Weathering the storm: How have IDA homebuyers fared in the foreclosure crisis?

Abstract

This study is the first to compare the homeownership outcomes of Individual Development Account (IDA) homebuyers with other low-income homebuyers. The IDA homebuyers purchased homes in 16 states with IDA funds between 1999 and 2007. We compare both loan terms and foreclosure outcomes for the IDA homebuyer sample to comparison groups of other low-income homebuyers who purchased homes in the same counties and during the same time period. We find that IDA homebuyers were more likely to receive government-insured loans and less likely to receive high interest rate or subprime loans than other low-income homebuyers. Further, we find that cumulative foreclosure rates for IDA homebuyers were one-half to one-third the rate for other low-income homebuyers in the same communities. Overall, the findings suggest that low-income IDA program participants have fared better in the foreclosure crisis than other low-income homebuyers.

Keywords:

• homeownership
• foreclosure
• assets
• counseling

Acknowledgments

The authors gratefully acknowledge the cooperation and work of program staff at the six IDA programs that participated in this research, and in particular, Gwendy Donaker Brown, Kerry Cook, Julian Huerta, Amy Jo Kennedy, Dan Kornelis, Karen Lyons Serna, Marcy Meyer and Deltrise Sanford. We thank Ashley Arrington and David Price for their excellent research assistance and Steve Crawford, Bob Friedman, Lindley Higgins, Robert Lerman, Leigh Tivol, and Douglas Wissoker for their helpful comments and suggestions. In addition, we value the input of participants at the 2009 Association for Public Policy Analysis and Management Fall Conference and an Urban Institute Opportunity and Ownership seminar. This research was made possible through support from NeighborWorks America and the Ford Foundation. The findings and conclusions presented are those of the authors, and do not necessarily reflect the opinions of CFED, the Urban Institute, their boards, or their sponsors.

Notes

¹The literature finds that maximum match amounts (or match caps) are positively related to increased participant savings, as measured through average monthly net deposits (Han and Sherraden 2009; Schreiner and Sherraden 2007a). The literature is less clear around the effect of match rates on participant savings (Zielewski et al. 2009). However, the amount individuals have for a down payment (their own IDA savings plus the IDA match) is expected to be higher with versus without participation in the IDA program.

²The six programs are: Opportunity Fund (San Jose, CA); La Casa of Goshen (Goshen, IN); New Hampshire Community Loan Fund (Concord, NH); New Century IDA Program (Forsyth County, NC); WECO Fund (Cleveland, OH); and Foundation Communities (Austin, TX).

³The IDA programs included in our study previously obtained signed release forms from each participant giving the IDA program consent to release their information for research purposes.

⁴The original sample included 856 homebuyers, but 25 homebuyers were dropped from the analysis because of missing information on purchase date or purchase location.

⁵Although living within an IDA program's local service area is a requirement for applying for an IDA program, three of the six programs permitted participants to purchase homes outside of their service area or out of state. This flexibility in program rules resulted in homebuyers in a total of 16 states, although the vast majority of homebuyers bought homes in the six states in which the IDA programs are located.

⁶Eleven of the 64 counties in which homes were purchased did not provide free or low-cost online access to public records, which precluded verification of status for 11 homeowners. We were also unable to verify an additional 17 homeowners. Note that two programs conducted some of the property searches themselves, one due to privacy issues and one as a part of the program's normal operating procedures. We verified that program's property searches were conducted in the same manner as those conducted by the authors.

⁷The Home Mortgage Disclosure Act was enacted in 1975 in an effort to identify discriminatory lending practices. The Act requires mid-size and large lending institutions to report basic information from loan applications. Overall, the HMDA data capture about 80 percent of all loans. HMDA data are less complete in nonmetropolitan areas and counties with small populations because institutions serving these areas are not required to report the geographic location of the property, and some are even exempt from filing under HMDA. While this is a downside of the HMDA data, they are the most complete data available on homebuyers' demographic characteristics and their loan terms. The majority of our IDA homebuyers bought homes in larger counties and metropolitan areas, so our analysis should not be significantly affected by this limitation.

⁸The high interest rate flag is based on an APR calculation, which takes into account the initial interest rate as well as the subsequent rates on the loan.

⁹The last HUD subprime list was published in 2005, so we use the 2005 list to identify subprime lenders in 2006 and 2007.

¹⁰We cannot match on month and year of home purchase because the HMDA files do not include month, only year, of the loan origination.

¹¹Our IDA homebuyer data set includes only income levels at the time of program entry, while the HMDA data provides home purchasers' incomes at loan origination.

¹²IDA homebuyers' loans may be present in the mortgage performance comparison sample but will represent such a small proportion they will not drive the overall sample characteristics.

¹³All dollar amounts are in real 2005 dollars.

¹⁴We limit the sample to homebuyers with FICO scores below 680 because automated underwriting software programs such as Fannie Mae's Desktop Underwriter and Freddie Mac's Loan Prospector have used this FICO score as a minimum threshold for conventional first mortgages for borrowers with higher loan-to-value ratios. Fannie Mae Eligibility Matrix, updated on January 14, 2010, accessed on February 10, 2010 ( https://www.efanniemae.com/sf/refmaterials/eligibility/index.jsp).

¹⁵The loan-to-value ratios in the IDA homebuyer sample and the three comparison samples are very similar – between 0.829 and 0.839 – which suggests some comparability across the IDA and comparison samples. These relatively low loan-to-value ratios likely result from the fact that we are examining first mortgages.

¹⁶The HMDA data provide year of home purchase, while the mortgage performance data provide information on both the month and year of home purchase.

¹⁷Marital status is captured at the time of enrollment. Only five of the six programs were able to provide us with this information.

¹⁸For the IDA sample, race, ethnicity and gender of the IDA participant are reported. This person is presumably on the loan application. In the HMDA data, race, ethnicity and gender of the primary applicant are reported. Race, ethnicity and gender of the co-applicant are also reported in HMDA but are not used in this analysis.

¹⁹In the HMDA sample, a homebuyer is considered an unmarried woman if the primary applicant is a woman. If some of these women are married, then the percentage of unmarried women in the HMDA comparison sample would be even lower.

²⁰As mentioned in the data section, the HMDA data only began to include a high interest rate flag in 2004. To construct a comparable high interest rate flag for the IDA sample, we use the IDA homebuyers interest rate reported by the IDA program and calculate whether that rate was three points above the prime rate at the time of purchase. IDA programs collected information on interest rates from paperwork they collected during the homebuying process, primarily the HUD 1 form.

²¹Recall that the homeowner status was verified for 803 of the 831 IDA homebuyers.

²²The cumulative foreclosure rates are roughly similar across the two time periods in both the IDA homebuyer sample and in the comparison samples. For example, the cumulative foreclosure rates for the IDA homebuyer sample are in the 3 percent range in both time periods. The similar foreclosure rates across the two periods could be due to the longer period over which we observe the loans that originate in the earlier years. We observe foreclosures through spring 2009, so we have 10 years to observe a foreclosure for a loan that originates in 1999 but only two years to observe a foreclosure for a loan that originates in 2007.