The Geography of the Recent Housing Crisis: The Role of Urban Form

Abstract

This study maps the geography of the recent housing crisis within and across American metropolitan areas, and evaluates how it is related to a series of spatial and socioeconomic variables at neighborhood and metropolitan levels. It finds that the spatial patterns of housing recessions vary widely by region. In general, fast-growing metropolitan areas in the Southwest and Florida experienced not only deeper but also longer housing recessions. In contrast, metropolitan areas in the South (except in Florida) saw shallower and shorter housing recessions. Metropolitan areas in the Midwest and Northeast had fewer price declines in the crisis, but their housing recessions tended to be longer. Housing recessions tend to be deeper and longer in larger metropolitan areas. Neighborhoods located closer to city centers experienced shallower and shorter recessions compared with those in fringe areas. Even after controlling for many other variables, automobile dependency is still a strong and positive predictor of housing recession depth and duration. The effects of other urban form variables, such as land-use density and mixed use, are mixed and vary by region. The significance of the effects of neighborhood demographic variables on recession depth is highly dependent on the inclusion of high-risk loan in the model, suggesting that predatory and high-risk lending is one major reason why lower income and minority neighborhoods were hit harder by the recent housing crisis. The effects of high-risk loan and neighborhood demographic variables on housing recession duration, however, are rather weak.

Keywords:

• housing recession
• spatial patterning
• urban form

Acknowledgments

This research was partially supported by a data grant awarded by the CoreLogic Academic Research Council. The authors would like to thank the editor and two anonymous reviewers for their valuable critiques and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

^1. The number varies by survey and source. For example, the 2004 National Survey on Communities reports that 55% of Americans indicated a preference for a smart-growth community over a conventional sprawl-style community (Ewing, 2007). The 1998 Vermonters Attitudes on Sprawl Survey shows that 74% of respondents preferred a home in the suburbs to one similarly priced near transportation, work, and shopping, and 48% of respondents claimed preferences for neighborhoods with mixed land use (Myers & Gearin, 2001). A survey by the National Association of Home Builders in 1999, however, indicated that 83% of respondents preferred a single-family detached home in the suburbs, even if it requires a longer commute than a town house in the city (Myers & Gearin, 2001).

^2. The monthly HPI data provided by CoreLogic are averaged to the quarter level in this study.

^3. Our data analyses indicate that among the 99 metropolitan areas considered in this study, 97 had housing prices reach a peak between 2005 (quarter 3) and 2007 (quarter 4). But we did find that in two metropolitan areas (Buffalo–Niagara Falls, New York, and Tulsa, Oklahoma), housing prices peaked in 2008. We thus decided to define our peak period as 2005 to 2008.

^4. We have the HPI data for the top 100 CBSAs or divisions in the United States, but the Honolulu Metropolitan area is not shown in Figure 1.

^5. In Figure 2, neighborhoods that are located three times farther than the median distance to city center are excluded as outliers because they tend to distort the shapes of the Loess curves.

^6. The city center is defined as the ZIP code with the highest employment density in the largest census principal city in each metropolitan area. Employment subcenters in a CBSA are not considered in this study.

^7. We use the term medium instead of small because all the metropolitan areas studied in this analysis are among the top 100 CBSAs in the United States. Our data show that the smallest metropolitan area in our data set is the Provo–Orem metropolitan area in Utah, with a population of 474,180.

^8. This is because American housing markets are regional, and these variables are more meaningful when they are measured in local contexts. For example, a 10-mile distance to the city center may mean quite different things in New York and Buffalo. Similarly, a New York neighborhood of 10 units per acre may be viewed quite differently, by local homebuyers, from a neighborhood in Houston with the same density, in terms of their crowdedness.

^9. The results of these models are available from the authors upon request.

^10. The absolute values of Pearson correlations between them are all below 0.40, except the correlation between land-use density and auto dependency, which is − 0.61.

Additional information

Notes on contributors

Hongwei Dong

Hongwei Dong is an assistant professor in the Department of Geography and City and Regional Planning at the California State University, Fresno. His research focuses on sustainable land use, transportation, and housing planning and policy.

J. Andrew Hansz

J. Andrew Hansz is a visiting professor of real estate in the Department of Risk Management at Pennsylvania State University, University Park. His research focuses on decision-making behaviors in real estate, real estate valuation, and real estate investment/portfolio issues.