This is the accessible text file for GAO report number GAO-02-773
entitled 'Mortgage Financing: Changes in the Performance of FHA-Insured
Loans' which was released on August 09, 2002.
This text file was formatted by the U.S. General Accounting Office
(GAO) to be accessible to users with visual impairments, as part of a
longer term project to improve GAO products’ accessibility. Every
attempt has been made to maintain the structural and data integrity of
the original printed product. Accessibility features, such as text
descriptions of tables, consecutively numbered footnotes placed at the
end of the file, and the text of agency comment letters, are provided
but may not exactly duplicate the presentation or format of the printed
version. The portable document format (PDF) file is an exact electronic
replica of the printed version. We welcome your feedback. Please E-mail
your comments regarding the contents or accessibility features of this
document to Webmaster@gao.gov.
Report to the Chairwoman, Subcommittee on Housing and Community
Opportunity, Committee on Financial Services, House of Representatives:
July 2002:
Mortgage Financing:
Changes in the Performance of FHA-Insured Loans:
GAO-02-773:
Contents:
Letter:
Results in Brief:
Background:
Early Performance of FHA Loans Originated during the Late
1990s Has Declined Slightly:
Program-and Market-Related Changes that Could Explain
Higher Foreclosure Rates:
Performance of Recent Loans Suggests that FHA’s Portfolio
May Be Riskier than Previously Estimated:
Agency Comments and Our Evaluation:
Appendixes:
Appendix I: Scope and Methodology:
Appendix II: Models Used to Forecast Defaults and Prepayments for
FHA-Insured Mortgages:
Data and Sample Selection:
Specification of the Model:
Estimation Results:
Appendix III: Data for Figures Used in This Report:
Appendix IV: Comments from the Department of Housing and Urban
Development:
Tables:
Table 1: Description of FHA’s Loss Mitigation Tools Available to
Lenders:
Table 2: Variable Names and Descriptions:
Table 3: Means of Predictor Variables:
Table 4: Coefficients from Foreclosure Equations and Summary
Statistics:
Table 5: Coefficients from Prepayment Equations and Summary Statistics:
Table 6: National 4-Year Cumulative Foreclosure Rates forAll FHA Loans
Originated during Fiscal Years 1990-1998 (Figure 1):
Table 7: National 4-Year Cumulative Foreclosure Rates for Long-Term,
Fixed Rate Loans Originated during Fiscal Years 1990-1998 (Figure 2):
Table 8: National 4-Year Cumulative Foreclosure Rates forFHA Fixed-and
Adjustable Rate Mortgage Loans Originated during Fiscal Years 1990-
1998(Figure 3):
Table 9: Adjustable Rate Mortgages as Share of All FHALoans Originated
during Fiscal Years 1990-1998(Figure 4):
Table 10: Share of FHA Long-Term, Fixed-Rate Loans Originated in
Selected States during Fiscal Years 1990-1998 (Figure 5):
Table 11: National 4-Year Cumulative Foreclosure Rates for FHA Long-
Term, Fixed-Rate Loans Originated inSelected States during Fiscal Years
1990-1998 (Figure 6):
Table 12: Share of FHA Adjustable Rate Mortgages Originatedin Selected
States during Fiscal Years 1990-1998(Figure 7):
Table 13: National 4-Year Cumulative Foreclosure Rates forFHA
Adjustable Rate Mortgages Originated in Selected States during Fiscal
Years 1990-1998(Figure 8):
Table 14: Distribution of LTV Categories for FHA Loans Originated
during
Fiscal Years 1990-1998 (Figure 9):
Table 15: National 4-Year Cumulative Foreclosure Rates for Selected
LTV
Classes of Long-Term, Fixed-Rate Mortgages Originated during Fiscal
Years 1990-1998 (Figure 10):
Table 16: Actual and Forecasted Cumulative Foreclosure Rates for FHA
Loans Insured during Fiscal Years 1996-2001, as of September 30, 2001
(Figure 11):
Figures:
Figure 1: National 4-Year Cumulative Foreclosure Rates for All FHA
Loans
Originated during Fiscal Years 1990-1998:
Figure 2: National 4-year Cumulative Foreclosure Rates for All FHA
Loans
Originated during Fiscal Years 1980-1998:
Figure 3: National 4-year Cumulative Foreclosure Rates for All FHA
Loans
Originated during Fiscal Years 1990-1998, by Loan Type:
Figure 4: Adjustable Rate Mortgages as Share of All FHA Loans
Originated
during Fiscal Years 1990-1998:
Figure 5: Share in Selected States of FHA Long-Term, Fixed-Rate Loans
Originated during Fiscal Years 1990-1998:
Figure 6: National 4-year Cumulative Foreclosure Rates in Selected
States for FHA Long-Term, Fixed-RateLoans Originated during Fiscal
Years 1990-1998:
Figure 7: Share of FHA Adjustable Rate Mortgages, in Selected States,
Originated during Fiscal Years 1990-1998:
Figure 8: National 4-year Cumulative Foreclosure Rates in Selected
States for FHA Adjustable Rate Mortgages Originated during Fiscal Years
1990-1998:
Figure 9: Share of FHA Loans within Various LTV Categories for Loans
Originated during Fiscal Years 1990-1998:
Figure 10: National 4-year Cumulative Foreclosure Rates for Selected
LTV
Classes of Long-Term, Fixed-Rate FHA Mortgages Originated during Fiscal
Years 1990-1998:
Figure 11: Actual and Forecasted Cumulative Foreclosure Rates for FHA
Loans Insured during Fiscal Years 1996-2001, as of September 30, 2001:
Figure 12: Cumulative Foreclosure Rates by Book of Business for 30-
Year,
Fixed-Rate, Nonrefinanced Mortgages, Actual and Predicted, Fiscal Years
1975-1995:
Figure 13: Cumulative Prepayment Rates by Book of Business for 30-Year,
Fixed-Rate, Nonrefinanced Mortgages, Actual and Predicted, Fiscal Years
1975-1995:
Abbreviations:
ARM: Adjustable rate mortgage:
Fannie Mae: Federal National Mortgage Association:
FHA: Federal Housing Administration:
Freddie Mac: Federal Home Loan Mortgage Corporation:
HUD: Department of Housing and Urban Development:
LTV: Loan-to-value:
Letter:
July 10, 2002:
The Honorable Marge Roukema
Chairwoman, Subcommittee on Housing
and Community Opportunity
Committee on Financial Services
House of Representatives:
Dear Madam Chairwoman:
The Department of Housing and Urban Development (HUD), through its
Federal Housing Administration (FHA), provides insurance for private
lenders against losses on home mortgages. The insurance program is
supported by the Mutual Mortgage Insurance Fund (Fund). To help place
the Fund on a financially sound basis, the Congress enacted legislation
in November 1990 that required the Secretary of HUD to, among other
things, take steps to ensure that the Fund achieve and maintain an
economic value of at least 2 percent of the Fund’s insurance-in-
force.[Footnote 1] In February 2001 we reported that a 2 percent
capital ratio appeared sufficient to withstand moderately severe
economic downturns that could lead to worse-than-expected loan
performance.[Footnote 2] However, we cautioned against concluding that
the Fund could withstand the specified economic scenarios regardless of
the future activities of FHA or the market. Specifically, we noted that
our estimates and those of others are valid only under a certain set of
conditions, including that recently insured FHA loans respond to
economic conditions similarly to the response of those insured in the
more distant past. At the end of fiscal year 2001, loans originated in
the most recent 4 fiscal years accounted for about 70 percent of FHA’s
portfolio.
Concerned about reported increases in FHA’s default and foreclosure
rates, you asked that we assess the performance of loans made in recent
years and the implications for the Fund of any worsening loan
performance. To address your concerns, we (1) describe how the early
performance of FHA loans originated in recent years differs from the
performance of loans originated earlier; (2) describe changes in FHA’s
program or the conventional mortgage market that could explain recent
loan performance; and (3) assess whether the overall riskiness of FHA’s
portfolio is greater than we previously estimated and assess the impact
that any increased riskiness might have on the ability of the Fund to
withstand worse-than-expected loan performance.
To meet these objectives, we used data provided by FHA to compare
foreclosure rates for FHA-insured loans over time by the type of loan,
the location of the property, and the amount of the loan as a
percentage of the property’s value (loan-to-value ratio). We reviewed
FHA guidance, trade literature, and publicly available information to
identify changes in the FHA and conventional mortgage market that could
explain any differences in loan performance for recently originated
loans. Finally, using the model that we developed for our prior report
and basing it on the experience of FHA loans insured from fiscal years
1975 through 1995, we also compared the estimated and actual
foreclosure rates through 2001 of loans insured from fiscal years 1996
through 2001. Appendix I provides a more detailed description of our
scope and methodology. Appendix II provides a technical description of
the model we used to assess estimated and actual loan performance.
We conducted our work from July 2001 through June 2002, in accordance
with generally accepted government auditing standards.
Results in Brief:
Although FHA loans made in recent years have experienced somewhat
higher foreclosure rates than loans made in the years immediately
preceding them, recent loans are performing much better than loans made
in the 1980s. Specifically, FHA loans made during the 1990s had lower
cumulative foreclosures by the fourth year after origination than
similarly aged loans made during the 1980s. However, foreclosure rates
were somewhat higher for loans originated during the latter 1990s than
they were earlier in the decade. Specifically, through their fourth
year, loans insured during fiscal years 1990 through 1994 had an
average cumulative foreclosure rate of 2.23 percent, while loans
originated later in the decade had an average foreclosure rate of 2.93
percent. Foreclosure rates were even higher for adjustable rate
mortgages and mortgages on properties located in California.
Specifically, between 1990 and 1994 the 4-year cumulative foreclosure
rate for adjustable rate mortgages, which nearly doubled in volume
during the 1990s, averaged 2.53 percent, as compared with a 3.90
percent average 4-year cumulative foreclosure rate for adjustable rate
mortgages originated between 1995 and 1998. California, which accounted
for 15 percent of the dollar value of all single-family loans FHA
insured during the 1990s, had an average foreclosure rate of 6.41
percent for both fixed rate and adjustable rate mortgages. In
comparison, the 4-year cumulative foreclosure rate for FHA loans
insured during the 1990s outside of California averaged 1.97 percent.
Part of the increase in the overall foreclosure rate during the 1990s
is attributable to the increasing number of loans with higher loan-to-
value ratios. However, regardless of the loan-to-value ratio of a loan,
foreclosure rates generally were higher for loans made later in the
decade.
Although economic factors such as house price appreciation are key
determinants of mortgage foreclosure, changes in underwriting
requirements as well as changes in the conventional mortgage market may
partly explain the higher foreclosure rates experienced later in the
1990s. Since 1995 there have been numerous changes to FHA’s
underwriting procedures, designed mainly to increase homeownership
opportunities. Generally, these changes have allowed more borrowers who
may not have met previous underwriting standards to qualify for loans,
or have increased the loan amounts for which these borrowers qualify.
In addition, since 1995 private mortgage insurers have been more likely
to insure loans with low down payments for borrowers whom the private
insurers identified as being relatively low risk. As a result of both
types of changes, the risk associated with FHA’s loan portfolio may
have increased since 1995. FHA also took steps to tighten underwriting
and to mitigate losses from foreclosures. Because of data limitations,
we were unable to directly estimate the effect of changes in FHA
underwriting and the conventional mortgage market on loan performance.
Specifically, the data that FHA collects at the individual loan level
on items such as credit scores and debt-to-income ratios, which would
allow such an analysis, have not been collected for a sufficient number
of years or are not sufficiently detailed to permit their inclusion in
a model that estimates the impact of economic variables on loan
performance.
Although more years of loan performance are necessary to make a
definitive judgment, our analysis suggests that factors not fully
captured in the model we used for our February 2001 report may be
affecting the performance of recent FHA loans and causing the overall
riskiness of FHA’s portfolio to be somewhat greater than we previously
estimated. These factors could include the changes in underwriting and
in the conventional mortgage market described above. In particular, we
found that foreclosure rates through the end of fiscal year 2001, for
books of business insured after fiscal year 1995, are greater than what
would be anticipated from a model based on the performance of loans
insured from 1975 through 1995.[Footnote 3] Thus the Fund may be
somewhat less able to withstand worse-than-expected loan performance
resulting from adverse economic conditions. We continue to urge caution
in concluding that the Fund can withstand specified economic scenarios
regardless of how recently insured loans respond to economic
conditions.
We presented a draft of this report to officials from HUD for their
review and comment. They provided written comments that are reprinted
in appendix IV. Generally, HUD officials agreed with the findings of
the report and commented that the underwriting changes made in 1995
allowed FHA to be successful in its mission of increasing homeownership
opportunities for underserved groups.
Background:
FHA was established in 1934 under the National Housing Act (P.L. 73-
479) to broaden homeownership, shore up and protect lending
institutions, and stimulate employment in the building industry. FHA
insures private lenders against losses on mortgages that finance
purchases of properties with one to four housing units. Many FHA-
insured loans are made to low-income, minority, and first-time
homebuyers.
Generally, lenders require borrowers to purchase mortgage insurance
when the value of the mortgage is large relative to the price of the
house. FHA provides most of its single-family insurance through a
program supported by the Mutual Mortgage Insurance Fund. The economic
value of the Fund, which consists of the sum of existing capital
resources plus the net present value of future cash flows, depends on
the relative size of cash outflows and inflows over time. Cash flows
out of the Fund from payments associated with claims on foreclosed
properties, refunds of up-front premiums on mortgages that are prepaid,
and administrative expenses for management of the program. To cover
these outflows, FHA deposits cash inflows--up-front and annual
insurance premiums from participating homebuyers and the net proceeds
from the sale of foreclosed properties--into the Fund. If the Fund were
to be exhausted, the U.S. Treasury would have to cover lenders’ claims
and administrative costs directly. The Fund remained relatively healthy
from its inception until the 1980s, when losses were substantial,
primarily because of high foreclosure rates in regions experiencing
economic stress, particularly the oil-producing states in the West
South Central section of the United States.[Footnote 4] These losses
prompted the reforms that were first enacted in November 1990 as part
of the Omnibus Budget Reconciliation Act of 1990 (P.L. 101-508). The
reforms, designed to place the Fund on an actuarially sound basis,
required the Secretary of HUD to, among other things, take steps to
ensure that the Fund attained a capital ratio of 2 percent of the
insurance-in-force by November 2000 and to maintain or exceed that
ratio at all times thereafter.[Footnote 5] As a result of the 1990
housing reforms, the Fund must meet not only the minimum capital ratio
requirement but also operational goals before the Secretary of HUD can
take certain actions that might reduce the value of the Fund. These
operational goals include meeting the mortgage credit needs of certain
homebuyers while maintaining an adequate capital ratio, minimizing
risk, and avoiding adverse selection. However, the legislation does not
define what constitutes adequate capital or specify the economic
conditions that the Fund should withstand.
The 1990 reforms also required that an independent contractor conduct
an annual actuarial review of the Fund. These reviews have shown that
during the 1990s the estimated value of the Fund grew substantially. At
the end of fiscal year 1995, the Fund attained an estimated economic
value that slightly exceeded the amount required for a 2 percent
capital ratio. Since that time, the estimated economic value of the
Fund continued to grow and always exceeded the amount required for a 2
percent capital ratio. In the most recent actuarial review, Deloitte &
Touche estimated the Fund’s economic value at about $18.5 billion at
the end of fiscal year 2001. This represents about 3.75 percent of the
Fund’s insurance-in-force.
In February 2001 we reported that the Fund had an economic value of
$15.8 billion at the end of fiscal year 1999. This estimate implied a
capital ratio of 3.20 percent of the unamortized insurance-in-force.
The relatively large economic value and high capital ratio reported for
the Fund reflected the strong economic conditions that prevailed during
most of the 1990s, the good economic performance that was expected for
the future, and the increased insurance premiums put in place in 1990.
In our February 2001 report we also reported that, given the economic
value of the Fund and the state of the economy at the end of fiscal
year 1999, a 2 percent capital ratio appeared sufficient to withstand
moderately severe economic scenarios that could lead to worse-than-
expected loan performance. These scenarios were based upon recent
regional experiences and the national recession that occurred in 1981
and 1982. Specifically, we found that such conditions would not cause
the economic value of the Fund at the end of fiscal year 1999 to
decline by more than 2 percent of the Fund’s insurance-in-force.
Although a 2 percent capital ratio also appeared sufficient to allow
the Fund to withstand some more severe scenarios, we found that three
of the most severe scenarios we tested would cause the economic value
of the Fund to decline by more than 2 percent of the Fund’s insurance-
in-force.[Footnote 6] These results suggest that the existing capital
ratio was more than sufficient to protect the Fund from many worse-
than-expected loan performance scenarios. However, we cautioned that
factors not fully captured in our economic models could affect the
Fund’s ability to withstand worse-than-expected experiences over time.
These factors include recent changes in FHA’s insurance program and the
conventional mortgage market that could affect the likelihood of poor
loan performance and the ability of the Fund to withstand that
performance.
In deciding whether to approve a loan, lenders rely upon underwriting
standards set by FHA or the private sector. FHA’s underwriting
guidelines require lenders to establish that prospective borrowers have
the ability and willingness to repay a mortgage. In order to establish
a borrower’s willingness and ability to pay, these guidelines require
lenders to evaluate four major elements: qualifying ratios and
compensating factors; stability and adequacy of income; credit history;
and funds to close.
In recent years, private mortgage insurers and conventional lenders
have begun to offer alternatives to borrowers who want to make small or
no down payments.[Footnote 7] Private lenders have also begun to use
automated underwriting as a means to better target low-risk borrowers
for conventional mortgages. Automated underwriting relies on the
statistical analysis of hundreds of thousands of mortgage loans that
have been originated over the past decade to determine the key
attributes of the borrower’s credit history, the property
characteristics, and the terms of the mortgage note that affect loan
performance. The results of this analysis are arrayed numerically in
what is known as a “mortgage score.” A mortgage score is used as an
indicator of the foreclosure or loss risk to the lender.
Early Performance of FHA Loans Originated during the Late 1990s Has
Declined Slightly:
During their early years, FHA loans insured from fiscal year 1995
through fiscal year 1998 have shown somewhat higher cumulative
foreclosure rates than FHA loans insured from fiscal year 1990 through
fiscal year 1994, but these rates are well below comparable rates for
FHA loans insured in the 1980s. To better understand how foreclosure
rates might vary, we compared the rates for different types of loans--
fixed-rate and adjustable rate mortgages (ARMs)--locations of
properties, and loan-to-value (LTV) ratios. For loans made in recent
years, FHA has been experiencing particularly high foreclosure rates
for ARMs and mortgages on properties located in California. One measure
of the initial risk of a loan, its LTV, can partly explain the
difference over time in foreclosure rates. That is, FHA insured
relatively more loans with high LTVs later in the decade than it
insured earlier in the decade. However, the same pattern of higher
foreclosure rates in the later 1990s exists even after differences in
LTV are taken into account.[Footnote 8]
Foreclosure Rates Are Somewhat Higher for FHA Loans Made Later in the
1990s, but Do Not Approach the Levels for Loans Made in the Previous
Decade:
We compared the four-year cumulative foreclosure rates across books of
business to measure the performance of FHA’s insured loans.[Footnote 9]
As shown in figure 1, the 4-year cumulative foreclosure rate for FHA-
insured loans was generally higher for loans originated later in the
1990s than for loans originated earlier in that decade.[Footnote 10]
Through their fourth year, loans originated during fiscal years 1990
through 1994 had an average cumulative foreclosure rate of 2.23
percent, while loans originated during fiscal years 1995 through 1998
had an average cumulative foreclosure rate of 2.93 percent.
Figure 1: National 4-Year Cumulative Foreclosure Rates for All FHA
Loans Originated during Fiscal Years 1990-1998:
[See PDF for image]
Note: Data for all figures are in appendix III.
Source: GAO analysis of FHA data.
[End of figure]
Although the 4-year cumulative foreclosure rates for loans that FHA
insured in the later part of the 1990s were higher than that for loans
that FHA insured earlier in that decade, those rates were still well
below the high levels experienced for loans that FHA insured in the
early-to mid-1980s, as shown in figure 2. The 4-year cumulative
foreclosure rates for FHA loans originated between 1981 and 1985, a
period of high interest and unemployment rates and low house price
appreciation rates, ranged between 5 and 10 percent, while the rates
for loans originated during the 1990s, when economic conditions were
better, have consistently been below 3.5 percent.
Figure 2: National 4-Year Cumulative Foreclosure Rates for All FHA
Loans Originated during Fiscal Years 1980-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
FHA Foreclosure Rates Have Been Particularly High for Adjustable Rate
Mortgages:
Since fiscal year 1993, FHA has experienced higher 4-year cumulative
foreclosure rates for ARMs than it has for long-term (generally 30-
year) fixed-rate mortgages, as shown in figure 3. In addition, between
1990 and 1994 the 4-year cumulative foreclosure rate for ARMs averaged
2.53 percent, as compared with a 3.90 percent average 4-year cumulative
foreclosure rate for ARMs originated between 1995 and 1998. These
higher foreclosures have occurred even though mortgage interest rates
have been generally stable or declining during this period.
Figure 3: National 4-Year Cumulative Foreclosure Rates for All FHA
Loans Originated during Fiscal Years 1990-1998, by Loan Type:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
In the early 1990s, when ARMs were performing better than fixed-rate
mortgages, the performance of ARMs had relatively little impact on the
overall performance of loans FHA insured because FHA insured relatively
few ARMs. However, as shown in figure 4, later in the decade ARMs
represented a greater share of the loans that FHA insured, so their
performance became a more important factor affecting the overall
performance of FHA loans. FHA is studying its ARM program and has
contracted with a private consulting firm to examine the program’s
design and performance.
Figure 4: Adjustable Rate Mortgages as Share of All FHA Loans
Originated during Fiscal Years 1990-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
FHA Foreclosure Rates Have Been Particularly High in California:
FHA insured a greater dollar value of loans in the 1990s in California
than in any other state. Among the states in which FHA does the largest
share of its business, 4-year cumulative foreclosure rates for both
long-term, fixed-rate mortgages and ARMs were typically highest in
California. California, which accounted for 15 percent of the dollar
value of all single-family loans that FHA insured during the 1990s, had
an average foreclosure rate of
6.41 percent for both fixed rate and ARMs. In comparison, the 4-year
cumulative foreclosure rate for FHA loans insured during the 1990s
outside of California averaged 1.97 percent. According to FHA, the poor
performance of FHA loans originated in California was attributable to
poor economic conditions that existed during the early-to mid-1990s,
coupled with the practice of combining FHA’s interest-rate buy-down
program with an ARM to qualify borrowers in California’s high-priced
housing market.[Footnote 11]
The five states with the greatest dollar value of long-term fixed-rate
mortgages insured by FHA during the 1990s were California, Texas,
Florida, New York, and Illinois. Loans insured in these states made up
about one-third of FHA’s business for this loan type from fiscal year
1990 through fiscal year 1998, with California alone accounting for
about 13 percent, as shown in figure 5. As a result, the performance of
loans insured in California can significantly affect the overall
performance of FHA’s portfolio of loans of this type.
Figure 5: Share in Selected States of FHA Long-Term, Fixed-Rate Loans
Originated during Fiscal Years 1990-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
For long-term fixed-rate mortgages that FHA insured in California from
fiscal year 1990 through fiscal year 1998, the 4-year cumulative
foreclosure rates averaged about 5.6 percent. As shown in figure 6,
Florida, Texas, and New York also had relatively high 4-year
foreclosure rates during the early 1990s. And Florida experienced
relatively high 4-year cumulative foreclosure rates again from 1995
through 1998. For states that were not among the five states with the
greatest share of fixed-rate mortgages, the 4-year cumulative
foreclosure rates for the same type of loan over the same period
averaged less than 2 percent.
Figure 6: National 4-Year Cumulative Foreclosure Rates in Selected
States for FHA Long-Term, Fixed-Rate Loans Originated during Fiscal
Years 1990-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
The four states with the highest dollar value of ARMs insured by FHA
during the 1990s were California, Illinois, Maryland, and Colorado.
Loans insured in these states made up about 42 percent of FHA’s
business for this loan type, with California alone accounting for about
21 percent, as shown in figure 7. As a result, the performance of ARMs
insured in California can significantly affect the overall performance
of FHA’s portfolio of loans of this type.
Figure 7: Share of FHA Adjustable Rate Mortgages, in Selected States,
Originated during Fiscal Years 1990-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
As shown in figure 8, the 4-year cumulative foreclosure rates for ARMs
that FHA insured in California were consistently higher than the rates
for any of the other three states with the largest dollar volume of
ARMs insured by FHA, as well as the average rate for the remaining 46
states and the District of Columbia combined. In fact, for ARMs that
FHA insured in California in fiscal years 1995 and 1996, the 4-year
cumulative foreclosure rate was about 10 percent, more than twice as
high as the rate for any of the other three states with the highest
dollar volume of loans or for the remaining 46 states and the District
of Columbia combined.
Figure 8: National 4-Year Cumulative Foreclosure Rates in Selected
States for FHA Adjustable Rate Mortgages Originated during Fiscal Years
1990-1998:
[See PDF for image]
Source: GAO analysis of FHA data.
[End of figure]
Difference in LTV Ratios Can Explain Part but Not All of the Difference
in Foreclosure Rates:
Although differences in the share of FHA-insured loans with high LTVs
(above 95 percent) may be a factor accounting for part of the
difference in cumulative foreclosure rates between more recent loans
and loans insured earlier in the 1990s, the same pattern exists even
when differences in LTV are taken into account. As shown in figure 9,
the share of FHA-insured loans with LTVs of 95 percent or more was
higher later in the 1990s.[Footnote 12]
Figure 9: Share of FHA Loans within Various LTV Categories for Loans
Originated during Fiscal Years 1990-1998:
[See PDF for image]
Note: Excludes loans whose LTV equals zero.
Source: GAO analysis of FHA data.
[End of figure]
Generally, as shown in figure 10, higher LTV ratios, which measure
borrowers’ initial equity in their homes, are associated with higher
foreclosure rates.[Footnote 13] However, figure 10 also shows that the
same general pattern over time for the 4-year cumulative foreclosure
rates that was shown in figure 1 continues to exist even when the loans
are divided into categories by LTV.[Footnote 14] Thus, differences in
LTV alone cannot account for the observed differences in foreclosure
rates.
Figure 10: National 4-Year Cumulative Foreclosure Rates for Selected
LTV Classes of Long-Term, Fixed-Rate FHA Mortgages Originated during
Fiscal Years 1990-1998:
[See PDF for image]
Note: Excludes loans whose LTV equals zero. These loans showed a
similar pattern of foreclosure rates.
Source: GAO analysis of FHA data.
[End of figure]
Finally, we also considered whether the differences in foreclosures
rates could be explained by differences in prepayment rates. Higher
prepayment rates might be associated with lower foreclosure rates: if a
higher percentage of loans in a book of business are prepaid, then only
a smaller share of the original book of business might be subject to
foreclosure. However, we found that during the 1990s, prepayment rates
showed the same pattern across the years as foreclosure rates and, if
anything, were generally higher when foreclosure rates were higher,
suggesting that less frequent prepayment was not a factor explaining
higher foreclosure rates in the late 1990s.
Program-and Market-Related Changes that Could Explain Higher
Foreclosure Rates:
Although economic factors such as house-price-appreciation rates are
key determinants of mortgage foreclosure, a number of program-and
market-related changes occurring since 1995 could also affect the
performance of recently insured FHA loans. Specifically, in 1995 FHA
made a number of changes in its single-family insurance program that
allow borrowers who otherwise might not have qualified for home loans
to obtain FHA-insured loans. These changes also allow qualified
borrowers to increase the amount of loan for which they can qualify.
According to HUD, these underwriting changes were designed to expand
homeownership opportunities by eliminating unnecessary barriers to
potential homebuyers. The proportion of FHA purchase-mortgages made to
first-time homebuyers increased from 65 percent in 1994 to 78 percent
at the end of March 2002 and the proportion of FHA purchase-mortgages
made to minority homebuyers increased from 25 percent to 42 percent. At
the same time, there has been increased competition from private
mortgage insurers offering mortgages with low down payments to
borrowers identified as relatively low risk. The combination of changes
in FHA’s program and the increased competition in the marketplace may
partly explain the higher foreclosure rates of FHA loans originated
since fiscal year 1995. FHA has since made changes that may reduce the
likelihood of mortgage default, including requiring that, when
qualifying an FHA borrower for an ARM, the lender use the ARM’s second
year mortgage rate rather than the first-year rate. In addition, FHA
has implemented a new loss-mitigation program.[Footnote 15] Because
certain data that FHA collects on individual loans have not been
collected for a sufficient number of years or in sufficient detail, we
were unable to estimate the effect of changes in FHA’s program and
competition from conventional lenders on FHA loan performance.
Changes in FHA’s Underwriting Guidelines Could Have Resulted in Higher
Foreclosure Rates:
FHA issued revised underwriting guidelines in fiscal year 1995 that,
according to HUD, represented significant underwriting changes that
would enhance the homebuying opportunities for a substantial number of
American families.[Footnote 16] These underwriting changes made it
easier for borrowers to qualify for loans and allowed borrowers to
qualify for higher loan amounts. However, the changes may also have
increased the likelihood of foreclosure. The loans approved with more
liberal underwriting standards might, over time, perform worse relative
to existing economic conditions than those approved with the previous
standards. The revised standards decreased what is included as
borrowers’ debts and expanded the definition of what can be included as
borrowers’ effective income when lenders calculate qualifying
ratios.[Footnote 17] In addition, the new underwriting standards
expanded the list of compensating factors that could be considered in
qualifying a borrower, and they relaxed the standards for evaluating a
borrower’s credit history.
FHA Has Changed How It Defines Long-Term Debt:
The underwriting changes that FHA implemented in 1995 can decrease the
amount of debt that lenders consider in calculating one of the
qualifying ratios, the debt-to-income ratio, which is a measure of the
borrower’s ability to pay debt obligations. This change results in some
borrowers having a lower debt-to-income ratio than they would otherwise
have, and it increases the mortgage amount for which these borrowers
can qualify. For example, childcare expenses were considered a
recurring monthly debt in the debt-to-income ratio prior to 1995, but
FHA no longer requires that these expenses be considered when
calculating the debt-to-income ratio.
Another change affecting the debt-to-income ratio is that only debts
extending 10 months or more are now included in the ratio; previously,
FHA required all debts extending 6 months or more to be included. As a
result of this change, borrowers can have short-term debts that might
affect their ability to meet their mortgage payments, but these debts
would not be included in the debt-to-income ratio. However, FHA does
encourage lenders to consider all of a borrower’s obligations and the
borrower’s ability to make mortgage payments immediately following
closing.
FHA Has Changed How It Defines Effective Income:
The 1995 changes not only decreased the amount of debt considered in
the debt-to-income ratio; they also increased the amount of income
consideredæincreasing the number of borrowers considered able to meet a
particular level of mortgage payments. When calculating a borrower’s
effective income, lenders consider the anticipated amount of income and
the likelihood of its continuance. Certain types of income that were
previously considered too unstable to be counted toward effective
income are now acceptable in qualifying a borrower. For example, FHA
previously required income to be expected to continue for 5 years in
order for it to be considered as effective income. Now income expected
to continue for 3 years can be used in qualifying a borrower.
Similarly, FHA now counts income from overtime and bonuses toward
effective income, as long as this income is expected to continue.
Before 1995, FHA required that such income be earned for 2 years before
counting it toward effective income.
FHA Uses Additional Compensating Factors to Qualify Borrowers:
If borrowers do not meet the qualifying ratio guidelines for a loan of
a given size, lenders may still approve them for an FHA-insured
mortgage of that size. FHA’s 1995 revised handbook on underwriting
standards adds several possible compensating factors or circumstances
that lenders may consider when determining whether a borrower is
capable of handling the mortgage debt. For example, lenders may
consider food stamps or other public benefits that a borrower receives
as a compensating factor increasing the borrower’s ability to pay the
mortgage. These types of benefits are not included as effective income,
but FHA believes that receiving food stamps or other public benefits
positively affects the borrower’s ability to pay the mortgage. Lenders
may also consider as a compensating factor a borrower’s demonstrated
history of being able to pay housing expenses equal to or greater than
the proposed housing expense. In FHA’s revised handbook, the section on
compensating factors now states, “If the borrower over the past 12 to
24 months has met his or her housing obligation as well as other debts,
there should be little reason to doubt the borrower’s ability to
continue to do so despite having ratios in excess of those
prescribed.”:
FHA Has Changed How It Evaluates Borrowers’ Past Credit History:
In addition to changes affecting borrowers’ qualifying ratios, the 1995
underwriting changes affected how FHA lenders are supposed to evaluate
credit history to determine a borrower’s willingness and ability to
handle a mortgage. As with qualifying ratios and compensating factors,
FHA relies on the lender’s judgment and interpretation to determine
prospective borrowers’ creditworthiness. The 1995 underwriting changes
affected FHA guidelines regarding unpaid federal liens as well as
credit and credit reports. Specifically, before 1995, borrowers were
ineligible for an FHA-insured mortgage if they were delinquent on any
federal debt or had any federal liens, including taxes, placed on their
property. Following the 1995 changes, borrowers may qualify for a loan
even if federal tax liens remain unpaid. FHA guidelines stipulate that
a borrower may be eligible as long as the lien holder subordinates the
tax lien to the FHA-insured mortgage. If the borrower is in a payment
plan to repay liens, lenders may also approve the mortgage if the
borrower meets the qualifying ratios calculated with these payments.
Finally, FHA expanded the options available to lenders to evaluate a
borrower’s credit history. The previous guidance on developing credit
histories mentions only rent and utilities as nontraditional sources of
credit history. Lenders can now elect to use a nontraditional mortgage
credit report developed by a credit reporting agency if no other credit
history exists.[Footnote 18] Lenders may also develop a credit history
by considering a borrower’s payment history for rental housing and
utilities, insurance, childcare, school tuition, payments on credit
accounts with local stores, or uninsured medical bills.[Footnote 19] In
general, FHA advises lenders that an individual with no late housing or
installment debt payments should be considered as having an acceptable
credit history.
Increased Competition and Changes in the Conventional Mortgage Market
Could Have Resulted in Higher FHA Foreclosure Rates:
Increased competition and recent changes in the conventional mortgage
market could also have resulted in FHA’s insuring relatively more loans
that carry greater risk. Homebuyers’ demand for FHA-insured loans
depends, in part, on the alternatives available to them. In recent
years, FHA’s competitors in the mortgage insurance market--private
mortgage insurers and conventional mortgage lenders--have increasingly
offered products that compete with FHA’s for those homebuyers who are
borrowing more than 95 percent of the value of their home. In addition,
automated underwriting systems and credit-scoring analytic software
such as those introduced by the Federal National Mortgage Association
(Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie
Mac) in 1996 are believed to be able to more effectively distinguish
low-risk loans for expedited processing. The improvement of
conventional lenders’ ability to identify low-risk borrowers might
increase the risk profile of FHA’s portfolio as lower-risk borrowers
choose conventional financing with private mortgage insurance, which is
often less expensive. In addition, by lowering the required down
payment, conventional mortgage lenders and private mortgage insurers
may have attracted some borrowers who might otherwise have insured
their mortgages with FHA. If, by selectively offering these low down
payment loans to better risk borrowers, conventional mortgage lenders
and private mortgage insurers were able to attract FHA’s lower-risk
borrowers, recent FHA loans with down payments of less than 5 percent
may be more risky on average than they have been historically. FHA is
taking some action to more effectively compete with the conventional
market. For example, FHA is attempting to implement an automated
underwriting system that could enhance the ability of lenders
underwriting FHA-insured mortgages to distinguish better credit risks
from poorer ones. Although this effort is likely to increase the speed
with which lenders process FHA-insured loans, it may not improve the
risk profile of FHA borrowers unless lenders can lower the price of
insurance for better credit risks.
FHA Has Taken Steps to Improve the Quality of Its Underwriting:
Since 1996, FHA has revised and tightened some guidelines, specifically
in underwriting ARMs, identifying sources of cash reserves and
requiring more documentation from lenders. These steps should reduce
the riskiness of loans that FHA insures. In a 1997 letter to lenders,
FHA expressed concern about the quality of the underwriting of ARMs,
particularly when a buy down is used, and reminded lenders that the
first-year mortgage-interest rate must be used when qualifying the
borrower (rather than the lower rate after the buy down). FHA also
stipulated that lenders should consider a borrower’s ability to absorb
increased payments after buy down periods. FHA also emphasized that
lenders should rarely exceed FHA’s qualifying ratio guidelines in the
case of ARMs. In 1998, seeing that borrowers were still experiencing
trouble handling increased payments after the buy down period, FHA
required borrowers to be qualified at the anticipated second-year
interest rate, or the interest rate they would experience after the buy
down expired, and it prohibited any form of temporary interest-rate buy
down on ARMs. These changes will likely reduce the riskiness of ARMs in
future books of business.
FHA has also required stricter documentation from lenders on the use of
compensating factors and gift letters in mortgage approvals. In a June
10, 1997, letter to lenders, FHA expressed concern about an increased
number of loans with qualifying ratios above FHA’s guidelines for which
the lender gave no indication of the compensating factors used to
justify approval of the loans. FHA emphasized in this letter that
lenders are required to clearly indicate which compensating factor
justified the approval of a mortgage and to provide their rationale for
approving mortgages above the qualifying ratios. Similarly, in an
effort to ensure that any gift funds a borrower has come from a
legitimate source, FHA has advised lenders of the specific information
that gift letters should contain and the precise process for verifying
the donor or source of the gift funds.
In 2000, FHA also tightened its guidelines on what types of assets can
be considered as cash reserves. Although cash reserves are not
required, lenders use cash reserves to assess the riskiness of loans.
FHA noticed that in some cases lenders considered questionable assets
as cash reserves. For example, lenders were overvaluing assets or
including assets such as 401(k)s or IRAs that were not easily converted
into cash. As a result, FHA strengthened its policy and required
lenders to judge the liquidity of a borrower’s assets when considering
a borrower’s cash reserves. The new policy requires lenders, when
considering an asset’s value, to account for any applicable taxes or
withdrawal penalties that borrowers may incur in converting the asset
to cash.
FHA Has Implemented a New Loss Mitigation Program that Could Reduce
Foreclosures and Foreclosure Losses:
In 1996 Congress passed legislation directing FHA to terminate its
Single-Family Mortgage Assignment Program.[Footnote 20] FHA ceased
accepting assignment applications for this program on April 26, 1996.
The same legislation authorized FHA to implement a new program that
included a range of loss mitigation tools designed to help borrowers
either retain their home’s or to dispose of their property in ways that
lessen the cost of foreclosure for both the borrowers and FHA.
Specifically, the loss mitigation program provides a number of options
for reducing losses, including special forbearance, loan modification,
partial claim, pre-foreclosure sale, and deed-in-lieu-of-foreclosure
(see table 1 for an explanation of these options). To encourage lenders
to engage in loss mitigation, FHA offers incentive payments to lenders
for completing each loss mitigation workout. In addition, lenders face
a variety of financial penalties for failing to engage in loss
mitigation. FHA’s loss mitigation program went into effect on November
12, 1996; however, use was initially fairly low, with only 6,764 loss
mitigation cases realized in fiscal year 1997, as lenders began to
implement the new approach. HUD experienced substantial growth in loss
mitigation claims over the next 4 fiscal years, with total claims
reaching 25,027 in fiscal year 1999 and 53,389 in fiscal year 2001. The
three loss mitigation tools designed to allow borrowers to remain in
their homesæspecial forbearance, loan modification, and partial
claimærealized the largest increase in use. In contrast, the use of
deed-in-lieu-of-foreclosure and pre-foreclosure sale, options
resulting in insurance claims against the Fund, declined.[Footnote 21]
Table 1: Description of FHA’s Loss Mitigation Tools Available to
Lenders:
Loss mitigation tool: Special forbearance; Type of action taken by
lender: The use of a long-term repayment plan that may provide for
reduced or suspended payments when there is a reasonable likelihood
that the borrower can resume normal payments.
Loss mitigation tool: Loan modification; Type of action taken by
lender: A permanent change in the term, interest rate, or loan type of
a mortgage to accommodate inclusion of the accumulated delinquency. The
new monthly payment may be higher or lower than the existing payment.
Loss mitigation tool: Partial claim; Type of action taken by lender:
Provides for funds to be advanced from the Fund to repay past amounts
due on the mortgage for a borrower. To be eligible for this option, a
borrower must have long-term financial stability to support the
mortgage debt but lack the resources to cure the delinquency.
Loss mitigation tool: Pre-foreclosure sale; Type of action taken by
lender: When the borrower is unable or unwilling to maintain ownership
and the market value of the property is less than the level of debt,
this option allows the borrower to sell the property and apply the
proceeds to retire the debt.
Loss mitigation tool: Deed-in-lieu-of-foreclosure; Type of action taken
by lender: If a pre-foreclosure sale is not feasible, the borrower may
deed the property to HUD to avoid foreclosure.
Source: An Assessment of FHA’s Single-Family Mortgage Insurance Loss
Mitigation Program: Final Report, Abt Associates Inc., November 30,
2000:
[End of table]
Existing Data Preclude a Full Assessment of the Impact of FHA Program
and Conventional Mortgage Market Changes on Mortgage Default Rates:
Existing FHA data are not adequate to assess the impact of both FHA
program changes and the changes in the conventional mortgage market on
FHA default rates. Adequately assessing the impact of those changes
would require detailed data on information used during loan
underwriting to qualify individual borrowers. Such data on qualifying
ratios, use of compensating factors, credit scores, and sources and
amount of income would allow FHA to assess how factors key to
determining the quality of its underwriting have changed over time. In
addition, these data could be used in a more comprehensive analysis of
the relationship among FHA foreclosures and FHA program design, the
housing market, and economic conditions. Some of the data required for
that type of assessment and analysis are not collected by FHA, while
other data elements have not been collected for a sufficient number of
years to permit modeling the impact of underwriting changes on loan
performance.
Since 1993, FHA has collected data on items such as payment-to-income
and debt-to-income ratios, monthly effective income, and total monthly
debt payments. However, FHA has not collected more detailed information
on individual components of income and debt, such as overtime, bonus
income, alimony and childcare payments, or length of terms for
installment debt. Nor does FHA collect information on the use by
lenders of compensating factors in qualifying borrowers for FHA
insurance. These data would be required, for example, to analyze the
impact on loan performance of underwriting changes that FHA implemented
in 1995.
One of the most important measures of a borrower’s credit risk is the
borrower’s credit score. Lenders began using credit scores to assess a
borrower’s likelihood of default in the mid-1990s. In March 1998, FHA
approved Freddie Mac’s automated underwriting system for use by lenders
in making FHA-insured loans and began collecting data on borrower
credit scores for those loans underwritten using the system. Similarly,
in August 1999 FHA approved the use of Fannie Mae’s and PMI Mortgage
Servicers’ automated underwriting systems, and it currently collects
credit scores on loans underwritten using these systems. According to
HUD officials, FHA plans to begin collecting credit score data on all
FHA-insured loans underwritten through either automated underwriting
systems or conventional methods.
Finally, because of the newness of FHA’s loss mitigation program and
the several years required for a loan delinquency to be completely
resolved, it is difficult to measure the impact that loss mitigation
activities will ultimately have on the performance of FHA loans. As
recently as 2000, substantial revisions to the program were made that
could improve the program’s effectiveness according to Abt Associates
Inc.[Footnote 22] A recent audit of the program by HUD’s Office of
Inspector General noted the large increase in usage of loss mitigation
strategies and concluded that the program is reducing foreclosures and
keeping families in their homes.
Performance of Recent Loans Suggests that FHA’s Portfolio May Be
Riskier than Previously Estimated:
The overall riskiness of FHA loans made in recent years appears to be
greater than we had estimated in our February 2001 report on the Mutual
Mortgage Insurance Fund, reducing to some extent the ability of the
Fund to withstand worse-than-expected loan performance.[Footnote 23]
Although more years of loan performance are necessary to make a
definitive judgment, factors not accounted for in the models that we
used for that report appear to be affecting the performance of loans
insured after 1995 and causing the overall riskiness of FHA’s portfolio
to be greater than we previously estimated. In that report we based our
estimate of the economic value of the Fund (as of the end of fiscal
year 1999), in part, on econometric models that we developed and used
to forecast future foreclosures and prepayments for FHA-insured loans
based on the historical experience of loans dating back to 1975.
However, a large share of the loans in FHA’s portfolio at that time
were originated in fiscal years 1998 and 1999, and therefore there was
little direct evidence of how those loans would perform. As a result,
at the time that we released that estimate we cautioned that recent
changes in FHA’s insurance program and the conventional mortgage
market, such as those discussed in the previous section, could be
causing recent loans to perform differently, even under the same
economic conditions, from earlier loans.
To estimate the potential impact of these changes, we first used our
previous model to develop estimates of the relationship between, on the
one hand, the probability of foreclosure and prepayment and, on the
other hand, key explanatory factors such as borrower equity and
unemployment for loans insured between fiscal years 1975 and
1995.[Footnote 24] On the basis of these estimates and of the actual
values beyond 1995 for key economic variables, such as interest and
unemployment rates and the rate of house price appreciation, we
forecasted the performance (both foreclosures and prepayments) of loans
that FHA insured from fiscal year 1996 through fiscal year 2001. We
then compared those forecasts with the actual experience of those
loans. (See app. II for a full discussion of our methodology.) As is
shown in figure 11, for each year’s book of business, we found that
cumulative foreclosure rates through the end of fiscal year 2001
exceeded our forecasted levels.[Footnote 25] For example, for the book
of business with the longest experience, loans insured in 1996, we
forecasted that the cumulative foreclosure rate through the end of
fiscal year 2001 would be 3.44 percent, but the actual foreclosure rate
was 5.81 percent. These results suggest that some factors other than
those accounted for in the model may be causing loans insured after
1995
to perform worse than would be expected based on the historical
experience
of older loans.[Footnote 26]
Figure 11: Actual and Forecasted Cumulative Foreclosure Rates for FHA
Loans Insured during Fiscal Years 1996-2001, as of September 30, 2001:
[See PDF for image]
Note: The number of years of data varies by book of business. For
example, there are up to 6 years of data on the performance of loans
originated in 1996, while there is only 1 year of data for loans
originated in 2001. Thus, the foreclosure rates for loans originated in
1996 represent 6-year cumulative foreclosure rates, while the
foreclosure rates for loans originated in 2001 represent 1-year
cumulative foreclosure rates.
Source: GAO analysis of FHA data.
[End of figure]
The fact that cumulative foreclosures for recent FHA-insured loans have
been greater than what would be anticipated from a model based on the
performance of loans insured from fiscal year 1975 through fiscal year
1995 suggests that the caution we expressed in our 2001 report about
the effect of recent changes in FHA’s insurance program and the
conventional mortgage market on the ability of the Fund to withstand
future economic downturns is still warranted. In particular, the
performance of loans insured in fiscal years 1998 and 1999, which
represented about one-third of FHA’s loan portfolio at the end of 1999,
could be worse than what we previously forecasted. In turn, lower
performance by these loans could affect the economic value of the Fund
and its ability to withstand future economic downturns.
To assess the extent of this effect, we would need to know the extent
to which the performance of loans insured in fiscal years 1998 and 1999
has been and will be worse than what we forecasted in developing our
previous estimate of the economic value of the Fund. Because loans
insured in fiscal years 1998 and 1999 have not completely passed
through the peak years for foreclosures,[Footnote 27] these loans’
foreclosures to date provide only a limited indication of their long-
term performance. We do, however, have a better indication of the long-
term performance of loans insured in fiscal years 1996 and 1997 because
they are older loans with more years of experience. The experience of
these loans suggests that changes that are not accounted for in our
models are causing these books of business to have higher foreclosure
rates than would be anticipated from a model based on the performance
of earlier loans. If loans insured in fiscal years 1998 and 1999 are
affected by changes that are not accounted for in our models in the
same way that loans insured in fiscal years 1996 and 1997 appear to be
affected, then the 1998 and 1999 loans will continue to have higher
cumulative foreclosure rates than we estimated. Higher foreclosure
rates, in turn, imply a lower economic value of the Fund, which is
generally estimated as a baseline value under an expected set of
economic conditions. With a lower baseline economic value of the Fund
under expected economic conditions, the Fund would be less able to
withstand adverse economic conditions.
To better understand the reasons for the increased risk of recently
originated FHA loans would require additional data on factors that
might explain loan performance--including qualifying ratios and credit
scores. Even if these historical data were available today, it is too
soon to estimate with confidence the impact that recent changes will
ultimately have on recently insured loans because many of these loans
have not yet reached the peak years when foreclosures usually occur.
Recently insured loans represent the majority of FHA’s portfolio. The
impact of underwriting changes and changes in the conventional mortgage
market on the riskiness of the portfolio is not fully understood.
Understanding this risk will give a better basis for determining
whether the Fund has an adequate capital ratio, and also whether
program changes are in order to adjust that level of risk.
Agency Comments and Our Evaluation:
We obtained written comments on a draft of this report from HUD
officials. The written comments are presented in appendix IV. Generally
HUD agreed with the report’s findings that the underwriting changes
made in 1995 likely increased the riskiness of FHA loans insured after
that year. HUD commented that fiscal year 1995 was the first year in
which FHA exceeded the 2 percent capital ratio mandated by the National
Affordable Housing Act of 1990. According to HUD, by making the 1995
underwriting changes FHA modestly increased the risk characteristics of
FHA loans and, by doing so, allowed FHA to achieve its mission of
increasing homeownership opportunities for underserved groups. HUD also
provided information, which has been incorporated into the final report
as appropriate, on the change in homeownership rates among underserved
groups since 1994.
As agreed with your offices, unless you publicly release its contents
earlier, we plan no further distribution of this report until 30 days
after its issuance date. At that time, we will send copies of this
report to the Ranking Minority Member of the House Subcommittee on
Housing and Community Opportunity and other interested members of
Congress and congressional committees. We will also send copies to the
HUD Secretary and make copies available to others upon request.
Please contact me or Mathew J. Scire at (202) 512-6794, or Jay Cherlow
at (202) 512-4918, if you or your staff have any questions concerning
this report. Key contributors to this report were Jill Johnson, DuEwa
Kamara, Mitch Rachlis, Mark Stover, and Pat Valentine.
Sincerely yours,
Richard J. Hillman
Director, Financial Markets and
Community Investment:
Signed by Richard J. Hillman:
[End of section]
Appendixes:
Appendix I: Scope and Methodology:
We initiated this review to determine (1) how the early performance of
FHA loans originated in recent years has differed from loans originated
in earlier years; (2) how changes in FHA’s program and the conventional
mortgage market might explain recent loan performance; and (3) if there
is evidence that factors affecting the performance of recent FHA loans
may be causing the overall riskiness of FHA’s portfolio to be greater
than what we previously estimated, and if so what effect this might
have on the ability of the Fund to withstand future economic downturns.
To address these objectives, we obtained and analyzed data on loans
insured by FHA from 1990 through 1998 by year of origination; by loan
type (fixed interest rates versus adjustable interest rates); by loan-
to-value ratio; and by location of the property, for selected states
that held the greatest share of FHA-insured loans. We compared the
foreclosure rates for the first 4 years of these loans. We selected a
4-year cumulative foreclosure rate as a basis for comparing books of
business because it best balanced the competing goals of having the
greatest number of observations and the greatest number of years of
foreclosure experience.[Footnote 28] We also interviewed HUD officials
and reviewed HUD mortgagee letters, trade literature, and publicly
available information on the conventional mortgage market. Finally,
using the model that we developed for our prior report and basing it on
the experience of FHA loans insured from fiscal years 1975 through
1995, we also compared the estimated and actual foreclosure rates
through 2001 of loans insured from fiscal years 1996 through 2001.
We worked closely with HUD officials and discussed the interpretation
of HUD’s data. Although we did not independently verify the accuracy of
the data, we did perform internal checks to determine (1) the extent to
which the data fields were coded; and (2) the reasonableness of the
values contained in the data fields. We checked the mean, median, mode,
skewness, and high and low values for each of the variables used.
We conducted our review in Washington, D.C., between July 2001 and June
2002 in accordance with generally accepted government auditing
standards.
[End of section]
Appendix II: Models Used to Forecast Defaults and Prepayments for FHA-
Insured Mortgages:
For an earlier report,[Footnote 29] we built econometric and cash flow
models to estimate the economic value of FHA’s Mutual Mortgage
Insurance Fund (Fund) as of the end of fiscal year 1999. In that
report, we acknowledged that factors not fully captured in our models
could affect the future performance of loans in FHA’s portfolio and,
therefore, the ability of the Fund to withstand worse-than-expected
economic conditions. In particular, we suggested that these factors
could include changes in FHA’s insurance program and the conventional
insurance market. For our current report we sought to assess whether
there is evidence that factors not captured in our previous model may
be causing the overall riskiness of FHA’s portfolio to be greater than
we previously estimated and, if so, would that have a substantial
effect on the ability of the Fund to withstand future economic
downturns. In this appendix, we describe how we conducted that
assessment.
Our basic approach was to (1) reestimate the econometric models built
for our previous report using the same specifications as before and
data on loans insured by FHA in all 50 states and the District of
Columbia, but excluding U.S. territories, from 1975 through 1995 (in
the previous report, we used data on loans originated through 1999);
(2) use the estimated coefficients and actual values of our explanatory
variables during the forecasted period to forecast foreclosures and
prepayments through fiscal year 2001 for loans insured from fiscal year
1996 through fiscal year 2001; and (3) compare the forecasted and
actual foreclosures and prepayments for these loans during that time. A
finding that our foreclosure model fit the data well for loans insured
from 1975 through 1995, but consistently underestimated foreclosure
rates for post-1995 loans, would suggest that there had been a
structural change in the post-1995 period not captured in our models
that might cause the future performance of FHA-insured loans to be
worse than we estimated for our previous report.
Our econometric models used observations on loan years--that is,
information on the characteristics and status of an insured loan during
each year of its life--to estimate conditional foreclosure and
prepayment probabilities.[Footnote 30] These probabilities were
estimated using observed patterns of prepayments and foreclosures in a
large set of FHA-insured loans. More specifically, our models used
logistic equations to estimate the logarithm of the odds
ratio,[Footnote 31] from which the probability of a loan’s payment (or
a loan’s prepayment) in a given year could be calculated. These
equations were expressed as a function of interest and unemployment
rates, the borrower’s equity (computed using a house’s price and
current and contract interest rates as well as a loan’s duration), the
loan-to-value (LTV) ratio, the loan’s size, the geographic location of
the house, and the number of years that the loan had been active. The
results of the logistic regressions were used to estimate the
probabilities of a loan being foreclosed or prepaid in each year.
We prepared separate estimates for fixed-rate mortgages, adjustable
rate mortgages (ARMs), and investor loans. The fixed-rate mortgages
with terms of 25 years or more (long-term loans) were divided between
those that were refinanced and those that were purchase money mortgages
(mortgages associated with home purchase). Separate estimates were
prepared for each group of long-term loans. Similarly, investor loans
were divided between mortgages that were refinanced and the loans that
were purchase money mortgages. We prepared separate estimates for each
group of investor loans (refinanced and purchase money mortgages). A
separate analysis was also prepared for loans with terms that were less
than 25 years (short-term loans).
A complete description of our models, the data that we used, and the
results that we obtained is presented in detail in the following
sections. In particular, this appendix describes (1) the sample data
that we used; (2) our model specification and the independent variables
in the regression models; and (3) the model results.
Data and Sample Selection:
For our analysis, we selected from FHA’s computerized files a 10
percent sample of records of mortgages insured by FHA from fiscal years
1975 through 1995 (1,046,916 loans). From the FHA records, we obtained
information on the initial characteristics of each loan, such as the
year of the loan’s origination and the state in which the loan
originated; LTV ratio; loan amount; and contract interest rates.
To describe macroeconomic conditions at the national and state levels,
we obtained data at the national level on quarterly interest rates for
30-year fixed-rate mortgages on existing housing, and at the state
level on annual civilian unemployment rates from DRI-WEFA.[Footnote 32]
We also used state level data from DRI-WEFA on median house prices to
compute house price appreciation rates by state. To adjust nominal loan
amounts for inflation, we used data from the 2000 Economic Report of
the President on the implicit price deflator for personal consumption
expenditures.
Specification of the Model:
People buy houses for consumption and investment purposes. Normally,
people do not plan to default on loans. However, conditions that lead
to defaults do occur. Defaults may be triggered by a number of events,
including unemployment, divorce, or death. These events are not likely
to trigger defaults if the owner has positive equity in his or her home
because the sale of the home with realization of a profit is preferable
to the loss of the home through foreclosure. However, if the property
is worth less than the mortgage, these events may trigger defaults.
Prepayments of home mortgages can also occur. These may be triggered by
events such as declining interest rates, which prompt refinancing, and
rising house prices, which prompt homeowners to take out accumulated
equity or sell the residence. Because FHA mortgages are assumable, the
sale of a residence does not automatically trigger prepayment. For
example, if interest rates have risen substantially since the time that
the mortgage was originated, a new purchaser may prefer to assume the
seller’s mortgage.
We hypothesized that foreclosure behavior is influenced by, among other
things, the (1) level of unemployment, (2) size of the loan, (3) value
of the home, (4) current interest rates, (5) contract interest rates,
(6) home equity, and (7) region of the country within which the home is
located. We hypothesized that prepayment behavior is influenced by,
among other things, the (1) difference between the interest rate
specified in the mortgage contract and the mortgage rates generally
prevailing in each subsequent year, (2) amount of accumulated equity,
(3) size of the loan, and (4) region of the country in which the home
is located.
Our first regression model estimated conditional mortgage foreclosure
probabilities as a function of a variety of explanatory variables. In
this regression, the dependent variable is a 0/1 indicator of whether a
given loan was foreclosed in a given year. The outstanding mortgage
balance, expressed in inflation-adjusted dollars, weighted each loan-
year observation.
Our foreclosure rates were conditional on whether the loan survives an
additional year. We estimated conditional foreclosures in a logistic
regression equation. Logistic regression is commonly used when the
variable to be estimated is the probability that an event, such as a
loan’s foreclosure, will occur. We regressed the dependent variable
(whose value is 1 if foreclosure occurs and 0 otherwise) on the
explanatory variables previously listed.
Our second regression model estimated conditional prepayment
probabilities. The independent variables included a measure that is
based on the relationship between the current mortgage interest rate
and the contract rate, the primary determinant of a mortgage’s
refinance activity. We further separated this variable between ratios
above and below 1 to allow for the possibility of different marginal
impacts in higher and lower ranges.
The variables that we used to predict foreclosures and prepayments fall
into two general categories: descriptions of states of the economy and
characteristics of the loan. In choosing explanatory variables, we
relied on the results of our own and others’ previous efforts to model
foreclosure and prepayment probabilities, and on implications drawn
from economic principles. We allowed for many of the same variables to
affect both foreclosure and prepayment.
Equity:
The single most important determinant of a loan’s foreclosure is the
borrower’s equity in the property, which changes over time because (1)
payments reduce the amount owed on the mortgage and (2) property values
can increase or decrease. Equity is a measure of the current value of a
property compared with the current value of the mortgage on that
property. Previous research strongly indicates that borrowers with
small amounts of equity, or even negative equity, are more likely than
other borrowers to default.[Footnote 33]
We computed the percentage of equity as 1 minus the ratio of the
present value of the loan balance evaluated at the current mortgage
interest rate, to the current estimated house price. For example, if
the current estimated house price is $100,000, and the value of the
mortgage at the current interest rate is $80,000, then equity is .2 (20
percent), or 1-(80/100). To measure current equity, we calculated the
value of the mortgage as the present value of the remaining mortgage,
evaluated at the current year’s fixed-rate mortgage interest rate. We
calculated the current value of a property by multiplying the value of
that property at the time of the loan’s origination by the change in
the state’s median nominal house price, adjusted for quality changes,
between the year of origination and the current year.[Footnote 34]
Because the effects on foreclosure of small changes in equity may
differ depending on whether the level of equity is large or small, we
used a pair of equity variables, LAGEQHIGH and LAGEQLOW,[Footnote 35]
in our foreclosure regression. The effect of equity is lagged 1 year,
as we are predicting the time of foreclosure, which usually occurs many
months after a loan first defaults.
We anticipated that higher levels of equity would be associated with an
increased likelihood of prepayment. Borrowers with substantial equity
in their homes may be more interested in prepaying their existing
mortgages, and may take out larger ones to obtain cash for other
purposes. Borrowers with little or no equity may be less likely to
prepay because they may have to take money from other savings to pay
off their loans and cover transaction costs.
For the prepayment regression, we used a variable that measures book
equity--the estimated property value less the amortized balance of the
loan--instead of market equity. It is book value, not market value,
that the borrower must pay to retire the debt.[Footnote 36]
Additionally, the important effect of interest rate changes on
prepayment is captured by two other equity variables, RELEQHI and
RELEQLO, which are sensitive to the difference between a loan’s
contract rate and the interest rate on 30-year mortgages available in
the current year. These variables are described below.
Loan-to-Value (LTV) Ratio:
We included an additional set of variables in our regressions related
to equity: the initial LTV ratio. We entered LTV as a series of dummy
variables, depending on its size. Loans fit into eight discrete LTV
categories. In some years, FHA measured LTV as the loan amount less
mortgage insurance premium financed in the numerator of the ratio, and
appraised value plus closing costs in the denominator. To reflect true
economic LTV, we adjusted FHA’s measure by removing closing costs from
the denominator and including financed premiums in the numerator.
A borrower’s initial equity can be expressed as a function of LTV, so
we anticipated that if LTV was an important predictor in an equation
that also includes a variable measuring current equity, it would
probably be positively related to the probability of foreclosure. One
reason for including LTV is that it measures initial equity accurately.
Our measures of current equity are less accurate because we do not have
data on the actual rate of change in the mortgage loan balance or the
actual rate of house price change for a specific house.
Loans with higher LTVs are more likely to foreclose. We used the lowest
LTV category as the omitted category. We expected LTV to have a
positive sign in the foreclosure equations at higher levels of LTV. LTV
in our foreclosure equations may capture the effects of income
constraints. We were unable to include borrowers’ income or payment to
income ratio directly because data on borrowers’ income were not
available.[Footnote 37] However, it seems likely that borrowers with
little or no down payment (high LTV) are more likely to be financially
stretched in meeting their payments and, therefore, more likely to
default. The anticipated relationship between LTV and the probability
of prepayment is uncertain.
For two equations--long-term refinanced loans and investor-refinanced
loans--we used down payment information directly, rather than the
series of LTV variables. We defined down payment to ensure that closing
costs were included in the loan amount and excluded from the house
price.
Unemployment:
We used the annual unemployment rates for each state for the period
from fiscal years 1975 through 1995 to measure the relative condition
of the economy in the state where a loan was made. We anticipated that
foreclosures would be higher in years and states with higher
unemployment rates, and that prepayments would be lower because
property sales slow down during recessions. The actual variable we used
in our regressions, LAGUNEMP, is defined as the logarithm of the
preceding year’s unemployment rate in that state.
Interest Rates:
We included the logarithm of the interest rate on the mortgage as an
explanatory variable in the foreclosure equation. We expected a higher
interest rate to be associated with a higher probability of foreclosure
because higher interest rates cause higher monthly payments. However,
in explaining the likelihood of prepayment, our model uses information
on the level of current mortgage rates relative to the contract rate on
the borrower’s mortgage. A borrower’s incentive to prepay is high when
the interest rate on a loan is greater than the rate at which money can
currently be borrowed, and it diminishes as current interest rates
increase. In our prepayment regression we defined two variables,
RELEQHI and RELEQLO. RELEQHI is defined as the ratio of the market
value of the mortgage to the book value of the mortgage, but is never
smaller than 1. RELEQLO is also defined as the ratio of the market
value of the mortgage to the book value, but is never larger than 1.
When currently available mortgage rates are lower than the contract
interest rate, market equity exceeds book equity because the present
value of the remaining payments evaluated at the current rate exceeds
the present value of the remaining payments evaluated at the contract
rate. Thus, RELEQHI captures a borrower’s incentive to refinance, and
RELEQLO captures a new buyer’s incentive to assume the seller’s
mortgage.
We created two 0/1 variables, REFIN and REFIN2, that take on a value of
1 if a borrower had not taken advantage of a refinancing opportunity in
the past, and 0 otherwise. We defined a refinancing opportunity as
having occurred if the interest rate on fixed-rate mortgages in any
previous year in which a loan was active was at least 200 basis
points[Footnote 38] below the rate on the mortgage in any year through
1994, or 150 basis points below the rate on the mortgage in any year
after 1994.[Footnote 39] REFIN takes a value of 1 if the borrower had
passed up a refinancing opportunity at least once in the past. REFIN2
takes on a value of 1 if the borrower had passed up two or more
refinancing opportunities in the past.
Several reasons might explain why borrowers passed up apparently
profitable refinancing opportunities. For example, if they had been
unemployed or their property had fallen in value, they might have had
difficulty obtaining refinancing. This reasoning suggests that REFIN
and REFIN2 would be positively related to the probability of
foreclosure; that is, a borrower unable to obtain refinancing
previously because of poor financial status might be more likely to
default.
Similar reasoning suggests a negative relationship between REFIN and
REFIN2 and the probability of prepayment; a borrower unable to obtain
refinancing previously might also be unlikely to obtain refinancing
currently. A negative relationship might also exist if a borrower’s
passing up one profitable refinancing opportunity reflected a lack of
financial sophistication that, in turn, would be associated with
passing up additional opportunities. However, a borrower who
anticipated moving soon might pass up an apparently profitable
refinancing opportunity to avoid the transaction costs associated with
refinancing. In this case, there might be a positive relationship, with
the probability of prepayment being higher if the borrower fulfilled
his or her anticipation and moved, thereby prepaying the loan.
Another explanatory variable is the volatility of interest rates,
INTVOL, which is defined as the standard deviation of the monthly
average of the Federal Home Loan Mortgage Corporation’s series of 30-
year, fixed-rate mortgages’ effective interest rates. We calculated the
standard deviation over the previous 12 months. Financial theory
predicts that borrowers are likely to refinance more slowly at times of
volatile rates because there is a larger incentive to wait for a still
lower interest rate.
We also included the slope of the yield curve, YC, in our prepayment
estimates, which we calculated as the difference between the 1-and 10-
year Treasury rates of interest. We then subtracted 250 basis points
from this difference and set differences that were less than 0 to 0.
This variable measured the relative attractiveness of ARMs versus
fixed-rate mortgages; the steeper the yield curve, the more attractive
ARMs would be. When ARMs have low rates, borrowers with fixed-rate
mortgages may be induced into refinancing into ARMs to lower their
monthly payments.
For ARMs, we did not use relative equity variables as we did with
fixed-rate mortgages. Instead, we defined four variables, CHANGEPOS,
CHANGENEG, CAPPEDPOS, and CAPPEDNEG to capture the relationship between
current interest rates and the interest rate paid on each mortgage.
CHANGEPOS measures how far the interest rate on the mortgage has
increased since origination, with a minimum of 0, while CHANGENEG
measures how far the rate has decreased, with a maximum of 0. CAPPEDPOS
measures how much further the interest rate on the mortgage would rise
if prevailing interest rates in the market did not change, while
CAPPEDNEG measures how much further the mortgage’s rate would fall if
prevailing interest rates did not change. For example, if an ARM was
originated at 7 percent and interest rates increased by 250 basis
points 1 year later, CHANGEPOS would equal 100 because FHA’s ARMs can
increase by no more than 100 basis points in a year. CAPPEDPOS would
equal 150 basis points, since the mortgage rate would eventually
increase by another 150 basis points if market interest rates did not
change, and CHANGENEG and CAPPEDNEG would equal 0. Because interest
rates have generally trended downward since FHA introduced ARMs, there
is very little experience with ARMs in an increasing interest rate
environment.
Geographic Regions:
We created nine 0/1 variables to reflect the geographic distribution of
FHA loans, and included them in both regressions. Location differences
may capture the effects of differences in borrowers’ incomes,
underwriting standards by lenders, economic conditions not captured by
the unemployment rate, or other factors that may affect foreclosure and
prepayment rates. We assigned each loan to one of the nine Bureau of
the Census (Census) divisions on the basis of the state in which the
borrower resided. The Pacific division was the omitted category; that
is, the regression coefficients show how each of the regions was
different from the Pacific division. We also created a variable,
JUDICIAL, to indicate states that allowed judicial foreclosure
procedures in place of nonjudicial foreclosures. We anticipated that
the probability of foreclosure would be lower where judicial
foreclosure procedures were allowed because of the greater time and
expense required for the lender to foreclose on a loan.
Loan Size:
To obtain an insight into the differential effect of relatively larger
loans on mortgage foreclosures and prepayments, we assigned each loan
to 1 of 10 loan-size categorical variables (LOAN1 to LOAN10). The
omitted category in our regressions was that of loans between $80,000
and $90,000, and results on loan size are relative to those loans
between $80,000 and $90,000. All dollar amounts are inflation adjusted
and represent 1999 dollars.
Number of Units:
The number of units covered by a single mortgage was a key determinant
in deciding which loans were more likely to be investor loans. Loans
were noted as investor loans if the LTV ratio was between specific
values, depending on the year of the loan or whether there were two or
more units covered by the loan. Once a loan was identified as an
investor loan, we separated the refinanced loans from the purchase-
money mortgages and performed foreclosure and payoff analyses on each.
For each of the investor equations, we used two dummy variables defined
according to the number of units in the dwelling. LIVUNT2 has the value
of 1 when a property has two dwelling units and a value of 0 otherwise.
LIVUNT3 has a value of 1 when a property has three or more dwelling
units and a value of 0 otherwise. The missing category in our
regressions was investors with one unit. Our database covers only loans
with no more than four units.
Policy Year and Refinance Indicator:
To capture the time pattern of foreclosures and prepayments (given the
effects of equity and the other explanatory variables), we defined
seven variables on the basis of the number of years that had passed
since the year of the loan’s origination. We refer to these variables
as YEAR1 to YEAR7 and set them equal to 1 during the corresponding
policy year and 0 otherwise. Finally, for those loan type categories
for which we did not estimate separate models for refinancing loans and
nonrefinancing loans, we created a variable called REFINANCE DUMMY to
indicate whether a loan was a refinancing loan.
Table 2 summarizes the variables that we used to predict foreclosures
and prepayments. Table 3 presents mean values for our predictor
variables for each mortgage type for which we ran a separate
regression.
Table 2: Variable Names and Descriptions:
Loan size dummy variables:
Variable name: LOAN1; Variable description: Loan size dummy variables:
1 if loan amount is less than $40,000, else 0.
Variable name: LOAN2; Variable description: Loan size dummy variables:
1 if loan amount is $40,000 or above but below $50,000, else 0.
Variable name: LOAN3; Variable description: Loan size dummy variables:
1 if loan amount is $50,000 or above but below $60,000, else 0.
Variable name: LOAN4; Variable description: Loan size dummy variables:
1 if loan amount is $60,000 or above but below $70,000, else 0.
Variable name: LOAN5; Variable description: Loan size dummy variables:
1 if loan amount is $70,000 or above but below $80,000, else 0.
Variable name: LOAN6; Variable description: Loan size dummy variables:
1 if loan amount is $80,000 or above but below $90,000, else 0.
Variable name: LOAN7; Variable description: Loan size dummy variables:
1 if loan amount is $90,000 or above but below $100,000, else 0.
Variable name: LOAN8; Variable description: Loan size dummy variables:
1 if loan amount is $100,000 or above but below $110,000, else 0.
Variable name: LOAN9; Variable description: Loan size dummy variables:
1 if loan amount is $110,000 or above but below $130,000, else 0.
Variable name: LOAN10; Variable description: Loan size dummy variables:
1 if loan amount is at least $130,000, else 0.
Economic Variables:
Variable name: LOGINT; Variable description: Loan size dummy variables:
Log of the contract mortgage interest rate.
Variable name: REFINANCE DUMMY; Variable description: Loan size dummy
variables: 1 if the loan is a refinancing loan, else 0.
Variable name: RELEQLO; Variable description: Loan size dummy
variables: The ratio of the market value of the mortgage to the book
value if the market value is below the book value, else 1.
Variable name: RELEQHI; Variable description: Loan size dummy
variables: The ratio of the market value of the mortgage to the book
value if the market value is above the book value, else 1.
Variable name: REFIN; Variable description: Loan size dummy variables:
1 if, in at least 1 previous year, the mortgage interest rate had been
at least 200 basis points below the contract rate in any year prior to
1995 or 150 basis points below the contract rate after 1994 and the
borrower had not refinanced, else 0.
Variable name: REFIN2; Variable description: Loan size dummy variables:
1 if, in at least 2 previous years the above situation prevailed, else
0.
Variable name: INTVOL; Variable description: Loan size dummy variables:
The volatility of mortgage rates, defined as the standard deviation of
30-year fixed-rate mortgage interest rates over the previous 12 months.
Variable name: YC; Variable description: Loan size dummy variables: The
slope of the yield curve, defined as the difference between 1-and 10-
year U.S. Treasury interest rates minus 250 basis points, but not less
than 0.
Variable name: LIVUNT2; Variable description: Loan size dummy
variables: 1 if the property has two housing units, else 0.
Variable name: LIVUNT3; Variable description: Loan size dummy
variables: 1 if the property has three or more housing units, else 0.
Variable name: LAGUNEM; Variable description: Loan size dummy
variables: The log of the previous year’s unemployment rate in each
state.
Variable name: JUDICIAL; Variable description: Loan size dummy
variables: 1 if state allowed judicial foreclosure (list of states
varies by year), else 0.
Policy Year Dummy Variables:
Variable name: YEAR1; Variable description: Loan size dummy variables:
1 if in loan’s first year, else 0.
Variable name: YEAR2; Variable description: Loan size dummy variables:
1 if in loan’s second year, else 0.
Variable name: YEAR3; Variable description: Loan size dummy variables:
1 if in loan’s third year, else 0.
Variable name: YEAR4; Variable description: Loan size dummy variables:
1 if in loan’s fourth year, else 0.
Variable name: YEAR5; Variable description: Loan size dummy variables:
1 if in loan’s fifth year, else 0.
Variable name: YEAR6; Variable description: Loan size dummy variables:
1 if in loan’s sixth year, else 0.
Variable name: YEAR7; Variable description: Loan size dummy variables:
1 if in loan’s seventh year, else 0.
Loan-to-value dummy variables:
Variable name: LTV0; Variable description: Loan size dummy variables: 1
if LTV equals 0, assumed missing data, else 0.
Variable name: LTV1; Variable description: Loan size dummy variables: 1
if LTV is above 0 and less than 60, else 0.
Variable name: LTV2; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 60, but less than 85, else 0.
Variable name: LTV3; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 85, but less than 92, else 0.
Variable name: LTV4; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 92, but less than 96, else 0.
Variable name: LTV5; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 96, but less than 98, else 0.
Variable name: LTV6; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 98, but less than 100, else 0.
Variable name: LTV7; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 100, but less than 102, else 0.
Variable name: LTV8; Variable description: Loan size dummy variables: 1
if LTV is greater than or equal to 102, but less than 106, else 0.
Equity variables:
Variable name: LAGEQLOW; Variable description: Loan size dummy
variables: The lagged value of market equity (defined as 1 minus the
ratio of the present value of the loan balance, evaluated at the
current mortgage interest rate, to the current estimated house price)
if equity is less than or equal to 20 percent, else .2.
Variable name: LAGEQHIGH; Variable description: Loan size dummy
variables: The lagged value of market equity (defined as 1 minus the
ratio of the present value of the loan balance, evaluated at the
current mortgage interest rate, to the current estimated house price
minus .2) if equity is greater than 20 percent, else 0.
Variable name: BOOKNEG; Variable description: Loan size dummy
variables: The lagged value of book equity (defined as 1 minus the
ratio of the amortized loan balance to the current estimated house
price) if equity is less than or equal to 20 percent, else .2.
Variable name: BOOKPOS; Variable description: Loan size dummy
variables: The lagged value of book equity (defined as 1 minus the
ratio of the amortized loan balance to the current estimated house
price minus .2) if equity is greater than 20 percent, else 0.
Variable name: CHANGEPOS; Variable description: Loan size dummy
variables: The amount by which the interest rate of an ARM has
increased since origination, with a minimum of 0.
Variable name: CHANGENEG; Variable description: Loan size dummy
variables: The amount by which the interest rate of an ARM has
decreased since origination, with a maximum of 0.
Variable name: CAPPEDPOS; Variable description: Loan size dummy
variables: The amount by which the interest rate of an ARM could still
rise, if prevailing interest rates in the market did not change, with a
minimum of 0.
Variable name: CAPPEDNEG; Variable description: Loan size dummy
variables: The amount by which the interest rate of an ARM could still
decline, if prevailing interest rates in the market did not change,
with a maximum of 0.
Variable name: DOWNPAY; Variable description: Loan size dummy
variables: The down payment, expressed as a percentage of the purchase
price of the house; closing costs were excluded from the house price
and included in the loan amount.
Census division dummy variables:
Variable name: DV_A[A]; Variable description: Loan size dummy
variables: 1 if the loan is in the Mid-Atlantic states (NY, PA, NJ),
else 0.
Variable name: DV_E; Variable description: Loan size dummy variables: 1
if the loan is in the East South Central states (KY, TN, AL, MS), else
0.
Variable name: DV_G; Variable description: Loan size dummy variables: 1
if the loan is in the West North Central states (MN, MO, IA, NB, KS,
SD, ND), else 0.
Variable name: DV_M; Variable description: Loan size dummy variables: 1
if the loan is in the Mountain states (CO, UT, AZ, NM, NV, ID, WY, MT),
else 0.
Variable name: DV_N; Variable description: Loan size dummy variables: 1
if the loan is in the New England states (MA, CT, RI, NH, ME, VT), else
0.
Variable name: DV_P; Variable description: Loan size dummy variables: 1
if the loan is in the Pacific states (CA, OR, WA), else 0.
Variable name: DV_R; Variable description: Loan size dummy variables: 1
if the loan is in the East North Central states (IL, MI, OH, IN, WI),
else 0.
Variable name: DV_S; Variable description: Loan size dummy variables: 1
if the loan is in the South Atlantic states (FL, GA, NC, SC, VA, MD,
DC, DE, WV), else 0.
Variable name: DV_W; Variable description: Loan size dummy variables: 1
if the loan is in the West South Central states (TX, OK, LA, AR), else
0.
[A] DV = Division :
Source: U.S. General Accounting Office. :
[End of table]
Table 3: Means of Predictor Variables:
[See PDF for image]
[End of table]
Estimation Results:
As previously described, we used logistic regressions to model loan
foreclosures and prepayments as a function of a variety of predictor
variables. We estimated separate regressions for fixed-rate purchase
money mortgages (and refinanced loans) with terms over and under 25
years, ARMs, and investor loans. We used data on loan activity
throughout the life of the loans for loans originated from fiscal years
1975 through 1995. The outstanding loan balance of the observation
weighted the regressions.
The logistic regressions estimated the probability of a loan being
foreclosed or prepaid in each year. The standard errors of the
regression coefficients are biased downward, because the errors in the
regressions are not independent. The observations are on loan years,
and the error terms are correlated because the same underlying loan can
appear several times. However, we did not view this downward bias as a
problem because our purpose was to forecast the dependent variables,
not to test hypotheses concerning the effects of independent variables.
In general, our results are consistent with the economic reasoning that
underlies our models. Most important, the probability of foreclosure
declines as equity increases, and the probability of prepayment
increases as the current mortgage interest rate falls below the
contract mortgage interest rate. As shown in tables 4 and 5, both of
these effects occur in each regression model and are very strong. These
tables present the estimated coefficients for all of the predictor
variables for the foreclosure and prepayment equations.
Table 4 shows our foreclosure regression results. As expected, the
unemployment rate is positively related to the probability of
foreclosure and negatively related to the probability of prepayment.
Our results also indicate that generally the probability of foreclosure
is higher when LTV and contract interest rate are higher. The overall
quality of fit was satisfactory: Chi-square statistics were significant
on all regressions at the 0.01-percent level.
Because the coefficients from a nonlinear regression can be difficult
to interpret, we transformed some of the coefficients for the long-
term, nonrefinanced, fixed-rate regressions into statements about
changes in the probabilities of foreclosure and prepayment. The overall
conditional foreclosure probability for this mortgage type is estimated
to be about 0.6 percent.[Footnote 40],[Footnote 41] In other words, on
average, there is a 6/10 of a 1 percent chance for a loan of this type
to result in a claim payment in any particular year.[Footnote 42] By
holding other predictor variables at their mean values, we can describe
the effect on the conditional foreclosure probability of changes in the
values of predictor variables of interest. For example, if the average
value of the unemployment rate were to increase by 1 percentage point
from its mean value (in our sample) of about 6 percent to about 7
percent, the conditional foreclosure probability would increase by
about 17 percent (from 0.6 percent to about 0.7 percent). Similarly, a
1 percentage-point increase in the mortgage contract rate from its mean
value of about 9.4 percent to about 10.4 percent would also raise the
conditional foreclosure probability by 17 percent (from about 0.6
percent to about 0.7 percent). Values of homeowners’ equity of 10
percent, 20 percent, 30 percent, and 40 percent result in conditional
foreclosure probabilities of 0.7 percent, 0.5 percent, 0.3 percent, and
0.2 percent, respectively, illustrating the importance of increased
equity in reducing the probability of foreclosure.
Table 5 shows our prepayment regression results. The overall
conditional prepayment probability for long-term, fixed-rate mortgages
is estimated to be about 5.0 percent. This means that, in any
particular year, about 5 percent of the loan dollars outstanding will
prepay, on average.[Footnote 43] Prepayment probability is quite
sensitive to the relationship between the contract interest rate and
the currently available mortgage rate. We modeled this relationship
using RELEQHI and RELEQLO. Holding other variables at their mean
values, if the spread between mortgage rates available in each year and
the contract interest rate widened by 1 percentage point, the
conditional prepayment probability would increase by about 78.5 percent
to about 8.9 percent.
Table 4: Coefficients from Foreclosure Equations and Summary
Statistics:
[See PDF for image]
[End of table]
Table 5: Coefficients from Prepayment Equations and Summary Statistics:
[See PDF for image]
[End of table]
Model Predictions for Historical Period:
To test the validity of our models, we examined how well they predicted
actual patterns of FHA’s foreclosure and prepayment rates through
fiscal year 1995. Using a sample of 10 percent of FHA’s loans made from
fiscal years 1975 through 1995, we found that our predicted rates
closely resembled actual rates.
To predict the probabilities of foreclosure and prepayment in the
historical period, we combined the models’ coefficients with
information on a loan’s characteristics and information on economic
conditions described by our predictor variables in each year from a
loan’s origination through fiscal year 1995. If our models predicted
foreclosure or prepayment in any year, we determined the loan’s balance
during that year to indicate the dollar amount associated with the
foreclosure or prepayment. We estimated cumulative foreclosure and
prepayment rates by summing the predicted claim and prepayment dollar
amounts for all loans originated in each of the fiscal years 1975
through 1995. We compared these predictions with the actual cumulative
(through fiscal year 1995) foreclosure and prepayment rates for the
loans in our sample. Figure 12 compares actual and predicted cumulative
foreclosure rates, and figure 13 compares actual and predicted
cumulative prepayment rates for long-term, fixed-rate, nonrefinanced
mortgages.[Footnote 44]
Figure 12: Cumulative Foreclosure Rates by Book of Business for 30-
Year, Fixed-Rate, Nonrefinanced Mortgages, Actual and Predicted, Fiscal
Years 1975-1995:
[See PDF for image]
Source: GAO analysis of HUD data.
[End of figure]
Figure 13: Cumulative Prepayment Rates by Book of Business for 30-Year,
Fixed-Rate, Nonrefinanced Mortgages, Actual and Predicted, Fiscal Years
1975-1995:
[See PDF for image]
Source: GAO analysis of HUD data.
[End of figure]
[End of section]
Appendix III: Data for Figures Used in This Report:
Foreclosure rates in the following tables are expressed as a percentage
of loan amounts. Specifically, for tables 6 through 15 we compute all
rates using the original loan amount of the foreclosed loans compared
to the original loan amount of like loans insured by FHA for the
corresponding year. For tables 16 we compute foreclosure rates using
the unpaid balance of foreclosed loans as a percentage of the total
value of mortgages originated.
Table 6: National 4-Year Cumulative Foreclosure Rates for All FHA Loans
Originated during Fiscal Years 1990-1998 (Figure 1):
Year of origination: 1990; Foreclosure rate: 2.87%; Original amount of
foreclosed loan: $1,468,904,919; Total loans originated:
$51,171,603,963.
Year of origination: 1991; Foreclosure rate: 2.78; Original amount of
foreclosed loan: 1,334,851,353; Total loans originated:
47,977,729,478.
Year of origination: 1992; Foreclosure rate: 1.86; Original amount of
foreclosed loan: 923,919,357; Total loans originated: 49,542,579,739.
Year of origination: 1993; Foreclosure rate: 1.69; Original amount of
foreclosed loan: 1,367,705,598; Total loans originated:
80,735,908,098.
Year of origination: 1994; Foreclosure rate: 2.24; Original amount of
foreclosed loan: 1,956,485,804; Total loans originated:
87,234,242,852.
Year of origination: 1995; Foreclosure rate: 3.30; Original amount of
foreclosed loan: 1,517,690,292; Total loans originated:
46,021,098,615.
Year of origination: 1996; Foreclosure rate: 3.34; Original amount of
foreclosed loan: 2,294,973,060; Total loans originated:
68,615,725,261.
Year of origination: 1997; Foreclosure rate: 3.16; Original amount of
foreclosed loan: 2,297,495,007; Total loans originated:
72,668,032,499.
Year of origination: 1998; Foreclosure rate: 2.28; Original amount of
foreclosed loan: 2,232,185,460; Total loans originated:
97,830,968,343.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 7: National 4-Year Cumulative Foreclosure Rates for Long-Term,
Fixed Rate Loans Originated during Fiscal Years 1990-1998 (Figure 2):
Year of origination: 1980; Foreclosure rate: 3.33%; Original amount:
$340,425,000; Total loans originated: $10,235,649,629.
Year of origination: 1981; Foreclosure rate: 7.42; Original amount:
578,087,000; Total loans originated: 7,788,823,419.
Year of origination: 1982; Foreclosure rate: 9.94; Original amount:
569,819,000; Total loans originated: 5,735,087,556.
Year of origination: 1983; Foreclosure rate: 5.02; Original amount:
1,200,882,000; Total loans originated: 23,930,937,692.
Year of origination: 1984; Foreclosure rate: 8.11; Original amount:
1,154,103,000; Total loans originated: 14,231,238,175.
Year of origination: 1985; Foreclosure rate: 7.85; Original amount:
1,782,238,000; Total loans originated: 22,708,988,850.
Year of origination: 1986; Foreclosure rate: 4.34; Original amount:
2,468,155,000; Total loans originated: 56,917,684,653.
Year of origination: 1987; Foreclosure rate: 2.74; Original amount:
1,914,245,000; Total loans originated: 69,782,899,762.
Year of origination: 1988; Foreclosure rate: 3.26; Original amount:
1,208,982,000; Total loans originated: 37,113,171,210.
Year of origination: 1989; Foreclosure rate: 3.07; Original amount:
1,209,371,000; Total loans originated: 39,405,607,204.
Year of origination: 1990; Foreclosure rate: 2.89; Original amount:
1,308,801,408; Total loans originated: 45,326,035,945.
Year of origination: 1991; Foreclosure rate: 2.84; Original amount:
1,149,372,455; Total loans originated: 40,464,875,909.
Year of origination: 1992; Foreclosure rate: 1.93; Original amount:
675,069,579; Total loans originated: 35,006,571,763.
Year of origination: 1993; Foreclosure rate: 1.61; Original amount:
901,944,638; Total loans originated: 55,892,535,448.
Year of origination: 1994; Foreclosure rate: 1.98; Original amount:
1,110,636,930; Total loans originated: 56,140,577,134.
Year of origination: 1995; Foreclosure rate: 2.91; Original amount:
820,737,707; Total loans originated: 28,195,589,414.
Year of origination: 1996; Foreclosure rate: 3.10; Original amount:
1,332,871,376; Total loans originated: 43,011,763,810.
Year of origination: 1997; Foreclosure rate: 3.02; Original amount:
1,201,681,220; Total loans originated: 39,805,525,095.
Year of origination: 1998; Foreclosure rate: 2.18; Original amount:
1,609,113,831; Total loans originated: 73,826,808,921.
Note: 1980-1989 loan amounts were estimated from a 10 percent sample.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 8: National 4-Year Cumulative Foreclosure Rates for FHA Fixed-and
Adjustable Rate Mortgage Loans Originated during Fiscal Years 1990-1998
(Figure 3):
Year of origination: 1990; FRM foreclosure rates: 2.89%; ARM
foreclosure rates: 1.79%.
Year of origination: 1991; FRM foreclosure rates: 2.84; ARM foreclosure
rates: 1.71.
Year of origination: 1992; FRM foreclosure rates: 1.93; ARM foreclosure
rates: 1.72.
Year of origination: 1993; FRM foreclosure rates: 1.61; ARM foreclosure
rates: 2.18.
Year of origination: 1994; FRM foreclosure rates: 1.98; ARM foreclosure
rates: 3.30.
Year of origination: 1995; FRM foreclosure rates: 2.91; ARM foreclosure
rates: 4.29.
Year of origination: 1996; FRM foreclosure rates: 3.10; ARM foreclosure
rates: 4.20.
Year of origination: 1997; FRM foreclosure rates: 3.02; ARM foreclosure
rates: 3.65.
Year of origination: 1998; FRM foreclosure rates: 2.18; ARM foreclosure
rates: 3.59.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 9: Adjustable Rate Mortgages as Share of All FHA Loans Originated
during Fiscal Years 1990-1998 (Figure 4):
Year of origination: 1990; Percentage: 1%; Amount: $376,394,573.
Year of origination: 1991; Percentage: 4; Amount: 1,968,220,459.
Year of origination: 1992; Percentage: 16; Amount: 7,976,055,601.
Year of origination: 1993; Percentage: 13; Amount: 10,509,318,684.
Year of origination: 1994; Percentage: 18; Amount: 15,670,591,954.
Year of origination: 1995; Percentage: 27; Amount: 12,411,803,262.
Year of origination: 1996; Percentage: 24; Amount: 16,806,552,046.
Year of origination: 1997; Percentage: 34; Amount: 24,479,889,799.
Year of origination: 1998; Percentage: 13; Amount: 12,498,114,087.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 10: Share of FHA Long-Term, Fixed-Rate Loans Originated in
Selected States during Fiscal Years 1990-1998 (Figure 5):
Selected states: California; Share of all loans: 13%; Total loans
originated: $55,168,696,004.
Selected states: Texas; Share of all loans: 8; Total loans originated:
33,963,938,873.
Selected states: Florida; Share of all loans: 6; Total loans
originated: 26,002,603,640.
Selected states: New York; Share of all loans: 4; Total loans
originated: 16,903,498,072.
Selected states: Illinois; Share of all loans: 3; Total loans
originated: 14,340,445,180.
Selected states: Remaining States; Share of all loans: 65; Total loans
originated: 271,291,101,670.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 11: National 4-Year Cumulative Foreclosure Rates for FHA Long-
Term, Fixed-Rate Loans Originated in Selected States during Fiscal
Years 1990-1998 (Figure 6):
Year of Origination: 1990; California: 4.36%; Texas: 4.41%; Florida:
4.04%; Illinois: 2.09%; New York: 2.78%; Remaining States: 2.40%.
Year of Origination: 1991; California: 6.99; Texas: 3.69; Florida:
3.40; Illinois: 2.50; New York: 2.93; Remaining States: 2.10.
Year of Origination: 1992; California: 6.18; Texas: 2.30; Florida:
2.40; Illinois: 1.73; New York: 2.56; Remaining States: 1.31.
Year of Origination: 1993; California: 6.00; Texas: 1.45; Florida:
1.78; Illinois: 1.27; New York: 1.63; Remaining States: 0.92.
Year of Origination: 1994; California: 6.87; Texas: 1.78; Florida:
2.23; Illinois: 1.40; New York: 1.90; Remaining States: 1.19.
Year of Origination: 1995; California: 7.14; Texas: 2.66; Florida:
4.44; Illinois: 2.44; New York: 2.25; Remaining States: 2.07.
Year of Origination: 1996; California: 7.20; Texas: 2.71; Florida:
4.80; Illinois: 2.61; New York: 2.45; Remaining States: 2.05.
Year of Origination: 1997; California: 5.86; Texas: 2.93; Florida:
4.81; Illinois: 2.72; New York: 2.22; Remaining States: 2.16.
Year of Origination: 1998; California: 3.67; Texas: 2.00; Florida:
4.16; Illinois: 1.65; New York: 1.35; Remaining States: 1.63.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 12: Share of FHA Adjustable Rate Mortgages Originated in Selected
States during Fiscal Years 1990-1998 (Figure 7):
Selected states: California; Share of all loans: 20.5%; Total loans
originated: $21,078,783,499.
Selected states: Illinois; Share of all loans: 9.6; Total loans
originated: 9,806,420,567.
Selected states: Maryland; Share of all loans: 6.4; Total loans
originated: 6,576,127,681.
Selected states: Colorado; Share of all loans: 5.5; Total loans
originated: 5,675,242,154.
Selected states: Remaining States; Share of all loans: 58.0; Total
loans originated: 59,560,366,564.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 13: National 4-Year Cumulative Foreclosure Rates for FHA
Adjustable Rate Mortgages Originated in Selected States during Fiscal
Years 1990-1998 (Figure 8):
Year of origination: 1990; California: 3.16%; Maryland: 0.00%;
Colorado: 1.56%; Illinois: 1.71%; Remaining States: 1.77%.
Year of origination: 1991; California: 4.51; Maryland: 1.72; Colorado:
1.04; Illinois: 1.28; Remaining States: 1.36.
Year of origination: 1992; California: 4.97; Maryland: 1.49; Colorado:
0.46; Illinois: 1.40; Remaining States: 1.28.
Year of origination: 1993; California: 5.85; Maryland: 1.36; Colorado:
0.49; Illinois: 1.46; Remaining States: 1.26.
Year of origination: 1994; California: 7.88; Maryland: 2.21; Colorado:
0.74; Illinois: 2.01; Remaining States: 1.59.
Year of origination: 1995; California: 10.20; Maryland: 3.22; Colorado:
1.29; Illinois: 3.02; Remaining States: 2.52.
Year of origination: 1996; California: 10.01; Maryland: 4.02; Colorado:
1.30; Illinois: 2.80; Remaining States: 2.57.
Year of origination: 1997; California: 8.78; Maryland: 3.63; Colorado:
1.01; Illinois: 2.45; Remaining States: 2.38.
Year of origination: 1998; California: 8.58; Maryland: 2.34; Colorado:
0.55; Illinois: 2.48; Remaining States: 2.53.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 14: Distribution of LTV Categories for FHA Loans Originated
during Fiscal Years 1990-1998 (Figure 9):
Year of origination: 1990; 0<90: 24%; 90<=LTV<95: 34%; LTV>= 95:
41%.
Year of origination: 1991; 0<90: 20; 90<=LTV<95: 28; LTV>= 95: 48.
Year of origination: 1992; 0<90: 17; 90<=LTV<95: 25; LTV>= 95: 46.
Year of origination: 1993; 0<90: 13; 90<=LTV<95: 18; LTV>= 95: 37.
Year of origination: 1994; 0<90: 12; 90<=LTV<95: 16; LTV>= 95: 38.
Year of origination: 1995; 0<90: 13; 90<=LTV<95: 23; LTV>= 95: 61.
Year of origination: 1996; 0<90: 13; 90<=LTV<95: 23; LTV>= 95: 54.
Year of origination: 1997; 0<90: 15; 90<=LTV<95: 25; LTV>= 95: 56.
Year of origination: 1998; 0<90: 15; 90<=LTV<95: 23; LTV>= 95: 47.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 15: National 4-Year Cumulative Foreclosure Rates for Selected
LTV Classes of Long-Term, Fixed-Rate Mortgages Originated during Fiscal
Years 1990-1998
(Figure 10):
Year of origination: 1990; LTV>= 97: 4.97%; 95<=LTV< 97: 3.50%;
90<=LTV<95: 2.69%; 0<90: 1.79%.
Year of origination: 1991; LTV>= 97: 3.26; 95<=LTV< 97: 3.20;
90<=LTV<95: 2.78; 0<90: 2.08.
Year of origination: 1992; LTV>= 97: 2.87; 95<=LTV< 97: 1.95;
90<=LTV<95: 1.83; 0<90: 1.60.
Year of origination: 1993; LTV>= 97: 1.58; 95<=LTV< 97: 1.63;
90<=LTV<95: 1.51; 0<90: 1.22.
Year of origination: 1994; LTV>= 97: 2.03; 95<=LTV< 97: 1.99;
90<=LTV<95: 1.75; 0<90: 1.38.
Year of origination: 1995; LTV>= 97: 3.23; 95<=LTV< 97: 3.01;
90<=LTV<95: 2.69; 0<90: 2.27.
Year of origination: 1996; LTV>= 97: 3.58; 95<=LTV< 97: 3.20;
90<=LTV<95: 2.91; 0<90: 1.99.
Year of origination: 1997; LTV>= 97: 3.59; 95<=LTV< 97: 3.24;
90<=LTV<95: 2.93; 0<90: 1.81.
Year of origination: 1998; LTV>= 97: 3.16; 95<=LTV< 97: 2.42;
90<=LTV<95: 2.11; 0<90: 1.34.
Source: GAO’s analysis of data obtained from HUD.
[End of table]
Table 16: Actual and Forecasted Cumulative Foreclosure Rates for FHA
Loans Insured during Fiscal Years 1996-2001, as of September 30, 2001
(Figure 11):
Year of origination: 1996; Actual foreclosure rate: 5.81%; Forecast
foreclosure rate: 3.44%.
Year of origination: 1997; Actual foreclosure rate: 4.20; Forecast
foreclosure rate: 2.62.
Year of origination: 1998; Actual foreclosure rate: 1.86; Forecast
foreclosure rate: 1.24.
Year of origination: 1999; Actual foreclosure rate: 1.05; Forecast
foreclosure rate: 0.56.
Year of origination: 2000; Actual foreclosure rate: 0.49; Forecast
foreclosure rate: 0.24.
Year of origination: 2001; Actual foreclosure rate: 0.01; Forecast
foreclosure rate: 0.01.
Note: The number of years of data varies by book of business. For
example, there are up to 6 years of data on the performance of loans
originated in 1996, while there is only 1 year’s data for loans
originated in 2001. Thus, the foreclosure rates for loans originated in
1996 represent 6-year cumulative foreclosure rates, while the
foreclosure rates for loans originated in 2001 represent 1-year
cumulative foreclosure rates.
Source: GAO analysis of data obtained from FHA.
[End of table]
[End of section]
Appendix IV: Comments from the Department of Housing and Urban
Development:
U. S. Department of Housing and Urban Development Washington, D.C.
20410-8000:
June 26, 2002:
OFFICE OF THE ASSISTANT SECRETARY:
FOR HOUSING-FEDERAL HOUSING COMMISSIONER:
Mr. Stanley J. Czerwinski:
Director, Physical Infrastructure Team United States General Accounting
Office Washington, DC 20548:
Dear Mr. Czerwinski:
Thank you for the opportunity to provide comments on the General
Accounting Office (GAO) draft report: MORTGAGE FINANCING: Information
on Changes in the Performance of FHA-Insured Loans (GAO -02-773). The
Department is especially interested in understanding the factors that
affect the performance of one-to four-family mortgages insured by the
Federal Housing Administration (FHA) because FHA makes homeownership
possible for hundreds of thousands of American families each year, and
the financial soundness of FHA’s Mutual Mortgage Insurance Fund (the
Fund) is vital to its ability to continue in this important role. Your
report provides valuable information to help ensure that FHA remains
strong.
The Department generally agrees with the findings of the report. We
would, however, offer some comments related to the finding that FHA
foreclosure rates appear to be somewhat higher for books insured after
Fiscal Year (FY) 1995 compared to the books of business insured between
FY 1990 and FY 1994.
FHA made several changes to its underwriting guidelines in FY 1995 in
order to promote increased homeownership opportunities among low-income
and minority homebuyers. By doing so, FHA modestly increased the risk
characteristics of its post 1995 books of business, but it succeeded in
raising FHA’s proportion of first-time homebuyers from 65.4 percent in
1994 to 77.6 percent at the end of the second quarter of FY 2002, and
in raising its share of minority homebuyers from 25.3 percent to 42.0
percent in the same time period. FY 1995 was the first year in which
FHA exceeded the 2.0 percent capital ratio mandated by the National
Affordable Housing Act of 1990. Since then, FHA’s capital ratio has
continued to increase. It reached 3.75 percent at the end of FY 2001
and is projected to exceed 4.0 percent beginning in FY 2002. These
figures suggest that FHA has been successful in its mission of opening
homeownership opportunities for underserved groups.
FHA is continually alert to its obligation to promote homeownership
while properly managing and pricing the risk that it assumes. By
promoting the use of loss mitigation, for example, FHA is
simultaneously helping homeowners to stay in their homes while reducing
losses to the FHA Fund. By reducing its mortgage insurance premium in
FY
2001, FHA lowered the cost of homeownership to potential homebuyers
while
maintaining a healthy capital ratio.
The Department is committed to constant improvement in FHA’s
performance, and appreciates your ongoing suggestions for ways to keep
FHA strong while providing greater homeownership opportunities to
American families.
Sincerely yours,
John C. Weicher:
Assistant Secretary for Housing-:
Federal Housing Commissioner:
Signed by John C. Weicher:
[End of figure]
FOOTNOTES
[1] The economic value of the Fund is the sum of existing capital
resources plus the net present value of future cash flows.
[2] These included scenarios that are based on recent regional
experiences and on the 1981 through1982 national recession. See U.S.
General Accounting Office, Mortgage Financing: FHA’s Fund Has Grown,
but Options for Drawing on the Fund Have Uncertain Outcomes, GAO-01-460
(Washington, D.C.: Feb. 28, 2001).
[3] A book of business represents all loans insured during a given
year.
[4] The West South Central region comprises Arkansas, Louisiana,
Oklahoma, and Texas.
[5] The Act defined the capital ratio as the ratio of the Fund’s
capital, or economic net worth, to its unamortized insurance-in-force.
However, the Act defined unamortized insurance-in-force as the
remaining obligations on outstanding mortgages--a definition generally
understood to apply to amortized insurance-in-force. FHA has calculated
the 2 percent capital ratio using unamortized insurance-in-force as it
is generally understood--which is the initial amount of mortgages.
[6] These scenarios included (1) a scenario in which the entire nation
experiences a downturn similar to the one New England experienced
during the late 1980s and early 1990s, (2) a scenario in which FHA
experiences foreclosure rates similar to those it experienced in the
late 1980s, and (3) a scenario in which 35.6 percent or more of FHA
loans experience foreclosure rates similar to those experienced by FHA
in the West South Central portion of the United States in the late
1980s.
[7] Conventional mortgage lenders, by offering second mortgages of up
to 23 percent of the value of the house, sometimes allow borrowers to
borrow more than the value of the house without obtaining mortgage
insurance.
[8] Later in this report we discuss in some detail the potential impact
that both changes in FHA’s program and competition from conventional
lenders may have on foreclosure rates for FHA-insured loans, and on the
riskiness of FHA’s portfolio.
[9] We selected a 4-year cumulative foreclosure rate because it best
balanced the competing goals of having the greatest number of recent
observations and the greatest number of years of experience. We also
examined a 3-year cumulative foreclosure rate across books of business
originated between 1990 and 1999 and found a similar pattern in
foreclosure rates. Therefore, we concluded that a 4-year cumulative
claim rate was a reasonable indicator of loan performance.
[10] These figures represent the original loan amount of the foreclosed
loans for which FHA paid a claim during the first 4 years of the life
of these mortgages as a percentage of the total value of mortgages
originated in that year.
[11] Buy downs allow sellers to pay a nominal amount to lower (or buy
down) the homebuyer’s interest rate for the first year. With lower
first-year payments, buyers can more easily qualify for a mortgage for
which they otherwise would have been ineligible. According to FHA, some
homebuyers, when faced with a large increase in mortgage payments after
the buy down period, had a greater likelihood of defaulting on their
mortgages.
[12] For this analysis and the one that follows, we do not include
loans for which the recorded LTV is zero.
[13] In previous modeling work we also found that even when the effects
of other factors are taken into account, higher LTVs are associated
with a greater likelihood of foreclosure.
[14] For the purpose of this analysis, we grouped FHA loans into four
categories by LTV: LTV greater than or equal to 97 percent; LTV at
least 95 percent but less than 97 percent; LTV at least 90 percent but
less than 95 percent; and LTV greater than zero but less than 90
percent.
[15] Loss mitigation refers to steps taken by the mortgage lender to
avoid foreclosure. In November 1996 FHA implemented a new loss
mitigation program that included a range of options that helped
homeowners to either retain their homes or dispose of them in ways that
reduced the costs of foreclosure for both the homeowners and FHA.
[16] In 1994, FHA established an Underwriting Working Group to review
FHA’s underwriting guidelines and recommend changes and modifications
that would eliminate unnecessary barriers to homeownership; provide the
flexibility to underwrite creditworthy nontraditional and underserved
borrowers; and, clarify certain underwriting requirements so that they
are not applied in a discriminatory manner. The group’s recommendations
formed the basis for underwriting changes made in fiscal year 1995.
[17] FHA uses two qualifying ratios to determine whether a borrower
will be able to meet the expenses involved in homeownership. The
payment-to-income ratio (not to exceed 29 percent) examines a
borrower’s expected monthly housing expenses as a percentage of monthly
income; the debt-to-income ratio (not to exceed 41 percent) looks at a
borrower’s expected monthly housing expenses plus long-term debt as a
percentage of monthly income. Both ratios can be exceeded if
significant compensating factors exist. Compensating factors are
conditions related to the borrower that may be used in justifying
approval of a mortgage with qualifying ratios exceeding FHA benchmark
guidelines.
[18] A nontraditional credit report is designed to assess the credit
history for borrowers without the credit references normally appearing
on a traditional credit report. In developing a nontraditional credit
report, credit agencies are to consider only the type of credit that
requires periodic payments, such as payments for rental housing,
utilities, telephone and cable service, insurance payments, school
tuition, and medical bills.
[19] Since 1992, a borrower’s lack of credit history cannot be used as
a basis for rejecting a loan application. At that time, FHA began
requiring lenders to use an alternate method of verifying credit (or
establishing an alternative credit history) for borrowers with no
credit history by documenting rent and utility payments.
[20] Until April 26, 1996, lenders servicing FHA loans were required to
either recommend that HUD accept assignment of FHA-insured mortgage
notes that had become 90 days delinquent or initiate foreclosure
proceedings. If loans were accepted for assignment, HUD paid an
insurance claim to the servicer and became the holder and servicer of
the loan.
[21] Incentive claims for taking loss mitigation actions are paid to
the mortgage lender or servicer for the three home retention tools--
special forbearance, loan modification, and partial claim. However, for
deed-in-lieu and pre-foreclosure sales, the loan is terminated and an
insurance claim, similar to that paid when an FHA loan is foreclosed,
is paid to the mortgage lender or servicer.
[22] See An Assessment of FHA’s Single-Family Mortgage Insurance Loss
Mitigation Program: Final Report, Abt Associates Inc., November 30,
2000.
[23] See U.S. General Accounting Office, Mortgage Financing: FHA’s Fund
Has Grown, but Options for Drawing on the Fund Have Uncertain Outcomes,
GAO-01-460 (Washington, D.C.: Feb. 28, 2001).
[24] For our previous work, we used data on loans insured through
fiscal year 1999.
[25] These figures represent the total value of unpaid balances on
loans for which FHA paid claims as a percentage of the total value of
mortgages originated.
[26] We view this evidence as merely suggestive because we have only a
few years of experience with these loans, particularly those insured
after 1997. After more years have passed, the evidence on the
performance of loans insured after 1995 will be more conclusive.
[27] The peak foreclosure years for a book of business are generally
the third through seventh years of the loans.
[28] We also examined 3 years of foreclosure experience and found that
the relative foreclosure rates for each book of business exhibited
similar patterns.
[29] See U. S. General Accounting Office, Mortgage Financing: FHA’s
Fund Has Grown, but Options for Drawing on the Fund Have Uncertain
Outcomes, GAO-01-460 (Feb. 2001).
[30] These probabilities are conditional, because they are subject to
the condition that the loan has remained active until a given year.
[31] If P is the probability that an event will occur, the “odds ratio”
is defined as P/(1-P). The logistic transformation is the natural
logarithm of the odds ratio, or ln[P/(1-P)], of which the logistic
regression provides an estimate. See G.S. Maddala, Limited Dependent
Variables and Qualitative Variables in Econometrics (Cambridge:
Cambridge University Press, 1983). Also see John H. Aldrich and Forrest
D. Nelson, Linear Probability, Logit, and Probit Models (SAGE
Publications: Beverly Hills, London, and New York, 1984), pp. 41-44.
[32] DRI-WEFA is a leading economic forecasting firm.
[33] When we discuss the likely effects of one of our explanatory
variables, we are describing the marginal effects of that variable,
while holding the effects of other variables constant.
[34] We revised the estimated rate of appreciation in nominal median
house prices downward by 2 percentage points per year to account for
depreciation and the gradual improvement in the quality of the existing
housing stock over time.
[35] Essentially, LAGEQHIGH takes the value of equity minus.2 if equity
is greater than 20 percent, or 0 if equity is less than or equal to 20
percent. LAGEQLOW takes the value of equity if equity is 20 percent or
less, and .2 if equity is greater than 20 percent.
[36] Similarly, for foreclosures within the ARM equations, we defined
equity as book equity (the estimated property value less the amortized
balance of the loan) and not market equity. The effects of interest
rate changes in the ARM equations were estimated using a separate
variable.
[37] We also did not know whether individual borrowers had subsequently
acquired second mortgages or other obligations that would affect
prepayment or foreclosure probabilities.
[38] A basis point equals 1/100 of a percentage point.
[39] Transaction costs associated with refinancing have fallen in
recent years, making it more profitable than before to refinance at a
smaller decrease in interest rates.
[40] The conditional foreclosure probability is calculated as F(Z) =
EXP(Z)/[1+EXP(Z)], where Z = Si (Xi*Bi), where Xi refers to the mean
value of the ith explanatory variable and Bi represents the estimated
coefficient for the ith explanatory variable.
[41] Conditional foreclosure probabilities for the other mortgage types
were estimated as follows: long-term, fixed-rate, refinancing mortgages
(0.4); short-term, fixed-rate mortgages (0.1); ARMs (0.4); investor,
nonrefinancing mortgages (0.6); and investor, refinancing mortgages
(0.3).
[42] This average is for the dollar worth of a loan, not the number of
loans.
[43] Conditional prepayment probabilities for the other mortgage types
were estimated as follows: long-term, fixed-rate, refinancing mortgages
(9.4); short-term, fixed-rate mortgages (4.0); ARMs (7.0); investor,
nonrefinancing mortgages (5.3); and investor, refinancing mortgages
(8.0).
[44] Although we present figures comparing actual and predicted rates
only for long-term, fixed-rate, nonrefinanced mortgages, the close
resemblance holds true for all loan types.
GAO’s Mission:
Obtaining Copies of GAO Reports and Testimony:
Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as “Today’s Reports,” on its
Web site daily. The list contains links to the full-text document
files. To have GAO e-mail this list to you every afternoon, go to
HYPERLINK http://www.gao.gov \* MERGEFORMAT www.gao.gov and select
“Subscribe to daily E-mail alert for newly released products” under the
GAO Reports heading.
Order by Mail or Phone:
U.S. General Accounting 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, Waste, and Abuse in Federal Programs:
Web site: www.gao.gov/fraudnet/fraudnet.htm
E-mail: fraudnet@gao.gov
Automated answering system: (800) 424-5454 or (202) 512-7470
Public Affairs:
Jeff Nelligan, managing director, NelliganJ@gao.gov (202) 512-4800:
U.S. General Accounting Office, 441 G Street NW, Room 7149:
Washington, D.C. 20548.