Thy Neighbor’s Mortgage: Does Living in a Supbrime Neighborhood Affect One’s Probability of Default?
Sumit Agarwal, Ph.D.
Federal Reserve Bank of Chicago
Brent W. Ambrose, Ph.D.
Penn State University
Souphola Chomsisengphet, Ph.D.
Office of the Comptroller of the Currency
Anthony B. Sanders, Ph.D.
George Mason University
During the early 2000s house prices rose dramatically, followed by flattening prices during 2006, and finishing with the burst of the bubble in 2007. At the same time this housing bubble was forming, the subprime mortgage market experienced exponential growth. The connection between house prices and subprime origination activity was not coincidental. In fact, subprime mortgages were tailored to markets with increasing house prices. Often lenders originated these loans with low initial payments in the expectation that the borrowers would refinance into more favorable mortgage terms in the future via increased home equity. With the softening of the housing market in 2006, however, refinancing became more difficult, especially for subprime borrowers that did not meet the traditional underwriting requirements of prime mortgages. The inability of borrowers to obtain new mortgages under more favorable terms led to a large increase in mortgage defaults and foreclosures. In our paper we attempt to answer two questions: did subprime mortgages cluster in certain areas, and if so, does being in a neighborhood with relatively high levels of subprime originations increase an individual’s probability of default. Stated differently, if a lot of my neighbors have subprime mortgages, am I more likely to default on my mortgage.
Financial economists generally model mortgage default within an option-based pricing framework. In these models, a necessary condition for default is that the value of the mortgage exceeds the value of the home. When house prices fall, more borrowers become “underwater” on their mortgage, and default rates increase. This explains why some of the markets with the largest price declines in recent years (i.e. California, Arizona, Florida, and Nevada) have also experienced the highest foreclosure rates.
Interestingly, subprime loans originations were not evenly distributed across the country. Areas with high subprime origination tended to be the same areas with high price appreciation leading up to the burst of the housing bubble. For example, from 2000 to 2006 house prices doubled within the Phoenix, AZ market. Over the same period, subprime originations ballooned, with nonprime loan origination volume increasing over tenfold. The run-up in home values, coupled with increased nonprime mortgage originations, makes Phoenix an ideal market to examine the relationship between alternative mortgage products and house price growth.
After narrowing our focus on Phoenix, AZ, we find significant variation in both house prices and alternative mortgage origination activity across zip codes. Although the volume of non-prime mortgage origination was highest in the urban-peripheral, or new-build areas, the ratio of non-prime to prime originations was highest in inner city areas. In other words, alternative mortgage origination was concentrated in low-income areas.
Research shows that foreclosure effects spillover to surrounding properties. For example, if a property goes to foreclosure sale, it may negatively affect the value of neighboring properties. If poor underwriting standards characterized the subprime mortgage market, and subprime loans clustered in certain areas, the spillover effects may be particularly strong in these locations. We try to determine whether borrowers in neighborhoods with high concentration of non-prime loans default more on average.
We use loan level information from a database that covers roughly 61% of the non-prime mortgage market. We focus on loans in Maricopa County, AZ from 2000 to 2007. The database includes both borrower characteristics (e.g. credit score, income) and loan characteristics (e.g. LTV, property zip code, ARM or FRM). The data clearly shows that higher risk lending increased over the sample period. For example, at the beginning of the sample period full income documentation loans accounted for roughly 75% of all non-prime loans. By 2004, only about 40% of borrowers provided full income documentation. We also see a large decrease in the proportion of loans with a fixed interest rate.
Every month we calculate the proportion of mortgage loans outstanding of different types in each zip code. These zip code concentration measures tell us the proportion of outstanding loans in an area that are of a certain type (i.e. subprime, ARM, No-Doc). We also calculate monthly mortgage default rates for each zip code, and to control for house price changes within a zip code, we create a repeat-sales index for each zip code. We find subprime origination is correlated with house price appreciation, which suggests that credit availability may have helped fuel the housing bubble in Phoenix.
In order to determine whether neighborhood subprime mortgage concentration affects an individual’s probability of default, we estimate a proportional hazard model of borrower default. The intuition behind a hazard model is that it estimates how control variables affect the likelihood of an outcome. In our case we investigate how individual borrower risk characteristics, loan characteristics, nearby foreclosure levels, house price changes, and zip-code mortgage concentration measures affect the likelihood of mortgage default.
Most of our results match a priori expectations. Borrower credit risk is inversely related to default risk, while higher LTV ratios increase default risk. No-doc loans are 2.4 times more likely to default than full documentation loans, and loans with prepayment penalties increase the odds of default by 12.6%. As expected, subprime loans contain more default risk than prime loans, with a 1.3 times greater probability of default. Surprisingly, we find that ARM loans in general are less likely to default. However, hybrid-ARMs, which were common in the subprime market, increase default risk. Consistent with prior research, loans on non-owner-occupied houses are more likely to default. We also find that when house prices are rising, default risk decreases.
Next we examine how mortgage activity in surrounding areas affects the odds of default. Although having a subprime loan increases an individual’s likelihood of default, living in an area with a high concentration of subprime loans does not increase the likelihood of default. Stated differently, subprime spillover effects do not increase default risk. A possible explanation is that areas with high levels of subprime activity also experienced high house price appreciation during the peak subprime years. Being located in an area with a high concentration of hybrid-arms, low-doc loans, or no-doc loans increases the likelihood of default. A 1% increase in no-doc concentration is associated with a 10% increase in the odds of default. Also, high levels of foreclosures nearby increase default risk on a loan. Aside from the subprime concentration ratio, these findings suggest that high concentration ratios of risky loans in a zip code increase and individual’s default risk.
In summary, we investigate the relationship between default and non-prime mortgage concentration at the local level. We find that subprime loans are concentrated near the central business district and other low-income neighborhoods. Borrower and loan characteristics affect an individual’s default risk, and we provide evidence that foreclosures have a spillover effect at the local level. Surprisingly we find that being located in an area with high sub-prime mortgage origination actually lowers the likelihood of default. However, these same areas tended to experience the largest price gains at the height of the subprime boom. This suggests that extending subprime credit may not have increased the general level of defaults, but rather helped fuel house price increases in high subprime concentration areas. Subprime and Alt-A loans, however, do increase the odds of default. So although being located in an area with a high concentration of non-prime loans does not increase an individual’s default risk, default risk remains higher on subprime and Alt-A mortgages.
 This article is adapted from Agarwal, Sumit, Brent W. Ambrose, Souphola Chomsisengphet, and Anthony B. Sanders, 2012, “Thy Neighbor’s Mortgage: Does Living in a Subprime Neighborhood Affect One’s Probability of Default?”, Real Estate Economics, 40(1): 1 – 22.