
within single-family residential neighborhoods. However, there are some cases where
developers submit proposals to rezone an area for apartment developments adjacent to
single-family residences. In these instances, homeowners often request that the planning
board vote against the proposal because they believe apartment complexes lower property
values, are aesthetically displeasing, create a greater demand for goods and services, and
increase noise pollution, crime rates, and traffic congestion in their neighborhoods.
A case study, which illustrates the opposition to apartments, comes from the July
and August 2000 Omaha Planning Board meetings. Residents of the Stonybrook
Homeowner’s Association were vehemently opposed to the idea of apartments and
instead favored a proposal for a light commercial development. Apparently, the
commercial developer told residents if they did not accept his residential and commercial
development, a high-density apartment complex would be in its place (Burbach 2000).
At any rate, residents told the Board that apartments would lower neighborhood property
values because of their poor aesthetic quality and the enhancement of traffic congestion.
The irony is that planners have used apartments to buffer single-family residences from
retail developments and major roads (Moudon and Hess, 2000).
Public hearings also suggest that homeowner activism at municipal planning
board meetings is fueled by an alteration of their spatial vision. Purcell’s (2001) study
concerning homeowner activism applies the concept of spatial visualization with a
suburban homeowner’s view of what their landscape should look like. A suburban
homeowner’s spatial vision can be described as a “bourgeois utopia” consisting primarily
4
of owner-occupied, single-family dwellings on large lots, occupied by nuclear families,
and living in “harmony with nature” (Purcell 2001, 181). Land uses introduced in these
areas such as malls, movie theatres, and multifamily dwellings disrupt the suburbanite’s
spatial vision o f a utopia landscape (Purcell 2001), which spurs their activism.
Nature of the Problem and Justification
The nature of the problem involves negative perception and litigation. The
negative perception is clearly defined by John S. Adams’s statement that, “homeowners
accept the idea of people living in apartments, but want them located someplace other
than next door” (Adams 1987, 25). For the most part, the literature focuses on this
perception, but fails to analyze the problem quantitatively. According to Omaha City
Acting Planning Manager Rod Phipps, there has been no quantitative study done for the
City of Omaha to determine if multifamily dwellings have an adverse affect on single
family dwellings (Phipps 2000). This research study examines existing apartment
complexes west of Interstate-680 and their impact on the sales price of single-family
dwellings. Quantifying these results would help justify and confirm the homeowner’s
negative perception, or dismiss it as a problem.
5
Fig. 2. A view of the Eagle Run Apartment Complex and adjacent to single-family dwellings.
In some cases homeowners who are near to higher-density housing or rental
properties file lawsuits against the state if they believe their homes were over assessed for
tax purposes. One such case involves a homeowner who filed an appeal with the
Wyoming State Board of Equalization challenging the appraiser’s market value
assessment of their home in 1994. The owner claimed that the value of their home was
depreciated by the lack of normal maintenance and its propinquity to condominiums
(Assessment Journal 1996). The county board declared that the assessor did not evaluate
all features influencing value and concluded that the homeowner’s property fair market
value was less than what was appraised (Assessment Journal 1996). Another case
involves homeowners in Texas who filed a class action suit in 1986. Homeowners
claimed that single-family rental properties present in their subdivision adversely affected
their property values (Wang et al. 1991).
6
The Omaha City Master Plan strives for contiguous suburban growth for Omaha,
where the development pattern:
“will be based on a series o f high-density mixed use areas that contain, at a minimum, a combination of employment, shopping, personal services,
' open space, and multi-family uses in order to help relieve traffic congestion, allow for a more efficient use of mass transit, and help reverse the current pattern of strip commercial development” (Omaha Master Plan: “Land Use Element” 1997, 5).
Moudon and Hess’s (2000) study suggests that the nucleation of multifamily complexes
should be viewed as positive suburban developments because high densities enable
growth management, promote mixed land use, and social diversity (Moudon and Hess
2000). Again, the density related factor poses a problem for homeowners, who believe
that apartment complexes require more goods and public services, thus increasing their
property tax (NAA 1997). This research study analyzes the negative externalities,
concerning high population densities to examine the effects they have on the selling
price.
Research Objectives
The research study has two main objectives concerning single-family dwellings
and their propinquity to apartment complexes west of Interstate-680 in Omaha, Nebraska.
Only existing apartment complexes were considered in this study and not condominiums
and townhouses. Condominiums and townhouses are omitted because they can be rented
or owned. Furthermore, this study does not take in consideration the potential effect
minor roads and commercial properties have on the sales price of single-family
dwellings. However, if the 914.4-meter (3,000-foot) radius of the apartment complex
7
extended across a major road where single-family dwellings sold, then it was assumed the
particular apartment complex did not influence the dwellings’ sales price. Therefore,
other studies are needed to differentiate the impacts of these land uses on the value of
single-family dwellings.
• The first objective is to determine if proximity to apartment complexes has a
negative impact on the selling price of single-family dwellings.
• The second objective is to determine if the number of apartment complexes
influences the selling price of single-family dwellings.
Hypotheses and Rationale
If the assumptions were correct that apartment complexes increase population
density, which result in a decrease of green space, promote traffic congestion, increase
noise, or attract crime thus creating a negative externality, then the first hypothesis would
be true (Refer to Figure 2). The second hypothesis should also be true because the
increased number of apartment complexes is associated with the negative externality
assumption described above (Refer to Figure 3).
• The selling price of single-family dwellings will increase with increasing
distance from the apartment complex.
• The selling price of single-family dwellings will decrease with increasing
numbers of apartment complexes.
8
Selling Price of Single-Family Dwellings
Distance from Apartments
Fig. 3. A conceptual model depicting reverse distance decay where the selling price of single-family dwellings increase with increasing distance from apartment complexes within a radius of 914.4 meters (3,000 feet).
Selling Price of Single-Family Dwellings
Number of Apartments
914.4 Meters
7ig. 4. A conceptual model where the selling price of single-family dwellings decrease with increasing numbers of apartment complexes within a radius of 914.4 meters (3,000 feet).
9
The rationale for the research is to test the hypotheses based on William Alonso’s
location and land rent theory and Masahisa Fujita’s urban economic theory. Alonso used
distance decay to illustrate bid-rent curves o f land values. He claimed that land values
are highest at the central business districts (CBDs), parks, or lakes, and would decrease
with increasing distance (Alonso 1964). However, reverse distance decay is used for
testing the hypotheses, where the sales price of single-family properties increases with
increasing distance from an apartment complex (Figure 3). This application of reverse
distance decay is justifiable, if it is assumed apartment development follows Fujita’s
urban economic theory concerning neighborhood externalities. Fujita explains that
households want low-density residential areas because they provide desired green space.
Conversely, higher densities decrease wanted green space, increase traffic congestion,
increase noise levels, and crime, which in effect create negative externalities (Fujita
1989). Therefore, in this research study, apartments are assumed to create negative
externalities, which are considered undesirable traits in a neighborhood, thus resulting in
lower property values.
Significance of the Research
The placement of particular housing structures is an emotional issue. Often,
purchasing a home is the most important investment of a lifetime (Adams 1984 and
Bourne 1981). Therefore this research contributes to the existing literature regarding
land use, land rents, and property values. If the hypotheses are correct, city planners
should regulate the development of apartment complexes in single-family neighborhoods
in terms of size, design, and the like. If the hypotheses are incorrect, city planners can
10
strengthen their case to distribute affordable housing throughout the city. Consequently,
there is a need for geographers to develop a theoretical model that can assist planners to
determine how apartment complexes influence land values, property values, and the tax
base.
General Definition of Terms
Address geocoding: A GIS procedure that finds a location on a map using an address
(ESRI1996).
Address matching: A GIS procedure that finds an address in a table, which corresponds
to an address attribute’s table of a theme (ESRI 1996).
Apartment complex: In this research study refers to as twenty or more dwelling units
under a single structure.
Collinearity: Where to two independent variables (multicollinearity more than two
variables) are highly correlated with each other, meaning their correlation
coefficients are close to 1.
Exclusionary zoning: Prevents the construction of mobile homes, or other types of
homes (multifamily housing) for middle income persons or racial minorities (Platt
1996).
Geographic Information Systems (GIS): Is the storage, retrieval, manipulation, and
display of spatial data.
Hedonic model: Similar to a multiple regression model, except that it applies to
housing (Eppli 2000). This model includes the structural and locational attributes
of housing to predict property values.
11
Multifamily dwellings: Five or more units under a single-structure (Van Vliet 1998).
Scripts: Programs written with Avenue, a programming language used by Arcview
(ERSI 1996).
Shapefile: A data format used in ArcView.
Standardized (Regression) Beta Coefficients: The transformation of raw data, where the
mean is zero, standard deviation is 1, and the constant, bo is 0. The purpose of the
standardized coefficient is to make variables with differing measuring units more
comparable to determine how each independent variable is influencing the
dependent variable (Hair 1987 and SPSS 2001).
Theme: A term referring to a spatial data layer containing a geographical feature and its
characteristics (ERSI 1996). The following themes generated for this research
study include: apartment complex and house location, streets, a city limit and a
study area boundary, and drainage.
Unstandardized Beta Coefficients: Are coefficients of the estimated regression model
(SPSS 2001).
12
Ida Street
\V. Dodge
Center |
Apartment Complexes and Single Family Dwellings West of Interstate-680
12 M iles
• Apartment Com plex• Single-fam ily D n d lin s
S treetA / Study Area Biuimtaiy
City Limits ■ Drainage
Fig. 5. The location of apartment complexes and single-family dwellings in the study area.
Study Area
The following streets and Interstate delimit the study area north, south, east, and
west are Ida Street, Harrison, Interstate-680, and 180th Street respectively (Refer to
Figure 5). This area was chosen because most of the development occurred in the mid
1950s, after Euclidean zoning had been long implemented, which supposedly protects the
residents and the valuation of single-family dwellings. Typical zoning in the study area
consists of low-density, multi-family dwellings designated as R6 (apartment complexes),
which are adjacent to high-density single-family dwellings, designated as R4. As seen in
13
Figure 6, two apartment complexes are buffering the negative externality effect of West
Maple Road and community commercial districts.
Blondo
Typical Zoning Surrounding Apartment Complexes
Apartment Complex
Single-family Dwellings
| CC Community Commercial DistrictDR Development R eserve District
■ | Q C GeneralCommercial District
■ B G O General Office DistrictLI Limited Industrial DistrictR3 SFD Residential District (Medium Density)
R4 SFD Residential District (High Density)
R5 Urban Family Residential DistrictR6 Low Density Multi-Family Residential District
1 Miles (Low Density)
Fig. 6. Typical zoning surrounding apartments and single-family dwellings in the study area.
The land use element of the Omaha City Master Plan (1997) explains that the
ideal location for apartments would be adjacent to offices, commercial centers, and
portions of civic centers designated for mixed-use (Refer to Figures 6 and 7). It
designates West Maple, West Dodge, and West Center Roads for unlimited multifamily
(hereafter referred to as apartment) development within a quarter-mile north or south of
these roads (Omaha City Master Plan 1997). If apartments are constructed outside of
these corridors, then they are to be limited to 100 dwelling units, except if they are farther
than a quarter-mile away from another apartment complex (Omaha City Master Plan
14
1997). Fifty apartment complexes were chosen in this region, some of which are adjacent
to single-family dwellings. Figure 5 shows the general spatial concentration of apartment
complexes are found along the western edge of Interstate-680 and the major roads.
However, Figure 7 depicts that the heaviest concentration of apartments occurs just south
of the Interstate-680 West Maple Road off ramp. Most of the apartments are within close
proximity to major thoroughfares and commercial developments because planners
consider them buffers, thus protecting single-family dwellings. Moreover, apartment
complexes situated next to these major thoroughfares provide “adequate transportation”
for increased residential density (Moudon and Hess 2000, 253).
and Zoning in the Study Area
Blondg
Heaviest Concentration of Apartment Complexes
Apartment Complex
Single-family Dwellings
CC Community Commercial DistrictDR Development Reserve District
| GC General Commercial District
^ G O General Office District
^ ^ L l Limited Industrial DistrictR3 SFD Residential District (Medium Density)
R4 SFD Residential District (High Density)R5 Urban Family Residential DistrictR6Low Density Multi-Family Residential District
1 Miles ■ ■ R7 Multi-Family Residential ■ Distict (Medium Density)
Figure 7. Depicts the heaviest concentration of apartments occurs just south of West Maple Road and Blondo.
15
Chapter Summary
The purpose of this thesis is to examine the impact of apartment complexes on the
selling price of single-family dwellings by using distance and structural density as
variables to test the hypotheses. Chapter 2 reviews the literature on housing value
concerning undesired residential properties, New Urbanist developments, price-distance
gradients, and the utilization of a GIS in Real Estate. Chapter 3 describes the
methodological design, data collection, and analyses performed in the research study.
Chapters 4 and 5 provide an extensive discussion of the results and conclusions of the
research study respectively.
16
Chapter 2
LITERATURE REVIEW
Minimal research has been done to examine the impact apartment complexes have
on the value of single-family dwellings. This literature review offers a detailed
description of the objectives, data characteristics, methodology, and conclusions found in
previous studies indirectly related to the research. The summaiy of this Chapter provides
justification why some of the procedures presented in this literature review support the
methodology used in this research study.
Undesired Residential Properties
Homeowners view single-family rental properties, mobile home parks, and
federally assisted housing as having a negative impact on the value of their homes. The
literature for the most part supports this view. Wang et al. (1991) examined the impact of
single-family rental properties on the market value of owner occupied single-family
dwellings in San Antonio, Texas. Wang et al. (1991) tested three hypotheses concerning
this issue. The first hypothesis suggested that higher percentages of rental properties
present in a neighborhood would lower the sales price of single-family dwellings. The
second hypothesis suggested sales price would decrease with increased numbers of
surrounding rental properties. The third hypothesis suggested that poorly maintained
properties decreased sales price.
The four main attributes that Wang et al. (1991) used which concern this research
study were: housing quantity, housing quality, locational, and rental property variables.
Housing quantity variables were subdivided into square footage of living area, lot size,
17
number of garages, bedrooms, bathrooms, and other room types. Housing quality
variables were categorized by the age of the structure, air conditioning, and number of
fireplaces. The locational variables were dummy variables based on area designation.
The rental property attribute was subdivided into the number of rental properties present
in 5-house and 8-house groupings, the sum of the rental properties in the 5 and 8-house
groupings, and a variable concerning the percentage o f rental properties for each
subdivision. The sources for housing characteristic data came from the multiple listing
services (MLS), Southwest Texas Data Bank, and Bexar County Appraisal District, while
the rental variables were based on Wang et al.’s (1991) calculations.
The methods employed by Wang et al. (1991) were Pearson correlations and the
following four hedonic models: a full attribute model, a second model eliminating the
number of bathrooms, bedrooms, and other room types; a third model including the rental
variables, that consisted of the number of rental properties within five and eight house
groupings; and the fourth model, which only used the rental property variable that was
the sum of the total rental properties within the five and eight house groupings.
Wang et al. (1991) regression results supported all three hypotheses, such that all
rental property variables had the expected negative signs and were significant at the 90
percent or greater level. They found that homebuyers would pay 2 percent more if
surrounding rental properties were maintained. They found that single-family rental
properties have the same adverse effects as apartments or other “undesired properties”
(W angetal. 1991, 164).
18
Wang et al. (1991) differ from Beny and Bednarz (1975), which found that the
percentage of apartments within census tracts have a positive effect on housing price.
They believed this to be a rational outcome because the presence of many apartment
buildings symbolizes a more intensive land use, thus increasing the land value.
Cadwallader (1996) compares Wang et al. (1991) and Berry and Bednarz (1975) and
suggests that multicollinearity in the independent variables probably caused the different
outcomes between the two studies.
Berry and Bednarz (1975) developed a multivariate housing model for Chicago.
The dependant variable was the selling price of 275 single-family dwellings in the study
area. The methods employed included simple correlation, a full attribute multiple
regression model, and stepwise regression to create a reduced variable model. The four
main attributes that Berry and Bednarz studied were housing characteristics, housing
improvements, neighborhood characteristics, and accessibility. Housing characteristics
were divided into floor area, age, and lot size. Housing improvements were also
subdivided into air conditioning, garage, improved attic, improved basement, and number
o f bathrooms. Dummy variables represented all of the housing improvements, except for
the number of bathrooms. Neighborhood characteristics contained the following
variables: median family income, percentage of multiple family dwellings, and migration,
which represented the percentage of family’s living in a different census tract five years
before.
The multiple regression results show that all of the variables had the expected
signs of the regression coefficients and were significant. Such that housing
19
characteristics floor space and lot size had a positive effect on sales price, while the effect
o f age was negative (Berry and Bednarz 1975). Neighborhood characteristics showed
that sales price increased with higher median family income as well as with higher
percentages of apartments.
Berry and Bednarz (1975) had problems with multicollinearity and attempted to
resolve the issue by employing a preliminary analysis that tested a larger set of variables
to determine which variables had collinearity. The variables having problems with
collinearity were removed from the data, leaving the remaining dataset used in the study.
However, Berry and Bednarz explained that not all of the collinearities could be
removed. For instance, large homes will usually occupy a large lot.
Munneke and Slawson (1999) examined the effect of proximity of mobile home
parks on single-family property values in Baton Rouge, Louisiana. Fifty-eight mobile
home parks were located in the study area. During the four-year study period, 3,025
single-family dwellings sales transactions were collected from a MLS.
Data used by Munneke and Slawson (1999) were price per square footage of
living area, lot area, square footage of living area, age, presence of central air, and several
straight-line distance measurements to central business districts, major intersections,
nearest shopping center, distance to airport, presence of mobile home parks in a half-
mile, and presence of mobile home parks greater than a half-mile away. The methods
employed were descriptive statistics, a full probit model on the entire sample and a
reduced-form probit equation. The results of the reduced-form probit model show that
20
mobile home parks had a negative effect on single-family dwellings less than one-half
mile away.
There is the belief that federally assisted housing has a negative effect on the
selling price of homes. Homeowners automatically assume that Section 8 housing will
lower their property values (Babb et al. 1980). Babb et al. (1980) gathered data from the
following sources: the 1970 U.S. Census of Population and Housing, 1980 Census of
Population and Housing, a computerized sales list, the location of public and federally
assisted housing from the Memphis Housing Authority and the Memphis Department of
Housing and Community Development.
Eleven control neighborhoods without public and federally assisted housing were
established to match the socioeconomic characteristics of those that did. Neighborhood
boundaries were established after examining aerial photographs and land use maps.
Three dependant variables used in the study were selling price, the number of sales, and a
ratio concerning neighborhood sale price to the home sale price for the city. The ratio
had two purposes: one it adjusted for inflation, two it showed whether the valuation
increased, lowered, and or remained stable during the 12-year time period. A t-test for
dependent samples was used.
The results show that the average sales price in both groups increased during the
time period. However, the rate of increase between both the control and federally
assisted housing sites were not noticeable. The t-tests showed that the intra
neighborhood differences in mean sales price and ratio were also not significant. Babb et
al. (1980) conclude there is no evidence that the introduction of federal assisted housing
21
lowered the market value of owner-occupied homes in Memphis between 1970 through
1980.
The Market Value of New Urbanism
New Urbanism is based on the traditional neighborhood design where high urban
density exists as a result of smaller lot sizes, thus supporting a mass transit orient design.
Other principles of design include the implementation of mixed land uses, such as
apartments above stores, different housing types for varied income groups, within close
proximity of each other; streets are based on the grid system, and the design promotes a
pedestrian friendly atmosphere; retail, restaurants, parks, and civic centers are within a
five minute walking distance.
Eppli and Tu (1999) conducted a study to determine if homebuyers would pay
more for homes in New Urban communities than for adjacent suburban housing
developments. The New Urban developments chosen for the study followed most of the
principles of design as described in addition to having surrounding suburban
developments for a control group. If the New Urban developments had the following
characteristics: a “significant” number of multifamily dwellings, the presence of
subsidized housing, were resort communities, and if either the New Urban or suburban
housing communities had less than 150 sale transactions during the 1994 to 1997 study
period, they were eliminated from the analysis (Eppli and Tu 1999, 22). It is important
to note that the Eppli and Tu failed to explain why New Urban developments that
contained a significant number of multifamily dwellings were excluded from the study;
however, all other elimination criteria were discussed thoroughly. Six communities were
22
chosen for the study: Laguna West, California; Kentlands, Maryland; Southern Village,
North Carolina; Northwest Landing, Washington; and Celebration, Florida. The total
transaction used in the study was 5,833 and 5,169 came from the suburban housing sales.
Eppli and Tu (1999) chose 32 attributes of housing characteristics collected from
First American Real Estate Solutions (FARES, previously known as Experian) and tax
assessors’ office to explain the anticipated price differential. Each of the developments
were analyzed separately and combined using several multiple regression models, with
one New Urban variable added to each of the equations (parameter estimate).
The regression models developed for all of the communities explained 85 to 92
percent of the variance in sales price of the single-family dwellings. The authors claim
that t-statistics was significant at the 99 percent level for the New Urban variable. They
conclude that homebuyers are willing to pay more for homes in New Urban
developments than suburban housing (Eppli and Tu 1999).
This study by Eppli and Tu (1999) that concerns New Urbanism was selected
because the design elements of this paradigm allows for single-family dwellings to be
within close proximity to a variety of land uses. Their results indicated homebuyers are
willing to pay more for houses in these developments than typical suburban
developments. If the hypotheses for this thesis are supported, then Omaha planners
should consider adopting New Urbanism design principles in the Master Plan.
Furthermore, future New Urbanism studies should analyze the communities with a
significant number of apartment complexes to determine their impact on sales price.
23
Price-Distance Gradients with the Impact of Externalities
Externalities are effects produced by a particular land use on adjacent areas (Platt
1994). Externalities are either positive or negative in relation to conforming land uses.
Platt (1994) notes that physical externalities include water and air pollution, noise, view,
impact on wildlife, and dumping of wastes. He characterizes socioeconomic externalities
as the loss of jobs, retail sales, taxes, increased use of goods and services, and traffic
congestion. The studies reviewed in this section describe the impact of externalities
using price-distance gradients on the value of homes.
Li and Brown (1980) examined micro-neighborhood externalities using a linear
model to estimate hedonic housing prices in the southeast Boston metropolitan area. The
sources of the 1970 data came from a MLS and census tract information. They sampled
781 sales of single-family homes in fifteen suburban communities. The variables were
structural and site attributes, which used MLS data for the houses’ characteristics such as
number of rooms, bathrooms, fireplaces, garage spaces, presence of basement, patio, and
age. Neighborhood (census tract) attributes included median income for the study area
and residential density. Local public services and cost attributes contained variables such
as aesthetics, noise and air pollution, and proximity to determine housing prices. The
dependant variable in the study was the selling price. Aesthetics was measured by people
ranking a view on an index of one to five, one for the lowest and five as the highest.
The multivariate analysis employed was nonlinear least squares regression to
lessen the sum of the squared residuals. Simple hedonic linear models were used to
determine house price and size. Their findings suggest that accessibility had a positive
24
effect on sales price value, while external diseconomies such as higher density, which
leads to traffic congestion, pollution, and poor aesthetics, were negative. The negative
effects on sales price diminished more quickly with distance than the positive effects. Li
and Brown’s (1980) findings support the rationale o f the first hypothesis in this research
study where sales price increases with increasing distance from an apartment complex.
Darling (1973) examined the impact of three California urban water parks on the
surrounding property values of vacant lots, single-family dwellings, high-rise apartments,
duplexes, and quadplexes. Darling’s main hypothesis was that property values would
decrease with increasing distance from the urban water parks. He also examined how
much property values would decrease with respect to distance and the rate this would
occur.
The data used for this study came from the 1960 California Census, municipal
records, county records, and MLS databases. Interviews were also used to construct a
demand curve for explaining the utilization of the park. In this review, only the
quantitative methods and results are discussed because these pertain to this research
study. The independent variables used in the study were square footage of living area,
square footage of lot area, number of rooms, and baths. Distance measurements were
calculated within a 3,000-foot “arbitrary” boundary from the urban parks’ shoreline to the
properties (Darling 1973, 25). Darling made the assumption that beyond the 3,000-foot
boundary, the urban water park would not influence the properties’ value. Both the
assessed value and sales price were used as dependent variables.
25
Full attribute regression models such as linear, double log transformation, and
quadratic models were employed to test the hypothesis. However, only the linear model
results were presented because of the space limitations in the journal (Darling 1973).
Resultant signs of the regression coefficients for distance were positive thus, proving
their hypothesis to be valid where property values would increase with increasing
distance from an urban water park.
GIS Applications in Real Estate
A GIS can be used in numerous applications pertaining to real estate and
marketing. Bible and Hsieh (1996) used a GIS to generate regional variables that would
affect the variation o f apartment rents. The Metropolitan Statistical Area of Shreveport-
Bossier City, Louisiana was the location used to study non-subsidized apartment
complexes.
The data characteristics for the apartment included age, number of bedrooms,
square footage for each unit, whether a pool, fireplace, and tennis courts were present,
and the number of total units. A database containing location and data characteristics
were imported from a spreadsheet into a GIS. The address matching procedure in a GIS
was used to locate the apartments on the TIGER Map files created by the United States
Census Bureau. The census information regarding housing and population on CD ROM
disks were downloaded onto a spreadsheet and imported in a GIS, and merged with
census tracts. On a separate theme, neighborhood characteristics were delineated with
one-mile radius circles (buffers) for each apartment building. After these regions were
identified, regional variables were generated. Bible and Hsieh (1994) explained that
26
numerous regional variables could be generated, however it would not be statistically
correct to use all of them. Instead eleven variables were chosen and tested using the
ordinary least squares regression analysis. The regression procedure employed by SAS
showed that multicollinearity existed. To rectify this problem, the variables were
grouped into five attributes of regional variables.
Six models were prepared to test for multicollinearity. Results of a correlation
analysis indicated that age was negatively significant indicating that the older apartments
had a lower rent per square foot. It also showed that apartments with a swimming pool
and or a fireplace had a positive and significant correlation with apartment rents.
Summary of Literature
In summary, only a paucity of the literature concerning land use and property
valuation directly examines the effect of apartment complexes on the selling price of
single-family dwellings. The following summarizes the methods and results utilized in
the reviewed literature and how it supports the methodological framework of this thesis
through data selection, GIS procedures, data analysis, and the expected results.
In many of the studies presented the dependent variable was sales price while the
independent variables were housing characteristics, various distance measurements, and
some variable representing rental properties (Berry and Bednarz 1975; Eppli and Tu
1999; Li and Brown 1980; Munneke and Slawson 1999; and Wang et al. 1991). This
research study selected similar housing structural characteristics and straight-line distance
measurements because much of the variation influencing sale price was accounted for in
27
the literature. The selected data for these reviewed studies came from the following
sources: county records, MLS, census data, Tiger map files, and data banks.
For the most part, many of the studies utilized the straight-line distance procedure
to calculate distance measurements (Bible and Hsieh 1996; Darling 1973; Li and Brown
1980; and Munneke and Slawson 1999). However, only studies conducted in the 1990s
employed a GIS, because GIS was not widely available earlier. All of the price distance
gradient studies in this review captured the variance within a half-mile, 3000-feet, and or
a mile. Their results justify that impact of both positive and negative externalities occur
at a relatively short distance assumed in the rationale o f this thesis.
The data analyses utilized in the literature were descriptive statistics (Li and
Brown 1980), Pearson correlation (Wang et al. 1991 and Berry and Bednarz 1975)
multiple regression or hedonic pricing models (Berry and Bednarz 1975; Bible and Hsieh
1994; Eppli and Tu 1999; Li and Brown, 1975; and Wang et al. 1991). These statistical
analyses were also used for this research.
The impact of undesired residential properties on the sales price of single-family
dwellings examined in the literature review are single-family rental properties,
apartments, mobile home parks, and Section 8 housing. Wang et al. (1991) studied the
impact of single-family rental properties on the value of owner occupied single-family
dwellings. They concluded that all o f the rental property variables were significant and
had the expected negative regression coefficient. Their study differed from the results of
Berry and Bednarz (1975) who developed a multivariate land use model for Chicago.
Their results indicated that higher percentages of multifamily dwellings within a census
28
tract had a positive effect on sales price. Munneke and Slawson (1999) examined the
impact of mobile home parks on the sales price with relation to proximity using probit
models. They explain that homes less one-half mile away from mobile home parks have
a lower sales price than those farther away. Babb et al. (1984) determined if the
introduction of Section 8 housing had a negative impact on sales price of existing owner-
occupied housing. Their results indicate there was no evidence that the introduction of
Section 8 had a negative effect on sales price. The differing methodologies and results
concerning the impact of these particular undesired land uses on sales price indicate that
more rigorous quantitative analysis needs to be done.
29
Chapter 3
METHODOLOGY
The objective of this research study was to determine if the sales price of single
family dwellings were influenced by proximity and structural density of apartment
complexes. The hypotheses state that the sales price of a single-family dwelling would
increase with increasing distance from apartment complex and decrease with increasing
numbers of apartment complexes. As seen in Figure 8, the stages of the methodological
design necessary to test the hypotheses were Data selection, Data Acquisition, Database
Organization and Evaluation, and Data Analysis. The subsequent sections provide the
justification and explanation of the data and methods used for this research study.
Data
Since the research was completed at a micro-scale level, detailed data were
selected to determine if distance and structural density of apartment complexes
influenced the sales price of single-family dwellings. The selling price o f the single
family dwellings, which sold in 1999 or 2000, were chosen as the dependent variable
because it reflected market conditions at the time better than assessed housing valuations.
The independent variables used for this study were housing characteristics, such as
square feet, number of total rooms, bathrooms, bedrooms, age, and garage. These
particular variables were selected because several studies have shown them to affect sales
price (Berry and Bednarz 1975, Eppli and Tu 1999, Li and Brown 1980, Munneke and
Slawson 1999, Richardson et al. 1974, and Wang et al. 1991). The independent variables
30
distance and number of apartments were essential in ultimately testing the hypotheses.
A list of the variables utilized in the study as well as their sources is found in Table 1.
Methodological Design
Quantitative and Statistical Presentation
Database Organization and Data Evaluation
Data Analysis
Data(Necessary to test hypotheses)
Conclusions(Supported Hypotheses)
GIS Procedures(Data Generation)
Data Acquisition(Douglas County: Access to MLS)
Fig. 8. Conceptual field methods adapted from Introduction to Scientific Geographic Research. Haring et al. Boston: WCB McGraw,1992.
31
Table 1. List of Variables (Dependent Variable is Sales Price)
Variable Description Sources
SP Sales Price (1999-2000) DCMLSSq_Ft Total Square feet o f living area DCMLST otR m Total rooms excluding bathrooms DCMLSBath Number of bathrooms DCMLSBed Number of bedrooms DCMLSAge Age of the dwelling in years (2001-year built) DCMLSGarage Number of garages DCMLSNum_Apts Number of apartment complexes located within
3,000 feet (914.4 meters) o f a single-family dwellingGIS
Dist_one Distance in meters from the first closest apartment complex to the SFDs GISDist_two Distance in meters from the second closest apartment complex to the SFDs GISDist_three Distance in meters from the third closest apartment complex to the SFDs GISDist_four Distance in meters from the fourth closest apartment complex to the SFDs GISDist_five Distance in meters from the fifth closest apartment complex to the SFDs GISDist_six Distance in meters from the sixth closest apartment complex to the SFDs GISDist_seven Distance in meters from the seventh closest apartment complex to the SFDs GISDist_eight Distance in meters from the eighth closest apartment complex to the SFDs GISBed_Rnt The lowest one bedroom rent for an apartment Prty ManApt_Age Age of the apartment complex in years (2001-year built) DCATot_Sqft Total square footage of the apartment DCATotJUnits Total number of dwelling units in the apartment DCA
Note: SFDs stands for single-family dwellings and Prty Man represents property managers.
Data Acquisition
The variables concerning sales price of single-family dwellings in 1999 or 2000 in
addition to the structural attributes came exclusively through a MLS used by the Douglas
County Assessors Office (DCA). The DCA also provided the structural attributes of the
apartment complexes located in the study area. The street, parcel, city limit boundary,
zoning, and drainage shapefiles needed for locating apartment complexes and single
family dwellings, data generation, and cartographic representations in a GIS came from
the Omaha City Planning Department. It was decided to use these shapefiles because
they were better maintained for address matching purposes as opposed to Tiger map files
provided by the U.S. Census.
32
Several studies utilize the straight-line distance procedure to calculate distance
measurements (Bible and Hsieh 1996; Darling 1973; Li and Brown 1980; and Munneke
and Slawson 1999). However, utilizing a GIS to measure a straight-line distance
provides the advantage of obtaining measurements between two absolute locations within
a radius designated by the user (Bible and Hsieh 1996). As a result, the straight-line
distance procedure in a GIS was chosen because of this advantage. In this study, a radius
(buffer) of 914.4 meters (3,000 feet) was generated around the center of each apartment
complex. The designation of this particular radius of was used because the results of
many price-distance gradient studies revealed that the impact of desired and undesired
adjacent land uses on the valuation of single-family dwellings occurred within a half-mile
to 3,000-feet (Darling 1973, Li and Brown 1980, and Munneke and Slawson 1999).
GIS Procedures
Fifty apartment complexes without the presence of condominiums or townhouses
were considered for analysis within the delimited study area. As a result, the apartment
complexes built in the year 2000 or before were geocoded by address using the street
shapefile to determine their geographic location as a point. However, out of the fifty
apartments only thirty-two were located using the geocoding procedure. This is because
the City does not digitize (draw) and maintain the street attribute table concerning
address ranges of private streets where many apartments are located as often as public
streets. Other missing cases involved incomplete address ranges for particular streets
where the apartments would be located. To rectify this problem, the parcel shapefile was
used to locate the property. Basically, a parcel (lot) contains the following attribute
33
information: the owner’s name, address, the apartment complex name, and or the
assessed valuation. Information from the parcel shapefile was then compared with the
multifamily dwelling paper files to physically locate and attribute the missing apartment
complex. Aalberts and Bible (1992) experienced similar problems with their research
and resorted to this particular solution.
Once the locations of the apartment complexes were determined, buffers with a
914.4-meter (3,000-feet) radius were created around all apartment complexes (Bible and
Hsieh 1994). Only the sales of single-family dwellings that occurred within the 914.4
meters of an apartment complex were included because it was assumed that beyond this
radius, the impact o f an apartment would have a minimal effect. The streets that were
encircled by the buffers were then selected using the shape button (Refer to Figure 9).
The street attribute table was opened to gather address ranges, which were entered at the
MLS online database to acquire the selling price and the necessary independent variables
needed of the homes, which sold during the two-year period. During 1999 through 2000,
1,670 single-family dwellings sold within 914.4 meters of an apartment complex.
However, statistical analysis was only performed using 1,665 dwellings because five of
the observations were outliers depicted in preliminary descriptive statistics and scatter
plots. The subsequent section explains the elimination of the data in greater detail. The
homes were geocoded by address using the street shapefile to determine their geographic
location as a point. All of the homes had a seventy-five percent or greater success rate in
being located in Arc View.
34
ES Ed* if«rr EFIeiv* U'4<>i'fCi 3^TdO*» U<* ..........
W && I B ! MSB] M M M ISO BO U It>l b ]> ? ^ 3 l& iy i i iM X I-J ____________________ \ \
Shape ButtonDistance
Pud s ftpC~3
914.4 m. Radius (Buffer)Themes
Lake Zorinsky
T unquiHriS i
f»> utMfe 15m.
f uub-»j*l5mj
Swl-* 1 | 57.511
Fig. 9. ArcView 3.2 interface depicting the necessary themes and tools needed to complete the research.
After locating apartment complexes and single-family dwellings, distance from
the apartment complex to the single-family dwelling was calculated. ArcView has the
capability to measure the distance between two point themes (spatial data layers). The
point themes used to calculate distance were the apartment complex theme and the home
sales theme. A sample Avenue script, provided by the software package, calculated the
distance in meters from the selected apartment complex to the single-family dwellings as
seen in APPENDIX A.
However, two problems with calculating distance using this script occurred.
Consequentially, modifications were developed which solved these problems described
below. The first problem dealt with calculating distance from one apartment complex to
35
the all of the single-family dwellings, which sold in the study area. Thus, distance was
calculated beyond the 914.4-meter buffer centering the apartment complex. The second
problem was many houses were influenced by more than one apartment complex. In
Figure 9 it is apartment that two or more apartment buffers encircled the houses.
Therefore, more than one distance measurement needed to be calculated. To rectify these
problems the home sales theme was subdivided into fifty shapefiles so that each group of
houses was within 914.4 meters. The attribute table for each group contained the
distance measurements, such as Dist one, representing the distance to the closest
apartment, Dist two the second closest apartment, and so forth. The other solution to the
distance problem as well as considering structural density was to determine the number
of apartments within 914.4 meters for each house.
Database Organization and Data Evaluation
The database organization and data evaluation was completed before any of the
statistical procedures were implemented. First, all 50 single-family dwelling shapefile
attribute tables were merged together in one database file. An address sort was
performed, which facilitated in the elimination of repetitive data entries. Distance
measurements for each dwelling were sorted. During the data evaluation process, five
unusual observations were eliminated because of outstanding age values, 71, 96, and 101;
a dwelling with six bathrooms, and another dwelling with a six car heated garage. The
abnormal data were eliminated because if included the statistical analyses it would have
distorted results.
36
Data Analysis
Several valuation studies have used multivariate analysis, such as multiple
regression, the hedonic model (Refer to General Definition of Terms), and factor analysis
for predicting house price by examining how independent variables, such as structural
characteristics of housing influence the dependent variable, sales price (Berry and
Bednarz 1975, Eppli and Tu 1999, Li and Brown 1980, Richardson et al. 1974, and Wang
et al. 1991). These data for this research are analyzed by means of descriptive statistics,
Pearson correlation matrices, multiple regression and factor analysis to test the
hypotheses whether distance or structural density influences the sales price of the single
family dwelling. In the preceding sections, an explanation and justification is given for
each statistical procedure utilized in the research.
Descriptive statistics were performed to summarize the continuous numeric
variables. This also assisted with understanding the nature of data structure in addition to
determining incorrect data entries within the maximum and minimum value range of the
variables.
Pearson correlation matrices were developed to assist with examining bivariate
correlations among the variables and to provide the direction o f the expected signs of the
regression coefficients in later analyses. Pearson correlation matrices were developed for
all the attributes of the single-family dwellings, including all the distance measurements,
in order to analyze the correlation coefficients linear relationship between two variables
within the range of ± 1 (Refer to the APPENDIX D for the full variable correlation
matrix). The value 1 in the Pearson correlation matrix signifies that there is always a
37
perfect linear relationship when a variable is compared with itself, such as square feet
compared with square feet. If the coefficient is significant at the 2-tail level (p-value <
0.05), then the correlation is significant. The p-value = 0.000 indicates that the
correlation is very significant.
Multiple regression analysis involves a single dependent variable that is related to
several independent variables (predictors) (Hair et al. 1987). The purpose of multiple
regression is to predict the response of the dependent variable through variations of the
independent variables (Hair et al. 1987). Multiple regression was used to predict the
housing price because it allows the measure of “consumer behavior” (Eppli and Tu 1999,
36). The distance from the apartment complex to the first single-family dwelling
(Dist one) and the number of apartments within a 914.4-meter radius (Num Apts) were
variables included in the multiple regression model that essentially tested the hypotheses.
The Dist one variable was the only distance variable utilized in the multivariate
analyses because it was the closest measurement from the apartment complex to each
single-family dwelling in the area. Referring to Table 2 in Chapter 4, it is evident that
two or more apartment complexes influenced very few single-family dwellings, thus
providing justification of using just the first distance measurement. Furthermore, it
would not be statistically accurate to implement multiple regression using all of the
distance measurements. It is suggested as a “rule of thumb” that researchers should have
10 times more observations than the number of attributes in a multiple regression model
(Hair et al. 1987, 33). However, in practice most studies use 5 times the number of
observations than attributes (Eppli and Tu 1999 and Hair et al. 1987).
38
In this research study, three multiple regression models were applied to the data
to test both hypotheses. First, a full attribute multiple regression model was applied to
the data (Berry and Bednarz 1973 and Wang et al. 1991). The first model multiple
regression equation was:
(3.1) SP = /(Sq_F t|, Tot_Rm2 , Bath3 , Bed4 , Age5 , Garage6, Num_Apt7 ?Dist_one8 ,)
The second model applied to the data was a reduced attribute model (Berry and
Bednarz 1973 and Wang et al. 1991). Some of the attributes were eliminated from Model
2 because of high correlations evident in the Pearson correlation matrix, unexpected signs
with the coefficients, and if the p-values were greater than 0.05. The second reduced
attribute model of the multiple regression equation was:
(3.2) SP = /(Tot_Rm j, Bed2, Age3 , Garage^ Num_Apt5 )
Factor analysis is a method used to analyze several independent variables and
decide whether they can be reduced and condensed into a smaller set of factors or
components with a minimum loss of information (Hair et al. 1987). This analysis serves
the purposes of identifying latent factors, reducing or removing multicollinearity, and the
factor scores can be used in subsequent multiple regression analysis (Hair et al. 1987).
As a result, factor analysis was performed in this research study because severe
multicollinearity existed in both the full and reduced attribute multiple regression models.
39
All of the variables were entered in the factor analysis and rotated with the orthogonal
Varimax rotation method. This rotation was utilized to uncorrelate all o f estimated factor
score coefficients to ameliorate the potential effect of multicollinearity (Hair et al. 1987
and SPSS 2001).
The factor scores were saved using the Anderson-Bartlet method to keep them
uncorrelated in SPSS (SPSS 2001) so they could be used in a multiple regression
analysis, as described in Model 3. The third model of multiple regression utilizing factor
scores was:
(3.3) SP = / (Factori, Facto r, F acto r)
For the model equation 3.3, Factor 1, represents square footage, total number of rooms,
bathrooms, and bedrooms; Factor 2, represents garage and age; while Factor 3 were the
hypotheses variables number of apartments and distance.
Summary of Methodology
Chapter 3 concerned the methodological design utilized for the thesis as shown in
Figure 8. Data were acquired from various sources such as Douglas County Assessor’s
office, Omaha City Planning Department, MLS, and GIS generated variables, such as
distance and number of apartments. The problems that were encountered with data
acquisition as well as the solution were discussed in this Chapter. Database organization
was concerned with the ordering of distance measurements from the apartment complex
to the first closest house, to the second closest house, and so forth. Data evaluation, such
as eliminating repetitive data and correcting typographical errors were essential before
performing statistical analyses. The Data processing consisted of descriptive statistics,
40
Pearson correlation matrices, multiple regression, and factor analysis. It was discussed
that descriptive statistics assisted understanding the nature of the data structure and aided
in the elimination of outlier data. Pearson correlation matrices were used to distinguish
where potential multicollinearity existed, with the full attribute regression model. The
reduced attribute multiple regression model was performed, but multicollinearity still
existed. Factor analysis was then implemented and the factor scores were saved and
utilized in a subsequent multiple regression model. The models of the multiple
regression equations utilized in this research study are defined in this Chapter. Chapter 4
provides a quantitative and statistical presentation of the data and an explanation of the
results.
41
Chapter 4
DISCUSSION OF RESULTS
In this Chapter an explanation is given for the results produced for the following
statistical analyses: descriptive statistics, Pearson correlation matrices, multiple
regression, factor analysis, and multiple regression utilizing factor scores. Each analysis
proved useful in testing the research hypotheses.
Descriptive Statistics
The descriptive statistics in Table 2 shows the number of observations, the
minimum and maximum, mean, and standard deviation values for each variable
concerning single-family dwellings. It is evident in Table 2 that all of the variables had a
broad range. For example, the sales price ranged from $66,150 to $752,500; square
footage from 920 to 7,900; and total rooms from 3 to 15. The age of the dwellings
ranged from 1 to 44 years with a mean age of 18.11 years. Furthermore, the number of
apartment complexes within 914.4 meters of a house ranged from 1 to 8. The scatter
plots in APPENDIX B clearly indicate that the maximum sales price $752,500 also
contains the maximum values for square footage, total rooms, bathrooms, and bedrooms.
The scatter plots in APPENDIX C that show the impact the hypotheses’ variables
(number of apartments and distance) on sales price indicate that the majority of homes
are within 914.4 meters (3,000 feet) of just one apartment. However, as the number of
apartment complexes increase within the 914.4 meters of a single-family dwelling, the
sales price decreases.
42
All of the homes in the study were within 914.4 meters of at least one apartment
complex because of the selection method. The distance from the apartment complex to
the closest group of houses was called Dist one. However, not all of the houses were
within the radius of 2 or more apartments. For instance, Table 2 shows that D isttw o has
667 single-family dwellings that sold, while Dist eight only has 17, which explains why
the mean number of apartments (Num Apts) within 914.4 meters of a single-family
dwelling is 1.65.
Table 2. Descriptive Statistics for Single-Family DwellingsVariable N Minimum Maximum Mean Std. DeviationSP 1665 $66,150 $752,500 148558.61 63169.307Sq_Ft 1665 920 7900 2050.54 696.104T o tR m 1665 3 15 7.77 1.537Bath 1665 1 8 2.73 .739Bed 1665 1 6 3.35 .602Age 1665 1 44 18.11 11.452Garage 1665 0 4 2.10 .522Num_Apts 1665 1 8 1.65 1.124D istone 1665 12.7216 913.5065 548.244062 207.7627452D isttw o 667 172.9067 912.9890 646.460752 180.7730136D istthree 232 314.6783 912.5189 720.297664 144.1123990Dist_four 72 450.3017 909.2468 696.196751 148.4509670D istfive 51 478.0857 913.2174 685.371157 138.2457610D ists ix 30 559.0625 881.1622 695.305003 70.1587240D istseven 23 634.9855 891.1474 825.035161 58.4735467Dist_eight 17 802.5654 911.2570 864.464706 28.4347109
Descriptive Statistics were also performed on all 50 apartment complexes located
in the study area concerning their one bedroom rents, age, total square footage of the
complex, and total units. The apartment variables were not implemented in multiple
regression or factor analysis because the objective of the research focused on distance
and structural density. However, the descriptive statistics and Pearson correlation results
43
for the apartment complexes found in Table 3 and APPENDIX D respectively provided
some useful information, particularly with age. As seen in Table 3 the age range of the
apartment complexes is from 1 to 35 years and the mean age is 16.84 years. When
comparing the mean age in years of single-family dwellings to the apartment complexes,
it is evident that apartment complexes are only slightly younger than all of the single
family dwellings. Both means depict that construction occurred more or less at the same
time.
Table 3. Descriptive Statistics for the Apartment ComplexesN Minimum Maximum Mean Std. Deviation
Bed_Rnt 50 435 895 540.88 89.233
Apt_Age 50 1 35 16.86 11.254
Tot_Units 50 36 624 199.48 128.561
Tot_SqFt 50 33654 584432 204827.74 125418.974
Pearson Correlation
Two Pearson correlation matrices were developed for this research study. The
first correlation matrix located in APPENDIX D contained all the distance measurements,
while Table 4 used just the closest distance (Dist one) to the apartment complex. Table 4
shows the Pearson correlation coefficients of all the variables to have the expected signs
and significant at the 0.01 level when compared with sales price. Such that the
correlation coefficients for square footage, total number of rooms, bathrooms, bedrooms,
garages, and distance are positive, while age and number of apartments are negative.
44
Table 4. Pearson Correlation Matrix
SP Sq_Ft TotRm Bath Bed Age Garage Num_Apts
Sq_FtPearsonCorrelation
Sig. (2-tailed)
.795**
.000
Tot_RmPearsonCorrelation
Sig. (2-tailed)
.485**
.000
.717**
.000
BathPearsonCorrelation
Sig. (2-tailed)
.561**
.000
.699**
.000
.571**
.000
BedPearsonCorrelation
Sig. (2-tailed)
.415**
.000
.607**
.000
.620**
.000
.510**
.000
AgePearsonCorrelation -.527** -.200** -.033 -.162** -.065**
Sig. (2-tailed) .000 .000 .182 .000 .008
GaragePearsonCorrelation .636** .480** .305** .383** .271** -.473**
Sig. (2-tailed) .000 .000 .000 .000 .000 .000
Num_AptsPearsonCorrelation
Sig. (2-tailed)
-.192**
.000
_141**
.000
-099**
.000
-.111**
.000
-.103**
.000
.130**
.000
-.090**
.000
D isto n ePearsonCorrelation .086** .099** .070** .077** .056* .016 .034 -.341**
Sig. (2-tailed) .000 .000 .004 .002 .023 .505 .162 .000
Note: Pearson correlation matrix stops with Dist one. ** Coefficient is significant at the 0.01 level. * Coefficient is significant at the 0.05 level. For the full matrix seeAPPENDIX D.
It is important to note that correlations between the square footage variable with
the number of total rooms, bathrooms, and bedrooms variables were relatively high. The
Pearson correlation coefficients between square footage and sales price, total rooms,
bathrooms and bedrooms were 0.795, 0.717, 0.699, and 0.607 respectively. Thus,
indicating that multicollinearity would present a problem in later multiple regression
analyses.
45
Multiple Regression
Multiple regression was first performed on a full attribute model to test the
hypotheses by including the number of apartments and distance variables. The t values
for the independent variables represented in Model 1 should be below or above 2 in order
to be considered good predictors (SPSS 2001). In Table 5 the t-values that did not meet
these requirements in the model are number of bathrooms and distance. These same
variables are also insignificant at the 0.05-level. In Model 1 the unstandardized beta
coefficients for total rooms and bedrooms have unexpected negative signs. It was
expected that these signs would be positive, since the Pearson correlation coefficients
depicts them as such in Table 4. The variables had unexpected signs because
multicollinearity existed in the model. The coefficient of determination R2, explains that
80 percent of the variation of sales price can be explained by the attributes used in this
model. The adjusted R is also significant at 79.9 percent.
46
Table 5. Regression Results for Model 1Variable Unstandardized Coefficients
BStd. Error Standardized Coefficients
Betat
(Constant) . 42778.102* 6024.669 7.100
Sq_Ft 65.884* 1.791 .726* 36.779
Tot_Rm -2896.283* (**) 699.407 -.070*(**) -4.141
Bath 62.867 1341.576 .001 .047
Bed -5795.447*(**) 1551.870 -.055*(**) -3.734
Age -1637.470* 70.329 -.297* -23.283
Garage 21638.556* 1693.732 .179* 12.776
Num_Apts -2567.602* 667.865 -.046* -3.844
Dist_one 1.515 3.575 .005 .424
R .894
R2 .800
Adj. R2 .799
Std. Error of The Estimate
2845.437
F Value 826.021*Note: The full attribute model used in multiple regression analysis had N= 1,665. * Significant 0.05 level. ** Coefficient is not with expected sign.
Table 6 shows the regression results for the reduced attribute model, which omits
variables square footage, bathrooms, and distance. Model 2 excludes the variable square
footage due to its high correlation with other housing characteristic variables evident in
Model 1. Bathrooms and distance were insignificant and were also removed from the
model. In Table 6 the results of Model 2 in show that all the variables have the expected
coefficient signs and are significant at the 0.05-level. The t-values were all well above
or below 2 suggesting the independent variables were good predictors. However, with
47
this model coefficient o f determination R2 only accounts for 60.1 percent of the total
variation of the selling price.
Table 6. Regression Results for Model 2.Variable Unstandardized
CoefficientsB
Std. Error StandardizedCoefficients
Beta
t
(Constant) -32210.358 7394.915 -4.356
TotRm 11982.617* 832.088 .292* 14.401
Bed 11413.341* 2088.077 .109* 5.466
Age -1838.226* 98.546 -.333* -18.653
Garage 42712.773* 2265.005 .353* 18.858
NumApts -4296.462* 883.527 -.076* -4.863
R 0.776
R2 0.601
Adj. R2 0.600
Std. Error of The Estimate
39940.632
F Value 500.664*Note: The reduced attribute model used in multiple regression analysis had N=l,665. * Significant 0.05 level. All coefficients had expected signs.
Collinearity statistics in Table 7 show the tolerance and variation o f inflation
factors for Models 1 and 2. The tolerance statistic represents the percentage of variation
in each predictor. If tolerances approach zero and if the variation of inflation factors are
greater than 2, then multicollinearity is considered a problem in the model. Table 7
shows that multicollinearity exists in both models, especially Model 1. For instance, in
48
Model 1 all of the tolerances are slightly under 1 and the variation of inflation factors for
the housing characteristics square footage, total rooms, and bathrooms are slightly greater
than 2. It is evident that the offending variable causing the multicollinearity in Model 1
was square footage, which has the highest variance of inflation factor at 3.220. In Table
7 Model 2 is somewhat better, but the tolerance for all the variables is still less than 1 and
less variation exists. The collinearity statistics performed in Model 2 explain that
multicollinearity was reduced, but not significantly. As a result, factor analysis was
performed on the data to remove multicollinearity.
Table 7. Collinearity Statistics for Model 1 and 2.Model 1 Model 2
Variables Tolerance VIF Tolerance VIFSq_Ft .311 3.220 — —
Tot_Rm .418 2.393 .586 1.706Bath .492 2.034 —
Bed .553 1.808 .606 1.649Age .744 1.343 .753 1.329Garage .618 1.619 .686 1.458Num_Apts .857 1.167 .972 1.029Dist one .875 1.143 -- --
Note: — Indicates that the variables were removed from the analysis. VIF represents variation of inflation factors.
Factor Analysis
Factor analysis was employed to address the issue of multicollinearity. It does
this by reducing and summarizing the data, such that original variables are grouped into
factors with a minimum loss o f information (Hair el al. 1987). All of the variables in
Model 1 were utilized in the factor analysis. The results of the factor analysis are
presented in the following manner: communalities, total variance explained, unrotated
49
and rotated component matrix, a component plot in rotated space, and a component
coefficient matrix. An explanation is given for each of the presented results.
Table 8. Communalities for Factor AnalysisInitial Extraction
Sq_Ft 1.000 .819Tot Rm 1.000 .764Bath 1.000 .664Bed 1.000 .659Age 1.000 .812Garage 1.000 .714Num_Apts 1.000 .673Dist one 1.000 .693
Note: Extraction Method: Principal Component Analysis.
Communalities in factor analysis reflect the amount of variance for all the
variables. For the principal component extraction method, all of the initial values are
equal to 1. The extraction communalities estimate the variance for the factors used in the
analysis. If the extraction values are small, then the factors should be removed from the
analysis (Hair et al. 1987 and SPSS 2001). Table 8 shows that in this research study all
of the extraction communalities are 0.659 or greater justifying the use of all the factors in
the analysis.
Table
9.
Total
Var
iance
Ex
plain
ed
using
Fa
ctor
Ana
lysi
s
50
Vi00"3COoJ
T30>c33crGOo
3GO3.2cS*->oPi
fV* r - ,-m VO*—< min VO r tvd id cnco m t~-
r - r f mOS
i n o oovd Ov voco I-1
,— 1 oo ooCN <NOV i n cqcn
'O<d co >
— VO mt"-. S O 'O (N
3IuW VO T VO
r-~ co -3- S r o co r-~Cw • • •
r O VO i nc3 ^ —>
<4-HO
O p" oco o VOCN cq CNcd —1
VO VO oo VO or - m Ov m ■<3- vo ocq r-; vo oo m oO vd CN cd vd CN oin r - oo oo OV Ov o
VO m vo 00 o Ov Tir - co co vo OV r ocq cq r-; CN - m VO vq ooO vd i n oo vd i n CN^ r
C/2
3c<c<ucoo,eou
3*3c’CDmT3O.g3
c_oosxw- K ija
<uJ33>3CD00W
0 s-
e4/SoQ.EoU
O t " O Ov CO i n Ovco o VO m OV m VO CNCN CO CN vq T t Tfr CO CNCO
MH CN CO T f m VO t - OO
51
In Table 9 the total variance explained provides the initial eigenvalues, extraction
sums of squared loadings, and the rotation sums of squared loadings. Table 9 shows that
in this analysis there are eight components that represent the variables. The total initial
eigenvalues determine the amount o f variance observed. Factor analysis is appropriate
when only a few components that have eigenvalues greater than or equal to 1 explain
most of the variance. As shown in Table 9, the first three components explain most of
the variance, such that they account for 72.456 percent of the cumulative variability of
the original variables. Since the principal component extraction method was used, the
total extraction sums of squared loadings, total percent of variance, and cumulative
percent column are the same as the initial eigenvalues.
In Table 10 the unrotated component matrix shows that 3 components were
extracted using principal component analysis. However, a Varimax orthogonal rotation
was implemented to uncorrelate factor scores, spread the variation among of the variables
evenly, and to assist with interpreting the results by grouping the factor loadings as seen
in Table 11 (Hair et al.1987 and SPSS 2001). In Table 11 it is evident that variables in
the rotated component matrix can be grouped by three distinct components and have their
expected signs. Factor 1 consisted of square footage, total rooms, bathrooms, bedrooms;
factor 2, age and garage; and factor 3, number of apartments and distance.
52
Table 10. Unrotated Component MatrixComponent
1 2 3Sq_Ft .899 9.294E-02 4.521E-02Tot Rm .808 .222 .247Bath .806 .110 5.344E-02Bed .752 .206 .225Age -.316 .367 .760Garage .613 -.178 -.554Num_Apts -.235 .747 -.246Dist one .155 -.686 .445
Note: Extraction Method: Principal Component Analysis. 3 components extracted.
Table 11. Rotated Component MatrixComponent
1 2 3Sq_Ft .864 .256 8.206E-02Tot Rm .873 2.788E-03 4.079E-02Bath .786 .209 5.587E-02Bed .811 7.109E-03 3.731E-02Age 2.136E-02 -.900 -4.443E-02Garage .362 .763 1.416E-02Num_Apts -5.834E-02 -.141 -.806Dist_one 5.623E-02 -8.130E-02 .826Note: Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization.
53
Component Plot in Rotated Space
garage
m
tot distol
0.0
0.0
Component 1o.o
-.5Component 3
Fig. 10. Component Plot in Rotated Space
The component plot in rotated space in Figure 10 also shows the rotated
component matrix. This plot serves three purposes, first it helps with interpretation of
how the components were grouped, it shows the separation o f age and garage, and thirdly
it can be used to justify the signs of variables that test the hypotheses, such that the
number of apartments variable is negative, while distance is positive. Garage and
number of apartments are in the same space as the other variables because of their signs.
54
Table 12. Component Score Coefficient MatrixComponent1 2 3
Sq_Ft .286 .045 -.003Tot Rm .332 -.138 -.025Bath .265 .024 -.016Bed .308 -.125 -.024Age .154 -.655 -.001Garage .017 .496 -.041Num_Apts .052 -.063 -.603Dist one -.018 -.099 .627Note: Extraction Method: Principal Component Analysis.Rotation Method: Varimax with Kaiser Normalization.Component Scores.
Regression Utilizing Factor Scores
The factor scores used in the multiple regression analysis for Model 3 were saved
in the factor analysis procedure using the Anderson-Rubins method. This method
estimates factor score coefficients and keeps them uncorrelated. If the factor scores were
saved as regression factor scores, then correlation could occur even with an applied
orthogonal rotation (SPSS 2001). The three factor scores consisted of the following:
factor 1: square footage, total rooms, bathrooms, and bedrooms; factor 2: age and garage;
0 * 0and factor 3: number of apartments and distance. The R and adjusted R are both at 70.6
and 70.5 percent respectively. As a result, the factors scores of Model 3 capture more of
the variance than Model 2, therefore creating a better predicting model of sales price.
The reason why the unstandardized and standardized beta coefficients are all positive is
because when the rotation occurred during factor analysis, variables that contained
positive coefficients were grouped with variables consisting of negative coefficients, thus
55
canceling each other out. However, the component plot illustrates that all of the variables
had the expected signs, including those, which test the hypotheses, number of apartments
and distance.
Table 13. Regression Coefficients of Factor Scores for Model 3.Unstandardized
CoefficientsB
Std. Error StandardizedCoefficients
Beta
t
(Constant) 148558.609 840.164 176.821
Factor 1: Housing Characteristics
36826.739* 840.416 0.583* 43.820
Factor 2: Age and Garage 37570.1758* 840.416 0.595* 44.704
Factor 3: Number of Apartments and Distance
7033.610* 840.416 0.111* 8.369
R 0.840
R2 0.706
Adj. R2 0.705
Std. Error of The Estimate
3482.385
F-Value 1329.558*Note: The multiple regression attribute model using factor scores saved using the Anderson-Rubins method had N =1,665. * Significant 0.05 level. Refer to Figure 10 and Table 11 for the signs of the factor score regression coefficients.
As shown in Table 14 the collinearity statistics for Model 3 show that all of the
factor groupings have a tolerance and variation of inflation factors of 1, indicating that
multicollinearity was removed from the analysis. When comparing the collinearity
statistics for Models 1 and 2 in Table 7 with Model 3 in Table 14 it is evident that
multicollinearity was clearly eliminated.
56
Table 14. Collinearity Statistics for Model 3.
Factors Model 3
Tolerance VIF
Factor 1: Housing Characteristics 1.000 1.000Factor 2: Garage and Age 1.000 1.000
Factor 3: Number of Apartments and Distance
1.000 1.000
Summary of Results
The first hypothesis, where sales price o f single-family dwellings increases with
increasing distance, was supported only after performing factor analysis. This is because
of severe multicollinearity existed among the variables in Models 1 and 2 as shown in
Table 7. Although single-family dwellings that are closer to apartment complexes sell for
a lower price and price increases with increasing distance, the relationship is strong at
0.627 only when analyzing component 3 in the component score coefficient matrix in
Table 12. The subsequent regression utilizing the factor scores saved in the Anderson-
Rubins method, indicated that factor 3: number of apartments and distance, are
significant at the 0.05 level (Refer to Table 13). Although they are significant, the
unstandardized and standardized coefficients are positive. This is because in the process
of rotation in factor analysis grouped the variables N um A pts and Dist one, with
differing signs together, thus canceling the negative coefficient sign of Num Apts.
However, using the component plot is essential to determine that the Num Apts variable
is in fact negative, while Dist one is positive.
57
The second hypothesis, that dealt with the greater number of apartment complexes
within 914.4 meters of a single-family dwelling, the lower the sales price was also
supported when examining all of the analyses. For instance, the results of the scatter plot,
the full and reduced attribute multiple regression model, the unrotated and rotated
components, and the factors analysis scores (when analyzing the component loading plot)
all had the expected negative coefficient. The results from testing the second hypothesis,
show that the number of apartments variable has a stronger relationship than the distance
variable, however only a few single-family dwellings are influenced by multiple numbers
o f apartment complexes as was seen in Table 2 and APPENDIX C.
58
Chapter 5
CONCLUSION
The American dream is to own a single-family dwelling. It is evident that the
federal government has supported this dream with the implementation of housing
programs, favorable tax incentives, and zoning regulations. Furthermore, the placement
of particular housing structures is an emotional issue. This is because owning a home
signifies a persons’ placement in the social fabric of the American society as well as
being an important investment.
The literature and case study presented in this thesis explained that homeowners
dislike apartment complexes because they lower property values, are aesthetically
displeasing, create a greater demand for goods and services, increase noise pollution and
crime rates, create traffic congestion, and disrupt their spatial visualization of what their
landscape should look like. For the most part, the literature exhaustively explains the
homeowners’ dilemma, but fails to analyze it quantitatively. The objective for this
research study is to analyze the impact apartment complexes have on the sales price of
single-family dwellings by using distance and structural density as factors. The two
hypotheses determined if the selling price of single-family dwellings increase with
increasing distance from an apartment complex and if the greater the number of
apartment complexes within 914.4 meters (3,000 feet) o f a single-family dwelling the
lower the selling price.
The first step in the methodological design determined the necessary data required
to test the hypotheses. The data used in this thesis came from the Douglas County
59
Assessor’s office, the Omaha City Planning Department, and data generation by a GIS to
obtain distance and number of apartment variables. The quantitative methods utilized
were descriptive statistics, Pearson correlation matrices, multiple regression, factor
analysis, and regression using factor scores. Descriptive statistics assisted with
understanding the nature of the data structure. Pearson correlation matrices explained the
bivariate relationship among all the variables to determine where multicollinearity
appeared severe, while the multivariate analyses: multiple regression, factor analysis, and
regression using factor scores were utilized to determine what model best predicted house
price by examining how independent variables, such as structural characteristics of
housing and the variables, which tested the hypotheses, number of apartments and
distance, influenced the dependent variable, sales price.
The results o f the quantitative analysis performed on the data indicated that both
the first and second hypotheses are supported. The selling price of single-family
dwellings increased with increasing distance, but only after performing factor analysis
and regression analysis utilizing factor scores. Regression using factor scores was
utilized because severe multicollinearity existed in both the full and reduced attribute
multiple regression models. However, the second hypothesis where selling price of
single-family dwellings decrease with increasing numbers o f apartment complexes was
supported by all of the multivariate analyses including the full and reduced attribute
multiple regression model, the Varimax rotated factor analysis scores, and regression
utilizing factor scores. The regression coefficient for the number o f apartments variable
was negative for Models 1 and 2 and were significant at the 0.05-level. The number of
60
apartments variable had a positive coefficient when using factor scores in regression
because it was grouped with distance. As a result, referring to both the rotated
component matrix and the rotated component plot are essential when analyzing the
factors scores because they showed that the number of apartments variable were in fact
negative and distance positive. In other words, both hypotheses are supported, but the
support of the second hypothesis is stronger. This is because the number of apartments
variable had a negative coefficient in Models 1 and 2 and in the results presented in the
tables for factor analysis. More research needs to be done to investigate the multifaceted
effect apartment complexes have on the value of single-family dwellings.
Future research could investigate various issues concerning the impact of
apartment complexes on the property value of single-family dwellings. For instance,
research could involve utilizing various distance measurements, such as a weighted
inverse distance squared, measuring distance via a network of roads, straight-line
distance, and zonal distance from an apartment complex to single-family dwellings to
determine which method of measurement is most accurately represented in multivariate
models to predict sales price. An additional study could differentiate the impact of
various land uses, such as commercial centers, major and minor roads, and apartment
complexes to determine which has the most influential effect on the sales price of single
family dwellings. Still another potential study could involve whether the introduction of
apartment complexes adjacent to existing neighborhoods comprised o f single-family
dwellings influenced sale prices.
61
APPENDIX A
ESRI Avenue Script
62
ESRI ArcView 3.2 Sample Script: View.CalculateDistanceThe following comments regarding this Avenue script are quoted from (ESRI 1999) Title: Calculates distances from points in one theme to points in another
Topics: Analysis
Description: This script will prompt you for two point themes in the active view. The first is the point theme containing the selected points that you wish to calculate the distance FROM. The second is the point theme containing points that you wish to calculate the distance TO.
You will also be prompted for an identifying field in the FROM theme. The value of this field will be used to name to distance field in the TO theme.
A distance field will be added to the TO theme for each point selected in the FROM theme. This distance field will be populated with the distance between the selected From point and the To point.
If the view is projected, the distance will be returned in distance units.
Requires: This script should be attached to a button in the view GUI. To prepare to use this script:
- 1. Add to your view two point themes.- 2. Determine which is to be theFromTheme and which is to be theToTheme.- 3. Determine which item is to be the identifying field. This is most likely
a field that contains a key number or name.- 4. Create a selected set o f points in theFromTheme, if you like.- 5. Attach this script to a button inthe Veiw GUI, if you like.
Self:
Returns:
theThemeList = av.GetActiveDoc.GetThemes.Clone thePrj = av.GetActiveDoc.GetProjection
' Use these lines to hard-code the input parameters...'theFromFtab = av.GetProject.FindDoc("Viewl ").FindTheme("FromTheme").GETFTab 'theFromlDField = theFromFtab.FindField(”Bnmb")’theToTheme = av.GetProject.FindDoc(”Viewl ”).FindTheme("ToTheme")’theToFtab = theToTheme.GetFTab
63
' Use these lines to prompt the user for input parameters...theFromTheme = (MsgBox.List(theThemeList,"to calculate the distance FROM.","Please select a theme..."))if (theFromTheme = nil) then exit end theFromFtab = theFromTheme.GetFTab theThemeList.RemoveObj(theFromTheme)theToTheme = (MsgBox.List(theThemeList,"to calculate the distance TO.","Please select a theme..."))if (theToTheme = nil) then exit end theToFtab = theToTheme. GetFTabtheFromlDField = MsgBox.ListAsString(theFromFtab.GetFields,"containing the point identifier.","Please select a field...")
if (theFromlDField = nil) then exit
end
theFromShapeField = theFromFTab.FindField("Shape") theToShapeField = theToFtab.FindField("Shape")
theT oFtab. SetEditable(true)
for each f in theFromFtab.GetSelection
theFromID Value = theFromFtab.RetumValueString(theFromIDField,f) theNewFieldName = "DistTo"+theFromIDValue
' If a distance field doesn't exist, add it... if (theToFtab.FindField(theNewFieldName) = nil) then
' Choose the field size commensurate with the view's projection, if (thePrj.IsNull) then theDistanceField = field.make(theNewFieldName,#field_decimal,8,4)
elsetheDistanceField = field.make(theNewFieldName,#field_decimal,8,2)
end
theT oFtab.addfields( {theDistanceField} )' If a distance field does exist, clear it...
elsetheToFtab. Calculate( "0 ",theToFtab.FindField(theNewFieldName)) theDistanceField = theToFTab.FindField(theNewFieldName)
end
64
' Get the point location you are measuring FROM.' If the view is projected, get the the point location in projected units.
theFromShape = theFromFTab.RetumValue(theFromShapeField,f) if (thePij.IsNull.Not) thentheFromShape = theFromShape.RetumProjected(thePij)
end
for each t in theToFtab ' Get the point location you are measuring TO.' Get it in projected units, if the view is projected. theT o Shape = theT oFT ab .Return V alue(theT oFtab .F indField(" Shape " ),t)
if (thePij.IsNull.Not) thentheToShape = theToShape.RetumProjected(thePij)
end
' Calculate the distance between the two points.' Add the value to the output (TO) branch table. theDistance = theFromShape.Distance(theToShape) theT oFtab. SetV alue(theDistanceF ield,t,theDistance)
end
end
theT oFtab. SetEditable(false)
av.ShowMsg(”Distance calculation complete")
65
APPENDIX B
Scatter Plots: Sales Price Versus Housing Characteristics
03C /D
800000
700000
600000
500000
400000
300000
200000
100000
0
□ D □ cP
1000 2000 3000 4000 5000 6000 7000 8000
Square Footage
800000
700000
600000
500000<uot—£ 400000 ■
_u CS
C /D300000 ■
200000
100000
12 14 162 84 6 10
Total Number of Rooms
800000
700000
600000
500000
400000
300000
200000
100000
00 2 4 6 8 10
Number of Bathrooms
800000
700000
600000
500000
400000
300000
200000
100000
02 3 4 70 1 5 6
Number of Bedrooms
800000
700000
600000
500000
400000 ■
□ §300000
200000 ■
qO □ □DS100000
Age of the Dwelling in Years
800000
700000 ■
600000 ■
500000 .
CL 400000
300000 ■
200000
100000
Number of Garages
69
APPENDIX C
Scatter Plots: Sales Price Versus Hypotheses Variables
800000
700000
600000
500000
O. 400000
on300000 «
200000
100000
Number of Apartments Influence Single-family Dwellings
800000 ■
700000 ■
600000 ■
500000 .
400000
300000
200000
100000
0 __________0 200 400 600 800 1000
Dist one
71
400000
300000 ■
200000 ■
100000
400000
300000
200000
osC / 3
100000 ®,nS.D*a3l'£6 £ 4 $%
300 400 500 600 700 800 900 1000
Dist three
180000
160000
140000
cn
120000 «□ m
100000 ■□ D
□ □
80000
500 800 900400 600 700 1000
Dist four
180000
160000
140000
120000 ■
100000 ■
80000
800 900 1000400 500 600 700
Dist five
73
140000
130000
120000□ □<DO
110000
100000
90000
80000800 900500 600 700
Dist six
140000 ’
□130000 ■
120000 • □<DO
*c
□ □ □□ □□ D □
M 110000 ■ □ □JU
□CZJa ° °
100000 ■ □□□
□
90000 « □
80000 1600 700 800 900
D istseven
120000 ■
110000
100000
860 880
Dist eight
75
APPENDIX D
Pearson Correlation Matrices
Pea
rson
C
orre
lati
on
Mat
rix
Dist
ei
ght
L ___
___
-.393 00 ^ r—sm *
^ W .006
-.47
3 inm©r-VOm ^
00©
r- 'S 'm ,— , CN *\ o * .0
08 r -
Dis
t_se
ven
V© CNVO tT iCNosCN m cr>CN
incn<N©ooCN
m<N -.45
6(*
) oCN©cnCN
r nsoCN
rfCNCNr^tCN
in ^ tj- * in •»
r*»©© CN
Dis
t_si
x
VOr-CNOsm Om mCN
OsCN
©minOs©
OOVO
©mCNr-rs ^r ©m
©CN OSCNin©m
os *inCN©©
om
Dis
t_fi
ve
2 5^in
oo© «Ti vO * OO inSOOs 00so in
VO ^-X *in
o©© in00oo©
00cnin msoSO©
mp in
Dis
tfo
ur
.426
(**) oop
CN
.326
(**) inOO
CNr-vsm ooinCN
CNr-os ^ O *rn
oo©©CNr-
r-m inCN
CNr-r© m
CN
Dis
tjhr
ee
O -p Vi * <N
oopCNm(N .2
12
(**) op
CN c«"s CN .135 (*)
pCNmCN
oo ^ © * c n *_s
©©CNmCN
r *© CN©
232 CN©
sorfr -CNcnCN
Dis
ttw
o V impOSso r*-soVO -.
049 inoCsj
r*-VOSO00m©
CNCNrnr *SOSO -.
036
.349 r^soso
r -OO P ' inCN©rSOSO
CN ^ 2
oo©©r *SOSO
Dis
t_on
e
\C>oo * ooop 16
65
.099
(»*) ©©© 1665
.070
(«) ©p 16
65 r - * ©CN©© 16
65 s oin * mCN© 16
65 s o
pin©in 16
65
Num
Apt
s
<NOs * —< *oop 16
65 —-©©© 16
65 Os ^ Os * © *
©©© 1665 * « *
©©© 1665m ^ © *
*©©© 16
65
(**)o
er
©©© 1665
Gar
age
.636
(**) o
op 1665 O 00 *
^ w©©© 16
65 ins /-n © * cn
©©© 16
65
.383
(M) ©©© 16
65
«*rCN
©©© 1665 cn ^ r - *Tf *
i’ w
©©© 1665
Age CN *•A *
oop 16
65 o ^ o *CN *
©©© 16
65 mm©
CN00
1665 CN y-s SO *
*—* *©©© 16
65
i n ^ s o -X- © *
OO©© 16
65 - 1665
Bed
^ t-S
oop 16
65
.607
(**
) ©©© 16
65
.620
(**) ©
©© 16
65 2 rir> *
©©© 16
65 - 1665
• ns o *® w
oo©© 16
65
Bat
h
SO *ooo 16
65
.699
(**
) ©©© 16
65
.571
(**) ©
©© 16
65
- 1665 2 5 P
m *_,
©©© 16
65 CN ^ SO * ^ «
©©© 16
65
To
tRm Vi p . OO *
T toop 16
65 £ i rr~-
©©© 16
65 - 1665
i n©©© 16
65 ofN * ^ *
©©© 1665 m©
CNOO
1665
Sq_F
t
V.OS *r-; *_,
ooo 16
65 - 1665 ! - ^ ©©© 1665 Os ^os *
v o *©©© 16
65 op
©©© 16
65 © ^ o *CN *
©©© 16
65
(XCO - 16
65
.795
(**) ©©© 1665 • r i - p '
o o * ©©© 1665 SO *
m©©© 16
65 ©©© 1665
r - ^ CN * i n *
»’ w©©© 16
65
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
Z
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
Z
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
z
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
Z
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
z
Pear
son
Cor
rela
tion
Sig.
(2-t
aile
d)
Z
CUc n
Sq_F
t
Tot
_Rm
Bat
h
Bed
Age
77
w r~- 'S' 1>-
-.537 (*)
.026 r-
*©SOp .0
19
.126
.631
.295
.250 r oor-*rs .280
-.263 rSCS cn<N .574
(**) OO cnrs
r-* ^ *r-** *©©© cnrs
**rTcnr-
©©© cnrs©r-rs
00OS cnrsNOcn©
©r*-00 cnrscnrsrs
Vi©cn
.120 O n<NVi ocn
O,,*roi’cnoo ©cn
r-©rsrsr-rs ©cn
©©rs©00rs ©cn ©
rsOs
©cn 2 r cn©p ©cn .699
(**) ©©p
.065 Os
© tnSOO 46 © *r s /
oop vsvsOs©
SO©vs ViSOSOrs
©SO© IT)OOcn * oo
©©© vs<N ^ ir> *O n
©©© V) -
.052 <Nvoso
rsr-rs ^ so * r- *
©©© rsr-r-so so rsr-
rr- r-cn rsr- .869
(**) ©©© rs - rsr- .952
(**) ©©©
9ZV so*nprscnrs
cn NTj* 46 p 46r w©©©
rscnrsoocn * rs
©©©rsmrs
4P46
Vi
©©©rscnrs -
rscnrs .869
(**) ©©© rsr .838 ©©©
-.027 00 r-SOSO
rs ^SO 46Tf *©©p
r-*SOSOSO © * r** +_s
©©pr-soso - SOSO
— 46vs©©© 23
2 rr- cn rs ©©rs©©©
.034 <NSO
1665 cn *
©©© 1665 - 1665
*46©©
©©©r-SOSO .2
38(*
*) ©©©rscnrs
r-SO © rsrV)Os©
©©p
o ^ o\ * ooop 16
65 - 1665 4Pcn ♦
©©© 1665
4646rsso
©©© vOsocn , s46© 46
©©©rscnrs
rs ^SO 46r- 46©©© rsr-»
NO ^ © 46 SO 46©©©
- 1665 o ^Os 46 O *
©©© 1665 cnp
rsSO
1665 r -rspr
00 r—sosoSOrs SO*np
rscnrsrsV)p
rsNOp rs ViSO©ONTT©
m ,—„ t~~ * rfoop 16
65 § r©©© 16
65 so©
vs©vs 1665
*46rs©
00©© soso rspsor*
rscnrscnr-© rJ"Vi rsr-*
NO©©cn©
^ iTfNo©p 16
65 o ♦ —< *©©© 16
65
.056 (*) cnrs© 1665
46
1 C=oo©V)rs©
r-soSO © rs© rscnrsr-m v>rs rsr-
OO00©oocnVi
.383
(**) oop 1665 — * —- *r -w/
©©o 1665 r-~ -—' r- I
=> ^rs©p 16
65 socnpOscn
r-soNOoo ^ © *rs ©p
rscnrsos° Irn
oo©© rs S iPV)©©p
.305 (**) ©op 16
65 Os ^Os 46 o 46©©© 16
65
.070
(**) tj-©© 1665 oocnp
rsrscnr-NO©
V im « ©rscnrs
vsm 00V irs rsr-
©O n00©
.480
(**) oop 1665 ^ 4P— 46
©©© 1665 CTv 'T '' Ov * O
©©© 1665 on
'T t©vs©<N
r-SOSO .212 (**)
oprscnrs .3
26 V i©©rsr- .4
67(**) ©©
SOr*1 *SO *ooo 16
65 rsOn 46 -1—* *©©© 16
65 VO00 £ o
©©© 1665 v>m©
Os©cnr-soso rs
©©©rscnrs
vofN * TT©©© rsr- 2 rin
©©©
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
Z
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
Z
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
z
Pear
son
Corr
elatio
n
Sig.
(2-ta
iled)
Z
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
Z
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
Z
Pear
son
Corr
elatio
n
Sig.
(2-ta
iled)
Gar
age
Num
_Apt
s
Dist
_one
Dis
ttw
o
Dis
tthr
ee
Dis
tfou
r
Dis
tfiv
e
78
-.211
.416 r-
.948
(M) ooo - r~-
** Co
rrela
tion
is sig
nific
ant
at the
0.0
1 lev
el (2
-taile
d).
* Co
rrela
tion
is sig
nific
ant
at the
0.0
5 lev
el (2
-taile
d).
a Ca
nnot
be
com
puted
be
caus
e at
least
one
of the
va
riabl
es
is co
nsta
nt.
23 -.026
.906 23 - 23 .948
(**)
.000
30 - 30 -.026
.906 23 -.211
.416 r -
inon r-- O' *V0 *
Ooo 30 -.223
.305 23 .278
.280
to 2 ST«o
mop 30 -.036
.870 23 .295 otoCN 17
oCNtoP 30 -.2
79
.198 23
921' .631 r -
tO .200
.289 OCO -.734
(**) OOO 23
oSO jp tor .0
19
tO .207
.272 30
r - ^Tf * oop 23 -.5
37 (*)
.026
oo
TTm 30 .574
(**) oo 23
tO .120 osCNto 30 -.263
.224 23 '55' r~-
tOon -r-Tt *in *_, .0
02 30
to ^ *to * .0
07 23
CO r— sCN * pOOop
r-
*0021'
OnCNto 30 -.263
.224 23
tO .272
.147 30 -.456 (*)
.029 23 -.567 (*) oo
or-
to .095 00
p 30 -.235
.280 23 -.473
.055 r -
IT) .231
.219 30 -.219
.314 ZZ r—NCO * SO * .006
tO .276
.139 30 3
.462 CO
CN -.393 OO
z
Pear
son
Corr
elatio
n
Sig.
(2-ta
iled)
z
Pear
son
Corr
elatio
n
Sig.
(2-ta
iled)
z
Pear
son
Cor
rela
tion
Sig.
(2-ta
iled)
Z
Dis
tsix
Dist
_sev
en
Dis
teig
ht
79
Pearson Correlation Matrix For Fifty Apartments in the Study AreaBedRnt Age AptSqFt TotUnits
Bed_Rnt Pearson Correlation 1 -.577** .184 -.052
Sig. (2-tailed) .000 .202 .718
N 50 50 50 50
Age Pearson Correlation -.577** 1 -.107 .006
Sig. (2-tailed) .000 .460 .967
N 50 50 50 50
Apt_SqFt Pearson Correlation .184 -.107 1 .945**
Sig. (2-tailed) .202 .460 .000
N 50 50 50 50
TotU nits Pearson Correlation -.052 .006 .945** 1
Sig. (2-tailed) .718 .967 .000
N 50 50 50 50Note: ** Correlation is significant at the 0.01 level (2-tailed).
80
BIBLIOGRAPHY
Aalberts, R. J. and D.S. Bible, “Geographic Information Systems: Applications for the Study of Real Estate.” The Appraisal Journal 60:4(1992): 483-492.
Adams, John S. “Residential Structure of the Midwestern Cities.” Annals o f the Association o f American Geographers 60, no.l (1970): 37-62
________ . “The Meaning of Housing in America.” Annals o f the Association o fAmerican Geographers 74, no.4 (1984): 515-26.
________ . Housing America in the 1980s. New York: Russell Sage Foundation,1987.
Alonso, William. Location and Land Use; Toward a General Theory o f Land Rent. Cambridge: Harvard University Press, 1964.
Asabere, Paul K. and Forrest E. Huffman. “Hierarchical Zoning, Incompatible Uses and Price Discounts.” Real Estate Economics 25, no. 3 (1997): 439-451.
Babb, Carol E., Louis G. Pol, and Rebecca F. Guy. “The Impact of Federally-Assisted Housing on Single-Family Housing Sales: 1970-1980.” Mid-South Journal o f Business 4, no. 3 (1984): 13-17.
Berry, Brian J.L. and Robert S. Bednarz. “A Hedonic Model of Prices and Assessments For Single-family Homes: Does the Assessor Follow the Market or the Market Follow the Assessor?” Land Economics 51, (1975): 21-40.
________ ,_______ , and H.M. Kim. “Challenges to the Monocentric Model.”Geographic Analysis 25, no.l (1993), 1-4.
Bible, Douglas S. and Cheng-Ho Hsieh. “Applications of Geographic Information Systems for the Analysis of Apartment Rents.” The Journal o f Real Estate Research 12, no. 1 (1996): 79-87.
Borchert, John R. “America’s Changing Metropolitan Region” Annals, Association o f American Geographers 62, no.2 (1972): 353-73.
Bourne, Larry S. Geography o f Housing. New York: Wiley, 1981.
Burbach, Chris. “Board Votes to Rezone Stonybrook.” Omaha World-Herald. 3 August 2000, Thursday Sunrise Edition.
81
Cadwallader, Martin. Urban Geography: An Analytical Approach. Upper Saddle River: Prentice Hall, 1996.
Can, A. “Measurement o f Neighborhood Dynamics in Urban House Prices.” Economic Geography 66 (1990): 254-272.
City of Omaha Planning Department. Omaha Master Plan. Omaha, 1997.
Crone, Theodore M. and Richard P. Voith. “Risk and Return within the Single-Family Housing Market.” Real Estate Economics 27, no. 1 (1999): 63-78.
Darling, Arthur H. “Measuring the Benefits Generated by Urban Water Parks.” Land Economics 49, (1973): 22-34.
Dobriner, William M. Class in Suburbia. Englewood Cliffs: Prentice Hall, 1963.
Downs, Anthony. Opening Up the Suburbs; An Urban Strategy For America. New Haven: Yale University Press, 1973.
________ . Stuck in Traffic: Coping With Peak-Hour Traffic Congestion.Washington, D.C.: Lincoln Institute of Land Policy, 1992.
Douglas County Assessors. “Douglas County Assessed Value Inquiry.”http://www.co.douglas.ne.us/dept/assessor/framevalinfo.htm Viewed April 17, 2001.
Environmental Systems Research Institute, Inc. (ESRI). ArcView 3.2:
Environmental Systems Research Institute, Inc. (ESRI). Using ArcView GIS. Redlands: Environmental Systems Research Institute, Inc., 1996.
Fujita, Masahisa. Urban Economic Theory: Land Use and City Size. Cambridge: Cambridge University Press, 1989.
Hair Joseph F. Jr., Rolph E. Anderson, and Ronald L Tatham. Multivariate DataAnalysis: With Readings. 2d ed. New York: Macmillan Publishing Company, 1987.
Haring, Lloyd L., John F. Lounsbury, and John W. Frazier. Introduction to Scientific Geographic Research. Boston: WCB McGraw-Hill, 1992.
Harris, Richard and Robert Lewis. “Constructing a Fault(y) Zone: Misrepresentations of American Cities and Suburbs, 1900-1950.” Annals o f the Association o f American Geographers 88, no. 4 (1998): 622-639.
82
Hartshorn, Truman A. Interpreting the City: An Urban Geography. 2nd edition. New York: John Wiley and Sons, Inc., 1992.
Keating, David C. “Exclusionary Zoning: In Whose Interests Should the Police Power Be Exercised?” Real Estate Law Journal 23, no. 4 (1995): 304-331.
Li, Mingche M. and H. James Brown. “Micro-Neighborhood Externalities and the Hedonic Housing Prices.” Land Economics 56, no. 2 (1980): 124-141.
Me Millen, Daniel P. and John Me Donald. “Population Density in Suburban Chicago: A Bid-Rent Approach.” Urban Studies 35, no.7 (1998): 119-131.
Moudon, Anne V. and Paul M. Hess. “Suburban Clusters: The Nucleation ofMultifamily Housing in Suburban Areas of the Central Puget Sound.” Journal o f the American Planning Association 66, no. 3 (2000): 243-264.
Mundy, Bill. “The Scientific Method and the Appraisal Process.” The Appraisal Journal 60, no. 4 (October 1992): 493-499.
Munneke, Henry J. and V. Carlos Slawson Jr. “A Housing Price Model with Endogenous Externalities and Location: A study of Mobile Home Parks.” Journal o f Real Estate Finance and Economics 19, no. 2 (1999): 113-131.
Muth, Richard F. Cities and Housing; The Spatial Pattern o f Urban Residential Land Use. Chicago: University of Chicago Press, 1969.
National Apartment Association. “National Apartment Association Zoning Moratorium White Paper.” (June 1997): 1-26.
Omaha Planning Board Meeting, 12 July 2000.
Omaha Planning Board Meeting, 2 August 2000.
Phipps, Rod. Omaha City Planning Manager. Interview by author, 9 August 2000.
Platt, Rutherford. Land Use and Society: Geography, Law, and Public Policy.Washington D.C.: Island Press, 1996.
Purcell, Mark. “Neighborhood Activism Among Homeowners as a Politics of Space.” Professional Geographer 5, no. 2 (2001): 178-194.
Richardson, Harry W., Joan Vipond, and Robert A. Furbey. “Determinants of Urban House Prices.” Urban Studies 11, (1974): 179-199.
83
Simons, Robert A., Roberto G. Quercia, and Ivan Marie. “The Value Impact of NewResidential Construction and Neighborhood Disinvestment on Residential Sales Price.” The Journal o f Real Estate Research 15, no. 15 (1998): 147-161.
Sweeny, J. L. “Housing Unit Maintenance Model of Tenure” Journal of Economic Theory 8, (1974): 111-138.
Smith, Betty E. “A Review of Monocentric Urban Density” Journal o f Planning Literature 12, no. 2 (1997): 115-136.
SPSS For Windows: Standard Evaluation Version, 11.0., 19 September 2001. [software on-line] available from: http://www.spss.com: Internet; accessed 1 May 2002.
Waddell, Paul, Brian J.L. Berry, and Irving Hoch. “Housing Price Gradients: TheIntersection of Space and Built Form.” Geographical Analysis 25, no. 1 (1993): 5- 19.
Wang, Ko, V. Grissom, James Webb, and Lewis Spellman. “The Impact of Rental Properties on the Value of Single-Family Residences.” Journal o f Urban Economics 30, no. 2 (1991): 152-166.
Van Vliet, Willem. The Encyclopedia o f Housing. Thousand Oaks: Sage, 1998.
