Credit risk data may help target foreclosure mitigation
Published September 1, 2007 | September 2007 issue
Total mortgage delinquencies and foreclosures have shot up in 2007, but where are they occurring? "Across much of the country," you might say, and basically you'd be right. But that answer isn't precise enough for many housing nonprofits, charitable foundations and local government agencies working to assist troubled borrowers and their communities. To target their programs, they want to know which local areas are being especially hard hit now and which ones are likely to be hard hit in the near future. Nonprofits want this information because they know that it's important to deliver outreach programs to at-risk borrowers early on, both to prevent problems and to ensure that defaulting borrowers seek reputable help before it's too late. Governments and foundations want advanced warning of foreclosure risks because directing foreclosure prevention and mitigation resources to the neediest areas takes time.
Despite extensive public records detailing each pending and completed foreclosure, good information on the geographic patterns of current and prospective foreclosures has been surprisingly hard for many nonprofits and agencies to get. In most counties, the information is scattered in numerous individual documents. Transferring a backlog of information to a database that supports aggregate analysis and foreclosure mapping is a tedious process. (Private firms also transfer and sell this information, often to property investors, but many local governments and nonprofits cannot afford to buy it.)
This article aims to fill some of that information gap by showing which parts of the district tend to face high consumer credit risks. This approach is based in part on a case study by the authors of how foreclosure rates in some Twin Cities neighborhoods in 2002 could have been predicted.1 The study showed that several easily implemented methods could have given fairly accurate advance warning of where foreclosure rates would be high.
Overall, the most accurate simple method we studied was to rank neighborhoods by the percentage of all households with significantly blemished credit histories. Specifically, neighborhoods (or census tracts) were ranked by the percentage of households with a very low credit score as of 1999, three years in advance. Credit scores condense the information in a consumer's credit file into a single number representing the likelihood that the consumer will repay a loan. Thus a household with repeated late payments and a fairly recent bankruptcy would tend to have a low score, whereas a household with a history of consistently paying debts in full and on time would have a high score. Also, households with little experience using credit may have no score at all. (For more information on credit histories and credit scores, including information on how to check the accuracy of your own credit history, see AnnualCreditReport.com.) For the Twin Cities core counties of Hennepin and Ramsey, this approach correctly identified, three years in advance, over three-fourths of the neighborhoods that experienced high foreclosure rates in 2002.
A drawback to this approach is that credit score data can be expensive. Most lenders and even many real estate investors have access to the data for their business operations, but local nonprofits and units of government typically don't have the budgets to obtain them.
A partial solution is now available. The Federal Reserve System recently acquired data from a large credit bureau on the distribution of credit scores across U.S. neighborhoods as of December 2004. Although the Federal Reserve banks are not allowed to release the raw data, they can share maps and other summary information. To assist district foreclosure mitigation efforts, the accompanying maps divide district census tracts into five categories (quintiles) based on how many households have very low credit scores (scores typical of subprime borrowers). The maps may also be useful for other purposes, such as designing general consumer financial counseling, education and assistance efforts.
The research on 2002 foreclosures suggests that the areas in red, where low credit scores were most prevalent three years ago, are at risk for high rates of foreclosures now. These areas tend to fall in and around medium to large cities, Indian reservations (shaded areas) and some rural amenity and forestry areas (such as western Montana, northwestern Wisconsin and the Upper Peninsula of Michigan). By contrast, the largely agricultural belt running from central Montana across much of the Dakotas and into Minnesota displays a low incidence of risky credit scores.
A few caveats are in order, however. First, credit score information can't predict high foreclosure areas perfectly. Additional sources of information should be used when they are available. For example, Hennepin County in Minnesota now makes its foreclosure records regularly available to policymakers in a convenient electronic format. This allows very timely monitoring of where actual foreclosures are occurring. If, or as, other district counties follow suit, better analysis of public foreclosure records will obviously supersede credit scores as an indicator of existing foreclosures and can supplement or possibly replace credit scores as a predictor of foreclosures.
Second, the 2002 research covered only one urban area, and at this point we lack direct evidence on how well a high incidence of low credit scores will predict foreclosures in other cities or in rural areas. For example, it's unclear whether the high rate of low and missing credit scores on Ninth District Indian reservations will predict high rates of foreclosures, given the special characteristics of housing and housing finance institutions there. To partially address such concerns, our web site provides supplementary information, such as the prevalence of low credit scores only among households with a mortgage. Still, all the data we provide should be interpreted in light of other information about local housing market conditions that may affect foreclosure rates.
Third, changes in the mortgage market since 2002 may have weakened the connection between credit scores and foreclosures relative to what our study found. In 2004-06, nontraditional mortgage contracts with monthly payments that started low but escalated steeply after a few years quickly became popular, including among borrowers with medium-level credit scores. Many borrowers, not just those with low credit scores, have subsequently become delinquent on these nontraditional mortgages, and further problems are expected. Although no significant change in the relationship between credit scores and foreclosures is obvious in the recent Hennepin County foreclosure data, practitioners and policymakers should keep in mind the possibility that the maps here could miss some clusters of foreclosures related to nontraditional mortgage products.1"Targeting Foreclosure Interventions: An Analysis of Neighborhood Characteristics Associated with High Foreclosure Rates in Two Minnesota Counties," Federal Reserve Bank of Minneapolis Community Affairs Report, No. 2007-1. [1,035K PDF]