THE EFFECT OF HOUSING MARKET SEGMENTATION ON COMMUTING : A GEOGRAPHICAL ANALYSIS

국문 초록 (Abstract)

본 연구의 목적은 도시공간구조가 통근에 어떤 영향을 미치는지 고찰하는 것이다. 도시구조의 포괄적 범위를 상기할 때, 도시공간구조를 그 구성요소로 나누어 고찰할 필요가 있다. 이 연구에서는 주택시장분화를 공간구조의 구성요소로 간주하여, 그것이 각 도시별 통근길이와 어떤 상관관계가 있는지 살펴본다. 미국 85개 대도시에서의 주택시장분화와 통근길이의 관계에 대한 경험적 분석에 따르면, 주택시장분화가 가속화될수록 통근길이가 연장되는 것으로 나타났다. 통근길이에 영향을 미치는 요인들을 밝혀내는 것은 교통수요를 예측하는데 뿐만 아니라 통근길이가 늘어남에 따라 가중되는 교통문제 (가령 교통체증, 안전 및 공해)를 해결하는 데 있어서도 유용한 단서가 된다. 이 연구는 도시에 따라 각기 다른 형태로 나타나는 주택시장분화현상이 그 도시의 통근길이 연장에 과연 기여를 하는지에 대한 가설을 검증하고자 한다. 도시별 주택시장분화현상을 측정하기 위해서 본연구는 주택시장분화지표를 정의한다. 지표는 주택하위시장형성에 기여하는 변수를 기반으로 하여 각 도시별 주택하위시장을 도출하는 것으로 시작되며, 이에는 퍼지클러스터링 알고리즘이 적용되었다. 퍼지클러스터링이 적합한 방법론인지 검증하기 위해 기존의 블린 연산에 기초한 클러스터링으로부터 도출된 결과를 비교하였고, F 테스트는 이 점을 확증해 주었다. 통근길이가 모든 통근모드에 해당되는 경우, 통근길이와 주택분화현상의 관계가 통계적으로 두드러지지 않은 반면, 차량에 국한되어 정의된 경우 주택분화현상이 통계적으로 현격한 영향을 미치는 것으로 드러났다. 이점은 주택분화현상 (또는 주거분리현상)이 지속가능한 교통에 부정적인 영향을 미칠 수 있음을 시사하는 것으로써 이에 대한 심층적인 연구가 필요하다고 하겠다.

다국어 초록 (Multilingual Abstract)

This study examines the role of urban spatial structure on commuting. Among the many forces that constitute the dimensions of urban structure, it is focused on the fragmentation of a metropolitan area into distinct housing submarkets. The study refers to this phenomenon as “housing market segmentation”. Empirical analysis of the relationship between housing market segmentation and commute length in 85 metropolitan areas shows that housing market segmentation (as a component of urban spatial structure) lengthens commute distances.
The operationalization of the IHMS rests on the definition of housing submarkets, which are derived by means of a fuzzy c-means algorithm. The performance of fuzzy clustering is evaluated in comparison with that of k-means methods. An F-test confirms that fuzzy clustering outperforms hard clustering. Fuzzy clustering is shown to be of great use in the classification of housing markets based on census data.
The index of housing market segmentation is measured for 85 metropolitan areas. Housing market tends to be more segmented in large metropolitan areas while housing market segmentation is not correlated with geographic region. Complexity of industrial structure and racialized process of residential development seems to contribute to a high degree of housing market segmentation.
Regression analysis of metropolitan-wide commute length shows that commute time is affected by labor market scale (e.g., the number of workers), labor market structure (e.g., composition of industry, specialization or diversification), labor market performance (e.g., unemployment), socioeconomic variables (e.g., income, age, ethnic composition), modal split, and urban form features (e.g., housing density). When commute length is defined in terms of vehicle miles, housing market segmentation is found to be an important contributing factor. The study indicates that disparity between residential neighborhoods (i.e., housing market segmentation as a component of urban structure) can impose negative impact on sustainable transportation.

목차 (Table of Contents)

• ACKNOWLEDGEMENTS = ⅳ
• Table of Contents = ⅴ
• ABSTRACT = xvi
• Ⅰ. INTRODUCTION = 1
• Ⅰ.1 BACKGROUND = 1
• Ⅰ.2 Organization of the Study = 3
• Ⅰ.3 Purpose and Research Questions = 6
• Ⅰ.3.1 Research Purposes = 6
• Ⅰ.3.2 Research Questions = 9
• Ⅰ.4 Significance of the Study = 11
• Ⅱ. LITERATURE REVIEW = 18
• Ⅱ.1 Sustainability = 19
• Ⅱ.1.1 Sustainable Development = 19
• Ⅱ.1.2 Defining Urban Sustainability = 20
• Ⅱ.1.3 Indicators of Urban Sustainability = 23
• Ⅱ.1.4 Sustainable Transportation = 26
• Ⅱ.2 The Link between Land use and Transportation = 28
• Ⅱ.2.1 Theory of the Land Use/Transportation Link = 30
• Ⅱ.2.2 Empirical Studies of the Land use-Transportation Link = 33
• Ⅱ.3 Housing Markets = 48
• Ⅱ.3.1 Characteristics of Housing Markets = 48
• Ⅱ.3.2 Hedonic Price Model of Housing Market = 50
• Ⅱ.3.3 Identifying Spatial Housing Submarkets = 53
• Ⅲ. METHODOLOGY = 56
• Ⅲ.1 Study Set = 57
• Ⅲ.2 Variables and Data Sources = 59
• Ⅲ.2.1 Census Tract Level Data: Determinants of House Price = 60
• Ⅲ.2.2 Metropolitan Level Data: Determinants of Commute Length = 65
• Ⅲ.3 Methodology = 71
• Ⅲ.3.1 Selecting Predictors of Census Tract-Level House Prices = 74
• Ⅲ.3.2 Delineating Spatial Housing Submarkets = 75
• Ⅲ.3.3 The Index of Housing Market Segmentation (IHMS) = 81
• Ⅲ.3.4 Effect of HMS on commute length = 82
• Ⅳ. IDENTIFYING SPATIAL HOUSING SUBMARKETS WITH FUZZY CLUSTERING = 85
• Ⅳ.1 Introduction = 85
• Ⅳ.2 Fuzzy c-means Algorithms for Housing Market Classification = 86
• Ⅳ.3 Parameters in Fuzzy c-means Algorithms = 89
• Ⅳ.3.1 Optimal number of clusters c = 89
• Ⅳ.3.2 Optimal fuzziness exponent m = 97
• Ⅳ.4 Case Study: housing submarkets in the Buffalo-Niagara Falls metropolitan area = 104
• Ⅳ.5 Evaluating the performance of fuzzy clustering = 112
• Ⅳ.5.1 Results of housing market classification based on K-means vs. Fuzzy c-means = 113
• Ⅳ.5.2 Comparing the performance between fuzzy clustering and hard clustering = 116
• Ⅳ.6 Conclusions = 118
• Ⅴ. HOUSING MARKET SEGMENTATION IN THE US METROPOLITAN AREAS = 121
• Ⅴ.1 Introduction = 121
• Ⅴ.2 Index of Housing Market Segmentation = 122
• Ⅴ.3 Inedx of Housing Market Segmentation in the US metropolitan areas: findings from the 2000 Census = 128
• Ⅴ.4 A Tale of Two Cities: IHMS illustrated in attribute space = 136
• Ⅴ.5 A Tale of Two Cities: The IHMS illustrated in geographic space = 141
• Ⅴ.6 Conclusions = 150
• Ⅵ. THE EFFECT OF HOUSING MARKET SEGMENTATION ON COMMUTING = 151
• Ⅵ.1 Introduction = 151
• Ⅵ.2 Factors Influencing Commute Time = 152
• Ⅵ.2.1 Correlation Analysis between Commute Time and IHMS = 152
• Ⅵ.2.2 Global Regression Model = 153
• Ⅵ.2.3 Local Regression Model = 165
• Ⅵ.3 Factors Influencing vehicle Miles of Commute (VMC) = 167
• Ⅵ.3.1 Correlation Analysis between VMC and IHMS = 167
• Ⅵ.3.2 Global Regression Analysis = 168
• Ⅵ.3.3 Local Regression Analysis = 171
• Ⅵ.4 Housing Market Segmentation and Commuting = 174
• Ⅵ.5 Conclusions = 178
• Ⅶ. CONCLUSION = 180
• Ⅶ.1 Summary = 180
• Ⅶ.2 Limitations to the Research = 183
• Ⅶ.3 Future Research = 184
• Ⅶ.4 Discussion = 187
• APPENDIX = 189
• REFERENCES = 258