|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 THE IMPLEMENTATION OF URBAN GROWTH BOUNDARY AND ITS EFFECTS ON LOCAL HOUSING AFFORDABILITY: PORTLAND, OR EGON AS A CASE STUDY By XING MA A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FU LFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF M A S TER OF A RTS IN U RBAN AND R EGIONAL PLANNING UNIVERSITY OF FLORIDA 2011
2 2011 Xing Ma
3 T o my beloved parents
4 ACKNOWLEDGMENTS First of all, I would like to express my gratitude to my thesis committee c hair, Andres D. Blanco, for all of the help, support, a nd encouragement he provides me throughout my graduate stud ies at the University of Florida helped and inspired me to develop this thesis. I also would like to e xpress appreciation to my thesis C o chair, Paul D. Z wick, for his contribut ion to my study, essential help with some statistic al difficulties They made the thesis research process less painful and more enjoyable for me. In addition, I would lik e to thank David C. Ling m y thesis committee member and mi nor coordinator in Real Estate, for providing valuable feedbacks and helping with real estate stud ies Secondly, I would like to express special appreciation and love to my family in C hina, especially to my mo ther and father Yanhong Song and Yang Ma Without their love, support, and encouragement I would not be able to finish my studies in US. They have made huge and unforgotten sacrifices to support me throughout my life. I would like to de d icate this documen t to all my family members. Additionally, the support and belief from my girlfriend, Jing Wang, strongly motivates me to overcome any difficulties and to achieve something big I am thankful that I was able to join the Urban and Regional Planning p rogram in 2009 fall, where I have been lucky to meet a lot of friendly professors and students from all over the world. I learned a lot from these individuals. They enhanc ed my knowledge and practical experiences in urban planning, allowing myself to improv e and grow over the past two years. Thank you for the wonderful time we spent together and the joys we shared.
5 In retrospect, I was like a child when I just graduated from university taking noth ing serious; whereas I am grown up now and about to step out from g raduate school in to a different life having a job, setting up my own family and sharing the burdens upon my parents I deeply appreciate what I have been through so far, and all the things make me the way I am now
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Problem Statement ................................ ................................ ................................ 13 Urban Growth Boundary and Housing Affordability ................................ ................ 14 Urban Growth Boundary ................................ ................................ ................... 14 Housing Affordability ................................ ................................ ........................ 16 Why Portland as A Case Study ................................ ................................ ............... 16 Overview of the Thesi s ................................ ................................ ........................... 18 2 LITERATURE REVIEW ................................ ................................ .......................... 20 The Implementation of UGB in Portland ................................ ................................ 20 Adoption History of UGB in Portland ................................ ................................ 21 Planning Administration ................................ ................................ .................... 23 Dispute Resolution ................................ ................................ ........................... 24 Unintended Effect of UGB ................................ ................................ ................ 24 The Conflicts between Growth Management and Housing Affordability ................. 25 Th Issue ................................ ................................ ................................ .................... 28 Theoretical Debates in Previous Studies ................................ .......................... 28 Statistical Analysis Indicators and Regression Models ................................ .. 33 Indicators HOI and H+T affordability i ndex ................................ .............. 34 Multiple regression mo dels ................................ ................................ ........ 38 Summary of Previous Studies ................................ ................................ ................. 43 3 METHODOLOGY ................................ ................................ ................................ ... 45 Correlat ion between Land Prices and Housing Prices ................................ ............ 46 Data Source and Study Area ................................ ................................ ............ 46 Prices T rends Analysis and Spearman Correlation Coef ficient ........................ 47 Median Housing Prices Regression Model ................................ ............................. 48 Data Source and Study Area ................................ ................................ ............ 48
7 Variables in the Regression Model ................................ ................................ ... 50 4 ANALYSIS AND RESULTS ................................ ................................ .................... 51 Correlation between Land and Housing Prices ................................ ....................... 51 Prices Trends Analysis ................................ ................................ ..................... 52 Spearman Correlation Coefficient ................................ ................................ .... 57 M edian Housing Prices Regression Model ................................ ............................. 58 Correlation Coefficient and Scatter Plot Analysis ................................ ............. 58 Median Housing Prices Regression M odels ................................ ..................... 65 The Price Effects of the UGB Variable ................................ ............................. 69 5 DISCUSSION ................................ ................................ ................................ ......... 71 Summary of the Correlation Analysis ................................ ................................ ...... 71 Inconsistent Prices Trends of San Diego and Phoenix ................................ ..... 72 Limitations of the Correlation S tudy ................................ ................................ .. 73 Summary of the Regression Analysis ................................ ................................ ..... 74 Uncorrelated Independent Variables in the Model ................................ ............ 76 Results of the Regression Model ................................ ................................ ...... 77 Limitations of the Regression Analysis ................................ ............................. 78 6 CONCLUSIONS ................................ ................................ ................................ ..... 80 APPENDIX : DATA USED IN THE REGRESSION MODEL ................................ .......... 82 LIST OF REFERENCES ................................ ................................ ............................... 85 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 89
8 LIST OF TABLES Table page 2 1 Median housing prices in major western metropolitan areas .............................. 31 2 2 Percentage change of six variables of urban development pattern in Portland, 1980 2000 (Jun, 2004) ................................ ................................ ....................... 33 3 1 List of the metropolitan areas for the p rices trend a nalysis and Spearman Correlation ................................ ................................ ................................ .......... 47 3 2 List of the u rbanized a reas in the m ultiple r egression m odel ............................. 49 3 3 List o f variables in the median housing prices r egression m odel ........................ 50 4 1 Ranking of the Spearman Correlation Coefficient of 10 metropolitan areas. ...... 57 4 2 Correlation coefficients of dependent variable and independent variables in the regression model ................................ ................................ .......................... 59 4 3 Summary of the vehicle density median housing prices regression model ........ 67 4 4 Summary of the housing density median housing prices regression model ....... 68 4 5 Summary of the job density median housing pr ices regression model ............... 68 4 6 Summary of the population density median housing prices regression model ... 69 4 7 Correlation coeffic ients of the UGB dummy variable and other independent variables in the regression model ................................ ................................ ....... 70 5 1 Summary of the correlation coefficients of independent variables in the regression model ................................ ................................ ................................ 75 A 1 Data used in the median housing prices regression model ................................ 82
9 LIST OF FIGURES Figure page 2 1 Map of the u rban g rowth b oundary as of May 2006 ................................ ........... 21 2 2 The estimated model of total transportation cost in the H+T Affordability Index ................................ ................................ ................................ ................... 36 2 3 Percentages of com munities considered affordable u s tandard politan areas ..................... 37 2 4 Map of the locations of afforda ble communities in Portland Metropolitan Area u ................................ ................................ ..... 38 4 1 Trends of land p rices of 11 metropolitan areas ................................ ................... 54 4 2 Trends of percentage of land share of 11 metropolitan are as ............................ 55 4 3 Trends of housing p rices of 11 metropolitan areas ................................ ............. 56 4 4 The scatter plot of median housing pri ces and average household size ............ 60 4 5 The scatter plot of median housing prices and UGB dummy variable ................ 60 4 6 The scatter plot of median housin g prices and unemployment rate. ................... 61 4 7 The scatter plot of median housing pric es and median household income ........ 62 4 8 The scatter plot of median housing prices and mean commu te time .................. 62 4 9 The scatter plot of median housing p rices and vehicle density. .......................... 63 4 10 The scatter plot of median hous ing prices and housing density ......................... 64 4 11 The scatter plot of median housing prices and job density ................................ 64 4 12 The scatter plot of median housing prices and population density ..................... 65
10 LIST OF ABBREVIATION S CNT Center for Neighborhood Technology CTD Center for Transit Oriented Development DLCD Department of Land Conversation and Development HOI Housing Opportunity Index H + T Index Housing and Transportation Housing Affordability Index HUD U.S. Department of Housing and Urban Development IGA Intergovernmental Agreement LCDC Land Conservation and Development Commission LUBA Land Use Board of Appeals NAHB National Association of Home Builders TDR Transfer of Development Rights UA Urbanized Area UGA Urban Growth Area UGB Urban Growth Boundary USA Urban Service Area
11 Abstrac t 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 IMPLEM ENTA TION OF URBAN GROWTH BOUNDARY AND ITS EFFECT S ON LOCAL HOUSING AFFORDABILITY: PORTLAND, OREGON, AS A CASE STUDY By Xing Ma May 2011 Chair: Andres Blanco Cochair: Paul Zwick Major: Urban and Regional Planning Since 1958 in the United States Urban Growth Boundary (UGB) has been adopted as an impor tant growth management tool that aim s to prevent urban sprawl and preserve the natural environment on the city s outskirts The UGB is an invisible line drawn around the city, which indicates the developable areas and farmland It compellingly control s the future urban development inside the boundary and favors higher density development, public transit and walk able urban spaces in city area. After the implementation of UGBs nationwide, there are increasing concerns on this mandate growth management tool and the l ocal housing affordability Based on some facts and previous studies, the boundary constrain s the supply of land and changes the urban development pattern which tends to increase housing prices. The purpose of t his thesis explores the price effec ts of the UGB constrain s land supply and changes urban development pattern on Portland s housing affordability As to Portland the city a dopt ed UGB in 1979 under the legislation and management of the state government. D ue to the unique implementation st atue and urban devel opment pattern,
12 Portland has been a classic UGB city for researches and a perfect research object for my study Therefore, this research is aimed to explore the relationship between the implementation of the Portland s and local housi ng affordability. It review s all the important previous studies on U GB and housing affordability, and describes t heir arguments, methodologies and limitation s Both theoretical and statistical methods are adopted i n this study. I n the analysis a nd results chapter, prices trend analysis and Spearman correlation coefficient are used to explore the relationship between land and housing prices between 1984 and 2009 Moreover, a regression model was conducted to tests the price effects of the UGB on m edian housing prices among 35 major urbanized areas in 2000 as well as the other nine independent variables I n conclusion the results showed that the UGB did not contribute to the median housing prices of the urbanized areas and the best median housing prices regression model consisted of median household income, job density, media n commute time which has a highest adjusted R square. The final part of the study provide s the explanations for the result and discuss es the limitations of the research
13 C HAPTER 1 INTRODUCTION Problem Statement Nowadays, more and more cities and counties have adopted some form of Urban Growth Boundary ( UGB) in their comprehensive planning which is a compelling requirement of local or state government. The growth boundary s ervices as an important growth management tool that intends to curb suburban or leap frog development, protect farm and forest land and encourage redevelop of the inner city. Compared to traditional growth management method UGB intends to encourage hig her density development, public transit and walk able urban spaces in city area. Moreover there are concerns on this mandate growth management tool that the boundary might constrain the future supply of land and affect the housing market which was suppor ted by many previous studies (Downs, 2004) This kind of mandate growth manag ement and planning tool may result in higher housing price s, leading to lower local housin g affordability. The relationship between the boundary and housing affordability is cont roversial and unclear. This study aims to explore the relationship between UGB and housing affordability in Portland, Oregon and find out whether the implementation of UGB tends to lower the housing affordability in the city If there is a certain relatio nship between these two research objects, this study would also answer how significan t this relationship is based on the statistical analysis.
14 Urban Growth Boundary and Housing Affordability As mentioned above, the research question is to explore the re lationship between growth boundar y and housing affordability. More importantly it i s necessary to introduce t hese two concepts and define them for the following study. Urban Growth Boundary A typical Urban Growth Boundary is an invisible line or boun dary drawn around the city area, which separates the developable lands for future urbanization from farm and forest land. Inside this boundary, the land can be used for future urban development; however, on the other side of the boundary, the land is pres erved and not able to be used for urban development for certain period of time, for example, 20 years. According to Staley et al. ( 1999) there are at least 6 goals of the growth boundary : p reserve open and farmland; m inimize the use of land generally by reducing lot sizes and increasing residential density; r educe infrastructure costs by encouraging urban revitalization, infill, and compact development; c learly separate urban and rural uses; e nsure the orderly transition of land from rural to urban uses; and p romote a sense of unified community. N ormally, the adoption of the growth boundary is under the planning requirement of local or s tate level government. F or instance, i n United States, Only Oregon and Washington ha ve very aggressive state legislatio n, which requires all the counties and cities to adopt growth boundaries (Anderson, 1999) In most of the cases, the boundary should leave enough land supply to accommodate future ur ban development, such as for residential indus trial and commercial uses etc In Oregon, according to the state
15 laws, all cities are required to have a 20 year supply of land for future develop ment with in their growth boundary areas ( Anderson, 1999 ). Moreover, these cities also need to review the boundary area every 5 years to ensure there is a sufficient supply of land for the future. T he size of the boundary is critical in the adoption and expansion process: if too much land is included in the boundary, the boundary might not effectively manage the urban growth; if the supply of land is less than the future demand, the land prices would be pushed up. T o be effective in the practical implementation process the UGB can be adopted in several different forms, such as urban service area ( USA ), urban growth area (UGA) and urban d emarcation lines, etc. These planning tools have similar goals which control the future urban development and encourage compact development. However, unlike traditional UGB s that control the supply of land, the urban service area manages the expansion of the urban infrastructure, such as water, electricity, sewer etc and is defined as an The expansion decision of urban infrastructure is made in orderly intervals and coordinated with land use decision s (Warken, 2003) In addition, there are several alternative planning tools of UGB, such as Intergovernmental Agreement (IGA) in Boulder County Colorado Transfer of Development Rights (TDR) program in Lanc ast er County, Pennsylvania According to the definition provided b y Anderson (1999) the goal of the IGA s are to represent binding agreements between the parties, creating a contractual obligation to comply with them While TDR is a type of zoning ordinance that allows the owners of p roperty
16 located in low density deve lopment or conservation use zone to sell the development rights to other owners (Roddewing & Inghram, 1987) I n sum i n this study, the UGB is defined as a land supply m anagement tool that controls the future supply of developable land. T he concept of U GB is used consistent ly throughout th e whole paper. Housing Affordability Generally speaking, housing affordability measures the ability to afford a housing unit based on one household or family annual income level It is a relationship between housing p rice and annual income. Traditionally, according to the standard provided by the U.S. Department of Housing and Urban Development (HUD ) the house market is considered as affordable only if households pay less than 30 % of their annual incomes for ho using Recently a more comprehensive approach has been created to test housing affordability the Housing and Transportation Index. This index is derived from the evidence that transportation cost is the second large st household cost after housing and stated tha t the combined portions of the cost of housing and transportation should be not exceed 45 percent of a household s annual income ( Center for Transit Oriented Development and Center for Neighborhood Technology 2006 ). T his approach will be elaborated in the following Analysis and Results Chapter. I n this study, both of the 30% and 45% standard s are applied to test the housing affordability. Why Portland as A Case Study I n this research, Portland is the case study used to demonstrate and explore the effe cts of UGB on local housing affordability. Based on the facts and statistic al analysis of Portland, I will not only represent the qualitative description of the effects of the
17 boundary on housing, but also test the significance of UGB on housing affordabil ity in a quantitative way. C ompared to other UGB cities in the US, Portland has some unique characteristics to be a n ideal research object for th is thesis First, the state of Oregon is one of three states in which the UGB is mandated, with the others c onsisting of Washington and Tennessee The efficiency of the implementation of this planning tool is ensured by the mandate power from the state level. T o some extent, the mandate d planning process make s it easier for researchers to gather evidence of th e effects of the UGB in the city, such as changed development pattern, constrain supply of land, etc. Second Portla nd has been well known for its Urban Growth B oundary planning policy for more than 30 years since it adopted the UGB back in 1979. As menti oned before, the growth boundary is focused on achieving the long term goals of compact development Even after the boundary was implemented the effects of UGB on the city still need a certain time period to emerge T he size of the boundary is critical issue: if it contains more land than future needs, then the future urban growth is not effectively contained; if it accommodates less land than needed there will not be enough l and for the future urban growth and population W hen the boundary in Portland was originally created the boundary was drawn slightly large and with little scientific data supported (Anderson, 1999). I t makes the boundary less effective in control ling urban growth. Since then, the boundary in Portland was little expanded. Third t he existence of abundant and available data resource s is another main reason that Portland is popular choice for urban growth management study Portland has attracted much attention from government officials research institutes, and many
18 scholars in urban studies since the city adopted the UGB in the late 197 0s. The city has become a textbook example for state or local government s who intend to adopt UGB s as growth management tool s in their jurisdiction As has been stated above these are the main reaso n s and advantages that I also choose Portland in my research Overview of the Thesis T he most important purpose of this thesis is to answer the question w hat are the effects caused by implementation of UGB on local housing affordability ? T he main body of the paper consists of qualitative description and quantitative analysis chapters to find out the answer to th is issue. I n Chapter 4 Analysis and Results prices trend and correlation analysis between land and housing prices and a median housing prices m ultiple regression model will be conducted in order to test the following hypotheses Hypothesis 1. D ue to the UGB constrain on the supply of land, the trends of land prices and land share in the Portland metropolitan area should be higher than other metr opolitan areas without UGBs in United States. Since land prices have a significant impact on housing prices, the housing prices in Portland should be also higher. I n addition the Spearman Correlation C oefficient should supports the argument that the land and housing prices in Portland are highly correlated. Hypothesis 2. Besides affecting the land prices, the UGBs also tend to change the urban development patterns, such as higher density development, high redevelopment and infill development rate of the i nner city, etc. T he regression model takes into account all of the factors that contribute to housing prices to measure the price effects of different predictors on t he median housing prices among 35 urbanized
19 areas. The result of the regression analysis s hould demonstrate the exact price effects of the UGB on the median housing prices. T he whole the sis is structured in six chapters. Chapter 1 is the brief introduction of the research topics, objectives questions and hypothesis Chapter 2 first reviews t he boundary adoption history and planning administration of UGB in Portland. Then it presents all the important theoretical debates and statistic al analysis done by previous studies, focusing on UGB and housing affordability in Portland Chapt er 3 provides the details about the statistical analysis methodology Chapter 4 demonstrates the analysis and the results in these two methods. Chapter 5 discusses the results of the analysis and the limitations of the research The l ast chapter draw s the conclusion fo r the research questions and make s recommendations from a planning perspective
20 CHAPTER 2 LITERATURE REVIEW T his chap ter present s the adoption history, and the implementation status of UGB in Portland, such as the manag ing authority planning administra tion planning policies, etc. Following that, the next part discusses the conflicts between growth management and housing affordability which is fundamental information to link the UGB in Portland with local housing affordability issue. T he third part is the most essential in the literature review which is the answer to my research question s provided by previous researchers It will be divided into two sections that demonstrate the arguments, methodologies and limitations in the previous debates and statis tical analysis done by other major researche r s T he theoretical debates focus on the relationship between the UGB, land prices and housing prices, while the statistical analysis concentrate on multiple regression of Portland s housing affordability that co ncludes all the factors contributing to housing prices A ll the previous studies in each section will be reviewed in chronological order. The Implementation of UGB in Portland Since the UGB was adopted by State government in 1979, it has been implemented as one important part of the statewide land use planning program in Portland for more than three decades T he purpose of a growth management tool is to contain the urban development within a certain boundary and to prevent sprawl Today, The UGB in Portlan d is considered as one of the most successful cases of managing urban growth in the nation, mainly because of its long implementation history and the mandated administration from Oregon State.
21 Figure 2 1 Map of the u rban g rowth b oundary as of May 2006 (Source: http://library.oregonmetro.gov/files/ugbmap0506.pdf Last accessed January, 2011 ) Adoption History of UGB in Portlan d Oregon State required all the cities within the state to ad opt the urban growth boundary in 1979. H owever, the state actually has begun to work on urban growth management and planning administration issues from the early 1970s. B efore the growth boundary adoption in 1979, Oregon State bega n its efforts to control the urban growth and development in the late 196 0s. I n 1969, Senate Bill 10 0 was adopted by the Oregon State L egislature, which required every city and county in the state to have a comprehensive land use planning
22 that meets state s standards (Oregon Go vernment 2010) Due to the lack of effective enforcement in this b ill, most of the cities and counties refused to develop such plans. Later in 1973, the Governor Tom McCall gave a famous speech to legislature and campaigned across the state to appeal for a statewide land use planning program. As a result, on the same year, Senate Bill 100 was approved, and it created Land Conservation and Development Commission (LCDC) the Department of Land Con servation and Development (DLCD ) and statewide protection for farm and forest land The task for LCDC, assis ted by DLCD were: adopt s state land use goals and implements rules, assures local plan compliance with the goals, coord inates state and local plan ning, and manages the coastal zone program ( LCDC, 2010) T he Statewide Planning Goals are the fundamental planning ordinances in Oregon between state and local governments According to the official Statewide Planning Goals statements, the UGB shall be established to identify and separate u r banizable land from rural land (Oregon Government, 2010) which apply to all cities in the state. More importantly, its enforcement from the state is one of the main reason s that assure efficiency of planning policie s at the local level Under this system, all the cities are required to submit their proposed UGBs to the LCDC and the LCDC will verify them according to the planning ordinances in the statewide planning goals ( Oates, 2006 ) I n 1979, Metro -a regional gov ernment, was created by Portland area voters, which is the first metropolitan council in the United States ( Oregon Government, 2010 ). I t is responsible for the management of the urban growth boundary for the Portland metropolitan area Multnomah, Clackamas, and Washington Counties.
23 Planning Administration As mentioned above, the Metro is one of the 240 UGB jurisdictions in Oregon which directly manages the urban growth boundary in the Portland metropolitan area ( The Oregon Encyclopedia, 2011) A ccording to the information on Metro s official website, Metro has some specific land use planning powers including: coordination between regional and local comprehensive plans in adopting a regional urban growth boundary; r equiring consistency of local comprehensiv e plans with statewide and regional planning goals; p lanning for activities of metropolitan significance including (but not limited to) transportation, water quality, air quality and solid waste According to Oregon State Law, the UGB in Portland Metrop olitan region is required to contain 20 year of supply of land for the future residential development inside the boundary. T he Metro Council is responsible for reviewing the supply of land for the future use every five years, if necessary, expand the bound ary. The UGB in Portland was originally approved to include 15.8% more land than expected to be development for 20 year (Nelson & Moore, 1993 ). However, since the first approval in 1981 the boundary has been expanded for more than thirty times, for exampl e, 3,500 acres in1998, 380 acres in 1999, and 1,956 acres in 2004; however, most of these expansions were less than 20 acres. The biggest expansion of UGB in history was recorded in 2002, which was 18,867 acres, providing 38,657 housing units and 2,671 ac res for additional jobs (Metro, 2011). I n addition according to the planning procedures in Oregon if Portland intends to expand its boundary, it must notify the state and hold hearings The expansion
24 application filed by local government will be handed to Land Use Board of Appeals by the state government (LUBA). Dispute Resolution I n Oregon s planning process, he state has established a series of procedures to solve disputes arose between specific landowners and planning ordinances. First, the dispute should be considered at the local level, before a hearing officer. A fter the hearing, the decision made by the officer may be appealed to the local government s city or board of county commissioners. A t the same time, a second appeal may be handed to a spe cial review board under the Land Use Board of Appeals (LUBA). L UBA is serviced as an e xclusive jurisdiction to review any land use decision made by the local government, special district or state agency ( Anderson, 1999). A nd the final decision would be mad e by state s circuit courts. Before LUBA was created, the land use appeals were reviewed by LCDC and the circuit court. T he establishment of LUBA is to simplify the appeal process, speed resolution of land use disputes and provide consistent interpretatio n of state and local land use laws ( Oregon Government, 2010). Moreover compared to the slow and costly regular appealing processes handled by LUBA the landowners can turn to alternative resolution like mediation and arbitration which are reviewed by LC DC T hese also are the very common dispute resolution methods in the Oregon State. I n the mediation methods, the mediators are chosen by both parties to handle the local land use dispute, whose service fee s are paid by the LCDC. Unintended Effect of UGB D uring the implementation of the UGB in Portland, an unintended side effect occurred in the boundary expansion process Hobby Farmer s. The Hobby Farmers,
25 which is a new interest group, consisted of non commercial farmers. According to the Staley and Mild ner s (1999) Urban Growth Boundaries and Housing Affordability study, they s tated that b y planting a field of Christmas trees or a large patch of strawberries, these landowners have been able to get rural homebuilding permits under the exemption for far mers These new houses built by theses farmers on some rural large parcels seem to create exurban sprawl outside the boundary The exurban development mainly occurred on the exception areas which are pre commitment to urban use or limited usefulness for f arm and forest land (Knaap & Nelson, 1992). On the other hand, the farmers group strongly oppose to the expansion of the UGB s worrying the future expanded urban development will destroy the quality of their lives in the rural areas T he normal expansion p rocess of the UGB s is interfered by these farm ers, which leads to constrain the supply of land more scarcely It is the least the city planners want to see in the UGB s implementation process. The Conflicts between Growth Management and Housing Affordabil ity A fter the adoption of UGB s nationwide there have been increasing concerns regarding the relationship between growth management and local housing affordability. Some scholars claim there are conflicts arising between g rowth boun d ar ies and housing affor dability, which cause s the local housing market to be less affordable. T his section will discuss these conflicts, which services as an overview of the relationship between UGBs and housing affordability. I n the dissertation Myths & Facts: About Growth Ma Morrill and Hodge (1991) analyzed myth s regarding growth management and housing affordability. They argued that contradic tory to the common understanding, the fact is that the increased housing prices is not resulted from the demand of newcomers in the city or the higher
26 construction cost, but the reduction of land sup ply caused by growth management. F or example, from 1983 to 1990, the construction cost in Seattle rose 16 % while the housing price rose roughly 60 % ( M orrill & Hodge, 1991 ) In compa ring the demand of newcomers and construction cost, they concluded that the growth management reduces the economically available supply of land, therefore raising its costs and the costs of houses on it ( Morrill & Hodge, 1991 ) Nevertheless t he limitation s in their study included that their arguments were not pr esented in detail and lacked sufficient data to support. O n the other hand, Gerrit Knaap and Lewis Hopkins (2001) considered UGBs as an instrument of inventory control that concentrate s on when UGB should be expanded and by how much to mitigate its prices pressure on housing market I f UGBs are implemented successfully to manage the growth, the supply of land must be constrained by the boundary for promoting higher density development. Since the land value is a main cost in the housing cost, the UGB should be expanded before the land value is being pushing up too much which caus es significant impact on the housing price (Knaap & Hopkins, 2001) T he ideal and simple solution is to know wh en to expand the boundary and how much land it should include. I n the real world, the time frame of expansion and the amount of land that should be included are almost impossible to predict. However, t hey insisted that the UGBs tend to constrain the land s upply and push up the land value, which eventually affec t s the housing affordability. In addition, K na a p and (2001) research p rovided three possible logical method s to avoid the negative impact s caused by UGB lead time, safety stock, and market
27 f actor inventory T hese methods explained the when and how much problem in different time periods during the expansion decision making process. Nelson et al. (2002) studied the link between growth management and housing aff ordability. They explained that the market demand, instead of land constrain s caused by growth management tools, is the primary determinant of housing prices. Furthermore, Nelson et al. (2002) believed that both the tr aditional land use regulations (i.e. zoning low density only develo pment etc ), and growth management policies can raise the price of housing, mainly by constraining the supply of land. I n their understanding, the effects of UGB in Portland suggested that UGBs can affect land values, however, their impacts on the housing affordability is still in dispute (Nelson, 2002). A recent study done by Quigley and Rosenthal (2005) reviewed and analyzed the previous research es on the effects of growth management on the price of housing They criticized that most of the previous stud ies ignored the endogeneity of regulation and price, and the complexity of local policymaking and regulatory behavior Quigley & Rosenthal, 2005). I n their perspective a systematic and national measurement of land use regulations should be developed a nd a regulatory survey should be conducted to collect the local level regulatory information. I n summary, these previous academic studies showed th at UGBs tend to constrain the supply of land and then increase the land prices In regards to housing prices to some extent, the increased land prices will lead to higher housing prices inevitably However, the exact magnitude and impacts caused by the UGBs on housing affordability remain uncertain
28 Theoretical and Statistical Analysis o f Portland s UGB and Ho using Affordability Issue T he last section of this ch apter reviews previous studies of Portland s UGB and the housing affordability issue, which will be introduced in two small parts theoretical debates and statistical analysis. Theoretical Debates in Pre vious Studies The following debates on the effects of the UGB on Portland s housing affordability fo cus on the theoretical evidence provided in the previous studies. B esides the theoretical debates, some related simple statistical analyses are showed to su pport their arguments, such as demographic housing or economic data etc. I will start by describing the conflict s between the UGBs and housing affordability at the state level I n the a rticle written by Knaap and Nelson (1999), they reviewed the statew ide land use planning program in Oregon and then focused on the implementation of the UGBs at the local level, from the political and planning perspectives T hey believed that after the establishment of UGBs in Oregon, the UGBs appeared to affect the land value by providing the information about when the boundary will be expanded and by how much (Knaap & Nelson, 1992). In the housing market in Oregon, s tate law requires the local governments to meet minimum development density, which facilitate s the constru ction of multiple family units These requirements actually increase the potential supply of housing in the cities. However, according to Knaap and Nelson s (1992) research, little evidence exist to suggest that neither UGBs have constrained the suppl y of developable land and housing nor have resulted in more affordable housing market due to the higher density development in the city.
29 Staley and Mildner (1999) studie d the effects of the UGB on Portland s housing market in four categories : housing cost and prices, housing density, development land and infill development, and consumer choices. B ased on the housing prices and income level analysis, in 1999, Portland was ranked among the 10 % of the least affordable housing market in the nation and on the West Coast (Staley & Mildner, 1999). Secondly, the average housing density increased from five homes per acre to eight homes per acre, and the living quality decreased due to the shrinking size of average house unit (Staley & Mildner, 1999) S ince the boundary constrained the supply of land in side the boundary, the city had to develop vacant land (infill) and redevelo p the existing properties within the boundary, which l ed to higher construction cost and higher housing prices (Staley & Mildner, 1999). Staley and Mildner (1999) concluded that the Portland s UGB certainly contribute d to higher housing costs, whereas the magnitude was uncertain. T he same year, Staley Mildner and Edgens (1999) conducted another study of Portland s UGB and housing affordability issue T hey emphasized that the amount of vacant land has decreased from 75,000 acres in 1985 to less than 55,000 in 1999, and the Portland metropolitan area would face an 8,59 0 housing units deficit even with the achieved de nsities recommended in the Metro 204 0 Plan (Staley et al. 1999 ) A s a result, the housing prices increased significantly after 1990, for example, the housing prices in 19 94 was approximately 140% of the housing prices in 1985 (Staley et al.,1999) In addition Staley et al. (1999) observed that the UGB had push ed i nvestment inward, and forced higher density and inner city development. Because of pressure from different special interest groups, such as environmental activists, zero
30 expansion advocates, etc., the expansion of the UGB proved d ifficult, which further increased housing costs. Due to these issues, the UGB in Portland tended to constrain the supply of land and housing which lower ed local housing affordability. In another study, Goodstein and Phillips (2000) discussed the impacts of the UGB in Portland on residential development pattern s, focusing on rising land values and increased housing density T hey argued that the land values have risen after the adoption of the UGB and the average lo t size o f new residential development decr eased 13.5% and 20% respectively in Clackamas and Multnomah counties between 1991 to 1995 ( Goodstein & Phillips, 2000) During the same time period, Portland also experience d high rate of redevelopment and infill development which consisted of 29% of res idential development, much higher than 1980s (Goodstein & Phillips, 2000). Furthermore besides smaller lot size s, the trend toward s higher density caused by the UGB led to smaller yards, fewe r open space and less privacy in new housing development. In c omp aring the median housing prices in the Portland metropolitan area to other western metropolitan areas the result s showed that the median housing price increased almost 70% during 19 91 to 1996 ( Table 2 1). However, due to other explanations, such as increa sed demands of the market speculative influences etc., the increase in land values and housing density did not necessarily indicate that the UGB led to increased housing prices in Portland To weigh all other factors, they conducted a regression model to test price effects of the UGB on Portland s housing prices, which will be discussed in the statistical analysis chapter
31 Table 2 1. Median housing prices in major w estern m etropolitan a reas Metropolitan Areas 1991 1993 1996 %Changed San Francisco $2 75,000 280,000 289,000 5.09% San Diego 171,000 163,000 165,000 3.90% Seattle 135,000 140,000 153,000 14.33% Salt Lake City 85,000 102,000 146,000 71.46% Portland 85,000 108,000 144,000 69.41% Sacramento Denver Las Vegas Phoenix 145,000 90,000 97,000 85,000 140,000 101,000 117,000 109,000 135,000 130,000 123,000 120,000 6.90% 44.44% 26.80% 41.18% Source: Goodstein and Philips, 2000. Squires (2002) studied the data from the U.S. Census and Metro, and claimed the UGB in Portland has slowed down the spr awling of the urbanized area and in creased the population density F or example the population (per sq. mile) totaled 4,517 in 1950, 2,940 in 1980 and 3,167 in 1994 at the meanwhile, the area (sq. miles) is 114 in 1950, 349 in 1980 and 388 in 1990 The UG B in Portland also created a dual land market inside and outside the growth boundary. Th e land outside the boundary lost its speculative value, which is limited to agricultural use, while the land inside the boundary remained or even gained speculative val ue. T hey also observed that the new housing development density increased, while the average new lot s ize for housing development decreased from 12,800 square feet in 1978 down to 6,200 square foot in 1998 (Squires, 2002). T hese are all the direct and indi rect effects of the UGB on the housing prices, which intended to lower the housing affordability in Portland. These are the critical comments about the effects of Portland s UGB on local housing affordab ility; however, the 1000 Friends of Oregon group hel d a different opinion regarding housing prices. They believed that the housing prices in Portland are similar to, or lower than comparable cities, which is considered as an affordable housing market for renters and buyers (1000 Friends of Oregon, 1999). As a response to the
32 constrain of the supply of land caused by the UGB, the 1000 Friends of Oregon (1999) claimed that the price of land is a small portion of the housing prices and the rapidly increasing housing prices in 1990s was mainly result from the e conomic upturn but not the UGB. T aking Los Angeles as a comparable case, the housing prices in Los Angeles cost $30,000 more than in the Portland region, which indicates r elatively unlimited land supply could not assure more affordable housing (1000 Frien ds of Oregon, 1999). The shrinking average lot size was not caused by the UGB, but mainly resulted from the demands from local people, as the age and families in Portland continued to get younger and smaller. T he Statewide Planning Program in Oregon, inclu ding the UGB in Portland, creat ed a more affordable housing market for the local people by reducing the amount of property taxes and infrastructure paid by every homebuyer. Contrast to the above studies, Jun Myung Jin (2004) focused on the effects of Po UGB on the urban development patterns based on Census data from 1980 to 2000 H e compared the urban development pattern in Portland to other 31 metropolitan areas for the following variables : urbanized population, urbanized area, population densi ty in urbanized area, employment in central city, housing units in the urbanized area, auto and transit users and mean commute time ( Table 2 2 ). T he results showed that Portland did not appear to have less suburbanization higher infill development or decr eased auto use, compared to other metropolitan areas (Jun, 2004). Furthermore he conducted a housing supply regression model to examine what factors affect ed the location of new housing and whether the UGB affect ed the urban residential development patter n T he most important result of the analysis indicated that the UGB variable was not statistically significant in both 1990 and 2000 models.
33 Therefore, he concluded that the location of new housing is not affected by the UGB variable (Jun, 2004). Table 2 2 Percentage c han ge of six v ariables of u rban d evelopment p attern in Portland, 1980 2000 (J un 2004). Variables Percentage changed, 1980 2000 Rank (out of 32) Urbanized Population (000s) 54.3 8 Urbanized Area (square miles) 35.8 9 Population Density 13.6 15 Employment in central city Housing units in urbanized area 70.8 54.4 6 16 Auto users 69.9 12 Public transit users 26.1 11 Mean comminuting time 14.5 15 Source: U.S. Bureau of Census, STF3, 1980 and 2000, (Jun, 2004). I n summary, most of the p revious studies observed that the UGB tended to increase the land value, mainly because of its constrain of supply of land and preference for higher density development. Moreover the urban development pattern was also changed by the implementation of the UGB such as higher redevelopment and infill development rate, increased housing density, etc. Demonstrated in the above discussion a ll these observed direct or indirect effects caused by the UGB seem ed to lower the hou sing affordability in Portland. H ow ever, the exact impacts is unclear and not well supported by statistical analysis. Sta tis tical Analysis Indicators and Regression Models In above theoretical analysis, the conclusions about effects of the UGB on Portland s housing affordability are lack of s tatistical analysis to support. To answer the question whether the UGB affects the housing prices and by how much, I will study the previous statistical researches which include simple indicators and several multiple regression models. T he indicators demonstrates the ranking of Portland s housing
34 affordability among major metropolitan areas while the regression model shows that whether the UGB in Portland contribute significantly to housing prices, as well as the factors that co ntributes to housing prices most. Indicators HOI and H+T Affordability Index The first indicator is the Housing Opportunity Index (HOI) provided by National Association of Home Builders (NAHB). The HOI, that includes local median income and housing cost, is defined as the share of housing prices that would have been affordable to median family income, based on the standard mortgage underwriting criteria (NAHB, 2011). According to the information from the websites of NAHB, the data of median family income o f metropolitan areas are collected from Department of Housing and Urban Development (HUD) and the data of housing costs are from Core logic. T o be consistent with the conventional assumption in the lending market, the NAHB assumes that a family can afford to spend 28 % of its gross income on housing (NAHB, 2011) I n the national HOI ranking, the Portland metropolitan area ranked the 181st among 226 metropolitan areas. Moreover between 1999 and 2002 2006 and 2010 the ranking range of Portland s HOI was f rom 127 th to 1 98th in the nation However, in the year 2003 2006, the housing affordability of Portland dramatically ranked 65 th to 109 th which was the highest ranking ever recorded (NAHB, 2011). T he reason of these sudden changes was unclear. C oncluded f rom the above the data, the housing market in Portland is considered unaffordable, approximately 180 th among more than 200 metropolitan areas. L ater, Center for Neighborhood Technology (CNT) developed another affordability index Housing and Transportation Affordability Index (H+T Index) Since the transportation cost is the second largest expenditure after housing of one family CNT
35 included the transportation cost into the affordability calculation. Therefore, H+T Affordability Index is calculated as the t otal expenditure of housing and transportation cost dividing annual household income. A s the transportation cost, it was calculated by a compl icated model (Figure 2 2 ) which includes not only the cost of commute to and from work, but also the daily trave l cost (CNT, 2011). Furthermore, the transportati on cost is highly related to location, characteristic s of the neighborhood and urban form. A ccording to the basic urban economics theory prices adjust to ensure location al equilibrium (Sullivan, 2009, p.28) the transportation cost is directly related to the housing location that i t is the trade off between the distance to work and housing prices or rent. C ompared to the traditional 30% standard, the H+T Affordability Index defined that the total expendit ure less than 45% of the income is considered as affordable In addition the index covers 80% of the total population 337 metropolitan areas and 161,000 neighborhoods (CNT, 2011). It is important to note that the H+T Affordability Ind exes of the metropol itan areas are based on neighborhood scale.
36 Figure 2 2 The e stimated m odel of t otal transportation c ost in the H+T Affordability Index. (Source: http://htaindex.cnt.org/method.php Last accessed Janua ry, 2011 Edited by Author ) According to the H+T Community Profiles (CNT, 2011), 69% and 39% of the selected communities are considered as affordable respectively based on the 30% and 45% standards Only sixty four percent and thirty seven percent of Portland s communities are considered affordable based on the same standard, which are both lower than the national average percentages. A mong the selective 14 metropo litan areas, Portland ranked 14th and 9th in these two different affordability standards and it is not considered as an affordable housing market More importantly t he percentages of affordable communities in all the metropolitan areas decreased when the affordability index takes account of the transportation cost. T he percentage changed bet ween 30% and 45 % standards indicates that in certain cities, the transportation cost caused more
37 significant impacts on local housing affordability for example, Charleston, Duluth Superior, Daytona Beach, etc. Showed in Figure 2 3, Boston, Washington D.C. and Chicago have a relative high and stable housing market than other metropolitan areas, while Portland s housing affordability index is slightly lower than the national average level. Figure 2 3 Percentages of communities considered a ffordable u sing 30% s tandard and 45% H+T Affordability Indexes of m ajor m etropolitan a reas. (Source: http://htaindex.cnt.org/metro profiles.php Created and edit by author ) T he Figure 2 4 illustrates the locations of affordable communities in the Portland Metropolitan area. T he icon 1 2 and 3 in the map stand for main city of counties in the metropolitan area, Portland, Cedar Hills in Washington County, Troutdale in Multnomah County and Clackamas C ounty. A s seen in the figure, most of the
38 neighborhoods located in the Portland City area are considered as affordable, based on both standards Figure 2 4 Map of the l ocat ions of affordable c ommunities in Portland Metropolitan Area under 30% and 4 5% Standards (Sour ce: http://htaindex.cnt.org/metro profiles.php L ast accessed January, 201 1. E dited by author ) In sum, HOI showed that Portland s housing market is not considered as affordable, far lower than the national average level; while the 30% and H+T Affordability Index analysis held the opinion that the Portland s housing affordability is approximately at the national average level. However compared to HOI analysis, the results of th e H+T Affordability Index is more comprehensive and reliable that it contains transportation cost which is a nother crucial factor that contributing to the housing affordability. Multiple regression models B esides the housing cost, annual income and trans portation cost variables mentioned above, I will review the previous s housing affordability regression
39 model, which take into account of the price effects of all other factors on local housing affordabi lity. Though all h ousing affordability model s focused on the Portland UGB and the housing affordability issue s therein the studies were conducted in different ways, mainly limited by available data and the conducting date of the research T he results provided us with the statistical analysis of eff ects of Portland s UGB on housing affordability from different perspectives Started from 1985, Gerrit Knaap conducte d a regression model to analyze the effects of the UGB on residential land values in the metropolitan Portland area (Clackamas, Multnomah and Washington Counties) T he data consisted of all the arm s length residential land transaction s within the Portland s Urban Growth B oundary in the fiscal year 1980 The selective variables captured the influences on the residential land values, includ ing characteristic s of the transaction, physical features of the land, constrain s of the zoning, etc., while a dummy variable was used to indicate the relationship of the parcels of land to the UGB (Knaap, 1985). I n this model, the dependent variable was t he sales price divided by the number acres in the sale. F rom the results of the analysis Knaap concluded that the land values varied significantly across Clackamas and Washington Counties, and it were lower outside the UGBs of these two counties while th e land values remained the same in Multnomah County. However, lack of additional information, the model did not test for the mechanism how the UGB caused impacts on the land values, for example on the values of land in the urban zones, of the land within 300 feet of a sewer line, etc. (Knaap, 1985). In addition, Knaap only tested the relationship between the location of land parcels to the UGB of Portland, but not focusing on the housing affordability.
40 Alkadi (1996) used a hedonic model to conduct a time s eries analysis to test the relationship between housing prices and the UGB in Washington County of the Portland metropolitan area. T he dependent variable was the sales prices of individual houses in Washing ton County before and after the adoption of Portla nd s UGB, and 29 independent variables were included in the housing prices model collected from supply d emand, accessibility factors, public service, structure and site factors of housing transactions etc. during 1978 to 1990. UGB was adopted as a dummy variable with UGB = 1 indicating the house was sold after the implementation of the UGB in Portland October, 1980 and TIME variable indicated the exact time of the house sold, from January 1978 = 1 to December 1990 =156 (Alkadi, 1996). According to the st atistical results, Alkadi (1996) concluded that there was no relationship between housing prices and the implementation of the UGB, wh ereas the increase rate of housing prices was found to be less after the implement of the UGB than before. It is important to note that the distance of sale to the UGB was the only variable tested to be related with a higher increase rate in housing prices. Following the previous theoretical debate, Goodstein and Phillips (2000) used a regression model to test the price effe cts of all the factors on the Portland s housing prices, which was a cross sectional analysis among 37 major cities throughout the country from 1991 to 1996. T he model contained 9 independent variables which reflected the supply and demand side s and specul ation in the housing market. The estimation of the price effects of the UGB was represented as a proxy variable in the regression model. C omparing the predicted median housing prices to the Portland actual median housing prices the results demonstrated t hat the UGB has probably
41 slightly increased the median housing prices in Portland, approximately less than $10,000(Goodstein & Phillips, 2000). T hey concluded that the increased housing pri ces during the study period mainly resulted from the rapid employme nt and income growth. T wo years later, their conclusions were supported by another study conducted by Anthony Downs Compared to previous Portland s housing affordability regression model s, Downs (2002) conducted a more comprehensive regression to test pr ice effects of 25 key independent variables on the dependent variable percentage incre ases in home prices during various time periods There were five time periods in the regression models, 1990 to 2000, 1990 to 1994, 1990 to 1996, 1994 to 2000 and 1996 to 2000, which were selected due to the available home prices data (Downs, 2002). I n the analysis process, the first step was to run the regression with all the variables which aimed to find out the variables that were stati sti cally significant to the depen dent variable. I n the second step, these statistically significant variables would be included into the best regression models for each time period More importantly once the best models were developed, the UGB dummy variable would be also included in to each of the model to test the price effects of the UGB in five different time periods. I n the results, after the dummy variabl es were included, the adjusted R square values of three time periods ( 1990 1994, 1990 1996, and 1990 2000 ) rose significantly, while the adjusted R square values of the other two time periods (1994 2000 and 1996 2000) dropped (Downs, 2002). Downs (2002) concluded that the UGB only had significant impacts on the increasing housing prices from 1990 to 1994, but not during other tim e periods between 199 0 and 2000. H e also gave the explanation that only in the time period 1990 1994, the job and income growth rate increased rapidly as well as the
42 home prices, and the estimated 20 year land supply did mitigate the UGB s constraining im pacts on the housing market. I n the same year, Fischel (2002) reviewed the studies conducted by Downs (2002), Goodstein and Phillips (2000), and Knaap and Nelson (1992) and made his own comments and opinion s. Downs, Goodstein and Phillips had the similar conclusion that the rapid rose in housing prices during early 1990s was mainly attributable to rapid job and income growth rather than the constrain of the UGB, while Knaap and Nelson reached a different conclusion (Fischel, 2002). H e provides the explanat ion that the densities within the UGB had been increased due to the statewide growth management requirements before the implementation of the UGB. As a result, the potential housing supply in late 1980s was high enough for the future increasing housing dem and. Myung Jin Jun (2006) used a hedonic prices regression analysis to assess the mean ho u sing prices in the Portland metropolitan area in the year 1990 and 2000 which took account of the building structure, housing market and accessibility factors Simi larly, t he UGB wa s presented as a dummy variable. T he results of the regression model showed that the UGB variable had no significant effect on the housing prices, inside or out the growth boundary areas in year s 1990 and 2000 (Jun, 2006) However further research question s were generated by the results in that the research should also consider the relationship between land values, housing density and housing prices, because these variables are highly correlated (Jun, 2006). Grout et al. ( 2009 ) conducted a regression discontinuity design to explore the price effects of Portland s UGB on the vacant land prices. T he data of the samples were the undeveloped parcels of land which were located in and around the Portland s Urban
43 Growth Boundary, and the distance to the UGB variable was measured as negative values inside the UGB and positive value outside the boundary (Grout et al., 2009). The whole study area was divided into 12 sections which indicated the different locat ion s of the vacant parcels in or around t he UGB. T he results demonstrated that only the vacant land in Western Portland was significantly influenced by the distance to the UGB variable which created a discontinuity point within Portland s UGB H owever, other vacant land within the Portland s UGB did not appear to be affected by the UGB variable, so there were not sufficient evidence to establish any relationship between the implementation of the UGB and the vacant land values Summary of Previous Studies on Portland s Issue The above studies exp lored and analyzed the relationship between Portland s UGB and local housing affordability issue both qualitatively and quantitatively. Th ough their research design s are not exactly the same, most of the studies reached the similar conclusion that the impl ementation of Portland s UGB did not seem to affect the local housing affordability. I n the theoretical analysis section most of the researchers insisted that the implementation of Portland s UGB cons train the supply of land for future urban development, which tends to increase urban density, redevelopment and in fill development rate, and eventually could push up the land prices. Even if the boundary contains 20 year supply of land for future development and encourages higher density, Portland still coul d have a housing deficit in 2040 (Staley et al. 1999 ). T here are the main theoretical opinions presented by previous studies UGB definitely caused price effects on local housing marke t.
44 However, the results of previous statistical studies reached the different conclusion that Portland s UGB did not affect local housing prices. On the contrary, housing prices were affected by the growth in income and employment opportunities. D ue to the fact that the statistical analyses were adopted in diff erent years between the time periods from 1985 to 2009, these studies had different research design and limitations such as different available data, research time frame methodology, dependent and independent variables etc. In the studies adopted in ear ly year s the Portland s UGB was tested to have slightly affected the land prices and housing prices. However, in the later studies, with more available data and comprehensive research design, the conclusion was that the UGB in Portland did not contribute to increasing local housing prices.
45 CHAPTER 3 METHODOLOGY The literature review from the previous chapter provided background information demonstrating the controversial relationship between UGB and housing affordability. Based on the previous arguments, the growth boundary intends to push up land prices and change urban development pattern s which leads to lower housing affordability. This study used both qualitative and quantitative analysis to test two research hypotheses. As mentioned before, the two research hypotheses are: (1) D ue to the UGB constrain on the supply of land, the trends of land prices and land share in the Portland metropolitan area should be higher than other metropolitan areas without UGBs in United States. Since land prices have a s ignificant impact on housing prices, the housing prices in Portland should be also higher. I n addition the land and housing prices in Portland are highly correlated; (2) besides affecting the land prices, the UGBs also tend to change the urban development patterns, such as higher density development, high redevelopment and infill development rate of the inner city, etc. T he regression model takes into account all of the factors that contribute to housing prices to measure the price effects of different pre dictors on the median housing prices among 35 urbanized areas. The details in these two statistic analysis are not similar. T he former one is to test the correlation between land prices and housing prices in major metropolitan areas in America; while the latter one is a multiple regression model of median housing prices among major cities.
46 Correlation between Land Prices and Housing Prices I n this meth od, the correlation analysis consists of depictive analysis of trends in land prices, housing prices per centage of land share in housing prices, and a Spearman correlation analysis between the two based on a data set of 10 major metropolitan areas. T he purpose of this analysis is to provide clear answers for the following research question s : According to the trends analysis from 1984 to 2009, a re the land prices and the percentage of land share (the proportion of land prices in the housing prices) higher in Portland than other 10 metropolitan areas without the control of the UGB and their average trend? Are the housing prices higher in Portland than other 10 places, as well as their average trend? B ased on the results of the Spearman correlation, is the significan ce of the correlation coefficient value higher in Portland than other metropolitan areas without the UGB ? Data Source and Study A rea The data about land prices and housing prices of different major metropolitan areas are assessed from the of Lincoln Insti tute of Land Policy, which contain s the information about land, structure and housing prices bac k from 19 84 to 2009 quarterly. T he recorded housing prices come from real estate trans ac tions however, the land prices are rarely recorded due to most of direct land sales are occurred in built up areas and new suburban development areas ( Lincoln Institu te of Land Policy, 201 0).Therefore, the land prices are calculated by residual method housing prices subtract ed the cost of structure and the percentage of land share was land prices dividing housing prices. The area of focus for this method was the metrop olitan area of the city due to all the data from Lincoln Institute of Land Policy are collected within this boundary. However, due to some metropolitan areas contain many non densely developed or undeveloped
47 lands the tested results intend to enhance the gap between Portland and other metropolitan areas. T he re are 10 comparable metropolitan areas used in the analysis (Table 3 1), which have he similar population size as Portland metropolitan area, but without UGBs. Table 3 1 List of the metropolitan a r eas for the p rices trend a nalysis and Spearman Correlation Name of Metropolitan Area s Total Population (000s) Portland OR 2,265,223 Average of 10 Metropolitan Area 2,323,848 Cincinnati, OH 1,979,202 Cleveland, OH 2,945,831 Indianapolis, IN 1,607,486 Milwaukee, WI 1,689,572 Phoenix, AZ 3,251,876 Pittsburgh PA 2,358,695 San Antonio, TX 1,592,383 San Diego, CA 2,813,833 St. Louis, MO 2,603,607 Tampa, FL 2,395,997 Source: U.S. Bureau of Census, Census 2000 Prices T rends A nalysis and Spearman C orrelation Coefficient To answer the above question s I will use the line charts which show the trends of land prices, percentage of land share and housing prices to find out the relationship between land and housing prices among these 11 major metropolita n area s. Moreover, the Spearman Correlation C oefficient was adopted to test the significant level between the land and housing prices. T he line charts give a qualitative description of the prices and land share trends, while the correlation coefficient pro vide more quantitative analysis for land and housing prices Finally I ranked these metropolitan areas according to their value of significance in the statistical result.
48 Median Housing Prices Regression Model C ompared to the first method that f ocused on the correlation between land and housing, this multiple regression model will consider all the factors that contribut ed to housing prices, which is a more comprehensive model. T he purpose of this regression is to provide answers to these research questions : A ccording to the results of correlation coefficient are all the selected independent variables statistically significant to the median housing prices? Followed the first two questions which variables are statically significant in the regression model and contributing to median housing prices most? Is the UGB variable causing impact s on the median housing prices? If not, is the UGB variable correlated to some of the independent variables, such as housing density, population density, median commute time, etc? B ased on the statistical results of the regression model, the best median housing prices model with the highest adjusted r will be discussed and explained. Data S ource and Study A rea In the regression, the dependent variable is the median housing prices of major cities in the states and 9 pred ic tors are considered in the model. Most of the data are downloaded from the Census 2000 from the website of U.S. Census Bureau T he basi c of the data will be described in the following section. T he sample size of the cities is 35, and the cities names are listed in detail in Table 3 2. Among these 35 samples, 10 cities have adopted some forms of growth boundary planning tool s, such as urban service area, intergovernmental agreement, etc.
49 Table 3 2 List of the u rbanized a reas in the m ultiple r egression m odel Location / Urbanized Areas West Coast (6) Southwest (7) Midwest (11) Southeast ( 4 ) East Coast ( 7 ) Portland, OR San Jose, CA Seattle, WA Boulder, CO Denver CO Twins Cities MI Lexin gton, KY Memphis, TN Miami, FL Virginia Beach, VA Los Angeles, CA San Diego, CA San Francisco, CA Dallas, TX Houston, TX Phoenix, AZ Salt Lake City, UT San Antonio, TX Buffalo, NY Chicago, IL Cincinnati, OH Cleveland, OH Detroit, MI Indiana polis, IN Milwaukee, WI Pittsburgh, PA St. Louis, MO Atlanta, GA Charlotte, NC Tampa, FL Baltimore MD Boston, MA New York, NY Philadelphia PA Washington DC Source: Edited by author. T o be noticed, differently from the prices trend analysis the resear ch area in the analysis in not metropolitan area, but the boundary of urbanized area (UA) of each city. D efining the study area among different cities is essential to the accuracy of the final results. A metropolitan area, according to the definition made by U.S. Census, is a substantial central nucleus, surrounded by the adjacent area s which have highly economic linkages with the central nucleus; while the urbanized area is referred to a central city and its surrounding highly built up area s (Greene and Pi ck, 2006). T he reason I choose urbanized area to be the study area is because the metropolitan area contain s many not highly built up areas surround ing the central cities, which intends to affect the final results of some predictors More importantly to th e cities controlled by growth boundary, the metropolitan area contains many areas outside the boundary which are not built up, mainly farmland and forestland. I n addition, the city boundary is
50 also not a n ideal study area due to it only consists of the cen tral city, ignoring some thickly developed areas around the central cities without the control of UGB. Variables in the Regression Model T his multiple regression model is cross sectional that it integrates the median housing prices of 35 urbanized areas in year 2000 which consisted of 9 independent variables (Table 3 3). A ll the variables were gathered from census 2000 database of the U.S. Bureau of Census and modified for the statistical purposes. T o be noticed, in the model the UGB variable is repre sented as a dummy variable that 0 stands for negative cities with no growth boundary, and 1 stands for positive city with growth boundary The purposes of the model aimed to explore the price effects of each independent variable on the median housing prices of the urbanized areas, especially focusing on the price effects caused by the implementation of the UGB. Table 3 3 List of v ariables in the median h ousing p rices r egression m odel Variables Expected Sign Data Measurement M HP, Median Housing Pr ices A HS, Average Household Size HD, Housing Density (per square miles) JD, Job Density MCT, Mean Commute time (per person) MHI, Median Annual Household Income PD, Population Density (per square miles) UGB, UGB Dummy UR, Unemployment Rate VD, Vehicle Density ( per square miles) Interval + Interval + Interval + Interval Interval + Interval + Interval + Ordinal + Ratio Interval Source: Edited by author.
51 CHAPTER 4 ANALYSIS AND RESULTS T his chapter presents two statistical analyses of the p rice effects of the Urban Growth Boundary ( UGB ) on Portland s housing affordability: price trend analysis to determine correlation between land prices and housing prices and the regression analysis to determine the prices effects of the UGB on the median h ousing prices through a regression analysis. These two methods are both cross sectional analyse s, comparing the Portland metropolitan area or urbanized area to other metropolitan area or urbanized areas in the United States. Furthermore, t he second part in this chapter discusses the above two statistical analyses of the UGB s price effect s in Portland presented and provides explanations for their results. The limitations of each method and the recommendation s to improve the research also will be discussed. Correlation between Land and Housing Prices T his method concentrated on the relationship between land prices and housing prices. According to the theoretical discussion in L iterature R eview C hapter Portland s UGB intends to affect housing prices by const rain ing the supply of land and increasing the density of development. There are two parts in this analysis process First the trends of land prices, land shares (the percentage of land prices in housing prices) and housing prices in Portland and 10 other metropolitan areas without UGB s. S econd the Spearman Correlation Coefficient is used to explore the signif icance level between land and housing prices in these areas.
52 Prices T rends Analysis Figure 4 1 and 4 2 show the trends of land prices and land shar e for the Portland metropolitan area and 10 metropolitan areas without the control of a UGB between 1984 and 2009 A s demonstrated in Figure 4 1, except for Phoenix, Portland and San Diego, the remaining metropolitan areas had similar land prices trend s b etween 1984 and 2003 However, among these metropolitan areas, the land prices in Milwaukee and Tampa increased faster than the other six metropolitan areas between 2004 and 200 8 Obviously the land prices in San Diego were not comparable to the other met ropolitan areas because it r emain ed much higher than all other areas during the whole time period B etween 1984 and 1989 in Portland the land prices trend was almost identical to the other areas excluding Phoenix and San Diego; the prices began to increa se beyond those from the other eights areas and the average of the 10 metropolitan areas after 1990, and following this trends f or the rest of the time period. Portla nd and Phoenix had similar land price trends between 1990 and 2007, but the land prices in Phoenix dropped significantly after the economic downturn in 2007. Except for extremely high land prices in San Diego, the Fi gure 4 1 illustrates that the land pric es in Portland continued growing and remain ed in a higher position even during the economic crisis from 2007 to 2009. Again in the land share trends, San Diego had a much higher land share percentage than all other metropolitan areas ; the value was more than 60% during the whole time period and peaked at about 82% in 2006. H owever, higher land prices were not necessarily a ssociated with higher land share. The land share percentage trends were below 40% of the housing prices in the remaining metropolitan areas excluding
53 Phoenix and Portland Notable among these low land share metropolitan areas Tampa exceeded 40% after 2000 and peaked at about 58% in 2005. A fter that, it dropped dramatically to 12 % in 2009. T he land share in Portland continued increasing steadily and exceeded 40% in 1990 peaking at 62% in 2006. I n 2007, d ue to the economic crisis the land share in Portland dropped to 46%, higher than all the other metropolitan areas (except San Diego) an d the average of these areas As seen in Figure 4 3, the housing prices trends of these 11 metropol itan areas illustrated a pattern simil ar to the land prices trends S an Diego was still leading among all the metropolitan areas with much higher housing prices from 1984 to 2009. T he other areas had a similar grow th pattern in that the housing price s increased slowly but steadily over time, w hereas housi ng prices in Portland started to grow faster from 1990, peaking at $ 384 404 in 2007. Even though the housing prices in Phoenix were higher than Portland between 2004 and 2007, Phoenix prices dropped dramatically during the economic crisis to t he point where they were, much lower than those in Portland by 2009. S imilar to the land price trend the housing prices in Portland grew constantly and remained at a higher level than other metropolitan areas during the whole time period, except for San D iego and Phoenix.
54 Figure 4 1 Trends of land p rices of 11 m etropolitan a reas. (Sour ce: Lincol n Instit ute of Land Policy Edited by author )
55 Figure 4 2. Trends of p ercentage of land s hare of 11 m etropolitan a reas. (Source: Linc o l n Institu te of Land Policy Edited by author)
56 Figure 4 3 Trends of h ousing p rices of 11 m etropolitan a reas. (Sour ce: Lincol n Institut e of Land Policy Edited by author )
57 Spearman Correlation C oefficient The above price trends analysis illustrated that there was a certain relationship between land and housing prices. Therefore, the Spearman Correlation Coefficient was utilized to measure the correlation between land and housing prices in each metropolitan area T he results of the Spearman two tail tests as seen in Table 4 1, indicated that land and housing prices in each metropolitan area were significa ntly correlated to each other Table 4 1 Ranking of the S pearman C orrelation C oefficient of 10 metropolitan a reas Name of Metropolitan Area s Spearman Correlation Coefficient (Sig. at 0.01 level) Ranking Portland, OR .997 1 San Diego, CA .992 2 Milwa ukee, WI .988 3 St. Louis, MO .927 4 Pittsburgh, PA .917 5 Cincinnati, OH .890 6 Cleveland, OH .803 7 Phoenix, AZ .752 8 San Antonio, TX .701 9 Tampa, FL .685 10 Indianapolis, IN .548 11 Source: Lin coln Institution of Land Policy, edited by author T he Spearman C orrelation C oefficient is a commonly used measure of the size of an effect and that value of .1 represent a small effect, .3 is a medium effect and .5 is a large effect (Field, 2005, p.111 ) In the results presented in Table 4 1, the coefficients w ere all larger than + .5, indic at ing that land prices had a large effect on housing prices in all the metropolitan areas Portland, San Diego and Milwaukee had the highest coefficient s among th e metropolitan areas ( close to + 1 ), whereas the coefficient for Indianapolis was just slight ly higher than + .5. Portland had the highest coefficient among all the areas at 0.997. Th ough Phoenix and Portland had similar
58 pattern s for land and housing price trends, the Phoenix coefficient was 0.752 This value was much less than 0.997 and r anked at 8 th among all the metropolitan areas. In su m m ary land prices, land share s and housing prices in Portland grew faster than other metropolitan areas after 1990 and remained at a high level even during the econom ic downturn. Only San Diego had a higher land prices, land share s and housing prices than Portland during the whole time period. M oreover, b ased on the results of the Spearman two tail tests, these two metropolitan areas had almost the same coefficient va l ues 0.997 in Portland and 0.992 in San Diego. However, there is no evidence showed that the rapid growth of the land and housing prices in Portland was caused by the UGB. Median Housing Prices Regression Model Compared to the correlation analysis between land prices and housing prices th is regression model is more comprehensive and accurate statistical method that it aims t o measure the price effects of eight other independent variables besides UGB dummy variable on the median housing prices among 35 ur banized areas. T he firs t part of the analysis is to find out which independent variables are statistically significa nt to the median housing prices. I n the second part, based on the results of the regression model the best model s to measure the median ho using prices will be explained in detail Correlation C oefficient and Scatter Plot Analysis Before presenting the best median housing prices regression models, the relationship between dependent variable and each of the independent variable s were tested to see if they were statistically significant to the median housing prices (Table 4 2). Since the interval or ratio data were not normally distributed, the Spearman
59 Correlation Coefficient was used in the test which i s a non parametric test and its reject ion zone is less than 0.05. Table 4 2 Correlation c oefficients of dependent v ariable and i ndependent v a riables in the r egression m odel Independent Variables Sig (two tail test at 0.01 level ) Correlation coefficient A HS, Average Household Size HD, H ousing Density (per square miles) JD, Job Density MCT, Mean Commute time (per person) MHI, Median Household Income (annually) PD, Population Density (per square miles) UGB, UGB Dummy UR, Unemployment Rate VD, Vehicle Density (per square miles) .198 ( .223 ) .0 02 .496 .0 01 552 .0 29 368 .00 0 780 .001 .535 .314 ( .175 ) .148 .000 ( .250 ) .615 Source: U.S. Bureau of Census, 2000 Census edit by author. T he results of the Spearman t ests indicated that only the average household size, UGB and unempl oyment rate were not statistically significant to the median housing prices dependent variable. The significance of the coefficients of these three variables was larger than 0.05, which meant that there is no relationship between median housing prices and average household size, UGB and unemployment rate. Furthermore, a s seen in Figure 4 4, 4 5 and 4 6, the dispersion of median housing prices and these three variables did not illustrate any correlation in the scatter plots maps. Th ese result s indicated that the UGB variables were not correlated to the median housing prices in the regression model. In addition the straight line in the scatter plots indicated the mean values of the all the sample point, but not necessarily indicating any linear relationship b etween the tested two variables.
60 Figure 4 4 The s catter p lot of m edian h ousing p rices and a verage h ousehold s ize. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author ) Figure 4 5 The s catter p lot of m edian h ousing p rices and UGB d ummy v ar iable. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author )
61 Figure 4 6 The s catter p lot of median h ousing p rices and u nemployment r ate. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author ) On the other hand, six independent variable s in the model were statistically significant to the dependent variable T he coefficient s of job density, median annual household income population density and vehicle density were larger than 0.5, which indicated that these variables had strong positive effects on median housing prices. T he coefficients of housing density and mean commute time indicated these two variables had medium effects on median housing prices. The median household income and the vehicle density variables had the highest coef ficient s among all the variables, 0.780 and 0.615 respectively; while the coefficient of median commutes time was 0.368, which was the lowest value In addition, the scatter plots of median household income and mean commute time also confirmed their correlation s with median housing prices (Figure 4 7 and 4 8).
62 Figure 4 7 The scatter plot of median h ousing prices and m edian h ousehold i ncome. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author ) Figure 4 8 The scatter plot of median housing p rices and mean c ommute time (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author )
63 T he coefficient s of housing density, job density and population density had very close significance value, 0.496, 0.552 and 0.535 respectively Moreover the scatter pl ots also illustrated that these three independent variables and the dependent variable had a very similar dispersion pattern ( Figure 4 10, 4 11 and 4 12 ), which showed that these variables were highly correlated to each other I n addition, the dispersion o f median housing prices and vehicle density in the scatter plot also showed similar pattern (Figure 4 9) T his kind of similarity might result in collinearity problem s in the following regression analysis. Figure 4 9 The s catter p lot of m edian h ousing p rices and v ehicle d ensity. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author )
64 Figure 4 10 The scatter p lot of m edian h ousing prices and h ousing d ensity. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author ) Figure 4 11 The s catter p lot of median h ousing p rices and j ob d ensity. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by author )
65 Figure 4 12 The s catter p lot of m edian h ousing p rices and p opulation d ensity. (Sour ce: U.S. Bureau of Census, 2000 Census Edit by au thor ) Median Housing Prices Regression Model s As discussed above, the similarity of the dispersion between vehicle density, housing density, job density and population density would cause collinearity problem s in the regression analysis, which would lower the credibility of the median housing prices model. If a model contained two of more of the above four variables, the VIF would be larger than 10 in the regression result s which indicated two or more variables in the model were highly correlated. T o avoid the coollinearity issues in the regression models, there were four different regression models generated to measure the housing prices in these 35 urbanized areas which consisted of t hree independent variables and a constant Among these four regression models, t he same variables in four models were median household income and mean commute time T he different variable was one of the four variables vehicle density housing density, job density and population density T hese four median housing prices model s are:
66 Median Housing Prices= a1*(Median Household Income ) + a2*(Vehicle Density)+ a3*( Mean Commute time )+ a4*(UGB Dummy)+ b(Constant) Median Housing Prices=a1*(Median Household Income ) + a2*(Housing Density)+ a3*( Mean Commute time )+ a4*(UGB Dummy)+ b(Cons tant) Median Housing Prices=a1*(Median Household Income ) + a2*(Job Density)+ a3*( Mean Commute time )+ a4*(UGB Dummy)+ b(Constant) Median Housing Prices=a1*(Median Household Income ) + a2*(Population Density)+ a3*( Mean Commute time )+ a4*(UGB Dummy)+ b(Const ant) T he following four tables showed detailed information about the median housing prices models with four different independent variables v ehicle density (Table 4 3) housing density (Table 4 4) job density (Table 4 5) and population density (Table 4 6 ) A s demonstrated in the tables, the median household income had coefficient s ranging from 0.617 to 0.708, which indicated a large positive effect on the median housing prices, whil e the mean commute time had a medium negative effect on the median housing prices, ranging from 0.259 to 0.195. All four density variables had a positive effect on the housing prices with close coefficient values Furthermore, the vehicle and job densities showed large effects on median housing prices (greater than 0.5 0 ), whil e housing and populati on densities had medium effects (close to 0.5 0 ). I n the regression models, median household income was the most crucial variable s in measuring the median housing prices, which had a highest R square change, approximately 0.61, while the R square change of the mean commute time ranged from 0.031 to 0.052. Among all four density variables, the vehicle density had a highest R square change (0.190) in the regression model, followed by job density at 0.187. As to the median housing prices regression model, all four models had a very close adjusted R square value, approximately 0.810, which could explain about 81% of the housing
67 prices cases. T he job density median housing prices model had a highest adjusted R square value at 0.831 The UGB variable is not statistically significant in all four models and it tends to lower the adjusted R square of the regression approximately by 0.005. In the models, in spite of the effects caused by the UGB, the coefficient of the median household income an d density variables remai ned almost the same while mean commute time variable was affected much significantly. For example, in Table 4 3, the results of the vehicle density housing prices model indicated that the mean commute time variable was not statist ically significant due to the influence of the UGB variable. Table 4 3 Summary of the v ehicle d ensity median housing prices regression m odel Model Independent Variables Sig. (Two Tail Test) Coefficient Adjusted R square R square change VD 1 Constant M edian Household Income .000 .000 .780 .597 .609 VD 2 Constant Median Household Income Vehicle Density .000 .000 .543 .496 .786 .190 VD 3 VD 4 Constant Median Household Income Vehicle Density Mean Commute time Constant Median Household Income Vehi cle Density Mean Commute time UGB .000 .000 .024 .000 .000 .053 .863 .617 .516 .195 .614 .512 .188 .015 .813 .807 .031 .005 Source: Generated by SPSS, e dited by author.
68 Table 4 4 Summary of the housing density median housing prices r egress ion m odel Model Independent Variables Sig. (Two Tail Test) Coefficient Adjusted R square R square change HD 1 Constant Median Household Income .000 .000 .780 .597 .609 HD 2 Constant Median Household Income H ousing Density .000 .000 .653 434 .768 .1 73 HD_ 3 HD 4 Constant Median Household Income Housing Density Mean Commute time Constant Median Household Income Housing Density Mean Commute time UGB .000 .000 .004 .000 .000 .013 .901 .748 .485 256 .746 .483 .250 .011 .816 .810 .0 51 .00 6 Source: Generated by SPSS, edited by author. Table 4 5. Summary of the job d ensity m edia n h ousing p rices regression m odel Model Independent Variables Sig. (Two Tail Test) Coefficient Adjusted R square R square change JD 1 Constant Median Household I ncome .003 000 .780 .597 .609 JD 2 Constant Median Household Income Job Density .000 .000 604 .4 67 783 187 JD 3 JD 4 Constant Median Household Income Job Density Mean Commute time Constant Median Household Income Job Density Mean Commute time UGB .000 .000 .003 .000 .000 .0 1 3 .731 694 516 .25 3 .694 .516 .239 0.28 831 .826 .05 0 .005 Source: Generated by SPSS, edited by author.
69 Table 4 6. Summary of the p opulation density median h ousing p rices regression m odel Model Independe nt Variables Sig. (Two Tail Test) Coefficient Adjusted R square R square change PD 1 Constant Median Household Income .000 .000 .780 .597 .609 PD 2 Constant Median Household Income Population Density .000 .000 .618 .4 38 .7 59 .1 65 PD 3 PD 4 Constan t Median Household Income Population Density Mean Commute time Constant Median Household Income Population Density Mean Commute time UGB .004 .000 .000 .005 .004 .000 .000 .19 .626 .708 .493 .25 9 .698 .485 .238 .42 .8 09 .804 .05 2 .005 Source: Generated by SPSS, edited by author. I n sum, the best median housing prices regression model is a 1 *(Median Household Income ) + a 2 *(Job Density ) + a 3 *( Mean Commute time ) + b (Constant), which was able to explain about 83% of the housing prices cases. T he most significant variable s wer e the median household income and job density which had a large positive price effects on the median housing prices. In addition, the R square change of the median household income was 0.609, which was much higher than job d ensity (0.165) in the model. The UGB variables are not statistically significant and affect s other independent variables in the models, which tend to lower the adjusted R square in each model. The Price Effect s of the UGB Variable T he above results of reg ression analysis showed that the UGB variable is not statistically significant in any of the four density model s which indicated that the U GB variable had no direct price effect s on median housing prices among 35 urbanized
70 areas. H owever, as showed above, the UGB variable would slightly affect price effects of other independent variable s in the models, such as housing density job density, mean commute time median house hold income, population density, etc., which might cause indirectly price effects on me dian housing price. A s showed in Table 4 7, all the correlation coefficients were larger than 0.05, which indicated that there were no correlation between the UGB variable and any of the independent variables in the models. Therefore, the implementation of the UGB had no direct or indirect price effects on the median housing prices in year 2000. Table 4 7. Correlation c oefficients of the UGB dummy v ariable and o ther i ndependent v a riables in the regression m odel Independent Variables Sig. (two tail test, at 0.01 level) Correlation coefficient HD, Housing Density (per square miles) JD, Job Density MCT, Mean Commute time (per person) MHI, Median Household Income (annually) PD, Population Density (per square miles) VD, Vehicle Density (per square miles) .234 .207 .219 .213 .064 .316 .775 .050 .264 .194 .056 .326 Source: U.S. Bureau of Census, 2000 Census edit by author.
71 CHAPTER 5 DISCUSSION This part draws the summaries of both the statistical analysis which are correlation analysis between land p rices and housing prices, and median housing prices regression model. In the summary, the results of the analysis will be discussed and explained in detail. I n addition t he limitations of the analysis and the recommendations for improvements of each metho d will be presented in the following paragraphs. Summary of the Correlation Analysi s T he prices trends and correlation analysis were aimed to test the first hypothesis due to the supply of land constrained by the UGB, Portland had higher land prices and h ousing prices than other metropolitan areas since th e adoption of the UGB in 1979, and the land prices are highly correlated with the housing prices T he results demonstrated that the land prices, percentage of land share and housing prices in Portland rem ained at a high level among the selected metropolitan areas during the whole study period, except for the San Diego and Phoenix metropolitan areas P ortland land prices, land shares and housing prices began to grow much faster than other metropolitan areas since 1990, and dropped less significant ly than other areas during the economic downturn that started from 2007 T he main reason of the faster growth of the housing prices in Portland was result from the rapid growth in income and employment, which was al so supported by Downs (2002) research. M oreover, the correlation coefficient of land prices and housing prices of Portland had the highest significant value among all the selected metropolitan areas which indicated that the land prices were highly correl ated with housing prices. However, the correlation
72 coefficient of San Diego was almost the same to Portland, 0.992 in San Diego and 0.997 in Portland. In the early 1990s, the land prices, land share and housing prices in Portland started to grow faster tha n most of the selective metropolitan area s approximately 10 years after the establishment of the UGB. However, ther e is little evidence to establish any relationship between the repaid growth in housing prices and the implementation of the UGB. Inconsist ent Prices T rends of San Diego and Phoenix I n 1984, land and housing prices in San Diego were higher than those in Portland, approximately 40% (higher) A fter that, San Diego s land and housing prices kept growing steadily and experienced a significant in crease between 1997 and 2006, approximately 400%. T he increase was mainly resulted from an economic upturn rapid growth in population employment and business. D uring the same time period, Portland s land and housing prices also grew steadily, but less dra matically. During the economic downturn the land and housing prices in San Diego also dropped severely which closed the gap with Portland. However, land and housing prices in San Diego were still about 200% and 150% of Portland in 2009. A s to the land sh are, Portland experienced a more signific ant growth than San Diego between 1984 and 2006. In 2009, the land share in San Diego was still about 130% of Portland s I n addition, the correlation coefficients of these two metropolitan areas had an almost the s ame correlation value Though Phoenix has high prices trends similar to Portland, there were some differences worth mentioning in the Phoenix s prices and land share trends. Land and
73 housing prices were st able between 1984 and 2004, while land and housing prices in Portland grew steadily. Between 2004 and 2006, Phoenix land and housing prices experienced a suddenly increase, after that, the land and housing prices dropped to a significantly l evel in 2009. The land and housing prices in P ortland also experi enced an increase and dec r ease during the same time period, but less significantly and remained at a high level in 2009. I n addition the land share in Phoenix continued dropping from 1984 to 2004, and then after experienc ing a sudden increase from 2004, the land share dropped dramatically to a low level (23%) among all the metropolitan areas in the economic downturn. H owever, the land share in Portland was growing steadily during almost the whole time period and dropped slightly to 45.8% during the econo mic downturn In sum, without the establishment of the UGB, San Diego had a higher land prices, land shares and housing prices than Portland, which is strong ly contradicted the hypothesis. O ne explanation for these results is that San Diego and Phoenix ar e the most populous areas among the eleven selective metrop olitan areas, and they have a much stronger demand in the housing market. There was also no evidence showed that the prices trends and the correlation between land and housing p rices were resulted from the implementation of the UGB. Without further information or analysis, it is difficult to conclude that the UGB affected the land prices, land shares and housing prices trends in Portland. Limitations of the Correlation S tudy I n th is cross sectional prices trends analysis, t he samples size total number of metropolitan areas, is not large enough to provide strong evidence that the metropolitan
74 area with UGB has a higher land prices, land shares and housing prices which should be expanded to include m ore metropolitan areas, with or without the control of the UGB. More important ly, unlike the UGB in Portland established with in the boundary of the metropolitan area, m ost of the UGBs area established within the boundaries of other municipalities There fore due to other metropolitan area contain many non highly built up areas, it is important to note that the metropolitan study area intend to enhance the gap between the metropolitan area with or without UGBs, which lower s the credibility and accuracy of the analysis results T he results did prove that there were h igher prices trends and higher correlation between land prices and housing prices in the Portland metropolitan area than other metropolitan area however, it provided little evidence that all t hese results were resulted from the implementation of the UGB. It is completely possible that the higher prices trends in Portland had nothing to do with the implementation of the UGB, which could be the result of increased household income, increased empl oyment opportunities, economics upturn, market inflation, etc. Unless other factors that contribute to the housing prices were included in this analysis, it is difficult to state that the higher prices trends and correlation coefficient of Portland were ca used by the implementation of the UGB. I n addition as mentioned in C hapter 3, the land prices are calculated by housing prices subtracting the construction cost, which leads to increasing the correlation coefficient between land and housing prices. Summa ry of the Regression Analysis T he results of the correlation coefficient analysis demonstrated that only six independent variables were statistically significant to the median housing prices (Table 4 8), and the UGB dummy had no direct price effects on med ian housing prices. All
75 these independent variables had positive large price effects on the dependent variable whereas the mean commute time variable had a negative medium effect on the median housing prices Furthermore vehicle density, housing density, job density and population density variables are highly correlated, which caused a very high VIF (much greater than 10) value in the models. T herefore, there were four regression models with these four density variables. The best model consisted of median household income, job density, mean commute, which have the highest adjusted R square value, approximately 0.83. In this model, the R square changes contributed by these three variables are about 0.61, 0.19 and 0.05 respectively. After including the UGB i nto these four models it tends to lower the adjusted R square of each model. I n addition the UGB dummy variable also did not have any correlation with the six independent variables that were statistically significant to the median housing prices, which i ndicated that the UGB variable had no indirect price effect on median ho u sing prices. It is important to note that, at least in 2000, the UGBs did not affect the urban development pattern, such as housing density, population density, median commute time, e tc, which is contradictory to the hypotheses. Table 5 1 Summary of the c orr elation coefficients of i ndependent v a riables in the r egression m odel Independent Variables Correlation Status Tested Sign Effect A HS, Average Household Size HD, Housing Density JD, Job Density MCT, Mean Commute time MHI, Median Household Income PD, Population Density UGB, UGB Dummy UR, Unemployment Rate VD, Vehicle Density Uncorrelated Uncorrelated -Correlated + Large Correlated + Large Correlated Large Correlated + Large Correlated + Large Uncorrelated Uncorrelated -Uncorrelated Correlated Uncorrelated + -Large Source: Analyzed in SPSS, edit by author.
76 Uncorrelated I ndependent V ariables in the M odel T he three uncorrelated independent variables are average household size, unemployment rate and the UGB dummy. T he average household size indicates the average people within one household. A mong 35 urbanized areas, the mean, range and standard variance of average househo ld size are 2.57, 0.78 and 0.18 which ind icates that there are no huge differences between each urbanized area The mean, range and standard variance of the unemployment rate are 0.054, 0.041 and 0.01, which indicates that the difference of unemployment rate in each urbanized area is very little. M oreover, in 2000, the housing market in each urb anized area was very stable and the unemployment rate was low. H owever, for certain year s the unemployment rate i s related to the housing prices, f or example, in the Goodstein and Philips (2000) study men tioned in the Literature Review chapter. I f the study time period of the regression model extended to multiple years, the unemployment rate would have had a larger price effect on the median housing prices. A s to the UGB dummy variable, there are several factors that influence its price effect s on the median housing prices. First there are different forms of UGBs adopted in different states, which do not have exact the same effects as the one in Portland, for example, the IGA in Colorado or urban service area in Tallahassee. Moreover in the implementation process, some of UGBs in other states do not have such strong management enforcement and legislation at the state level, while only Oregon, Washington, and Tennessee requires the cities to establish UGBs Third the UGBs need a certain amount of time to take effects after the adoption. I n the case of Portland after approximately 10 years, the land and housing prices started to growth faster than before at 1990. F or other urbanized areas with UGBs, the im plementation history of the
77 UGBs did not have such a long time period as Portland s which had a 30 year implementation hist ory, which also intend ed to low er the prices effect of the UGB Lastly under the state law, the Portland is requires have a 20 year supply of land for future urban development within its growth boundary area to help mitigate the price pressure s of the UGB on the land and housing prices. The vacant land in the boundary should be sufficient for the future development until 202 0. In addit ion in the Portland s UGB history, there have been several times of expansions to accommodate more land into the boundary area. Results of the Regression M odel I n the final regression models, the most important independent variable is the median househol d income Even without other variables the median household income model is able to explain about 61% of the all the housing prices samples. T he results indicated that the higher median household income in one urbanized area is accomplished the higher med ian housing prices. Though four density variables are highly correlated in the regression analysis, they are equally important to measure the median housing prices in different urbanized areas, contributing approximately 0.18 R square changes to the models They all have positive large prices effects on the median housing prices. Similar to Jun s (2006) study, the housing density is highly related to the median housing prices. Moreover consistent with Downs (2002) study, the best model consisted of averag e household income and job density, which contribute d most to the median housing prices. O nly the median commute time had a negative medium price effect on the median housing prices, which measures the total commute time to and from work daily I n the cla ssic urban economic theory, prices adjust to ensure location al equilibrium which
78 indicates the housing prices in the market is equilib rium between the housing cost and the transportation cost ( O Sulliv an, 2009 p.28 ) Consistent with this economics theo ry, the finding in this analysis showed that longer median commute time spent the low er the housing price. However this variable just contribute s approximately 0.05 R square changes to regression model. Limitations of the R egression A nalysis T he existin g regression model focused on the social and economic factors of the housing market ; however other factors regarding the physical characteristic of hou ses also contribute to the median housing prices. For example, a s demonstrate d in the previous literatur e some independent variables should be included into the regression model, such as mean construction cost, construction index variable, land prices, or average house footage, etc. The sample size of 35 urbanized areas is not large enough for multiple regr ession analysis which requires larger sample size to provide more accurate and reliable results A s a r u le of thumb the required sample size for a larger effect should be 40 for 3 predictors and 80 for medium effect (Field, 2005). I f more predictors are included in the regression model, the bigger sample size is required to achieve a better predictive regression model. Since t he data in this regression analysis are based on a single year, in this kind of cross sectional study, certain independent variabl es had no correlation with the median ho using prices; however, as shown in the literature review, these independent variables were correlated to the housing prices in the time series study. It is one of the main reasons the time period of the regression an alysis should be expanded to multiple years. Moreover if the time period of the model expanded to multiple years the results the
79 price effects of different independent v ariables on the median housing prices will be more comprehensive and accurate.
80 CHAP TER 6 CONCLUSIONS T his research is aimed to explore the relationship between the UGB in Portland and the local housing affordability. The first chapter introduced the research problems, concept of the UGB and housing affordability, and the second chapter r eviewed all the previous studies related to this issue. Both qualitative and quantitative methods are adopted to describe and measure the price effect of the UGB on local housing prices. T here we re two hypotheses in this study: Hypothesis 1. D ue to the UGB constrain on the supply of land, the trends of land prices and land share in the Portland metropolitan area should be higher than other metropolitan areas without UGBs in United States. Since land prices have a significant impact on housing prices, the ho using prices in Portland should be also higher. I n addition the Spearman Correlation C oefficient should supports the argument that the land and housing prices in Portland are highly correlated. Hypothesis 2. Besides affecting the land prices, the UGBs al so tend to change the urban development patterns, such as higher density development, high redevelopment and infill development rate of the inner city, etc. T he regression model takes into account all of the factors that contribute to housing prices to mea sure the price effects of different predictors on the median housing prices among 35 urbanized areas. The result of the regression analysis should demonstrate the exact price effects of the UGB on the median housing prices. F or the first hypothesis, the results of prices trends and correlation coefficient demonstrated that the land prices, land shares and housing prices of Portland is higher than some of the metropolitan areas without UGBs, except for Phoenix and San Diego.
81 Moreover, San Diego and Portlan d have a very close correlation coefficient value. The above results did not indicate any evidence that the higher housing prices in Portland are resulted from the implementation of the UGB bet ween 1984 and 2009. A s the second hypothesis, the best median housing regression model consisted of median household income, job density and median commute time. F urthermore, the vehicle density, housing density and population density had a similar price effect on the median housing prices as job density. H owever, th e results of the regression model, and the correlation coefficients between the UGB and other independent variables indicated that UGB dummy variable had no direct or indirect price effects on Portland s median housing prices. I n conclusion, as show n in th e literature review, UGB did increase the density of new development, and redevelopment and infill development rate within the Portland metropolitan area throughout the entire implementation period of the UGB H owever, in 2000, the correlation coefficient between the UGB and other independent variables show that the UGB did not affect the urban development pattern, such as housing density, population density, job density, median commute time, etc. Though the housing prices in Portland did increase more than some of the other metropolitan areas without further statistical analysis, it is still difficult to prove that the increase in housing prices are resulted from the implementation of the UGB. T here is also no evidence in the analysis indicating that the U GB cause d prices pressures on the housing affordability in the year 2000
82 APPENDIX A APPENDIX DATA USED IN THE REGRESSI ON MODEL Table A 1 Data used in the median housing prices regression model Urbanized Area Median Housing Prices Average Househol d S ize Housing Density (per square miles) Mean Commute time (minutes) Median Househol d Income Job Density (per square miles) Populatio n Density (per square miles) UGB Unemplo yment Rate Vehicle Density (per square miles) Atlanta, GA 138700 2.7 692 31.1 52 512 1360 1783 0 0.052 900 Baltimore, MD 138700 2.5 1279 29.1 46931 2369 3041 0 0.055 1439 Boston, MA 219900 2.5 924 28.3 53908 1842 2323 0 0.043 1158 Boulder, Colorado 302300 2.2 1467 19.2 48518 2931 3416 1 0.062 1931 Bu ffalo, NY 87900 2.4 1189 20.6 37060 2092 2664 0 0.075 1178 Charlotte, NC 141200 2.5 727 26.1 50731 1346 1745 0 0.051 909 Chicago, IL 161400 2.7 1490 31.2 50747 2975 3914 0 0.064 1789 Cincinnati, OH 116700 2.5 948 23.3 4448 5 1709 2238 0 0.043 1080 Cleveland, OH 121000 2.4 1195 24.4 41920 2143 2761 0 0.055 1276 Dallas, TX 97900 2.7 1144 26.8 46993 2206 2946 0 0.049 1427 Denver, Colorado 173800 2.5 1622 28.4 50372 3068 3979 1 0.040 2074 Detr oit, MI 131500 2.6 1257 26.0 48541 2362 3094 0 0.061 1382 Houston, TX 89700 2.8 1113 28.4 44658 2181 2951 0 0.062 1324 Indianapoli s, IN 113800 2.5 949 23.5 45503 1675 2205 0 0.046 1103 Lexington, Kentucky 110200 2.3 1617 19 .4 39269 2915 3609 1 0.051 1899
83 Table A 1 Continued Urbanized Area Median Housing Prices Average Househol d Size Housing Density (per square miles) Mean Commute time (minutes) Median Househol d Income Job Density (per square miles) Popul atio n Density (per square miles) UGB Unemplo yment Rate Vehicle Density (per square miles) Los Angeles, C A 220500 3.0 2395 28.7 44735 5299 7068 0 0.077 2934 Memphis, Tennessee 90000 2.6 981 23.5 39309 1810 2431 1 0.067 1091 Miami, Flori da 127900 2.6 1897 28.1 40214 3486 4407 1 0.066 1868 Milwaukee, WI 128300 2.5 1121 21.8 43727 2058 2688 0 0.057 1280 Twins Cities, Minnesota 140400 2.5 1071 22.6 53242 2047 2671 1 0.036 1439 New York, NY 209700 2.7 2042 35. 1 49648 4143 5309 0 0.070 2301 Philadelphi a, PA 120100 2.6 1157 28.2 47265 2218 2816 0 0.063 1300 Phoenix, AZ 126400 2.7 1498 26.0 44623 2760 3638 0 0.048 1669 Pittsburgh, PA 86400 2.3 925 24.9 38142 1654 2057 0 0.058 932 Portland, Oregon 167800 2.5 1396 23.6 46360 2591 3340 1 0.057 1676 Salt Lake city, UT 156200 3.0 1329 22.5 48130 2817 3847 0 0.045 1880 San Antonio, TX 73700 2.8 1221 23.6 38237 2432 3257 0 0.057 1426 San Diego, CA 224800 2.7 1265 25.0 46613 2630 3419 0 0.057 1586 San Jose, CA 443000 2.9 2039 25.9 74133 4603 5914 1 0.037 2920
84 Table A 1. Continued Urbanized Area Median Housing Prices Average Househol d Size Housing Density (per square miles) Mean Commute time (minutes) Median Househol d Income Job Density (per square miles) Populatio n Density (per square miles) UGB Unemplo yment Rate Vehicle Density (per square miles) St.Louis, MO 101000 2.5 1064 24.7 44221 1924 2506 0 0.058 1191 Tampa, FL 93900 2.3 1228 25.0 37864 2064 2571 0 0.049 1176 Virginia Beach, V irginia 108900 2.6 1035 23.3 42176 2024 2647 1 0.051 1290 Washington DC 189300 2.6 1351 32.2 63558 2643 3401 0 0.045 1770 Source: U.S. Bureau of Census, 2000 Census edit by author.
85 LIST OF REFERENCES Anderson, H.A. (1999). Use and Implementation of Urban Growth Boundaries. Center for Regional and Neighborhood Action. Accessed November 29, 2010. Alkadi, A (1996). Hedonic analysis of housing prices near the Port land urban growth boundary, 1978 1990. Ph.D. dissertation, Portland State University, United States -Oregon. Retrieved October 25, 2010, from Dissertations & Theses: Full Text. (Publication No. AAT 9628853). Avin U & B a y e r M. (2003). Right sizing Urb an Growth Boundaries. Planning [serial online]. February 2003; 69(2):22. Available from: Academic Search Premier, Ipswich, MA. Accessed November 29, 2010. Center for Neighborhood Technology. (2010). H+ T Affordability Index Metro Report. Official websites : http://htaindex.cnt.org/ Assessed Dec ember 2010. Center for Transit Oriented Development and Center for Neighborhood Technology. (2006). The Affordability Index: A New Tool for Measuring the True Affordability o f a Housing Choice. Washington, D.C.: Brookings Institution. Downs, A (2002). Housing Policy Debate, Vol.13, Issue 1. Fannie Mae Foundation 2002, p.7 31. Downs, A (2004). Growth management and affordable housing: Do they conflict? Washington, D.C: Brookings Institution Press. Easley, G. (1992). Staying inside the lines: Urban growth boundaries Chicago, IL: American Planning Association. Field, A. P. (2005). Discovering statistics using S PSS: (and sex, drugs and rock 'n' roll). London: Sage Publications, p.111. Fischel, W A. (2002). Housing Policy Debate, Vol.13, Issue 1. Fannie Mae Foundation Goo dstein, E & Phillips, J (2000). Contemporary Economic Policy, Vol.18, No.3. July 2000, P.334 344(ISSN 1074 3529). Greene, P. R. & Pick B. J. (2006). Exploring the urban community a GIS approach. Pearson Prentice Hall. Grout, C. A. e t al (2009). Land use regulations and property values in portland, oregon: A regression discontinuity design approach. Regional Science and Urban
86 Economics, In Press, Corrected Proof doi:DOI: 10.1016/j.regsciurbeco.2010.09.002. Jun, M (2006). The Effects of Portland's Urban Growth Boundary on Housing Prices. Journal of the American Planning Association [serial online]. Spring2006 2006;72(2):239 243. Available from: Academic Sea rch Premier, Ipswich, MA. Accessed November 29, 2010. Knaap, G. (1985). The price effects of urban growth boundaries in metropolitan Portland, Oregon. Land Economics 61 (1): 26 35. Knaap, G & Hopkins, L. (2001). The inventory approach to urban growth bou ndaries. Journal of the American Planning Association 67(3): 314 326. Knaap, G & Nelson, A. (1992). The regulated landscape: Lessons on state land use planning from Oregon. Cambridge, MA: Lincoln Land Institution. Ling, D & Wayne, A (2005). Real Estate Principals A Value Approach New York, NY. McGraw Hill/Irwin 2005. Lincoln Institute of Land Use Policy. (2010). Land and Property Values in the U.S. Assessed January, 2011. http://www. lincolninst.edu/subcenters/land values/ Metro. (2002). Last updated January 28, 2006. Assessed December 30th, 2006. http://www.metroregion.o rg/article.cfm?articleID=277 Metro. (2002). Urban growth boundary : frequently asked questions. Assessed January, 2011. h ttp://topaz.metro region.org Morrill, R & David C. H (1991). Myths & Facts About G rowth Management. Seattle : University of Washington. National Association of Home Builders. (2010). Opportunity Ind ex: Complete History by Metropolitan Area (1991 Assessed March 2006. http://www.nahb.org/page.aspx/category/sectionID=135 Nelson, A C et al (2002). T he Link Between Growth Management and Housing Affordability: The Academic Evidence. Washington, DC: Brookings Institution Center on Urban and Metropolitan Policy. Nelson, A.C., & Moore T (1995). Assessing Urban Growth Management: The Case of n dary. Land Use Policy 10(1995):293 302.
87 Oates, D (2005). ndary Oregon State University Press. Oates, D. (2010). Urban Growth Boundary The Oregon Encyclopedia. Accessed January 16th, 2011. http://www.oregonencyclopedia.org/entry/ view/urban_growth_boundary/ Oregon Government. (2010). Accessed January 2011. http://www.oregon.gov/LCD/history.shtml Oregon Government, Oregon Departme nt of La nd Conservation and Development (2010). Accessed January 2011. http://www.oregon.gov/LCD/goals.shtml#Statewide_Planning_Goals ivan, A. (2009). Urban economics. McGraw Hill Education, p.28. Quigley, J & Larry R (2005). Cityscape, 8(1): 69 137. Roddewig, R. J., & Inghram, C. A. (198 7). Transferable development rights programs. Chicago, IL (1313 E. 60th St., Chicago 60637: American Planning Association. Squires, G. D. (2002). Urban sprawl: Causes, consequences, & policy responses. Washington, D.C: Urban Institute Press. Staley, S. et al. (1999). growth Boundaries, Smart Growth, and Housing Affordability. Policy Study 263. Wa shington, DC: Reason Public Policy Institute. Warnken, C G. (2003). The price effects of the urban service area boundary in Tallahas see, Florida. Ph.D. dissertation, The Florida State University, United States -Florida. Retrieved October 25, 2010, from Dissertations & Theses: Full Text.(Publication No. AAT 3109535). U.S. B ureau of Census (2010) Census 2000 Assessed January, 201 1. http://www.census.gov/main/www/cen2000.html U.S. Department of Housing and Urban Development. (2010). Assessed January, 2011. http://www.huduser.org/portal/publications/pdf/brd/15Dolbeare.pdf U.S. Department of Housing and Urban Development. (2010). Trends in Housing Costs: 1985 2005 and the 30 Percent of Income S Assessed January, 2011. http://www.huduser.org/portal/publications/pdf/Trends_hsg_costs_85 2005.pdf
88 1000 friends of Oregon. (1999). Oregon Housing Cost St udy: Myths & Facts About Oregon s Urban Growth Boundary. Assessed Dec ember 2010. http://www.onethousan dfriendsoforegon.org/resources/myths.html pdf
89 BIOGRAPHICAL SKETCH Xing Ma (Max) was born in 1986 in Xi an, China At three years old his family moved to G uangzhou, where he was raised He received his bachelor degree as Urban Planning at the South Ch ina University of Technology which was five year undergraduate program focusing on architecture and urban design. In 2009, he entered t he Urban and Regional Planning m aster p rogr am at the University of Florida located in Florida, USA. During the 2 year program, he concentrated on economi c development planning and got a minor in the business school. He graduated from the College of Design, Construction and Planning in May 2011.