Description
The purpose of this paper is to examine whether the underserved area requirements for
Fannie Mae and Freddie Mac (the government-sponsored enterprises [GSEs]) and the community needs
requirements of the Community Reinvestment Act (CRA) contributed to the house price run-up in the
USA
Journal of Financial Economic Policy
Was there a regulatory approval market for mortgages?
Peggy Crawford J oetta Forsyth
Article information:
To cite this document:
Peggy Crawford J oetta Forsyth , (2015),"Was there a regulatory approval market for mortgages?",
J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 354 - 365
Permanent link to this document:
http://dx.doi.org/10.1108/J FEP-06-2015-0032
Downloaded on: 24 January 2016, At: 21:53 (PT)
References: this document contains references to 20 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 25 times since 2015*
Users who downloaded this article also downloaded:
Muhammad Umar, Gang Sun, (2015),"Country risk, stock prices, and the exchange rate of the
renminbi", J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 366-376 http://dx.doi.org/10.1108/
J FEP-11-2014-0073
Muhammad Ali Nasir, Mushtaq Ahmad, Ferhan Ahmad, J unjie Wu, (2015),"Financial and economic
stability as ‘two sides of a coin’: Non-crisis regime evidence from the UK based on VECM", J ournal of
Financial Economic Policy, Vol. 7 Iss 4 pp. 327-353 http://dx.doi.org/10.1108/J FEP-01-2015-0006
J ames R. Barth, Tong Li, Wen Shi, Pei Xu, (2015),"China’s shadow banking sector: beneficial or
harmful to economic growth?", J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 421-445 http://
dx.doi.org/10.1108/J FEP-07-2015-0043
Access to this document was granted through an Emerald subscription provided by emerald-
srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
*Related content and download information correct at time of
download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Was there a regulatory approval
market for mortgages?
Peggy Crawford and Joetta Forsyth
Graziadio School of Business and Management, Pepperdine University,
Malibu, California, USA
Abstract
Purpose – The purpose of this paper is to examine whether the underserved area requirements for
Fannie Mae and Freddie Mac (the government-sponsored enterprises [GSEs]) and the community needs
requirements of the Community Reinvestment Act (CRA) contributed to the house price run-up in the
USA.
Design/methodology/approach – This paper predicts the incidence of “Rebounds”, which indicate
that a mortgage had been previously denied, to provide evidence on whether certain regulations caused
excessively risky mortgage originations. As a different lender rejected the loan given the interest rate
that they were willing to charge and information on the borrower, a higher incidence of Rebounds
provides evidence that lenders were more frequently disagreeing about loans. This can indicate
differences in regulatory pressure or oversight across lenders.
Findings – This paper provides evidence that the GSEs were purchasing fewer Rebounds directly
fromlenders. However, evidence suggests that indirectly, the securitization market served as a conduit
for Rebounds to the GSEs that needed to satisfy regulatory underserved area requirements. The
necessity of complying with the CRA was found to increase Rebounds. Among regulators, the Federal
Reserve was found to have been particularly associated with Rebounds.
Originality/value – The paper’s contribution comes from linking Rebounds to legislative and
regulatory infuences. This contributes to the literature on excess credit and fraud, as well as the effect
of underserved area requirements and the CRA. Also, this paper adds a newdimension to the literature
on securitization, by showing the infuence of regulation on the securitization of risky mortgages.
Keywords Mortgages, Government policy and regulation, Financial institutions and services,
Corruption, Financial meltdown, Housing supply and markets
Paper type Research paper
Introduction
There are a number of factors that could have contributed to the house price run-up and
subsequent fnancial crisis in the USA. This paper focuses on several legislative and
regulatory triggers which may have impacted the behavior of government-sponsored
enterprises (GSEs) and banks as potential contributors.
The GSEs were subject to increasingly stringent requirements to serve
“underserved areas” by the US Department of Housing and Urban Development
(HUD). These were areas with low-income and minority residents. Table I provides
the GSE underserved area requirement over time, as reported by Barth (2009).
However, the GSEs were also required by the Offce of Federal Housing Enterprise
Oversight (OFHEO) to have underwriting standards that prevented them from
The authors wish to thank the Graziadio School of Business and Management for funding from
Funds for Excellence and the Julian Virtue Award.
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1757-6385.htm
JFEP
7,4
354
Received1 June 2015
Revised6 July2015
Accepted6 July2015
Journal of Financial Economic
Policy
Vol. 7 No. 4, 2015
pp. 354-365
©Emerald Group Publishing Limited
1757-6385
DOI 10.1108/JFEP-06-2015-0032
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
directly purchasing loans of insuffcient credit quality – an issue with some
underserved area mortgages. In other words, the GSEs had conficting regulations
that both encouraged and prevented them from purchasing mortgages from
underserved areas.
A “work-around” was created that allowed the GSEs to use a middleman with
different underwriting standards. In fact, Buttimer (2011) argues that the GSEs and
other lenders were encouraged by regulation to purchase securitized mortgages. The
GSEs were not allowed to directly purchase low credit-quality mortgages from lenders,
but they were allowed to purchase themfromsecuritizers, acting as middlemen, who did
not have the same restrictions. The GSEs may have been willing to pay a premium to
securitizers because of the additional beneft of satisfying regulatory requirements,
encouraging the origination of low quality loans.
Another factor which may have contributed to the house price run-up was the
Community Reinvestment Act (CRA) of 1977, combined with a merger and expansion
wave among banks. The CRA required that banks help meet the credit needs of their
local community consistent with safe and sound operations. To enforce the Act, federal
regulatory agencies were instructed to examine banks for CRAcompliance and consider
this information when approving applications for mergers or acquisitions. Bank
expansion was still limited (primarily within a state) until the Riegle-Neal Interstate
Banking and Branching Effciency Act of 1994 eliminated restrictions on interstate
banking. In addition, it required that regulatory agencies consider the CRA rating of
the out-of-state bank when approving interstate branches. It is also possible that these
lenders bought loans from the securitization market to satisfy the CRA.
There are a number of reasons banks may have an incentive to take on excessive risk.
For instance, deposit insurance may create the incentive for a bank to take on excessive
risk by protecting depositors in case of bank default. Any time mortgages are sold-off or
insured by an outside party; there may be an incentive for the originator to reduce
screening of the loans. Securitization, purchases by the GSEs and guarantees by the
Federal Housing Administration and Veteran’s Administration may have provided
these. It should be noted that none of these issues were new to the house price run-up.
However, regulatory oversight of underwriting standards may have been loosened,
and this may have been especially problematic if other entities that had an incentive to
screen mortgages were unaware of the shift. If regulators had previously provided
strong oversight of mortgage underwriting practices, then purchasers of mortgages,
bank creditors and others may have come to rely on regulators to screen loans. If the
screening unexpectedly stopped, these entities may not have known to increase their
own screening to fll the void. In a speech after housing prices collapsed, Alan
Greenspan admitted that he did not have the Federal Reserve regulate the mortgage
practices of banks carefully because he believed that it was not in a bank’s self-interest
to make excessively risky loans[1].
Intricately tied up with the question of what caused the house price run-up is the
question of what lenders knew. If lenders were caught up with market forces or investor
Table I.
GSE underserved
area requirement
over time
Year Before 2001 2001-2004 2005 2006-2007 2008
GSE underserved area requirement (%) 24 31 37 38 39
355
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
exuberance, then it can be argued that they did not have reason to believe that the
housing market would collapse and felt their loans were appropriately compensated for
risk given available market information. On the other hand, if they were responding to
regulatory pressure to make poor-quality loans or, conversely, freed to make poor
quality loans from a lack of oversight, it can be argued that they were aware that the
interest rate that they charged did not fully compensate the suppliers of capital for the
risk of the loans.
This paper uses “Rebound” to examine this issue. Rebound refers to originated
mortgage loans that were previously denied by another lender. One reason these loans
may have been denied is that a different lender considered the loan to be insuffciently
compensated for risk given the interest rate that they were willing to charge and the
information available to them. Lenders may have different policies regarding the
amount of risk (and subsequent interest rate) that they are willing to bear. However, if
some, but not other lenders, are put under pressure to make certain loans or released
from oversight by their regulator, then more frequent “disagreements” across lenders
about loans can be the result. Rebound can therefore be used to investigate whether
some lenders were either subject to increased regulatory pressure to make certain loans,
released from regulatory oversight, or both.
Literature review
There are a number of papers that address excessive credit and securitization. Mian and
Suf (2009) fnd that areas with high latent demand, as measured by previous loan
denials, had bigger house price run-ups and, subsequently, more defaults. They
conclude that securitization was the problem, although they also question why
securitization had not been a problem before. Purnanandam (2010) discusses how the
“originate-to-distribute” model associated with securitization gives the seller the
incentive to reduce screening and pass on risk to the ultimate purchasers of mortgages.
He fnds higher defaults at banks that were selling loans. Keys et al. (2010) fnd that at a
credit score of 620, defaults on mortgages jumped. As this is a “rule of thumb” cut-off
point for securitizers to make a loan, they suggest that this is evidence that securitizers
allowed excessively risky loans. Nadauld and Sherlund (2011) also blame securitization
for reduced screening. This paper contributes to this literature by examining whether
regulatory-driven demand by the GSEs caused securitizers to buy poor quality loans.
Buttimer (2011) argues that the GSEs had “bifurcated” regulation. They had
underwriting standards imposed by OFHEO that prevented them from purchasing
subprime loans directly fromoriginators. However, due to the pressure by HUDto make
loans to underserved areas and low-income borrowers, they purchased securitized
subprime loans. Our paper investigates these arguments.
Others have pointed to fraud. Ben-David (2011) fnds that seller down-payment
schemes contributed to the increase in house prices. Garmaise (2015) fnds that the
frequency of reported asset values on mortgage applications are clustered just above
multiples of $100,000, with fewer reported assets just below multiples of $100,000.
Piskorski et al. (2013) show that securitizers misrepresented crucial information about
the underlying mortgages they bundled. Mian and Suf (2015) fnd that in areas where
reported income on mortgage applications outstrips reported income as independently
reported to the internal revenue service (IRS), mortgage fraud is more frequent and there
are higher default rates. In each of these schemes, to some extent, the bank was in a
JFEP
7,4
356
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
position to perform further screening at a minimum or put a halt to the practice
altogether. This paper contributes to this strand of the literature by examining
Rebounds. One possibility that Rebounds can account for is that the originating lender
was in possession of adverse (although not necessarily fraudulent) information about
the borrower.
Turning to the CRA and underserved area requirements of the GSEs, Bostic et al.
(2005) fnd that banks that had more CRA lending were more likely to subsequently
acquire another lender. They fnd that the effect was biggest for large banks which, they
hypothesized, were more likely to get public scrutiny. Agarwal et al. (2012) fnd that
around bank CRA examination dates, mortgage lending is elevated. Manchester (2008)
and Frame (2008) fnd that the GSEs purchased securitized mortgages at a higher
frequency, when the mortgages satisfed affordable housing goals, indicating that the
GSEs were using the securitization market to meet underserved area requirements.
Ghent et al. (2015) use a regression discontinuity approach to determine if lenders
increased subprime originations or altered the pricing of loans to satisfy GSEaffordable
housing mandates or CRArequirements. They do not fnd evidence that these mandates
had an effect. DiLellio and Forsyth (2014) fnd that clusters of reported income at certain
levels (suggesting purposefully picked income numbers) are higher for incomes just
above the jumbo loan cut-off point than below, where the greater documentation
requirements of the GSEs are hypothesized to suppress income falsifcation. Jumbo
loans were too big to legally qualify for purchase by the GSEs. This paper contributes to
this discussion by looking at whether these regulatory requirements affected the
incidence of Rebounds.
DiLellio and Forsyth (2015) also make use of the Rebound measure. They fnd, not
surprisingly, that in 2001, reported income on a Rebound loan is lower than for a typical
loan at a lender, that was not a Rebound. However, during the house price run-up, they
fnd that these Rebound applications quickly had reported income that was signifcantly
higher than reported income on applications that had not been denied elsewhere, which
they suggest indicates increased income falsifcation.
The data
Data were collected on mortgage applications, reported as part of the Home Mortgage
Disclosure Act (HMDA). The year 2006 was selected, which was when the mortgage
market was at its peak. These data cover most mortgage applications in the USA, as
most residential mortgage lenders, including fnancial institutions that are not banks,
are legally required to report. The data were restricted to applications for home
purchases only, eliminating rental or vacation properties. All property types were kept,
including manufactured housing, but excluded multifamily properties. Only mortgages
secured by a frst or subordinate lien were kept. Lastly, observations were eliminated
where the purpose was not the origination of a loan, such as a sale of a mortgage from
one institution to another or a pre-approval request.
The regulator for a lender was included in the HMDA data. The National Credit
Union Administration (NCUA) was excluded as the null regulator, leaving the Offce of
the Comptroller of the Currency (OCC), Federal Reserve System (FRS), Federal Deposit
Insurance Corporation (FDIC), Offce of Thrift Supervision (OTS) and HUD. The FDIC
was listed as the regulator for institutions that had insured deposits and that were not
regulated by the NCUA, OCC, FRS or OTS. HUD is a “blanket” designation and was
357
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
listed for institutions, such as mortgage brokers, that were not regulated by one of the
other regulators or insured by the FDIC. These institutions were typically regulated at
the state level.
The HMDA data had gender and demographic information of applicants and
co-applicants. Variables indicating the state, metropolitan statistical area (MSA) and
census tract of the property were available. Loans were grouped outside an MSA in the
same state into a “rural” region. Borrower income was reported, and the natural log was
used. All continuous variables were winsorized in the HMDA data by 1 per cent to
reduce the undue infuence of outliers.
IRS data for the 50 US states and District of Columbia were obtained. These data
reported average adjusted gross income at the ZIP code level based on IRS forms 1040,
1040Aand 1040 EZ. The IRS data are by zip code, but the census tract is reported in the
HMDA data. To merge the IRS data with the HMDA data, a mapping was developed
between ZIP codes and state-county-census tracts.
An underserved area indicator was constructed for each unique census tract, from
data obtained from Freddie Mac. Annual unemployment rates were obtained from the
US Bureau of Labor Statistics at the county level. A jumbo loan indicator was
constructed based on the mortgage amount.
The Rebound measure indicates that a loan application appears to have been
previously denied by another lender for the same property. To identify Rebound
applications, applications with a matching denied application at a different lender in the
same year were identifed, based on the same census tract, loan amount, race, ethnicity
and gender of the applicant and co-applicant. In some instances, there was more than
one matching loan application or previously denied application. The matching
applications (accepted and denied) were then numbered. If an application had a larger
number than the maximum number for the denied applications, it was not coded as a
Rebound. A denied loan could be matched more than one time for applications at
different lenders, as applicants could apply at multiple lenders. Once Rebounds were
identifed, denied applications were removed from the dataset, and originated loans
focused on.
To measure expansion by lenders, the natural log of the change in the number of
census tracts from the previous year was taken. If the change was not positive, the
variable was set equal to zero and included a binary indicator that the lender did not
expand. The natural log of the total accepted applications of the lender was taken, as a
measure of size. An indicator was created for whether borrower reported income on the
application was in the lowest 25th percentile of reported income, to capture loans that
were more likely to satisfy CRArequirements (Bhutta and Canner, 2013 who also broke
out borrowers by percentile). Finally, the indicator that the loan was in the lowest 25 per
cen of reported income was interacted with the natural log of the increase in lender
census tracts. Only lenders that were expanding, and therefore concerned about the
CRA, would be expected to have an increased incentive to make loans to low-income
borrowers.
Empirical results
Table II provides summary statistics for Rebound. As can be seen for all mortgages
originated, there are 5,296,951 observations. Rebound applications represent 6.28 per
cent of the sample. There are 374,170 observations that are jumbo loans, which is 7.06
JFEP
7,4
358
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
per cent of the entire sample. Of all jumbo loans, Rebounds are 11.55 per cent. One
explanation for this result is that the GSEs were preventing lenders fromselling to them
mortgage applications that had been disapproved by a previous lender, as the incidence
of Rebounds is lower for the overall sample than for jumbos, which are ineligible for
purchase by the GSEs. The GSEs have extensive loan eligibility and documentation
requirements, which seemto have had at least some effect in reducing the acceptance of
problematic loans. It is also possible that there are more initial denials of jumbo loans
because of uneven standards across lenders, leading to a higher frequency that later get
accepted.
However, the results for underserved areas suggest a possible back-channel for the
acquisition of loans by the GSEs. There are 2,066,510 observations for underserved
areas, representing 39.01 per cent of the sample. Rebounds are 7.91 per cent of these
observations. The effect of underserved areas appears to be to increase the incidence of
loan disagreements between lenders. As the previous results with jumbo loans provide
evidence that the approval requirements of the GSEs suppressed the origination of these
loans, a back-channel transmission of these loans is suggested – securitizers bought
these loans and sold them to the GSEs. These results support Buttimer (2011) and are
consistent with the fndings of Manchester (2008) and Frame (2008) that the GSEs
purchased a disproportionate amount of securitized loans from underserved areas.
There are 1,848,249 mortgage originations by lenders that failed to expand, which is
34.89 per cent of the sample. These lenders were presumably unconcerned about
satisfying CRA requirements to merge or expand into new areas. Rebound mortgages
account for 5.86 per cent of mortgage loans by these lenders, which is less than for the
entire sample. One explanation of this result is that the absence of regulatory pressure
from the CRA resulted in lower lender disagreement about loans.
Finally, there are 1,236,282 mortgage applications in the lowest 25th reported income
percentile of all applications. This represents 23.34 per cent of accepted applications,
rather than 25 per cent. This is not surprising because it is expected that lower-income
applications will get rejected at a higher rate. Interestingly, Rebounds are 5.60 per cent
of these observations, which is lower than for the overall sample. The tendency of
lenders is to reject loans of low-income applicants that were previously rejected by a
different lender. However, in regulatory underserved areas, whose designation is
partially based on low-income, Rebounds are higher.
Summary statistics are presented in Table III for variables used in our subsequent
regression analysis, except for state indicators and borrower demographics. Looking at
variables that were not already presented in Table II, the mean of the natural log of the
Table II.
Mean rebounds by
loan characteristics
Variable No. of observations Mean rebounds
All mortgages 5,296,951 0.0628
Jumbo mortgage indicator (JM) 374,170 0.1155
Underserved area indicator (USR) 2,066,510 0.0791
Non-expansion indicator (NE) 1,848,249 0.0586
Indicator for low 25 application income (LOWI) 1,236,282 0.0560
Notes: The sample is for mortgage loans in 2006, reported in the HMDA data; a rebound loan was
denied by a different lender
359
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
increase in number of census tracts for a lender is ?0.1472, indicating that the typical
lender that expanded in 2006 had a slightly positive increase in the number of census
tracts covered, but averaged less than one census tract. The mean of the natural log of
total accepted mortgage applications for a lender is 10.0153, which corresponds to
22,366 applications. The mean of the natural log of reported income for the borrower is
4.3366, which corresponds to reported income of $76,447.
Ordinary least squares (OLS) regressions are presented in Table IV, where Rebound
is the dependent variable. Because of the state indicators and other binary indicators,
there are a large number of fxed effects. Therefore, to avoid an incidental parameters
problemin non-linear maximumlikelihood estimation, OLS was used (Abrevaya, 1997).
Borrower demographics and state indicators were included but not reported.
Huber-White standard errors were used, as recommended by Petersen (2009), to control
for the possible presence of missing lender-region specifc explanatory variables.
The jumbo loan coeffcient (JM) is 0.03872, and is signifcant at the 1 per cent level.
This implies that the underwriting requirements of the GSEs suppressed the origination
of loans for the purpose of resale to the GSEs, as the jumbo loan cutoff delineates which
loans are eligible for sale directly to the GSEs. This effect increases the incidence of
Rebounds by 61.7 per cent compared to the mean of Rebound loans.
However, the coeffcient for the underserved area indicator (USR) is positive and
signifcant at the 1 per cent level. If a loan was for a property in an underserved area, the
likelihood of a Rebound rises by 0.01750, which represents 27.9 per cent of the mean
number of Rebounds.
The results for the underserved area indicator support Buttimer (2011), suggesting
that a back-channel mechanism was in place through which the GSEs acquired
mortgages from underserved areas. As shall be seen, income (in addition to race) is
controlled for. Therefore, the underserved area indicator is expected to pick up the effect
Table III.
Sample summary
statistics
Variable Mean SD
Rebound 0.0628 0.2427
Jumbo mortgage indicator (JM) 0.0706 0.2562
Underserved area indicator (USR) 0.3901 0.4878
Ln (Number increase in census tracts) (LCTs) ?0.1472 1.7407
Non-expansion indicator (NE) 0.3489 0.4766
Ln (total accepted mortgage applications) (LTOT) 10.0153 2.3298
Ln (reported income [$k]) (LI) 4.3366 0.6156
Indicator for low 25 application income (LOWI) 0.2334 0.4230
Low 25 ?Ln (Number increase in census tracts) (LOW?CT) ?0.0599 0.9267
Average AGI ($k) (AVGAGI) 60.6531 51.1392
Unemployment (%) (UE) 4.5095 1.2314
OCC 0.2920 0.4547
FRS 0.1367 0.3435
FDIC 0.0712 0.2571
OTS 0.1128 0.3163
HUD 0.3665 0.4818
Notes: The sample had 5,296,951 mortgage loan observations in 2006; a rebound loan was denied by
a different lender
JFEP
7,4
360
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
of the regulatory requirement that relates specifcally to underserved areas. It was only
the GSEs that were subject to this requirement. Given that the jumbo loan coeffcient
implied that the underwriting requirements of the GSEs suppressed their ability to
purchase Rebound loans directly, and given that the GSEs were allowed to purchase
securitized mortgages to fulfll the underserved area requirement, these results are
suggestive that the securitization market served as a back-channel that funneled
underserved area loans to the GSEs.
The natural log of the increase in census tracts by a lender (LCTs) is insignifcant.
Expansion is not related to a willingness to accept problematic loans. There are two
potential opposing effects. When a lender wishes to expand, they may be more willing to
accept marginal loans to establish a presence in a new area. However, a lender may be
more cautious about lending in a new area because they do not know local conditions.
These effects seem to cancel each other out, although as shall be seen below, expansion
still plays a role, when in conjunction with regulatory requirements.
The non-expansion indicator (NE) coeffcient is positive and signifcant at the 1 per
cent level. Lenders that are not expanding (or conversely contracting) accept mortgage
applications that had been rejected by a previous lender with more frequency. The
natural log of total accepted mortgage applications has a coeffcient of 0.00536 and is
signifcant at the 1 per cent level. Larger lenders accept previously rejected loans with
more frequency. While several explanations are plausible, it is possible that LTOT is
picking up a “too big to fail” effect, which deserves further investigation, or that larger
banks are more subject to scrutiny under the CRA, corresponding to Bostic et al. (2005).
Table IV.
Predictors of rebound
Explanatory variable
Intercept ?0.11125 (?16.35)***
Jumbo mortgage indicator (JM) 0.03872 (14.77)***
Underserved area indicator (USR) 0.01750 (21.86)***
Ln (#increase in census tracts) (LCTs) ?0.00016 (?0.29)
Non-expansion indicator (NE) 0.00368 (2.94)***
Ln (total accepted mortgage applications) (LTOT) 0.00536 (13.68)***
Ln (reported income in $k) (LI) 0.01096 (14.90)***
Indicator for low 25 application income (LOWI) 0.00108 (1.35)
Low 25 ?Ln(Number increase in census tracts) (LOW?CT) 0.00128 (2.89)***
Average AGI ($M) (AVGAGI) 0.06821 (9.01)***
Unemployment (%) (UE) ?0.00002 (?0.04)
OCC ?0.02003 (?10.01)***
FRS 0.02875 (7.80)***
FDIC 0.01207 (6.01)***
OTS 0.00128 (0.60)
HUD 0.00380 (2.41)**
Adjusted R
2
0.0366
Notes: This is an OLS regression with Rebound as the dependent variable; a rebound loan was denied
by a different lender; there are 5,296,951 observations of accepted mortgages for 2006; borrower
demographics and state indicators are included but are not reported; cluster t-statistics are in
parentheses; ***and **denotes signifcance at the 1 and 5% levels, respectively
361
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The natural log of the reported income of the mortgage applicant is positive and
signifcant at the 1 per cent level. The higher the applicant income, the more frequently
a lender accepts applications that had been denied elsewhere. As some lenders were
found to be complicit in income falsifcation with borrowers, this effect may come from
income falsifcation by either the borrower or lender. Turning to LOWI, once borrower
income is included, the indicator that applicant income is in the lowest 25 per cent of
applicants is not signifcant.
However, when the effect of the CRA is included, the results are quite different for
low-income borrowers. Low ?CT was created from interacting LOWI with LCTs. For
mortgage applications in the lowest 25th income percentile only, it equals the natural log
of the increase in the number of census tracts of a lender. This variable is intended to
capture the effect of the CRA on lenders that wished to expand. These lenders were
required to show that they were serving community needs, including originating
mortgage loans to low-income borrowers.
The coeffcient for LOW ? CT is positive and signifcant at the 1 per cent level,
indicating that when a lender is expanding and the applicant has low income, the
frequency of accepted loans that had been denied elsewhere is higher. One explanation
is that the CRA was infuencing lenders to accept loans and contrasts sharply with the
results for LOWI and LCTs taken alone, which are both insignifcant. Without infuence
fromthe CRA, this indicates that the presence of a low-income borrower alone, or the fact
that a lender is expanding alone, do not have an effect. However, when a lender is
expanding, then the presence of a low-income borrower is signifcant.
The coeffcient is 0.00128. FromTable III, the standard deviation for LCTs is 1.7407.
Multiplying this by 0.00128, the effect is to increase the incidence of Rebound by 0.0022
for low-income borrowers, given a one standard deviation increase in LCTs. This
represents a 3.5 per cent increase in the incidence of Rebounds at the mean level of
Rebound. This effect is signifcant, but smaller than the effect of USR. However, it
should be expected that LCTs is a noisy measure in capturing the desire of a lender to
expand (unlike USR which is a more precise measure) and therefore is expected to be
biased downward.
Lenders in higher income areas more frequently accepted mortgage applications that
had been denied by a different lender. The coeffcient for AVGAGI is positive and
signifcant at the 1 per cent level. However, the coeffcient for unemployment is not
signifcant. This may refect that unemployment was generally lowin 2006, while house
prices had been undergoing large increases at the same time, which could have created
the false impression with lenders that there would be suffcient equity if the borrower
defaulted.
Turning to the regulatory indicators, the coeffcient for the Federal Reserve is the
highest regulatory coeffcient and is 0.02875. It is signifcant at the 1 per cent level. This
implies that the Federal Reserve may have been lax, thereby allowing for more
disagreement among banks about loans. It should be noted that as the sample covers one
year, it cannot be determined if the effect changed over time, indicating that regulatory
policy changed. However, the Federal Reserve regulates bank holding companies, and
these banks accepted a higher frequency of Rebound loans, all else constant. At the
mean level of Rebounds, the effect is to increase Rebounds by 45.8 per cent. This effect
could be due to the fact that for structural reasons, bank holding companies were more
accepting of risky loans. Another possibility is that the Federal Reserve was especially
JFEP
7,4
362
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
lax in monitoring the mortgage lending practices of its banks. This would be consistent
with Alan Greenspan’s remarks.
The FDIC has the second highest coeffcient. The FDIC indicator corresponds to
banks that were not regulated by a national regulator, yet received insured deposits.
Perhaps this coeffcient indicates that the FDICwas overstretched in overseeing lenders.
The OCC has the lowest coeffcient.
Conclusion
The Rebound variable is used to examine the frequency with which lenders were
accepting loans that other lenders had denied. A Rebound is a mortgage previously
denied by a different lender. As the lenders are disagreeing on the loan, it can potentially
indicate that some lenders were responding to regulatory pressure of the relaxation of
regulatory oversight.
It is found that jumbo loans, which were ineligible for sale to the GSEs, had more
Rebounds than other mortgages. The effect of a jumbo loan was to increase the incidence
of rebounds by 61.7 per cent compared to the mean of Rebound loans. One explanation
of this result is that the underwriting requirements of the GSEs prevented them from
directly purchasing low credit-quality mortgages from lenders.
However, the results are very different when looking at underserved areas. If a loan
was for a property in an underserved area, the likelihood of a Rebound rises by 27.9 per
cent in comparison to the mean number of Rebounds. The GSEs were increasingly
required to make loans to underserved areas, with low-income borrowers. There are
signifcantly more rebound loans to these areas, even after the reported income on the
mortgage and area income are included. As there is evidence that the GSEs were
constrained in which mortgages they could purchase directly fromlenders, these results
lend support for an alternative mechanism. The GSEs appeared to satisfy their
underserved area requirement by purchasing securitized mortgages. Though they could
not purchase low credit-quality loans directly, they used securitizers as middlemen to
purchase these loans. This may explain why the securitization market mushroomed
with poor credit-quality mortgages during the house price run-up, when it had not done
so before.
A smaller effect is found indicating that the CRA may have contributed to the house
price run-up. For those lenders that were expanding, and therefore potentially concerned
with satisfying the community needs requirements of the CRA, there is a higher
incidence of Rebounds when a borrower is in the bottom 25 per cent in income. This
effect is signifcant, but is of a lesser magnitude than for the underserved area
requirements. A one standard deviation increase in the explanatory variable for
low-income borrowers leads to a 3.5 per cent increase in the incidence of Rebounds at the
mean level of Rebound.
Finally, the Federal Reserve has the largest effect of lender regulators. The effect of
the Federal Reserve is to increase Rebounds by 45.8 per cent compared to the mean level
of Rebounds. It is possible that bank holding companies were more accepting of risky
loans for structural reasons. However, the other possibility is that the Federal Reserve
was especially lax in monitoring the mortgage lending practices of its banks. This
would be consistent with Alan Greenspan’s mea culpa that he did not monitor the
mortgage lending practices of banks carefully. The FDIC has the second largest
regulatory coeffcient, and the OCC has the smallest.
363
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
While this paper does not dispute that a variety of factors came into play to create the
house price run-up in the USA, this paper focuses on legislative and regulatory triggers.
It is found that several were associated with Rebounds and, therefore, potentially caused
an excess supply of credit that contributed to the house price run-up.
Note
1. See http://economix.blogs.nytimes.com/2008/10/23/greenspans-mea-culpa/ (accessed April
2013).
References
Abrevaya, J. (1997), “The equivalence of two estimators of the fxed effects logit model”,
Economics Letters, Vol. 55 No. 1, pp. 41-44.
Agarwal, S., Benmelech, E., Bergman, N. and Seru, A. (2012), “Did the Community Reinvestment
Act CRA) lead to risky lending?”, Kreisman Working Paper Series in Housing Law and
Policy, available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id?2172549
Barth, J. (2009), The Rise and Fall of the US Mortgage and Credit Markets, A Comprehensive
Analysis of the Market Meltdown, John Wiley & Sons, Hoboken, NJ.
Ben-David, I. (2011), “Financial constraints and infated home prices during the real-estate boom”,
American Economic Journal: Applied Economics, Vol. 3 No. 3, pp. 55-78.
Bhutta, N. and Canner, G. (2013), “Mortgage market conditions and borrower outcomes: evidence
from the 2012 HMDA data and matched HMDA – credit record data”, Federal Reserve
Bulletin, Vol. 99 No. 4, pp. 1-58.
Bostic, R.W., Paulson, A., Mehran, H. and Saidenberg, M. (2005), “Regulatory incentives and
consolidation: the case of commercial bank mergers and the community reinvestment act”,
Advances in Economic Analysis & Policy, Vol. 5 No. 1.
Buttimer, R. (2011), “The fnancial crisis: imperfect markets and imperfect regulation”, Journal of
Financial Economic Policy, Vol. 3 No. 1, pp. 12-32.
DiLellio, J. and Forsyth, J. (2014), “Government-sponsored enterprises and income falsifcation on
mortgage applications”, International Journal of Business, Accounting, and Finance, Vol. 8
No. 1, pp. 34-48.
DiLellio, J. and Forsyth, J. (2015), “Income falsifcation on mortgage applications during the
housing bubble”, Working Paper, Pepperdine University.
Frame, W. (2008), “The 2008 federal intervention to stabilize Fannie Mae and Freddie Mac”,
Journal of Applied Finance, Vol. 18 No. 2, pp. 124-136.
Garmaise, M. (2015), “Borrower misreporting and loan performance”, Journal of Finance, Vol. 70
No. 1, pp. 449-484.
Ghent, A., Hernandez-Murillo, R. and Owyang, M. (2015), “Did affordable housing legislation
contribute to the subprime securities boom?”, Real Estate Economics, forthcoming.
Keys, B., Mukherjee, T., Seru, A. and Vig, V. (2010), “Did securitization lead to lax screening?
Evidence fromsubprime loans”, Quarterly Journal of Economics, Vol. 125 No. 1, pp. 307-362.
Manchester, P. (2008), “Goal performance and characteristics of mortgages purchased by Fannie
Mae and Freddie Mac, 2001-2005”, Working Paper No. HF-017, US Department of Housing
and Urban Development.
Mian, A. and Suf, A. (2009), “The consequences of mortgage credit expansion: evidence from the
US mortgage default crisis”, The Quarterly Journal of Economics, Vol. 124 No. 4,
pp. 1449-1496.
JFEP
7,4
364
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Mian, A. and Suf, A. (2015), “Fraudulent income overstatement on mortgage applications during
the credit expansion of 2002 to 2005”, NBER Working Paper Series No. 20947, available at:
www.nber.org/papers/w20947
Nadauld, T. and Sherlund, S. (2011), “The impact of securitization on the expansion of subprime
credit”, Review of Radical Political Economics, available at: http://papers.ssrn.com/sol3/
papers.cfm?abstract_id?1410264
Petersen, M. (2009), “Estimating standard errors in fnance panel data sets: comparing
approaches”, Review of Financial Studies, Vol. 22 No. 1, pp. 435-480.
Piskorski, T., Seru, A. and Witkin, J. (2013), “Asset quality misrepresentation by fnancial
intermediaries: evidence from RMBS market”, Social Science Research Network, available
at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id?2215422
Purnanandam, A. (2010), “Originate-to-distribute model and the subprime mortgage crisis”, The
Review of Finance and Statistics, Vol. 24 No. 6, pp. 1881-1915.
Corresponding author
Joetta Forsyth can be contacted at: [email protected]
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
365
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_155460090.pdf
The purpose of this paper is to examine whether the underserved area requirements for
Fannie Mae and Freddie Mac (the government-sponsored enterprises [GSEs]) and the community needs
requirements of the Community Reinvestment Act (CRA) contributed to the house price run-up in the
USA
Journal of Financial Economic Policy
Was there a regulatory approval market for mortgages?
Peggy Crawford J oetta Forsyth
Article information:
To cite this document:
Peggy Crawford J oetta Forsyth , (2015),"Was there a regulatory approval market for mortgages?",
J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 354 - 365
Permanent link to this document:
http://dx.doi.org/10.1108/J FEP-06-2015-0032
Downloaded on: 24 January 2016, At: 21:53 (PT)
References: this document contains references to 20 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 25 times since 2015*
Users who downloaded this article also downloaded:
Muhammad Umar, Gang Sun, (2015),"Country risk, stock prices, and the exchange rate of the
renminbi", J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 366-376 http://dx.doi.org/10.1108/
J FEP-11-2014-0073
Muhammad Ali Nasir, Mushtaq Ahmad, Ferhan Ahmad, J unjie Wu, (2015),"Financial and economic
stability as ‘two sides of a coin’: Non-crisis regime evidence from the UK based on VECM", J ournal of
Financial Economic Policy, Vol. 7 Iss 4 pp. 327-353 http://dx.doi.org/10.1108/J FEP-01-2015-0006
J ames R. Barth, Tong Li, Wen Shi, Pei Xu, (2015),"China’s shadow banking sector: beneficial or
harmful to economic growth?", J ournal of Financial Economic Policy, Vol. 7 Iss 4 pp. 421-445 http://
dx.doi.org/10.1108/J FEP-07-2015-0043
Access to this document was granted through an Emerald subscription provided by emerald-
srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
*Related content and download information correct at time of
download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Was there a regulatory approval
market for mortgages?
Peggy Crawford and Joetta Forsyth
Graziadio School of Business and Management, Pepperdine University,
Malibu, California, USA
Abstract
Purpose – The purpose of this paper is to examine whether the underserved area requirements for
Fannie Mae and Freddie Mac (the government-sponsored enterprises [GSEs]) and the community needs
requirements of the Community Reinvestment Act (CRA) contributed to the house price run-up in the
USA.
Design/methodology/approach – This paper predicts the incidence of “Rebounds”, which indicate
that a mortgage had been previously denied, to provide evidence on whether certain regulations caused
excessively risky mortgage originations. As a different lender rejected the loan given the interest rate
that they were willing to charge and information on the borrower, a higher incidence of Rebounds
provides evidence that lenders were more frequently disagreeing about loans. This can indicate
differences in regulatory pressure or oversight across lenders.
Findings – This paper provides evidence that the GSEs were purchasing fewer Rebounds directly
fromlenders. However, evidence suggests that indirectly, the securitization market served as a conduit
for Rebounds to the GSEs that needed to satisfy regulatory underserved area requirements. The
necessity of complying with the CRA was found to increase Rebounds. Among regulators, the Federal
Reserve was found to have been particularly associated with Rebounds.
Originality/value – The paper’s contribution comes from linking Rebounds to legislative and
regulatory infuences. This contributes to the literature on excess credit and fraud, as well as the effect
of underserved area requirements and the CRA. Also, this paper adds a newdimension to the literature
on securitization, by showing the infuence of regulation on the securitization of risky mortgages.
Keywords Mortgages, Government policy and regulation, Financial institutions and services,
Corruption, Financial meltdown, Housing supply and markets
Paper type Research paper
Introduction
There are a number of factors that could have contributed to the house price run-up and
subsequent fnancial crisis in the USA. This paper focuses on several legislative and
regulatory triggers which may have impacted the behavior of government-sponsored
enterprises (GSEs) and banks as potential contributors.
The GSEs were subject to increasingly stringent requirements to serve
“underserved areas” by the US Department of Housing and Urban Development
(HUD). These were areas with low-income and minority residents. Table I provides
the GSE underserved area requirement over time, as reported by Barth (2009).
However, the GSEs were also required by the Offce of Federal Housing Enterprise
Oversight (OFHEO) to have underwriting standards that prevented them from
The authors wish to thank the Graziadio School of Business and Management for funding from
Funds for Excellence and the Julian Virtue Award.
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1757-6385.htm
JFEP
7,4
354
Received1 June 2015
Revised6 July2015
Accepted6 July2015
Journal of Financial Economic
Policy
Vol. 7 No. 4, 2015
pp. 354-365
©Emerald Group Publishing Limited
1757-6385
DOI 10.1108/JFEP-06-2015-0032
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
directly purchasing loans of insuffcient credit quality – an issue with some
underserved area mortgages. In other words, the GSEs had conficting regulations
that both encouraged and prevented them from purchasing mortgages from
underserved areas.
A “work-around” was created that allowed the GSEs to use a middleman with
different underwriting standards. In fact, Buttimer (2011) argues that the GSEs and
other lenders were encouraged by regulation to purchase securitized mortgages. The
GSEs were not allowed to directly purchase low credit-quality mortgages from lenders,
but they were allowed to purchase themfromsecuritizers, acting as middlemen, who did
not have the same restrictions. The GSEs may have been willing to pay a premium to
securitizers because of the additional beneft of satisfying regulatory requirements,
encouraging the origination of low quality loans.
Another factor which may have contributed to the house price run-up was the
Community Reinvestment Act (CRA) of 1977, combined with a merger and expansion
wave among banks. The CRA required that banks help meet the credit needs of their
local community consistent with safe and sound operations. To enforce the Act, federal
regulatory agencies were instructed to examine banks for CRAcompliance and consider
this information when approving applications for mergers or acquisitions. Bank
expansion was still limited (primarily within a state) until the Riegle-Neal Interstate
Banking and Branching Effciency Act of 1994 eliminated restrictions on interstate
banking. In addition, it required that regulatory agencies consider the CRA rating of
the out-of-state bank when approving interstate branches. It is also possible that these
lenders bought loans from the securitization market to satisfy the CRA.
There are a number of reasons banks may have an incentive to take on excessive risk.
For instance, deposit insurance may create the incentive for a bank to take on excessive
risk by protecting depositors in case of bank default. Any time mortgages are sold-off or
insured by an outside party; there may be an incentive for the originator to reduce
screening of the loans. Securitization, purchases by the GSEs and guarantees by the
Federal Housing Administration and Veteran’s Administration may have provided
these. It should be noted that none of these issues were new to the house price run-up.
However, regulatory oversight of underwriting standards may have been loosened,
and this may have been especially problematic if other entities that had an incentive to
screen mortgages were unaware of the shift. If regulators had previously provided
strong oversight of mortgage underwriting practices, then purchasers of mortgages,
bank creditors and others may have come to rely on regulators to screen loans. If the
screening unexpectedly stopped, these entities may not have known to increase their
own screening to fll the void. In a speech after housing prices collapsed, Alan
Greenspan admitted that he did not have the Federal Reserve regulate the mortgage
practices of banks carefully because he believed that it was not in a bank’s self-interest
to make excessively risky loans[1].
Intricately tied up with the question of what caused the house price run-up is the
question of what lenders knew. If lenders were caught up with market forces or investor
Table I.
GSE underserved
area requirement
over time
Year Before 2001 2001-2004 2005 2006-2007 2008
GSE underserved area requirement (%) 24 31 37 38 39
355
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
exuberance, then it can be argued that they did not have reason to believe that the
housing market would collapse and felt their loans were appropriately compensated for
risk given available market information. On the other hand, if they were responding to
regulatory pressure to make poor-quality loans or, conversely, freed to make poor
quality loans from a lack of oversight, it can be argued that they were aware that the
interest rate that they charged did not fully compensate the suppliers of capital for the
risk of the loans.
This paper uses “Rebound” to examine this issue. Rebound refers to originated
mortgage loans that were previously denied by another lender. One reason these loans
may have been denied is that a different lender considered the loan to be insuffciently
compensated for risk given the interest rate that they were willing to charge and the
information available to them. Lenders may have different policies regarding the
amount of risk (and subsequent interest rate) that they are willing to bear. However, if
some, but not other lenders, are put under pressure to make certain loans or released
from oversight by their regulator, then more frequent “disagreements” across lenders
about loans can be the result. Rebound can therefore be used to investigate whether
some lenders were either subject to increased regulatory pressure to make certain loans,
released from regulatory oversight, or both.
Literature review
There are a number of papers that address excessive credit and securitization. Mian and
Suf (2009) fnd that areas with high latent demand, as measured by previous loan
denials, had bigger house price run-ups and, subsequently, more defaults. They
conclude that securitization was the problem, although they also question why
securitization had not been a problem before. Purnanandam (2010) discusses how the
“originate-to-distribute” model associated with securitization gives the seller the
incentive to reduce screening and pass on risk to the ultimate purchasers of mortgages.
He fnds higher defaults at banks that were selling loans. Keys et al. (2010) fnd that at a
credit score of 620, defaults on mortgages jumped. As this is a “rule of thumb” cut-off
point for securitizers to make a loan, they suggest that this is evidence that securitizers
allowed excessively risky loans. Nadauld and Sherlund (2011) also blame securitization
for reduced screening. This paper contributes to this literature by examining whether
regulatory-driven demand by the GSEs caused securitizers to buy poor quality loans.
Buttimer (2011) argues that the GSEs had “bifurcated” regulation. They had
underwriting standards imposed by OFHEO that prevented them from purchasing
subprime loans directly fromoriginators. However, due to the pressure by HUDto make
loans to underserved areas and low-income borrowers, they purchased securitized
subprime loans. Our paper investigates these arguments.
Others have pointed to fraud. Ben-David (2011) fnds that seller down-payment
schemes contributed to the increase in house prices. Garmaise (2015) fnds that the
frequency of reported asset values on mortgage applications are clustered just above
multiples of $100,000, with fewer reported assets just below multiples of $100,000.
Piskorski et al. (2013) show that securitizers misrepresented crucial information about
the underlying mortgages they bundled. Mian and Suf (2015) fnd that in areas where
reported income on mortgage applications outstrips reported income as independently
reported to the internal revenue service (IRS), mortgage fraud is more frequent and there
are higher default rates. In each of these schemes, to some extent, the bank was in a
JFEP
7,4
356
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
position to perform further screening at a minimum or put a halt to the practice
altogether. This paper contributes to this strand of the literature by examining
Rebounds. One possibility that Rebounds can account for is that the originating lender
was in possession of adverse (although not necessarily fraudulent) information about
the borrower.
Turning to the CRA and underserved area requirements of the GSEs, Bostic et al.
(2005) fnd that banks that had more CRA lending were more likely to subsequently
acquire another lender. They fnd that the effect was biggest for large banks which, they
hypothesized, were more likely to get public scrutiny. Agarwal et al. (2012) fnd that
around bank CRA examination dates, mortgage lending is elevated. Manchester (2008)
and Frame (2008) fnd that the GSEs purchased securitized mortgages at a higher
frequency, when the mortgages satisfed affordable housing goals, indicating that the
GSEs were using the securitization market to meet underserved area requirements.
Ghent et al. (2015) use a regression discontinuity approach to determine if lenders
increased subprime originations or altered the pricing of loans to satisfy GSEaffordable
housing mandates or CRArequirements. They do not fnd evidence that these mandates
had an effect. DiLellio and Forsyth (2014) fnd that clusters of reported income at certain
levels (suggesting purposefully picked income numbers) are higher for incomes just
above the jumbo loan cut-off point than below, where the greater documentation
requirements of the GSEs are hypothesized to suppress income falsifcation. Jumbo
loans were too big to legally qualify for purchase by the GSEs. This paper contributes to
this discussion by looking at whether these regulatory requirements affected the
incidence of Rebounds.
DiLellio and Forsyth (2015) also make use of the Rebound measure. They fnd, not
surprisingly, that in 2001, reported income on a Rebound loan is lower than for a typical
loan at a lender, that was not a Rebound. However, during the house price run-up, they
fnd that these Rebound applications quickly had reported income that was signifcantly
higher than reported income on applications that had not been denied elsewhere, which
they suggest indicates increased income falsifcation.
The data
Data were collected on mortgage applications, reported as part of the Home Mortgage
Disclosure Act (HMDA). The year 2006 was selected, which was when the mortgage
market was at its peak. These data cover most mortgage applications in the USA, as
most residential mortgage lenders, including fnancial institutions that are not banks,
are legally required to report. The data were restricted to applications for home
purchases only, eliminating rental or vacation properties. All property types were kept,
including manufactured housing, but excluded multifamily properties. Only mortgages
secured by a frst or subordinate lien were kept. Lastly, observations were eliminated
where the purpose was not the origination of a loan, such as a sale of a mortgage from
one institution to another or a pre-approval request.
The regulator for a lender was included in the HMDA data. The National Credit
Union Administration (NCUA) was excluded as the null regulator, leaving the Offce of
the Comptroller of the Currency (OCC), Federal Reserve System (FRS), Federal Deposit
Insurance Corporation (FDIC), Offce of Thrift Supervision (OTS) and HUD. The FDIC
was listed as the regulator for institutions that had insured deposits and that were not
regulated by the NCUA, OCC, FRS or OTS. HUD is a “blanket” designation and was
357
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
listed for institutions, such as mortgage brokers, that were not regulated by one of the
other regulators or insured by the FDIC. These institutions were typically regulated at
the state level.
The HMDA data had gender and demographic information of applicants and
co-applicants. Variables indicating the state, metropolitan statistical area (MSA) and
census tract of the property were available. Loans were grouped outside an MSA in the
same state into a “rural” region. Borrower income was reported, and the natural log was
used. All continuous variables were winsorized in the HMDA data by 1 per cent to
reduce the undue infuence of outliers.
IRS data for the 50 US states and District of Columbia were obtained. These data
reported average adjusted gross income at the ZIP code level based on IRS forms 1040,
1040Aand 1040 EZ. The IRS data are by zip code, but the census tract is reported in the
HMDA data. To merge the IRS data with the HMDA data, a mapping was developed
between ZIP codes and state-county-census tracts.
An underserved area indicator was constructed for each unique census tract, from
data obtained from Freddie Mac. Annual unemployment rates were obtained from the
US Bureau of Labor Statistics at the county level. A jumbo loan indicator was
constructed based on the mortgage amount.
The Rebound measure indicates that a loan application appears to have been
previously denied by another lender for the same property. To identify Rebound
applications, applications with a matching denied application at a different lender in the
same year were identifed, based on the same census tract, loan amount, race, ethnicity
and gender of the applicant and co-applicant. In some instances, there was more than
one matching loan application or previously denied application. The matching
applications (accepted and denied) were then numbered. If an application had a larger
number than the maximum number for the denied applications, it was not coded as a
Rebound. A denied loan could be matched more than one time for applications at
different lenders, as applicants could apply at multiple lenders. Once Rebounds were
identifed, denied applications were removed from the dataset, and originated loans
focused on.
To measure expansion by lenders, the natural log of the change in the number of
census tracts from the previous year was taken. If the change was not positive, the
variable was set equal to zero and included a binary indicator that the lender did not
expand. The natural log of the total accepted applications of the lender was taken, as a
measure of size. An indicator was created for whether borrower reported income on the
application was in the lowest 25th percentile of reported income, to capture loans that
were more likely to satisfy CRArequirements (Bhutta and Canner, 2013 who also broke
out borrowers by percentile). Finally, the indicator that the loan was in the lowest 25 per
cen of reported income was interacted with the natural log of the increase in lender
census tracts. Only lenders that were expanding, and therefore concerned about the
CRA, would be expected to have an increased incentive to make loans to low-income
borrowers.
Empirical results
Table II provides summary statistics for Rebound. As can be seen for all mortgages
originated, there are 5,296,951 observations. Rebound applications represent 6.28 per
cent of the sample. There are 374,170 observations that are jumbo loans, which is 7.06
JFEP
7,4
358
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
per cent of the entire sample. Of all jumbo loans, Rebounds are 11.55 per cent. One
explanation for this result is that the GSEs were preventing lenders fromselling to them
mortgage applications that had been disapproved by a previous lender, as the incidence
of Rebounds is lower for the overall sample than for jumbos, which are ineligible for
purchase by the GSEs. The GSEs have extensive loan eligibility and documentation
requirements, which seemto have had at least some effect in reducing the acceptance of
problematic loans. It is also possible that there are more initial denials of jumbo loans
because of uneven standards across lenders, leading to a higher frequency that later get
accepted.
However, the results for underserved areas suggest a possible back-channel for the
acquisition of loans by the GSEs. There are 2,066,510 observations for underserved
areas, representing 39.01 per cent of the sample. Rebounds are 7.91 per cent of these
observations. The effect of underserved areas appears to be to increase the incidence of
loan disagreements between lenders. As the previous results with jumbo loans provide
evidence that the approval requirements of the GSEs suppressed the origination of these
loans, a back-channel transmission of these loans is suggested – securitizers bought
these loans and sold them to the GSEs. These results support Buttimer (2011) and are
consistent with the fndings of Manchester (2008) and Frame (2008) that the GSEs
purchased a disproportionate amount of securitized loans from underserved areas.
There are 1,848,249 mortgage originations by lenders that failed to expand, which is
34.89 per cent of the sample. These lenders were presumably unconcerned about
satisfying CRA requirements to merge or expand into new areas. Rebound mortgages
account for 5.86 per cent of mortgage loans by these lenders, which is less than for the
entire sample. One explanation of this result is that the absence of regulatory pressure
from the CRA resulted in lower lender disagreement about loans.
Finally, there are 1,236,282 mortgage applications in the lowest 25th reported income
percentile of all applications. This represents 23.34 per cent of accepted applications,
rather than 25 per cent. This is not surprising because it is expected that lower-income
applications will get rejected at a higher rate. Interestingly, Rebounds are 5.60 per cent
of these observations, which is lower than for the overall sample. The tendency of
lenders is to reject loans of low-income applicants that were previously rejected by a
different lender. However, in regulatory underserved areas, whose designation is
partially based on low-income, Rebounds are higher.
Summary statistics are presented in Table III for variables used in our subsequent
regression analysis, except for state indicators and borrower demographics. Looking at
variables that were not already presented in Table II, the mean of the natural log of the
Table II.
Mean rebounds by
loan characteristics
Variable No. of observations Mean rebounds
All mortgages 5,296,951 0.0628
Jumbo mortgage indicator (JM) 374,170 0.1155
Underserved area indicator (USR) 2,066,510 0.0791
Non-expansion indicator (NE) 1,848,249 0.0586
Indicator for low 25 application income (LOWI) 1,236,282 0.0560
Notes: The sample is for mortgage loans in 2006, reported in the HMDA data; a rebound loan was
denied by a different lender
359
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
increase in number of census tracts for a lender is ?0.1472, indicating that the typical
lender that expanded in 2006 had a slightly positive increase in the number of census
tracts covered, but averaged less than one census tract. The mean of the natural log of
total accepted mortgage applications for a lender is 10.0153, which corresponds to
22,366 applications. The mean of the natural log of reported income for the borrower is
4.3366, which corresponds to reported income of $76,447.
Ordinary least squares (OLS) regressions are presented in Table IV, where Rebound
is the dependent variable. Because of the state indicators and other binary indicators,
there are a large number of fxed effects. Therefore, to avoid an incidental parameters
problemin non-linear maximumlikelihood estimation, OLS was used (Abrevaya, 1997).
Borrower demographics and state indicators were included but not reported.
Huber-White standard errors were used, as recommended by Petersen (2009), to control
for the possible presence of missing lender-region specifc explanatory variables.
The jumbo loan coeffcient (JM) is 0.03872, and is signifcant at the 1 per cent level.
This implies that the underwriting requirements of the GSEs suppressed the origination
of loans for the purpose of resale to the GSEs, as the jumbo loan cutoff delineates which
loans are eligible for sale directly to the GSEs. This effect increases the incidence of
Rebounds by 61.7 per cent compared to the mean of Rebound loans.
However, the coeffcient for the underserved area indicator (USR) is positive and
signifcant at the 1 per cent level. If a loan was for a property in an underserved area, the
likelihood of a Rebound rises by 0.01750, which represents 27.9 per cent of the mean
number of Rebounds.
The results for the underserved area indicator support Buttimer (2011), suggesting
that a back-channel mechanism was in place through which the GSEs acquired
mortgages from underserved areas. As shall be seen, income (in addition to race) is
controlled for. Therefore, the underserved area indicator is expected to pick up the effect
Table III.
Sample summary
statistics
Variable Mean SD
Rebound 0.0628 0.2427
Jumbo mortgage indicator (JM) 0.0706 0.2562
Underserved area indicator (USR) 0.3901 0.4878
Ln (Number increase in census tracts) (LCTs) ?0.1472 1.7407
Non-expansion indicator (NE) 0.3489 0.4766
Ln (total accepted mortgage applications) (LTOT) 10.0153 2.3298
Ln (reported income [$k]) (LI) 4.3366 0.6156
Indicator for low 25 application income (LOWI) 0.2334 0.4230
Low 25 ?Ln (Number increase in census tracts) (LOW?CT) ?0.0599 0.9267
Average AGI ($k) (AVGAGI) 60.6531 51.1392
Unemployment (%) (UE) 4.5095 1.2314
OCC 0.2920 0.4547
FRS 0.1367 0.3435
FDIC 0.0712 0.2571
OTS 0.1128 0.3163
HUD 0.3665 0.4818
Notes: The sample had 5,296,951 mortgage loan observations in 2006; a rebound loan was denied by
a different lender
JFEP
7,4
360
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
of the regulatory requirement that relates specifcally to underserved areas. It was only
the GSEs that were subject to this requirement. Given that the jumbo loan coeffcient
implied that the underwriting requirements of the GSEs suppressed their ability to
purchase Rebound loans directly, and given that the GSEs were allowed to purchase
securitized mortgages to fulfll the underserved area requirement, these results are
suggestive that the securitization market served as a back-channel that funneled
underserved area loans to the GSEs.
The natural log of the increase in census tracts by a lender (LCTs) is insignifcant.
Expansion is not related to a willingness to accept problematic loans. There are two
potential opposing effects. When a lender wishes to expand, they may be more willing to
accept marginal loans to establish a presence in a new area. However, a lender may be
more cautious about lending in a new area because they do not know local conditions.
These effects seem to cancel each other out, although as shall be seen below, expansion
still plays a role, when in conjunction with regulatory requirements.
The non-expansion indicator (NE) coeffcient is positive and signifcant at the 1 per
cent level. Lenders that are not expanding (or conversely contracting) accept mortgage
applications that had been rejected by a previous lender with more frequency. The
natural log of total accepted mortgage applications has a coeffcient of 0.00536 and is
signifcant at the 1 per cent level. Larger lenders accept previously rejected loans with
more frequency. While several explanations are plausible, it is possible that LTOT is
picking up a “too big to fail” effect, which deserves further investigation, or that larger
banks are more subject to scrutiny under the CRA, corresponding to Bostic et al. (2005).
Table IV.
Predictors of rebound
Explanatory variable
Intercept ?0.11125 (?16.35)***
Jumbo mortgage indicator (JM) 0.03872 (14.77)***
Underserved area indicator (USR) 0.01750 (21.86)***
Ln (#increase in census tracts) (LCTs) ?0.00016 (?0.29)
Non-expansion indicator (NE) 0.00368 (2.94)***
Ln (total accepted mortgage applications) (LTOT) 0.00536 (13.68)***
Ln (reported income in $k) (LI) 0.01096 (14.90)***
Indicator for low 25 application income (LOWI) 0.00108 (1.35)
Low 25 ?Ln(Number increase in census tracts) (LOW?CT) 0.00128 (2.89)***
Average AGI ($M) (AVGAGI) 0.06821 (9.01)***
Unemployment (%) (UE) ?0.00002 (?0.04)
OCC ?0.02003 (?10.01)***
FRS 0.02875 (7.80)***
FDIC 0.01207 (6.01)***
OTS 0.00128 (0.60)
HUD 0.00380 (2.41)**
Adjusted R
2
0.0366
Notes: This is an OLS regression with Rebound as the dependent variable; a rebound loan was denied
by a different lender; there are 5,296,951 observations of accepted mortgages for 2006; borrower
demographics and state indicators are included but are not reported; cluster t-statistics are in
parentheses; ***and **denotes signifcance at the 1 and 5% levels, respectively
361
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The natural log of the reported income of the mortgage applicant is positive and
signifcant at the 1 per cent level. The higher the applicant income, the more frequently
a lender accepts applications that had been denied elsewhere. As some lenders were
found to be complicit in income falsifcation with borrowers, this effect may come from
income falsifcation by either the borrower or lender. Turning to LOWI, once borrower
income is included, the indicator that applicant income is in the lowest 25 per cent of
applicants is not signifcant.
However, when the effect of the CRA is included, the results are quite different for
low-income borrowers. Low ?CT was created from interacting LOWI with LCTs. For
mortgage applications in the lowest 25th income percentile only, it equals the natural log
of the increase in the number of census tracts of a lender. This variable is intended to
capture the effect of the CRA on lenders that wished to expand. These lenders were
required to show that they were serving community needs, including originating
mortgage loans to low-income borrowers.
The coeffcient for LOW ? CT is positive and signifcant at the 1 per cent level,
indicating that when a lender is expanding and the applicant has low income, the
frequency of accepted loans that had been denied elsewhere is higher. One explanation
is that the CRA was infuencing lenders to accept loans and contrasts sharply with the
results for LOWI and LCTs taken alone, which are both insignifcant. Without infuence
fromthe CRA, this indicates that the presence of a low-income borrower alone, or the fact
that a lender is expanding alone, do not have an effect. However, when a lender is
expanding, then the presence of a low-income borrower is signifcant.
The coeffcient is 0.00128. FromTable III, the standard deviation for LCTs is 1.7407.
Multiplying this by 0.00128, the effect is to increase the incidence of Rebound by 0.0022
for low-income borrowers, given a one standard deviation increase in LCTs. This
represents a 3.5 per cent increase in the incidence of Rebounds at the mean level of
Rebound. This effect is signifcant, but smaller than the effect of USR. However, it
should be expected that LCTs is a noisy measure in capturing the desire of a lender to
expand (unlike USR which is a more precise measure) and therefore is expected to be
biased downward.
Lenders in higher income areas more frequently accepted mortgage applications that
had been denied by a different lender. The coeffcient for AVGAGI is positive and
signifcant at the 1 per cent level. However, the coeffcient for unemployment is not
signifcant. This may refect that unemployment was generally lowin 2006, while house
prices had been undergoing large increases at the same time, which could have created
the false impression with lenders that there would be suffcient equity if the borrower
defaulted.
Turning to the regulatory indicators, the coeffcient for the Federal Reserve is the
highest regulatory coeffcient and is 0.02875. It is signifcant at the 1 per cent level. This
implies that the Federal Reserve may have been lax, thereby allowing for more
disagreement among banks about loans. It should be noted that as the sample covers one
year, it cannot be determined if the effect changed over time, indicating that regulatory
policy changed. However, the Federal Reserve regulates bank holding companies, and
these banks accepted a higher frequency of Rebound loans, all else constant. At the
mean level of Rebounds, the effect is to increase Rebounds by 45.8 per cent. This effect
could be due to the fact that for structural reasons, bank holding companies were more
accepting of risky loans. Another possibility is that the Federal Reserve was especially
JFEP
7,4
362
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
lax in monitoring the mortgage lending practices of its banks. This would be consistent
with Alan Greenspan’s remarks.
The FDIC has the second highest coeffcient. The FDIC indicator corresponds to
banks that were not regulated by a national regulator, yet received insured deposits.
Perhaps this coeffcient indicates that the FDICwas overstretched in overseeing lenders.
The OCC has the lowest coeffcient.
Conclusion
The Rebound variable is used to examine the frequency with which lenders were
accepting loans that other lenders had denied. A Rebound is a mortgage previously
denied by a different lender. As the lenders are disagreeing on the loan, it can potentially
indicate that some lenders were responding to regulatory pressure of the relaxation of
regulatory oversight.
It is found that jumbo loans, which were ineligible for sale to the GSEs, had more
Rebounds than other mortgages. The effect of a jumbo loan was to increase the incidence
of rebounds by 61.7 per cent compared to the mean of Rebound loans. One explanation
of this result is that the underwriting requirements of the GSEs prevented them from
directly purchasing low credit-quality mortgages from lenders.
However, the results are very different when looking at underserved areas. If a loan
was for a property in an underserved area, the likelihood of a Rebound rises by 27.9 per
cent in comparison to the mean number of Rebounds. The GSEs were increasingly
required to make loans to underserved areas, with low-income borrowers. There are
signifcantly more rebound loans to these areas, even after the reported income on the
mortgage and area income are included. As there is evidence that the GSEs were
constrained in which mortgages they could purchase directly fromlenders, these results
lend support for an alternative mechanism. The GSEs appeared to satisfy their
underserved area requirement by purchasing securitized mortgages. Though they could
not purchase low credit-quality loans directly, they used securitizers as middlemen to
purchase these loans. This may explain why the securitization market mushroomed
with poor credit-quality mortgages during the house price run-up, when it had not done
so before.
A smaller effect is found indicating that the CRA may have contributed to the house
price run-up. For those lenders that were expanding, and therefore potentially concerned
with satisfying the community needs requirements of the CRA, there is a higher
incidence of Rebounds when a borrower is in the bottom 25 per cent in income. This
effect is signifcant, but is of a lesser magnitude than for the underserved area
requirements. A one standard deviation increase in the explanatory variable for
low-income borrowers leads to a 3.5 per cent increase in the incidence of Rebounds at the
mean level of Rebound.
Finally, the Federal Reserve has the largest effect of lender regulators. The effect of
the Federal Reserve is to increase Rebounds by 45.8 per cent compared to the mean level
of Rebounds. It is possible that bank holding companies were more accepting of risky
loans for structural reasons. However, the other possibility is that the Federal Reserve
was especially lax in monitoring the mortgage lending practices of its banks. This
would be consistent with Alan Greenspan’s mea culpa that he did not monitor the
mortgage lending practices of banks carefully. The FDIC has the second largest
regulatory coeffcient, and the OCC has the smallest.
363
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
While this paper does not dispute that a variety of factors came into play to create the
house price run-up in the USA, this paper focuses on legislative and regulatory triggers.
It is found that several were associated with Rebounds and, therefore, potentially caused
an excess supply of credit that contributed to the house price run-up.
Note
1. See http://economix.blogs.nytimes.com/2008/10/23/greenspans-mea-culpa/ (accessed April
2013).
References
Abrevaya, J. (1997), “The equivalence of two estimators of the fxed effects logit model”,
Economics Letters, Vol. 55 No. 1, pp. 41-44.
Agarwal, S., Benmelech, E., Bergman, N. and Seru, A. (2012), “Did the Community Reinvestment
Act CRA) lead to risky lending?”, Kreisman Working Paper Series in Housing Law and
Policy, available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id?2172549
Barth, J. (2009), The Rise and Fall of the US Mortgage and Credit Markets, A Comprehensive
Analysis of the Market Meltdown, John Wiley & Sons, Hoboken, NJ.
Ben-David, I. (2011), “Financial constraints and infated home prices during the real-estate boom”,
American Economic Journal: Applied Economics, Vol. 3 No. 3, pp. 55-78.
Bhutta, N. and Canner, G. (2013), “Mortgage market conditions and borrower outcomes: evidence
from the 2012 HMDA data and matched HMDA – credit record data”, Federal Reserve
Bulletin, Vol. 99 No. 4, pp. 1-58.
Bostic, R.W., Paulson, A., Mehran, H. and Saidenberg, M. (2005), “Regulatory incentives and
consolidation: the case of commercial bank mergers and the community reinvestment act”,
Advances in Economic Analysis & Policy, Vol. 5 No. 1.
Buttimer, R. (2011), “The fnancial crisis: imperfect markets and imperfect regulation”, Journal of
Financial Economic Policy, Vol. 3 No. 1, pp. 12-32.
DiLellio, J. and Forsyth, J. (2014), “Government-sponsored enterprises and income falsifcation on
mortgage applications”, International Journal of Business, Accounting, and Finance, Vol. 8
No. 1, pp. 34-48.
DiLellio, J. and Forsyth, J. (2015), “Income falsifcation on mortgage applications during the
housing bubble”, Working Paper, Pepperdine University.
Frame, W. (2008), “The 2008 federal intervention to stabilize Fannie Mae and Freddie Mac”,
Journal of Applied Finance, Vol. 18 No. 2, pp. 124-136.
Garmaise, M. (2015), “Borrower misreporting and loan performance”, Journal of Finance, Vol. 70
No. 1, pp. 449-484.
Ghent, A., Hernandez-Murillo, R. and Owyang, M. (2015), “Did affordable housing legislation
contribute to the subprime securities boom?”, Real Estate Economics, forthcoming.
Keys, B., Mukherjee, T., Seru, A. and Vig, V. (2010), “Did securitization lead to lax screening?
Evidence fromsubprime loans”, Quarterly Journal of Economics, Vol. 125 No. 1, pp. 307-362.
Manchester, P. (2008), “Goal performance and characteristics of mortgages purchased by Fannie
Mae and Freddie Mac, 2001-2005”, Working Paper No. HF-017, US Department of Housing
and Urban Development.
Mian, A. and Suf, A. (2009), “The consequences of mortgage credit expansion: evidence from the
US mortgage default crisis”, The Quarterly Journal of Economics, Vol. 124 No. 4,
pp. 1449-1496.
JFEP
7,4
364
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Mian, A. and Suf, A. (2015), “Fraudulent income overstatement on mortgage applications during
the credit expansion of 2002 to 2005”, NBER Working Paper Series No. 20947, available at:
www.nber.org/papers/w20947
Nadauld, T. and Sherlund, S. (2011), “The impact of securitization on the expansion of subprime
credit”, Review of Radical Political Economics, available at: http://papers.ssrn.com/sol3/
papers.cfm?abstract_id?1410264
Petersen, M. (2009), “Estimating standard errors in fnance panel data sets: comparing
approaches”, Review of Financial Studies, Vol. 22 No. 1, pp. 435-480.
Piskorski, T., Seru, A. and Witkin, J. (2013), “Asset quality misrepresentation by fnancial
intermediaries: evidence from RMBS market”, Social Science Research Network, available
at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id?2215422
Purnanandam, A. (2010), “Originate-to-distribute model and the subprime mortgage crisis”, The
Review of Finance and Statistics, Vol. 24 No. 6, pp. 1881-1915.
Corresponding author
Joetta Forsyth can be contacted at: [email protected]
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
365
Regulatory
approval
market
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
3
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_155460090.pdf