Fortunately, I finally found, buried away in a San Francisco Federal Reserve Bank report, some hints on who defaulted in California, the home to a sizable majority of all defaulted mortgage dollars.
Lending in Low- and Moderate-Income Neighborhoods in California: The Performance of CRA Lending During the Subprime MeltdownTwo economists at the San Francisco Fed made a study of mortgages originated in California in the Housing Bubble years of 2004-2006. They painstakingly matched federal Home Mortgage Disclosure Act data about mortgage originations, which track whether minorities get enough mortgage money but don't track whether they pay it back, with private (and expensive) Lender Processing Services (LPS) data, which don't care about ethnicity but do care about borrowers paying what they owe. The Fed economists came up with 239,101 successful matches of mortgages that show up in both databases. About 10,000 of them had entered the foreclosure process by the end of September 2007.
Elizabeth Laderman Carolina Reid Federal Reserve Bank of San Francisco November 26, 2008
This is a good time period for looking at foreclosures, since these aren't 2009 foreclosures that were caused by the recession; instead, these are the foreclosures that caused the recession. Most of this Fed report consists of a defense of the Community Reinvestment Act by showing lower foreclosure rates for banks (e.g., Washington Mutual) than for independent mortgage companies (e.g., Countrywide). Of course, that's not a very informative comparison since the Clinton Administration warned Countrywide-like firms that they would have the CRA extended to cover them legislatively unless they behaved like they already were under the CRA. Thus, Angelo Mozilo signed a deal with HUD secretary Henry Cisneros to lend more to minorities and lower income borrowers, and then put Cisneros on Countrywide's board, and even named Cisneros consultant to Mozilo's CRA-like trillion dollar pledge of January 2005.
No, what's really interesting is on pp 12-14. I put the most interesting stuff in bold at the end:
In Table 3, we present a very simple model where we predict the likelihood of foreclosure, controlling for borrower risk factors including income, race, and credit score. We present the findings as odds ratios to assist in interpreting the coefficients. We also control for neighborhood characteristics that may influence the underwriting decision, including the CAP rate, the age of the housing stock, and the percent of owner-occupied housing. Given the importance of house values in predicting foreclosures, we control for house price appreciation in each of the model iterations.So, in the economists’ simple multiple regression model, after adjusting for income and FICO, minorities in California still had substantially higher foreclosure rates than whites:
Several findings from even this simple model stand out. First, metropolitan house price changes do have a significant effect on the likelihood of foreclosure. Rapid house price appreciation in the 2 years preceding origination significantly increases the likelihood of foreclosure. This is consistent with previous research that has linked foreclosures and delinquencies to local housing market conditions, particularly in California where house prices rose quickly in relation to fundamentals and where subsequent corrections have been quite dramatic (Doms, Furlong and Krainer 2007). The tract’s capitalization rate is significant only at the 10 percent level, but also increases the foreclosure rate as expected. A higher percent of owner occupied housing in a tract and more recent construction both also seem to increase the likelihood of foreclosure, but only slightly.
Second, and not surprisingly, FICO scores matter. A borrower with a FICO score of less than 640 is 12.6 times more likely to be in foreclosure than a borrower with a FICO score of more than 720; for borrowers with a FICO score between 640 and 720, the odds ratio is 4.7 times compared to borrowers with the highest credit scores. We also find that race has an independent effect on foreclosure even after controlling for borrower income and credit score. In particular, African American borrowers were 3.3 times as likely as white borrowers to be in foreclosure, whereas Latino and Asian borrowers were 2.5 and 1.6 times respectively more likely to be in foreclosure as white borrowers.
- blacks 3.3X - Latinos 2.5X - Asians 1.6X
(These adjusted gaps are all statistically significant at the 0.01 level.)
Presumably, the raw differences in foreclosure rates are even greater. Unfortunately, the actual raw numbers aren't listed in the report, and the authors refused my repeated email requests to release the unadjusted numbers by ethnicity.
The raw ratios are important for estimating the overall share of defaulted dollars by ethnicity in California. We know from the federal HMDA data that minorities accounted for 77% of subprime home purchase dollars borrowed in California in 2006 (the worst vintage for defaults) and 56% of all home purchase dollars. You can see the graphs here. (I'm excluding borrowers of unknown ethnicity and mixed ethnicity couples).
Request 1: Would it be possible to reverse engineer the actual raw ratios from the numbers that do appear in this report?
Request 2: Also, is the Fed subject to the Freedom of Information Act?
According to Google, even though this report is eight months old, the part about the race differences in foreclosure rates has only quoted once before, by E. Scott Reckard, an LA Times reporter who has done a lot of good work on the mortgage meltdown.