«Urban Problems and sPatial methods VolUme 17, nUmber 1 • 2015 U.S. Department of Housing and Urban Development | Office of Policy Development and ...»
One significant limitation of the CTS data is its coverage of the mortgage market—in particular, the lack of coverage of prime loans and loans held by banks in portfolio. Nevertheless, given that subprime mortgages account for a significant share of all foreclosures and that most subprime loans that led to the crisis were privately securitized, this sample provides important insights into the performance of loan modifications for this segment of the market. Also, given the potential that modifications are more challenging among privately securitized loans (meaning loans not managed by Fannie Mae, Freddie Mac, or Ginnie Mae), this sample is particularly relevant for policymakers. Finally, one other strength of this data and study is their national coverage; as noted previously, other studies that have examined detailed loan modification outcomes by borrower race and ethnicity have tended to focus on borrowers in New York City (Been et al., 2013; Voicu et al., 2012) or in select states (Collins and Reid, 2010). Expanding the analysis to a national sample of loans enables us to determine the extent to which these more geographically targeted findings on loan default resolutions by race and ethnicity are nationally representative.
Findings In this section, we present the results of our analysis. First, we assess whether Black, Hispanic, and Asian borrowers receive less aggressive loan modifications, contingent on receiving a modification. Second, we explore the effectiveness of loan modifications in preventing redefault, again with a focus on differences across racial and ethnic groups.
Loan Modification Terms by Race and Ethnicity Because the success of modification is likely shaped by the type of modification a borrower receives, it is important to understand whether, conditional on modification, different borrowers receive differently structured modifications. In exhibit 2, we present three separate models to assess whether racial or ethnic differences influence the type of modification received. The
Total Observations 33,383 33,383 29,896 ARM = adjustable-rate mortgage. HAMP = Home Affordable Modification Program.
*p 0.05. **p 0.01. ***p 0.001.
Notes: Interest rate decreased and loan balance decreased are dummy variables regarding the type of modification. Payment change reflects the change in monthly payment before and after modification, recorded in percentage terms. Black, Hispanic, and Asian are based on Home Mortgage Disclosure Act data. Non-Hispanic White is the excluded group. ARM is a dummy for an adjustable rate mortgage. Income (logged) is at the time of application. HAMP-eligible is determined at the time of modification. House price change measures the difference in house prices at the ZIP Code level between date of modification and origination. More than 60 days behind before modification marks the delinquency status the month before modification.
The model also controls for FICOTM credit score and combined loan-to-value ratio quartiles, a no-documentation dummy, a prepayment penalty dummy, and metropolitan statistical area-level fixed effects.
Source: Corporate Trust Services
Cityscape 175Collins, Reid, and Urban
first two columns present the findings from a linear probability model assessing whether a modification entailed (1) a reduction in the loan interest rate and (2) a reduction in the loan principal or balance. The third column presents an ordinary least squares model in which the dependent variable is the percent change in the monthly mortgage payment. In addition to the variables in the exhibit, each model controls for FICO and CLTV quartiles, the percent change in house prices at the ZIP Code level from origination to modification (logged), the borrower’s income at origination (logged), a no-documentation indicator variable, a prepayment penalty indicator variable, and MSA-level fixed effects.
Importantly, we find very few differences in the likelihood of either interest rate reductions or balance decreases by race or ethnicity. Compared with White borrowers, Black, Hispanic, and Asian borrowers are equally likely to receive a modification that decreases the interest rate or principal balance, after controlling for a wide range of factors. When we examine the amount of change in monthly payments, we find that Black, Hispanic, and Asian borrowers all receive a slightly larger reduction in their monthly payments than White borrowers. The differences are slight, however, with the difference ranging from a 1.1-percent greater decrease in monthly payments for Black borrowers to 2.5 percent for Hispanic borrowers.
In terms of the other control variables, we find that loans with adjustable interest rates are more likely to result in interest rate and principal balance decreases and also a greater decrease in the monthly payments. We find that higher income borrowers are slightly more likely to receive a modification that entails a principal decrease, but again the effect is small.
We also find that living in ZIP Codes with lower house price declines (or house price increases) between origination and modification significantly decreases the likelihood of a borrower receiving a principal reduction, but it increases the likelihood of the borrower receiving a lowered interest rate and a greater reduction in monthly payments. For some servicers, a rising market may change their net present value, or NPV, calculation and reduce the likelihood that they would be willing to forgo principal. We also find that borrowers who are more than 60 days delinquent at the time of modification are slightly more likely to see interest rate reductions than borrowers who are only 2 months behind on their payments; however, serious delinquency appears to have no effect on the likelihood of principal decreases or the amount of payment relief the borrower is offered.
The model also shows that the HAMP-eligible indicator variable has a significant, positive effect on the type of modification a borrower receives. Conditional on receiving a modification, borrowers who fit within HAMP-eligibility criteria—including receiving their modification after HAMP was launched—are more likely to receive an interest rate or principal reduction on their loan. They also receive a much greater decrease in their monthly payments—on average, borrowers who could have received a HAMP modification see their monthly payments go down 15.6 percent compared with borrowers who receive a modification before HAMP was launched or who may not be eligible because their loan balance is too high.
Modifications and Loan Performance Our second assessment relates to the effectiveness of loan modifications and whether the modifications granted have been successful at keeping borrowers in their homes. We begin
by examining differences in loan performance in a descriptive framework. Exhibit 3 displays summary statistics for our sample of modified loans in the top panel and, for a comparable sample of nonmodified loans in the CTS data, in the bottom panel.17 Approximately 22 percent of the delinquent loans in our sample were modified between December 2006 and December
2012. Without controlling for other borrower or loan characteristics, we find that modification rates are higher for Hispanic and Black borrowers; approximately 33 percent of Black borrowers and 25 percent of Hispanic borrowers received a loan modification compared with 19 percent of Non-Hispanic White borrowers and 17 percent of Asian borrowers. Most loan modifications involve a rate reduction; 80 percent of all modifications include some form of interest forgiveness, with an average reduction of monthly payments of 29 percent. In contrast, only 18 percent of modifications include a reduction in the loan balance. Consistent with other studies, in these descriptive results, we do not find significant racial or ethnic differences in the types of modifications received. Overall, the incidence of various loan modification types and terms are remarkably consistent across racial and ethnic categories. We also find that approximately Exhibit 3
The nonmodified sample includes loans originated between 2004 and 2006 that were at least 60 days delinquent and active in June 2009 but that were not modified between December 2006 and December 2012.
one-third of modified loans end in foreclosure, despite the modification.18 Asian borrowers have the lowest levels of foreclosure after modification (26 percent), with White, Black, and borrowers classified as “Other” having slightly higher foreclosure rates than average.
For comparison, in the bottom panel of exhibit 3, we present the foreclosure rate for loans that were 60 days or more delinquent but not modified during our observation period. Overall, modified loans perform better than unmodified loans; 38 percent of unmodified loans in our sample end in foreclosure. The data also show that modifications reduced the foreclosure rate for minority borrowers, but not for non-Hispanic White borrowers. For instance, 42 percent of Black borrowers who did not receive a modification lost their home to foreclosure compared with 32 percent of Black borrowers who received a modification. In contrast, the foreclosure rate for non-Hispanic White borrowers is similar for delinquent loans that underwent modification and those that did not.
In exhibit 4, we examine serious delinquency rates for modified loans by race and ethnicity, measured at 6 and 12 months after modification. Approximately 7 percent of loans have missed at least three payments 3 months after modification (and are therefore 90 or more days late).
Within a year, nearly 18 percent of borrowers have missed at least three payments. In comparison with White and Black borrowers, Asian and Hispanic borrowers have slightly lower rates of being 90 or more days delinquent, at 6 months and 1 year after modification.
Although these descriptive statistics can help reveal overall trends in loan performance after modification, they do not control for borrower, loan, or housing market characteristics that might influence the likelihood that a borrower can continue to pay his or her mortgage, even with a modification. For this reason, we explore the effectiveness of loan modifications in a multivariate framework. Exhibit 5 presents a cross-sectional linear probability model that assesses loan performance 6 and 12 months after modification. As with the previous models, nonreported controls include FICO and CLTV quartiles, a no-documentation indicator variable, a prepayment penalty indicator variable, house price changes at the ZIP Code level from the time of modification to 6 or 12 months after origination, and MSA-level fixed effects.
Focusing first on differences by race and ethnicity, we find that only Black borrowers are slightly more likely to experience delinquency than White borrowers after receiving a modification. Specifically, 12 months after modification, Black borrowers are more likely to be 60 days delinquent, but the size of the effect is small. We also find that Hispanic borrowers are slightly less likely to be at least 30 days delinquent 6 months after modification. Overall, after controlling for a wide range of factors, we find very little variation in the effectiveness of modifications by race or ethnicity. Borrowers with higher incomes at origination are slightly less likely to become delinquent or go into foreclosure after modification.
HAMP eligibility also reduces the likelihood of redefault. Modifications that were HAMP eligible reduce the likelihood of foreclosure after 12 months by 6 percent, even after controlling for a wide range of other characteristics. We do not find a significant effect of house price changes on the likelihood of redefault; in part, this may be because of the fact that we are measuring house price change between the month of modification and 6 or 12 months later, which results in smaller differentials than our previous measure of house price change, which captured the higher price levels among loans originated before 2007 and the subsequent rapid drop in values during the crisis.
Perhaps counterintuitively, we find that borrowers who are seriously delinquent (90 days or more) at the time of modification are less likely to redefault after modification. Although this finding may in part be due to a selection effect (for example, servicers may be effectively identifying borrowers who need a modification to stay in their homes), it also suggests that modifications can be effective even for borrowers who are several months behind in their payments.
In terms of the modifications themselves, we find that interest rate reductions tend to reduce delinquencies (this is consistent at both 6 and 12 months after modification), but that only principal forgiveness reduces the likelihood of foreclosure a year after modification. When we run the same model but control for the percentage change in monthly payments, we find that a decrease in monthly payments reduces the likelihood of redefault and foreclosure across the board (exhibit 6). As we discuss in the conclusion that follows, these results point to the importance of understanding both the nature of modifications and the outcomes that are tracked post-modification to assess which modifications are the most effective at keeping borrowers in their home over the long term.
Total observations 27,973 27,973 27,973 27,940 22,645 22,645 22,645 22,411 ARM = adjustable-rate mortgage. HAMP = Home Affordable Modification Program.
*p 0.05. **p 0.01. ***p 0.001.