Mortgage Financing: Actions Needed to Help FHA Manage Risks from 
New Mortgage Loan Products (11-FEB-05, GAO-05-194).		 
                                                                 
The U.S. Department of Housing and Urban Development (HUD),	 
through its Federal Housing Administration (FHA), insures	 
billions of dollars in home mortgage loans made by private	 
lenders. FHA insures low down payment loans and a number of	 
parties have made proposals to either eliminate or otherwise	 
change FHA's borrower contribution requirements. GAO was asked to
(1) identify the key characteristics of existing low and no down 
payment products, (2) review relevant literature on the 	 
importance of loan-to-value (LTV) ratios and credit scores to	 
loan performance, (3) report on the performance of low and no	 
down payment mortgages supported by FHA and others, and (4)	 
identify lessons for FHA from others in terms of designing and	 
implementing low and no down payment products.			 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-194 					        
    ACCNO:   A17537						        
  TITLE:     Mortgage Financing: Actions Needed to Help FHA Manage    
Risks from New Mortgage Loan Products				 
     DATE:   02/11/2005 
  SUBJECT:   Data collection					 
	     Homeowners loans					 
	     Housing programs					 
	     Insurance premiums 				 
	     Lending institutions				 
	     Mortgage loans					 
	     Mortgage programs					 
	     Program evaluation 				 
	     Risk management					 
	     Statistical data					 
	     Mortgage protection insurance			 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-05-194

United States Government Accountability Office

GAO	Report to the Chairman, Subcommittee on Housing and Community
Opportunity,

           Committee on Financial Services, House of Representatives

February 2005

                                    MORTGAGE
                                   FINANCING

    Actions Needed to Help FHA Manage Risks from New Mortgage Loan Products

                                       a

GAO-05-194

[IMG]

February 2005

MORTGAGE FINANCING

Actions Needed to Help FHA Manage Risks from New Mortgage Loan Products

                                 What GAO Found

FHA and many other mortgage institutions provide many low and no down
payment products with requirements that vary in terms of eligibility,
borrower investment, underwriting, and risk mitigation. While these
products are similar, there are some important differences, including that
FHA has lower loan limits, allows closing costs and the up-front insurance
premium to be financed in the mortgage, and permits the down payment funds
to come from nonprofits that receive funds from sellers. FHA also differs
in that it does not require prepurchase counseling.

A substantial amount of research GAO reviewed indicates that LTV ratio and
credit score are among the most important factors when estimating the risk
level associated with individual mortgages. GAO's analysis of the
performance of low and no down payment mortgages supported by FHA and
others corroborates key findings in the literature. Generally, mortgages
with higher LTV ratios (smaller down payments) and lower credit scores are
riskier than mortgages with lower LTV ratios and higher credit scores.

Some practices of other mortgage institutions offer a framework that could
help FHA manage the risks associated with introducing new products or
making significant changes to existing products. Mortgage institutions may
impose limits on the volume of the new products they will permit and on
who can sell and service these products. FHA officials question the
circumstances in which they can limit volumes for their products and
believe they do not have sufficient resources to manage a product with
limited volumes. Mortgage institutions sometimes require additional credit
enhancements, such as higher insurance coverage; and sometimes require
stricter underwriting, such as credit score thresholds, when introducing a
new low or no down payment product. FHA is authorized to require an
additional credit enhancement by sharing risk through co-insurance but
does not currently use this authority. FHA has used stricter underwriting
criteria but this has not included credit score thresholds.

Average Four-Year Default Rates for FHA Insured Loans Originated in 1998,
1999, and 2000 (by LTV)

Source: FY 2003 Actuarial Review of the Mutual Mortgage Insurance Fund.

United States Government Accountability Office

Contents

  Letter

Results in Brief

Background

Characteristics and Standards of Low and No Down Payment Products Vary by
Mortgage Institution

Research Shows LTV Ratio and Credit Score Are Important When Estimating
Risk of Individual Mortgages

Our Analysis Indicated That Mortgages with Higher LTV Ratios and Lower
Credit Scores Pose Greater Risks

Several Practices Mortgage Institutions Use in Designing and Implementing
Low and No Down Payment Products Could Be Instructive for FHA

Conclusions

Matters for Congressional Consideration

Recommendations for Executive Action

Agency Comments and Our Evaluation 1 3 7

12

23

28

39 46 47 48 48

Appendixes

Appendix I:

Appendix II:

Appendix III:

Appendix IV:

Scope and Methodology

Papers Identified in Literature Search and Included in Analysis

Comments from the Department of Housing and Urban Development

GAO Contacts and Staff Acknowledgments

GAO Contacts
Staff Acknowledgments

51

57

60

62 62 62

Tables   Table 1: Eligibility Limitations of FHA Compared with RHS, VA, 
                                                                       and 
                           Selected Products of Fannie Mae and Freddie Mac 15 
             Table 2: LTV Calculations for FHA, RHS, VA, Fannie Mae, and   
                                     Freddie Mac                           18 
Figures Figure 1: Evolution of Low and No Down Payment Products Figure   9 
           2: Percentage of Home Purchase Mortgages with a LTV             
                    Ratio Higher Than 95 Percent for HUD (FHA), VA, Fannie 
                         Mae, and Freddie Mac for Loans Originated in 2000 10 

Contents

Figure 3:	Percentage of Loans with LTV Ratios of 95 Percent or Higher That
Fannie Mae and Freddie Mac Purchased, 1997-2000 11

Figure 4: Four-year Relative Default Rates by LTV Ratio for FHA-Insured
Mortgages (1992, 1994, and 1996) 30 Figure 5: Four-year Relative Default
Rates by Credit Score for FHA-Insured Mortgages (1992, 1994, and 1996) 31

Figure 6:	Four-year Relative Default Rates by LTV Ratio and Credit Score
for FHA-insured Mortgages (1992, 1994, and 1996) 32

Figure 7: Four-Year Relative Default Rates by LTV Ratio for Conventional
Mortgages (1997, 1998, and 1999) 34 Figure 8: Four-Year Relative Default
Rates by Credit Score for Conventional Mortgages (1997, 1998, and 1999) 35

Figure 9:	Four-Year Relative Default Rates by LTV Ratio and Credit Score
for Conventional Mortgages (1997, 1998, and 1999) 37

Abbreviations

ARM adjustable rate mortgage
FASAB Federal Accounting Standards Advisory Board
FHA Federal Housing Administration
GSE government-sponsored enterprise
HECM Home Equity Conversion Mortgage
HUD U.S. Department of Housing and Urban Development
LTV loan-to-value
MMI Mutual Mortgage Insurance
OFHEO Office of Federal Housing Enterprise Oversight
RHS Rural Housing Service
USDA U.S. Department of Agriculture
VA U.S. Department of Veterans Affairs

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

A

United States Government Accountability Office Washington, D.C. 20548

February 11, 2005

The Honorable Bob Ney
Chairman
Subcommittee on Housing and Community Opportunity
Committee on Financial Services
House of Representatives

Dear Mr. Chairman:

Every year, the U.S. Department of Housing and Urban Development
(HUD), through its Federal Housing Administration (FHA), insures billions
of dollars in home mortgage loans made by private lenders, very often with
low down payments. FHA mortgage insurance helps homebuyers with
limited funds to obtain a home mortgage. Homebuyers with FHA-insured
loans need to make a 3 percent contribution toward the purchase of the
property and may finance some of the closing costs associated with the
loan. As a result, an FHA-insured loan could equal nearly 100 percent of
the
property's value or sales price-commonly called loan-to-value (or LTV)
ratio.1 In recent years, various mortgage industry participants, such as
lenders, private mortgage insurers, and government-sponsored enterprises
(the Federal National Mortgage Association [Fannie Mae] and the Federal
Home Loan Mortgage Corporation [Freddie Mac]) have begun to support
mortgage products that require very little or no down payment. Among
these products, some allow third-party provision of gift down payment
assistance. Recently, in a HUD contractor study of a national sample of
FHA loans, of those loans that received down payment assistance, 29
percent received assistance from a nonprofit down payment assistance
provider. In addition, the FHA and others have proposed eliminating the
borrower contribution requirement for FHA-insured loans. At the same
time, the mortgage industry has moved toward greater use of automated
systems assessing the risk level of mortgages. These automated
underwriting systems rely, in part, on individuals' credit scores or
credit
history, and these systems have played an integral role in the provision
of
low and no down payment mortgage products.2

1For purposes of this report, we define loans with LTV ratios of greater
than 97 percent as having a high LTV ratio.

2Credit scores are a single numerical score, based on an individual's
credit history, that measures that individual's creditworthiness.

In light of recent changes in the composition of FHA-insured mortgage
products and the proposal to eliminate the borrower contribution
requirement for FHA-insured mortgages, you asked us to evaluate low and no
down payment lending. Specifically, this report examines (1) the key
characteristics and standards of mortgage products-supported by FHA and
others-that require low or no down payments; (2) what published research
indicates about the importance of variables such as LTV ratios and credit
scores in estimating the risk level associated with individual mortgages;
(3) the performance of low and no down payment mortgages supported by FHA
and others; and (4) what lessons FHA might learn from others that support
low and no down payment lending in terms of designing and implementing
such products.

To address the objectives we interviewed officials at FHA, the U.S.
Department of Agriculture (USDA), and U.S. Department of Veterans Affairs
(VA); and staff at selected conventional mortgage providers3; private
mortgage insurers; two government-sponsored enterprises (GSE); the Office
of Federal Housing Enterprise Oversight (OFHEO); selected state housing
finance agencies; and nonprofit down payment assistance providers. We
reviewed descriptions of low and no down payment mortgage products
supported by selected mortgage industry participants and compared the
standards used by these entities. To determine what published research
indicates about the variables that are most important when estimating the
risk level associated with individual mortgages, we reviewed recent and
relevant papers that we identified through a systematic search of economic
literature. To describe the low and no down payment performance of loans
supported by FHA and others, we examined the relationship among mortgage
performance, LTV ratio, and credit score using 4-year default rates for a
special research sample of over 400,000 mortgages insured by FHA during
1992, 1994, and 1996 and for all conventional loans originated in 1997,
1998, and 1999 and purchased by Fannie Mae or Freddie Mac. We chose these
years because, for FHA, these data are the only significant data set of
FHA-insured loans that includes credit scores and that had at least 4
years of loan performance activity. For Fannie Mae and Freddie Mac, in
1997, these institutions began purchasing an increasing number of loans
with the highest LTV ratios; loans originated after 1999 would have less
than 4 years of experience to analyze. The GSEs provided us data that they
considered to be proprietary. We did not disclose

3Conventional lenders provide mortgages that do not carry government
insurance or guarantees.

information that could be considered proprietary, but this did not limit
our overall findings. We assessed the reliability of the FHA, Fannie Mae,
and Freddie Mac data by discussing the data with knowledgeable FHA
officials and Fannie Mae and Freddie Mac officials; reviewing recent audit
reports that evaluated the information systems at each entity; and
comparing the data with similar publicly available data. We determined
that the data are sufficiently reliable to use in our analysis of the
performance of low and no down payment mortgages. To determine what
lessons FHA might learn from others that support low and no down payment
lending, we obtained testimonial information from mortgage industry
participants about the steps they take to design and implement low and no
down payment lending products. We did not verify that these institutions,
in fact, used these practices.

We performed our audit work from January 2004 to December 2004 in
accordance with generally accepted government auditing standards. Appendix
I provides a full description of our scope and methodology.

Results in Brief	FHA and other mortgage institutions provide products that
enable homebuyers to purchase homes using little of their own funds. While
similar, the products offered by FHA and others have important differences
in terms of eligibility, borrower investment, underwriting, and risk
mitigation. With respect to eligibility, for example, FHA loan limits are
lower than those in the conventional market. In terms of borrower
investment, FHA's product differs from others in that it allows some of
the closing costs and the up-front insurance premium to be financed in the
mortgage. Although FHA requires a 3 percent borrower contribution, it can
come from sources other than the borrower. Many conventional low and no
down payment products also permit down payment funds to come from others
but generally stipulate that down payment funds, either directly or
indirectly, cannot come from an interested or seller-related party. FHA
also does not permit down payment funds to come directly or indirectly
from sellers but does permit nonprofits that receive contributions from
sellers to provide down payment assistance to homebuyers. With respect to
underwriting, many mortgage institutions use automated systems to some
extent. These systems allow lenders to quickly assess the riskiness of
mortgages by simultaneously considering multiple factors including the
credit score and credit history of borrowers. With respect to risk
mitigation, FHA differs from conventional mortgage institutions that
provide low and no down payment products. For example, while FHA does not
require prepurchase counseling, some institutions require borrowers to

receive counseling for their no down payment products. Additionally, some
mortgage institutions require higher private mortgage insurance coverage
on their no down payment products. While FHA already provides nearly 100
percent insurance, it does have the authority to share risk but does not
currently use this authority.

A substantial body of economic research indicates that LTV ratio and
credit score are among the most important factors when estimating the risk
level associated with individual mortgages. We reviewed 45 economic
research papers that examined multiple factors that could be important; of
these, 37 examined if LTV ratio was important. Almost all of these papers
(35) found the LTV ratio of a mortgage important when estimating the risk
level associated with individual mortgages. For example, one study found
that the default rates for mortgages with an LTV ratio above 95 percent
are three to four times higher than default rates for mortgages with an
LTV ratio of 90 to 95 percent. Of the 45 papers we reviewed, 19 examined
whether credit score was important. All but one of these papers found the
borrower's credit score important when estimating the risk level
associated with individual mortgages. For example, one study found that a
mortgage with a credit score of 728 (indicating an applicant with
excellent credit) had a default probability of 1.26 percent, while the
default probability of a mortgage with a credit score of 642 was more than
two times higher-3.41 percent. The research also indicated that additional
factors-such as characteristics of the borrower, mortgage, and
property-may help in estimating the risk level associated with individual
mortgages.

Our analysis of FHA and conventional mortgage data indicated that,
generally, mortgages with higher LTV ratios (smaller down payments) and
lower credit scores are riskier than mortgages with lower LTV ratios and
higher credit scores.4 For example, FHA-insured mortgages with LTV ratios
greater than 80 percent and low credit scores (below 660), had a default
rate above the FHA average default rate.5 There is a similar relationship
for conventional mortgages. For example, conventional mortgages with LTV

4While this analysis is useful in determining the extent to which LTV
ratios and credit scores are helpful in predicting the risk level
associated with individual mortgages, the FHA and conventional relative
default rates are not strictly comparable because they are for different
time periods and because FHA has a higher overall default rate.

5For this analysis, we define default as a credit event that includes
foreclosed mortgages, as well as mortgages that did not experience a
foreclosure, but that would typically lead to a credit loss, such as a
"short sale" or a "third party sale" termination of the mortgage.
Delinquency was not considered to be default.

ratios greater than 80 percent and credit scores below 700 had a default
rate above the conventional average default rate.

FHA and other mortgage institution officials report using a number of
similar practices in designing and implementing low and no down payment
products but FHA does not typically follow some practices that could help
it to manage the risks associated with introducing new products or making
significant changes to existing products. Mortgage institutions we spoke
with, such as Fannie Mae, Freddie Mac, and the private mortgage insurers,
told us they often initially analyze the risk of products that are similar
to those they are considering by both using internal data and data they
acquire externally. For example, Freddie Mac officials said that Freddie
Mac obtained external loan data when they purchased loans through a
structured transaction to provide insight into low and no down payment
lending when designing its low and no down payment products.6 These loans
had characteristics of loans they were considering purchasing on a routine
basis. FHA officials also said they have purchased loan performance and
other data when designing a new product or studying changes to an existing
product but rely more heavily on internal data. Mortgage institutions also
sometimes require additional credit enhancements or stricter underwriting
when introducing a new low or no down payment product.7 FHA has the
authority, but does not currently require an additional credit
enhancement, and has made adjustments to mortgage underwriting features
but this has not included credit score thresholds such as those used by
other mortgage institutions.8 FHA does adjust premiums. In implementing
new products, mortgage institutions may impose limits on the volume of the
new products they will permit and on who can sell and service these
products. Fannie Mae and Freddie Mac officials described new low and no
down payment products that they first introduced as part of a pilot or
with certain limits on how many of the new products they would commit to
purchase. However, limits on the availability of new or revised FHA
mortgage insurance are sometimes set

6Structured transaction is a broad term that covers any of several methods
of dividing cash flows among several investors in a pool of mortgages.

7A credit enhancement is provided when a party agrees to assume risks
associated with a loan. For example, mortgage insurance is a type of
credit enhancement.

8FHA also has the authority to obtain credit enhancements through the use
of co-insurance. Co-insurance requires lenders to share in the risks of
insuring mortgages by assuming some percentage of the losses on the loans
that they originated.

through legislation and focus on the volume of loans that FHA may insure.9
FHA officials questioned the circumstances in which they can use pilots or
limit volumes when it is not required by Congress and told us that they
lack the resources to effectively manage a program with limited volumes.
Finally, mortgage institutions, including FHA, establish enhanced
monitoring and oversight for new low and no down payment products and
revise new products as they improve their understanding of these products.
In addition to reviewing routine data on the characteristics and
performance of all loans, some institutions may regularly review the
underwriting of all loans within a new product line as compared with a
sample of loans for their established products (for which they better
understand performance). FHA typically reviews a sample of loans for a new
product and says that they make changes to products based on their
acquired understanding.

This report includes matters for congressional consideration and
recommendations to HUD. We suggest that Congress consider limiting the
initial availability of any new single-family insurance product it may
authorize. We recommend that FHA consider using pilots for new products
and for making significant changes to its existing products and that, when
doing so, FHA develop a framework for when to use pilots and how they
should be implemented. We also recommend that FHA explore various
techniques for mitigating risks when implementing new products that have
greater risk or for which risk is not well understood.

We provided a draft of this report to HUD, Fannie Mae, Freddie Mac, USDA,
and VA. We received written comments from HUD, which are discussed later
in this report and reprinted in appendix III. We also received technical
comments from HUD, Fannie Mae, and USDA which have been incorporated where
appropriate. VA did not have comments on the draft. HUD stated that in
developing its proposed zero down payment product, it considered all of
the items that we recommended they consider including piloting. However,
it is not clear under what circumstances HUD believes piloting or limiting
the availability of a changed or new product would be appropriate or
possible. During the course of our work, HUD officials told us that they
face challenges in administering a pilot program because of the difficulty
of selecting only a limited number of lenders and borrowers. HUD officials
also held that they may not have the authority to limit products

9For example, Congress established volume limits with the introduction of
FHA's Home Equity Conversion Mortgage program.

and that they lacked sufficient resources to adequately manage products as
part of a pilot or with limited volumes. We believe that HUD needs to
further consider piloting or limiting volume of new or changed products.
There are several available techniques for limiting an initial product,
including limiting the time period in which it is available. Further we
believe that in some circumstances the potential costs of making widely
available a product with risk that is not well understood could exceed the
cost of initially implementing such a product on a limited basis. We
therefore recommend that HUD consider a wide range of options for
mitigating risk, including piloting or limiting the volume of new or
changed products. To the extent HUD believes it does not have the
authority for exercising the options we describe, it should seek the
authority from Congress.

Background	Mortgage insurance, a commonly used credit enhancement,
protects lenders against losses in the event of default. Lenders usually
require mortgage insurance when a homebuyer has a down payment of less
than 20 percent of the value of the home. FHA, VA, the USDA's Rural
Housing Service (RHS), and private mortgage insurers provide this
insurance. In 2003, lenders originated $3.8 trillion of single-family
mortgage loans, of which more than 60 percent were for refinancing. Of all
the insured loans including refinancings originated in 2003, private
companies insured about 64 percent, FHA insured about 26 percent, VA
insured about 10 percent, and RHS insured a very small number. Private
mortgage insurers generally offer first loss coverage- that is, they will
pay all the losses from a foreclosure up to a stated percentage of the
claim amount.10 Generally, these insurers limit the coverage that they
offer to between 25 percent and 35 percent of the claim amount. The
insurance offered by the government varies in the amount of lender
incurred losses it will cover. For example, VA guarantees losses up to 25
percent to 50 percent of the loan, while FHA's principal single-family
insurance program insures almost 100 percent.11 FHA plays a particularly
large role in certain market segments, including low-income and first-time
homebuyers. During fiscal years 2001 to 2003, FHA insured a total of about
3.7 million mortgages with a total value of about $425 billion. FHA
insures most of its mortgages for single-family

10Claim amount includes outstanding loan amount plus other costs including
legal fees.

11Single-family loans insured by FHA may be used to finance the purchase
of new or existing one-to-four-family properties. 12 U.S.C. S:1709(b).

housing under its Mutual Mortgage Insurance Fund. To cover lenders'
losses, FHA collects insurance premiums from borrowers. These premiums,
along with proceeds from the sale of foreclosed properties, pay for claims
that FHA pays lenders as a result of foreclosures.

Fannie Mae and Freddie Mac are government-sponsored private corporations
with stated public missions chartered by Congress to provide a continuous
flow of funds to mortgage lenders and borrowers. Fannie Mae and Freddie
Mac purchase mortgages from lenders across the country and finance their
mortgage purchases through borrowing or issuing mortgage-backed securities
that are sold to investors. They purchase single-family mortgages up to
the "conforming loan limit," which for 2005 was set at $359,650.12 Their
purchase guidelines and underwriting standards have a dominant role in
determining the types of loans that primary lenders will originate in the
conventional conforming market.

Members of the conventional mortgage market (such as private mortgage
insurers, Fannie Mae, Freddie Mac, and large private lenders) have been
increasingly active in supporting low and no down payment mortgage
products. Many private mortgage insurers will now insure a mortgage up to
100 percent of the value of the housing being purchased. Fannie Mae and
Freddie Mac, working together with the private mortgage insurers, have
become more aggressive in developing high LTV products that target low-and
moderate-income or first-time homebuyers while also developing high LTV
products designed for use by borrowers across the income spectrum. Figure
1 shows the history of the introduction of low and no down payment
mortgage products at three LTV levels. FHA and VA have been backing low
and no down payment mortgages for many years, and Fannie Mae and Freddie
Mac permitted conventional lenders to sell them mortgages with an LTV of
97 percent in 1994 and 1998, respectively. Freddie Mac and Fannie Mae's no
down payment mortgage products were introduced in 2000.

12Referred to as the conforming loan limit because the mortgages conform
to underwriting standards established by Fannie Mae and Freddie Mac. The
limit is higher for single-family mortgages secured by two-, three-, and
four-unit dwellings.

Figure 1: Evolution of Low and No Down Payment Products

          1944 1948 1957 1972 1974 1991 1994 1998 2000 2002 2003 2004

LTV: 95 FHA Fannie Freddie Mae Mac

Fannie Freddie97 FHA Mae Mac

100 Veterans Fannie Mae Affairs USDA Freddie Mac

Sources: VA, USDA, FHA, Fannie Mae, and Freddie Mac.

As shown in figure 2, a greater proportion of the FHA-insured and
VA-guaranteed mortgage loans had low down payments than was the case for
loans purchased by Fannie Mae and Freddie Mac. Further, the number of
loans FHA insured in 2000 that had LTVs greater than 95 percent exceeded
the total number of loans with such LTVs that were guaranteed by VA and
purchased by Fannie Mae and Freddie Mac combined.

Figure 2: Percentage of Home Purchase Mortgages with a LTV Ratio Higher
Than 95 Percent for HUD (FHA), VA, Fannie Mae, and Freddie Mac for Loans
Originated in 2000

Number of loans

1,200,000

1,000,000

800,000

600,000

400,000

200,000

0 HUD VA Fannie Mae Freddie Mac

0-95 LTV

>95 LTV

Sources: HUD and VA.

While relatively few loans purchased by Fannie Mae or Freddie Mac had low
or no down payments, in recent years the GSEs have purchased relatively
more of these loans than in the past. As shown in figure 3, both Fannie
Mae and Freddie Mac, during the years 1997-2000, acquired a higher
proportion of mortgages with a high LTV than in previous years. To do
this, they increased the number of product options available to borrowers
with limited down payment funds.

Figure 3: Percentage of Loans with LTV Ratios of 95 Percent or Higher That
Fannie Mae and Freddie Mac Purchased, 1997-2000

Percentage

6

5

4

3

2

1

0 1997 1998 1999 2000

Fannie Mae

Freddie Mac

Source: Harold Bunce, "GSE's Funding of Affordable Loans: A 2000 Update,"
Department of Housing and Urban Development, April 2002.

The mortgage industry is increasingly using mortgage scoring and automated
underwriting. During the 1990s, private mortgage insurers, the GSEs, and
larger financial institutions developed automated underwriting systems.
Mortgage scoring is a technology-based tool that relies on the statistical
analysis of millions of previously originated mortgage loans to determine
how key attributes such as the borrower's credit history, the property
characteristics, and the terms of the mortgage note affect future loan
performance. Automated underwriting refers to the process of collecting
and processing the data used in the underwriting process. FHA has
developed and recently implemented a mortgage scoring tool, called the FHA
TOTAL Scorecard, to be used in conjunction with existing automated
underwriting systems. More than 60 percent of all mortgages were being
underwritten by an automated underwriting system, as of 2002,

and this percentage continues to rise.13 The mortgage industry also uses
credit scoring models for estimating the credit risk of individuals-these
methodologies are based on information such as payment patterns.
Statistical analyses identifying the characteristics of borrowers who were
most likely to make loan payments have been used to create a weight or
score associated with each of the characteristics. According to Fair,
Isaac and Company sources, credit scores are often called "FICO scores"
because most credit scores are produced from software developed by Fair,
Isaac and Company.14 FICO scores generally range from 300 to 850 with
higher scores indicating better credit history. The lower the credit
score, the more compensating factors lenders might require to approve a
loan. These factors can include a higher down payment and greater borrower
reserves.

Characteristics and Standards of Low and No Down Payment Products Vary by
Mortgage Institution

The characteristics and standards for low and no down payment mortgage
products vary among mortgage institutions. Standards to determine a
borrower's eligibility differ from lender to lender. For example, one
mortgage institution might have a limit on household income where another
might not. Each of these mortgage products requires some form of borrower
investment. Most mortgage institutions use automated systems to underwrite
loans but differ on how they consider factors such as the borrower's
credit score and credit history. Finally, mortgage institutions also try
to mitigate the increased risk associated with these products by employing
tools like prepurchase counseling and greater insurance coverage.

13Susan Wharton Gates, Vanessa Gail Perry, and Peter Zorn, "Automated
Underwriting in Mortgage Lending: Good News for the Underserved," Housing
Policy Debate, 13, no. 2, 2002.

14Fair, Isaac and Company, "Understanding your Credit Score," July 2002.

Eligibility Standards Are Not Uniform throughout the Mortgage Industry

Each mortgage institution we studied limits in some way the mortgages or
the borrowers that may be eligible for their low and no down payment
products, but the specific limits and criteria differ among institutions.
Fannie Mae and Freddie Mac are constrained in the size of the mortgages
they may purchase. Specifically, the Housing and Community Development Act
of 1980 requires a limit (conforming loan limit) on the size of mortgages
that can be purchased by either Fannie Mae or Freddie Mac. In 2005, the
conforming mortgage limit for Fannie Mae and Freddie Mac is $359,650 for
most of the nation.15 FHA is also limited in the size of mortgages it may
insure. The FHA loan limit varies by location and property type, depending
on the cost of homes in an area and the number of units in a property.
Thus, FHA's loan limit may be as high as 87 percent of the conforming loan
limit, or $312,895 in 2005; or as low as 48 percent of the conforming loan
limit, or $172,632 in 2005. In addition, FHA also has higher limits in
Alaska, Hawaii, Guam, and the U.S. Virgin Islands because these are
considered to be high cost areas. Although VA does not have a mortgage
limit, lenders generally limit VA mortgages to four times the VA guaranty
amount, which is now set at 25 percent of the conforming loan limit. Since
the maximum guaranty currently is legislatively set at $89,913,
VA-guaranteed mortgages will rarely exceed $359,650.

15The conforming loan limit is 50 percent higher for Alaska, Hawaii, Guam,
and the U.S. Virgin Islands.

Moreover, while FHA does not restrict eligibility to borrowers with
certain income, other mortgage institutions may limit eligibility by
borrower income and other measures. Most state housing finance agencies
target their low and no down payment products to first-time homebuyers.16
Some mortgage institutions providing affordable low and no down payment
products also limit the loans to households with income at or below area
median levels. For example, USDA's RHS, in its section 502 Guaranteed Loan
program, does not guarantee loans to individuals with incomes exceeding
115 percent of the area median income or 115 percent of the median family
income of the United States. We also found that Web sites of many state
housing finance agencies show that their mortgage products include income
limits as well as sales price limits and in some cases designated
"targeted areas" within a state.17 Table 1 illustrates some of the major
similarities and differences in the eligibility criteria of FHA and other
mortgage institutions.

16Often state housing finance agencies define first-time homebuyers as
individuals who, during the past three consecutive years have not had
ownership in a primary residence.

17State housing finance agencies have also been actively involved in low
and no down payment mortgage lending. Using primarily mortgage revenue
bonds that are sold to investors, they are able to originate, fund, or
self-insure below-market interest rate mortgages.

  Table 1: Eligibility Limitations of FHA Compared with RHS, VA, and Selected
                     Products of Fannie Mae and Freddie Mac

                                                      Fannie Mae  
                                                     MyCommunity- Freddie Mac 
                                                       Mortgage    Affordable 
                                                                         Gold 
     Eligibility    FHA        RHS            VA     program(TM)   programs   
       criteria                                                   
                           Resident in    Veteran or                          
    Borrower type   N/Aa                    active       N/A          N/A
                         rural-designated    duty                 
                               area                               

Income  N/A    Cannot exceed 115 N/A   Cannot exceed 100 Cannot exceed 100 
                  percent of area         percent of area    percent of area  
                   median income or        median incomeb,c median incomeb,c  
                  the median family                         
                 income of the U.S.                         

Property   1-4 unit,       1-unit,       1-4 unit,      1-4 unit,       1-unit,     
  type                                                               
         owner-occupied owner-occupied owner-occupied owner-occupied owner-occupied 
                        principal      principal      principal           principal 
           principal    residence      residence      residence           residence 
           residenced                                                

  Mortgage type Up to 30-year 30-year fixed rate Up to 30-year fixed Up to 30
                        years fixed Up to 30-year fixed

fixed rate or adjustable rate mortgages (ARM)

                         rate or ARMs rate or ARMs rate

Legend

N/A = not applicable

Sources: HUD, VA, USDA, Fannie Mae, and Freddie Mac.

aWhile FHA does not set specific requirements for the type of borrower, it
tends to serve low-income and first-time homebuyers.

bCan be higher in high cost areas.

c Waiver of income limit may apply in targeted areas.

dUnder certain circumstances FHA may insure loans on second residences,
investment properties, and properties owned by nonprofits and state and
local governments.

Fannie Mae and Freddie Mac affordable mortgage products primarily target
low-to-moderate income and first-time homebuyers. Freddie Mac and RHS
allow a borrower to purchase a home containing one unit, while FHA, VA,
and Fannie Mae allow a borrower to purchase properties that have up to
four units with one mortgage. VA stipulates that if the veteran must
depend on rental income from the property to qualify for the mortgage, the
borrower must show proof that he or she has the background or
qualifications to be successful as a landlord and have enough cash
reserves to make the mortgage payments for at least 6 months without help
from the rental income. With regard to mortgage type, many mortgage
institutions permit 30-year fixed-rate mortgages. Some also permit
adjustable rate mortgages (ARM).

Most Low and No Down Payment Products Require Some Form of Borrower
Investment

Most low and no down payment mortgage products require some form of
borrower investment, either a borrower contribution or cash reserve, as a
way of reducing risk and assuring that the borrower has a stake in the
property. Low down payment products offered by FHA, Fannie Mae, Freddie
Mac and private insurers require a cash investment of at least 3 percent
from the borrower. No down payment mortgage products offered by VA, RHS,
Fannie Mae, Freddie Mac, and some private insurers require either no down
payment or a minimum amount (such as $500 in Fannie Mae's
MyCommunityMortgage program).

Many institutions permit down payment assistance. FHA stipulates that the
gift donor may not be a person or entity with an interest in the sale of
the property, such as the seller, real estate agent or broker, builder, or
entity associated with them. FHA mortgagee letters state that "gifts from
these sources (seller, builder, etc.) are considered inducements to
purchase and must be subtracted from the sales price." However, FHA allows
nonprofit agencies that may receive contributions from the seller to
provide down payment assistance to the borrower. In contrast, Fannie Mae,
Freddie Mac, and some of the private insurers generally do not allow down
payment funds, either directly or indirectly, from an interested or
seller-related party to the transaction. Fannie Mae and Freddie Mac
officials told us that such seller-related contributions could contribute
to an overvaluation of the price of the property.

Even where borrowers pay no down payment they very often must pay a
minimum percentage of closing costs from their own funds.18 FHA requires
that borrowers pay 3 percent of the total loan amount toward the purchase
of the home. This contribution may be used for down payment or closing
costs. Thus, FHA borrowers may finance closing costs, within limits. FHA
borrowers may also finance their insurance premium. Unlike FHA, some
mortgage institutions do not allow financing of the closing costs and the
insurance premiums in the first mortgage. VA generally allows payment of
all closing costs to be negotiated while restricting those that may be
charged to the borrower. VA allows borrowers to finance their insurance
premium, called the funding fee. In the section 502 Guaranteed Loan
program for RHS, borrowers may pay closing costs but they are not required
to do so and may be allowed to finance the closing costs and their

18Closing costs could include a loan origination fee, a mortgage
recordation fee, a title transfer tax, appraisal fees, attorney fees, and
title insurance.

insurance premium, called the Guarantee Fee.19 Freddie Mac in its no down
payment product requires a 3 percent borrower contribution to be used for
closing costs, financing costs, or prepaids and escrows, all of which can
come from gifts or property seller contributions.

FHA, RHS, VA, Fannie Mae, and Freddie Mac differ somewhat in terms of
their maximum allowable LTV ratios and how they calculate this ratio. LTV
ratios are important because of the direct relationship that exists
between the amount of equity borrowers have in their homes and the
likelihood of risk of default. The higher the LTV ratio, the less cash
borrowers will have invested in their homes and the more likely it is that
they may default on mortgage obligations, especially during times of
economic hardship.

The Omnibus Budget Reconciliation Act of 1990 (Pub. L. No. 101-508),
established LTV limits for FHA-insured mortgages of 98.75 percent if the
home value is $50,000 or less, or 97.75 percent if the home value is in
excess of that. However, because FHA allows financing of the up-front
insurance premium, borrowers can receive a mortgage with an effective LTV
ratio of close to 100 percent.

In table 2, we calculate the effective LTV ratio for selected low and no
down payment products. The example assumes a $100,000 purchase price
(appraisal value) and a 30-year fixed-rate mortgage. It also assumes
average closing costs of about 2.1 percent of sales price. FHA has a
formula to calculate the maximum loan amount based on a percentage of the
purchase price of the home. FHA does not have a down payment requirement
but instead has what FHA calls a minimum cash investment requirement. This
investment requirement can be used to pay either the down payment and in
some cases the closing costs. Not shown are the actual out-of-pocket
expenses to the borrower which could vary based on the individual
transaction and whether the investment requirement was split among the
closing costs and down payment, as well as whether the borrower opted to
finance their up-front premium.20

19According to USDA officials, a borrower may finance closing costs and
the Guarantee Fee as long as they do not exceed the property's appraised
value.

20Out-of-pocket expenses can include expenses such as funds required to
establish an escrow account.

    Table 2: LTV Calculations for FHA, RHS, VA, Fannie Mae, and Freddie Mac
                   Government Conventional Mortgage elements

FHAa 203b

                                                  USDA 502 guaranteed program

VA zero down

Fannie Mae and Freddie Mac 100 LTV products

              Purchase price            $100,000  $100,000  $100,000 $100,000 
       Loan amount before up-front       97,750b   100,000  100,000       N/A 
                insurance                                            
Plus up-front insurance premium/fee   +1,466c   +2,000d  +2,150e       N/A 
                                         (1.5%)        (2%) (2.15%)  
              Total mortgage             99,216f   102,000  102,150   100,000 
           Effective LTV ratiog           99.2%     102%      102%       100% 

Sources: HUD, VA, USDA, Fannie Mae, and Freddie Mac.

aThis is the existing FHA mortgage insurance product that requires the
least amount of a down payment. This product is not a "no down" mortgage
product.

bWe are assuming, for this example, that the mortgage is in a state with
average closing costs of above 2.1 percent of sales price. In this case,
the maximum mortgage (not including a financed up-front insurance premium)
would be $100,000 x 97.75 percent.

cUp-front insurance premium = $97,750 x 1.5 percent = $1,466.

dGuarantee fee = $100,000 x 2 percent = $2,000. Loans can be guaranteed up
to 102 percent LTV if doing so is necessary to allow the 2 percent
guarantee fee to be financed by the borrower. A 100 percent LTV threshold
may only be exceeded to allow the guarantee fee to be financed. If a
borrower chooses not to finance the guarantee fee, loans are limited to
100 percent LTV. Closing costs are allowed but closing costs may not be
financed if the borrower is already financing the guarantee fee such that
they have reached the maximum allowed LTV of 102 percent, not including
closing costs.

eFunding fee = $100,000 x 2.15 percent = $2,150.

fThe borrower is required to pay 3 percent (of the contract sales
price-called a minimum cash investment requirement) toward closing costs
and down payment.

gCalculation = total mortgage / purchase price.

In addition, some of the affordable conventional mortgage products allow
for subordinate financing in the form of secondary mortgages to pay for a
down payment and/or closing costs. These secondary mortgages allow for a
total effective LTV of up to 105 percent.

Some Underwriting When underwriting mortgages, FHA and other mortgage
institutions Standards Differ between require that lenders examine a
borrower's ability and willingness to repay FHA and Other Mortgage the
mortgage debt. Lenders for low and no down payment mortgages may

use automated underwriting systems examining the borrower's credit score

Institutions 	or creditworthiness, qualifying ratios, and cash reserves.
In some cases, they use manual underwriting to accommodate nontraditional
credit histories. By screening the majority of applications with automated

systems, underwriters have more time to review special cases with manual
underwriting.

Many mortgage institutions use credit scores in assessing mortgage
applicants through their automated underwriting systems. For standard
products, institutions tend to rely on automated underwriting, which
develops a mortgage score based on various factors including credit score
and, based on this, they make a decision on the loan. However, in some
instances, mortgage institutions set credit score minimums for some low
and no down payment products. In some instances, these credit score
minimums exist within the automated underwriting system. In other
instances, the credit score minimums exist only in products that are
underwritten using manual underwriting. FHA does not require a credit
score minimum, nor do VA and RHS. These three governmental agencies
examine the overall pattern of credit behavior rather than rely on one
particular credit score. All three agencies allow a good deal of judgment
and interpretation on the part of the underwriters in determining the
creditworthiness of the prospective borrower. Fannie Mae does not use
externally derived credit scores for its loan products that use automated
underwriting but instead relies on the credit history of the borrower.
Based on a review of Web sites of private mortgage insurers, products with
no down payment that are insured by these private mortgage insurers have
minimum credit score requirements ranging from 660 to 700. Individual low
and no down payment products that use credit score minimums use a variety
of cutoff scores. Many mortgage industry sources consider borrowers with
credit scores of 720 or higher as having excellent credit. One study that
focused on issues related to homeownership and cited extensive interviews
with leading experts in government and industry found that mortgage
applicants with scores above 660 are likely to have acceptable credit.21
On the other hand, for applicants with FICO scores between 620 and 660,
mortgage institutions typically perform more careful underwriting,
scrutinizing many factors. FICO scores under 620 indicate higher risk and
are unlikely to be approved by conventional lenders unless accompanied by
compensating factors.

Some of these mortgage institutions may, under some circumstances, accept
a lower credit score, if the borrower provides additional compensating
factors (such as 2 months cash reserve) that would indicate

21Michael Collins, "Pursuing the American Dream: Homeownership and the
Role of Federal Housing Policy," January 2002.

a lower risk on the part of the borrower. Mortgage institutions might also
accept a lower credit score if they were receiving additional compensation
for the risk, such as a mortgage originator receiving a higher interest
rate or a mortgage insurer getting a higher insurance premium. Some
mortgage institutions state in their underwriting guidance that FICO
scores together with the LTV determine in part the borrower's minimum
contribution. For example, one private mortgage insurer allows borrowers
with credit scores equal or greater than 700 to have a minimum borrower
contribution of 0 percent on a 100 LTV loan. For this same insurer, a
borrower with a credit score between 660 and 699 would have a minimum
borrower contribution of 3 percent on a 100 LTV loan.

Many mortgage institutions use two qualifying ratios as factors in
determining whether a borrower will be able to meet the expenses involved
in homeownership. The "housing-expense-to-income ratio" examines a
borrower's expected monthly housing expenses as a percentage of the
borrower's monthly income, and the "total-debt-to-income ratio" looks at a
borrower's expected monthly housing expenses plus long-term debt as a
percent of the borrower's monthly income. Lenders who do business with
Fannie Mae or Freddie Mac place more emphasis on the total-debt-to-income
ratio. Total debt includes monthly housing expenses and the total of other
monthly obligations, such as auto loans, credit cards, alimony, or child
support. The guidelines for manual underwriting are discussed below;
automated underwriting systems weight the qualifying ratios, as well as
numerous other factors, in assessing the borrower's ability to meet the
expenses involved in homeownership.

Unless there are compensating factors, FHA monthly
housing-expenseto-income ratio is set at a maximum of 29 percent, while
the monthly "total-debt-to-income ratio" is, at most, 41 percent of the
borrower's stable monthly income. The requirements set by Fannie Mae,
Freddie Mac, and the private insurers on the monthly housing
expense-to-income ratio vary greatly. Some have set lower thresholds, such
as Freddie Mac, which uses as a guideline that the monthly housing
expense-to-income ratio should not be greater than 25 percent to 28
percent, with exceptions for some products. Others, such as some private
insurers, have set higher thresholds than FHA has set, such as 33 percent.

Some mortgage institutions set thresholds on the "total-debt-to-income"
ratio that are lower than FHA's threshold. Conventional mortgages that are
manually underwritten to Fannie Mae or Freddie Mac standards are set at a
benchmark total-debt-to-income ratio of 36 percent of the borrower's
stable

monthly income, compared with FHA's 41 percent. However, Fannie Mae and
Freddie Mac state that they occasionally specify a higher allowable
debt-to-income ratio for a particular mortgage loan if compensating
factors are present.22

Cash reserves represent the amount of funds a borrower has after closing
on the loan. Generally the reserves required of borrowers are expressed in
terms of the numbers of monthly mortgage payments they may comprise.
Conceptually they represent the ability of the borrower to repay the
mortgage out of accumulated funds. Many mortgage institutions including
FHA consider it a compensating factor that reduces the risk of
delinquency. FHA, unlike conventional lenders who do business with the
GSEs and the private insurers, does not require cash reserves for its low
down payment product. VA and RHS also do not require cash reserves.
Generally the GSEs and the private insurers with whom we spoke required
cash reserves of either 1 or 2 months of monthly mortgage payments for low
and no down payment products.

Mortgage Institutions Use Various Risk Mitigation Tools

Some of the mortgage institutions we spoke with used various tools to
mitigate risk. For example, most mortgage institutions offering affordable
low and no down payment mortgages to first-time homebuyers require
prepurchase counseling, and some require postpurchase counseling. These
include lenders working with Fannie Mae, Freddie Mac, private insurers,
and state housing finance agencies. Homeownership counseling for
first-time homebuyers takes a variety of forms. There are counseling
programs administered by government agencies, lenders, nonprofit
organizations, and the private insurers, among others. These programs are
delivered through many different avenues including classroom, home study,
individual counseling, and telephone. The content of the counseling
programs also varies significantly across each of these administrative and
delivery mechanisms, as does the timing of the counseling-which can be
either prior to closing or postpurchase (when the borrower becomes
delinquent on a payment).

More specifically, Freddie Mac in each of its Affordable Gold products
(intended for first-time homebuyers who generally earn 100 percent or less
of area median income) requires that at least one qualifying borrower in
the

22Fannie Mae officials noted that Fannie Mae's automated underwriting
system allows for a higher total-debt-to-income ratio and factors the
ratio in its evaluation of the loan.

transaction must receive prepurchase counseling. Lenders must document the
organization that administered the counseling and how the counseling was
delivered. Freddie Mac exempts those borrowers who have cash reserves
after closing equal to at least two monthly mortgage payments from the
counseling requirement. Similarly, Fannie Mae in its MyCommunityMortgage,
requires prepurchase counseling for first-time homebuyers when they are
purchasing a one-unit property. If they are purchasing a two to four unit
property, landlord counseling is required. Fannie Mae also requires
postpurchase counseling for borrowers under certain low down payment
programs who become delinquent on their payments early in the mortgage.

Some private insurers require pre-and postpurchase counseling, but some
only recommend it. For example, two private insurers require pre-and
postpurchase counseling with all of its affordable low and no down payment
products, and they provide most of this counseling themselves. On the
other hand, another private insurer recommends, but does not require,
prepurchase counseling for first-time homebuyers in its low and no down
payment products. However, this insurer's underwriting guidance states
that prepurchase counseling is considered a positive underwriting factor.
It also recommends postpurchase counseling, particularly for borrowers who
are experiencing financial difficulties but have a good chance of
overcoming their financial problems and maintaining homeownership.

FHA, unlike most low and no down payment mortgage institutions serving
affordable first-time homebuyers, does not require prepurchase counseling.
VA also does not require prepurchase counseling, but considers it to be a
compensating factor in improving creditworthiness. RHS encourages lenders
to offer or provide for homeownership counseling and lenders may require
first-time homebuyers to undergo such counseling if it is reasonably
available in the local area.

FHA, VA, RHS, and the private insurers also differ in the amount of
insurance or guaranty they provide to protect lenders against the losses
associated with mortgages that go to foreclosure. While FHA essentially
protects against almost 100 percent of the losses associated with a
foreclosed mortgage, VA, RHS and the private insurers protect against a

portion of the loss.23 Private insurers generally provide protection to
lenders for only a portion of losses. This protection is usually expressed
as a percentage of the claim amount. For example, an insurer may provide
insurance coverage of 30 percent. This means that the insurer will cover
losses up to 30 percent of the claim amount. In exchange for offering this
insurance, the insurer charges borrowers a premium.

Some of the insurers with whom we spoke, as well as the GSEs, noted that
they require higher insurance coverage for mortgages with lower down
payments. For example, one insurer said that the amount of insurance
coverage tends to be 35 percent for no down payment mortgages, in contrast
to 30 percent for low down payment mortgages. Private insurers noted that
they charge higher premiums or require more stringent underwriting when
they provide higher insurance coverage. For example, one private insurer
stated that its monthly premium rates to a borrower increase about 15
percent for every 5 percentage point increase in insurance coverage
between 20 and 35 percent.

Research Shows LTV Ratio and Credit Score Are Important When Estimating
Risk of Individual Mortgages

Economic research we reviewed indicated that LTV ratios and credit scores
are among the most important factors when estimating the risk level
associated with individual mortgages.24 We identified and reviewed 45
papers that examined factors that could be informative.25 Of these, 37
examined if the LTV ratio was important and almost all of these papers
(35) found the LTV ratio of a mortgage important and useful. Nineteen
research papers evaluated how effective a borrower's credit score was in
predicting loan performance, and all but one reported that the credit
score was

23FHA covers 100 percent of the mortgage balance but does not cover all of
the costs of the foreclosure.

24Research we reviewed includes articles, reports, and papers that were
made available to us from economic journals, the internet, libraries, or
were provided to us by various entities (e.g., HUD, Fannie Mae, Freddie
Mac). For the purposes of this report, we refer to these documents as
"papers."

25Many papers employ multiple models to analyze the importance of
variables; as a result, summing the number of papers that found a variable
important and the number of papers that found a variable not important
will not equal the total number of papers that analyzed the importance of
a specific variable. For example, two of the papers that assessed the
importance of LTV found it important in some circumstances but not in
others.

important and useful.26 In addition, a number of the papers reported that
other factors were useful when estimating the risk level. For example,
characteristics of the borrower-such as qualifying ratios-were cited in
several of the papers we reviewed. Finally, other research evaluated
additional factors; however, we identified very few papers that
investigated the same variables or corroborated these findings.
Collectively, the research we reviewed appeared to concur that considering
multiple factors was important and useful in estimating the risk level of
individual mortgages. For example, some of the papers (7) reported that
considering LTV ratio and credit score concurrently was important and
useful when estimating the risk level of individual mortgages.27

LTV Ratio Helps Estimate Risk of Individual Mortgages

Many studies found that a mortgage's LTV ratio was an important factor
when estimating the risk level associated with individual mortgages. In
theory, LTV ratios are important because of the direct relationship that
exists between the amount of equity borrowers have in their homes and the
likelihood of risk of default. The higher the LTV ratio, the less cash
borrowers will have invested in their homes and the more likely it is that
they may default on mortgage obligations, especially during times of
economic hardship (e.g., unemployment, divorce, home price depreciation).
And, according to one study, "most models of mortgage loan performance
emphasize the role of the borrower's equity in the home in the decision to
default."28 We identified 45 papers that examined the relationship between
default and one or more predictive variables; of these, 37 examined if LTV
ratio was important and useful. Almost all of these papers (35) determined
that LTV ratio was effective in predicting loan performance-specifically,
when predicting delinquency, default, and foreclosure. Several papers
reported that there was a strong positive relationship between LTV ratio
and default. Specifically, one paper reported that the default rates for
mortgages with an LTV ratio above 95 percent were three to four times
higher than default rates for mortgages

26Of the 45 papers identified, 13 identified both LTV ratio and credit
score as important and useful when estimating the risk level associated
with individual mortgages.

27Of the papers we reviewed, researchers used several measures of loan
performance (e.g., default, delinquency, severity). Please see each paper
for the particular loan performance measure it used.

28Robert B. Avery, Raphael W. Bostic, Paul S. Calem, and Glenn B. Canner.
"Credit Risk, Credit Scoring, and the Performance of Home Mortgages,"
Federal Reserve Bulletin (July 1996).

with an LTV ratio between 90 to 95 percent.29 Another paper found that, at
the end of 5 years, the cumulative probability of default risks for
mortgages with an LTV ratio less than 95 percent was 2.48 percent;
however, the cumulative probability of default for mortgages with an LTV
ratio greater than or equal to 95 percent was 3.53 percent.30 While the
majority of the empirical research found that LTV ratio mattered, 4 of the
research efforts did not find that LTV ratio is important when estimating
the risk level associated with individual mortgages.31 For example, one
paper found that, for subprime loans, delinquency rates were relatively
unaffected by the LTV ratio.32 Generally, subprime loans are loans made to
borrowers with past credit problems at a higher cost than conventional
mortgage loans. Additionally, some (7) research efforts examined the
relationship between the LTV ratio and severity (losses), and all found
that there was a positive relationship between the LTV ratio and severity.
For a detailed list of the economic research that addresses the
relationship between LTV ratio and mortgage performance, see appendix II.

Credit Score Helps Estimate Despite the relatively recent use of credit
score information in the
Risk of Individual mortgage industry, several studies found that credit
score was an important
Mortgages and useful factor when estimating the risk level associated with
individual

mortgages.33 In general, credit scores represent a borrower's credit
history.

29Yongheng Deng, John M. Quigley, and Robert Van Order, Mortgage Default
and Low Downpayment Loans: The Costs of Public Subsidy, National Bureau of
Economic Research: Working Paper No. 5184 (Cambridge, Mass.: July 1995).

30Yongheng Deng and Stuart Gabriel, Modeling the Performance of
FHA-Insured Loans: Borrower Heterogeneity and the Exercise of Mortgage
Default and Prepayment Options, a report submitted to the Office of Policy
Development and Research, U.S. Department of Housing and Urban
Development, May 2002.

31Generally, the studies that found the LTV ratio to be important had
sample sizes greater than 6,000 (and in some cases in the millions). In
comparison, three of the four studies that did not find LTV to be
important had smaller sample sizes and the fourth study found LTV to be
important for the prime market, but not for the subprime market.
Additionally, one study used national aggregate data rather than loan
level data. This study found LTV important in some specifications, but not
in others.

32Amy Crews Cutts and Robert Van Order, "On the Economics of Subprime
Lending." Freddie Mac Working Paper Series #04-01(January 2004)
(http://freddiemac.com/corporate/reports/).

33Generally, the mortgage industry began widely using credit score
information in the late 1990s; therefore, considering credit score in
empirical loan performance analysis is very recent. Further, there was a
particularly strong housing market during this period.

Credit histories consist of many items, including the number and age of
credit accounts of different types, the incidence and severity of payment
problems, and the length of time since any payment problems occurred. The
credit score reflects a borrower's historic performance and is an
indication of the borrower's ability and willingness to manage debt
payments. Of the 45 papers we reviewed, 19 evaluated how effective a
borrower's credit score was in predicting loan performance. Eighteen
research efforts evaluated how effective a borrower's credit score was in
predicting delinquency, default, and foreclosure; all of these efforts
found that a borrower's credit score was important. Generally, the papers
reported that higher credit scores were associated with lower levels of
defaults. Specifically, one study found that a mortgage with a credit
score of 728 (indicating an applicant with excellent credit) had a default
probability of 1.26 percent, while a mortgage with a credit score of 642
had a default probability of 3.41 percent-or more than two times higher.34
Additionally, four research efforts examined the relationship between
credit score and severity (losses), and three reported that there was a
negative relationship between credit score and severity. For example, one
study found that credit scores were also helpful in predicting the amount
of losses resulting from foreclosed mortgages. In particular, the paper
reported the loss rate for defaulted mortgages with high credit scores was
lower than foreclosed mortgages with low credit scores.35 For a list of
the economic research, that we reviewed, that addresses the relationship
between credit score and mortgage performance, see appendix II.

Other Factors May Help in Estimating the Risk of Individual Mortgages

Many of the papers we reviewed identified factors that, in addition to LTV
ratios and credit scores, were important and useful determinants of credit
risk for home mortgages. Of these, the most widely analyzed
factor-accumulation of equity in the home-was a subject of 26 studies we
reviewed. Some factors were the subject of far fewer papers. Yet other
factors were the subject of a single paper only. The most widely assessed
factors included borrower characteristics such as accumulation of equity
in

34Robert F. Cotterman. Analysis of FHA Single-family Default and Loss
Rates, a report submitted to the Office of Policy Development and
Research, U.S. Department of Housing and Urban Development, March 25,
2004.

35Cotterman. Analysis of FHA Single-family Default and Loss Rates.

the home, qualifying ratios, and income.36 Additionally, characteristics
of the area in which the property was located included variables such as
unemployment rates and income levels. Finally, characteristics of the
mortgage included variables such as mortgage age and term of the mortgage
(e.g., 15 year vs. 30 year). The extent to which the authors agreed on the
importance of the other factors varied. For example, nearly all of the
papers that looked at equity accumulation (a factor which is not known at
the time of loan origination), the unemployment rate of the area in which
the property is located, and mortgage age, found that these factors were
important. However, the research was less certain as to the importance of
qualifying ratios and income. That is, several of the papers found that
the qualifying ratios and income were important in estimating risk;
however, some found that qualifying ratios and income were not an
important factor.

The economic research we reviewed also indicated that considering factors
in combination was helpful in estimating the risk level of individual
mortgages. Of all 45 papers we reviewed, more than half conducted
multivariate analyses. For example, seven studies found that using credit
score information in combination with the LTV ratio was helpful in
estimating the risk level of individual mortgages. Specifically, one study
found that the "foreclosure rate is particularly high for borrowers with
both low credit scores and high LTV ratios-almost 50 times higher than
that for borrowers with both high credit scores and low LTV ratios."37
Other studies examined several aspects of a mortgage concurrently. For
example, in one study, the authors controlled for certain loan
characteristics, such as credit history and LTV, and they found that
borrower income is useful in estimating risk levels of mortgages.38 In
another study, the authors controlled for house price appreciation (10
percent) and unemployment rates (8 percent) and examined loan
performance-after 15 years-in terms of LTV ratio and a borrower's relative
income. Regardless of income, default was higher for zero down payment
mortgages. Specifically, under

36Qualifying ratios evaluated in the studies we identified include ratios
such as payment-to-income, debt-to income, personal savings as percentage
of disposable income, and household liabilities divided by household
assets.

37Avery, Bostic, Calem, and Canner. "Credit Risk, Credit Scoring, and the
Performance of Home Mortgages."

38Robert Van Order and Peter Zorn. "Performance of Low-Income and Minority
Mortgages," A paper prepared for the Joint Center for Housing Studies'
Symposium on Low-Income Homeownership as an Asset-Building Strategy,
Working Paper: LIHO-01.10 (September 2001).

these conditions, the authors reported that zero down payment mortgages of
borrowers with incomes below 60 percent of the metropolitan statistical
area's (MSA) median level would have cumulative default rates about twice
as high as mortgages that required a 10 percent down payment made to
borrowers with similar incomes. Similarly, the zero down payment mortgages
of borrowers with incomes greater than one-and-a-half times the MSA's
median level would have cumulative default rates about 50 percentgreater
than mortgages that required a 10 percent down payment made to borrowers
with similar incomes.39

Our Analysis Indicated That Mortgages with Higher LTV Ratios and Lower
Credit Scores Pose Greater Risks

Consistent with studies we reviewed, our analysis of FHA and conventional
mortgage data indicated that mortgages with high LTV ratios (smaller down
payments) and low credit scores generally are riskier than mortgages with
low LTV ratios and high credit scores.40 For example, FHA-insured
mortgages with LTV ratios greater than 80 percent and low credit scores
(below 660) had a default rate above the FHA average default rate.
Similarly, conventional mortgages with LTV ratios greater than 80 percent

39Deng, Quigley, and Van Order, Mortgage Default and Low Downpayment
Loans: The Costs of Public Subsidy.

40The data used in the analysis include a sample of FHA-insured mortgages
originated in calendar years 1992, 1994, and 1996 and conventional
mortgages originated in calendar years 1997, 1998, and 1999. These data
represent the most recently available data for each entity that includes
variables necessary to conduct the analysis (such as LTV ratio and credit
score). We did not include mortgages guaranteed by the USDA and VA in our
analysis because credit score information for these mortgages was not
readily available. Fannie Mae and Freddie Mac provided the conventional
data.

In our analysis, we specified six LTV ratio categories and six credit
score categories. We defined default as a credit event that includes
foreclosed mortgages, as well as mortgages that did not experience
foreclosure, but that would typically lead to a credit loss, such as a
"short sale" or a "third party sale" termination of the mortgage.

We calculated average 4-year default rates (by dollar amount) for
FHA-insured and conventional mortgages within specified LTV ratio and
credit score categories. The average 4-year default rates for each LTV
ratio and credit score categories were calculated as follows: the total
dollar amount of mortgages originated (in all three cohorts) and defaulted
within 4 years of origination, divided by the total dollar amount of
mortgages originated (in all three cohorts). We then classified the
default rates for each LTV ratio and credit score category relative to the
average default rate for the FHA-insured and conventional mortgages,
respectively. The relative default rates for each LTV ratio and credit
score category were calculated as follows: the average default rate for
each category, divided by the average default rate for all FHA-insured or
conventional mortgages as appropriate. For a more detailed description of
our analysis, see appendix I: Scope and Methodology.

and somewhat low credit scores (below 700) had a default rate above the
conventional average default rate. While this analysis is useful in
determining the extent to which LTV ratios and credit scores can help
predict the risk level associated with individual mortgages, care should
be taken when comparing the FHA with the conventional relative default. In
particular, the relative default rates are derived from different calendar
years (that is, a sample of FHA mortgages insured in 1992, 1994, and 1996
and conventional mortgages originated in 1997, 1998, and 1999 and
purchased by Fannie Mae or Freddie Mac). Also the average default rate for
FHA-insured mortgages is higher than the average default rate for
conventional mortgages.

FHA-Insured Mortgages with Higher LTV Ratios and Lower Credit Scores Are
Riskier Than Other FHA-Insured Mortgages

When considering LTV alone, FHA-insured mortgages with higher LTV ratios
(smaller down payments) generally perform worse than FHA-insured mortgages
with lower LTV ratios. As figure 4 illustrates, our analysis indicates
that the incidence of default increases as LTV ratios increase. When
considering the LTV ratio alone, the default rate for sampled FHA-insured
mortgages, with an LTV of 70 percent or less, is no more than half the
average FHA default rate. In contrast, the default rate for mortgages with
LTV ratios greater than 90 percent, as a group, surpasses the average FHA
default rate. For the highest LTV ratio group-greater than 97 to 100
percent-the default rate for these mortgages is about 1.75 times the
average FHA default rate.

Figure 4: Four-year Relative Default Rates by LTV Ratio for FHA-Insured
Mortgages (1992, 1994, and 1996)

Source: GAO analysis of FHA data.

Note: Because of the sensitive nature of the data, we have chosen to
illustrate relative 4-year default rates for each LTV ratio and credit
score category. These relative 4-year default rates are defined as
follows: the 4-year default rate for each category, divided by the average
4-year default rate for a sample of mortgages insured by FHA in 1992,
1994, and 1996. For example, if the 4-year default rate for a particular
LTV ratio and credit score category was 3 percent, and the 4-year default
rate for all FHA mortgages sampled was 2 percent, the relative 4-year
default rate for this category would be 3 divided by 2, or 1.5 times the
average 4-year default rate. To generate the average 4-year default rate,
we merged mortgage volume and performance data for the sample of FHA
mortgages insured in 1992, 1994, and 1996. Loan data are measured in
dollars.

FHA-insured mortgages with lower credit scores generally perform worse
than FHA-insured mortgages with higher credit scores, regardless of LTV
ratio. As figure 5 illustrates, our analysis indicated that the incidence
of default increases as credit scores decrease. Considering the credit
score, the default rate for sampled FHA-insured mortgages with credit
scores 700 and above is no more than half the average FHA default rate for
all sampled mortgages. The default rate for mortgages with a credit score
below 660, as a group, surpasses the average FHA default rate. For the
lowest credit

score group-less than 620-the default rate for these mortgages is almost
twice the average FHA default rate.

Figure 5: Four-year Relative Default Rates by Credit Score for FHA-Insured
Mortgages (1992, 1994, and 1996)

Source: GAO analysis of FHA data.

Note: For description of relative default rates, please see note with
figure 4.

As expected, FHA-insured mortgages with both high LTV ratios (smaller down
payments) and low credit scores generally perform worse than mortgages
with both low LTV ratios and high credit scores. Our analysis indicates
that the incidence of default increases as LTV ratios increase and credit
scores decrease. As figure 6 illustrates, mortgages with lower LTV ratios
and higher credit scores (those at the bottom of the figure) have lower
default rates than mortgages with higher LTV ratios and lower credit
scores (at the top of the figure). FHA-insured mortgages with LTV ratios
greater than 80 percent and low credit scores (below 660) had a default
rate above the FHA average default rate. FHA-insured mortgages, with LTV

ratios greater than 90 percent and credit scores below 620, had a default
rate more than double the FHA average.

Figure 6: Four-year Relative Default Rates by LTV Ratio and Credit Score
for FHA-insured Mortgages (1992, 1994, and 1996)

Source: GAO analysis of FHA data.

Notes: For description of relative default rates, please see note with
figure 4.

We do not present relative default rates for categories with fewer than
1,000 observations as the performance information may not be reliable when
there are too few observations. In the figure, these instances are noted
as "N/A." For a more detailed description of our analysis, see appendix I.

Conventional Mortgages with Higher LTV Ratios and Lower Credit Scores Are
Riskier Than Other Conventional Mortgages

Generally, the performance relationships that exist for FHA-insured
mortgages also exist for conventional mortgages originated in the late
1990s. As figure 7 illustrates, our analysis indicates that conventional
mortgages with higher LTV ratios (smaller down payments) generally perform
worse than conventional mortgages with lower LTV ratios. When considering
LTV ratio alone, the default rate for the group of conventional mortgages
with LTV ratios below 80 percent was no more than half the average
conventional default rate. Generally, the default rates then increase with
higher categories of the LTV ratio. In fact, the default rate for
conventional mortgages with an LTV ratio greater than 90, but less than 97
percent, as a group, is more than twice the average conventional default
rate. One notable exception to this general pattern is that conventional
mortgages in the highest LTV ratio category (that is, greater than 97 to
100 percent) appear to have a lower risk of default than do conventional
mortgages in some of the lower LTV ratio categories. According to GSE
officials, this may be explained by a number of possible factors. The GSEs
had just begun to purchase an increasing number of mortgages with very
high LTV ratios during the late 1990s and the GSEs took steps to limit the
risks associated with these mortgages. For example, some of these loans
were part of negotiated deals with individual lenders. These negotiated
transactions may have required the use of manual underwriting and minimum
credit scores, and the GSEs may have used specific servicers for these
loans. GSE officials told us that lenders and servicers operating as part
of negotiated deals with them tend to be more conservative in their
approach to these loans. GSE officials also told us that the borrowers
during this time period would have been the very best segment of the
applicant pool. Agency officials indicate that, for more recent loans
where volumes are higher and lenders are reaching deeper into the
applicant pool, default rates on loans in these categories are higher than
they were in the 1997-1999 period and are now consistent with the
relationship we would expect between LTV and default rates. We discuss
these practices in greater depth later in the report.

Figure 7: Four-Year Relative Default Rates by LTV Ratio for Conventional
Mortgages (1997, 1998, and 1999)

Sources: GAO analysis of Fannie Mae and Freddie Mac data.

Note: Because of the sensitive nature of the data, we have chosen to
illustrate relative 4-year default rates for each LTV ratio category.
These relative 4-year default rates are defined as follows: the 4-year
default rate for each category, divided by the average 4-year default rate
for all conventional mortgages originated in 1997, 1998, and 1999 and
purchased by Fannie Mae or Freddie Mac. For example, if the 4-year default
rate for a particular LTV ratio category was 3 percent, and the 4-year
default rate for all conventional mortgages sampled was 2 percent, the
relative 4-year default rate for this category would be 3 divided by 2, or
1.5 times the average 4-year default rate. To generate the average
conventional 4-year default rate, we combined Fannie Mae and Freddie Mac
mortgage volume and mortgage performance data; we also combined the sample
data for mortgages originated in 1997, 1998, and 1999. Loan data are
measured in dollars.

When considering credit score alone, conventional mortgages with lower
credit scores generally perform worse than conventional mortgages with
higher credit scores. As figure 8 illustrates, our analysis indicates that
the incidence of default generally increases as credit scores decrease.
The average default rate for mortgages with credit scores of 740 and
higher is no more than 20 percent that of the average default rate for
conventional loans and loan performance declines for each lower category
of credit score. In fact, the default rate for mortgages with a credit
score below 700,

as a group, surpasses the average default rate. Ultimately, the average
default rate for the lowest credit score category (below 620) is more than
4 times the average conventional default rate.

Figure 8: Four-Year Relative Default Rates by Credit Score for
Conventional Mortgages (1997, 1998, and 1999)

Sources: GAO analysis of Fannie Mae and Freddie Mac data.

Note: For description of relative default rates, please see note with
figure 7.

As expected, conventional mortgages with both high LTV ratios (smaller
down payments) and low credit scores generally perform worse than
mortgages with both lower LTV ratios (larger down payments) and higher
credit scores. Our analysis indicates that the incidence of default
generally increases as LTV ratios increase and credit scores decrease. As
figure 9 illustrates, mortgages with lower LTV ratios and higher credit
scores (those at the bottom of the figure) have much smaller default rates
than mortgages with higher LTV ratios and lower credit scores (at the top
of the figure).

Specifically, as a group, mortgages with LTV ratios greater than 80
percent and credit scores below 700 have default rates greater than the
average conventional default rate. Further, conventional mortgages with
LTV ratios greater than 80 percent and credit scores below 660 had a
default rate more than twice the conventional average.

Figure 9: Four-Year Relative Default Rates by LTV Ratio and Credit Score
for Conventional Mortgages (1997, 1998, and 1999)

Sources: GAO analysis of Fannie Mae and Freddie Mac data.

Notes: For description of what we mean by relative default rates, please
see note with figure 7.

We do not present relative default rates for categories with fewer than
3,000 mortgages as the performance information may not be reliable when
there are too few observations. In the figure, these instances are noted
as "N/A." For a more detailed description of our analysis, see appendix I.

One notable exception to this general pattern is that the group of
conventional mortgages with the highest LTV ratios (that is, greater than
97 to 100 percent) appears to have a lower risk of default than do the
group of conventional mortgages with lower LTV ratios for loans originated
during these years. For example, of conventional loans with credit scores
of 740 and higher, those that had LTV ratios greater than 97 percent, as a
group, performed better than those with LTV ratios greater than 90 to 97
percent. Similarly, of conventional loans with credit scores below 620,
those with the highest LTV ratio performed better than those with LTV
ratios greater than 90 to 97 percent. This anomaly, where the highest LTV
mortgages appear to perform better than the lower LTV loans, may reflect
that the GSEs had just begun to purchase an increasing number of mortgages
with very high LTV ratios in the years we analyzed and that the GSEs took
steps to limit the risks associated with these mortgages. Likewise,
lenders may perform more rigorous underwriting when first originating a
new loan product.

While this analysis is useful in determining the extent to which LTV
ratios and credit scores are helpful in predicting the risk level
associated with individual mortgages insured by FHA and for mortgages
purchased by the GSEs during specific years, there are several reasons why
care should be taken when comparing the FHA with the conventional relative
default rates. The relative default rates are derived from different years
(that is, FHA mortgages insured in 1992, 1994, and 1996; and conventional
mortgages originated in 1997, 1998, and 1999 and purchased by Fannie Mae
or Freddie Mac). Also, the actual average default rate for FHA-insured
mortgages is higher than the actual average default rate for conventional
mortgages. Finally, the distribution among LTV categories for FHA-insured
loans and conventional loans differs. Generally, over half of the loans
that the GSEs purchase have LTV ratios at or below 80 percent. In
comparison, loans insured by FHA generally have LTV ratios greater than 95
percent.

Several Practices Mortgage Institutions Use in Designing and Implementing
Low and No Down Payment Products Could Be Instructive for FHA

Mortgage institutions we spoke with used a number of similar practices in
designing and implementing new products, including low and no down payment
products. Some of these practices could be helpful to FHA in its design
and implementation of new products. When considering new products,
mortgage institutions focused their initial efforts on identifying other
products with similar enough characteristics to their new product so that
data on these products could be used to understand the potential issues
and performance for the proposed product. Some mortgage institutions,
including FHA, said they may acquire external loan performance data and
other data when designing new products. Moreover, mortgage institutions
often establish additional requirements for new products such as
additional credit enhancements or underwriting requirements. FHA has less
flexibility in imposing additional credit enhancements but it does have
the authority to seek co-insurance, which it is not currently using. FHA
makes adjustments to underwriting criteria and to its premiums, but is not
currently using any credit score thresholds. Mortgage institutions also
use different means to limit how widely they make available a new product,
particularly during its early years. FHA does sometimes use practices for
limiting a new product but usually does not pilot products on its own
initiative, and FHA officials question the circumstances in which they can
limit the availability of a program and told us they do not have the
resources to manage programs with limited availability. According to
officials of mortgage institutions, including FHA, they also often put in
place more substantial monitoring and oversight mechanisms for their new
products including lender oversight, but we have previously reported that
FHA could improve oversight of its lenders.

Mortgage Institutions Initially Analyze the Risk of Products Similar to
the Product They Are Seeking to Develop

Mortgage institutions, such as Fannie Mae, Freddie Mac, the private
mortgage insurers, and FHA first identify what information, including
data, they already have that would allow them to understand the
performance of a potential product. When these institutions do not have
sufficient data, they may purchase external data that allows them to
conduct their own analysis of loans that are related to a type of loan
product that they are considering. For example, Freddie Mac purchased
structured transactions of Alt A and subprime loans in order to learn more
about the underwriting characteristics and performance of high LTV and low
credit score loans.41 Freddie Mac officials reported that these data were
very helpful to them in considering how to best structure some of their
high LTV products. Moreover, the accounting standards related to the
Federal Credit Reform Act of 1990, which requires federal agencies to
estimate the budget cost of federal credit programs, suggest that federal
agencies making changes to programs should consider external sources of
data. FHA officials told us that FHA has purchased such loan performance
data. According to FHA officials, FHA relies more heavily on data that it
has collected internally from the approximately 1 million loans it
endorses each year and its single-family data warehouse, which contains
data on approximately 30 million loans. FHA officials stated that, when
possible, they use these internal data to create a proxy for how a loan
product with certain characteristics might perform. FHA officials said
they used these data to create a "virtual zero down loan" when FHA was
considering how it might implement a proposed no down payment product.

The mortgage institutions with whom we spoke noted that any loan
performance data they develop or produce when implementing new products
are also used to enhance their automated underwriting systems. The data
improve the statistical models used in their automated underwriting
systems. In May 2004, FHA implemented a statistical model for evaluating
mortgage risk that may be used in lenders' automated underwriting systems,
called the FHA TOTAL Scorecard. In developing the TOTAL Scorecard, FHA
purchased external data (credit score data), which they merged with their
existing FHA data to try to better understand the loan performance of
FHA-insured loans.

41Structured transaction is a broad term that covers any of several
methods of dividing cash flows among several investors in a pool of
mortgages. Alt A is a broad term that describes mortgages that fall just
outside of the underwriting guidelines that govern the regular mortgage
purchase business of the GSEs. Alt A mortgages are loans to borrowers with
relatively minor credit problems.

Mortgage Institutions Require Additional Credit Enhancements or Stricter
Underwriting for New Low and No Down Payment Products

Some mortgage institutions require additional credit
enhancements-mechanisms for transferring risk from one party to another-on
low and no down payment products and set stricter underwriting
requirements for these products. Mortgage institutions such as Fannie Mae
and Freddie Mac mitigate the risk of low and no down payment products by
requiring additional credit enhancements such as higher mortgage insurance
coverage. Fannie Mae and Freddie Mac require credit enhancements on all
loans they purchase that have LTVs above 80 percent. Typically, this takes
the form of private mortgage insurance. Fannie Mae and Freddie Mac also
require higher levels of private mortgage insurance coverage for loans
that have higher LTV ratios. For example, Fannie Mae and Freddie Mac
require insurance coverage of 35 percent for loans that have an LTV
greater than 95 percent. This means that, for any individual loan that
forecloses, the mortgage insurer will pay the losses on the loan up to 35
percent of the claim amount. Fannie Mae and Freddie Mac require lower
insurance coverage for loans with LTVs below 95 percent. Fannie Mae and
Freddie Mac believe that the higher-LTV loans represent a greater risk to
them and they seek to partially mitigate this risk by requiring higher
mortgage insurance coverage.

Although FHA is required to provide up to 100 percent coverage of the
loans it insures, FHA may engage in co-insurance of its single-family
loans. Under co-insurance, FHA could require lenders to share in the risks
of insuring mortgages by assuming some percentage of the losses on the
loans that they originated (lenders may use private mortgage insurance).
FHA has used co-insurance before, primarily in its multifamily programs,
but does not currently use co-insurance at all.42 FHA officials told us
they tried to put together a co-insurance agreement with Fannie Mae and
Freddie Mac and, while they were able to come to agreement on the sharing
of premiums, they could not reach agreement on the sharing of losses and
it was never implemented.

FHA could also benefit from other means of mitigating risk such as
stricter underwriting or increasing fees. Fannie Mae officials also stated
that they would charge higher guarantee fees on low and no down payment
loans if they were not able to require the higher insurance coverage.
Fannie Mae

42According to FHA, FHA discontinued the multifamily co-insurance program
after experiencing significant losses. Since then, Congress provided FHA
authority to enter into risk sharing agreements with GSEs and housing
finance agencies on certain multifamily loans.

and Freddie Mac charge guarantee fees to lenders in exchange for
converting whole loans into mortgage-backed securities, which transfer the
credit risk from the lender to Fannie Mae or Freddie Mac. Within statutory
limits, the HUD Secretary has the authority to set up-front and annual
premiums that are charged to borrowers who have FHA-insured loans. In
fact, in the administration's 2005 budget proposal for a zero down payment
product, it included higher premiums for these loans. The Secretary has
the authority to establish an up-front premium, which may be up to 2.25
percent of the amount of the original insured principal obligation of the
mortgage. Within statutory limits, the Secretary may also require payment
of an annual premium. Under the Administrative Procedures Act, the
Secretary would generally follow a process in which the change to premiums
would include issuing a proposed rule, receiving public comments, and then
issuing a final rule.

Additionally, mortgage institutions such as Fannie Mae and Freddie Mac
sometimes introduce stricter underwriting standards as part of the
development of new low and no down payment products (or products about
which they do not fully understand the risks). Institutions can do this in
a number of ways, including requiring a higher credit score threshold for
certain products, or requiring greater borrower reserves or more
documentation of income or assets from the borrower. Freddie Mac officials
stated that they believed limits on allowing ARMs or multiple-unit
properties were also reasonable, at least initially. Once the mortgage
institution has learned enough about the risks that were previously not
understood, it can change the underwriting requirements for these new
products to align with its standard products. Although FHA sometimes has
certain standards set for it through legislation, there exists some
flexibility in how it implements a newly authorized product or changes to
an existing product. The HUD Secretary has latitude within statutory
limitations in changing underwriting requirements for new and existing
products and has done this many times. Examples included the decrease in
what is included as borrower's debts and an expansion of the definition of
what can be included as borrower's effective income when lenders calculate
qualifying ratios. In the context of the new zero down product, the
Federal Housing Commissioner at HUD has stated that all loans being
considered for a zero down loan would go through FHA's TOTAL Scorecard,
and borrowers would be required to receive prepurchase counseling.

Before Fully Implementing New Products, Some Mortgage Institutions May
Limit Their Availability

Fannie Mae and Freddie Mac sometimes use pilots, or limited offerings of
new products, to build experience with a new product type or to learn
about particular variables that can help them better understand the
factors that contribute to risk for these products. Freddie Mac and Fannie
Mae also sometimes set volume limits for the percentage of their business
that could be low and no down payment lending. Fannie Mae and Freddie Mac
officials provided numerous examples of products that they now offer as
standard products but which began as part of underwriting experiments.43
These include the Fannie Mae Flexible 97(R) product, as well as the
Freddie Mac 100 product. FHA has utilized pilots or demonstrations as well
when making changes to its single-family mortgage insurance but generally
does this in response to legislation that requires a pilot and not on its
own initiative. One example in which FHA might have opted to do a pilot,
or otherwise limited volumes, for a product is with allowing nonprofit
down payment assistance. Concerns have been raised about the performance
of FHA loans that have down payment assistance. FHA might have benefited
from setting some limits on this type of assistance such that they could
study its implications before allowing its broader use.

FHA's Home Equity Conversion Mortgage (HECM) insurance program is an
example of an FHA program that started out as a pilot. HECM was initiated
by Congress in 1987 and is designed to provide elderly homeowners a
financial vehicle to tap the equity in their homes without selling or
moving from their homes. Homeowners borrow against equity in their home
and receive payments from their lenders (sometimes called a "reverse
mortgage"). Through statute, HECM started out as a demonstration program
that authorized FHA to insure 2,500 reverse mortgages. Through subsequent
legislation, FHA was authorized to insure 25,000 reverse mortgages, then
50,000, and then finally 150,000 when Congress made the program permanent
in 1998. Under the National Housing Act, the HECM program was required to
undergo a series of evaluations and it has been evaluated four times since
its inception. FHA officials told us that administering this demonstration
for only 2,500 loans was difficult because of the challenges of selecting
only a limited number of lenders and borrowers. FHA ultimately had to
limit loans to lenders drawn through a lottery.

43The GSE officials did not tell us the numeric extent to which they
limited products' issuance during its pilot phase.

The appropriate size for a pilot program depends on several factors. For
example, the precise number of loans needed to detect a difference in
performance between standard loans and loans of a new product type depends
in part on how great the differences are in loan performance. If
delinquencies early in the life of a mortgage were about 10 percent for
FHA's standard high LTV loans, and FHA wished to determine whether loans
in the pilot had delinquency rates no more than 20 percent greater that
the standard loans (delinquency no more than 12 percent), a sample size of
about 1,000 loans would be a sufficient size to detect this difference
with 95 percent confidence. If delinquency rates are different, or FHA's
desired degree of precision were different, a different sample size would
be appropriate. FHA officials with whom we spoke told us they could use
pilots or otherwise limit availability when implementing a new product or
making changes to an existing product, but they also questioned their
authority and the circumstances under which they would do so. FHA
officials also said that they lacked sufficient resources to be able to
appropriately manage a pilot.

Some mortgage institutions may also limit the initial implementation of a
new product by limiting the origination and servicing of the product to
their better lenders and servicers, respectively. Mortgage institutions
may also limit servicing on the loans to servicers with particular product
expertise, regardless of who originates the loans. Fannie Mae and Freddie
Mac both reported that these were important steps in introducing a new
product and noted that lenders tend to take a more conservative approach
when first implementing a new product. FHA officials agreed that they
could, under certain circumstances, envision piloting or limiting the ways
in which a new or changed product would be available but pointed to the
practical limitations in doing so. FHA approves the sellers and services
that are authorized to support FHA's single-family product. FHA officials
told us they face challenges in offering any of their programs only in
certain regions of the country or in limiting programs to certain approved
lenders or servicers. They generally offer their products on a national
basis and, when they do not, specific regions of the county or lenders may
question why they are not able to receive the same benefit (even on a
demonstration or pilot basis). These officials did, though, provide
examples in which their products had been initially limited to particular
regions of the country or to particular lenders, including the rollout of
the HECMs and their TOTAL Scorecard.

Mortgage Institutions Establish Enhanced Monitoring and Oversight for New
Low and No Down Payment Products and Make Changes Based on What They Learn

Mortgage institutions, including FHA, may take several steps related to
increased monitoring of new products and then make changes based on what
they learn. Fannie Mae and Freddie Mac officials described processes in
which they monitor actual versus expected loan performance for new
products, sometimes including enhanced monitoring of early loan
performance. FHA officials told us they also monitor more closely loans
underwritten under revised guidelines. Specifically, FHA officials told us
that FHA routinely conducts a review of underwriting for approximately 6
to 7 percent of loans it insures. FHA officials told us that, as part of
the review, it may place greater emphasis on reviewing those aspects of
the insurance product that are the subject of a recent change. Some
mortgage institutions, such as Fannie Mae, told us that they may conduct
rigorous quality control sampling of new acquisitions, early payment
defaults, and nonperforming loans. Depending on the scale of a new
initiative, and its perceived risk, these quality control reviews could
include a review of up to 100 percent of the loans that are part of the
new product.

Fannie Mae and Freddie Mac also reported that they conduct more regular
reviews at seller/servicer sites for new products. In some cases, Fannie
Mae and Freddie Mac have staff who conduct on-site audits at the sellers
and servicers to provide this extra layer of oversight. FHA officials also
reported that they have staff that conduct reviews of lenders that they
have identified as representing higher risk to FHA programs. However, we
recently reported that HUD's oversight of lenders could be improved and
identified a number of recommendations for improving this oversight.44
Mortgage institutions may issue a lender bulletin, announcement, or
seller/servicer guidelines to clarify instructions for new products or
changes to existing products. FHA does this through the mortgagee letters
it issues to all of its approved lenders. Mortgage institutions may also
issue a lender bulletin, announcement, or seller/servicer guidelines to
communicate required additional controls, practices, procedures,
reporting, and remitting. Importantly, changes can be made to the
structure of a product, including the automated underwriting systems used
to approve individual loans, based on information learned from monitoring
of new products or from other sources.

44GAO, Single-Family Housing: Progress Made, but Opportunities Exist to
Improve HUD's Oversight of FHA Lenders, GAO-05-13 (Washington, D.C.: Nov.
12, 2004).

FHA officials told us that they routinely analyze the changing performance
of loans they insure as part of the annual process for estimating and
re-estimating subsidy costs. The Federal Credit Reform Act of 1990
requires that federal government programs that make direct loans or loan
guarantees (including insuring loans) account for the full cost of their
programs on an annual budgetary basis. Specifically, federal agencies must
develop subsidy estimates of the net cost of their programs that include
estimates of the net costs and revenues over the projected lives of the
loans made in each fiscal year. FHA's Mutual Mortgage Insurance Fund has
historically been self-sufficient (not requiring subsidy). When preparing
cost estimates for loan guarantee programs, agencies are expected to
develop a plan to establish the appropriate information, models, and
documentation to better understand the new product and to be able to make
changes based on what they learn.45 FHA officials state that they have a
process in which changes to their model are made to reflect the
incorporation of new programs and policies and that they review the
performance of a new program in the context of their annual development of
subsidy estimates, as well as their annual actuarial study.46

Conclusions	While credit score is an effective predictor of default, LTV
remains an effective predictor of default. Loans with lower or no down
payments carry greater risk. Without any compensating measures such as
offsetting credit enhancements and increased risk monitoring and oversight
of lenders, introducing a new FHA no down payment product would expose FHA
to greater credit risk. The administration's proposal for a zero down
product included increased premiums to help compensate for an increase in
the cost of the FHA program, and the Federal Housing Commissioner stated
that borrowers would be required to go through prepurchase counseling. The
extent to which increased cost for one program could effect the overall
performance of FHA's Mutual Mortgage Insurance (MMI) fund depends, in

45The Federal Accounting Standards Advisory Board (FASAB) is responsible
for promulgating accounting standards for the U.S. Government, and these
standards are recognized as generally accepted accounting principles for
the federal government. FASAB developed standards for agencies that
describe the types of analysis that would be expected for a change to an
existing program, including relevant historical data and modeling
capabilities.

46The Cranston Gonzales National Affordable Housing Act requires an
independent actuarial analysis of the economic net worth and soundness of
FHA's MMI Fund.

part, on the scale of any new product, its relative cost, and how the new
product affects demand for FHA's existing products.

Although FHA appears to follow many key practices used by mortgage
institutions in designing and implementing new products, several practices
not currently or consistently followed by FHA stand out as appropriate
means to manage the risks associated with introducing new products or
significantly changing existing products. Moreover, these practices can be
viewed as part of a framework used by some mortgage institutions for
managing the risks associated with new or changed products. The framework
includes techniques such as limiting the availability of a new product
until it is better understood and establishing stricter underwriting
standards-all of which would help FHA to manage risk associated with any
new product it may introduce. For example, FHA could set volume limits or
limit the initial number of lenders participating in the product. Further,
changes in FHA's premiums, an important practice used by FHA, within
statutory limits, permits FHA to potentially offset additional costs
stemming from a new product that entails greater risk or not well
understood risk.

FHA officials believe that the agency does not have sufficient resources
to implement products with limited volumes, such as through a pilot
program. However, when FHA introduces new products or makes significant
changes to existing products with risks that are not well understood, such
actions could introduce significant risks when implemented broadly.
Products that would introduce significant risks can impose significant
costs. We believe that FHA could mitigate these costs by using techniques
such as piloting.

Matters for 	If Congress authorizes FHA to insure no down payment products
or any other new single-family insurance products, Congress may want to

Congressional consider a number of means to mitigate the additional risks
that these

Consideration	loans may pose. Such means may include limiting the initial
availability of such a new product, requiring higher premiums, requiring
stricter underwriting standards, or requiring enhanced monitoring. Such
risk mitigation techniques would serve to help protect the Mutual Mortgage
Insurance Fund while allowing FHA the time to learn more about the
performance of loans using this new product. Limits on the initial
availability of the new product would be consistent with the approach
Congress took in implementing the HECM program. The limits could also come
in the form of an FHA requirement to limit the new product to better

performing lenders and servicers as part of a demonstration program or to
limit the time period during which the product is first offered.

Recommendations for Executive Action

If Congress provides the authority for FHA to implement a no down payment
mortgage product or other products about which the risks are not well
understood, we recommend that the Secretary of HUD direct the Assistant
Secretary for HUD-Federal Housing Commissioner to consider the following
three actions:

o 	incorporating stricter underwriting criteria such as appropriate credit
score thresholds or borrower reserve requirements,

o 	piloting the initial product or limiting its initial availability and
asking Congress for the authority if HUD officials determine they
currently do not have this authority, and

o 	utilizing other techniques for mitigating risks including use of credit
enhancements and prepurchase counseling.

Regardless of any new products Congress may authorize, when making
significant changes to its existing products or establishing new products,
we recommend that the Secretary of HUD direct the Assistant Secretary for
HUD-Federal Housing Commissioner to consider the following two actions:

o 	limiting the initial availability of the product and when doing so, the
Commissioner should establish the conditions under which piloting should
be used, the techniques for limiting the initial availability of a
product, and the methods of enhanced monitoring that would be connected to
predetermined measures of success or failure for the product.; and

o 	asking Congress for the authority to offer its new products or
significant changes to existing products on a limited basis, such as
through pilots, if HUD officials determine they currently lack sufficient
authority.

Agency Comments and 	We provided a draft of this report to HUD, Fannie
Mae, Freddie Mac, USDA, and VA. We received written comments from HUD,
which are reprinted in

Our Evaluation appendix III. We also received technical comments from HUD,
Fannie Mae,

Freddie Mac, and USDA, which have been incorporated where appropriate. VA
did not have comments on the draft.

HUD stated that it is in basic agreement with GAO that all policy options,
implications, and implementation methods should be evaluated when
considering or proposing a new FHA product. HUD also stated that in
designing its zero down payment program it considered the items that we
recommended it consider, including piloting. HUD stated that it adopted
the prepurchase counseling requirement as a component of a proposed zero
down program and that it determined that structuring the mortgage
insurance premium in such a way as to minimize risk represents the most
appropriate tool for managing the risk of this proposed program.

However, it is not clear under what circumstances HUD believes that
piloting or limiting the availability of a changed or new product would be
appropriate or possible. As we noted in our draft report, HUD officials
told us that they face challenges in administering a pilot program because
of the difficulty of selecting only a limited number of lenders and
borrowers. HUD officials also held that they may not have the authority to
limit products and that they lacked sufficient resources to adequately
manage products as part of a pilot or with limited volumes.

We believe that HUD needs to further consider piloting or limiting volume
of new or changed products because, as we state in the report, it is a
practice followed by others in the mortgage industry and could assist HUD
in mitigating the risks and costs associated with new or changed products,
while still allowing HUD to meet its goal of providing homeownership
opportunities. Difficulties in selecting a limited number of lenders and
questions about a lack of authority could both be addressed by seeking
clear authority from Congress on these matters, if HUD officials determine
they currently lack sufficient authority. As we note in our report, when
considering the resources necessary to implement products with limited
volumes, if FHA does not use pilots or limit the availability of certain
new or changed products, FHA may face costs due to the significant risks
that can be associated with products that are implemented broadly and
about which the risks are not well understood. We do not believe that
implementing products with initial limits is appropriate or necessary in
all cases. To ensure that piloting or limiting the initial availability is
given sufficient consideration, we continue to recommend that HUD consider
establishing the conditions under which piloting should be used and the
techniques for limiting the initial availability of a product, as well as
the

methods of enhanced monitoring that would be connected to predetermined
measures of success or failure for the product.

As agreed with your office, unless you publicly announce the contents of
this report earlier, we plan no further distribution until 30 days from
the report date. At that time, we will send copies of this report to the
appropriate Congressional Committees and the Secretaries of Housing and
Urban Development, Agriculture, and Veterans Affairs. We also will make
copies available to others upon request. In addition, the report will be
available at no charge on the GAO Web site at http://www.gao.gov.

If you or your staff have any questions concerning this report, please
contact me at (202) 512-8678 or [email protected] or Mathew Scire, Assistant
Director, at (202) 512-6794 or [email protected]. Key contributors to this
report are listed in appendix IV.

Sincerely yours,

William B. Shear Director, Financial Markets and

Community Investment

Appendix I

Scope and Methodology

To describe key characteristics and standards of mortgage products, we
interviewed officials at the Federal Housing Administration (FHA), U.S.
Department of Agriculture (USDA), and U.S. Department of Veterans Affairs
(VA); as well as staff at a conventional mortgage providers (Bank of
America); private mortgage insurers (for example, The PMI Group, Inc.;
Mortgage Guarantee Insurance Corporation); government-sponsored
enterprises (GSE) (such as Fannie Mae and Freddie Mac); Office of Federal
Housing Enterprise Oversight (OFHEO); various state housing finance
agencies; and nonprofit down payment assistance providers (for example,
Nehemiah Corporation of America and Ameridream, Inc.). We reviewed
descriptions of various mortgage products and compared the standards used
across entities including FHA, USDA, and VA regulations and program
guidance and the GSEs seller/servicer guides. We reviewed Web sites of
state housing finance agencies and if we identified zero down payment
programs, we corroborated some of the Web site information through
interviews of agency officials. To report on the volume of mortgage
products, we reviewed relevant reports including reports from the U.S.
Department of Housing and Urban Development (HUD).

To determine what economic research indicates about the variables that are
most important when estimating the risk level associated with individual
mortgages, we conducted a literature search. To identify recent and
relevant papers, we used various Internet search engines (such as Online
Computer Library Center, FirstSearch: EconLit; HUD USER) and inquired with
various mortgage industry participants (for instance, FHA, Fannie Mae,
Freddie Mac, and Nehemiah). Research we reviewed includes articles,
reports, and papers that were made available to us from economic journals,
the Internet, libraries, or were provided to us by various entities (e.g.,
HUD, Fannie Mae, Freddie Mac). For the purposes of this report, we refer
to these documents as "papers."

To facilitate the search we developed several criteria. For example, we
used the following search terms: mortgage, performance, default, LTV
ratio, credit score, and down payment assistance. We excluded the
following terms from our search: multifamily and commercial. We limited
our search to papers published or issued from 1999 to 2004; however, we
did include some papers relevant to our inquiry that were published or
issued prior to 1999 that we determined were significant to our research
objectives. We identified 151 papers. There may be some relevant research
that our search did not identify.

Appendix I Scope and Methodology

For the papers we identified, we conducted a multistep review. Initially,
we determined which papers to include in our analysis. Papers included in
the analysis were those that (1) were relevant to our inquiry, (2)
included empirical analysis, and (3) utilized satisfactory methodologies.
Papers that were not relevant were excluded from our analysis (for
example, subject of paper was off-point-car loans; or analysis of loans in
foreign country). Additionally, we determined if the paper included
empirical analysis. If the paper did not include empirical analysis, we
did not include it. However, we did review the paper to determine if it
talked about papers that we had not yet identified that appeared to have
empirical analysis. If the paper did identify an additional paper that
appeared to be relevant to our inquiry, we attempted to obtain it.
Finally, we excluded papers with weak methodologies. GAO economists
conducted the evaluations of economic models. During this review, we
excluded 106 papers leaving 45 for the second-stage review. Many of the
papers we excluded were excluded for lack of relevance or because they did
not include empirical analysis. The second review consisted of documenting
the findings of the papers that were relevant, had empirical analysis, and
used satisfactory methodologies. To facilitate this analysis, we developed
and maintained an Access database to document our analysis-cataloging the
specific factors these papers identified as being important to estimating
the risk level associated with individual mortgages. Finally, for these
papers, we synthesized the literature by determining how many papers found
each variable to be important. For a bibliography of the 45 papers
included in our analysis, see appendix II.

To examine the relationship between mortgage performance and two key
underwriting variables, loan-to-value (LTV) ratio and credit score, we
calculated 4-year default rates for several categories of mortgages with
various LTV ratios and credit scores. We selected 4-year default rates
because it best balanced the competing goals of having recent loans and
the greatest number of years of default experience. To perform this
analysis, we first obtained mortgage volume and performance data from
three mortgage institutions: FHA (government mortgages) and Fannie Mae and
Freddie Mac (conventional mortgages).1 The FHA mortgage data consist of a
stratified random sample of over 400,000 FHA-insured

1We did not include mortgages guaranteed by USDA and the VA in our
analysis because credit score information for these mortgages was not
readily available.

Appendix I Scope and Methodology

mortgages originated in calendar years 1992, 1994, and 1996.2 We used
these data because they are the only significant data set of FHA loans
that includes credit scores and that had at least 4 years of loan
performance activity. The data come from a sample built by FHA for
research purposes. The Fannie Mae and Freddie Mac data consist of all
purchase-money mortgages originated in calendar years 1997, 1998, and 1999
and purchased by Fannie Mae or Freddie Mac. The data provided by Fannie
Mae and Freddie Mac exclude government-insured mortgages. We selected
these loan years because they include loans that aged at least 4 years and
because, during these years, the GSEs began to purchase an increasing
number of loans with higher LTVs. The GSEs provided us data that they
considered to be proprietary. Although we limited the reporting of our
analysis to that which was considered nonproprietary, this did not limit
our overall findings for this objective. A comparison of results from the
FHA and the conventional mortgage performance analysis should be done with
care, for a number of reasons because the data are from different years,
FHA and the GSEs calculated LTVs differently, and FHA's average 4-year
default rate is higher than for the GSEs.

For this analysis, we used the LTV ratio contained in the data system for
each mortgage institution. FHA defines the LTV ratio as the original
mortgage balance, excluding the financed mortgage insurance premium,
divided by the appraised value of the house. For the GSEs, LTV ratio is
defined as the original mortgage balance divided by the lesser of the sale
price of the house or the appraised value of the house.

For this analysis, the credit score is the Fair Isaac score contained in
each institution's data system. The mortgage institutions obtain credit
scores in various ways. FHA has only recently begun to collect credit
scores in its single-family data warehouse. However, for research
purposes, FHA purchased historic credit score information for the sample
of mortgages originated in 1992, 1994, and 1996. On the other hand, Fannie
Mae obtains credit score information in two ways. For some mortgages, the
lender obtained the borrower's credit score information when it originated
the mortgage, and upon Fannie Mae's purchase of the mortgage, the lender
provides this credit score information to Fannie Mae. In some cases,
however, lenders do not obtain borrower's credit scores; when Fannie Mae
purchases the mortgage, it obtains a credit score for the borrower. For

2Foreclosed mortgages were over-sampled in 1992 and 1994. The figures
presented in the text are weighted according to the sample weights
provided by HUD.

Appendix I Scope and Methodology

some mortgages, the institutions indicated that a credit score for a
particular mortgage was unknown. Within the FHA data, about 8 percent of
the mortgages had unknown scores; within the GSE data, about 3 percent of
the mortgages had unknown scores. We included mortgages with unknown
credit scores in our analysis and presented the loan performance results.

We carried out several actions to ensure that data provided by FHA, Fannie
Mae, and Freddie Mac were sufficiently reliable for use in our analysis.
For the FHA sample data, we met with FHA staff involved in generating the
sample data set. We also discussed data quality procedures with
appropriate FHA staff. Based on these discussions in which FHA officials
described their policies and procedures and the results of external audits
of their data systems, we determined that the FHA data were sufficiently
reliable to use in our analysis. FHA officials indicated that their data
systems contain data entry edit checks and that data submitted by lenders
was reviewed by FHA. FHA's data system was audited by external auditors,
and no major issues concerning data quality were raised. We also discussed
data quality procedures with appropriate Fannie Mae and Freddie Mac staff.
These procedures included data entry edit checks, exception reports, and
checks for reasonableness. Additionally, we reviewed reports from audits
of Fannie Mae and Freddie Mac. These audits included an assessment of the
Fannie Mae information systems that generated the data used in this
report. The audits also assessed Freddie Mac information systems that
generated the data used in this report. We also compared the data with
similar publicly available data. Based on these discussions and reviews of
audit reports, we determined that the data Fannie Mae and Freddie Mac
provided were sufficiently reliable to use in our analysis.

With these data, we generated FHA and conventional 4-year default rates
for several combinations of LTV ratios and credit scores. To do this, we

o 	defined default as a credit event that includes foreclosed mortgages,
as well as mortgages that did not experience foreclosure, but that would
typically lead to a credit loss, such as a "short sale" or a "deed-in-lieu
of foreclosure" termination of the mortgage;

o  selected six LTV ratio categories;

o  selected six credit score categories;

Appendix I Scope and Methodology

o 	combined Fannie Mae and Freddie Mac mortgage volume and performance
data;

o 	combined mortgage volume and performance data for the sample (of
mortgages insured by FHA in 1992, 1994, and 1996; and conventional
mortgages originated in 1997, 1998, and 1999 and purchased by Fannie Mae
or Freddie Mac);

o 	calculated the average 4-year default rate for FHA (weighted average)
and for all conventional loans separately by dividing the total dollar
amount of mortgages experiencing a credit event by the total dollar amount
of mortgages originated (for FHA) or purchased (for conventional);

o 	calculated the average 4-year default rates for sampled FHA loans and
for conventional loans that fell within each LTV ratio and credit score
category; and

o 	calculated the relative 4-year default rates for each LTV ratio and
credit score categories for FHA loans and for conventional loans by
dividing the average 4-year default rate for each specific LTV and credit
score category by the average 4-year default rate for sampled FHA loans
and all conventional loans, respectively.

For example, if the merged average 4-year default rate for FHA loans
within a particular LTV ratio and credit score category was 3 percent, and
the average 4-year default rate for all FHA loans was 2 percent, the
relative 4year default rate for FHA loans within this particular category
would be 3 divided by 2, or 1.5 times the average FHA default rate.

We do not present relative default rates for categories with small numbers
of mortgages because the performance information may not be reliable when
there are too few observations. In the figures, these instances are noted
as "N/A." For the FHA analysis, we used a cutoff of about 1,000 mortgages
to determine whether there were sufficient observations to reliably
measure the relative default rate. For the conventional analysis, we used
a cutoff of about 3,000 mortgages to determine whether the relative
default rate was reliable. We chose a higher cutoff for the GSE analysis
because the GSEs have a lower default rate, and analysis of less frequent
events requires a larger sample size.

Appendix I Scope and Methodology

To determine what lessons FHA might learn from others that support low and
no down payment lending we obtained testimonial information from the
mortgage industry (for example, FHA, GSEs, private mortgage insurers, and
a private lender) about the steps they take to design and implement low
and no down payment lending. We selected these entities based on the
parallels to FHA, as well as their significance in the mortgage industry.
Where available, we reviewed industry and academic information relevant to
these steps in carrying out low and no down payment lending.

We performed our audit work from January 2004 to December 2004 in
accordance with generally accepted government auditing standards.

Appendix II

Papers Identified in Literature Search and Included in Analysis

Credit score

Other factor(s)

                                   Paper LTV

Brent W. Ambrose and Charles A. Capone. "The Hazard Rates of First and
Second Defaults,"

X

    Journal of Real Estate Finance and Economics, vol. 20 no. 3 (May 2000).

Brent W. Ambrose and Charles A. Capone. "Modeling the Conditional
Probability of Foreclosure Xin the Context of Single-Family Mortgage
Default Resolutions," Real Estate Economics, vol. 26no. 3 (1998).

Richard Anderson and James VanderHoff. "Mortgage Default Rates and
Borrower Race," The Journal of Real Estate Research, vol. 18 no. 2
(Sep/Oct 1999).

Robert B. Avery, Raphael W. Bostic, Paul S. Calem, and Glenn B. Canner.
"Credit Risk, Credit Scoring, and the Performance of Home Mortgages,"
Federal Reserve Bulletin (July 1996).

XX

  James A. Berkovec, Glenn B. Canner, Stuart A. Gabriel and Timothy H. Hannan.
                                     "Race,

X

Redlining, and Residential Mortgage Loan Performance," Journal of Real
Estate Finance and Economics, vol. 9 no. 1 (July 1994).

Paul S. Calem and Susan M. Wachter. "Community Reinvestment and Credit
Risk: Evidence from an Affordable-Home-Loan Program," Real Estate
Economics, vol. 27 no. 1 (1999).

                                       X

Paul S. Calem and James Follain. The Asset-Correlation Parameter in Basel II for
                                  Mortgages on

XX

Single-Family Residences, a report prepared as background for public comment on
  the Advance Notice of Proposed Rulemaking on the Proposed New Basel Capital
                           Accord, November 6, 2003.

Paul S. Calem and Michael LaCour-Little. Risk-based Capital Requirements
for Mortgage Loans, November 2001.

XX Charles A. Calhoun and Yongheng Deng. "A Dynamic Analysis of Fixed- and
                                Adjustable-Rate

X

Mortgage Terminations," Journal of Real Estate Finance and Economics, vol.
24 no. 1/2 (Jan 2002).

Dennis R. Capozza, Dick Kazarian, and Thomas A. Thomson. "Mortgage Default
in Local Markets," Real Estate Economics, vol. 25 no. 4 (Winter 1997).

X

  Richard L. Cooperstein, F. Stevens Redburn, and Harry G. Meyers. "Modelling
                                    Mortgage

X

    Terminations in Turbulent Times," AREUEA Journal, vol. 19 no. 4 (1991).

Robert F. Cotterman. Analysis of FHA Single-Family Default and Loss  X X 
                         Rates, a report prepared                           
    for the Office of Policy Development and Research, U.S. Department      
                           of Housing and Urban                             
                       Development, March 25, 2004.                         
    Robert F. Cotterman. New Evidence on the Relationship Between Race  X X 
                          and Mortgage Default:                             
     The Importance of Credit History Data, a report prepared for the       
                       Office of Policy Development                         
and Research, U.S. Department of Housing and Urban Development, May      
                                23, 2002.                                   
Robert F. Cotterman. Neighborhood Effects in Mortgage Default Risk,  X X X 
                        a report prepared for the                           
      Office of Policy Development and Research, U.S. Department of         
                            Housing and Urban                               
                         Development, March 2001.                           
Robert F. Cotterman. Assessing Problems of Default in Local Mortgage X   X 
                            Markets, a report                               
     prepared for the Office of Policy Development and Research, U.S.       
                        Department of Housing and                           
                      Urban Development, March 2001.                        
Amy Crews Cutts and Robert Van Order. "On the Economics of Subprime  X X X 
                          Lending." Freddie Mac                             
                Working Paper Series # 04-01(January 2004)                  
               (http://freddiemac.com/corporate/reports/).                  

  Appendix II Papers Identified in Literature Search and Included in Analysis

                         (Continued From Previous Page)

                                                             Credit     Other 
                           Paper                         LTV score  factor(s) 
      Ralph DeFranco. "Modeling Residential Mortgage       X      X 
         Termination and Severity Using Loan Level                  
       Data." (Ph.D diss., University of California,                
                     Berkeley, 2002).                               
Yongheng Deng, John M. Quigley. Woodhead Behavior and   X        
                the Pricing of Residential                          
                Mortgages, (December 2002).                         

        Yongheng Deng and Stuart Gabriel. Enhancing Mortgage Credit       X X 
                       Availability Among Underserved                       
      and Higher Credit-Risk Populations: An Assessment of Default and      
                         Prepayment Option Exercise                         
Among FHA-Insured Borrowers, a report prepared for the U.S. Department   
                               of Housing and                               
                      Urban Development, August 2002.                       

       Yongheng Deng and Stuart Gabriel. Modeling the Performance of      X X 
                             FHA-Insured Loans:                             
      Borrowers Heterogeneity and the Exercise of Mortgage Default and      
                           Prepayment Options, a                            
     report submitted to the Office of Policy Development and Research,     
                         U.S. Department of Housing                         
                      and Urban Development, May 2002.                      

Yongheng Deng, John M. Quigley, and Robert Van Order. "Mortgage
Terminations, X Heterogeneity and the Exercise of Mortgage Options,"
Econometrica, vol. 68 no. 2 (March 2000).

Yongheng Deng, John M. Quigley, and Robert Van Order, Mortgage Default and
Low XDownpayment Loans: The Costs of Public Subsidy, National Bureau of
Economic Research: Working Paper No. 5184 (Cambridge, Mass.: July 1995).

Peter J. Elmer and Steven A Seelig. "Insolvency, Trigger Events, and
Consumer Risk Posture in Xthe Theory of Single-Family Mortgage Default,"
Journal of Housing Research, vol. 10 no. 1 (1999).

Robert M. Feinberg and David Nickerson. "Crime and Residential Mortgage
Default: An Empirical Analysis." Applied Economics Letters, vol. 9 (2002).

      Dan Feshbach and Michael Simpson. "Tools for Boosting Portfolio     X X 
                           Performance," Mortgage                           
                           Banking, October 1999.                           
Dan Feshbach and Pat Schwinn. "A Tactical Approach to Credit Scores,"  X X 
                             Mortgage Banking,                              
                               February 1999.                               
Gerson M. Goldberg and John P. Harding. "Investment Characteristics of X 
                             Low-and Moderate-                              
Income Mortgage Loans," Journal of Housing Economics, vol. 12 (2003).    
    Government Accountability Office. Mortgage Financing: Changes in the  X 
                            Performance of FHA-                             
         Insured Loans, GAO-02-773. Washington, D.C. July 10, 2002.         
    Government Accountability Office. Mortgage Financing: FHA's Fund Has  X 
                           Grown, but Options for                           
    Drawing on the Fund Have Uncertain Outcomes, GAO-01-460. Washington,    
                             D.C. February 28,                              
                                   2001.                                    

Valentina Hartarska, Claudio Gonzalez-Vega, and David Dobos. Credit
Counseling and the XIncidence of Default on Housing Loans by Low-Income
Households, a paper prepared as part ofa collaborative research program
between Ohio State University and Paul Taylor and Associates, of Columbus,
Ohio. (February 2002).

Abdighani Hirad and Peter M. Zorn (corresponding author). A Little
Knowledge Is a Good Thing: XEmpirical Evidence of the Effectiveness of
Pre-Purchase Homeownership Counseling, May 22, 2001.

The Department of Housing and Urban Development, Office of Inspector
General. Follow-up of XDown Payment Assistance Programs Operated by
Private Nonprofit Entities. 2002-SE-0001,Seattle, Washington, September
25, 2002.

Appendix II Papers Identified in Literature Search and Included in
Analysis

                         (Continued From Previous Page)

                                                             Credit     Other 
                           Paper                         LTV score  factor(s) 
     The Department of Housing and Urban Development,               
         Office of Inspector General. Final Report                  
of Nationwide Audit: Down Payment Assistance Programs            
             (Office of Insured Single Family                       
     Housing), 2000-SE-121-0001, Seattle, Washington,               
                      March 31, 2000.                               
        Michael Lacour-Little and Stephen Malpezzi.        X        
        "Appraisal Quality and Residential Mortgage                 
      Default: Evidence From Alaska," Journal of Real               
        Estate Finance and Economics, vol. 27 no. 2                 
                          (2003).                                   
      Andrey D. Pavlov. "Competing Risks of Mortgage       X        
          Termination: Who Refinances, Who Moves,                   
and Who Defaults," Journal of Real Estate Finance and            
            Economics, vol. 23 no. 2 (September                     
                          2001).                                    
     Anthony Pennington-Cross. "Credit History and the            X 
             Performance of Prime and Nonprime                      
      Mortgages," Journal of Real Estate Finance and                
             Economics, vol. 27 no. 3 (2003).                       

Anthony Pennington-Cross. "Subprime and Prime Mortgages: Loss
Distributions." X X

Office of Federal Housing Enterprise Oversight Working Paper Series 03-01

(May 27, 2003).

Anthony Pennington-Cross. "Patterns of Default and Prepayment for Prime
and Nonprime X X Mortgages." Office of Federal Housing Enterprise
Oversight Working Paper 02-1 (March 2002).

     Roberto G. Quercia, Michael A. Stegman, Walter R. Davis, and Eric    X X 
                              Stein. Community                              
Reinvestment Lending: A Description and Contrast of Loan Products and    
                            Their Performance, a                            
report prepared for the Joint Center for Housing Studies' Symposium on   
                                 Low-Income                                 
        Homeownership as an Asset-Building Strategy, September 2000.        

Roberto G. Quercia, George W. McCarthy, and Michael A. Stegman. "Mortgage
Default Among X Rural, Low-Income Borrowers," Journal of Housing Research
vol. 6 no. 2 (1995).

Stephen L. Ross. "Mortgage Lending, Sample Selection and Default," Real
Estate Economics, X vol. 28 no. 4 (Winter 2000).

Robert A. Van Order and Peter M. Zorn. The Performance of Low Income X X 
                               and Minority                                 
              Mortgages: A Tale of Two Options, August 2001.                
      Robert Van Order and Peter Zorn. Performance of Low-Income and    X X 
                       Minority Mortgages, a report                         
     prepared for the Joint Center for Housing Studies' Symposium on        
                         Low-Income Homeownership                           
              as an Asset-Building Strategy, September 2001.                
     Robert Van Order and Peter Zorn. "Income, Location, and Default:   X   
                          Some Implications for                             
     Community Lending," Real Estate Economics, vol. 28 no. 3 (2000).       
    Economic Systems Inc., ORC Macro, and The Hay Group. Evaluation of  X   X 
                              VA's Home Loan                                
Guaranty Program: Final Report. A report prepared for the Department     
                        of Veterans Affairs. (July                          
                                  2004).                                    

                                  Source: GAO.

Appendix III

Comments from the Department of Housing and Urban Development

Appendix IIIComments from the Department of Housing and Urban Development

Appendix IV

                     GAO Contacts and Staff Acknowledgments

GAO ContactsWilliam B. Shear, (202) 512-8678 Matthew Scire, (202) 512-6794

Staff In addition to those individuals named above, Anne Cangi, Rudy
Chatloss, Bert Japikse, Austin Kelly, Marc Molino, Andy Pauline, Roberto
Pinero, and

Acknowledgments Mitch Rachlis made key contributions to this report.

GAO's MissionThe Government Accountability Office, the audit, evaluation
and investigative arm of Congress, exists to support Congress in meeting
its constitutional responsibilities and to help improve the performance
and accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or PhoneThe first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone:Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

CongressionalGloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public AffairsPaul Anderson, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548
*** End of document. ***