The Evolution Of The Subprime Mortgage Market
The Evolution of the Subprime Mortgage Market
Souphala Chomsisengphet and Anthony Pennington-Cross
This paper describes subprime lending in the mortgage market and how it has evolved through
time. Subprime lending has introduced a substantial amount of risk-based pricing into the mortgage
market by creating a myriad of prices and product choices largely determined by borrower credit
history (mortgage and rental payments, foreclosures and bankruptcies, and overall credit scores)
and down payment requirements. Although subprime lending still differs from prime lending in
many ways, much of the growth (at least in the securitized portion of the market) has come in the
least-risky (A–) segment of the market. In addition, lenders have imposed prepayment penalties
to extend the duration of loans and required larger down payments to lower their credit risk
exposure from high-risk loans.
Federal Reserve Bank of St. Louis Review, January/February 2006, 88(1), pp. 31-56.
Of course, this expanded access comes with
INTRODUCTION AND MOTIVATION
a price: At its simplest, subprime lending can be
described as high-cost lending.
Borrower cost associated with subprime
Homeownership is one of the primary
ways that households can build wealth.
In fact, in 1995, the typical household
lending is driven primarily by two factors: credit
held no corporate equity (Tracy, Schneider, and
history and down payment requirements. This
Chan, 1999), implying that most households find
contrasts with the prime market, where borrower
it difficult to invest in anything but their home.
cost is primarily driven by the down payment
Because homeownership is such a significant
alone, given that minimum credit history require-
economic factor, a great deal of attention is paid
ments are satisfied.
to the mortgage market.
Because of its complicated nature, subprime
lending is simultaneously viewed as having great
Subprime lending is a relatively new and
promise and great peril. The promise of subprime
rapidly growing segment of the mortgage market
lending is that it can provide the opportunity for
that expands the pool of credit to borrowers who,
homeownership to those who were either subject
for a variety of reasons, would otherwise be denied
to discrimination or could not qualify for a mort-
credit. For instance, those potential borrowers who
gage in the past.1 In fact, subprime lending is most
would fail credit history requirements in the stan-
dard (prime) mortgage market have greater access
1
to credit in the subprime market. Two of the major
See Hillier (2003) for a thorough discussion of the practice of “redlin-
ing” and the lack of access to lending institutions in predominately
benefits of this type of lending, then, are the
minority areas. In fact, in the 1930s the Federal Housing Authority
increased numbers of homeowners and the oppor-
(FHA) explicitly referred to African Americans and other minority
groups as adverse influences. By the 1940s, the Justice Department
tunity for these homeowners to create wealth.
had filed criminal and civil antitrust suits to stop redlining.
Souphala Chomsisengphet is a financial economist at the Office of the Comptroller of the Currency. Anthony Pennington-Cross is a senior
economist at the Federal Reserve Bank of St. Louis. The views expressed here are those of the individual authors and do not necessarily
reflect the official positions of the Federal Reserve Bank of St. Louis, the Federal Reserve System, the Board of Governors, the Office of
Comptroller of the Currency, or other officers, agencies, or instrumentalities of the United States government.
© 2006, The Federal Reserve Bank of St. Louis. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in
their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made
only with prior written permission of the Federal Reserve Bank of St. Louis.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
3 1
Chomsisengphet and Pennington-Cross
prevalent in neighborhoods with high concentra-
subprime lending and how it has evolved, to aid
tions of minorities and weaker economic condi-
the growing literature on the subprime market
tions (Calem, Gillen, and Wachter, 2004, and
and related policy discussions. We use data from
Pennington-Cross, 2002). However, because poor
a variety of sources to study the subprime mort-
credit history is associated with substantially more
gage market: For example, we characterize the
delinquent payments and defaulted loans, the
market with detailed information on 7.2 million
interest rates for subprime loans are substantially
loans leased from a private data provider called
higher than those for prime loans.
LoanPerformance. With these data, we analyze
Preliminary evidence indicates that the
the development of subprime lending over the
probability of default is at least six times higher
past 10 years and describe what the subprime
for nonprime loans (loans with high interest rates)
market looks like today. We pay special attention
than prime loans. In addition, nonprime loans
to the role of credit scores, down payments, and
are less sensitive to interest rate changes and, as
prepayment penalties.
a result, subprime borrowers have a harder time
The results of our analysis indicate that the
taking advantage of available cheaper financing
subprime market has grown substantially over
(Pennington-Cross, 2003, and Capozza and
the past decade, but the path has not been smooth.
Thomson, 2005). The Mortgage Bankers Associa-
For instance, the market expanded rapidly until
tion of America (MBAA) reports that subprime
1998, then suffered a period of retrenchment, but
loans in the third quarter of 2002 had a delin-
currently seems to be expanding rapidly again,
quency rate 51/2 times higher than that for prime
especially in the least-risky segment of the sub-
loans (14.28 versus 2.54 percent) and the rate at
prime market (A– grade loans). Furthermore,
which foreclosures were begun for subprime loans
lenders of subprime loans have increased their
was more than 10 times that for prime loans (2.08
use of mechanisms such as prepayment penal-
versus 0.20 percent). Therefore, the propensity
ties and large down payments to, respectively,
of borrowers of subprime loans to fail as home-
increase the duration of loans and mitigate losses
owners (default on the mortgage) is much higher
from defaulted loans.
than for borrowers of prime loans.
This failure can lead to reduced access to
financial markets, foreclosure, and loss of any
WHAT MAKES A LOAN SUBPRIME?
equity and wealth achieved through mortgage
From the borrower’s perspective, the primary
payments and house price appreciation. In addi-
distinguishing feature between prime and sub-
tion, any concentration of foreclosed property can
prime loans is that the upfront and continuing
potentially adversely impact the value of property
costs are higher for subprime loans. Upfront costs
in the neighborhood as a whole.
include application fees, appraisal fees, and other
Traditionally, the mortgage market set mini-
fees associated with originating a mortgage. The
mum lending standards based on a borrower’s
continuing costs include mortgage insurance
income, payment history, down payment, and the
payments, principle and interest payments, late
local underwriter’s knowledge of the borrower.
fees and fines for delinquent payments, and fees
This approach can best be characterized as using
levied by a locality (such as property taxes and
nonprice credit rationing. However, the subprime
special assessments).
market has introduced many different pricing tiers
Very little data have been gathered on the
and product types, which has helped to move the
extent of upfront fees and how they differ from
mortgage market closer to price rationing, or risk-
prime fees. But, as shown by Fortowsky and
based pricing. The success of the subprime market
LaCour-Little (2002), many factors, including
will in part determine how fully the mortgage
borrower credit history and prepayment risk, can
market eventually incorporates pure price ration-
substantially affect the pricing of loans. Figure 1
ing (i.e., risk-based prices for each borrower).
compares interest rates for 30-year fixed-rate loans
This paper provides basic information about
in the prime and the subprime markets. The
3 2
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 1
Interest Rates
Interest Rate at Origination
12
Subprime
Subprime Premium
10
Prime
8
6
4
2
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
NOTE: Prime is the 30-year fixed interest rate reported by the Freddie Mac Primary Mortgage Market Survey. Subprime is the average
30-year fixed interest rate at origination as calculated from the LoanPerformance data set. The Subprime Premium is the difference
between the prime and subprime rates.
Figure 2
Foreclosures In Progress
Rate Normalized to 1 in 1998:Q1
5
LP-Subprime
MBAA-Subprime
4
MBAA-Prime
3
2
1
0
1998
1999
2000
2001
2002
2003
2004
NOTE: The rate of foreclosure in progress is normalized to 1 in the first quarter of 1998. MBAA indicates the source is the Mortgage
Bankers Association of America and LP indicates that the rate is calculated from the LoanPerformance ABS data set.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
3 3
Chomsisengphet and Pennington-Cross
Table 1
Underwriting and Loan Grades
Credit history
Premier Plus
Premier
A–
B
C
C–
Mortgage delinquency
0 x 30 x 12
1 x 30 x 12
2 x 30 x 12
1 x 60 x 12
1 x 90 x 12
2 x 90 x 12
in days
Foreclosures
>36 months
>36 months
>36 months
>24 months
>12 months
>1 day
Bankruptcy,
Chapter
7
Discharged Discharged Discharged Discharged Discharged Discharged
>36 months
>36 months
>36 months
>24 months
>12 months
Bankruptcy,
Chapter
13
Discharged Discharged Discharged Discharged
Filed
Pay
>24 months
>24 months
>24 months
>18 months
>12 months
Debt ratio
50%
50%
50%
50%
50%
50%
SOURCE: Countrywide, downloaded from www.cwbc.com on 2/11/05.
prime interest rate is collected from the Freddie
tial increases. For example, from the beginning
Mac Primary Mortgage Market Survey. The sub-
of the sample to their peaks, the MBAA meas-
prime interest rate is the average 30-year fixed-
ure increased nearly fourfold and the
rate at origination as calculated from the
LoanPerformance measure increased threefold.
LoanPerformance data set. The difference between
Both measures have been declining since 2003.
the two in each month is defined as the subprime
These results show that the performance and ter-
premium. The premium charged to a subprime
mination profiles for subprime loans are much
borrower is typically around 2 percentage points.
different from those for prime loans, and after
It increases a little when rates are higher and
the 2001 recession it took nearly two years for
decreases a little when rates are lower.
foreclosure rates to start declining in the sub-
From the lender’s perspective, the cost of a
prime market. It is also important to note that,
subprime loan is driven by the loan’s termination
after the recession, the labor market weakened
profile.2 The MBAA reports (through the MBAA
but the housing market continued to thrive (high
delinquency survey) that 4.48 percent of subprime
volume with steady and increasing prices). There-
and 0.42 percent of prime fixed-rate loans were
fore, there was little or no equity erosion caused
in foreclosure during the third quarter of 2004.
by price fluctuations during the recession. It
According to LoanPerformance data, 1.55 percent
remains to be seen how subprime loans would
of fixed-rate loans were in foreclosure during the
perform if house prices declined while unemploy-
same period. (See the following section “Evolution
ment rates increased.
of Subprime Lending” for more details on the
The rate sheets and underwriting matrices
differences between these two data sources.)
from Countrywide Home Loans, Inc. (download
Figure 2 depicts the prime and subprime loans
from www.cwbc.com on 2/11/05), a leading lender
in foreclosure from 1998 to 2004. For comparison,
and servicer of prime and subprime loans, provide
the rates are all normalized to 1 in the first quarter
some details typically used to determine what
of 1998 and only fixed-rate loans are included.
type of loan application meets subprime under-
The figure shows that foreclosures on prime
writing standards.
loans declined slightly from 1998 through the
Countrywide reports six levels, or loan
third quarter of 2004. In contrast, both measures
grades, in its B&C lending rate sheet: Premier Plus,
of subprime loan performance showed substan-
Premier, A–, B, C, and C–. The loan grade is deter-
mined by the applicant’s mortgage or rent payment
2
history, bankruptcies, and total debt-to-income
The termination profile determines the likelihood that the borrower
will either prepay or default on the loan.
ratio. Table 1 provides a summary of the four
3 4
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Table 2
Underwriting and Interest Rates
LTV
Loan grade
Credit score
60%
70%
80%
90%
100%
Premier Plus
680
5.65
5.75
5.80
5.90
7.50
660
5.65
5.75
5.85
6.00
7.85
600
5.75
5.80
5.90
6.60
8.40
580
5.75
5.85
6.00
6.90
8.40
500
6.40
6.75
7.90
Premier
680
5.80
5.90
5.95
5.95
7.55
660
5.80
5.90
6.00
6.05
7.90
600
5.90
5.95
6.05
6.65
8.45
580
5.90
6.00
6.15
6.95
500
6.55
6.90
8.05
A–
680
660
6.20
6.25
6.35
6.45
600
6.35
6.45
6.50
6.70
580
6.35
6.45
6.55
7.20
500
6.60
6.95
8.50
B
680
660
6.45
6.55
6.65
600
6.55
6.60
6.75
580
6.55
6.65
6.85
500
6.75
7.25
9.20
C
680
660
600
6.95
7.20
580
7.00
7.30
500
7.45
8.95
C–
680
660
600
580
7.40
7.90
500
8.10
9.80
NOTE: The first three years are at a fixed interest rate, and there is a three-year prepayment penalty.
SOURCE: Countrywide California B&C Rate Sheet, downloaded from www.cwbc.com on 2/11/05.
underwriting requirements used to determine
slowly relaxed for each loan grade: the Premier
the loan grade. For example, to qualify for the
grade allows one payment to be 30-days delin-
Premier Plus grade, the applicant may have had
quent; the A– grade allows two payments to be
no mortgage payment 30 days or more delinquent
30-days delinquent; the B grade allows one pay-
in the past year (0 x 30 x 12). The requirement is
ment to be 60-days delinquent; the C grade allows
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
3 5
Chomsisengphet and Pennington-Cross
one payment to be 90-days delinquent; and the
quoted is 9.8 percentage points for a C– grade loan
C– grade allows two payments to be 90-days
with the lowest credit score and a 30 percent down
delinquent. The requirements for foreclosures
payment.
are also reduced for the lower loan grades. For
The range of interest rates charged indicates
example, whereas the Premier Plus grade stipu-
that the subprime mortgage market actively price
lates no foreclosures in the past 36 months, the
discriminates (that is, it uses risk-based pricing)
C grade stipulates no foreclosures only in the past
on the basis of multiple factors: delinquent pay-
12 months, and the C– grade stipulates no active
ments, foreclosures, bankruptcies, debt ratios,
foreclosures. For most loan grades, Chapter 7 and
credit scores, and LTV ratios. In addition, stipu-
Chapter 13 bankruptcies typically must have been
lations are made that reflect risks associated with
discharged at least a year before application;
the loan grade and include any prepayment penal-
however, the lowest grade, C–, requires only that
ties, the length of the loan, the flexibility of the
Chapter 7 bankruptcies have been discharged
interest rate (adjustable, fixed, or hybrid), the lien
and Chapter 13 bankruptcies at least be in repay-
position, the property type, and other factors.
ment. However, all loan grades require at least a
The lower the grade or credit score, the
50 percent ratio between monthly debt servicing
larger the down payment requirement. This
costs (which includes all outstanding debts) and
requirement is imposed because loss severities
monthly income.
are strongly tied to the amount of equity in the
Loan grade alone does not determine the cost
home (Pennington-Cross, forthcoming) and price
of borrowing (that is, the interest rate on the loan).
appreciation patterns.
Table 2 provides a matrix of credit scores and
As shown in Table 2, not all combinations of
loan-to-value (LTV) ratio requirements that deter-
down payments and credit scores are available
mine pricing of the mortgage within each loan
to the applicant. For example, Countrywide does
grade for a 30-year loan with a 3-year fixed interest
not provide an interest rate for A– grade loans
rate and a 3-year prepayment penalty. For exam-
with no down payment (LTV = 100 percent).
ple, loans in the Premier Plus grade with credit
Therefore, an applicant qualifying for grade A–
scores above 680 and down payments of 40 per-
but having no down payment must be rejected.
cent or more would pay interest rates of 5.65
As a result, subprime lending rations credit
percentage points, according to the Countrywide
through a mixture of risk-based pricing (price
rate sheet for California. As the down payment
rationing) and minimum down payment require-
gets smaller (as LTV goes up), the interest rate
ments, given other risk characteristics (nonprice
increases. For example, an applicant with the
rationing).
same credit score and a 100 percent LTV will be
In summary, in its simplest form, what makes
charged a 7.50 interest rate. But, note that the
a loan subprime is the existence of a premium
interest rate is fairly stable until the down pay-
above the prevailing prime market rate that a
ment drops below 10 percent. At this point the
borrower must pay. In addition, this premium
lender begins to worry about possible negative
varies over time, which is based on the expected
equity positions in the near future due to appraisal
risks of borrower failure as a homeowner and
error or price depreciation.
default on the mortgage.
It is the combination of smaller down pay-
ments and lower credit scores that lead to the
A BRIEF HISTORY OF SUBPRIME
highest interest rates. In addition, applicants in
lower loan grades tend to pay higher interest rates
LENDING
than similar applicants in a higher loan grade.
It was not until the mid- to late 1990s that the
This extra charge reflects the marginal risk asso-
strong growth of the subprime mortgage market
ciated with missed mortgage payments, foreclo-
gained national attention. Immergluck and Wiles
sures, or bankruptcies in the past. The highest rate
(1999) reported that more than half of subprime
3 6
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Table 3
Total Originations—Consolidation and Growth
Total B&C
Top 25 B&C
Top 25
B&C
originations
originations
market
share Total market
share
Year
(billions)
(billions)
of B&C
originations
of total
1995
$65.0
$25.5
39.3%
$639.4
10.2%
1996
$96.8
$45.3
46.8%
$785.3
12.3%
1997
$124.5
$75.1
60.3%
$859.1
14.5%
1998
$150.0
$94.3
62.9%
$1,450.0
10.3%
1999
$160.0
$105.6
66.0%
$1,310.0
12.2%
2000
$138.0
$102.2
74.1%
$1,048.0
13.2%
2001
$173.3
$126.8
73.2%
$2,058.0
8.4%
2002
$213.0
$187.6
88.1%
$2,680.0
7.9%
2003
$332.0
$310.1
93.4%
$3,760.0
8.8%
SOURCE: Inside B&C Lending. Individual firm data are from Inside B&C Lending and are generally based on security issuance or
previously reported data.
refinances3 originated in predominately African-
securitization rates calculated as the ratio of the
American census tracts, whereas only one tenth
total number of dollars securitized divided by the
of prime refinances originated in predominately
number of dollars originated in each calendar year.
African-American census tracts. Nichols,
Therefore, this number roughly approximates
Pennington-Cross, and Yezer (2005) found that
the actual securitization rate, but could be under
credit-constrained borrowers with substantial
or over the actual rate due to the packaging of
wealth are most likely to finance the purchase of
seasoned loans.4 The subprime loan securitiza-
a home by using a subprime mortgage.
tion rate has grown from less than 30 percent in
The growth of subprime lending in the past
1995 to over 58 percent in 2003. The securitiza-
decade has been quite dramatic. Using data
tion rate for conventional and jumbo loans has
reported by the magazine Inside B&C Lending,
also increased over the same time period.5 For
Table 3 reports that total subprime or B&C origina-
example, conventional securitization rates have
tions (loans) have grown from $65 billion in 1995
increased from close to 50 percent in 1995-97 to
to $332 billion in 2003. Despite this dramatic
more than 75 percent in 2003. In addition, all or
growth, the market share for subprime loans
almost all of the loans insured by government
(referred to in the table as B&C) has dropped from
loans are securitized. Therefore, the subprime
a peak of 14.5 percent in 1997 to 8.8 percent in
mortgage market has become more similar to the
2003. During this period, homeowners refinanced
prime market over time. In fact, the 2003 securi-
existing mortgages in surges as interest rates
tization rate of subprime loans is comparable to
dropped. Because subprime loans tend to be less
that of prime loans in the mid-1990s.
responsive to changing interest rates (Pennington-
Cross, 2003), the subprime market share should
4 Seasoned loans refers to loans sold into securities after the date of
tend to drop during refinancing booms.
origination.
The financial markets have also increasingly
5 Conventional loans are loans that are eligible for purchase by
securitized subprime loans. Table 4 provides the
Fannie Mae and Freddie Mac because of loan size and include
loans purchased by Fannie Mae and Freddie Mac, as well as those
held in a portfolio or that are securitized through a private label.
3 A refinance is a new loan that replaces an existing loan, typically
Jumbo loans are loans with loan amounts above the government-
to take advantage of a lower interest rate on the mortgage.
sponsored enterprise (conventional conforming) loan limit.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
3 7
Chomsisengphet and Pennington-Cross
Table 4
Securitization Rates
Loan type
Year
FHA/VA
Conventional
Jumbo
Subprime
1995
101.1%
45.6%
23.9%
28.4%
1996
98.1%
52.5%
21.3%
39.5%
1997
100.7%
45.9%
32.1%
53.0%
1998
102.3%
62.2%
37.6%
55.1%
1999
88.1%
67.0%
30.1%
37.4%
2000
89.5%
55.6%
18.0%
40.5%
2001
102.5%
71.5%
31.4%
54.7%
2002
92.6%
72.8%
32.0%
57.6%
2003
94.9%
75.9%
35.1%
58.7%
NOTE: Subprime securities include both MBS and ABS backed by subprime loans. Securitization rate = securities issued divided by
originations in dollars.
SOURCE: Inside MBS & ABS.
Many factors have contributed to the growth
the value of their homes. In fact, slightly over one-
of subprime lending. Most fundamentally, it
half of subprime loan originations have been for
became legal. The ability to charge high rates
cash-out refinancing.7
and fees to borrowers was not possible until the
In addition to changes in the law, market
Depository Institutions Deregulation and Monetary
changes also contributed to the growth and mat-
Control Act (DIDMCA) was adopted in 1980. It
uration of subprime loans. In 1994, for example,
preempted state interest rate caps. The Alternative
interest rates increased and the volume of origi-
Mortgage Transaction Parity Act (AMTPA) in 1982
nations in the prime market dropped. Mortgage
permitted the use of variable interest rates and
brokers and mortgage companies responded by
balloon payments.
looking to the subprime market to maintain vol-
These laws opened the door for the develop-
ume. The growth through the mid-1990s was
ment of a subprime market, but subprime lending
funded by issuing mortgage-backed securities
would not become a viable large-scale lending
(MBS, which are sometimes also referred to as
alternative until the Tax Reform Act of 1986 (TRA).
private label or as asset-backed securities [ABS]).
The TRA increased the demand for mortgage debt
In addition, subprime loans were originated
because it prohibited the deduction of interest on
mostly by nondepository and monoline finance
consumer loans, yet allowed interest deductions
companies.
on mortgages for a primary residence as well as
During this time period, subprime mortgages
one additional home. This made even high-cost
were relatively new and apparently profitable,
mortgage debt cheaper than consumer debt for
but the performance of the loans in the long run
many homeowners. In environments of low and
was not known. By 1997, delinquent payments
declining interest rates, such as the late 1990s
and defaulted loans were above projected levels
and early 2000s, cash-out refinancing6 becomes
and an accounting construct called “gains-on sales
a popular mechanism for homeowners to access
7 One challenge the subprime industry will face in the future is the
need to develop business plans to maintain volume when interest
6 Cash-out refinancing indicates that the new loan is larger than the
rates rise. This will likely include a shift back to home equity
old loan and the borrower receives the difference in cash.
mortgages and other second-lien mortgages.
3 8
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Table 5
Top Ten B&C Originators, Selected Years
Rank
2003
2002
1
Ameriquest Mortgage, CA
Household Finance, IL
2
New Century, CA
CitiFinancial, NY
3
CitiFinancial, NY
Washington Mutual, WA
4
Household Finance, IL
New Century, CA
5
Option One Mortgage, CA
Option One Mortgage, CA
6
First Franklin Financial Corp, CA
Ameriquest Mortgage, DE
7
Washington Mutual, WA
GMAC-RFC, MN
8
Countrywide Financial, CA
Countrywide Financial, CA
9
Wells Fargo Home Mortgage, IA
First Franklin Financial Corp, CA
10
GMAC-RFC, MN
Wells Fargo Home Mortgage, IA
2001
2000
1
Household Finance, IL
CitiFinancial Credit Co, MO
2
CitiFinancial, NY
Household Financial Services, IL
3
Washington Mutual, WA
Washington Mutual, WA
4
Option One Mortgage, CA
Bank of America Home Equity Group, NC
5
GMAC-RFC, MN
GMAC-RFC, MN
6
Countrywide Financial, CA
Option One Mortgage, CA
7
First Franklin Financial Corp, CA
Countrywide Financial, CA
8
New Century, CA
Conseco Finance Corp. (Green Tree), MN
9
Ameriquest Mortgage, CA
First Franklin, CA
10
Bank of America, NC
New Century, CA
1996
1
Associates First Capital, TX
2
The Money Store, CA
3
ContiMortgage Corp, PA
4
Beneficial Mortgage Corp, NJ
5
Household Financial Services, IL
6
United Companies, LA
7
Long Beach Mortgage, CA
8
EquiCredit, FL
9
Aames Capital Corp., CA
10
AMRESCO Residential Credit, NJ
NOTE: B&C loans are defined as less than A quality non-agency (private label) paper loans secured by real estate. Subprime mortgage
and home equity lenders were asked to report their origination volume by Inside B&C Lending. Wholesale purchases, including loans
closed by correspondents, are counted.
SOURCE: Inside B&C Lending.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
3 9
Chomsisengphet and Pennington-Cross
accounting” magnified the cost of the unantici-
one of the nation’s largest thrifts. United
pated losses. In hindsight, many lenders had
Companies filed for bankruptcy, and Aames
underpriced subprime mortgages in the competi-
Capital Corporation was delisted after significant
tive and high-growth market of the early to mid-
financial difficulties. Household Financial
1990s (Temkin, Johnson, and Levy, 2002).
Services, one of the original finance companies,
By 1998, the effects of these events also spilled
has remained independent and survived the
over into the secondary market. MBS prices
period of rapid consolidation. In fact, in 2003 it
dropped, and lenders had difficulty finding
was the fourth largest originator and number two
investors to purchase the high-risk tranches. At
servicer of loans in the subprime industry.
or at about the same time, the 1998 Asian financial
crisis greatly increased the cost of borrowing and
again reduced liquidity in the all-real-estate mar-
THE EVOLUTION OF SUBPRIME
kets. This impact can be seen in Table 4, where
LENDING
the securitization rate of subprime loans drops
from 55.1 percent in 1998 to 37.4 percent in 1999.
This section provides a detailed picture of
the subprime mortgage market and how it has
In addition, the volume of originations shown in
evolved from 1995 through 2004. We use indi-
Table 3 indicates that they dropped from $105.6
vidual loan data leased from LoanPerformance.
billion in 1999 to $102.2 billion in 2000. Both of
The data track securities issued in the secondary
these trends proved only transitory because both
market. Data sources include issuers, broker
volume and securitization rates recovered in
dealers/deal underwriters, servicers, master ser-
2000-03.
vicers, bond and trust administrators, trustees,
Partially because of these events, the structure
and other third parties.
of the market also changed dramatically through
As of March 2003, more than 1,000 loan pools
the 1990s and early 2000s. The rapid consolidation
were included in the data. LoanPerformance
of the market is shown in Table 3. For example,
estimates that the data cover over 61 percent of
the market share of the top 25 firms making sub-
the subprime market. Therefore, it represents the
prime loans grew from 39.3 percent in 1995 to
segment of the subprime market that is securitized
over 90 percent in 2003.
and could potentially differ from the subprime
Many firms that started the subprime industry
market as a whole. For example, the average rate
either have failed or were purchased by larger
of subprime loans in foreclosure reported by the
institutions. Table 5 shows the top 10 originators
LoanPerformance data is 35 percent of the rate
for 2000-03 and 1996. From 2000 forward the list
reported by the MBAA. The MBAA, which does
of top originators is fairly stable. For example,
indicate that their sample of loans is not represen-
CitiFinancial, a member of Citigroup, appears
tative of the market, classifies loans as subprime
each year, as does Washington Mutual and
based on lender name. The survey of lenders of
Countrywide Financial. The largest firms increas-
prime and subprime loans includes approximately
ingly dominated the smaller firms from 2000
140 participants. As will be noted later in the
through 2003, when the market share of the top
section, the LoanPerformance data set is domi-
25 originators increased from 74 percent to 93
nated by the A–, or least risky, loan grade, which
percent.
may in part explain the higher rate of foreclosures
In contrast, many of the firms in the top 25
in the MBAA data. In addition, the demand for
in 1996 do not appear in the later time periods.
subprime securities should impact product mix.
This is due to a mixture of failures and mergers.
The LoanPerformance data set provides a host
For example, Associated First Capital was acquired
of detailed information about individual loans
by Citigroup and at least partially explains
that is not available from other data sources. (For
Citigroup’s position as one of the top originators
example, the MBAA data report delinquency and
and servicers of subprime loans. Long Beach
foreclosure rates but do not indicate any informa-
Mortgage was purchased by Washington Mutual,
tion about the credit score of the borrower, down
4 0
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
payment, existence of prepayment penalties, or
originations peaking (and surpassing FRMs) at
interest rate of the loan.8) The data set includes
over 866,000.9
many of the standard loan application variables
The subprime market took a temporary
such as the LTV ratio, credit score, loan amount,
downturn when the total number of FRM sub-
term, and interest rate type. Some “cleaning” of
prime originations declined during the 1998-2000
the data is conducted. For example, in each tab-
period; this observation is consistent with our
ulation, only available data are used. Therefore,
earlier brief history discussion and the down-
each figure may represent a slightly different
turn in originations reported by Inside Mortgage
sample of loans. In addition, to help make the
Finance (2004) and shown in Table 3. Since 2000,
results more comparable across figures, only
however, the subprime market has resumed its
adjustable- and fixed-rate loans to purchase or
momentum. In fact, from 2002 to 2003 the
refinance a home (with or without cash out) are
LoanPerformance data show a 62 percent increase
included from January 1995 through the December
and the Inside Mortgage Finance data show a 56
of 2004. But because of the delay in data reporting,
percent increase in originations.
the estimates for 2004 will not include all loans
During the late 1990s, house prices increased
from that year.
and interest rates dropped to some of the lowest
rates in 40 years, thus providing low-cost access
Volume
to the equity in homes. Of the total number of
subprime loans originated, just over one-half
Although the subprime mortgage market
were for cash-out refinancing, whereas more than
emerged in the early 1980s with the adoption of
one-third were for a home purchase (see Figure 4).
DIDMCA, AMTPA, and TRA, subprime lending
In 2003, for example, the total number of loans for
rapidly grew only after 1995, when MBS with
cash-out refinancing was over 560,000, whereas
subprime-loan collateral become more attractive
the number of loans for a home purchase totaled
to investors. Figure 3 illustrates this pattern using
more than 820,000, and loans for no-cash-out
our data (LoanPerformance) sample. In 1995, for
refinancing loans amounted to just under 250,000.
example, the number of subprime fixed-rate mort-
In the prime market, Freddie Mac estimated that,
gages (FRMs) originated was just slightly above
in 2003, 36 percent of loans for refinancing took
62,000 and the number of subprime adjustable-
at least 5 percent of the loan in cash (downloaded
rate mortgages (ARMs) originated was just
from the Cash-Out Refi Report at
above 21,000. Since then, subprime lending has
www.freddiemac.com/news/finance/data.html
increased substantially, with the number of FRM
on 11/4/04). This estimate is in contrast with
originations peaking at almost 780,000 and ARM
typical behavior in the subprime market, which
always has had more cash-out refinancing than
8
no-cash-out refinancing.
An additional source of information on the subprime market is a
list of lenders published by the United States Department of Housing
Given the characteristics of an application,
and Urban Development (HUD) Policy Development and Research
lenders of subprime loans typically identify bor-
(PD&R). This list has varied from a low of 51 in 1993 to a high of
256 in 1996; in 2002, the last year available, 183 subprime lenders
rowers and classify them in separate risk cate-
are identified. The list can then be matched to the Home Mortgage
gories. Figure 5 exhibits four risk grades, with
Disclosure Act (HMDA) data set. The list is compiled by examining
A– being the least risky and D being the riskiest
trade publications and HMDA data analysis. Lenders with high
denial rates and a high fraction of home refinances are potential
grade.10 The majority of the subprime loan origi-
candidates. The lenders are then called to confirm that they special-
ize in subprime lending. As a result, loans identified as subprime
using the HUD list included only firms that specialize in subprime
9 Similarly, Nichols, Pennington-Cross, and Yezer (2005) note that
lending (not full-service lenders). As a result, many subprime loans
the share of subprime mortgage lending in the overall mortgage
will be excluded and some prime loans will be included in the
market grew from 0.74 percent in the early 1990s to almost 9 percent
sample. Very little detail beyond the interest rate of the loan and
by the end of 1990s.
whether the rate is adjustable is included. For example, the existence
10
of prepayment penalties is unknown—a unique and key feature
Loan grades are assigned by LoanPerformance and reflect only the
of subprime lending. Still this lender list has proved useful in
rank ordering of any specific firm’s classifications. Because these
characterizing the neighborhood that these loans are originated
classifications are not uniform, there will be mixing of loan qualities
in. See, for example, Pennington-Cross (2002) and Calem, Gillen,
across grades. Therefore, these categories will likely differ from the
and Wachter (2004).
Countrywide examples used earlier.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
4 1
Chomsisengphet and Pennington-Cross
Figure 3
Number of Loans Originated
Number
1,000,000
Adjustable Rate
Fixed Rate
800,000
600,000
400,000
200,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 4
Number of Loans Originated by Purpose
Number
1,000,000
Purchase
Refinance—Cash Out
Refinance—No Cash Out
750,000
500,000
250,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
SOURCE: LoanPerformance ABS securities data base of subprime loans.
4 2
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 5
Number of Loans Originated by Grade
Number
700,000
A–
B
600,000
C
D
500,000
400,000
300,000
200,000
100,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
SOURCE: LoanPerformance ABS securities data base of subprime loans.
nations in this data set are classified into the low-
created by Fair Isaac Corporation to measure
est identified risk category (grade A–), particularly
consumer credit worthiness) for FRMs is almost
after 1998. In addition, the proportion of grade
50 points lower than for ARMs (623 versus 675).
A– loans to the total number of loans has contin-
During the 1990s, average credit scores tended to
uously increased from slightly over 50 percent
decline each year, particularly for ARM borrow-
in 1995 to approximately 84 percent in 2003. On
ers; but since 2000, credit scores have tended to
the other hand, the shares of grades B, C, and D
improve each year. Hence, it appears that sub-
loans have all declined since 2000. Overall, these
prime lenders expanded during the 1990s by
observations illustrate that, since 1998-99, the
extending credit to less-credit-worthy borrowers.
subprime market (or at least the securitized seg-
Subsequently, the lower credit quality unexpect-
ment of the market) has been expanding in its
edly instigated higher delinquency and default
least-risky segment. It seems likely then that the
rates (see also Temkin, Johnson, and Levy, 2002).
move toward the A– segment of subprime loans
With the improved credit quality since 2000,
is in reaction to (i) the events of 1998, (ii) the dif-
the average FICO has jumped from just under 622
ficulty in correctly pricing the higher-risk seg-
in 2000 to just over 651 in 2004 (closing in on
ments (B, C, and D credit grades), and, potentially,
the 669 average conventional FICO reported by
Nichols, Pennington-Cross, and Yezer, 2005). As
(iii) changes in the demand for securities for sub-
shown in Figure 7, lenders of subprime loans are
prime loans in the secondary market.
increasing the number of borrowers with scores
Credit Scores
in the 500-600 and 700-800 ranges and decreasing
the number with scores below 500. Specifically,
On average, ARM borrowers have lower credit
from 2000 to 2003, the share of borrowers with
scores than FRM borrowers (see Figure 6). In 2003,
FICO scores between 700 and 800 rose from
for example, the average FICO (a credit score
approximately 14 percent to 22 percent.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
4 3
Chomsisengphet and Pennington-Cross
Figure 6
Average Credit Score (FICO)
FICO
800
Adjustable Rate
Fixed Rate
700
600
500
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 7
Share of Loans by Credit Score
Percentage
80
FICO 500
500 < FICO 600
600 < FICO 700
700 FICO < 800
60
800 FICO
40
20
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
4 4
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 8
Loan Amounts by Credit Score
Dollars
250,000
FICO 500
500 < FICO 600
600 < FICO 700
200,000
700 FICO < 800
800 FICO
150,000
100,000
50,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 9
House Prices by Credit Score
Dollars
350,000
FICO 500
500 < FICO 600
300,000
600 < FICO 700
700 FICO < 800
800 FICO
250,000
200,000
150,000
100,000
50,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
4 5
Chomsisengphet and Pennington-Cross
Figure 10
Loan to Value Ratio (LTV)
Loan to Value Ratio
90
Adjustable Rate
Fixed Rate
85
80
75
70
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Moreover, lenders have on average provided
look like lenders are adding more risk by originat-
smaller loans to higher-risk borrowers, presumably
ing more ARMs with higher LTVs; however, this
to limit risk exposure (see Figure 8). As noted pre-
change primarily reflects borrowers with better
viously, these changes in underwriting patterns are
credit scores and more loans classified as A–.
consistent with lenders looking for new ways to
Therefore, this is additional evidence that lenders
limit risk exposure. In addition, although loan
of subprime loans reacted to the losses sustained
amounts have increased for all borrowers, the
in 1998 by moving to less-risky loans—primarily
amounts have increased the most, on average,
to borrowers with higher credit scores.
for borrowers with better credit scores. Also, as
As shown in Figure 11, this shift in lending
expected, borrowers with the best credit scores
strategy was accomplished by (i) steadily reducing
purchased the most expensive houses (see
loans with a large down payment (LTV
70), (ii)
Figure 9).
decreasing loans with negative equity (LTV > 100),
and (iii) increasing loans with a 10 percent down
Down Payment
payment. Overall, lenders of subprime loans have
Figure 10 depicts average LTV ratios for sub-
been increasing loan amounts, shifting the distri-
prime loan originations over a 10-year period. The
bution of down payments, and increasing credit
primary finding here is that down payments for
score requirements, on average, since 2000.
FRMs were reduced throughout the 1990s but have
In general, borrowers with larger down pay-
increased steadily since. (Note that the change in
ments tend to purchase more expensive homes
business strategy occurs just after the 1998 crisis.)
(Figure 12). By tying the amount of the loan to
In contrast, over the same period, down payments
the size of the down payment, lenders limit their
for ARMs were reduced. On first inspection, it may
exposure to credit risk.
4 6
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 11
Share of Loans by LTV
Percentage
50
LTV 70
70 < LTV 80
80 < LTV 90
40
90 < LTV 100
100 < LTV
30
20
10
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 12
House Prices by LTV
Dollars
450,000
LTV 70
400,000
70 < LTV 80
80 < LTV 90
90 < LTV 100
350,000
100 < LTV
300,000
250,000
200,000
150,000
100,000
50,000
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
4 7
Chomsisengphet and Pennington-Cross
Figure 13
LTV by Credit Score
Loan to Value Ratio
100
FICO 500
500 < FICO 600
600 < FICO 700
700 FICO < 800
90
800 FICO
80
70
60
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
The LTV-FICO Trade-off
INTEREST RATES
In Figure 13, we observe that borrowers with
This section examines patterns in the interest
the best credit scores tend to also provide the
rate that borrowers are charged at the origination
largest down payments. But, beyond this obser-
of the loan. This does not reflect the full cost of
vation, there seems little correlation between
borrowing because it does not include any fees
credit scores and down payments.
and upfront costs that are borne by the borrower.
In contrast, Figure 14 shows a clear ordering
In addition, the borrower can pay extra fees to
of down payments (LTV ratios) by loan grade.
lower the interest rate, which is called paying
Loans in higher loan grades have smaller down
points.
payments on average. In fact, over time, especially
Despite these stipulations, we are able to find
after 2000, the spread tends to increase. This find-
relationships between the observed interest rates
ing is consistent with the philosophy that loans
and underwriting characteristics. There is not
identified as being more risky must compensate
much difference in the average interest rate (the
lenders by providing larger down payments. This
interest rate on the loan excluding all upfront
helps to reduce credit risk associated with trigger
and continuing fees) at origination for FRMs and
events, such as periods of unemployment and
ARMs (see Figure 16). But, both product types
changes in household structure, which can make it
have experienced a large drop in interest rates,
difficult for borrowers to make timely payments.
from over 10 percent in 2000 to approximately 7
Consistent with the loan grade classifications,
percent in 2004.
Figure 15 shows that lower-grade loans have lower
Underwriting standards usually rely heavily
credit scores. Therefore, as loans move to better
on credit history and LTVs to determine the appro-
grades, credit scores improve and down payments
priate risk-based price. In Figures 17 and 18 we
decrease.
see evidence of risk-based pricing based on bor-
4 8
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 14
LTV by Loan Grade
Loan to Value Ratio
90
A–
B
C
D
80
70
60
50
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 15
Credit Score by Loan Grade
Credit Score
700
A–
B
C
D
650
600
550
500
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
4 9
Chomsisengphet and Pennington-Cross
Figure 16
Interest Rates
Interest Rate
12
Adjustable Rate
11
Fixed Rate
10
9
8
7
6
5
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
rower credit scores and, to some small extent, on
result, the lender must increase the interest rate
borrower down payments. For example, borrowers
to decrease its loss if a default occurs.
with the highest FICO scores tend to receive a
Figure 19 shows the average interest rate by
lower interest rate. In 2004, average interest rates
loan grade. The riskiest borrowers (Grade D)
vary by over 2 percentage points from the highest
receive the highest interest rate, whereas the least-
to the lowest FICO scores.
risky borrowers (Grade A–) receive the lowest
This range of interest rates does not hold
interest rate. Interestingly, although interest rates
when pricing is based solely on down payments.
overall changed dramatically, the spread between
In fact, the striking result from Figure 18 is that,
the rates by grade have remained nearly constant
on average, the pricing of subprime loans is very
after 1999. This may indicate that the risks, and
similar for all down-payment sizes, except for
hence the need for risk premiums, are in levels,
loans with LTVs greater than 100, which pay a
not proportions, across risk grades.
substantial premium. One way to interpret these
results is that lenders have found good mecha-
Prepayment Penalties
nisms to compensate for the risks of smaller down
It is beyond the scope of this paper to define
payments and, as a result, down payments in
specific examples of predatory lending, but pre-
themselves do not lead to higher borrower costs.
payment penalties have been associated with
However, if the equity in the home is negative,
predatory practices. A joint report by the U.S.
no sufficient compensating factor can typically
Department of Housing and Urban Development
be found to reduce expected losses to maintain
(HUD) and the U.S. Department of Treasury
pricing parity. The borrower has a financial
(Treasury) (2002) defined predatory lending as
incentive to default on the loan because the loan
lending that strips home equity and places bor-
amount is larger than the value of the home. As a
rowers at an increased risk of foreclosure. The
5 0
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 17
Interest Rates by Credit Score
Interest Rate
12
FICO 500
500 < FICO 600
11
600 < FICO 700
700 FICO < 800
800 FICO
10
9
8
7
6
5
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 18
Interest Rates by LTV
Interest Rate
14
LTV 70
70 < LTV 80
80 < LTV 90
90 < LTV 100
12
100 < LTV
10
8
6
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
5 1
Chomsisengphet and Pennington-Cross
Figure 19
Interest Rates by Loan Grade
Interest Rate
14
A–
B
C
D
12
10
8
6
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
characteristics include excessive interest rates
rate or reduced fees, (ii) the borrower is provided
and fees, the use of single-premium credit life
an alternative mortgage choice, (iii) the nature of
insurance, and prepayment penalties that provide
the penalty is disclosed to the borrower, and (iv)
no compensating benefit, such as a lower interest
the penalty cannot be charged if the borrower
rate or reduced fees. In addition, some public
defaults on the loan and the note is accelerated
interest groups such as the Center for Responsible
(www.fanniemae.com/newsreleases/2000/
Lending believe that prepayment penalties are in
0710.jhtml).11 Therefore, we may expect to see a
their very nature predatory because they reduce
decline in the use of prepayment penalties starting
borrower access to lower rates (Goldstein and Son,
in 2000 and 2002, at least in part due to changes
2003).
in the demand for subprime securities.
Both Fannie Mae and Freddie Mac changed
Despite these concerns, prepayment penalties
their lending standards to prohibit loans (i.e.,
have become a very important part of the sub-
they will not purchase them) that include some
prime market. When interest rates are declining
types of prepayment penalties. On October 1, 2002,
or steady, subprime loans tend to be prepaid at
Freddie Mac no longer allowed the purchase of
elevated rates compared with prime loans
subprime loans with a prepayment penalty after
(Pennington-Cross, 2003, and UBS Warburg, 2002).
three years. However, loans originated before
In addition, subprime loans tend to default at
that date would not be affected by the restriction
elevated rates. As a result, the expected life of an
(see www.freddiemac.com/singlefamily/
average subprime loan is much shorter than that
ppmqanda.html downloaded on 2/14/05). If a
subprime loan stipulates a prepayment penalty,
11 When a borrower defaults, the lender typically will send an accelera-
Fannie Mae will consider the loan for purchase
tion note informing the borrower that the mortgage contract has
been violated and all of the remaining balance and fees on the
only if (i) the borrower receives a reduced interest
loan are due immediately.
5 2
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
Figure 20
Share of Loans with a Prepayment Penalty
Percentage
100
Adjustable Rate
Fixed Rate
80
60
40
20
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
of a prime loan. Therefore, there are fewer good
tation of the 1982 AMTPA by the Office of Thrift
(nonterminated) loans to generate income for an
and Supervision (OTS). Before 1996, the OTS
investor to compensate for terminated (defaulted
interpreted AMTPA as allowing states to restrict
and prepayed) loans. One mechanism to reduce
finance companies (which make many of the sub-
the break-even price on these fast-terminating
prime loans) from using prepayment penalties,
loans is to use prepayment penalties (Fortowsky
but the OTS exempted regulated federal deposi-
and LaCour-Little, 2002). Although this same
tory institutions from these restrictions. In 1996,
mechanism is used in the prime market, it is not
the OTS also allowed finance companies the
as prevalent.
same exemption. However, this position was
Figure 20 shows that, prior to 2000, the use
short lived and the OTS returned to its prior
interpretation in 2002.
of prepayment penalties grew quickly. Substan-
In 2003 and 2004, prepayment penalties
tially more ARMs than FRMs face a prepayment
declined for ARMs and held steady for FRMs.
penalty. For loans originated in 2000-02, approx-
This was likely caused by (i) the introduction of
imately 80 percent of ARMs were subject to a pre-
predatory lending laws in many states and cities
payment penalty compared with approximately
(typically these include ceilings on interest rates
45 percent of FRMs. Equally important, the share
and upfront fees, restrictions on prepayment
of ARMs and FRMs subject to a prepayment
penalties, and other factors)12; (ii) the evolving
penalty rose dramatically from 1995 to 2000. In
position of Fannie Mae and Freddie Mac on pre-
fact, at the end of the five-year period, ARMs were
five times more likely and FRMs twice as likely
12 For more details on predatory lending laws that are both pending
to have prepayment penalties.
and in force, the MBAA has a “Predatory Lending Law Resource
This rapid increase can at least partially be
Center” available at www.mbaa.org/resources/predlend/ and the
Law Offices of Herman Thordsen also provide detailed summaries
attributable to regulatory changes in the interpre-
of predatory laws at www.lendinglaw.com/predlendlaw.htm.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
5 3
Chomsisengphet and Pennington-Cross
Figure 21
Share of Loans with a Prepayment Penalty by Credit Score
Percentage
100
FICO 500
500 < FICO 600
600 < FICO 700
80
700 FICO < 800
800 FICO
60
40
20
0
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
Figure 22
Length of Prepayment Penalty
Months
45
Adjustable Rate
Fixed Rate
40
35
30
25
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
SOURCE: LoanPerformance ABS securities data base of subprime loans.
5 4
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Chomsisengphet and Pennington-Cross
payment penalties; and (iii) the reversed OTS
prepayment penalties has declined in the past few
interpretation of AMTPA in 2002 (see 67 Federal
years because the securities market has adjusted
Register 60542, September 26, 2002), which again
to public concern about predatory lending and
made state laws apply to finance companies just
the regulation of finance companies has changed.
as they had prior to 1996.
The evidence also shows that the subprime
The share of loans containing a prepayment
market has provided a substantial amount of risk-
penalty is lowest among borrowers with the
based pricing in the mortgage market by varying
highest, or best, FICO scores (see Figure 21). In
the interest rate of a loan based on the borrower’s
2003, for instance, about 20 percent of borrowers
credit history and down payment. In general, we
with a FICO score above 800 were subject to a
find that lenders of subprime loans typically
prepayment penalty, whereas over 60 percent of
require larger down payments to compensate for
borrowers with a FICO score below 700 faced
the higher risk of lower-grade loans. However, even
such a penalty.
with these compensating factors, borrowers with
To understand the prevalence of these penal-
low credit scores still pay the largest premiums.
ties, one must know how long prepayment penal-
ties last. Figure 22 shows that the length of the
penalty has generally been declining since 2000.
REFERENCES
Again, the introduction and threat of predatory
lending laws and Freddie Mac purchase require-
Calem, Paul; Gillen, Kevin and Wachter, Susan. “The
ments (that the term of a prepayment penalty be
Neighborhood Distribution of Subprime Mortgage
no more than three years) is likely playing a role
Lending.” Journal of Real Estate Finance and
in this trend. In addition, FRMs tend to have much
Economics, 2004, 29(4), pp. 393-410.
longer prepayment penalties. For example, in
2003, the average penalty lasted for almost three
Capozza, Dennis R. and Thomson, Thomas A.
years for FRMs and a little over two years for
“Subprime Transitions: Long Journey into
ARMs, both of which meet current Freddie Mac
Foreclosure.” Presented at the American Real
guidelines.
Estate and Urban Economics Annual Meeting,
Philadelphia, PA, January 2005.
CONCLUSION
Fortowsky, Elaine B. and LaCour-Little, Michael.
“An Analytical Approach to Explaining the
As the subprime market has evolved over the
Subprime-Prime Mortgage Spread.” Presented at
past decade, it has experienced two distinct
the Georgetown University Credit Research Center
periods. The first period, from the mid-1990s
Symposium Subprime Lending, 2002.
through 1998-99, is characterized by rapid growth,
with much of the growth in the most-risky seg-
Goldstein, Debbie and Son, Stacey Strohauer. “Why
ments of the market (B and lower grades). In the
Prepayment Penalties are Abusive in Subprime
second period, 2000 through 2004, volume again
Home Loans.” Center for Responsible Lending
grew rapidly as the market became increasingly
Policy Paper No. 4, April 2, 2003.
dominated by the least-risky loan classification
(A– grade loans). In particular, the subprime mar-
Hillier, Amy E. “Spatial Analysis of Historical
ket has shifted its focus since 2000 by providing
Redlining: A Methodological Exploration.” Journal
loans to borrowers with higher credit scores,
of Housing Research, November 2003, 14(1), pp.
allowing larger loan amounts, and lowering the
137-67.
down payments for FRMs. Furthermore, the sub-
prime market had reduced its risk exposure by
Immergluck, Daniel and Wiles, Marti. Two Steps Back:
limiting the loan amount of higher-risk loans and
The Dual Mortgage Market, Predatory Lending, and
imposing prepayment penalties on the majority
the Undoing of Community Development. Chicago:
of ARMs and low credit-score loans. The use of
The Woodstock Institute, 1999.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
J A N UA R Y / F E B R UA R Y
2 0 0 6
5 5
Chomsisengphet and Pennington-Cross
Inside Mortgage Finance. The 2004 Mortgage Market
Statistical Annual. Washington, DC: 2004.
Nichols, Joseph; Pennington-Cross, Anthony and Yezer,
Anthony. “Borrower Self-Selection, Underwriting
Costs, and Subprime Mortgage Credit Supply.”
Journal of Real Estate Finance and Economics,
March 2005, 30(2), pp. 197-219.
Pennington-Cross, Anthony. “Subprime Lending in
the Primary and Secondary Markets.” Journal of
Housing Research, 2002, 13(1), pp. 31-50.
Pennington-Cross, Anthony. “Credit History and the
Performance of Prime and Nonprime Mortgages.”
Journal of Real Estate Finance and Economics,
November 2003, 27(3), pp. 279-301.
Pennington-Cross, Anthony. “The Value of Foreclosed
Property.” Journal of Real Estate Research
(forthcoming).
Temkin, Kenneth; Johnson, Jennifer E.H. and Levy,
Diane. Subprime Markets, the Role of GSEs, and
Risk-Based Pricing. Washington, DC: U.S. Department
of Housing and Urban Development, Office of
Policy Development and Research, March 2002.
Tracy, Joseph; Schneider, Henry and Chan, Sewin.
“Are Stocks Over-Taking Real Estate in Household
Portfolios?” Current Issues in Economics and
Finance, Federal Reserve Bank of New York, April
1999, 5(5).
UBS Warburg. “Credit Refis, Credit Curing, and the
Spectrum of Mortgage Rates.” UBS Warburg
Mortgage Strategist, May 21, 2002, pp. 15-27.
U.S. Department of Housing and Urban Development
and U.S. Department of Treasury, National Predatory
Lending Task Force. Curbing Predatory Home
Mortgage Lending. Washington, DC: 2002.
5 6
J A N UA R Y / F E B R UA R Y
2 0 0 6
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
