A new study shows that African-American and Latino borrowers were much more likely to get subprime loans, even after important underwriting criteria have been taken into account. A few years later, subprime loans are six times more likely to fall into foreclosure. The authors conclude that the short and tragic lifespan of subprime products should compel legislators to draft a CFPA that protects housing wealth, and indirectly, our nation's economy.
"Foreclosure in the Nation's Capital: How Unfair and Reckless Lending Undermines Homeownership" gathers loan data from 2004 to 2007 in the Washington, DC metro area. A unique feature of NCRC's research is that it has linked HMDA data, a common source for many mortgage studies, with loan performance data available through a proprietary set of servicing records.
The chart at left shows trends in foreclosures in DC on a quarterly basis. The blue line represents the inventory of foreclosed homes. From a low in 2006, DC now has more foreclosed homes than at any time in the last ten years. Moreover, these numbers may be false positives. It is possible that with a glut of pre-existing REO properties on the market, that lenders have been holding off on starting new foreclosures.
The research used a (ALERT: Data talk ahead!) logistic regression model to identify subprime loans. No one factor made a loan subprime in their definitions. Instead, a set of loan terms, borrower credit, and interest rates were used as independent variables that contributed to a nominal label of prime or subprime.
Key findings include:
- African-Americans and Latinos were 80 and 70 percent more likely, respectively, to get a subprime loan than were white borrowers, after controlling for the credit score, income, loan-to-value, and neighborhood characteristics.
- Mortgages made to African-Americans and Latino borrowers were 20 and 90 percent, respectively, more likely to enter into foreclosure.
- Loans purchased by the GSEs were half as likely to enter foreclosure as those held by private MBS investors.
- The most telling loan terms for gauging subprime: the presence of either a balloon payment (72 percent subprime) or a prepayment penalty (54 percent subprime).
Other data points confound what might be expected. For example, the share of subprime loans was highest in moderate-income
neighborhoods. The highest share of foreclosed homes are in middle-income neighborhoods.
Inferences for Policy
A new CFPA bureau may be in the works.
Nobel Prize-winning economist Joseph Stiglitz has rightfully critiqued the helter-skelter response to the crisis. We have given banks hundreds of billions of dollars, but we have done little to draft rules that would prevent the causes of the crisis from happening again:
But how can there be a restoration of confidence when all we have done is to pour more money into the banks? We have changed neither the regulatory structures, the incentive systems, nor even those who are running these institutions. As we taxpayers are pouring money into these banks, we have even allowed them to pour out money to their shareholders and to their executives in the form of bonuses, and to acquire other institutions. We imposed no conditions that would lead to more lending, and not surprisingly, we have not gotten more lending.
NCRC is going to push for the CFPA. Their core issue - a redrafting of the Community Reinvestment Act - could be drawn into the CFPA legislation. BankTalk has discussed some of those needs. I expect that the new bill will include a few updates. Certainly, the bankers would be more than happy to see to it that mortgage brokers are put on the same playing field as they are with regard to the CRA. I think there is a good chance that the new bill will also require mortgage lenders to report new data on their activities. Obama is a fan of transparency. Besides, the data is meant to be relevant. It is not an accident that NCRC had to buy additional data to make their case. The cost of LPS data is not cheap, and for the most part, few of the intended consumers of HMDA data could afford to supplement their own research with LPS data.