«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
We include, as a potential control variable for each regression equation, the size of the MSA housing market, measured as the log of the total number of active loans as of October 2008. In the equation for the gradient component, we include the ratio of maximum to median ZIP Code delinquency rate in the MSA to control for the potential effect of an outlier neighborhood.30 Economic Factors Deteriorating economic conditions also affect delinquency patterns. For instance, we expect higher average delinquency rates, on average, in cities with more rapidly increasing unemployment during 2008, or in cities with higher unemployment levels.
Spatial patterns of mortgage delinquency in an MSA may reflect the spatial distribution of borrowers’ incomes. We describe the spatial characteristics of borrowers’ median incomes within an MSA using the distributional moments—mean, standard deviation, skewness, and kurtosis (weighting by active loan count)—along with the two spatial autocorrelation measures.31 Regression Results Appendix D lists the economic and housing market variables that we ultimately selected for inclusion in one or more of the regression equations based on consideration of statistical significance and robustness.32 The mean value and standard deviation of each variable across the 91 metropolitan areas included in the study also appear in appendix D.
The neighborhood delinquency gradient factor may reflect idiosyncratic factors that determine the maximum neighborhood delinquency rate, rather than economic or housing market conditions affecting the broader metropolitan area.
We calculate the borrowers’ median income for each ZIP Code relative to MSA median income from pooled 2005, 2006, and 2007 HMDA data.
We employed stepwise regression as a first pass to develop baseline specifications, which we then evaluated for robustness by testing each variable excluded by the stepwise procedure.
262 Refereed Papers Geographic Patterns of Serious Mortgage Delinquency: Cross-MSA Comparisons Exhibit 7 summarizes the regression results. Results for the skewness, mean, gradient, and spatial autocorrelation components appear in columns 1 through 4, respectively.
One broad conclusion that emerges from the analysis is that the shape (skewness and kurtosis) of the neighborhood delinquency rate distribution and the spatial autocorrelation of neighborhood delinquency rates are closely tied to spatial patterns of subprime lending activity during 2005 and
2006.33 In the regression equations for the skewness/kurtosis and autocorrelation components, the estimated coefficient of the subprime lending counterpart of the dependent variable is the strongest explanatory variable. Thus, the regression analysis supports our previous contention that highdelinquency pockets in metropolitan areas characterized by high positive skewness or high spatial autocorrelation will often be neighborhoods with high subprime concentrations.
A second, broad, and not particularly surprising, conclusion is that economic conditions are at least as important as the subprime share mean/standard deviation component in influencing the delinquency rate mean/standard deviation component. This conclusion is consistent with our previous observation that metropolitan areas in Group 2 experienced harsher housing market or economic declines.
Third, spatial autocorrelation of neighborhood delinquency rates is strongly influenced by spatial autocorrelation of market decline during the “bust” period. Specifically, neighborhood delinquency spatial autocorrelation is positively related to the spatial autocorrelation of percent change in home purchase loan originations from the third quarter of 2006 through the third quarter of 2007.
The gradient component exhibits a somewhat eclectic set of associations. It is positively related to the subprime gradient component and inversely related to subprime spatial autocorrelation.
In addition, the neighborhood delinquency gradient component exhibits a positive association with MSA house price appreciation from the third quarter of 2005 through the third quarter of 2006 and an inverse association with housing affordability as of the third quarter of 2006. The latter relationships are consistent with rapidly rising house prices triggering overdevelopment that subsequently generated high-foreclosure pockets.34 Conclusion We first classified metropolitan areas into six groupings distinguished by their geographic patterns of serious mortgage delinquency. Understanding these patterns and their contributing factors may be informative for assessing local and neighborhood effects of the mortgage crisis and for developing appropriate strategies to mitigate the effects on communities.
We also estimated a regression equation for the Gini coefficient and found that it is very closely tied to the Gini coefficient of subprime lending (relative to total lending) activity.
We observe various additional results specifically for the gradient component. It is inversely related to the log of MSA active loan count, indicating that steeper gradients occur in smaller cities. It is positively related to the ratio of maximumto-median ZIP Code delinquency rate, confirming the importance of controlling for idiosyncratic neighborhood effects. The stepwise regression for the delinquency gradient component also yields three variables that are statistically significant at the 10-percent level in the equation: the subprime autocorrelation component, the third quarter 2008 unemployment rate, and the skewness of 2006 borrowers’ median income relative to MSA median family income across ZIP Codes. F-tests indicate joint and pairwise significance at the 5-percent level for these three variables.
264 Refereed Papers Geographic Patterns of Serious Mortgage Delinquency: Cross-MSA Comparisons Second, we examined some housing market and economic conditions associated with the different spatial patterns. Although overall delinquency rates are highest in cities with large home price declines or high unemployment, the examination in this article highlights how most other cities have high-delinquency pockets, mostly because of subprime lending concentrations.
The first cluster consists of MSAs with high spatial autocorrelation and low- or moderate-delinquency rate means. These MSAs contain a modest number of high- or moderately high-delinquency neighborhoods that are clustered together or comprise a distinct pocket of neighborhoods within the MSA. The second grouping exhibits a high mean and standard deviation for delinquency rates.
These MSAs have wide variation across neighborhoods, with most delinquencies occurring in distressed neighborhoods.
A third grouping is distinguished by a highly positively skewed, long, or fat-tailed distribution.
Metropolitan areas in the fourth cluster are characterized by low-to-moderate mean delinquency rates, high positive skewness, and a steep gradient around the peak delinquency neighborhood, whereas those in the fifth cluster are distinguished specifically by their steep gradient. The sixth group consists of metropolitan areas that have low-to-moderate scores for all components.
These classifications are potentially useful for understanding the effects of the mortgage crisis on the dynamics of housing market decline and recovery. For instance, home prices appear to be stabilizing in some metropolitan areas despite little reduction in the inventory of foreclosed properties. Most likely, the foreclosures are concentrated in specific neighborhoods that are lagging behind the overall market recovery, as negative spillover effects tend to diminish with distance.35 We believe the analysis has practical applications for selecting or adapting appropriate strategies and policy responses to stabilize neighborhoods and contain foreclosure spillover effects. For example, NSP funds might be most effective for reversing or containing problems associated with foreclosure when spatially targeted to neighborhoods detached from or on the perimeters of broader areas of elevated delinquency and foreclosure. Metropolitan areas with low-to-moderate delinquency means and highly skewed delinquency distributions (Groups 3 and 4, and some cities in Group 6) are those where strategic deployment of NSP funds could be particularly effective at containing neighborhood decline.
Finally, we recognize that this study relies on data from 2008 and that housing markets have further deteriorated in many cities since then, so some cities may need to be reclassified. Although looking back has value, we wish to emphasize the role of this study as an example or template for ongoing analysis.
See Frame (2010) and Lee (2008) for reviews of the literature on price-related spillover effects.
Authors Lariece M. Brown is a senior financial economist at the Federal Deposit Insurance Corporation.
Hui-Chin Chen is a former financial analyst at Freddie Mac.
Melissa T. Narragon is an economic research senior at Freddie Mac.
Paul S. Calem is a senior economist at the Federal Reserve System.
Acknowledgments The authors thank Calvin Schnure and two anonymous referees for their thoughtful feedback.
References Can, Ayse. 1992. “Residential Quality Assessment: Alternative Approaches Using GIS,” The Annals of Regional Science 26: 97–110.
Can, Ayse, and Isaac F. Megbolugbe. 1997. “Spatial Dependence and House Price Construction,” Journal of Real Estate Finance and Economics 14: 203–222.
Derudder, B., P.J. Taylor, F. Witlox, and G. Catalano. 2003. “Hierarchical Tendencies and Regional Patterns in the World City Network: A Global Urban Analysis of 234 Cities,” Regional Studies 37 (9): 875–886.
Edmiston, Kelly. 2009. “Characteristics of High-Foreclosure Neighborhoods in the Tenth District,” Economic Review 94: 51–75.
Fleischman, Daniel. 2010. “Nonprofit Strategies for Returning REO Properties to Effective Use.” In REO & Vacant Properties: Strategies for Neighborhood Stabilization, a joint publication of the Federal Reserve Banks of Boston and Cleveland and the Federal Reserve Board: 107–114.
Fortin, Marie-Jos’ee, Mark R.T. Dale, and Jay ver Hoef. 2002. “Spatial Analysis in Ecology.” In Encyclopedia of Environmetrics, edited by Abdel H. El-Shaarawi and Walter W. Piegorsch. Chichester, United Kingdom: Wiley: 2051–2058.
Frame, W. Scott. 2010. “Estimating the Effect of Mortgage Foreclosures on Nearby Property Values: A Critical Review of the Literature,” Economic Review 95 (3): 1–9.
Goldstein, Ira. 2010. “Maximizing the Impact of Federal NSP Investments Through the Strategic Use of Local Market Data.” In REO & Vacant Properties: Strategies for Neighborhood Stabilization, a joint publication of the Federal Reserve Banks of Boston and Cleveland and the Federal Reserve Board: 65–75.
Immergluck, Dan. 2010a. “Neighborhoods in the Wake of the Debacle: Intrametropolitan Patterns of Foreclosed Properties,” Urban Affairs Review 46 (1): 3–36.
———. 2010b. “The Accumulation of Lender-Owned Homes During the U.S. Mortgage Crisis:
Examining Metropolitan REO Inventories,” Housing Policy Debate 20: 619–635.
———. 2009. Intrametropolitan Patterns of Foreclosed Homes: ZIP-Code-Level Distributions of Real-Estate-Owned (REO) Properties During the U.S. Mortgage Crisis. Community Affair discussion paper No. 01-09. Federal Reserve Bank of Atlanta.
Kingsley, G. Thomas, Robin E. Smith, and David Price. 2009. “The Impacts of Foreclosures on Families and Communities: A Primer.” Urban Institute: http://www.urban.org/url.cfm?ID=411909 (accessed August 12).
Lee, Kai-yan. 2008. Foreclosure’s Price-Depressing Spillover Effects on Local Properties: A Literature Review. Community Affairs discussion paper, 2008-01. Federal Reserve Bank of Boston.
Lembo, Jr., Arthur J. 2008. “Spatial AutoCorrelation: Moran’s I, Geary’s C.” http://faculty.salisbury.
edu/~ajlembo/419/lecture15.pdf (accessed April 3).
Litchfield, Julie A. 1999. “Inequality: Methods and Tools.” http://siteresources.worldbank.org/ INTPGI/Resources/Inequality/litchfie.pdf (accessed March 2, 2009).
Newburger, Harriet. 2010. “Acquiring Privately Held REO Properties With Public Funds: The Case of the Neighborhood Stabilization Program.” In REO & Vacant Properties: Strategies for Neighborhood Stabilization, a joint publication of the Federal Reserve Banks of Boston and Cleveland and the Federal Reserve Board: 101–106.
272 Refereed Papers Geographic Patterns of Serious Mortgage Delinquency: Cross-MSA Comparisons Office of Management and Budget (OMB). 2006. Update of Statistical Area Definitions and Guidance on Their Uses. Bulletin No. 07-01. White House website: http://www.whitehouse.gov/sites/default/ files/omb/assets/omb/bulletins/fy2007/b07-01.pdf (accessed March 2, 2009).
Ryan, Jr., Jay N. 2010. “REO Disposition and Neighborhood Stabilization: A Servicer’s View.” In REO & Vacant Properties: Strategies for Neighborhood Stabilization, a joint publication of the Federal Reserve Banks of Boston and Cleveland and the Federal Reserve Board: 95–100.
Vicino, Thomas J., Bernadette Hanlon, and John Rennie Short. 2007. “Megalopolis 50 Years On:
The Transformation of a City Region,” International Journal of Urban and Regional Research 31 (2):
Additional Reading Hatcher, Larry. 1994. A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural Equation Modeling. Cary, NC: SAS Institute.
Weisstein, Eric W. n.d. “Gradient.” MathWorld—A Wolfram Web Resource. http://mathworld.
wolfram.com/Gradient.html (accessed March 2, 2009).