1. Report No. SWUTC/06/167552-1

2. Government Accession No.

3. Recipient's Catalog No.

5. Report Date April 2006

4. Title and Subtitle Household Location Choices: The Case of Homebuyers and Apartment Dwellers in Austin, Texas 6. Performing Organization Code

7. Author(s) Michelle Bina and Kara M. Kockelman

8. Performing Organization Report No. Report 167552-1 10. Work Unit No. (TRAIS)

9. Performing Organization Name and Address Center for Transportation Research University of Texas at Austin 3208 Red River, Suite 200 Austin, Texas 78705-2650

11. Contract or Grant No. 10727

13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address Southwest Region University Transportation Center Texas Transportation Institute Texas A&M University System College Station, Texas 77843-3135

14. Sponsoring Agency Code

15. Supplementary Notes Supported by general revenues from the State of Texas. 16. Abstract This paper explores the issue of residential location choice for apartment dwellers and recent homebuyers in the Austin, Texas area. An understanding of residential location choice is fundamental to behavioral models of land use and, ultimately, travel demand. Surveys of over 200 apartment dwellers and 900 recent homebuyers offer valuable data on movers and their reasons for moving, as well as their priorities in home type and location selection and tradeoffs from such decisions. Survey results show that apartment dwellers have different reasons for moving than homebuyers. The top reasons for moving for homebuyers are wanting to own a home and wanting a newer/bigger/better home; many apartment dwellers also reported wanting a newer/bigger/better home (the most frequently chosen response), but they were also much more likely to be moving for an easier commute, new job or job transfer, and planning to attend or graduate from college. Both apartment dwellers and recent homebuyers reported their level of importance for various housing and location attributes; several of which described various types of access, such as commute time, proximity to shopping availability of bus services, and access to bus services. These features accounted for approximately 40% of the average reported importance for all housing and location attributes for apartment dwellers, but only 25% for homebuyers, suggesting that apartment dwellers place a higher priority on access than homebuyers. Lifecycle variables, such as living situation and marital status, are important indicators in binary choice experiments and the importance of various measures of accessibility, especially in the case of apartment dwellers. Families in apartments are more likely to choose a location with plenty of parking over a downtown location with limited parking and homebuyers with children favor larger lots over proximity to shopping facilities. Although lifecycle variables are statistically significant in models of stated preferences for homebuyers, current home and location features are more practically and statistically significant, on average. In many instances, homebuyers are more likely to favor improvements that reflect their location choice. For example, households with homes far from the Austin Central Business District (CBD) tend to favor home enhancements over improvements in accessibility despite the fact these locations have lower levels of accessibility, in general. Overall, binary logit and ordered probit model results show that women and non-Caucasians (apartment dwellers and homebuyers) tend to be more concerned with all types of access – commute time, proximity to shopping, availability of bus services, and access to major freeways. Predictive models of monthly rent, home value, and location offer important insights, while controlling for many key factors. For instance, centrality is valued by residents, so monthly rents fall rapidly within 3 miles of the CBD but taper at further distances, and home values fall linearly by $8,000 for each additional mile away from the CBD, ceteris paribus. As predicted rents and home values decrease with less accessibility (i.e., being further from the CBD), these values increase with apartment/home size (i.e., interior square footage), recognizing the tradeoffs that homes make between various attributes for a given cost constraint. Additional models of rent and home value isolate structural features of the home from location features. For apartments, location attributes provided higher predictive power (adjusted R2=0.624) when compared to the physical attributes (adjusted R2=0.339). The opposite results occur when looking at homes: the predictive power of the structural aspects (adjusted R2=0.666) is slightly higher than the location information (adjusted R2=0.601). This further supports that apartment dwellers are more concerned with location features than home features, especially in comparison to homebuyers. Cross-tabulations and a multinomial logit model explore home type choice, for various demographic groups. Understandably, households with many children are more likely to purchase large homes, and households with high incomes are more likely to purchase new homes and larger lots. Attached housing is more prominent for low-income and single-person households. A multinomial logit location choice model for recent homebuyers shows home affordability and centrality are important to households, as well as the size of homes in the neighborhood. Model segmentation reveals that households with children are more sensitive home affordability and less attracted to central locations than households without children. These results and many others are explained by various model specifications. 17. Key Words Location Choice, Land Use, Travel Demand

18. Distribution Statement No restrictions. This document is available to the public through NTIS: National Technical Information Service 5285 Port Royal Road Springfield, Virginia 22161

19. Security Classif.(of this report) Unclassified

20. Security Classif.(of this page) Unclassified

21. No. of Pages 117

22. Price

HOUSEHOLD LOCATION CHOICES: THE CASE OF HOMEBUYERS AND APARTMENT DWELLERS IN AUSTIN, TEXAS

By Michelle Bina

Kara M. Kockelman

Research Report SWUTC/06/167552-1

Southwest Region University Transportation Center

Center for Transportation Research

University of Texas at Austin

Austin, Texas 78712

April 2006

iv

DISCLAIMER The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the Department of Transportation, University Transportation Centers Program, in the interest of information exchange. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

ACKNOWLEDGEMENTS The authors recognize that support for this research was provided by a grant from the U.S. Department of Transportation, University Transportation Centers Program to the Southwest Region University Transportation Center which is funded 50% with general revenue funds from the State of Texas.

I would like to express my gratitude to Dr. Kara Kockelman for her guidance and support, and Dr. Randy B. Machemehl for his suggestions as the second reader. Without the help of Ahmed Qatan, Shadi Hakimi, Nick Lownes, Valdemar Warburg, and Shashank Gadda, data of apartment dwellers would not have been obtained. Undergraduates Robin Lynch and Daniel Villalobos also aided in the collection process, as well as Stacey Bricka for her help during the survey instrument design process. I would like to thank the following persons for their invaluable help in mailing surveys to recent homebuyers as well as entering the results into a database: Annette Perrone, Laura Narat, Kristin Donnelly, Caroline Soo, Robin Lynch, and Prenon Islam. Ahmed Qatan and David Suescun made the surveys available on the Internet. Also, I am especially thankful to Valdemar Warburg and David Suescun for their help collecting data as well as modeling and interpreting data results.

v

ABSTRACT

This paper explores the issue of residential location choice for apartment dwellers and recent homebuyers in the Austin, Texas area. An understanding of residential location choice is fundamental to behavioral models of land use and, ultimately, travel demand. Surveys of over 200 apartment dwellers and 900 recent homebuyers offer valuable data on movers and their reasons for moving, as well as their priorities in home type and location selection and tradeoffs from such decisions.

Survey results show that apartment dwellers have different reasons for moving than homebuyers. The top reasons for moving for homebuyers are wanting to own a home and wanting a newer/bigger/better home; many apartment dwellers also reported wanting a newer/bigger/better home (the most frequently chosen response), but they were also much more likely to be moving for an easier commute, new job or job transfer, and planning to attend or graduate from college.

Both apartment dwellers and recent homebuyers reported their level of importance for various housing and location attributes; several of which described various types of access, such as commute time, proximity to shopping availability of bus services, and access to bus services. These features accounted for approximately 40% of the average reported importance for all housing and location attributes for apartment dwellers, but only 25% for homebuyers, suggesting that apartment dwellers place a higher priority on access than homebuyers.

Lifecycle variables, such as living situation and marital status, are important indicators in binary choice experiments and the importance of various measures of accessibility, especially in the case of apartment dwellers. Families in apartments are more likely to choose a location with plenty of parking over a downtown location with limited parking and homebuyers with children favor larger lots over proximity to shopping facilities. Although lifecycle variables are statistically significant in models of stated preferences for homebuyers, current home and location features are more practically and statistically significant, on average. In many instances, homebuyers are more likely to favor improvements that reflect their location choice. For example, households with homes far from the Austin Central Business District (CBD) tend to favor home enhancements over improvements in accessibility despite the fact these locations have lower levels of accessibility, in general. Overall, binary logit and ordered probit model results show that women and non-Caucasians (apartment dwellers and homebuyers) tend to be more concerned with all types of access – commute time, proximity to shopping, availability of bus services, and access to major freeways.

Predictive models of monthly rent, home value, and location offer important insights, while controlling for many key factors. For instance, centrality is valued by residents, so monthly rents fall rapidly within 3 miles of the CBD but taper at further distances, and home values fall linearly by $8,000 for each additional mile away from the CBD, ceteris paribus. As predicted rents and home values decrease with less accessibility (i.e., being further from the CBD), these values increase with apartment/home size (i.e., interior square footage), recognizing the tradeoffs that homes make between various attributes for a given cost constraint.

Additional models of rent and home value isolate structural features of the home from location features. For apartments, location attributes provided higher predictive power (adjusted R2=0.624) when compared to the physical attributes (adjusted R2=0.339). The opposite results occur when looking at homes: the predictive power of the structural aspects (adjusted R2=0.666)

vi

is slightly higher than the location information (adjusted R2=0.601). This further supports that apartment dwellers are more concerned with location features than home features, especially in comparison to homebuyers.

Cross-tabulations and a multinomial logit model explore home type choice, for various demographic groups. Understandably, households with many children are more likely to purchase large homes, and households with high incomes are more likely to purchase new homes and larger lots. Attached housing is more prominent for low-income and single-person households.

A multinomial logit location choice model for recent homebuyers shows home affordability and centrality are important to households, as well as the size of homes in the neighborhood. Model segmentation reveals that households with children are more sensitive home affordability and less attracted to central locations than households without children. These results and many others are explained by various model specifications.

vii

EXECUTIVE SUMMARY

The research team was given the opportunity to investigate the reasoning used by apartment dwellers and homebuyers when selecting home locations. The past 40 years have seen significant shifts in urban land use patterns and travel behaviors. Rising income and vehicle ownership have made it possible for many families to rent apartments and purchase homes in suburban areas and travel longer distances, resulting in minimal transit use and decentralization of metropolitan areas. Sprawling regions and automobile dependence are an issue of great interest to policy-makers, planners, transportation engineers, environmentalists, and many others. To address these concerns, integrated models of land use and transportation are being developed to better predict future travel patterns. Analysis of residential location choice can inform such models.

In order to generate informative models of residential location choice behavior, surveys of recent movers were undertaken in the Austin, Texas region. The surveys asked households about their primary reasons for moving, the importance of various housing and location attributes, their current home and location, current travel patterns, and basic demographic information. The results of these models revealed more than simply who is more attracted to what and what they are willing to pay. For example, the binary logit results suggest that if a developer and/or community wish to attract high-income and single-person households, it might best focus on building apartments close to downtown, while improving access to shopping. In order to attract families with children, however, they should build large apartment complexes or homes in the suburbs with access to recreation facilities.

While the home choice decision is very complex, these new data sets and their many associated behavioral models offer various insights. The reasons for a move and priorities in home selection, the hedonic models of home and rent value, the paired comparisons of potential home enhancements, the importance scores of various attributes, and the logit models of home type and location choice should allow researchers, planners, and developers to more accurately characterize the tradeoffs households make in their home/location choices. When coupled with models of life cycle changes, land development and population growth, as well as travel demand, vehicle ownership and other behaviors, such models facilitate a more integrated look at our communities and their futures.

viii

ix

TABLE OF CONTENTS

Chapter 1. Introduction..........................................................................................................................1

1.1 Study Objectives..................................................................................................................1

Chapter 2. Literature Review.................................................................................................................3

2.1 Residential Location Choice................................................................................................3

2.2 Reasons for Moving.............................................................................................................4

2.3 Priorities of Households ......................................................................................................5

2.4 Tradeoffs Made by Households...........................................................................................5

Chapter 3. Methodology and Data Sets .................................................................................................7

3.1 Survey of Realtors ...............................................................................................................7

3.2 Survey of Apartment Dwellers ............................................................................................9

3.3 Survey of Recent Homebuyers ..........................................................................................10

Chapter 4. Apartment Dwellers ...........................................................................................................15

4.1 Reasons for Moving...........................................................................................................15

4.2 Priorities During Housing Search......................................................................................16

4.3 Model Results ....................................................................................................................17

4.3.1 Monthly Rent Linear Regression..............................................................................18

4.3.2 Binary Logit Results for Scenarios...........................................................................20

4.3.3 Ordered Probit Analysis of the Importance of Access .............................................24

Chapter 5. Recent Homebuyers ...........................................................................................................27

5.1 Reasons for Moving...........................................................................................................27

5.2 Priorities During Housing Search......................................................................................28

5.3 Model Results ....................................................................................................................30

5.3.1 Home Value Linear Regression................................................................................31

5.3.2 Binary Logit Results for Scenarios...........................................................................34

5.3.3 Ordered Probit Analysis of the Importance of Access .............................................37

5.3.4 Home Type Multinomial Logit Model .....................................................................39

5.3.5 Multinomial Logit Models of Location Choice........................................................45

Chapter 6. Comparison of Apartment Dwellers and Recent Homebuyers..........................................49

6.1 Commute Time..................................................................................................................52

x

6.2 Distance to Shopping.........................................................................................................52

6.3 Access to Bus Services ......................................................................................................52

6.4 Access to Freeways ...........................................................................................................53

Chapter 7. Conclusions........................................................................................................................55

7.1 Limitations and Future Work ............................................................................................56

Appendix A..........................................................................................................................................59

Appendix B..........................................................................................................................................79

Appendix C..........................................................................................................................................83

Appendix D..........................................................................................................................................95

References .........................................................................................................................................101

xi

LIST OF FIGURES

Figure 3.1 Recent homebuyer respondent locations............................................................................11

Figure 4.1 Average scores of access attributes as a percentage of total score for all attributes for various demographic groups of apartment dwellers............................................................................17 Figure 4.2 Predicted rent values versus distance to CBD for varying number of bedrooms ..............20 Figure 5.1 Average scores of access attributes as a percentage of total score for all attributes for various demographic groups of homebuyers.......................................................................................29 Figure 5.2 Change in predicted home value versus number of bedrooms and bathrooms..................32 Figure 5.3 Predicted home value versus distance to CBD for various levels of interior square footage, dwelling age, and school quality ...........................................................................................33 Figure 5.4 Home type (structure and age) versus household income..................................................41 Figure 5.5 Home type (interior and lot size) versus household income..............................................41 Figure 5.6 Probability of home type choice for varying household income .......................................45 Figure 6.1 Importance of access-related attributes for apartment dwellers and homebuyers .............51

xii

LIST OF TABLES Table 3.1 Realtor survey sample characteristics ...................................................................................8 Table 3.2. Realtors’ mean scores of importance of access ....................................................................8 Table 3.3 Apartment dweller sample characteristics...........................................................................10 Table 3.4 Recent homebuyer sample characteristics...........................................................................12 Table 4.1 Primary reason for moving: (Apartment dwellers versus CPS results)...............................15 Table 4.2 Mean score of importance of housing and location attributes for apartment dwellers .......16 Table 4.3 Weighted least squares regression of monthly rent .............................................................19 Table 4.4 Final specification for scenario questions (1-3) for apartment dwellers .............................21 Table 4.5 Final specification for scenario questions (4-6) for apartment dwellers .............................22 Table 4.6 Ordered probit results for importance of commute time and distance/travel time to shopping to apartment dwellers ...........................................................................................................24 Table 4.7 Ordered probit results for importance of access to bus services and major freeways to apartment dwellers...............................................................................................................................25 Table 5.1 Primary reasons for moving for recent homebuyers ...........................................................27 Table 5.2 Mean score of importance of housing and location attributes for recent homebuyers........28 Table 5.3 Ranking of importance of housing and location attributes for segmented homebuyers .....30 Table 5.4 Ordinary least squares regression of home value ................................................................31 Table 5.5 Final specifications for scenario questions for recent homebuyers.....................................35 Table 5.6 Final specifications for scenario questions for recent homebuyers, cont. ...........................36 Table 5.7 Ordered probit results for importance of commute time and distance/travel time to shopping for recent homebuyers..........................................................................................................37 Table 5.8 Ordered probit results for importance of access to bus services and access to major freeways for recent homebuyers..........................................................................................................38 Table 5.9 Home type versus household composition and income ......................................................40

xiii

Table 5.10 Home type choice model results (using MNL) .................................................................43 Table 5.11 Home type choice probabilities for given household characteristics ................................44 Table 5.12 Pooled residential location choice model results (using MNL) ........................................46 Table 6.1 Comparison of sample characteristics for apartment dwellers and recent homebuyers......50 Table B. 1 Recent homeowner sample characteristics (including weighting schemes), compared to PUMS data.......................................................................................................................................82 Table C. 1 Ordered probit results for importance of price and perception of crime rates to apartment dwellers...............................................................................................................................84 Table C. 2 Ordered probit results for importance of attractive neighborhood appearance and noise to apartment dwellers .................................................................................................................85 Table C. 3 Ordered probit results for importance of social composition and neighborhood amenities to apartment dwellers ..........................................................................................................86 Table C. 4 Ordered probit results for importance of views and closeness to friends or relatives to apartment dwellers...............................................................................................................................87 Table C. 5 Ordered probit results for importance of commute to school, quality of local public schools, and distance to local public schools to apartment dwellers...................................................88 Table C. 6 Ordered probit results for importance of investment potential, perception of crime rates, and number of bedrooms to recent homebuyers ........................................................................89 Table C. 7 Ordered probit results for importance of noise, lot size, and social composition of the neighborhood to recent homebuyers....................................................................................................90 Table C. 8 Ordered probit results for importance of views, neighborhood amenities, and closeness to friends or relatives to recent homebuyers........................................................................91 Table C. 9 Ordered probit results for importance of quality of local publice schools and distance to local public schools to recent homebuyers ......................................................................................92 Table C. 10 Ordered probit results for importance of physical disability accommodations and distance to medical facilities to recent homebuyers ............................................................................93 Table D. 1 MNL location choice model results, segmented for households with and without children ................................................................................................................................................96 Table D. 2 MNL location choice model results, segmented for single-person households and those married with children .................................................................................................................97

xiv

Table D. 3 MNL location choice model results, segmented on the basis of household income .........98 Table D. 4 MNL location choice model results, segmented for two-worker, one-worker, and zero-worker households.......................................................................................................................99

1

HOUSEHOLD LOCATION CHOICES: THE CASE OF HOMEBUYERS AND APARTMENT DWELLERS IN AUSTIN, TEXAS

CHAPTER 1. INTRODUCTION The past 40 years have seen significant shifts in urban land use patterns and travel behaviors. Rising income and vehicle ownership have made it possible for many families to rent apartments and purchase homes in suburban areas and travel longer distances, resulting in minimal transit use and decentralization of metropolitan areas. Sprawling regions and automobile dependence are an issue of great interest to policy-makers, planners, transportation engineers, environmentalists, and many others. Shaw (1994) notes a few examples of these interests: policy-makers and planners worry about reliance on oil; transportation engineers worry about deficits incurred because of increasing transportation infrastructure; environmentalists worry about pollutants and air quality due to automobile emissions; and many Americans worry about stress, loss of time, and costs due to increasing traffic congestion. To address these concerns, integrated models of land use and transportation are being developed to better predict future travel patterns. Analysis of residential location choice can inform such models.

1.1 Study Objectives In order to generate informative models of residential location choice behavior, surveys of recent movers were undertaken in the Austin, Texas region. The surveys asked households about their primary reasons for moving, the importance of various housing and location attributes, their current home and location, current travel patterns, and basic demographic information.

This work analyzes two key populations: apartment dwellers and recent homebuyers. According to the Census of Population Survey (CPS), renters comprised 62.7% of total movers in the U.S. (over 40 million) during 2002 and 2003, representing a majority of movers and a key demographic in regards to residential mobility (Schachter 2004). And they are a demographic group that has not previously been studied in much detail. However, according to the 2000 Census of Population, only 33.8% of all U.S. households rent their dwelling unit; therefore, a separate survey effort examined choices of homebuyers. Surveys of recent movers can provide better data since such respondents can more accurately recall their motivations for moving and their characteristics at the time of the move. The data from these surveys are used to evaluate characteristics of their move, such as their reasons for moving and their priorities and preferences in choosing a home and location.

A household’s choice to move and where to move is a complex and costly decision. “When people buy or rent housing, they are obtaining a bundle of goods that includes interior living space; housing services such as schools and parks; and externalities like neighborhood image, noise, and smog” (Parsons Brinkerhoff Quade & Douglas, Inc. 1999, p.96). For virtually every household, a residence cannot be found in which all of these housing and location attributes are optimized; and size, cost, accessibility, or other features may be compromised. This paper focuses on accessibility as a key component of the housing choice and evaluates households’ relative preferences and tradeoffs made while focusing on transport accessibility.

The results from this study are expected to be of value to practitioners and researchers of various disciplines. For instance, since this paper focuses on access, proponents of transit-oriented

2

development (i.e., planners and developers) may have interest in this research. Knowing the types of persons and households who value high accessibility may allow developers to better market their communities to maximize their profit returns. In turn, successful transit-oriented communities may also increase transit ridership, benefiting the local transit agency. This is especially relevant to Austin developers since a commuter rail is currently in design and is planned to begin operation in 2008. Additionally, apartment complex managements and residential housing developers may use results of hedonic pricing models to predict and/or set rent values and home purchase prices. Information obtained from these survey efforts and model results will be useful to a variety of persons and can be applied for different purposes.

This thesis is organized as follows: Chapter 2 provides a review residential location choice studies and related literature. Chapter 3 provides an overview of the survey methodologies employed and describes the data sets used for analysis. Using the data obtained, various regression models were developed in order to quantify the preferences and priorities of the two demographics groups and their locations. The results of these models are discussed in Chapters 4 and 5, for apartment dwellers and recent homebuyers, respectively. Chapter 6 compares the two data sets and discusses similarities between model results. Conclusions, limitations, and potential extensions of this research are summarized Chapter 7.

3

CHAPTER 2. LITERATURE REVIEW The relationship between access and housing choices is not well understood but is gaining more attention. This chapter presents a brief summary of the evolution of residential location choice modeling and the resulting behavior framework that has been widely adopted by many researchers. Also, this research focuses on three unique (and important) features of the housing choice: reasons for moving, priorities/preferences of the household, and tradeoffs made by the household. Existing related literature is also described in the chapter, in terms of the studies’ results and weaknesses.

2.1 Residential Location Choice Residential location choice models have been used to model the choices made by households in regard to their housing situation. The standard framework for residential location choice models hypothesizes a sequence of decisions that begins with a decision to move and ends with a chosen home and location (e.g., Grigg 1982, Weisbrod et al. 1980, Guiliano 1989, Ben-Akiva and Bowman 1998). Studies have examined various aspects of residential location choice, such as residential mobility (Speare et al. 1975), market search (Clark 1982), dwelling type (Boehm 1982, Tu and Goldfinch 1996, Cho 1997), and location choice (Gabriel and Rosenthal 1989, Wadell 1996). Economic theory supports a utility-maximization approach to models of residential location choice via tradeoffs between access to goods and services, home quality and size, and housing prices (Giuliano 1989).

Rosen (1974) first presented a theoretical work on hedonic prices that has motivated the specification of models to relate housing market prices to housing characteristics (see, e.g., Huh and Kwak 1997, Kockelman 1997, Orford 2000). However, it has been argued that hedonic price functions offer limited information regarding consumer behavior (Ellickson 1981). A need to reflect taste variations among households has motivated the application of logit models in the analysis of housing markets (Cho 1997). Studies most relevant to this work have incorporated access variables into the analysis.

Since accessibility is a major theme in residential location theories, transportation has been a focus of many models (e.g., Alonso 1964, Anas 1982, Weisbrod et al. 1980). Such studies provide a good basis for understanding the connections between transportation and land use; however, empirical data suggests that many models are incomplete (Giuliano 1989). Two key weaknesses in many older studies include the assumption of a single-worker household and a monocentric city (in which all jobs occupy the central business district) (Giuliano 1989). Feminization of the workforce and decentralization of metropolitan areas have invalidated these assumptions (Waddell 1996). Efforts to model dual-worker households include studies by Waddell (1996), Sermons and Koppelman (2001), and Freedman and Kern (1997), among others.

Also, many studies are based on the data of static households (see, e.g., Bhat and Guo 2004). Surveys of recent movers can provide better data since such respondents can more accurately recall their motivations for moving and their characteristics at the time of the move. Since household location models seek to identify the determinants of the move decision and the chosen dwelling unit (including location), there is strong interest in identifying priorities during the search process and in quantifying all tradeoffs.

4

2.2 Reasons for Moving The decision to move is based on some type of dissatisfaction of one’s current living situation or new opportunities that may lie elsewhere. In response to such interests, the U.S. Census Bureau recently published a couple Current Population Reports (Schachter 2001 and Schachter 2004) that contain only cross-tabulations and raw distributions of the 2000 Current Population Survey (CPS) and the 2003 Annual Social and Economic Supplement to the CPS evaluating households’ reasons to move. The results of these studies include a high number of reason-to-move responses in “other” categories (suggesting that there are some unexpected yet important reasons for moving), followed by those wanting a newer/better home and wanting to own/not rent. Other results include a comparison between long- and short-distance moves: long-distance moves are usually made because of work-related reasons and short-distance moves are mainly related to housing, where long-distance is considered intracounty and short-distance is intercounty. It was also concluded that highly educated movers are more likely to move for job-related reasons, and even more so for a long-distance moves. This makes sense intuitively as highly educated households generally make higher incomes, placing a higher premium on the importance of their job. This is also affirmed in the study by showing that high-income individuals are more likely to move for job-related reasons than low-income individuals, suggesting a link between employment and residential mobility. Understandably, economists generally believe that a job-related move can be considered an economic investment and those movers are more willing to bear the moving costs. These study conclusions demonstrate that there are links between household characteristics and reasons for moving. However, the studies do not quantify correlations between multiple demographic factors and responses nor do they identify the type of housing structure or tenure choice, which is a major limitation.

Over 30 years ago, Murie (1974) explored the reasons for household movement and related them to a few housing and demographic variables, but his data is out-dated and only summarizes basic variable statistics segmented on the basis of household composition.1 Murie (1974, p.115) utilized life-cycle variables and concluded that each lifestyle group has similar characteristics, values, and goals that make their choices similar, even across social class groups; and “although the survey evidence presented in this study indicates that ‘family cycle’ and ‘life-style’ factors are important, it has been evident that these are not sufficient explanations of residential mobility.” Unfortunately, the basis of his analysis (as well as the different types of tenure and housing in England) provide few transferable results to this survey effort.

Filion et al. (1999) surveyed and analyzed residents of Kitchener-Waterloo, Canada. The study addressed households’ reasons for moving, where a nicer dwelling, a nicer neighborhood, and a more convenient location are the top reasons. An important hypothesis explored by the researchers stated that households rarely move for one reason, and so they allowed multiple reasons to be reported. They used exploratory factor analysis to determine interrelated moves and found that the strongest occurrence involved a nicer dwelling, a nice neighborhood, and new children. However, they did not relate these reasons to home qualities or demographic characteristics, and the data is based on static households. To the authors’ knowledge, no recent study exists that isolates apartment dwellers and recent movers while exploring their reasons for moving.

5

2.3 Priorities of Households Another important aspect of residential location choice involves the housing search process, particularly the importance of factors that determine household priorities. Filion et al. (1999) asked residents about their level of importance for access to various destinations (i.e., stores, expressways, and bus routes) and location characteristics (i.e., attractive area, safety, and noise). Living in an area that was safe, quiet, and private were the most important (believed to be very important by 72%, 45%, and 44% of the weighted sample, respectively); and all place variables were found to be more important than access variables (in which none of the variables were found to be very important by more than 31% of the weighted sample). The authors concluded (through focus group analysis) that many services and locations are easily accessible throughout the studied city, which would explain the lack of preference towards locations with high accessibility. These results may not be expected for urban areas with greater congestion and other access limitations.

The 2004 American Community Survey (ACS), conducted for Smart Growth America and the National Association of Realtors, examined Americans’ preferences in a residential location. Being within a 45-minute commute to work was the highest priority (79%), followed by easy access to the highway (75%), which reveal that commute and accessibility are among the highest priorities for U.S. residents. However, no “other” category was included in the question, which may just reveal the relative importance of the attributes listed, not the true priorities of a household. Another major limitation of the study is the small sample size (1,130 adults) that was used to make conclusions about the entire U.S. population. And again, neither study (Belden et al. 2004 and Filion et al. 1999) explored explanatory variables that affect these relationships (such as income and household size).

2.4 Tradeoffs Made by Households Moves are costly, with sellers generally paying 6% (of their home’s value) in realtor fees and 1 to 2% in other transaction costs, and all parties typically paying several hundred to thousands of dollars for transport of furnishings. Because of this costly decision, many households find it necessary to tradeoff a variety of location, size, quality, and cost factors when selecting a home and location. Weisbrod et al. (1980) used logit models to determine the contribution of various attributes to a household’s chosen residential location, using variables relating to housing price, building type, housing size, previous location, income, demographic characteristics, taxes, crime, school quality, and trip accessibility. Household composition was found to be more significant than any other tradeoff analyzed, especially in determining housing type. Teacher/pupil ratio was not significant in location choice. The obvious weakness of the study is the date of the data. Today, there are higher costs associated with longer commutes (in terms of monetary value and time), increased vehicle ownership as more women have joined the workforce, and a higher importance of school quality.

The 2004 ACS used stated preference questions to explore a few tradeoffs, but provided only raw statistics. One question compared commute time to lot size; and respondents were evenly split on the choice between a longer commute (more than 45 minutes) and a larger lot size (at least an acre), where women, African-Americans, and homebuyers were the most likely to chose a shorter commute.

Using the behavioral framework introduced, this paper integrates discrete choice and regression models with residential location and home attributes. The analysis relates choices and

6

preferences to key demographic variables, including life-cycle variables. Data used to calibrate these models was obtained from surveys of apartment dwellers and recent homebuyers, along with an initial survey of realtors. The methodologies employed in this research effort are discussed in the next chapter.

ENDNOTES 1 Murie surveyed two types of households: “new” and “continuing,” where a “new” household involves a new housewife or a housewife who had previously lived elsewhere. Analysis of data obtained was presented for new and continuing households, separately.

7

CHAPTER 3. METHODOLOGY AND DATA SETS In order to examine residential location choices for residents of the Austin, Texas area, three key groups of people were surveyed: realtors, apartment dwellers, and recent homebuyers. Austin was chosen in order to minimize survey costs. Also, Austin is considered the most congested medium-sized U.S. city, in terms of travel delay and congestion costs (Schrank and Lomax 2005); therefore, residents of this area are likely to be concerned about accessibility to various destinations. The survey instruments were individually developed for each of the three respondent types and can be found in Appendix A.2 This chapter presents a brief discussion of the data acquisition processes for each of the surveys (detailed methodologies can be found in Appendix B) and sample characteristics for each data set obtained in this research effort.

3.1 Survey of Realtors Since realtors understand what is most important to movers, realtors were interviewed and surveyed. Three realtors were interviewed at length in order to provide an in-depth look at the moving process and help formalize the survey instruments. A survey was distributed to 229 realtors in the Austin area via electronic mail during December of 2004, which yielded 22 completed surveys. Each initial email was followed up by an average of 3 reminder emails; however, the response rate remained low. Table 3.1 shows some sample characteristics for the 22 realtors surveyed and their clientele.

Since not all movers employ the help of realtors, comparing the statistics of recent residents of the Austin area to the typical clientele of these realtors will determine any biases present in the realtors’ perspective. Using the 5% Public-use Microdata Sample (PUMS) from the 2000 Census, only 38.3% of recent homebuyers in Travis County (the county containing the City of Austin) had at least one child living in their household, which shows a clear bias in the realtor data where the average realtors’ clientele contained a majority (70.6%) of households with children. It is expected that households with children and households without children have very different characteristics and preferences in the realm of housing and residential location choices. Since this is such a small sample and the realtors may present a biased perspective on movers, the results of the interviews and the survey were used only to provide some background information about the moving process, suggest some hypotheses, and serve as an aid in the development of the household survey.

8

Table 3.1 Realtor survey sample characteristics

Years working in the real estate business 10.30 9.23 Estimated number of clients per year 35.91 18.70 Estimated average yearly sales $5.18 M $4.10 M % of clients living in Austin metro area before move 74.14% 0.16 % of clients living in other parts of Texas before move 10.86% 0.10 % of clients living outside Texas before move 15.00% 0.13 % of clients living in central Austin (within 4 miles of downtown) after move

25.09% 0.29

% of clients living in Travis Co. after move 50.45% 0.28 % of clients living in Hays Co. after move 3.84% 0.06 % of clients living in Williamson Co. after move 26.18% 0.27 % of homes less than $150k 19.48% 0.18 % of homes between $150k and $250k 40.86% 0.25 % of homes between $250k and $500k 31.76% 0.27 % of homes greater than $500k 7.19% 0.09 % of lots less than 1/4 acre 71.10% 0.31 % of lots between 1/4 and 1/2 acre 21.52% 0.23 % of lots greater than 1/2 acre 7.62% 0.10 % of clients without children 29.43% 0.24 % of clients that have young children 33.52% 0.25 % of clients that have children of various ages 35.62% 0.19 Although there do seem to be biases in the realtor data set, some statistics of the survey responses were used as a starting point for various hypotheses. For instance, since this research effort focuses on transportation and accessibility, the survey asked realtors how important access (to jobs, shopping, good schools, and other amenities) is to various demographic groups, on a scale of 1 to 5, where 5 is “very concerned.” Table 3.2 reveals their average responses.

Table 3.2. Realtors’ mean scores of importance of access3 Household characteristic Mean score

City dwellers 3.50 Low-income clients 3.24 Clients who rent 3.12 Not highly educated clients 3.07 Out-of-state movers 3.06 Older (45+ years) clients 2.78 Married clients 2.69 Younger (<45 years) clients 2.59 Clients who have children 2.44 Clients who own 2.40 Highly educated (Masters+) clients 2.38 Single clients 2.33 Local movers 2.27 Suburban dwellers 2.25 Clients who do not have children 2.17 High-income clients 2.00

Variable Mean Std Deviation

9

The results suggest that central-city dwellers (those closer to the central business district [CBD]) are the most concerned with accessibility, which may be because city dwellers are much more likely to be driving in congested conditions and/or because they have chosen their locations in order to improve access. The clients that are reported to be least concerned with accessibility are high-income clients, those without children, and, as expected, suburban dwellers. Clients that do not have children may face fewer time constraints, and so accessibility may not be as important. These results seem to be fairly intuitive and provide a basis for the expected results of the household survey. However, the low sample size of the survey and biases of realtor clientele require that any conclusions be made with caution and examined via household survey results.

3.2 Survey of Apartment Dwellers A location choice survey for apartment dwellers was designed and administered by graduate students at the University of Texas at Austin during the spring semester of 2005 as part of a collective effort between researchers and students in a graduate course. The survey was designed as a self-completion survey, for door-to-door and Internet distribution. Several revisions and a pilot test were executed in order to develop a comprehensive survey, which can be found in Appendix A.

Survey data was collected during late February and early March of 2005. Approximately 1600 apartments were visited (knocked on); and out of these, 28 % answered the door. The sample loss of around 1150 consists of the apartments where nobody was home, nobody lived, or the residents did not want to answer the door. In order to minimize respondent bias, surveys were distributed and collected on Saturdays and Sundays, where individuals of various characteristics (including employment status) have similar probabilities of being at home.

Approximately 450 residents answered their doors, which constitute the net sample size. Of these, 260 accepted to fill out a survey, which corresponds to a 58% response rate. However, due to missing responses, only 232 of received surveys were used for analysis. Several imputation techniques were utilized to mitigate the number of missing values for rent, square footage, age, and household income. The final data set was weighted to account for age, gender, and household income based on the 2000 Census PUMS data for renters who live in apartment buildings within the Austin area4. Supplementary data was obtained to describe each observation’s location. The region’s Metropolitan Planning Organization (MPO), Capital Area Metropolitan Planning Organization (CAMPO), provided information on land area, population, households, and employment for all traffic serial zones (TSZs); and Census tract information on housing characteristics was matched to the TSZs. Detailed information about the sampling procedure and data imputation techniques is given in Appendix B.

Before analysis can occur, sample characteristics of the data should be examined for possible biases. Table 3.3 provides some weighted summary statistics, compared to 2000 Census PUMS data for apartment dwellers in the City of Austin. The results of the comparison show that the sample is fairly representative of the area.

10

Table 3.3 Apartment dweller sample characteristics

Variable Mean sample

results (N=240)

Mean Census results (N=6013)

Household size 2.08 2.08 Number of children in household 0.49 0.37 Presence of children (at least one child) 0.26 0.20 Married 0.28 0.24 Married & have at least one child 0.16 0.11 Age 33.13 33.72 Male 0.57 0.57 Number of vehicles available in household 1.38 1.23 Number of vehicles per household member 0.77 0.74 No vehicles in household 0.06 0.14 Household income ($/year) $37,930 $35,996 Caucasian (base) 0.48 0.54 Hispanic/Latino 0.26 0.15 African-American 0.10 0.11 Asian 0.10 0.07 Other ethnicity 0.05 0.13 Less than high school 0.05 0.15 High school 0.37 0.16 Associate's or technical degree 0.14 0.33 Bachelor's degree 0.31 0.25 Master's degree or higher 0.14 0.11 Number of bedrooms 1.63 1.32 Monthly rent (dollars) $680 $620

Many practitioners and researchers are interested in why a household chooses to rent versus own a dwelling unit. The survey asked the respondents to indicate their main reason for choosing to live in an apartment. 44% indicated affordability, 18% needed a short-term residence, 15% appreciated the size, relative to their needs, 13% wanted low maintenance, and 9.5% chose “other” as a response. Virtually every “other” reason relates to the location of the apartment complex, including being close to work or school. Based on all responses, one might hypothesize that lower income and smaller households tend to live in apartments. 2000 Census PUMS data for the Austin area confirms this hypothesis, indicating that the average household income of those living in apartments is $36,000 – which is less than half the household income of non-apartment dwellers ($74,200). Moreover, the average household size for those residing in apartments is 2.08 persons, whereas an average of 2.63 persons live in other dwelling types.

3.3 Survey of Recent Homebuyers A survey of recent homebuyers was conducted very shortly after the survey of apartment dwellers. The survey was designed as a self-completion survey and was intended for mail-out mail-back and Internet distribution. Several revisions and a pilot test were executed in order to develop a comprehensive survey which can be found in Appendix A.

USA Data Inc. assembled a sample frame of all home buyers (identified via deed purchases and transfers) in the Austin three-county region (Hays, Travis, and Williamson Counties) between March 2004 and February 2005 (a one-year period). A random sample of just over half of these

11

identified households was purchased, providing 4,451 names and addresses. Surveys were mailed to all 4,451 households in April 2005, and reminders were sent four to six weeks later. (The survey was available on-line as well, for those who had not retained the original survey form, but the reminders generated only 100 or so responses, so no further reminders were sent.) The sample yielded 965 complete surveys, or a 21.7% response rate. Since some households were not appropriate for the sample frame5, the actual response rate, from the pool of qualified survey recipients, is believed to be somewhat higher (25% or higher). Those who did respond but did not qualify as “recent movers” were not included in the final data set used for analysis. Figure 3.1 represents the locations of all survey respondents.

TRAVIS

HAYS

BURNET

BASTROP

WILLIAMSON

BLANCO

CALDWELL

Figure 3.1 Recent homebuyer respondent locations

As can be seen in the figure, few observations are located outside of Travis County. The sampling frame was reduced to include only Travis County residents because of few Hays and Williamson County residents in the sample created by USA Data Inc. (A detailed discussion of sampling issues can be found in Appendix B.) The final data set used for analysis included 943 recent homebuyers in Travis County, and was not weighted because of the limited size (1,069

12

observations) of appropriate PUMS data (a more detailed discussion can be found in Appendix B). Essentially, the PUMS sample of recent homebuyers is really no larger than the sample obtained here. So it was not felt that the weights would provide much, if any, reductions in bias. Table 3.4 provides some summary statistics, compared to 2000 Census PUMS data.

Table 3.4 Recent homebuyer sample characteristics

Variable

Mean sample results

(N=943)

Mean Census results

(N=1069) Household size 2.27 2.68 Number of children in household 0.51 0.71 Presence of children (at least one child) 0.31 0.38 Married 0.55 0.60 Married & have at least one child 0.25 0.32 Age (head of household) 39.58 40.20 Male (head of household) 0.56 0.69 Number of vehicles available in household 1.95 1.86 Number of vehicles per household member 0.94 0.83 No vehicles in household 0.01 0.02 Household income ($/year) $93,256 $94,056 White (base) 0.84 0.80 Hispanic 0.08 0.15 Black 0.02 0.07 Asian 0.03 0.05 Other ethnicity 0.03 0.01 Two or more workers 0.42 0.68 One worker 0.50 0.02 No workers 0.08 0.30 Total number of workers in household 1.43 1.91 Full-time student 0.04 0.09 Retired 0.05 0.06 Single-family home 0.90 0.80 Number of bedrooms 3.12 2.99 Age of dwelling 25.18 15.22 Brand new home (indicator) 0.03 0.32

Even though weights were not applied to the data set obtained, it does seem to be fairly representative, according to Census PUMS data, especially in terms of demographic characteristics. Any bias that might be incurred probably relates to their homes. Our survey results reflect households living in older homes and fewer occurrences of brand new homes. However, the skew in age may be attributed to a few outlying data points of very old homes (three recently purchased homes were built 98, 101, and 147 years ago). This effect might be mitigated through the use of quadratic specifications for the dwelling age variable.

The data obtained from the three surveys provides the basis for analysis. As discussed, no model specifications were calibrated due to the small sample and bias of the realtors’ perspective, providing only knowledge of the moving process and some initial hypotheses to be tested. For example, it is expected that those living close to the CBD, low-income households, and renters are more likely to be concerned about access in general. The survey of renters, weighted to

13

represent the apartment dwelling population of Austin, show that they tend to have lower incomes and smaller household sizes, as opposed to homebuyers in the area. Data on recent homebuyers, although not weighted, does seem to be representative of homebuyers (who moved within the past year). The following two chapters provide analysis of the data sets obtained for apartment dwellers and recent homeowner.

ENDNOTES 2 The surveys have been reformatted for this publication. The questions are exactly the same as presented to the respondent, but have slightly different formatting. 3 Characteristics are rated on a scale of 1-5, where 1 is “not at all concerned” and 5 is “very concerned.” 4 This sampling area is the 787xx Zip Code Tabulation Area (ZCTA), which has a population of 777,789. 5 Several survey recipients called to report that they had recently refinanced, rather than purchased their home. Others indicated that they were not actually living at the property but had purchased it as an investment.

14

15

CHAPTER 4. APARTMENT DWELLERS This chapter discusses the analysis of the apartment dweller data set, including final specifications and regression model results. Summary statistics are provided for survey results, in comparison to Census results for all recent movers, for the reasons for moving. After the choice has been made, movers have apartment and location preferences and survey results are revealed in this chapter (and ordered probit models highlight the characteristics related to these priorities). Also, tradeoffs made by the households are explored through the use of hedonic price modeling for rent values and binary logit specifications for stated preference questions. These survey results and model specifications help to identify the determinants of an apartment dweller’s residential location choice.

4.1 Reasons for Moving Simply knowing why people move can be very helpful in developing residential choice models. The survey asked apartment dwellers to indicate their primary reason for moving to their current apartment. Table 4.1 compares these corrected (population weighted) sample results to those of the 2003 U.S. Current Population Survey (CPS), which sampled over 40 million recently relocated households across the U.S.

Table 4.1 Primary reason for moving: (Apartment dwellers versus CPS results) Survey results for

apartment dwellers only (N=232)

CPS results for all U.S. movers

(N=40M) Reason for moving

Frequency Percent Percent Wanted newer/better apartment 45 18.7% 19.8% Easier commute 41 16.8% 3.2% Other 35 14.5% 27.5% New job/job transfer 30 12.5% 8.8% Wanted/needed less expensive housing 24 9.8% 6.5% Planned to attend or graduate from college 15 6.2% 2.5% Marriage or divorce 15 6.0% 6.7% Wanted to rent 14 5.9% N/A Birth/adoption 7 3.0% N/A Change of climate 6 2.4% 0.4% Retiring 1 0.5% 0.3% Health reasons 1 0.5% 1.4%

The comparisons suggest that Austin’s apartment dwellers differ from recent U.S. movers in several ways. The greatest difference (13.6%) between the two exists because of the high percentage of surveyed apartment dwellers that move for an easier commute. This may be attributed to Austin’s heavy congestion and limited freeway corridors (Schrank and Lomax 2005).

Another difference relates to those moving for a new job or job transfer: 3.7% more apartment dwellers state this as their primary reason for moving. A new job or job transfer often signals a long-distance move; and the CPS results support this by indicating that the most common single reason for an inter-county or international move is a new job or job transfer (Schachter 2004). Long-distance movers may be more inclined to choose an apartment, in order to become more familiar with the area before buying a home.

16

A third difference is the higher percentage of apartment dwellers seeking less expensive housing, which is intuitive since apartments are generally a less expensive housing option. Also, a higher percentage of apartment dwellers move to begin college studies, which also is intuitive, since many college students rent apartments, and Austin has a relatively high population of college students (13.7% vs. 8.32% in the US).

Finally, fewer apartment dwellers chose an “other” reason for moving, which may suggest that the list of reasons here is more complete. Common “other” reasons include wanting to establish one’s own household (e.g., move out of parent’s home), separating from previous household member (such as a roommate), needing to be closer to medical facilities, wanting to be closer to family, and wanting to move to different city.

4.2 Priorities During Housing Search Once a household has chosen to move, the process of searching for a new apartment/location begins. During this search, a household has priorities for key features. So respondents were asked to report the level of importance (on a scale of 1 to 4) of several housing and location attributes. Table 4.2 lists these attributes, along with the average level of importance for the corrected (population weighted) sample.

Table 4.2 Mean score of importance of housing and location attributes for apartment dwellers

Housing/location attributes Average score of

level of importance6

% “Very Important” or “Somewhat Important”

Price 3.68 96.84% Commute time to work 3.27 86.65% Perception of crime rate 3.22 84.63% Attractive neighborhood appearance 3.17 83.88% Commute time to school 3.13 77.57% Access to major freeways 3.08 80.02% Noise 2.99 72.40% Distance/travel time to shopping 2.64 55.95% Neighborhood amenities/recreational facilities 2.62 59.12% Social composition of the neighborhood 2.62 56.15% Access to public transportation 2.61 56.08% Views 2.49 47.57% Closeness to friends or relatives 2.42 45.41% Quality of local public schools 2.25 40.44% Distance to local public schools 2.21 38.91% Predictably, price is the most important attribute to apartment dwellers. Of course, price is a key criterion in virtually any choice, for most people. Moreover, lower income households tend to rent (as discussed earlier), and therefore may be more concerned with this attribute. Surprisingly, the quality of and distance to local public schools attributes are scored least important. Perhaps this is because apartment dwelling households tend to contain fewer children. The 2000 Census suggests that 20.4% of U.S. households living in an apartment have children, as compared to 30.3% among non-apartment households.

Commute time is second most important attribute, on average, which, as explained earlier, may be credited in part to Austin’s traffic congestion. Of course, commute time is just one of several access attributes that were included in the survey. If one sums the average scores of all access

17

attributes considered (commute time to work, commute time to school, access to major freeways, distance/travel time to shopping, access to public transportation, and distance to local public schools), they amount to about 40% of the total average score. Thus, one might argue that price, apartment size and quality, and other key non-travel-related attributes play a somewhat greater role in one’s apartment search − but access is still quite key. In fact, in the open-ended responses of primary reasons for moving, several persons indicated that distance to medical facilities is also important, especially elderly persons, suggesting that access may be even more important and account for more than 40% of the total if included in the list of attributes. Figure 4.1 shows the average weight of scores for access attributes for various demographic groups.

38.970 38.837 38.646 38.503 38.306 38.203 37.969 37.367 37.05935.620

26.447

0

5

10

15

20

25

30

35

40

45

Ages 1

8-35

Non-C

auca

sian

Married

with

child

ren

High in

come (

>$50k

)

Ages 3

6-55

Low in

come (

<$25

k)

All obs

ervati

ons

Medium

inco

me (>=$

25k&

<$50k

)

Cauca

sian

Single-

perso

n hou

seho

ld

Ages 5

6+

Perc

ent

Figure 4.1 Average scores of access attributes as a percentage of total score for all attributes for various

demographic groups of apartment dwellers

Over all 240 respondent observations, the percentage weight that access carries ranges from 13% to 71%, with very few (less than 5% of the observations) exhibiting an access valuation over 50%. Although there is little difference between average percentages for various groups of interest, there is a dramatic drop in the importance of access, relative to other attributes, for older persons (ages 56 or higher). To further explore this issue, ordered probit models were used for analysis of the underlying factors that influence the individual attributes scores. With a general idea of apartment dwellers’ reasons for moving and their priorities, model regressions (presented in the next section) provide more insight by utilizing multivariate analyses of monthly rent values and the preferences of households.

4.3 Model Results Weighted least squares (WLS), binary logit, and ordered probit regression models were used to analyze responses to various survey questions posed. The results are as follows:

18

4.3.1 Monthly Rent Linear Regression A linear regression model (Table 4.3), weighted by population correction factors, was used to examine how rent relates to various housing and location variables. This is valuable information in determining where to build and zone for multifamily apartment complexes (as well as how to price such units). The results also provide a sense of the tradeoffs that households make in terms of cost (rent) and benefits (e.g., interior square footage). All variables that were expected to have an impact were included in the initial specifications, and the final model specification emerged from a systematic procedure of eliminating statistically insignificant variables, combined with intuitive considerations.

Location attributes were obtained by matching geocoded apartment complex addresses to several data sources including CAMPO’s zonal file (which provided information on zones’ areas, population, households, and employment), the 2000 Census of Population (which provided information on median home values, housing units, housing units’ median number of rooms, and average commute times for employed people), and the work by Kalmanje and Kockelman (2004). It is this last work that provided accessibility indices, calibrated from logsums emerging from travel demand models of home-based-work and home-based non-work trips. The logsum used here is the expected maximum utility derived across all modes, departure times, and destinations available to a trip maker. Kalmanje et al. (2004) calibrated multinomial logit models for Austin area trips using the 1996 Austin Travel Surveys. They considered four modes and five times of day, along with the region’s 1074 TSZ as destinations. Calibrations were determined using the following equation:

)ln(,

,,,,,,∑ ++=tm

ptmjiCostpcjiTimeptijp eLOGSUM βββ

where βt , βc and βm,t are the coefficients on time, cost, and the alternative-specific constants in the joint mode-departure time choice model.

In addition to the estimated coefficients (β) and calculated statistical significance, standardized beta coefficients are also given. These values were calculated by multiplying the β-values by the standard error of the associated explanatory variable (Xi) and the response variable (Y). These standardized coefficients, as well as the calculated elasticities, give a sense of the practical significance for each variable. Elasticities were calculated at the means (by multiplying the calibrated beta coefficients by the average value of the dependent [rent] variable, and dividing by the average value of the associated independent variable). The final adjusted R2 value is near 0.8, suggesting a reasonable fit. Additionally, it seems that the predictive power of the apartment-specific aspects (adjusted R2 = 0.339) is much lower than the location information (adjusted R2 = 0.624), when these explanatory attributes are examined separately.

19

Table 4.3 Weighted least squares regression of monthly rent WLS regression of monthly rent (final specification)

β Std. β p-value Elasticities Constant 389.68 85.83 0.000 (Number of bedrooms)^2 41.55 4.42 0.000 0.192 Interior square footage 8.49E-02 0.04 0.028 0.108 Suburban indicator (base) Urban indicator -402.11 42.16 0.000 -0.418 Distance to CBD (miles) 136.02 69.21 0.051 1.317 (Distance to CBD) ^2 -7.68 2.45 0.002 -0.593 Natural logarithm of distance to CBD -589.34 200.41 0.004 -1.534 Number of bus stops per sq. mile 4.32 0.61 0.000 0.437 Median TSZ household income (dollars) 2.54E-02 0.00 0.000 1.270 Employment density (jobs per sq. mile) 4.68E-02 0.01 0.000 0.105 Logsum for home-based work trips 6.32E-02 0.02 0.000 -0.455

Number of observations 236 Adjusted R2 0.781

As would be expected, larger apartments are more expensive (in terms of square footage and number of bedrooms). One would think that the number of bedrooms would be negatively correlated with rent values when controlling for square footage (since increasing the number of bedrooms with the same overall interior size would translate to smaller, cramped bedrooms), which is not the case here. Perhaps, this is because space is already quite limited in apartments and residents would prefer sectioned off areas for added personal space, especially in situations with multiple roommates.

Location characteristics in the model give intuitive results, such as high rent values for apartments closer to the Central Business District (CBD)7, high density areas (in terms of employment), and more accessible destinations. Using quadratic and logarithmic transformations, in addition to the linear form, of the distance-to-CBD variable reveal rapidly decreasing rent values within 3 miles of the CBD, then a slower decrease for further distances. It seems that apartment dwellers are willing to pay more for accessibility. Interestingly, school quality was not found to be significant.

Calculated elasticities and standardized beta coefficients indicate that distance to the CBD and median household income are the most practically significant. As shown in Table 4.2, attractive neighborhood appearance is important to apartment dwellers, and median household income for the neighborhood may be a reflection of this attribute.

As hypothesized earlier, households often make tradeoffs when choosing where to live; one such tradeoff may be space versus accessibility. The two features are added utility for the household, but costs often constrain the household from maximizing space and accessibility. The WLS regression model confirms this by reflecting larger rent values for large and centrally located apartments. Figure 4.2 graphically represents the predicted rent values as distance to the CBD increases, for 1-, 2-, and 3-bedroom apartments.

20

0

500

1000

1500

2000

2500

0 1 2 3 4 5 6 7 8 9 10

Distance to CBD (miles)

Pred

icte

d re

nt v

alue

s ($

)

1 bedroom2 bedrooms3 bedrooms

Figure 4.2 Predicted rent values versus distance to CBD for varying number of bedrooms

Note: In calculating the predicted rent values, all other explanatory variables were assumed to hold their average values from the data set, as shown in Table 6.1.

The figure confirms the hypothesis that households may tradeoff a variety of attributes, such as accessibility, subject to a given cost constraint. For instance, theoretically, a household could choose either a 1-bedroom apartment approximately 1.5 miles from the CBD or a 3-bedroom apartment 3.7 miles from the CBD for the same monthly rent, everything else constant. Although this regression provides an estimation of the monetary values for some housing attributes, more insight may be provided by asking households which features are preferable to them, when comparing different options.

4.3.2 Binary Logit Results for Scenarios The six stated preference questions were developed in order to appreciate which apartment respondents prefer. All six scenarios presented a choice between an enhanced apartment or neighborhood feature and a transportation improvement. Respondents were asked to consider their current apartment and location with the following enhanced options: 1) an addition of 200 square feet of apartment floor space or freeway proximity that reduces their commute time by half; 2) an apartment with 2 close friends and 1 relative nearby or an apartment near a light rail station that can take the respondent to work or school; 3) an apartment with plenty of parking or a downtown apartment with only one parking space (and additional parking spaces cost $60 per month per space); 4) an apartment close to a shopping center or one with a larger kitchen and living room; 5) an apartment close to a bus stop with a route that contains a stop close to their place of employment or one offering a park view but not near a bus route; 6) a brand new apartment and complex or an older apartment that is within 5 miles of a shopping mall. (Refer to Appendix A for full text of scenario questions.) Tables 4.4 and 4.5 show the model results for the six scenarios. In every comparison, Apartment 2 is the base choice, meaning that the parameter

21

estimates represent the additional utility of Apartment 1, as compared to Apartment 2. As before, elimination of statistically insignificant variables and intuitive considerations have been used to obtain the final specifications. A p-value of 0.20 was generally accepted as the upper limit of statistical significance. However, the relatively small sample sizes make it difficult to obtain statistical significance on all variables of interest.

Table 4.4 Final specification for scenario questions (1-3) for apartment dwellers

Add 200 SF of apartment size vs. 50% reduction in

commute time

Friends & relatives nearby

vs. light rail access to job/school

Plenty of parking vs. downtown location with

limited parking Variable

β p-value β p-

value β p-value

Constant 1.095 0.026 3.511 0.000 5.777 0.000 Monthly rent ($) -1.49E-03 0.065 Urban -2.247 0.000 -1.075 0.007 Distance to CBD 0.064 0.157 -0.279 0.000 Median household income (dollars) 1.58E-05 0.011 Density employment per sq. mile 4.66E-04 0.002 4.85E-04 0.005 Logsum for home-based work trips 2.29E-04 0.102 -0.001 0.000 Number of workers in household 0.374 0.084 Number of licensed drivers -0.659 0.009 Living alone 0.326 0.038 Living with friend(s) 1.080 0.050 Living with family 0.830 0.050 1.325 0.006 Living with significant other 0.987 0.039 Married 0.374 0.021 2.109 0.011 Married & have at least one child -1.757 0.067 Male (survey respondent) 0.295 0.017 -0.560 0.084 Employed full-time 0.309 0.048 Retired 0.938 0.080 #Observations 230 233 232 Cox & Snell R Square 0.142 0.121 0.197 Nagelkerke R Square 0.189 0.161 0.272 Market shares (home 1 vs. home 2) 45% vs. 55% 55% vs. 45% 65% vs.35%

22

Table 4.5 Final specification for scenario questions (4-6) for apartment dwellers

Shopping access vs. larger kitchen and

living room Bus access vs. view New complex vs.

shopping access Variable

β p-value β p-

value β p-value

Constant -3.541 0.003 -0.334 0.584 0.122 0.898 Number of bedrooms -0.464 0.160 -0.695 0.023 -0.796 0.034 Interior size (sq. ft.) 1.39E-03 0.036 1.82E-03 0.031 Urban 2.184 0.000 -1.246 0.006 Distance to CBD 0.196 0.020 Number of bus stops per sq mile 7.04E-03 0.100 Density population per sq. mile 2.81E-04 0.019 3.15E-04 0.096 Density employment per sq. mile -4.41E-04 0.002 5.25E-04 0.008 Number of workers in household -0.344 0.105 Number of children in household 0.401 0.055 -0.802 0.018 Presence of children (at least one child) 1.815 0.011 No vehicles in household 1.562 0.054 1.776 0.047 Household income ($1000/year) 1.36E-02 0.038 -2.30E-02 0.002 Living with friend(s) 1.158 0.013 Living with family 1.112 0.026 Married 0.671 0.114 Married & have at least one child 1.011 0.124 Retired -2.189 0.009 Less than high school -1.308 0.113 Non-Caucasian 0.663 0.036 -0.639 0.080 #Observations 222 229 224 Cox & Snell R Square 0.181 0.177 0.154 Nagelkerke R Square 0.244 0.237 0.231 Market shares (home 1 vs. home 2) 41% vs. 59% 48% vs. 52% 76% vs. 24%

Because respondents were asked to consider their current apartments and were given the option of two improvements to their existing conditions, the models control for their current apartment and location features. Some residents may not opt for certain features if they already have that feature (i.e., households with large apartments may not chose to increase their apartment size, since they already have enough space) and some may chose to further improve upon an existing attributes because is important to them (i.e., those living in apartments close to the CBD may place higher value on access to shopping and other facilities). The latter hypothesis relates to residential self-selection, which hypothesizes that households chose to live in their neighborhoods because of their preferences and attitudes towards trip-making (i.e. length of commute). Therefore, existing conditions were included in the specifications; and results should help determine which attributes lend themselves to the residential self-selection theory.

For apartment improvements, a respondent’s current number of bedrooms and apartment square footage are found to be significant in two scenarios, but with opposite effects. The results indicate that those living in apartments with smaller bedrooms are more likely,8 relative to those with larger bedrooms, to choose increased shopping access over a new apartment/complex. This may be because of residential self-selection: a household that has chosen to live in an apartment with less desirable apartment features but potentially greater access may seek further

23

improvements in shopping access since it is valuable to them. However, those same types of households (who live in apartments with smaller bedrooms) are more likely to choose a view with a nearby park over bus access. As suggested earlier, apartments with smaller bedrooms may indicate more persons in the household and chose apartments with a greater number of bedrooms as opposed to larger apartments with fewer bedrooms in order to have more privacy; this increase in persons may be children and, therefore, a nearby park may be more important than bus access.

Location-specific variables further illuminate the relationship between revealed and stated preferences. Those living in an urban setting are more likely than those in suburban settings to choose the improved access option in each scenario for which this control variable was found to be statistically significant, further supporting the theory that those who chose high access areas are more likely to prefer those related features in stated preference questions. However, in the case of a new apartment/complex versus improved shopping access, the magnitude of the positive coefficients on population and employment density are large enough to counteract the urban indicator variable.9 For the case of increased home size versus shorter commute times, those living further from the CBD and those living in high-income neighborhoods (a variable possibly controlling for attractive neighborhood appearance) are more likely than those living close to the CBD and in low-income neighborhoods to chose the apartment improvement over a short commute time.

Demographic characteristics were used to determine which types of persons are more likely to prefer access improvements. Those living with family and are married are more likely (than those single, living alone, living with roommates, or living with a significant other) to choose apartment enhancements (e.g., friends and relatives living nearby, plenty of parking, and new apartment complex) over access improvements, as one might expect. Households with two or more children are more likely than those with no children or only one child to opt for a nearby park (where their children can play, ostensibly) than transit access.

Higher-income households tend to value a park view over bus stop proximity and nearby shopping over a larger apartment, ceteris paribus, perhaps because travel costs (including parking) are of less importance to them and they have more money to spend on shopping. Interestingly, those employed full-time are more likely, ceteris paribus, to choose increased apartment size over a shorter commute.

Concerning ethnicity, the results suggest that non-Caucasian households are more interested in shorter travel times (to shopping and bus services) than enhanced apartment features. This may indicate that these demographic groups depend more on public transportation or other non-single-occupancy vehicle modes, or it may be that hey are more time-constrained in their activities. Households with no vehicles were also more likely to choose bus service access (as would be expected, since they are likely to be dependent on transit), as well as shopping access.

Calculated elasticities for each variable in the models provided very interesting results. Elasticities were calculated by multiplying the calibrated beta coefficients by the market share of the scenario, and divided by the average values of the independent variables; these values give a sense of the practical significance of each variable. For instance, an elasticity value of 0.5 represents a 5% in the probability of choosing Apartment 1 over Apartment 2 for a 1% increase in the independent variable. For each apartment- and location-specific variable, the elasticities were greater than 0.9 and the majority of demographic variables have elasticities of less than 0.5,

24

suggesting that the revealed preferences of the households’ chosen apartment are more significant in determining their preferences than their demographic characteristics.

4.3.3 Ordered Probit Analysis of the Importance of Access Ordered probit models were used to explore priorities during the housing search process. Since the variables of primary interest concern accessibility and its impact on location choice, explanatory variables like commute time, distance/travel time to shopping, and access to major freeways and public transportation were studied. Final model specifications, obtained by systematic elimination of statistically insignificant (p-values less than 0.2) and practically insignificant variables, are shown in Tables 4.6 and 4.7, which provide some interesting results. Adjusted Likelihood Ratio Index (LRI) values give a sense of fit for the model; these values were calculated according to the following equation:

only constantsat

covergenceat

likelihood Logvariables# likelihood Log

-1 LRI Adjusted−

=

Table 4.6 Ordered probit results for importance of commute time and distance/travel time to shopping to apartment dwellers

Commute Time to Work

Distance/Travel Time to Shopping Variable

β p-value β p-value Constant 1.438 0.000 0.903 0.000

Marriage or divorce -0.517148 0.21438 Birth/adoption in household 0.952704 0.02501 New job/job transfer 0.453 0.021 Easier commute 0.839 0.003

Reasons for Moving

Newer/bigger/better apartment 0.597 0.011 Number of workers in household -0.240 0.011 Number of children in household 0.588 0.005 0.589 0.000 Presence of children (at least one child) -0.700 0.087 -0.782 0.035

Household Characteristics

Household income ($/year) 6.33E-03 0.043 Living with family 0.822 0.000 Living

Situation Living with significant other 0.446 0.032 Married & have at least one child -0.827 0.038 -1.009 0.001 Age of survey respondent 0.021 0.000 Male (survey respondent) -0.281 0.136

Survey Respondent

Characteristics Non-Caucasian 0.677 0.000 μ (0) 0 N/A 0 N/A μ (1) 0.836 0.000 1.292 0.000 Thresholds μ (2) 2.386 0.000 2.586 0.000 Number of observations 191 226 Loglikelihood at convergence -182.489 -260.366 Loglikelihood: constants only -202.962 -285.708

Adjusted LRI 0.057 0.057

25

Table 4.7 Ordered probit results for importance of access to bus services and major freeways to apartment dwellers

Access to Bus Services Access to Major Freeway(s) Variable

β p-value β p-value Constant 2.374 0.000 2.235 0.000

New job/job transfer -0.482 0.024 Reasons for Moving Less expensive housing -0.735 0.000

Number of workers in household -0.287 0.009 Number of children in household -0.662 0.003 Household has no vehicles 0.367 0.003 Number of vehicles per household member -1.086 0.000 -0.918 0.000

Household Characteristics

Household income ($/year) -8.50E-03 0.006 -3.59E-03 0.169 Living alone (base) Living with friend(s) -0.628 0.000 Living with family 0.357 0.092

Living Situation

Living with significant other 0.340 0.058 Married & have at least one child -0.287 0.185 Full-time student 0.368 0.033 Bachelor's degree or higher -0.307 0.090

Survey Respondent

Characteristics Non-Caucasian 0.446 0.003 0.342 0.025 μ (0) 0 N/A 0 N/A μ (1) 0.713 0.000 0.921 0.000 Thresholds μ (2) 1.723 0.000 2.290 0.000 Number of observations 212 227 Loglikelihood at convergence -254.805 -245.899 Loglikelihood: constants only -291.012 -260.373

Adjusted LRI 0.090 0.029

Model results show an increase in probability of viewing commute time as important for females, non-Caucasians, and those without children. Looking at the magnitudes of these variables, it is clear that number of children in the household is significant. To quantify the combined effects, calculations can be compared between households with a couple of children and many children. For a married couple with two children, the total effect is a 0.65-point reduction (on a scale from 0 to 3) in the importance of commute as opposed to a married couple with six children who experience a 1.46-point increase in the importance of commute. This effect is further magnified for the case of single-parents: a single-parent with two children is likely to increase their importance of commute by 0.27 points, and a single-parent with six children views commute as even more important (an increase of 2.58 points). Therefore, it can be concluded that increases in children and decreases in adults relates to greater concern with commute, which is intuitive because these households probably have greater time constraints.

Specifications for the importance of shopping access provided very similar results, where some characteristics of families (i.e. number of children and those living with family) reflect increases in importance of shopping access, but other family-related variables (i.e. presence of children and married with children) show the opposite effect. The same calculations as previously made (in the case of importance of commute) were made to determine the overall effect of household size on importance of shopping access. For a married couple with two children, the total effect is

26

a 0.74-point reduction (on a scale from 0 to 3)in the importance of commute as opposed to a married couple with six children who experience a 1.52-point increase in the importance of commute. Again, this effect is further magnified for the case of single-parents: a single-parent with two children is likely to increase their importance of commute by 0.37, and a single-parent with six children views commute as even more important (an increase of 2.77 points). Therefore, it can similarly be concluded that increases in children and decreases in adults relates to greater concern with access to shopping.

Transit access provides more intuitive results - and a higher LRI, suggesting a better fit. It is rated as more important by students, non-Caucasians, and those with fewer vehicles, lower levels of education, and lower household income. Freeway access provides less intuitive results – and an LRI of just 0.029. Those who rated freeway access higher tend to be non-Caucasians, unmarried persons without children, and those from lower-income households. Overall, similarities between the models show that non-Caucasian survey respondents are more interested in access (except for shopping access), everything else constant. However, the predictive power of these models (all LRIs are below 0.10) suggests that it is difficult to explain the priorities of households during the moving process at a disaggregate level.

In summary, isolating apartment dwellers from other movers has proven to be valuable. Renters have different reasons for moving and, correspondingly, have different demographic, housing, and location characteristics. They are more likely to move for an easier commute, a new job/job transfer, and cheaper housing. Clearly, they place high value on accessibility if moving for an easier commute is a top reason; also, commute is a high priority when looking for an apartment, second only to price. Surprisingly, quality of and distance to local public schools is not a high priority for the average apartment resident (however, the majority of apartment residents do not have children). Discrete choice model results revealed that the number of children in the household is highly influential on the households’ relative preference to access, especially to their employment and shopping facilities. Models of monthly rent revealed the relative monetary contribution of various housing and location characteristics. Apartment size and centrality (close to CBD and high employment density) increase rents, indicating that households need to make tradeoffs between space and accessibility for a given cost constraint. Survey results and regression analysis highlight key attributes of residential choice for apartment dwellers; these characteristics are further explored by comparing these results to those of homebuyers. The following chapter provides results on reasons for moving, housing and location preferences, and household tradeoffs for recent homebuyers.

ENDNOTES 6 Attributes rated on a scale of 1-4, where 1 is “not at all important” and 4 is “very important”. 7 Distance to the CBD is defined as the Euclidean distance from the TSZ centroid to 6th Street and Congress Avenue. 8 In this context, “more likely” refers to an increase in the probability of choosing a given option, while simultaneously controlling for all other variables included in the model. This terminology is used throughout the thesis. 9 Calculations assume 3,486 persons per square mile and 1,965 jobs per square mile, which are the average values for the urban observations in the sample.

27

CHAPTER 5. RECENT HOMEBUYERS This chapter discusses the analysis of the recent homebuyers’ data set, including final specifications and results of various regression models. Summary statistics are provided for reasons for moving and correlations between those reasons are explored. After the choice has been made, movers have home and location preferences; survey results of their priorities are revealed in this chapter (and ordered probit models highlight the characteristics related to these priorities). Tradeoffs made by the households are explored through the use of hedonic price modeling for home value and binary logit specifications for stated preference questions. Additionally, cross-tabulations and a multinomial logit model of home type quantify probabilities of choosing dwelling types for various demographics, and multinomial location choice models quantify the relative tradeoffs of location attributes in choosing a home. These survey results and model specifications help to identify the determinants of a recent homebuyer’s residential location choice.

5.1 Reasons for Moving Respondents were asked to indicate their “primary reason(s) for moving” to their current home. Although most other surveys (e.g., Murie (1974) and the U.S. Current Population Survey) ask respondents to indicate a single primary reason for moving, it is believed that many households move for multiple reasons (as suggested by Filion et al. [1999]). The results of the survey confirm this hypothesis: almost half of all respondents (48.23%) indicated more than one “primary” reason for moving. Table 5.1 provides these sample results.

Table 5.1 Primary reasons for moving for recent homebuyers Primary Reason for Moving (Survey Results) Frequency Percent

Wanted to own home 481 51.17% Newer/bigger/better home 226 24.04% Other reason 208 22.13% New job/job transfer 201 21.38% Easier commute 177 18.83% Marriage or divorce 96 10.21% Higher quality schools 85 9.04% Less expensive housing 48 5.11% Birth/adoption in household 43 4.57% Change of climate 40 4.26% Attending or graduating from college 30 3.19% Retiring 27 2.87% Member of household moving out of home/need smaller home 12 1.28% Health reasons 10 1.06%

It is interesting to note that the most reported reason for moving was to own a home, suggesting that approximately 51% of all recent homebuyers in Travis County are first-time homebuyers or previous renters. Although this may seem high, the 2000 American Housing Survey suggests that 45.8% of recent buyers had previously lived in rented housing, suggesting that 51% is a reasonable estimate.

Simple bivariate correlations indicate statistically significant associations (correlations are significant at the 0.01 level) between (1) birth/adoption and wanting a newer/bigger/better home

28

and one that is closer to quality schools, as well as (2) retirement and wanting a change of climate, closer access to family and medical facilities, and having an “other” reason for moving. Common “other” reasons include wanting to be closer to family, simply wanting to move to the Austin area, and, to a lesser extent, wanting a larger yard and moving as an investment. Multivariate models in the following section further explore the reasons for moving and their influence on households’ (who had purchased Austin-area homes in the past year) priorities.

5.2 Priorities During Housing Search Households were also asked to consider various housing and location attributes and their importance during their recent search. Table 5.2 offers summary statistics of these results. Consistent with the realtor survey results, price is most important. The quality of and distance to local public schools are less important, which differs in a significant way from what was expected based on the realtor survey results. However, as previously discussed (in Section 3.1), the realtors represent a bias towards households with children, who are more likely to be concerned with school quality.

Table 5.2 Mean score of importance of housing and location attributes for recent homebuyers

Housing/location attributes Mean score10 % Indicating

“Very Important” or “Important”

Price 3.72 99.3% Attractive neighborhood appearance 3.59 96.6% Investment potential or resale 3.40 89.4% Perception of crime rate in the neighborhood 3.36 89.8% Number of bedrooms 3.29 89.9% Commute time to work (or school for full-time students) 3.12 79.1% Noise levels 3.08 80.5% Lot size / yard size 2.86 69.3% Access to major freeway(s) 2.70 64.8% Social composition of the neighborhood 2.69 60.8% Distance/travel time to shopping 2.53 52.5% Quality of local public schools 2.52 50.5% Views 2.49 45.4% Neighborhood amenities / recreational facilities 2.45 49.5% Closeness to friends or relatives 2.25 39.7% Distance to medical services 2.11 31.4% Distance to local public schools 2.04 34.0% Access to bus services 1.57 14.3% Physical disability accommodations 1.47 9.8% Of interest is the relative importance of housing versus location attributes in the data set. By summing the average scores for each attribute, access-related attributes (commute time and access to major freeway(s), bus services, shopping, medical services, and local public schools) account for only 25% of the importance scores for Travis County homebuyers. Figure 5.1 shows this percentage of scores for access attributes averaged across respondents falling into various demographic groups.

29

25.8854 25.168 24.4817 24.1923 24.039 23.7498 23.6460 23.4392 23.43722.4738 21.996

0

5

10

15

20

25

30

Non-C

auca

sian

Married

with

child

ren

Low in

come (

<$50k

)

Ages 3

6-55

Ages 1

8-35

All obs

ervati

ons

High in

come (

>$100

k)

Medium

inco

me (>=$5

0k&<$1

00k)

Cauca

sian

Single-

perso

n hou

seho

ld

Ages 5

6+

Perc

ent

Figure 5.1 Average scores of access attributes as a percentage of total score for all attributes for various

demographic groups of homebuyers Over all observations, the percent weight that access carries, over all scores, ranges from 0% to 50%, with very few (fewer than 3% of the observations) offering a percentage above 35%. However, summing the importance scores for home-specific attributes (i.e., price, investment potential, number of bedrooms, lot size, views, and physical disability accommodations) only accounts for 35% of the average household’s priorities. The rest depends on the neighborhood (i.e., attractive neighborhood appearance, perception of crime rate, noise levels, social composition of the neighborhood, neighborhood amenities, and closeness to friends or relatives). Therefore, the overall location (neighborhood and access attributes) appears to account for 65% of the average household’s importance valuation across all these attributes.

The relative priorities may vary for different types of households on the basis of household size, income, number of workers, and other characteristics. Table 5.3 shows a few of these results. The most significant shifts in priorities occur for households with children, especially households with a married couple, who are more concerned about investment potential, social composition, and quality of and distance to local public schools.

30

Table 5.3 Ranking of importance of housing and location attributes for segmented homebuyers

Housing/location attributes Single-person

household

Married with

children

High-income

(>$100k)

Low-income (<$50k)

Price 1 1 2 1 Attractive neighborhood appearance 2 2 1 2 Investment potential or resale 3 5 3 4 Perception of crime rate in the neighborhood 4 4 4 3 Number of bedrooms 6 3 5 5 Commute time to work (or school for full-time students) 5 7 6 6 Noise levels 7 8 7 7 Lot size / yard size 8 9 8 8 Access to major freeway(s) 9 14 11 10 Social composition of the neighborhood 10 11 10 9 Distance/travel time to shopping 11 13 13 11 Quality of local public schools 16 6 9 16 Views 13 15 12 12 Neighborhood amenities / recreational facilities 14 12 14 14 Closeness to friends or relatives 12 17 16 13 Distance to medical services 15 16 17 15 Distance to local public schools 18 10 15 18 Access to bus services 17 18 19 17 Physical disability accommodations 19 19 18 19 For further analysis, ordered probit models, presented and discussed in the upcoming Section 5.3.3, offer a multivariate look at such priorities, simultaneously controlling for a variety of other household characteristics as well as reasons for moving. With a general idea of homebuyers’ reasons for moving and their priorities, model regressions (presented in the next section) provide more insight by utilizing multivariate analyses of home value, the preferences of households, home type choice, and location choice.

5.3 Model Results Four different types of models were used to analyze the recent mover data set. An ordinary least squares (OLS) hedonic regression model of home value reveal marginal market valuations of various housing features, as well as effects of several location characteristics. Logit models were used to analyze stated preferences in binary experiments, and ordered probit models were used to track levels of associated importance in search criteria. A multinomial logit models helps explain significant factors in choosing a particular dwelling type and location within the region.

Models may be regressed on various home features, location (neighborhood/zonal) attributes, and demographic variables. Zonal attributes were obtained by matching geocoded home addresses to several data sources including CAMPO’s zonal file (which provided information on the zones’ areas, population, number of households, and employment), the 2000 Census of Population (which provided information on median home values, housing units, housing units’ median number of rooms, and average commute times for employed people), and the work by Kalmanje and Kockelman (2004). It is this last work that provided accessibility indices, calibrated from logsums emerging from travel demand models of home-based work and home-based non-work trips11. Because of high correlation between the two indices (0.998), only one of the accessibility variables was used in analysis.

31

5.3.1 Home Value Linear Regression The OLS hedonic model of home purchase prices allows one to quantify (in dollars) many tradeoffs that households make in home selection, across home and location attributes. Table 5.4 provides the final model specification, which was developed based on a process of stepwise addition and deletion and a maximum p-value of 0.10.

Table 5.4 Ordinary least squares regression of home value OLS regression of home value (final specification)

Variables β Std. β p-value Elasticities Constant -127,037 0.01

Attached housing -32,066 -0.06 0.00 -0.024 Number of bedrooms 41,834 0.24 0.05 1.103 (Number of bedrooms)^2 -6,979 -0.24 0.05 -0.606 Number of bathrooms -46,359 -0.23 0.03 -0.838 (Number of bathrooms)^2 19,691 0.40 0.00 0.816 Number of living areas (including studies) 10,846 0.07 0.00 0.165 Age of dwelling (2005 base) -1,402 -0.24 0.00 -0.298 (Age^2) 20.71 0.29 0.00 0.179 Interior square footage 39.86 0.31 0.00 0.701

Hou

sing

-spe

cific

cha

ract

eris

tics

Lot size (acres) 52,762 0.09 0.00 0.171 Rural 12,584 0.03 0.09 0.013 Distance to CBD -8,001 -0.26 0.00 -0.524 Number of bus stops per sq mile 44.29 0.08 0.00 0.022 Mean travel time to work for workers in the area -4,666 -0.18 0.00 -0.961 Median home value 0.33 0.30 0.00 0.483 Logsum for home-based work trips -26.85 -0.20 0.00 1.233 Lo

catio

n-sp

ecifi

c ch

arac

teris

tics

Mean SAT score for local high school 149.79 0.13 0.00 1.311

Number of observations 729 Adjusted R2 0.823 Several physical features of the home are found to be statistically significant. Everything else constant – including square footage, both the number of bedrooms and bathrooms are found to be statistically significant with quadratic effects. Perhaps this is because having additional bathrooms might be considered a luxury and translates to a more valuable home. Because of these quadratic effects, Figure 5.2 illustrates how changes in home size (number of bedrooms and bathrooms) affect the predicted home value.

32

-$40,000

-$20,000

$0

$20,000

$40,000

$60,000

$80,000

$100,000

$120,000

$140,000

0 1 2 3 4

Pred

icte

d C

hang

e in

Hom

e Va

lue

($)

Number of bathroomsNumber of bedrooms

Figure 5.2 Change in predicted home value versus number of bedrooms and bathrooms

A sharp increase in the predicted home value for households with 3 or more bathrooms indicates many bathrooms may be a luxury factor and may also be proxying for other costly attributes not included in the model (such as vaulted ceilings or quality construction). On the other hand, extra bedrooms do not dramatically affect home value, when controlling for the overall interior size of the home.

In addition to the number of bedrooms and bathrooms, other physical features of the home were found to affect home value. As would be expected, attached housing (i.e. duplexes and condominiums) tends to reduce a home’s value by $32,000, which is probably attributed to lack of privacy and possibly increased noise levels, especially since noise was a top priority for movers. A somewhat surprising result is that much older homes (over 68 years old) enjoy higher values. This may be attributed to quality of construction (including hardwood floors, crown molding, or other attributes) and neighborhood design diversity. It also may result from age proxying for other key variables, such as location. Older homes are more central, and the distance-to-central business district (CBD) and other variables somehow may not capture all these effects. Access considerations are not easy to quantify.

Many neighborhood (Traffic Serial Zone [TSZ]) and other location features also are significant. Results indicate that increases in distance to CBD and mean commute times in the area are associated with lower valued homes, as is intuitively expected. One of the most practically significant variables is the logsum measure of regional accessibility (which are based on discrete-choice models of travel demand [as calibrated by Kalmanje and Kockelman 2005]). Higher access for home-based work (HBW)-type trips is estimated to increase home value (elasticity of +1.2%). Even after controlling for several accessibility indices, the neighborhood’s median home value is estimated to have a positive effect, which may proxy for neighborhood appearance, views, the quality of public infrastructure, as well as other variables that are difficult to control for. It is encouraging to see that access to bus service increases the value of the home;

33

although, this variable is the least practically significant. School quality, in terms of mean SAT scores for the local public high school, is associated with higher home values, as would be expected, and is highly elastic.

Overall, the model’s predictive power is quite reasonable (adjusted R2 = 0.823), except in the case of high valued homes (especially for those valued around $1,000,000 or more), probably due to the coding of the data. The survey contained categorical responses for home value, and homes valued at $500,000 or more were coded as $500,000 for analysis. Additionally, it seems that the predictive power of the structural aspects (adjusted R2 = 0.671) is slightly higher than the location information (adjusted R2 = 0.614), when examined separately.

With both home structure and location attributes affecting home value, it is helpful to examine the relationship between the two. For example, Figure 5.3 examines the affect of distance to the CBD on home value, determined from model estimates. These changes are shown for the 25th, 50th, and 75th percentile values for interior size, dwelling age, and school quality (mean SAT scores for a local high school).

$150,000

$175,000

$200,000

$225,000

$250,000

$275,000

$300,000

0 1 2 3 4 5 6 7 8 9 10Distance to CBD (miles)

Pred

icte

d ho

me

valu

e ($

)

Interior size = 1250 sq. feetInterior size = 1750 sq. feetInterior size = 2750 sq. feetAge = 9 yearsAge = 21 yearsAge = 37 yearsSAT = 950SAT = 1050SAT = 1100

Figure 5.3 Predicted home value versus distance to CBD for various levels of interior square footage, dwelling

age, and school quality

Note: In calculating the predicted home values, all other explanatory variables were assumed to hold their average values from the data set, as shown in Table 6.1.

The figure confirms the hypothesis that households may tradeoff a variety of attributes, including increased accessibility, subject to a given cost constraint. For instance, theoretically, a household could choose either a 1250 square-foot home approximately 3 miles from the CBD or a 2750 square-foot home 10.5 miles from the CBD for the same home value, everything else remaining constant, since households value both access and size. While predicting home values gives a sense of the monetary tradeoffs of various attributes, analyzing stated preferences between two home enhancements offers additional insight on households’ relative preferences.

34

5.3.2 Binary Logit Results for Scenarios While the hedonic model for home value offers valuable metrics of revealed preferences using market prices, a great many factors are at play. Stated preference scenarios allow one to control for a host of such potentially confounding variables. Six hypothetical scenarios, each offering two home-choice options, were presented in the survey. (And all other features of each respondent’s current residence were assumed to apply, in order to permit a clear and relatively realistic choice situation.) The scenarios compared pairs of the following attributes: easy freeway access (being within 1 mile of one of Austin’s two major freeways and a 50% commute-time reduction), increased home size (a larger kitchen and living room), toll road access (within 1 mile of a major toll road and a 50% commute-time reduction), transit access (bus stops within a ¼ mile from the home and workplace, or other frequent destination), larger lot/yard, and easy access to shopping facilities (within 1 mile of a shopping center). (For the specific wording of each scenario, the survey is found in Appendix A.)

Binary logit models were calibrated to ascertain household preferences, as a function of a variety of demographic and other control variables, including information concerning their current residence (since these characterized the choice alternatives). Tables 5.5 and 5.6 show the final specifications for all six scenarios. Several variables are not shown in the final specifications because they were not statistically (or practically) significant; but they were considered initially. (These include occupation and type of dwelling unit, for example.)

35

Table 5.5 Final specifications for scenario questions for recent homebuyers Freeway access and ½ commute time vs. larger

kitchen and living room

Larger kitchen and living room

vs. toll road access and ½

commute time

Toll road access vs. bus access Variable

β p-value β p-

value β p-value

Constant 0.502 0.032 0.192 0.674 0.532 0.202 Age of dwelling (years) 1.25E-02 0.029 -1.66E-02 0.000 Down payment (%) 0.024 0.011 -0.019 0.073 Home

features Interior size (sq. ft.) 2.33E-04 0.109 -2.44E-04 0.089 3.31E-04 0.005 All or most friends/family live nearby -1.115 0.036

Suburban 0.320 0.066 Distance to CBD -0.080 0.000 0.077 0.006

Neighborhood features

Median income of neighborhood -6.40E-06 0.074 6.11E-06 0.108 Number of licensed drivers -0.506 0.000 Age -1.32E-02 0.054 Male 0.386 0.043 Number of vehicles available in household 0.204 0.111

Bus use (times per month) 6.00E-02 0.090 -0.056 0.012 HH Income -3.66E-06 0.099 6.88E-06 0.000

Household/ respondent information

Non-white -0.525 0.021 0.655 0.036 -0.443 0.051 Number of observations 721 667 768 Cox & Snell R2 0.05 0.059 0.113

Nagelkerke R2 0.066 0.086 0.239 Market shares (home 1 vs. home 2) 47% vs. 53% 72% vs. 28% 48% vs. 52%

36

Table 5.6 Final specifications for scenario questions for recent homebuyers, cont.

Bus access vs. larger lot size

Larger lot size vs. shopping access

Shopping access vs. freeway

access and ½ commute time Variable

β p-value β p-

value β p-value

Constant 0.4439 0.362 -0.6336 0.092 -0.4846 0.089 Number of bedrooms -0.4723 0.00 Number of bathrooms 0.3928 0.09 Home value ($) -4.18E-06 0.00 -1.46E-06 0.08 Down payment (%) 0.0227 0.04

Home features

Lot size (acres) 1.4336 0.00 Commute time to grocery store 0.0462 0.01 Rural 0.4955 0.11 Suburban 0.3274 0.05 -0.4756 0.000

Neighborhood features

Distance to CBD -0.1073 0.00 0.0718 0.00 0.047 0.010 Number of children in household 0.2723 0.01 Married 0.4113 0.05 Age 2.59E-02 0.00 -3.78E-02 0.00 Male 0.2468 0.14 0.2784 0.07 Number of vehicles available in household -0.2848 0.03 0.2507 0.02

Number of vehicles per licensed driver -0.426 0.05 Bus use (times per month) 0.0895 0.00 No workers in household 1.5614 0.00

Household/ respondent information

Non-white 0.4819 0.02 Number of observations 699 711 819

Cox & Snell R2 0.122 0.118 0.057

Nagelkerke R2 0.171 0.157 0.077 Market shares (home 1 vs. home 2) 32% vs. 68% 48% vs. 52% 38% vs. 62%

In examining the results of the models, there appear to be many similarities between those who favor commute-time reductions via freeway and toll road access. Those in smaller homes are more likely to choose increased freeway and toll road access (over increased home size). Those who live in suburban areas that are relatively close to the CBD, as well as those in lower income neighborhoods, tend to favor commute reductions. Households with homes far from the CBD show similar characteristics: they are more likely to prefer increases in home size (over commute reductions), a larger lot (over increased bus and shopping access), and value shopping more than freeway access, even though these are already likely to be larger homes on larger lots. These results provide further support for residential self-selection: these households chose to live far from the CBD because they place very high value on home attributes rather than reducing trip lengths by living in more central, accessible areas.

Demographically, men, higher income households, and those making a larger down payment are estimated to be less concerned with such access. In terms of transit access, demographic distinctions seem more apparent. Women, older persons, and frequent bus riders are more likely to prefer bus access over toll road access and increased lot size. Non-Caucasians and older persons are more likely to prefer easy shopping access, as opposed to larger lot size and/or easy

37

freeway access. Overall, the two scenarios related to transit access provide the highest predictive power. In both these cases, household income and vehicle ownership are very helpful predictors of response.

5.3.3 Ordered Probit Analysis of the Importance of Access Ordered probit models were calibrated in order to ascertain the importance of various attributes during the housing search process. Only reasons for moving and demographic/personal control variables were used (rather than home/structural and location attributes). Tables 5.7 and 5.8 provide parameter estimates for final specifications of access-related variables; these specifications were developed based on a process of stepwise addition and deletion and a maximum p-value of 0.10 (although variables of practical significance where allowed with p-values slightly greater than 0.10). Ordered probit model specifications for other attributes can be found in Appendix C.

Table 5.7 Ordered probit results for importance of commute time and distance/travel time to shopping for recent homebuyers

Commute Time Distance/Travel Time to Shopping

Variable

β p-value β p-value

Constant 1.632 0.000 1.615 0.000 Marriage or divorce 0.333 0.044 New job/job transfer 0.424 0.000 0.227 0.030 Easier commute 0.956 0.000 0.473 0.000

Reasons for Moving

Newer/bigger/better home 0.297 0.001 Two full-time workers or one full-time, one part-time worker 0.318 0.171

Two workers -0.486 0.017 One full-time worker -0.508 0.009 One part-time worker -0.823 0.002

Employment Status

Full-time student 1.045 0.005 Presence of children (at least one child) -0.490 0.005 Number of licensed drivers -0.152 0.096 Married 0.366 0.000 Married & have at least one child 0.523 0.008 Age (head of household) 0.008 0.034 Male (head of household) -0.301 0.000 -0.331 0.000 Number of vehicles per licensed driver -2.12E-01 0.017 Household income ($/year) 1.72E-06 0.043

Respondent and Household

Characteristics

Non-Caucasian 0.207 0.098 μ (0) 0 N/A 0 N/A μ (1) 1.135 0.000 1.297 0.000 Thresholds μ (2) 2.497 0.000 2.675 0.000 Number of observations 743 806 Loglikelihood at convergence -771.196 -927.375 Loglikelihood: constants only -833.246 -973.439

Adjusted LRI 0.064 0.035

38

Table 5.8 Ordered probit results for importance of access to bus services and access to major freeways for recent homebuyers

Access to Bus Services*

Access to Major Freeway(s) Variable

β p-value β p-value Constant -0.496 0.008 0.974 0.000

Easier commute 0.191 0.063 Member(s) of household moving out of the home/needed smaller home -0.685 0.009

Wanted to own home 0.222 0.029 0.203 0.026 Newer/bigger/better home 0.176 0.053 Attending or graduating from college 0.617 0.026

Reasons for Moving

Health reasons 1.254 0.000 0.559 0.038 Total number of workers in household 0.178 0.087 Full-time student 1.114 0.000 Employment

Status Retired 0.715 0.011 Presence of children (at least one child) -0.140 0.134 Number of licensed drivers 0.156 0.127 Age (head of household) 0.006 0.082 Male (head of household) -0.157 0.106 -0.212 0.012 Number of vehicles available in household -0.273 0.000 No vehicles in household 1.481 0.000 -0.806 0.064 Household income ($/year) 1.92E-06 0.023

Respondent and Household

Characteristics

Non-Caucasian 0.343 0.006 0.281 0.007 μ (0) 0 N/A 0 N/A μ (1) 0.858 0.000 1.042 0.000 Thresholds μ (2) 1.665 0.000 2.634 0.000 Number of observations 673 804 Loglikelihood at convergence -629.252 -903.472 Loglikelihood: constants only -682.506 -924.751

Adjusted LRI 0.062 0.011 * Household income is highly correlated with the variable of interest and was not included in initial or final specifications in order to observe effects of other variables. As hypothesized earlier, knowing the reason(s) for a household’s move can provide insight into the movers’ final home choices (e.g., retirees may locate closer to children or medical facilities, and expecting parents may be interested in larger homes and good schools). As one would expect, those who moved for an easier commute are more likely to indicate that access (of all types except bus services) is a priority. Those who move for a new job (or job transfer) view commute time and shopping access as important.

Demographically, several characteristics were estimated to play important roles in respondents’ valuations of access. For example, in every one of the four models, men are estimated to be less concerned with access than women. Higher income households tend to place greater value on access to shopping and freeways. This seems to contradict the realtor survey results, which indicated that lower-income households are more concerned with accessibility.

39

Additionally, the employment status of household members was also found to be significant. For commute and access to transit, more workers in the household indicated an increased likelihood of valuing accessibility; and the opposite effect seemed to occur for importance of access to shopping; however, household income was not a control variable in the commute and bus service models so the number of workers could be a proxy for wealth. Full-time students are estimated to be more likely to value commute time and transit access. Retirees are more likely to value transit access, which is comforting to see, since such persons may need to start considering other travel options (as age takes its toll on driving abilities).

Married persons are estimated to place greater value on commute and shopping access (than unmarried persons). Those owning more vehicles are less likely to value shopping, freeway, and transit access, while those owning no vehicles place greater value on bus access, as expected. Interestingly, non-Caucasians are more likely to place value on commute and access to freeway and bus services, which is consistent with Cervero and Duncan (2002) who found that those with ethnic backgrounds have a predisposition towards transit use.

Although there was insufficient variation for the importance of price and attractive neighborhood appearance to create an order probit model, other non-access attribute models provided some interesting results. The presence of children plays a vital role in priorities (i.e. quality and distance to public schools models have more than twice the predictive power as the other models), as concluded earlier when attribute rankings were segmented for various demographic characteristics. Specifications for the non-access attribute models can be found in Appendix C.

While these models’ results suggest it is difficult to predict the level of importance that recent movers assign to various access features (all four likelihood ratio index values lie below 0.07), they do illuminate some of the general trends at play. And these trends play a role in home type choice, as described in the following section.

5.3.4 Home Type Multinomial Logit Model Homebuyers have many choices in home type, based on the physical structure (i.e. detached or attached housing), age of dwelling, interior size, and lot size, for example. Simple cross-tabulations for home type versus various demographic characteristics provide a sense as to which types of households are choosing which types of homes. Table 5.9 shows these results. The home type alternatives were categorized by detached versus attached; older (>20 years) versus younger (≤20 years); large interior (>2500 square feet), medium interior (≤2500 and >1500 square feet), versus small interior (≤1500 square feet); and large lot (>0.5 acres) versus small lot (≤0.5 acres). For each cell of home type and demographic group, percentages are provided for the given row/home type (top) and column/demographic (bottom).

40

Table 5.9 Home type versus household composition and income Household income

Home type Total frequency

Single-person household

Married with children ≤$50k >$50k &

≤$100k >$100k

42.86% 8.57% 42.86% 34.29% 22.86% Attached 35 8.93% 1.52% 8.67% 3.68% 2.74% 4.55% 40.91% 4.55% 4.55% 90.91% Detached, older, large

size, large lot 22 0.60% 4.55% 0.58% 0.31% 6.85%

12.12% 36.36% 6.06% 39.39% 54.55% Detached, older, large size, small lot 33

2.38% 6.06% 1.16% 3.99% 6.16% 4.35% 26.09% 13.04% 43.48% 43.48% Detached, older, medium

size, large lot 23 0.60% 3.03% 1.73% 3.07% 3.42%

15.43% 22.36% 16.05% 42.59% 41.36% Detached, older, medium size,

small lot 162

14.88% 18.18% 15.03% 21.17% 22.95%

40.12% 4.82% 44.31% 44.31% 11.38% Detached, older, small size, small lot 167

39.88% 4.04% 42.77% 22.70% 6.51% 8.89% 56.82% 2.22% 8.89% 88.89% Detached, newer, large

size, large lot 45 2.38% 12.63% 0.58% 1.23% 13.70% 5.66% 58.10% 4.67% 30.84% 64.49% Detached, newer, large

size, small lot 107 3.57% 30.81% 2.89% 10.12% 23.63%

22.22% 16.67% 16.67% 72.22% 11.11% Detached, newer, medium size, large lot 18

2.38% 1.52% 1.73% 3.99% 0.68% 19.29% 23.91% 20.98% 54.55% 24.48% Detached, newer,

medium size, small lot

143 16.07% 16.67% 17.34% 23.93% 11.99%

38.89% 5.71% 36.11% 52.78% 11.11% Detached, newer, small size, small lot 36

8.33% 1.01% 7.51% 5.83% 1.37% The most commonly purchased homes (for our database of recent movers in Travis County) are detached, older, smaller-sized homes and detached, medium-sized homes on small lots (both old and new). Attached homes (i.e. townhouses, duplexes, and condominiums) are far less common and more likely to be purchased by single-person and low-income households (although these types of households are most likely to purchase older, small, detached homes). Understandably, married couples with children and high-income households are more likely to occupy large, detached homes (both old and new). Additionally, Figures 5.4 and 5.5 graphically show how the probabilities (sample percentages) of choosing various home types, by individual characteristics, vary by household income.

41

0

10

20

30

40

50

60

70

80

90

100

Less than$15k

$15k -$24,999

$25k -$49,999

$50k -$74,999

$75k -$99,999

$100k -$149,999

$150k -$199,999

$200k ormore

Annual Household Income ($)

% C

hoos

ing

Hom

e Ty

pe)e

Single-family homeAttached housingYounger Home (<20 yrs)Older Home (>=20yrs)

Figure 5.4 Home type (structure and age) versus household income

0

10

20

30

40

50

60

70

80

90

100

Less than$15k

$15k - $24,999 $25k - $49,999 $50k - $74,999 $75k - $99,999 $100k -$149,999

$150k -$199,999

$200k or more

Annual Household Income

% C

hoos

ing

Hom

e Ty

pe

Larger Interior (>2500 sq.ft.)Medium Interior (<=2500,>1500)Small Interior (<=1500)Large lot (>0.5acres)Small lot (<0.5acres)

Figure 5.5 Home type (interior and lot size) versus household income

The figures show some interesting trends as to the relationship between various household income levels and home type. The likelihood of a household choosing a detached home increases as household income increases, where lower incomes (less than $50,000) are the most affected since the changes in probability are smaller at greater levels of household income. Also, for these lower levels of income, choosing an older home is anticipated, whereas the percentage distribution is more even for higher incomes.

42

Relationships between home size and household income exhibit more variability across income levels. A sharp increase in the probability of purchasing a home on a large lot occurs for households with annual incomes over $150,000; and, a slight increase occurs at very low (less than $15,000) incomes, which is surprising. (However, there are very few observation points (n=5) at this lowest income level that reported lot sizes; so the estimate is measured with significant error.) For interior size, the percentage of households choosing small homes steadily decreases with income. Large homes are increasingly likely for higher income households (with little variation between households with incomes between $100,000 and $200,000), and no observed occurrences for households with less than $25,000 annual incomes.

To further investigate the relationship between demographic characteristics and home type choice, a multinomial logit (MNL) model was calibrated. Home type alternatives are consistent with the types of homes shown in Table 5.9, characterized by the home’s physical structure, age, interior size, and lot size. Model specifications were calibrated by stepwise elimination of statistically insignificant variables (p-values less than 0.1) from an initial specification which included 5 alternative-specific variables interacted with various demographic characteristics (i.e. household size, income, number of children, ethnicity, vehicle ownership, respondent age, and number of workers). Table 5.10 gives the final specification for the model, with attached housing units as the base alternative. Disaggregate elasticities were calculated using the following equation, where Xi is the value of the explanatory variable of interest , Pi is the probability of home type, and Vi is the systematic utility associated with alternative i.

i

iiiiii PE

XV

where),1(X∂∂

=−= ββ

The average value over all observations was calculated to determine the average marginal effect of the variable on each of the alternatives, and then averaged over all alternatives (since little variation [standard deviations less than 0.4% of the average] occurred between elasticity values for the alternatives). These elasticity values, calculated at the disaggregate level and averaged over all observations and alternatives, are shown in Table 5.10.

43

Table 5.10 Home type choice model results (using MNL) Variables β p-value Elasticities

Detached (constant) -24.920 0.001 Old (constant) 6.456 0.000 Large interior (constant) -3.339 0.000 Medium interior (constant) -0.974 0.000 Large lot (constant) 12.353 0.077 Household size (specific to large lot) 0.278 0.015 0.470 Household income (specific to detached) -3.96E-05 0.001 -3.388 Household income (specific to large interior) 3.63E-05 0.000 3.122 Household income (specific to medium interior) 1.83E-05 0.000 1.553 Household income (specific to large lot) 2.53E-05 0.000 2.284 Natural logarithm of household income (specific to detached) 2.597 0.001 26.532

Natural logarithm of household income (specific to old) -0.606 0.000 -6.146

Natural logarithm of household income (specific to large lot) -1.511 0.025 -16.463

Number of children (specific to large interior) 0.756 0.000 0.060 Married with children indicator (specific to old) 1.37E-02 0.073 Age (specific to large interior) 8.52E-03 0.010 0.232 Number of workers (specific to old) 0.380 0.002 0.493 Number of workers (specific to large interior) -0.624 0.000 -0.847

Number of observations 786 Loglikelihood at convergence -1540.103 Loglikelihood: constants only -1720.962

Adjusted LRI 0.098 Results show that household income is an important predictive variable, for all alternatives and its home type choice probabilities are highly elastic with respect to income. Interestingly, ethnicity and vehicle ownership were not statistically significant for any of the alternatives. Using these estimated parameters, the following equations represent the utilities of each of the alternatives (where HHSize = household size, HHIncome = household income, MChild = married with children, NChildren = number of children, Age = age of respondent, and NWorkers = number of workers):

1) Vattached=0

2) Vdetached, old, large size, large lot = -9.45 + 0.28*HHSize + 2.21*10-5*HHIncome + 0.48*ln(HHIncome) + 1.37*10-2*MChild + 0.76*NChildren + 8.52*10-3*Age – 0.24*NWorkers

3) Vdetached, old, large size, small lot = -21.80 – 3.24*10-6*HHIncome + 1.99*ln(HHIncome) + 0.76*NChildren + 8.52*10-3*Age – 0.24*NWorkers

4) Vdetached, old, medium size, large lot = -7.08 + 0.28*HHSize + 4.09*10-6*HHIncome + 0.48*ln(HHIncome) + 1.37*10-2*MChild + 0.38*NWorkers

5) Vdetached, old, medium size, small lot = -19.44 - 2.12*10-5*HHIncome + 1.99*ln(HHIncome) +

1.37*10-2*MChild + 0.38*NWorkers

44

6) Vdetached, old, small size, small lot = -18.46 – 3.96*10-5*HHIncome + 1.99*ln(HHIncome) +

1.37*10-2*MChild + + 0.38*NWorkers

7) Vdetached, new, large size, large lot = -15.91 + 0.28*HHSize + 2.21*10-5*HHIncome + 1.09*ln(HHIncome) + 0.76*NChildren + 8.52*10-3*Age – 0.62*NWorkers

8) Vdetached, new, large size, small lot = -28.26 – 3.24*10-6*HHIncome + 2.60*ln(HHIncome) + 0.76*NChildren + 8.52*10-3*Age – 0.62*NWorkers

9) Vdetached, new, medium size, large lot = -13.54 + 0.28*HHSize + 4.09*10-6*HHIncome + 1.09*ln(HHIncome)

10) Vdetached, new, medium size, small lot = -25.89 - 2.21*10-5*HHIncome + 2.60*ln(HHIncome)

11) Vdetached, new, small size, small lot = -24.92 - 2.21*10-5*HHIncome + 2.60*ln(HHIncome)

To determine the probability of a household choosing alternative i (over the set of all alternatives, J), the following equation can be used:

∑=

= J

jj

i

V

ViP

1)exp(

)exp()(

Table 5.11 gives the estimated probabilities for each alternative for three types of households.

Table 5.11 Home type choice probabilities for given household characteristics Input Values

Household size 4 1 8 Household income $70k $40k $200k Number of children 2 0 6 Married with children (indicator) 1 0 1 Age (years) 35 22 40 Number of workers 2 1 2

Probabilities 1) Attached 2.88% 6.78% 0.02% 2) Detached, old, large interior, large lot 2.61% 0.26% 34.95% 3) Detached, old, large interior, small lot 13.20% 2.80% 10.57% 4) Detached, old, medium interior, large lot 4.50% 2.10% 0.27% 5) Detached, old, medium interior, small lot 22.76% 22.45% 0.08% 6) Detached, old, small interior, small lot 16.72% 28.57% 0.01% 7) Detached, new, large interior, large lot 1.63% 0.17% 41.22% 8) Detached, new, large interior, small lot 8.24% 1.85% 12.46% 9) Detached, new, medium interior, large lot 2.81% 1.38% 0.32% 10) Detached, new, medium interior, small lot 14.22% 14.80% 0.10% 11) Detached, new, small interior, small lot 10.44% 18.84% 0.01%

From these calculations and as expected, a household with many children and high income is predicted to be much more likely to chose a detached, single-family home with a large interior and large lot (either old or new) than the two other households shown in this table. The single-person household is likely to choose a small home (and is most likely to choose attached

45

housing, relative to the other two types of households). Figure 5.6 graphically represents howthese probabilities change with household income, everything else held constant.

0%

5%

10%

15%

20%

25%

30%

35%

40%

$0 $50,000 $100,000 $150,000 $200,000

Household Income ($)

Pro

ba

bil

ity

of

ch

oo

sin

gh

om

ety

pe

(%)

Attached

Detached, old, largeinterior,large lot

Detached, old, largeinterior,small lot

Detached, old, mediuminterior,large lot

Detached, old, mediuminterior,small lot

Detached, old, smallinterior, small lot

Detached, new, largeinterior,large lot

Detached, new, largeinterior,small lot

Detached, new, mediuminterior,large lot

Detached, new, mediuminterior,small lot

Detached, new, smallinterior, small lot

Figure 5.6 Probability of home type choice for varying household income

Note: In calculating the predicted probabilities, all other explanatory variables were assumed to hold their averagevalues from the data set, as shown in Table 6.1.

As shown by the figure, household income has a striking effect on choice of home type. Asincome decreases, a household is much more likely to choose attached housing, everything elseheld constant. For instance, households with an annual income of $10,000 has an estimated 31%chance of choosing an attached home (assuming the average values from the data set). High-income households have a greater probability of choosing a large, new, detached home on asmall lot. For example, households with an annual income of $200,000 have an estimated 26%chance of choosing a new, large, detached home on a small lot - though this probability willchange with other factors such as household size and number of children. This information isvery helpful in obtaining a sense of which households choose which home types. But, of course,it is also very helpful to investigate differences in their location choices.

5.3.5 Multinomial Logit Models of Location Choice

A location choice model was calibrated for recent movers, using Travis County’s 544 TSZs. Themovers’ choice set consisted of ten alternatives: nine randomly drawn from the set of TZSs, plusthe chosen option. Model specifications were calibrated by eliminating statistical insignificantvariables (p-values less than 0.1) from an initial specification which included all zonal attributesthat may define a household’s location. Zone “size” was quantified via a natural-log-of-number-of-housing-units control variable (in order to help ensure proportionality between choiceprobabilities and home availability, everything else constant).

First, a pooled model was calibrated, recognizing all sampled households at once. Then thehouseholds were segmented, based on a number of demographic attributes, resulting in a seriesof models for purposes of parameter comparisons. Elasticity values for each variable were

46

calculated at the disaggregate level and averaged over all observations. Table 5.12 presents the pooled model results, and specifications for the segmented models can be found in Appendix D.

Table 5.12 Pooled residential location choice model results (using MNL) Pooled Model

β p-value Elasticities Suburban location 0.476 0.001 0.192 Urban location 0.351 0.038 0.096 Distance to CBD (miles) -0.071 0.000 -0.431 Median household income (dollars) 1.10E-05 0.004 0.579

Ratio of median home value in TSZ to surveyed household income -0.311 0.000 -0.581

Median number of rooms in TSZ per dwelling unit 0.403 0.000 1.806 Population density (persons per sq. mile) 1.14E-04 0.000 0.297 Employment density (jobs per sq. mile) 1.86E-05 0.000 0.030 Logsum for home-based work trips -2.73E-04 0.001 -1.189 Natural logarithm of the number of housing units in TSZ 0.814 0.000 4.333

Number of observations 811 Loglikelihood at convergence -1541.511

Psuedo adjusted R2 0.173

The pooled model results suggest that central locations (closer to the CBD) are preferred, everything else constant – including the logsum measure of regional accessibility. This indicates that centrality offers something more than travel preferences alone reveal.12 While access is no doubt valued by many households, so is a quiet residential neighborhood. The need for balance between these competing objectives is a challenge, for planners, policymakers, developers and others who want to meet households’ preferences – while mitigating congestion, emissions, car dependence, and other associated impacts of longer-distance trip-making.

Median home values, divided by respondent household incomes, were used to describe neighborhood affordability. As expected, more expensive locations are less likely. Also as expected, neighborhoods offering larger homes (a higher median number of rooms per home) are preferred. Finally, the coefficient on the natural logarithm of housing units in a zone is near one, as anticipated on theoretical grounds (as mentioned earlier).

Elasticities were calculated for each of the variables and are reflections of practical significance. For instance, a 1% increase in the median number of bedrooms for the TSZ is found to increase the probability of location choice by 1.81%; location choice is also highly elastic for changes in the accessibility index for home-based work trips.

Data segmentation permits a closer look at behavioral tendencies across demographic groups. Final specifications for models segmented on the basis of presence of children, household size, household income, and number of workers is presented in Appendix D. Variations in parameter values across segmented models suggest that higher income households (i.e. those with annual incomes over $100,000) are more attracted to high-income neighborhood; conversely, lower income households (i.e. those with annual incomes less than $50,000) are more likely to choose denser, urban locations with lower-income households. CBD access and home affordability are estimated to be more important for households with children than those without, which is intuitive since households with children may have tighter budgets and spending costs are likely

47

to increase as children mature. On the other hand, households without children are attracted to denser (in terms of population and employment) locations – even while controlling for distance to CBD; perhaps these households choose to live in denser areas for added social interactions.

Recent homebuyer survey responses reveal their reasons for moving and their preferences, which provide variables of interest for various regression models. Homebuyers’ highest priorities during the housing search process include price, attractive neighborhood appearance, and investment potential. Location characteristics account for the majority (67%) of their priorities, and 25% (or 38% of location attributes) relate to access, suggesting that access is definitely a factor in the decision, but not primarily. The presence of children in the household strongly influences the household’s relative priorities – especially in regard to local public schools.

Ordered probit models analyze multivariate affects on the importance of these attributes. In particular, model results show that the reasons for moving do aid in predicting households’ valuation of various attributes, especially those who move for an easier commute or new job/job transfer who are more likely to be concerned about commute and freeway access, for example. Also, women and non-Caucasians are more likely to be concerned about all types of access; and high-income homebuyers value proximity to shopping and freeways.

Binary logit models for stated preferences scenarios highlight these differences in priorities by allowing respondents to compare attributes to each other. Controlling for revealed preferences (chosen home and location features) reveal that households often choose to enhance specific attributes further, even if they already have more of that particular attribute than the average homeowner. For example, households with large homes still choose increased home size over improved access to freeways and toll roads which would have reduced their commute by half - despite the fact that they probably have longer commutes than those living in smaller homes (which tend to closer to the CBD).

An OLS regression of home value and MNL models of location choice quantify the tradeoffs made by a household, in terms of dollars and probabilities. Hedonic price modeling for home value reveals that structural aspects of the home contribute slightly more predictive power (adjusted R2 value of 0.671 vs. 0.614) than location attributes. The home features included were attached versus detached (single-family) homes, quadratic effects of number of bedrooms and bathrooms, number of living areas, age of dwelling, interior home size, and lot size. Location variables, access-related and zonal attributes, provide some of the most practically significant variables – accessibility index and local school quality. Additionally, the model specifications show the tradeoffs that homes reflect for a given home price – such as home size and accessibility.

Cross-tabulations and an MNL model of home type choice reveal demographic characteristics that correlate with various dwelling types. Household income is statistically significant and yields high elasticity values. Model results show that number of children and workers in the household and age of the respondent also factor but are not as practically significant.

Location choice models reveal that central locations are preferable, even while controlling for home affordability. Zonal attributes for median household income and number of rooms are also attractive. These models provide an in-depth sense for the determinants of home choice for different household characteristics. The next chapter compares these results to those of apartment dwellers to determine the effects of tenure.

48

ENDNOTES 10 Attributes rated on a scale of 1-4, where 1 is “not at all important” and 4 is “very important”. 11 The logsum used here is the expected maximum utility derived across all mode, departure time, and destination combinations available to a trip maker. Kalmanje et al (2004) calibrated nested logit models for Austin area trips using the 1996 Austin Travel Surveys. They considered four modes and five times of day, along with the 1074 TSZ destinations. 12 Though simple in nature, distance-to-CBD measures almost always prove helpful, even in the face of other, more comprehensive variables. See, e.g., hedonic models by Kockelman (1997) and land use change models by Zhou and Kockelman (2005).

49

CHAPTER 6. COMPARISON OF APARTMENT DWELLERS AND RECENT HOMEBUYERS Although separating recent movers by tenure allows for more tailored survey design and analysis, this was done on the assumption that there are significant differences between apartment dwellers and homebuyers. This chapter provides a comparison of the two samples’ characteristics and their preferences. Previous model results show relative preferences for various characteristics; therefore, comparing their specifications will allow similarities among demographic groups to be emphasized. Since access is the main focus of this research effort, preferences towards commute times, proximity to shopping, bus services, and freeway access are discussed here.

In order to compare apartment dwellers and recent homebuyers, it can be helpful to first compare their average characteristics. Table 6.1 shows some key average attributes of the two groups, in terms of demographics and chosen home and location characteristics.

50

Table 6.1 Comparison of sample characteristics for apartment dwellers and recent homebuyers

Variable Mean values for apartment

dwellers*

Mean values for recent

homeowners Household size 2.08 2.27 Number of children (under 16 years old) living in home 0.49 0.51 Presence of children (at least one child) 0.26 0.31 Number of licensed drivers in household 1.53 1.83 Married 0.28 0.55 Married & have at least one child 0.16 0.25 Age of respondent 33.13 39.58 Male (survey respondent) 0.57 0.56 Number of vehicles available to household 1.38 1.95 Number of vehicles per licensed driver 0.88 1.08 Number of vehicles per household member 0.77 0.94 No vehicles in household 6.11E-02 7.95E-03 Household income ($/year) $37,930 $93,256 Caucasian 0.48 0.84 Total number of workers in household 1.30 1.43 Full-time student 0.19 3.51E-02 Retired 4.77E-02 4.79E-02 Attached housing 1.00 9.01E-02 Number of bedrooms 1.63 3.12 Number of bathrooms 1.47 2.14 Number of living rooms (X) 1.80 Age of dwelling (X) 25.18 Interior size (square feet) 862 2083 Lot size (acres) (X) 0.39 Value of home ($) (X) $220,675 Monthly rent ($) $680 (X) Rural (X) 0.12 Suburban 0.28 0.53 Urban 0.72 0.33 CBD (X) 1.55E-02 Distance to CBD 6.71 7.75 Number of bus stops per sq mile 69.98 58.64 Average travel time to work for workers in TSZ (minutes) 23.07 24.40 Median household income (dollars) in TSZ $34,562 $51,037 Median home value/surveyed household income in TSZ 1.22 2.30 Median home value (dollars) in TSZ 34,562 176,182 Median number of rooms per dwelling unit in TSZ 3.43 5.63 Population density (persons per sq. mile) 3,359 3,286 Employment density (jobs per sq. mile) 1,551 1,896 Logsum for home-based work trips -4,989 -5,440 Mean SAT score for local high school 951 1037 *Apartment dweller sample characteristics are weighted.

This comparison shows apartment dwellers to be younger and less likely to be married and/or have children. They have lower incomes, on average, and are more to be a no-vehicle household, when compared to homebuyers. Apartment dwellers are also more ethnically diverse and have

51

greater occurrences of full-time students, as concluded earlier.

Concerning their chosen locations, apartment dwellers tend to live in areas of greater access (i.e. closer to the central business district [CBD] and higher availability of bus services). Interestingly, even though apartment residents tend to live in areas of greater land use intensity, homebuyers and renters, on average, live in areas of approximately equal population density (in part due to larger household sizes homes in less dense purely residential neighborhoods). However, calculated 25th and 75th percentile values for apartment dwellers (2,714 and 3,558 persons per square mile) and homebuyers (973 and 5,004 persons per square mile) reveal that homebuyers’ neighborhoods are much more diverse, in terms of population density, than apartment dwellers’ neighborhoods. For employment density, calculated 25th and 75th percentile values for apartment dwellers (461 and 2,881 jobs per square mile) and homebuyers (135 and 1,388 persons per square mile) reveal that the number of jobs per square mile in renters’ neighborhoods is skewed towards greater employment density, though the two groups’ averages are not too different (1,896 vs. 1,551 jobs per square mile). This suggests that employers are locating in areas further from the CBD (assuming that homebuyers are located further from the CBD than apartment dwellers, on average, as stated earlier), as employment in Austin decentralizes somewhat (in a relative sense, since more jobs are also being attracted downtown). Possibly due to this decentralization, the mean commute times for the average neighborhoods for renters and homebuyers are approximately equal. These are all important characteristics that may help explain preference differences, especially in terms of access.

Overall, apartment dwellers tend to be more concerned with access of all types, as shown in Figure 6.1, which compares the two groups’ average reported levels of importance for access-related attributes. To explore the underlying characteristics of these differences, previously discussed regression model results for the two groups are discussed in terms of four key types of access: commute time, distance to shopping, access to bus services, and proximity to freeways.

0.00 0.50 1.00 1.50 2.00 2.50

Commute

Access to major freeways

Distance to shopping

Distance to schools

Bus services

Mean score

Recent homebuyersApartment dwellers

Figure 6.1 Importance of access-related attributes for apartment dwellers and homebuyers

52

6.1 Commute Time Overall, apartment dwellers tend to be more concerned about commute time than recent homebuyers. An easier commute is the second most common reason for moving and their second highest priority in choosing their current residential location, whereas an easier commute is the fifth most reported reason for moving and is ranked the sixth highest priority for homebuyers. Ordered probit model results show similarities between the two groups. In both cases, those who moved for a new job/job transfer or an easier commute are more likely to view commute time as important, as would be expected. Demographically, women and non-Caucasians place higher importance on commute time. Binary logit results from the presented scenarios provide further support: in both cases, women are more likely to choose commute reductions (including toll road access, for homebuyers13) over increased home size. However, non-Caucasian homebuyers were more likely to choose home enhancement over a shorter commute. As discussed earlier, non-Caucasians are more likely to be apartment dwellers, and perhaps they choose this type of dwelling to increase their levels of accessibility, since apartments tend to be in locations closer to the CBD and number bus stops per square mile, as discussed earlier.

6.2 Distance to Shopping A different type of access is distance to retail and commercial facilities, which is of interest to developers who are considering building stores near residential areas (or vice versa). Apartment dwellers and homebuyers seem to place similar importance on shopping accessibility (mean scores of importance of 2.64 and 2.53 , ranking this feature eighth out of 15 options and eleventh out of 19 options for apartment dwellers and homebuyers, respectively). Ordered probit models suggest different characteristics for renters and owners that value this type of access. For instance, presence of children was found to be very significant in the case of homebuyers but not apartment dwellers. However, in both cases, elderly persons (due to the low magnitude of their associated coefficients) are found to value shopping access more than others; this characteristic also carries practical significance due to high elasticity values in each case. Results of the binary logit models do not show any clear tendencies either. Families with children are more likely to choose home improvement over shopping access, though in the case of homebuyers, elasticities are not very practically significant. Model results suggest that it is difficult to categorize relative importance of shopping access among various demographic groups. Perhaps this is because the value of shopping access differs by various retail facility type. (For example, residents may want to be in close proximity to a grocery store but not a mall because of the added noise and traffic that major retail sites attract.)

6.3 Access to Bus Services Access to bus services is more of a priority to apartment dwellers (mean score of importance of 2.61 vs. 1.57 for homebuyers), as would be expected since renters tend to be lower-income and may not be able to afford a car and must rely on public transportation. Ordered probit model results show that full-time students (both renters and homebuyers) are more likely to value access to transit, as well as those moving to attend college. This is most likely attributed to the lack of parking at the University of Texas at Austin and availability of well-serviced shuttle routes to and from various Austin locations. Again, it is found that non-Caucasians are more likely to value accessibility and have relatively high elasticity values in both models. As would be expected, lower-income households and those with fewer vehicles are also more likely to consider bus service a priority. Binary logit models also support these consistencies: lower-

53

income households, those with fewer vehicles, and non-Caucasians favored increased bus services over the stated home or apartment enhancement.

6.4 Access to Freeways Residents moderately value freeway access (mean scores of importance of 3.08 and 2.70 and ranking this sixth out of 15 options and ninth out of 19 options for apartment dwellers and homebuyers, respectively). Ordered probit models suggest inconsistencies between the models: both found the presence of children, married respondents, and household income to be significant, but specifications reflect opposite effects. The only common element between the apartment renter and homebuyer importance and stated preference models is the positive effect of non-Caucasian status. However, this was not the case in the binary logit models for homebuyers. Both ordered probit specifications contain the lowest adjusted likelihood ratio test values, which may explain why there are few similarities. The following chapter summarizes the important results from the two data sets and their analysis.

ENDNOTES 13 Toll road access was not included in survey of apartment dwellers.

54

55

CHAPTER 7. CONCLUSIONS This work provides new insights into location and dwelling choices by residents of the Austin area. One particularly valuable aspect of the research lies in the data sets themselves. Renters and homebuyers were surveyed separately, and questions ranged from reasons for moving, to rent and home value and apartment/home attributes, to tradeoffs between pairs of key qualities, and to ratings of individual attributes. A focal point of the surveys is access tradeoffs with dwelling and other neighborhood qualities.

A common element between apartment dwellers and recent homebuyers is a recent move. For homebuyers, top reasons for these moves are a desire to own one’s home, wanting a newer/bigger/better home structure, facing a new job or job transfer, and wanting an easier commute. One finds that apartment dwellers typically have very different reasons for moving; for example, a new job (or job transfer) and easier commutes are far more common motivations.

Hedonic models of monthly rent and home value highlight the monetary tradeoffs that households make in choosing their home. For example, logarithm and quadratic model specifications suggest that rents fall sharply for the first three miles from the central business district (CBD) and then decrease less rapidly for further distances; whereas, home values fall $8,000, linearly, per mile. Moreover, the rent model also reveals some counter effects; apartments in an urban area are $400 less per month but higher rents are associated with densely populated areas (in terms of jobs per square mile). Both rent and home value models indicate that increased access to bus services (more bus stops per square of mile) adds value. Physical characteristics of the dwelling unit were also found to be significant. As expected, larger lots and additional bedrooms, bathrooms, and living rooms increase value. Attached housing decreases home value by $32,000, on average. Quantifying such tradeoffs is important to developers when estimating prices and profitability.

Binary logit models of stated preferences were used to evaluate preferences in home versus access improvements. For apartment dwellers, results suggest that multi-person households, married couples, and those with children tend to prefer larger and newer apartments as well as better recreational facilities and suburban locations, while single persons are more likely to choose a shorter commute and more central locations. Life-cycle variables are not as significant in predicting preferences for homebuyers as compared to apartment dwellers. Current home and location variables were found to be statistically and practically significant in many of the models. For instance, households that chose to live farther from the CBD were more likely to choose the larger home and lot size options, as would be expected.

Respondents also indicated the level of importance they placed on a variety of home attributes. Ordered probit models of accessibility ratings and stated preferences across pairs of housing alternatives reveal that a variety of demographic characteristics, including recent reasons for moving, factor prominently. For example, full-time students, those with fewer vehicles, and lower income households tend to place greater importance on bus services. Women and non-Caucasians place a higher value on commute time.

A multinomial logit model of home type choice revealed that household income factors prominently, as well as the number of children and other household members. As expected, increases in household size, number of children, and household income are predicted to result in greater probabilities of choosing larger homes and lots. For example, a married couple with six

56

children and an annual income of $200,000 have an estimated 76% chance of choosing a large, detached single-family home on a large lot; whereas, a single-person household with an income of $30,000 is predicted to choose that home type almost never (less than 1% of the time).

Logit models of location choice were calibrated for recent homebuyers. Results of the pooled model suggest that households prefer central and accessible locations. Other characteristics that were found to be significant include high-income neighborhoods, neighborhood affordability, and higher population and employment densities. Segmentation based on a number of demographic attributes allowed parameter comparisons. Results suggest that as income increases, households are more sensitive to home size (preferring larger homes) – everything else constant. Home affordability is estimated to be more important for married couples with children than single-person households.

The results of these models reveal more than simply who is more attracted to what and what they are willing to pay. For example, the binary logit results suggest that if a developer and/or community wish to attract high-income and single-person households, it might best focus on building apartments close to downtown, while improving access to shopping. In order to attract families with children, however, they should build large apartment complexes or homes in the suburbs with access to recreation facilities.

Another possible goal of communities is greater ethnic and racial integration. Since non-Caucasian respondents appear to value public transit access, improvements in bus and/or additions of light rail service in neighborhoods dominated by Caucasian households may serve such objectives. Additionally, transit-oriented developers may find it best to market their developments towards non-Caucasians and single-persons households (especially women and younger persons) in order to ensure low vacancies. This may be especially useful for upcoming transit-oriented developments in light of the new Austin commuter rail being planned. These examples are just a few possibilities as to how results from this research can be applied. Some potential extensions of this work are described below.

7.1 Limitations and Future Work Although this study offers valuable insights, several extensions would be helpful. Ideally, for the case of apartment dwellers, more persons in more locations would be surveyed, producing greater variety in spatial as well as demographic characteristics. A random (rather than choice-based) sample would permit calibration of a location choice model, to more formally determine the neighborhood, price, and access factors that are at play in apartment choice. For homebuyers, a random sample of residents of the three-county region (as originally intended) would be helpful, especially since the availability of land outside of Austin makes Hays and Williamson counties more prone to residential development. Also, a greater number of realtor responses would allow more analysis of their perspectives. Seeking help with Austin’s realtor association (possibly soliciting via newsletters or attending their meetings) might help improve responses.

Concerning survey design, a few enhancements would make analysis more complete. From responses in “other” reasons for moving, the following reasons could be incorporated into the list of possible responses (in order to give respondents more options): to establish one’s own households (move out of parent’s home), to separate from a previous household member (such as a roommate), to move to different city, to have a larger yard, to be closer to family or friends, and as investment. Analysis could also benefit if respondents were to rank their reasons for moving, in order to provide a better sense of order and magnitude of preference. Also, an “other”

57

category should be included in the list of priorities; otherwise, respondent rankings only indicate relative preferences for listed attributes. “Other” responses would also determine if any key characteristics are missing. Data could also be improved by asking questions about residents’ previous residential locations and dwellings.

As for model specifications, supplementary data on grammar school quality and neighborhood crime rates may have improved the hedonic price and location choice models. School quality variables were added to the final data set manually near the end of the research effort. This information would have been helpful earlier, during the formation of the multinomial logit location choice model for the nine randomly drawn alternatives. Regressing for local school quality might have improved those models. With such data sets and models in hand, prediction of future land use patterns as well as the viability of new forms of residential design will be greatly enhanced.

In summary, while the home choice decision is very complex, these new data sets and their many associated behavioral models offer various insights. The reasons for a move and priorities in home selection, the hedonic models of home and rent value, the paired comparisons of potential home enhancements, the importance scores of various attributes, and the logit models of home type and location choice should allow researchers, planners, and developers to more accurately characterize the tradeoffs households make in their home/location choices. When coupled with models of life cycle changes, land development and population growth, as well as travel demand, vehicle ownership and other behaviors, such models facilitate a more integrated look at our communities and their futures.

58

59

Appendix A

Survey Instrument

60

Survey Title: Location Choice and Transportation Demand U.T. Austin Internal Review Board # 2004-10-0100 Dear Resident, The University of Texas at Austin is undertaking an important study to collect information on people who live in apartments and how they decide where to live. The information you provide will be an important step in understanding the reasons and motivations for selecting a particular apartment location. This information is critical to help planners and engineers better understand the residential and transportation needs of the community. You are one of a small number of apartment dwellers randomly selected to represent all apartment dwellers in the Austin area. For this reason, it is extremely important that we receive a response for each apartment dweller selected. You are not obligated to participate in the survey and you can tell us at anytime if you do not wish to take part. However, your input and opinions are very important to us, since it is critical that all the opinions of all apartment dwellers be considered as part of our study. Your individual responses are confidential and will only be used to produce statistical summaries. We will not publish or otherwise release information identifying any person to any other government agency (federal, state, or local) or to any private organization. We estimate that, for the average household, this form will take about 15 minutes to complete. For your convenience, we are offering two methods for completing the survey: If you are unable to complete the written survey in the coming half hour, please visit http://ce397.hypermart.net/index.htm to complete the online version. Thank you very much for your participation. If you have any questions, please contact me directly at (512) 471-0210 or kkockelm@mail.utexas.edu, or my research assistant Ms. Michelle Bina at mbina@utexas.edu. If you are interested in learning more about me and the kind of research I do, please visit my website at: http://www.ce.utexas.edu/prof/kockelman/. Sincerely, Dr. Kara Kockelman C.B. Luce Professor of Civil Engineering & Faculty Sponsor

61

The questions in this section are about your current apartment and neighborhood. By “neighborhood” we mean the residential area in which your residence is located (for example, within the nearest half mile or so). 1. When did you move into your current apartment? Month_______________ Year_______________ 2. How many bedrooms are in your apartment? � One � Three � Two � Four or more 3. How many bathrooms are there in your apartment? � One � Two � One and a half � Three or more 4. How did you choose your apartment? � Solely responsible/primary decision maker � Equal partner or played a significant role � Had some, but not significant role � Did not play any role in choosing my current home location 5. What is the main reason you live in an apartment? (select only one) � An apartment is affordable � An apartment has low maintenance � An apartment is the right size for my needs � I needed a short-term residence � Other (please specify)____________________________ 6. What is the zip code of your apartment? _____________________________________ 7. What is the total monthly rent for your apartment (NOT including utilities, garage costs or other extras)? If you share your apartment, please note the total rent paid, not just your share of the rent. � $0 - $299 � $1,000 - $1,199 � $300 - $499 � $1,200 - $1,399 � $500 - $799 � $1,400 - $1,599 � $800 - $999 � $1,600 or more

8. What is the approximate interior square footage of your apartment? � Less than 400 � 1,200 – 1,399 SF � 400 – 599 SF � 1,400 – 1,599 SF � 600 – 799 SF � 1,600 – 1,799 SF � 800 – 999 SF � 1,800 or greater SF � 1,000 – 1,199 SF � Don’t know 9. How many of your close friends and/or relatives live in your current neighborhood? � All � Some � Most � None 10. Please indicate your PRIMARY reason for moving from your previous residence to your current apartment (please select only one). � Marriage or divorce � Birth/adoption in household � New job/job transfer for either you or another household member � To be closer to work or school/easier commute for either you or another household member � Retiring � Wanted to rent � Wanted new/better apartment � Wanted/needed less expensive housing � Planned to attend or graduate from college � Change of climate � Health reasons � Other:________________________(please specify)

Section 1 - Your Current Apartment

62

11. Please mark the bubble that corresponds to your agreement with the following statements about your apartment.

Strongly Agree Agree Neutral Disagree Strongly

Disagree

Not Applicable/

Don’t Know

A. My apartment is the size I want. Ο Ο Ο Ο Ο Ο

B. My neighborhood has frequent bus service. Ο Ο Ο Ο Ο Ο

C.

There is a commercial center (with shops, stores, and/or restaurants) that is within walking distance (half-mile) of my home.

Ο Ο Ο Ο Ο Ο

D. Biking and/or walking in my neighborhood feels safe and enjoyable.

Ο Ο Ο Ο Ο Ο

E.

My apartment is conveniently located to where I work or go to school.

Ο Ο Ο Ο Ο Ο

F. I think that my apartment is affordable. Ο Ο Ο Ο Ο Ο

G.

My apartment is relatively new (for example, built within the past 10 years, approximately).

Ο Ο Ο Ο Ο Ο

H. My neighborhood seems quiet to me. Ο Ο Ο Ο Ο Ο

I.

Recreational amenities (such as a park, pool or tennis courts) are available nearby (for example, within a mile).

Ο Ο Ο Ο Ο Ο

J. My neighborhood is safe for children to play outdoors.

Ο Ο Ο Ο Ο Ο

K. I enjoy interacting with many of my neighbors. Ο Ο Ο Ο Ο Ο

L. I have easy access to the freeway. Ο Ο Ο Ο Ο Ο

M.

High quality public grammar, middle, and/or high schools are near my apartment.

Ο Ο Ο Ο Ο Ο

63

The questions in this section ask about your daily travel – for example, trips from home to work or to the store. We are interested in your trips only, not those of other members of your household. If you are NOT employed and do not go to school, please skip to Question 17. 12. Which of the following types of transportation is YOUR PRIMARY means of getting to work or school? � Walking � Taxi � Bicycle � Car, Truck, SUV, or Van � Bus 13. Approximately, HOW OFTEN, on average, do you currently take a BUS to get to work or school? � Daily � 2 to 5 times per week � Several times per month � Once per week � Once per month � Less than once per month � Never 14. What is your average, ONE-WAY commute time to work or school? � less than 5 minutes � 5 – 10 minutes � 11 – 20 minutes � 21 – 30 minutes � 31 – 40 minutes � 41 – 50 minutes � 51 – 60 minutes � 61 – 90 minutes � Greater than 90 minutes

15. I feel that my ONE-WAY commute to work or school is: � Too short � About right � Too long 16. What is your average travel time to the grocery store that you use the most? � less than 5 minutes � 5 – 10 minutes � 11 – 20 minutes � 21 – 30 minutes � 31 – 40 minutes � 41 – 50 minutes � 51 – 60 minutes � 61 – 90 minutes � Greater than 90 minutes 17. What is your average travel time to a mall or large shopping area? � less than 5 minutes � 5 – 10 minutes � 11 – 20 minutes � 21 – 30 minutes � 31 – 40 minutes � 41 – 50 minutes � 51 – 60 minutes � 61 – 90 minutes � Greater than 90 minutes

Section 3 - Housing Market Search 18. How long did you spend looking before you found your current apartment? � Did not look � Four to six months � Less than a week � Seven months to a year � One week to three weeks � Greater than one year � One to three months 19. Which sources did you consult during your housing search? (Please check all that apply). � Real estate agent/apartment locator � Knew landlord � Employer � Property management group � Newspaper or other type of publication � Looked around and saw signs � Internet � Did not search because moved in with � Friends or relatives friends or relatives �Other:________________________(please specify)

Section 2 - Your Daily Travel

64

20. How many apartment complexes did you look at (visit), either by walking through or driving by? � None � 16 – 30 � 1 – 15 � More than 30 � 6 – 15 Section 3 – Finding Your Current Apartment 21. During your apartment search process, how important were the following apartment and neighborhood features

22. Were you familiar with your current neighborhood prior to moving here? � Yes � No

23. Are you considering moving within the next year? � Yes � No

Very

Important Somewhat Important

Not too important

Not at all important

No Opinion

A. Price Ο Ο Ο Ο Ο

B. Attractive neighborhood appearance Ο Ο Ο Ο Ο

C. Views Ο Ο Ο Ο Ο

D. Commute time to work Ο Ο Ο Ο Ο

E. Commute time to school Ο Ο Ο Ο Ο

F. Distance/travel time to shopping Ο Ο Ο Ο Ο

G. Social composition of the neighborhood Ο Ο Ο Ο Ο

H. Perception of crime rates in the neighborhood Ο Ο Ο Ο Ο

I. Noise Ο Ο Ο Ο Ο

J. Quality of local public schools Ο Ο Ο Ο Ο

K. Closeness to friends or relatives Ο Ο Ο Ο Ο

L. Access to public transportation Ο Ο Ο Ο Ο

M. Neighborhood amenities / recreational facilities Ο Ο Ο Ο Ο

N. Access to major freeway(s) Ο Ο Ο Ο Ο

O. Distance to local public schools Ο Ο Ο Ο Ο

65

24. If you answered “yes” to Question 23, please indicate why you are considering moving? Please mark all that apply. � Marriage or divorce � Birth/adoption in household � Possibility of new job/job transfer for either you or another household member � To be closer to work or school/easier commute for either you or another household member � Retiring � Want to own home/not rent � Want new/better home � Want/need less expensive housing � Plan to attend or graduate from college � Want a change of climate � Health reasons � Other:________________________(please specify) Section 5 – Scenarios The following are scenarios of two potential residences. Please select the apartment you would prefer to live in, assuming that rent and all other attributes of the apartments and their neighborhoods are the same except those indicated. 25. Apartment 1 has an extra 200 square feet (compared to Apartment 2). However, Apartment 2 is within 2 miles of a major freeway which reduces your commute time (to work or school) by half. Which apartment would you prefer to live in? � Apartment 1 � Apartment 2 26. Apartment 1 is located in a neighborhood where two close friends and one relative live. However, Apartment 2 is located within a half mile of a proposed light rail stop that would take you to work/school?. Which apartment would you prefer to live in? � Apartment 1 � Apartment 2 27. Apartment 1 is located in a complex with plenty of parking for your household and your guests. Apartment 2 is located close to downtown but has only one parking spot for your household (an additional parking spot is available for $60 per month). Which apartment would you prefer to live in?

28. Apartment 1 is located within 1 mile of a large shopping center with a supermarket, restaurants, and retail shops. However, Apartment 2 has a significantly larger kitchen and living room. Which apartment would you prefer to live in? � Apartment 1 � Apartment 2 29. Apartment 1 is located within ½ mile of a public bus stop, which can take you to within ½ mile of your workplace or school. Apartment 2 has a view of a large park but is not on or near a bus route. Which apartment would you prefer to live in? � Apartment 1 � Apartment 2 30. Apartment 1 is located in a brand new complex with new appliances, new carpet, etc. Apartment 2 is older but is located within 5 miles of a shopping mall. Which apartment would you prefer to live in? � Apartment 1 � Apartment

� Apartment 1 � Apartment 2

66

31. Please mark the bubble that corresponds to YOUR agreement with the following

statement Section 7 – Demographics Please complete all of the questions below. Your accurate responses to these questions will help us classify the survey results we obtain. Please keep in mind that all answers will be kept confidential. 32. Including yourself, how many total persons live in your household? � One � Three � Two � Four or more 33. Including yourself, how many workers live in your household? � None � Three � One � Four or more � Two 34. How many persons under 16 years of age live in your household? � None � Three � One � Four or more � Two 35. Including yourself, how many total persons have a driver’s license? � None � Three � One � Four or more � Two 36. Are you married?

� Yes � No 37. Which of the following best describes your living situation? � Living alone � Living with family � Living with friend(s) � Living with significant other 38. What is your ethnicity? � Caucasian � African American � Hispanic � Asian � Other: _______________________(please specify) 39. What is your age? ____________ 40. What is your gender? � Male � Female

Section 6 – Your Opinions

Strongly Agree Agree Neutral Disagree Strongly

Disagree

A. I always look at the closest store first when I want to buy something. Ο Ο Ο Ο Ο

B. There needs to be more highways in order to reduce traffic congestion. Ο Ο Ο Ο Ο

C. Driving a car is my only option in order to get to where I need to go. Ο Ο Ο Ο Ο

D. I think that there should be more bus or other transit options available. Ο Ο Ο Ο Ο

E. I prefer to make as few car trips as possible. Ο Ο Ο Ο Ο

F. Getting to work/school without a car is difficult. Ο Ο Ο Ο Ο

G. It would be hard for me to drive less. Ο Ο Ο Ο Ο

H. I often look for ways to reduce my travel time. Ο Ο Ο Ο Ο

I. I enjoy driving. Ο Ο Ο Ο Ο

67

41. How many vehicles (cars, trucks, vans, SUVs, or motorcycles) does your household own, lease, or have available for daily use?________ 42. How many days per week do you typically drive (for any reason)?________ (0 to 7) 43. What is the highest level of education you have completed? � Less than high school � High school (or equivalent) � Associate’s or technical degree (or equivalent) � Bachelor’s degree � Master’s degree or higher

44. What is your employment status? (Please select only one.) � Employed full-time � Employed part-time � Full-time student � Homemaker - please skip to Question 47 � Unemployed - please skip to Question 47 � Retired - please skip to Question 47 45. What is the zip code for your workplace or school (full-time students only)? Please see the attached map for a listing of Austin zip codes. ____________________

46. Which of the following best describes your occupation? � Office and Administrative Support � Healthcare practitioners, technicians, and support � Sales � Installation, maintenance, and repair � Management, business, and financial operations � Building and grounds cleaning and maintenance � Production, transportation, and material moving � Professional (computer, legal, engineering, architecture, IT, etc.) � Education, training, and library � Protective services (firefighters, police officers, etc.) � Construction and extraction � Personal care and service � Food preparation and serving � Other (please specify) _______________________________ 47. What is your approximate household annual income (before taxes)? Please consider the income for all persons in your household (Income data is very important for the planning process. Please note that all information will be kept highly confidential.) � Less than $15,000 � $75,000 to $99,999 � $15,000 to $24,999 � $100,000 to $149,999 � $25,000 to $49,999 � $150,000 to $199,999 � $50,000 to $74,999 � $200,000 or more 48. What is YOUR approximate hourly wage (if salary, divide annual salary by 2000 to approximate)? (Income data is very important for the planning process. Please note that all information will be kept highly confidential.) � Less than $7.50 per hour � $37.50 to $49.99 per hour � $7.50 to $12.49 per hour � $50.00 to $74.99 per hour � $12.50 to $24.99 per hour � $74.99 to $99.99 per hour � $25.00 to $37.49 per hour � Greater than $100.00 per hour

68

Is there anything else you’d like to tell us regarding your choices about where to live and your choices about daily travel? Please provide comments in the space below. _______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_________________________________

Thank you for your help in this important study. By completing this survey, you are helping planners and decision-makers understand what is most important to apartment dwellers in deciding where to live. For questions or more information, please contact please contact me directly at (512) 471-0210 or kkockelm@mail.utexas.edu, or my research assistant Ms. Michelle Bina at mbina@utexas.edu.

69

Survey Title: Location Choice and Transportation Demand U.T. Austin Internal Review Board # 2004-10-0100 Dear Resident, The University of Texas at Austin is undertaking an important study to collect information on recent movers in the Austin area and how they decide where to live. The information you provide will be an important step in understanding the reasons and motivations for selecting a particular location. This information is critical to help planners and engineers better understand the residential and transportation needs of the community. You are one of a small number of recent movers randomly selected to represent all recent movers in the Austin area. For this reason, it is extremely important that we receive a response for each household selected. You are not obligated to participate in the survey and you can tell us at anytime if you do not wish to take part. However, your input and opinions are very important to us, since it is critical that all the opinions of all residents be considered as part of our study. Your individual responses are confidential and will only be used to produce statistical summaries. We will not publish or otherwise release information identifying any person to any other government agency (federal, state, or local) or to any private organization. We estimate that, for the average household, this form will take about 15 minutes to complete. We recommend that the head of the household that is most knowledgeable about your decision to move be the one to take the survey. Please return the survey to UT Austin as soon as possible, using the enclosed reply envelope. Thank you very much for your participation. If you have any questions, please contact me directly at (512) 471-0210 or kkockelm@mail.utexas.edu, or my research assistant Ms. Michelle Bina at mbina@utexas.edu. If you are interested in learning more about me and the kind of research I do, please visit my website at: http://www.ce.utexas.edu/prof/kockelman/. Sincerely, Dr. Kara Kockelman C.B. Luce Professor of Civil Engineering & Faculty Sponsor

70

The questions in this section are about your current residence and neighborhood. By “residence” we mean the house or any other type of housing unit in which you live. By “neighborhood” we mean the residential area in which your residence is located – the area you consider to be your neighborhood (for example, within the nearest half mile or so). 1. When did you move to your residence? Month_______________ Year_______________ 2. How would you describe your residence? � Single family home � Duplex � Townhouse � Condominium � Other: _______________(please specify) 3. How many bedrooms are there in your residence? � One � Three � Two � Four or more 4. How many bathrooms are there in your residence? � One � Two and a half � One and a half � Three or more � Two 5. How many family/living areas (including studies) are there in your residence? � None � Two � One � Three or more 6. In what year was your residence built?

_____________

7. Were you familiar with your current neighborhood prior to moving here? � Yes � No 8. How many of your close friends and/or relatives live in your current neighborhood? � All � Some � Most � None

9. What was the value of your residence at the time of purchase? � Less than $50,00 � $50,000 - $74,999 � $75,000 - $99,999 � $300,000 - $349,999 � $100,000 - $149,999 � $150,000 - $199,999 � $200,000 - $249,999 � $250,000 - $299,999 � $350,000 - $499,999 � $500,000 or more � I do NOT own my residence 10. What percentage of your new home’s price was your down payment? � 0% - 2% � 11% - 15% � 3% - 5% � 16% - 20% � 6% - 8% � 21% - 25% � 9% - 10% � 25% or more � I did not make a down payment. 11. What is the interior square footage of your residence (not including garages or porches)? � Less than 1,000 SF � 3,000 – 3,499 SF � 1,000 – 1,499 � 3,500 – 3,999 � 1,500 – 1,999 � 4,000 – 4,499 � 2,000 – 2,499 � 4,500 – 4,999 � 2,500 – 2,999 � 5,000 or greater 12. What is your lot size? (For reference, a football field is 1.3 acres, and a 75 ft x 145 ft lot is exactly 0.25 acres.) � Less than 0.25 acres � Between 0.25 acres and 0.49 acres � Between 0.50 acres and 0.99 acres � 1 acre or greater � Not applicable

Section 1 - Your Current Residence

71

13. Please indicate your PRIMARY reason(s) for moving from your previous residence to your current residence. Please select ALL that apply. � Marriage or divorce � Birth/adoption in household � New job/job transfer for either you or another household member � To be closer to work or school/easier commute for either you or another household member � Retiring � Member(s) of household moving out of the home/ needed smaller home

� Wanted to own home � Wanted to live near higher quality schools � Wanted newer/bigger/better home � Wanted/needed less expensive housing � Planned to attend or graduate from college � Change of climate � Health reasons � Other:__________________(please specify)

14. Please mark the bubble that corresponds to your/your household’s agreement with the given statements about your current residence.

Strongly Agree Agree Neutral Disagree Strongly

Disagree

Not Applicable/ Don’t Know

A My/our residence is the size I/we want. Ο Ο Ο Ο Ο Ο

B My/our yard is the size I/we want. Ο Ο Ο Ο Ο Ο

C My/our neighborhood offers frequent bus service. Ο Ο Ο Ο Ο Ο

D

There is a commercial center (with shops, stores, and/or restaurants) that is within walking distance (half-mile) of my/our home.

Ο Ο Ο Ο Ο Ο

E Biking and/or walking in my neighborhood feels safe and enjoyable.

Ο Ο Ο Ο Ο Ο

F

My/our residence is conveniently located to where I (or another member of my household) work or go to school.

Ο Ο Ο Ο Ο Ο

G My residence offers good investment potential. Ο Ο Ο Ο Ο Ο

H I/we consider my/our neighborhood to be quiet. Ο Ο Ο Ο Ο Ο

I

Recreational amenities (such as a park, pool or tennis courts) are available nearby (for example, within a mile).

Ο Ο Ο Ο Ο Ο

J My neighborhood is safe for children to play outdoors. Ο Ο Ο Ο Ο Ο

K I/we enjoy interacting with many of my/our neighbors. Ο Ο Ο Ο Ο Ο

L I/we have easy access to the freeway. Ο Ο Ο Ο Ο Ο

M

High quality public grammar, middle, and/or high schools are near my/our home.

Ο Ο Ο Ο Ο Ο

72

The questions in this section ask about your daily travel – for example, trips from home to the store. 15. What is your average ONE-WAY travel time to the grocery store that you use the most? � Less than 5 minutes � 21 – 30 minutes � 5 – 10 minutes � 31 – 45 minutes � 11 – 20 minutes � More than 45 minutes 16. What is your average ONE-WAY travel time to a mall or other large shopping center? � Less than 5 minutes � 41 – 50 minutes � 5 – 10 minutes � 51 – 60 minutes � 11 – 20 minutes � 61 – 90 minutes � 21 – 30 minutes � More than 90 minutes � 31 – 40 minutes 17. Approximately, HOW OFTEN, on average, do you currently take a BUS to get to work or school or other frequent destination? � Daily � Once per month � 2 to 5 times per week � Less than once per month � Several times per month � Never � Once per week

73

18. Please tell us about the WORKERS in your household. If no one in your household is employed, please skip this question.

Household member

You/Person 1 Person 2 Person 3 Person 4

� Employed full-time � Employed full-time � Employed full-time � Employed full-time a. What is the employment status for the workers in your household?*

� Employed part-time

� Employed part-time

� Employed part-time

� Employed part-time

� Drive alone � Drive alone � Drive alone � Drive alone � Carpool � Carpool � Carpool � Carpool � Bicycle/walking � Bicycle/walking � Bicycle/walking � Bicycle/walking � Bus � Bus � Bus � Bus

b. Which of the following is your/their PRIMARY means of getting to work? � Telecommute/

work at home � Telecommute/ work at home

� Telecommute/ work at home

� Telecommute/ work at home

� Less than 5 min. � Less than 5 min. � Less than 5 min. � Less than 5 min. � 5 – 10 min. � 5 – 10 min. � 5 – 10 min. � 5 – 10 min. � 11 – 20 min. � 11 – 20 min. � 11 – 20 min. � 11 – 20 min. � 21 – 30 min. � 21 – 30 min. � 21 – 30 min. � 21 – 30 min. � 31 – 40 min. � 31 – 40 min. � 31 – 40 min. � 31 – 40 min. � 41 – 50 min. � 41 – 50 min. � 41 – 50 min. � 41 – 50 min. � 51 – 60 min. � 51 – 60 min. � 51 – 60 min. � 51 – 60 min. � 61 – 90 min. � 61 – 90 min. � 61 – 90 min. � 61 – 90 min.

c. What is your/their average, ONE-WAY commute time to work?

� More than 90 min. � More than 90 min. � More than 90 min. � More than 90 min. d. What is the street address or major cross streets of your/their workplace?

__________________

__________________

__________________

__________________

__________________

__________________

__________________

__________________

e. What is the zip code of your/their workplace? Please see the attached map for a listing of Austin zip codes.

__________________

__________________

__________________

__________________

__________________

__________________

__________________

__________________

f. What is the level of education for you/them?

� Less than high school � High school (or equivalent) � Associate’s or technical degree � Bachelor’s degree � Master’s degree or higher

� Less than high school � High school (or equivalent) � Associate’s or technical degree � Bachelor’s degree � Master’s degree or higher

� Less than high school � High school (or equivalent) � Associate’s or technical degree � Bachelor’s degree � Master’s degree or higher

� Less than high school � High school (or equivalent) � Associate’s or technical degree � Bachelor’s degree � Master’s degree or higher

g. What your/their gender?

� Male � Female

� Male � Female

� Male � Female

� Male � Female

* Anyone working 40 or more hours per work is employed full-time. For those holding more than one job, please report the commute time and location for the PRIMARY workplace.

Section 2 – Workers in Your Household

74

19. During your housing search, how important were the following housing and neighborhood features?

Section 3 – Finding Your Current Residence

Very

Important Somewhat Important

Not too important

Not at all important

No Opinion / Not

Applicable A Investment potential or resale Ο Ο Ο Ο Ο B Price Ο Ο Ο Ο Ο

C Attractive neighborhood appearance Ο Ο Ο Ο Ο

D Number of bedrooms Ο Ο Ο Ο Ο E Views Ο Ο Ο Ο Ο

F Commute time to work (or school for full-time students) Ο Ο Ο Ο Ο

G Distance/travel time to shopping Ο Ο Ο Ο Ο

H Social composition of the neighborhood Ο Ο Ο Ο Ο

I Perception of crime rate in the neighborhood Ο Ο Ο Ο Ο

J Noise levels Ο Ο Ο Ο Ο K Lot size / yard size Ο Ο Ο Ο Ο

L Physical disability accommodations Ο Ο Ο Ο Ο

M Quality of local public schools Ο Ο Ο Ο Ο N Closeness to friends or relatives Ο Ο Ο Ο Ο O Access to bus services Ο Ο Ο Ο Ο

P Neighborhood amenities / recreational facilities Ο Ο Ο Ο Ο

Q Access to major freeway(s) Ο Ο Ο Ο Ο R Distance to local public schools Ο Ο Ο Ο Ο S Distance to medical services Ο Ο Ο Ο Ο

75

The following are scenarios of two potential residences. Please indicate which home you would prefer to live in, assuming that all other attributes of the homes and their neighborhoods are the same except those indicated. Some scenarios include reducing your commute time. If you do not commute, please answer this question considering your household’s primary commuter’s needs. If no one is employed in your household, please answer this question for any other frequent trip that you make. 20. House A and House B are exactly like your current residence. However, House A is within 1 mile of the MoPac freeway, which reduces your commute time by half. House B has a larger kitchen and living room. Which house would you prefer to live in? � House A � House B

21. House A and House B are exactly like your current residence. However, House A has a larger kitchen and living room. House B is the same size as your current residence, but it is located within 1 mile of a major toll road that costs 10¢ per mile and reduces your commute time by half. Which house would you prefer to live in? � House A � House B

22. House A and House B are exactly like your current residence. However, House A is near a major toll road that costs $2 to use per day (roundtrip) and makes your commute 15 minutes; but, House A is not near a bus line. House B is located 1/4 mile (1200 feet) from a bus stop with a routing that drops you off 1/4 mile from your workplace (or most frequent destination). Which house would you prefer to live in? � House A � House B 23. House A and House B are exactly like your current residence. However, House A is located 1/4 mile (1200 feet) from a bus stop with a routing that drops you off 1/4 mile from your workplace (or most frequent destination). House B has a larger lot/yard than House A but is not near a bus line. Which house would you prefer to live in? � House A � House B

24. House A and House B are exactly like your current residence. However, House A has a larger lot/yard than House B but is located 5 miles from any shopping. House B is located within a 1 mile of a shopping center that includes a supermarket, a couple of restaurants, and a few retail stores. Which house would you prefer to live in? � House A � House B 25. House A and House B are exactly like your current residence. However, House A lies 2 miles from a major shopping center that has many retail stores, several restaurants, and a movie theater. House B is located within 1 mile of the MoPac freeway, which reduces your commute time by half. Which house would you prefer to live in? � House A � House B

Please complete all of the questions below. Your accurate responses to these questions will help us classify the survey results we obtain. Please keep in mind that all answers will be kept confidential. 26. Including yourself, how many total persons live in your household? � One � Three � Two � Four or more

27. How many persons under 16 years of age live in your household? � None � Three � One � Four or more � Two

Section 4 – Scenarios

Section 5 – Demographics

76

28. Including yourself, how many total persons have a driver’s license? � None � Three � One � Four or more � Two 29. Are you married? � Yes � No 30. What is your ethnicity? � Caucasian � African American � Hispanic

� Asian � Other: _______________________(please specify) 31. What is your age? ____________ 32. What is your gender? � Male � Female 33. How many vehicles (cars, trucks, vans, SUVs, or motorcycles) does your household own, lease, or have available for daily use? ____________

34. What is YOUR employment status? (Please select only one.) � Employed full-time � Employed part-time � Full-time student � Homemaker - please skip to Question 36 � Unemployed - please skip to Question 36 � Retired - please skip to Question 36 35. Which of the following best describes YOUR occupation? � Office and administrative support � Professional (doctors, engineers, architects, attorneys) � Healthcare practitioners & technicians � Installation, maintenance, and repair � Sales � Building and grounds cleaning and maintenance � Management, business, and financial operations � Personal care and services � Production, transportation, and material moving � Military � Food preparation and serving � Education, training, and library � Self-employed � Construction and extraction � Full-time student � Other (please specify) ________________________________ 36. What is your approximate HOUSEHOLD annual income (before taxes)? Please consider the income for all persons in your household. (Income data is very important for the analysis process. Please note that all information will be kept highly confidential.) � Less than $15,000 � $75,000 to $99,999 � $15,000 to $24,999 � $100,000 to $149,999 � $25,000 to $49,999 � $150,000 to $199,999 � $50,000 to $74,999 � $200,000 or more 37. What is YOUR approximate hourly wage (if salary, divide annual salary by 2000 to approximate)? (Income data is very important for the analysis process. Please note that all information will be kept highly confidential.) � Less than $7.50 per hour � $37.50 to $49.99 per hour � $7.50 to $12.49 per hour � $50.00 to $74.99 per hour � $12.50 to $24.99 per hour � $74.99 to $100 per hour

77

� $25.00 to $37.49 per hour � More than $100 per hour Is there anything else you’d like to tell us regarding your choices about where to live and your choices about daily travel? Please provide comments in the space below. _______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

_______________________________________________________________________________

______________________________

Thank you for your help in this important study. By completing this survey, you are helping planners and decision-makers understand what is most important to residents in deciding where to live. If you would like to know the results of this study, we would be very happy to email you a working paper (in August) and then the final paper (in December). Please simply provide your email address: Email Address (for receipt of research papers on this topic): ____________________________________________________ For questions or more information, please contact me directly at (512) 471-0210 or kkockelm@mail.utexas.edu, or my research assistant Ms. Michelle Bina at mbina@utexas.edu.

78

79

Appendix B

Detailed Survey Methodology

80

Apartment Dweller Data Set The target population for the survey was all apartment dwellers within the City of Austin. Because the survey was intended for door-to-door distribution, there were significant limitations in reference to costs and time and not all apartment dwellers could be surveyed. Apartment complexes were randomly selected based on stratification by population and “size” (total rentable units) of the complex. The reasoning for stratifying on the basis of “size” is based on the idea that complexes that are close in “size” are also similar in terms of amenities, which may be important to renters and hard to quantify with supplementary data. All complexes were assigned to 4 groups in which the population of those areas are approximately equal (200,000 persons) in order to create a sample that was representative of Austin. Within each of those areas, six specific apartments selected were chosen randomly with equal numbers of “small” (80 or less rentable units), “medium” (81 to 250 rentable units) and “large” (greater than 250 rentable units) complexes. However, since data collectors were required to receive only 40 completed surveys, some of the randomly selected complexes were not surveyed and only 16 complexes were actually surveyed. Supplementary data sets were obtained from the Census.

Data was collected by going “door-to-door” on Saturdays and Sundays during late February and early March of 2005. The survey was delivered directly to the first adult answering the door and collected from respondents around 30 minutes later. The reasons for choosing this survey method are several: This method permitted faster distribution and response times, as well as higher response rates (Richardson et al 1995). It also permitted better data quality by allowing respondents to get their questions answered directly. Residents were also referred to the internet website if they were not currently available to complete the survey. Candy bars and maps were offered as incentives, and cards advertising the website URL were posted at unopened doors.

The data obtained included 260 observations. Because several responses contained missing data, techniques were used to impute some missing values where feasible. Some missing monthly rent values were imputed by comparing rent values of other apartment units of similar size from the same apartment complex. Square footage was imputed similarly, recognizing the number of rooms and rent levels within each apartment complex. However, since the variation of square footage within each apartment is much greater than rent variations (possibly due to the respondents’ ignorance of exact square footage as compared to rent), some values could not be imputed with sufficient certainty and remained missing.

Missing values for respondent age were imputed using ordinary least squares (OLS) regression techniques. A two-sample t-test, using data from records of those with age information and those without, suggested that age values were missing at random across observations. Stochastic regression imputation was used, which is a technique that uses a random draw to impute the data by adding this random term to a regression model’s central estimate of age. Additionally, as with many surveys, many household income responses were missing. Since this variable was reported categorically (i.e., as “grouped data”), a multi-threshold variation of the tobit model was used in LimDep software in order to provide an underlying continuous model for income prediction. These continuous values were then used for missing values, while category mid-points were used for all reporting households.

Despite these imputation efforts, many observations contained too many missing responses, and only 232 of received surveys were used for analysis. The final data set was weighted to account for age, gender, and household income based on the Census Bureau’s 2000 5% PUMS data for renters (not including institutionalized group housing units or those under the age of 18) who live

81

an a structure classified as an apartment building within the Austin area (more specifically, the 787xx Zip Code Tabulation Area [ZCTA]). The sample weights were created for 18 groups of people characterized by 3 age groups (18-35, 36-55, 56+), 3 income groups ($0-$24,999, $25,000-$49,999, $50,000+), and gender. It is believed the sample, after using these imputation techniques and applying observation weights, reflects the population of interest.

Recent Mover Data Set Initially, the target population of interest was intended to be the three county region (Hays, Travis, and Williamson Counties) surrounding (and including) the City of Austin. USA Data Inc. was to have access to a comprehensive list (containing 8,541 names and addresses) of all recent movers in these counties, based on county assessor records and utility activations. A “random” sample containing just over half of these 8541 identified households was purchased, providing 4,451 names and addresses. Census population data reveals populations for Hays, Travis, and Williamson Counties to be 97,589, 812,280, and 249,967, respectively; and a random sample is expected to approximately reflect these population distributions (8%, 70%, and 22%). However, the purchased list only contained 22 addresses for Hays County and 22 addresses for Williamson County (while containing 4244 addresses for Travis county or 99% of valid addresses), suggesting that a random sample of the complete frame of movers in the three counties was not available to USA Data. Therefore, the target population was reduced to only Travis County and surveys received from Hays and Williamson Counties were omitted from analysis, resulting in 943 observations.

As with the apartment dwellers data set, the sample was compared to Census PUMS data to determine any biases. A PUMS data was created for Travis County homebuyers that had moved within the past year, which only revealed 1,069 observations. A comparison between these two data sets revealed few major discrepancies; nevertheless, various formulations of observations were created on the bases of various combinations of age, income, vehicle ownership, and presence of children. However, none of the attempted weights seemed to bring the sample closer to the “true population” and no weights were used in the analysis. The following table shows characteristics for the unweighted sample, the sample with two different weightings, and PUMS data.

82

Table B. 1 Recent homeowner sample characteristics (including weighting schemes), compared to PUMS data

Variable Sample data (unweighted) N=43

Sample data (weighted [1])

N=858

Sample data (weighted [2])

N=858

PUMS data N=1069

Age of home 25.25 25.52 24.13 15.22 Bedrooms 3.11 3.08 3.25 2.99 Number of children 0.51 0.48 0.62 0.71 Presence of children 30.70% 29.08% 36.70% 38.26% Number of vehicles 1.95 1.88 2.10 1.86 Household income $93,200 $84,360 $117,144 $94,056 Number of workers

no workers 7.88% 8.41% 6.06% 30.03% 1 worker 41.66% 43.80% 37.68% 2.15%

2 or more workers 50.36% 47.79% 56.26% 67.82% Number of acres

less than 1 acre 94.67% 99.95% 92.68% 77.50% 1 acre or more 5.33% 0.05% 7.32% 22.50%

Building Type mobile home N/A N/A N/A 7.48%

boat, RV, van, etc. N/A N/A N/A 0.56% SF detached 89.84% 89.80% 92.36% 79.61% SF attached 3.21% 9.33% 1.95% 5.71%

condominium 6.11% 6.73% 4.74% 6.64% other 0.83% 0.57% 0.95%

[1] weighted on the basis of vehicle ownership, age, and income [2] weighted on the basis of presence of children, age, and income The applied weights seem to move sample average values of household income (even though it was one of the control variables) and the number of single-family homes further from the average values for the PUMS data. Although the first sampling scheme better reflects vehicle ownership, the number and presence of children in the household are less accurate, in comparison to PUMS data. The second sampling scheme corrects for this difference in children in the household but overestimates vehicle ownership and household income. Therefore, neither weighting scheme was applied.

Although the number and presence of children in the household are underestimated in the unweighted sample (in comparison to average PUMS statistics), the survey here asks for the number of children under 16 years of age, whereas the Census asks for the number of children under 18 years of age in the household, which suggests that the estimation may not be as underrepresented as shown by the table. There also seems to be a bias in the age of the dwelling unit. However, age of dwelling unit from PUMS data was obtained by using midpoints of only 9 categories provided (in terms of the year in which the dwelling was built). For homes built during year 1939 or earlier, an age of 65 years was used for the calculation of the average dwelling age, which may skew the PUMS data towards younger homes, especially considering that the sample data has two households whose homes are near 100 years and one that is 147 years old. Therefore, this age bias may not be as extreme as suggested by the table. Because weighting the data does not seem to reduce biases and the PUMS sample is really no larger than the sample obtained in this research effort, the unweighted sample is used for all analyses.

83

Appendix C

Importance of Non-access Related Attributes

84

Apartment Dwellers

Table C. 1 Ordered probit results for importance of price and perception of crime rates to apartment dwellers

Price Perception of Crime Rates Variable

β p-value β p-value

Constant 3.546 0.000 2.533 0.000 Newer/bigger/better apartment 0.585 0.003 Reasons for

Moving Less expensive housing 1.295 0.027 Number of workers in household 0.236 0.041 Number of licensed drivers -0.250 0.150 Number of children in household -0.195 0.074 Presence of children (at least one child) 0.802 0.022 Number of vehicles per household member -0.481 0.021 Household income ($/year) -5.65E-03 0.072

Household Characteristics

Living with friend(s) 0.352 0.152 -0.814 0.000 Married 0.850 0.010 Married & have at least one child -1.109 0.026 Male (survey respondent) -0.455 0.014 Full-time student -0.471 0.009 Bachelor's degree or higher -0.372 0.038 African-American 1.051 0.013

Survey Respondent

Characteristics

Non-Caucasian 0.344 0.046 0.435 0.007 μ (0) 0 N/A 0 N/A μ (1) 1.259 0.000 0.992 0.000 Thresholds μ (2) 2.772 0.000 2.522 0.000 Number of observations 229 223 Loglikelihood at convergence -144.285 -204.319 Loglikelihood: constants only -163.869 -244.420

Adjusted LRI 0.065 0.127

85

Table C. 2 Ordered probit results for importance of attractive neighborhood appearance and noise to

apartment dwellers

Attractive Neighborhood

Appearance Noise

Variable

β p-value β p-value

Constant 2.106 0.000 1.403 0.000 Birth/adoption in household -0.696 0.047 1.094 0.004 New job/job transfer -0.382 0.082 0.392 0.037 Easier commute -0.438 0.043 Newer/bigger/better apartment 0.488 0.013 0.324 0.061

Reasons for Moving

Less expensive housing Number of children in household 0.210 0.113 0.221 0.092 Number of vehicles available in household -0.157 0.127 Household

Characteristics Number of vehicles per licensed driver 0.424 0.011 Living with friend(s) -0.578 0.010 Living Situation Living with family -0.404 0.076 Married 0.689 0.006 0.713 0.003 Married & have at least one child -0.600 0.104 -0.963 0.018 Age of survey respondent 0.020 0.000 Male (survey respondent) -0.310 0.053 -0.610 0.000 Employed part-time 0.734 0.006 Full-time student -0.626 0.000 Unemployed 0.898 0.022

Survey Respondent

Characteristics

African-American 0.441 0.066 μ (0) 0 N/A 0 N/A μ (1) 1.015 0.000 1.176 0.000 Thresholds μ (2) 2.627 0.000 2.427 0.000 Number of observations 226 2219 Loglikelihood at convergence -216.791 -250.341 Loglikelihood: constants only -244.640 -278.018

Adjusted LRI 0.061 0.060

86

Table C. 3 Ordered probit results for importance of social composition and neighborhood amenities to

apartment dwellers Social Composition of

the Neighborhood Neighborhood

Amenities Variable

β p-value β p-value Constant 1.299 0.000 1.483 0.000

Reasons for Moving Newer/bigger/better apartment 0.488 0.001

Number of licensed drivers -0.283 0.003 Number of children in household 0.402 0.003 Number of vehicles per household member -0.357 0.034

Household Characteristics

Living with family 0.453 0.002 Married 0.356 0.131 Married & have at least one child -0.845 0.017 Male (survey respondent) -0.225 0.116 Full-time student -0.565 0.000 Bachelor's degree or higher 0.196 0.188

Survey Respondent

Characteristics

Non-Caucasian 0.383 0.006 μ (0) 0 N/A 0 N/A μ (1) 1.101 0.000 0.893 0.000 Thresholds μ (2) 2.327 0.000 2.234 0.000 Number of observations 223 224 Loglikelihood at convergence -270.345 -270.749 Loglikelihood: constants only -285.863 -286.428

Adjusted LRI 0.033 0.037

87

Table C. 4 Ordered probit results for importance of views and closeness to friends or relatives to apartment

dwellers

Views Closeness to Friends or Relatives

Variable β p-value β p-value Constant 0.858 0.000 1.147 0.000

Birth/adoption in household 0.792 0.124 Newer/bigger/better apartment 0.420 0.007 Attending or graduating from college 0.548 0.129

Reasons for Moving

Change of climate 0.734 0.132 Number of licensed drivers 0.194 0.080 Number of children in household 0.331 0.013 0.484 0.003 Presence of children (at least one child) -1.093 0.001

Household Characteristics

Number of vehicles per licensed driver 0.254 0.188 Living alone (base) Living with friend(s) -0.443 0.066 Living with family -0.434 0.042

Living Situation

Living with significant other -0.394 0.055 Married 0.602 0.021 Married & have at least one child -0.819 0.030 Male (survey respondent) -0.311 0.043 Full-time student -0.494 0.008 -0.568 0.002 Homemaker -1.454 0.089 Bachelor's degree or higher -0.513 0.001

Survey Respondent

Characteristics

Non-Caucasian 0.527 0.001 μ (0) 0 N/A 0 N/A μ (1) 1.331 0.000 1.021 0.000 Thresholds μ (2) 2.468 0.000 2.026 0.000 Number of observations 221 211 Loglikelihood at convergence -261.272 -261.432 Loglikelihood: constants only -279.736 -285.848

Adjusted LRI 0.030 0.050

88

Tab

le C

. 5 O

rder

ed p

robi

t res

ults

for

impo

rtan

ce o

f com

mut

e to

scho

ol, q

ualit

y of

loca

l pub

lic sc

hool

s, an

d di

stan

ce to

loca

l pub

lic sc

hool

s to

apar

tmen

t dw

elle

rs

Com

mut

e Ti

me

to

Scho

ol

Qua

lity

of L

ocal

Pub

lic

Scho

ols`

D

ista

nce

to L

ocal

Pub

lic

Scho

ols

Var

iabl

e

β p-

valu

e β

p-va

lue

β p-

valu

e

Con

stan

t 1.

184

0.00

2 0.

565

0.13

5 -0

.288

0.

514

New

job/

job

trans

fer

-0.7

45

0.00

5

Ea

sier

com

mut

e 0.

580

0.04

3

N

ewer

/big

ger/b

ette

r apa

rtmen

t 0.

487

0.06

3 0.

430

0.04

3

R

easo

ns fo

r M

ovin

g Le

ss e

xpen

sive

hou

sing

0.

636

0.07

9

N

umbe

r of w

orke

rs in

hou

seho

ld

-0.2

34

0.03

4

Pr

esen

ce o

f chi

ldre

n (a

t lea

st o

ne c

hild

)

0.

647

0.00

2 0.

644

0.00

2 H

ouse

hold

C

hara

cter

istic

s N

umbe

r of v

ehic

les p

er h

ouse

hold

mem

ber

-0.6

32

0.05

9 -0

.546

0.

031

Livi

ng a

lone

(bas

e)

Livi

ng w

ith fr

iend

(s)

-0.4

91

0.08

5 Li

ving

with

fam

ily

0.61

8 0.

002

0.68

3 0.

001

0.55

1 0.

009

Livi

ng S

ituat

ion

Livi

ng w

ith si

gnifi

cant

oth

er

0.47

2 0.

037

Age

of s

urve

y re

spon

dent

-0

.023

0.

030

0.02

1 0.

038

Full-

time

stud

ent

-0.5

13

0.01

7

B

ache

lor's

deg

ree

or h

ighe

r 0.

458

0.02

4

-0

.354

0.

078

His

pani

c/La

tino

0.65

7 0.

001

Surv

ey

Res

pond

ent

Cha

ract

eris

tics

Non

-Cau

casi

an

0.53

9 0.

004

0.48

8 0.

007

μ (0

) 0

N/A

0

N/A

0

N/A

μ

(1)

0.52

9 0.

000

0.75

9 0.

000

0.77

4 0.

000

Thre

shol

ds

μ (2

) 1.

551

0.00

0 1.

507

0.00

0 1.

459

0.00

0 N

umbe

r of o

bser

vatio

ns

169

173

170

Logl

ikel

ihoo

d at

con

verg

ence

-1

78.9

49

-189

.650

-1

87.0

42

Logl

ikel

ihoo

d: c

onst

ants

onl

y -2

03.0

45

-233

.053

-2

26.7

25

A

djus

ted

LRI

0.07

4 0.

156

0.14

4

89

Recent Homebuyers

Table C. 6 Ordered probit results for importance of investment potential, perception of crime rates, and number of bedrooms to recent homebuyers

Investment Potential*

Perception of Crime Rates

Number of Bedrooms

Variable β p-value β p-value β p-value

Constant 3.468 0.000 2.237 0.000 2.558 0.000 Birth/adoption in household 0.690 0.011 New job/job transfer 0.231 0.035 Easier commute 0.183 0.099 -0.217 0.042 Member(s) of household moving out of the home/needed smaller home 0.845 0.022

Wanted to live near high quality schools 0.407 0.017 Newer/bigger/better home 0.226 0.025 0.339 0.002

Reasons for Moving

Less expensive housing -0.337 0.059 0.317 0.067 Total number of workers in household -0.124 0.074 Employment

Status Full-time student 0.565 0.021 Number of children in household -0.199 0.008 -0.267 0.003 Presence of children (at least one child) 0.415 0.024 0.571 0.000 Number of licensed drivers -0.287 0.014 Married & have at least one child 0.309 0.040 Age (head of household) -0.016 0.000 0.013 0.000 0.011 0.003 Male (head of household) -0.232 0.007 -0.199 0.022 Number of vehicles available in household 0.312 0.009 0.242 0.000 Number of vehicles per licensed driver -0.522 0.000 Number of vehicles per household member -0.513 0.017 Asian -0.485 0.029 Other Ethnicity -0.389 0.084

Household Characteristics

Non-Caucasian 0.244 0.034 μ (0) 0 N/A 0 N/A 0 N/A μ (1) 1.362 0.000 1.029 0.000 1.648 0.000 Thresholds μ (2) 2.669 0.000 2.422 0.000 3.353 0.000 Number of observations 817 804 807 Loglikelihood at convergence -765.058 -770.348 -708.567 Loglikelihood: constants only -784.121 -799.733 -771.395

Adjusted LRI 0.014 0.022 0.068 *Household income was not included in initial specifications (because of high correlation) in order to observe other effects.

90

Table C. 7 Ordered probit results for importance of noise, lot size, and social composition of the

neighborhood to recent homebuyers

Noise Levels Lot Size Social Composition of the NeighborhoodVariable

β p-value β p-value β p-value Constant 1.133 0.000 1.144 0.000 1.309 0.000

Birth/adoption in household 0.433 0.008 Wanted to live near high quality schools 0.375 0.020 Newer/bigger/better home 0.152 0.096 0.169 0.083 Change of climate 0.359 0.096

Reasons for Moving

Health reasons 0.649 0.061 Total number of workers in household 0.185 0.013 Full-time student -0.579 0.003 Employment

Status Retired 0.578 0.002 Number of children in household -0.281 0.003 0.090 0.104 Presence of children 0.459 0.010 Married 0.137 0.131 Age (head of household) 0.025 0.000 0.006 0.139 Male (head of household) -0.253 0.001 -0.168 0.043 -0.119 0.117 Number of vehicles per licensed driver 0.269 0.010 Household income ($/year) 1.49E-06 0.048 Hispanic/Latino 0.351 0.028

Household Characteristics

Asian -0.409 0.049 -0.484 0.022 μ (0) 0 N/A 0 N/A 0 N/A μ (1) 1.156 0.000 1.453 0.000 1.159 0.000 Thresholds μ (2) 2.663 0.000 2.922 0.000 2.413 0.000 Number of observations 825 791 830 Loglikelihood at convergence -859.434 -852.752 -1005.176 Loglikelihood: constants only -903.441 -879.216 -1018.206

Adjusted LRI 0.041 0.019 0.007

91

Table C. 8 Ordered probit results for importance of views, neighborhood amenities, and closeness to friends or relatives to recent homebuyers

Views Neighborhood Amenities

Closeness to Friends/Relatives

* Variable

β p-value β p-value β p-value

Constant 0.910 0.012 0.811 0.000 1.071 0.000 New job/job transfer -0.323 0.001 Easier commute 0.189 0.063 Retiring 0.491 0.033 Member(s) of household moving out of the home/needed smaller home 0.717 0.030

Newer/bigger/better home 0.161 0.087 0.169 0.074

Reasons for Moving

Less expensive housing 0.365 0.016 Total number of workers in household 0.140 0.068 Employment

Status Retired 0.524 0.009 0.691 0.000 Number of children in household -0.130 0.010 0.131 0.006 Number of licensed drivers -0.400 0.033 -0.105 0.084 Married 0.155 0.100 0.272 0.004 Age (head of household) 0.011 0.001 Male (head of household) -0.224 0.005 Number of vehicles available in household 0.368 0.031 -0.238 0.000 Number of vehicles per licensed driver -0.462 0.105

Household income ($/year) 4.15E-06 0.000 1.65E-

06 0.043

Household Characteristics

Non-Caucasian 0.365 0.002 μ (0) 0 N/A 0 N/A 0 N/A μ (1) 1.489 0.000 1.058 0.000 0.987 0.000 Thresholds μ (2) 2.605 0.000 2.299 0.000 2.043 0.000 Number of observations 800 799 783 Loglikelihood at convergence -927.215 -990.434 -995.697 Loglikelihood: constants only -965.867 -1018.020 -1023.598

Adjusted LRI 0.027 0.015 0.019 *Household income was not included in initial specifications (because of high correlation) in order to observe other effects.

92

Table C. 9 Ordered probit results for importance of quality of local publice schools and distance to local

public schools to recent homebuyers Quality of Local Public Schools*

Distance to Local Public Schools* Variable

β p-value β p-value

Constant 0.480 0.047 -0.252 0.001 New job/job transfer 0.231 0.047 0.229 0.045 Retiring -0.574 0.052 -0.645 0.040 Wanted to own home -0.202 0.059

Reasons for Moving

Wanted to live near high quality schools 1.261 0.000 0.817 0.000 One part-time worker or no workers 0.558 0.007 Employment

Status Full-time student -0.504 0.073 Presence of children (at least one child) 1.414 0.000 1.476 0.000 Married 0.489 0.000 0.205 0.064 Married & have at least one child -0.420 0.108 -0.393 0.082 Age (head of household) -0.008 0.109 African-American 0.599 0.071

Household Characteristics

Non-Caucasian 0.230 0.047 μ (0) 0 N/A 0 N/A μ (1) 0.844 0.000 0.963 0.000 Thresholds μ (2) 1.757 0.000 2.320 0.000 Number of observations 675 689 Loglikelihood at convergence -770.880 -734.764 Loglikelihood: constants only -935.319 -872.157

Adjusted LRI 0.164 0.150 *Household income was not included in initial specifications (because of high correlation) in order to observe other effects.

93

Table C. 10 Ordered probit results for importance of physical disability accommodations and distance to medical facilities to recent homebuyers

Physical Disability Accommodations*

Distance to Medical Facilities* Variable

β p-value β p-value

Constant -0.680 0.154 -0.146 0.470 New job/job transfer -0.279 0.037 0.173 0.087 Wanted to live near high quality schools 0.335 0.033 Newer/bigger/better home Less expensive housing 0.459 0.046 Attending or graduating from college 0.600 0.008

Reasons for Moving

Health reasons 0.798 0.058 Full-time student 0.550 0.082 Employment

Status Retired 0.620 0.010 Presence of children 0.228 0.021 Number of licensed drivers -0.709 0.005 Married 0.236 0.089 0.300 0.001 Age (head of household) 0.034 0.000 0.023 0.000 Male (head of household) -0.197 0.097 -0.187 0.030 Number of vehicles available in household 0.571 0.014 Number of vehicles per licensed driver -0.934 0.019 -0.344 0.003

Household Characteristics

Non-Caucasian 0.732 0.000 0.443 0.000 μ (0) 0 N/A 0 N/A μ (1) 1.076 0.000 1.294 0.000 Thresholds μ (2) 1.985 0.000 2.566 0.000 Number of observations 605 753 Loglikelihood at convergence -483.330 -839.812 Loglikelihood: constants only -539.190 -907.947

Adjusted LRI 0.079 0.060 *Household income was not included in initial specifications (because of high correlation) in order to observe other effects.

94

95

Appendix D

Multinomial Logit Location Choice Segmentation Model Specifications

96

T

able

D. 1

MN

L lo

catio

n ch

oice

mod

el r

esul

ts, s

egm

ente

d fo

r ho

useh

olds

with

and

with

out c

hild

ren

H

ouse

hold

s with

chi

ldre

n H

ouse

hold

s with

out c

hild

ren

β

p-va

lue

Elas

ticiti

es

β p-

valu

e El

astic

ities

Su

burb

an

0.31

3 0.

054

0.14

0 0.

668

0.00

0 0.

257

Urb

an

0.

599

0.00

4 0.

186

Dis

tanc

e to

CB

D

-0.0

84

0.01

3 -0

.539

-0

.060

0.

006

-0.3

49

Med

ian

hous

ehol

d in

com

e (d

olla

rs)

2.31

E-05

0.

001

1.39

0

Rat

io o

f med

ian

hom

e va

lue

in T

SZ to

hou

seho

ld

inco

me

of su

rvey

ed h

ouse

hold

-0

.519

0.

000

-0.8

62

-0.2

14

0.00

0 -0

.411

Hou

sing

uni

t med

ian

room

s 0.

355

0.01

3 1.

698

0.49

9 0.

000

2.14

5 Po

pula

tion

dens

ity

(per

sons

per

sq. m

ile)

9.50

E-05

0.

024

0.18

4 1.

15E-

04

0.00

0 0.

326

Empl

oym

ent d

ensi

ty

(jobs

per

sq. m

ile)

2.

33E-

05

0.00

0 0.

043

Logs

um fo

r hom

e-ba

sed

wor

k tri

ps

-3.4

0E-0

4 0.

017

-1.4

98

-2.4

9E-0

4 0.

011

-1.0

67

Nat

ural

loga

rithm

of t

he n

umbe

r of h

ousi

ng u

nits

in

TSZ

0.71

5 0.

000

3.64

7 0.

852

0.00

0 4.

555

Num

ber o

f obs

erva

tions

24

0 57

3 Lo

glik

elih

ood

at c

onve

rgen

ce

-444

.695

-1

078.

683

Psue

do a

djus

ted

R2

0.19

2 0.

181

97

T

able

D. 2

MN

L lo

catio

n ch

oice

mod

el r

esul

ts, s

egm

ente

d fo

r si

ngle

-per

son

hous

ehol

ds a

nd th

ose

mar

ried

with

chi

ldre

n

Si

ngle

-per

son

hous

ehol

d M

arrie

d w

ith c

hild

ren

in h

ouse

hold

β

p-va

lue

Elas

ticiti

es

β p-

valu

e El

astic

ities

R

ural

(bas

e)

Subu

rban

0.

728

0.01

5 0.

241

0.37

5 0.

036

0.17

5 U

rban

0.

879

0.00

5 0.

318

D

ista

nce

to C

BD

-0.1

17

0.00

2 -0

.751

M

edia

n ho

useh

old

inco

me

(dol

lars

)

2.20

E-05

0.

006

1.37

0 R

atio

of m

edia

n ho

me

valu

e in

TSZ

to h

ouse

hold

in

com

e of

surv

eyed

hou

seho

ld

-0.1

78

0.00

1 -0

.425

-0

.510

0.

000

-0.8

15

Hou

sing

uni

t med

ian

room

s 0.

417

0.00

0 1.

668

0.38

6 0.

017

1.87

6 Po

pula

tion

dens

ity (p

erso

ns p

er sq

. mile

) 1.

32E-

04

0.00

0 0.

411

7.64

E-05

0.

107

0.13

7 Em

ploy

men

t den

sity

(job

s per

sq. m

ile)

3.42

E-05

0.

000

0.09

8

Logs

um fo

r hom

e-ba

sed

wor

k tri

ps

-4

.43E

-04

0.00

4 -1

.957

N

atur

al lo

garit

hm o

f the

num

ber o

f hou

sing

uni

ts

in T

SZ

0.82

3 0.

000

4.32

6 0.

687

0.00

0 3.

467

Num

ber o

f obs

erva

tions

18

7 19

7 Lo

glik

elih

ood

at c

onve

rgen

ce

-341

.657

-3

62.6

05

Psue

do a

djus

ted

R2

0.20

3 0.

197

98

Tab

le D

. 3 M

NL

loca

tion

choi

ce m

odel

res

ults

, seg

men

ted

on th

e ba

sis o

f hou

seho

ld in

com

e

H

igh

inco

me

hous

ehol

ds

(>

$100

k)

Med

ium

inco

me

hous

ehol

ds

(<

=$10

0k a

nd >

$50k

) Lo

w in

com

eho

useh

olds

(<

=$50

k)

β

p-va

lue

Elas

ticiti

es

β p-

valu

e El

astic

ities

β

p-va

lue

Elas

ticiti

es

Rur

al (b

ase)

Subu

rban

0.

623

0.00

3 0.

284

0.46

2 0.

001

0.18

1 0.

749

0.00

8 0.

249

Urb

an

0.57

7 0.

038

0.12

2

1.07

8 0.

001

0.36

5 C

BD

1.

494

0.00

4 0.

028

Dis

tanc

e to

CB

D

-0.1

33

0.00

0 -0

.813

N

umbe

r of b

us st

ops p

er sq

mile

-0.0

01

0.06

5 -0

.046

Mea

n tra

vel t

ime

to w

ork

for w

orke

rs in

th

e ar

ea

0.04

9 0.

020

0.90

3

Med

ian

hous

ehol

d in

com

e (d

olla

rs)

1.32

E-05

0.

004

0.85

1

-3.7

8E-0

5 0.

000

-1.5

34

Rat

io o

f med

ian

hom

e va

lue

in T

SZ to

ho

useh

old

inco

me

of su

rvey

ed

hous

ehol

d

-0.5

53

0.00

0 -0

.846

Hou

sing

uni

t med

ian

room

s 0.

196

0.09

1 0.

963

0.70

6 0.

000

2.97

6 0.

949

0.00

0 3.

719

Popu

latio

n de

nsity

(per

sons

per

sq.

mile

)

1.63

E-04

0.

000

0.41

5 1.

33E-

04

0.00

0 0.

400

Empl

oym

ent d

ensi

ty (j

obs p

er sq

. mile

)

6.11

E-05

0.

001

0.09

5

Logs

um fo

r hom

e-ba

sed

wor

k tri

ps

-5.5

9E-0

4 0.

000

-2.5

29

Nat

ural

loga

rithm

of t

he n

umbe

r of

hous

ing

units

in T

SZ

0.75

2 0.

000

3.96

6 0.

858

0.00

0 4.

407

0.85

6 0.

000

4.39

6

Num

ber o

f obs

erva

tions

30

5 34

7 21

6 Lo

glik

elih

ood

at c

onve

rgen

ce

-595

.588

-6

24.6

39

-376

.820

Ps

uedo

adj

uste

d R

2 0.

149

0.21

6 0.

240

99

T

able

D. 4

MN

L lo

catio

n ch

oice

mod

el r

esul

ts, s

egm

ente

d fo

r tw

o-w

orke

r, o

ne-w

orke

r, a

nd z

ero-

wor

ker

hous

ehol

ds

Tw

o-w

orke

r hou

seho

ld

One

-wor

ker h

ouse

hold

Ze

ro-w

orke

r hou

seho

ld

β

p El

astic

ities

β

p El

astic

ities

β

p El

astic

ities

Su

burb

an

0.23

7 0.

053

0.09

8

1.03

9 0.

001

0.46

7 D

ista

nce

to C

BD

-0

.046

0.

011

-0.2

85

-0.0

92

0.00

1 -0

.543

Maj

or ro

ads p

er sq

mile

0.

012

0.00

2 0.

057

Med

ian

hous

ehol

d in

com

e (d

olla

rs)

1.

84E-

05

0.00

2 0.

967

1.21

E-05

0.

015

0.64

6 R

atio

of m

edia

n ho

me

valu

e in

TSZ

to

hous

ehol

d in

com

e of

surv

eyed

hou

seho

ld

-0.3

09

0.00

0 -0

.473

-0

.420

0.

000

-0.8

54

Hou

sing

uni

t med

ian

room

s 0.

568

0.00

0 2.

565

0.40

9 0.

000

1.81

3

Popu

latio

n de

nsity

(per

sons

per

sq. m

ile)

1.06

E-04

0.

000

0.27

9 9.

64E-

05

0.00

2 0.

239

1.48

E-04

0.

000

0.38

2 Em

ploy

men

t den

sity

(job

s per

sq. m

ile)

1.

83E-

05

0.00

6 0.

031

Lo

gsum

for h

ome-

base

d w

ork

trips

-1.9

4E-0

4 0.

102

0.83

2 -2

.88E

-04

0.04

3 -1

.262

N

atur

al lo

garit

hm o

f the

num

ber o

f hou

sing

un

its in

TSZ

0.

804

0.00

0 4.

320

0.91

8 0.

000

4.79

9 0.

667

0.00

0 3.

454

Num

ber o

f obs

erva

tions

42

4 33

1 69

Lo

glik

elih

ood

at c

onve

rgen

ce

-815

.776

-6

,147

.564

-1

29.6

40

Psue

do a

djus

ted

R2

0.16

3 0.

191

0.17

6

100

101

References

Alonso, W. (1964). Location and Land Use. Cambridge, Massachusetts: Harvard University Press. Anas, A. (1982). Residential Location market and Urban Transportation. New York, New York: Academic Press, Inc. Belden, Russonello, and Stewart Research and Communications. (2004). “2004 American Community Survey: National Survey on Communities.” Washington, D.C. http://www.realtor.org/SG3.nsf/files/NAR-SGA%20Final%20(2004).pdf/$FILE/NAR-SGA%20Final%20(2004).pdf Ben-Akiva, M., and J.L. Bowman. (1998). “Integration of an Activity-based Model System and a Residential Location Model.” Urban Studies, Volume 35, No. 7, p.1131-1153. Bhat, C.R., and J.Y. Guo. (2004) "A Mixed Spatially Correlated Logit Model: Formulation and Application to Residential Choice Modeling," Transportation Research Part B, Vol. 38, No. 2, pp. 147-168. Boehm, T.P. (1982). “A Hierarchical Model of Housing Choice.” Urban Studies. Volume 19, p.17-31. Cervero, R., and M. Duncan. (2002). Residential Self Selection and Rail Commuting: A Nested Logit Analysis. Working Paper for the University of California Transportation Center. Clark, W.A.V. (1982). Modelling Housing Market Search. London, Great Britain: Biddles Ltd, Guildford and King’s Lynn. Cho, C. (1997). “Joint Choice of Tenure and Dwelling Type: A Multinomial Logit Analysis for the City of Chongju.” Urban Studies. Volume 34, No. 9, p.1459-1473. Ellickson, B. (1981). “An alternative test of a joint model of residential mobility and housing choice.” Journal of Urban Economics, Volume 9, p. 56-79. Filion, P., T. Bunting, and K. Warriner. (1999). “The Entrenchment of Urban Dispersion: Residential Preferences and Location patterns in the Dispersed City.” Urban Studies. Vol. 36, No. 8, p.1317-1347. Freedman, O., and C.R. Kern. (1997). “A model of workplace and residence choice in two-worker households.” Regional Science and Urban Economics, 27, 241-260. Friedman, J. (1981). “A Conditional Logit Model of the Role of Local Public Services in Residential Choice.” Urban Studies, 18, 347-358.

102

Gabriel, S.A., and S.S. Rosenthal. (1989). “Household Location and Race: Estimates of a Multinomial Logit Model.” The Review of Economics and Statistics. Vol. 71, No. 2, p. 240-249. Grigg, T.J. (1982). “Residential Location Choice Modeling: Gaussian Distributed Stochastic Utility Functions.” Research Report No. CE 33, University of Queensland. Guiliano, G. (1989). “New Directions for Understand Transportation and Land Use”; Environment and Planning A, 21: p. 145-159. Green, W. (2003). Econometric Analysis. Prentice Hall, 5th edition. Huh, S., and S-J. Kwak. (1997). “The choice of a functional form and variables in the hedonic price model in Seoul.” Urban Studies, Volume 34, No.7, p. 989-998. Kockelman, K. (1997). “The Effects of Location Elements on Home Purchase Prices and Rents: Evidence from the San Francisco Bay Area.” Transportation Research Record No. 1606: 40-50. Kalmanje, S. and K. Kockelman (2004). “Credit-Based Congestion Pricing: Travel, Land Value, and Welfare Impacts.” Transportation Research Record No. 1864: 45-53. Little, R.J.A., and D.B. Rubin. (1987). Statistical Analysis with Missing Data. New York: John Wiley. Mills, E. (1967). “An aggregate model of resource allocation in a metropolitan area.” American Economic Review, Volume 57, p. 197-210. Muth, R.F. (1969). Cities and Housing. Chicago: The University of Chicago Press. Murie, A. (1974). Household Movement and Housing Choice. Centre for Urban and Regional Studies, University of Birmingham. Orford, S. (2000). “Modelling spatial structures in local housing market dynamics: A multilevel perspective.” Urban Studies, Volume 37, No.9, p. 1643-1671. Parsons Brinkerhoff Quade & Douglas, Inc. (1999) NCHRP Report 423A: Land Use Impacts of Transportation: A Guidebook. Washington, DC: National Academy Press. Richardson, A.J., E.S. Ampt, and A.H. Meyburg. (1995). Survey Methods for Transport Planning. Parkville, Australia: Eucalyptus Press. Rosen, S. (1974). “Hedonic Prices and Implicit markets: Product differentiation in pure competition.” Journal of Political Economy, Volume 82, p. 34-55. Schachter, J. (2001). “Why People Move: Exploring the March 2000 Current Population Survey.” Washington, D.C.: US Census Bureau. <http://www.census.gov/prod/2001pubs/p23-204.pdf>

103

Schachter, J. (2004). “Geographic Mobility: 2002 to 2003.” . Washington, D.C.: US Census Bureau. http://www.census.gov/prod/2004pubs/p20-549.pdf Sermons, M. W., and F. S. Koppelman. (2001). “Representing the Differences between Female and Male Commute Behavior in Residential Location Models”. Journal of Transport Geography, Vol. 9, No. 2, p.101-110. Schrank, D., and T. Lomax. (2005). The Urban Mobility Report. Texas Transportation Institute, College Station, TX. Shaw, J. (1994). “Transit-based Housing and Residential Satisfaction: Review of Literature and Methodological Approach.” Transportation Research Record. Volume 1400, p.82-89. Speare, A., S. Goldstein, and W.H. Frey. (1974). Residential Mobility, Migration, and Metropolitan Change. Cambridge, Massachusetts: Ballinger Publishing Company. Tu, Y., and J. Goldfinch. (1996). “A Two-stage Housing Choice Forecasting Model.” Urban Studies. Vol. 33, No. 3, p.517-537. Waddell, P. (1996). “Accessibility and Residential Location: The Interaction of Workplace, Residential Mobility, Tenure, and Location Choice.” Presented at the Lincoln Land Institute TRED Conference. Weisbrod, G., M. Ben-Akiva, and S. Lerman. (1980). “Tradeoffs in Residential Location Decisions: Transportation versus Other Factors.” Transportation Policy and Decision-Making, Vol. 1, No. 1, p. 13-26. Wheaton, W. (1977). Income and Urban Residence: An Analysis of Consumer Demand for Location. The American Economic Review, 67(4), 620-631. Zhou, B., and K. Kockelman. (2005). “Neighborhood Impacts on Land Use Change: A Multinomial Logit Model of Spatial Relationships.” Submitted for presentation at the 85th Annual Meeting of the Transportation Research Board. Washington, D.C. Zondag, B., and M. Pieters. (2004). “Influence of accessibility on residential location choice.” Paper submitted for presentation at the 84th Annual Meeting of the Transportation Research Board. Washington, D.C.