Does Bankruptcy Reduce Foreclosure by pengxiang

VIEWS: 5 PAGES: 31

									                     Does Bankruptcy Reduce Foreclosure?

                                           Wenli Li,
                              Federal Reserve Bank of Philadelphia

                                     Michelle J. White,
                        UC San Diego, Cheung Kong G.S.B., and NBER


                                          September 2011



                                             Abstract

     In the paper, we examine whether filing for bankruptcy delays or prevents foreclosure,
using new household-level panel datasets that combine large samples of prime and subprime
mortgages with information on homeowners’ non-mortgage debt. Bankruptcy is predicted to
delay foreclosure both because legal actions against debtors are stayed for several months during
the bankruptcy process and because homeowners gain financially from discharge of unsecured
debt in bankruptcy and may use their gains to save their homes. Our main result is that
bankruptcy reduces the probability of both foreclosure starting and ending with sale of the
house—the elasticity of the start of foreclosure with respect to bankruptcy is -.05 and -.12 for
homeowners with prime and subprime mortgages, respectively; while the elasticity of
foreclosure sale with respect to bankruptcy is -.03 and -.11, respectively. All of these results are
strongly statistically significant. For prime mortgages, we also find that bankruptcy delays
foreclosure by longer after the start of the financial crisis in 2008 than before and that bankruptcy
delays foreclosure by longer for homeowners who behave strategically.


*The views expressed here are the authors’ and do not represent those of the Federal Reserve
Bank of Philadelphia or the Federal Reserve System. Michelle White is grateful for research
support and hospitality from Cheung Kong Graduate School of Business, Beijing.




                                                 1
Introduction

   Mortgage default by homeowners generally results in lenders initiating foreclosure. But
homeowners may file for bankruptcy in order to prevent or delay foreclosure. There are several
reasons why bankruptcy delays foreclosure and why homeowners gain from filing because of the
delay. First, bankruptcy filings stay all legal actions against filers—including foreclosure. To
proceed with foreclosure, lenders must petition the bankruptcy court to lift the stay, which takes
at least a few months. Bankruptcy particularly delays foreclosure in states that do not require a
judicial proceeding (a judge’s order) for foreclosure to occur, since the foreclosure process in
these states is very quick. Second, filing for bankruptcy improves homeowners’ financial
positions by discharging some unsecured debt and/or ending wage garnishment. This increases
homeowners’ ability to pay and—if they wish to save their homes—they may use the additional
funds to repay their mortgage arrears. If they plan to give up their homes, they also gain from
delaying foreclosure, because doing so allows them to live in their homes rent-free for longer.
Third, lenders may respond to mortgage default by quickly foreclosing or by delaying
foreclosure; this variation means that filing for bankruptcy may delay foreclosure by more for
some types of homeowners than others.
    This paper explores whether and to what extent bankruptcy delays or reduces foreclosure.
To test the bankruptcy-foreclosure relationship, we construct new household-level datasets that
combine large samples of prime and subprime mortgages with information on homeowners’ non-
mortgage debt. Our main result is that bankruptcy is negatively related to both the start of
foreclosure and sale of the property in foreclosure—the elasticities of foreclosure start and
foreclosure sale with respect to bankruptcy are -.05 and -.03 for prime mortgages, respectively,
and both figures are -.11 for subprime mortgages. Thus bankruptcy delays foreclosure by longer
for subprime than prime mortgages, perhaps because bankruptcy trustees tend to be more active
on behalf of homeowners who may have been victims of predatory lending. We also find that
lenders identify and respond differently to defaults by different types of homeowners: the delay
effect of bankruptcy is larger for homeowners who behave strategically. But the evidence
suggests that lenders cannot distinguish between the two types of homeowners in the subprime
mortgage sample and they therefore respond in the same way to defaults by both types. We also

                                                 2
argue that delaying foreclosure is efficiency-enhancing at least to some extent, because mortgage
lenders bear only some of the costs of foreclosure and therefore they tend to foreclose too often.
    Section II of the paper discusses U.S. bankruptcy and foreclosure law and our hypotheses
concerning how bankruptcy affects mortgage default and foreclosure. Section III discusses our
data and section IV gives results. Section V concludes and discusses policy implications.



Foreclosure and Bankruptcy Law

     In this section, we discuss the mortgage foreclosure process in the U.S. and how bankruptcy
affects homeowners’ incentives to default on their mortgages and lenders’ incentives to
foreclose. We also consider how the financial crisis that began in 2008 changed homeowners’
incentives.
    Foreclosure. Lenders have the right to foreclose when homeowners default on their mortgage
payments. The foreclosure process is governed by state laws, which determine how long and
formal it is. In some states, mortgage lenders must get a court order to foreclose, while in others,
they can proceed without court involvement. If homeowners do not repay their mortgage
arrears, property is eventually sold in foreclosure. The proceeds of the sale after expenses are
used, first, to repay the first mortgage in full, including principal and interest, fees, and penalties.
Next, the second and third mortgages, if any, are repaid in full in order. Third, homeowners
receive an amount up the state’s homestead exemption. Homestead exemptions vary widely
across states, from zero in a few states to unlimited in seven states. If anything is left, it goes to
unsecured creditors. In most foreclosures, the sale price of the house is insufficient to repay the
mortgage(s)—otherwise homeowners would have sold the house themselves outside of
foreclosure. Some states allow mortgage lenders to pursue “deficiency judgments” against
former homeowners, which are unsecured claims for the difference between the amount owed on
the mortgage(s) and the foreclosure sale price. 1 Most states also allow homeowners to reclaim
the property for a period after foreclosure by repaying the mortgage in full. These periods can
last up to a year.


1
 To prevent mortgage lenders from selling foreclosed homes for less than market value, some states allow lenders to
claim deficiency judgments only if the foreclosure sale has received court approval. See Elias (2008) for discussion
of foreclosure law.


                                                         3
     Homeowners can remain in their homes for varying periods after default. Some states force
them to leave before the foreclosure sale, while others allow them to stay. In the latter case, they
become tenants and the new owner must go through an eviction procedure to force them to leave.
Once foreclosure is complete, ex-homeowners do not owe any mortgage or rent payments. The
period from default to eviction ranges from a few months to more than a year—and homeowners
can extend it by filing for bankruptcy.
     Consider homeowners’ decisions to default. Homeowners may default because they
experience economic stress that reduces their ability-to-pay or they may default because doing so
makes them better off. Homeowners gain financially from defaulting if the present value of the
future cost of owning (PVCO) exceeds the present value of the future cost of renting alternative
housing (PVCR), or if PVCO > PVCR. This condition is more likely to hold if home equity is
negative, if homeowners expect house values to fall in the future, or if the terms of the mortgage
contract are unfavorable to the homeowner—such as a high interest rate or a “teaser” rate that is
about to increase sharply.
    Homeowners that default on their mortgages also have an incentive to default on non-
mortgage debt and vice versa, because default on any type of debt triggers a sharp drop in
homeowners’ credit scores and the additional effect of a second type of default is much smaller.
Default on non-mortgage debt causes lenders to pursue collection techniques, including calling
debtors at home and at work to demand payment and taking legal action to garnish debtors’
wages if debtors are employed and to seize money in debtors’ bank accounts if these accounts
can be located. Debtors are protected by Federal and state laws that prevent garnishment of
more than 25% of wages and sometimes more; while debtors’ assets are protected by state asset
exemptions, which apply regardless of whether debtors have filed for bankruptcy.
    Bankruptcy. How does filing for bankruptcy affect homeowners’ gain from defaulting on
their mortgages and on their non-mortgage debts? 2 There are two separate personal bankruptcy
procedures in the U.S., called Chapters 7 and 13, and most homeowners are allowed to choose
between them. Filing for bankruptcy under either procedure stops creditors’ collection efforts,
ends wage garnishment, and stops foreclosure for at least a few months. Many types of


2
 The discussion of bankruptcy law reflects the changes adopted as part of the 2005 bankruptcy reform. See White
(2009), White and Zhu (2010), Morgan et al (2010), and Li, White and Zhu (2010) for discussion of the 2005
bankruptcy reform and its effects on mortgage default. Berkowitz and Hynes (1999) and Lin and White (2001)
examine the relationship between mortgage default and bankruptcy prior to the 2005 reform.

                                                       4
unsecured debt—including credit card debts, installment loans, medical bills, and unpaid rent—
are discharged. 3
    Chapter 7. The most commonly-used bankruptcy procedure is Chapter 7. Most unsecured
debts are quickly discharged in Chapter 7, but mortgage contracts cannot be changed. 4 Debtors
are obliged to use assets above the exemptions in their states to repay unsecured debt, but they
are not obliged to use of their future earnings to repay. States have separate exemptions for
different types of assets, but the exemption for home equity—the “homestead” exemption—is
nearly always the largest. Debtors with high incomes are not allowed to file under Chapter 7, but
the restriction is generally only binding on debtors whose incomes are above the 90th percentile
of the income distribution. 5 Thus in states with high homestead exemptions, even debtors with
high incomes and high assets can use bankruptcy to avoid repaying their unsecured debt.
    Homeowners’ gain from filing under Chapter 7 can be expressed as:
                                    GainCh7 = U 7 + H 7 + ΔQ7 − A7 − C7

U 7 is the value of unsecured debt discharged in Chapter 7. Homeowners receive this subsidy in

bankruptcy regardless of whether they keep their homes or not. H 7 is the reduction in the

present value of future housing costs when homeowners file under Chapter 7. H 7 = 0 if

homeowners save their homes in Chapter 7, but it equals (PVCO − PVCR ) .if homeowners give
                                                             7      7


up their homes and become renters. Because filing for bankruptcy delays foreclosure, H 7 is
larger for homeowners who file for bankruptcy, since they are allowed to live in their homes
cost-free for longer. ΔQ7 denotes the change in the value of home equity when homeowners file

for bankruptcy. For homeowners that have negative home equity, give up their homes, and live
in states that allow lenders to pursue deficiency judgments, ΔQ7 equals the increase in the value
of home equity due to discharge of deficiency judgments in bankruptcy. Now suppose home
equity is positive, but less than the homestead exemption. Outside of bankruptcy, homeowners

3
  Unsecured debts not discharged in bankruptcy include unpaid child support, tax obligations, debts incurred by
fraud, student loans, and costs of bankruptcy itself.
4
  The prohibition on changing mortgage terms in bankruptcy is based on the Supreme Court’s decision in Nobleman
v. American Savings Bank, 508 US 324 (1993) and on 11 U.S.C. § 1322(b)(2), which prevents bankruptcy judges
from discharging mortgage debt that is secured only by a primary residence, even if the value of the house is below
the mortgage principle. See Levitin and Goodman (2008) for discussion. Note that bankruptcy law in the U.S. is
Federal law, so that it is uniform all over the country. But U.S. bankruptcy law allows states to set their own
exemptions for home equity and other assets. The state asset exemptions also apply outside of bankruptcy.
5
  The means test for Chapter 7 was adopted as part of the bankruptcy reform of 2005.

                                                         5
generally lose their home equity when they default on their mortgages, because homes sold in
foreclosure auctions generally sell for the amount of the mortgage. But because bankruptcy
delays foreclosure, these homeowners are often able to keep their home equity, either by selling
their homes outside of foreclosure or by repaying their mortgage arrears in full—homeowners’
ability to repay mortgage arrears increases in bankruptcy since unsecured debt is discharged.
However if home equity exceeds the homestead exemption, then homeowners are forced to give
up their homes even in bankruptcy, so that filing for bankruptcy does not increase their home
equity. A7 is the value of non-exempt assets other than home equity that homeowners must use
to repay unsecured debt in bankruptcy. In practice, this term is nearly always zero, since most
homeowners can convert their non-exempt assets into exempt home equity or some other exempt
asset before filing. C7 is homeowners’ cost of filing for bankruptcy under Chapter 7.

     Chapter 13. Chapter 13 is an alternate bankruptcy procedure that is mainly used by
homeowners trying to save their homes. 6 In order to file under Chapter 13, homeowners must
have regular earnings and must propose a plan to repay their mortgage arrears over 3 to 5 years.
Lenders cannot foreclose during the plan period if homeowners are making regular payments
and, if homeowners repay all of their mortgage arrears and make all of their regular mortgage
payments during the plan period, then the original mortgage contract is reinstated. 7 Filing under
Chapter 13 thus benefits homeowners who have large mortgage arrears but wish to save their
homes, since it allows them to repay the arrears over several years. The terms of first mortgages
cannot otherwise be changed in Chapter 13. But second mortgages and home equity loans can
be partially or fully discharged in Chapter 13 if they are underwater and sometimes bankruptcy
trustees also challenge fees and penalties that lenders add to mortgages following default. 8
    Filing under Chapter 13 also benefits homeowners who plan to give up their homes, since it
delays foreclosure and allows them to live in their homes cost-free for longer. These
homeowners often propose repayment plans, but quickly default on the payments. Lenders then
must wait until the bankruptcy judge lifts the stay on legal actions before they can foreclose.



6
  See Zhu (2010) and White and Zhu (2010) for discussion of Chapter 13 and how it is used by homeowners.
Carroll and Li (2008) present data showing that homeowners who try to save their homes in bankruptcy often fail.
This discussion of Chapter 13 is based on the law following the 2005 bankruptcy reform.
7
  11 U.S.C. § 1322(c)(1) allows debtors to cure defaults on their mortgages in Chapter 13.
8
  See Porter (2009) and Elias (2009) for discussion.

                                                        6
Homeowners may go through the process of proposing and defaulting on Chapter 13 repayment
plans several times, each time delaying foreclosure for six months or more.
   Homeowners’ gain from filing under Chapter 13 can be expressed as:
                        GainCh13 = U13 + H13 + ΔQ13 − max[E13 , A13 ] − C13.

Here U13 , the value of unsecured debt discharged in bankruptcy, is the same under both

Chapters. H13 , the reduction in the cost of housing when homeowners file under Chapter 13, is

larger than H 7 for homeowners who shift to rental housing, because Chapter 13 delays

foreclosure for longer than Chapter 7. ΔQ13 , the increase in the value of home equity when

homeowners file under Chapter 13, exceeds ΔQ7 for homeowners who have positive home
equity and large mortgage arrears, since these homeowners would not be able to save their
homes in Chapter 7, but may be able to save them in Chapter 13 because of the long repayment
period. E13 is the value of future earnings that homeowners must use to repay unsecured debt

in Chapter 13. This amount equals the difference between homeowners’ future earnings and a
formula-determined earnings exemption that varies across homeowners. For most homeowners,
E13 equals zero because earnings exemptions are quite high and mortgages are paid before

unsecured debts. A13 is the value of non-exempt assets that homeowners must use to repay

unsecured debt; this amount is the same as A7 . However homeowners who have both non-

exempt future earnings and non-exempt assets are required to repay the maximum of the two in
Chapter 13, not the sum. Finally, C13 is the cost of filing for bankruptcy under Chapter 13; it is

higher than the cost C7 of filing under Chapter 7. Because of the higher costs of Chapter 13,

this chapter only attracts homeowners if their financial gain from filing is higher.
   Overall, filing for bankruptcy benefits two separate groups of homeowners: those who plan
plan to save their homes because they have positive home equity, but are in arrears on their
mortgage payments, and those who plan to give up their homes, but gain from delay.
Homeowners in both situations can increase their gain by filing for bankruptcy.
     The housing bubble and the financial crisis. During the housing bubble, most homeowners
had positive home equity because home prices were rising. This suggests that they mainly filed
for bankruptcy in order to save their homes. But the mortgage crisis and the subsequent financial


                                                 7
crisis and recession caused home values to fall and wiped out many homeowners’ home equity.
This suggests that after the crisis, most homeowners filed for bankruptcy in order to increase
their financial gain from giving up their homes. In our empirical work, we examine whether the
bankruptcy-foreclosure relationship changes after the beginning of the financial crisis.
   Strategic behavior and the delay effect of bankruptcy. Suppose there are two types of
homeowners: those who behave strategically versus non-strategically. Non-strategic
homeowners are assumed to default on their mortgages only when their ability-to-pay falls;
while strategic homeowners default if they gain financially from giving up their homes. Does
bankruptcy have a different effect on foreclosure depending on whether homeowners behave
strategically or non-strategically?
    Following mortgage default, lenders must choose between foreclosing quickly versus
delaying foreclosure and, if they delay, they must choose between renegotiating the terms of the
mortgage contract or not renegotiating. Suppose first that lenders never renegotiate (this was
common following the financial crisis). Also suppose lenders can identify individual
homeowners’ types. Then we assume that lenders prefer to delay foreclosure in response to
default by non-strategic homeowners, because these homeowners will repay their mortgage
arrears in full (“self-cure” in industry parlance) if their ability-to-pay increases. But lenders
prefer to foreclose quickly when strategic homeowners default, because these homeowners will
not self-cure unless house values rise. In this situation, strategic homeowners file for bankruptcy
following mortgage default, since filing is their only means of delaying foreclosure. In contrast,
non-strategic homeowners do not file to delay foreclosure; their filing decisions are made for
other reasons and their filings have little effect on the timing of foreclosure. Under these
assumptions, bankruptcy is predicted to delay foreclosure by longer for strategic than for non-
strategic homeowners. An empirical finding of this type would therefore suggest both that
lenders can identify individual homeowners’ types and that lenders do not renegotiate mortgages
following default.
   An alternative assumption is that is that lenders do renegotiate mortgage contracts, but only
for strategic homeowners. This is because strategic homeowners will resume making mortgage
payments only if their contracts are changed to make keeping the house financially worthwhile.
In contrast, non-strategic homeowners are likely to re-default even if their mortgage payments
are reduced, so that renegotiation is not worthwhile and lenders prefer to foreclose quickly.

                                                  8
Under these assumptions, non-strategic homeowners are likely to file for bankruptcy following
default because bankruptcy is their only method of delaying foreclosure, but strategic
homeowners do not file because their mortgage contracts are renegotiated. Therefore bankruptcy
is predicted to delay foreclosure by longer for homeowners who behave non-strategically. An
empirical finding that bankruptcy delays foreclosure by more for non-strategic homeowners thus
would suggest both that lenders can identify individual homeowners’ types and that they
renegotiate only the mortgage contracts of strategic homeowners. 9
      A third possibility is that lenders cannot identify individual homeowners’ types and they
therefore respond in the same way to defaults by both types. Information concerning
homeowners’ types is likely to be asymmetric since mortgage lenders do not receive updated
information concerning homeowners’ incomes and they therefore do not know whether
homeowners have suffered job loss or health problems since the mortgage originated. For
homeowners with subprime mortgages, they also do not know whether the initial income
information supplied on the mortgage application was fraudulent. In this situation, lenders’ best
strategy following default is to play mixed by sometimes delaying foreclosure, sometimes
renegotiating, and sometimes foreclosing quickly. In this situation, bankruptcy is predicted to
have the same delay effect on foreclosure for both types of homeowners. Thus if we find that
bankruptcy delays foreclosure by the same amount for both types of homeowners, the result will
provide evidence that lenders cannot identify individual homeowners’ types.


Data and summary statistics

    We construct separate datasets of prime and subprime mortgages, each combining several
sources of information. 10 For our prime mortgage dataset, we start with a large sample of prime
mortgages from LPS Applied Analytics, Inc. Only first mortgages are included. LPS provides
information from homeowners’ mortgage applications concerning their financial situation,
characteristics of the property, terms of the mortgage contract, and information about
9
  A recent New York Times article suggests that Bank of America forecloses quickly on non-strategic homeowners,
while renegotiating mortgage terms with strategic homeowners. See Schwartz (2011). For evidence that
homeowners often self-cure if their mortgages are not renegotiated and also re-default even if renegotiation occurs,
see Adelino et al (2009).
10
   Most of the related literature models mortgage default rather than foreclosure and uses only subprime mortgage
data. See, for example, Jiang et al (2010), Demyanyk and van Hemet (forthcoming), Mayer et al (2009), and Keys
et al (2010). For comparison of prime versus subprime mortgage default behavior, see Li et al (2010). Foote et al
(2008), Gerardi et al (2007) and Morgan et al (2010) examine foreclosure.

                                                          9
securitization, plus monthly updates on whether homeowners paid in full or defaulted, whether
they filed for bankruptcy, whether lenders started foreclosure and whether the home was sold in
foreclosure. We also use information from the FRB Consumer Credit Panel Data, which is a 5%
random sample of all individuals in the U.S. that have credit bureau files. It provides
information on credit card loans, installment loans, car loans, student loans, and first and second
mortgages. For each loan, quarterly updates are provided concerning the loan principle, credit
limits where applicable, and whether default or bankruptcy occurred. 11 Because the FRB data
are quarterly and the LPS data are monthly, we convert all data to quarterly. We also use
information from the Home Mortgage Disclosure Act (HMDA), which covers all mortgage
originations in the U.S. HMDA provides information on homeowners’ race, sex, age, and
income at the time of the mortgage application, plus whether the mortgage was co-signed. For
our subprime mortgage dataset, we use subprime first mortgages from CoreLogic rather than
LPS, because CoreLogic has better subprime coverage. 12 We follow the same procedure to
combine the three sources of data. 13
     The resulting datasets each consist of individual prime and subprime first mortgages that
originated in any of the years 2004, 2005 or 2006, combined with information on homeowners’
non-mortgage debts. 14 We follow these mortgages each quarter, starting from the first quarter of
2006 or the date of mortgage origination, whichever is later, and ending with the last quarter of
2010 or the date of mortgage termination, whichever is earlier. Our sample period starts in 2006

11
   Our bankruptcy variable equals one if either the FRB or the LPS/CoreLogic data indicate that a borrower filed for
bankruptcy during a particular quarter. The bankruptcy indicators from the two datasets mostly coincide, but
variable reporting lags mean that filing dates sometimes differ by one quarter.
12
   CoreLogic covers nearly all mortgages that were in non-agency subprime mortgage securitizations, except that
subprime mortgages with principle greater than $417,000 (jumbo mortgages) are excluded. Around 72% of all
subprime mortgages issued during our period were included in non-agency securitizations, making our sample fairly
representative of all subprime mortgages. (See data given in table 1 of Ashcraft and Schuermann, 2008.) There are
some slight differences in the variables between the LPS and CoreLogic databases—see discussion below.
13
   We merge the HMDA data with LPS/CoreLogic, the FRB Consumer Credit Panel with LPS/CoreLogic and then
merge the two datasets with each other. The match between HMDA and LPS/CoreLogic is done by linking
mortgages based on the zipcode of the house, the date of origination of the mortgage (within 5 days), the origination
amount (within $500), the purpose of the loan (purchase, refinance or other), the type of mortgage (conventional,
VA guaranteed, FHA guaranteed or other), occupancy type (owner-occupied or non-owner-occupied), and lien
status (first lien or other). We delete observations with missing information on the age of the borrower, age of the
loan, or appraisal amount, or if the mortgage principle was less than $2,000 at origination. The match rates between
LPS/CoreLogic and the FRB Consumer Credit Panel are 3-4%, depending on year. Note that the match rates must
be below 5% because the FRB Panel covers only 5% of consumers with credit bureau files.
14
   We take a 50% random sample of both matched datasets. We also delete mortgages from counties that were
affected by Hurricanes Katrina and Rita, which occurred in August and September 2005. This is because many
homeowners in these counties delayed making mortgage payments after the hurricanes and their delinquencies were
recorded as defaults.

                                                         10
so that all observations are after the 2005 bankruptcy reform, which went into effect during the
last quarter of 2005. Mortgage terminations occur because the home is sold, refinanced,
transferred to a different servicer, or due to foreclosure. It should be noted that mortgages in our
sample originated near the peak of the housing bubble.
        Our prime and subprime samples each include approximately 200,000 separate mortgages,
for which we have about 1.7 and 2.1 million quarterly observations for the two samples,
respectively. In our prime sample, 8.5% of mortgages start foreclosure and 3.5% are sold in
foreclosure. The fraction of prime mortgage-holders that files for bankruptcy at some point
during the sample period is 4.2%. In our subprime sample, all of these figures are substantially
higher: 24% of mortgages start foreclosure and 11% of homes are sold in foreclosure. The
fraction of subprime mortgage-holders that files for bankruptcy is 6.0%.15 When we divide the
samples according to whether homeowners filed for bankruptcy, we find that foreclosure and
bankruptcy are positively related. For our prime sample, foreclosure starts for 46% for
homeowners who file for bankruptcy, compared to only 8.4% of homeowners not in bankruptcy.
For our subprime mortgage sample, these figures are 72% and 21%, respectively. Similarly,
20% of homes with prime mortgages are sold in foreclosure when homeowners file for
bankruptcy, compared to only 2.7% when homeowners do not file. For subprime mortgages,
these figures are 33% and 9.4%, respectively. Thus homeowners who file for bankruptcy are
much more likely both to start foreclosure and to proceed to the end of the process. See Table
1. 16


     Specification and Results
     Now turn to our empirical specification. Suppose Eit denotes an event for mortgage i that

occurs in quarter t, where the event can be 90-day default, the start of foreclosure, or sale of the
property in foreclosure. Also suppose Bit equals 1 if homeowner i files for bankruptcy in quarter

t. We estimate the following model using probit:
                           Eit = a − bBit − cXit + dTt + fSi + μit                             (1)




16
  The only other paper that uses matched data that includes both mortgage and non-mortgage debt is Elul et al
(2010). Their paper focuses on explaining mortgage default.

                                                       11
Xit denotes a vector of control variables (some of which may be lagged), Tt denotes quarter

fixed effects, and Si denotes state fixed effects. We drop mortgages in the quarter following

event Eit or, if homeowners file for bankruptcy, in the quarter following the bankruptcy filing.

The hypothesis being tested is therefore that event Eit is less likely to occur in quarter t if the
homeowner files for bankruptcy in the same quarter. (Below, we also test whether bankruptcy
filings affect whether default or foreclosure occurs one quarter later.)
     Because of concerns that many of the same financial factors explain both bankruptcy filings
and default or foreclosure, we instrument for homeowners’ bankruptcy decisions. Our
instruments are the aggregate bankruptcy filing rate in the homeowner’s bankruptcy court district
in the previous quarter and the homestead and personal property exemption levels in the
homeowner’s state of residence. Bankruptcy court districts are co-terminus with states for small
states, but larger states have up to four districts. 17 Previous research has shown that both the
homestead exemption and the lagged district-level bankruptcy filing rate are positively related to
bankruptcy filings—see Fay, Hurst and White (2003). The lagged aggregate bankruptcy filing
rate in the district is related to district-level economic conditions, to the level of bankruptcy
stigma in the district, and to variations in local bankruptcy court practices that make bankruptcy
more or less debtor-friendly. The home equity and personal property exemptions are state-level
variables which measure variations in the financial attractiveness of filing for bankruptcy—more
unsecured debt is discharged in states that have higher exemption levels. We use the results from
the first stage regression to calculate homeowners’ predicted probability of filing for
           ˆ                    ˆ
bankruptcy Bit , and then enter Bit in the second stage model.
     Table 2 gives summary statistics for all variables, using our base case sample which drops
mortgages in the quarter after the start of foreclosure. Control variables include the age and age
squared of the mortgage, 18 homeowners’ total revolving debt balance lagged one quarter, a
liquidity-constraint dummy that equals one if homeowners are currently using more than 80% of
their available revolving credit, a dummy variable that equals one if homeowners have high

17
   In states that have unlimited homestead exemptions, we set the value of the homestead exemption at one million
dollars. Data concerning bankruptcy exemptions is taken from Elias (2008b), various editions. There are 94
bankruptcy districts. We are grateful to Ted Eisenberg for providing us with a program that assigns counties to
bankruptcy districts.
18
   See Demyanyk and van Hemert (forthcoming) and Jiang et al (2010) for discussion of how mortgage age affects
default decisions.

                                                        12
income—defined as more than three times the median income level in their states, a dummy
variable that equals one if home equity is negative during the current quarter, 19 homeowners’
demographic characteristics, their risk score at the time of the mortgage application, a set of
mortgage and property characteristics taken from the mortgage application, information
concerning how the mortgage originated (whether the bank acquired the mortgage from a
correspondent bank or wholesale; the omitted category is mortgages originated directly by the
lender), whether the mortgage became part of a private mortgage securitization, whether the
mortgage was a jumbo, and whether the mortgage was guaranteed by FHA or VA. We also
include a measure of homeowners’ gain from refinancing at the current interest rate each
quarter. 20 Regional macroeconomic variables include the unemployment rate in the metropolitan
area lagged one quarter, the rate of growth of household income lagged one quarter (both from
the Bureau of Labor Statistics), the rate of growth of housing prices in the metropolitan area
lagged one quarter (from the Federal Housing Finance Agency), and the divorce rate in the state
lagged one year (from the Division of Vital Statistics, National Center for Health Statistics). 21
Variables that are updated each quarter are marked with asterisks in Table 2. 22
     Among the interesting figures in Table 2 is the fact that homeowners with subprime
mortgages are less likely to be married (as indicated by the lower probability of a co-signer), less
likely to be high income, more likely to be liquidity-constrained, and more likely to live in
districts with higher aggregate bankruptcy filing rates. Also, subprime mortgage-holders have
lower average risk scores than prime mortgage-holders, but were more likely to provide full
documentation of income and assets in their mortgage applications (although the information
was probably fraudulent in many cases). Homeowners with subprime mortgage also have less



19
   Home equity is recalculated each quarter by updating the value of the house at the time of mortgage origination
using the average increase in housing values in the metropolitan area since the mortgage originated. Then the
mortgage principle in the current quarter is subtracted.
20
   The measure is {r0[1-(1+rt)t-M]}/{ rt[1-(1+r0)t-M]}, where r0 is the interest rate on the homeowner’s existing
mortgage, rt is the interest rate currently available on new mortgages, and M is the remaining term of the mortgage.
See Richard and Roll (1989). When this measure is higher, the gain from refinancing is greater.
21
   We use non-MSA house price index of the state for zipcodes that do not fall into the MSAs covered by Federal
Housing Finance Agency or the state index when non-MSA house price indexes are not available. The divorce states
are available for most years in our sample. When missing, we take the average of the two adjacent years or the
adjacent year for 2005 and 2010.
22
   Because our subprime sample does not include jumbo mortgages, mortgages with FHA or VA guarantees, or
mortgages that were not in private securitizations, these variables are unavailable for the sample. Also there is no
correspondent bank variable in the subprime sample.

                                                        13
non-mortgage debt than homeowners with prime mortgages, but are more likely to be liquidity-
constrained. Subprime mortgages are also much less likely to be fixed-rate.
      Table 3 shows the results of the first-stage regressions explaining whether homeowners file
for bankruptcy. 23 Figures given in the tables are marginal effects, with p-values in parentheses.
Errors are clustered by mortgage. The results show that the lagged district-level bankruptcy
filing rate is positive and strongly statistically significant in explaining bankruptcy in both the
prime and subprime samples. The homestead exemption is positive as expected in both
regressions, but approaches significance only in the subprime sample (p = .206 and 0.088 in the
prime and subprime samples, respectively). The personal property exemption is insignificant in
both regressions, but this is not surprising because the exemption itself is small and not very
variable. As expected, homeowners in both samples more likely to file for bankruptcy if they are
liquidity-constrained on their non-mortgage debt, but the amount of revolving debt is positive
and significant only for the sample of prime mortgage-holders. Homeowners in both samples are
also more likely to file for bankruptcy if they have negative home equity. Homeowners with
prime mortgages are less likely to file for bankruptcy if they had high income at the time of the
mortgage application, but income is not significantly related to bankruptcy filings in the
subprime sample—suggesting that income figures given on subprime mortgage applications
were often fraudulent. The other individual-level variables are mainly taken from homeowners’
mortgage applications and the results indicate that some of the variables used by lenders to
predict mortgage default also predict bankruptcy. Both groups of homeowners are less likely to
file if their home equity is higher, if they had higher risk scores at origination, and if they
provided full documentation on their mortgage applications. But—surprisingly—homeowners
with subprime mortgages are more rather than less likely to file if they have fixed rate mortgages
and if they would gain financially from refinancing.
     Table 4 shows the coefficients of the bankruptcy variable when we estimating the models of
foreclosure start and foreclosure sale both with and without instruments. In the uninstrumented
models, the bankruptcy variable equals one if homeowners filed for bankruptcy in the same
quarter as event Eit occurred; while in the IV models, the bankruptcy variable is the predicted

probability of bankruptcy from the first stage regressions. Figures in the table are marginal
23
  The results shown in Table 3 are the first stage regressions for the model explaining the start of foreclosure.   We
run the first-stage regressions separately for the model explaining foreclosure sale, because the sample drops
mortgages following foreclosure sale rather than foreclosure start. The results are similar and are not shown.

                                                          14
effects, with p-values in parentheses. Without instruments, the bankruptcy variable always has a
positive sign. But when we instrument, the bankruptcy coefficient changes from positive to
negative. The positive signs in the models without instruments reflect the fact financially
distressed homeowners who default on their mortgages are also more likely to file for
bankruptcy, and vice versa. But with instruments, the homeowner-level financial characteristics
that lead to the positive relationship between bankruptcy and foreclosure are eliminated, leaving
the negative relationship that results from the bankruptcy stay on legal actions. The results of the
IV estimation show that the elasticity of bankruptcy with respect to foreclosure start is -.05 for
prime mortgages and -.12 for subprime mortgages and the elasticities of bankruptcy with respect
to foreclosure sale are -.03 and -.11 for the two types of mortgages, respectively. All of these
results are significant at the 0.1% level (p < .000). These results imply that filing for bankruptcy
reduces the probability that foreclosure starts and that foreclosure sale occurs for both prime and
subprime mortgages, but the delay effect of bankruptcy is much larger for subprime than prime
mortgages.
      Table 5 shows the full results of the foreclosure start and foreclosure sale models, estimated
using IV. The top and bottom panels are for prime and subprime mortgages, respectively. The
results for the predicted bankruptcy variable have already been discussed. Among the other
variables, the liquidity-constraint indicator is positive and significant in all regressions,
indicating that foreclosure is more likely when homeowners are liquidity-constrained. 24 As in
the bankruptcy regressions, high income is negatively associated with the start of foreclosure for
prime mortgage-holders. But for subprime mortgage-holders, high income positively associated
with both foreclosure start and foreclosure sale—additional evidence that income figures on
subprime mortgage applications were often fraudulent. Again the models fit better in the prime
than the subprime mortgage sample, suggesting that lenders’ behavior—like homeowners’—is
more difficult to predict in the subprime sample.
      We also re-ran the benchmark model using a sub-sample of mortgages that have been
delinquent by 30 days or more at some point during our sample period—these homeowners are
known to be financially shaky. The IV results for the predicted bankruptcy variable are shown in
table 6. The elasticities of foreclosure start with respect to bankruptcy are -.14 and -.18 for the
prime and subprime samples, respectively, and the figures for foreclosure sale are -.20 and -.06,

24
     Elul et al (2010) found that being liquidity-constrained is also an important determinant of mortgage default.

                                                            15
respectively. All of these results are strongly statistically significant and, except for the last
figure, all are more negative than in the full sample. These results suggest that filing for
bankruptcy is even more effective in delaying foreclosure when homeowners have a known
history of delinquency on debt payments. The larger effect of bankruptcy may reflect either
lenders’ decisions to slow the foreclosure process or that bankruptcy trustees provide additional
help to these homeowners.
     We also reran our base case model separately for the sub-periods before versus after the
start of the financial crisis. Our pre-crisis period covers the beginning of 2006 to the middle of
2008, while our post-crisis period covers the last half of 2008 to the end of 2010. We use the
middle of 2008 as the dividing line, since credit markets began to freeze up and the financial
crisis began around this time. The decline in housing prices also accelerated at this time,
although prices had begun to fall earlier. The results are shown in table 7, top panel. All of the
bankruptcy coefficients remain negative and strongly statistically significant. For the subprime
sample, the delay effect of bankruptcy remains about the same in both periods. However for the
prime mortgage sample, the delay effect of bankruptcy is much larger in the post-crisis than the
pre-crisis period. These results suggest that although bankruptcy judges had no formal power to
change the terms of mortgage contracts, they nonetheless used their power to delay foreclosure
as a means of pressuring lenders to renegotiate mortgages and they focused their efforts on prime
rather than subprime mortgages.
     Now turn to the question of whether the delay effect of bankruptcy on foreclosure differs
depending on whether homeowners behave strategically or non-strategically. As an indicator for
strategic behavior, we use a dummy variable that equals one if homeowners satisfy three
conditions: their income at the time of the mortgage application exceeds $100,000, their loan-to-
value ratio (current mortgage principle divided by predicted house value in the current quarter)
exceeds 90%, and they are not delinquent by 60 days or more on any of their credit cards. 25
Homeowners with high incomes have more to gain from behaving strategically and those with
high loan-to-value ratios gain more from giving up their homes. Also the fact that these
homeowners are not in default on any of their credit card debts suggests that they have not lost
their jobs or had other financial reverses in the recent past, so that their mortgage defaults are


25
  To estimate home value each quarter, we update the value of the house at the time of mortgage origination using
the increase in the housing price index for the relevant metropolitan area since the date of mortgage origination.

                                                        16
likely to be strategic. About 10% and 7% of homeowners with prime and subprime mortgages,
respectively, satisfy these conditions. We reran the regressions discussed in table 5, adding a
strategic homeowner dummy and an interaction between this dummy and the predicted
bankruptcy variable. In this specification, the coefficient of the predicted bankruptcy variable
(interpreted as a marginal effect) equals the effect of bankruptcy on the probability of foreclosure
for non-strategic homeowners and the coefficient of the interaction term (interpreted as a
marginal effect) equals the difference between the effect of bankruptcy on the probability of
foreclosure for strategic relative to non-strategic homeowners. The results are reported in table
8.
     The results show that the delay effect of bankruptcy differs strongly for strategic versus non-
strategic homeowners in the prime mortgage sample, but hardly differs at all in the subprime
mortgage sample. In the prime mortgage sample, the delay effect of bankruptcy on the start of
foreclosure is 80% larger for strategic versus non-strategic homeowners and the delays effect of
bankruptcy on the sale of homes in foreclosure is twice as large. These results support our
hypotheses that lenders can identify individual homeowners’ types if they have prime mortgages
and that lenders do not renegotiate mortgages in response to default by strategic homeowners.
In contrast, our results for the subprime sample suggest that the delay effect of bankruptcy is
virtually the same for both types of homeowners (although the small differences are statistically
significant). This suggests that lenders cannot identify individual homeowners’ types among
subprime mortgage-holders and they therefore respond in the same way to defaults by both
types. Given the widespread fraud in subprime mortgage applications, it is not surprising that
we find evidence that asymmetric information affected lenders’ response to default.



Conclusion
     In this paper, we examine whether bankruptcy filings by homeowners delay mortgage
foreclosure. We find that the relationships between bankruptcy and whether foreclosure starts or
whether homes are sold in foreclosure in the current quarter are uniformly negative and strongly
statistically significant. The elasticities of foreclosure with respect to bankruptcy are around -.03
to -.05 for our prime mortgage sample and around -.11 for our subprime mortgage sample. The
elasticities are even larger (more negative) in a subsample of mortgages where homeowners have


                                                 17
previously been in default on their credit card debt. Overall, these results suggest that
bankruptcy helps homeowners by delaying both the beginning and the end of the foreclosure
process.
     Delaying the foreclosure process may be efficiency-enhancing for a number of reasons.
Because displaced homeowners sometimes become homeless, delaying foreclosure gives them
more time to save for the costs of renting alternative housing, including paying for the move
itself, the security deposit and the first month’s rent. 26 Even if displaced homeowners avoid
homelessness, forcing them to move is inefficient since the social cost of foreclosure exceeds
lenders’ private cost. Among the extra costs not borne by lenders are that children of
homeowners are often forced to shift schools, causing their school performance to suffer. Also,
foreclosures cause neighboring homes fall in value, resulting in loss of home equity for
neighboring homeowners who therefore become more likely to default themselves. Foreclosures
homes often remain vacant and deteriorate, which causes neighborhoods to become blighted, and
local governments are forced to cut public services when property values fall and property tax
payments decline. 27 Delaying foreclosure also increases the probability that lenders and
homeowners renegotiate mortgages, rather than lenders selling the property in foreclosure. Thus
to the extent that filing for bankruptcy increases homeowners’ probability of remaining in their
homes, it increases the efficiency of the housing market.




26
   A recent study estimates that 10% of homeless people helped by social service agencies in 2008-09 became
homeless due to foreclosure of their homes. However some of these households previously rented rather than
owned and lost their homes because their landlords gave up the properties; for these households, bankruptcy would
not have delayed foreclosure. See “National Homelessness Advocacy Groups Release ‘Foreclosure to
Homelessness,’” June 27, 2009, www.prweb.com/releases/End _Homelessness/Foreclosure/prweb2581984.html and
Goodman (2009).
27
   Los Angeles has recently sued Deutsche Bank for being a slumlord and causing neighborhood blight by allowing
foreclosed properties to deteriorate. See “Los Angeles Sues Deutsche Bank Over Foreclosure Blight,” Insurance
Journal, May 5, 2011 www.insurancejournal.com/news/west2011/05/05/197373.html. See Campbell et al
(forthcoming) for an estimate of the decline in value of neighboring houses when foreclosures occur nearby.

                                                       18
                                         References


Adelino, Manuel, Kristopher Gerardi, and Paul Willen. 2009. “Why Don’t Lenders Renegotiate
More Home Mortgages? Redefaults, Self-Cures, and Securitization,” FRB of Boston Public
Policy Discussion Paper No 0904.

Ashcraft, Adam B., and Til Schuermann. 2008. “Understanding the Securitization of Subprime
Mortgage Credit,” Staff Report no. 318, NY Federal Reserve Bank, March 2008.
www.newyorkfed.org/research/staff_reports/sr318.pdf.

Berkowitz, Jeremy, and Richard Hynes. 1999. “Bankruptcy Exemptions and the Market for
Mortgage Loans,” J. of Law & Economics, 42: 809-830.

Campbell, John, Stefano Giglio, and Parag Pathak, “Forced Sales and Housing Prices,” Am.
Economic Rev., forthcoming.

Carroll, Sarah, and Wenli Li. 2011. Forthcoming. “The Homeownership Experience of
Households in Bankruptcy,” Cityscape.

Demyanyk, Yulia, and Otto van Hemert. Forthcoming. “Understanding the Subprime
Mortgage Crisis,” Review of Financial Studies.

Elias, Stephen. 2008. The New Bankruptcy: Will It Work for You? 1st edition. Berkeley: Nolo
Press.

Elias, Stephen. 2008. The Foreclosure Survival Guide, 1st edition. Berkeley: Nolo Press.

Elul, Ronel, Nicholas S. Souleles, Souphala Chomsisengphet, Dennis Glennon, and Robert
Hunt. 2011. “What Triggers Mortgage Default?” American Economic Review, 100(2): 490-94.

Elul, Ronel. 2009. “Securitization and Mortgage Default: Reputation versus Adverse Selection.”
Federal Reserve Bank of Philadelphia Working Paper 09-21.

Experian. 2011. “Experian Announces Strategic Defaults Remain High,” PR Newswire, July
23, 2011.

Foote, Christopher L., Kristopher S. Gerardi, Paul S. Willen. 2008. “Negative Equity and
Foreclosure: Theory and Evidence.” J. of Urban Econ., vol. 64: 234-45.

Ghent, Andra, and Marianna Kudlyak, 2011, “Recourse and Residential Mortgage Default:
Evidence from U.S. States,” Federal Reserve Bank of Richmond Working paper No. 09-10R.



                                              19
Gerardi, Kristopher, Adam Hale Shapiro, and Paul S. Willen (2007), “Subprime Outcomes:
Risky Mortgages, Homeownership Experiences and Foreclosures,” Federal Reserve Bank of
Boston Working Paper 0715.

Goodman, Peter S., “Foreclosures Force Ex-Homeowners to Turn to Shelters,” New York Times,
October 19, 2009.

Jiang, Wei, Ashlyn Aiko Nelson, and Edward Vytlacil. 2010. “Liar’s Loan? Effects of
Origination Channel and Information Falsification on Mortgage Delinquency,” Working paper,
Columbia University.

Keys, Benjamin J., Tanmoy K. Mukherjee, Amit Seru, and Vikrant Vig. 2010. “Did
Securitization Lead to Lax Screening? Evidence from Subprime Loans,” Quarterly Journal of
Economics, 125: 307-362.

Levitin, Adam, and Joshua Goodman, 2008. The Effect of Bankruptcy Strip-down on Mortgage
Markets. Presented at the American Law and Economics Association Conference, 2008.
www.law.georgetown.edu/faculty/levitin/Publications.

Li, Wenli, and Michelle J. White. 2009. “Mortgage Default, Foreclosures and Bankruptcy.”
NBER working paper 15472.

Li, Wenli, Michelle J. White and Ning Zhu, 2009. “Did Bankruptcy Reform Cause Mortgage
Defaults to Rise?” NBER working paper 15968 (2009). Forthcoming, American Economic
Journal: Economic Policy.

Lin, Emily, and Michelle J. White, 2001. “Bankruptcy and the Market for Mortgage and Home
Improvement Loans,” J. of Urban Econ., 50:138-162.

Mayer, Christopher, Karen Pence and Shane Sherlund. 2009. “The Rise in Mortgage Defaults,”
Journal of Economic Perspectives, 23(1): 27–50.

Morgan, Donald P., Benjamin Iverson, and Matthew Botsch. 2011. “Did the 2005 Bankruptcy
Reform Increase Subprime Foreclosures?” Federal Reserve Bank of New York Economic Policy
Review, forthcoming.

Pence, Karen M., 2006. “Foreclosing on Opportunity: State Laws and Mortgage Credit.” Rev.
of Econ. & Stat., vol. 88: 177-82.

Porter, Katherine. 2008. “Misbehavior and Mistake in Bankruptcy Mortgage Claims,” Texas Law
Review, 87(1): 121-182.

Richard, Scott F., and Richard Roll. 1989. “Prepayments on Fixed-rate Mortgage-backed
Securities,” Journal of Portfolio Management, 15(3): 73-82.




                                             20
Schwartz, Nelson D. 2011. “Bank’s Deal Means More Will Lose Their Homes,” New York
Times, July 12, 2011, p. B1.

White, Michelle J. 1998. “Why It Pays to File for Bankruptcy: A Critical Look at the Incentives
Under U.S. Personal Bankruptcy Law and a Proposal for Change.” Univ. of Chicago Law Rev.,
vol. 65: 685-732.


White, Michelle J. 2009. “Bankruptcy: Past Puzzles, Recent Reforms, and the Mortgage
Crisis,” Am. Law and Economics Review, Spring 2009.

White, Michelle J., and Ning Zhu. 2010. “Saving Your Home in Chapter 13 Bankruptcy,”
Journal of Legal Studies, 39(1): 33-61.

Zhu, Ning. 2010. “Household Consumption and Personal Bankruptcy,” J. of Legal Studies, .




                                              21
                                          Table 1:
                     Probability of Bankruptcy, Default and Foreclosure

                                Full Sample (2006q1 – 2010q4)
                         Full sample          If homeowner files for     If no bankruptcy filing
                                                    bankruptcy
Probability of:       Prime      Subprime       Prime     Subprime        Prime       Subprime
 Bankruptcy           0.042        0.060        1.000       1.000         0.000         0.000
 30-day               0.245        0.483        0.731       0.901         0.223        0.457
delinquency
 Foreclosure start    0.084        0.242        0.463        0.723         0.068        0.211
 Foreclosure sale     0.035        0.109        0.205        0.333         0.027        0.094

Notes: the sample covers 2006, quarter 1, through 2010, quarter 4. Mortgages are coded as one
if the event occurred at any time during the period and the figures given are averages over all
mortgages.




                                              22
                                        Table 2: Summary Statistics

                                                     Prime Mortgages             Subprime Mortgages
     District bankruptcy filing rate,                0.00071 (0.00084)            0.00110 (0.00072)
     lagged
     Homestead exemption ($000)                      216.829 (352.250)             223.783 (363.205)
     Personal property exemption ($000)               10.005 (10.747)                9.639 (10.067)
     Dummy for negative home equity,                   0.139 (0.146)                 0.195 (0.396)
     lagged
     Dummy for income > three times                     0.077 (0.266)                 0.029 (0.167)
     the state median
     Revolving debt balance, lagged                    24.607 (65.209)              13.164 (37.315)
     ($000)*
     If liquidity-constrained, lagged*                  0.085 (0.279)                 0.196 (0.397)
     Homeowner’s age (years)                           47.049 (12.898)               46.740 (12.241)
     Mortgage age (quarters)                             7.756 (091)                 11.670 (6.425)
     If mortgage was co-signed                          0.547 (0.500)                 0.389 (0.487)
     Income at origination ($000)                     96.678 (116.072)              73.947 (74.234)
     Risk score at origination                        721.579 (71.511)              649.437 (82.971)
     Full documentation                                 0.344 (0.475)                 0.686 (0.464)
     If mortgage was privately                          0.216 (0.411)                 1.000 (0.000)
     securitized
     Primary residence                                  0.906 (0.292)                 0.953 (0.212)
     Single family                                      0.792 (0.406)                 0.793 (0.405)
     If mortgage originated by                          0.183 (0.386)                 0.000 (0.000)
     correspondent bank
     If mortgage acquired wholesale                     0.171 (0.376)                 0.106 (0.308)
     Fixed rate mortgage                                0.762 (0.426)                 0.241 (0.428)
     If refinance (versus purchase)                     0.485 (0.500)                 0.661 (0.473)
     Homeowner’s gain from                              0.986 (0.102)                 0.830 (0.120)
     refinancing*
     If FHA/VA guarantee                                0.061 (0.240)                 0.000 (0.000)
     Jumbo mortgage                                     0.110 (0.313)                 0.000 (0.000)
     Unemployment rate, lagged* (%)                     5.589 (2.411)                 6.289 (2.561)
     Income growth rate, lagged* (%)                    1.007 (0.011)                 1.007 (0.011)
     House price growth rate, lagged*                   1.001 (0.034)                 0.999 (0.020)
     (%)
     Divorce rate, lagged* (%)                          2.767 (1.652)                 2.783 (1.627)
Notes: The sample used to calculate these figures drops mortgages in the quarter after foreclosure starts or a
bankruptcy filing occurs; this is the same sample used for the regressions explaining the start of foreclosure. All
figures are averages over all mortgage-quarter observations. Standard deviations are given in parentheses. All lags
are one quarter, except for the divorce rate which is lagged one year. Asterisks indicate variables that are updated
each quarter. Variable definitions are given in the text. We set the homestead exemptions to be $1 million for states
that have unlimited homestead exemptions. Due to our agreements with data vendors, we do not report mean values
or regression coefficients for borrowers’ gender or race.


                                                         23
           Table 3: Results of First Stage Regressions Explaining Bankruptcy Filings

                                                    Prime mortgages      Subprime mortgages
District bankruptcy filing rate lagged (%)               0.227 (0.000)       0.554 (0.000)
Homestead exemption (thousands $)                     1.64e-6 (0.206)       3.19e-6 (0.088)
Personal property exemption (thousands $)             6.90e-6 (0.313)       -1.16e-6 (0.896)
Dummy for negative home equity                       0.000716 (0.000)       0.000167 (0.045)
Dummy for high income at origination                 -0.000347 (0.000)      0.000190 (0.267)
Revolving debt balance, lagged (thousands $)          1.02e-6 (0.000)       -1.32e-6 (0.026)
If liquidity-constrained, lagged                      0.00233 (0.000)       0.00126 (0.000)
Homeowner’s age (years)                               2.75e-5 (0.004)       -4.31e-6 (0.754)
Homeowner’s age squared (years)                      -2.37e-7 (0.013)       -5.13e-9 (0.970)
Mortgage age (quarters)                              0.000123 (0.000)       -7.99-6 (0.743)
Mortgage age squared (quarters)                      -5.30e-6 (0.000)       -7.51e-6 (0.000)
Mortgage co-signed                                    6.27-5 (0.159)        0.000329 (0.000)
Interaction of negative home equity and liquidity    -0.000110 (0.284)      0.000107 (0.456)
constraint
Risk score at origination                            -7.10e-6 (0.000)       -2.52e-6 (0.000)
If mortgage had full documentation                   -0.000105 (0.014)     -0.000401 (0.000)
If mortgage was privately securitized                0.000249 (0.000)              --
If primary residence                                 0.000154 (0.028)       6.03e-5 (0.649)
If property is single family                          2.47e-5 (0.644)       4.66e-5 (0.528)
If mortgage acquired wholesale                        9.84e-5 (0.072)       3.67e-5 (0.699)
If mortgage originated by correspondent              0.000204 (0.001)              --
If fixed rate mortgage                               -0.000381 (0.000)      0.000292 (0.000)
If mortgage was for refinance (versus purchase)      0.000309 (0.000)       5.26e-5 (0.394)
Homeowner’s gain from refinancing                    -0.00139 (0.000)       0.00187 (0.000)
If mortgage guaranteed by FHA/VA                     0.000302 (0.001)              --
If jumbo mortgage                                    -0.000334 (0.000)             --
Unemployment rate, lagged                             4.64e-5 (0.000)       7.54e-5 (0.063)
House price growth rate, lagged                      -0.00544 (0.000)       -0.00494 (0.043)
Income growth rate, lagged                           -0.00263 (0.394)       -0.00146 (0.749)
Divorce rate, lagged                                  8.91e-5 (0.622)      -0.000143 (0.465)
State dummies?                                                Y                    Y
Quarter dummies?                                              Y                    Y
Pseudo R-squared                                            0.0967              0.0399
N                                                         1,728,556            2,129,206


                                                    24
Note: The sample used here drops mortgages in the quarter after foreclosure starts or a bankruptcy filing occurs.
Equations are estimated using probit. Figures given are marginal effects with p-values in parentheses.




                                                         25
                                             Table 4:

                    Results Explaining the Effect of Bankruptcy Filings on
                            Foreclosure Start and Foreclosure Sale
                               With and Without Instruments

                                           Full Sample


                              Foreclosure start                        Foreclosure sale
                     No instruments With instruments          No instruments With instruments
Prime mortgages        0.0173 (0.000)      -0.117 (0.000)      0.00101 (0.000)         -0.023 (0.000)
                        ε = 0.0066          ε = -0.045           ε = 0.0013              ε = -0.029
Subprime               0.103 (0.000)       -0.716 (0.000)     0.000207 (0.627)         -0.179 (0.000)
mortgages                ε = 0.015           ε = -0.117         ε = 0.00014             ε = -0.107

Note: figures given are marginal effects (p-values in parentheses) and elasticities.




                                                26
                                 Table 5:
        IV Results Explaining Foreclosure Start and Foreclosure Sale

                                     Prime Mortgages
                                          Foreclosure start     Foreclosure sale
Predicted bankruptcy                         -0.117 (.000)        -0.0231 (0.000)
Liquidity-constrained, lagged               0.00428 (0.000)      0.000866 (0.000)
Revolving debt balance, lagged ($000)        1.04e-7 (0.740)     -8.66e-7 (0.000)
Dummy for negative home equity               0.00195 (0.000)     -3.72e-5 (0.136)
Dummy for high income at origination        -0.000501 (0.000)    -6.02e-5 (0.092)
Interaction of negative home equity and     -0.000122 (0.279)     3.77e-5 (0.456)
liquidity constraint
Mortgage age (quarters)                     0.000239 (0.000)     0.000106 (0.000)
Mortgage age squared (quarters)             -1.18e-5 (0.000)     -4.77e-6 (0.000)
If mortgage co-signed                       -0.000757 (0.000)    -0.000200 (0.000)
Risk score                                  -1.77e-5 (0.000)     -3.92e-6 (0.000)
If mortgage had full documentation          -0.000205 (0.000)    -4.52e-5 (0.012)
If mortgage was privately securitized       0.00114 (0.000)      0.000332 (0.000)
If primary residence                        0.000450 (0.000)     -4.40e-6 (0.882)
If property is single family                0.000125 (0.016)     -2.88e-8 (0.999)
If mortgage acquired wholesale              0.000121 (0.039)      4.35e-5 (0.066)
If mortgage originated by correspondent      0.000351 (.000)      8.88e-5 (0.001)
If fixed rate mortgage                      -0.00181 (0.000)     -.000579 (0.000)
If mortgage was for refinance (versus       -1.85e-5 (0.692)     -0.000138 (0.000)
purchase)
Homeowner’s gain from refinancing            -0.0112 (0.000)     -0.00300 (0.000)
If mortgage guaranteed by FHA/VA             0.00168 (0.000)     0.000753 (0.000)
If jumbo mortgage                           0.000250 (0.003)      5.74e-5 (0.085)
Homeowner’s age (years)                     -4.96e-5 (0.000)     -2.38e-5 (0.000)
Homeowner’s age squared (years)              4.15e-7 (0.000)      2.13e-7 (0.000)
Unemployment rate, lagged                   0.000148 (0.000)      3.80e-5 (0.000)
House price growth rate, lagged              -0.0158 (0.000)     -0.00517 (0.000)
Income growth rate, lagged                  -0.00340 (0.302)     -0.00198 (0.129)
Divorce rate, lagged                        -7.38e-5 (0.696)     0.000141 (0.068)
State dummies?                                      Y                   Y
Quarter dummies?                                    Y                   Y
Pseudo R-squared                                  0.193               0.199
Number of observations                           1,728,556          1,756,043



                                            27
                                              Subprime Mortgages
                                                     Foreclosure start          Foreclosure sale
           Predicted bankruptcy                           -0.716 (0.000)            -0.179 (0.000)
           If liquidity-constrained, lagged              0.00487 (0.000)          0.000946 (0.000)
           Revolving debt balance, lagged                1.27e-6 (0.506)          -1.82e-6 (0.116)
           Dummy for negative home equity                0.000868 (0.000)         -0.00116 (0.000)
           Dummy for high income at origination          0.00275 (0.000)          0.000796 (0.000)
           Interaction of negative home equity and       -0.000276 (.421)        -0.000452 (0.010)
           liquidity constraint
           Mortgage age (quarters)                      -0.000175 (0.000)         0.000603 (0.000)
           Mortgage age squared (quarters)               -1.17e-6 (0.000)         -4.78e-5 (0.000)
           If mortgage co-signed                         -0.00149 (0.000)        -0.000775 (0.000)
           Risk score                                    -2.08e-5 (0.000)         -5.02e-6 (0.000)
           If mortgage had full documentation            -0.00236 (0.000)         -.000869 (0.000)
           If primary residence                         -0.000503 (0.096)         -0.00111 (0.000)
           If property is single family                 -0.000379 (0.020)         -.000245 (0.002)
           If mortgage acquired wholesale               -0.000483 (0.031)        -0.000398 (0.000)
           If fixed rate mortgage                        -0.00202 (0.000)         -0.00109 (0.000)
           If mortgage was for refinance (versus         -0.00383 (0.000)         -0.00206 (0.000)
           purchase)
           Homeowner’s gain from refinancing             -0.0275 (0.000)           -0.0139 (0.000)
           Homeowner’s age (years)                      -0.000175 (0.000)         -9.14e-5 (0.000)
           Homeowner’s age squared (years)               1.17e-6 (0.000)           7.02e-7 (0.000)
           Unemployment rate, lagged                     0.00141 (0.000)          0.000630 (0.000)
           House price growth rate, lagged               -0.0729 (0.000)           -0.0238 (0.000)
           Income growth rate, lagged                    -0.0781 (0.000)           -0.0292 (0.000)
           Divorce rate, lagged                          -0.00236 (0.000)        -0.000910 (0.000)
           State dummies?                                       Y                        Y
           Quarter dummies?                                     Y                        Y
           Pseudo R-squared                                   0.0730                   0.0893
           Number of observations                            2,129,206               2,325,823
Note: Equations are estimated using probit. Figures given are marginal effects with p-values in parentheses.




                                                        28
                                             Table 6:

             IV Results Explaining the Effect of Current Bankruptcy Filings on
                      Foreclosure Start and Foreclosure Liquidation


                         Sample of Mortgages Ever in 30-day Default

                                            Foreclosure start       Foreclosure sale
         Prime Mortgages                       -0.644 (0.000)          -0.215 (0.000)
                                                 ε = -0.142              ε = -0.197
         Subprime Mortgages                     -1.57 (0.000)          -0.104 (0.000)
                                                 ε = -0.175             ε = -0.0594

Note: figures given are marginal effects (p-values in parentheses) and elasticities.




                                                29
                                          Table 7:
             IV Results Explaining the Effect of Current Bankruptcy Filings on
                          Foreclosure Start and Foreclosure Sale:
                             Pre- versus Post-Financial Crisis

                                          Full Sample


                                      Pre-Financial Crisis
                                            Foreclosure start       Foreclosure sale
         Prime Mortgages                        -0.069 (< 0.000)     -0.0104 (< 0.000)
                                                   ε = -0.021            ε = -0.012
         Subprime Mortgages                     -0.993 (< 0.000)      -0.178 (< 0.000)
                                                   ε = -0.126            ε = -0.114


                                      Post-Financial Crisis
                                             Foreclosure start      Foreclosure sale
         Prime Mortgages                        -0.250 (< 0.000)     -0.060 (< 0.000)
                                                   ε = -0.101           ε = -0.078
         Subprime Mortgages                     -0.618 (< 0.000)     -0.150 (< 0.000)
                                                   ε = -0.117           ε = -0.108

Note: The pre-financial crisis period is 2006q1 to 2008q2 and the post-financial crisis period is
2008q2 to 2010q4. Figures given are marginal effects (with p-values in parentheses) and
elasticities.




                                                30
                                        Table 8:

        IV Results Explaining the Delay Effect of Bankruptcy on Foreclosure
                  For Strategic versus Non-strategic Homeowners


                                       Full Sample
                                              Foreclosure start      Foreclosure sale
Prime Mortgages
Delay effect for non-strategic                   -0.103 (< 0.000)    -0.0209 (< 0.000)
homeowners                                          ε = -.040            ε = -.027
Change in delay effect for strategic            -0.0837 (< 0.000)     -0.0207 (< 0.000)
homeowners                                          ε = -.032             ε = -.027

Subprime Mortgages
Delay effect for non-strategic                   -0.642 (< 0.000)     -0.160 (< 0.000)
homeowners                                          ε = -.096             ε = -0.11

Change in delay effect for strategic            -0.00259 (< 0.000)    -0.000450 (0.012)
homeowners                                         ε = -.00039           ε = -.00031




                                           31

								
To top