«Urban Problems and sPatial methods VolUme 17, nUmber 1 • 2015 U.S. Department of Housing and Urban Development | Office of Policy Development and ...»
For example, as high unemployment persisted and housing prices nationally continued to fall, HAMP added features to try to address the situation of unemployed homeowners and underwater borrowers. Still, the program has struggled to reach its intended scale. As of November 2013, 1.3 million borrowers had received modifications under HAMP, fewer than Treasury’s initial estimate of 3 to 4 million (GAO, 2014). In addition, since peaking in early 2010, the monthly volume of new modifications made under the program has largely trailed off. Despite HAMP also provides a bonus incentive of $1,500 to lenders/investors and $500 to servicers for modifications made while a borrower is still current on mortgage payments but at imminent risk of default. To help servicers make a determination if a modification would help to protect the investors’ interests in the loan, HAMP uses a standardized net present value, or NPV, model to compare expected cashflows from a modified loan to the same loan with no modification, using certain assumptions.
not reaching its volume target, some evidence shows that HAMP has been successful in extending beneficial terms to struggling homeowners. The program has led to significant reductions in payments—an average of $544 each month, or approximately 40 percent of their premodification payment—for borrowers who obtained relief (Treasury, 2014), and a study in New York City found that HAMP modifications outperformed non-HAMP modifications after 1 year (Voicu et al., 2012). In addition to HAMP modifications, the Office of the Comptroller of the Currency (OCC) estimates that an additional 2 million homeowners have received proprietary modifications (OCC, 2014), although very little is known about the terms of these modifications.
Although OCC and Treasury release regular reports on loan modification activity and redefaults, still only a few studies have examined the factors that influence the effectiveness of modifications in a multivariate framework, and even fewer studies consider differences across demographic groups. This study helps to fill that gap. In the next section, we review the existing literature on loan modifications, focusing specifically on studies that seek to understand loan modification trends by borrowers’ race and ethnicity.
Literature Review Although, in theory, the borrower and investor are each better off if a foreclosure is avoided, in practice, it has proven to be much more difficult to modify loans. Research has identified several institutional factors that may influence servicer practices, including servicer incentives and capacity, mortgage securitization and the associated pooling and servicing agreements, information asymmetries, and lack of borrower contact (Adelino, Gerardi, and Willen, 2013;
Cordell et al., 2010; Eggert, 2007; Gelpern and Levitin, 2009; Levitin and Twomey, 2011;
Piskorski, Seru, and Vig, 2010).
One of the biggest barriers to modifying loans has been the lack of incentives for servicers. Loan modifications are costly: they are labor and time intensive and cannot be easily automated. Unlike the costs associated with foreclosure, neither the labor nor the overhead costs associated with modifications are billable back to investors (Levitin and Twomey, 2011). If the modified loan redefaults before the servicer has recouped the cost of the modification, then the modification is a money loser for the servicer. As a result, until HAMP was put into place, most servicers had very little financial incentive to undertake loan modifications. Moreover, very few servicers invested in either the staff or the technological capacity to respond to the volume of distressed borrowers at the height of the foreclosure crisis (Cordell et al., 2010).
Researchers have also posited that the investor pooling and servicing agreements (PSAs) that govern privately securitized loans may limit a servicer’s ability to offer a loan modification. Although PSAs vary for different mortgage pools, in general, they require servicers to manage the loans in a way that maximizes the returns to the investor. A loan modification may be more difficult for servicers to undertake if they need to consider multiple investors with competing interests (Cordell et al., 2010). A handful of papers have found that loans in private-label securities were less likely to be modified than loans held in portfolio (Agarwal et al., 2011; Been et al., 2013; Piskorski, Seru, and Vig, 2010). In contrast, Adelino, Gerardi, and Willen (2013) argued that no differences in loan modification rates exist between loans
Cityscape 167Collins, Reid, and Urban
held in portfolio and those held in private-label securities. The extent to which securitization influences modifications is still unclear, however, because all four of these studies use different data, methods, and model specifications, making it difficult to compare results.4 A third explanation for differences in modification rates may lie in individual servicers’ institutional responses to the foreclosure crisis. One option for a servicer is to implement a highly automated process of default management, which enables the servicer to keep the costs of managing delinquencies low (Levitin and Twomey, 2011). In contrast, other servicers have created loss-mitigation units to work with distressed borrowers, often in concert with housing counselors or foreclosure prevention specialists. Experts have also described the renegotiation process as “more art than science”; ex ante it is difficult to know whether a modification will actually lead to a cure or whether it merely postpones delinquency (Adelino, Gerardi, and Willen, 2013). Given that a significant percentage of loans self-cure, servicers must also make a judgment as to whether the modification is really necessary for any individual borrower. The extent to which the servicer is willing to invest in staff and time to perfect this “art” may lead to different determinations about the benefits of offering a borrower a modification and on what terms.
All these factors have material effects for a borrower who is seeking to obtain a loan modification and stay in his or her home. Borrowers have very little control over the ownership or administration of their loan after origination, however; they cannot decide whether their loan will be securitized, who their servicer will be, or what contractual provisions will govern the servicing of their loan (Levitin and Twomey, 2011). Consumer rights regarding loss mitigation are fairly narrow, and the process by which loss-mitigation decisions are made is often opaque. As a result, advocates and housing counselors have raised the concern that the loan modification process may lead to unequal outcomes for certain protected classes (California Reinvestment Coalition, 2011). For example, race or perceived race could serve as a proxy that servicers use for decisionmaking on modifications, especially if these borrowers are deemed more time consuming and, therefore, more costly to serve.
In this article, we contribute to a growing literature on the effectiveness of loan modifications and specifically build on studies that examine loan modification terms and subsequent loan performance for different racial and ethnic groups. The lack of public data on individual loan modifications, coupled with the fact that most loan performance datasets do not include any information about the borrower with the exception of a FICOTM credit score, means that we still have a limited understanding of whether loan modifications help prevent foreclosures, and, if so, for whom.5 Given the importance of homeownership for asset building and community development, research on how to improve outcomes in the default resolution process is especially relevant for public policy.
Empirical studies that have examined the effectiveness of loan modifications have found that the terms of the modification are important in predicting redefault. In one of the first studies For a discussion of the points of contention and differences in methodology, see Adelino, Gerardi, and Willen (2013).
Treasury released the first loan-level data on HAMP in 2011. Mayer and Piven (2012) attempt to use these data to identify racial differences in modification outcomes, although in 79 percent of active permanent modification records and 82 percent of trial modification records no information on borrower race or ethnicity is in the data file.
to examine loan modification terms, White showed that most pre-HAMP modifications typically increased a borrower’s monthly payment and the principal owed on the loan (White, 2009a, 2009b). He argued that the high redefault rates of early modifications reflected the fact that the loan renegotiation process did little to increase the affordability of the mortgage. Subsequent studies have shown that the most successful loan modifications are those that result in a significant decrease in either the monthly payments or the principal of the loan (Cordell et al., 2009; Cutts and Merrill, 2008; Haughwout, Okah, and Tracy, 2010; Quercia and Ding, 2009). Quercia and Ding (2009), for example, found that loans with greater payment reductions have lower redefault risks and that loans have an even lower risk of redefault when payment reduction is accompanied by principal reduction. The authors suggest that, among the different types of modifications, the principal forgiveness modification has the lowest redefault rate. Cutts and Merrill (2008) similarly showed that the success rate of modified loans varies by the amount of arrearage capitalized into the loan modification; they found a direct relationship between a lower arrearage and a lower redefault rate.
Missing from these studies, however, is an analysis of how these factors might differ for different types of borrowers. Four studies post-crisis have used loan performance datasets merged with HMDA and other data sources to examine differences in loan modification rates by borrowers’ race and ethnicity.6 None of these studies found significant disparities in loan modification outcomes for Black or Hispanic borrowers. In an early study on loan modifications, Collins and Reid (2010) examined data on subprime and Alt-A loans originated in 2005 in California, Oregon, and Washington, analyzing loan modification outcomes through 2010.
The results for these three states showed no evidence of lower modification rates for minority borrowers than for White borrowers, conditional on being delinquent.
The other three studies focused on borrowers’ outcomes in New York City, which has data systems that enabled the authors to build comprehensive datasets with a large number of control variables. In the first study, Been and her colleagues (2013) used a sample of first lien, prime, and subprime mortgages in New York originated between 2004 and 2008 and found that the race or ethnicity of the borrower has no significant impact on the likelihood that a seriously delinquent loan was modified between 2008 and 2010. They also found that neighborhoods with large shares of Black residents are more likely to receive modifications (even after controlling for other neighborhood-level factors that might influence delinquent loan outcomes). Chan et al. (2014), using a sample of subprime and Alt-A privately securitized loans originated in New York between 2003 and 2008 observed through 2010, found a higher loan modification propensity for Black and Hispanic borrowers, after controlling for a wide range of factors. In the third study, Voicu and his colleagues (2011) used a sample of New York loans from the OCC Mortgage Metrics database (which covers nine of the largest mortgage A couple of studies before the foreclosure crisis examined the influence of borrowers’ race on postdelinquency outcomes.
For example, using a large sample of Federal Housing Administration (FHA) loans, Ambrose and Capone (1996) investigated whether racial differences influence the resolution of loans that enter default. They found that minority borrowers remain in default longer than White borrowers, suggesting that lenders may actually have been more lenient toward minority borrowers. They also found that the foreclosure rate is consistent for both minority and White borrowers, conditional on being delinquent. These previous studies relied almost exclusively on FHA data, however, and do not include other factors (for example, credit score or equity position) that might influence postdelinquency borrower outcomes.
Cityscape 169Collins, Reid, and Urban
servicers) and found that HAMP loans are more effective at preventing default than proprietary loan modifications, after controlling for a wide range of variables. While they found that borrowers who receive HAMP modifications are less likely to redefault compared with those who receive proprietary modifications, Voicu et al. (2011) also found that the borrower’s race or ethnicity is not significantly correlated with the odds of redefault.
As Collins and Reid (2010) pointed out, however, it is hard to use datasets on loan performance to determine whether racial or ethnic differences influence the incidence of loan modifications, because the data do not enable researchers to see the number of borrowers filing applications to have their loan modified. Without application data, determining differences in the incidence of modifications ultimately is difficult. A study by Mayer and Piven (2012) used the publicly released HAMP data to assess whether racial minorities and Hispanics, women, and low-income homeowners benefited equally from HAMP. They concluded that race, ethnicity, gender, and income have “very little” impact on borrowers’ successful participation in HAMP. A subsequent study conducted by the General Accounting Office (GAO) using nonpublic HAMP data on four servicers found some differences in the incidence of HAMP modifications across protected classes, but these differences were, in large part, because of differences in servicers’ determination of borrowers’ eligibility related to their debt-to-income ratio and the completeness of their modification request (GAO, 2014).