pathways after default: what happens to distressed mortgage borrowers and their homes?
TRANSCRIPT
Pathways After Default: What Happens to DistressedMortgage Borrowers and Their Homes?
Sewin Chan & Claudia Sharygin & Vicki Been &
Andrew Haughwout
# Springer Science+Business Media New York 2013
Abstract We use a detailed dataset of seriously delinquent mortgages to examine thedynamic process of mortgage default—from initial delinquency and default to finalresolution of the loan and disposition of the property. We estimate a two-stage competingrisk hazard model to assess the factors associated with post-default outcomes, includingwhether a borrower receives a legal notice of foreclosure. In particular, we focus on aborrower’s ability to avoid a foreclosure auction by getting a modification, by refinancingthe loan, or by selling the property. We find that the outcomes of the foreclosure processare significantly related to: loan characteristics including the borrower’s credit history,current loan-to-value and the presence of a junior lien; the borrower’s post-defaultpayment behavior, including the borrower’s participation in foreclosure counseling;neighborhood characteristics such as foreclosure rates, recent house price depreciationand median income; and the borrower’s race and ethnicity.
Keywords Mortgage . Default . Modification . Foreclosure . REO
J Real Estate Finan EconDOI 10.1007/s11146-012-9400-1
The views represented here are those of the authors and do not necessarily reflect those of the FederalReserve Bank of New York or the Federal Reserve System.
S. Chan (*)Robert F. Wagner School of Public Service, New York University, 295 Lafayette Street, New York,NY 10012, USAe-mail: [email protected]
C. SharyginUrban Institute, 2100 M Street, N.W., Washington, DC 20037, USAe-mail: [email protected]
V. BeenNew York University School of Law, 40 Washington Square South, New York, NY 10012, USAe-mail: [email protected]
A. HaughwoutFederal Reserve Bank of New York, 33 Liberty Street, New York, NY 10045, USAe-mail: [email protected]
Introduction
During the recent mortgage crisis, discussions of the impact of mortgage default andforeclosures on homeowners, neighborhoods and the housing market have oftenassumed that the process is relatively deterministic—if borrowers do not makemortgage payments, their properties are foreclosed, vacated and repossessed bylenders. However, mortgage default and foreclosure are less a single event than aprocess with uncertain duration and an eventual resolution that can vary widelyacross lenders, borrowers and neighborhoods. In this paper, we use a dataset of loansin New York City to examine how the post-default outcomes of non-prime mortgagesvary with borrower, loan and neighborhood characteristics. The richness of our databoth provides a more accurate and comprehensive set of controls than previousresearch and allows us to paint a clearer picture of what happens to distressedborrowers and their homes.
The complete process of mortgage default—from initial delinquency to finalresolution of the loan and disposition of the property—is relatively unexplored inthe literature, primarily because researchers have been unable to follow individualborrowers from loan origination to default to the beginning of the foreclosure processand through to specific resolutions such as a modification, the borrower refinancingthe mortgage or selling the property, or the lender demanding an auction sale of theproperty. Moreover, few post-default studies use data spanning the recent foreclosurecrisis in the United States, which saw markedly different types of borrowers and loanproducts than in previous years. By contrast, the dataset we have assembled containsdetailed recent information that allows us to examine the entire dynamic process. Weexplore a wide range of factors that are associated with a loan’s path after default,examining both the factors relating to a foreclosure notice, and those relating to theultimate resolution of any foreclosure. We are particularly interested in factors thatrelate to a household’s ability to avoid foreclosure by getting a modification,refinancing the loan, or selling the property.
The relevance of our research is underscored by the fact that despite significantefforts over the last few years to develop interventions to help borrowers keep theirhomes, there has been widespread dissatisfaction with the success of these measures.Policy-making, both at the government and at the lender and servicer level, shouldbegin with a better understanding of the factors associated with the various outcomesof defaults as the harm resulting from mortgage default will differ depending on howthe default is resolved.
To analyze the pathways from mortgage distress to ultimate resolution, we havecombined New York City data from a variety of sources, including: loan applicationand underwriting information such as borrowers’ credit scores and loan-to-valueratios (LTV); dynamic mortgage servicing records that include detailed loan terms,payment and termination information, including modifications; legal documents forthe property, including mortgages, foreclosure notices (lis pendens), and deed recordsspanning origination to termination of the loan and resale of the property; and avariety of property-specific and local neighborhood characteristics. In addition, wecalculate current LTV using repeat sale house price indices for community districts(political jurisdictions within New York City), that average just over four squaremiles each.
S. Chan et al.
Importantly, the data allow us to examine the entire complex sequence of decisionsdistressed borrowers and their lenders face. There are many possible courses of eventsonce a borrower is delinquent, even before a legal foreclosure notice is received. Someborrowers are able to catch up on payments or obtain a loanmodification. Others may beable to repay by refinancing. Some borrowers may choose to sell the property, whichforces them to move, but avoids foreclosure and the lowered credit rating and higherfuture borrowing costs that it entails. Borrowers might also ask to turn the property overto the lender before foreclosure proceedings are initiated (deed in lieu transfers). Finally,the lender may choose to forbear on the enforcement of its rights, and allow the borrowerto stay in the house without immediately pressing for repayment of the delinquentamounts. These possible post-default but pre-foreclosure outcomes are illustrated inFig. 1 as the first stage of the foreclosure process.
Borrowers who receive a lis pendens, a legal notice of foreclosure, then facemultiple possible outcomes. The borrower may still have the option to refinancethe loan or to sell the property and repay the lender. The borrower and lender may stillagree to modify the mortgage, or the lender may agree to accept a deed in lieu offoreclosure. The lender may choose to force the property to be offered at a foreclosureauction, and if the property does not sell for more than the lender’s reserve price, thelender may take back the property so that it becomes real estate owned (REO).Finally, the property may essentially stay in limbo, with the lender taking no actionto bring the property to auction. Figure 1 illustrates these outcomes in the secondstage of the foreclosure process. Each outcome may have different consequences forthe borrower, the lender, and the neighborhood.
We analyze the path to ultimate outcomes using dynamic competing risk modelsthat allow us to examine factors that may change along the path from mortgagedistress to resolution (for example, LTV may change as housing prices evolve overtime.) We implement a two-stage approach, first estimating the likelihood of a lispendens and pre-foreclosure resolutions such as a modification, refinance or sale,then estimating the likelihood of competing post-lis pendens outcomes, includingforeclosure auction. Such a model allows us to distinguish between factors associatedwith outcomes that occur before a foreclosure notice is served, and those associatedwith outcomes that occur afterwards.
Our empirical results show that the outcomes of the foreclosure process areassociated with a mix of many factors. Here, we briefly outline some of the mostimportant findings.
First stage of the foreclosure process Second stage of the foreclosure process
Default ModificationRefinanceSellDeed in lieuReceive lis pendens ModificationUnresolved (includes cure, Refinance
forebearance, remain delinquent) SellDeed in lieuForeclosure auction Auction saleUnresolved (includes cure, REO
forebearance, remain delinquent)
Fig. 1 The foreclosure process and possible post-default outcomes
What Happens to Distressed Mortgage Borrowers and Their Homes?
1. Loan characteristics. We find that borrower risk variables matter a great deal,but in ways that may at first seem counter-intuitive. After controlling for loanterms, borrowers with worse credit scores at origination are less likely to receivelis pendens and go to auction, and more likely to receive modifications orrefinance the loan on their own. This seemingly surprising result may signal thatafter controlling for loan terms, non-prime borrowers with higher credit scoresare adversely selected in ways that are unobservable to the researcher, but areobservable to servicers and underwriters so that lenders avoid giving theseborrowers modifications, and the borrowers find it difficult to obtain a newrefinance loan elsewhere. The result could also be due to the greater declines incredit scores that are experienced by borrowers with higher credit scores atorigination.Current LTV also has an important relationship to post-default outcomes.Underwater borrowers face a discontinuous hurdle in refinancing or selling becausethey either have to provide additional cash at the time of transfer or persuade thelender to accept less than the balance due. Consistent with this, we find thatborrowers with higher current LTVs are much less likely to avoid foreclosure byrefinancing their mortgage or selling their homes, but are more likely to receive amodification. This likely reflects both the lower likelihood that the lender will beable to recoup losses by going through the foreclosure process when LTVis high, aswell as the probability that lower LTV borrowers who cannot refinance areadversely selected and would not be able to repay, even with a modifica-tion. Having a junior lien is correlated both with a lower likelihood ofmodification, and increased likelihoods that the borrower will receive a lispendens and that the property will go to auction.
2. Borrower behaviors. Several borrower behaviors are associated with whether aloan is modified. Borrowers who make some payments post-default are morelikely to get a modification, but the effect is non-monotonic: the chance of amodification declines as the fraction of payments exceeds one half, presumablybecause these borrowers have demonstrated an increased ability to cure the loanon their own. Loans that are older by the time they default are more likely to bemodified, perhaps because the borrower has demonstrated a longer period ofprompt repayment. We also find that borrowers who received foreclosurecounseling are more likely to receive a modification, although we are not ableto discern what part of this effect is due to selection.
3. Neighborhood characteristics. Our dataset allows us to analyze neighborhoodcharacteristics at the census tract level, a level of geographic detail few studiesprovide. We find that high rates of foreclosure in the surrounding census tractduring the recent past are associated with an increased likelihood of a lis pendensand of a foreclosure auction, and a reduced likelihood of modification. We alsofind a positive association between the likelihood of modification and whetherthe community district in which the property securing the loan is located hassuffered recent house price depreciation. Foreclosure rates may be acting asleading indicators for further house price depreciation or may reflect greaterdepreciation than is captured in the community district level house price indices.If that is the case, while modifications are more likely in depreciating areas,
S. Chan et al.
lenders may be reluctant to modify loans in neighborhoods with high rates offoreclosure for fear that prices in those neighborhoods may be especially likely todecline further. For those neighborhoods, lenders may rather move through theforeclosure process quickly before prices fall further. Additional census tractlevel characteristics that are significantly associated with post-default outcomesinclude the share of non-prime mortgages originated in the 2 years before theloan’s origination, median income, and the tracts’ composition of residents byrace, ethnicity and immigrant status.
4. Borrower race and ethnicity. We find that Hispanic and Asian borrowers aremuch more likely to end up at a foreclosure auction. Both black and Hispanicborrowers have a greater likelihood of modification.
Our results are largely descriptive. It is important to keep in mind that each of thepost-default outcomes is the result of the interaction between the lender, the servicerof the loan and the borrower. As such, the interaction is likely complex, with servicerspotentially having incentives unknown to the borrower (and possibly at odds with theinterests of the lender).1 This may lead to involved game theoretic interactions thatare beyond the scope of this paper. We have not attempted to model this interaction,2
but rather have documented the relevant facts about the outcomes that should beuseful immediately to borrowers, lenders, servicers, non-profit organizations andpolicymakers and may serve to motivate more theoretical modeling of post-defaultprocesses in the future.
Previous Literature
There is a great deal of research about the factors that determine whether and when aborrower defaults. Our focus, however, is on the determinants of post-default outcomes.Early, pre-housing crisis research on the factors associated with post-default outcomesfocused primarily on describing the characteristics of households whose homes wereforeclosed. Lauria et al. (2004), for example, surveyed borrowers who suffered between1985 and 1990, and found that neither the borrowers’ race nor the racial composition oftheir neighborhoods affected the length of time between initial delinquency andforeclosure. Ambrose and Capone (1996), on the other hand, found that lenderstended to offer non-white borrowers more time before initiating foreclosure, butthat the eventual foreclosure rates of white and nonwhite borrowers werevirtually equal. Other early work also focused on the characteristics of themortgages that determined post-delinquency outcomes. Ambrose and Capone(1998), for example, analyzed a sample of defaulted FHA loans and found thatLTV is a major predictor of whether the delinquent loan is reinstated, sold,assigned to HUD, or foreclosed. Capozza and Thomson (2006) analyzed asample of defaulted subprime mortgages issued by a single lender and found
1 Further, borrowers typically are faced with mortgage distress on only a single or a very small number ofoccasions in their lifetimes, and therefore may have limited knowledge as to the best possible course ofaction.2 For promising advances in modeling the modification decision, see, for example, Haughwout et al.(2009).
What Happens to Distressed Mortgage Borrowers and Their Homes?
that loans with a higher interest rate premium were less likely to enter REO,and that foreclosure is more likely for loans with fixed interest rates, standarddocumentation and high LTVs.
The current economic crisis has spawned a larger literature exploring the characteristicsassociated with different post-default outcomes. Gerardi et al. (2007) used property-levelinformation (mortgage documents and transaction deeds) to compare the likelihood of saleand foreclosure auction for subprime and prime mortgages. They found that subprimeborrowers are five to six times more likely to lose their homes to foreclosure auction thanprime borrowers, and that high initial LTVs coupled with declines in city-level housingvalues are significantly associated with foreclosure auction. In an update, Gerardi andWillen (2009) matched mortgage documents with Home Mortgage Disclosure Act(HMDA) data, and found, focusing on multi-family (2–4 unit) properties, that black andHispanic borrowers’ properties are more likely to go to auction. Subprime borrowers andnon-owner occupiers also were more likely to sell their homes. The authorscould not distinguish, however, between forced sales resulting frommortgage default orthe initiation of foreclosure proceedings and voluntary sales that may have netted profitsfor the owner.
Another approach to analyzing the determinants of post-default outcomes usesinformation from loan servicers on the actions taken by borrowers and lenders in eachmonth over the life of the loan. Danis and Pennington-Cross (2005), for example,used a sample of fixed rate mortgages (FRMs) originated between 1996 and 2003from LoanPerformance and found that higher credit scores and longer periods ofdelinquency are associated with higher probabilities of prepayment than of foreclosure,while negative equity is associated with a greater probability of entering foreclosure thanprepayment. Using a larger sample over the same time period, Danis and Pennington-Cross (2008) found that delinquent borrowers with higher credit scores are less likely toenter foreclosure, but (contrary to their earlier paper) also are less likely to prepay. Theyalso found that local house price depreciation and volatility are importantpredictors of entering foreclosure, but that unemployment rates are not.Pennington-Cross and Ho (2010) found that the factors associated with theinitiation of foreclosure or becoming REO for adjustable rate mortgages(ARMs) are similar to those for FRMs, but differ in magnitude, and that thesize of payment shocks when ARMs adjust, not surprisingly, is significantlyassociated with foreclosure, REO and prepayment.
More recently, researchers have explored a fuller range of outcomes that mayoccur once a borrower becomes seriously delinquent. Pennington-Cross (2010) useda sample of subprime FRMs originated between 2001 and 2005 to estimate thecompeting probabilities that the loans in the foreclosure process would be paid off,become current, improve from “foreclosed” to delinquent status, remain in foreclosure,or enter REO. He found that loans with a higher fraction of months in delinquency priorto foreclosure, and fewer months spent in foreclosure, are relatively more likely tobecome REO than to prepay. Voicu et al. (2011) merged LoanPerformance with HMDAdata to analyze the factors associated with a variety of pre- and post-foreclosureresolutions of loans in default, including prepayment, cure, partial cure and REO.They found that black borrowers in default are less likely to enter foreclosure than arewhites, and Asian borrowers are less likely to cure and more likely to have theirproperties become REO. They also found that default outcomes are associated with a
S. Chan et al.
variety of loan characteristics, local legal, economic and housing market conditions, aswell as equity in the home.
Another set of papers have focused on the factors associated with loan modification.Some of these papers are limited by the lack of a direct indicator from the servicer dataabout whether a mortgage has been modified.3 Using LoanPerformance, which doeshave this information, Haughwout et al. (2009) analyze how the characteristics ofborrowers, loans, and modifications are associated with the probability that a borrowerreceiving a modification redefaults. Their descriptive statistics suggest that pre-HAMPmodifications were more likely (without controlling for other variables) to go toborrowers with lower credit scores and higher origination LTVs and debt-to-incomeratios (DTIs), and to those holding ARMs. Collins and Reid (2010) merged HMDA datawith that from a servicer that includes modification information, and found that loanswith high relative interest rates at origination and black borrowers are associated with agreater likelihood of modification, while borrowers with ARMs and higher DTIs areassociated with a lower likelihood. Been et al. (2011) analyzed a New York Citysubsample of the OCC-OTS Mortgage Metrics database (which also provides directinformation from servicers onmodifications) merged with borrower characteristics fromHMDA, property and neighborhood information. Among many results, they found thatloans with higher current LTVs are associated with a greater likelihood of modification,while loans with junior liens and borrowers with higher current credit scores areassociated with a lower likelihood. Agarwal et al. (2011) used the national OCC-OTSdatabase to study the determinants of liquidation versus “renegotiation” (which theydefine as modification, repayment plans, and refinancing), and found that seriouslydelinquent securitized loans are significantly less likely to be renegotiated than similarbank-held loans.
Finally, researchers also have begun to examine the efficacy of efforts many non-profitorganizations and local governments are making to provide counseling to borrowers atrisk of foreclosure. That research generally finds that counseling is associated withimproved outcomes, but it is difficult to distinguish between the impact of counselingand self-selection into counseling in these results. Collins and O’Rourke (2011)summarize the literature well; in addition to the studies in that review, Been etal. (2011) found that counseling is associated with a greater likelihood that aborrower will modify the loan and avoid liquidation.
In sum, the evidence about post-default outcomes has some consistent themes, butis often hard to harmonize and reconcile because of different definitions of the variouspost-default outcomes, differences and limitations in the data and methodologies used,and differences between the types of loans studied. While recent studies that havedetailed information on borrowers’ and lenders’ decisions in every month over the lifeof the loan improve upon studies that only observe at origination and at the eventualoutcome, even those more recent studies are limited by the lack of information recordedby the servicer on the actual property. For example, they often cannot distinguishbetween loans that were refinanced and properties that were sold, and cannot controlfor property- or neighborhood-level characteristics that may be important. Our analysisadvances the study of post-default outcomes bymerging dynamic loan level information
3 For example, Piskorski et al. (2010); Foote et al. (2009); Adelino et al. (2009, 2010).
What Happens to Distressed Mortgage Borrowers and Their Homes?
with extensive data about the borrowers, the properties, and the neighborhoods in whichthe properties are located.
Data Description
Mortgage Data from New York City
The starting point for our analysis is the mortgage dataset first described inChan et al. (2010). The dataset consists of all first-lien subprime and alt-Asecuritized mortgages originated between 2003 and 2008 in New York City thatare in CoreLogic’s LoanPerformance, a commercial database that covered about90 % of all non-prime securitized mortgages in the United States during thetime period we study. LoanPerformance provides information on borrowercharacteristics at origination, loan terms, monthly repayments and balances,and loan modifications.4 These LoanPerformance records were matched toland parcels using real property deeds from the New York City Departmentof Finance (DOF)’s Automated City Register Information System (ACRIS).5
That match gave a unique geographic identifier for each mortgaged propertythat allowed additional data from other sources to be merged on, including: (i)additional borrower characteristics (such as race and ethnicity) from HMDA,(ii) ACRIS information on whether the borrower took on additional mortgagedebt following loan origination, (iii) property characteristics from DOF’s RealProperty Assessment Database, (iv) data from the Center for New York CityNeighborhoods on whether the borrower received free foreclosure preventioncounseling,6 (v) repeat sales house price indices for 56 different communitydistricts of New York City from the Furman Center for Real Estate and UrbanPolicy, (vi) the rates of mortgage foreclosures and properties owned by lenders(REOs) among all 1–4 family properties in each census tract, and (vii) a varietyof census tract level variables from the 2000 Census and HMDA. Afterdropping loans that did not match to ACRIS property identifiers, a few loanswith missing values of key variables, and loans that were already in default bythe time they appeared in the LoanPerformance database, the resulting datasethas 140,033 mortgages and a wealth of information on the loan terms, theborrowers, the properties and their neighborhoods.
4 Haughwout et al. (2009) find that 92 % of modified loans in the complete LoanPerformance modificationsdatabase were reported by the servicer and 8%were inferred by LoanPerformance based on changes in the loanterms.5 The hierarchical matching algorithm used is described in more detail in Chan et al. (2010). Of the loans inthe LoanPerformance database, 93 % were matched back to the deed records. The deeds data do not includeStaten Island, which therefore was dropped from the analyses.6 The Center for New York City Neighborhoods (CNYCN) is a non-profit organization that coordinatesforeclosure counseling from a variety of non-profit providers to homeowners at risk of losing their home toforeclosure. Distressed mortgage borrowers who call 311 (New York City’s widely known phone numberfor government information and non-emergency services) are directed to CNYCN. It is likely that thecounseling reported by CNYCN represents the vast majority of foreclosure counseling taking place in NewYork City.
S. Chan et al.
In our analysis, we define borrower default as being at least 90 days past due onpayments. To determine whether and when the property entered foreclosure, we mergedon property-specific foreclosure information that allows us to tell whether any refinance,sale to a third party or transfer to the lender occurred before or after the initiation offoreclosure proceedings. The ACRIS deed records are used to classify all post-defaultproperty transfers as arms-length transactions to a third party, deed transfers to the lenderin lieu of foreclosure proceedings, auction sales to a third party, or REO, and the ACRISmortgage records are used to distinguish loan terminations associated with a refinance ofthe loan from those associated with a sale of the property.7 We are thus able to track eachof the post-default outcomes displayed in Fig. 1. By the end of the observation period inOctober 2010, about a third of all borrowers had been 60 days or more delinquent onpayments at some point in our study period, and the vast majority of these had falleneven further behind and entered 90-day delinquency. In our estimation sample of loanrecords over time, loans are followed from their first 90 day delinquency to foreclosureresolution, or until October 2010. This gives us 40,218 loans and a total of 632,345 loan-months. A subsample of these (30,258 loans) also matched to HMDA data.
In New York State, foreclosure is a judicial process that begins when the lender or otherholder of the mortgage files a complaint and a lis pendens, or legal notification thatforeclosure proceedings have begun.8 After the lis pendens is issued, the borrower has achance to answer the complaint and dispute the lender’s allegations of default. But in about90 % of foreclosures, the borrower does not answer, and thereby defaults (in a legal sense).Upon default (or if the borrower does answer and disputes the lender’s right to foreclose, butthe lender is able to prove the default), the judge appoints a referee to determine the amountof money the borrower owes, and the lender can then apply for a judgment directingforeclosure and sale. In contrast, in non-judicial foreclosure states, lenders can issue a noticeof default, then proceed quickly to seize the property for auction if the borrower is unable toremedy the default.9 Unlike several states, New York does not have a right of redemptionafter the foreclosure sale, and the purchaser of the property gains possession at the sale. BySeptember 2010, the average time that a foreclosed mortgage had spent in delinquency was16 months nationally, compared to 19 months in New York State.10
While our data is from New York City where Manhattan is a fairly atypical housingmarket, the other counties we study—Brooklyn, the Bronx and Queens—are where mostof NewYork City’s foreclosures have occurred. These other counties are like many otherlarge cities in the U.S. in terms of the nature and quality of the housing stock, density, andneighborhood demographics. Where New York City may differ from the rest of thenation is in the length of the foreclosure process, as described above.
7 The deed records are categorized by the DOF as standard transactions, deeds-in-lieu, debt free gifts ordivorce settlements, estate sales, other judgments, and referee sales (or auction sales). We identify auctionedproperties that became REO by flagging deeds that transfer to the lender.8 The New York legislature amended the process in 2008 to require that the process begin with a 90 daynotice of intent to file a foreclosure action for certain “high-cost”, “subprime” and “non-traditional” homeloans, effective September 1, 2008. That requirement was extended to all home loans in 2009, effectiveJanuary 14, 2010. We found little difference in our results below when we compared the periods before andafter these changes went into effect.9 For descriptions of judicial and non-judicial foreclosure systems, see, e.g., Crews Cutts and Merrill(2008).10 Lender Processing Services, 2010. This timeline is likely to have extended since the end of our studyperiod.
What Happens to Distressed Mortgage Borrowers and Their Homes?
Distribution of Post-Default Outcomes
Table 1 summarizes the distribution of post-default outcomes. Just over half of the40,218 defaulted loans received a lis pendens by October 2010. About one fifth of theloans were modified; the majority of these were modified before receiving a lispendens. One fifth of the defaulted loans prepay through refinances and sales; themajority does so by sale of the property, with over two thirds of these selling afterreceiving a lis pendens. The smaller number of refinances that occur, in contrast, doso mostly before receiving a lis pendens.
Of the properties that received a lis pendens, 14%went to auction. The majority of these(86 %) failed to sell and reverted to bank ownership. Because so few of the foreclosureauction properties actually sell, in our analyses below, we combine the auction sale andREOoutcomes into one foreclosure auction outcome. A negligible share of properties wasrecordedwith a deed in lieu transfer, so we drop the pre-lis pendens deed in lieu observationsand group the post-lis pendens observations with the REO properties in our analysis.
The “unresolved” category in Table 1 includes loans that were not matched to anoutcome using recorded documents from ACRIS. This includes loans that defaulted latein the observation period, as well as a very small number of loans that are right censoredin the LoanPerformance data before the end of the observation period. It also includesdefaults that may have been resolved by the borrower or the lender in a way not recorded inour loan records. In particular, we are unable to observe forbearance or other workouts thatallow the borrower to catch up on missed payments. To briefly delve into this category ofloans, we further classify them in Table 1 as “cured”, meaning the borrower becomes currenton the loan after default and remains current through the end of the observation period inOctober 2010 (3 % of all loans); “remained delinquent”, meaning the borrower never
Table 1 Distribution of post-default outcomes
All loans in default Received lis pendens
No Yes
Post-default outcome
Modified 8,650 21.5 % 5,697 30.5 % 2,953 13.7 %
Refinance 1,566 3.9 % 1,097 5.9 % 469 2.2 %
Arms-length sale 6,061 15.1 % 1,771 9.5 % 4,290 19.9 %
Deed-in-lieu 8 0.0 % 1 0.0 % 7 0.0 %
Foreclosure auction: 2,950 7.3 % 2,950 13.7 %
Auction sale 424 1.1 % 424 2.0 %
REO 2,526 6.3 % 2,526 11.7 %
Unresolved: 20,983 52.2 % 10,103 54.1 % 10,880 50.5 %
Cured 1,034 2.6 % 782 4.2 % 252 1.2 %
Cycled 2,368 5.9 % 1,184 6.3 % 1,184 5.5 %
Remained delinquent 17,581 43.7 % 8,137 43.6 % 9,444 43.8 %
Total number of loans 40,218 100 % 18,669 100 % 21,549 100 %
Sample: LoanPerformance first lien non-prime securitized mortgages originated 2003–2008 in New YorkCity that are ever 90 days delinquent by October 2010
S. Chan et al.
became current again after first entering default (44%); and “cycled”, meaning the borrowerbecame current after initial default, but subsequently returned to 90 day delinquency (6 %).While other analyses treat cure, or partial cure, as a terminal outcome, we are reluctant to call“cured” a final outcome because of the large number of loans that temporarily cure but endup delinquent or in default again, even in our limited sample timeframe. These “cured” loanswill likely end up in one of the other categories given time.11
Descriptive Statistics
Table 2 displays summary statistics for our sample of defaulted loans. We split these resultsby whether the loan received a lis pendens, and by the unconditional post-default outcome—modification, refinance, sale, auction, and unresolved. Unless otherwise noted, the tablerecords the fraction of loans within the column outcome that have the row characteristic.
About one quarter of the loans are interest-only, another quarter are non-interestonly FRMs, and one third are non-interest only ARMs. There are roughly equalnumbers of home purchase and refinance loans, and 92 % of borrowers claim to beowner-occupiers. The borrowers’ average FICO score at origination is 652,12 and theaverage DTI is 42 %, although only 29 % of the borrowers provided full documen-tation. One third of the loans have a junior lien; the average combined LTV is 92 % atthe time of initial default, and grows to 109 % by the last observation month.13 Theaverage loan balance at initial default is $421,750. On average, borrowers make apayment in 42 % of the months after entering default and very few (just 2.5 %) havereceived foreclosure counseling by the last month of observation.
Foreclosure rates in the mortgaged property’s census tract are, on average, twopercent in the 6 months prior to initial default.14 House prices in the property’scommunity district declined by an average of 8 % in the year preceding initial default.The average share of black residents in a property’s census tract was 48 % (based on the2000 Census), reflecting both a higher proportion of blacks in our sample of non-primeloans in default, and the relatively high levels of residential racial segregation in NewYork City such that blacks are more likely to locate in predominantly black
11 Our findings can be roughly compared with several other studies: Agarwal et al. (2011) observed that1 year after a loan becomes seriously delinquent, about half of borrowers in a national sample of the OCC-OTS Mortgage Metrics database are in liquidation, and about one quarter each were renegotiated or had nofurther action. Been et al. (2011) followed loans in a New York City subsample of the OCC-OTS databaseand found that 9 % received a modification, 8 % other workouts, about 15 % were cured, 5 % experiencedliquidation, and almost two thirds remained in serious delinquency during the study period. Capozza andThomson (2006) observed 6,000 mortgages from a single servicer, and found that of the mortgages thatwere delinquent but not in bankruptcy at the beginning of the study period, after 8 months 38 % remained indefault but did not fall further behind on payments, 21 % remained in default with worsening delinquency,6 % had cured, 11 % had entered bankruptcy proceedings and 24 % had become REO.12 The widely used Fair Isaac Corporation (FICO) credit score depends on credit and payment history,credit use and recent searches for credit. According to FICO, 27 % of the general population have scoresbelow 650.13 The numerator of the combined LTV measure includes the balance for the first lien (the focus of ouranalysis) as well as any other liens in existence at the time of origination. While we know the size of anyadditional liens taken out afterwards, some are home equity lines of credit, and so would be misleading toinclude in our LTV measure.14 Specifically, the number of foreclosure notices issued on 1–4 family buildings in a census tract during thepreceding six-months, divided by the stock of 1–4 family buildings.
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le2
Descriptiv
estatistics
Allloansin
default
Receivedlis
pendens:
Post-defaultoutcom
e:
No
Yes
Modification
Refinance
Sale
Auctio
nUnresolved
Atorigination:
Loantype:
FRM
(not
interestonly)
0.24
0.30
0.19
0.28
0.30
0.20
0.11
0.25
ARM
(not
interestonly)
0.35
0.31
0.38
0.27
0.55
0.49
0.44
0.30
Interestonly
0.23
0.22
0.24
0.21
0.08
0.17
0.24
0.26
Other
loan
type
0.18
0.16
0.19
0.22
0.06
0.13
0.21
0.17
FRM
relativ
einterestrate
atoriginationa
0.30
0.34
0.27
0.35
0.50
0.34
0.20
0.27
ARM
relativ
einterestrate
atoriginationb
1.09
0.84
1.30
0.75
2.85
2.15
1.62
0.72
Average
FICO
score
652
649
653
638
606
643
663
661
Average
debt
toincome
4242
4343
4142
4442
Fulldocumentatio
n0.29
0.33
0.26
0.37
0.44
0.29
0.21
0.26
Non-cashoutrefinanceloan
0.05
0.07
0.04
0.06
0.05
0.03
0.02
0.06
Cash-outrefinanceloan
0.46
0.55
0.39
0.59
0.66
0.39
0.22
0.46
Owner-occupier
0.92
0.92
0.92
0.95
0.94
0.91
0.94
0.91
From
HMDAc :
Fem
ale
0.41
0.40
0.42
0.43
0.44
0.42
0.40
0.41
Has
co-borrower
0.21
0.27
0.16
0.25
0.37
0.20
0.10
0.20
Hispanic
0.19
0.18
0.19
0.18
0.09
0.18
0.21
0.19
Black
0.38
0.35
0.39
0.40
0.38
0.38
0.37
0.36
Asian
0.09
0.09
0.08
0.07
0.07
0.08
0.12
0.09
Average
shareof
non-prim
eloansin
tract
0.22
0.22
0.23
0.23
0.20
0.22
0.24
0.22
S. Chan et al.
Tab
le2
(con
tinued)
Allloansin
default
Receivedlis
pendens:
Post-defaultoutcom
e:
No
Yes
Modification
Refinance
Sale
Auctio
nUnresolved
Inthedefaultmonth:
Average
monthsbetween
defaultandorigination
2632
2128
1717
1330
Average
loan
balance$
421,750
409,608
432,269
418,957
311,858
399,354
434,081
435,832
Average
combinedLT
V%
9293
9294
6681
8998
Has
junior
lien
0.35
0.29
0.41
0.29
0.14
0.33
0.55
0.37
Average
grow
thin
ARM
paym
ents
sinceorigination
25–50%
0.03
0.03
0.03
0.05
0.03
0.03
0.02
0.03
50%
0.01
0.02
0.01
0.01
0.01
0.01
0.01
0.02
Average
past6monthsforeclosure
rate
intract%
0.020
0.019
0.020
0.020
0.013
0.017
0.020
0.021
Inthelastobservationmonth:
Average
monthssincedefault
1713
2112
910
1822
Average
shareof
paym
entsmade
sincedefault
0.42
0.40
0.45
0.42
0.61
0.59
0.57
0.35
Average
pastyear
localhouse
pricegrow
th−0
.08
−0.08
−0.07
−0.12
0.07
0.02
−0.07
−0.10
Average
combinedLT
V%
109
104
114
111
6684
101
120
Receivedmortgage
counselin
g0.025
0.026
0.024
0.036
0.004
0.004
0.006
0.031
Censustractaverages
from
2000:
Share
ofblackresidents
0.48
0.46
0.49
0.51
0.52
0.48
0.52
0.45
Share
ofHispanicresidents
0.23
0.22
0.23
0.21
0.20
0.23
0.23
0.23
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le2
(con
tinued)
Allloansin
default
Receivedlis
pendens:
Post-defaultoutcom
e:
No
Yes
Modification
Refinance
Sale
Auctio
nUnresolved
Share
ofAsian
residents
0.06
0.07
0.06
0.06
0.06
0.07
0.06
0.07
Share
offoreignborn
0.37
0.37
0.37
0.37
0.35
0.36
0.35
0.37
Medianincome
40,434
41,719
39,322
41,747
41,501
39,959
39,805
40,039
Num
berof
loans
40,218
18,669
21,549
8,650
1,566
6,061
2,950
20,983
Sam
ple:
LoanP
erform
ance
firstliennon-prim
esecuritized
mortgages
originated
2003–200
8in
New
YorkCity
that
areever
90days
delin
quentby
Octob
er20
10.The
table
indicatesthefractio
nof
loanswith
inthecolumnou
tcom
ethat
have
therow
characteristic,un
less
otherw
iseno
ted
aThe
loan’sinterestrate
minus
theFreddie
Mac
averagerate
forprim
e30
-yearfixedrate
mortgages
during
themon
thof
origination
bThe
loan’sinitial
interestrate
minus
thesix-month
LondonInterbankOffered
Rate(LIBOR)at
origination
cThese
statisticsarefrom
thesubsam
pleof
30,258
loansthat
matched
toHMDA
S. Chan et al.
neighborhoods. The data for borrower race and ethnicity in Table 2 come from theHMDA subsample: 38 % of borrowers are black, 19 % are Hispanic, and 9 % are Asian.
Exploring the Factors Affecting Post-Default Outcomes
Empirical Strategy
Our empirical strategy employs a discrete time proportional hazard framework with com-peting risks to explore the factors associatedwith post-default outcomes. In each stage of ouranalyses, the probability that mortgage iwill transition to outcome j at time t, conditional onnot having previously transitioned to any outcome is given by the following hazard H:
Hitðoutcome ¼ jÞ ¼ MNL ðθj months since default þ bjloan and borrower characteristicsit
þ g jborrower behaviorsit þ @jneighborhood characteristicsit
þ ajcalendar and origination year fixed effectsÞ
where MNL is the standard multinomial logit functional form, and αj , βj , γj , δj and θjrepresent vectors of coefficient estimates for each outcome j.
We include time since default among the covariates to allow the hazard rate to betime-dependent. To control for city-, state-, or nation-wide macroeconomic factorssuch as unemployment rates and city-wide house price movements, we includecalendar year fixed effects. The origination year fixed effects are intended to pickup any city-wide systematic changes in mortgage characteristics over time, includingaverage borrower risk and underwriting standards. To control for unobserved hetero-geneity and possible dependence among observations for the same loan, we use arobust variance estimator that allows for clustering by loan.
We start by estimating a two stage process, as summarized in Fig. 1. The first stagecorresponds to the time up to the filing of a legal foreclosure notice (lis pendens) by thelender. Five outcomes are possible: (i) the loan is modified, (ii) the loan is refinanced,(iii) the property is sold in an arms-length transaction, (iv) a lis pendens is filed, and (v)no resolution (none of the above outcomes). The second stage is conditional on receiptof a lis pendens. The five possible outcomes are: (i) the loan is modified, (ii) the loan isrefinanced, (iii) the property is sold in an arms-length transaction, (iv) the lender sells theproperty at auction, or retains as REO, and (v) no resolution (none of the aboveoutcomes). We should note that the goal of these empirical models is descriptive anddo not necessarily allow causal inference. In particular, the sample in the second stage isonly loans that have received a lis pendens; clearly, a selected group.15
We also estimate a reduced form collapsed model whereby the five unconditionaloutcomes correspond to those in the second stage above. Thus, relative to the twostage process, the collapsed model amalgamates the pre and post–lis pendens
15 Incorporating a selection bias correction in the second stage model is difficult or impossible in theabsence of a compelling variable that affects the receipt of a lis pendens, but that does not belong in theultimate outcome model.
What Happens to Distressed Mortgage Borrowers and Their Homes?
versions of the outcomes: (i) modification, (ii) refinance, (iii) sale and (iv) noresolution. Lender sale at auction is the fifth unconditional outcome.
Results
Table 3 presents our results. The first set of columns gives estimates for the first stagetransitions. In the second set of columns are the estimates for the second stage, which isconditional on having received a lis pendens. Relative risk ratios (and p-values) arereported, relative to the no resolution outcome.16 The third set of columns containsanalogous estimates for the reduced form collapsed model. In Table 4, we present resultsof the same set of models, but include the HMDA variables. Although this is thesubsample of loans that matched to HMDA, the coefficients on the non-HMDAvariables are very similar to those in Table 3. Throughout this section, we will describea variable as associated with a greater likelihood of a particular outcome (or as “morelikely”) when the estimated relative risk ratio for the variable is greater than one.17 Ourresults here are descriptive and it is important to note that while post-default outcomeswill depend on both observable and unobservable factors, we can only incorporate thosethat are observable in this analysis.
Loan Characteristics
The first set of rows in Table 3 shows the association between the various outcomes and theloan type interacted with the relative initial interest rate on the loan.18 FRMs and ARMswith relatively lower interest rates at origination, as well as interest only mortgages, are lesslikely to receive a lis pendens than are non-standard mortgages (the reference group) andloans with relatively higher interest rates at origination. These results may reflect a belief onthe part of lenders that borrowers who have chosen or qualified for less exotic and lessexpensive mortgages may be more likely to cure or refinance, which could lead lenders todelay filing a lis pendens even when faced with serious delinquency. Similarly, the lowestinterest rate FRMs and ARMs are significantly and substantially less likely to end up at aforeclosure auction.19 Interest only loans and FRMs are more likely to be modifiedcompared to other product types, and among FRMs, higher interest rates are associatedwith a greater likelihood of modifying, particularly after receiving a lis pendens. Lower
16 The relative risk ratio is the exponentiated value of the multinomial logit coefficient.17 A relative risk ratio greater than one in the MNL framework implies a greater probability of thatparticular outcome as compared to the reference outcome. It does not necessarily imply a greater absoluteprobability of that particular outcome, because the absolute probability will also depend on the baselineprobabilities of the other outcomes.18 The relative interest rate at origination for FRMs is calculated as the loan’s interest rate minus the FreddieMac average rate for prime 30-year FRMs during the month of origination. For ARMs, it is the loan’s initialinterest rate minus the six-month London Interbank Offered Rate (LIBOR) at origination.19 Our results are contrary to Capozza and Thomson (2006) who find that loans with high interest ratepremia are less likely to be foreclosed, and that FRMs, loans with standard documentation and high LTVsare more likely to be foreclosed. The difference may reflect the different time periods studied (2001–2002originations for Capozza and Thomson versus 2003–2008 here), the different populations studied (6,000subprime loans from one lender, versus most non-prime loans originated in New York City here), or the factthat we look at the full range of post-default outcomes that may occur (rather than the more restrictedoutcomes they study), during a time when the federal, state and local governments were incentivizingmodifications.
S. Chan et al.
Tab
le3
Foreclosure
models—
fullsample
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Loancharacteristics:
Interestonly
loan
1.364**
1.231
0.847
0.816**
1.984**
0.738
1.001
0.850*
1.574**
1.072
0.915
0.872+
(0.072)
(0.237)
(0.090)
(0.022)
(0.149)
(0.215)
(0.068)
(0.066)
(0.069)
(0.168)
(0.052)
(0.066)
FRM
relativ
erate
atorigination:
<1%
1.113*
1.894**
0.907
0.783**
1.679**
1.480
0.909
0.648**
1.314**
1.916**
0.886+
0.588**
(0.058)
(0.362)
(0.110)
(0.028)
(0.139)
(0.386)
(0.081)
(0.087)
(0.059)
(0.291)
(0.062)
(0.079)
1–2%
1.130*
1.755**
1.017
0.892**
1.739**
1.552+
0.941
1.019
1.333**
1.748**
0.925
1.008
(0.065)
(0.342)
(0.126)
(0.034)
(0.141)
(0.410)
(0.081)
(0.110)
(0.063)
(0.273)
(0.065)
(0.106)
2–3%
1.170+
1.597*
1.036
0.888*
2.013**
1.596+
1.022
1.232+
1.469**
1.623**
1.001
1.127
(0.100)
(0.334)
(0.151)
(0.047)
(0.209)
(0.450)
(0.095)
(0.152)
(0.097)
(0.272)
(0.079)
(0.141)
>3%
1.417**
1.460
0.935
0.907
2.258**
0.696
0.987
0.937
1.800**
1.197
0.956
0.879
(0.168)
(0.376)
(0.186)
(0.070)
(0.338)
(0.257)
(0.128)
(0.202)
(0.165)
(0.252)
(0.103)
(0.192)
ARM
relativ
erate
atorigination:
<1%
0.785**
1.385
0.819
0.615**
1.290+
0.160+
0.852
0.736+
0.982
1.082
0.795*
0.674*
(0.068)
(0.365)
(0.135)
(0.030)
(0.186)
(0.170)
(0.114)
(0.122)
(0.073)
(0.265)
(0.081)
(0.109)
1–2%
0.921
1.259
0.817
0.896**
1.055
1.086
0.972
0.864+
1.001
1.266
0.914
0.838*
(0.055)
(0.274)
(0.104)
(0.030)
(0.085)
(0.315)
(0.072)
(0.066)
(0.048)
(0.217)
(0.059)
(0.064)
2–3%
0.860*
1.556*
1.002
1.005
1.076
0.929
0.955
0.997
0.967
1.319+
0.966
1.007
(0.053)
(0.278)
(0.106)
(0.032)
(0.077)
(0.223)
(0.061)
(0.068)
(0.044)
(0.187)
(0.053)
(0.070)
>3%
0.930
1.561*
1.123
1.053
0.942
1.437
1.017
1.049
0.970
1.540**
1.057
1.093
(0.100)
(0.293)
(0.135)
(0.048)
(0.109)
(0.359)
(0.077)
(0.106)
(0.076)
(0.230)
(0.067)
(0.110)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Current
loan
balancequintile:
2nd
1.099+
1.123
1.214*
1.050
1.031
0.652**
1.158**
1.167*
1.079+
0.948
1.201**
1.249**
(0.059)
(0.123)
(0.101)
(0.032)
(0.084)
(0.095)
(0.064)
(0.092)
(0.049)
(0.082)
(0.055)
(0.099)
3rd
1.105+
1.004
1.428**
1.139**
1.017
0.542**
1.159*
1.243**
1.067
0.823+
1.281**
1.401**
(0.063)
(0.145)
(0.134)
(0.036)
(0.085)
(0.098)
(0.073)
(0.105)
(0.051)
(0.092)
(0.067)
(0.119)
4th
1.234**
1.363+
1.417**
1.305**
1.199*
0.432**
1.170*
1.267**
1.177**
0.946
1.313**
1.502**
(0.074)
(0.222)
(0.151)
(0.044)
(0.107)
(0.101)
(0.082)
(0.116)
(0.059)
(0.125)
(0.076)
(0.138)
Top
1.361**
12.904**
3.723**
1.520**
1.711**
1.558*
2.434**
1.823**
1.391**
7.335**
3.018**
2.401**
(0.085)
(1.531)
(0.399)
(0.054)
(0.154)
(0.333)
(0.186)
(0.184)
(0.072)
(0.769)
(0.185)
(0.238)
FICO
atorigination:
680–
720
1.061
0.829
1.048
0.980
1.215*
0.812
0.942
0.971
1.100*
0.841
0.969
0.973
(0.057)
(0.136)
(0.096)
(0.027)
(0.104)
(0.256)
(0.056)
(0.062)
(0.050)
(0.119)
(0.049)
(0.063)
650–
680
1.243**
1.165
0.892
0.953+
1.505**
0.962
1.011
0.897+
1.312**
1.100
0.987
0.928
(0.067)
(0.175)
(0.086)
(0.027)
(0.125)
(0.284)
(0.060)
(0.058)
(0.059)
(0.145)
(0.050)
(0.061)
620–
650
1.360**
1.318+
0.986
0.949+
1.738**
1.532
0.963
0.917
1.466**
1.372*
0.975
0.953
(0.074)
(0.195)
(0.096)
(0.028)
(0.144)
(0.422)
(0.058)
(0.061)
(0.066)
(0.176)
(0.050)
(0.064)
590–
620
1.573**
1.231
1.012
0.841**
2.260**
1.435
1.104
0.711**
1.782**
1.416*
1.051
0.714**
(0.099)
(0.202)
(0.114)
(0.031)
(0.209)
(0.428)
(0.081)
(0.063)
(0.092)
(0.199)
(0.065)
(0.063)
560–
590
1.958**
1.564**
1.022
0.858**
2.593**
2.543**
1.199*
0.586**
2.135**
1.986**
1.126+
0.598**
(0.138)
(0.260)
(0.126)
(0.037)
(0.271)
(0.744)
(0.095)
(0.066)
(0.125)
(0.279)
(0.076)
(0.068)
530–
560
2.004**
1.819**
1.035
0.752**
3.086**
2.817**
1.232*
0.447**
2.354**
2.321**
1.140+
0.437**
(0.158)
(0.319)
(0.137)
(0.037)
(0.354)
(0.862)
(0.109)
(0.066)
(0.154)
(0.343)
(0.083)
(0.064)
S. Chan et al.
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
<530
2.144**
1.850**
1.016
0.745**
3.473**
3.481**
1.324**
0.633**
2.604**
2.614**
1.180*
0.542**
(0.199)
(0.337)
(0.151)
(0.042)
(0.466)
(1.068)
(0.131)
(0.107)
(0.198)
(0.397)
(0.097)
(0.093)
Current
combinedLT
V:
80–90%
1.104+
0.766*
0.733**
0.956
1.232*
0.507**
0.631**
0.824*
1.152**
0.733**
0.661**
0.808**
(0.066)
(0.083)
(0.059)
(0.027)
(0.107)
(0.110)
(0.037)
(0.066)
(0.057)
(0.072)
(0.031)
(0.063)
90–100
%1.268**
0.464**
0.491**
0.940*
1.542**
0.252**
0.464**
0.843+
1.360**
0.419**
0.473**
0.822*
(0.076)
(0.077)
(0.055)
(0.029)
(0.136)
(0.090)
(0.037)
(0.075)
(0.068)
(0.064)
(0.030)
(0.070)
100–
110%
1.291**
0.277**
0.448**
0.887**
1.458**
0.392**
0.410**
0.711**
1.350**
0.282**
0.423**
0.716**
(0.082)
(0.067)
(0.060)
(0.032)
(0.142)
(0.138)
(0.038)
(0.072)
(0.072)
(0.057)
(0.032)
(0.070)
110–
120%
1.525**
0.284**
0.382**
0.868**
1.780**
0.283**
0.464**
0.757*
1.614**
0.260**
0.445**
0.779*
(0.104)
(0.075)
(0.061)
(0.037)
(0.187)
(0.130)
(0.047)
(0.085)
(0.093)
(0.060)
(0.037)
(0.083)
>120%
1.773**
0.202**
0.445**
0.807**
3.031**
0.678
0.574**
0.720**
2.143**
0.264**
0.548**
0.757*
(0.125)
(0.046)
(0.065)
(0.035)
(0.329)
(0.203)
(0.057)
(0.087)
(0.127)
(0.049)
(0.044)
(0.086)
Has
junior
lien
0.840**
0.888
1.058
1.074**
0.752**
0.684*
1.126*
1.182**
0.799**
0.848+
1.100*
1.132*
(0.035)
(0.088)
(0.076)
(0.024)
(0.041)
(0.126)
(0.053)
(0.065)
(0.026)
(0.077)
(0.043)
(0.060)
Fulldocumentatio
n&
DTI
<=45
%1.238**
1.242+
1.212*
0.984
1.271**
1.459*
0.948
1.186*
1.230**
1.355**
1.002
1.086
(0.066)
(0.153)
(0.114)
(0.032)
(0.098)
(0.257)
(0.061)
(0.100)
(0.054)
(0.138)
(0.053)
(0.093)
>45
%1.362**
1.030
1.164
0.976
1.210*
1.130
1.047
1.107
1.277**
1.150
1.060
1.014
(0.069)
(0.132)
(0.112)
(0.031)
(0.091)
(0.210)
(0.066)
(0.091)
(0.054)
(0.121)
(0.056)
(0.084)
Non-cash-outrefinanceloan
1.274**
1.355+
0.753*
0.709**
1.930**
1.405
0.852+
0.690**
1.559**
1.524**
0.767**
0.594**
(0.080)
(0.216)
(0.094)
(0.029)
(0.174)
(0.383)
(0.078)
(0.095)
(0.081)
(0.201)
(0.057)
(0.081)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Cash-outrefinanceloan
1.308**
1.498**
0.847**
0.797**
1.667**
2.211**
0.836**
0.734**
1.518**
1.696**
0.817**
0.691**
(0.051)
(0.128)
(0.054)
(0.017)
(0.085)
(0.319)
(0.036)
(0.042)
(0.047)
(0.123)
(0.029)
(0.040)
Prepaym
entpenalty
ineffect
0.804**
0.730**
0.856+
1.044
0.876
0.861
0.900+
0.794*
0.832**
0.779*
0.878*
0.778**
(0.066)
(0.086)
(0.074)
(0.028)
(0.111)
(0.170)
(0.057)
(0.076)
(0.057)
(0.080)
(0.045)
(0.071)
Owner-occupier
1.372**
1.075
0.677**
0.874**
1.298**
1.236
0.861*
0.972
1.405**
1.115
0.796**
1.004
(0.090)
(0.147)
(0.056)
(0.028)
(0.104)
(0.278)
(0.053)
(0.081)
(0.071)
(0.129)
(0.039)
(0.086)
Changein
ARM
paym
ents:
25–50%
0.580**
1.329+
1.144
1.025
0.567**
1.466*
1.057
1.102
0.577**
1.432**
1.093
1.159+
(0.058)
(0.214)
(0.148)
(0.047)
(0.065)
(0.250)
(0.078)
(0.089)
(0.043)
(0.166)
(0.070)
(0.094)
>50
%0.443**
0.871
1.390+
1.252**
0.723+
2.017+
1.358*
1.097
0.538**
1.066
1.350**
1.180
(0.069)
(0.258)
(0.260)
(0.077)
(0.130)
(0.756)
(0.163)
(0.192)
(0.063)
(0.258)
(0.139)
(0.200)
Borrower
behavior:
No.
ofmonthsbefore
default
1.015**
0.971**
0.979**
0.973**
1.030**
0.991
0.987**
0.952**
1.023**
0.979**
0.980**
0.941**
(0.002)
(0.004)
(0.003)
(0.001)
(0.003)
(0.006)
(0.002)
(0.004)
(0.002)
(0.003)
(0.002)
(0.004)
Paymentssincedefault:
<half
2.780**
0.991
0.939
1.024
3.247**
1.204
1.045
0.989
2.763**
1.015
1.001
1.080
(0.131)
(0.160)
(0.110)
(0.039)
(0.223)
(0.247)
(0.073)
(0.081)
(0.108)
(0.125)
(0.059)
(0.086)
>half
1.675**
0.841*
0.837**
0.697**
2.688**
0.629**
1.154**
0.843**
1.943**
0.747**
1.004
0.877*
(0.069)
(0.072)
(0.053)
(0.015)
(0.178)
(0.096)
(0.057)
(0.055)
(0.069)
(0.056)
(0.038)
(0.055)
MonthsafterARM
rate
adjustment
1–3
1.124
1.228
1.318+
0.853*
0.945
1.203
1.000
1.237+
1.026
1.162
1.088
1.262*
(0.148)
(0.207)
(0.194)
(0.055)
(0.159)
(0.262)
(0.089)
(0.136)
(0.106)
(0.153)
(0.083)
(0.139)
S. Chan et al.
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
4–6
1.394**
1.009
1.241
1.028
1.353*
1.260
0.980
1.182
1.355**
1.093
1.040
1.223+
(0.156)
(0.199)
(0.207)
(0.066)
(0.184)
(0.286)
(0.100)
(0.135)
(0.117)
(0.163)
(0.090)
(0.138)
7–12
1.284**
1.030
1.002
1.148**
1.095
1.294
1.117
1.137
1.182**
1.110
1.108
1.166+
(0.108)
(0.199)
(0.165)
(0.060)
(0.113)
(0.240)
(0.092)
(0.103)
(0.077)
(0.150)
(0.081)
(0.105)
Receivedforeclosurecounselin
g1.262**
0.512
0.400*
0.987
1.462**
1.172
0.510**
0.692
1.317**
0.785
0.477**
0.669
(0.101)
(0.365)
(0.181)
(0.077)
(0.157)
(0.609)
(0.125)
(0.172)
(0.084)
(0.327)
(0.103)
(0.167)
Neighborhoodcharacteristics:
Recenthousepricedepreciatio
n:
0–10
%1.248**
0.832
0.957
1.017
1.102
0.818
1.005
1.303**
1.192**
0.827+
0.995
1.292**
(0.104)
(0.115)
(0.105)
(0.031)
(0.116)
(0.146)
(0.070)
(0.111)
(0.078)
(0.089)
(0.059)
(0.109)
>10
%1.253**
0.924
1.162
0.886**
0.969
0.908
1.139
1.447**
1.156*
0.962
1.140+
1.404**
(0.104)
(0.158)
(0.143)
(0.032)
(0.103)
(0.229)
(0.092)
(0.137)
(0.076)
(0.135)
(0.077)
(0.131)
Recentforeclosurerate:
1–2%
0.923*
1.065
1.031
1.065**
0.908
0.780+
1.068
1.109
0.888**
0.890+
1.107**
1.198**
(0.037)
(0.088)
(0.068)
(0.024)
(0.055)
(0.102)
(0.048)
(0.070)
(0.029)
(0.061)
(0.041)
(0.076)
2–3%
0.810**
0.882
1.120
1.151**
0.781**
0.899
1.084
1.188*
0.760**
0.806*
1.161**
1.325**
(0.036)
(0.103)
(0.091)
(0.031)
(0.051)
(0.138)
(0.059)
(0.083)
(0.028)
(0.073)
(0.052)
(0.093)
3–4%
0.704**
0.870
0.926
1.173**
0.689**
1.090
1.025
1.216*
0.651**
0.904
1.065
1.400**
(0.039)
(0.140)
(0.105)
(0.038)
(0.052)
(0.207)
(0.069)
(0.098)
(0.029)
(0.107)
(0.062)
(0.114)
>4%
0.588**
0.841
0.802+
1.326**
0.586**
0.642+
1.088
1.443**
0.537**
0.678**
1.083
1.692**
(0.037)
(0.151)
(0.102)
(0.045)
(0.048)
(0.152)
(0.077)
(0.115)
(0.027)
(0.095)
(0.067)
(0.136)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Share
ofnon-prim
eloans:
10–20%
1.188*
1.027
1.024
1.112*
1.015
1.140
1.235*
1.136
1.120+
1.063
1.147*
1.143
(0.087)
(0.151)
(0.104)
(0.047)
(0.107)
(0.246)
(0.103)
(0.134)
(0.068)
(0.129)
(0.075)
(0.136)
>20
%1.230**
1.251
0.990
1.168**
1.029
1.035
1.238*
1.385**
1.147*
1.186
1.130+
1.403**
(0.093)
(0.193)
(0.106)
(0.050)
(0.112)
(0.236)
(0.107)
(0.167)
(0.071)
(0.151)
(0.077)
(0.169)
Medianincome:
<$30,0000
0.936
0.865
0.914
0.979
0.801**
0.842
0.757**
0.497**
0.895**
0.807*
0.806**
0.489**
(0.042)
(0.091)
(0.070)
(0.025)
(0.049)
(0.124)
(0.038)
(0.031)
(0.032)
(0.071)
(0.034)
(0.031)
$30–
40,000
1.010
0.930
0.995
1.047*
0.863**
0.821
0.850**
0.706**
0.951+
0.862*
0.906**
0.732**
(0.038)
(0.084)
(0.065)
(0.022)
(0.044)
(0.101)
(0.036)
(0.037)
(0.029)
(0.062)
(0.032)
(0.038)
Share
ofBlack
residents:
20–40%
1.127*
1.165
0.840+
1.020
1.052
0.896
0.973
0.877+
1.094*
1.115
0.939
0.875+
(0.060)
(0.164)
(0.084)
(0.032)
(0.083)
(0.174)
(0.062)
(0.070)
(0.049)
(0.125)
(0.050)
(0.069)
40–60%
1.062
1.424**
1.027
1.082*
1.325**
1.104
1.033
0.863+
1.154**
1.355**
1.037
0.893
(0.064)
(0.194)
(0.104)
(0.037)
(0.110)
(0.221)
(0.068)
(0.074)
(0.056)
(0.152)
(0.057)
(0.076)
60–80%
1.239**
1.385*
1.059
1.078*
1.423**
0.970
0.933
0.814*
1.298**
1.281*
0.968
0.836*
(0.071)
(0.188)
(0.107)
(0.037)
(0.119)
(0.187)
(0.065)
(0.070)
(0.062)
(0.142)
(0.056)
(0.073)
>80
%1.180**
1.434**
1.002
1.080*
1.430**
1.202
0.978
1.042
1.261**
1.442**
0.983
1.093
(0.067)
(0.188)
(0.098)
(0.037)
(0.120)
(0.226)
(0.067)
(0.091)
(0.059)
(0.154)
(0.056)
(0.095)
Share
ofHispanicresidents:
20–40%
0.988
1.204+
1.102
1.155**
1.026
1.053
1.033
1.300**
0.990
1.116
1.061
1.365**
(0.045)
(0.131)
(0.085)
(0.031)
(0.067)
(0.170)
(0.055)
(0.086)
(0.037)
(0.101)
(0.047)
(0.090)
S. Chan et al.
Tab
le3
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalo
nlis
pendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
>40
%1.074
1.198
1.161
1.177**
1.396**
0.947
1.127+
1.485**
1.162**
1.132
1.147*
1.575**
(0.062)
(0.167)
(0.113)
(0.039)
(0.114)
(0.191)
(0.074)
(0.125)
(0.055)
(0.127)
(0.063)
(0.134)
Share
ofAsian
residents:
20–40%
0.962
0.853
1.007
0.981
1.080
1.133
1.121
1.289**
1.006
0.953
1.062
1.254*
(0.056)
(0.141)
(0.105)
(0.034)
(0.096)
(0.242)
(0.079)
(0.118)
(0.049)
(0.124)
(0.063)
(0.113)
>40
%1.053
0.656
1.710**
0.900
1.143
0.609
1.124
1.247
1.090
0.669
1.304*
1.039
(0.141)
(0.239)
(0.322)
(0.072)
(0.232)
(0.387)
(0.175)
(0.321)
(0.122)
(0.211)
(0.161)
(0.267)
Share
offoreignborn:
40–60%
0.977
0.963
0.911
0.989
1.046
0.985
1.053
0.699**
1.006
0.946
1.010
0.700**
(0.031)
(0.075)
(0.054)
(0.019)
(0.048)
(0.107)
(0.040)
(0.034)
(0.026)
(0.059)
(0.032)
(0.034)
>60
%0.955
0.798
0.897
1.024
0.811*
1.015
1.009
0.657**
0.903+
0.811
0.982
0.682**
(0.061)
(0.141)
(0.105)
(0.037)
(0.077)
(0.265)
(0.073)
(0.062)
(0.048)
(0.116)
(0.060)
(0.064)
Num
berof
loan-m
onths
292,378
358,813
632,345
Com
petin
grisk
mod
elswith
relativ
erisk
ratio
sreported.S
tand
arderrorsarein
().Statisticalsign
ificance
isindicatedby
:+10
%,*5%,and
**1%.S
ample:LoanP
erform
ance
first
liennon-prim
esecuritized
mortgages
originated
2003–200
8in
New
YorkCity
thatareever90
days
delin
quentb
yOctob
er20
10.A
llmod
elsalso
include:nu
mberof
mon
thssince
default;indicatorsforno
orlow
documentio
ninteracted
with
DTI,localhousepriceappreciatio
n>5%,o
ther
race;indicatorsformissing
values
ofFICO,D
TI,andprepayment
penalty
;fixedeffectsforcalendar
year
andoriginationyear
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le4
Foreclosure
models—
HMDA
matched
sample
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Loancharacteristics:
Interestonly
loan
1.393**
1.063
0.828
0.804**
1.952**
0.755
1.017
0.847+
1.595**
1.013
0.917
0.872
(0.082)
(0.224)
(0.100)
(0.025)
(0.164)
(0.237)
(0.078)
(0.073)
(0.078)
(0.173)
(0.059)
(0.073)
FRM
relativ
erate
atorigination:
<1%
1.092
1.489+
0.869
0.788**
1.636**
1.598
0.912
0.653**
1.290**
1.664**
0.881
0.613**
(0.063)
(0.317)
(0.117)
(0.032)
(0.147)
(0.457)
(0.092)
(0.095)
(0.063)
(0.281)
(0.070)
(0.088)
1–2%
1.146*
1.629*
0.960
0.884**
1.729**
1.595
1.025
0.992
1.350**
1.699**
0.960
0.949
(0.072)
(0.347)
(0.134)
(0.039)
(0.154)
(0.459)
(0.101)
(0.124)
(0.070)
(0.289)
(0.077)
(0.117)
2–3%
1.189+
1.400
0.998
0.894+
1.914**
1.770+
1.079
1.207
1.474**
1.590*
1.024
1.157
(0.111)
(0.330)
(0.170)
(0.053)
(0.213)
(0.534)
(0.116)
(0.170)
(0.106)
(0.293)
(0.094)
(0.161)
>3%
1.455**
1.195
0.779
0.907
2.178**
0.736
1.159
0.931
1.818**
1.100
0.996
0.848
(0.180)
(0.352)
(0.188)
(0.077)
(0.357)
(0.295)
(0.172)
(0.233)
(0.177)
(0.258)
(0.126)
(0.213)
ARM
relativ
erate
atorigination:
<1%
0.766**
1.223
0.818
0.601**
1.305+
0.163+
0.886
0.694*
0.973
0.993
0.817+
0.649*
(0.073)
(0.356)
(0.149)
(0.032)
(0.205)
(0.175)
(0.129)
(0.127)
(0.080)
(0.266)
(0.091)
(0.116)
1–2%
0.928
1.170
0.822
0.899**
1.060
1.202
0.999
0.835*
1.011
1.257
0.937
0.833*
(0.060)
(0.274)
(0.116)
(0.033)
(0.093)
(0.362)
(0.083)
(0.070)
(0.052)
(0.231)
(0.067)
(0.070)
2–3%
0.895+
1.316
1.004
0.966
1.059
0.954
0.998
0.994
0.991
1.205
0.988
1.009
(0.060)
(0.257)
(0.118)
(0.033)
(0.082)
(0.249)
(0.071)
(0.074)
(0.050)
(0.185)
(0.060)
(0.077)
>3%
0.868
1.271
1.101
1.050
0.959
1.257
1.063
1.061
0.946
1.316+
1.075
1.120
(0.106)
(0.262)
(0.150)
(0.053)
(0.118)
(0.343)
(0.089)
(0.117)
(0.081)
(0.215)
(0.076)
(0.125)
Current
loan
balancequintile:
2nd
1.088
1.178
1.162
1.005
1.082
0.601**
1.188**
1.123
1.087+
0.943
1.189**
1.149
(0.064)
(0.144)
(0.109)
(0.034)
(0.095)
(0.100)
(0.072)
(0.095)
(0.053)
(0.090)
(0.060)
(0.100)
3rd
1.097
0.958
1.459**
1.121**
1.020
0.575**
1.161*
1.183+
1.056
0.804+
1.274**
1.321**
S. Chan et al.
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
(0.068)
(0.157)
(0.153)
(0.039)
(0.093)
(0.109)
(0.080)
(0.107)
(0.055)
(0.100)
(0.073)
(0.122)
4th
1.229**
1.400+
1.376**
1.300**
1.238*
0.412**
1.200*
1.210+
1.183**
0.893
1.311**
1.426**
(0.082)
(0.260)
(0.165)
(0.048)
(0.121)
(0.108)
(0.092)
(0.120)
(0.065)
(0.133)
(0.084)
(0.142)
top
1.338**
11.724**
3.529**
1.462**
1.816**
1.746*
2.540**
1.682**
1.414**
6.516**
2.999**
2.169**
(0.093)
(1.646)
(0.427)
(0.057)
(0.179)
(0.398)
(0.214)
(0.184)
(0.081)
(0.785)
(0.205)
(0.236)
FICO
atorigination:
680–720
1.063
0.884
1.030
0.987
1.125
0.848
0.853*
0.953
1.075
0.857
0.902+
0.971
(0.064)
(0.162)
(0.105)
(0.030)
(0.104)
(0.299)
(0.056)
(0.068)
(0.054)
(0.137)
(0.050)
(0.070)
650–680
1.228**
1.229
0.859
0.953
1.366**
1.045
0.906
0.869+
1.262**
1.158
0.904+
0.894
(0.073)
(0.210)
(0.093)
(0.030)
(0.123)
(0.348)
(0.059)
(0.063)
(0.063)
(0.173)
(0.051)
(0.066)
620–650
1.383**
1.288
0.943
0.954
1.659**
1.732+
0.901
0.913
1.463**
1.378*
0.928
0.959
(0.084)
(0.217)
(0.102)
(0.031)
(0.148)
(0.536)
(0.060)
(0.067)
(0.073)
(0.202)
(0.052)
(0.072)
590–620
1.587**
1.207
1.003
0.821**
2.071**
1.704
1.028
0.741**
1.759**
1.466*
1.003
0.744**
(0.110)
(0.227)
(0.125)
(0.033)
(0.206)
(0.571)
(0.082)
(0.072)
(0.100)
(0.235)
(0.068)
(0.073)
560–590
1.870**
1.513*
1.003
0.848**
2.396**
3.061**
1.143
0.616**
2.029**
2.011**
1.091
0.636**
(0.146)
(0.282)
(0.137)
(0.040)
(0.269)
(1.004)
(0.100)
(0.077)
(0.130)
(0.319)
(0.080)
(0.080)
530–560
1.957**
1.788**
1.024
0.731**
2.993**
3.577**
1.116
0.480**
2.314**
2.439**
1.073
0.453**
(0.170)
(0.354)
(0.150)
(0.040)
(0.363)
(1.220)
(0.109)
(0.076)
(0.164)
(0.406)
(0.087)
(0.072)
<530
2.246**
1.772**
0.970
0.755**
3.180**
4.247**
1.279*
0.678*
2.599**
2.604**
1.137
0.602**
(0.228)
(0.373)
(0.164)
(0.048)
(0.458)
(1.471)
(0.140)
(0.123)
(0.215)
(0.454)
(0.104)
(0.111)
Current
combinedLT
V:
80–90%
1.092
0.704**
0.783**
0.970
1.261*
0.464**
0.638**
0.835*
1.158**
0.665**
0.679**
0.802**
(0.071)
(0.085)
(0.069)
(0.030)
(0.117)
(0.110)
(0.041)
(0.073)
(0.061)
(0.073)
(0.035)
(0.068)
90–100
%1.272**
0.430**
0.492**
0.926*
1.603**
0.231**
0.474**
0.864
1.385**
0.390**
0.478**
0.811*
(0.083)
(0.079)
(0.060)
(0.032)
(0.152)
(0.093)
(0.041)
(0.083)
(0.075)
(0.066)
(0.033)
(0.075)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
100–110%
1.291**
0.165**
0.467**
0.899**
1.499**
0.469*
0.418**
0.762*
1.363**
0.225**
0.434**
0.741**
(0.090)
(0.053)
(0.069)
(0.036)
(0.158)
(0.169)
(0.043)
(0.084)
(0.079)
(0.054)
(0.036)
(0.079)
110–120%
1.550**
0.290**
0.404**
0.865**
1.889**
0.346*
0.492**
0.765*
1.665**
0.284**
0.473**
0.755*
(0.115)
(0.081)
(0.072)
(0.040)
(0.215)
(0.162)
(0.054)
(0.094)
(0.104)
(0.069)
(0.043)
(0.089)
>120%
1.741**
0.216**
0.470**
0.818**
3.201**
0.827
0.579**
0.765*
2.169**
0.311**
0.556**
0.768*
(0.135)
(0.053)
(0.076)
(0.040)
(0.377)
(0.263)
(0.064)
(0.101)
(0.141)
(0.062)
(0.049)
(0.096)
Has
junior
lien
0.846**
0.885
1.012
1.046+
0.760**
0.735
1.103+
1.154*
0.809**
0.864
1.070
1.121+
(0.038)
(0.100)
(0.081)
(0.026)
(0.046)
(0.142)
(0.058)
(0.069)
(0.029)
(0.087)
(0.046)
(0.066)
Fulldocumentatio
n&
DTI
<=45
%1.207**
1.190
1.227*
1.067+
1.231*
1.553*
0.897
1.296**
1.184**
1.282*
0.985
1.181+
(0.071)
(0.163)
(0.127)
(0.038)
(0.104)
(0.316)
(0.065)
(0.120)
(0.057)
(0.146)
(0.059)
(0.115)
>45
%1.347**
0.910
1.134
1.054
1.132
1.366
1.024
1.160+
1.222**
1.085
1.050
1.095
(0.076)
(0.129)
(0.121)
(0.037)
(0.094)
(0.283)
(0.071)
(0.104)
(0.057)
(0.126)
(0.061)
(0.098)
Non-cash-outrefinanceloan
1.330**
1.503*
0.829
0.736**
2.012**
1.519
0.858
0.735*
1.610**
1.635**
0.804*
0.647**
(0.093)
(0.276)
(0.116)
(0.033)
(0.198)
(0.451)
(0.092)
(0.111)
(0.091)
(0.246)
(0.068)
(0.096)
Cash-outrefinanceloan
1.350**
1.476**
0.831**
0.801**
1.701**
1.996**
0.842**
0.722**
1.557**
1.643**
0.815**
0.679**
(0.058)
(0.145)
(0.060)
(0.019)
(0.095)
(0.316)
(0.041)
(0.046)
(0.053)
(0.135)
(0.033)
(0.043)
Prepaym
entpenalty
ineffect
0.802*
0.681**
0.900
1.063*
0.855
0.893
0.887+
0.755**
0.822**
0.752*
0.890*
0.781*
(0.069)
(0.090)
(0.084)
(0.031)
(0.110)
(0.187)
(0.060)
(0.078)
(0.059)
(0.086)
(0.049)
(0.077)
Owner-occupier
1.375**
1.079
0.678**
0.882**
1.196*
1.136
0.851*
0.971
1.357**
1.086
0.792**
0.997
(0.101)
(0.173)
(0.063)
(0.031)
(0.103)
(0.277)
(0.060)
(0.090)
(0.076)
(0.145)
(0.044)
(0.095)
Changein
ARM
paym
ents:
25–50%
0.531**
1.286
1.096
1.031
0.531**
1.421+
1.053
1.038
0.533**
1.398**
1.068
1.090
(0.058)
(0.236)
(0.156)
(0.051)
(0.064)
(0.271)
(0.084)
(0.089)
(0.043)
(0.182)
(0.074)
(0.094)
>50
%0.455**
0.850
1.129
1.269**
0.709+
2.146+
1.285+
1.145
0.546**
1.064
1.208
1.234
S. Chan et al.
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
(0.079)
(0.277)
(0.250)
(0.084)
(0.143)
(0.979)
(0.172)
(0.210)
(0.071)
(0.285)
(0.141)
(0.224)
Borrower
behavior:
no.o
fmonthsbefore
default
1.012**
0.972**
0.984**
0.973**
1.030**
0.990
0.987**
0.954**
1.020**
0.980**
0.982**
0.942**
(0.002)
(0.004)
(0.003)
(0.001)
(0.004)
(0.007)
(0.003)
(0.005)
(0.002)
(0.004)
(0.002)
(0.005)
Paymentssincedefault:
<half
2.760**
0.876
0.912
1.022
3.038**
1.294
0.954
0.969
2.695**
1.007
0.930
1.055
(0.143)
(0.166)
(0.121)
(0.043)
(0.230)
(0.285)
(0.075)
(0.087)
(0.115)
(0.138)
(0.062)
(0.094)
>half
1.633**
0.806*
0.831**
0.705**
2.504**
0.540**
1.133*
0.810**
1.870**
0.698**
0.991
0.856*
(0.073)
(0.078)
(0.058)
(0.016)
(0.181)
(0.088)
(0.061)
(0.058)
(0.072)
(0.058)
(0.041)
(0.059)
MonthsafterARM
rate
adjustment
1–3
1.107
1.287
1.303
0.851*
0.989
1.200
1.003
1.170
1.043
1.193
1.087
1.177
(0.157)
(0.241)
(0.214)
(0.060)
(0.173)
(0.292)
(0.099)
(0.142)
(0.115)
(0.176)
(0.091)
(0.143)
4–6
1.425**
0.885
1.136
0.982
1.359*
1.238
0.973
1.241+
1.389**
1.012
1.010
1.266*
(0.169)
(0.216)
(0.218)
(0.070)
(0.197)
(0.312)
(0.109)
(0.149)
(0.127)
(0.177)
(0.097)
(0.150)
7–12
1.300**
0.925
0.992
1.101+
1.081
1.258
1.071
1.138
1.198**
1.048
1.063
1.150
(0.117)
(0.214)
(0.179)
(0.062)
(0.119)
(0.262)
(0.099)
(0.110)
(0.083)
(0.162)
(0.087)
(0.110)
Receivedforeclosurecounselin
g1.225*
0.286
0.280*
1.026
1.398**
1.387
0.564*
0.810
1.275**
0.747
0.478**
0.799
(0.107)
(0.288)
(0.163)
(0.086)
(0.165)
(0.728)
(0.148)
(0.203)
(0.089)
(0.342)
(0.114)
(0.200)
Neighborhoodcharacteristics:
Recenthousepricedepreciatio
n:
0–10
%1.224*
0.917
1.007
1.027
1.089
0.893
1.001
1.313**
1.170*
0.894
1.010
1.315**
(0.113)
(0.142)
(0.122)
(0.034)
(0.124)
(0.174)
(0.076)
(0.121)
(0.084)
(0.107)
(0.065)
(0.120)
>10
%1.259*
1.080
1.160
0.886**
0.993
0.990
1.148
1.416**
1.167*
1.076
1.150+
1.396**
(0.115)
(0.205)
(0.159)
(0.035)
(0.115)
(0.273)
(0.102)
(0.146)
(0.083)
(0.167)
(0.085)
(0.142)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Recentforeclosurerate:
1–2%
0.919+
1.090
0.991
1.053*
0.928
0.735*
1.090+
1.081
0.891**
0.888
1.101*
1.150*
(0.040)
(0.103)
(0.073)
(0.027)
(0.061)
(0.107)
(0.055)
(0.075)
(0.032)
(0.069)
(0.045)
(0.079)
2–3%
0.814**
0.848
1.084
1.141**
0.815**
0.845
1.075
1.175*
0.774**
0.772*
1.131*
1.289**
(0.040)
(0.114)
(0.099)
(0.033)
(0.058)
(0.144)
(0.065)
(0.089)
(0.031)
(0.080)
(0.057)
(0.098)
3–4%
0.708**
0.962
0.928
1.155**
0.700**
1.115
1.024
1.175+
0.656**
0.976
1.052
1.329**
(0.043)
(0.171)
(0.116)
(0.042)
(0.058)
(0.227)
(0.076)
(0.104)
(0.032)
(0.126)
(0.067)
(0.118)
>4%
0.604**
0.942
0.760+
1.282**
0.623**
0.631+
1.083
1.410**
0.559**
0.721*
1.058
1.613**
(0.042)
(0.185)
(0.108)
(0.048)
(0.055)
(0.160)
(0.085)
(0.122)
(0.030)
(0.111)
(0.072)
(0.141)
Share
ofnon-prim
eloans:
10–20%
1.145+
1.024
1.023
1.117*
1.007
1.160
1.230*
1.176
1.095
1.066
1.139+
1.161
(0.091)
(0.165)
(0.112)
(0.051)
(0.116)
(0.265)
(0.111)
(0.153)
(0.072)
(0.139)
(0.080)
(0.150)
>20
%1.191*
1.198
0.969
1.147**
0.996
1.001
1.243*
1.359*
1.114
1.125
1.123
1.326*
(0.098)
(0.203)
(0.112)
(0.053)
(0.119)
(0.243)
(0.117)
(0.181)
(0.076)
(0.155)
(0.082)
(0.174)
Medianincome:
<$30,000
0.968
0.848
0.914
0.994
0.806**
0.809
0.799**
0.497**
0.914*
0.786*
0.839**
0.496**
(0.048)
(0.105)
(0.079)
(0.028)
(0.053)
(0.133)
(0.045)
(0.034)
(0.036)
(0.079)
(0.040)
(0.034)
$30–40,000
1.059
1.065
1.072
1.041+
0.866**
0.833
0.868**
0.676**
0.979
0.938
0.936+
0.705**
(0.043)
(0.106)
(0.078)
(0.024)
(0.048)
(0.111)
(0.041)
(0.039)
(0.032)
(0.074)
(0.037)
(0.041)
Share
ofBlack
residents:
20–40%
1.113+
1.103
0.845
1.020
1.006
0.970
0.966
0.920
1.063
1.124
0.935
0.922
(0.066)
(0.178)
(0.095)
(0.036)
(0.087)
(0.206)
(0.069)
(0.080)
(0.052)
(0.141)
(0.056)
(0.079)
40–60%
0.988
1.478*
1.056
1.086*
1.252*
1.075
0.994
0.972
1.085
1.378*
1.020
1.022
(0.067)
(0.238)
(0.123)
(0.041)
(0.114)
(0.255)
(0.074)
(0.092)
(0.059)
(0.182)
(0.064)
(0.097)
60–80%
1.192**
1.443*
1.030
1.083*
1.295**
0.980
0.924
0.890
1.226**
1.292*
0.950
0.911
S. Chan et al.
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
(0.079)
(0.231)
(0.123)
(0.043)
(0.121)
(0.227)
(0.073)
(0.086)
(0.066)
(0.167)
(0.063)
(0.089)
>80
%1.133+
1.444*
0.944
1.079+
1.293**
1.341
0.946
1.197+
1.188**
1.498**
0.936
1.224*
(0.076)
(0.230)
(0.110)
(0.043)
(0.122)
(0.309)
(0.075)
(0.118)
(0.064)
(0.192)
(0.062)
(0.121)
Share
ofHispanicresidents:
20–40%
0.956
1.222
0.968
1.144**
1.017
1.168
0.986
1.296**
0.967
1.188+
0.985
1.308**
(0.049)
(0.151)
(0.086)
(0.034)
(0.073)
(0.215)
(0.059)
(0.094)
(0.040)
(0.120)
(0.049)
(0.095)
>40
%1.023
1.202
1.083
1.179**
1.357**
1.181
1.085
1.512**
1.125*
1.197
1.085
1.560**
(0.068)
(0.196)
(0.122)
(0.044)
(0.124)
(0.273)
(0.081)
(0.143)
(0.060)
(0.155)
(0.068)
(0.149)
Share
ofAsian
residents:
20–40%
0.947
0.874
0.983
1.005
1.125
1.245
1.120
1.210+
1.011
0.981
1.050
1.166
(0.063)
(0.156)
(0.118)
(0.040)
(0.113)
(0.299)
(0.090)
(0.128)
(0.056)
(0.139)
(0.071)
(0.122)
>40
%1.049
0.494
1.565*
0.893
1.105
0.477
1.236
1.155
1.084
0.515+
1.319+
0.941
(0.162)
(0.219)
(0.350)
(0.085)
(0.271)
(0.357)
(0.218)
(0.370)
(0.141)
(0.190)
(0.187)
(0.301)
Share
offoreignborn:
40–60%
0.965
0.979
0.914
0.990
1.059
0.965
1.027
0.700**
1.000
0.955
0.998
0.700**
(0.034)
(0.086)
(0.059)
(0.021)
(0.052)
(0.118)
(0.043)
(0.037)
(0.029)
(0.068)
(0.035)
(0.037)
>60
%0.921
0.875
0.815
1.028
0.870
1.258
0.954
0.659**
0.895+
0.945
0.921
0.695**
(0.067)
(0.169)
(0.110)
(0.043)
(0.091)
(0.340)
(0.078)
(0.070)
(0.053)
(0.144)
(0.064)
(0.073)
Borrower
characteristicsfrom
HMDA:
Hispanic
1.176**
0.820
0.989
1.002
1.201*
0.695
1.127+
1.159*
1.184**
0.792+
1.083
1.187*
(0.067)
(0.123)
(0.098)
(0.032)
(0.095)
(0.156)
(0.070)
(0.087)
(0.054)
(0.097)
(0.057)
(0.091)
Black
1.106+
0.981
0.974
0.990
1.255**
0.852
1.025
0.849*
1.158**
0.938
1.007
0.879+
(0.057)
(0.115)
(0.084)
(0.029)
(0.088)
(0.131)
(0.056)
(0.057)
(0.048)
(0.084)
(0.047)
(0.061)
Asian
1.126+
1.200
1.093
0.954
0.983
0.875
1.072
1.450**
1.078
1.140
1.040
1.416**
(0.077)
(0.190)
(0.122)
(0.037)
(0.104)
(0.234)
(0.085)
(0.124)
(0.062)
(0.154)
(0.068)
(0.124)
What Happens to Distressed Mortgage Borrowers and Their Homes?
Tab
le4
(con
tinued)
Firststageof
theforeclosureprocess
Secondstageof
theforeclosureprocess(conditio
nalon
lispendens)
Collapsed
model
Modification
Refinance
Sale
Lispendens
Modification
Refinance
Sale
Auctio
nModification
Refinance
Sale
Auctio
n
Fem
ale
1.003
1.006
0.901+
1.010
1.183**
1.316*
1.036
0.913*
1.065*
1.117+
0.988
0.922+
(0.032)
(0.078)
(0.053)
(0.019)
(0.052)
(0.145)
(0.038)
(0.041)
(0.028)
(0.070)
(0.031)
(0.042)
Has
co-borrower
1.142**
1.244*
0.789**
0.648**
1.303**
1.194
1.100
0.862+
1.260**
1.442**
0.892*
0.629**
(0.042)
(0.120)
(0.060)
(0.018)
(0.073)
(0.188)
(0.066)
(0.068)
(0.039)
(0.119)
(0.041)
(0.050)
Num
berof
loan-m
onths
235,664
298,913
518,942
Com
petin
grisk
mod
elswith
relativ
erisk
ratio
sreported.S
tand
arderrorsarein
().Statisticalsign
ificance
isindicatedby
:+10
%,*5%,and
**1%.S
ample:LoanP
erform
ance
first
liennon-prim
esecuritized
mortgages
originated
2003–200
8in
New
YorkCity
thatareever90
days
delin
quentb
yOctob
er20
10.A
llmod
elsalso
include:nu
mberof
mon
thssince
default;indicatorsforno
orlowdocumentio
ninteracted
with
DTI,localhouse
priceappreciatio
n>5%,otherrace;indicatorsformissing
values
ofFICO,D
TI,prepaymentpenalty,
race,andgender;fixedeffectsforcalendar
year
andoriginationyear
S. Chan et al.
relative interest rate FRMs are more likely to be refinanced, likely reflecting a selectioneffect whereby borrowers that are better risks in ways that are unobservable in our data,were originally able to get more competitive rates and are able to find another lender torefinance their loan following default.20
Loans with larger current balances are more likely than those with smaller balancesto experience one of the terminal outcomes rather than remain in delinquency. Thosein the highest quintile of loan balances have a 50 % greater relative risk of receiving alis pendens and a twice as large relative risk of auction than loans in the lowestquintile.21 The smallest loans may be associated with a lower chance of arriving atauction because of fixed costs faced by a lender in the foreclosure process. Theassociation between the current loan balance and modification is also monotonicallyincreasing: loans with balances in the highest quintile have a relative risk of modi-fication that is over one third higher than those in the lowest quintile. This couldreflect a greater incentive for these borrowers to seek a modification, as well as thesubstantial fixed costs associated with modification making larger loans more worth-while for lenders to modify. A large outstanding loan balance is one of the strongestpredictors of refinance and sale, especially before receiving a lis pendens, which mayreflect that borrowers who originally qualified for a large mortgage have a greaterability to pursue other options, as well as a greater incentive to do so.
Importantly, we find a generally monotonic relationship between borrower FICOscore and the likelihood of receiving a lis pendens and going to auction, with better(higher) credit scores associated with a higher likelihood of a lis pendens and auction.This finding is consistent with Pennington-Cross (2010) using a much differentsample. While we cannot observe contemporaneous FICO scores, the Fair IsaacCorporation has shown that borrowers with higher original scores suffer a largerdecline in scores following delinquency, default and foreclosure (Christie 2010).Thus, our result likely implies that borrowers who have suffered a larger decline inFICO score since origination are more likely to receive a lis pendens or go to auction.It is important to note that we have already controlled for loan type and some loanterms, which will have been driven in large part by the borrower’s risk characteristicsat origination, so this FICO score result (and analogously, those that follow) shouldbe interpreted as conditional on these loan characteristics.
One possibility is that lenders may think that higher credit score borrowers who are90 days delinquent may be having trouble repaying as a result of some fundamentalchange like divorce, job loss or poor health, because they have a history of regularrepayment compared with lower credit score borrowers, with more erratic paymenthistories. Since our sample consists entirely of non-prime loans, it is also likely thatconditional on loan type and terms, the higher FICO score borrowers are adverselyselected in ways that are known to lenders, but that are unobserved in our data. This
20 We also tried specifications that included the current rate premium (the current rate on the loan minus theFreddie Mac average rate on 30 year FRMs), because it is usually an important predictor of prepaymentbehavior. However, this was not significant in explaining refinances and sales, probably because it is highlycorrelated with the relative rates at origination, and because we also have calendar year dummies in ourmodel.21 This finding is in contrast to Pennington-Cross (2010), who found that the relative risk of REO decreasesas the unpaid balance increases.
What Happens to Distressed Mortgage Borrowers and Their Homes?
would similarly explain the otherwise puzzling result of lower FICO score borrowersbeing more likely to refinance.22
Modifications are much more likely among borrowers with lower FICO scores atorigination. 23 Overall, borrowers with scores below 590 have a relative risk ofmodification that is twice that of borrowers with scores above 720 (the referencegroup). This might suggest that lenders are reluctant to modify loans for borrowerswho have a fundamental problem (as above) or who have strategically defaulted, thatis, chosen to miss payments despite having the ability to pay.24
Loans with higher LTVs have a lower relative risk of receiving a lis pendens andgoing to auction. For these higher LTV loans, lenders may believe that the net presentvalue of the mortgage if the borrower continues to try to make payments is higherthan the lender’s expected return from selling the property, or they may be concernedwith potential liability issues for maintenance, taxes and injuries while the property isvacant.25 For the lower LTV loans, the failure of the borrower to sell the property andrepay the mortgage is seemingly irrational, and lenders may interpret this inaction as aproblematic signal and proceed quickly with the foreclosure process.
The loan’s current LTV has a strong and monotonically increasing association withthe likelihood of modification, with loans greater than 120 % LTV having twice therelative risk of modification compared to loans with less than 80 % LTV. This isconsistent with other research and likely reflects a greater willingness of lenders tomodify loans for which repossession would result in the greatest losses.
Loans with higher current LTV are associated with a lower likelihood of arefinance or sale, reflecting the difficulty of either course of action when equity isnegative or nearly so. The estimates suggest that the relative risk of refinancing orselling the property when the loan is underwater is less than half that for those at 80 %LTVor lower. Having an additional lien is associated with an increased probability ofa lis pendens and auction. Conditional on combined LTV, a loan with a junior lien atorigination will have a lower first-lien-only LTV, and thus an auction sale is morelikely to cover the balance owed. A junior lien also makes it less likely a borrowerwill get a modification, which is not surprising given the well-known difficulty ofnegotiating a modification when more than one lender is involved.
Full documentation loans are associated with a higher likelihood of modificationrelative to loans with no or low documentation. Among these full-documentationloans, those with DTIs over 45 are even more likely to modify, possibly because thereis more scope for reducing the DTI as part of the modification. Full documentationborrowers who had lower DTIs at origination were more likely to be able to avoidforeclosure by refinancing.
22 The adverse selection of higher FICO score borrowers in default is also consistent with Brevoort andCooper (2010) who find that post-foreclosure borrowers with originally higher FICO scores take longer torecover their scores than those with originally lower scores.23 This is consistent with Haughwout et al. (2009) who observed, based upon a national sample, thatborrowers who eventually modify had lower credit scores at origination.24 Been et al. (2011), who are able to observe contemporaneous FICO scores, find that borrowersexperiencing greater declines in FICO since origination are less likely to receive modifications. As notedabove, since FICO scores tend to fall following default, our results are also consistent with their findings.25 Some have argued that some banks have delayed foreclosure sales to avoid realizing loan losses. Banksmight equally well delay even beginning the foreclosure process on underwater properties for the samereasons.
S. Chan et al.
Compared to home purchase loans, refinances are less likely to receive a lispendens, less likely to go to auction, and more likely to be modified or to refinance.Purchasers, which includes all first-time homebuyers, are likely to have shorterhomeownership and mortgage borrowing experience than refinancers and thus maybe more likely to end up at auction. Sales are less likely among refinance loans,possibly reflecting a longer housing tenure and attachment among those who wereoriginally refinance borrowers.
Owner-occupiers have a lower likelihood of receiving a lis pendens and a higherlikelihood of modification than do investors, consistent with other literature. Owner-occupants also have a lower likelihood of sale, again possibly reflecting attachment orpossibly the fact that non-occupants were not included in any programs to helpborrowers avoid foreclosure.
Larger ARM payment shocks are associated with a lower relative probability ofmodification, with loans whose payments are at least 25 % higher since originationhaving a relative risk of modification that is approximately half that of borrowerswhose payments increased by less than 25 %. ARMs with monthly payments that areat least 50 % larger than they were at origination are also much more likely to receivea lis pendens.
Borrower Behaviors
Modifications are more likely, and lis pendens and auctions less likely, the greater thenumber of months between origination and the initial default date. This could reflect agreater willingness of lenders to work with borrowers who have a longer track recordof payment. The length of time between origination and default also reduces thelikelihood of refinancing or sale, the latter perhaps reflecting greater borrowerattachment to the home.
We control for the fraction of monthly payments made by the borrower since theirinitial default, and find that those who have made more than half of these payments areless likely to receive a lis pendens than those who have made no payments—unsurpris-ingly, lenders are less likely to foreclose on borrowers who continue to repay, even someof the time. Borrowers who have made some payments are also more likely to obtain amodification, although this effect is non-monotonic. Borrowers who have made fewerthan half of the monthly payments since defaulting have a relative risk of modificationthat is over two and a half times greater than those making no payments (the referencegroup), while those making more than half the payments have relative risks that are lessthan two times greater. Lenders may be more likely to modify those who can show someability to continue making payments, while not wanting to modify (as much) those whoseem able to catch up on their own. This latter group of borrowers may also be less likelyto seek a modification.
For ARMs, modifications pre-lis pendens are more probable shortly after the initialrate adjustment, and this effect decreases over time. Defaults that occur within thiswindow are more likely due to the borrower being unable to deal with the highermortgage payments than are defaults outside of this window, which may be morelikely due to other financial distress, such as job loss.
The indicator variable for foreclosure counseling is defined to be equal toone in the months after a borrower receives the counseling. We find that the
What Happens to Distressed Mortgage Borrowers and Their Homes?
receipt of counseling is significantly associated with an increased relative riskof modification, and that the effect is more pronounced after the receipt of a lispendens. This result suggests that counseling may help borrowers better nego-tiate modifications with lenders, although some part of this effect could be dueto selection. Other studies of foreclosure counseling (see the review in Collinsand O’Rourke (2011)) tend to find that counseling helps borrowers avoidforeclosure, even though there is always evidence of selection bias when thatcounseling is voluntary.
Neighborhood Characteristics
Conditional on current LTV, properties in community districts with house pricedepreciation greater than 10 % in the past year are associated with a lower probabilityof receiving a lis pendens.26 But once a lis pendens is filed, greater neighborhoodprice depreciation is associated with a greater likelihood of ending up at auction, evenafter controlling for current LTV. Loans that are in community districts with anyhousing price depreciation in the past year are more likely to be modified, again, aftercontrolling for current LTV. Lenders might target declining house price areas formodification because borrowers in appreciating areas will have more options in termsof selling or refinancing the property to avoid repossession. Borrowers in declininghouse price areas with positive equity or with high psychological or other costs ofmoving might be more likely to seek a modification to help them ride out the decline.We find that borrowers in neighborhoods with house price growth in the past year aremore likely to refinance their loans, but these coefficients are not statisticallysignificant.
We find a positive and monotonically increasing association between the rate offoreclosure notices in the neighborhood within the prior 6 months and the probabilityof receiving a lis pendens and of going to auction. A foreclosure rate of over 4 % inthe surrounding census tract increases the relative risk of a lis pendens by one third,compared with mortgages in tracts where the foreclosure rate is less than 1 % (thereference group). These foreclosure rates are plausibly exogenous to each borrower,and they may serve as a proxy for reduced stigma associated with going through theforeclosure process when one’s neighbors also are doing so (see Haughwout et al.2009). To some extent, the foreclosure rates may also proxy for very local neighbor-hood economic conditions, such as unemployment, that affect the borrower’s abilityto repay loans. Further, neighborhood foreclosure rates may reflect expectationsabout future house price depreciation that are not captured in price indices based onrecent sales transactions. We already control for house price depreciation in the pastyear and found that foreclosure proceedings are less likely to be initiated in areaswhere prices have already declined. If high foreclosure rates signal further houseprice declines, lenders may want to move loans in those neighborhoods throughforeclosure more rapidly in order to avoid further erosion of their collateral.
On the other hand, the likelihood of modification falls monotonically with recentforeclosure rates in the property’s census tract. Because foreclosure rates are mea-sured at the census tract level, while the house price indices are at the larger
26 Note that we already control for origination year and calendar year.
S. Chan et al.
community district level, greater house price depreciation in higher foreclosure tractscompared with the community district average would lead to systematically over-estimated housing values and underestimated current LTVs for loans in high foreclo-sure neighborhoods.27 Thus, higher recent foreclosure activity in the neighborhoodcould be acting as a proxy for the extent of very localized depreciation. Takentogether, the effect on recent house price depreciation and foreclosure rates maymean that while modifications are more likely in generally depreciating areas, lendersavoid modifying loans in the neighborhoods with the very worst house priceperformance.
The rate of foreclosures in the surrounding census tract has a generally negativeeffect on the probability of a refinance. The possibility that foreclosure notices are aforward indicator of falling house prices may make it more difficult for borrowers tosecure new loans in these neighborhoods; or borrowers may simply not demandrefinances in these circumstances and rather, just give up.
We also included in the specification the non-prime share of mortgages originatedin the census tract during the 2 years prior to the loan’s origination. The non-primeshare is positively associated with receiving a lis pendens and arriving at auction.Loans in tracts where this non-prime share was more than 20 % have a relative risk ofreceiving a lis pendens that is 17 % higher, and a relative risk of auction that is 39 %higher, than loans where the non-prime share was less than 10 % (the referencegroup). The non-prime share of mortgages at origination also increases the likelihoodof modification before lis pendens. There are several possible explanations for theseresults. It may be the case that non-prime shares at origination are a leading indicatorof high foreclosure rates and lower housing values, prompting lenders to push loansinto foreclosure before even more value is lost. It may also be that particular lendersare concentrated in these areas and that the model is picking up different behavior bythose particular lenders. Another possibility is that areas with high non-prime shareswere “targeted” by aggressive mortgage lenders for loans that were inappropriate inways unobserved in our data. Servicers now dealing with these distressed loans maysee the pattern, decide that there is little chance that such loans will ever be profitablein their current form, and therefore offer a modification, or alternatively, proceedthrough the foreclosure process as quickly as possible. If this is true, then the share ofnon-prime originations in a neighborhood is acting as a proxy for poor underwritingstandards or other inappropriate practices at origination that are unobserved by theresearcher, but are observed by the servicer.
For neighborhood income and racial and ethnic composition, we find that loans inpoorer neighborhoods are associated with a lower likelihood of a lis pendens orauction. This is possibly due to these being more modest properties that do not justifythe lender’s fixed costs of foreclosure proceedings. Conditional on neighborhoodincome and all the other factors above, loans in census tracts with higher shares ofblack residents are associated with a higher likelihood of a lis pendens than loans intracts with lower shares of black residents. In contrast, black residential composition
27 Foreclosures could be causing diminished house price appreciation by increasing the housing supply onthe market and driving down prices. And, they may generate negative externalities such as the visibledeterioration of properties that lead to lower property values in high foreclosure tracts relative to others inthe same community district.
What Happens to Distressed Mortgage Borrowers and Their Homes?
has no clear association with auctions. Loans in tracts with higher shares of Hispanicresidents are more likely both to receive a lis pendens and go to auction, while thosein neighborhoods with relatively more immigrants are no less likely to receive lispendens, but are much less likely to go to auction.
Loans in tracts with median annual income less than $30,000 are less likely toreceive modifications than those in neighborhoods with median income over $40,000(the reference group). However, conditional on this, larger shares of black residentsare associated with a greater likelihood of modification. Overall, loans in tracts thatare over 60 % black have a relative risk of modification that is at least one quarterhigher than loans in tracts that are less than 20 % black. This magnitude is slightlylower when we control for the borrower’s own race, as Table 4 shows, but the effect isstill statistically significant. Modifications are also more likely for loans in neighbor-hoods that are more than 40 % Hispanic. We did not find any effect of the share ofAsian or immigrant residents on the relative risk of modification.
Borrower Characteristics from HMDA
Table 4 shows that overall, Hispanic and Asian borrowers have a relative risk ofending up at a foreclosure auction that is 19 and 42 % higher respectively, comparedwith other borrowers. Black and Hispanic borrowers are 16 and 18 % more likelythan other borrowers to have their loans modified. However, the likelihood that aborrower will sell or refinance is independent of her race and ethnicity.
Loans with a co-borrower are less likely to receive a lis pendens and go to auction,and more likely to be modified than those without co-borrowers. Co-borrower loansare more likely to refinance, possibly reflecting a smaller variance in income whenthere are two borrowers, and thus a higher likelihood of obtaining new credit. On theother hand, loans with co-borrowers are less likely to sell, perhaps reflecting gener-ally lower mobility rates among couples. The remaining non-HMDA variables havecoefficients that are very similar to those in Table 3.
Conclusion
Using a rich data set, we have shown how post-default outcomes (including theforeclosure notice, modification, refinance, sale, and foreclosure auction) are associ-ated with the type of loan and its terms, borrower characteristics and behavior, and thecharacteristics of the neighborhood. As noted at the outset, our results are largelydescriptive and we have not tried to model either the complex set of incentives thelender, servicer and borrower face, or the interactions among the parties.Nevertheless, we have made an important contribution by documenting, with verydetailed data, associations that researchers, policymakers and practitioners shouldconsider in trying to craft better solutions to the still-unfolding foreclosure crisis.
The outcomes of borrower defaults and foreclosures have implications not just for thelender and the household’s financial well-being, but also for the health of the surround-ing neighborhood. Homes in financial distress may depress local house prices throughdeterioration of the property, and foreclosed properties may become vacant and lead tomore crime, vermin, or other negative effects. Properties that sell at auction or become
S. Chan et al.
REO may also lower house prices in the surrounding area by increasing market supplyand by lowering nearby appraisals. For these reasons, the various outcomes of distressedmortgage borrowers have far-reaching consequences for lenders, households and neigh-borhoods. Understanding the factors associated with more favorable outcomes willfacilitate more targeted and effective interventions.
Acknowledgments We thank Amy Crews Cutts, Michael Gedal, Josiah Madar, Mary Weselcouch, andthe participants in the Association for Public Policy Analysis and Management Fall 2010 and AmericanReal Estate and Urban Economics Mid-Year 2011 conferences for their thoughtful suggestions. We are alsograteful to the staff of the Federal Reserve Bank of New York for their help with the LoanPerformance data,as well as to the Public Data Corporation and the New York City Department of Finance for their assistancein providing the additional data needed for this project.
References
Adelino,M., Gerardi, K., &Willen, P. S. (2009). Why don’t lenders renegotiate more homemortgages? Redefaults,self-cures, and securitization. Federal Reserve Bank of Boston Public Policy Discussion Papers, No. 09-4.
Adelino, M., Gerardi, K., & Willen, P. (2010). What explains differences in foreclosure rates? A response toPiskorski, Seru, and Vig. Federal Reserve Bank of Boston Working Paper, No. 10-2.
Agarwal, S., Amromin, G., Ben-David, I., Chomsisengphet, S., & Evanoff, D.D. (2011). The role ofsecuritization in mortgage renegotiation. Federal Reserve Bank of Chicago Working Paper, No. 2011-02.
Ambrose, B. W., & Capone, C. A., Jr. (1996). Do lenders discriminate in processing defaults? Cityscape: AJournal of Policy Development and Research, 2(1), 89–98.
Ambrose, B. W., & Capone, C. A., Jr. (1998). Modeling the conditional probability of foreclosure in thecontext of single-family mortgage default resolutions. Real Estate Economics, 26(3), 391–429.
Been, V., Weselcouch, M., Voicu, I., & Murff, S. (2011). Determinants of the incidence of loan modifica-tions. http://ssrn.com/abstract=1941915. Accessed 1 July 2012.
Brevoort, K. P., & Cooper, C.R. (2010). Foreclosure’s wake: the credit experiences of individuals followingforeclosure. Board of Governors of the Federal Reserve System Finance and Economics DiscussionSeries, No. 2010-59.
Capozza, D., & Thomson, T. (2006). Subprime transitions: lingering or malingering in default? Journal ofReal Estate Finance and Economics, 33(3), 241–258.
Chan, S., Gedal, M., Been, V., & Haughwout, A. (2010). The role of neighborhood characteristics in mortgagedefault risk: Evidence from New York City. http://ssrn.com/abstract=1631944. Accessed 1 July 2012.
Christie, L. (2010). How foreclosure impacts your credit score. CNNMoney.com. http://money.cnn.com/2010/04/22/real_estate/foreclosure_credit_score/index.htm. Accessed 1 July 2012.
Collins, J., & Reid, C. (2010). Who receives a mortgage modification? Race and income differentials inloan workouts. Federal Reserve Bank of San Francisco Working Paper.
Collins, J. M., & O’Rourke, C. (2011). Homeownership education and counseling: Do we know whatworks? Research Institute for Housing America. http://www.housingamerica.org/RIHA/RIHA/Publi-cations/76378_10554_Research_RIHA_Collins_Report.pdf. Accessed 1 July 2012.
Collins, J. & Reid, C. (2010). Who Receives a Mortgage Modification? Race and Income Differentials inLoan Workouts. Federal Reserve Bank of San Francisco Working Paper 2010-07.
Crews Cutts, A., & Merrill, W. A. (2008). Interventions in mortgage default: Policies and practices toprevent home loss and lower costs. In N. P. Retsinas & E. S. Belsky (Eds.), Borrowing to live:Consumer and mortgage credit revisited (pp. 203–254). Washington DC: Brookings Institution.
Danis, M. A., & Pennington-Cross, A. N. (2005). A dynamic look at subprime loan performance. Journalof Fixed Income, 15(1), 28–39.
Danis, M. A., & Pennington-Cross, A. N. (2008). The delinquency of subprime mortgages. Journal ofEconomics and Business, 60(1–2), 67–90.
Foote, C. L., Gerardi, K. S., Goette, L., & Willen, P. S. (2009). Reducing foreclosures: No easy answers.NBER Macroeconomics Annual, 24, 89–138.
Gerardi, K., Shapiro, A. H., & Willen, P. S. (2007). Subprime outcomes: Risky mortgages, homeownershipexperiences, and foreclosures. Federal Reserve Bank of Boston Working Paper, No. 07-15.
Gerardi, K. S., & Willen, P. S. (2009). Subprime mortgages, foreclosures, and urban neighborhoods. TheB.E. Journal of Economic Analysis & Policy, 9(3), Article 12.
What Happens to Distressed Mortgage Borrowers and Their Homes?
Haughwout, A., Okah, E., & Tracy, J. (2009). Second chances: Subprime mortgage modification and re-default. Federal Reserve Bank of New York Staff Reports, No. 417.
Lauria, M., Baxter, V., & Bordelon, B. (2004). An investigation of the time between mortgage default andforeclosure. Housing Studies, 19(4), 581–600.
Lender Processing Services. (2010). LPS Mortgage Monitor Report, October 2010. Jacksonville: LenderProcessing Services, Inc.
Pennington-Cross, A. N. (2010). The duration of foreclosures in the subprime mortgage market: Acompeting risks model with mixing. Journal of Real Estate Finance and Economics, 40(2), 109–129.
Pennington-Cross, A. N., & Ho, G. (2010). The termination of subprime hybrid and fixed-rate mortgages.Real Estate Economics, 38(3), 399–426.
Piskorski, T., Seru, A., & Vig, V. (2010). Securitization and distressed loan renegotiation: Evidence fromthe subprime mortgage crisis. Journal of Financial Economics, 97(3), 369–397.
Voicu, I., Jacob, M., Rengert, K., & Fang, I. (2011). Subprime loan default resolutions: Do they vary acrossmortgage products and borrower demographic groups? Journal of Real Estate Finance and Economics.doi:10.1007/s11146-011-9305-4.
S. Chan et al.