<%BANNER%>

Sub-Prime to Suboptimal

Permanent Link: http://ufdc.ufl.edu/UFE0043813/00001

Material Information

Title: Sub-Prime to Suboptimal Realized Effects of the Foreclosure Crisis on Neighborhood Quality in Hillsborough County, Florida
Physical Description: 1 online resource (80 p.)
Language: english
Creator: Gibbons, Charles M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: foreclosure -- neighborhoods -- sub-prime
Urban and Regional Planning -- Dissertations, Academic -- UF
Genre: Urban and Regional Planning thesis, M.A.U.R.P.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With its roots in risky subprime lending practices, the foreclosure crisis of the late 2000s gutted neighborhoods across the United States. As foreclosed properties sit vacant, they can begin to negatively affect the actual and perceived quality of a neighborhood. The State of Florida sits at the epicenter of this crisis, with many communities incurring thousands of home foreclosures. This study looks to understand the local impacts of foreclosures on neighborhood quality in Hillsborough County, FL. In utilizing home price as a proxy for neighborhood quality, small geography data allowed for analyses at a fine level of detail. Using both Ordinary Least Squares and Geographically Weighted Regression techniques, the impacts of foreclosure and vacancy on home value were modeled with spatial context. To measure physically tangible effects, the employment of cluster analysis allowed clusters of foreclosures and code violations to be compared. The results of this study show that home foreclosures have had the greatest and most disproportionate effect on home values in high income neighborhoods. However, these impacts have not yet translated to noticeable physical effects in wealthyneighborhoods. Low income neighborhoods have seen a lesser impact on home values, but seemed to have reached a tipping point of negative impacts; here physical deterioration is found to cluster near foreclosures. The results of this study highlight the need for planning practitioners and local officials to focus their community stabilization efforts on these low income neighborhoods racked by foreclosure.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Charles M Gibbons.
Thesis: Thesis (M.A.U.R.P.)--University of Florida, 2011.
Local: Adviser: Blanco, Andre.
Local: Co-adviser: Zwick, Paul D.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043813:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043813/00001

Material Information

Title: Sub-Prime to Suboptimal Realized Effects of the Foreclosure Crisis on Neighborhood Quality in Hillsborough County, Florida
Physical Description: 1 online resource (80 p.)
Language: english
Creator: Gibbons, Charles M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: foreclosure -- neighborhoods -- sub-prime
Urban and Regional Planning -- Dissertations, Academic -- UF
Genre: Urban and Regional Planning thesis, M.A.U.R.P.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With its roots in risky subprime lending practices, the foreclosure crisis of the late 2000s gutted neighborhoods across the United States. As foreclosed properties sit vacant, they can begin to negatively affect the actual and perceived quality of a neighborhood. The State of Florida sits at the epicenter of this crisis, with many communities incurring thousands of home foreclosures. This study looks to understand the local impacts of foreclosures on neighborhood quality in Hillsborough County, FL. In utilizing home price as a proxy for neighborhood quality, small geography data allowed for analyses at a fine level of detail. Using both Ordinary Least Squares and Geographically Weighted Regression techniques, the impacts of foreclosure and vacancy on home value were modeled with spatial context. To measure physically tangible effects, the employment of cluster analysis allowed clusters of foreclosures and code violations to be compared. The results of this study show that home foreclosures have had the greatest and most disproportionate effect on home values in high income neighborhoods. However, these impacts have not yet translated to noticeable physical effects in wealthyneighborhoods. Low income neighborhoods have seen a lesser impact on home values, but seemed to have reached a tipping point of negative impacts; here physical deterioration is found to cluster near foreclosures. The results of this study highlight the need for planning practitioners and local officials to focus their community stabilization efforts on these low income neighborhoods racked by foreclosure.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Charles M Gibbons.
Thesis: Thesis (M.A.U.R.P.)--University of Florida, 2011.
Local: Adviser: Blanco, Andre.
Local: Co-adviser: Zwick, Paul D.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043813:00001


This item has the following downloads:


Full Text

PAGE 1

1 SUB PRIME TO SUBOPTIMAL : REALIZED EFFECTS OF THE FORECLOSURE CRISIS ON NEIGHBORHOOD QUALITY IN HILL S B O ROUGH COUNTY, FLORIDA By CHARLES MYRON GIBBONS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS IN URBAN AND REGIONAL PLANNIN G UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Charles Myron Gibbons

PAGE 3

3 To my incredible parents, Jane and Gary Gibbons Thank you f or always

PAGE 4

4 ACKNOWLEDGMENTS I would first like to extend a sincere thanks to my Chair, Dr. Andres Blanco, and Co Chair, Dr. Paul Zwick. This research would have never come to fruition without the countless hours I spend in their courses and office hours. Dr. Zwick opened my eyes to so many new and exciting applications of GIS, while constantly pushing me to conceive of new methods and approaches in geospatial technologies. Dr. Blanco has been an incredible resource in the synthesis of my research. Without his help, I would surely be lost in the conceptualizing and implementation of my research methods. To my fellow classmates, I owe you a large measure of thanks. You all were there for serious academic discussions as we for a better group new of friends. To my co workers at T he Shimberg Center, thank you for all of your help. Thanks to Caleb and my Bugaboo, Vince, for making me laugh and helping me through the thesis p rocess. Also, to Mark for his sweet ring tone. And a huge thanks to Bill and Liz, for allowing me my opportunity at T he Center. You have both pushed me to grow as an academic and GIS professional. I have to extend a monumental thanks to my family, Krist ie, Jane, & Gary Gibbons. through the trial and tribulations of life It has been your love that has sustained me in good times cause of my fa mily. And finally to my love, Nicole. Without your support, care, needed distractions, perfectly timed jokes, sounding board discussions, back rubs, encouragement, and love I would have never made it through this grueling process. Thank you.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATI ONS ................................ ................................ ............................. 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 Trying Econo mic Times ................................ ................................ .......................... 12 The Foreclosure Issue ................................ ................................ ............................ 12 Research Questions ................................ ................................ ............................... 13 Organization ................................ ................................ ................................ ........... 14 2 LITERATURE REVIEW ................................ ................................ .......................... 15 Foreclosure Background ................................ ................................ ......................... 15 Subprime Mortgages ................................ ................................ ........................ 15 Foreclosure Basics ................................ ................................ ........................... 17 Housing and Ne ighborhood Quality ................................ ................................ ........ 19 What Makes a Neighborhood? ................................ ................................ ......... 19 Hedonic Regressions ................................ ................................ ....................... 20 Linking Price with Quality ................................ ................................ ................. 21 Broken Windows ................................ ................................ .............................. 24 Code Violations ................................ ................................ ................................ 27 Impact on Price ................................ ................................ ................................ 27 3 METHODOLOGY ................................ ................................ ................................ ... 30 Methodology Overview ................................ ................................ ........................... 30 Datasets ................................ ................................ ................................ .................. 30 Foreclosure Data ................................ ................................ .............................. 30 Code Violation Data ................................ ................................ ......................... 32 Dataset Mast er List ................................ ................................ .......................... 33 Regression Analyses ................................ ................................ .............................. 33 Model Basics ................................ ................................ ................................ .... 33 Data Preparation ................................ ................................ .............................. 37 Ordinary Least Squares Regression ................................ ................................ ....... 38

PAGE 6

6 Theory ................................ ................................ ................................ .............. 38 Method ................................ ................................ ................................ ............. 39 Geographically Weighted Regression ................................ ................................ ..... 39 Theory ................................ ................................ ................................ .............. 40 Method ................................ ................................ ................................ ............. 42 Cluster Analysis ................................ ................................ ................................ ...... 43 Data Preparation ................................ ................................ .............................. 43 Theory ................................ ................................ ................................ .............. 44 Method ................................ ................................ ................................ ............. 45 Non Parametric Statistics ................................ ................................ ....................... 45 Theory ................................ ................................ ................................ .............. 45 Method ................................ ................................ ................................ ............. 46 4 FINDINGS AND RESULTS ................................ ................................ ..................... 47 OLS Results ................................ ................................ ................................ ............ 47 Model Reliability ................................ ................................ ............................... 47 Residuals ................................ ................................ ................................ .......... 47 Coefficients ................................ ................................ ................................ ....... 50 GWR Results ................................ ................................ ................................ .......... 51 Model Improvements ................................ ................................ ........................ 5 1 Coefficients in Space ................................ ................................ ........................ 52 Foreclosure & Code Violation Clusters ................................ ................................ ... 53 Old West Tampa ................................ ................................ .............................. 53 Westchase ................................ ................................ ................................ ........ 54 5 DISCUSSION ................................ ................................ ................................ ......... 64 Plunging Home Values ................................ ................................ ........................... 64 A Broken Link? ................................ ................................ ................................ ....... 65 Realized Effects on Neighborhood Quality ................................ ............................. 67 6 CONCLUSIONS AND RECOMMENDATIONS ................................ ....................... 69 Conclusions ................................ ................................ ................................ ............ 69 Planning Policy ................................ ................................ ................................ ....... 70 Study Limitations ................................ ................................ ................................ .... 70 Further Resear ch ................................ ................................ ................................ .... 71 APPENDIX: DOR LAND USE CODES ................................ ................................ .......... 72 LIST OF REFERENCES ................................ ................................ ............................... 76 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 80

PAGE 7

7 LIST OF TABLES Table page 3 1 R egression variable specifications ................................ ................................ ..... 34 4 1 Adapted OLS regression coefficient table ................................ .......................... 48 4 2 Adapted OLS regression diagnostic table ................................ .......................... 49 4 3 Adapted GWR supplementary results table ................................ ........................ 52

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Subprime share of the mortgage marke t with home ownership rate ................... 16 2 2 U.S. properties in foreclosure 2007 2010. ................................ .......................... 18 2 3 Foreclosure vacancy trajectories by income ................................ ...................... 24 2 4 Model of foreclosure impac t proximity on property values ................................ .. 28 3 1 Regression Conceptual Model ................................ ................................ ............ 36 3 2 Representation of ordinary least squares model ................................ ................ 39 3 3 Model of a spatial kernel. ................................ ................................ .................... 41 3 4 Adaptive spatial kernels in GWR. ................................ ................................ ....... 41 4 1 2010 block group geographies, Hillsborough County, FL ................................ .. 55 4 2 Hillsborough County 2009 ACS median income and 2006 2009 home foreclosure rate. ................................ ................................ ................................ .. 56 4 3 Hill sborough County OLS residuals ................................ ................................ 57 4 4 Hillsborough County GWR local R 2 and GWR residuals. ................................ 58 4 5 Hillsborough County GWR standardized foreclosure coefficient surface and vacancy coefficient surface. ................................ ................................ ............... 59 4 6 Old West Tampa Home foreclosure and code violations. ................................ 60 4 7 Old West Tampa Code violations Getis Ord Gi* Z scores and foreclosure Getis Ord Gi* Z scores. ................................ ................................ ...................... 61 4 8 Westchase Home foreclosure and code violations. ................................ ......... 62 4 9 Westchase Code violations Getis Ord Gi* Z scores and for eclosure Getis Ord Gi* Z sc ores ................................ ................................ ................................ 63 5 1 Hillsborough County Per centage of homes vacant in 201 ................................ 66

PAGE 9

9 LIST OF ABBREVIATION S ACS American Community Survey ARM Adjustable Rate Mortgage ESRI Environmental Systems Research Institute FGDL Florida Geographic Data Library GIS Geographic Information System(s ) GWR Geographically Weighted Regression HOA Home Owners Association OLS Ordinary Least Squares REO Real Estate Owned SPSS Statistical Package for the Social Sciences

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts in Urban and Regional Planning SUB PRIME TO SUBOPTIMAL: REALIZED EFFECTS OF THE FORECLOSURE CRISIS ON NEIGHBORHO OD QUALITY IN HILL S BOROUGH COUNTY, FLOR I DA By Charles Myron Gibbons December 2011 Chair: Andres Blanco Cochair: Paul Zwick Major: Urban and Regional Planning With its roots in risky subprime lending practices, the foreclosure crisis of the late 2000s gutted neighbor hoods across the United States. As foreclosed properties sit vacant they can begin to negatively affect the actual and perceived quality of a neighborhood. The State of Florida sits at the epicenter of this crisis, with many communities incurring thousands of home foreclosures This study looks to understand the local impacts of foreclosures on neighborhood quality in Hillsborough County, F lorida In u tilizing home price as a proxy for ne ighborhood quality, small geography data allowed for analyses at a fine level of detail. Using both Ordinary Least Squares and Geographically Weighted R egression techniques, the impacts of foreclosure and vacancy on home value were modeled with spatial context. To measure physically tangible effects, the employment of cluster analysis allowed clusters of foreclosures and code violations to be compared. The results of this study show that home foreclosures have had the greatest and most disproportionate effect on home values in high income neighborhoods. However,

PAGE 11

11 these impacts have n ot yet translated to noticeable physical eff ects in wealthy neighborhoods. Low income neighborhoods have seen a lesser impact on home values, but seemed to have reached a tipping point of negative impacts; here physical deterioration is found to cluster n ear foreclosures. The results of this study highlight the need for planning practitioners and local officials to focus their community stabilization efforts on these low income neighborhoods racked by foreclosure.

PAGE 12

12 CHAPTER 1 INTRODUCTION Trying Economic Times The economic crisis of the late 2000s has become a defining event for many American citizens. T his so called Great Recession has been describ 2010). Starting in the United States in late 2007, the decline has hampered growth across most sectors of the economy. As of mid 2011, the national unemployment rate remai ns above 9% while financial markets continue to be wildly volatile (BLS, 2011). One of the most telling economic indicators is continued stagnation in the United States housing market E ven as housing prices have fallen, a lack of available financing options has kept American houses from being sold. With its economy based heavily on real estate and growth, the State of Florida has been particularly hard hit by economic adversity since the burst of the housing bubble in 2007 and 2008 The dangerous co mbination of predatory lending, lax planning policy, intense land speculation, and rampant development created a housing catastrophe waiting to happen. Only in hindsight are policy makers beginning to see the error of their ways; the scars on the economic landscape apparent. The Foreclosure Issue A central impetus of the global economic downturn is rooted in the risky lending practices associated with the US housing bubble. Credit worthiness became an unnecessary factor when applying for a home loan, as sub prime and adjustable rate construction and housing market fueled many lenders to offer these dangerous loans to homeowners who could not carry the financial burden. By 2005, approximately six out

PAGE 13

13 of every 100 households in metropolitan areas of Florida were financed via subprime loans, nearly twice the national average (Mayer & Pence, 2008). When the housing bubble burst many Floridian homeowners were unprepared. Wide spread job loss and rising payment rates on ARMs forced thousands of homeowners to default on their mortgage; most of these homes would eventually go in to foreclosure. Fo reclosure rates have continued to remain at record high s since the crisis began in 2 008. Major metropolitan areas in Florida have since been decimated by foreclosures. Often cited are the negative effects of foreclosure on the overall housing market. Less understood is the impact of these foreclosures on the neighborhoods where they ta ke place. These impacts on neighborhoods go beyond effects on housing sales. F oreclosure s and resulting housing vacancies also have the potential to change the perception of a neighborhood and ultimately its quality. As city planners look to serve the pu blic good, understanding the nature the relationship between foreclosure and neighborhood quality will be paramount in their success. Research Question s This research looks to explore and quantify the effects of foreclosure on neighborhood quality. Using a variety of Geographic Information Systems (GIS) techniques, this work will examine the foreclosure crisis in the context of Hillsborough County, Florida, a major county in the greater Tampa Bay area. This research will look to answer the following ques tions: Has the foreclosure crisis negatively affected neighborhood quality in Hillsborough County, FL? If so, t o what extent? Are low income neighborhoods being disproportionately affected? Are local clusters of foreclosures compounding negative neighborh ood effects?

PAGE 14

14 Organization This work will be divided into the following six chapters. Chapter 2 gives additional background detail as well as highlighting relevant literature. This includes an overview of foreclosure, the link between neighborhood quality and housing value, and the external effects of foreclosure on neighborhood quality. Chapter 3 explains the theory behind the spatial processes and analysis used. It also contains a detailed methodology used to conduct the research. Chapter 4 presents t he results of the GIS analyse s. Chapter 5 is a discussion of the results given the current knowledge and research reviewed in the literature review. Chapter 6 provides a conclusion with recommendations for further research.

PAGE 15

15 CHAPTER 2 LITERATURE REVIEW Foreclosure Background Subprime Mortgages At the root of foreclosure crisis is the subprime mortgage. Whereas a prime mortgage is a high quality loan offered to borrowers based on their ability to pay back the than had previously missed a mortgage payment, filed for bankruptcy, or had enough delinquen t debts to warrant a low credit score. These mortgages typically feature higher interest rates that may be variable; subprime mortgages have been known to have low initial interest rates that are raised after their original teaser periods. These Adjustab le Rate Mortgages (ARM ) have become known as some of the most notorious subprime loans (Foote et al., 2008) A global excess in capital created an influx of financial investments in the early 2000s, allowing banks to offer mortgage loans in a greater quanti ty than the 1990s. (Taylor, 2009) Not only was there an availability of credit, but an increasing percentage of the associated loans were considered to be high risk or subprime, as seen in Figure 2 1. Many of these sub prime loans were also sold to homeow ners who bought other homes under speculation. Under normal circumstances, most banks would be unwilling to take on this kind of risk. During the housing boom, these subp rime loans were sold in securities to investors, along with more stable loans that ma de them more palatable; this in turn allowed lenders to issue even more risky loans. The availability of credit allowed lifetime renters to become homeowners, typically beyond their means. Many

PAGE 16

16 Floridians utilized these loans in two ways unique to the stat e: by using the equity in their houses to Figure 2 1 Subprime share of the m ortgage m arket with home o wnership r ate (Source: U.S. Census Bureau) to properties. The vast majority of these were risky loans, with as many as 40% of currently foreclosed properties putting no money down on their homes. (Foote et al., 2008) When interest rates on existing sub prime mortgages exploded with a simultaneous plummet of housing prices in 2007 investors began to realize the toxicity of their assets. By 2008 these toxic debt assets in the United States had completely eroded away the stability of the global financial markets, beginning with the failure of Washington Mutual and Lehman Brothers banks. The government sponsored mortgage corporations Federal National Mortgage Association (Fannie Mae) and the Federal

PAGE 17

17 Home Loan Mortgage Cor poration (Freddie Mac) were only saved by a controversial government bailout, preventing a total global economic meltdown. (Glaeser et al., 2010). It was during this time that borrowers began to default on their payments; the wave of foreclosure was soon to follow. Foreclosure Basics When a property enters foreclosure, Merriam mortgaged es (2010) The process begin s with a payment default, wherein the borrower misses one or more scheduled payment s At this point the lender seeks to rectify the default through loss mitigation methods, but if payment arrangement s are not made within a 90 day period a foreclosure suit is filed and the case winds its way through the Court system; this continues until a Final Judgment of Foreclosure is entered and the Court schedules a public auction of the property In most instances, the property is purchased by the foreclosing lender who bids a credit bid for the property at the foreclosure sale If the property goes unsold to a third party it becomes Real property has either been sold or becomes bank owned its occupants are legal ly evicted. (Fogler, 2010) Since the sharp decline in housing prices in 2007, foreclosures have dramatically increased as indic ated in Figure 2 2. The Flori d a real estate market has experienced the foreclosure crisis in a unique way; the availability of land has fueled an ever growing housing market that felt the crash particularly hard. Borrowers with suspect credit used these available subprime loans to either buy houses which were out of their means or refinance their current home to gain liquid cas h. There was excessive investing in real

PAGE 18

1 8 estate as well during this time of rising prices, which in hindsight, were too good to be true. The foreclosure rates in Central and South Florida have been astounding, with nearly one out of every 41 homes in som e phase of the foreclosure process versus one out of every 570 nationwide (Olorunnipa, 2010) With this staggering rate, many Figure 2 2 U.S. p roperties in f oreclosure 2007 2010. ( Source: U.S. Foreclosure Market Report) neighborhoods have become inund ated with pockets of foreclosed properties that lie vacant. Currently the lending market is much more stringent in terms of available credit, stricter than before the housing boom. The Federal Housing Administration has created a new set of guidelines th at make it even more difficult for potential borrowers with bad credit to obtain a home loan. Increases in both minimum credit scores and monthly premiums will continue to keep credi t tight and houses on the market (Browning, 2010). Without access to cr edit, potential homeowners have no way to purchase these foreclosed properties which remain vacant.

PAGE 19

19 Housing and Neighborhood Quality What Makes a Neighborhood? Before accessing the quality of a neighborhood, the very concept of a neighborhood itself must be defined. Typically individuals intuitively understand what a neighborhood feels like and they may very well be able to identify the neighborhood they are in Beyond basic intuition, the concept of a neighborhood becomes murkier and more difficult to articulate. Many researchers have, in the past, focused on the idea admini strative control, political, and social functions (Hunter, 1979). This purely functional basis forced the examination of a neighborhood primarily on its actions. Besides being difficult measure, these categorical delineations could often come into confli ct; such as when the economic identity of a neighborhood is in direct violation of its social identity e.g., an industrial area that is not viewed as such in its social context (Martin, 2003). These nebulous representations of neighborhoods have given way to more measureable means. Planning and neighborhood expert George Galster defines a neighborhood in 01, p.2112) This definition covers an assortment of complex attributes ranging from structural, infrastructural, demographic, class, public service, environmental, proximal, political, social and sentimental characteristics that together, help to create a quantifiable portrait of a ne ighborhood. The key variable which unifies the relationship between variables is their proximity in space (Galster, 2001). In terms of being quantifiable, this definition excels at being able to put neighborhood characterist ics into discrete units. Given their

PAGE 20

20 spatial orientation, attributes within the Galster model of a neighborhood naturally lend themselves to analysis with a geospatial reference i.e., a Geographic Information System (GIS). Hedonic Regressions The concept of hedonic regression is usually associated with valuing housing and real estate markets. Its historic genesis actually comes from the work of Andrew Court in the 1930s where these models were primarily used to develop price indices for early automobile s (Goodman, 1998). Since then, the central theory behind hedonic deconstruct the use some form of ordinary least squares re gression analysis to examine how each individual piece uniquely c ontributes p. 1201). In its simplest form, a hedonic regression model takes the form a simple linear regression. ) (2 1) Where Y is the summed value of the asset in question (e.g. home value) X is a value of an attribute affecting the value of y (e.g. structural or neighborhood attributes) These values are calculated for each independent variable, which when summed, affect the regression output (Franklin & Waddell, 2002). It is this division of an asset into separate utilities that makes housing prices so well suited to the hedonic regression methodology.

PAGE 21

21 Using this theory the value of a home can be modeled by using descriptive characteristics that explain the variability in home prices. Common categories of depe ndent variables used to describe housing markets include: S TRUCTURAL F EATURES Physical makeup of the property including property size, total living area, number of rooms, structural type, building age, existence of garage, etc. N EIGHBORHOOD C HARACTERISTICS Quality of the surrounding neighborhood measured by school quality, neighborhood demographics, income levels, crime levels, etc. S PATIAL L OCATION Location within a housing market. Can also includes proximity measures to other features O T HER C ONDITIONS Miscellaneous conditions affecting value including unfinished housing, foreclosure/short sale, etc. Proper model specification is achieved by choosing the variables which have strong theoretical links with housing quality (Sopranzetti, 2010 ). The ideal outcome of a hedonic regression model is to have a favorable coefficient of determination, or R 2 value. This measure of goodness of fit is one of the primary ways to account for how well the independent, explanatory variables explain differe nces in the dependent variable, in this case housing value (Amemiya, 1980). Further evaluation of hedonic models quickly becomes much more complex and unnecessary to explain at this juncture. Linking Price with Quality While they are not a perfect indi cation of neighborhood quality, housing prices can be a useful readily available proxy for researchers to use in a variety of analyses. Following in the footsteps of Ellen & Turner, Ding and Knaap (2002) argue for and the people living within them ; that

PAGE 22

22 attainment, criminal involvement, teen sexual activity, and employment, though the mechanisms of causation remain difficult to p. 3 ) By the same token, housing price and quality are equally affected by these neighborhood characteristics. When all other things are held constant, housing price should be indicative of relative neighborhood quality alone. Goodman (1977) gives an example of comparing two houses within a geography whose physical assets are identical. If the descriptive social and economic characteristics of the geography are the same, the difference between housing units must not be accounted for s omewhere in the variable. However, if we find that there are socioeconomic differences at a smaller geography, like that of a traditional neighborhood, then the price difference can be attributed to the quality of the neighborhood (Goodman, 1977, p. 486). Ding and Knaap find that high quality neighborhoods are difficult to specifically exhibit high or stable property values, low outmigration rates, high household incomes, ra cial cohesion, and high quality public services 2002 p. 2 ) This is not to say that low income neighborhoods are necessarily decidedly correlated with high neighborhood quality rates. The argument is not made f or the dogmatic approach of taking housing prices as a direct index of neighborhood quality; the opposite is true, as these values are better suited to identity patterns and relative relationships (Ding and Knaap, 2002). Home values are therefore seen to b e a reasonable proxy by which one can estimate neighborhood quality. This can be especially useful for planners as low home values can indicate area of critical need and concern.

PAGE 23

23 The selection of variables in a hedonic regression model of housing prices c an be an important factor when tailoring a model. Choosing model variables more heavily focused towards the neighborhood characteristics will thereby increase the sensitivity to neighborhood factors (Can, 1990) Some physical characteristics of housing mus t be included, but a reliance on socioeconomic factors (e.g. percentage minority, female headed households with children and income), while shying away from the physical (number of bathrooms, size of garage, etc.) will skew a model towards a neighborhood q uality bias. This is not an inherent flaw in the model, as long as these assumptions are being explicitly noted. The Vacancy Link When a property completes the foreclosure process, its occupants are forced to leave w hen the title of the property is transferred. When a macro or micro market force causes the property to remain vacant, then a property becomes a risk to the overall quality of a neighborhood. In the wake of the foreclosure crisis, the lack of available credit has continued to keep potential homebuyers at all income levels from being able to purchase a new home (Browning, 2010 ). When a property stays vacant for an extended period of time it can become abandoned and even blighted; this can have a profound effect on the perception of a neighborhood (Immergluck & Smith, 2006). Compounding the large scale market issues associated with vacancy is the added strain of income level. Foreclosures which take place in high income areas are much less likely to remain vacant than those in low income areas. As seen in F igure 2 3, low in come neighborhoods are simply less desirable places for buyers to invest in a

PAGE 24

24 home; thereby, fewer homes are sold in the initial phases following a foreclosure. Many will likely sit vacant for long periods, once again inducing blight (Immergluck &Smith, 2 005). As vacancy rates climb a neighborhood runs the risk of falling into the broken windows trap. Figure 2 3 Foreclosure vacancy trajectories by i ncome ( Source: Adapted from Immergluck & Smith, 2005 ) Broken Windows The idea of the Broken Windows Theo ry was initially codified as a theoretical criminology, Wilson and Kelling understand the universality in the link between the people and community. They introduce the conce pt with the example of an actual broken window: Social psychologists and police officers tend to agree that if a window in a building is broken and is left unrepaired, all the rest of the windows will soon

PAGE 25

25 be broken. This is as true in nice neighborhoods as in rundown ones. Window breaking does not necessarily occur on a large scale because some areas are inhabited by determined window breakers whereas others are populated by window lovers; rather, one unrepaired broken window is a signal that no one cares and so breaking more windows costs nothing. (It has always been fun.) (Wilson & Kelling, 1982, p. 2). When a property incurs some sort of physical deterioration (be it damage, dilapidation, abandonment, etc.) it is crucial that the problem be quickly rem edied; the Broken Windows cycle is said to affect both low income high crime, as well as high income low crime neighborhoods. The sociologic/criminological theory implies that the perceived safety and order of a neighborhood are crucially important to its actual quality (Wilson & Kelling, 1982). When neighbors, citizens, and passersby perceive a neighborhood to be safe and orderly their actions and demeanor are likely to reflect this; the same can be said for the negative relationship in poorly perceived neighborhoods. Relatively minor faults, a broken window; a wall with graffiti; shingles peeling off a house, can eventually snowball into full disillusioned with the state of a neighborhood, they begin to become less active in its policing. In high quality neighborhoods, citizens have shown a sense of social responsibility and investment in their neighborhoods. The opposite is true in low quali ty, high crime neighborhoods; neighbors show a fend for themselves attitude towards crime and disorder (Wilson and Kelling, 1982). Further research has shown the link between the physical disorder in a neighborhood and the quality of life of its residents (Chappell, Monk Turner, & Payne, 2011). The vacancy associated with foreclosure leaves those properties open to the ed

PAGE 26

26 properties, homeowners saw vacant and abandoned buildi Shilling (2009, p.110) draws the direct correlations between vacancy and broken windows: Direct increases in crime adjacent to vacant/abandoned properties Nearly $73 million in damage per year due to fires in abandoned properties High cost of demolition/cleaning leaves properties as they are Adjacent properties lose value Adjacent homeowners risk raises in insurance premiums/loss of coverage, furt her exacerbating negative effects These neighborhoods with vacancies will often see these negative impacts after a certain critical mass of vacant and abando ned properties are reached. Each neighborhood has a sort of tipping point which will bring about w idespread crime and neighborhood change, with rega rds to racial makeup. After a certain percentage of a neighborhood becomes African American, a tipping point is reached and t he concept of white flight takes hold. This tipping point concept is further extended to vacancy and decline should greatly increase (Galster, 2000). The longer a h ome si ts vacant, the further said property will continue to degrade towards a point of disorder ; at this point municipalities are then forced to issue code violation citations (Accordino & Johnson, 2002)

PAGE 27

27 Code Violations Local governments use the issuance of code violations to maintain the statues held in their housing and building codes. The Hillsborough County Code Enforcement responsible for the enforcement of the Minimum Housing Code, which defines standards for the maintenance of existing properties and structures, portions of the Land Development Code These issues are reported by either a code enforcement office on patrol duty or by the assessment of a property after it is reported by another party. Code enforcement hotlines exist for this in purpose in many jurisdictions (Hillsborough County, 2011). In this way a municipality can inform a property owner of their need for compliance and levy any penalties and fines against him/her. Typical code violations associated with vacant and abandoned properties include: overgrowth of vegetation, graffiti, vandalism, improper pool maintenance, collection of trash/refuse, noticeable building degradation; these are typically commonplace amongst all violations. With that in mind, researche r s have used the code violation as an indicator for neighborhood quality/stability The overwhelming number of recent foreclosures has forced larger cities to act accordingly, insofar as creating vacant foreclos ure registration systems to monitor code the enforcement issues associated foreclosed properties (Shilling, 2009). Impact on Price Given all the previously discussed negative impacts associated with foreclosure activity, the direct parallel betw een foreclosure and home value decreases has been well documented. Foreclosed properties, which convert to REOs, are generally offered

PAGE 28

28 2005). When properties in a nei ghborhood become foreclosed and vacant, market forces alone will drive down nearby prices. As vacancies increase, losses in home value s diminish at the highest rates as seen in the housing market models of William Wheaton (1990). Beyond a given rate (uniq ue for each neighborhood) further vacancy has less impact on housing values as most of the vacancy impacts are already absorbed into the market. Figure 2 4 Model of f oreclosure i mpact p roximity on p roperty v alues ( Source: Adapted from Immergluck & Smith, 2006 ) Immergluck and Smith have led the majority of resea rch studies in the past decade. T heir recent (2006) piece highlights the early effects of the foreclosure crisis on housing price. Both direct (i.e., lower valued home s for sale) and indire ct (i.e., lowering of neighborhood quality) effects of foreclosure tend to be related to the proximity of the event. The closer a property is to a foreclosure, the greater the effect o n home value. As evidenced in F igure 2 4, buffer rings around a proper ty can

PAGE 29

29 approximate a zone of influence from a foreclosure activity. Significant impacts are seen within one quarter of a mile in Z one B, but the greatest impacts occur within one eighth of a mile in zone A. Using small scale hedonic regression, th e Immer gluck study suggests in an average Census Tract, that each foreclosure within Zone A would decrease property values at a rate of 1.136% per year (Immergluck & Smith, 2006, p. 69). In low median income tracts, the Immergluck & Smith (2006) model showed ev en greater impacts on disadvantages households with a 1.80% per year loss in Zone A. This has been attributed to the negative neighborhood externalities which are further depressing low income neighborhoods. The current debate amongst academics is to whe ther or not these negative impacts on home value and neighborhood quality are worsening in both low and high income neighborhoods. In the past, most high income research will look to shed light on the ways in which the current foreclosure crisis is impacting both high and low income neighborhoods.

PAGE 30

30 CHAPTER 3 METHODOLOGY Methodology Overview To answer the questions posed in the introduction of this work the methodology uses a variety of geospatial techniques to investigate the impacts of foreclosure in Hillsborou gh County, Florida There are two general types of analysis used to answer these questions: One is the use of regression techniques to investigate the relatio nship between foreclosure and home value/neighborhood quality at the block group level; both an Ordinary Least Squares (OLS) Regression and Geographically Weighted Regressions (GWR) are used. Two is the use of the Getis Ord Gi* cluster analysis at a neigh borhood level to investigate the relationship between foreclosure and code violations; non parametric statistical techniques will be used to evaluate significant unless ot herwise noted. The findings of this study demonstrate how the Foreclosure Crisis of the late 2000s has significantly impacted neighborhood quality in communities in Florida. By investigating neighborhoods of both high and low socioeconomic statuses, this work is designed to help planners and elected officials make the best use of limited resources to combat foreclosures and stabilize at risk neighborhoods. Datasets Foreclosure Data An initial challenge in undertaking research of this nature was locating accurate and useable foreclosure data. In coordination with the Shimberg Center for Housing Studies at the University of Florida the foreclosure data was acquired from ForeclosuresDa ily.com. Although this data is primarily used by real estate

PAGE 31

31 professionals, care was taken by the research staff at the Shimberg Center to validate the data against court records. Additional small sample sources of foreclosure data were acquired at high cost and compared against the ForeclosuresDaily data. It was found that approximately 70 80% of the actual foreclosures taking place are represented in the ForeclosuresDaily data; it should be noted that there was no discernable selective bias by which fo reclosed properties were omitted. Given the monetary and opportunity costs associated with acquiring more accurate data, the ForeclosuresDaily data was deemed to be accurate enough for research purposes (personal communication, August 4, 2011) The fore change in the certificate of title. It was decided that the years 2006 through 2009 would be aggregat ed to create a dataset encompassing the cumulative foreclosures since the onset of the foreclosure crisis. These closed foreclosures are then joined to a GIS shapefile of all parcels in Hillsborough County based on a common Parcel Tax Folio number. Hills borough County was chosen as the study area primarily on the basis of join effectiveness; over 95% of the foreclosures joined to the property parcel shapefile without any issue. In a final quality control step, foreclosure parcels are filtered to only incl ude those parcels which could be considered an owned home. For this study, an owned home falls into either Single Family Residential or Condominium land use categories. These were the two primary land uses in which an average property owner could take ou t a home mortgage based on equity. Condominiums were included under the assumption

PAGE 32

32 that foreclosures and vacancies affect the intake of C ondo Association fees, which then impacts the upkeep of the property; thus the Broken Windows theory still applies. T o achieve this end, the previously joined Closed Foreclosure shapefile has a Select by Attribute function performed so that only Single Family Homes (Land Use Code 001) and Condominia (Land Use Code 004). See Appendix A for a full listing of Land Use Cod es. Using the Feature to Point tool, these selected parcels are then converted to a point shapefile based on the parcel centroids resulting in a 2006 2009 Foreclosure Point shapefile. Code Violation Data The other unique data source used in this research was code violation data for Hillsborough County. Code enforcement in Hillsborough County is undertaken by individual municipalities, including City of Tampa, City of Plant City, Temple Terrace, and Unincorporated Hillsborough County. Plant City and Temp le Terrace were unable to provide usable data for this project; fortunately, their relative sizes within the county are small. A Code Violation dataset was created by combining the violations for the City of Tampa and Unincorporated Hillsborough County for the year 2009. This year was chosen to reflect code violations (and thereby negative neighborhood effects) in a time period following extended foreclosure activity. These code violations were then joined to a GIS shapefile of all parcels in Hillsborough County based on a common Parcel Tax Folio number. Due to variations in the reporting of violations, the type of violation was not considered in any analysis. Each incident is valued equally as a negative influence on the neighborhood.

PAGE 33

33 Dataset Master Lis t The following is a list of all of the datasets used in the analyses to follow. It should be noted that some datasets comprise many variables, while others have a single variable. 2006 2009 Closed Foreclosure Shapefile Acquired from ForeclosuesDaily.co m 2009 City of Tampa & Hillsborough County Code Violations Shapefile Acquired respectively 2010 Hillsborough Parcels Shapefile Acquired from the Florida Geographic Database Libr ary (FGDL) 2010 SF1 Hillsborough County Census Shapefile Acquired from the US. Census Bureau Regression Analyses Model Basics To evaluate the impact of foreclosure on neighborhood quality, a hedonic style regression model was chosen. While based on a re al estate focused hedonic regression, the model chosen focuses less on the physical particulars of each property and more on the surrounding neighborhood characteristics. In this way the model attempts to predict housing price, primarily with neighborhood quality ind icators. Since this paper has established the theoretical relationship between housing price and neighborhood quality, these regressions are created to further understand the impact that foreclosure has on neighborhoods. The following T able 3 1 lists the variable names, types, expected relationship (to increase in Just Value), data sources, and geographies chosen for regression analysis. The causal relationships between these variables can be seen in F igure 3 1, a conceptual model for the r egression analyses.

PAGE 34

34 Table 3 1. Regression v ariable s pecifications Variable Time Period Variable Type Expected Relationship Source Original Geography Just Value of Home 2010 Dependent Parcel Data Parcel Home Foreclosure Rate 2006 2009 Explanatory Negative Foreclosures Daily Parcel Percent Vacant 2010 Explanatory Negative Census Block Group Population Density 2010 Explanatory Variable Census Block Group Percentage of Female Headed Households with Children 2010 Explanatory Negative Census Block Group Percent Minority 2010 Explanatory Negative Census Block Group Total Living Area of Home 2010 Explanatory Positive Parcel Data Parcel The Just Value of a Home represents a fair market value for a piece of property as determined by the local property appraiser. In this regression the Home foreclosure Rate is a measure of the cumulative percentage of homes within a block group which have experienced a fo reclosure between 2006 and 2009; it is expected to have a negative relationship with the de pendent variable. As previously described in the literature review, high foreclosure rates are associated with lowered home values.

PAGE 35

35 Percent Vacant is a traditional measure of vacancy based on the number of occupied housing units divided by the total nu mber of housing units; it too is expected to have a negative relationship as described in the literature review Population Density is the measure of persons per square mile. The relationship for this value can vary based on the type of neighborhood en countered; e.g. high valued homes can exist in both high density, mixed use areas as well as traditional low density subdivisions. Percentage of Female Headed Households with Children represents the rate of households which are headed by a single mother; this was best available proxy for income level based on data available This is expected to have a negative value as low income neighborhoods demand low housing values as the poor cannot afford high rents/home values. Percent minority measures the per centage of non white persons Relevant literature has consistently shown that high minority areas have low values. Total living area is the measureable living area, measured in square feet, in which a family can live. Larger living areas are consistentl y positively related to home value; potential buyers value having additional space. As an indicator area median income is strongly correlated with other measurements of neighborhood quality (Immergluck & Smith, 2006). Unfortunately, the US Census Bureau will not be releasing socioeconomic data, including median income, with the 2010 Census. It will instead be included in the 2010 release of the American Community Survey (ACS); this data was unfortunately not availabl e at the time of publication. The 2009 ACS data was also unusable for the regression analyses as it was in the incongruent 2000 Census geographies.

PAGE 36

36 Figure 3 1 Regression Conceptual Model (Source: Charles Gibbons, 2011)

PAGE 37

37 Data Preparation In order to perform regression analyses in an ArcGIS environment, all of the variables must exist within a single shapefile. In many instances, this requires the manipulation of variables so that they have matching geographies. In this study, Just Value, Home Foreclosures, and Total Living Area all exist at the parcel level; to ameliorate this problem, the parcel level data must be aggregated to the US Census Block Group geographies. Aggregation to the 2010 US Census Block Group geo graphy is achieved via the following steps: 1. U sing Select by Attributes on the Hillsborough County Parcels layer to select out only Owned Home land uses, 001 and 003 (in the same way this was achieved for Home foreclosures). 2. Using the Feature to Point t ool, these selected parcels are then converted to a point shapefile based on the parcel centroids. This step assures that large parcels, which straddle multiple clock groups, are not counted more than once. 3. Using a Spatial Join. The 2010 US Census Blo cks should be the target layer and the 2010 Homes Parcels as the join layer. Parcel data should be summarized as averages. This will give an average Just Value and Total Living Area for each block group The resulting file only requires a join with the f oreclosure data to be complete. This is achieved using another spatial join with the 2010 US Census Blocks as the target layer and the 2006 2009 Foreclosure Points as the join layer. Foreclosure points should be summarized using basic summation. The res ulting Regression Master shapefile contains all of the pertinent regression data at a Block Group geography.

PAGE 38

38 Ordinary Least Squares Regression An Ordinary Least Squares (OLS) Regression is first used to explore and define the relationships between the exp lanatory variables and home value. This initial regression is a global model which is evaluated aspatially. Theory The OLS regression takes its shape from a simple linear regression (see Equation 2 1 in Chapter 2). It uses known values to calculate an equation which best fits as a model for a dependant variable, as seen in F igure 3 2. What makes an OLS more powerful than a traditional linear model is the use of statistical checks to minimize the residuals or squared deviations (ESRI, 2009). The OLS performs the following (3 1) values. The use of statistical software packages and GIS applications has made these calculations more manageable. Care should be taken in choosing the specifications of a model. Improperly specified models will yield results which may not be trustworthy. Beyond the value of R 2 (measure of goodness of fit), the OLS tool in ArcGIS gives a report which includes a variety of measurements used to evaluate and calibrate a model The user should also residuals. Clustered residuals are indicative of a misspecified specified model which is missing one or more key explanatory variables (Zwick, 2010 ).

PAGE 39

39 Method In ArcGIS the Ordinary Least Squares (OLS) Tool is opened via ArcToolbox. The Regression Master shapefile will act as the Input Feature Class, as it contains all of the relevant spatial data. Dependent and Explan atory Variables (as defined in T able 3 1) are entered into the appropriate portion of the UI. After the model is run, it will return an OLS output shapefile as well as a report which displays the model results. Figure 3 2 Representation of ordinary least squares m odel (Source: ESR I, 2011) used. The OLS output shapefile acts as the Input Feature class; the Residuals field as the Input Field. Default settings are employed, except that Row Standardiza tion is necessary to avoid the inherent bias associated with Census Block Groups (Zwick, 2010). These geographies are controlled to have equal population which requires the use of Row Standardization. This function will also generate a report for evaluat ion. Geographically Weighted Regression A Geographically Weighted Regression (GWR) is used to further explore relationships between the explanatory variables and home value in a spatial context. This type of regression is a local model which takes spati al variation into account.

PAGE 40

40 Theory One of the criticisms of the OLS model is that it only gives one set of coefficients; it is in fact a global model which shows the relationships between variable across an entire study area. In practice, homogeneity is often not encountered across a study area. Problems in the real world are often complex and have varying relationships which change in space, a concept known as spatial non stationarit y. The development of GWR in the last 15 years has allowed resear chers to produce regression models which more accurately represent changes over space (Fotheringham, 2002). In practice, a GWR differs from an OLS by running the regression locally. The term Geographically Weighted Regression can almost be considered a m isnomer. It is not just one regression model, but actually the summary of many regressions in space. A GWR runs an OLS type model at every data point within a dataset. In the case of Hillsborough County, this amounts to 876 regressions; one for each blo ck group. This is achieved by only using nearby neighbors in the calculation of each individual regression model. To make this possible nearby datapoints are weighted based on their distance away from the regression point; this concept, refered to as a Spatial Kernel, is shown in F igure 3 3 Fotheringham (2002) describes the Spatial Kernel: For a given regression point, the weight of a data point is at a maximum when it shares the same location as the regression point. This weight decreases continuously as the distance between the two points increases. In this way, a regression model is calibrated locally simply by moving the regression point across the region (p. 44) everything is related in space, but near things are more related than others.

PAGE 41

41 Figure 3 3 Model of a spatial k ernel ( Source: Fotheringh am, 2002 ) Figure 3 4 Adaptive spatial k ernels in GWR ( Source: Fotheringham, 2002 ) Choosing the size of the kernel can have a large impact on the performance of a GWR. Use of an adaptive spatial kernel helps to account for varying distances betwe en regression points. Seen in F igure 3 4 the Adaptive Spatial Kernel changes the size of

PAGE 42

42 Without this feature, more sparsely located data points would lack the critical numb er of points to create an effective and significant model (Fotheringham, 2002). Given its inherent spatial constructs, the GWR has quickly become a popular tool amongst researchers. Like any model, it can only act as a representation of the real stationarity into account makes it more robust than an OLS when trying to understand how models vary in space. Method In ArcGIS the Geographically Weighted Regression (GWR) Tool is opened via ArcToolbox. The Regression Master shapefile will act as the Input Feature Class, as it again contains all of the relevant spatial data. Dependent an d Explanatory Variables (as defined in T able 3 1) are entered into the appropriate portion of the UI. Bandwidth Method should be left at the default, AICc. Kernel type should be set to ADAPTIVE; this will account for the variable size and locations of th e block groups. After the model is run, it will return a GWR output shapefile as well as a dbf file which displays the model results. The same method will be used to evaluate clustering of residuals as in the OLS. The Spatial Autocorrelation (Global Mora acts as the Input Feature class; the Residuals field as the Input Field. Default settings are employed, except that Row Standardization should be used. calculated manually using the following formula (Bring, 1994): (3 2)

PAGE 43

43 deviation to that of uses the changes in standard deviations to show the relative importance of the independent variables. Standardized betas should be calculated for each coefficient directly in the ArcGIS field calculator. Cluster Analysis The regression models used in the primary analysis are intended to explore the relationship between foreclosure rates and the just value of homes. Given the link between home value and neighborhood quality, the regre ssions model the negative effects of foreclosure on neighborhood quality by proxy only. To measure the physical effects on neighborhoods directly, a secondary cluster analysis will be employed at the neighborhood level. Using Getis Ord Gi* Hotspot analys is, clusters of foreclosure incidents will be compared against clusters of land development code violations. The object is to determine if foreclosure clusters are indicative of code violation clusters. This degree of clustering will then be evaluated usi ng traditional non parametric statistical methods. Using the results from the GWR analysis, two neighborhoods were chosen for the cluster analysis; one low income and one high income (based on 2009 ACS income data) neighborhood were selected. This will a on neighborhoods of different income levels. Data Preparation The process of selecting neighborhoods is undertaken using the coefficient surfaces created in the GWR. To find areas in which foreclosures had t he greatest effect on home value, block groups with extremely negative Home Foreclosure Rate

PAGE 44

44 coefficient values are deemed favorable candidates for selection. From this subset, one low income and one high income neighborhood were chosen with a high forecl osure and vacancy rates. This effectively selects neighborhoods which have high foreclosure and vacancy rates, where foreclosures have an large impact on housing value Based on this rationale, the following criteria were used in determining the neighbo rhoods of choice: Bottom 50% of negative Home Foreclosure Rate Coefficients Home Foreclosure rate greater than 2.5% Vacancy rate greater than 14% The two neighborhoods which best fit these specifications were Old West Tampa (low income) and the southeast quadrant of Westchase (high income). To prepare the data for analysis the parcels within the two neighborhoods were selected out based on their des cribed locations as census designated places. The quadrant subsection of the entire Westchase community was selected to make the neighborhood sizes more even. The selected parcels are then spatially joined to the foreclosure point and code violation poin t shapefiles. The final shapefile for each neighborhood contains a count of both foreclosure activities and code violations for each parcel. The ory The Getis Ord Gi Hotspot analysis tool is a common method of determining clustered values. It evaluates the weighted values in a dataset in space to determine where like values are clustered. The result of this process is the Gi* statistic, actually a Z score. This Z score indicates where statistically significant clusters of hot/high values

PAGE 45

45 (positive Z scores) and cold/low values (negative Z scores) exist in space (ESRI, 2011). These Z scores are included in the output shapefile. Method The Getis Ord Gi* statist ic can be calculated using the Hot Spot Analysis (Getis Ord Gi*) toolset in ArcGIS. The entire process must be completed for both neighborhoods, with iterations of foreclosures and code violations as the input field. Choosing a Conceptualization of Spati al Relationships is an important step in specifying the model. For this analysis the Inverse Distance relationship was chosen; the spatial impacts from foreclosures and code violations should decrease drastically as distance increases. The distance band should be set to 1/8 of a mile. This is the distance supported by Immergluck & Smith (2006) in which a foreclosure has having the greatest impact on the surrounding neighborhood. Row Standardization should also be employed to account for the imposed aggr egation scheme associated with parcel data. Non Parametric Statistics Beyond a simple visual interpretation of the Z scores associated with foreclosure each nei ghborhood. This will effectively measure the correlation between clusters of foreclosures and clusters of code violations. Theory Because the Z Scores returned by the Hotspot analysis are in the form of ratio Moment Correlat paired values, in this case Z scores for foreclosure and code violation clustering, to

PAGE 46

46 measure the linear dependence between two variables (Burt & Barber, 1996). The 3. (3 3) The value of r can range from 1 to 1. The following guide (Quin nipiac, 2011) indicates the strength of a relationship based on the value of r: +.70 or higher Very strong positive relationship +.40 to +.69 Strong positive relationship +.30 to +.39 Moderate positive relationship +.20 to +.29 weak positive relationship +.01 to +.19 No or negligible relationship .01 to .19 No or negligible relationship .20 to .29 weak negative relationship .30 to .39 Moderate negative relationship .40 to .69 Strong negative relationship .70 or higher Very strong negative relationship For the r value to be statically significant, further significance tests must also be undertaken. Method suite SPSS. Before loading the Z Scores directly int o SPSS, the Z scores for foreclosure and code violations clusters must first be matched together Microsoft Excel based on a common parcel ID; they are then loaded into SPSS. The Bivariate hould be checked. Upon completion, a report is generated with the requisite r statistic and significance value.

PAGE 47

47 CHAPTER 4 FINDINGS AND RESULTS OLS Results Model Reliability As a global model, an Ordinary Least Squares regression looks at the strength of a model to predict an outcome across an entire dataset. As per the results in T able 4 2, in this study the OLS performed reasonably well. With an Adjusted R 2 of 0.70, on a s cale of 0 1, the chosen explanatory variables are seen to be good indicators of home value overall. With p values of nearly zero, the Joint F and Wald statistics in T able 4 2 indicate that the model is in fact significant. However, t he significant Koenke r (BP) Statistic provides an indication that this model should be evaluated in the context of a GWR, once the global coefficients are understood (ESRI, 2011). Residuals The residuals in a regression model represent that which cannot be explained by the explanatory variables chosen in the model. Careful examination of model residuals con tain residual values, it is their relationships in space which are deemed important. The Jarque Bera Statistic evaluated in T able 4 2, indicates whether or not the residuals are normally distributed. In this case, the nearly zero p value indicates non n ormal residuals and therefore, model bias. Spatial Autocorrelation tool. If high or low residuals are clustered in space then a model

PAGE 48

48 Table 4 1 Adapted OL S regression coefficient t able Variable Coefficient StdError t Statistic Probability Robust StdError Robust t Robust Prob Intercept 77239.8839 10615.6200 7.2761 0.0000 18154.1868 4.2547 0.0000 Vacancy Rate 637.7891 293.0880 2.1761 0.0298 251.9920 2.5310 0.0115 Density 2555.2891 373.6703 6.8384 0.0000 455.2410 5.6130 0.0000 Percent Minority 835.9340 115.1266 7.2610 0.0000 86.8348 9.6267 0.0000 Total Living Area 120.5626 3.5797 33.6794 0.0000 8.1648 14.7661 0.0000 Home Foreclosure Rate 386164.4201 144842.1339 2.6661 0.0078 81606.6001 4.7320 0.0000 Female Headed Households 43763.0049 28689.9333 1.5254 0.1275 22646.6851 1.9324 0.0536

PAGE 49

49 Table 4 2 Adapted OLS regression diagnostic t able Diag_Name Diag_Value Definition AIC 21609.68 Akaike's Information Criterion : A relative measure of performance used to compare models; the smaller AIC indicates the superior model. AICc 21609.85 Corrected Akaike's Information Criterion: second order correction for small sample sizes. R2 0.7011 R Squared, Coefficient of Determination: The proportion of variation in the dependent variable that is explained by the model. AdjR2 0.6991 Adjusted R Squared: R Squared adjusted for model complexity (number of variables) as it relates to the data. F Stat 339.79 Joint F Statistic Value: Used to assess overall model significance. F Prob 0 Joint F Statistic Probability (p value): The probability that none of the explanatory variables have an effect on the dependent variable. Wald 614.93 Wald Statistic: Used to assess overall robust model significance. Wald Prob 0 Wald Statistic Probability (p value): The computed probability, using robust standard errors, that n one of the explanatory variables have an effect on the dependent variable. K(BP) 99.03 Koenker's studentized Breusch Pagan Statistic: Used to test the reliability of standard error values when heteroskedasticity (non constant variance) is present. K(BP) Prob 0 Koenker (BP) Statistic Probability (p value): The probability that heteroskedasticity (non constant variance) has not made standard errors unreliable. JB 22682.67 Jarque Bera Statistic: Used to determine whether the residuals deviate from a normal distribution. JB Prob 0 Jarque Bera Probability (p value): The probability that the residuals are normally distributed. Sigma2 2995763352.11 Sigma Squared: OLS estimate of the variance of the error term.

PAGE 50

50 is seen to be misspecified; there are likely some key explanatory variables which are returned a Z score of 60.60 (2.58 being the 99 th percentile threshold), indicating clustering with an infinitesimally small chance of this value being caused by random chance. This is also visually evident in Figure 4 3; the map indicates severe clustering of both high and low values in space. This indication of misspecification can also be aided by the used of GWR to remove nonstationarity and decrease the clustering of residuals (ESRI, 2011). The variable which could have helped to reduce the clustering of residuals is m edian income. As when compared against the American Community Survey (ACS) Median Income map in F igure 4 2, the OLS residuals correlate in a visual inspection. Unfortunately, Median income data was not released in the 2010 Census. The 2009 ACS data rema ins in the 2000 Census Block geographies, therefore it was unusable in the regression. Geographies must be identical to be used in a GIS based regression model. Coefficients The OLS regression coefficients returned results that were mostly expected from a theoretical perspective. Based o n the expected coefficients in T able 3 1, most of the actual coefficients found in the model match their theoretical counterparts accordingly. To check these coefficients for significance, calculated p values are consulted. Based on the significant Koenker (BP) Statistic the Robust p values are used instead of the ll of the variables have p values below that of the 0.05, meaning they are significant at the 95 th percentile. At this level there is statistically only less than a 5% chance that the coefficients are due to random chance. Only Female Headed Households h ad a higher

PAGE 51

51 value of 0.0536, making it significant at 94.6 th percentile. This is deemed acceptable for this research, but these coefficients are viewed within the light of their borderline p values. The two coefficients of greatest interest, Home Forec losure and Vacancy Rate, returned differing values. The coefficient for Home Foreclosure Rate was highly negative as was expected; home foreclosures are indicative of lower home values. However, vacancy rates were found to be positive; this is in direct disagreement with the concept that high vacancy rates are indicative of lower home values. This unexpected result could be a response to other market forces not accounted for in the initial conceptual model. These possible influences will be discussed fu rther in Chapter 5. GWR Results Model Improvements The use of Geographically Weighed Regression (GWR) improved upon the results found in the OLS. As seen in T able 4 3, the Adjusted R 2 value was increased substantially, to 0.91, while the AICc was lowered; both of these indicators show a model framework that is better suited to answering the questions asked. The local R 2 surface in F igure 4 4 shows the majority of the study area has values above 0.85, showing good overall performance. Spatial Autocorrelation of the GWR residuals were still clustered, but to a much lesser extent as seen in F igure 4 4. The resulting Z score of 3.41 still indicates clustering, but this is much closer to an acceptable level than the OLS model.

PAGE 52

52 Table 4 3 Adapted GWR supplementary r esults t able Variable Name Value Definition Neighbors 245 Residual Squares 705027831911 Effective Number 78.47 Sigma 29732.41 AICc 20578.74 R2 0.9191 R2Adjusted 0.9112 Dependent Field 0 Home Just Value Explanatory Field 1 Home Foreclosure Rate Explanatory Field 2 Vacancy Rate Explanatory Field 3 Female Headed Households Explanatory Field 4 Density Explanatory Field 5 Total Living Area Explanatory Field 6 Percent Minority Coefficients in Space The defining feature of a GWR is the creation of coefficient surfaces; these allow researchers to look at how coefficients vary over space. As in the OLS, the two variables of greatest importance are the coefficients for Home Foreclosure Rate and Home Vac ancy Rate. These coefficients are pictured in F igure 4 5. Home Foreclosure Rate produced negative coefficients across the entire study area. The most negative coefficients appear to be in coastal areas with high incomes, when compared against 2009 ACS data. Even low income areas with these highly negative coefficients are bordering high income neighborhoods. Vacancy Rate showed unexpected variation over space. While the global model produced a positive coefficient, the GWR produced a range of values from 8.51 to

PAGE 53

53 86.86. It was expected that vacancy would produce negative coefficients, as higher vacancies would result in lower home prices. This was only true within a few areas in the county. The lowest negative values are clustered in the red shaded portions of the map in F igure 4 5. This area represents primarily low income residents, along with a high level of transient college students associated with the University of South Florida. There is also a swath of slightly negative coefficients to the south and east, including the higher income developments of Fishhawk and Lithia. Foreclosure & Code Violation Clusters Old West Tampa Old West Tampa is an older, ethnically diverse neighborhood which has a relatively low median income; approximately $ 22,000 based on overlaying 2009 ACS da ta (average for Hillsborough County was approximately $40,000). The combination of high Home Foreclosure Rates and low income put the neighborhood of Old West Tampa at a high risk for negative neighborhood effects and code violations. Seen in terms of incidents in F igure 4 6, foreclosures and code violations do tend to coincide visually. After running the Hotspot analysis, the maps in F igure 4 7 returned the following hotspot clusters, with red signifying clusters of high values and blue signifying clusters of low values. revealed a correlation value of 0.354 at a 99% confidence interval. This indicates a moderately positive correlation between the clustering of foreclosure incidents and land development code violations in the neighborhood of Old West Tampa.

PAGE 54

54 Westchase Built in the late 1990s, Westchase is a newer more affluent community with a much higher median income than Old West Tampa; approximately $85,000 based on overlaying 2009 ACS data. The neighborhood experienced a large number of completed foreclosures, but there very few code violations; this can be seen in F igure 4 8. Running the cluster analysis on this neighborhood, seen in F igure 4 9, shows the same red to blue patterns as the previously done. 0.291 at a 99% confidence interval. This is indicative of a weak negative relationship between the two variables. In short, there was barely any discernable relationship between foreclosure incid ents and land development code violations in the neighborhood of Westchase. The few code violations which did occur had seemingly no connection to the pockets of foreclosure activity in the area.

PAGE 55

55 Figure 4 1 2010 block group g eographies, Hillsborough County, FL ( Source: Charles Gibbons )

PAGE 56

56 Figure 4 2 Hillsborough County 2009 ACS median income and 2006 2009 home foreclosure r ate ( Source: Charles Gibbons )

PAGE 57

57 Figure 4 3 Hillsborough County OLS r esiduals ( Source: Charles Gibbons )

PAGE 58

58 F igure 4 4 Hillsborough County GWR l ocal R 2 and GWR r esiduals ( Source: Charles Gibbons )

PAGE 59

59 Figure 4 5 Hillsborough County GWR standardized foreclosure c oefficient surface and vacancy c oefficient s urface ( Source: Charles Gibbons )

PAGE 60

60 Figure 4 6 Old West Tampa Home f oreclosure and code v iolations ( Source: Charles Gibbons )

PAGE 61

61 Figure 4 7 Old West Tampa Code v iolati ons Getis Ord Gi* Z scores and foreclosure Getis Ord Gi* Z s cores ( Source: Charles Gibbons )

PAGE 62

62 Figure 4 8 Westchase Home f ore closure and code v iolations ( Source: Charles Gibbons )

PAGE 63

63 Figure 4 9 Westchase Code violations Getis Ord Gi* Z s cores and f oreclosure Getis Ord Gi* Z s cores ( Source: Charles Gibbons )

PAGE 64

64 CHAPTER 5 DISCUSSION This work was originally intended to examine foreclosures and factors of neighborhood change since the burst of the housing bubble. The initial plan was to do a short longitudinal/cross sectional analysis from 2006 through 2010. The lack of available dat a for such a short time period forced a change in direction, along with a change in scope. By combining in depth regression analyses with local clustering, this research design was able to triangulate on the actual effects of foreclosure on neighborhoods in a meaningful way. The following discussion highlights the results and findings of this quality. Plunging Home Values Since the onset of The Great Recession, the US housing market has taken a major downturn. While much of the loss of home value is associated with stagnations in the markets, a great deal of degradation can be attributed to nearby home foreclosures. The results of both regression analyses proved the incredible impact that the percentage of foreclosed homes can have on an area. The OLS model returned a coefficient value of 386164.4201 Based on this global model, a mere 1% increase in the number of homes being foreclosed upon would lower the average home price by nearly $3,900 within a block group. The GWR provided several key insights into the relationship between foreclosures and home values. The largest negative coefficients exist in block groups which are predominately high income; including gr eater Westchase, Town and Country, greater South Tampa, and continuing towards the wealthy retirement communities in

PAGE 65

65 Apollo Beach/Sun City Center. In these neighborhoods a one percent increase in the foreclosure rate could decrease the average home price from anywhere between $60,000 and $177,000 on the high end. While the extremely high coefficients represent losses in property value to the extremely wealthy, huge losses in value are possible in moderate and even low income housing. Areas which seem mos t susceptible to these losses are low income neighborhoods which border higher income neighborhoods. Evidence has previously shown that high home prices can often spill over into adjacent low quality/income neighborhoods (Immergluck & Smith, 2006). This effect has been seen in terms of loss of home value due to foreclosure. Low income neighborhoods which abut high income neighborhoods, like Carver City and Old West Tampa, are seeing largely negative Home Foreclosure coefficients. Other traditionally poo r neighborhoods which are situated further from high income areas are seeing much smaller negative Home Foreclosure coefficients. It seems as though higher income neighborhoods have more home value to lose from nearby foreclosures and they are losing it disproportionately. Extremely poor neighborhoods often had high rates o f foreclosure, but there was already so little actual home value to lose that the effects of foreclosure have had little impact on them. A Broken Link? The link between foreclo sure and vacancy rates has been well documented in economic and planning literature. This case was actually not different from theory in terms of the actual foreclosure and vacancy rates. Where there was a high rate of foreclosure there was often a high rate of vacancy, especially in low income neighborhoods as evidenced in F igure 5 1.

PAGE 66

66 Figure 5 1 Hillsborough County Percentage of homes v acant in 2010 ( Source: Charles Gibbons ) Strangely there seemed to be a disconnect between Vacancy Rate coefficie nts and Home Foreclosure Rate coefficients. In the OLS regression, vacancy coefficients were positive while foreclosure coefficients were highly negative. This pattern was repeated in the GWR; except in some low income areas, like those with negative vac ancy coefficients near the Un iversity of South Florida (see F igure 4 5). Aside from these locations, negative foreclosure coefficients are associated with positive vacancy coefficients. The combination of high foreclosure rates with high vacancy in low

PAGE 67

67 income neighborhoods seems to have a synergistic effect in creating negative coefficient values. Poor neighborhoods seem to be closer to a tipping point of neighborhood quality. Already physically and socially troubled, the rash of foreclosures followi ng the bubble burst pushed low income neighborhoods into disrepair; this allows the broken windows cycle to take hold. High and middle income neighborhoods have no t seen these effects for a variety of possible reasons. In many high/middle income neighbor hood s, Homeowners Associations (HOA ) exist to maintain a uniform character within a community and more vigilant enforcement of community maintenance standards. The fees associated with membership may be used towards the physical upkeep of properties 2011) If homes become foreclosed and vacant, HOAs can temporarily maintain a Problems can aris e for HOAs if enough prop erties are foreclosed upon, and budget shortfalls are created due to lack of paid fees; this represents another tipping point which many communities are on the verge of reaching (Perkins, 2010). If a neighborhood does not reach a point of deterioration, most of the loss in housing value is caused purely by market forces. Foreclosed homes go on the market at a lower rate and therefore, they have the effect of lower ing housing prices in the immediate area. The higher the quality of a neighborhood, the more resistant it is to the effects of foreclosure. Realized Effects on Neighborhood Quality The local cluster analyses confirmed the idea that low income areas have a greater chance of seeing the physical deterioration of a neighborh ood. Old West

PAGE 68

68 Tampa sits in close proximity to wealthier neighborhoods to the south; its largely negative foreclosure coefficient is likely caused by a spillover effect. The cluster analysis revealed that foreclosures clusters are correlated with code vi olation clusters. In low income neighborhoods, the influx of foreclosures has caused the broken windows cycle to manifest itself in the form of code violations. These physical symptoms are a Unlike Ol d West Tampa, the neighborhood of Westchase saw the opposite effect. The Westchase area saw a high rate of foreclosure and vacancy, but these indicators did not manifest many code violations. The few code v iolations which did exist did not seem to be rel ated to the foreclosures in the neighborhood; this was reflected in the correlation statistic. There was effectively no correlation between the foreclosure s and code violation s in this high income neighborhood. This supports the previous claims made abou t high quality neighborhoods, while also bolstering the argument that HOAs decrease physical effects of foreclosure upon neighborhoods.

PAGE 69

69 CHAPTER 6 CONCLUSION S AND RECOMMENDATION S As the United States remains in the throes of a deep economic downturn, understanding how to deal with the issue s created by foreclosure s will be a key to the recovery of neighborhoods and the housing market as a whole. Planners and community officials should work to understand the impacts that foreclosure activitie s have had on neighborhoods of all varieties. Given the limited resources of local government s every dollar should be specifically targeted to help neighborhoods which have suffered most from the foreclosure crisis. Conclusions The effects of foreclos ures on neighborhood quality have been felt across the entire nation. This study has shown that in Hillsborough County, Florida, the burden has not been equally shared. Home values have been dramatically slashed since the onset of the foreclosure crisis in 2007. Still, high income neighborhoods have seen the greatest prop ortional decline in home value. E ach foreclosure incident has a greater impact on home values in a high income neighborhood than one in low income neighborhood. These results, contrary to traditional literature, could likely be due to the rampant investor speculation taking place in the early 2000s. High income neighborhoods had highly inflated home values which were acting under abnormal market forces. These losses in home value ne ver translated into measurable physical decline within high income neighborhoods. Here, land development code violations were not very prevalent, even with high incidents of foreclosure. In lower income neighborhoods the incidents of code violations crea ted a noticeable impact on the community. Given

PAGE 70

70 the overall quality of the neighborhood, a greater number of code violations are to be expected in a lower income neighborhood. The location of these code violations gives a link of causality between them a nd the foreclosure incidents. The 1/8 th mile clusters of foreclosures and code violations produced a statistically significant correlation between the two. Planning Policy This research primarily focused on the identification of areas which are vulnera ble to the impacts of foreclosure. Ideally, local agencies should use this type of analysis to guide their actions in their jurisdictions. Programs and policies should be enacted by local government to not only curb foreclosure, but also to manage existing foreclosed properties. One such device is the use of land banking which are governmental or nongovernmental nonprofit entities that focus on the conversion of vacant, abandoned properties into productive use f Cleveland, Ohio has recently set up a land bank by which they have acquired nearly 1000 properties. Striking a deal with lenders, the land bank offers to buy off unwanted foreclosed properties as long as the lender pays for demolition (Saito, 2011). Typ ically, these extremely distressed houses cost more in terms of maintenance than their actual worth. By acting in this way, a local government owned land bank can acquire and hold a property to stabilize a neighborhood for future redevelopment. These progr ams are not only good planning for foreclosures, but they act as a show of good faith on the part of a local government entity. Study Limitations This study was primary limited by data availability. Income data at the 2010 Census Block Group geography wi ll not become available until 2012. The most

PAGE 71

71 accurate foreclosure data is only available at a high price premium; therefore the cheaper, less accurate ForeclosuresDaily.com data was used. Sales data for these years was spotty in terms of availability and quality. In using Just Value, an inherent error is associated with property appraiser data. An appraisal is just an approximation data would be a better approximation of real value, especially in a poor housing market. Crime data would have been a great indicator of neighborhood quality, but it too was unavailable at a small geography. With more time and greater resources, these regression models could be improved by using better indicators of neighborhood quality. Also, given the small study area only six explanatory variables could be used in the GWR. With a larger study area, more descriptive variables could be used to strengthen the model in space. Further Res earch At the conclusion of this study, the avenues for future research in foreclosure are abundant. As previously expressed, this research was initially intended to be a time series. Given the lack of available small geography data for the period of 2006 2011, this was not possible. To better understand the ways in which the foreclosure crisis has created neighborhood change, high quality small area data could be used. The full extent of damage from the foreclosure crisis w ill not truly be known until t he United States reaches a full economic recovery.

PAGE 72

72 APPENDIX: DOR LAND USE CODES USE CODE Residential Property 000 Vacant Residential 001 Single Family 002 Mobile Homes 003 Multi family 10 units or more 004 Condominia 005 Cooperatives 006 Retirement Homes not eligible for exemption. Others shall be given an Institutional classification 007 Miscellaneous Residential (migrant camps, boarding homes, etc.) 008 Multi family less than 10 units 009 Undefined R eserved for Use by Department of Revenue Commercial Property 010 Vacant Commercial 011 Stores, one story 012 Mixe d use store and office or store and residential or residential combination 013 Department Stores 014 Supermarkets 015 Regional Shopping Centers 016 Community Shopping Centers 017 Office buildings, non professional service buildings, one story 018 Office buildings, non professional service buildings, multi story 019 Profession al service buildings 020 Airports (private or commercial), bus terminals, marine terminals, piers, marinas. 021 Restaurants, cafeterias 022 Drive in Restaurants 023 Financial institutions (banks, saving and loan companies, mortgage companies, credit services) 024 Insurance company offices 025 Repair service shops (excluding automotive), radio and T.V. repair, refrigeration service, electric repair, laundries, laundromats 026 Service stations 027 Aut o sales, auto repair and storage, auto service shops, body and fender shops, commercial garages, farm and machinery sales and services, auto rental, marine equipment, trailers and related equipment, mobile home sales, motorcycles, construction vehicle sales 028 Parking lots (commercial or patron) mobile home parks 029 Wholesale outlets, produce houses, manufacturing outlets 030 Florist, greenhouses

PAGE 73

73 031 Drive in theaters, open stadiums 032 Enclosed theaters, enclosed auditoriums 033 Nightclubs, cocktail lounges, bars 034 Bowl ing alleys, skating rinks, pool halls, enclosed arenas 2009 NAL SDF NAP User Guide 13 035 Tourist attractions, permanent exhibits, other entertainment facilities, fairgrounds (privately owned). 036 Camps 037 Race tracks; horse, auto or dog 038 Golf courses, driving ranges 039 Hotels, motels Industrial Property 040 Vacant Industrial 041 Light manufacturing, small equipment manufacturing plants, small machine shops, instrument manufacturing printing plants 042 Heavy industrial, heavy equipment manufacturing, large machine shops, foundries, steel fabricating plants, auto or aircraft plants 04 3 Lumber yards, sawmills, planing mills 044 Packing plants, fruit and vegetable packing plants, meat packing plants 045 Canneries, f ruit and vegetable, bottlers and brewers distilleries, wineries 046 Other food processing, candy factories, bakeries, potato chip factories 047 Mineral processing, phosphate processing, cement plants, refineries, clay plants, rock and gravel plants. 048 Warehousing, distribution term inals, trucking terminals, van and storage warehousing 049 Open storage, new and used building supplies, junk yards, auto wrecking, fuel storage, equipment and material s torage Agricultural Property 050 Improved agricultural 051 Cropland soil capability Class I 052 Cropland soil capability Class II 053 Cropland soil capability Class III 054 Timberland site index 90 and above 055 Timberland site index 80 to 89 056 Timberland site index 70 to 79 057 Timberland site index 60 to 69 058 Timberland site index 50 to 59 059 Timberland not classified by site index to Pines 060 Grazing land soil capability Class I

PAGE 74

74 061 Grazing land soil capability Class II 062 Grazing land soil capability Class III 063 Grazing land soil capability Class IV 064 Grazing land soil capability Class V 065 Grazing land soil capability Class VI 066 Orchard Groves, Citrus, etc. 067 Poultry, bees, tropical fish, rabbits, etc. 068 Dairies, feed lots 069 Ornamentals, miscellaneous agricultural 2009 NAL SDF NAP User Guide 14 Institutional Property 070 Vacant 071 Churches 072 Private schools and colleges 073 Privately owned hospitals 074 Homes for the aged 075 Orphanages, other non profit or charitable services 076 Mortuaries, cemeteries, crematoriums 077 Clu bs, lodges, union halls 078 Sanitariums, convalescent and rest homes 079 Cultural organizations, facilities Government Property 080 Undefined Reserved for future use 081 Military 082 Forest, parks, recreational areas 083 Public county schools include all property of Board of Public Instruction 084 Colleges 085 Hospitals 086 Counties (other than public schools, colleges, hospitals) including non municipal government. 087 State, other than military, forests, parks, recreational areas, colleges, hospitals 088 Federal, other than military, forests, parks, recreational areas, hospitals, colleges 089 Municipal, other tha n parks, recreational areas, colleges, hospitals Miscellaneous Property 090 Leasehold interests (government owned property leased by a non governmental lessee) 091 Utility, gas and electricity, telephone and telegraph, locally assessed railroads, water and sewer service, pipelines, canals, radio/television communication 092 Mining lands, petroleum lands, or gas lands 093 Subsurface rights 094 Right of way, streets, roads, irrigation channel, ditch, etc. 095 Rivers and lakes, submerged lands 096 Sewage disposal, solid waste, borrow pits, drainage reservoirs,

PAGE 75

75 waste land, marsh, sand dunes, swamps 097 Outdoor recreational or parkland, or high water recharge subject to classified use assessment. Centrally Assessed Property 098 Centrally assessed Non Agricultural Acreage Property 099 Acreage not zoned agricultural

PAGE 76

76 LIST OF REFERENCES Accordino, J., & Johnson, G. T. (2000). Addressing the vacant and abandoned property problem. Journal of Urban Affairs, 22 (3), 301 315. doi:10.1111/0735 2166.00058 Amemiya, T. (1980). Selection of regressions. International Econ omic Review 21(2). 331 354. Retrieved from http://www.jstor.org/pss/2526185 mortgage foreclosure boom. Home Preservation Foundation. 1 25. Retrieved from http://www.hpfonline.org/content/pdf/Apgar_Duda_Study_Short_Version.pdf Bureau of Labor Statistics. (2011). Labor force statistics from the current population survey. Retrieved from http://www.bls.gov/cps/ Bring, J. (1994). How to standardize regression coefficients. The American Statistician. 48(3). 209 213 Retrieved from http://www.psych.umn.edu/faculty/waller/classes/mult11/readings/bring1994.pdf Browning, L. (2010, December 2). When borrowers default on second homes. The New York Times. Retrieved from http://www.nytimes.com/2010/12/05/realestate/mortgages/05Mort.html Burt, J.E. & Barber,G.M.(1996). Elementary statistics for geographers (2 nd ed.) Gulliford Press. Can, A. (199 0). The measurement of neighborhood dynamics in urban prices. Economic Geography. 66(3). 254 272. Retrieved from http://www.jstor.org.lp.hscl.ufl.edu/stable/pdfplus/143400.pdf Ch appell, A., Monk Turner, E, & Payne, B. (2011). Broken windows or window breakers: The influence of physical and social disorder on quality of life. Justice Quarterly. 28(3). 522 540 Retrieved from http://dx.doi.org/10.1080/07418825.2010.526129 Ding, C. & Knaap, G. J. (2002). Property values in inner city neighborhoods: The effects of homeownership, housing investment, and economic development. University of Maryland. Retrieved from http://www.smartgrowth.umd.edu/research/pdf/neighborhoodstability 816.pdf Elsby, M. W., Hobijn, B., & Sahin, A. (2010). The lab or market in the great recession. NBER. Retrieved from http://www.nber.org.lp.hscl.ufl.edu/papers/w15979.pdf?new_window=1

PAGE 77

77 ESRI. (2011). How hot spot analysis: getis ord gi* (spatial statistics) works. Retrieved from http://resources.esri.com/help/9.3 /arcgisengine/java/gp_toolref/spatial_statistics_t ools/how_hot_spot_analysis_colon_getis_ord_gi_star_spatial_statistics_works.ht m ESRI. (2011). Interpreting OLS results. Retrieved from http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//005p0000003000 0000.htm ESRI. (2009). Ordinary least squares (spatial statistics). Retrieved from http://webhelp.esri.com/arcgisdesktop/9.3/index.cfm?id=2173&pid=2169&topicna me=Ordinary_Least_Squares_(Spatial_Statistics) Foote, C. L., Gerardi, K., Goette, L., & Willen, P. Sub prime facts: What (we think) we know about the subprime crisis and what we don't. Federal Reserve Bank of Boston Retrieved from http://ssrn.com/abstract=1153411 Forecl osure. (2010). Merriam Webster. Retrieved December 3, 2010 from http://www.merriam webster.com/ Fotheringham, A., Brunsdon, C., & Charlton, M. (2002). Geographically Weighted Regression : The Analysis of Spatially Varying Relationships Joh n Wiley & Sons, Ltd. (UK) Franklin, J. P. & Waddell, P. (2002). A hedonic regression of home prices in king county, washington, using activity specific accessibility measures. Retrieved from http://nexus.umn.edu/courses/pa8202/waddell%20 %20hedonic.pdf Galster, G. (2000). Threshold effects and neighborhood change. Journal of Planning Education and Research 20(2). 146 162. Retrieved from http://jpe.sagepub.com.lp.hscl.ufl.edu/content/20/2/146.full.pdf Galster, G. (2001). On the nature of neighbourhood. Urban Studies, 38 (12), 2111 2124. Retrieved from http://usj.sagepub.com.lp.hscl.ufl.edu/content/38/12/2111.full.pdf+html Glaeser, E. L., Gottlieb, J. D., & Gyourko, J. (2010). Can cheap credit explain the housing boom?. Retrieved from http://economics.stanford.edu.lp.hscl.ufl.edu/files/Glaeser5_24_Credit.pdf Goodman, A. C. (1977). A comparison of block group and census tract data in a hedonic housing price model. Land Economics. 54(3). 483 487. Retrieved from http://www.jstor.org.lp.hscl.ufl.edu/stable/pdfplus/3145991.pdf?acceptTC=true

PAGE 78

78 Goodman, A. C. (1997). A ndrew court and the invention of hedonic price analysis. Journal of Urban Economics. 44(2). 291 298. Retrieved from http://www.sciencedirect.com/science/article/pii/S009411909 7920714 Hillsborough County. (2011). Frequently asked questions code enforcement. Retrieved from http://www.hillsboroughcounty.org/faq/departmentDetail.cfm?chvDepartment=Co de%20Enforcement HUD. (2011). Land banks. U.S. Department of Housing and Urban Development. Retrieved from http://www.hud.gov/offices/cpd/about/conplan/foreclosure/landbanks.cfm Hunter, A. (1979). The urban neighborhood its analytical and social contexts. Urban Affairs Review. 14(3). 267 288. Retrieved from http://uar.sagepub.com/content/14/3/267.full.pdf Immergluck, D. & Smith, G. (2005). The impact of single family mortgage foreclosures on neighborhood crime. Housing Stuides 21(6). 851 866. Retrieved from http://www.prism.gatech.edu.lp.hscl.ufl.edu/~di17/HousingStudies.pdf Immergluck, D. & Smith, G. (2005). The external costs of foreclosure: the impact of single family mortgage foreclosures on propert y values. Fannie Mae Foundation. 17(1). 57 79. Retrieved from http://findaforeclosurecounselor.net/network/neighborworksProgs /foreclosuresolu tions/pdf_docs/hpd_4closehsgprice.pdf Investopedia. (2010). Subprime mortgage. Investopedia. Retrieved from http://www.investopedia.com/terms/s/subprime_mortgage.asp Martin, D. G. (2003). Enacting neighborhood. Urban Geography 24(5). 361 385. Retrieved from http://bellwether.metapress.com.lp.hscl.ufl.edu/content/x5p11423r 35tt55m/fulltext .pdf Mayer, C. & Pence, K. (2008). Subprime mortgages: What, where, and to whom?. Federal Reserve Board. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.167.5371&rep=rep1&ty pe=pdf Nelson, R. H. (2011). Homeowners associations in historical perspective. Public Administration Review. 71(4). 546 549. Retrieved from http://onlinelibrary.wiley.com.lp.hscl.ufl.edu/doi/10.1111/j.1540 6210.2011.02384.x/full

PAGE 79

79 Olorunnipa, T. (2010, October 28). South florida foreclosure rate highest. The Miami Herald. Retrieved from http://www.miamiherald.com/2010/10/28/1895871/south florida foreclosure rate.html Perkins, C. (2010). Privatopia in distress: The impact of the foreclosure crisis on Nevada Law Journal. 10(561). 561 585. Retrieved from http://scholars.law.unlv.edu/cgi/viewcontent.cgi?article=1022&context=nlj&sei redir=1#search=%22homeowner%20associations%20foreclosure%22 Quinnipiac. (2011). http://faculty.quinnipiac.edu/libarts/polsci/Statistics.html NPR. Retrieved from http://www.npr.org/2011/08/29/139971310/land bank knocks out some foreclosure problems Shilling, J. (2009). Code enforcement and community stabilization: The forgotten first responders t o vacant and foreclosed homes. Albany Government Law Review. 2. 101 162. Retrieved from http://community wealth.com/_pdfs/articles publications/cdcs/article sc hilling.pdf Sopranzetti, B. J. (2010). Hedonic regression analysis in real estate markets: A primer. In C. Lee, Handbook of Quantitative Finance and Risk Management. 1201 1207. Retrieved from http://www.springerlink.com.lp.hscl.ufl.edu/content/n144n14n031432m6/fulltext.p df Tobler, W. R. (1970). A computer movie simulating urban growth in the detroit region. Economic Geography. 46. Retrieved from http://www.jstor.org/stable/143141 Wheaton, W. C. (1990). Vacancy, search, and prices in a housing market matching model. Journal of Political Economy. 98(6). 1270 1292. Retrieved from http://www.jstor.org.lp.hscl.ufl.edu/stable/pdfplus/2937758.pdf?acceptTC=true Wilson, J. Q., & Kelling, G. (1982). The police and neighborhood safety: Broken windows. Atlantic Monthly, 127 29 38. Retrieved from http://www.forestry.gov.uk/website/pdf.nsf/b591cb1aa3d9d9ac802570ec004f557 d/7e15282335cea36b802575e4004c96b7/$FILE/BrokenWindowTheory.pdf Zwick, P (2010). Basic spatial statistics: Analyzing patterns. [PowerPoint slides].

PAGE 80

80 BIOGRAPHICAL SKETCH Charles Gibbons was born in 1987 in Tampa, Florida. He is the son of Jane and Gary Gibbons. He has one sister, Kristie. Charles studied at H.B. Plant High school before attending the University of Florida, where he earned his Bachelor of Science in geog raphy in 2009. During his tenure, he would support his Florida Gator football team by playing alto saxophone in T he Pride of the Sunshine The University of Florida In the fall of 2009, Charles began his studies in Urban an d Regional Planning at the University of Florida. As a graduate assistant, Charles began to work as a GIS Analyst at the Shimberg Center for Housing Studies at UF. Here he provided valuable assistance in creating and maintaining spatial datasets related to the study of affordable and special needs housing in Florida. Charles will graduate with his Master of Arts in Urban and Regional Planning in December 2011. Charles hopes to continue his work with geospatial technologies and planning in the future. Outside of his academic career, and exploration has kept him infected by the travel bug. He also enjoys playing home chef and considers himself a foodie