<%BANNER%>

The Effect of the Closure and Subsequent Redevelopment of a Military Base on Surrounding Single-Family House Prices

Permanent Link: http://ufdc.ufl.edu/UFE0042279/00001

Material Information

Title: The Effect of the Closure and Subsequent Redevelopment of a Military Base on Surrounding Single-Family House Prices a Case Study of a U.S. Navy Master Jet Base, Naval Air Station Cecil Field, Jacksonville, Florida
Physical Description: 1 online resource (131 p.)
Language: english
Creator: Jennings, Gregory
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: 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: Politicians frequently paint a picture of socioeconomic doom and gloom when a military base closure announcement is made that affects their constituency area. While a base closure can certainly have a negative impact on a surrounding community; a community s chance of surviving a base closure is very much dependent on the context of the area, its economic diversity and the local base redevelopment planning effort. Recent research on the effects of base closure indicates that a closure typically has a short-lived socioeconomic impact on a community until the community can redevelop the base and reap benefits from the redevelopment. Though communities can typically prosper shortly after a base closure, the same research indicates that the lingering negative effect of base closure is on the housing market. While the negative effect of base closure may very well be on the housing market, little research exists that demonstrates the housing price was actually affected by the base closure versus other economic impacts or shocks. This thesis tests the effect of the closure of Cecil Field Naval Air Station in Jacksonville, Florida on the surrounding communities single-family housing prices. This research tests the effects of the closure of Cecil Field on the Jacksonville single-family housing market through the use of a hedonic pricing model which is run through ordinary least squares regression and geographically weighted regression. Jacksonville, Florida is the home to three large Navy bases and other military support annexes. Cecil Field is within a ten minute drive from Jacksonville Naval Air Station which remains open today. Because of the significance of Jacksonville Naval Air Station in the terms of military member family support with its commissary, exchange, large hospital and easy access to major highways, most military members of Cecil Field and their families chose to live closer to Jacksonville Naval Air Station. The results of this research indicate that the closure of Cecil Field had no impact on single-family housing prices due to its rural location and the fact that most of its members lived closer to Jacksonville Naval Air Station.
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 Gregory Jennings.
Thesis: Thesis (M.A.U.R.P.)--University of Florida, 2010.
Local: Adviser: Jourdan, Dawn.
Local: Co-adviser: Zwick, Paul D.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042279:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042279/00001

Material Information

Title: The Effect of the Closure and Subsequent Redevelopment of a Military Base on Surrounding Single-Family House Prices a Case Study of a U.S. Navy Master Jet Base, Naval Air Station Cecil Field, Jacksonville, Florida
Physical Description: 1 online resource (131 p.)
Language: english
Creator: Jennings, Gregory
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: 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: Politicians frequently paint a picture of socioeconomic doom and gloom when a military base closure announcement is made that affects their constituency area. While a base closure can certainly have a negative impact on a surrounding community; a community s chance of surviving a base closure is very much dependent on the context of the area, its economic diversity and the local base redevelopment planning effort. Recent research on the effects of base closure indicates that a closure typically has a short-lived socioeconomic impact on a community until the community can redevelop the base and reap benefits from the redevelopment. Though communities can typically prosper shortly after a base closure, the same research indicates that the lingering negative effect of base closure is on the housing market. While the negative effect of base closure may very well be on the housing market, little research exists that demonstrates the housing price was actually affected by the base closure versus other economic impacts or shocks. This thesis tests the effect of the closure of Cecil Field Naval Air Station in Jacksonville, Florida on the surrounding communities single-family housing prices. This research tests the effects of the closure of Cecil Field on the Jacksonville single-family housing market through the use of a hedonic pricing model which is run through ordinary least squares regression and geographically weighted regression. Jacksonville, Florida is the home to three large Navy bases and other military support annexes. Cecil Field is within a ten minute drive from Jacksonville Naval Air Station which remains open today. Because of the significance of Jacksonville Naval Air Station in the terms of military member family support with its commissary, exchange, large hospital and easy access to major highways, most military members of Cecil Field and their families chose to live closer to Jacksonville Naval Air Station. The results of this research indicate that the closure of Cecil Field had no impact on single-family housing prices due to its rural location and the fact that most of its members lived closer to Jacksonville Naval Air Station.
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 Gregory Jennings.
Thesis: Thesis (M.A.U.R.P.)--University of Florida, 2010.
Local: Adviser: Jourdan, Dawn.
Local: Co-adviser: Zwick, Paul D.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042279:00001


This item has the following downloads:


Full Text

PAGE 1

1 THE EFFECT OF THE CLOSURE AND SUBSEQUENT REDEVELOPMENT OF A MILITARY BASE ON SURROUNDING SINGLEFAMILY HOUSE PRICES: A CASE STUDY OF A U.S. NAVY MASTER JET BASE, NAVAL AIR STATION CECIL FIELD, JACKSONVILLE, FLORIDA By GREGORY P. JENNINGS 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 PLANNING UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Gregory P. Jennings

PAGE 3

3 To my sons

PAGE 4

4 ACKNOWLEDGMENTS First and foremost a most deserving thanks goes to my wife for taking a leap and starting this great adventure with me. A special thanks to my sons for understanding when D addy w as not available due to class assignments and this research S incere gratitude is owed to my chairperson, Dawn Jourdan, for her unending support, patience, and for pushing me to the end as she quickly learned my propensity for proc r astination. Thank s go out to my committee members; Paul Zwick and Andres Blanco, for always having your door open and helping me guide my research from the unmanageable to the manageable. A great deal of appreciation is owed to the staff at Metro Mark et Trends, Inc. for allowing access to yo ur sales data. This research could not have been conducted without it. Finally thanks go out to the staff at Jacksonv ille Economic Developme nt Commission for supporting numerous fact finding vis its and for always answering questions.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES .......................................................................................................... 9 LIST OF ABBREVIATIONS ........................................................................................... 11 ABSTRACT ................................................................................................................... 13 CHAPTER 1 INTRODUCTION .................................................................................................... 15 Problem St atement ................................................................................................. 17 Professional Significance of the Problem ................................................................ 18 Methodology ........................................................................................................... 18 Organization of Thesis ............................................................................................ 20 2 THEORETICAL FRAMEWORK .............................................................................. 21 Effects of Base Closure .......................................................................................... 21 Base C lo sure and M ultipliers ............................................................................ 21 Base Closure and Its Affect on the Housing Market ......................................... 24 Housing Prices ........................................................................................................ 24 Housing Prices at the National Level ................................................................ 26 Housing Prices at the Regional Level ............................................................... 28 Housing Prices at the Local Level .................................................................... 32 Proximity to a specific externality and its effect on housing prices ............. 37 Regression Methods ............................................................................................... 39 3 METHODOLOGY ................................................................................................... 43 History of Cecil Field ............................................................................................... 43 Data ........................................................................................................................ 49 Dependent Variable .......................................................................................... 49 Explanatory Variables ...................................................................................... 51 Model ...................................................................................................................... 54 Hedonic Pricing Model ..................................................................................... 54 Regression Analysis of the Hedonic Pricing Model .......................................... 55 4 RESULTS ............................................................................................................... 56

PAGE 6

6 Geographically Weighted Regression (GWR): Trials and Tribulations ................... 56 Variables and Variance in Their Values (Local Multicollinearity) ...................... 56 Variables and Global Multicollinearity ............................................................... 58 Issues Not Related to Geographically Weighted Regression .................................. 59 Impact Zones .................................................................................................... 59 Structural Variables .......................................................................................... 60 Percent of High School Dropouts in C ensus T ract ( PCTDROP) V ariable ........ 60 Prime I n terest Rate at Time of Transaction ( RATE) V ariable ........................... 60 Central Business District .................................................................................. 61 The Revised Hedonic Pricing Model ....................................................................... 61 Results Using Revised Methodology ...................................................................... 62 Zip Codes with Cecil Personnel ........................................................................ 62 1992 Dataset .................................................................................................... 63 1993 Dataset .................................................................................................... 66 1994 Dataset .................................................................................................... 69 1999 Dataset .................................................................................................... 73 2000 Dataset .................................................................................................... 75 2009 Dataset .................................................................................................... 78 Acreage of Lot ( LOTSIZE ) Variable ........................................................................ 81 Property Age in Years ( AGE) Variable .................................................................... 82 Square Footage of House ( SF ) Variable ................................................................. 83 Median Household Income in the Census Tract ( INCOME ) Variable ..................... 85 Median Age in the Census Tract ( MEDAGE) Variable ............................................ 86 Eucli dean Distance to Cecil Field ( DISTCEC ) Variable .......................................... 87 Euclidean D istance to the Central Business District ( DISTCBD) Variable .............. 88 Euclidean Distance to an Industrial Land Use ( DISTINDU) Variable ...................... 88 Euclidean Distance to a Commercial Land Use ( DISTCOMM ) Variable ................. 88 Euclidean Distance to a Hospital ( DISTHPTL ) Variable .......................................... 89 Euclidean Distance to a School ( DISTSCHL ) Variable ........................................... 89 Euclidean Distance to a Major Water Body ( DISTWATR) Variable ........................ 89 Euc lidean Distance to a Major Road ( DISTROAD) Variable ................................... 90 Euclidean Distance to a Know n Zip Code w ith Cecil Personnel ( DISTZIP) Variable ............................................................................................................... 90 5 CONCLUSION ........................................................................................................ 91 Summary of Findings .............................................................................................. 91 Recommendations for Future Studies .................................................................... 93 Limitations of this Study .......................................................................................... 93 Areas of Future Research ....................................................................................... 93 APPENDIX A THE MODERN BASE CLOSURE PROCESS ........................................................ 95 B KNOWN ZIP CODES WITH CECIL FIELD PERSONNEL ...................................... 97

PAGE 7

7 C 1992 DUVAL GWR RESULTS ................................................................................ 98 D 1993 DUVAL GWR RESULTS .............................................................................. 103 E 1994 DUVAL GWR RESULTS .............................................................................. 108 F 1999 DUVAL GWR RESULTS .............................................................................. 113 G 2000 DUVAL GWR RESULTS .............................................................................. 118 H 2009 DUVAL GWR RESULTS .............................................................................. 123 LIST OF REFERENCES ............................................................................................. 128 BIOGRAPHICAL SKETCH .......................................................................................... 130

PAGE 8

8 LIST OF TABLES Table page 3 1 2006 Direct e conomic i mpact of the Cecil Commerce Center ............................ 48 3 2 2006 Direct and i ndirect economic i mpact of the Cecil Commerce Center ......... 48 3 3 2015 Direct economic i mpact of the Cecil Commerce Center ............................ 48 3 4 2015 Direct and indirect economic i mpact of the Cecil Commerce Center ......... 49 3 5 Description of dependent variable used in the hedonic price m odel ................... 49 3 6 Major events i nvolving Cecil Field ...................................................................... 50 3 7 Description of explanatory variables used in the hedonic price models .............. 51 3 8 Vector category and source of explanatory variables used in the hedonic p rice models ....................................................................................................... 52 4 1 Revised e xplanatory v ariables ............................................................................ 61

PAGE 9

9 LIST OF FIGURES Figure page 3 1 Map showing the location of Cecil Field and proximity to other military bases ... 45 4 1 Initial 1992 Ordinary Least Squares ( OLS ) Results ............................................ 59 4 2 1992 OLS Results .............................................................................................. 63 4 3 1992 Geographically Weighted Regression ( GWR ) Results .............................. 65 4 4 1993 OLS Results .............................................................................................. 67 4 5 1993 GWR Results ............................................................................................. 68 4 6 1994 OLS Results .............................................................................................. 70 4 7 1994 GWR Results ............................................................................................. 71 4 8 1999 OLS Results .............................................................................................. 73 4 9 1999 GWR Results ............................................................................................. 74 4 10 2000 OLS Results .............................................................................................. 76 4 11 2000 GWR Results ............................................................................................. 77 4 12 2009 OLS Results .............................................................................................. 78 B 1 Naval Air Station ( NAS ) Cecil Field Personnel Zip Codes .................................. 97 C 1 Duval 1992 Property Age in Years ( AGE) GWR Output ..................................... 98 C 2 Duval 1992 Square Footage of House ( SF ) GWR Output .................................. 99 C 3 Duval 1992 Acreage of Lot ( LOTSIZE ) GWR Output ....................................... 100 C 4 Duval 1992 Median Household Income in the Census Tract ( INCOME) GWR Output ............................................................................................................... 101 C 5 Duval 1992 Median Age in the Census Tract ( MEDAGE) GWR Output ........... 102 D 1 Duval 1993 AGE GWR Output ......................................................................... 103 D 2 Duval 1993 SF GWR Output ............................................................................ 104 D 3 Duval 1993 LOTSIZE GWR Output .................................................................. 105

PAGE 10

10 D 4 Duval 1993 INCOME GWR Output ................................................................... 106 D 5 Duval 1993 MEDAGE GWR Output ................................................................. 107 E 1 Duval 1994 AGE GWR Output ......................................................................... 108 E 2 Duval 1994 SF GWR Output ............................................................................ 109 E 3 Duval 1994 LOTSIZE GWR Output .................................................................. 110 E 4 Duval 1994 INCOME GWR Output ................................................................... 111 E 5 Duval 1994 MEDAGE GWR Output ................................................................. 112 F 1 Duval 1999 AGE GWR Output ......................................................................... 113 F 2 Duval 1999 SF GWR Output ............................................................................ 114 F 3 Duval 1999 LOTSIZE GWR Output .................................................................. 115 F 4 Duval 1999 INCOME GWR Output ................................................................... 116 F 5 Duval 1999 MEDAGE GWR Output ................................................................. 117 G 1 Duval 2000 AGE GWR Output ......................................................................... 118 G 2 Duval 2000 SF GWR Output ............................................................................ 119 G 3 Duval 2000 LOTSIZE GWR Output .................................................................. 120 G 4 Duval 2000 INCOME GWR Output ................................................................... 121 G 5 Duval 2000 MEDAGE GWR Output ................................................................. 122 H 1 Duval 2009 AGE GWR Output ......................................................................... 123 H 2 Duval 2009 SF GWR Output ............................................................................ 124 H 3 Duval 2009 LOTSIZE GWR Output .................................................................. 125 H 4 Duval 2009 INCOME GWR Output ................................................................... 126 H 5 Duval 2009 MEDAGE GWR Output ................................................................. 127

PAGE 11

11 LIST OF ABBREVIATIONS AFB Air Force Base BEA Bureau of Economic Analysis BRAC Base Realignment and Closure CBD Central Business District dbA Decibals on the A Scale DOD Department of Defense EIS Environmental Impact Statement FHA Federal Housing Administration FTP File Transfer Protocol GIS Geographic Information System GWR Geographically Weighted Regression JAA Ja cksonville Aviation Authority JEDC Jacksonville Economic Development Commission MIRS Mortgage Interest Rate Survey MMT Metro Market Trends, Inc. MPF Maritime Prepositioning Force MSA Metropolitan Statistical Areas NAR National Association of Realtors NAS Naval Air Station NLS Non linear Least Squares OLS Ordinary Least Squares PIN Property Identification Number REIS Regional Economic Information System RIMS Regional Input Output Multiplier System

PAGE 12

12 TSLS Two Stage Least Squares USMC United States Marine Corps VIF Variance Inflation Factor

PAGE 13

13 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 THE EFFECT OF T HE CLOSURE AND SUBSEQUENT REDEVELOPMENT OF A MILITARY BASE ON SURROUNDING SINGLEFAMILY HOUSE PRICES: A CASE STUDY OF A U.S. NAVY MASTER JET BASE, NAVAL AIR STATION CECIL FIELD, JACKSONVILLE, FLORIDA By Gregory P. Jennings August 2010 Chair: Dawn Jourdan Major: Urban and Regional Planning Politicians frequently paint a picture of socioeconomic doom and gloom when a military base closure announcement is made that affects their constituency area. While a base closure can certainly have a negative impact on a surrounding community ; a communitys chance of surviving a base closure is very much dependent on the context of the area, its economic diversity and the local base redevelopment planning effort. Recent research on the effects of base closure indicates that a closure typically has a short lived socioeconomic impact on a community until the community can redevelop the base and reap benefits from the redevelopment. Though communities can typically prosper shortly after a base closure, the same re search indicates that the lingering negative effect of base closure is on the housing market. While the negative effect of base closure may very well be on the housing market, little research exists that demonstrates the housing price was actually affected by the base closure versus other economic impacts or shocks. This thesis tests the effect of the closure of Cecil Field Naval Air Station in Jacksonville, Florida on the surrounding communities singlefamily hous ing prices

PAGE 14

14 This research tests the effect s of the closure of Cecil Field on the Jacksonville singlefamily hous ing market through the use of a hedonic pricing model which is run through ordinary least squares regression and geographically weighted regression. Jacksonville, Florida is the home to three large Navy bases and other military support annexes. Cecil Field is within a ten minute drive from Jacksonville Naval Air Station which remains open today. B ecause of the significance of Jacksonville Naval Air Station in the terms of military member family support with its commissary, exchange, large hospital and easy access to major highways, most military members of Cecil Field and their families chose to live closer to Jacksonville Naval Air Station. The results of this research indicate that the c losure of Cecil Field had no impact on singlefamily housing prices due to its rural location and the fact that most of its members lived closer to Jacksonville Naval Air Station.

PAGE 15

15 CHAPTER 1 INTRODUCTION The United States has seen a decline in defense spending since the Cold War and has reshaped its military base footprint to adapt to a rapidly changing world and better match facilities to forces (U.S. Department of Defense, n.d.). To guide this process, Congress passed the Base Closure and Realignment A ct of 1988, also known as Base Realignment and Closure (BRAC), to provide a bipartisan avenue for the Department of Defense (DOD) to reduce its footprint and realign remaining military bases and assets (U.S. Department of Defense, n.d.). The Base Closure a nd Realignment Act is intended to minimize political play in the decision making process (Dardia, McCarthy, Malkin, & Vernez, 1996, p. 1); however, when the initial BRAC slate is released by the Secretary of Defense for consideration by the presidentially appointed Base Closure and Realignment Commission lawmakers often hurl unfounded predictions of dire consequences as a result of an impending military base closure in the community they represent. (Bradshaw, 1999, p. 1) The United States has been through five BRACs since the inception of the law with the most recent BRAC decision in 2005. Case studies conducted on closures since the late 1960s indicate that the dire predictions of the politicians and the cities affected are, with some exceptions, often exaggerated, and that the communities surrounding the base and the host city return to status quo or prosper soon after the base closes (Bradshaw, 1999, p. 2). Though the effects of base closure are argued to be minor or short lived, Dardia et al. (1996) and Bradshaw (1999) found that the most negatively affected indicator of a base closure is commonly the sales and rental values of housing surrounding the base due to the mass outmigration of military members and their

PAGE 16

16 families; leaving a void in demand wit hin the housing market. Although base closure case studies exist that examine socioeconomic indicators such as population, labor force size, unemployment rates, school enrollment, housing values and municipal revenues ; in depth research examining solely th e base closures effect on the housing market is lacking The author was able to find one doctoral dissertation completed in 2009 which studies the effect of two base closures on housing prices Because of the absence of such research, t he author sought literature with a focus outside of the military base closure arena but closely synonymous with the community effects of a military base closure. Interestingly, literature was found that examines the building of sports stadiums and their effect on the surro unding housing market. While not completely synonymous with a base closure ( the community does not experience the loss of thousands of military members and their families) much can be learned and applied to this study in terms of the methodology utilized. This research will add a housing market focused component to existing base closure research and will provide more housing market data to future affected communities in order to assist them with their redevelopment plans and predictions of base closure eff ects on their community. This thesis is a longitudinal case study of the closure of a U.S. Navy Master Jet Base, Naval Air Station (NAS) Cecil Field, located in Jacksonville (Duval County) Florida. NAS Cecil Field was slated for closure as a result of the 1993 BRAC decision and officially c eased air operations on September 30, 1999 (GlobalSecurity.org, n.d.). This thes is will study the effects of the closure on the surrounding community housing market; focusing on singlefamily house sales values Though other indicators play into

PAGE 17

17 the housing market, t he author chose to focus on singlefamily house sales price s due to the ready availability of this data. Other data such as rental vacancies or vacant singlefamily homes was either unavailable for the study area or cost prohibitive to procure. Jacksonville is unique as compared to bases in formerly conducted case studies because, prior to the closure of NAS Cecil Field, it was home to three sizable naval bases which include NAS Jacksonville and Naval Station Mayport It is also home to U.S. Marine Corps (USMC) Blount Island Command; a key logistics hub for the U.S. Marine Corps Maritime Prepositioning Force (MPF). This high naval concentration contributes to existing literature that is primarily based on Air Force and Army base closures which are typically solo bases in a surrounding community with or without a diverse economy. Jacksonvilles extensive land area, diverse economy and high naval concentration conveys a different perspective to existing literature. Problem Statement This research is to address the problem that no in depth housing market analysis exists that attempts to ascertain where military member domicile concentrations exist in the surrounding community ; and whether or not a military base closure had an effect on the single family house housing market in the surrounding community The research objective of this study is to measure the impact of the closure of master jet base, NAS Cecil Field on singlefamily house sales prices to determine if the Jacksonville housing market benefitted, suffered, or was indifferent due to the base closure. The questions to be asked are: 1) When the closure of Cecil Field was announced in 1993, was there an anticipatory effect on single family house sales prices from the announcement to the actual closure in 1999?

PAGE 18

18 a. Parcels occupied by singlefamily houses with a sales date between 1993 1994 i n Duval County 2) After Cecil Field closed and was turned over to the City of Jacksonville in 1999, what effect did the closure and redevelopment have on single family house sales price s after the closure and in 2009? a. Parcels occupied by singlefamily homes w ith a sales date between 19992000 and 2009 in Duval County Professional Significance of the Problem This study has professional significance because of its contribution to communities similar to Jacksonville, Florida that must endure and plan for the effects of the closure of a military base in their community in the future. Jacksonville s previously discussed uniqueness and this studys focus on housing prices in military concentration areas outside of the base fence line adds an additional dimension to existing case studies Using this information, communities in a similar context to Jacksonville can better plan and prepare for an impending base closure. Methodology This research is a longitudinal case study on the former Naval Air Station Cecil Field which is primarily located in the southwest of the city of Jacksonville. The city of Jacksonville is in Duval County and the two are a consolidated government under one mayor. A small portion of the southern tip of the former base extends south of the Duval County border into Clay County This research will be conducted utilizing data of public record and data purchased from a vendor. All data will be imported into a geographic information system (GIS) software package for anal ysis and presentation. Parcel data for Duval Count y, to

PAGE 19

19 include property identification number (PIN) and geographic shape file, were obtained from the Florida Department of Revenue file transfer protocol (FTP) website. Additional parcel data, to include PI N and effective year built, were obtained from the Florida Geographic Data Library maintained at the Unive rsity of Florida. The ZIP code boundary geographic shape file was obtained from the United States Census Bureau website. Proprietary t ransactional data, to include all singlefamily home sales recorded, w ere purchased from Metro Market Trends, Inc. located in Pensacola, Florida. This research will test a single dependent variable which is the singlefamily ho use sales price There are t hree cause s, o r independent variables. The first independent variable is the announcement in 1993 of the impending closure of NAS Cecil Field. The second independent variable is the closure of NAS Cecil Field in 1999. The third independent variable is the subsequent redevelopment of NAS Cecil Field. Many explanatory variables act on the dependent variable and they are: lot size, bathrooms, bedrooms, structure age, income, median age, percent of high school dropouts, financing cost, distance to major arterial roads, dist ance to schools, distance to waterfront, distance to hospitals, distance to commercial land uses, distance to industrial land uses, distance to Cecil Field and zip codes that Cecil Field personnel lived at. The dependent variable will be tested: (1) one c alendar prior to the base closure; (2) the calendar year of and the calendar year following the announcement of the base closure; (3) the calendar year of and the calendar year following the final air operations at NAS Cecil Field; (4) the year prior to the authoring of this thesis to determine the impact of Jacksonvilles redevelopment of Cecil Field on the dependent variable.

PAGE 20

20 Organization of Thesis This thesis focuses on measuring the impacts on the Duval County singl e family housing market from the closure of NAS Cecil Field to determine if the single family housing market benefitted, suffered losses or was indifferent to the base closure. In Chapter Two the theoretical framework is assembled by reviewing literature on previous base closure case studies to ascertain the economic effects of base closure. Then literature on housing pricing methodologies is analyzed in order to build the methodology for this thesis. Chapter Three details the history of Cecil Field and the data and methodology f or this case study. C hapter Four presents the results and the significant findings Chapter Five discusses the results and analyzes them as they pertain to Cecil Field. Chapter Six concludes the study with a review of the research questions and findings, l imitations of the research and suggestions for future research.

PAGE 21

21 CHAPTER 2 THEORETICAL FRAMEWORK As this research attempts to answer the question of what effect the closure of Cecil Field had on the singlefamily housing market in the surrounding community it is necessary to engage the literature associated with the socio economic impacts of military base closures which le a d to the focus question of this research. From there literature on housing market pricing and the various methodologies used by researchers will be analyzed in search of a methodology fi tting for the research question of this thesis. Effects of Base Closure The economic impact of the closure of a military base can potentially be positive, negative or neutral on the surrounding community. Dardia et al. (1996, p. 14) argue that communities with diverse agglomeration economies fare better than communities with a military base as a singlesource economic base. No matter if their economy is diverse or not, politicians and local officials often argue through economic predictions that the closure of a military base in their community will have a negative socioeconomic impact. Others not tied to the communities affected argue through post closure case studies that a base closure can stimulate the economy through redevelopment of the closed base and that often the community is left better off than when the base was operational or that any negative impacts are localized immediately around the base and dissipate within the next few years. Base C losure and M ultipliers As lawmakers and local officials attempt to persuade the Base Closure and Realignment Commission hereinafter Commission, to spare their districts military base

PAGE 22

22 from closure, the socioeconomic pic ture forecasted is often bleak. The bleak picture is typically painted by the communities through the use of multiplier s (Dardia et al., 1996, p. 9). A multiplier is a ratio that represents the direct an d indirect effects on employment or income when one j ob is created or lost and assumes that goods and services are being provided from within the community to support that one job, therefore creating or removing additional jobs (Dardia et al., 1996, p. 9). Assumptions about the extent to which incomes are s pent within a community can lead to very different assessments of the impacts from the loss of that income (Cowan & Webel, 2005, p. 2). For example, co mmunities that choose a higher multiplier (indicating that more goods and services are being provided to the base from inside the community or region) will predict a more negative effect on overall employment and income in the community if one job is lost on the military base However, a rural agr icultural community typically cannot supply the goods and services necessary to support a military base and the base must look outside of the community for those goods and services (Bradshaw, 1999, p. 9). Therefore a rural community that chooses multiplier s based on goods and services being supplied from within the community will artificially inflate the forecasted negative effects of base closure on their community. Based on the authors experience, dependent on the type of base and its mission, military b ases may receive most of their supplies through logistics chains with hubs outside of the region. This can affect the income or employment multiplier chosen for an urbanized community as well. If an urbanized community assumes the base supplies are coming from within the community then the multiplier and negative impact prediction of the potential base closure is again inflated.

PAGE 23

23 Based on the authors experience, most economic activity for military members and their spouses is internalized to the base. Military bases typically provide housing for lower ranking junior enlisted members and occasionally for higher ranking members, hav e commissaries to buy groceries at discounted rates in comparison to the local economy, have exchanges to buy retail items, and provide medical and dental care at no cost. Though, many military members typically live outside of the base on the economy, mos t choose to depend on the base for their daily needs due to lower or no costs which can alter any nonmilitary multiplier chosen by a community to predict the socioeconomic effects of a closure. Military retirees within the community typically rely on the base for daily needs as well. When a base closes, some retirees will leave the community in order to be near another military base to use its services or for other reasons (Fagan, 2001, p. 18). Retirees that choose not to leave the community revert their s pending from the base to the local economy which is a positive impact on the community (Bradshaw, 1999, p. 5). Though the impact can be minor, communities must account for military retirees when predicting the impact of a base closure. Military spouses that are employed in the community will likely leave with their spouse when the base closes (Renski, 2007, p. 50). The job that the spouse held within the community then becomes vacant and provides an opportunity to any person with the same skill set that los t employment due to the base closure (Renski, 2007, p. 50). If the job vacancy is not filled from within the community, the business looking to fill the vacancy will reach outside of the community to fill the position which promotes inmigration to the com munity.

PAGE 24

24 Base Closure and Its Affect on the Housing Market Military members and their families that live outside of the military base will most likely cause a void in the local housing market demand upon their departure (Dardia et al., 1996, p. 8). Existing case studies show that housing is the most impacted indicator among all of the economic indicators (Dardia et al., 1996, p. 8). Rental and sales pr ices fall as there is suddenly less demand (Bradshaw, 1999, pp. 67). This effect can be relieved quickly or linger dependent on the communitys redevelopment efforts or industry diversity. The effect is null if most of the military members and their families live within the military installation. As previously mentioned, earlier studies recognize that the housing market surrounding the closed base is affected; however, almost no literature exists that explains just how much the closure of the base has to do with housing mar ket. For example, Dardia et al. (1996, pp. 2829) show informative graphs that follow trends in affected areas surrounding the closing bases from the announcement of the closure through the actual closure. These graphs clearly show an increase in vacancies and a decrease in sales prices during the closure time period. While this trend is import ant to note, there is no statistical analysis that attrib utes the vacancies and h ouse prices to the base closure. Though the vacancy and housing price variables most likely were affected by the closures, other variables were very much at play Housing Prices R esearch demonstrates that there are national, regional and local attributes to house prices. Each house in a neighborhood has its own distinct attributes. OSullivan (2009) argues that housing stock is heterogeneous, with each dwelling offering a

PAGE 25

25 diff erent bundle of housing services (p. 339). OSullivan continues with his argument that: Dwellings differ in size, layout, style, utilities (heating and electrical), and the quality of the interior and exterior. As we saw in the chapter on neighborhood choice, when you choose an apartment or house, you also choose a neighborhood, with its own bundle of housing services. Neighborhoods differ in accessibility to jobs and social opportunities, local public goods and taxes, and environmental quality. There are currently two techniques for analyzing home prices: hedonic models and repeat sales models (Quigley, 1995, p. 2). Repeat sales models measure the price of the same house sold multiple times (Quigley, 1995, p. 2). Quigley (1995, p. 2) argues that the repeat sales model drastically reduces an empirical studys sample size because it relies on the characteristics of the home to have remained the same between sales and also the sales may not be representative of the local housing market. Hedonic models relate t he selling price to structural and locational characteristics of the home and are routinely estimated from repeated cross sectional samples of dwellings (Quigley, 1995, p. 2). Quigley argues that in a hedonic model neither the functional form of the rel ationship nor the set of variables is known with certainty which limits the generality of the procedure when applied across markets or time periods (Quigley, 1995, p. 2). Small sample sizes and the argument that several researchers have found that repeat sales tend to involve lower priced and homogenous starter homes which are more frequently sold than more expensive homes (Quigley, 1995, p. 5) tends to guide researchers toward the use of hedonic models for m ost empirical analysis of home prices. For this reason, the following subsections on housing prices will look at previous hedonic pricing research and examine the national, regional and local components of the house price in an attempt to dissect the key v ariables at each level.

PAGE 26

26 Housing Prices at the National Level Peek and Wilcox (1991) conduct a regression analysis utilizing three national data sources ( Federal Housing Administration (FHA), US Census Bureau and Mortgage Interest Rate Survey (MIRS)) from 1 9631989 and s ix national data sources ( FHA, National Association of Realtors (NAR), US Census Bureau, MIRS, Bureau of Economic Analysis (BEA) and Freddie Mac ) from 19701989 in order to understand the components and their strength of effect on house price s The reas on for splitting the analysis i s due to the additional latter organizations not recording data duri ng the 19631970 time period. Housing prices, as with any product or commodity are a function of supply and demand. Peek and Wilcox (1991) identi fy the explanatory variables of both housing supply and housing demand. They hypothesize that housing supply (HS) responds positively to price ( + P) and negatively to the real price of construction materials ( RPCON) and that housing demand (HD) responds negatively to price ( P) positively to the real income per household (+INC), positively to size and age distribution of the population (+HH), negatively to the cyclical component of the unemployment rate (UGAP) negatively to homeowners real aftertax borrowing costs ( RATMR) and negatively to household heads age 2029 ( POP20s) They then equate supply and demand producing a reducedform equation for real house prices (p. 366) : P=H( UGAP, RATMR, + INC, + HH, POP20s, + RPCON) After running their regression analysis for the 19701989 period, Peek and Wilcox (1991) f i nd a lack of compelling evidence that transitory unemployment affects the prices of longterm assets (p. 373). A majority of their data series show that housing prices respond negatively to increasing RATMR For the remaining explanatory variables the coefficients were uniformly positive or negative and were statistically

PAGE 27

27 significant : an increase in INC and HH increases housing prices, an increase in POP20s decreases housing pr ices and an increase in RPCON increases housing prices and accounts for 42% (.422 coefficient) of the overall house price which reflects real world conditions where materials generally reflect about 50% of a homes construction cost. After performing a regression analysis for both the 19631989 and the 19701989 periods Peek and Wilcox (1991) conclude: Real house prices are estimated to decline with increases in real after tax interest rates, and rise with both cyclical and more permanent income increases and increases in the relative cost of materials. Demographic factors such as the size and age distribution of the population are also significant determinants of house prices (p. 378) Peek and Wilcoxs study is significant because of their use of data seri es from multiple national data sources versus relying on one source While not all of the data series agreed on the explanatory variables effect (positive or negative) on housing prices ; the explanatory variables with all of the data series in agreement ( the same variable is either all positive or all negative through all of the data series) a high t statistic (rejects the null hypothesis that the coefficient value is zero) and similar coefficient values provides significant backup to Peek and Wilcox s co nclusion above. While the explanatory variables they chose to include do not explain the entire makeup of housing prices, they do provide a good insight as to the effects of income, lending costs, age and distribution of the population, unemployment and construction material costs on housing prices. Some of the explanatory variables they chose, namely construction material costs and lending costs, tend to follow national trends and the cost is similar throughout the nation. However; variables such as incom e, age and distribution of the population, and unemployment react differently at a regional or local level. For example, unemployment could be well above the national average in a certain

PAGE 28

28 city and its contribution (coefficient) and statistical significance to housing prices could be much greater than in Peek and Wilcoxs (1991) study where there was little statistical evidence of its effect. Housing Prices at the Regional Level Hwang and Quigley (2006) conduct an analysis at the regional level utilizing U.S. m etropolitan r egions. In the study they analyze the inter relationship of housing prices, vacancies and residential construction activity in response to the exogenous factors, which affect the fortunes of the regional economy and also take into account local land use and building regulations (p. 426). There analysis is of 74 Metropolitan Statistical Areas (MSA) from 19871999. They first illustrate the key relationships being explored in their analysis. They find a strong positive r elationship between current annual real price changes as a function of their lagged values and suggest that lags and slow adjustment to market conditions are crucial to understanding the course of prices (p. 427). They find little relationship between ho using prices decreasing with a vacancy rate increase and a slightly greater relationship between an increase in building permits and an increase in housing prices (p. 428). In defining their model, Hwang and Quigley (2006, p.430) argue that housing demand is a function of prices, incomes, and demographic variables and housing supply is a function of profitability. They define profitability as depending on housing prices and input prices, including the costs of labor, materials, financing, and regulations i nhibiting new construction. They define vacancy rate as a difference between aggregate supply and demand in any market period (p. 430). The data series and variables at the MSA level that Hwang and Quigley (2006) use to conduct their analysis are as fol lows: metropolitan housing price indices

PAGE 29

29 published by the U.S. Office of Federal Housing Enterprise Oversight which is defined by the weighted repeat sales method of all singlefamily houses financed through Freddie Mac and Fannie Mae, homeowner vacancy published by the U.S. Bureau of the Census, building permits published by the U.S. Bureau of the Census mortgage interest rate published by Freddie Mac, median tax rate for each metropolitan area as a percentage of house values, annual rents at the 40thHwang and Quigleys (2006) empirical results for housing prices found that an increase in new housing stock negatively impacts housing prices and that the housing stock coefficients were unaffected when the housing vacancy variable (which an increase negatively impacted housing prices as well) was removed. They argue that this suggests an independent role between new housing stock and vacancies in housing price s. They found that an increase in rental prices increases housing prices but that the rental variable coefficient was insignificantly different from zero (unable to r eject the null hypothesis that the rental price coefficient was any different from zero). A n increase in user costs negatively impacts housing prices and they found this coefficient to be very significant in all five of their models. Their lagged price variable had a coefficient of per centile of distribution from the U.S. Department of Housing and Urban Development, labor costs as average earnings per worker in the construction industry from the Regional Economic Information System (REIS) maintained by the Bureau of Economic Analysis, proprietary metropolitan data on material costs for residential construction, financing costs for housing suppliers from the DRI database, index of stringency regulation from the author Malpezzi; and three exogenous variables to include per capita income, employment and per capita transfer payments for unemployment which are all from REIS (pp.437439)

PAGE 30

30 approximately .5 which Hwang and Quigley (2006) argue signifies that half of the discrepancy between the market clearing price and the observed price is eliminated within a year (p. 440), therefore demonstrating the lag in the housing market. Utilizing the results from their regression analysis, Hwang and Quigley (2006) simulated an unexpected exogenous income shock on three MSA housing markets. Within the first year they found approximately a .1 to .4 percent increase in housing prices within the three MSAs. Houstons prices peaked at a .4 percent increase in the first year after the initial shock and dissipated to below the zero increase level w ithin ten years. San Jose and Tucson saw a lagged peak in response to the shock of approximately .8 percent and 1.3 percent increases r espectively five years from the initial shock. The prices then dissipated but were still well above the zero increase level at ten years (.6 percent and 1.2 percent respectively). Hwang and Quigley (2006) make an interesting argument to explain the differences in the three MSAs housing prices in response to the income shock. They tie it to the strong relationship between building activities and regulation (p. 446). Houstons building permits increased by 3500 in the first year after the simulated income shock. Tucsons increased by 1200 and San Joses increased by 650. The building permit increases quickly fell to status quo for all three MSAs at year three. They conclude that the simulation, based on their regression models discussed earlier, shows that a housing market with more stringent regulation has a more persistent price appreciation arising from an endogenous shock. The authors do not show the level of regulation (based on Malpezzis index of stringency regulation which was one of their variables ) for the three MSAs. They also do not indicate what the local building regulations are composed of. However; to further

PAGE 31

31 argue the impact of building regulations on housing prices, Hwang and Quigley (2006) conduct another simulation, this time on D enver, based on their regression analysis results (p. 447). They shock Denver with an exogenous increase of income and demonstrate the change in housing prices based on Denvers building regulations. They then impose San Franciscos building regulations on Denver. The impact i s dramatic as the San Francisco building regulations bring Denver a slow and steady increase in housing prices, still increasing at year 10 by slightly more than .5 percent. Denvers current regulations show an initial decrease in housing prices with only a .01 increase by year 10. Denvers current building regulations also brought higher building permits numbers as well which increased the amount of vacancies in comparison to San Franciscos regulations. Hwang and Quigley (2006) have successfully demonstrated through regression analysis of empirical MSA data that: (1) housing prices are more localized and react differently based on regulation, income, new housing stock, vacancies and user costs ; (2) the change in housing prices is not immediate when expose d to an exogenous shock and the lag in the market is dependent upon the local context ; (3) Dardia et al.s (1996, pp. 2829) argument that base closure increases vacancies and decreases home prices has statistical validity based on Hwang and Quigleys (2006) analysis of MSA data. However; Hwang and Quigleys (2006) study is conducted more at a regional level. Of course this depends on the area that the MSA is comprised of and smaller MSAs could have a more localized effect. But, micro neighborhood e xtern alities (Li & Brown, 1980) such as proximity to landfills, construction of new stadiums, construction of major arterial roads, noise levels, or even proximity to military bases, are diluted

PAGE 32

32 when engulfed in the economic activity of a large MSA when attempting to determine the components of house prices This is not to point out a flaw in Hwang and Quigleys study, as they chose to focus at the MSA level due to the consistency of data availability across many MSAs, but to point out that there are more local ized variables that Hwang and Quigley did not account for that will be discussed in the following subsection. Housing Prices at the Local Level When attempting to explain the puzzle that is the house price at the local level, researchers tend to focus more on local accessibility criteria such as proximity to the construction of a major road, a stadium, a bridge, a grocery store, a school, a river and conservation land (Li & Brown, 1980, p. 126) Researchers also focus on the proximity to congestion, noise pollution and air pollution (Li & Brown, 1980, p. 126). Li and Brown (1980) conducted a study of 781 sales of singlefamily houses in 15 suburban towns located in the southeast sector of the Boston metropolitan area that combines both the positive and negative impacts of micro proximity which they claim, at the time of the publishing of their research, had never been done before (p. 126). The authors classify micro neighborhood variables into three types: aesthetic attributes, pollution levels and proximity (Li & Brown, 1980, p. 125). They group the variables into five categories and sum the values: (1) structural and site characteristics; (2) neighborhood (census tract) characteristics; (3) local public services and costs; (4) macro accessibility to CBD; (5) micro neighborhood characteristics such as aesthetics, pollution levels, and proximity to nonresidential activities. (Li & Brown, 1980, p. 126) The structural and site characteristics include: number of rooms number of bathrooms, number of fireplaces, number of garage spaces, presence of basement, presence of a patio, and age of the structure (Li & Brown, 1980, p. 126). They provide

PAGE 33

33 both a linear and nonlinear version of number of bathrooms and age. In the nonlinear versions they square the number of bathrooms and age, hypothesizing that the number of bathrooms has a nonlinear effect on sales price and that the square of age with a positive coefficient would measure the valuation of older houses relativ e to newer houses (Li & Brown, 1980, pp. 127128). The neighborhood (census tract) characteristics measure median income, residential density (number of units per square mile), percentage of persons between 16 and 21 years old who are high school dropouts (provides measure of vandalism and crime), and air pollution levels indexed by mean values of total suspended particles (Li & Brown, 1980, p. 128). The authors hypothesize that median income will become less significant when they add the remaining neighborhood characteristics just mentioned (Li & Brown, 1980, p. 128). Local public services and costs are measured by school quality and property taxes The authors find it difficult to measure the quality of the school so they devise their own measure of quality: an input variable, expenditure per pupil; and an output variable, the standard test scores for fourthgrade pupils (Li & Brown, 1980, pp. 128129). For property taxes they measure the actual taxes paid for each unit. Macro accessibility is measured as the distance to the Boston central business district (CBD) and is intended to provide a gross measure of relative locational advantage (Li & Brown, 1980, p. 129). Micro neighborhood characteristics are a measure of aesthetic characteristics of a site and its view and of noise levels (Li & Brown, 1980) Li and Brown (1980, p. 129) devise an index from one to five lowest to highest visual quality respectively. For noise

PAGE 34

34 levels they utilize decibels on the A scale (dbA) and assume that an increase in 10 dbA is perceived as a doubling of the noise level (Li & Brown, 1980, p. 129). The authors empirical results show that the number of rooms, age of structure, number of garage spaces, number of fireplaces, a basement, presence of a patio, and land area are the most significant determinants of the sales price (Li & Brown, 1980, p. 133). The number of rooms squared coefficient demonstrates that adding rooms to a house has a significant but declining effect on housing price (Li & Brown, 1980, p. 133). The authors argue that this is most likely due to the declining marginal value of rooms and the economy of scale in housing construction (Li & Brown, 1980, p. 133). Median income was significant at the 95% confidence level but when the authors introduced aes thetic quality and other desirable attributes that are highly correlated with income the median income coefficient lowered significantly and became statistically insignificant (Li & Brown, 1980, p. 134). Li & Brown (1980) also discovered that the introduc tion of the microneighborhood variables did very little to the structural attribute coefficients which they argue is because construction costs are independent of location (p. 134). Also, initially, variables such as percentage of 1621 years old who are high school dropouts, residential density and test scores were not significant at the 95% confidence level. However, when the authors introduce the microneighborhood variables the coefficients and significance for the aforementioned variables increase, re sidential density changes from a positive to a negative sign, and the test scores variable almost became significant at the 95% level (Li & Brown, 1980, p. 134). The micro neighborhood variables have an effect on the distance to CBD coefficient as well. Wh en the model is run without the microneighborhood variables each mile from the

PAGE 35

35 CBD had an increasing effect on housing prices and the t statistic was insignificant However; when the microneighborhood variables are introduced, the CBD coefficient sign ch anges to negative and the t statistic becomes significant, indicating that each mile from the CBD decreases home prices. The air pollution coefficient has a negative effect on house prices as would be expected; however, the significance is very low (Li & B rown, 1980, p. 135). Li and Brown (1980) argue that this is because the variation in air pollution across the suburbs is small and the air pollution level is low (p. 135). They also make an interesting argument that there is a high correlation between air pollution levels and microneighborhood characteristic s where it may measure closely associated factors such as congestion, noise pollution, and visual disorder (Li & Brown, 1980, p. 135). The microneighborhood characteristics of visual quality and noise level are statistically significant with the properties with the highest visual index commanding a premium over the lowest visual index (Li & Brown, 1980, p. 135). However; noise level only measured significant at the 90% confidence level but did demonstrate a price drop for each doubling of the noise level (Li & Brown, 1980, p. 135). The coefficients for proximity, specifically closeness to the ocean, rivers, and expressway interchanges, are highly valued as revealed by their large negative values (Li & Brown, 1980, p. 135). Proximity to conservation land and schools is insignificant. Li and Brown argue that this is because most houses are within a reasonable distance of schools, and buses are provided at no charge (Li & Brown, 1980, p. 135). The second proximity group, specifically proximity to industry, commercial and major thruways, shows an interesting phenomenon. The authors found that for industry and commercial, accessibility dominates over the externality (noise and pollution) (Li & Brown, 1980, p.

PAGE 36

36 135). For thruways they found the opposite pattern of that for proximity to industries the externality dissipates slower than the positive accessibility effect (Li & Brown, 1980, p. 135). W hen examining the age and squared age coefficients, Li and Brown (1980) find that age does subtract from the house value; however, the older a house gets in the Boston suburbs, the more historical significance it takes on. In this case, at 264 years of age, the house begins to increase in value due to the squared age coefficient (Li & Brown, 1980, p. 133). Li and Brown (1980) make a significant argument through hedonic analysis of empirical data that there is certainly a local element, even a microneighborhood element to house prices. The introduction of mi cro neighborhood attributes to their latter models had a significant effect on several of their variables and changed the coefficient sign (positive or negative) to the sign that was expected. To truly explain the attributes of housing prices, researchers must take into account microneighborhood variables that also take into account accessibility and externalities immediate to the area, or even parcel being studied. Li and Brown make a compelling argument that accessibility trumps negative externalities when examining proximity to industrial and commercial entities. This compelling argument deserves a more in depth analysis as the theoretical framework of this thesis delves deeper into the effect of the closure of a military base on housing prices. Three pieces of literature dealing with the construction of major sports stadiums and the closure of military bases and their effects on housing prices will be examined in the next subsection in an attempt to get closer to the effect of the closure of a military base on singlefamily house prices.

PAGE 37

37 Proximity to a specific externality and its effect on housing prices Tu (2005) ; Dehring, Depken, & Ward ( 2006); and Hiebert (2009) st udied the effects of the announcement of the construction of the Washington Redskins stadium ( Fed Ex Field) on housing prices; the construction of the Dallas Cowboys stadium on housing prices; and the closure of Reese Air Force Base (AFB) and Red River Army Depot on housing prices; respectively. The summation of variabl es and methodology in this subsection will not be as in depth as the national, regional and local sections as key variables have already been discussed in the previous sections There are two variable s in particular in these three studies that are not inc luded in the national, regional and local examples and they are dummy variables that represent a point in time and proximity to a specific entity such as a stadium or military base (not general proximity such as proximity to commercial or industrial). Dehr ing et al. (2006); and Hiebert (2009) use the point in time dummy variable that represent s if the public announcement of the construction of the stadiums or the closure of the military bases has any influence on housing prices. They set the date of the var iable back 30 days prior to closing to represent when the contract to purchase was agreed upon by both buyer and seller. The Dehring et al. (2006) study saw movements in price but all announcement variables were statistically insignificant Only when they totaled the accumulated effects of the several stadium announc ements did they see a small effect on price at a 95% confidence level. The Hiebert (2009) dissertation analyzed the announcement of the closure of an AFB and an Army base in two different counti es of Texas and saw much greater price movement in the negative direction that was statistically significant at the 95% confidence level and 90% confidence level, with some statistically insignificant results What is interesting is that the Hiebert (2009) study

PAGE 38

38 utilized 60 observations at the most for one town, 36 for a town with fewer significant results, and achieved much higher statistical significance ; while the Dehring et al. (2006) study utilized thousands of observations with almost all of the announcement variables achieving no significance. It is this authors opinion that the difference in the two studies is that the closure of a military base represents the possibility of the movement of thousands of people out of the areas affected which potenti ally creates a large void in the demand for housing. Also, as earlier mentioned in the introduction of this thesis, Bradshaw (1999) argues that the closure of a military base is politically charged with sometimes imprecise ass ertions of dire economic consequences from the base closure. A stadium, while creating a positive or negative externality to its neighbors and also politically a charged issue, has far less of an effect on the demand for housing. Also, there could have been a lag in the markets response (as Hwang and Quigley (2006) demonstrated in their model) to the stadium announcements that Dehring et al.s (2006) research does not capture. Tu (2005) and Hiebert (2009) both use a dummy distance variable that indicates t he propertys distance from the stadium or military base respectively. In Tus (2005) initial study of the impact of the stadium, he also draws rings around the stadium at one mile radius intervals. Tu (2005) f inds that the impact of the stadium becomes st atistically insignificant outside of three miles so he label s the ring three miles out as IMPACT, the ring two miles out as IMPACT2, and the ring one mile out as IMPACT1 (Tu, 2005, p. 387). Tu (2005) then adds two interactive variables; IMPACT times DISTAN CE (I_DISTANCE) and IMPACT times DISTANCE squared (I_DISTANCE2) (p. 387). Within IMPACT Tu (2005) shows that the properties sold at a reduction in price compared to

PAGE 39

39 properties outside of IMPACT (p. 387) The interactive variable I_DISTANCE shows that each additional mile reduces the properties price a certain percentage (Tu, 2005, p.387) The interactive variable I_DISTANCE2 shows that the reduction in price for each mile is nonlinear because of its negative sign and its high significance (Tu, 2005, p. 3 87) Hieberts (2009) differs from Tu (2005) in that he does not alter the distance variable and measures more than just the distance from the parcel to the military base. Hiebert (2009) also includes dummy variables that measure the distance to major roads, hospitals, transportation nodes, schools, and colleges (p. 72). When considering the effect of the closure of a military base it is this authors opinion that Tus (2005) method of demonstrating the diminishing effect of the base on house prices as the distance from the base increases is a more effective method because it gives a perspective of how localized the effects are. The linear method that Hiebert (2009) employs demonstrates that the two base closures have an effect on house prices but does not define a boundary to the effect. Regression Methods National, regional and local variables were discusse d in the previous section in an attempt to identify pertinent variables to include in a hedonic pricing model. While there are hundreds of variable types that can be included in a hedonic pricing model, there are also many methods employed by researcher s to conduct the regression analysi s in order to estimate the variables coefficients. While this section will not cover all of the methods available to the researcher, it will evaluate the more commonly used methods in an attempt to understand and identify a method more suitable for the hedonic pricing model that will be used in this thesis.

PAGE 40

40 Hwang and Quigley (2006) estimate the coefficients for each vari able through the use of twostage least squares (TSLS) regression in an error components framework versus ordinary least squares (OLS). Garson (n.d.), a professor of Public Administration at N orth Carolina State University, states that twostage least squares regression is used to cover models which violate ordinary least squares (OLS) regression's assumption of recursivity, specifically models where the researcher must assume that the disturbance term of the dependent variable is correlated with the cause (s) of the independent variable(s). Garson (n.d.) describes the two stages as: (1) a stage in which new dependent or endogenous variables are created to substitute for the original ones, and (2) a stage in which the regression is computed in OLS fashion, but using the newly created variables. The purpose of the first stage is to create new dependent variables which do not violate OLS regression's recursivity assumption. Similar to the Hwang and Quigley (2006) analysis, Li and Brown (1980) utilize a two sta ge regression equation because of the simultaneity between sales price and its property tax (p. 131); however, in the second stage they utilize nonlinear least squares (NLS) and term this method the search method. NLS allows the authors to choose the parameters so as to minimize the sum of the squared residuals ( Li & Brown, 1980, pp. 129130). Tobler (1970) argues that Everything is related to everything else, but near things are more related than distant things. While Hwang and Quigley (2006) and Li and Brown (1980) account for data correlation and simultaneity respectively, they do not account for spatial structure in the residuals from the model (Charlton & Fotheringham, 2009, p. 3). Charlton and Fotheringham (2009) argue that this will lead t o inefficient estimates of the parameters, which in turn means that the standard errors of the parameters will be too large and that there are implications for interference where

PAGE 41

41 potentially significant parameter estimates may appear not to be so (p. 3) They continue their argument that spatial structure in the data means: that the value of the dependent variable in one spatial unit is affected by the independent variables in nearby units. This leads to parameter estimates which are both biased and inefficient. A biased estimates [sic] is one that is either too high or too low as an estimate of the unknown true value. (p. 3) Spatial heterogeneity is another characteristic of data with spatial aspects that is not accounted for in a basic regression model (Charlton & Fotheringham, 2009, p. 3). Spatial heterogeneity occurs when the relationships being modeled are not homogeneous spatially and vary across space (Charlton & Fotheringham, 2009, p. 3). In an effort to account for these spatial regression anomalies, Geographically Weighted Regression (GWR) was developed as a fairly recent contribution to modeling spatially heterogeneous processes (Charlton & Fotheringham, 2009, p. 4). The basic concept of GWR when testing a unit at location x,y is that obse rvations which are nearer that location should have a greater weight in the estimation than observations which are further away (Charlton & Fotheringham, 2009, p. 4). As the bandwidth gets larger the weights approach unity and the local GWR model approac hes the global OLS model (Charlton & Fotheringham, 2009, p. 6) In other words, as the distance increases from the unit being tested, the weighting decreases eventually to zero and the GWR model becomes a standard OLS regression model at those distances. G ao, Asami, and Chung (2002) conduct An Empirical Evaluation of Hedonic Regression Models and evaluate a simple linear regression model that does not account for spatial regression anomalies, as well as, a spatial dependency model and a GWR. A housing and land price dataset from Tokyo i s used for the illustrating the prediction power of the models (Gao et al., 2002). The results of their study indicate

PAGE 42

42 that regression models generally predicted a higher percentage of samples correctly; however, at up to the 70% mark of well predicted samples, ordinary simple linear regression was able to predict at the same error rate as the spatial regression models. It was only past the 70% mark of well predicted samples that the error rate slightly increased over spati al regression models. Gao et al., (2002) stress that a different data set could provide different results. The spatial aspect of data certainly introduces an additional aspect to consider when constructing a regression model While Gao et al. (2002) may demonstrate that simple linear regression models are as accurate as spatial regression models up to a certain point, it is important to consider their comment that a different data set may produce a different result. It is this authors opinion that the Toky o dataset may have met the OLS assumption of recursivity T herefore, the error rate would be low and the prediction ability of the simple linear regression model would be as accurate as the spatial regression model (GWR). Other data sets may show the predi ction ability of simple linear regression to be more fraught with error and GWR would become the researchers choice when dealing with spatial data.

PAGE 43

43 CHAPTER 3 METHODOLOGY This thesis examines the effect of the NAS Cecil Field base closure announce ment, the actual base closure and the subsequent redevelopment of the base on singlefamily home price s in the Jacksonville communit y surrounding the base. The closure of Cecil Field was r ecommended by the BRAC Commission to the President in July 1993 and approved by the President and Congress in the same month. It was officially decommissioned as an operational naval air station in September 1999 and closed its gates in 2000 with redevelopment efforts beginning soon after. The author chose Cecil Field for this case study for four key reasons: 1 ) Jacksonville is unique when compared to other base closures because it was home (prior to Cecil Fields closure) to three sizable naval bases and a USMC MPF logistics hub which fall within a 40 mile diameter circle centering approximately on Naval Air Station Jacksonville; 2 ) Cecil Field employed over 7,000 military and almost 1,500 civilians and was home port to 17 fixedwing squadrons of carrier aircraft with approximately 20,000 acres of real estate; 3 ) Housing s ales data to include all recorded sales back to 1992, not just Multiple Listing Service sales, wa s available through a private vendor; 4) Jacksonvilles proximity to the University of Florida greatly enhanced the authors ability to gather data and the Cit y of Jacksonville Economic Development Commission was very cooperative in giving their time and allowing the author access to closure and redevelopment documents History of Cecil Field N AS Cecil Field (now Cecil Commerce Center) opened in June 1941 on 2600 acres of land in Duval County (GlobalSecurity.org, n.d.) in the southwest of Duval

PAGE 44

44 County Following the attack on Pearl Harbor Cecil Field quickly began operations and began training replacement combat pilots for the war eventually being commissioned a Naval Auxiliary Air Station (GlobalSecurity.org, n.d.) In 1943 the base transitioned from training fighter s to becoming the principle war at sea and dive bombing training center for the Navy and was the pilots last stop before assignment to combat in either the Atlantic or Pacific fleet (GlobalSecurity.org, n.d.). After World War II Cecil Field was disestablished and reestablished several times until it was finally commissioned as a Naval Air Station on June 30, 1952 ( GlobalSecurity.org, n.d.). In the 1950s Cecil Field was chosen to be one of four bases to be used fo r the operation of jet aircraft and grew to 4,600 acres (GlobalSecurity.org, n.d.) Through the years Cecil Field continued to grow to approximately 20,000 acres and was the home to 17 fixed wing carrier aircraft squadrons (GlobalSecurity.org, n.d.) 7,000 military and almost 1,500 civilians before its closure was announced.

PAGE 45

45 Figure 31. Map showing the location of Cecil Field and proximity to other militar y bases After the closure of Cecil Field in late 1999 and early 2000, the federal government eventually turned over 16,583 acres and 642 acres to Duval and Clay Counties respectively and the development of Cecil Commerce Center in Duval County began. Curre ntly, the location of Cecil Commerce Center in southwest Duval County is mostly rural in close proximity to Interstate 295 and major north south and east west corridor s, Interstate 95 and Interstate 10 respectively. Cecil Commerce Center was directly connected to Interstate 10 in October 2009 when the Cecil Commerce Center Parkway was completed by the Florida Department of Transportation. The Cecil Commerce Center Parkway is the first section of the First Coast Outer Beltway which will connect Interstate 10 in Duval County to Interstate 95 in St. Johns County.

PAGE 46

46 Cecil Commerce Center consists of Cecil north and Cecil south which occupy over 6,000 acres and are under the co ntrol of the Jacksonville Economic Development Commission (JEDC). The aviation portion of the Cecil Commerce Center occupies an additional 6,081 acres and is run by the Jacksonville Aviation Authority (JAA). The remaining 5,000 acres is under a conservation easement and will never be developed. Due to a series of encroachment issues at NAS Oceana in Virginia Beach, Virginia (the only remaining master jet base on the east coast after the closure of NAS Cecil Field ) and the City of Virginia Beach s unwillingn ess to purchase thousands of homes and businesses in order to condemn them and relieve the encroachment, the 2005 BRAC Commission voted to close Oceana and transfer the Oceana jets to Cecil Field, reopening NAS Cecil Field The Governor of Florida and Mayo r of Jacksonville were initially supportive of this idea and testified before the 2005 BRAC Commission that they would welcome back the Navy to Cecil Field even though they had already started demolition and construction at Cecil and had spent city and state taxpayer money The public outcry to this decision was large as the majority of the public supported the further development of Cecil Commerce Center The offer to the Department of Defense was reversed by the city and state and Jacksonville remains the owner of Cecil Commerce Center today. Since Duval County has taken over Cecil Field over $200 million in public funds have been invested on capital improvements at Cecil Field, including an equestrian center, recreational facilities, and improvements made by JAA, JEA, the City of Jacksonville and FCCJ (Fishkind and Associates, 2006). In 2006 the Jacksonville Chamber of Commerce commissioned Fishkind and Associates, Inc. to conduct an

PAGE 47

47 economic analysis of Cecil Commerce Center in order to determine if the Navy moving back to Cecil Field would benefit Duval County more economically than continuing to cultivate Cecil Commerce Center. This was in response to an initiative by a proNavy advocacy group called VoteJacksonville.com who made an unsuccessful effor t to get the return of the Navy to Cecil Field added as a referendum for the cit izenry to vote on after the state and city reversed their offer to the DoD Fi shkind and Associates (2006) argue: While a large portion of this facility is currently vacant land, the Cecil Commerce Center represents a critically important and primary future economic engine for the City of Jacksonville and northeast Florida. The extensive investments and initial successes of the location, occurring in under a decade indicate the economic viability and demand for this location. Through the use of the Regional Input Output Multiplier System II (RIMS) multipliers for northeast Florida, 2003, Fishkind and Associates, Inc. conduct an economic analysis estimating the 2006 (current at the time of Fishkind publication) direct and indirect economic impact to Cecil Commerce Center and also project the analysis to 2015 and 2030. Assuming the Fishkind and Associates, Inc. analysis is accur ate; t he 2006 and 2015 tables are provided in this thesis to support the hypothesis that single family home values within a certain proximity to Cecil Field could be affected by both the closure of NAS Cecil Field and redevelopment of Cecil Field as Cecil Commerce Center

PAGE 48

48 Table 3 1 2006 Direct Economic Impact of the Cecil Commerce Center Direct Impacts (2006) Economic Earnings Employment Manufacturing (mfg) $ 94,412,788 $ 26,813,456 592 Logistics & Dist (log) $ 2,036,001 $ 872,278 22 Office (off) $ 24,727,987 $ 10,630,024 248 Pub & edu use (pu) $ 12,991,611 $ 6,327,956 269 General Aviation (GA) $ 89,290,671 $ 19,318,818 474 Mixed use (mix) $0 $0 0 Commercial (com) $ 1,188,578 $ 254,947 7 Park (prk) $0 $0 0 Military (mil) $ 41,179,093 $27,627,054 668 Recreation (rec) $0 $0 0 TOTAL $ 265,826,729 $ 81,979,599 2,280 Source: Fishkind and Associates, Inc. Table 3 2 2006 Direct & Indirect Economic Impact of the Cecil Commerce Center Direct & Indirect Impacts (20 06) Economic Earnings Employment Manufacturing (mfg) $ 186,153,695 $ 57,808,950 1,439 Logistics & Dist (log) $4,538,450 $ 1,432,327 47 Office (off) $55,252,215 $ 20,524,230 554 Pub & edu use (pu) $29,883,303 $ 11,067,553 431 General Aviation (GA) $169,152,246 $ 49,520,606 1,327 Mixed use (mix) $0 $ 0 0 Commercial (com) $2,339,477 $ 686,522 17 Park (prk) $0 $ 0 0 Military (mil) $86,336,087 $ 27,627,054 1,099 Recreation (rec) $ 0 $ 0 0 TOTAL $533,655,474 $ 168,667,242 4,914 Source: Fishkind and Associates, Inc. Table 3 3 2015 Direct Economic Impact of the Cecil Commerce Center Direct Impacts (2015) Economic Earnings Employment Manufacturing (mfg) $645,579,337 $183,346,064 4,048 Logistics & Dist (log) $82,920,768 $35,525,504 896 Office (off) $394,650,704 $169,651,754 3,958 Pub & edu use (pu) $27,045,733 $13,173,440 560 General Aviation (GA) $160,685,532 $34,765,721 853 Mixed use (mix) $143,627,688 $36,113,525 1,037 Commercial (com) $85,407,784 $18,319,763 503 Park (prk) $387,573 $160,722 6 Military (mil) $15,164,756 $6,541,140 246 Recreation (rec) $581,360 $241,083 9 TOTAL $1,556,051,234 $497,838,716 12,116 Source: Fishkind and Associates, Inc.

PAGE 49

49 Table 3 4 2015 Direct & Indirect Economic Impact of the Cecil Commerce Center Direct & Indirect Impacts (2015) Economic Earnings Employment Manufacturing (mfg) $1,272,888,779 $395,288,228 9,838 Logistics & Dist (log) $184,838,685 $58,334,760 1,911 Office (off) $881,807,532 $327,560,084 8,834 Pub & edu use (pu) $62,210,595 $23,040,260 897 General Aviation (GA) $304,402,671 $89,116,196 2,389 Mixed use (mix) $274,716,679 $89,910,933 2,319 Commercial (com) $168,108,140 $49,331,536 1,253 Park (prk) $797,199 $ 247,620 10 Military (mil) $31,794,427 $ 10,174,035 405 Recreation (rec) $1,195,799 $ 371,431 15 TOTAL $3,182,760,506 $ 1,043,375,083 27,870 Source: Fishkind and Associates, Inc. Data Dependent Variable Table 3 5 Description of Dependent Variable Used in the Hedonic Price Model Variable Description PRICE Sale price recorded for the property LOGPRICE Natural logarithm of Price The data utilized for the dependent variable, single family home sales price, is transactional real estate data for all singlefamily home sales transactions at the county level. The data is proprietary data and was obtained from Metro Market Trends, Inc. (MMT ). MMT records all of the real estate transactions recorded in all of the counties of the state of Florida. T he author was able to obtain every deed type issued for both counties singlefamily sales transactions ; however only singlefamily homes issued a warranty deed at the time of sale will be included in this study. The reason being that homes issued a warranty deed versus a quit claim deed, agreement deed or special warranty deed truly demonstrate the market value of the property Other types of deeds issued such as a quit claim deed could have been issued due to a tax foreclosure sale

PAGE 50

50 where the grantor gives no warranty to the grantee that the title is free and clear of liens or when a property is gifted from one family member to another for little to no cost. Using deed types other than warranty deeds has the potential to increase outliers and skew the market sales data in the sample. To further decrease the potential for outliers the top and bottom 2% of the total count of the s ample for each calendar year of sales data were removed. The dependent variable will be tested by the calendar year during specific years January 1992 is the farthest back that MMT kept pricing data records in the State of Florida. This will allow one calendar year of data to be tested prior to the DoD announcement of bases to be considered for closure and realignment on March 15, 1993. The following major events involving Cecil Field are noted in order from March 1993 through November 2009: Table 3 6 Major Events Involving Cecil Field Event Month/Year Initial DoD 1993 BRAC Announcement March 1993 BRAC Commission recommendation to President July 1993 Approval of BRAC Commission closure recommendation by President July 1993 Approval of President closure recommendation by Congress July 1993 Jacksonville develops goals and objectives for Cecil November 1993 Jacksonville initiates Cecil reuse planning process September 1994 Navy issues Final EIS with preferred alternative which is the City of Jacksonvilles reuse plan October 1998 NAS Cecil Field ceases operations September 1999 BRAC Commission recommends to President to close NAS Oceana, Virginia and transfer assets to and reopen NAS Cecil Field September 2005 Mayor of Jacksonville turns down DoD attempt to reopen NAS Cecil Field; Cecil Commerce Centers fate secure October 2005 Bridgestone/Firestone announces 1 million square feet distribution center at north Cecil Commerce Center June 2006 Interstate 10 off ramp and Cecil Commerce Center Parkway open October 2009 SAFT announce s construction of lithium ion battery factory at Cecil Commerce Center November 2009

PAGE 51

51 The dependent variable will be tested in the following calendar years: (1) 1992; (2) 1993 and 1994; (3) 1999 and 2000; (4) 2009. The reason for this is: (1) 1992 establishes the bases impact on singlefamily house prices prior to any knowledge that the base may close ; (2) 1993 is when the closure of the base is announced and is included to catch any impact on singlefamily house prices and 1994 is included to account for a lag in t he market (as was discussed earlier in Chapter 2 when examining Hwang and Quigley (2006)) after the closure announcement ; (3) 1999 is when the base officially ceased air operations closing in early 2000, and is included to catch any impact on singlefamily house prices and calendar year 2000 is included to account for a lag in the market after the base officially ceased air operations ; (4) 2009 attempts to capture any impact the re development of Cecil Field NAS as Cecil Field Commerce Center has had on singlefamily house prices. Explanatory Variables Table 3 7 Description of Explanatory Variables Used in the Hedonic Price models Variable Description LOTSIZE Acreage of Lot SF Square footage of house BATH Number of bathrooms (available only in data starting in 1999) ROOM Number of bedrooms (available only in data starting in 1999) AGE Property age in years AGE2 AGE squared INCOME Median household income in the census tract MED AGE Median age in the census tract PCTDROP Percent of high school dropouts in the census tract RATE Prime interest rate at time of transaction DISTROAD Euclidean distance to major or minor arterial road in kilometers DISTSCHL Euclidean distance to school in kilometers DISTWATR Euclidean distance to body of water in kilometers DISTHPTL Euclidean distance to hospital in kilometers DISTCOMM Euclidean distance to commercial landuse DISTINDU Euclidean distance to industrial landuse ZIPMIL Dummy variable 1, if the property is located in a zip code where it is known that a higher percentage of Cecil personnel lived, 0 otherwise IMPCEC Dummy variable 1, if the property is located in the 5 kilometer radius impact area of Cecil Field, 0 otherwise

PAGE 52

52 IM PCEC3 Dummy variable 1, if the property is located within a 3 kilometer radius of Cecil Field, 0 otherwise IMPCEC 1 Dummy variable 1, if the property is located within a 1 kilometer radius of Cecil Field, 0 otherwise DISCEC Euclidean distance from the property to Cecil Field I_DISCEC Interactive variable, IMPCEC times DISCEC I_DISCEC2 Interactive variable, IMPCEC times DISCEC squared Table 37. Continued Table 3 8 Vector Category and Source of Explanatory Variables Used in the Hedonic Price models Variable Vector Source of Data LOTSIZE Structural Florida Department of Revenue (parcel geometry) SF Structural MMT, Inc. BATH Structural MMT, Inc. (available only in data sets starting in 1999) ROOM Structural MMT, Inc. (available only in data sets starting in 1999) AGE Structural Florida Geographic Data Library, University of Florida AGE2 Structural ARCGIS INCOME Demo graphic U.S. Census Bureau 2000 decennial data MED AGE Demo graphic U.S. Census Bureau 2000 decennial data PCTDROP Demo graphic U.S. Census Bureau 2000 decennial data RATE Financial Wall Street Journal DISTROAD Proximity Buffer distance in 1 kilometer increments, ARCGIS DISTSCHL Proximity Buffer distance in 1 kilometer increments, ARCGIS DISTWATR Proximity Buffer distance in 1 kilometer increments, ARCGIS DISTHPTL Proximity Buffer distance in 1 kilometer increments, ARCGIS DISTCOMM Proximity Buffer distance in 1 kilometer increments, ARCGIS DISTINDU Proximity Buffer distance in 1 kilometer increments, ARCGIS ZIPMIL Zone Geometry from U.S. Census Bureau, Data of military member location from NAS Cecil Field Final Base Reuse Plan created by the City of Jacksonville IMPCEC Zone 5 kilometer buffer distance, ARCGIS IMPCEC3 Zone 3 kilometer buffer distance, ARCGIS IMPCEC1 Zo ne 1 kilometer buffer distance, ARCGIS DISCEC Proximity Buffer distance in 1 kilometer increments, ARCGIS I_DISCEC Interactive ARCGIS I_DISCEC2 Interactive ARCGIS The explanatory variables are a conglomeration of national, regional, and local attributes based on the studies by Peek and Wilcox (1991), Hwang and Quigley (2006), and Li and Brown (1980), respectively. They were selected based on the aforementioned autho rs research, what is applicable to a military base closure, and what data is readily available to the author. The structural vector is the structural

PAGE 53

53 characteristics of the property. The BATH and ROOM variables were not available in the datasets until cal endar 1999 and therefore will not be utilized in any models previous to that year. An increase in LOTSIZE, SF, BATH, and ROOM coefficients is expected to have a positive effect on the house price. An increase in AGE is expected to have a negative effect on the house price to some point in time when the effect will be positive. AGE2 is a nonlinear variable that represents the curve and point in time that AGE would become positive due to the houses historical impact to the community as argued by Li and Brow n (1980) The demographic vector is the demographic characteristics of the property. An increase in INCOME and MEDAGE is expected to have a positive impact on the house price. An increase in PCTDROP is expected to have a negative impact on the house price. This is due to the perception that in an increase in high school dropouts is causal to an increase in crime as argued by Li and Brown (1980). The financial vector is the prime interest rate that the national lending rates are tied to. An increase in RATE will have a negative impact on the house price. The proximity vector is the Euclidean distance from the property to the attribute. An increase in distance for DISTROAD, DISTSCHL, DISTWATR, DISTHPTL, DISTCOMM, and DISTINDU is expected to have a negative impact on the house price. The Li and Brown (1980) study argued that the overall net effect (proximity vs. negative externalities) of proximity to commercial and industrial uses was positive. This author is hypothesizing that the house price will react similarly to commercial and industrial uses as it did in the Li and Brown (1980) study. Essentially, the farther you get from the attribute, the more negative the impact on the house price. DISCCEC will most likely be

PAGE 54

54 positive and negative to the house price as the status of Cecil Field changes through the calendar years tested. The zone vector indicates if the property is in a specified zone spatially. The ZIPMIL zone represents known zip codes where Cecil Field personnel (military and civilian) were known to li ve in higher percentages. IMPCEC, IMPCEC3, and IMPCEC1 are buffers drawn around Cecil Field at five, three, and one kilometers respectively. This methodology mirrors the methodology utilized by Tu (2005) and will demonstrate that properties inside IMPEC wi ll sell at a reduction in price compared to properties outside of IMPCEC. The interactive variable I_DISCEC will demonstrate that each additional kilometer will reduce the properties price a certain percentage. The interactive variable I_DISCEC2 will demo nstrate that the reduction in price for each mile will be nonlinear. Model Hedonic Pricing Model In order to examine the impact of the announcement of the closure of NAS Cecil Field, the actual closure, and the subsequent redevelopment of the base as Cecil Commerce Center, a hedonic pricing model is employed. As discussed earlier in Chapter 2, the hedonic pricing model relates the price to the structural and locational attributes of the home (Quigley, 1995, p. 2) The number and type of attributes embodied in a particular property distinguish it from others and determine its price (Tu, 2005). This hedonic model is a hybrid of sorts, including national, regional, and local attributes The hedonic equation is written as P = f( S,D, F,R ,Z,I) where P is the sal e price of the property, S is a vector of structural characteristics, D is a vector of demographic characteristics, F is a vector of the financial characteristic, R is a vector of

PAGE 55

55 proximity characteristics, Z is a vector of zonal characteristics, and I is a vector of interactive characteristics. Regression Analysis of the Hedonic Pricing Model The coefficients for the hedonic model will initially be estimated by utilizing the Ordinary Least Squares function in the Spatial Statistics tools available in ArcGI S. If ArcGIS warns the author to check to ensure that the residuals are not spatially autocorrelated, the Spatial Autocorrelation ( Morans I) function in the Analyzing Patterns tools available in ArcGIS will be utilized. If significant clustering is discove red by the Spatial Autocorrelation function, the GWR function in the Modeling Spatial Relationships will be utilized to conduct the regression analysis and estimate the coefficients.

PAGE 56

56 CHAPTER 4 RESULTS This chapter presents the results of the regression analysis utilizing the hedonic pricing methodology as applied to the singlefamily housing surrounding Cecil Field and in Duval County. It begins with the difficulties of using geographically weighted regression ( GWR ) with the variables as defined in Chapter 3 and gives an overview of fine tuning the hedonic pricing model in order to better work with GWR and other discoveries It concludes with the results of the revised methodol ogy. Geographically Weighted Regression: Trials and Tribulations GWR is an extremely powerful tool to the researcher when attempting to determine the localized effects of an explanatory variable on the dependent variable being tested. The author utilized ESRIs ArcGIS to conduct both the ordinary least squares ( OLS ) and GWR regression analysis for this thesis. The author cannot attest to other software packages, but in ArcGIS, GWRs ability to give the researcher a visual picture, versus a table like OLS, of the explanatory variables effect on the dependent variable all ows the researcher to see that the explanatory variable does not act the same globally and can act differently on the dependent variable within a radius of just blocks. Variables and Variance in Their Values (Local Multic ollinearity) The localized result of GWR requires the values for the explanatory variable being applied to vary almost to the parcel level in order to avoid local multi collinearity Local multi collinearity is the spatial clustering of identical values. For example, the initial methodology for this study contains a dummy variable (ZIPMIL) for the parcels falling within a zip code identified as having Cecil personnel living there. If the parcel falls

PAGE 57

57 within such a zip code the parcel receives a value of one (1). If the parcel falls outside of such a zip code the parcel receives a value of zero (0). It was discovered very soon into the analysis process that while OLS will accept a variable with only two values, if OLS suggests testing for spatial autocorrelation and the author discovers spatial autocorrelation and attempts to run GWR, GWR will not run with a variable with only two values. This is because GWR estimates the variables coefficient based on it examining the values of the same variable for different parcels surrounding the parcel (i ts neighbors) being tested. A variable with only two values does not provide enough variance and the software keeps seeking more neighbors in order to find more variance. The maximum number of neighbors the software will seek is 1,000. Anymore than 1,000 n eighbors essentially means the variable is being applied globally to the entire geographical space (such as how OLS functions) versus locally to the parcel being tested and nullifies the reason for utilizing GWR. The same principle applies to data at the c ensus tract level such as for the proposed variables INCOME, MEDAGE and PCTDROP. The parcels that fall within the same tract will all have the same values for the aforementioned variables respectively. The researcher has to direct the software to search beyond the tract to adjacent tracts to find differing values for each of the variables. This minimizes the local intent of GWR. When setting up buffers to determine proximity the author hypothesized that the same issue could surface depending on the width of the buffer. The author did not necessarily experience such an issue in this study but in order to insure enough variance in the proximity variables DISTROAD, DISTSCHL, DISTWATR, DISTHPTL, DISTINDU, DISTCOMM and DISCEC, the buffers were replaced with 5 meter grid

PAGE 58

58 matrices for each variable. An additional grid matrix was added to replace the dummy variable value in ZIPMIL The new grid matrix measures the parcels distance from each zip codes centroid. It was discovered that the matrices were much quicker to generate than the previously created buffers and have the potential to work better with GWR After replacing the buffers with the grid matrices, the author discovered an almost insurmountable issue with the proximity variables and local multicollinearity When GWR is used to calculate the regression model it is already accounting for the spatial aspect of the variable. That plus the proximity measurement for the variable causes a local multicollinearity that quite fr equently does not allow the GWR function in ArcGIS to proceed. It is a painstaking process to attempt to isolate which proximity explanatory variables or combination of variables is causing the program not to run. The author started to do so with the 1992 dataset but realized there was just not enough time in the span of this study to do so. For this reason, no proximity or zone variables are included in the GWR regression analysis. Variables and Global Multicollinearity A product of the Ordinary Least Squares function of ArcGIS is the variance inflation factor (VIF) for each variable. The VIF measures the redundancy among explanatory variables (ARCGIS.COM, n.d.). For example, the AGE and AGE2 variables AGE represents the AGE of the structure. AGE2 repres ents the squared age of the structure. An initial OLS regression that includes both variables indicate s a high VIF for both variables Any variable with a VIF over 7.5 must be discarded due to global multicollinearity and the possibility of inflating the v alue of the variables coefficient.

PAGE 59

59 Figure 4 1. Initial 1992 OLS Results As can be seen in the initial OLS results for the 1992 Duval dataset all of the variables with squared values significantly raise the VIF value due to redundancy. The squared variables were initially included to indicate a curve of the nonsquared variable. The removal of the squared variables significantly reduce s the VIF for their counterpart variables to within normal limits. Variables with a VIF under 7.5 can be used in GWR For this reason the squared variables were removed for both OLS and GWR. Issues Not Related to GWR Impact Zones The author set up impact zones to replicate the methodology used by Tu (2005) when he identified the farthest impact that was statistically significant that the Washington Redskins stadium had on surrounding house prices. This study was initially set up with arbitrary impact zones at one, three, and five kilometers from Cecil Field. The author then decided to forgo the arbitrary impact zones and test Tus (2005) methodology setting up impact zones up to 13 kilometers from Cecil Field in two kilometer increments using the 1992 dataset. The reason the farthest zone is 13

PAGE 60

60 kilometers is because NAS Jacksonville is located due east of Cecil Field abo ut 13 kilometers away. It was discovered that the three kilometer impact zone was statistically significant as were the seven, nine, and eleven kilometer impact zones. The three latter mentioned impact zones are all closer to NAS Jacksonville rather than C ecil Field. The author decided that it was not clear whether NAS Jacksonville or Cecil Field was having greater influence on the farther impact zones so the impact zones were removed from the methodology of this study. Structural Variables As mentioned ear lier in Chapter 3, the proprietary sales data does not include the number of bathrooms (BATH) and bedrooms (ROOM) until 1998. When formatting the 1999 dataset the author discovered that only twothirds of the warranty deed sa mples in the dataset contained the number of bathrooms and bedrooms which means that just over 3,000 samples would have had to be deleted. It was decided by the author to forgo including BATH and ROOM as structural variables starting with the 1999 dataset. Percent of High School Dropout s in Census Tract ( PCTDROP ) Variable When formatting the 2000 decennial U.S. Census Bureau data for number of 1825 year olds that have not obtained their high school diploma the researcher noticed an anomaly in the data for two tracts when calculating the dropout percentage rate. Two of the tracts had a higher number of 1825 year olds who had not obtained their high school diploma than the population for 1825 year olds in the tract. For this reason the PCTDROP variable was removed from the model. Prime I nterest Rate at Time of Transaction ( RATE) Variable The prime interest rate (RATE) variable had to be removed because the OLS regression would not run with a variable that had no variance. Being a national level

PAGE 61

61 variable, the interest rate was the same for every parcel in the specific dataset being tested. This was identified as an issue early on. All of the parcels in the 1992 dataset test run had a RATE of six (6). When the author altered just one parcel from a RATE of 6 to a RATE of 6.1 the OLS regression then ran. It is impossible for the software to estimate a coefficient for a variable that that does not vary. Central Business District It was discovered during this study that the central business district (CBD) was having an effect on singlefamily house sales prices. For this reason a proximity explanatory variable called DISTCBD was added to the hedonic pricing model. The Revised Hedonic Pricing Model Because of all of the earlier mentioned reasons, the hedonic pricing m odel was revised. Due to a statistically significant JarqueBera statistic (indicates spatial autocorrelation) in OLS for every dataset and statistically significant Morans I for clustering of the residuals in the 1992 dataset, the author decided that OLS and GWR would be run for every dataset. The OLS regression includes all of the revised hedonic pricing model explanatory variables and the GWR includes only the variables that do not account for proximity. Table 41 Revised Explanatory Variables Variable OLS GWR LOTSIZE Yes Yes SF Yes Yes AGE Yes Yes INCOME Yes Yes MEDAGE Yes Yes DISTROAD Yes No DISTSCHL Yes No DISTWATR Yes No DISTHPTL Yes No DISTCOMM Yes No DISTINDU Yes No

PAGE 62

62 DISTCEC Yes No DISTZIP Yes No DISTCBD Yes No Table 41. Continued Results Using Revised Methodology The following results use the revised hedonic pricing model just discussed is broken down by the year of the dataset tested. For all datasets the dependent variable PRICE is in dollars, therefore the coefficients are in dollars. The results discussed in this chapter are brief, touching on the results of the OLS and GWR regression analysis. The authors interpretation of the resu lts will follow in Chapter 5. Zip Codes with Cecil Personnel Figure B 1 is a graphic depicting which zip codes Cecil Field personnel primarily lived at and was derived from the NAS Cecil Field Final Base Reuse Plan (1996) published by the Cecil Field Devel opment Commission. The two main zip codes where Cecil Personnel lived at the time of the publication of the Final Base Reuse Plan were 32210 and 32244 with Cecil Field personnel populations of approximately 1,509 ( 20.3% of the Cecil personnel) and 1,420 (1 9.1% of the Cecil personnel) respectively plus their dependents. The 2005 U.S. Census Bureau population for 32210 was 60,807 and 32244 was 54,451. Both Cecil Field zip code populations represent approximately 2.5% of the total population for each zip code. These two zip codes will hereinafter be referred to as the main zip codes. The main zip codes are extremely close to NAS Jacksonville and provide easy access to Cecil Field, NAS Jacksonville, the CBD, and Interstates 95 and 10. When Cecil Field was stil l operational as a NAS, NAS Jacksonville provided dependents with more robust services than NAS Cecil Field. The

PAGE 63

63 main Naval Hospital for the south east is at NAS Jacksonville and the larger Exchange (department store) and Commissary are at NAS Jacksonville. 1992 Dataset The 1992 dataset is representative of the Jacksonville housing market while NAS Cecil Field was an operational jet base The OLS results are shown in Figure 42. Figure 42. 1992 OLS Results Due to a statistically significant Koenker (BP) v alue the robust estimates must be used. The adjusted R Squared is .671 which means that approximately 67% of the housing price story is told with the current model. The LOTSIZE variable is statistically significant to the .00 level (100% confidence ) that an increase in the s ize of the lot will result in a price increase of $301.71 for each acre increased. The A GE variable is statistically significant to the .00 level that an increase in the age of the structure by one year will result in a price decrease of $333.44 for each year of increase in structure age.

PAGE 64

64 The SF variable is statistically significant to the .00 level that an increase in the square footage of the structure by one square foot will result in a price increase of $44.28 for each square foot i ncrease. The INCOME variable is statistically significant to the .00 level that an increase by one dollar of the median income in the census tract will result in a price increase of $0.29 for each dollar increase in median income. The MEDAGE variable is st atistically significant to the .00 level that an increase in the median age of the census tract by one year will result in a price increase of $608.75 for each year increase. The DISTCEC variable is statistically significant to the .75 level (only 25% conf idence ), therefore, the author could not reject the null hypothesis that the coefficient was different from zero (0) and that the distance from Cecil field had any effect on the 1992 single family house prices. The DISTCBD variable is statistically significant to the .00 level that an increase in the distance from the central business district by one meter will result in a price increase of $0.45 for each one meter increase in distance from the CBD. The DISTINDU variable is statistically significant to the .00 level that an increase in the distance from an industrial land use by one meter will result in a price increase of $1.02 for each one meter increase from the industrial land use. The DISTCOM M variable is statistically significant to the .11 leve l (89% confidence), therefore, the author could not reject the null hypothesis that the coefficient was different from zero (0) and that the distance from a commercial land use had any effect on the 1992 single family house prices. The DISTHPTL variable is statistically significant to the .00 level that an increase in the distance from the hospital by one meter will result in a price decrease of $0.66 for each one meter increase in distance from the hospital. The DISTSCHL variable is statistically significa nt to the .10 level ( 90% confidence), therefore, the author

PAGE 65

65 could not reject the null hypothesis that the coefficient was different from zero (0) based on a statistical significance to the .05 level as the software is set for. However, the t statistic is 2.02, just barely over the threshold of 2.0 which leads the author to believe that an increase in the distance from a school by one meter will result in a price decrease of $0.85 for each one meter increase in distance from the school at a 90% confidence l evel. The DISTWATR variable is statistically significant to the .00 level that an increase in the distance from a major water body by one meter will result in a price decrease of $0.87 for each one meter increase in distance from the major water body. The DISTROAD variable is statistically significant to the .00 level that an increase in the distance from a major or minor arterial road by one meter will result in a price increase of $3.79 for each one meter increase in distance from the major or minor arter ial road. The DISTZIP variable is statistically significant to the .42 level (only 58% confidence), therefore, the author could not reject the null hypothesis that the coefficient was different from zero (0) and that the distance from the centroid of any z ip code identified as having Cecil Field personnel living there had any effect on the 1992 single fami ly house prices. The GWR result maps for the 1992 dataset are located in Appendix C. Figure 43 demonstrates that each parcel was compared to 700 of its neighbors and that the R Squared Adjusted is .447, or in other words the five variables tell 45% of the housing price story Figure 43. 1992 GWR Results

PAGE 66

66 Recall that the AGE coefficient was statistically significant in the OLS results and was negative for each year the age of the structure increased. The AGE coefficient in GWR varies by area and is negative in most areas but is positive just west of the CBD where predominantly older structures are clustered. This is because the older structures could b e considered historic and actually gain value with age. The SF coefficient in OLS was positive and statistically significant. The GWR results in the main zip codes where Cecil personnel lived and all of Duval County depict a positive increase in house pric es with an increase in square footage. The LOTSIZE coefficient in OLS was positive and statistically significant. The GWR results in the main zip codes depict a neutral to increase in house prices if the size of the lot were to increase in those areas. The INCOME coefficient in OLS was positive and statistically significant. The 2000 decennial U.S. Census Bureau median income for the main zip codes ranges from $29,000 to $50,000. The GWR results in the main zip codes depict a neutral to increase in house p rices if the median income were to increase in those areas. The MEDAGE coefficient in OLS was positive and statistically significant. The 2000 decennial U.S. Census Bureau median age for the main zip codes is very diverse and represents the full range of m edian ages. The GWR results in the main zip codes depict a neutral to positive increase in house prices if the median age were to increase in those areas. 1993 Dataset The 1993 dataset is representative of the Jacksonville housing market when the announcem ent was made that NAS Cecil Field was selected by the Base Closure and Realignment Committee and approved for closure by the President and Congress. The OLS results are shown in Figure 44

PAGE 67

67 Figure 44. 1993 OLS Results Due to a statistically significant Koenker (BP) value the robust estimates must be used. The adjusted R Squared increased slightly from .671 in 1992 to .697 in 1993. Similar to the 1992 dataset, the coefficients for DISTCEC, DISTSCHL, and DISTZIP are all s tatistically insignificant due to thei r robust t S tatistics being under two (2). Though it is still statistically insignificant, the robust t Statistic for DISTCEC did increase from .75 to 1.5 and the coefficient increased slightly as well. The DISTCOMM coefficient went from being statistically insignificant in 1992 to statistically s ignificant at the .00 level (100% confidence ) in 1993. An increase in distance from a commercial land use will increase the house price by $3.71 for each meter in increased distance. The DISTINDU coefficient decreased from $1.02 in 1992 to $0.86 in 1993. T he LOTSIZE coefficient increased drastically from $301.70 in 1992 to a $2, 988.08 per acre increase in 1993.

PAGE 68

68 Another drastic change is that the DISTROAD coefficient went from positive to negative and indicates that an increase in distance from a major or mi nor arterial road will decrease the house sale price by $1.10 for each one meter increase in distance from the arterial r oad. The remaining coefficients saw a slight increase or decrease in their coefficient value; however, all of their signs and statistic al significance remain unchanged. The GWR result maps for the 1993 dataset are located in Appendix D. Figure 45 shows that each parcel was compared to 700 of its neighbors and that the R Squared Adjusted decreased from .447 in 1992 to .295 in 1993. Figure 45. 1993 GWR Results The AGE coefficient remained statistically significant in the OLS results and was negative for each year the age of the structure increased. The AGE coefficient in GWR varies by area and remained negative in most areas but remained positive just west of the CBD where predominantly older structures are clustered. The high and low coefficient values decreased slightly in comparison to the 1992 values The SF coefficient in the OLS results remained positive and statistically significa nt. The GWR results in the main zip codes and all of Duval County still depict a positive increase in house prices with an increase in square footage. The high and low coefficient values increase d slightly in comparison to the 1992 values. The LOTSIZE coef ficient in the OLS results remained positive and statistically significant ; however, it was mentioned

PAGE 69

69 earlier that the coefficient increased drastically The GWR results in the main zip codes changed from a neutral to positive increase in house prices in 1992 to a neutral to negative decrease in house prices in 1993 if the size of the lot were to increase in those areas. Similar to the OLS results, t he high and low coefficient values significantly increased and significantly decreased respectively which co uld be due to the significantly lower R Squared Adjusted which indicates missing variables in the model. The LOTSIZE coefficients could have significantly changed to compensate for variables that were not defined in the model. The INCOME coefficient in the OLS results remained positive and statistically significant. The GWR results in the main zip codes still depict a neutral to positive increase in house prices if the median income were to increase in those areas. The influence area for the positive increase in house prices has grown farther south than the main zip codes; more toward NAS Jacksonville. Also, the area west of Cecil Field has shifted from neutral to a positive increase in house prices if the median income increased in that area. The MEDAGE coefficient in the OLS results remained positive and statistically significant. The GWR results in the area of the main zip codes depict ed a neutral to positive increase in house prices in 1992 but changed to a neutr al to negative decrease in house prices in 1993 if the median age were to increase in those areas. 1994 Dataset The 1994 dataset is representative of the Jacksonville housing market a year after the NAS Cecil Field closure announcement was made in order t o attempt to capture any lag in the housing market reaction to the announcement The OLS r esults are shown in Figure 46.

PAGE 70

70 Figure 46. 1994 OLS Results Due to a statistically significant Koenker (BP) value the robust estimates must be used. The adjusted R Squared increased almost unnoticeably from .697 in 1993 to .698 in 1994. In comparison to the 1993 dataset DISTCEC and DISTZIP remain statistically insignificant due to their robust TStatistic being under two (2). DISTSCHL became statistically significa nt with a robust TStatistic at 3.49 and a probability at the .00 level (100% confidence). DISTWATR was statistically significant in 1993 but became statistically insignificant in the 1994 dataset. The coefficients for DISTSCHL, DISTCOMM, and DISTINDU nea rly doubled in value when compared to 1993. While LOTSIZE experienced a drastic increase in 1993, it experienced a drastic decrease in 1994 from $2,988.08 to $1,340.02 per acre increase in lot size. LOTSIZE also became statistically insignificant in 1994 using the robust TStatistic. It remained statistically significant using the nonrobust T Statistic, however, as mentioned earlier, the

PAGE 71

71 statistically significant Koenker (BP) value requires the researcher to use the robust values. All of the signs for the statistically significant coefficients remained the same as 1993. It is interesting to note that the SF robust t Statistic has remained very high through all three datasets so far suggesting that square footage contributes the most to a house s price; not in dollars but in consistency of the estimated coefficient statistical significance, and surety that an increase in a houses square footage will increase its value. The GWR result maps for the 1994 dataset are located in Appendix E. Figure 47 shows that each parcel was compared to 700 of its neighbors and that the R Squared Adjusted increased drastically from .295 in 1993 to 738 in 1994 with no additional variables added and the neighbors remaining at 700. Figure 47 1994 GWR Results The AGE coefficient remained statistically significant in the OLS results and was negative for each year the age of the structure increased. It is interesting to note that while the age coefficient decreased from 1993 to 1994 in the OLS results, the high and low coefficient values increased in comparison to 1993 by as much as $128.00 for the high coefficient value. The GWR patterns for AGE remain the same between 1993 and 1994 but the 1994 patterns contain more intense reds indicating higher coefficient values for each year increase in the structure s age. The SF coefficient in the OLS results remained positive and statistically significant. The GWR results in the main zip codes and all of Duval County still depict a positive increase in house prices with an

PAGE 72

72 incre ase in square footage; however, the northern area of the main zip codes and directly on the east side of Cecil Field experienced a decrease in the value of the SF coefficient while the area directly west of NAS Jacksonville experienced an increase in the v alue of the SF coefficient. The high and low coefficient values increased slightly in comparison to the 1993 values. The LOTSIZE coefficient in the OLS results decreased dramatically and lost its statistical significance when using the robust values. The G WR results in the main zip codes changed from a neutral to negative decrease in house prices in 1993 to a neutral to very positive increase in house prices in 1994 if the size of the lot were to increase in those areas. The area outside of Cecil Field saw no change and remains negative. T he fact that the LOTSIZE variable is statistically insignificant in the OLS results brings into question the accuracy of the LOTSIZE GWR results. The INCOME coefficient in the OLS results remained positive and statistically significant. The GWR results in the main zip codes still depict a neutral to positive increase in house prices if the median income were to increase in those areas ; however, the positive influence is much less intense The patterns remain relatively simil ar to the 1993 patterns, but the area west of Cecil Field has shifted back to a neutral effect on house prices if the median income increased in that area. The MEDAGE coefficient in the OLS results remained positive and statistically significant. The GWR r esults in the area of the main zip codes depicted a neutral to negative decrease in house prices in 1993 but changed to a neutral to positive increase in house prices in 1994 if the median age were to increase in those areas. Also, the area west of Cecil Field saw a dramatic change from a negative to a positive influence on house prices with an increase in the median age.

PAGE 73

73 1999 Dataset The 1999 dataset is representative of the Jacksonville housing market in the year that NAS Cecil Field officially ceased air operations. The OLS results are shown in Figure 48. Figure 48. 1999 OLS Results Due to a statistically significant Koenker (BP) value the robust estimates must be used. The adjusted R Squared decreased from .698 in 1994 to .684 in 1999. Had the nonro bust estimates been allowed, all of the variables would have been statistically significant. However, because robust estimates must be used, LOTSIZE changed to an unexpected negative sign and remained statistically insignificant and AGE changed from being statistically significant in 1994 to statistically insignificant in 1999. DISTZIP, DISTWATR, and DISTCEC all changed from being statistically insignificant in 1994 to being statistically significant in 1999. DISTWATRs sign made an unexpected change

PAGE 74

74 from n egative to positive meaning that an increase in distance from a major water body will increase the house price. DISTCEC and DISTZIP both became significant on the year that NAS Cecil Field ceased operations. Both coefficient values were positive indicating that an increase in distance from Cecil Field and from the centroid of a zip code with known Cecil Personnel living in it will increase a singlefamily houses sales price. MEDAGE, DISTROAD, and DISTCBD nearly doubled their values when compared to 1994. T he GWR result maps for the 1999 dataset are located in Appendix F. Figure 49 shows that each parcel was compared to 700 of its neighbors and that the R Squared Adjusted decreased from .738 in 1994 to 729 in 1999. Figure 49 1999 GWR Results The AGE coefficient became statistically insignificant in the OLS results. In the GWR results the high coefficient value in creased from $132.00 in 1994 to $943.00 in 1999. The GWR patterns for AGE remain the same between 1994 and 1999 but the 1999 patterns contain less intense reds indicating neutral coefficient values for each year increase in the structures age. The SF coefficient in the OLS results remained positive and statistically significant. The GWR results in the main zip codes and all of Duval Co unty still depict a positive increase in house prices with an increase in square footage; however, the area of the main zip codes and the area surrounding the circumference of Cecil Field up to NAS Jacksonville experienced a decrease in the value of the SF coefficient The high and values increased by $25.00 per square foot in

PAGE 75

75 comparison to the 1993 value. The LOTSIZE coefficient in the OLS results remained statistically insignificant when using the robust values. The high value in the GWR results doubled t o $166,000. The results in the main zip codes remain neutral to a positive increase in house prices in 1999 if the size of the lot were to increase in those areas. The area outside of Cecil Field saw no change and remains negative. The fact that the LOTSIZE variable is statistically insignificant in the OLS results and its sign in the OLS results is opposite of what it should be brings into question the accuracy of the LOTSIZE GWR results The INCOME coefficient in the OLS results remained positive and stat istically significant. The GWR results in the main zip codes still depict a neutral to positive increase in house prices if the median income were to increase in those areas; however, the positive influence is less intense. The area n orthwest of Cecil Field increased from a neutral to a positive increase in house prices if the median income were to increase in that area. This area has consistently flipflopped from neutral to positive and back with each dataset tested since the initial 1992 dataset. T he MEDAGE coefficient in the OLS results remained positive and statistically significant. The high coefficient in the GWR MEDAGE results increased drastically from $5,300 in 1994 to $49,000 in 1999. The GWR results in the area of the main zip codes depicted a neutral to positive increase in house prices in 1994 and remained that way in 1999. Also, the area west of Cecil Field that saw a dramatic change from a negative to a positive influence on house prices in 1994 shifted to the northwest edge of Cecil Field in 1999. 2000 Dataset The 2000 dataset is representative of the Jacksonville housing market in the year that NAS Cecil Field ceased to exist as a NAS and is one year after air operations officially ended. The purpose of this dataset is to capture any lag in the market in

PAGE 76

76 reaction to the end of air operations at Cecil Field. The OLS results are shown in Figure 4 10. Figure 410. 2000 OLS Results Due to a statistically significant Koenker (BP) value the robust estimates must be used. The adjusted R Squared decreased from .684 in 1999 to .674 in 2000. LOTSIZE continues to be statistically insignificant and have the opposite sign than expected. DISTINDU and DISTSCHL changed from being statistically significant in 1999 to statistically insignificant in 2000 DISTWATR continues to have the opposite sign expected. No statistically significant variables experienced major changes in their coefficients values with exception of DISTCOMM which increased by $8.00 and seems to fluctuate in each dataset

PAGE 77

77 The GWR res ult maps for the 2000 dataset are located in Appendix G. Figure 411 shows that each parcel was compared to 7 50 of its neighbors, an increase of 50 neighbors from 1999, and that the R Squared Adjusted increased from .740 in 1999 to 795 in 2000. Figure 4 11. 2000 GWR Results The AGE coefficient became statistically significant in the 2000 OLS results. In the GWR results the high coefficient value decreased from $943.00 in 1999 to $769.00 in 2000. The GWR patterns for AGE remained the same between 1999 a nd 2000 and the overall results are relatively unchanged between the two datasets. The SF coefficient in the OLS results remained positive and statistically significant. The GWR results in the main zip codes and all of Duval County remain relatively unchanged from 1999 to 2000. There was only a slight increase in the SF high coefficient from $100.00 in 1999 to $112.00 in 2000. The LOTSIZE coefficient in the OLS results remained statistically insignificant when using the robust values. The high value in the GWR results increased from $166,000 in 1999 to $219,000 in 2000. The results in the main zip codes and all of Duval County remain relatively unchanged from 1999 to 2000. The fact that the LOTSIZE variable is statistically insignificant in the OLS results and its sign in the OLS results is opposite of what it should be continues to bring into question the accuracy of the LOTSIZE GWR results. The INCOME coefficient in the OLS results remained positive and statistically significant. The GWR r esults in the main zip codes still depict a

PAGE 78

78 neutral to positive increase in house prices if the median income were to increase in those areas. The area of positive coefficients northwest of Cecil Field decreased in actual area, not quite going back to a neutral state. The MEDAGE coefficient in the OLS results remained positive and statistically significant. The high coefficient in the GWR MEDAGE results decreased from $49,000 in 1999 to $30,000 in 2000. The GWR results throughout Duval County remained relat ively unchanged. 2009 Dataset The 2009 dataset is representative of the Jacksonville housing market nine years after the closing of NAS Cecil Field and nine years into the redevelopment of Cecil Field. The purpose of this dataset is to capture any effects of the redevelopment of Cecil Field as Cecil Commerce Center. The OLS results are shown in Figure 412. Figure 412. 2009 OLS Results

PAGE 79

79 Due to a statistically significant Koenker (BP) value the robust estimates must be used. The adjusted R Squared decreas ed from .674 in 2000 to .587 in 2009. LOTSIZE, DISTINDU, and DISTSCHL continued to be statistically insignificant. INCOME, DISTCOMM, DISTWATR, DISTROAD, and DISTZIP changed from being statistically significant in 2000 to statistically insignificant in 2009. The AGE variable unexpectedly changed signs from negative to positive which is contrary to the hypothesized negative sign. An explanation for the drastic decrease in statistically significant variables could be the drop in the number of observations from 9818 in 2000 to 1748 in 2009 which reflects the national recession in the real estate market. The low adjusted R Squared illustrates a lack of variables. The GWR result maps for the 2009 dataset are located in Appendix H. Figure 413 shows that each parce l was compared to 1000 of its neighbors, an increase of 2 50 neighbors from 2000, and that the R Squared Adjusted decreased from .795 in 2000 to 661 in 200 9 Figure 413. 2009 GWR Results The AGE remained statistically significant in the 2009 OLS results In the GWR results the high coefficient value remained unchanged and the low coefficient value increased by $2,300 to $160.00. The GWR coefficient pattern for AGE changed drastically west of the St. Johns River as the entire area west of the river cont ains neutral to negative coefficients. The SF coefficient in the OLS results remained positive and statistically significant. The GWR results in the area of the main zip codes show an increase in

PAGE 80

80 positive coefficients for an increase in SF. The high and low coefficients remained relatively unchanged. The LOTSIZE coefficient in the OLS results remained statistically insignificant. The low value in the GWR results decreased drastically from $50,000 in 2000 to $289,000 in 2009. The coefficients in the main z ip codes remained relatively unchanged from 2000 to 2009 The fact that the LOTSIZE variable is statistically insignificant in the OLS results and its sign in the OLS results is opposite of what it should be continues to bring into question the accuracy of the LOTSIZE GWR results The INCOME coefficient in the OLS results changed from statistically significant to statistically insignificant. The GWR results in the main zip codes still depict a neutral to positive increase in house prices if the median incom e were to increase in those areas. The area of positive coefficients northwest of Cecil Field decreased in actual area, not quite going back to a neutral state. The MEDAGE coefficient in the OLS results remained positive and statistically significant. The area immediately northeast of Cecil Field, the area of the main zip codes, and the area immediately west of NAS Jacksonville all saw an increase in positive coefficients for MEDAGE. The rest of Duval County remained relatively unchanged.

PAGE 81

81 CHAPTER 5 DISCUSSION This chapter analyzes the regression results in more detail as they pertain to Cecil Field and its effects on Duval County house prices. E ach variable will be discussed in relation to the OLS results. If the variable was included in the GWR analysis the GWR results will be discussed as well. Acreage of Lot ( LOTSIZE ) Variable In previous regression studies conducted by other authors the size of the lot was generally a key variable representing a large portion of the house price. In the 1992 and 1993 datasets the LOTSIZE variable was trustworthy as an indicator of house price due to it having a global positive effect on house prices and its statistical significance. However, in 1994 t he variable lost its statistical significance when using the required robust values. In 1999, 2000 and 2009 the sign changed from positive to negative which does not make sense globally. A general rule is to discard any variables found to be statistically insignificant in the OLS results prior to running GWR. The author made the decision keep the variables consistent that were run in the GWR equation to see if there were any patterns that could be attributed to Cecil Field. The highly variable high and low coefficient values for LOTSIZE cannot be trusted as the values s eemed to have become inflated when the variable became statistically insignificant in the OLS results. The pattern around the CBD remained fairly consistent in that an increase in the size of the lot in the central business district would increase the pric e of the house. This makes sense as the lots are generally constrained within the CBD. The pattern at the main zip codes identified earlier seemed inconsistent and it is difficult to attribute the patterns to any action at

PAGE 82

82 Cecil Field. The pattern surrounding Cecil Field remained consistent through all of the datasets in that an increase in the size of the lot would have a neutral to negative effect on the house price. This makes sense as Cecil Field is in a rural location. The negative coefficient also mak es sense as an increase in the size of a lot in a rural area does not necessarily mean that the price will increase proportionately to the increase in the size of the lot. While the size of the lot may play a large factor in the price of a house, after rev iew of the OLS and GWR results it is not possible to utilize the LOTSIZE variable as an indicator that the actions at Cecil Field had any effect on surrounding house prices. Previous research by other authors reviewed in Chapter 2 indicates that the additi on of more structural and local variables often settles the tendency for a variable to switch signs and bounce between statistical significance and insignificance. While the author would have liked to add additional structural variables, the data was not available to do so. Property Age in Years ( AGE) Variable With the exception of the 1999 dataset, the AGE v ariable was consistent in terms of being statistically significant; however, the sign of the variable unexpectedly switched to positive in 2009. Prev ious research reviewed in Chapter 2 indicates that an increase in the age of the structure generally decreases the price of the house up to a certain point where the house is deemed historic and the coefficient slowly curves upward in the positive direct ion. Researchers often square the age of the structure to determine at what point the age of the structure actually increases the houses value, if there is any increase at all. The squared values were removed from this studys pricing model for earlier di scussed reasons. The global coefficients in the OLS results make sense with

PAGE 83

83 the exception of the sign change in 2009 which could be explained by the lack of structural variables as was discussed with the variable LOTSIZE. The increase from the 300s to under 100 in 1999 and 2000 could be a result of the housing boom that the nation experienced when investors were buying old houses, renovating them and flipping them for a quick profit. When reviewing the GWR results for AGE the focus is really around the CBD and directly across the St. Johns River from the CBD in the San Marco area. It is evident by reviewing the structure ages in the AGE result maps that these two areas contain the older structures in Jacksonville and the age of these structures increases their value. The spike in the high coefficient value in 1999 can be attributed to the variables statistical insignificance in the OLS results or to the housing boom; however, the 2000 datasets high coefficient value only decreased by 200 so the latter i s assumed to have influenced the value. There seems to be a cloud of light red that gravitates west toward Cecil Field from NAS Jacksonville as the dataset years progress. This cloud represents neutral to negative influences on the house pr ice. The yello w and green surrounding Cecil Field is all negative influence on the house prices in regard to the AGE variable. This advancing cloud could be attributed to actions at Cecil Field or could be attributed to the housing boom in 1999 and 2000. The red almost disappears in every area west of the St. Johns River in 2009. This is most likely attributed to the cooling of the housing market. Square Footage of House ( SF ) Variable The SF variable is the most stable variable in the model and its OLS result is consiste ntly statistically significant with a consistent positive sign. The value of the coefficient in the OLS results consistently increases through each dataset and generally

PAGE 84

84 follows the regional real estate market trend; however, the 2009 value may be overtly high due to the statistical insignificance of the other variables in the model. It is important to note that research reviewed in Chapter 2 indicates that generally the square footage influence on the house price is on a curve and the influence of square f ootage is less on the price as the square footage of the house increases. This may or may not have been evident had the square footage squared variable (SF2) been kept in the pricing model. Both the high and low values are consistently positive in the GWR results and the high value generally trends upward through the dataset years following the regional market trend The GWR result maps for each dataset year are showing a general trend opposite to what previous research ers have found using OLS regression. The Duval County results are showing that the SF variable has a greater potential to increase the house price in higher square footage homes east of the St. Johns River and in the immediate vicinity of NAS Jacksonville than in lower square footage homes For the 1993 and 1994 datasets (announcement of the closure) Cecil Field shows a slightly higher increase in price in comparison to 1992 if the square footage was increased in a house around Cecil Field. The 1993 dataset indicates that on the northeast corner of Cecil Field there is even more of a positive price impact should the square footage of the home increase. The main zip codes experience a more positive impact toward the southwest of 32210 which is right next to NAS Jacksonville. The 2009 dataset dem onstrates a more positive impact in main zip code 32205 which borders Interstate 10 and has Normandy Boulevard running through the center providing a direct route to the Cecil Commerce Center. With 2009 being a cool housing market,

PAGE 85

85 there is a real possi bility that the potential for redevelopment at Cecil Field may have contributed to the SF variables positive influence on price in this area. The other possibility is that NAS Jack sonville is influencing the prices within this area. Median Household Incom e in the Census Tract ( INCOME ) Variable The INCOME variable was consistently statistically significant in OLS results with the exception of the 2009 dataset. It consistently had a positive influence on price in the OLS results and the coefficient values st ayed fairly consistent as well. The consistent coefficient values may have been influenced by the fact that only the 2000 decennial U.S. Census Bureau data was used to define the median income. Had the median income varied according to the year of the dataset being tested, the coefficient results may have varied more depending on if there was a great variance in the median income for the census tract being tested. The only reason the author can think of that would have changed the median income greatly is t he recent national recession which would have influenced the 2009 dataset or possibly the closure of Cecil Field. In the GWR results the median income consistently influences the CBD in all of the dataset years where an increase in the median income would provide a significant boost to house prices. This makes sense as the CBD of Jacksonville generally houses lower income individuals and families. An increase in median income in a lower income neighborhood will most likely increase house prices. Also, the m ain zip codes and basically the entire area surrounding NAS Jacksonville would experience increase in house prices with an increase in median income. Though the Cecil Field population and the NAS Jacksonville population comprise a small percentage of the overall population of this area, the median income of the military members and nonmilitary members

PAGE 86

86 living in these areas is most likely lower than other areas of Jacksonville and therefore an increase in median income will increase house prices in these ar eas. The house prices in the area directly to the west of Cecil Field would seem to benefit from an increase in median income as well. However, it is difficult to tell whether the closure of Cecil Field had an influence on the INCOME variable as the author is unsure if there was a loss of income in that area attributed to the closure of Cecil Field due to the use of 2000 census data. A lso, data at the census block level, versus the tract level, may have created more concise patterns in the GWR results and w ould have allowed the author to use fewer neighbors in the GWR settings due to the greater variance in median incomes at the parcel level. Median Age in the Census Tract ( MEDAGE) Variable The MEDAGE is consistently statistically significant through all of the dataset years with a consistent positive sign. The value of the coefficient in the OLS results consistently increases through each dataset and generally follows the regional real estate market trend; however, the 2009 value may be overtly high due to t he statistical insignificance of the other variables in the model. In the GWR results an increase in median age has a consistent positive influence on house prices in the main zip codes and the area surrounding NAS Jacksonville. Starting in 1994 and carrying throughout the remaining dataset years, there is a positive MEDAGE influence at the border of Cecil Field that shifts around the border slightly at each dataset year but predominantly settles around the northern portion of Cecil Field. The military population is generally a younger population. It is evident from the GWR maps that, according to the 2000 census tract data, younger populations tend to live around Cecil Field and NAS Jacksonville, to include the main zip codes. The GWR

PAGE 87

87 results are indicating that an increase in the median age in these areas would increase house prices in the respective areas. As with the INCOME variable, it is difficult to tell whether the closure of Cecil Field had an influence on the MEDAGE variable as the author is unsure if there was an increase, decrease, or no change to the median age as a result of the closure and redevelopment of the base due to the use of 2000 census data. The same holds true with MEDAGE as it did with INCOME with regard to the use of census block dat a versus tract data and the variance of the variables at the parcel level. Euclidean Distance to Cecil Field ( DISTCEC) Variable The DISTCEC variable was far from being statistically significant in the OLS results until 1999 when the base closure occurred and the variable then become very significant at a .00 (100% confidence) level and never went below the .02 (98% confidence) level for the remaining dataset years. The first statistically significant coefficient is .31 in 1999 which increases to .41 in 2000 and increases again to .59 in 2009. This indicates, for example in 1999, that each meter increase from Cecil Field will increase a houses price by $0.31. So a property 1000 meters from Cecil will theoretically be worth $310.00 more than a property right on the border of the base. The fact that the coefficient for 2009 is the highest could be due to the high amount of statistically insignificant variables in the 2009 dataset or may indicate that the redevelopment of Cecil Field has been slower than expect ed. Cecil Field was clearly having a negative impact on house prices according to this global variable. It would be interesting to see the reaction of this variable at the local GWR level if the model could have been run in the GWR function of ArcGIS This variable possibly indicates why a majority of the Cecil Field personnel lived in the main zip codes where the property

PAGE 88

88 values could theoretically be approximately $4650.00 greater in 1999 as compared to living right next to Cecil Field. Euclidean Distance to the Central Business District ( DISTCBD) Variable The DISTCBD variable OLS results remained statistically significant and positive through all of the datasets. The variable indicates that an increase in distance away from the CBD will increase the house price which is mainly due to more poverty in the downtown area than in the suburbs. The coefficient in the 2009 dataset is most likely elevated due to the large amount of statistically insignificant variables in the dataset. Euclidean Distance to an Industrial Land Use ( DISTINDU ) Variable The OLS results for the DISTINDU variable remained statistically significant until 2000 and 2009. The coefficient remained positive through all of the datasets. The types of land uses classified as industrial for this model are 1510 Food processing, 1520 Timber processing, 1523 Pulp and paper mills, 1530 Mineral processing, 1540 Oil and gas processing, 1550 Other light industry, 1560 Other heavy industrial, 1561 Ship building and repair, 1562 Prestressed concrete plants, 1563 Metal fabrication plants, and 1590 Industrial under construction. The OLS results for this variable indicate that an increase in distance away from an industrial land use will increase the house price. This finding contradicts Li and Browns (1980) ar gument that accessibility outweighs the negative externalities of such a land use. Euclidean Distance to a Commercial Land Use ( DISTCOMM ) Variable The OLS results for the DISTCOMM variable were statistically significant with the exception of the 1992 and 2009 datasets. The coefficient remained positive through all of the datasets and increased substantially as the dataset years increased. The types of land uses classified as commercial for this model are 1460 Oil and gas storage: except

PAGE 89

89 those areas associat ed with industrial use or manufacturing, 1480 Cemeteries, 1490 Commercial and services under construction. It would have been interesting to see whether commercial land uses such as grocery stores or department stores would have had the opposite effect. E uclidean Distance to a Hospital ( DISTHPTL ) Variable The OLS results for the DISTHPTL variable were statistically significant and negative for all of the datasets. The negative sign indicates that an increase in distance away from a hospital will reduce the house price. Euclidean Distance to a School ( DISTSCHL ) Variable The OLS results for the DISTSCHL variable were statistically significant only for the 1994 and 1999 datasets. The sign was negative for all of the datasets. The negative sign indicates that an increase in distance away from a school will reduce the house price. This variable may have had improved results had the schools been ranked in terms of the education quality or standardized testing results. The variable in this model treated all schools equally. Euclidean Distance to a Major Water Body ( DISTWATR) Variable The OLS results for the DISTWATR variable varied in statistical significance. The sign for the coefficient varied from negative to positive. Based on common knowledge of the real estate market, an increase in distance from a water body should decrease the house price unless the water body is being used for an industrial use such as shipbuilding. The results of this variable are inconclusive. This variable may have had improved results had the water bodies been classified by their use such as recreational use or industrial use. The variable in this model treated all water bodies equally.

PAGE 90

90 Euclidean Distance to a Major Road ( DISTROAD) Variable The OLS results for the DISTROAD variable were s tatistically significant with the exception of the 2009 dataset. The coefficients signs were negative with the exception of the 1992 dataset which was positive. Overall the results of this variable indicate that house owners prefer accessibility to major/ minor arterial roads and that the house price decreases the farther a property is from a major/minor arterial road. Euclidean Distance to a Known Zip Code with Cecil Personnel ( DISTZIP) Variable The OLS results for DISTZIP were statistically insignificant with the exception of the 1999 and 2000 datasets. The signs varied from positive to negative but the 1999 and 2000 datasets both had positive coefficient signs. The zip codes identified in this variable are all of the zip codes identified in the Final Base Reuse Plan where Cecil Personnel lived. Several of the zip codes were right next to Cecil Field or included Cecil Field as illustrated in Figure B 1. The DISTZIP variable was only statistically significant during the years that Cecil Field closed and indicates that an increase in distance from a Cecil Personnel zip code will increase the house price. The addition of a second zip code variable that included only the two main zip codes were Cecil P ersonnel lived may have had more telling results.

PAGE 91

91 CHAPTER 6 CONCLUSION This chapter contains an analysis of this study and recommendations for future studies involving base closure and hedonic pricing models. First, this chapter begins with a summary of findings with an emphasis on how the results attempt to answer the research question. The chapter then provides recommendations for future studies and how this methodology can be improved. Third, it discusses the limitations of this research. Finally, this chapter closes with a brief discussion of areas of future research. The overall emphasis of this chapter lies in allowing future researchers to refine the process of determining the effects of a base closure on surrounding single family house prices. Summary of Findings The results of this study attempt to answer the research questions which are: 1) When the closure of Cecil Field was announced in 1993, was there an anticipatory effect on singlefamily house sales prices from the announcement to the actual closure in 1999? 2) After Cecil Field closed and was turned over to the City of Jacksonville in 1999, what effect did the closure and redevelopment have on singlefamily house sales prices after the closure and in 2009? In regard to question one (1) the OLS results are unclear as the two variables specifically tied to Cecil Field, DISTCEC and DISTZIP, were not statistically significant during the 1992 period when no closure was announced and during the 1993 to 1994 period when the closure was announced and the year following, respectively. The most telling variable in the GWR analysis is SF. The 1992 dataset indicates that an increase

PAGE 92

92 in square feet will provide the lowest increase in the house price (a green color band) The 1993 dataset indic ates that an increase in square footage for houses located on the northeast of Cecil Field will provide a higher price increase while the rest of the area remains slightly above the 1992 values (yellow color band) The 1994 dataset shows a slightly elevate d increase in price should the square footage of the house increase (yellow color band) This provides clear evidence that home values immediately around Cecil Field could potentially be the slightest amount higher after the announcement of the base closur e in comparison to before the base closure. This could possibly be explained by the fact that the market knew that the base would eventually be redeveloped; however, the final reuse plan was not published until 1996 and the final use of the land in 1994 was undetermined. In regard to question two (2), the OLS results are clear as the two variables specifically tied to Cecil Field, DISTCEC and DISTZIP, are both statistically significant in 1999 and 2000 and indicate that an increase in distance away from Cecil Field will increase the house price. The GWR results are clear as the SF variable remains at the lowest level (green band) for the 1999, 2000 and 2009 datasets immediately around C ecil Field but increases slightly to the yellow band as the distance increases east toward NAS Jacksonville from Cecil Field. Both the OLS and GWR analysis provide inconclusive evidence that there was any anticipatory effect on house prices after the announcement of NAS Cecil Fields closure. However, both analyses provide conclusive evidence that the actual closure of NAS Cecil Field had a negative effect on house prices immediately surrounding the base but that the effect diminishes rather quickly as dist ance increases from the base. This could

PAGE 93

93 be due to an obvious influence from NAS Jacksonville that seems to correct any minimal effect that Cecil Field had. The evidence is inconclusive that the redevelopment of Cecil Field has had any effect on house pric es. Recommendations for Future Studies In future studies there is certainly room for improvement of the methodology used for this study. First the inclusion of more structural variables may settle the flipflopping of signs and statistical significance. This was argued by authors reviewed in Chapter 2 and this author believes the argument is valid. Unfortunately, the structural variables needed were not available in the data accessible by the author. Variables such as number of bedrooms, number of bathrooms, garage, pool, and exterior faade construction (brick, vinyl, stucco, etc.). Second, attempt to get demographic data at the block level and according to the year of the dataset. It takes much more time to harvest the data but the results of the labor m ay be more conclusive. Third, as seen in this study, GWR is a powerful tool that provides the researcher with pertinent local level data. If the researcher plans to use GWR, build a model with as few proximity variables as possible and still be able to tel l the story of what composes the house price. Be familiar with your tools before building your model. Limitations of this Study The lack of block level data and structural data hampered the results of the study. Use of such data might have provided mor e conclusive results. Areas of Future Research There is certainly an opportunity to carry this research farther. Though much h as happened since the closure of NAS Cecil Field, Jacksonville has only just started to tap the economic potential of Cecil Commerce Center. The recent hiring of a new developer

PAGE 94

94 that has successfully redeveloped bases in the past may pay dividends to the pace of redevelopment at Cecil Commerce Center. Also, the recent completion of the Cecil Commerce Center Parkway and its connection to Interstate 10 has great potential to spur commercial development in the Center It would be very interesting to c ontinue t his study in another ten years, with a carefully developed pricing model, to see if the development of Cecil Commerce Center has any kind of effect on surrounding house prices.

PAGE 95

95 APPENDIX A THE MODERN BASE CLOSURE PROCESS The modern base closure process began when the first Base Closure and Realignment Act was enacted in October 1988 and allowed for only one wave of closures (Collins, 2008, p. 4). Prior to the 1988 Act, decisions made by the Secretary of Defense in regard to base closures during and after the Vietnam War were often made without consulting the military service or Congress (Collins, 2008, p. 4). Concerns that decisions to close bases were politically motivated surfaced in the 1970s by Congressmen representing northeastern and Midwestern states, adding to opposition of future base closures (Wilson & Weingarter, 1993, p. 33). The 1988 Base Closure and Realignment Act allowed the Secretary of Defense to appoint an eight member Base Closure and Realignment Commission, who was not independent and worked for the Secretary of Defense, and all discussions were held behind closed doors (Collins, 2008, p. 4). Upon completion of the commissions deliberations, a recommendation was made to the President of the United States and the Presidents decision coul d only be reversed through a joint resolution in Congress which could be vetoed by the President (Collins, 2008, p. 4). The end of the Cold War created additional need for the Department of Defense to close additional bases and in 1990 the Base Closure and Realignment Act was reenacted to support closures in 1991, 1993, and 1995 (Collins, 2008, p. 5). Complaints on the 1988 Acts lack of transparency saw major changes in the 1990 version. The eight member Commission, appointed by the President, must be subm itted to Congress for advice and consent. The President should consult with the Speaker of the House of Representatives for two members, the majority leader of the Senate for two members,

PAGE 96

96 the minority leader of the House of Representatives for one member, and the minority leader of the Senate for one member. All proceedings are open to designated committees within the Senate and House of Representatives and with the exception of meetings with classified information the public with opportunities for the publ ic to be heard. The Secretary of Defense is required to publish in the Federal Register the criteria to be used by the Commission and the recommended list of base closures and realignments (with a summary of the selection process) to be cons idered by the C ommission. After visiting the sites, holding public hearings, and deliberating, the Commission is required to submit their recommendations to the President and copy both Armed Services Committees in Congress. They must explain any deviations from the Secretary of Defenses recommendations. The President can approve or disapprove the recommendations as a whole. If disapproved, the President must provide reason for disapproval to the Commission and Congress and the Commission must resubmit a revised recommendation to the President. Upon approval the President must submit the approved list to Congress and Congress can either approve or disapprove the list via joint resolution. The President can veto the joint resolution which will require Congress to override t he veto with a twothirds majority. The 1990 Base Closure and Realignment Act provides greater transparency to the public and removes some of the power from the Secretary of Defense when making closure decisions. The Act was amended in 2004 to support addi tional required closures and realignments in 2005.

PAGE 97

97 APPENDIX B KNOWN ZIP CODES WITH CECIL FIELD PERSONNE L Figure B 1 NAS Cecil Field Personnel Zip Codes

PAGE 98

98 APPENDIX C 1992 DUVAL GWR RESUL TS Figure C 1. Duval 1992 Property Age in Years ( AGE) GWR Output

PAGE 99

99 Figure C 2. Duval 1992 Square Footage of House ( SF ) GWR Output

PAGE 100

100 Figure C 3 Duval 1992 Acreage of Lot ( LOTSIZE ) GWR Output

PAGE 101

101 Figure C 4. Duval 1992 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 102

102 Figure C 5 Duval 1992 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 103

103 APPENDIX D 1993 DUVAL GWR RESULTS Figure D 1. Duval 1993 Property Age in Years ( AGE) GWR Output

PAGE 104

104 Figure D 2. Duval 1993 Square Footage of House ( SF ) GWR Output

PAGE 105

105 Figure D 3 Duval 1993 Acre age of Lot ( LOTSIZE ) GWR Output

PAGE 106

106 Figure D 4. Duval 1993 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 107

107 Figure D 5 Duval 1993 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 108

108 APPENDIX E 1994 DUVAL GWR RESULTS Figure E 1. Duval 1994 Property Age in Years ( AGE) GWR Output

PAGE 109

109 Figure E 2. Duval 1994 Square Footage of House ( SF ) GWR Output

PAGE 110

110 Figure E 3 Duval 1994 Acre age of Lot ( LOTSIZE ) GWR Output

PAGE 111

111 Figure E 4. Duval 1994 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 112

112 Figure E 5 Duval 1994 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 113

113 APPENDIX F 1999 DUVAL GWR RESULTS Figure F1. Duval 1999 Property Age in Years ( AGE) GWR Output

PAGE 114

114 Figure F2. Duval 1999 Square Footage of House ( SF ) GWR Output

PAGE 115

115 Figure F 3 Duval 1999 Acre age of Lot ( LOTSIZE ) GWR Output

PAGE 116

116 Figure F 4. Duval 1999 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 117

117 Figure F 5 Duval 1999 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 118

118 APPENDIX G 2000 DUVAL GWR RESULTS Figure G 1. Duval 2000 Property Age in Years ( AGE) GWR Output

PAGE 119

119 Figure G 2. Duval 2000 Square Footage of House ( SF ) GWR Output

PAGE 120

120 Figure G 3 Duval 2000 Acre age of Lot ( LOTSIZE ) GWR Output

PAGE 121

121 Figure G 4. Duval 2000 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 122

122 Figure G 5 Duval 2000 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 123

123 APPENDIX H 2009 DUVAL GWR RESULTS Figure H 1. Duval 2009 Property Age in Years ( AGE) GWR Output

PAGE 124

124 Figure H 2. Duval 2009 Square Footage of House ( SF ) GWR Output

PAGE 125

125 Figure H 3 Duval 2009 Acre age of Lot ( LOTSIZE ) GWR Output

PAGE 126

126 Figure H 4. Duval 2009 Median Household Income in the Census Tract ( INCOME ) GWR Output

PAGE 127

127 Figure H 5 Duval 2009 Median Age in the Census Tract ( MEDAGE) GWR Output

PAGE 128

128 LIST OF REFERENCES Bradshaw, T. K. (1999). Communities not fazed. Journal of the American Planning Association, 65 (2), 193206. Charlton, M. & Fatheringham, A. S. (2009). Geographically Weighted Regression [White Paper]. Retrieved from http://ncg.nuim.ie/ncg/GWR/GWR_WhitePaper.pdf Collins, M. (2008). The Economic Impact of Base Realignment and Closure on Local Communities. Unpublished masters thesis, Georgetown University, Washington D.C. Cowan, T. & Baird, W. (2005). Military Base Closure: Socioeconomic Impacts (CRS Report for Congress RS22147, May 18, 2005). Fort Belvoir, VA: Defense Acquisition University. Retrieved S eptember 30, 2009, from http://www.nationalaglawcenter.org/assets/crs/RS22147.pdf Dardia, M., McCarthy, K. D., Malkin, J., Vernez, G., & RAND C orp S anta M onica, CA (1996). The effects of military base closures on local communities: A short term perspectiv e. Dehring, C. A, Depkin, C. A., & Ward, M. R. (2006). The Impact of Stadium Announcements on Residential Property Values: Evidence from a Natural Experiment in Dallas Fort Worth. Retrieved March 19, 2010, from http://papers.ssrn.com/sol3/papers.cfm?abstr act_id=993250 Fagan, M. & Reader, R. (1996). Communities May Lose Military Retirees Along with Their Bases. Rural Development Perspectives, 11(3), 1722. Fishkind & Associates, Inc. (2006, October 24). Cecil Commerce Center Economic Impact Analysis: Prepared for Jacksonville Regional Chamber of Commerce Gao, X., Asami, Y., & Chang J. C. (2002). An Empirical Evaluation of Hedonic Regression Models. In Proceedings of the ISPRS Technical Commission IV Symposium 2002. Bejing, China: International Society for Photogrammetry and Remote Sensing. Retrieved from http://www.isprs.org/proceedings/XXXIV/part4/pdfpapers/302.pdf Garson, D. (n.d.). Two State Least Squares (2SLS) Regression Analysis. Retrieved April 28, 2010, from North Carolina State University, PA 765766 Quantitative Research in Public Administration Web site: http://faculty.chass.ncsu.edu/garson/PA765/2sls.htm GlobalSecurity.org. (n.d.). Military: Naval Air Station Cecil Field. Retrieved October 4, 2009, from http://www.globalsecurity.org/military/facility/cecil field.htm

PAGE 129

129 Hiebert, T. E. (2009). How is the Value of Real Estate Affected by the Department of Defense Base Realignment and Closure Process? Dissertation Abstracts International, 70(09), A. (UMI No. 337 2054) Hwang, M. & Quigley, J. M. (2006). Economic Fundamentals in Local Housing Markets: Evidence from U.S. Metropolitan Regions. Journal of Regional Science, 46(3), 425453. Li, M. M. & Brown, J. H. (1980). Micro Neighborhood Externalities and Hedonic H ousing Prices. Land Economics, 56(2), 125141. O'Sullivan, A. (2009). Urban economics Boston: McGraw Hill/Irwin Peek, J. & Wicox, J. (1991). The Measurement and Determinants of SingleFamily House Prices. AREUEA Journal, 19 (3), 353382. Quigley, J. M. (1995). A Simple Hybrid Model for Estimating Real Estate Price Indexes. Journal of Housing Economics, 4, 1 12. Renski, H. & Reilly, C. (2007). Understanding the Impact: Closing Naval Air Station Brunswick. Augusta, Maine: Maine State Planni ng Office. Retrieved August 4, 2009, from http://www.maine.gov/tools/whatsnew/attach.php?id=97954&an=1 Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography 46(2), 23424. Tu, C. (2005). How Does a New Sports Stadium Affect Housing Values? The Case of FedEx Field. Land Economics, 81(3), 379395. U.S. Department of Defense. (n.d.). Base Realignment and Closure 2005: Frequently Asked Questions. Retrieved April 5, 2009, from http://www.defenselink.mil/brac /faqs001.html Wilson, C.L. & Weingartner, J. L. (1993). Blame Proof Policy Making: Congress and Base Closures. Unpublished masters thesis, Naval Postgraduate School, Monterey, California.

PAGE 130

130 BIOGRAPHICAL SKETCH Lieutenant Commander Greg Jennings reported aboard the N aval Reserve Officers Training Corps, University of Florida, as a postgraduate student in the Urban and Regional Planning program in December of 2008 after completing a tour with the 25th Naval Construction Regiment in Gulfport, Missis sippi. A native of Virginia Beach, Virginia, Lieutenant Commander Jennings graduated from Savannah College of Art and Design in 1996 with a Master of Architecture. Prior to joining the United States Navy he worked as an intern architect for four years in Florida. He was commissioned a United States Naval Officer and member of the Civil Engineer Corps on 07 July 2000. Lieutenant Commander Jennings first tour was as Assistant Resident Officer in Charge of Construction at Norfolk Naval Station, Virginia, from December 2000 to September 2002. While serving as Assistant Resident Officer in Charge of Construction, Lieutenant Commander Jennings managed major airfield military construction projects and job order contracts. In October 2002, Lieutenant Commander Je nnings reported to Public Works Center Pearl Harbor to serve as intern architect in the Civil Engineer Corps Intern Architect Development Program. He then was assigned as the Navy Region Hawaii Historic Preservation program manager. In that capacity he w as responsible for consulting with parties around the nation on treatment of the Pearl Harbor National Historic Landmark and all historic structures within the Navy Region Hawaii AOR. During his tenure Navy Region Hawaii won the FY 2003 Chief of Naval Operations Cultural Resources Management Award for Installations.

PAGE 131

131 In September 2005, Lieutenant Commander Jennings reported to Naval Mobile Construction Battalion Seven, Gulfport, Mississippi. At Naval Mobile Construction Battalion Seven he served as Charlie Company Commander; Detail Whiskey field exercise Officer in Charge; Detail Buehring, Kuwait Officer in Charge; homeport training officer; and the Detail Pacific Partnership Officer in Charge aboard the USS Peleliu. In September 2007, Lieutenant Commander Jennings reported to the 25th Naval Construction Regiment, Gulfport, Mississippi. At the 25th Naval Construction Regiment he served as the Future Operations officer in charge of mission planning and exercise planni ng for Seabees deploying to the United St ates Southern Command and United States Northern Command areas of operations and was in charge of disaster preparedness planning for the command. Lieutenant Commander Jennings is a licensed arch itect in the State of Florida.