Originating lender localness and mortgage sustainability: an evaluation of delinquency and foreclosure in Indiana's mortgage revenue bond program

Abstract

Originating lenders play a vital role in selecting and preparing borrowers for homeownership, directly and through partnerships with community entities. While previous research demonstrates the importance of originating lenders for mortgage access to low- and moderate-income borrowers, this analysis evaluates the influence of the originating lender, and in particular the localness of the lender, on mortgage sustainability (reduced delinquency and foreclosure). Employing data on more than 5,000 low- and moderate-income borrowers participating in Indiana's Mortgage Revenue Bond (MRB) program from 2004–2006, this analysis finds that the localness of the originating lender is significantly predictive of mortgage sustainability. After controlling for borrower, mortgage, and market characteristics, an increase in the localness of the lender is associated with a decrease in the probability of delinquency and foreclosure, particularly for higher risk (lower credit score) borrowers participating in the MRB program.

Keywords:

• homeownership
• low-income housing
• mortgages
• financial institutions and industry

Acknowledgments

This research was funded in part by an Early Doctoral Student Research Grant (EDSRG) from the US Department of Housing and Urban Development. The author would like to thank David Reingold, David Good, Michael Collins, and the anonymous reviewers for their helpful suggestions and comments on this article. In addition, the author would like to thank David Kaufmann and the staff of the Indiana Housing and Community Development Authority for their assistance during this analysis. Any errors, omissions or opinions expressed in this article are those of the author.

Notes

¹In this analysis, lender localness is measured as a continuous variable as the percentage of a lender's loan volume in the county where a borrower purchases a home. Other analyses measure localness as whether or not a lender has an office in the county where the borrower purchased a home (Avery et al. 1999; Hoffman 2001), or the distance from the lender's office to the location of purchase (Ergungor 2007a; 2007b). Regardless of the measurement, localness is consistently conceptualized as a lender's presence (measured by lending volume or physical office) in the area where a borrower purchased a home.

²This research has generally found that non-local lenders (and in particular less regulated non-bank mortgage companies and broker institutions) tend to lend as much or more (in terms of total volume) to LMI borrowers as do local banks, due in part to “informational returns to scale.” Larger banks may have more resources to target, appraise and service LMI borrowers (Campen 1993; Avery, Beeson, and Sniderman 1999; Avery et al. 1999; Williams and Nesiba 1997; Williams, McConnel, and Nesiba 2001). There is some evidence that regulated banks operating in communities with Community Reinvestment Act (CRA) agreements in place increase their lending activity to LMI borrowers (Bostic and Robinson 2005; Shlay 1999; Schwartz 1998). Further, emerging research finds that local lenders (in particular CRA-regulated lenders lending within their assessment areas) may originate more affordable mortgages (lower interest rates) to LMI borrowers (Avery et al. 2005; Ergungor 2007a; Traiger and Hinckley 2008).

³Borrower risk may be approximated using credit score information, where borrowers with credit scores below 660 are generally considered “higher risk” than borrowers with credit scores at or above 660 (Barakova et al. 2003).

⁴Increasingly, HFAs couple the reduced interest rate financing with other forms of borrower assistance, including downpayment subsidies or second mortgages, reduced private mortgage insurance, and/or homebuyer education; however, the reduced interest rate is the hallmark of the MRB program (Goldberg and Harding 2003). MRB subsidized mortgages are currently restricted by Congress to first-time homebuyers (who have not purchased a home in the past three years), earning less than area median income, or less than 115 percent of area median income for families of three or more, or less than 140 percent of area median income in targeted underserved areas. Further, the price of homes to be purchased with MRBs is limited to 90 percent of the average purchase price. In 2006, the median income of borrowers assisted with MRB mortgages nationwide was $31,703, which is 65 percent of the national median of $48,451. The average purchase price was $132,939, 62 percent of the national median purchase price of $222,000 (NCSHA 2008).

⁵Lender servicers may be different entities than lender originators; while loan originators initiate the mortgage transaction with a borrower, lender servicers collect mortgage payments from the borrower on a monthly basis. In the present analysis, while MRB mortgages are originated by different lenders, the servicing (monthly collection of mortgage payments) is provided by one lending institution for all MRB mortgages.

⁶These perspectives are not mutually exclusive, indeed studies on mortgage sustainability often incorporate elements of multiple perspectives; however, each emphasizes the potential importance of different contributors to mortgage delinquency and foreclosure.

⁷Foreclosure is a legal process whereby a lender (or their agent) takes action to claim a borrower's collateral property when the borrower has defaulted on their mortgage obligation, usually by failing to make timely payment on the debt. The foreclosure process typically starts at 120 days of delinquency (at the lender's discretion) and is governed by state laws that dictate the terms of the foreclosure process and its minimum duration. As of 2007, expected minimum foreclosure process timelines, from referral to sheriff sale, ran from 33 days (Tennessee) to 312 days (Iowa), while actual timelines varied from 63 days (Virginia) to 448 days (Maine) (Cutts and Merrill, 2008). See also Quercia and Stegman (1992) and Cutts and Green (2005).

⁸Under the Community Reinvestment Act of 1977, depository lending institutions that are federally insured are subject to evaluations of the degree to which they meet community needs. Current regulations (see NCRC 2007) stipulate a three-tiered lending, investment and service test for large depository institutions with assets greater than $1 billion, a scaled back evaluation in the three areas for midsize institutions ($250 million to $1 billion), and a lending test only for small depository institutions (assets less than $250 million). Mortgage company affiliates of depository institutions are not directly responsible under the CRA; however, their parent depository institution may elect to have their lending activity considered on CRA evaluations. By contrast, independent mortgage companies and broker institutions, not affiliated with a depository institution, are exempt from CRA evaluations. While the evaluation is completed by the respective federal regulator, the community can play an important role in the evaluation and its ramifications, by providing information to regulators on the lending institution and its service to the community and/or using information from evaluations to protest approvals for bank mergers and acquisitions. The efficacy of this “community regulation” is dependent in part on the sophistication of the community to hold lenders accountable, for example, through the development of CRA agreements (Bostic and Robinson 2005).

⁹Preliminary empirical evidence suggests that certain types of pre-purchase counseling and education reduce the risk of delinquency (Hartarska, Gonzalez-Vega, and Dobbs 2002; Hirad and Zorn 2002). Tracking the borrowers' mortgage payment patterns for two to seven years after home purchase, Hirad and Zorn found that individual counseling and face to face classroom education reduced the rate of delinquency 41 percent and 23 percent respectively.

¹⁰Although the downpayment assistance in Indiana is not funded through the sale of bonds (rather through HOME funds), the downpayment assistance through IHCDAs downpayment program was only available with the MRB loan product from 2004–2006, thus adding an additional incentive to the reduced interest rate for LMI borrowers to receive MRB mortgage financing.

¹¹An example of guidelines for IHCDA's participating lenders in 2008 is available on IHCDA's website, at http://www.in.gov/ihcda/files/2008_MRB_MOSA.pdf. In general, upon the sale of the mortgage to the master servicer, the master servicer assumes the primary credit risk for the mortgage. However, participating lenders may be required to “buy back” mortgages from the master servicer if they become delinquent within a set period of months after closing (if the originator cannot provide sufficient documentation of due diligence at the time of origination). From 2004–2006, lenders were potentially required to re-purchase mortgages that became delinquent within 3 months of closing; for MRB mortgages originated in Indiana in 2008, participating lenders are potentially required to re-purchase mortgages that become delinquent within 9 months of closing.

¹²At the time of this analysis, IHCDA limited the originator to a 1 percent origination fee and $450 in lender closing costs.

¹³Typically, 60 or 90 days late is considered severe delinquency; the threshold of 60 days (or more) is considered severe delinquency in this analysis (Avery et al. 1996; Hirad and Zorn 2002). In this dataset, loans are identified as “ever in foreclosure” (1) if they were currently in the process of foreclosure as of March 23, 2008, or (2) if the foreclosure process had been completed (resulting in REO or lender sale of the property) prior to March 23, 2008. Loans that entered the foreclosure process, but cured prior to March 23, 2008, would not be considered to be in the foreclosure process in this dataset, but would instead be included as “ever 60-days delinquent.”

¹⁴Alternative model specifications have been developed using multi-level modeling (often referred to as hierarchical linear modeling), to account for the leveled nature of the data. The intraclass correlations (between group variation) is around 11 percent, suggesting that HLM provides little additional value to standard logistic or probit regression. The results found using the multivariate probit regression adjusting for clustering (using STATA's cluster command) are robust under the HLM specification. Results of the HLM models are available from the author.

¹⁵Data are not available to estimate the primary independent variable, lender localness measured as a percentage of loan volume in a county, for lending institutions not reporting under HMDA. Therefore, MRB borrowers with loans originated by lending institutions not reporting under HMDA are not included in the primary analysis. As demonstrated in Table 1, sample borrowers for whom lender data are available are slightly more likely to become ever 60-days delinquent (22.07 percent) or ever in foreclosure (7.9 percent) than the total population of MRB borrowers from 2004–2006; while t-tests demonstrate that the difference in outcomes is statistically significant between sample and non-sample MRB borrowers (p < 0.01), the small number of excluded borrowers (less than 6 percent of the total) and the non-random nature of their exclusion (lenders not reporting under HMDA) reduces concerns about sample bias for accurately representing Indiana's MRB program.

¹⁶Under the CRA, the lender's “community” (and thus its assessment area) is generally defined at the county level, as the county within which the lender operates a branch office (Apgar and Duda 2003; Ergungor 2007b). Each individual county-branch office combination becomes a unique assessment area for the lending entity.

¹⁷Because of uneven distributions of continuous numerical independent variables (including lender size), the natural logs of such indicators are utilized in the modeling specifications in this analysis to normalize the distributions and increase the efficiency of the estimates.

¹⁸Principal component factor analysis is a commonly prescribed approach for dealing with multicollinearity (see Kennedy 2003 for a more complete discussion).

¹⁹Depository banks include lending institutions that provide banking services (checking and savings accounts) in addition to originating mortgages. Depository commercial banks and their affiliates are regulated by the Office of the Comptroller of the Currency (OCC), the Federal Deposit Insurance Corporation (FDIC) or the Federal Reserve System (FRS). Depository thrift institutions (and their affiliates) are regulated by the Office of Thrift Supervision (OTS).

²⁰The median gross household income of borrowers in the MRB program is $33,150, or approximately 70 percent of county AMI adjusted for household size. This represents approximately 74 percent of the three year average Indiana median income of $44,806 from 2004–2006 (according to the US Census Bureau).

²¹The debt ratio is calculated using the borrower's total monthly financed debt at the time of origination, including but not limited to the mortgage payment.

²²This is the only type of “second mortgage” permitted through Indiana's MRB program. The LTV variable does not include the loan amount of this second mortgage; it only includes the loan amount of the first mortgage.

²³Information on changes in single family home prices over time (quarterly) for counties located within MSAs is available through the Federal Housing Finance Agency's House Price Index. Similar to Stegman et al. (2007), an alternative modeling specification limits the sample to those borrowers purchasing homes within MSA counties (n = 4,463) and incorporates the change in home price information.

²⁴In an analysis of mortgage denial rates, Harvey et al. (2001) employ factor analysis to create a neighborhood quality factor comprised of changes in median household income, rental rates, vacancy rates, percentage of minority residents, percentage of households on public assistance, and percentage headed by females, at the census tract level.

²⁵In a probit model, the standardized beta coefficient represents the standard deviation change in the dependent variable (the underlying latent “continuous” dependent variable) for a one unit increase in binary independent variables, or a one standard deviation increase in continuous independent variables.

²⁶ΔPr represents the predicted change in the probability of the outcome (delinquency or foreclosure) for a change in the independent variable, (1) from its minimum value to maximum value (for binary independent variables) or (2) for a one standard deviation change around its mean (for continuous independent variables). To compute the predicted probabilities, all continuous variables are held at their means, and dummy variables are held at their modal values (except for the specified value of X). Lender type is held at Commercial Bank (all other = 0); Debt ratio is less than 36 percent (0); Borrower is not a minority (0); Loan-to-value ratio exceeds 97 percent (1), FHA (1), USDA (0) and DPA (1).

²⁷This makes sense, as lower credit risk borrowers may be less likely to become delinquent or enter foreclosure due to “triggering events” (because of their lower risk), and may be more responsive to market factors (like home values).

²⁸Rather than including categories for lender size, a nonlinear relationship can also be modeled by including squared and cubic terms for lender size with the original variable. An additional model (not shown) includes the squared and cubic terms for lending volume, and demonstrates a statistically significant nonlinear relationship (demonstrated by the statistical significance of all three terms for lending volume). The results of this nonlinear transformation are less directly interpretable than the categorical specification of lender size; thus, the results of the categorical specification are presented here.

²⁹While this may initially seem contradictory, it may be that both small and large lenders have informational advantages when originating mortgages to LMI borrowers, allowing them to select borrowers less likely to become delinquent or enter foreclosure; smaller lenders may benefit from soft information and local knowledge, while larger lenders may benefit from informational returns to scale (Bostic and Robinson 2004; Williams and Nesiba 1997; Nakamura 1994). Further, it may be that both small and very large lenders face the most pressure from external entities to provide for longer term mortgage outcomes, resulting in reduced delinquency and foreclosure among their MRB borrowers. Smaller institutions may be pressured by their local communities on whom they are dependent for survival, and larger institutions (with assets in excess of $1 billion) face the most stringent evaluations under CRA from their regulators.

³⁰It is important to point out that while originating lender characteristics, particularly localness, appear to be associated with mortgage sustainability, borrower characteristics are still more important (in terms of magnitude) predictors of mortgage delinquency and foreclosure – even for higher risk borrowers. Most notably, the credit score of MRB borrowers is the most substantial and consistent predictor of mortgage delinquency and foreclosure, even when the sample is limited to borrowers with credit scores below 660 (higher risk) or equal to or above 660 (lower risk), suggesting that even among high (or low) credit risk borrowers, there is still substantial variation in credit risk between the borderline and very low (or very high) credit scores.

³¹To the extent that such omitted variables are also associated with lender localness, there is potential for biased estimates of the localness effect. While this is possible, it is not likely given the inclusion of critical borrower and neighborhood characteristics (income, credit score, affordability, population size, neighborhood quality, lending saturation) that would be most likely associated with both localness and mortgage sustainability.

³²For example, in 1980, 72 percent of all mortgages were originated by depository institutions (thrifts and commercial banks); however, by 1997, such depository institutions accounted for only 43 percent of all originations (of which 43 percent were originated by their mortgage company subsidiaries), with independent and affiliated mortgage companies assuming the lion's share of originations. Further, an increase in lender consolidation and mergers has driven the volume of lending from small, community banks to larger institutions; more than half (52 percent) of all mortgages in 2000 were originated by the largest 25 home purchase lenders (Apgar and Duda 2003).