University of Southern California Dissertations and Theses

Reshaping Los Angeles: housing affordability and neighborhood change

(USC Thesis Other)

Reshaping Los Angeles: housing affordability and neighborhood change

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Reshaping Los Angeles Housing Affordability and Neighborhood Change By Sarah Louise Mawhorter Committee: Richard K. Green, chair Dowell Myers, chair Manuel Castells George C. Galster A Dissertation Presented to the Faculty of the Graduate School University of Southern California In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Urban Planning and Development Degree Conferral December 2016 i For Matt Young. Copyright 2016 by Sarah Louise Mawhorter. All rights reserved. ii Abstract In this dissertation, I set out to lay an empirical foundation for understanding why the housing stock in the Los Angeles region has been constrained since the building boom of the 1980s, and why regional demand for housing has continued to grow over the period of the housing boom and bust. I measure the declines in number of housing units available on the market, arising from both reduced construction and reduced turnover. I also examine the sources of growth in potential demand for housing, from migration and from natural increases of the existing population. I then compare the consequences of the housing shortfall for different types of neighborhoods and for different groups of people by age, race, and education level. I build on filtering theory with new ideas about how to conceptualize and measure changes in housing supply and demand, and I introduce new techniques to analyze the connections between regional shifts in supply and demand and local changes in neighborhoods. By combining well- established filtering theory with innovative demographic methods, I find empirical evidence of the powerful, usually hidden forces of the regional housing market and the way they impact various types of neighborhoods and groups of people. I make two main contributions to housing research. First, I argue that the housing supply that matters for growth is the housing available on the market, and this includes both newly built units and existing units that have been turned over by their previous occupants. Second, the underlying sources of growth in housing demand include (1) natural increases in the adult population through larger generations and longer life expectancies, in addition to the more commonly-studied (2) migration, (3) household formation, and (4) changes in income. I find that the number of vacant housing units available on the market fell 12.3 percent from 1990 to 2012 as a result of reduced construction and reduced turnover, while the regional adult population grew 29.2 percent. I also find that most of the growth in housing demand during the boom and bust periods was simply the result of natural increases rather than from immigration or migration to the region. From 2000 to 2012, the adult population grew 15.8 percent through natural increases, and only grew 1.4 percent through migration. In the context of this housing shortfall, with far fewer housing units available on the market and with steadily growing demand, I find stark disparities in housing and neighborhood outcomes for more and less advantaged groups. Middle-aged and older adults fared better than younger newcomers trying to make their way in the housing market, Whites and Asians fared better than Blacks and Hispanics, and young college graduates fared better than young people without college degrees. Keywords Housing affordability, neighborhood change, demography, regional housing market, urban development iii Contents Acknowledgments........................................................................................................................................... viii Introduction ........................................................................................................................................................ 1 No Vacancy: Reduced Housing Availability in the Los Angeles Region ............................... 9 Chapter 1. 1. Introduction ............................................................................................................................................... 9 2. Housing Supply Constraints .................................................................................................................. 11 3. Measuring Housing Availability ............................................................................................................ 12 4. Reduced Construction ............................................................................................................................ 12 5. Reduced Residential Mobility ................................................................................................................ 13 6. Reduced Attrition of Older Cohorts .................................................................................................... 14 7. Reduced Housing Availability ............................................................................................................... 15 8. Conclusion ............................................................................................................................................... 17 The Parts of Its Sum: Regional Housing Demand Pressures and Neighborhood Change 18 Chapter 2. 1. Introduction ............................................................................................................................................. 18 2. Research Methods ................................................................................................................................... 19 3. Regional Housing Market Conditions .................................................................................................. 21 3.1. Growing Demand for Housing ..................................................................................................... 21 3.2. Constrained Housing Supply ......................................................................................................... 27 3.3. Rising Housing Prices ..................................................................................................................... 27 4. Neighborhood Housing Market Conditions ....................................................................................... 28 4.1. Neighborhood Typology ................................................................................................................ 28 4.2. Growing Demand for Housing ..................................................................................................... 32 4.3. Constrained Housing Supply ......................................................................................................... 33 4.4. Rising Housing Prices Across the Region ................................................................................... 34 4.5. The Housing Stock Constrains and Enables Growth ................................................................ 35 iv 4.6. Housing Constraints and Neighborhood Migration by Race ................................................... 37 5. Conclusion ............................................................................................................................................... 39 Appendix A. Neighborhood Typology Cluster Analysis Technique ................................................... 41 Losing Their Place in Line: Diverging Neighborhood Outcomes for Young People ....... 45 Chapter 3. 1. Introduction ............................................................................................................................................. 45 2. Motivation and Background .................................................................................................................. 46 2.1. Millennials ......................................................................................................................................... 47 2.2. Economic Disparities and the Divergent Fortunes of Places ................................................... 48 3. Research Approach and Methods......................................................................................................... 49 3.1. Data ................................................................................................................................................... 49 3.2. Density Quintiles ............................................................................................................................. 50 4. Regional Changes in Young College Graduates and Non-Graduates ............................................. 51 4.1. Rising Income Inequality for Young People by Educational Attainment .............................. 51 4.2. Changing Populations of Young People by Educational Attainment ..................................... 52 5. Changes in the Neighborhoods Where Young College Graduates and Non-Graduates Live .... 53 5.1. Multivariate Analysis of Neighborhood-Level Increases and Decreases in Young College Graduates and Non-College Graduates .............................................................................................. 54 5.2. Dependent Variables: Tract Level Changes in Young Adults by Educational Attainment . 55 5.3. Neighborhood Characteristics ....................................................................................................... 57 5.4. Newly Built and Recently Vacated Housing Available on the Market .................................... 67 5.5. Housing Tenure ............................................................................................................................... 68 6. Policy Implications .................................................................................................................................. 71 7. Conclusion ............................................................................................................................................... 72 References ......................................................................................................................................................... 74 v Figures Figure 1.1 Home price and housing permit indices for the Los Angeles MSA compared with the Case-Shiller twenty-city composite and the U.S., 1980-2014 (2000 = 100) ............................................ 12 Figure 1.2 Housing construction by tract density quintile in the Los Angeles region ........................... 13 Figure 1.3 One-year residential mobility rates by tenure in the Los Angeles region ............................. 14 Figure 1.4 Los Angeles regional cohort growth and attrition rates in the 1990s and 2000s and additional households remaining in the region in 2010 because of lower attrition rates in the 2000s. Cohorts are identified by their age at the start of decade. ......................................................................... 15 Figure 1.5 Changes since 1990 in the number of housing units available on the market by tenure and the adult population in the Los Angeles region .......................................................................................... 16 Figure 1.6 Housing availability by tenure and density quintile in the Los Angeles region .................... 17 Figure 2.1 Population by age and race in the Los Angeles region in 2000, 2007, and 2012 ................. 22 Figure 2.2 Expected growth in adults through aging and mortality, actual growth, and net migration by age and race in the Los Angeles region from 2000 to 2012 ................................................................. 23 Figure 2.3 Growth due to natural increases, in-migration, and out-migration by race in the Los Angeles region during boom and bust .......................................................................................................... 26 Figure 2.4 Map of neighborhood typology in the Los Angeles metropolitan area, with inset showing the five-county region ..................................................................................................................................... 31 Figure 2.5 Expected adult population growth through aging and mortality in the Los Angeles region during boom and bust, 2000 to 2012 ............................................................................................................ 32 Figure 2.6 Increases in the housing stock by neighborhood type in the Los Angeles region during boom and bust .................................................................................................................................................. 33 Figure 2.7 House prices changes by neighborhood type in the Los Angeles region during boom and bust ..................................................................................................................................................................... 34 Figure 2.8 Actual growth in adults compared with expected growth and housing growth by neighborhood type in the Los Angeles region during boom and bust .................................................... 36 Figure 2.9 Housing supply response and migration rates of young adults by race for neighborhood types in the Los Angeles region, 2000 to 2012 ............................................................................................ 37 Figure 2.10 Correlation between neighborhood net migration rates and the housing supply response to expected demand ......................................................................................................................................... 38 vi Figure 3.1 Map of census tracts in the Los Angeles-Long Beach Combined Statistical Area by 1990 population density quintile ............................................................................................................................. 50 Figure 3.2 Change in population aged 25-34 and ratio of non-graduates’ incomes to college graduates’ personal incomes in the Los Angeles region ............................................................................. 51 Figure 3.3 Percent change in college graduates and non-college graduates aged 25-34 in the U.S., California, and Los Angeles region ............................................................................................................... 52 Figure 3.4 Changes since 1990 in the total population, households, renters, and owners of people aged 25-34 by educational attainment in the Los Angeles region............................................................. 53 Figure 3.5 Net changes in population aged 25-34 by educational attainment and density quintile in the Los Angeles region .................................................................................................................................... 54 Figure 3.6 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by percent college graduates interacted with tract homeownership rate ...................................................... 60 Figure 3.7 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by employment density ......................................................................................................................................... 61 Figure 3.8 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by percent foreign born ........................................................................................................................................ 64 Figure 3.9 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by recently vacated housing units interacted with tract homeownership rate .............................................. 68 Figure 3.10 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by recently built housing units interacted with tract homeownership rate ................................................... 69 Figure 3.11 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by homeownership rate ........................................................................................................................................ 70 vii Tables Table 1.1 Housing construction permits in the Los Angeles region, 1980-2014 ..................................... 9 Table 2.1 Immigration to the Los Angeles region and internal growth through the aging of existing residents in the 1990s and boom and bust periods ..................................................................................... 21 Table 2.2 Growth rates through natural increases and net migration by age and race in the Los Angeles region, 2000 to 2012 ......................................................................................................................... 25 Table 2.3 Percent of the regional population of each group living within each type of neighborhood in 1990 and 2012 .............................................................................................................................................. 30 Table 3.1 Descriptive Statistics and Correlations with Density ................................................................ 56 Table 3.2 Estimated Marginal Effects of Neighborhood Characteristics on Census Tract Changes in College Graduates and Non-Graduates Aged 25-34, 2007 to 2012 ......................................................... 58 viii Acknowledgments Each of my committee members has earned my deep gratitude for their longstanding commitment to my development as a scholar and person. Dowell Myers, your enthusiasm for research made the work fun, and your kindness kept me going when it wasn’t. Thank you for inspiring me to study the dynamics of change, and for showing me how to build an argument based on evidence made tangible through graphs and tables. Richard Green, thank you for your unfailing support throughout graduate school, from the course where I met you through the tail end of my dissertation. You have always asked for my best, and I could not ask for more. George Galster, thank you for engaging with my research and helping me to clarify my thinking. Our conversations are a joy to me, and I always welcome your perceptive insights into work, art, and life. Manuel Castells, thank you for sharing your keen understanding of the complex social world. You have pushed me to broaden my research, and I dearly value the advice and encouragement you have given me along the way. Jan Breidenbach, when we first met you showed me how I would like to teach, and since then you have taught me far more in the realms of research, teaching, and life. Thank you. You each gave your time generously, and I relied on you for professional and personal guidance and thoughtful feedback on my work. I thoroughly enjoy working with each of you, and I am honored to call you mentors and friends. Beyond my committee, the faculty and staff of the USC Price School and the broader university have supported me throughout graduate school. Gary Painter, Lisa Schweitzer, Liz Falletta, Kathy Kolnick, Roberto Suro, and Elizabeth Currid-Halkett all deserve my thanks for their mentorship. It has been a great pleasure to work with Jennifer Ailshire over the past few years. David Sloane, Tridib Banerjee, Marlon Boarnet, Debbie Natoli, and Nicole Esparza taught excellent courses that have shaped my understanding of planning, and Julie Kim helped me navigate through both the masters and PhD programs. Aubrey Hicks and Donna Jean Ward shared their creative and joyful approach to engaging with policy. My graduate education would not have been possible without Chris Wilson’s extraordinary work. Thank you for all you do for students and for the school, Chris. I am grateful to the Price School and USC for financial support: my PhD studies were supported by the USC Provost’s fellowship, and my dissertation research was funded by the Oakley Endowed Fellowship. My fellow PhD students are steadfast friends. Ray Calnan, Mi Young Kim, Jung Hyun Choi, Denise Payan, Seva Rodnyansky, Jennifer Candipan, Jovanna Rosen, Janna Goldberg, and Brettany Shannon, thank you for the rambling conversations, inspiration, reflection, and encouragement. My thanks to comrades further afield as well: Meagan Ehlenz, Kelly Kinahan, Mary Rocco, Simon Mosbah, and Whitney Airgood-Obrycki. Jake Wegmann, thank you for your generous invitation to collaborate when I had barely finished my coursework, and all the fun we have had since then puzzling through the intricacies of informal construction. Rolf Pendall, Katrin Anacker, and Ed Goetz, thank you for taking the time to offer me advice and mentorship from afar. My family and friends have sustained me throughout graduate school. To my parents, thank you for passing along your love of learning, and for raising me with the freedom to be curious, creative, and stubborn in pursuing the things I care about. Thank you for finding new ways to support me as I enter new phases of life. To my brothers Peter, John, and Ross, and sister Jane, thank you for sharing your lives with me. Your creativity and excitement about the world refresh and inspire me. James and Lisa, thank you for your constant love and support through all these years of study. From ix my first paper in my masters program strewn all over your office, to the last pages of my dissertation written at your dining room table, you have offered me your home as a writer’s retreat and a refuge from graduate student life. I cannot say how much I have enjoyed spending this time with you. Louise, Dorchen, and Bruce, thank you for cheering me on. To the community of St. John’s Cathedral, thank you for all the celebration and comfort you have offered me over the years. To my neighbors and dear friends Nicole Guillen, Vincent Selhorst- Jones, Anna Colby, and Evan Colby, thank you for all the walks, all the meals, all the games, all the laughter, and all your many kindnesses as I labored away. Hayley Tyler, Beth Kracum, Caitlin Goldbaum, Brandynn Martin, and Russell Graham thank you for listening, and for all our great conversations and adventures. Ellen Valkevich and Liesl McCormick, thank you for understanding exactly how difficult it is to write a dissertation, and for reminding me how much fun it is to write a dissertation. Matt Young, thank you for telling me that I could do it - until I could. You enabled me to write this dissertation with your persistent belief in me, your thoughtfulness, your patience, and your time and energy. When I needed to think about only one thing, you thought about everything else. Thank you for your partnership in life. 1 Introduction Why is it so hard to find an affordable place to live in Los Angeles? We know the basic answers to this question: many people want to live in the Los Angeles area and there is not enough housing to go around, a straightforward case of a mismatch between demand and supply. But the reasons that so many people continue to live in Los Angeles (in the face of high and rising prices), and the reasons that not enough housing is built (despite such high prices) are more complex. And the impacts of the housing shortfall are uneven for different groups of people and different neighborhoods within the region. In this dissertation, I set out to lay an empirical foundation for understanding why the housing stock in the Los Angeles region 1 has been constrained since the building boom of the 1980s, and why regional demand for housing has continued to grow over the period of the housing boom and bust. I measure the declines in number of housing units available on the market, arising from both reduced construction and reduced turnover. I also examine the sources of growth in potential demand for housing, from migration and from the natural increase of the existing population. I then compare the consequences of the housing shortfall for different types of neighborhoods and for different groups of people by age, race, and education level. As part of the investigation, I uncover evidence that sheds light on the local processes of gentrification, immigrant suburbanization, and the Latinization of historically Black neighborhoods, and how these changes in neighborhoods are related to the regional housing shortfall. Yet even with these changes, the overall socioeconomic hierarchy of neighborhoods within the region remained largely unchanged. 1 The Los Angeles region is defined throughout this dissertation as the five-county Los Angeles- Long Beach, California Combined Statistical Area, or CSA, encompassing Los Angeles County, Orange County, Riverside County, San Bernardino County, and Ventura County. 2 Findings I find that the number of vacant housing units available on the market fell 12.3 percent from 1990 to 2012, while the regional adult population grew 29.2 percent. The drop in the vacancies is not just the result of reduced construction, but also reduced turnover. The large majority of vacancies are the result of people moving out of existing housing units. Even in 1990, with relatively robust housing construction, newly built housing accounted for only 23.7 percent of units for sale on the market and 6.7 percent of units for rent. Mobility rates for homeowners fell from 10.4 percent per year in 1990 to 5.2 percent per year in 2012, and mobility rates for renters fell from 34.5 percent in 1990 to 26.3 percent in 2012. Reduced turnover led to an overall 2.5 reduction in the number of vacated units available on the market. With tenure conversions of owner-occupied housing to rentals, vacated units for sale decreased by 16.3 percent from 1990 to 2012, and vacated units for rent increased by a slight 1.6 percent. The reductions in mobility are not simply a matter of slowing down the pace of turnover; they are the result of people occupying housing units for longer periods, leaving fewer openings for newcomers to establish households. At the same time, construction fell even more dramatically. Regional construction fell from an average of 96 thousand units per year during the 1980s to 40 thousand during the 1990s, back up to 69 thousand during the housing boom from 2000 to 2007, and finally down to 25 thousand after the boom from 2008 to 2014. This led to a 90.8 percent decline in the number of newly built units on the market from 1990 to 2012. Taken together, declines in turnover, construction, and tenure conversions led to a 30.9 percent reduction in the number of units available for sale and a 5.4 percent reduction in the number of units available for rent, during a time of regional growth in the demand for housing. The growth of housing demand stems primarily from population growth, often thought of as in- migration of new residents to the region. Yet I find that most of the growth in housing demand over the boom and bust periods was simply the result of the natural increases from people who already lived in the region - settled immigrants and their children, baby boomers and their children, and others - rather than from immigration or migration to the region, which was the major source in previous decades. During the housing boom from 2000 to 2007, the adult population only grew by 0.8 percent through migration, but it grew 9.3 percent from natural increases as a large generation of young people grew into adulthood and older people lived longer. During the bust from 2007 to 2012, the adult population grew only 0.8 percent through migration, but it grew 6.5 percent through natural increases. This internal growth in demand from existing residents explains why so many people continue to live in the Los Angeles region even when it is difficult to afford housing. Los Angeles is already their home. For migrants considering coming to Los Angeles, expensive housing may prompt them to seek another place to move. But for people who grew up in the Los Angeles region or have established lives, relationships, and jobs, moving away is a much more difficult prospect. Demand growth through natural increases is also less obvious than migration: this is a pervasive, invisible pressure throughout the housing market. In the context of the housing shortfall, with far fewer housing units available on the market and with steadily growing demand, I find stark disparities in outcomes for more and less advantaged groups. Younger adults entering the housing market were more affected by the housing shortfall than older 3 adults who had already established households. Among younger adults, Hispanics and Blacks were more strongly affected by the housing shortfall than Whites and Asians. Hispanic migration into and out of neighborhoods was highly correlated with the intensity of the neighborhood housing shortfall. They moved away from neighborhoods with intense demand pressure and little housing construction, and they moved into neighborhoods where lots of housing was built even though most of the housing was built in suburban and exurban areas far from urban amenities and employment opportunities. In contrast, White young adults were able to afford to move into desirable urban neighborhoods where the housing supply was especially constrained. Their neighborhood migration rates were not at all correlated with neighborhood housing shortfalls. And White in-migration coincided with higher levels of out-migration for Hispanic young adults, a replacement that is evidence of gentrification. Black young adults migrated out of the Los Angeles region overall, losing 10.1 percent to net out-migration, and they declined the most in neighborhoods with the most severe housing shortfalls. The one type of neighborhood where Black young adults increased substantially was in the far suburban reaches of the region. Asian young adults had the highest levels of in-migration to the region, and like Whites, they were able to move into the more desirable, more housing- constrained neighborhoods in addition to the exurban neighborhoods where most of the housing was being built. Finally, I find disparities in the neighborhood attainment of young college graduates and young adults without college degrees during the period since the housing crisis. Young college graduates increased in the region, though less than in the state and the nation. Young college graduates increased the most in neighborhoods that already had more college graduates, better access to employment, lower vacancy rates, more public transit access, and less subprime lending. On the other hand, young people without college degrees declined overall, partially as a result of higher college graduation rates, partially as a result of declining immigration to the region, and partially as a result of out-migration. They declined most in the same high quality neighborhoods where young college graduates increased, but they were able to remain in the places that had lost the most jobs during the economic restructuring of the 1990s, the places with the most subprime lending during the boom, and the places hit the hardest by the housing bust. Both newly built housing and recently vacated existing units turned over by previous occupants made a difference in the places where young people could move. Housing construction mattered more for college graduates, and vacancies mattered more for non-college graduates, but turnover was the largest source of housing for both groups. Housing tenure played an important role in the places where young people moved: young college graduates increased the most in rental neighborhoods and places where more rental housing was built, while young non-graduates remained at higher rates in mixed tenure neighborhoods and places where more housing was built for sale. This tenure difference reflects two opposing trends in the aftermath of the housing crisis. Young non-college graduates continued to increase in the exurban areas where investors bought up foreclosed homes and rented them out. And young college graduates flooded into rental housing in urban neighborhoods as mortgage access was pulled back after the bust and it became increasingly difficult for them to purchase homes. The disparities between groups make it seem as though there are clear winners and losers in the regional housing shortfall - old win out over young, White over Black, Asian over Hispanic 4 immigrants, and college graduates over those without college degrees. And the housing shortfall does seem to heighten the disparities between relatively advantaged and less advantaged groups. In truth, all are worse off, whether they are moving away from their home because they cannot afford their neighborhood (or region) anymore, or paying more and more for housing. Only investors in real estate win in this scenario. That includes landlords, longtime homeowners (as long as Proposition 13 protects them from rising property taxes), and capital investors, who profited from home buying during the boom and then acquired homes to rent out during the bust as housing values corrected but rents continued to rise. And the long-term social and economic health of the region suffers, as neighborhood social fabrics are fragmented when people move away, and as Los Angeles grows less able to attract the talented newcomers who have built this varied and vibrant metropolis over the past decades. Background This dissertation is grounded in filtering theory as a framework for understanding the role of the regional population and the housing stock in neighborhood change. As originally formulated in the 1950s and 60s, filtering theory describes how new housing was built in the suburbs for relatively high income households, who then moved out and left their previous urban housing available for lower income households to move in (Lowry 1960; Olsen 1969). In this way, the construction of new housing allowed relatively high quality housing to “filter down” and become more affordable. The concept of filtering has been most commonly used to describe this specific phenomenon of downward filtering. And downward filtering was happening when filtering theory was first developed. Yet in recent years the filtering process has led to rising prices, as demand for housing in older urban neighborhoods has been revived (Brueckner and Rosenthal 2009). Filtering theory has been enriched by scholars working to understand the dynamic relationship of the population and the housing stock over time (Myers 1983; Grigsby et al. 1987; Baer and Williamson 1988; Rothenberg et al. 1991; Galster and Rothenberg 1991; Galster 1996). They have developed a set of concepts for understanding how the housing market allocates housing through competition within submarkets, and how neighborhood housing prices change over time. Durable Housing. One key insight of filtering theory is that housing is durable: once built, it lasts long beyond its initial occupants. This means that over time, many different households will live in the same housing unit over its lifespan. Housing usually deteriorates over time until it is demolished to make way for new housing construction (Corgel and Smith 1981), but can also be maintained or renovated (Galster 1987). Housing Submarkets. The second major concept is based on the idea that there are different types of housing (such as rental or for sale housing, and single family or multifamily housing, and different numbers of bedrooms), and different levels of housing quality. Housing is grouped into submarkets of similar type, arranged in a loose hierarchy by quality within the regional housing market. Housing units of similar type and quality are often (but not always) spatially clustered together in neighborhoods, especially since neighborhood characteristics determine much of the desirability of housing (Bostic, Longhofer, and Redfearn 2007). 5 Housing Demand. Just as the housing stock has a variety of characteristics and quality levels, different people prefer different types of housing, and are willing and able to pay for different levels of housing quality. Housing preferences are influenced by factors such as age and lifestyle, family size, children, and a host of other personal factors. The ability to pay depends mainly on income and wealth, though people may choose to spend more or less of their income on housing. Often people with similar socioeconomic and demographic characteristics have similar preferences for housing, and similar ability to pay. While housing generally deteriorates over time, people generally increase their ability and willingness to pay for housing as they age, and prefer different types of housing as they move through different stages in their lives (Myers 1983; Clapham 2005). Matching of the Population and the Housing Stock. Housing is allocated via prices, and people compete for housing on the market according to their preferences and relative ability and willingness to pay. This is a constant process that unfolds over time, as housing is added and removed and preferences shift and change. At any one time, most of the housing stock was built in previous decades to meet slightly different needs, and it takes time for developers to build new housing in response to changing demand. Mismatches between housing supply and demand often arise because housing is durable and slow to change but the population is constantly changing. Competition and Prices. Prices in each submarket rise or fall depending on the supply and demand for housing of that type and quality. When demand in a submarket falls, prices fall and housing becomes more affordable for a lower income group to move in, as in the classic case of higher income households leaving older neighborhoods for new homes in the suburbs. When demand for a specific type of housing increases, housing developers must scramble to adjust. When supply in that submarket does not quickly meet demand, some people are priced out and seek housing in a different submarket, which in turn impacts the level of housing demand and prices in the downstream submarket. In this way, changes in one submarket also affect other similar submarkets. This theory describes processes of change over time rather than one specific outcome, with a recognition that the filtering process does not always lead to the production of affordable housing. The idea that housing filters downward and becomes more affordable over time rests on the assumption that enough housing is built to satisfy growing demand. When less new housing is built, growth in demand leads to increasing competition for existing housing units and rising prices. Demand for housing in different submarkets changes over time as different segments of the population move through their life cycles, with different housing needs and preferences at different stages of life (Clapham 2005). On the whole, each group occupies housing for a certain period of time, then as their incomes rise and preferences change they move on to a different submarket, which was vacated by the previous group who lived there. If the previous group remains longer in that submarket, they delay the release of their homes onto the market for the next occupant, the number of households competing for housing in that submarket increases, and prices rise. If a 6 generation is larger than the previous generation, competition for the types of housing they prefer at each stage of their lives will increase, as demonstrated by Campbell (1966) and Myers and Pitkin (2009). Depending on the incomes of the growing groups, and the type of housing they prefer, competition will rise and prices increase in certain submarkets (or neighborhoods) more than others. But as others are priced out of those neighborhoods, they will move into other neighborhoods and increase demand pressure and prices there. In cases with strong demand growth where the housing supply is especially constrained, the lowest income people may be priced out of the housing market, unable to form households or moving out of the region. The Los Angeles region is an ideal setting to observe the filtering process and neighborhood change in the context of a severely constrained housing supply and rapidly growing demand. Aging immigrants and baby boomers, and the children of immigrants and baby boomers, have contributed to strong growth in adults and transformed the demographic landscape (Myers et al. 2010). These key segments of the population have buoyed demand for aging housing in dense urban neighborhoods. In a reversal, higher income households are out-competing lower income households in newly desirable urban neighborhoods. House prices and rents have increased rapidly, with effects rippling through local labor markets, transportation systems, and development patterns. And many lower income people have been priced out of the housing market entirely. Contributions In this dissertation, I build on filtering theory with new ideas about how to conceptualize and measure changes in housing supply and demand, and I introduce new techniques to analyze the connections between regional shifts in supply and demand and local changes in neighborhoods. By combining well-established filtering theory with innovative demographic methods, I find empirical evidence of the powerful, usually hidden forces of the regional housing market and the way they impact various types of neighborhoods and groups of people. I make two main points about the housing market. First, the housing supply that matters for growth is the housing available on the market, and this includes both newly built units and existing units that have been turned over by their previous occupants. Second, the underlying sources of growth in housing demand include (1) natural increases in the adult population through larger generations and longer life expectancies, in addition to the more commonly-studied (2) migration, (3) household formation, and (4) changes in income. To understand the causes of the housing shortfall, planners must pay attention to both sources of housing available on the market, and all four sources of demand growth. Because filtering does not seem to “work” to produce affordable housing, many of its concepts have been set aside by planners. This detailed analysis of the way filtering works in practice offers planners an understanding of the regional context for the changes they see in local neighborhoods. This can help planners working in local cities or neighborhoods understand how what happens in another neighborhood might affect theirs, or how development in their area is related to other parts of the region. 7 Beyond its relevance for planners, this housing market-based perspective is also highly pertinent to the academic literature on neighborhood change. Neighborhood change happens through the medium of the housing market, as people find places to live based on the housing options available on the market, their preferences for housing and neighborhoods, and their budget constraints. The housing market is key to translating disparities in parents’ incomes to disparities in their children’s futures through unequal access to schools and employment opportunities, as highlighted by Owens (2010) and more recently Sharkey (2013). As this dissertation shows, housing shortfalls can amplify the disparities between groups, and filtering theory can help explain why segregation is so persistent, relevant to the work of Crowder, South, Sampson, and other sociologists studying residential mobility and neighborhood change (Crowder, South, and Chavez 2006; Sampson 2009; Sampson 2011; Hwang and Sampson 2014). Local mismatches between supply and demand also help explain large influxes of a new racial or ethnic group in a neighborhood (Crowder, Hall, and Tolnay 2011; Sharkey 2012; Bader and Warkentien 2016). Many neighborhood change scholars include some consideration of the housing market in their work. Yet the context of the regional housing market is often absent in studies of neighborhood-level changes, even though this context is often crucial for understanding the sources of change and mobility. Policy Implications This research does not lead to recommendations for specific policies to solve the housing shortfall, but it contributes crucial understanding of the colliding urban processes that have led to the severe housing shortfall in the Los Angeles region, and it reveals the connections between regional housing market dynamics and changes in local neighborhoods and specific groups of people. This research underscores the deep consequences of the housing shortfall for certain people and places and for the region as a whole, adding to the urgent calls for planners, policymakers, and residents to respond in support of action to relieve the affordability crisis. The public conversation around housing affordability and neighborhood change has long coalesced around two main positions: the viewpoint that we simply need to build more housing versus calls to protect and preserve neighborhoods and vulnerable residents. Recently more nuanced voices have gained ground, calling for construction to address the housing shortfall coupled with a recognition of the need to do more than build to protect those who suffer most from the shortfall (Jacobus 2016; Badger 2016; Chiang 2016). There has also been a renewed public acknowledgment that opposition to development in order to protect neighborhoods from change often exacerbates inequalities of access to housing and neighborhoods rather than preserving affordable housing for those who need it most (Phillips 2016; Dougherty 2016). This research supports the idea that while building more housing is the long-term solution to the housing affordability crisis, the shortfall is driven by more than simply a lack of construction, and the responses must involve more than just construction. Building enough housing to alleviate the shortfall is unlikely to happen anytime soon given the scale of the shortfall and the physical, political, and economic obstacles to development. In the meantime, planners must pursue additional strategies to protect vulnerable populations whose access to housing and neighborhoods is threatened by the housing shortfall and rising prices. 8 The housing shortfall is based on longer occupancy and less mobility as well as reduced construction. Proposition 13 incentivizes homeowners to remain in their homes longer than they otherwise might. One provision of Proposition 13 works by limiting property tax increases so that no matter how quickly prices rise, the assessed value for tax purposes can only increase by a maximum of 2 percent per year after purchase. Homeowners who remain longer in their homes pay less property tax relative to the value of their home, as long as prices are rising (Myers 2009). If they were to purchase a different home, they would have to pay property taxes in line with current housing prices again, almost always higher than before. If Proposition 13 were revisited, it would make better financial sense for aging homeowners to relocate and downsize, leaving their homes in prime locations available for younger people to move in. This research also suggests that it matters a great deal where housing is built in the region, and whether those places are high quality neighborhoods with access to employment, good schools, and other amenities and public services. Housing construction in exurban areas may allow some people to remain in the region who would otherwise be priced out, but it does not lessen the housing shortfall in more desirable urban and suburban neighborhoods. Since not enough housing can be built in high quality neighborhoods, planners can invest in infrastructure, schools, and transportation in disinvested neighborhoods throughout the region so that even the least desirable neighborhoods become safe and supportive places to live. Overview This dissertation consists of three essays, addressing housing supply constraints, demand growth, and changes in the neighborhoods where different groups of people live in the Los Angeles region from 1990 onward, with a focus on the housing boom of the early 2000s and the subsequent bust period. In the first essay, “No Vacancy: Reduced Housing Availability in the Los Angeles Region,” I examine reductions in the amount of housing available on the market as a result of both decreasing construction and declining turnover. In the second essay, “The Parts of Its Sum: Regional Housing Demand Pressures and Neighborhood Change,” I explore the sources of growth in housing demand, and analyze the neighborhood-level migration of different groups within the region in relation to the supply constraints in each neighborhood. In the third essay, “Losing Their Place in Line: Diverging Neighborhood Outcomes for Young People,” I take up the question of how the housing shortfall impacts the neighborhoods where young college graduates and non-graduates live within the region. 9 No Vacancy: Chapter 1. Reduced Housing Availability in the Los Angeles Region 1. Introduction Suburban growth has been the main source of affordable housing in the Los Angeles area since World War II — for residents of both suburban and urban neighborhoods. As new housing was built in the suburbs, middle and upper income households migrated away from older urban neighborhoods to the suburbs. Demand declined in urban neighborhoods, and prices fell. In this way, high quality older housing became affordable for low income households (Baer and Williamson 1988; Galster and Rothenberg 1991). Yet housing construction in the Los Angeles region slowed dramatically after the 1980s, as shown in Table 1.1. With the sharp regional economic downturn in the early 1990s, permitted housing units fell to 208 thousand over five years, less than a third of the number permitted in the previous five years. Construction slowed even further in the late 1990s, to 196 thousand. Housing development picked up during the housing boom of the early 2000s, to 342 thousand, but regional permits were still only slightly more than half of what they had been during the late 1980s. Table 1.1 Housing construction permits in the Los Angeles region, 1980-2014 Data Source: U.S. Census Bureau SOCDS, Texas A&M. Even with strong demand for housing and ample financing for construction and mortgages, housing construction was limited by land availability. As a practical matter, less open land was available for new construction in the suburbs, at least within easy striking distance of employment centers. The ocean and surrounding mountains limit outward expansion, except for in far-flung exurban areas. 1980-84 1985-89 1990-94 1995-99 2000-04 2005-09 2010-14 Single Family Permits 157,981 320,206 136,961 150,266 242,832 159,105 57,441 Multifamily Permits 148,667 333,446 70,985 45,251 98,971 88,042 77,369 Total Permits 306,648 653,652 207,946 195,517 341,803 247,147 134,810 10 And in existing urban and suburban neighborhoods, developers faced the strong local opposition to growth and restrictive regulations that make infill and redevelopment projects so difficult and expensive to build (Green, Malpezzi, and Mayo 2005). Financing for new construction pulled back after the housing crisis, and by the 2010-14 period construction permits fell to 135 thousand, only a fifth of the level of construction during the late 1980s. With the land constraints in the suburbs and an upswing in renting in the aftermath of the housing crisis, development patterns began to shift away from suburbanization toward urban infill and redevelopment (Jackson 1985), but at a much lower rate of construction than before. By 2012, there were only 9.2 percent as many newly built housing units on the market than there had been in 1990. Newly built housing only accounts for a small portion of the vacant housing units available on the market in any given year. Even with relatively robust construction, in 1990 only 23.7 percent of the units for sale and 6.7 percent of the units for rent were newly built. The rest of the units on the market were in older buildings with prior occupants moving out. However, by 2012, turnover rates in Los Angeles housing were much lower than in 1990. Mobility rates for homeowners were cut in half, falling from 10.4 percent per year in 1990 to 5.2 percent per year in 2012, and mobility rates for renters fell from 34.5 percent in 1990 to 26.3 percent in 2012. With some housing units converted from owner-occupied to rental units, this led to a 16.3 percent reduction from 1990 to 2012 in vacated units for sale, and an 1.6 percent increase in vacated units for rent. Reduced mobility is not simply a matter of slowing down the pace of turnover; it is a result of people occupying housing units for longer periods, leaving fewer openings and less housing available for others to move in. As a result of falling mobility, turnover supplied fewer units onto the housing market. On top of declining construction, which reduced the number of units on the market by 10.1 percent between 1990 and 2012, lower mobility rates reduced the units on the market by an additional 2.2 percent. When falling construction and reduced mobility are considered together, vacant units on the market declined by 12.3 percent between 1990 and 2012, during a period when the regional adult population grew 29.2 percent. In this chapter, I appraise the housing supply and demand trends in the Los Angeles region that have reduced the amount of housing available on the market since 1990. I analyze changes in the number of vacancies for the region as a whole, by tenure, and in both dense and sparsely populated parts of the region. The declines in the number of housing units made available by new construction and turnover of the existing stock are particularly dramatic when compared with the growth in demand over the boom and bust periods from 2000 onward, detailed in the following chapter. For every 100 expected additional households, only 69 housing units were added during the boom, and only 52 units were added during the bust. The setting for this research is the Los Angeles-Long Beach Combined Statistical Area (CSA), which encompasses 33,955 square miles stretching from the Pacific Ocean to the Arizona and Nevada border, and includes Los Angeles, Orange, Riverside, San Bernardino, and Ventura counties. The Los Angeles region serves as a relevant example of the changes in population and housing that have led to affordable housing shortfalls in similar urban areas with growing populations and constrained housing development. Los Angeles’ housing supply is highly constrained, and it has high and growing demand for housing. Los Angeles is also a large region with a great deal of variation in its population, housing stock, and types of neighborhoods. In Yin’s (2008) framework for case study 11 selection, Los Angeles offers a critical and extreme case for the study of the consequences of a housing shortfall. Because these trends began earlier in Los Angeles than most areas in the U.S., it may hold clues about future development patterns in other cities. 2. Housing Supply Constraints The geography, existing built environment, and regulations of the Los Angeles region constrain new housing construction. The Pacific Ocean on one side and the mountains on the other are natural barriers to urban growth. In his study of the relationship between geographical constraints and housing supply elasticities, Saiz (2010) measured the percentage of the area within a 50-km radius of the central city that is undevelopable because of either steep slopes or water features. Of the 95 metropolitan statistical areas (MSAs) in the sample, Ventura ranks first with 79.6 percent undevelopable land, Los Angeles-Long Beach ranks fourteenth with 52.5 percent undevelopable land, and Riverside-San Bernardino ranks twenty-fifth with 37.9 percent undevelopable land. The relative density of the existing built environment in Los Angeles’ developable areas also makes it more difficult and expensive to find land to build new housing. Despite Los Angeles’ reputation for sprawl, Galster et al. (2001) found that the Los Angeles urbanized area had relatively high housing density, second only to New York among 13 major urban areas in the country (even without taking undevelopable land into account). Similarly, Malpezzi and Guo (2001) defined and tested various measures of urban density in 35 large U.S. metro areas. Los Angeles ranks third behind New York and San Francisco on their preferred metric, the density of the census tract containing the median person in the metro area. The regulatory environment in the Los Angeles region is generally restrictive to new housing construction. All three MSAs in the Los Angeles region have comparatively high scores on the Wharton Regulation Index developed by Gyourko, Saiz, and Summers (2008) and adapted to MSAs by Saiz (2010). Pendall, Puentes, and Martin (2006) found that cities in the Los Angeles area often use growth management strategies such as infrastructure fees and permit caps to restrict development despite relatively permissive zoning ordinances and strong affordable housing programs. All three factors – geography, density, and regulations – make it difficult and expensive to build housing in the Los Angeles area, especially in the most desirable parts of the region. Taking geographic and regulatory factors into account, Saiz (2010) estimated housing supply elasticities for the 95 MSAs in his sample. Housing supply elasticities measure the amount of new construction relative to price changes in an area, in this case measuring the response of housing suppliers to increasing demand. When prices rise and developers build accordingly, the housing supply is considered to be elastic; when prices rise and little new construction follows, the housing supply is inelastic. In Saiz’s estimates, the housing supply in Los Angeles region is fairly inelastic: Los Angeles- Long Beach’s elasticity of 0.63 is the second-lowest in his sample, Ventura’s elasticity of 0.75 is the eighth lowest, and Riverside-San Bernardino’s elasticity of 0.94 is the eighteenth. These figures also show that the housing supply is more constrained in some parts of the region than others; construction is more constrained in the central urban areas and mountainous northern areas and less constrained in the wide suburban valleys to the east. Green, Malpezzi, and Mayo (2005) also found that Los Angeles already had a relatively low supply elasticity during the period of 1979 through 12 1996. In replicating their elasticity estimates for the Los Angeles combined statistical area with more recent data, I find that the housing supply was roughly 45 percent as responsive to changes in prices during the 1996 to 2013 period as it was between 1980 and 1996. 3. Measuring Housing Availability Though the housing supply is typically measured as the total number of housing units in the region, only a small portion of those housing units are available on the market in any given year. I identify housing units available on the market within the past year using IPUMS microdata for the 1990 and 2000 Census, 2005-07, 2008-10, and 2011-13 ACS. I define housing available on the market as units that were (a) built during the past year or (b) currently vacant, recently rented, or recently sold, or (c) occupied by a household who moved in during the past year, assuming that before they moved in the unit was on the market. This method results in a slight undercount of housing available on the market, since some units may be turned over multiple times in a single year. To account for the difference between the 15-month period (January to the following March) used in the Census residential mobility questions and the 12-month period used in the ACS residential mobility questions, I adjust the 15-month Census figures downward to match the 12-month ACS figures using SIPP data on the seasonality of moves since residential mobility is lower during the January- March period. 4. Reduced Construction Los Angeles’ relatively weak supply response to rising prices during the 2000s is illustrated in Figure 1.1, which compares house price increases and construction permits for the Los Angeles MSA, a national composite of twenty cities, and the nation. Both the S&P/Case-Shiller Home Price Index and the annual housing permit totals are indexed to equal 100 in the year 2000. Figure 1.1 Home price and housing permit indices for the Los Angeles MSA compared with the Case-Shiller twenty-city composite and the U.S., 1980-2014 (2000 = 100) Data Sources: FRED, Federal Reserve Bank of St. Louis; US Census Bureau SOCDS, Texas A&M. 13 Although house prices in Los Angeles increased faster during the boom years of the early 2000s than the twenty-city composite and the nation, construction increases lagged slightly behind. The real building boom in the Los Angeles region happened during the mid- to late-1980s, when an average of 96 thousand units were built per year. During the 1990s, an average of only 40 thousand units were built per year. Construction picked up to 69 thousand units per year during the boom from 2000 to 2007, and then fell again to merely 25 thousand units per year after the housing crisis. As a result, 90.8 percent fewer newly built units were available on the market in 2012 than there had been in 1990. As shown in Figure 1.2, housing construction declined the most in the densest parts of the region after the building boom of the 1980s, and housing construction continued to decline everywhere but the most sparsely populated quintile in the 2000s, despite the housing price boom. Figure 1.2 Housing construction by tract density quintile in the Los Angeles region Data Sources: Census 1990 and 2000; ACS 2008-12. 5. Reduced Residential Mobility As housing construction declined, residential mobility rates for both owners and renters also fell during this period, as shown in the left-hand graph in Figure 1.3. The one-year mobility rate for renters was 34.5 percent in 1990, and fell to 26.3 percent by 2012. Likewise, the one-year mobility rate for owners fell from 10.4 percent to 5.2 percent from 1990 to 2012. This mobility rate measures turnover in housing units, and as such it reflects directly on supply availability. The fewer movers in a given year, the fewer existing units are released into the housing market and become available for newcomers to move in. As a result of reduced mobility, combined with tenure conversions from owning to renting, the number of existing units that were vacated and put on the market for sale fell 16.3 percent between 1990 and 2012, while the number of existing units that were vacated and put on the market for rent rose slightly by 1.6 percent. 14 Figure 1.3 One-year residential mobility rates by tenure in the Los Angeles region Data Sources: IPUMS Census 1990 and 2000; ACS 2005-07, 2008-10, 2011-13. 6. Reduced Attrition of Older Cohorts The declines in residential mobility were in part the result of older generations remaining longer in the Los Angeles region and occupying housing units for longer periods, measured by cohort growth or attrition rates. Cohort growth or attrition rates follow groups of people as they age, and are based on the net change between the number of people at a certain age at the beginning of the decade and the number of people who are ten years older ten years later(Myers 1992; Myers 2011). For example, this is calculated as the difference between the population aged 15-24 in 2000 and the population aged 25-34 in 2010. If people in that cohort migrated away from the region during that time, there would be a net decrease over the decade, and vice versa. This is a net change, so there is likely more in-migration and out-migration which somewhat balance each other out. And cohort growth or attrition rates do not include any estimate of movement within the region, so they do not represent the amount of housing available on the market as a result of their moves. However, they are a good indicator of increasing or decreasing demand pressure in the rental and for sale markets from each age group. As shown in the left of Figure 1.4, cohort attrition rates were reduced for all older cohorts among both owners and renters, due to both longer lifespans and reduced out-mobility from the region. For example, this means that the cohort of owners aged 55-64 in 1990 declined by 16 percent from 1990 to 2000, but owners aged 55-64 in 2000 only declined by 10 percent from 2000 to 2010. As a result, by the end of the 2000s, 29 thousand more owners aged 65-74 remained in the Los Angeles region than there would have been if that cohort had declined at 1990s attrition rates, as illustrated in the right of Figure 1.4. 15 Figure 1.4 Los Angeles regional cohort growth and attrition rates in the 1990s and 2000s and additional households remaining in the region in 2010 because of lower attrition rates in the 2000s. Cohorts are identified by their age at the start of decade. Data Sources: Census 1990, 2000, and 2010. When differences in attrition for all the cohorts aged 25-34 and over are added together, 158 thousand more owners remained, and 176 thousand more renters remained than if 1990s attrition rates had held steady. 71 percent of the regional growth in owners and 78 percent of the growth in renters during the 2000s can be attributed to declining cohort attrition rates. The significance of these findings is that when so much growth in households is the result of older cohorts simply remaining longer in the region or longer in the housing market, it becomes particularly difficult for young people and other newcomers to break into the housing market. Whereas we usually look for signs of an influx from outside the region to explain growing demand, this reveals the equally powerful effects of prolonged residence. 7. Reduced Housing Availability Both the reduction in new housing construction and the decline in residential turnover meant that considerably fewer housing units were available on the market in 2012 than in 1990. In addition, tenure conversions also affected the number of units for rent and for sale. Figure 1.5 shows the change in the total number of housing units available on the market from 1990 onward as well as the change in units for sale and for rent. During the 1990s, there was a dramatic 19.1 percent reduction in the number of rental units available on the market, while the number of units for sale declined slightly by 3.5 percent. From 2000 to 2006, there was an overall 3.8 percent increase in the number of units available on the market, though not enough to make up for the steep 14.9 percent overall decline of the 1990s. As some vacated units were converted from owner-occupied housing to rental units, the number of units for rent increased 5.6 percent while the number of units for sale declined 0.4 percent. After the housing crisis, the number of units for sale took a nose dive, declining 20.1 percent from 2006 to 2009 and another 10.1 percent from 2009 to 2012, for an overall 30.9 percent 16 decrease in the number of units for sale on the market from 1990 to 2012. Tenure conversions accelerated in the aftermath of the housing crisis, and the number of rental units on the market increased by 11 percent from 2006 to 2009, then leveled off and declined 1.0 percent from 2009 to 2012. During the whole period from 1990 to 2012 there was a 30.9 percent drop in the number of housing units available for sale, and a 5.4 percent drop in the number of housing units available for rent. Even significant tenure conversions from owner-occupied units to rentals did not bring the number of units for rent back up to 1990 levels. Figure 1.5 Changes since 1990 in the number of housing units available on the market by tenure and the adult population in the Los Angeles region Data Sources: IPUMS Census 1990 and 2000; ACS 2005-07, 2008-10, 2011-13. The location of available units within the region shifted as well, as shown in Figure 1.6. With the increases in demand over the 2000s, the only places where the overall number of units available increased were the least dense tracts. In all the other density quintiles, the number of units for sale decreased dramatically over the 2000s. Fewer for-sale units were built, rapidly rising prices and then rapidly falling prices led fewer people to vacate their homes. Foreclosed units were taken off the market, and previously owner-occupied homes were purchased by investors and converted to rentals (Myers et al. 2016). Note that very little housing is available for sale in the densest quintiles. Aspiring homebuyers face limited options and stiff competition in those places, and may need to move to the suburbs in order to buy a home regardless of their neighborhood preferences. Rental housing turns over more often, and is more evenly distributed across the region. But while rentals held even or increased slightly in many parts of the region, rental demand was climbing as a larger generation of millennials entered the housing market (discussed in Chapters 2 and 3), foreclosures, and reduced access to mortgage financing after the housing crisis. 17 Figure 1.6 Housing availability by tenure and density quintile in the Los Angeles region Data Sources: Census 1990 and 2000; ACS 2008-12. 8. Conclusion The steep declines in the amount of housing available on the market, while the adult population grew steadily and housing prices rose, reveal a severe housing shortfall in the Los Angeles region. Conversions of previously owner-occupied housing to rentals meant that the number of units for sale declined especially sharply by 30.9 percent, while rental housing availability declined more steeply during the 1990s and recovered somewhat during the 2000s, for an overall decrease of 5.4 percent. The only places where housing construction and overall housing availability grew were in the least dense exurban parts of the region. The housing shortfall was the result of both declines in residential mobility and the abrupt drop in housing construction after the 1980s, which was only somewhat reversed during the housing boom period of the early 2000s, and which tumbled even further after the housing crisis. The declines in residential mobility were the result of older cohorts remaining longer in their homes, so they were not released onto the market for occupancy by other would-be movers. Measuring the amount of housing available on the market presents a more precise picture of the housing supply than simply measuring the aggregate housing number of housing units, since it accounts for both housing construction and the turnover of existing units. This view reflects the difficult situation faced by millennials trying to enter the housing market in 2012 as compared with their parents in 1990 over 20 years earlier in 1990. The pressures of the housing market had intensified, and many young people were not able to establish households - some continued to live with their parents, some doubled up, and some moved away from the Los Angeles region entirely. In this environment of scarcity and competition, only advantaged young people and newcomers were able to survive and thrive. 18 The Parts of Its Sum: Chapter 2. Regional Housing Demand Pressures and Neighborhood Change 1. Introduction In neighborhoods across the Los Angeles region, people are encountering rapid change. Starting in the late 1990s, new groups began to show up in neighborhoods that had been relatively stable for decades: young White hipsters in urban immigrant neighborhoods, middle-aged Hispanic immigrants in historically Black neighborhoods, and Asian immigrants in traditionally White suburbs. Other groups suddenly seemed to go missing: where were the newly-arrived Hispanic immigrants, and where were the Black young adults? And everyone paid more and more for housing, whether buying a home or renting, whether living in Mid-City Los Angeles or the farthest reaches of the San Gabriel Valley. This last point may seem out of place, but it is highly relevant: each of the demographic shifts in different types of neighborhoods can be traced back to the deep regional housing shortfall that makes it so very difficult to find an affordable place to live. In this chapter, I examine the underlying sources of regional growth in adults, including natural increases in the adult population through larger generations and longer life expectancies, in addition to migration. I trace the connections between groups of people and between neighborhoods in the regional housing market. I observe how regional demand pressures play out at the local level, and I show how the housing stock constrains and enables growth in different types of neighborhoods. In such a tight housing market, how do different groups of people manage to find space for themselves? Or do they simply find themselves squeezed out of the places they have called home? As a long-established city of immigrants, the Los Angeles region is accustomed to population growth from outside. But I find that most of the growth in adults since 2000 came simply from the aging of people who already lived in the region - not only White baby boomers and millennials, but middle-aged Hispanic and Asian immigrants and their children. This internal, invisible growth put tremendous pressure on the housing market, especially in urban neighborhoods that started out with large populations of younger adults and children. 19 Housing construction lagged far behind population growth. For every 100 expected additional households, only 69 housing units were added during the boom, and only 52 units were added during the bust. And most of this housing was built in suburban and exurban neighborhoods where undeveloped land was readily available, not in the urban neighborhoods with the most demand pressure. I show how people moved out of the urban neighborhoods facing intense demand pressure without much housing construction, and into the neighborhoods where housing was built, however far from employment opportunities or other amenities. I also find stark disparities between the housing and neighborhood outcomes for different groups in such a tight housing market: young people fared worse than older people, Blacks fared worse than Whites, and Hispanic immigrants fared worse than Asian immigrants. Finally, even with all these changing neighborhoods, some things remained the same: the basic socioeconomic hierarchy of the region was unaltered. 2. Research Methods I begin by analyzing the sources of regional growth in demand for housing over the course of the housing boom of the early 2000s and the subsequent great recession and housing bust. I use Census and American Community Survey (ACS) data to analyze population and housing shifts during the boom (2000 Census to 2005-09 ACS) and bust (2005-09 ACS to 2010-14 ACS) periods. 2 The underlying sources of growth in housing demand include (1) natural increases in the adult population through larger generations and longer life expectancies, in addition to the more commonly-studied (2) migration, (3) household formation, and (4) changes in income. In this chapter I focus on growth in the adult population through natural increases and migration. Because household formation is directly influenced by housing market conditions, changes in the adult population are a better measure of changes in demand pressure than changes in the number of households. Even if unable to form households because of high housing prices, people who attempt to form households put pressure on the housing market. In-migration and out-migration are also influenced by housing market conditions to some extent, but they depend on labor market conditions and factors outside the region as well. And generational size differences are the legacy of population patterns and fertility decisions made decades in the past, independent of any recent changes in housing market conditions. Changes in life spans are similarly independent of current housing market conditions. Changes in incomes largely depend on labor market conditions and are outside the scope of this dissertation. During the study period, incomes in the Los Angeles region stagnated and declined (the median household income fell 11.4 percent from 1990 to 2012). The 2 In order to analyze changes at the neighborhood as well as regional level, I use 5-year averaged ACS data, which are available at the census tract level. The 2005 to 2009 collection period is centered on 2007, a mid point that I will use to describe the data. The full ACS was first conducted in 2005. Though the end of the 2007 period includes the beginning of the downturn and fall 2008, the crisis, the 2007 averages provide the best available census tract level data on demographic and housing conditions at the height of the housing boom. 20 substantial growth in housing demand is all the more striking since it was driven primarily by population growth, and undermined by declines in both household formation and incomes. I measure growth in adults from natural increases of the existing population by starting with the detailed age profile of the population at the beginning of each period. I calculate expected growth for the boom and the bust using 5-year age category data for the start of each period, adding 7 years and 3 months to the age of the population from the April 1, 2000 Census to the middle of 2007 for the 2005-09 ACS, and adding 5 years to the age of the population from the middle of 2007 to the middle of 2012 for the 2010-14 ACS. I subtract deaths using age- and race-specific mortality counts for the Los Angeles CSA during each period (CDC 2015). After I estimate the adult population growth expected due to aging and mortality for detailed age groups by race and ethnicity, I subtract expected growth from the actual growth that occurred to measure net migration for each group. First I examine detailed changes by age and race over the entire period, then I compare growth rates during the boom and the bust. After appraising demand growth and net migration for different groups at the regional level, I turn to the neighborhood level. In order to study how regional demographic shifts and housing market conditions affect different types of places, I must first develop a typology of neighborhoods within the region. I define neighborhood types using 1990 Census measures interpolated to 2010 census tract boundaries using the Longitudinal Tract Database (Logan, Xu, and Stults 2012). Setting the neighborhood definitions in 1990 with 2010 census tract boundaries allows me to measure any changes that occur from 1990 onward in consistent geographic units over time, combining 2000 Census data with 2005-09 and 2010-14 ACS data, all interpolated to 2010 census tract boundaries as necessary. I use K-means cluster analysis to define neighborhood types, after testing various hierarchical and partitioning clustering techniques and determining that the K-means method produces the most distinct neighborhood types in terms of high within-cluster similarity and high between-cluster variation. I include a range of demographic and socioeconomic criteria in the cluster analysis: age structure, race and ethnicity, immigration, educational attainment, median household income, and homeownership rates, all standardized and weighted according to their level of clustering within the region. The resulting neighborhood types are not necessarily spatially contiguous, though they tend to be closely grouped following the clustered spatial structure of the region. A full explanation of the methods used to determine the neighborhood types is included in Appendix A. I describe the different types of neighborhoods at the beginning of the section of neighborhood change, with a map to illustrate their spatial patterns. For each neighborhood type, I measure expected growth through aging and the growth of the housing stock. I use the difference between expected growth in the population and the actual growth of the housing stock as an indicator the housing supply response to demand pressure, a concept similar to housing price elasticity. I then assess the relationship between neighborhood-level net migration rates of various groups and the tightness of the neighborhood housing market. 21 3. Regional Housing Market Conditions When a new and unfamiliar group of people enters a neighborhood, local residents, journalists and scholars often ask why they came to this place. What suddenly drew them to move to a place where other similar people had not lived before? This question is important especially for urban planners working to design better neighborhoods for people. But behind the question, “Why did they come?” lurks another question: “Why did they leave the place where they were before?” This question is less straightforward - it may have little to do with the neighborhood where they live now. The answer could be that they could not afford their old neighborhood anymore. Or the answer could be that they did not move away from other neighborhoods, that there were simply more people in the group than before, and some needed to strike out for new places. The push factors are almost always more murky and diffuse than the pull factors that attracted people to a certain place. In this section, I analyze the regional housing market conditions in the Los Angeles CSA over boom and bust, with a focus on the demographic sources of growth in demand for housing. Steady, strong growth in demand without much growth in housing meant that people were met with push factors at every turn, in one of the most constrained housing markets in the country. Usually, such high prices would mean that people would move away from the region, or stop coming in the first place. And both of these things did happen. But even with very low in-migration rates, the regional adult population continued to grow at a steady clip as middle-aged immigrants and baby boomers aged, and as their children grew up and entered the housing market. 3.1. Growing Demand for Housing Though incomes were stagnant, regional population growth fueled increasing demand for housing over the course of the boom and bust. Regional growth accelerated from an average of 114 thousand adults added per year during the 1990s to 172 thousand per year during the boom and 179 thousand per year during the bust. From 2000 to 2012, the adult population grew from 12.4 million to 14.6 million, a robust 17.3 percent increase in a little over a decade. Growth increased during the boom and bust despite lower immigration: immigration fell from an annual average increase of 173 thousand immigrants during the 1990s down to 151 thousand per year during the boom and 82 thousand per year during the bust. As immigration slowed, the main source of regional growth shifted from immigration to internal growth as larger generations of young regional residents came of age and older regional residents lived longer. Table 2.1 Immigration to the Los Angeles region and internal growth through the aging of existing residents in the 1990s and boom and bust periods Data Sources: Census 1990, 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 1990-2012. With natural increases through aging driving growth, it is no surprise that most of the growth was concentrated in the older age groups. As shown in Figure 2.1, middle-aged adults in their late 40s through early 60s increased 40.7 percent from 2000 through 2012. Seniors also grew at a rapid pace; 1990s 2000-07 2007-12 Immigration, Annualized 173,000 151,000 82,000 Internal Growth in Adults, Annualized 107,000 159,000 163,000 22 the region gained 30.3 percent more seniors from 2000 to 2012. Youth aged 15-24 grew more modestly, adding 17.0 percent. The one group that did not grow was young adults aged 25-44, who held steady with a slight decline of 1.0 percent. Demand for housing is strongest among middle-aged adults and seniors, and the sizeable increases in those age groups without declines in younger age groups put an especially strong pressure on the regional housing market. Figure 2.1 Population by age and race in the Los Angeles region in 2000, 2007, and 2012 Data Sources: Census 2000; ACS 2005-09, 2010-14. Growth Through Natural Increases and Net Migration The relative contributions of natural increases and migration can be roughly distinguished by comparing the growth expected through aging and mortality of the population from the start of each period with the actual growth that occurred. The difference between expected natural increase and actual growth is due to the impact of net migration into and out of the region. The changes expected through aging and mortality from 2000 to 2012 are shown as outlined bars in the middle graph in Figure 2.2. The actual changes that occurred are shaded bars, and the difference - net migration - is shown below at the same scale. This view allows for comparisons of the relative magnitude of natural increases and migration in the growth of each racial group and their contributions to the growth of the total population. The most growth came through the aging of Hispanic young and middle-aged adults, who increased the regional population by 1.22 million adults over age 35 between 2000 and 2012, even while losing 50 thousand through migration. Many were immigrants who had arrived as young people over the previous several decades, now entering middle age and beyond. Their children came of age at the same time, contributing another 393 thousand adults to regional growth: Hispanic youth and young adults aged 15 to 34 increased by 183 thousand through aging alone, and added another 210 thousand through migration. Across all ages, the region gained 1.62 million Hispanic adults over the boom and bust periods, 1.46 million of whom already lived there and were simply aging into adulthood. 23 Figure 2.2 Expected growth in adults through aging and mortality, actual growth, and net migration by age and race in the Los Angeles region from 2000 to 2012 Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. 24 In comparison, White baby boomers only added 357 thousand adults over age 50, their expected growth of 505 thousand dampened by net out-migration of 148 thousand. Their growth was countered by declines in younger White adults as baby boomers aged out of the category, followed by a much smaller Generation X. White adults in their 30s and 40s declined by 506 thousand, mostly through aging with slight out-migration. Even with the millennials entering the housing market, their expected growth of 46 thousand youth in their late teens and 20s was cut in half by net out- migration of 24 thousand, leaving an increase of only 22 thousand White youth aged 15 to 29. Though the regional population of White adults had been expected to increase by 57 thousand, or 1.1 percent, net out-migration of 183 thousand White adults, mostly at older ages, meant that White adults declined by 127 thousand, or 2.4 percent. Migration to the Los Angeles region was strongest for Asians, who added 416 thousand adults through migration and another 165 thousand through expected aging, for a total increase of 581 thousand. About half of their increases came through expected aging as immigrants who had arrived during the past several decades entered middle-age and older age groups. Despite expected declines in younger age groups, robust in-migration of Asian adults of all ages boosted the population. Asian adults were by far the fastest-growing group: they increased by 40.8 percent between 2000 and 2012. In contrast, Black adults of all ages migrated away from the Los Angeles region. As Black millennials came into adulthood and Black baby boomers entered middle age, Black adults were expected to increase by 146 thousand, or 15.9 percent. However, with net out-migration of 75 thousand (a full 8.2 percent of their population in 2000), they increased by only 70 thousand, or 7.7 percent, mainly adults over 45. Natural Increases Versus Migration Rates On the whole, natural increases from existing residents were a much larger source of regional growth than migration, and increases in youth under age 24 and middle-aged and older adults over 45 put pressure on specific segments of the housing market. The relative contributions of expected aging and net migration are reflected in the raw numbers above, but can be most clearly seen by comparing the aging growth rates and net migration rates by age and race shown in Table 2.2. While growth through expected aging was concentrated in the older age groups, growth through migration was strongest for younger adults, who generally tend to migrate to metropolitan regions at higher rates than older adults. Middle-aged and older adults of all racial and ethnic groups increased remarkably through aging tempered by minor out-migration. Adults aged 45 and over were expected to increase by 41.7 percent from 2000 to 2012. With net out-migration of -4.5 percent, middle-aged and older adults increased by 37.2 percent overall. Youth aged 15 to 24 were expected to increase by 13.5 percent, and positive net migration of 3.4 percent mostly Asian and Hispanic youth added to their growth for a total increase of 17.0 percent. Young adults aged 25 to 34 were expected to decline by -8.3 percent, but picked up the slack through positive net migration of 10.4 percent and increased by 2.1 percent overall. Asian young adults had an especially strong positive net migration rate of 35.1 percent, and Hispanic and White young adults had 10.3 percent and 8.3 percent positive net migration rates respectively, while Black young adults decreased by -11.6 percent through negative net migration. Positive net migration of 1.6 percent was not enough to avert expected declines for adults aged 35 to 25 44, who declined by -4.0 percent. Over age 35, only Asians had substantial positive net migration, and most other groups declined slightly through net out-migration. Table 2.2 Growth rates through natural increases and net migration by age and race in the Los Angeles region, 2000 to 2012 Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. Aging and Migration in Boom and Bust Growth in the adult population was relatively constant over the boom and bust periods, with minor decreases in aging growth rates and an increase in out-migration rates between the boom and the bust periods. To compare the sources of growth across different time periods, I calculate expected growth rates due to aging and mortality and rates of net in-migration and out-migration by race, shown in Figure 2.3. I adjust the growth rates for the 7 year and 3 month boom period for comparability with the 5-year bust period. 3 Net in-migration and out-migration rates are based on the sum of net in-migration or net out-migration across all age groups of a particular race. This measures the relative magnitude of positive or negative net migration across groups rather than measuring the gross amount of in-migration or out-migration that occurred, which is not possible with this data. The regional adult population grew at a 5-year rate of 6.4 percent during the boom and 6.0 percent during the bust through a combination of aging and mortality. Hispanic adults had the highest 5-year expected growth rates through aging: 12.7 percent during the boom and 11.8 percent during the bust. Black adults were expected to increase at a rate of 6.7 percent during the boom and 5.2 percent during the bust. Asian adults were expected to increase at a rate of 5.2 percent during the boom and 3 The boom period in this analysis starts at April 1, 2000 Census and runs through the middle of 2007 for the 2005-09 ACS average. The bust period starts in the middle of 2007 and runs through the middle of 2012 for the 2010-14 ACS average. Aging Mig All Aging Mig All Aging Mig All Aging Mig All Aging Mig All 15-24 13.5 3.4 17.0 20.3 7.0 27.2 8.7 -8.3 0.4 -16.1 32.9 16.8 24.5 -14.2 10.3 25-34 -8.3 10.4 2.1 -4.5 10.3 5.8 -16.6 8.3 -8.3 -13.5 35.1 21.6 4.1 -11.6 -7.5 35-44 -5.7 1.6 -4.0 27.4 0.0 27.4 -31.6 -1.6 -33.2 -6.5 27.7 21.2 -19.1 -5.4 -24.5 45-54 30.4 -3.4 27.0 82.4 -6.3 76.0 4.5 -6.7 -2.2 14.8 25.2 40.0 31.6 -7.6 24.0 55-64 67.4 -4.1 63.3 115.2 -4.5 110.7 45.2 -8.3 36.9 76.9 28.8 105.6 54.0 -4.2 49.9 65-74 40.4 -5.2 35.1 72.5 -1.7 70.8 24.2 -8.8 15.4 57.9 24.4 82.3 38.1 -5.7 32.4 75+ 31.9 -7.1 24.9 96.2 0.7 96.9 3.4 -2.4 1.0 86.2 19.7 105.8 29.6 2.3 31.9 All>15 15.8 1.4 17.3 32.4 3.5 36.0 1.1 -3.5 -2.4 11.6 29.2 40.8 15.9 -8.2 7.7 Percent Change, 2000 to 2012 Population Hispanic Non-Hisp. White Asian & Pac. Isl. Black 26 4.3 percent during the bust. White growth through aging barely registered by comparison. White adults were expected to increase at a rate of 0.3 percent through aging during both the boom and the bust, though with considerable variation in growth at different age groups, as discussed above. Figure 2.3 Growth due to natural increases, in-migration, and out-migration by race in the Los Angeles region during boom and bust Note: Scaled for comparability of 87-month and 60-month boom and bust periods. Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. For the adult population as a whole, net in-migration rates held steady over the boom and the bust, but net out-migration rates increased during the bust. In-migration contributed less than half as many adults as aging, at a 5-year rate of 2.7 percent during the boom and 2.6 percent during the bust. During the boom, less than half as many adults were lost through out-migration as were added through in-migration: net out-migration reduced the adult population at a 5-year rate of -1.2 percent. The out-migration rate nearly doubled during the bust, with a net loss of -2.2 percent of adults over 5 years. Migration rates varied considerably between racial and ethnic groups, and changed between the boom and the bust periods. Both Hispanic and White in-migration rates declined and out-migration rates increased from the boom to the bust. During the boom, the Hispanic in-migration rate of 2.6 percent out-paced their out-migration rate of -0.6 percent, for a net positive migration rate of 2.0 percent. But during the bust, the Hispanic in-migration rate fell to 1.8 percent and their out- migration rate increased to -1.5 percent, and their net migration rate fell to only 0.3 percent. White in-migration rates were somewhat lower: 1.3 percent during the boom, falling to 0.9 percent during the bust. And White in-migration rates did not keep up with their out-migration rates, which started at -1.8 percent during the boom and almost doubled to -3.4 percent during the bust. All told, White adults had a net negative migration rate of -0.5 percent during the boom and -2.4 percent during the bust. 27 These declines in Hispanic and White in-migration were offset by increases in Asian in-migration from the boom to the bust. During the boom, the Asian in-migration rate of 10.0 percent was already about twice as high as the Asian growth rate through aging, and increased to 11.2 percent during the bust, with little-to-no net out-migration. In contrast, Black adults had very low in- migration rates, and had the highest rates of out-migration: a -3.3 percent 5-year loss rate during the boom and a -3.5 percent loss rate during the bust. Regardless of the economic conditions, the dominant trend throughout the boom and the bust was the growth of middle-aged and older adults of all races through the aging of residents who already lived in the region. This growth was set in motion by previous waves of immigration and migration to the region, and supplemented by a relatively small amount of migration from younger adults. Growth through natural increases can put far more pressure on the regional housing market than migration. Migration applies external demand pressure, whereas natural increases apply internal pressure. For migrants from outside the region, if housing prices are too high they may choose to move elsewhere instead. With natural increases, the people involved already live in the region, and they need to find somewhere to live regardless of prices. They may decide to move elsewhere if they cannot afford housing in the region, but at a steep cost. And in some ways natural increases apply invisible pressure, manifesting itself not through an influx of newcomers but through residents simply continuing to occupy the homes where they already live, leaving little left over for younger residents and newcomers. 3.2. Constrained Housing Supply Housing construction in the Los Angeles region is limited by the mountains on one side and the ocean on the other, a shortfall of undeveloped land in between, and strong local resistance to new development. After the housing crisis, the tight restrictions on financing for new construction added an additional obstacle to housing development. As a result, housing construction lagged far behind population growth during both the boom and the bust. The regional housing stock grew 7.3 percent during the boom and only 3.6 percent during the bust, while the adult population grew 10.1 percent during the boom and 6.5 percent during the bust. Assuming headship rates by age stayed constant, only 69 housing units were added for every 100 potential added households during the boom, and only 52 housing units were added for every 100 potential added households during the bust. 3.3. Rising Housing Prices With rising demand pressure and little new construction, housing prices rose swiftly. During the boom, the median house value in the region nearly doubled, rising 95.6 percent, from $275 thousand to $538 thousand in 2015 inflation-adjusted dollars. Home buying was boosted by easy access to mortgage credit, and house values rose faster than rents. Even so, the median rent rose 19.7 percent during the boom, from $1,043 to $1,248. During the bust, the pattern was reversed: house values corrected sharply downward by 28.5 percent while the median rent held steady with a 1.3 percent increase. Even with the correction of the bust, house values were 39.8 percent higher in 2012 than they had been in 2000, and rents were 21.2 percent higher. 28 4. Neighborhood Housing Market Conditions Growth in demand pressure and housing stock constraints were highly uneven between different parts of the region. In many cases, the neighborhoods with the highest expected demand pressure also had the most constrained housing stock. In this section, I analyze expected demand growth, supply constraints, and the resulting shifts in population, household formation, and homeownership across thirteen types of neighborhoods in the Los Angeles region. I uncover signs of gentrification in some neighborhoods, signs of the loss of certain groups in other neighborhoods, and signs of competition between different groups with unequal sway in finding places to live in an increasingly unaffordable housing market. 4.1. Neighborhood Typology The cluster analysis reveals sharp separation between the types of places where different groups of people lived. Table 2.3 shows the percentage of the regional population of various groups of people living in each neighborhood type in 1990, the year the typology is based on, and in 2012 over two decades later. This measure reflects the distribution of each group across neighborhood types, whether concentrated or dispersed. Keep in mind that the neighborhood clusters are not equally sized; they include different numbers of census tracts and different population totals to begin with. As a result, groups would not have equal percentages in each neighborhood type even if they were evenly distributed across the region. Still, comparisons between the neighborhood concentrations of different groups show stark contrasts between the types of neighborhoods where different groups live, and comparisons over time show changes in concentrations within certain neighborhood types. The percentage of the population in each neighborhood type is a good baseline for comparison. It is all the more striking that 24 percent of the Black population lived in N4 when that neighborhood type encompassed only 3 percent of the overall regional population. As has been well documented in the segregation literature, 4 different racial and ethnic groups lived in different types of neighborhoods. In 1990, 48 percent of the regional Black population lived in just two of the thirteen neighborhood types: N1 and N4, which were low and lower middle income 4 Group distribution is a slightly different concept than segregation, which measures separation between different groups of people. But group distribution measures also allow for analyzing separation between groups using comparisons between the places where different groups are most and least concentrated within the region. A disadvantage is that distributions do not lead to a single summary measure of separation between groups, and are best suited to detailed analysis. An advantage of measuring distributions rather than separation between groups is that the distribution is measured independently for each group, and changes in the distribution of one group do not affect the measurement of the distribution of another group. This also allows for comparisons between groups of various types, such as immigrants vs. renters vs. the population. 29 neighborhoods mostly located in the center of the region, in South Los Angeles and the cities of Inglewood and Compton. By 2012, they had become slightly more dispersed across the region, as their population declined in N1 and N4 and increased in N8 and N11. Still, 37 percent of the Black population lived in N1 and N4 in 2012. The White population was also concentrated mainly in just a few types of neighborhoods, but on the other end of the socioeconomic scale. In 1990, 40 percent of Whites lived in two neighborhood types: N11 and N13, with relatively high income homeowners in places such as Lakewood, Chino Hills, and San Dimas, and very high income homeowners in places such as the Palos Verdes Peninsula, much of Orange County, Glendora, and other suburban cities in the foothills. Another 22 percent of Whites lived in high income rental neighborhoods in places such as Santa Monica and nearby coastal cities along the bay, Pasadena, West Hollywood, and Newport Beach. By 2012, the White population had become even more concentrated: 48 percent lived in the highest income suburbs of N11 and N13. In 1990 only 20 percent of Whites lived in the lower income neighborhood types N1-N7, as compared with 44 percent of the overall population. By 2012 only 17 percent of Whites lived in lower income neighborhoods. In 1990, Hispanics were especially concentrated with 45 percent of their population in N3 and N5, lower income neighborhoods such as Vernon, Bell, Santa Ana, and San Fernando. Another 22 percent of Hispanics lived in N6, middle income rental neighborhoods in places such as Alhambra and Monterey Park, and N8, mostly suburban neighborhoods in Riverside and San Bernardino counties. Their rapid growth in the suburban neighborhoods of N8 and N11 meant that they were slightly more dispersed throughout the region by 2012, but still only 13 percent of Hispanics lived in N9, N10, N12, and N13, the high income neighborhoods dominated by Whites. Since the majority of immigrants in the region in 1990 were Hispanic, similar concentrations can be seen when looking at the distribution of immigrants across neighborhood types, though immigrants were more slightly dispersed than Hispanics. Immigrants became notably more dispersed between 1990 and 2012 through increases in higher-income neighborhood types. These increases came mostly through Asian immigration: Asians were less concentrated than other groups in 1990, and unlike other groups they lived in neighborhoods throughout the socioeconomic scale, with the highest concentrations in N6, N12, N13, and N2. Asians increased their presence in N8, N9, N10, N11, and N13, skewing towards higher income neighborhoods. Neighborhoods were divided by tenure as well. In 1990, 68 percent of homeowners in the region lived in just five of the thirteen neighborhood types, mostly in higher-income areas: N11, N13, N8, N6, and N5. Renters were slightly more dispersed between different neighborhoods. In 1990, 60 percent of renters lived in their top five neighborhoods, a slightly different set of places: N9, N6, N10, N3, and N8. Owners became slightly more concentrated in those five neighborhoods by 2012, while renters became slightly more dispersed, mostly through increased in renters in the suburban neighborhoods of N11 and the highest-income neighborhoods of N13. 30 Table 2.3 Percent of the regional population of each group living within each type of neighborhood in 1990 and 2012 Data Sources: Cluster analysis with 1990 Census measures; Census 1990; ACS 2010-14. N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 1990 5 4 10 3 10 11 2 12 9 6 14 4 12 2012 4 3 8 2 9 10 2 15 8 5 17 4 12 1990 4 4 7 3 8 11 3 12 11 8 14 4 12 2012 3 4 6 2 7 10 3 13 10 8 17 4 13 1990 2 1 4 3 8 8 4 14 8 5 19 6 19 2012 2 1 3 2 7 7 4 14 7 5 22 5 19 1990 6 8 10 3 8 14 1 9 14 12 8 2 5 2012 5 7 9 3 7 12 2 11 13 11 10 2 7 1990 6 4 13 3 12 10 1 14 7 2 14 4 10 2012 5 3 10 2 10 10 2 17 7 3 17 3 11 1990 4 4 9 3 9 11 1 11 9 7 14 4 12 2012 4 4 8 2 9 10 2 14 9 6 16 4 11 1990 3 4 5 3 7 10 8 11 12 7 13 4 13 2012 2 3 5 3 7 9 6 11 9 6 17 5 17 1990 8 7 20 3 16 12 2 11 7 2 8 2 3 2012 8 5 18 2 15 12 2 16 6 1 9 2 3 1990 3 3 4 3 7 11 3 14 11 6 18 5 12 2012 3 3 6 3 8 10 3 16 9 4 20 4 10 1990 1 3 1 2 3 8 2 6 12 14 14 6 27 2012 1 3 2 2 4 8 2 7 11 12 17 6 25 1990 6 9 20 2 14 14 1 7 8 4 6 5 7 2012 5 6 13 2 12 14 1 12 9 4 10 5 8 1990 1 2 1 0 4 9 3 14 13 9 21 4 19 2012 0 2 1 0 2 6 4 11 11 10 25 3 23 1990 24 3 5 24 7 8 0 11 4 3 5 3 2 2012 14 3 3 18 5 8 1 18 6 3 13 3 4 1990 2 10 4 1 9 19 0 6 8 5 9 13 12 2012 1 7 3 1 7 16 1 8 10 6 14 12 15 1990 7 6 25 2 20 12 1 11 5 2 6 3 3 2012 7 3 16 2 15 11 2 19 6 2 12 2 3 1990 $34 39 42 57 59 62 50 64 66 78 87 99 129 2012 32 36 37 48 51 52 48 54 61 77 79 83 115 1990 $881 964 923 1,037 1,127 1,187 1,032 1,110 1,225 1,391 1,382 1,505 1,641 2012 1,004 1,099 1,027 1,159 1,223 1,255 1,205 1,244 1,349 1,630 1,563 1,623 1,856 N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 Neighborhood Type Neighborhood Type Median Gross Rent (adjusted 2015 $) Tenure Age Education Race and Ethnicity Immigrants Median Hh Income (adj 2015 $, thousands) Children Renters Owners Households Population Percent of Regional Population by Neighborhood Type Non-Hsp White Non-Hsp Black NH Asian & PI Hispanic Colleges Grads HS Grads Less Than HS Seniors Adults 31 Figure 2.4 Map of neighborhood typology in the Los Angeles metropolitan area, with inset showing the five-county region Data sources: Cluster analysis with 1990 Census data interpolated to the 2010 census tract level. 32 Despite slight changes in the distributions of certain groups, the overall sociodemographic structure of the region remained remarkably stable over more than two decades, and the economic hierarchy of neighborhoods was basically unchanged. The median household income fell for neighborhoods across the region, but only two neighborhood types switched places in the order: N4 dropped from 9 th to 10 th as its median household income fell 16.0 percent while the median household income in N7 only fell 5.2 percent. Even though the neighborhood types were not defined using measures of spatial location or proximity, neighborhoods with similar demographic and socioeconomic profiles tended to be located near each other, or at least in places with similar urban or suburban characteristics. Figure 2.4 shows the spatial distribution of neighborhood types in the region, labeled with names based on the predominant groups that lived in the neighborhood type and their relative income level. The clearest spatial pattern is that neighborhoods with more rental housing tended to be located near urban centers, while neighborhoods with more owner-occupied housing tended to be located in the suburbs. 4.2. Growing Demand for Housing Expected growth varied a great deal across neighborhood types, from places where the adult population was expected to grow by nearly a third to places where the adult population was expected to decline. Figure 2.5 shows the expected percent change through aging and mortality during the boom and bust. The neighborhood types with the most expected growth were places with many children and few middle-aged and older adults to begin with. Conversely, the neighborhood types with the least expected growth (or even expected declines) were places with established populations of middle-aged and older adults and few children poised to enter adulthood. Figure 2.5 Expected adult population growth through aging and mortality in the Los Angeles region during boom and bust, 2000 to 2012 Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. 33 4.3. Constrained Housing Supply Rising demand pressure did not necessarily lead to corresponding increases in housing construction. The construction shortfall was particularly severe in urban areas with little undeveloped land for new construction and strong local resistance to growth, as well as neighborhoods with lower housing prices and less potential for profit. As shown in Figure 2.6, 57.3 percent of all the units added were located in just three types of neighborhoods with strong construction activity: N11, N8, and N7, low-density middle- and high-income suburbs that make up most of San Bernardino and Riverside counties and the further reaches of Los Angeles County. Across these three types of neighborhoods, the housing stock grew 21.0 percent over the boom and bust, slightly more than the regional adult population growth of 17.3 percent. Housing growth was also strong in the highest-income N13 neighborhoods located mainly in Los Angeles and Orange counties, where housing increased 11.5 percent over the boom and bust. In the rest of the region, the housing stock grew only 5.8 percent. Figure 2.6 Increases in the housing stock by neighborhood type in the Los Angeles region during boom and bust Data Sources: Census 2000; ACS 2005-09, 2010-14. 34 In general, new housing was built only in the most expensive places, with people who could afford to pay high prices, and in places where land was plentiful and inexpensive, but which were far from employment opportunities and amenities of urban centers. For people who could not afford housing in N13, “drive ‘till you qualify” was the order of the day, and the drive kept getting longer and longer over the course of the boom. Home buying played an important role in this shift towards the suburban and exurban areas of N7, N8, and N11. Many of the older urban neighborhoods offered mostly rental housing, with few opportunities to buy. In order to take advantage of the expansion of mortgage financing during the boom, people bought the new homes built in places such as Lancaster, Palmdale, Antelope Valley, and parts of San Bernardino and Riverside counties. Unfortunately, these places were the hardest hit by the housing bust, and continue to struggle. 4.4. Rising Housing Prices Across the Region Though both expected demand pressure and the growth of the housing stock varied across neighborhood types, the patterns of home value and rent changes during the boom and bust are Figure 2.7 House prices changes by neighborhood type in the Los Angeles region during boom and bust Data Sources: Census 2000; ACS 2005-09, 2010-14. 35 strikingly similar across neighborhoods, and for the 25 th and 75 th percentiles as well as the medians. Comparisons between house value changes and rent changes yield a sharper contrast than comparisons between house price changes in higher and lower income neighborhoods. This reflects the strong distinctions between rental and owner-occupied housing, as well as the tight connections between housing submarkets within the region. Growing regional demand put pressure on every last corner of the housing market, from urban to suburban areas, wealthy to poor neighborhoods, and luxury apartments to economy studios. 4.5. The Housing Stock Constrains and Enables Growth Many of the places where the most growth could be expected were the least able to accommodate growth through expansion of the housing stock. Figure 2.8 compares expected growth in adults, actual change in the housing stock, and actual growth in adults by neighborhood type during the boom and the bust periods. Conceptually, it measures potential demand pressure in a neighborhood, the severity of the housing shortfall, and the resulting population growth. In most neighborhoods, far less housing was built than the expected growth of the population. In those neighborhoods, actual growth was far lower than expectations. In the few neighborhood types where housing was built at a rate that met or exceeded expected growth, actual growth far exceeded expected growth, as growing demand spilled over from all the neighborhoods where housing was constrained. The scatterplots in Figure 2.8 show the importance of housing growth in translating expected growth into actual growth. The horizontal axis represents the difference between housing growth and expected growth in adults; higher values represent a stronger supply response to demand pressure. The vertical axis represents the difference between actual growth and expected growth; higher values indicate net migration into the neighborhood above expected growth, and lower values indicate net migration out of the neighborhood, reducing growth below expected levels. Neighborhood-level migration was strongly positively correlated with housing production: even without including N7, the most extreme value. Migration was correlated with the supply response by .912 during the boom and .837 during the bust. Expected increases in demand through aging remained mostly steady during the boom and the bust. Less housing was produced during the bust, though most neighborhood types had similar relative levels of housing production in both periods. The places with the most construction during the boom also had the most construction during the bust, and vice versa. The one neighborhood type that shifted considerably between the bust and the boom was N2, where housing construction fell 4.7 percentage points short of expected growth during the boom, but exceeded expected growth by 2.6 percentage points during the bust. 36 Figure 2.8 Actual growth in adults compared with expected growth and housing growth by neighborhood type in the Los Angeles region during boom and bust Note: Scaled for comparability of 87-month and 60-month boom and bust periods. Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. 37 4.6. Housing Constraints and Neighborhood Migration by Race Housing supply growth mattered more for some groups of people than others. Figure 2.9 shows the relationship between the housing supply response to expected demand growth in a neighborhood and net migration of young adults aged 25 to 44 by race. Migration rates of Hispanic young adults were tightly linked with the housing supply response (with a correlation of .878). In places where more housing was built compared with expected demand, more Hispanic young adults moved in, and they tended to migrate away from neighborhood types where the housing stock was more constrained. The migration of White young adults, on the other hand, was not necessarily tied to Figure 2.9 Housing supply response and migration rates of young adults by race for neighborhood types in the Los Angeles region, 2000 to 2012 Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. 38 housing supply growth (with a correlation of -.025). White young adults were able to compete for housing even in the very tight markets of N2 and N4, where little new housing was produced. Unlike other groups, White young adults barely increased at all in the suburban and exurban neighborhoods of N7, N8, and N11 , where most of the new housing in the region. Asian young adults did increase in the areas with the most construction, but they migrated to other areas as well, and migration in Asian young adults was correlated with the housing supply response by .771. Notably, Asian young adults increased through net positive migration into every neighborhood type but N1, the lowest-income neighborhood type. Black young adults fared quite differently, with negative net migration in most parts of the region, and positive net migration only in the places where lots of housing was built. Their net migration into N7 was off the chart, and their migration was correlated with the housing supply response by .921, and remains quite strong at .917 when N7 is not included in the correlation. The relationship between the housing supply response and neighborhood migration for particular groups can be expressed in terms of correlations, shown in Figure 2.10. The correlations between the housing supply response and migration rates fell slightly from the boom and the bust, in part because housing construction rates fell. Still, the basic patterns remain: neighborhood migration was Figure 2.10 Correlation between neighborhood net migration rates and the housing supply response to expected demand Data Sources: Census 2000; ACS 2005-09, 2010-14; CDC Multiple Cause of Death Files, 2000-2012. Boom Bust Boom Bust Boom Bust Boom Bust Boom Bust 15-24 .492 .326 .015 -.197 .647 .839 .584 .562 .871 .446 25-44 .856 .713 .160 .056 .934 .807 .843 .129 .825 .616 45-64 .835 .641 .578 .760 .842 .602 .808 -.017 .767 .622 65+ .938 .668 .707 .841 .882 .674 .803 .276 .878 .275 All >15 .966 .918 .408 .397 .941 .798 .928 .381 .876 .645 Correlation of Neighborhood Migration and Housing Supply Response Population NH White Black Asian & PI Hispanic 39 least correlated with the housing supply response for White young adults, and most correlated for Blacks of all ages. The correlation fell most sharply between the boom and bust for Asians, who began to move out of the exurban N7 neighborhoods where most of the construction had been during the boom. Why is the correlation stronger for the population overall than for specific groups by age and race? This is evidence of competition between groups for housing within neighborhoods. For example, if there is an influx of White young adults in a housing-constrained neighborhood, and they out-bid other groups for housing, the net migration of those other groups will be lower, but the overall level of migration to the neighborhood may remain the same. The correspondence between migration to the neighborhood and the housing supply response remains constant, but the relationship is altered for each group. The clearest instance of this occurred during the bust in N2 Low Income Hispanic, Asian, & White Immigrant Renters. During the bust, the overall population of N2 was expected to grow 3.8 percent through aging and mortality alone. The housing stock grew 6.4 percent, the second-highest level of construction during the bust. Over the boom and the bust, White young adults under 40 added 33 thousand through net in-migration, amounting to 6.6 percent of the total population of N2 in 2000. Asian young adults under 35 added another 16 thousand, and Black young adults added 3 thousand. (In fact, N2 was the only housing-constrained neighborhood type where Blacks had positive net migration). In contrast to this influx of 52 thousand young Whites, Asians, and Blacks, Hispanics of all ages migrated away from N2, losing 76 thousand to net negative migration, and with expected growth Hispanic young adults under 40 declined by 34 thousand. The overall population of N2 grew by 31 thousand, or 6.2 percent, as the in-migration of young Whites, Asians, and Blacks was offset by out-migration of Hispanics. 5. Conclusion The first part of this chapter highlights the sheer scale of demand growth that is possible through the aging of residents who already live in the region, especially as compared with migration. Migration and income growth are commonly considered as the main sources of growth in demand for housing. Over the period of boom and bust, incomes fell 7.8 percent, but the number of adults increased 10.1 percent through a combination of 9.3 percent growth through expected aging and 0.8 percent growth through net migration. This illustrates the intense demand pressure on housing within the region, despite falling incomes. Demand growth in concert with the declines in the amount of housing available on the market (described in Chapter 1) led to a severe housing shortfall, which was disastrous for housing affordability. In 2000, 14.7 percent of owners with a mortgage paid more than half of their incomes in monthly owner costs. With the rising prices of the boom, by 2007, 24.1 percent of owners were paying over half their incomes. As prices fell and some were able to refinance their mortgages, by 2012, 21.2 percent of owners paid more than half their incomes for housing costs. In 2000, 22.2 percent of renters paid more than half their income for rent. By 2007, 28.5 percent of renters paid more than half their income, and by 2012, the proportion paying over half their incomes for rent had climbed to 31.9 percent. By the end of the boom and bust period, the number of households, 40 renters or owners, who were paying over half their income for housing costs had increased by 67 percent - from 822 thousand to 1.37 million, nearly a quarter of the total households in the region. The boom period and the bust period were not as different as expected in terms of the severity of the housing shortfall. This was in part because most of the growth in demand in both periods came through aging of the existing population rather than through aging. Housing was already so expensive during the boom period, and so little housing was built in response to rising prices, that migration to the Los Angeles region fell during the boom. The biggest differences between the housing boom and bust were related to housing tenure, starting with the foreclosure crisis and continuing as investors purchased homes and converted them to rental units. These changes are outside the scope of this paper. But the surge in demand from young people just entering the housing market also increased the demand for rental housing (Campbell 1966). Reduced access to mortgage credit and falling housing prices also encouraged a shift toward renting. The second part of this chapter highlights the disparities between the neighborhoods where different groups moved in the context of the housing shortfall. Different types of neighborhoods had different levels of expected demand growth, and different amounts of housing built over the boom and the bust. Some of the urban neighborhoods where the most demand growth was expected also had the most constrained housing growth. By far the most housing was built in the suburban and exurban neighborhoods with low populations to begin with. The neighborhood migration rates of Hispanic and Black young adults were highly correlated with the severity of the housing shortfall: they consistently moved away from places with tight housing markets and toward places where housing was built. This suggests that their housing and neighborhood outcomes were highly constrained by the housing stock and competition in those neighborhoods. However, the neighborhood migration rates of Asian and White young adults were barely correlated with neighborhood-level housing shortfalls: Asian and White young adults were able to out-compete other groups for housing, even in very tight submarkets. There is some evidence of gentrification in N2, where White young adults increased suddenly, with a large outflow of Hispanic residents. These differences between the neighborhood migration rates of various racial and ethnic groups are signs of the degree to which regional housing shortfalls worsen the disparities between groups. Some can afford to compete for housing in high-quality neighborhoods even if they have to pay more for it. Others are priced out, and are left to find housing in less desirable neighborhoods. Once less advantaged groups end up in neighborhoods that offer worse access to employment, schools, and other amenities and public services, they fall further behind the advantaged groups who live in more supportive environments. Everyone bears the cost of a housing shortfall, but the people who are most heavily penalized are those who can least afford to pay the price. 41 Appendix A. Neighborhood Typology Cluster Analysis Technique I used cluster analysis to map the urban socio-spatial structure in 1990 in order to analyze subsequent changes in various types of neighborhoods. I included a range of demographic and socioeconomic variables used to define neighborhoods: age structure, race and ethnicity, immigration, educational attainment, median household income, and homeownership rates. To the extent that these variables are correlated with each other, they provide a stronger delineation of neighborhood types. I generated sets of clusters using a range of methods: average, centroid, and Ward's method hierarchical techniques and K-means and K-medians partitioning techniques. I compared the resulting sets of clusters from the various techniques and assess how well they delineated neighborhood types using the intraclass correlations of the variables included in the cluster analysis to maximize within-cluster similarities and between-cluster differences. Variable Weighting I first standardized each of the variables so that their various units of measurement would not lead to various levels of influence over the cluster results. Though standardizing the variables removes the problem of measurement units giving certain variables undue weight, it is not necessarily appropriate that each of the variables involved should have equal weight in defining neighborhood types. Indeed, some of the variables are likely more important than others in distinguishing neighborhoods, and should be weighted accordingly. Though there is no single standard method to determine variable weights for cluster analysis, Gnanadesikan et al. (2007) lay out a series of methods and divided them into weighting schemes that equalize the impact of different variables and weighting schemes that highlight the importance of variables that more clustered in the data. For each of the clustering techniques, I tested (1) unweighted variables, (2) a weighting scheme based on the average correlation of each variable with the other variables, which emphasizes variables that tend to cluster with other variables, and (3) a weighting scheme based on the inverse of the average correlation of each variable with the other variables, which emphasizes variables which are less clustered with other variables. Across clustering techniques, the variable weights based on average correlations outperformed both the unweighted variables and the variable weights based on the inverse of the average correlations. Hierarchical Clustering Techniques Both average-linkage and centroid-linkage hierarchical techniques produced sets of clusters with a large majority of census tracts in one cluster, most of the remaining tracts in one or two mid-size clusters, and then the last few tracts in many clusters with only a handful of tracts each. The large and mid-size clusters in these sets represent broad ranges of values in the variables included, while the much smaller clusters represent narrow extremes. These lopsided sets of clusters are not intuitively helpful in distinguishing between the diverse types of neighborhoods in the region, and also have lower intraclass correlation coefficients than the more balanced sets of clusters produced by other methods. Ward’s method of hierarchical clustering produced a more balanced set of clusters, which I then compared with the clusters produced by the K-means cluster technique. 42 Partition Clustering Techniques I also used the K-means and K-medians partitioning techniques to define neighborhoods. Partitioning methods start with a certain number of observations, or ‘seeds’, defined by the user, and find the best solution of clusters based on those seeds in terms of minimizing within-cluster variation and maximizing between-cluster variation. These partitioning methods produce slightly different clusters depending on the seeds that they begin with, and the user defines the number of clusters desired. I compared the results based on a range of 7 clusters to 15 clusters, as discussed below. I allowed for random selection of the seeds, and iterated the cluster analysis 100 times for each number of clusters tested. Then I measured the intraclass correlation coefficients for each of the variables included in the cluster analysis, and for each number of clusters I selected the iteration with the highest intraclass correlation across all variables. Across all numbers of clusters, the K- means clustering approach out-performed the K-medians approach in terms of the intraclass correlations of the variables involved. Choosing a Number of Clusters For both Ward’s method and K-means clustering approaches, the user must decide how many clusters provide the best solution. With Ward’s method, I used the Calínski–Harabasz pseudo-F index and the Duda–Hart Je(2)/Je(1) index stopping rules to determine the optimal number of clusters. Using this technique, 13 clusters produced the best balance between within-cluster and between-cluster variation. For the K-means clustering method, I tested cluster solutions for 7 through 15 clusters, a range of clusters that had reasonably good Calínski–Harabasz pseudo-F and Duda–Hart Je(2)/Je(1) statistics when using Ward’s method. For each number of clusters, I compared the intraclass correlation coefficients for each of the variables included in the analysis, as shown in the graph below. Most of the intraclass correlations were relatively high to begin with, and stayed consistent across the various numbers of clusters. However, the percent Black and the percent Asian and Pacific Islander started with low intraclass correlations, which improved as the number of clusters 43 increased. The intraclass correlation coefficient for the percent Black increased from .17 with 7 clusters to .79 with 11 clusters, and leveled off from there. The intraclass correlation coefficient for the percent Asian and Pacific Islander started even lower, at .10 with 7 clusters, and increased to .41 with 13 clusters, leveling off afterward. I settled on 13 clusters as the optimal solution: 13 is the minimum number of clusters with reasonably good intraclass correlations for all the variables involved in the analysis, and increasing the number of clusters beyond 13 does not markedly improve the results. Across all numbers of clusters, the K-means method produced consistently better results than Ward’s method in terms of the intraclass correlations of the variables. Robustness Though the k-means method performed better than Ward’s method, both methods resulted in similar neighborhood typologies. Most of the tracts from each Ward’s method cluster fall into just one or two of the clusters generated through the K-means method, as shown in the table below. And where Ward’s method clusters span multiple K-means clusters, there are intuitive similarities between those clusters. For example, Ward’s cluster N2 is mostly made up of K-means clusters N8 and N11, which are both suburban areas with mostly White and Hispanic high school graduate homeowners with children. In the K-means cluster solution, N8 and N11 are distinguished from each other by income level (N8 is middle income, whereas N11 is high income) and by their racial and ethnic composition (N8 has more Hispanic residents; N11 has more White residents). In another example, Ward’s cluster N9 includes tracts from K-means cluster N1 and N4, the two clusters with significant Black populations. In the K-means typology, N1 and N4 are distinguished from each other by tenure (N1 has more renters; N4 more homeowners) and income (N1 has lower income levels than N4). But still, N1 and N4 are similar and physically contiguous on the map, and N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 Total N1 0 0 0 0 0 0 4 10 59 0 78 15 2 168 N2 0 0 0 0 1 0 48 1 0 0 0 0 120 170 N3 0 0 0 0 0 0 0 0 0 233 41 22 49 345 N4 0 0 38 0 0 0 9 0 52 0 0 0 0 99 N5 0 0 0 0 0 0 33 34 0 0 0 257 31 355 N6 0 1 2 0 27 0 220 130 0 0 0 0 5 385 N7 0 0 0 72 27 0 0 0 0 0 0 0 0 99 N8 0 279 0 0 0 0 0 218 0 0 0 0 0 497 N9 2 61 0 0 203 39 3 31 0 0 0 0 0 339 N10 4 0 0 0 29 189 2 0 0 0 0 0 0 224 N11 78 503 0 0 7 14 0 0 0 0 0 0 0 602 N12 26 28 75 0 8 2 15 1 0 0 0 0 0 155 N13 420 0 1 0 6 42 0 0 0 0 0 0 0 469 530 872 116 72 308 286 334 425 111 233 119 294 207 3907 K-means Method Total Ward's Method 44 it is not surprising that they might be lumped together using a different clustering technique. The overlaps and similarities between the clusters produced through the Ward’s and K-means methods give me confidence that the neighborhood typology developed using the better of the two approaches is a meaningful representation of the socio-spatial structure of the Los Angeles region. 45 Losing Their Place in Line: Chapter 3. Diverging Neighborhood Outcomes for Young People 1. Introduction As less new housing was built and older generations remained longer in their homes, far fewer housing opportunities were available for millennials relative to previous generations trying to enter the housing market in the Los Angeles region. And the competition for housing was fierce in such a large generation, as the children of immigrants and baby boomers surged into adulthood. In this tight housing market, the widening gap between the incomes of college graduates and people without college degrees may well have translated to a widening gap between the neighborhoods where they lived. In this chapter, I address two main research questions: First, did young college graduates and non-college graduates increasingly live in different types of neighborhoods? Second, did the housing stock play a role in influencing the places where young people lived? I focus on the Los Angeles region, which has attracted millennial college graduates despite an especially constrained housing supply. I analyze how the availability of newly built and recently vacated housing is related to differences between the neighborhoods where millennials and the previous wave of young people lived. I separately assess the changing residential locations of young adults with and without college degrees, which exposes the divergent trajectories of relatively advantaged and less advantaged young people. In the Los Angeles region, declining construction and residential turnover reduced the number of housing units available on the market by 12.3 percent between 1990 and 2012 – while the adult population grew by 29.2 percent. Young college graduates aged 25-34 contributed to the growing demand for housing: their numbers declined slightly in the 1990s and then increased sharply in the 2000s for an overall increase of 29.7 percent from 1990 through 2012. In contrast, young adults without a college degree declined in both the 1990s and 2000s, and there were 18.4 percent fewer young non-college graduates in 2012 than in 1990. These opposing trends are the result of several factors, including a) a larger millennial generation with b) higher college graduation rates, c) in- 46 migration of young college graduates, d) out-migration of young non-college graduates, and e) declining immigration to the Los Angeles region. With such large increases, young college graduates filled up and spilled over the edges of the neighborhoods where young graduates already lived. In the neighborhood-level analysis, I find that young college graduates increased in the most desirable neighborhoods, whereas young non-college graduates were more likely to remain in disadvantaged neighborhoods and declined nearly everywhere else. These findings are consistent with the idea that relatively advantaged people are becoming more advantaged over time by living in comparatively better neighborhoods, while disadvantaged people are increasingly priced out of better neighborhoods and often priced out of the housing market entirely. I also find that the amount of housing available on the market in a neighborhood, whether in the form of newly built units or recently vacated units, served as an enabling or constraining factor for changes in young college graduates and non-graduates alike. And housing tenure played an important role, with pronounced increases in young college graduates in neighborhoods with more rental housing. The Los Angeles region can offer lessons for other places where demand pressure confronts a constrained housing stock. The detailed empirical observation of the intersection of housing and population in this paper offers evidence to help planners understand the complex dynamics of neighborhood change and address affordable housing shortfalls. Most of all, the findings reinforce the urgency of building new housing and adapting existing housing to meet current and future housing needs. In the following section, I discuss the motivation and background for this research. Then I explain the methods and data I use to address the research questions. I examine the regional shifts in the population of young people by educational attainment. Then I present the model and results of a multivariate analysis of neighborhood-level changes in young people by educational attainment. To conclude the paper, I consider the implications of the findings in the context of policies that constrain the availability of housing for young people. 2. Motivation and Background This research is motivated by the desire to better understand how both the increasing number of college graduates and declines in the amount of housing available on the market impact the process of neighborhood change. I rely on filtering theory to understand how a constrained housing supply and demographic and economic shifts have led to reduced housing affordability, especially for young people, and to understand how young people manage to meet their housing needs. Filtering theory and the conceptual model that undergirds my approach are described in detail in the introduction to this dissertation. The filtering process produces affordable housing when new construction satisfies growing demand for housing. However, it is clear that when housing construction does not keep up with growing demand for certain types of housing and neighborhoods, housing will become less affordable over time and filter up to higher income households. The key point is that in the context of a constrained housing supply, higher income groups will pay more for housing, and lower income groups will end up in lower quality housing and neighborhoods or crowded out of the housing market altogether. 47 I study changes in young people because they are just entering the housing market, and are last in line for housing after previous generations, who on the whole have already established households. And since there are substantially more millennials than the previous generation, they must find places to live beyond the housing and neighborhoods where young people before them already lived. Their efforts to find housing throw the consequences of the housing shortfall into sharp relief. Millennials have been on the receiving end of unusually difficult economic conditions since the housing crisis, and now they are on the cutting edge of shifts that will impact cities and neighborhoods going forward. 2.1. Millennials As the first wave of millennials enters their late 20s and early 30s, their choices about where to live are transforming urban neighborhoods. From children of baby boomers to children of immigrants to immigrants themselves, millennials are more racially diverse, more educated, and more technologically savvy than previous generations (Pew Research Center 2014b). Millennials are less likely to drive (Dutzik, Inglis, and Baxandall 2014), more likely to live with their parents (Fry 2015), less likely to marry (Goodman, Pendall, and Zhu 2015), and less likely to buy homes (Xu et al. 2015) – at least so far. And the growing millennial presence in both dense urban neighborhoods and suburban neighborhoods is fueling rapid change in local housing markets (Moos 2015). Millennials’ impact on neighborhoods is amplified by their sheer numbers; across the country, there were 4.1 million more people aged 25-34 in 2014 than in 2004, an increase of 10.3 percent in a decade. Millennials have already demonstrated that they can reshape neighborhoods, and urban scholars, planners, and policymakers are paying close attention to the ways in which millennials’ choices differ from previous generations (Pendall 2012; Bostic 2015). Some of these differences have been framed as the result of changing preferences for housing and neighborhoods (Lachmann and Brett 2015). Yet as Pendall, Bostic, and even Lachmann and Brett point out, millennials’ choices may reflect their constraints more than their preferences (Bleemer et al. 2014; Myers 2016). Millennials face an array of challenges, including unprecedented levels of student debt (Stiglitz 2013; Fry 2014), stunted careers and stagnant wages in the wake of the Great Recession (Elsby, Shin, and Solon 2013), rising income inequality (Pew Research Center 2014a), tightened mortgage underwriting standards (Courchane, Kiefer, and Zorn 2015), and an ongoing affordable housing shortfall (Leopold et al. 2015). Millennials are often considered as a single group. There is much that this generation shares, but there are large and growing disparities between relatively advantaged and less advantaged millennials. In this paper, the contrast between young college graduates and non-college graduates reveals diverging neighborhood trajectories, and leads to insights about the consequences of the housing shortfall. Both groups of young people matter for the future of the region, in slightly different ways. Young college graduates bring skills and advantages that have been shown to spill over into wage growth in the places where they live, even for local residents without college degrees (Moretti 2004). Regions with declining shares of young college graduates are likely to become less economically competitive over time. While the housing and neighborhood attainments of young college graduates are telling indicators of regional economic health, non-college graduates make up 67 percent of the working- aged labor force in the Los Angeles region, and their prospects are also a central concern for the 48 region’s economic future. Considering young college graduates and non-college graduates separately is particularly important because the full cost of the housing shortfall can only be seen in the outcomes for less advantaged groups. 2.2. Economic Disparities and the Divergent Fortunes of Places A growing body of research has traced the increasing disparities between the places where more and less advantaged people live over the last several decades. As the divide between the knowledge and service economies has widened, people have increasingly sorted themselves by skill level into growing, expensive cities or declining but more affordable cities (Moretti 2013). In what Moretti terms the “Great Divergence”, this in turn increases the gap between cities with stronger and weaker economies. The rich cities get richer and the poor get poorer. As Diamond (2013) finds, the sorting of people into diverging places compounds the inequalities between high-skilled and low-skilled people: the already-advantaged people moving to more successful cities gain further advantages from the improving amenities and economic conditions in those places. A related process is occurring at the submetropolitan scale, in neighborhoods. In this case, there has been a shift in the location of growth and decline: suburbs have shown signs of decline after decades of growth and prosperity, and some central urban neighborhoods and inner ring suburbs have rebounded, especially in growing regions. Millennials have played a starring role in this new phase of urban development, as documented by Moos (2015) and his description of the recent “youthification” of cities. This hearkens back to Campbell’s (1966) idea that when there is a generational surge in the number of young people, rental housing in particular will be in high demand. Myers and Pitkin (2009) have taken up this vein of research, describing how the rise and fall of numbers of people reaching age 25 not only impacts rental markets but corresponds to turning points in relative neglect or revitalization of U.S. cities over the past sixty years. Demand shifts clearly impact neighborhoods. But what is the role of the housing supply? Jackson (1985) and others have established how suburban land development - specifically the availability of affordable housing outside urban neighborhoods - spurred the extended cycle of suburban growth that continues in many U.S. cities. Several more recent studies suggest how newly built housing in urban neighborhoods might attract young people to central cities. Using a filtering model, Brueckner and Rosenthal (2009) predict that redevelopment will attract higher-income households back to downtown areas from the suburbs. And Pfeiffer (2015) argues that new construction plays an important role in both attracting and retaining millennials in central urban neighborhoods. Access to housing and neighborhoods is stratified by income and race and ethnicity (Briggs 2005), and disparities in housing outcomes perpetuate economic and racial inequality. Wachter (2015) calls for increased attention to the ways in which tightened mortgage underwriting standards may intensify existing inequalities. Foster and Kleit (2015) find that at the county level, uneven access to housing through housing affordability, homeownership, and subprime lending can lead to increasing economic inequality. In this paper, I study the reflexive relationship between housing availability and inequality at the neighborhood level. 49 3. Research Approach and Methods I assess regional changes in the population of young people aged 25-34 by educational attainment, and changes in their household formation rates and homeownership rates. After evaluating regional shifts in housing opportunities and outcomes for young people, I use multivariate regression to examine neighborhood-level changes in the number of young college graduates or non-college graduates as a function of housing opportunities, tenure, neighborhood characteristics, and indicators of the 1990s economic restructuring and the 2000s housing boom and bust. The time period for the changes in the multivariate analysis, 2007 to 2012, reflects changes in the aftermath of the housing crisis, and uses the latest available data at the census tract level. I study changes in the neighborhoods where young adults aged 25-34 live for two main reasons. First, because the millennial generation is larger than previous generations, they represent a growing segment of housing demand. Second, because young adults are starting out in their housing careers, their housing outcomes offer a window into the consequences of recent changes in housing markets; their outcomes are less influenced by the legacy of past housing market conditions than the housing outcomes of older generations. Dividing young adults into college graduates and non-college graduates provides insights about the neighborhood-level impacts of increasing income inequality. Using educational attainment as a proxy for relative advantage and disadvantage is inspired by the research on the divergence of cities described above. 3.1. Data For the regional-level analysis I use 1990, 2000, and 2010 Census and 2005-09, 2008-12, and 2009-13 American Community Survey (ACS) tables from the U.S. Census Bureau, downloaded from Social Explorer. I supplement the tables with CSA-level microdata from the 1990 and 2000 Census and the 2005-07, 2009-10, and 2011-13 ACS, downloaded from IPUMS. For the neighborhood-level analysis I use 1990 and 2000 Census and 2005-09 and 2010-14 ACS tables, interpolated into 2010 Census Tract boundaries using the Longitudinal Tract Database (Logan, Xu, and Stults 2012). I also include measures of employment density from the U.S. Census Bureau’s ZIP Code Business patterns dataset, and the housing boom and bust indicators are provided by the U.S. Department of Housing and Urban Development’s Neighborhood Stabilization Program. 50 3.2. Density Quintiles As shown in Figure 3.1, I divide census tracts in the Los Angeles region into quintiles by their 1990 population density, which I use to characterize urban form. Though population density is only one dimension of urban form (Galster et al. 2001), Malpezzi and Guo (2001) have found that it is a fairly good proxy for other elements of urban form on the regional level. Here I use population density to compare the density of tracts within a single region. I define the density quintiles in 1990 and then use these consistent geographic boundaries to measure subsequent changes in various areas of the region by density level. I use this measure of relative population density to understand the changing spatial patterns of housing opportunities and the residential locations of young people. Density quintiles are defined by a characteristic that varies over space rather than by relative location within the region, so they fit Los Angeles’ polycentric urban structure better than measures of distance from a central point. Density quintiles also offer graduated categories for comparison instead of the simple dichotomy between the central city and the suburbs. The central city/suburb distinction is especially problematic in the case of Los Angeles, where the City of Los Angeles encompasses over a fifth of the population of the region, with neighborhoods as diverse as the urban core and the suburban fringe. Figure 3.1 Map of census tracts in the Los Angeles-Long Beach Combined Statistical Area by 1990 population density quintile 51 4. Regional Changes in Young College Graduates and Non-Graduates As the larger millennial generation entered their late 20s and early 30s, the number of young people grew in Los Angeles, but not nearly as much as the growth in California and the country (shown in Figure 3.2). From 2000 through 2007, Los Angeles’ 0.21 percent annual growth in young people aged 25 to 34 surpassed California’s annual rate of 0.15 percent and kept pace with the 0.20 percent national average. But after the crisis, Los Angeles’ 0.12 percent annual growth in young adults lagged far behind California’s 0.87 percent growth and the nation’s 0.92 percent growth. Figure 3.2 Change in population aged 25-34 and ratio of non-graduates’ incomes to college graduates’ personal incomes in the Los Angeles region Data Sources: Census 2000; ACS 2005-09, 2010-14; IPUMS Census 1990, 2000; ACS 2005-07, 2008-10, 2011-13. 4.1. Rising Income Inequality for Young People by Educational Attainment Consistent with national trends, incomes for young people in the Los Angeles region declined over the past several decades, and income inequality increased between young people with and without college degrees. Between 1990 and 2012, average personal incomes for both groups declined substantially, due to a combination of lower wages and decreasing employment. The average personal income for college graduates aged 25-34 declined 22 percent, from $59,000 to $46,000 in inflation-adjusted 2013 dollars. The average personal income for young adults without college degrees declined 31 percent, from $31,000 to $22,000. In 1990, people aged 25-34 without college degrees made an average of 53 percent of the income of young college graduates. By 2012, young adults without college degrees earned an average of only 48 percent of the income of young college graduates. As a result, the non-college educated young suffered both an absolute decline in income and a further decrease relative to the college-educated. Even this may be an understatement of their inequality, because the trends are calculated only for those still residing in the region; others who are most impacted may have already departed to opportunities other states that fared better in the recession. 52 4.2. Changing Populations of Young People by Educational Attainment In the early part of the 2000s, the Los Angeles region surpassed the state and the nation in both steeper increases in the number of young college graduates and deeper decreases in non-college graduates. After the housing crisis, the growth in young college graduates faltered relative to the state and the nation, and the declines in young non-college graduates continued in Los Angeles despite growth elsewhere. Figure 3.3 Percent change in college graduates and non-college graduates aged 25-34 in the U.S., California, and Los Angeles region Data Sources: Census 2000; ACS 2005-09, 2010-14. Notice that the percent changes in the non-college graduates are smaller than the percent changes in the college graduates, but they still represented larger aggregate changes since non-college graduates outnumber college graduates by such a large ratio. This contrast between the in-migration of young college graduates and out-migration of young non- college graduates falls in line with Moretti (2013) and Diamond’s (2013) work on the divergence of skilled and unskilled workers in metropolitan areas: young college graduates could make it in the overheated housing market, and young people without college degrees simply could not keep up. This divergence is due in part to higher college graduation rates of existing residents, but also to increasing in-migration of young college graduates and increasing out-migration of young non- graduates, as well as the decline in immigration and shifts in immigrant composition. These shifts add up to considerable changes in the average educational attainment levels of young people in the Los Angeles region, accompanied by striking changes in household formation and homeownership rates. Among people aged 25-34, college graduates increased slightly from 21.5 percent to 23.2 percent between 1990 and 2000, but by 2012 a full 30.3 percent of people aged 25- 34 were college graduates. Figure 3.4 shows changes since 1990 in population, households, renters, and owners for young college graduates and non-college graduates. Household formation rates fell for both young college graduates (from 49.9 percent in 1990 to 42.9 percent in 2012) and non- college graduates (from 38.3 percent in 1990 to 30.3 percent in 2012). Homeownership rates fell 53 even more sharply, from 41.7 percent of young college graduates in 1990 to 31.0 percent in 2012, and from 30.4 percent of young non-graduates in 1990 to 21.4 percent in 2012. Figure 3.4 Changes since 1990 in the total population, households, renters, and owners of people aged 25-34 by educational attainment in the Los Angeles region For young college graduates, these changing rates add up to a 29.7 percent increase in population from 1990 through 2012 and a more modest 11.6 percent increase in households. Young college graduate owners decreased by 16.8 percent, and renters increased by 31.9 percent. For young non- graduates, their population decreased by 18.4 percent, households decreased by 35.5 percent, owners decreased by 54.7 percent, and renters decreased by 27.1 percent over the same period. Taken together, the number of renters aged 25-34 decreased by 13.5 percent or roughly 105,000 from 1990 through 2012, and the number of owners aged 25-34 decreased by 42.3 percent, a loss of 165,000 owners. 5. Changes in the Neighborhoods Where Young College Graduates and Non- Graduates Live Figure 3.5 shows annual changes in the number of college graduates and non-college graduates aged 25-34 by density quintile. Young college graduates increased in all density quintiles in both periods. In the early 2000s, the number of young college graduates increased the most in both the densest and the least dense parts of the region. From 2007 to 2011, annual increases were much lower in all density quintiles, and the biggest increases came in the least dense areas. Young non-college graduates declined in both periods in all parts of the region except the least dense areas. Decreases were particularly steep in the densest quintile, with deeper declines from 2007 to 2011. These patterns of growth in young college graduates and decline in young non-college graduates add up to dramatic shifts in neighborhood populations, especially in the densest neighborhoods with large populations of young adults. 54 Figure 3.5 Net changes in population aged 25-34 by educational attainment and density quintile in the Los Angeles region 5.1. Multivariate Analysis of Neighborhood-Level Increases and Decreases in Young College Graduates and Non-College Graduates It is clear that on a basic level, some of the largest increases in young college graduates occurred in the same types of places where some of the largest decreases in young non-graduates happened. This goes only partway towards addressing my first research question: Q1. Did young college graduates and non-college graduates increasingly live in different types of neighborhoods? It also does not address my second question: Q2. Did the housing stock play a role in influencing the places where young people lived? I use multivariate regression in order to investigate how changes in the residential locations of young college graduates and non-graduates were related fine-grained differences in neighborhood characteristics and housing availability. I analyze census tract-level changes in young college graduates or non-college graduates as a function of housing opportunities, lagged neighborhood characteristics, and indicators of the 1990s economic restructuring and 2000s housing boom and bust. I use ordinary least squares regression to estimate the following lagged dependent variable model: ∆𝑌 𝑖 2007 𝑡 𝑜 2012 = 𝛽 0 + 𝛽 1 𝑌 𝑖 2007 + 𝛽 2 𝑃 𝑖 2007 + 𝛽 3 𝐻 𝑖 2007 + 𝛽 4 𝑁 𝑖 1990 + 𝛽 5 𝑅 𝑖 1990𝑠 + 𝛽 6 𝐵 𝑖 2000𝑠 + 𝜀 𝑖 Where ∆𝑌 𝑖 2007 𝑡𝑜 2012 is the change in the number of college graduates or non-college graduates aged 25-34 in a census tract between 2007 and 2012; 𝑌 𝑖 2007 is the initial number of young college graduates or non-college graduates in a tract; 𝑃 𝑖 2007 is the total population of a tract in 2007; 𝐻 𝑖 2007 is the number of housing units made available on the market during the initial time period, preceding the change measured in the dependent variable; 𝑁 𝑖 1990 is a vector of neighborhood characteristics, lagged to 1990, which includes indicators of the built environment, housing and labor market conditions, socioeconomic status, and demographic characteristics; 𝑅 𝑖 1990𝑠 includes 55 indicators of the 1990s economic restructuring; 𝐵 𝑖 2000𝑠 includes indicators of the 2000s housing boom and bust; and 𝜀 𝑖 is an error term. Table 3.1 offers summary statistics for the dependent and independent variables in the model, as well as the correlations of each variable with population density. 5.2. Dependent Variables: Tract Level Changes in Young Adults by Educational Attainment The dependent variables for the multivariate analyses are the census tract-level changes in the number of college graduates or non-college graduates aged 25-34 between 2007 and 2012, in separate regressions. Change is measured by counts: the simple increase or decrease in the number of young graduates or non-graduates in a census tract. Measuring change in counts has the advantages that (a) unlike percentage point changes, counts are not influenced by changes occurring in other population groups, and (b) unlike percent changes, counts are not influenced by the initial population of the group in a tract, which can lead to misleadingly high percent changes when the initial population is small. The disadvantage of counts is that they are not normalized by census tract size, which I deal with by including the population of the census tract as a control variable in the regression. Changes in both young graduates and non-graduates are generally normally distributed, with a slightly higher concentration of values around the mean. Changes in both young grads and non-grads are generally normally distributed, with a slightly higher concentration of values around the mean (not enough to violate the assumptions of normality in the OLS model and warrant a different modeling strategy). Post-estimation tests do not indicate that extreme values are driving the results. I use robust standard errors clustered at the PUMA level to account for the spatial dependencies in the data. The changes measured by the dependent variables represent shifts for people and for places: changes in the neighborhoods where young people live, and shifts in the populations of young people within neighborhoods. When interpreting the results, it is also crucial to consider that young college graduates are a growing group during this period, and young non-graduates are a declining group. As an increasing group, young graduates are likely to fill up the neighborhoods where previous generations of young graduates lived and spill out into new neighborhoods. Filtering theory leads us to expect that some of the places with the biggest increases in young graduates may also have the largest decreases in young non-graduates; young college graduates may crowd out young non-graduates by taking up housing that would have been available to them earlier. And just as young graduates are likely to increase in the most desirable neighborhoods available, near more advantaged neighborhoods (Guerrieri, Hartley, and Hurst 2013), young non-graduates are likely to remain in the least desirable neighborhoods with relatively affordable housing (Glaeser, Gyourko, and Saks 2006). Since attracting young graduates and retaining young non-graduates are in some ways opposing dynamics, I expect the results to reflect those differences. In the case of an increasing group, a positive coefficient can be interpreted as a sign that young college graduates are more likely to move into new neighborhoods with the corresponding characteristic. In the case of a decreasing group, though, a positive coefficient can be interpreted as a sign that young people are less likely to move away from those neighborhoods, or that there is still space for them to move into those neighborhoods as compared with neighborhoods where there is no longer space for young people without college degrees. I still expect housing opportunities to matter for both groups, but with a slightly different inflection; housing opportunities enable young college graduates to move into new 56 Table 3.1 Descriptive Statistics and Correlations with Density Geography: Los Angeles-Long Beach-Riverside Combined Statistical Area Unit of Analysis: 2010 Census Tracts, N = 3798 Mean Std. Dev. Min. Max. Correl. with Ln Pop. Dens. Percent College Graduates Aged 25-34, 2005-09 4.4 4.6 0.0 55.0 .13 Change in Percent College Graduates Aged 25-34, 2005-09 to 2010-14 0.2 2.7 -55.0 21.4 .05 Change in College Graduates Ages 25-34, 2005-09 to 2010-14 16 133 -2092 1601 -.03 Percent Non College Graduates Aged 25-34, 2005-09 10.6 5.0 0.0 27.2 .31 Change in Percent Non College Grads Aged 25-34, 2005-09 to 2010-14 -0.7 3.6 -21.0 16.2 -.07 Change in Non College Graduates Ages 25-34, 2005-09 to 2010-14 -13 200 -1323 1406 -.18 Percent college graduates, 1990 20.8 14.5 0.4 73.0 -.12 Employment density, 2000 (logged) 24.6 2.2 3.6 29.8 .19 Labor force participation rate, 1990 67.8 8.6 7.7 89.7 .12 Unemployment rate, 1990 7.0 3.9 0.0 26.0 .29 Median gross rent, 1990 1,235 296 18 1,786 -.11 Housing vacancy rate, 1990 7.2 8.0 0.0 73.6 -.46 Population density, 1990 (logged) 8.4 1.5 -3.4 11.6 — Housing age, 1990 Percent built in past decade 24.6 22.9 0.0 99.5 -.48 Percent built 11-30 years ago 37.6 20.7 0.4 100.0 -.04 Percent built over 30 years ago 37.9 27.0 0.0 99.3 .44 Percent buildings with 10 or more units, 1990 17.8 19.7 0.0 94.9 .42 Percent public transit commutes, 1990 4.8 7.3 0.0 62.9 .49 Percent foreign born, 1990 26.2 17.0 1.6 81.8 .56 Race and ethnicity, 1990 Percent Non Hispanic White or Other 51.8 29.4 0.2 97.5 -.49 Percent non-Hispanic Black or African American 7.5 14.4 0.0 93.4 .18 Percent non-Hispanic Asian or Pacific Islander 8.4 9.4 0.0 80.9 .23 Percent Hispanic 31.6 25.2 1.5 98.6 .39 Age, 1990 Percent aged under 18 26.5 7.4 0.1 43.3 -.01 Percent aged 18-64 63.3 6.5 9.2 97.1 .17 Percent aged 65 or over 10.2 7.1 0.1 90.7 -.15 1990s economic restructuring indicators Change in percent manufacturing workers, 1990 to 2000 -4.4 4.7 -34.5 24.2 -.14 Change in average household wage or salary income, 1990 to 2000 -1,925 14,531 -92,006 132,668 -.09 2000s housing boom and bust indicators Units built 2000 to 2004 (logged) 2.9 2.1 0.0 8.4 -.37 Percent high-cost mortgages, 2004 to 2006 24.4 12.4 0.0 80.0 .07 Percent house price change, 2006 to 2009 -30.2 10.6 -54.6 -6.0 .30 Units vacated 2005 to 2009 (logged) 5.7 0.7 0.0 8.0 .25 Units built 2005 to 2009 (logged) 1.7 1.9 0.0 7.7 -.34 Homeownership rate, 1990 55.9 24.7 0.2 98.4 -.55 57 neighborhoods, while they constrain the neighborhoods where young non-graduates can afford to live within the region. Table 3.2 shows estimates of the marginal effects of neighborhood characteristics and housing opportunities on changes in college graduates and non-graduates aged 25-34. For models 1 and 2, the dependent variable is the tract-level change in college graduates aged 25-34 between 2007 and 2012. For models 3 and 4, the dependent variable is the tract level change in non-college graduates aged 25-34 during the same period. Models 2 and 4 include interactions and quadratic terms which improve model fit, while models 1 and 3 do not. When controlling for the initial population of the tract, the initial population of young college graduates or non-graduates represents their relative proportion of the population within the tract. Since any measurement error in the initial 2007 population of young college graduates or non- graduates is also incorporated in the measurement of the dependent variable, this coefficient simply represents a baseline control in the lagged dependent variable model rather than a meaningful relationship. 5.3. Neighborhood Characteristics The neighborhood characteristics included in the analysis consist of variables related to socioeconomic status (educational attainment), labor market conditions (proximity to jobs, labor force participation, and unemployment), housing market conditions (vacancy rates and rents), the built environment (population density, housing age, housing in multi-unit buildings, and public transportation), and demographic characteristics (immigration, race and ethnicity, and age). Like homeownership rates, other neighborhood characteristics are measured in 1990 to reduce endogeneity, and many of these measures are remarkably stable over time. This analysis is limited by the lack of school and crime data, though those dimensions are to some extent captured in both rents and the proportion of neighborhood residents with college degrees. In addition to standard built environment, housing and labor market, socioeconomic, and demographic neighborhood characteristics, I also include indicators of the 1990s economic restructuring and 2000s housing boom and bust to see if these events have repercussions from 2007 to 2012. To address the first question about whether college graduates’ and non-graduates’ neighborhood trajectories diverged, I examine the contrasts between the results for young college graduates and non-graduates. Some of the neighborhood characteristics are pull factors which are likely to attract people to the neighborhood, and others are push factors which are likely to repel people. It appears that pull factors play a bigger role in neighborhood increases in college graduates, while push factors have stronger relationships with changes in non-graduates. The pull factors in the models (including socioeconomic status, recently built housing, employment density, and public transit use) all have the expected, and rather strong, positive relationships with changes in young graduates. For non-graduates, these pull factors are associated with weaker (recently built housing), insignificant (employment density, public transit use), and even negative effects (socioeconomic status). Similarly, some of the push factors that appear to repel or have little effect on young graduates have the opposite effect on changes in non-graduates. Places that were most negatively impacted by the 1990s economic restructuring, high-cost mortgage lending during 58 Table 3.2 Estimated Marginal Effects of Neighborhood Characteristics on Census Tract Changes in College Graduates and Non-Graduates Aged 25-34, 2007 to 2012 Dependent Variables: Tract Level Change in: 2007 to 2012 Unit of Analysis: 2010 Census Tracts ME SE ME SE ME SE ME SE Initial level of DV, 2007 -0.31 *** 0.06 -0.39 *** 0.07 -0.63 *** 0.03 -0.67 *** 0.03 Total population, 2007 0.00 0.00 0.00 0.00 0.04 *** 0.01 0.05 *** 0.01 Percent college graduates, 1990 50.13 41.94 606.84 *** 140.27 -361.03 *** 43.31 -332.01 *** 79.39 Interaction with homeownership rate -776.76 *** 160.18 -64.63 108.07 Employment density, 2000 (logged) 2.58 1.31 -24.94 ** 7.78 0.69 1.49 -13.44 9.01 Squared employment density 0.69 *** 0.19 0.37 0.23 Labor force participation rate, 1990 91.69 69.30 80.65 59.32 42.45 92.65 22.93 91.51 Unemployment rate, 1990 144.54 105.62 98.00 99.11 47.37 182.36 5.92 175.21 Median gross rent, 1990 0.00 0.02 -0.01 0.01 0.01 0.02 0.01 0.02 Housing vacancy rate, 1990 -156.10 ** 53.88 -170.40 *** 50.03 -211.83 *** 49.36 -218.96 *** 48.41 Population density, 1990 (logged) -10.32 ** 3.26 -10.08 ** 3.13 -11.65 * 5.40 -10.13 5.31 Housing age, 1990, Reference category: Percent built 11-30 years ago Percent built in past decade 34.39 17.81 33.54 17.94 -1.68 22.21 11.91 21.04 Percent built over 30 years ago 41.53 ** 13.64 25.38 * 11.08 -1.56 15.26 -12.37 15.32 Percent 10 or more units, 1990 39.78 27.94 -2.22 25.60 12.10 28.57 -1.60 28.59 Percent public transit commutes, 1990 -32.25 81.94 241.75 * 96.52 -179.95 * 81.33 24.27 100.97 Percent foreign born, 1990 22.98 53.37 306.31 *** 72.01 50.47 49.13 127.78 85.40 Squared percent foreign born, 1990 -363.17 *** 96.70 -86.88 114.71 Race and ethnicity, 1990, Reference category: Percent Non-Hispanic White or Other Percent non-Hsp. Black or African American 16.18 20.52 -11.77 19.25 -21.53 32.35 -38.10 29.85 Percent non-Hsp. Asian or Pacific Islander 36.38 46.93 -22.04 40.46 -62.04 40.69 -90.18 * 39.98 Percent Hispanic -9.73 29.54 -44.33 22.71 36.20 33.55 31.07 34.13 Age, 1990, Reference category: Percent aged 18-64 Percent aged under 18 -374.32 ** 141.66 -94.56 130.93 -322.06 ** 105.80 -197.70 105.71 Percent aged 65 or over -112.43 107.76 -37.53 93.13 -220.56 113.09 -167.77 109.09 1990s economic restructuring indicators Change in pct manufacturing workers, 1990s -46.29 42.41 -38.58 43.17 139.88 * 63.77 132.67 * 61.78 Change in avg hh wages or salaries, 1990s 0.01 0.15 0.04 0.14 -0.43 * 0.21 -0.48 * 0.23 2000s housing boom and bust indicators Units built 2000 to 2004 (logged) 1.17 1.17 1.47 1.03 1.53 1.71 1.16 1.70 Percent high-cost mortgages, 2004 to 2006 -97.38 ** 34.74 -136.98 *** 37.70 339.83 *** 61.12 333.05 *** 60.27 Percent house price change, 2006 to 2009 6.78 41.46 28.37 41.89 -252.19 *** 63.45 -259.55 *** 64.84 Units vacated 2005 to 2009 (logged) 14.97 * 6.49 54.82 *** 14.15 27.04 *** 7.44 129.32 *** 17.79 Interaction with homeownership rate -64.14 ** 20.06 -171.08 *** 23.84 Units built 2005 to 2009 (logged) 6.87 *** 1.53 14.19 *** 4.20 4.11 * 1.99 -3.94 4.53 Interaction with homeownership rate -12.77 * 6.07 14.39 7.47 Homeownership rate, 1990 -54.44 28.79 396.37 ** 146.33 -83.03 ** 29.82 1182.00 *** 170.38 Squared homeownership rate 102.97 * 51.33 -264.15 *** 70.27 Controls for group quarters and public housing yes yes yes yes N 3798 3798 3798 3798 adjusted R 2 0.131 0.175 0.280 0.297 Robust standard errors clustered by Public Use Microdata Area in parentheses. * p<0.05, ** p<0.01, *** p<0.001 Model 1 Model 2 Model 3 Model 4 College Graduates Non College Graduates 59 the boom, and falling house prices during the bust retained non-graduates relative to less affected neighborhoods. These contrasts support the idea that the landscape of opportunity and choice for young graduates looks more like a landscape of scarcity and constraint for young non-graduates. While young college graduates can often pay the higher prices to have a choice of neighborhoods, young non-graduates are more constrained in their ability to move towards better neighborhoods. These findings suggest that young non-graduates may be able to find footholds only in the leftover housing in the least desirable neighborhoods, and some cannot even find that. Divergent trajectories of young graduates and non-graduates are likely to lead to divergent trajectories of neighborhoods as well. When the most advantaged young people increasingly move to the most advantaged neighborhoods, those neighborhoods are swept up in a cycle of redevelopment and renewal that attracts even more advantaged young people. When the least advantaged young people end up in the least desirable neighborhoods, those neighborhoods may become even more disinvested as their residents’ struggle to hang on with declining wages in an overheated housing market. Socioeconomic Status Percent College Graduates. Following the metro-level divergence literature, I use neighborhood educational attainment levels as a proxy for neighborhood advantage and socioeconomic status. Based on Moretti (2013) and Diamond’s (2013) findings that people are increasingly sorting into different metro areas by education level, I expect that places with higher educational attainment levels in 1990 will have increases in young college graduates and decreases in young non-graduates from 2007 to 2012, in a divergence of trajectories for neighborhoods and young people. Neighborhoods with higher levels of educational attainment in 1990 saw increases in college graduates and decreases in young non-graduates, demonstrating neighborhood-level dynamics that echo the metro-level findings of Moretti (2013) and Diamond (2013). As neighborhoods that start with higher education levels gain highly educated young people and lose less educated young people, neighborhoods in the region become more segregated by education level. And young college graduates and non-graduates become more segregated from each other by neighborhood, pulling further away from each other in a linked pair of positive and negative feedback loops. Housing tenure plays a role in this as well; increases in young graduates were only associated with higher levels of educational attainment in neighborhoods with high proportions of rental housing, as shown in Figure 3.6. This is related to other variables discussed below, which show strong increases of young college graduates in neighborhoods with more rental housing. 60 Figure 3.6 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by percent college graduates interacted with tract homeownership rate Labor Market Conditions Employment and access to jobs are important pull factors for neighborhoods. I measure the relationship between changes in young people and labor market conditions using indicators of employment density, labor force participation, and unemployment. Employment Density is a measurement of the relative proximity to jobs in a census tract. Employment density is calculated using an inverse distance weighted gravity model, using U.S. Census Bureau ZIP Code Business Patterns data, interpolated to census tracts using a U.S. Census Bureau ZIP code to census tract crosswalk file, with the matrix of distances between census tracts provided by the National Bureau of Economic Research. Because of data limitations, employment density is calculated in 2000 rather than 1990. Employment density is highly stable over time (Redfearn 2009), and even when measurements of endogenous changes in employment density are included in the model, they have no measureable relationship with the dependent variables and make no difference to the other results. Employment density has a u-shaped quadratic relationship with changes in young college graduates, as shown in Figure 3.7. Increases are the steepest (and most precisely predicted) in neighborhoods with higher levels of access to employment. Together with the result for public transit use below, this suggests that millennial college graduates have a higher tendency to live closer to jobs than previous young college graduates. Employment density does not have a statistically significant association with changes in young non-graduates in either Model 3 or 4. This is a case where young 61 college graduates can be seen responding to a neighborhood pull factor which seems to have little effect on the locations of young non-graduates. Figure 3.7 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by employment density Housing Market Conditions Median Gross Rent. Rent levels are indicators of the relative market value of a neighborhood’s housing and amenities. As mentioned above, rents may capture the value of some of the variables that are missing from the model: crime rates, school quality, and other neighborhood characteristics that are even more difficult to observe. Rents are highly correlated with home values, and I include rents rather than values because less than a third of people aged 25-34 are homeowners. To the extent that rents measure the desirability of a neighborhood, they may encourage young people to move in, but higher rents also mean lower affordability. Median gross rent has no measurable relationship on changes in either young college graduates or non-graduates. This is not a surprising result, since many of the factors that determine rents are included in the model. But it does assuage concerns that omitted variables such as school quality or crime rates, also capitalized in rents, might have strong associations with the dependent variables. Housing Vacancy Rates. High vacancy rates can signal an oversupplied housing market, rapid turnover, or the presence of a high proportion of low quality and substandard housing units. The first possibility is unlikely in the Los Angeles region housing market during this period, though there may be places on the exurban fringe where overbuilding could lead to oversupply. Renters move 62 more often than homeowners, so controlling for homeownership rates is essential when analyzing the effects of vacancy rates. Higher turnover rates often indicate neighborhood instability, though they can also mean greater housing availability (a moot point given that this variable is lagged to 1990, and housing availability is included in the model). On the whole, these three possible reasons for high vacancy rates are associated with negative neighborhood characteristics, and I expect high vacancy rates to be associated with net decreases in both young graduates and non-graduates. Housing vacancy rates have a negative association with changes in both young college graduates and non-graduates, across all models. Built Environment Population Density is the main indicator of the urban (or suburban, or exurban) environment. Population density can also be thought of as a proxy for location within the region, since on a wide scale the Los Angeles region is densest at the center. As discussed above, density better accounts for smaller-scale polycentricity than measures of centrality. Preferences for walkability and dense urban environments are often used to explain the rapid increases in young people in relatively dense neighborhoods (Lachmann and Brett 2015), though on the whole millennials have increased more in the suburbs than in urban neighborhoods. As shown in Table 2, many of the other neighborhood characteristics are correlated with density: recently vacated units, rental housing, multifamily housing, older housing, public transit use, unemployment, the intensity of the housing bust, immigrants, and Hispanic, Asian, and Black residents. Conversely, recent construction, homeownership, housing vacancy rates, and Non- Hispanic Whites are correlated with less dense environments. Even the presence of young non- graduates is correlated with density, though changes in young graduates or non-graduates are not particularly correlated with density. Given all these interrelated factors, it is essential that any argument about preferences for density, urban neighborhoods, or suburban life take these other variables into account. What looks like a preference for density could instead be a preference for renting, or multifamily housing, or public transit. Accordingly, I do not expect to see strong associations between population density and changes in young graduates and non-graduates. When controlling for other variables, changes in young graduates are negatively associated with density. This result holds for young college graduates even when tracts from the least dense quintile are excluded from the analysis, so the relationship is not merely an artifact of rapid growth on the exurban fringe. This negative relationship may point to a preference for open space or larger lot sizes, and runs counter to the narrative that young people have increasing preferences for denser urban environments. Changes in young non-graduates are negatively associated with density to a similar degree, though the coefficient is not quite statistically significant. The relationship disappears entirely for non- graduates when the least dense quintile tracts are excluded, which suggests that any relationship between density and changes in young non-graduates is driven by exurban growth rather than a consistent pattern throughout the region. Housing Age. Though a younger housing stock is generally an advantage for neighborhoods relative to middle-aged housing, old housing is likely to spur increases in young college graduates. This somewhat counterintuitive prediction is in part due to the potential for redevelopment of older 63 housing delineated by Brueckner and Rosenthal (2009). Old housing is also associated with gentrification and renovation by homeowners (Myers 1984; Galster 1987). And old housing that remains may be of higher quality than middle-aged housing due to survivor bias: the housing that is not demolished as it ages is likely to be well constructed and better maintained (Corgel and Smith 1981). Variables related to the year when housing was built are included in the model in two distinct ways, measuring different aspects of the housing stock. First, the amount of recently-built housing in a census tract is measured by the log number of housing units built from 2005 to 2009 (as described above), and the amount of housing built during the housing boom is measured by the log number of housing units built from 2000 to 2004. Second, housing age is included as a lagged neighborhood characteristic, to measure the age structure of the housing stock in 1990. There are moderate correlations among these two types of measures (the strongest is .39, between the amount of housing built between 2000 to 2004 and the percentage of housing built less than 10 years old in 1990; the weakest is -.27, between the amount of housing built between 2005 and 2009 and the percentage of housing over 30 years old in 1990). However, there is enough variation between these concepts so that the coefficients for each of these measures remain stable when the other measures involving year built are removed from the models. As compared with middle-aged housing, the presence of old housing is associated with increases in young college graduates. This fits with the idea that old housing in growing regions provides opportunities for upgrading and gentrification (Glaeser, Gyourko, and Saks 2006; Brueckner and Rosenthal 2009). Changes in young non-graduates do not appear to be associated with the age of housing. Public Transit Use. Millennials use public transit more than previous generations, for reasons from the difficulty of affording cars to higher regulatory barriers to obtaining drivers’ licenses to changing preferences for driving (Dutzik, Inglis, and Baxandall 2014). This would suggest increases in both young graduates and non-graduates in neighborhoods with more public transit. However, the relationship between changes in young non-graduates and public transit brings up an important complication in the dependent variable. A good proportion of non-college graduates aged 25-34 are immigrants, but immigration to Los Angeles decreased during this period. Recent immigrants use transit at high rates, but their usage decreases as they remain longer in the U.S. (Myers 1996; Blumenberg and Shiki 2007). And if young non-graduates include fewer immigrants, the importance of transit may decline for this group, consistent with the findings of Blumenberg and Evans (2010). In future research, I plan to compare the housing and neighborhood outcomes of recent immigrants with native-born young people. The percentage of commutes made by public transit rather than by car or by other modes is not a perfect measurement of public transit access in a neighborhood: people may take public transit even where bus or rail routes are scarce, and people may drive to work even in transit-dense neighborhoods. Still, on the wider regional scale, transit use is tightly linked with access to transit. Transit use data has the advantage of being readily available in the census and ACS, unlike more precise measures of transit access. Use data also gives a sense of the quality as well as the density of transit access. A neighborhood may have a transit station with frequent stops, but if the buses or trains do not connect people with the places where they work, or if they are considered dangerous, expensive, or inconvenient, use will be lower. 64 In Model 1, there is no visible relationship between public transit use and changes in young college graduates. The inclusions of interactions and quadratics in Model 2 reveals that young graduates increased more in neighborhoods with more public transit use. In Model 3, neighborhoods with higher rates of public transit use appear to have greater decreases in young non-graduates. In model 4, with interactions and quadratics included, the coefficient for public transit use is much reduced and no longer even approaches significance. It seems that in this case, a better fitting model reveals the weakness of the relationship. Demographic Characteristics Immigration. Neighborhoods with immigrant populations have been associated with subsequent gentrification (Vicino, Hanlon, and Short 2011; Hwang 2015), as immigrants have often revived housing demand in disinvested neighborhoods (Myers and Liu 2005; Saiz 2007) and started businesses that enrich a neighborhood’s retail landscape (Logan, Alba, and Stults 2003; Brown- Saracino 2009). I expect that this pattern will hold in the Los Angeles region, and immigrant neighborhoods will see larger increases in young college graduates. On the other hand, Saiz and Wachter (2011) and Crowder, Hall, and Tolnay (2011) have found native-born residents are more likely to move away from neighborhoods with influxes of immigrants or large immigrant populations. Increases in young college graduates (but not non-graduates) are associated with the presence of immigrants in 1990, though the relationship becomes indeterminate in neighborhoods with very Figure 3.8 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by percent foreign born 65 high proportions of immigrants (see Figure 3.8). This quadratic relationship of immigrant populations with changes in young college graduates may reflect the competing influences of immigrants on neighborhood change found in the previous studies discussed above. While the presence of immigrants in a neighborhood might make it more attractive for young college graduates, at some point large concentrations of immigrants represent a more ambiguous factor in their neighborhood choices. Though at least in Los Angeles, neighborhoods with high concentrations of immigrants are not a clear deterrent to increases in young college graduates, some of whom may be immigrants or the children of immigrants themselves. Race and Ethnicity. The relationship of immigration and neighborhood change is mediated by the related dimension of race and ethnicity. The persistent disparities between White and Black populations form a backdrop for the newer disparities between Whites and the growing Hispanic and Asian populations. There is evidence from Chicago that neighborhoods with predominantly Black and Latino populations are less likely to gentrify (Hwang and Sampson 2014). However, the racial and ethnic relationship with neighborhood change is less clear-cut in Los Angeles. Bader and Warkentian (2016) find that Los Angeles has a higher proportion of durably integrated neighborhoods than other major cities, perhaps due to Los Angeles’ long history of immigration and racial diversity. Alba, Logan, and Stults (2000) point to smaller distinctions between urban and suburban contexts to explain the slightly more fluid connections between race and place in Los Angeles as compared with New York and Chicago. The rapid declines of Black residents and transition of Black neighborhoods to Hispanic neighborhoods suggests a racial dynamic to the process of neighborhood change in Los Angeles, which is an avenue for future research. That said, racial and ethnic residential mobility patterns have been remarkably stable over time, and generally reinforce existing inequalities in neighborhood residential locations (Crowder and South 2005; Sampson 2009; Sharkey 2012). Race and ethnicity at first appear to have little to do with changes in young people, with the exception of declines in young people without college degrees in neighborhoods with a higher proportion of Asian or Pacific Islanders. This may reflect the widespread growth of Asian immigrant enclaves in Los Angeles suburbs and the high education levels of the more recent Asian immigrants who are replacing previous generations of White residents with lower education levels (Li 2009; Bader and Warkentien 2016). In addition, the initial levels of young college graduates are positively correlated with White and Asian populations, and the initial levels of young non-graduates are positively correlated with Black and Hispanic populations. When the initial population is removed from the model, racial composition still shows no significant effect on changes in young college graduates. However, without the initial population, declines in young non-graduates in neighborhoods with larger Hispanic and Black populations show up. The racial dimensions of the neighborhood change are mostly outside the scope of this paper, but I plan to use a similar framework to assess racial inequalities in neighborhood trajectories and housing opportunities. 1990s Economic Restructuring In addition to neighborhood characteristics, I include indicators of two major shifts - the economic restructuring of the 1990s and the housing boom and bust of the 2000s. Beginning as early as the 1950s, the U.S. economy underwent a transition away from manufacturing and towards a knowledge and service-based economy. This shift intensified during the 1990s, as employment in manufacturing fell from 17.7 to 14.1 percent of workers nationwide, and employment in knowledge and service 66 sectors increased from 39.6 to 51.9 percent of workers. Los Angeles had a strong manufacturing sector in 1990 (19.6 percent of workers), thanks in part to the aerospace industries which were about to be hit hard by the end of the Cold War. In the Los Angeles region, employment in manufacturing fell from 1.36 million jobs in 1990 to 1.02 million jobs in 2000, a 24.8 percent decline in the course of a decade. Knowledge and service jobs grew by nearly a third, from 2.81 million in 1990 to 3.76 million in 2000, all while employment growth stagnated and wages failed to keep up with inflation. To measure the relative effects of the economic restructuring in neighborhoods across the Los Angeles region, I use census tract level changes in the percent of workers employed in manufacturing, and changes in the average household wage or salary income between 1990 and 2000. I expect that neighborhoods more affected by economic restructuring may have continued to decline afterward, and could experience declines in young people even a decade later. The echoes of the 1990s economic restructuring appear to affect changes in young non-graduates but not college graduates. As expected, there is a direct relationship between losses of manufacturing jobs during the 1990s and the loss of young non-graduates. But places with the largest decreases in wages during the 1990s had relative gains in non-graduates from 2007 to 2012. This is another piece of evidence that young college graduates were able to afford housing only in lower quality neighborhoods. 2000s Housing Boom and Bust I also include housing boom and bust indicators, because repercussions are likely to linger in the 2007 through 2012 period. To measure the effects of the housing boom in a neighborhood, I use the log number of housing units built between 2000 and 2004, and also the percentage of mortgages that were high-cost loans between 2004 and 2006. The log number of housing units built between 2000 and 2004 varies a great deal, and is moderately inversely correlated with density (-.37). The U.S. Department of Housing and Urban Development’s Neighborhood Stabilization Program provides census tract-level data on high-cost mortgages in this period, which they define as mortgages “where the rate spread is 3 percentage points above the Treasury security of comparable maturity”, calculated using Home Mortgage Disclosure Act data. During this period, 24.4 percent of mortgages were defined as high-cost loans, and the rate of high-cost lending had little correlation with density. High-cost mortgage lending activity during the boom period has been associated with neighborhoods with higher levels of property risk, lower average credit scores, and low income and minority neighborhoods (Calem, Gillen, and Wachter 2004), as well as with higher foreclosure rates after the housing crisis. To measure the severity of the housing bust, I use the percentage change in housing prices between 2006 and 2009 at the PUMA level, the smallest geographic area for which annual house price data is available in the ACS. The mean drop in house prices during the bust was -30.2 percent, and the smallest drop was still substantial at -6.0 percent. House price changes were moderately positively correlated with density (.30); denser places may have been relatively protected from the foreclosure crisis because of lower homeownership rates. In terms of the housing boom indicators, high-cost mortgage lending during the boom had opposite associations with changes in young college graduates and non-graduates. In the period after the boom, young non-graduates tended to remain in neighborhoods with intense high-cost mortgage lending, while young college graduates tended to leave those neighborhoods. These opposing effects 67 point to the divergent neighborhood trajectories of young college graduates and non-college graduates. Nonetheless, the amount of housing built during the boom did not have a measurable association with changes in either young graduates or non-graduates in subsequent years. The repercussions of the housing crisis also appear to have different effects on changes in young college graduates and non-graduates. Changes in housing prices during the bust have an indeterminate relationship with changes in young graduates, though the sign on the coefficient is positive. In contrast, the strength of the housing bust has a strong inverse relationship with changes in young non-graduates. This inverse relationship means that young non-graduates remained or even increased in neighborhoods that were hardest-hit by the housing crisis. 5.4. Newly Built and Recently Vacated Housing Available on the Market I expect that in a tight housing market with low vacancy rates, the relative number of units available on the market in a neighborhood should act as a constraining (or enabling) factor in where young people move. I define housing opportunities in a neighborhood as the average number of housing units available on the market within the past year, measured in the 2005-09 ACS. This includes currently vacant units for sale or for rent and units with residents who moved in within the past year. I measure the number of available housing units in a tract, logged to account for a skewed distribution. By this definition, units that turned over multiple times within the past year are only counted once; however, this leads to conservative estimates of the effects of available units because of the underestimation in places with higher mobility rates. I measure recently built and vacated units separately to consider construction and moving out of existing housing as distinct processes. Recently built units account for only a small portion of all available units (averaging 8.0 percent in the Los Angeles region in 2007). While I expect housing availability to be associated with relative increases in both young graduates and non-graduates, I expect each group to have a slightly different relationship with recently built and vacated units. Changes in young graduates should have a stronger association with new construction, usually built for the upper end of the housing market, whereas changes in young non-graduates should have a stronger association with recently vacated existing units. Neighborhoods with more units available on the market, whether newly built or recently vacated, were associated with relative increases in young graduates and non-graduates. This suggests that in such a constrained market, with a growing population of young people looking for housing, housing availability serves as a constraining or enabling factor regardless of the push or pull factors of neighborhood characteristics. As expected, newly built units are more strongly associated with increases in young graduates than non-graduates, and recently vacated units have a stronger association with young non-graduates than graduates. When recently vacated and built units are interacted with neighborhood homeownership rates, a pattern emerges (illustrated in Figure 3.9 and Figure 3.10). For young graduates, the effects of recently vacated and built housing only hold in neighborhoods with low homeownership rates. This indicates a shift among young college graduates from owning towards renting, possibly related to the combination of tightened underwriting standards since the housing crisis and student loan debt. 68 For young non-graduates, the effects of recently vacated housing are also strongest in neighborhoods with low to medium homeownership rates. The interaction of recently built units with homeownership rates is not quite significant for non-graduates, but from the predicted values graph it appears that young non-graduates remain or even increase in neighborhoods where more housing is built with higher homeownership rates, and consistently decrease in neighborhoods with the lowest homeownership rates regardless of housing construction. This contrast suggests that young college graduates are taking up the recently built units in mainly rental neighborhoods, crowding out young non-graduates who are still able to buy recently-built homes in neighborhoods with higher homeownership rates. 5.5. Housing Tenure Given the reductions in homeownership rates for young people over this period, I expect shifts away from neighborhoods with high homeownership rates for both young college graduates and non-graduates. Housing tenure is tightly linked with both structure type and urban form: homeownership is more prevalent in single family homes in suburban neighborhoods, while rental housing is more prevalent in multifamily buildings in dense urban neighborhoods. This means that someone looking to buy is less likely to find a home in an urban neighborhood, and someone interested in renting is less likely to find a place to live in a suburban neighborhood. Figure 3.9 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by recently vacated housing units interacted with tract homeownership rate 69 Figure 3.10 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by recently built housing units interacted with tract homeownership rate To understand the role of renting or homeownership in the context of housing opportunities, I also interact the number of recently built and vacated units with neighborhood homeownership rates, lagged to 1990. I use this interaction rather than directly measuring the tenure of recently built and vacated units because a neighborhood with more recently-built housing for rent is also likely to have more recently-built housing for sale. While housing construction and moving out of existing housing can be reasonably considered separate processes, building rental units and building for sale units are likely to be linked in a neighborhood, and vacating rental units and vacating owner-occupied units may also be linked. Tenure is also somewhat fluid over time; a vacated owner-occupied unit may be purchased by an investor and rented out. Recent conversions between owning and renting may be endogenous to shifts in young people, and the significant lag of the 1990 homeownership rate reduces the potential for endogeneity. To this end, it is also an advantage to study young people: the earliest people involved in the dependent variable, those aged 25-34 in 2005, would have been 10-19 in 1990, barely beginning to enter the housing market. Though homeownership rates can shift through additions of new rental or for sale units or conversions between renting and owning, the homeownership rate in 2012 has a correlation of .897 with the homeownership rate in 1990, over two decades earlier. To the extent that homeownership rates changed in the intervening years, it would serve to dampen the signal and create conservative estimates of the effect of tenure. 70 Figure 3.11 Fitted estimates of changes in college graduates and non-college graduates aged 25-34 by homeownership rate Consistent with the patterns shown in the interactions, the largest increases in young college graduates are predicted in neighborhoods with mainly rental housing (shown in Figure 3.11). Homeownership rates have a generally negative relationship with changes in young graduates, though at high homeownership rates there is more variation in the effect. The association of changes in young graduates with homeownership rates only becomes marginally positive above 73.7 percent homeowners, well above the mean homeownership rate of 55.9 percent. Changes in young non-graduates have a quite different relationship with homeownership: the biggest decreases in young non-graduates are predicted in neighborhoods with the lowest and highest homeownership rates, and the smallest decreases are predicted in places with a mix of rental and owner-occupied housing. Young non-college graduates continued to increase in the exurban areas where investors bought up foreclosed and inexpensive homes and rented them out (as documented by Pfeiffer and Lucio (2015) in Phoenix, Arizona). And young college graduates flooded into rental housing in urban neighborhoods as mortgage access was pulled back after the bust and it became increasingly difficult for them to purchase homes. 71 6. Policy Implications Though Los Angeles is still a growing region with a vibrant economy, Los Angeles has fallen behind the state and the nation in attracting young college graduates. And Los Angeles has continued to lose young non-college graduates even as they have increased elsewhere. Despite the region’s overall growth, its less advantaged neighborhoods are worsening. Economic theory might suggest that the supply of opportunities should expand to meet demand, but political and financial institution constraints on suppliers have led to shortfalls that reduce availability and drive up prices to consumers. This research suggests that reductions in housing opportunities may further constrain the growth of young college graduates and hasten the loss of young non-college graduates in all but the least desirable neighborhoods. Yet it also offers a way forward: greater housing availability can make a positive difference in enabling young people to find places to live. Some of the factors that affect housing availability are beyond the reasonable influence of planners and policymakers. The ocean, mountains, and protected open spaces that constrain land availability for housing are valuable natural resources. Lower marriage rates make for more households competing for housing, and longer lifespans mean that older cohorts occupy housing longer. But in addition to natural, cultural, and demographic trends, existing policies and planning regulations conspire to constrain housing availability, whether by suppressing residential turnover and mobility or by constraining the construction of new housing. In California, Proposition 13 gives homeowners strong incentives to remain longer in their homes. Under Proposition 13, which has been in place since 1978, house value assessments for property taxes are constrained to rise 2 percent per year after a home is purchased, regardless of changes in prices or inflation. The incentive for owners to remain in place is stronger the faster prices rise, which might help explain lower cohort attrition rates in the 2000s compared with the 1990s. This policy leads recent buyers to pay far higher property taxes than people who bought their homes in the past, and young people are disproportionately burdened by property taxes in California (Myers 2009). Though Proposition 13 is naturally popular with established homeowners, and they hold considerable voting power, researchers and policymakers have called for a reconsideration of Proposition 13 because of its destructive impacts on both housing markets and the fiscal health of local governments. This research adds further motivation to revisit the policy so that more housing options are made available in neighborhoods throughout California. As described above, an array of policies and regulations limits housing construction in the Los Angeles region. The impetus behind these regulations is strong public resistance to new housing development in existing neighborhoods. A timely example of this phenomenon is the Neighborhood Integrity Initiative in the City of Los Angeles, slated for inclusion on the local ballot in 2016, and now postponed to 2017. A coalition of anti-development groups has proposed an initiative intent on halting large-scale housing development projects. Further limits to housing construction would only exacerbate the housing shortfall that is already affecting the region’s ability to attract and retain young people. But opposition to growth remains a powerful political force in such a large and congested region where residents are used to a relatively strong economic outlook and expect continued growth. In a faltering economy (such as that predicted by the Los Angeles 2020 Commission’s 2014 report, which begins, “Los Angeles is barely treading water while the rest of the world is moving forward”) residents might have more welcoming attitudes towards newcomers and better tolerate new construction in their neighborhoods. 72 Even if the tide of NIMBYism were to turn, redevelopment and infill construction would still be more difficult and expensive than greenfield development. And since NIMBY attitudes are unlikely to wane, planners must find creative strategies to facilitate housing construction in existing neighborhoods. Subsidized affordable housing construction serves only a small, specific slice of the growing demand for housing. Then again, there are indications that extensive informal housing construction already contributes a sizeable amount of new housing to accommodate growing demand at the lower end of the housing market (Wegmann and Mawhorter 2016). And planning scholars have demonstrated the potential for the construction of accessory dwelling units (ADUs) in single-family neighborhoods to expand the capacity of the existing housing stock without the visible disruptions of demolitions and redevelopment (Wegmann and Chapple 2014). Local resistance to adding new and possibly lower-income residents applies to ADU construction as well, and both planning restrictions and a lack of flexible financial instruments inhibit the legal construction of ADUs. Nevertheless, this is one area where the interests of established homeowners and would-be renters align: housing opportunities for renters would be potential income streams for homeowners. On the demand side, young people are limited in their ability to buy homes because of tightened access to mortgage credit since the housing crisis, as well as reduced employment, lower incomes, smaller households, and higher student loan burdens (Xu et al. 2015). Wachter (2015) emphasizes the uneven consequences for young people of diminished access to mortgage financing. This research implies that there are uneven consequences for neighborhoods as well: the surges of young college graduates in urban neighborhoods with rental housing are related to their inability to purchase homes in suburban neighborhoods. If restricted mortgage financing accelerates the divergence of people and neighborhoods, then expanding mortgage financing may be useful policy tool to combat rising inequality. The call to expand mortgage financing to a wider range of people and neighborhoods is complicated by the necessity of avoiding the predatory lending and risky loans that drove the housing boom and bust. Even so, expanded access to mortgage credit would mean expanded access to neighborhoods for young college graduates and non-college graduates alike. 7. Conclusion In this analysis, I find evidence that young college graduates and non-college graduates are increasingly living in different types of neighborhoods, as young college graduates are expanding their presence in higher-quality neighborhoods and young non-graduates only manage to find housing in lower-quality neighborhoods, if at all. Young college graduates are increasingly living in urban neighborhoods with lots of rental housing, while young non-graduates are increasingly living in the exurban neighborhoods where the bulk of new housing was built during the housing boom of the early 2000s. Millennials are clearly not a homogenous group, and those with the advantage of a college education are winning out in the competition for housing. The divergence of the types of neighborhoods where young college graduates and non-graduates live exacerbates their disparities of education and income. I also find that greater amounts of housing available in a neighborhood allow for increases in both young college graduates and non-graduates. To the extent that the housing available on the market makes a difference for where young people move, their movement to new and different neighborhoods is influenced by housing market constraints as well as their preferences for certain types of neighborhoods. And since so much less housing is available than for previous generations, 73 as discussed in Chapter 1, the constraints millennials face are likely to play a larger role in their housing and neighborhood outcomes. Proposition 13, the Neighborhood Integrity Initiative, ADU restrictions, and limited access to mortgage credit are merely a few examples of policies that constrain housing availability for young people in the Los Angeles region. In the current environment of resistance to growth, reversing these policies is far from politically expedient. Existing residents have a claim to neighborhoods and a voice in local political processes, while aspiring newcomers do not. Planners speak for the future, and they have an uphill battle to cultivate the political will to encourage the construction of housing in existing neighborhoods. This research provides planners with ammunition for that challenge by underscoring the steep costs for young people when little new housing is built. The Los Angeles region is already losing young people without college degrees. Los Angeles may also lose its ability to attract young college graduates if it does not build enough housing. In the short term, young people are paying the high costs of the housing shortfall, but in the long run the housing shortfall threatens the social and economic health of the region. 74 References Alba, Richard D., John R. Logan, and Brian J. Stults. 2000. “The Changing Neighborhood Contexts of the Immigrant Metropolis.” Social Forces 79 (2): 587–621. doi:10.2307/2675510. Bader, Michael D. M., and Siri Warkentien. 2016. “The Fragmented Evolution of Racial Integration since the Civil Rights Movement.” Sociological Science 3 (March): 135–66. doi:10.15195/v3.a8. Badger, Emily. 2016. “How to Make Expensive Cities Affordable for Everyone Again.” Washington Post Wonkblog. February 19. https://www.washingtonpost.com/news/wonk/wp/2016/02/19/how-to-make-expensive- cities-affordable-for-everyone-again/. Baer, William C., and Christopher B. Williamson. 1988. “The Filtering of Households and Housing Units.” Journal of Planning Literature 3 (2): 127–52. doi:10.1177/088541228800300201. Bleemer, Zachary, Meta Brown, Donghoon Lee, and Wilbert Van der Klaauw. 2014. “Debt, Jobs, or Housing: What’s Keeping Millennials at Home?” Staff Report 700. Rochester, NY: Federal Reserve Bank of New York. http://papers.ssrn.com/abstract=2530691. Blumenberg, Evelyn, and Alexandra Evans. 2010. “Planning for Demographic Diversity: The Case of Immigrants and Public Transit.” Journal of Public Transportation 13 (2). doi:http://dx.doi.org/10.5038/2375-0901.13.2.2. Blumenberg, Evelyn, and Kimiko Shiki. 2007. “Transportation Assimilation: Immigrants, Race and Ethnicity, and Mode Choice.” In Transportation Research Board 86th Annual Meeting. Washington, D.C. http://trid.trb.org/view.aspx?id=802238. Bostic, Raphael W. 2015. “The Millennials Are Finally Moving, But Will They Find Home?” Home Matters. October. http://www.homemattersamerica.com/the-millennials-are-finally-moving- but-will-they-find-home/. Bostic, Raphael W., Stanley D. Longhofer, and Christian L. Redfearn. 2007. “Land Leverage: Decomposing Home Price Dynamics.” Real Estate Economics 35 (2): 183–208. Briggs, Xavier de Souza. 2005. The Geography of Opportunity: Race and Housing Choice in Metropolitan America. Washington D.C.: Brookings Institution Press. Brown-Saracino, Japonica. 2009. A Neighborhood That Never Changes: Gentrification, Social Preservation, and the Search for Authenticity. Chicago: University of Chicago Press. Brueckner, Jan K, and Stuart S Rosenthal. 2009. “Gentrification and Neighborhood Housing Cycles: Will America’s Future Downtowns Be Rich?” Review of Economics and Statistics 91 (4): 725–43. doi:i: 10.1162/rest.91.4.725</p>. 75 Calem, Paul S., Kevin Gillen, and Susan Wachter. 2004. “The Neighborhood Distribution of Subprime Mortgage Lending.” The Journal of Real Estate Finance and Economics 29 (4): 393–410. doi:10.1023/B:REAL.0000044020.67401.51. Campbell, Burnham O. 1966. Population Change and Building Cycles. University of Illinois Bulletin 91. Urbana, IL: University of Illinois Press. Centers for Disease Control and Prevention, National Center for Health Statistics. 2015. “Underlying Cause of Death 1999-2014 on CDC WONDER Online Database, Released 2015. Data Are from the Multiple Cause of Death Files, 1999-2014, as Compiled.” http://wonder.cdc.gov/ucd-icd10.html. Chiang, John. 2016. “How to Address California’s Housing Affordability Crisis.” Los Angeles Daily News, June 9. http://www.dailynews.com/opinion/20160609/how-to-address-californias- housing-affordability-crisis-john-chiang. Clapham, David. 2005. The Meaning of Housing: A Pathways Approach. Bristol, UK: The Policy Press. Corgel, John B., and Halbert C. Smith. 1981. The Concept and Estimation of Economic Life in the Residential Appraisal Process. Society of Real Estate Appraisers Foundation. Courchane, Marsha J., Leonard C. Kiefer, and Peter M. Zorn. 2015. “A Tale of Two Tensions: Balancing Access to Credit and Credit Risk in Mortgage Underwriting.” Real Estate Economics 43 (4): 993–1034. doi:10.1111/1540-6229.12105. Crowder, Kyle, Matthew Hall, and Stewart E. Tolnay. 2011. “Neighborhood Immigration and Native Out-Migration.” American Sociological Review 76 (1): 25–47. doi:10.1177/0003122410396197. Crowder, Kyle, Scott J. South, and Erick Chavez. 2006. “Wealth, Race, and Inter-Neighborhood Migration.” American Sociological Review 71 (1): 72–94. doi:10.2307/30038976. Crowder, Kyle, and Scott J. South. 2005. “Race, Class, and Changing Patterns of Migration between Poor and Nonpoor Neighborhoods.” American Journal of Sociology 110 (6): 1715–63. doi:10.1086/428686. Diamond, Rebecca. 2013. “The Determinants and Welfare Implications of US Workers’ Diverging Location Choices by Skill: 1980-2000.” Working Paper, December. http://web.stanford.edu/~diamondr/jmp_final_121813.pdf. Dougherty, Conor. 2016. “How Anti-Growth Sentiment, Reflected in Zoning Laws, Thwarts Equality.” The New York Times, July 3. http://www.nytimes.com/2016/07/04/business/how-anti-growth-sentiment-reflected-in- zoning-laws-thwarts-equality.html. Dutzik, Tony, Jeff Inglis, and Phineas Baxandall. 2014. “Millennials in Motion: Changing Travel Habits of Young Americans and the Implications for Public Policy.” Boston, MA: U.S. PIRG Education Fund and Frontier Group. http://trid.trb.org/view.aspx?id=1328287. 76 Elsby, Michael W., Donggyun Shin, and Gary Solon. 2013. “Wage Adjustment in the Great Recession.” Working Paper 19478. Cambridge, MA: National Bureau of Economic Research. http://www.nber.org/papers/w19478. Foster, Thomas B., and Rachel Garshick Kleit. 2015. “The Changing Relationship Between Housing and Inequality, 1980–2010.” Housing Policy Debate 25 (1): 16–40. doi:10.1080/10511482.2014.933118. Fry, Richard. 2014. “Young Adults, Student Debt and Economic Well-Being.” Social and Demographic Trends Project. Washington, D.C.: Pew Research Center. http://www.pewsocialtrends.org/2014/05/14/section-1-student-debt-and-overall- economic-well-being/. ———. 2015. “More Millennials Living With Family Despite Improved Job Market.” Washington, D.C.: Pew Research Center. http://www.pewsocialtrends.org/2015/07/29/more- millennials-living-with-family-despite-improved-job-market/. Galster, George C. 1987. Homeowners and Neighborhood Reinvestment. Durham: Duke University Press. ———. 1996. “William Grigsby and the Analysis of Housing Sub-Markets and Filtering.” Urban Studies 33 (10): 1797–1805. doi:10.1080/0042098966376. Galster, George C., Royce Hansen, Michael Ratcliffe, Harold Wolman, Stephen Coleman, and Jason Freihage. 2001. “Wrestling Sprawl to the Ground.” Housing Policy Debate 12 (4): 681–717. Galster, George C., and Jerome Rothenberg. 1991. “Filtering in Urban Housing: A Graphical Analysis of a Quality-Segmented Market.” Journal of Planning Education and Research 11 (1): 37– 50. doi:10.1177/0739456X9101100106. Glaeser, Edward L., Joseph Gyourko, and Raven E. Saks. 2006. “Urban Growth and Housing Supply.” Journal of Economic Geography 6 (1): 71–89. doi:10.1093/jeg/lbi003. Gnanadesikan, R., J. R. Kettenring, and Srinivas Maloor. 2007. “Better Alternatives to Current Methods of Scaling and Weighting Data for Cluster Analysis.” Journal of Statistical Planning and Inference, Special Issue: In Celebration of the Centennial of The Birth of Samarendra Nath Roy (1906-1964), 137 (11): 3483–96. doi:10.1016/j.jspi.2007.03.026. Goodman, Laurie, Rolf Pendall, and Jun Zhu. 2015. “Headship and Homeownership: What Does the Future Hold?” Washington, D.C.: The Urban Institute. http://www.urban.org/research/publication/headship-and-homeownership-what-does- future-hold. Green, Richard K., Stephen Malpezzi, and Stephen K. Mayo. 2005. “Metropolitan-Specific Estimates of the Price Elasticity of Supply of Housing, and Their Sources.” The American Economic Review 95 (2): 334–39. doi:10.2307/4132843. 77 Grigsby, William G., Morton Baratz, George C. Galster, and Duncan Maclennan. 1987. The Dynamics of Neighborhood Change and Decline. Vol. 28, part 1. Progress in Planning. Oxford, Oxfordshire  ; Elmsford, New York: Pergamon Journals. Guerrieri, Veronica, Daniel Hartley, and Erik Hurst. 2013. “Endogenous Gentrification and Housing Price Dynamics.” Journal of Public Economics 100: 45–60. Gyourko, Joseph, Albert Saiz, and Anita Summers. 2008. “A New Measure of the Local Regulatory Environment for Housing Markets: The Wharton Residential Land Use Regulatory Index.” Urban Studies 45 (3): 693–729. doi:10.1177/0042098007087341. Hwang, Jackelyn. 2015. “Pioneers of Gentrification: Transformation in Global Neighborhoods in Urban America in the Late Twentieth Century.” Demography 53 (1): 189–213. doi:10.1007/s13524-015-0448-4. Hwang, Jackelyn, and Robert J. Sampson. 2014. “Divergent Pathways of Gentrification Racial Inequality and the Social Order of Renewal in Chicago Neighborhoods.” American Sociological Review 79 (4): 726–51. doi:10.1177/0003122414535774. Jackson, Kenneth T. 1985. Crabgrass Frontier: The Suburbanization of the United States. New York: Oxford University Press. Jacobus, Rick. 2016. “Why We Must Build.” Shelterforce, Published by the National Housing Institute. March 9. http://www.shelterforce.org/article/4408/why_we_must_build/. Lachmann, M. Leanne, and Deborah L. Brett. 2015. “Gen Y and Housing: What They Want and Where They Want It.” Washington, D.C.: Urban Land Institute. Leopold, Josh, Liza Getsinger, Pamela Blumenthal, Katya Abazajian, and Reed Jordan. 2015. “The Housing Affordability Gap for Extremely Low-Income Renters in 2013.” Washington, D.C.: The Urban Institute. http://www.urban.org/research/publication/housing-affordability- gap-extremely-low-income-renters-2013. Li, Wei. 2009. Ethnoburb: The New Ethnic Community in Urban America. University of Hawaii Press. Logan, John R., Richard D. Alba, and Brian J. Stults. 2003. “Enclaves and Entrepreneurs: Assessing the Payoff for Immigrants and Minorities.” International Migration Review 37 (2): 344–88. Logan, John R., Zengwang Xu, and Brian J. Stults. 2012. “Longitudinal Tract Database.” Spatial Structures in the Social Sciences, Brown University. http://www.s4.brown.edu/us2010/Researcher/LTDB.htm. Lowry, Ira. 1960. “Filtering and Housing Standards: A Conceptual Analysis.” Land Economics 36 (4): 362–70. Malpezzi, Stephen, and Wen-Kai Guo. 2001. “Measuring ‘Sprawl:’ Alternative Measures of Urban Form in U.S. Metropolitan Areas.” University of Wisconsin Center for Urban Land Economics Research Working Paper. 78 Moos, Markus. 2015. “From Gentrification to Youthification? The Increasing Importance of Young Age in Delineating High-Density Living.” Urban Studies, September, 42098015603292. doi:10.1177/0042098015603292. Moretti, Enrico. 2004. “Estimating the Social Return to Higher Education: Evidence from Longitudinal and Repeated Cross-Sectional Data.” Journal of Econometrics, Higher education (Annals issue), 121 (1–2): 175–212. doi:10.1016/j.jeconom.2003.10.015. ———. 2013. The New Geography of Jobs. Reprint edition. Boston, Mass.: Mariner Books. Myers, Dowell. 1983. “Upward Mobility and the Filtering Process.” Journal of Planning Education and Research 2 (2): 101–12. doi:10.1177/0739456X8300200206. ———. 1984. “Turnover and Filtering of Postwar Single-Family Houses.” Journal of the American Planning Association 50 (3): 352–58. doi:10.1080/01944368408976603. ———. 1992. Analysis with Local Census Data: Portraits of Change. San Diego, CA: Academic Press. ———. 1996. “Changes over Time in Transportation Mode for the Journey to Work: Aging and Immigration Effects.” In Conference on Decennial Census Data for Transportation Planning Sponsored by the Transportation Research Board, Federal Highway Administration, Federal Transit Administration, and the Bureau of Transportation Studies. Irvine, CA. ———. 2009. “The Demographics of Proposition 13: Large Disparities Between the Generations and the Unsustainable Effects of House Prices.” Los Angeles, CA: USC Population Dynamics Research Group. ———. 2011. “Attrition of Homeownership in California in the 2000s: Now Seeking Generational Replacements.” Special Report. Los Angeles, CA: Population Dynamics Research Group. ———. 2016. “Peak Millennials: Three Reinforcing Cycles That Amplify the Rise and Fall of Urban Concentration by Millennials.” Housing Policy Debate 0 (0): 1–20. doi:10.1080/10511482.2016.1165722. Myers, Dowell, Janna Goldberg, Sarah Mawhorter, and Seong Hee Min. 2010. “Immigrants and the New Maturity of Los Angeles.” In Los Angeles 2010: State of the City, edited by Ali Modarres. Pat Brown Institute. Myers, Dowell, and Cathy Yang Liu. 2005. “The Emerging Dominance of Immigrants in the US Housing Market 1970–2000.” Urban Policy and Research 23 (3): 347–66. doi:10.1080/08111470500197920. Myers, Dowell, Gary Painter, Hyojung Lee, and JungHo Park. 2016. “Diverted Homeowners, the Rental Crisis and Foregone Household Formation.” Research Institute for Housing America. https://www.mba.org/2016-press-releases/may/new-report-demographics-and- the-economy-drive-reduced-rental-housing-affordability. 79 Myers, Dowell, and John Pitkin. 2009. “Demographic Forces and Turning Points in the American City, 1950-2040.” The Annals of the American Academy of Political and Social Science 626: 91–111. Olsen, Edgar O. 1969. “A Competitive Theory of the Housing Market.” The American Economic Review 59 (4): 612–22. Owens, Ann. 2010. “Neighborhoods and Schools as Competing and Reinforcing Contexts for Educational Attainment.” Sociology of Education 83 (4): 287–311. Pendall, Rolf. 2012. “The Next Big Question Facing Cities: Will Millennials Stay?” The Atlantic CityLab. September 11. http://www.theatlanticcities.com/jobs-and- economy/2012/09/next-big-question-facing-cities-will-millennials-stay/3229/. Pendall, Rolf, Robert Puentes, and Johnathan Martin. 2006. “From Traditional to Reformed: A Review of the Land Use Regulations in the Nation’s 50 Largest Metropolitan Areas.” Research Brief. Washington, D.C.: The Brookings Institution. Pew Research Center. 2014a. “The Rising Cost of Not Going to College.” Social and Demographic Trends Project. Washington, D.C. http://www.pewsocialtrends.org/2014/02/11/the-rising- cost-of-not-going-to-college/. ———. 2014b. “Millennials in Adulthood: Detached from Institutions, Networked with Friends.” Washington, D.C. http://www.pewsocialtrends.org/2014/03/07/millennials-in-adulthood/. Pfeiffer, Deirdre. 2015. “Is the Real Estate Industry Cementing Millennials’ Residence in Central Cities? Insights from Phoenx and Houston During the 2000s and 2010s.” presented at the Association of Collegiate Schools of Planning Annual Conference, Houston, Texas, October 24. Pfeiffer, Deirdre, and Joanna Lucio. 2015. “An Unexpected Geography of Opportunity in the Wake of the Foreclosure Crisis: Low-Income Renters in Investor-Purchased Foreclosures in Phoenix, Arizona.” Urban Geography 36 (8): 1197–1220. doi:10.1080/02723638.2015.1053201. Phillips, Shane. 2016. “The Disconnect Between Liberal Aspirations and Liberal Housing Policy Is Killing Coastal U.S. Cities.” Better Institutions. April 20. http://www.betterinstitutions.com/blog/2016/3/27/liberal-cities-housing-policy-hypocrisy. Redfearn, Christian L. 2009. “Persistence in Urban Form: The Long-Run Durability of Employment Centers in Metropolitan Areas.” Regional Science and Urban Economics 39 (2): 224–32. doi:10.1016/j.regsciurbeco.2008.09.002. Rothenberg, Jerome, George C. Galster, Richard V. Butler, and John Pitkin. 1991. The Maze of Urban Housing Markets. Chicago: University of Chicago Press. http://www.press.uchicago.edu/ucp/books/book/chicago/M/bo3630137.html. Saiz, Albert. 2007. “Immigration and Housing Rents in American Cities.” Journal of Urban Economics 61 (2): 345–71. doi:10.1016/j.jue.2006.07.004. 80 ———. 2010. “The Geographic Determinants of Housing Supply.” The Quarterly Journal of Economics 125 (3): 1253–96. doi:10.1162/qjec.2010.125.3.1253. Saiz, Albert, and Susan Wachter. 2011. “Immigration and the Neighborhood.” American Economic Journal: Economic Policy 3 (2): 169–88. Sampson, Robert J. 2009. “Racial Stratification and the Durable Tangle of Neighborhood Inequality.” Annals of the American Academy of Political and Social Science 621 (January): 260–80. doi:10.2307/40375842. ———. 2011. Great American City: Chicago and the Enduring Neighborhood Effect. Chicago: The University of Chicago Press. Sharkey, Patrick. 2012. “Temporary Integration, Resilient Inequality: Race and Neighborhood Change in the Transition to Adulthood.” Demography 49 (3): 889–912. doi:10.1007/s13524- 012-0105-0. ———. 2013. Stuck in Place: Urban Neighborhoods and the End of Progress toward Racial Equality. Chicago: University Of Chicago Press. Stiglitz, Joseph E. 2013. “Student Debt and the Crushing of the American Dream.” May 12. http://opinionator.blogs.nytimes.com/2013/05/12/student-debt-and-the-crushing-of-the- american-dream/. Vicino, Thomas J., Bernadette Hanlon, and John Rennie Short. 2011. “A Typology of Urban Immigrant Neighborhoods.” Urban Geography 32 (3): 383–405. Wachter, Susan. 2015. “Housing America: The Unequal Geography of Risk and Opportunity.” Housing Policy Debate 25 (4): 813–16. doi:10.1080/10511482.2015.1043701. Wegmann, Jake, and Karen Chapple. 2014. “Hidden Density in Single-Family Neighborhoods: Backyard Cottages as an Equitable Smart Growth Strategy.” Journal of Urbanism: International Research on Placemaking and Urban Sustainability 7 (3): 307–29. doi:10.1080/17549175.2013.879453. Wegmann, Jake, and Sarah Mawhorter. 2016. “How Does Anyone Afford to Live Here? Quantifying the Rate of Informal Housing Construction in California’s Cities.” Working Paper (Available on Request). Xu, Yilan, Carrie Johnson, Suzanne Bartholomae, Barbara O’Neill, and Michael S. Gutter. 2015. “Homeownership Among Millennials: The Deferred American Dream?” Family and Consumer Sciences Research Journal 44 (2): 201–12. doi:10.1111/fcsr.12136. Yin, Robert K. 2008. Case Study Research: Design and Methods. 4th edition. Los Angeles, CA: SAGE Publications, Inc.

Abstract (if available)

Abstract In this dissertation, I set out to lay an empirical foundation for understanding why the housing stock in the Los Angeles region has been constrained since the building boom of the 1980s, and why regional demand for housing has continued to grow over the period of the housing boom and bust. I measure the declines in number of housing units available on the market, arising from both reduced construction and reduced turnover. I also examine the sources of growth in potential demand for housing, from migration and from natural increases of the existing population. I then compare the consequences of the housing shortfall for different types of neighborhoods and for different groups of people by age, race, and education level. I build on filtering theory with new ideas about how to conceptualize and measure changes in housing supply and demand, and I introduce new techniques to analyze the connections between regional shifts in supply and demand and local changes in neighborhoods. By combining well-established filtering theory with innovative demographic methods, I find empirical evidence of the powerful, usually hidden forces of the regional housing market and the way they impact various types of neighborhoods and groups of people. I make two main contributions to housing research. First, I argue that the housing supply that matters for growth is the housing available on the market, and this includes both newly built units and existing units that have been turned over by their previous occupants. Second, the underlying sources of growth in housing demand include (1) natural increases in the adult population through larger generations and longer life expectancies, in addition to the more commonly-studied (2) migration, (3) household formation, and (4) changes in income. I find that the number of vacant housing units available on the market fell 12.3 percent from 1990 to 2012 as a result of reduced construction and reduced turnover, while the regional adult population grew 29.2 percent. I also find that most of the growth in housing demand during the boom and bust periods was simply the result of natural increases rather than from immigration or migration to the region. From 2000 to 2012, the adult population grew 15.8 percent through natural increases, and only grew 1.4 percent through migration. In the context of this housing shortfall, with far fewer housing units available on the market and with steadily growing demand, I find stark disparities in housing and neighborhood outcomes for more and less advantaged groups. Middle-aged and older adults fared better than younger newcomers trying to make their way in the housing market, Whites and Asians fared better than Blacks and Hispanics, and young college graduates fared better than young people without college degrees.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Three essays on housing demographics: depressed housing access amid crisis of housing shortage

PDF

Insights into residential mobility and pricing of rental housing: the role of gentrification, home-ownership barriers, and market concentrations in low-income household welfare

PDF

A better method for measuring housing affordability and the role that affordability played in the mobility outcomes of Latino-immigrants following the Great Recession

PDF

The impact of mobility and government rental subsidies on the welfare of households and affordability of markets

PDF

Affordable south Los Angeles: survival, support, and different futures

PDF

Housing market crisis and underwater homeowners

PDF

Environmental justice in real estate, public services, and policy

PDF

Risks, returns, and regulations in real estate markets

PDF

Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

PDF

The impact of demographic shifts on automobile travel in the United States: three empirical essays

PDF

Avoiding middle-class planning 2.0: media arts and the future of urban planning

PDF

The interactions between housing and business

PDF

Living arrangements of young Latino couples: getting by in high-cost Los Angeles

PDF

Diverging pathways to citizenship and immigrant integration in the U.S.

PDF

Choice neighborhoods: a spatial and exploratory analysis of Housing Authority City of Los Angeles public housing

PDF

Emissions markets, power markets and market power: a study of the interactions between contemporary emissions markets and deregulated electricity markets

PDF

Three essays on aging, wealth, and housing tenure transitions

PDF

Urbanization beyond the metropolis: three papers on urbanization patterns and their planning implications in the contemporary global south

PDF

Suitable California Opportunity Zone (COZ) locations for affordable housing development in the cities of Bakersfield, Los Angeles, and Palmdale

PDF

Residential care in Los Angeles: policy and planning for an aging population

Asset Metadata

Creator Mawhorter, Sarah Louise (author)

Core Title Reshaping Los Angeles: housing affordability and neighborhood change

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Urban Planning and Development

Degree Conferral Date 2016-12

Publication Date 10/01/2017

Defense Date 08/12/2016

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag demography,housing affordability,neighborhood change,OAI-PMH Harvest,regional housing market,urban development

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Green, Richard K. (committee chair), Myers, Dowell (committee chair), Castells, Manuel (committee member), Galster, George C. (committee member)

Creator Email sarah.mawhorter@gmail.com,smawhort@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC11256097

Unique identifier UC11256097

Identifier etd-MawhorterS-4861.pdf (filename)

Legacy Identifier etd-MawhorterS-4861

Dmrecord 312287

Document Type Dissertation

Format theses (aat)

Rights Mawhorter, Sarah Louise

Internet Media Type application/pdf

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu