The Untold Costs of Subprime Lending: The Impacts of Foreclosure on Communities of Color in California Carolina Reid Federal Reserve Bank of San Francisco April 10, 2009 The views expressed herein are those of the author and do not necessarily reflect those of the Federal Reserve Bank of San Francisco or the Federal Reserve System.
Overview Methods Matching HMDA with Lender Processing Services, Inc. Applied Analytics (formerly McDash) Characteristics of Loan Sample Variable Selection and Coding Key Findings Policy Implications Caveats and Future Research
CRA Paper We found that loans made by CRA regulated institutions in their assessment area were half as likely to be in foreclosure as loans made by independent mortgage companies We also found that borrower characteristics, loan terms, and origination channel significantly influenced loan performance Key finding: even after controlling for a wide range of borrower, loan, and lender characteristics, African Americans and Latinos were significantly more likely to be in foreclosure than Whites Research that I would like to present today explores loan outcomes by race in more detail
Research Questions Using updated data, do we still observe differences in loan performance among Whites, African Americans and Latinos in California? What factors can explain help to explain observed differences? Were there differences in market access and product choice? How do outcomes for Whites, Latinos and African Americans vary geographically? Are there qualitative differences in outcomes if we look at older, inner city neighborhoods versus the boomburbs of California s Central Valley and Inland Empire?
Key Findings Differences in borrower characteristics, especially FICO score, are associated with higher rates of default House price declines, and their attendant effect on LTV and equity, increase the likelihood of default But, differential access to mortgage markets and products can help to explain a significant share of observed ed differences in foreclosure rates among whites and minorities, even after controlling for borrower and housing market characteristics
Matched LAR records from HMDA with loan records in Lender Processing Services, Inc. (formerly McDash) Created crosswalk between zipcodes and census tracts using block groups and weighted probability by housing units Variables that were used for matching Loan Closing Date Loan Amount Lien Status Loan Type (e.g. conventional, FHA) Loan Purpose (e.g. purchase, refinance) Occupancy Status Added lender information from Board s HMDA lender file Matched file includes loan performance data through January 2009 (previous file went through Dec. 2007)
We limit our analysis to a sample of Conventional, first-lien, owner-occupied loans originated in California between een January ar of 2004 and December of 2006 Home purchase loans Eliminated extreme/nonsense observations Sample size: 525,990 loan records Sample corresponds well to other studies analyzing HMDA data regarding the distribution of loans and their characteristics No significant differences or bias found between matched and unmatched observations in HMDA and McDash May underestimate impact due to lower coverage of subprime market in McDash California effect is observable: higher incomes, higher house values, higher FICO scores
30.00 27.46 27.61 25.00 20.00 00 15.00 10.00 7.36 Percent High Cost 5.00 0.00 White, Non Hispanic Hispanic Black
12.0 10.0 9.50 9.79 8.0 6.0 4.0 8.45 7.52 3.72 2.65 Percent of Loans 20 2.0 - White, Non Hispanic Hispanic Black Delinquent (30-90+) Foreclosure/REO
60.0 0 55.7 50.0 40.0 35.2 30.0 22.2 2 20.0 14.3 13.5 11.9 30.6 26.0 23.1 23.3 24.0 Percent of Borrowers 10.0 8.0 4.7 5.4 2.1 00 0.0 Less than 600 600-639 640-679 680-719 More than 720 White, Non Hispanic Hispanic Black
70.0 61.0 60.0 58.1 by FICO 52.2 49.2 50.0 44.2 40.0 36.7 32.0 30.0 25.0 21.9 20.0 18.9 12.8 10.0 5.9 2.6 13.5 12.8 Percent of Borrowers with High-Cost Loan 00 0.0 Less than 600 600-639 640-679 680-719 More than 720 White, Non Hispanic Hispanic Black
80 70 70.3 60 50 40 30 54.6 45.2 29.2 56.6 42.9 Percent of Borrowers 20 10 0 White, Non Hispanic Hispanic Black Independent Mortgage Co. Federally Regulated Financial Institution
Models Binomial logit model predicting the likelihood of obtaining a loan through an independent mortgage company (IMC=1) versus a federally regulated financial institution Multinomial logit model predicting the likelihood of obtaining different loan types (Prime Fixed, Prime ARM, Subprime Fixed, Subprime ARM) Binomial logit models predicting the likelihood of default (90 days or more delinquent and foreclosure) Explore differences by whether loan was originated through an IMC or not
Variables used in the analysis Borrower characteristics: Race, income, FICO score Neighborhood characteristics: FFIEC Income Classification, percent Black, percent Hispanic, percent College educated Housing market characteristics: ti Boom/Bust housing market, Herfindahl- Hirschman Index Loan characteristics: Higher-priced loan, LTV, fixed v. ARM, prepayment penalty, DTI, level of documentation, loan amount Lender characteristics: Independent mortgage company v. federally regulated financial institution, market presence Loan Source: Origination channel (wholesale/mortgage broker, correspondent lender)
Likelihood that Borrower Obtains Loan through Independent Mortgage Company Estimates Odds Ratio Estimates Odds Ratio Borrower Neighborhood Black 0.1512 1.16*** Percent Black 0.0016 1.00*** Hispanic 0.4334 1.54*** Percent Hispanic 0.0020 1.00*** Income 0.0004 1.00*** Percent College 0.0035 1.00*** FICO Score at Origination 0.0048 1.00*** Housing and Mortgage Market Income Level Herfindahl Hirschman Index 0.0003 1.00*** Low Income 0.3734 1.45*** Market Presence 2.7135 0.07*** Moderate Income 0.1774 1.19*** House Price Appreciation 0.4431 1.56*** Middle Income 0.0678 1.07*** (2 years before purchase)
Likelihood that Borrower Obtains Loan through Independent Mortgage Company Estimates Odds Ratio Estimates Odds Ratio Borrower Neighborhood Black 0.1512 1.16*** Percent Black 0.0016 1.00*** Hispanic 0.4334 1.54*** Percent Hispanic 0.0020 1.00*** Income 0.0004 1.00*** Percent College 0.0035 1.00*** FICO Score at Origination 0.0048 1.00*** Housing and Mortgage Market Income Level Herfindahl Hirschman Index 0.0003 1.00*** Low Income 0.3734 1.45*** Market Presence 2.7135 0.07*** Moderate Income 0.1774 1.19*** House Price Appreciation 0.4431 1.56*** Middle Income 0.0678 1.07*** (2 years before purchase)
Likelihood that Borrower Obtains Loan through Independent Mortgage Company Estimates Odds Ratio Estimates Odds Ratio Borrower Neighborhood Black 0.1512 1.16*** Percent Black 0.0016 1.00*** Hispanic 0.4334 1.54*** Percent Hispanic 0.0020 1.00*** Income 0.0004 1.00*** Percent College 0.0035 1.00*** FICO Score at Origination 0.0048 1.00*** Housing and Mortgage Market Income Level Herfindahl Hirschman Index 0.0003 1.00*** Low Income 0.3734 1.45*** Market Presence 2.7135 0.07*** Moderate Income 0.1774 1.19*** House Price Appreciation 0.4431 1.56*** Middle Income 0.0678 1.07*** (2 years before purchase)
Likelihood of Selecting Alternative Loan Product in Comparison to Prime Fixed Rate Mortgage Product = Borrower Characteristics (Black, Hispanic, FICO, Income) * Mortgage Market Characteristics (IMC, Wholesale, Correspondent, Conforming Loan) Borrower Characteristics Estimates Odds Ratio Black Fixed ARM 0.0241 0241 102 1.02 Subprime Fixed 1.1948 3.30*** Subprime ARM 1.1556 3.18*** Hispanic Fixed ARM 0.0529 0.95*** Subprime Fixed 0.8888 2.43*** Subprime ARM 1.3055 3.69*** FICO at Origination Fixed ARM 0.0036 0.99*** Subprime Fixed 0.0054 0 0054 099*** 0.99 Subprime ARM 0.0143 0.98*** Income Fixed ARM 0.0008 1.00*** Subprime Fixed 0.0051 0.99*** Subprime ARM 0.0030 1.00***
Likelihood of Selecting Alternative Loan Product in Comparison to Prime Fixed Rate Mortgage (continued) Product = Borrower Characteristics (Black, Hispanic, FICO, Income) * Mortgage Market,, g Characteristics (IMC, Wholesale, Correspondent, Conforming Loan) Mortgage Market Characteristics Estimates Odds Ratio Independent Mortgage Company d Fixed ARM 0.0094 0094 101 1.01 Subprime Fixed 1.7895 5.99*** Subprime ARM 3.2286 25.24*** Wholesale/Broker Originated Fixed ARM 0.8220 228*** 2.28 Subprime Fixed 0.0145 1.02 Subprime ARM 2.3356 10.34*** Correspondent Originated Fixed ARM 0.9388 0.39*** Subprime Fixed 0.0253 1.03 Subprime ARM 0.5656 0.57*** Conforming Loan Amount Fixed ARM 1.5535 0.21*** Subprime Fixed 0.0409 0409 104 1.04 Subprime ARM 1.4868 0.23***
Likelihood of Default (90+ and Foreclosure) Independent Mortgage Companies Federally Regulated Institutions Estimates Odds Ratio Estimates Odds Ratio Black 0.1490 1.2*** 0.0736 1.1 Hispanic 0.4076 1.5*** 0.3212 1.4*** FICO 0.0006 1.0** 0.0082 1.0*** LTV 0.0629 1.1*** 0.0698 1.1*** DTI 0.0020 1.0* 0.0061 1.0*** High Cost 1.7209 5.6*** 0.7298 2.1*** ARM 0.2648 1.3*** 0.1624 1.2*** Interest Only Loan 0.5500 1.7*** 0.7832 2.2*** Prepayment Penalty 1.0458 2.8*** 0.3181 1.4*** No Documentation 0.4059 1.5*** 0.4943 1.6*** Wholesale 0.1198 11*** 1.1*** 0.2411 13*** 1.3*** House Price Declines 0.5477 1.7*** 1.2443 3.5*** Percent College 0.0024 1.0** 0.0123 1.0***
Likelihood of Default (90+ and Foreclosure) Independent Mortgage Companies Federally Regulated Institutions Estimates Odds Ratio Estimates Odds Ratio Black 0.1490 1.2*** 0.0736 1.1 Hispanic 0.4076 1.5*** 0.3212 1.4*** FICO 0.0006 1.0** 0.0082 1.0*** LTV 0.0629 1.1*** 0.0698 1.1*** DTI 0.0020 1.0* 0.0061 1.0*** High Cost 1.7209 5.6*** 0.7298 2.1*** ARM 0.2648 1.3*** 0.1624 1.2*** Interest Only Loan 0.5500 1.7*** 0.7832 2.2*** Prepayment Penalty 1.0458 2.8*** 0.3181 1.4*** No Documentation 0.4059 1.5*** 0.4943 1.6*** Wholesale 0.1198 11*** 1.1*** 0.2411 13*** 1.3*** House Price Declines 0.5477 1.7*** 1.2443 3.5*** Percent College 0.0024 1.0** 0.0123 1.0***
Likelihood of Default (90+ and Foreclosure) Independent Mortgage Companies Federally Regulated Institutions Estimates Odds Ratio Estimates Odds Ratio Black 0.1490 1.2*** 0.0736 1.1 Hispanic 0.4076 1.5*** 0.3212 1.4*** FICO 0.0006 1.0** 0.0082 1.0*** LTV 0.0629 1.1*** 0.0698 1.1*** DTI 0.0020 1.0* 0.0061 1.0*** High Cost 1.7209 5.6*** 0.7298 2.1*** ARM 0.2648 1.3*** 0.1624 1.2*** Interest Only Loan 0.5500 1.7*** 0.7832 2.2*** Prepayment Penalty 1.0458 2.8*** 0.3181 1.4*** No Documentation 0.4059 1.5*** 0.4943 1.6*** Wholesale 0.1198 11*** 1.1*** 0.2411 13*** 1.3*** House Price Declines 0.5477 1.7*** 1.2443 3.5*** Percent College 0.0024 1.0** 0.0123 1.0***
Key conclusions African Americans and Latinos in California had access to very different mortgage markets Geographic distance doesn t seem to be defining feature: even in relatively small area like Oakland lenders targeted different neighborhoods This differential access led to very different mortgage products, even after controlling for borrower characteristics And, the foreclosure crisis has hit minority households much harder, with significant implications for the minority wealth gap
Policy Implications Findings support the need to reevaluate the regulatory landscape to ensure access to responsible credit Consider extending supervisory regime to non-regulated financial institutions Fair Lending Regulations Do we need to move to a more systemic evaluation of fair lending violations? Local matters Link disconnected actors, fill information gaps, provide reputational controls, shape preferences Mediating institutions (e.g. nonprofits) may be an important way of ensuring that both borrowers and lenders make the right credit decisions
Caveats and Future Research Limitations of study Limited to California experience and home purchase loans Additional attention to specifying the relationships between the types of lenders, borrower characteristics, and loan performance Future research What are the implications of this for the long-term wealth of Latinos and African Americans? What are the other possible negative spillover effects for minority communities?
Carolina Reid Federal Reserve Bank of San Francisco Carolina.reid@sf.frb.org reid@sf frb org Federal Reserve Bank of San Francisco 101 Market Street San Francisco, CA 94105