oversight

Homeownership: Information on Foreclosed FHA-Insured Loans and HUD-Owned Properties in Six Cities

Published by the Government Accountability Office on 1997-10-08.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                United States General Accounting Office

GAO             Report to the Chairman, Subcommittee
                on Housing and Community Opportunity,
                Committee on Banking and Financial
                Services, House of Representatives

October 1997
                HOMEOWNERSHIP
                Information on
                Foreclosed
                FHA-Insured Loans
                and HUD-Owned
                Properties in Six Cities




GAO/RCED-98-2
      United States
GAO   General Accounting Office
      Washington, D.C. 20548

      Resources, Community, and
      Economic Development Division

      B-276286

      October 8, 1997

      The Honorable Rick A. Lazio
      Chairman, Subcommittee on Housing
        and Community Opportunity
      Committee on Banking
        and Financial Services
      House of Representatives

      Dear Mr. Chairman:

      Through its Federal Housing Administration (FHA), the Department of
      Housing and Urban Development (HUD) provides federally backed
      mortgage insurance to hundreds of thousands of homeowners annually.
      However, each year, lenders foreclose on a portion of the FHA-insured
      mortgages that go into default and file insurance claims with HUD for their
      losses. With few exceptions, HUD takes ownership of the foreclosed
      properties, which generally remain vacant until HUD sells them. Critics of
      FHA contend that the unsound underwriting of FHA-insured loans in
      low-income urban communities has contributed to large numbers of
      foreclosures and vacant HUD-owned homes in these areas. They further
      contend that these homes remain vacant for long periods, attracting crime,
      reducing local property values, and contributing to neighborhood blight.

      To provide some insights into the concerns raised by FHA’s critics, we
      examined “early foreclosures”—those occurring within 18 months of the
      loan endorsement date.1 As agreed with your office, we did not attempt to
      evaluate the soundness of mortgage underwriting decisions or the impact
      of vacant homes on neighborhood conditions because of the
      methodological difficulties that a broad examination of these issues would
      present.

      We looked at early foreclosures because, according to FHA, they are an
      indicator of potentially unsound underwriting practices (e.g., lending to
      unqualified borrowers), whereas foreclosures occurring later are more
      likely to result from unforeseen circumstances that impair the ability of




      1
       After making a loan to a borrower, a lender seeks FHA’s approval to insure the loan. The date when
      FHA formally approves mortgage insurance for the loan is termed the “loan endorsement date.”



      Page 1                                                            GAO/RCED-98-2 Homeownership
                   B-276286




                   borrowers to make mortgage payments (e.g., job loss).2 In addition, we
                   examined the length of time HUD-owned single-family properties remained
                   unsold. To provide perspective on the types of neighborhoods where early
                   foreclosures and unsold properties may be of greatest concern, we made
                   comparisons across low-, medium-, and high-income areas. You requested
                   that we include Chicago, Illinois, and Washington, D.C., in our analysis,
                   and we selected four additional cities—Atlanta, Georgia; Baltimore,
                   Maryland; Dallas, Texas; and San Bernadino, California—because they
                   provided geographic diversity and had relatively high levels of FHA loan
                   activity during the past few years.3

                   Specifically, you asked us to (1) compare early foreclosure rates on
                   FHA-insured single-family loans made in low-, medium-, and high-income
                   areas nationwide and in the six cities; (2) compare across income areas
                   the proportion of loans made in the six cities by FHA-approved mortgage
                   lenders with and without early foreclosures; (3) identify factors that
                   influence early foreclosure rates; and (4) compare the length of time
                   HUD-owned single-family properties remained unsold in low-, medium-, and
                   high-income areas in the six cities.


                   Our analysis of the FHA-insured single-family loans made during calendar
Results in Brief   years 1992 through 1994 nationwide and in the six cities showed that early
                   foreclosures occurred infrequently but that early foreclosure rates were
                   higher for low-income areas than for either medium- or high-income areas.4
                   The early foreclosure rate for low-income areas nationwide was
                   0.45 percent (i.e., 4.5 early foreclosures occurring for every 1,000

                   2
                    For this report, we considered an early foreclosure to be both a loan on which the lender foreclosed
                   within 18 months of the loan endorsement date and a loan on which the lender did not actually
                   foreclose but on which HUD paid an insurance claim to the lender within 18 months of the loan’s
                   endorsement. The latter accounted for about 33 percent of the early foreclosures in our data set and
                   were part of HUD’s mortgage assignment program, which was terminated in 1996. This program gave a
                   borrower who defaulted on an FHA-insured loan the opportunity to avoid foreclosure by petitioning
                   HUD to take assignment (i.e., ownership) of the loan and provide forbearance to the borrower. In
                   taking assignment of a loan, HUD paid the mortgage debt and assumed responsibility for servicing the
                   loan.
                   3
                    The nationwide data reflect loans made in all of the metropolitan statistical areas (MSA), which
                   include central cities and surrounding suburbs, while the data for the six cities reflect loans made
                   within the formal boundaries of these cities but not loans made in the surrounding suburbs. We
                   defined an area’s income level as “low” if the per capita income was at or below 80 percent of the per
                   capita income for the MSA/city, “medium” if the per capita income was greater than 80 percent but at or
                   below 120 percent of the MSA’s/city’s level, and “high” if the per capita income was greater than
                   120 percent of MSA’s/city’s level.
                   4
                    We examined loans made during calendar years 1992 through 1994 because HUD’s database did not
                   have complete demographic information for loans made before 1992 and because 1994 was the last full
                   year we could include in an analysis examining the performance of loans over an 18-month period.
                   Approximately 32 percent of the loans for the six cities were taken out to refinance existing
                   mortgages. Comparable data for loans insured by private mortgage insurers were not available.


                   Page 2                                                             GAO/RCED-98-2 Homeownership
B-276286




mortgages insured) compared with 0.30 percent and 0.21 percent for
medium- and high-income areas, respectively. Although this pattern
prevailed in the six cities, there were also differences from one city to
another. For example, among the six cities, the early foreclosure rates for
low-income areas ranged from 0.47 percent in Washington, D.C., to
1.45 percent in Dallas.5

For four of the cities—Atlanta, Baltimore, Dallas, and Washington, D.C.—
lenders with early foreclosures6 made a larger proportion of their loans for
properties in low- and medium-income areas and a smaller proportion of
their loans for properties in high-income areas than did lenders that did
not experience early foreclosure. In San Bernadino, however, lenders with
early foreclosures made a smaller proportion of their loans for properties
in low-income areas and a larger proportion of their loans for properties in
high-income areas than lenders without early foreclosures. Also, in
Chicago, lenders with early foreclosures made a smaller share of their
loans in medium-income areas than lenders without early foreclosures.

Various factors influence the probability of early foreclosure. Our analysis
of the FHA-insured loans made in calendar years 1992 through 1994 in the
six cities indicated that loans made for homes in poorer census tracts,
smaller loans, and loans with higher loan-to-value ratios7 or higher interest
rates were associated with higher probabilities of early foreclosure.

As of December 31, 1996, HUD held a total of 1,374 properties in its
inventory in the six cities we reviewed. Our analysis did not identify a
pattern in the median time that these properties remained in HUD’s
inventory in different income areas.8 For example, in Atlanta the median
time in inventory was higher in low-income areas than in high-income
areas, while in Chicago the median time in inventory was about the same
in both of these income areas. However, in five of the six cities and for the
six cities combined, the proportion of properties that had been in
inventory for more than 6 months was greater in low-income areas than in
either medium- or high-income areas.

5
 The statements made in this report reflect what we observed in HUD’s data on loans approved during
calendar years 1992 through 1994 nationwide and in the six cities. The foreclosure patterns we
observed may be different from the patterns we might have observed for loans from a different time
period or under different economic conditions.
6
 We defined lenders with early foreclosures as lenders with one or more early foreclosures during the
time periods we reviewed.
7
 This indicator expresses the amount of the loan as a percentage of the property’s value.
8
 The median is a value in an ordered set of values below and above which the number of values is
equal.



Page 3                                                             GAO/RCED-98-2 Homeownership
             B-276286




             Lenders usually require mortgage insurance when a home buyer has a
Background   down payment of less than 20 percent of the value of the home because
             foreclosures are more likely on these loans than on those with higher
             down payments. As the principal provider of federally backed mortgage
             insurance, FHA insured 32 percent of the insured mortgages originated in
             1995. However, FHA fulfills an even larger role in providing insurance for
             some groups of borrowers, particularly low-income home buyers,
             minorities, and central city residents.

             FHA provides most of its single-family mortgage insurance through the
             Section 203(b) program, which covers loans for purchasing a new or
             existing one- to four-family home. The 203(b) program, among other
             programs, is supported by the Mutual Mortgage Insurance Fund (MMI
             Fund), which is funded by revenue from insurance premiums and
             foreclosed property sales. By law, the fund must meet or endeavor to meet
             statutory capital ratio requirements: that is, it must contain sufficient
             reserves to cover the estimated future payments of claims on foreclosed
             mortgages and other costs. Other FHA insurance programs for single-family
             home loans include the Section 203(k) program, for purchasing or
             refinancing and rehabilitating a home at least 1 year old, and the Section
             234(c) program, for purchasing a unit in a condominium project.

             A mortgage loan is commonly considered “in default” when the borrower
             misses three consecutive monthly payments and a fourth payment is due.
             At that point, foreclosure proceedings against the borrower become a
             serious possibility. In the case of FHA-insured loans, once the foreclosure
             process is completed, the lender files an insurance claim with HUD for its
             losses (unpaid mortgage balance and interest, along with the costs of
             foreclosure and other expenses). After the claim is paid, the lender
             transfers the title to the home to HUD, which is responsible for managing
             and selling the property. HUD-owned properties generally remain vacant
             until they are resold.

             At the end of fiscal year 1996, HUD had about 24,700 single-family
             properties in its inventory. The purpose of HUD’s property disposition
             program is to reduce the inventory of acquired property in a manner that
             expands homeownership opportunities, strengthens neighborhoods and
             communities, and ensures a maximum return to the mortgage insurance
             fund. Although FHA has always received enough in premiums from
             borrowers and other revenues to cover the costs of foreclosed MMI Fund
             loans, losses totaled about $12.8 billion in 1994 dollars, or about $24,400




             Page 4                                            GAO/RCED-98-2 Homeownership
                        B-276286




                        for each foreclosed and subsequently sold single-family home over the
                        19-year period ending in 1993.

                        To mitigate losses to FHA and hold lenders accountable for the quality of
                        the loans they make, FHA performs several activities related to the
                        approval, monitoring, and recertification of mortgage lenders participating
                        in FHA’s programs. For example, FHA monitors, by mortgage lender, the
                        percentage of loans in default or on which FHA has paid the lender a claim.
                        FHA also conducts on-site reviews of the loan origination and servicing
                        practices of selected lenders. In addition, in 1996, FHA issued guidelines
                        intended to promote the use of special forbearance plans, mortgage
                        modifications, and other tools to help FHA borrowers in default remain in
                        their homes whenever possible and to mitigate losses to FHA resulting from
                        loan foreclosures.


                        Nationwide, early foreclosures did not occur for 99.68 percent of the
Early Foreclosure       FHA-insured single-family loans made during calendar years 1992 through
Rates Were Highest in   1994.9 However, early foreclosure rates were higher for low-income areas
Low-Income Areas        than for either medium- or high-income areas.10 Nationwide, the early
                        foreclosure rate for low-income areas was 0.45 percent (i.e., 4.5 early
                        foreclosures occurring for every 1,000 mortgages insured) compared with
                        0.30 percent and 0.21 percent for medium- and high-income areas,
                        respectively. Federal regulations require FHA to monitor the performance
                        of FHA-insured loans by mortgage lender but not by income area.
                        Consequently, FHA does not have criteria for determining what would
                        constitute excessively high early foreclosure rates for low-, medium-, or
                        high-income areas nationwide or in a specific geographic region.

                        Consistent with the nationwide pattern, early foreclosure rates in the six
                        cities were highest for low-income areas, but these rates and the
                        proportion of early foreclosures occurring in each income area varied by
                        city. Within 18 months, foreclosures occurred on 254 of the 50,323 loans
                        made in the six cities, for an early foreclosure rate of 0.50 percent. For the
                        six cities combined, the early foreclosure rates for low-, medium-, and
                        high-income areas were 0.80 percent, 0.45 percent, and 0.30 percent,
                        respectively.

                        9
                         Early foreclosures also represent a small share of the foreclosures that will eventually occur. For
                        example, Price Waterhouse has forecasted that foreclosures will eventually occur on 6.97 percent of
                        the 30-year fixed-rate mortgages made in fiscal year 1994 that are supported by FHA’s MMI Fund.
                        10
                          We calculated the number of early foreclosures by identifying loans on which the lender had
                        foreclosed and/or on which FHA had paid a claim within 18 months of the loan endorsement date. We
                        divided this number by the total number of loans to arrive at an early foreclosure rate.



                        Page 5                                                            GAO/RCED-98-2 Homeownership
                                         B-276286




                                         Among the individual cities, the early foreclosure rates for low-income
                                         areas ranged from 0.47 percent in Washington, D.C., to 1.45 percent in
                                         Dallas. For medium-income areas, they ranged from 0.15 percent in
                                         Chicago to 1.02 percent in San Bernadino, and for high-income areas, they
                                         ranged from zero percent in Washington, D.C., to 0.86 percent in San
                                         Bernadino. San Bernadino had the highest early foreclosure rate
                                         (1.05 percent) for all income areas combined. According to HUD and San
                                         Bernadino city officials, job losses associated with military base closings
                                         and corporate downsizing have been a primary cause of foreclosures on
                                         FHA-insured mortgages in San Bernadino. Chicago had the lowest early
                                         foreclosure rate (0.26 percent) for all income areas combined. Table 1
                                         shows early foreclosure rates in the six cities by income areas.

Table 1: Early Foreclosure Rates for
FHA-Insured Loans Made in Calendar                                                          Income level of areasa
Years 1992-94 in Six Cities, by Income   City                                   Low            Medium                   High                 All
Areas
                                         Atlanta                                1.40                0.41                0.23             0.63
                                         Baltimore                              0.78                0.73                0.59             0.66
                                         Chicago                                0.48                0.15                0.12             0.26
                                         Dallas                                 1.45                0.79                0.17             0.63
                                         San Bernadino                          1.14                1.02                0.86             1.05
                                         Washington, D.C.                       0.47                0.21                    0            0.28
                                         Six cities combined                    0.80                0.45                0.30             0.50
                                         a
                                          We defined an area’s income level as “low” if the per capita income was at or below 80 percent
                                         of the per capita income for the city, “medium” if the per capita income was greater than
                                         80 percent but at or below 120 percent of the city level, and “high” if the per capita income was
                                         greater than 120 percent of city level.

                                         Source: GAO’s analysis of data from HUD and the Bureau of the Census.



                                         For the six cities combined, the percentage of early foreclosures occurring
                                         for low-income areas was disproportionately high relative to the
                                         percentage of loans made for homes in these areas. As shown in appendix
                                         I, for the six cities combined, low-income areas accounted for 44.5 percent
                                         (113 of 254) of the early foreclosures, compared with 27.9 percent (14,050
                                         of 50,323) of the loans made.11

                                         This pattern also held true for the six cities individually. Among the six
                                         cities, the proportion of early foreclosures occurring for low-income areas
                                         ranged from 8.9 percent (5 of 56 early foreclosures) in Baltimore to


                                         11
                                          Seventy-one early foreclosures—42 fewer than actually occurred—would have represented a
                                         proportionate number of early foreclosures.



                                         Page 6                                                            GAO/RCED-98-2 Homeownership
                        B-276286




                        66.7 percent (6 of 9 early foreclosures) in Washington, D.C., while the
                        corresponding proportions of loans made for properties in these areas
                        were 7.5 percent and 39.4 percent, respectively.

                        In two of the six cities—Baltimore and Dallas—the percentage of early
                        foreclosures for medium-income areas was disproportionately high
                        relative to the percentage of loans made for homes in these areas. For
                        example, in Baltimore, medium-income areas accounted for 46.4 percent
                        of the early foreclosures, compared with 42.2 percent of the loans made. In
                        high-income areas in each of the six cities, the percentage of early
                        foreclosures was smaller than the percentage of loans made for properties
                        in these areas.

                        Appendix I provides additional details on early foreclosure rates in the six
                        cities.


                        For the six cities combined, lenders with early foreclosures made a larger
Lenders With Early      percentage of their loans for properties in low- and medium-income areas
Foreclosures Made a     and a smaller percentage of their loans for properties in high-income areas
Larger Share of Their   than lenders without early foreclosures. Lenders with early foreclosures
                        made 30.3, 43.1, and 26.6 percent of their loans for properties in low-,
Loans in Low- and       medium-, and high-income areas, respectively, while the corresponding
Medium-Income Areas     figures for lenders without early foreclosures were 24.7, 40.7, and
                        34.5 percent. This pattern also prevailed in four of the individual
Than Lenders Without    cities—Atlanta, Baltimore, Dallas, and Washington, D.C. In San Bernadino,
Early Foreclosures      however, lenders with early foreclosures made a smaller proportion of
                        their loans for properties in low-income areas and a larger proportion of
                        their loans for properties in high-income areas than lenders without early
                        foreclosures. Also, in Chicago, lenders with early foreclosures made a
                        smaller share of their loans in medium-income areas than lenders without
                        early foreclosures. The relative proportions of loans made for properties in
                        the different income areas of each city by lenders with and without early
                        foreclosures are shown in table 2.




                        Page 7                                           GAO/RCED-98-2 Homeownership
                                       B-276286




Table 2: Proportion of FHA-Insured
Loans Made in Calendar Years 1992-94                                                                     Income level of areasa
for Properties in Low-, Medium-, and   City                              Type of lender                  Low        Medium              High
High-Income Areas in Six Cities, by
                                       Atlanta                           With early
Lenders With and Without Early
                                                                         foreclosures                     34.9           45.0           20.1
Foreclosures
                                                                         Without early
                                                                         foreclosures                     24.1           41.5           34.3
                                       Baltimore                         With early
                                                                         foreclosures                      8.4           43.1           48.5
                                                                         Without early
                                                                         foreclosures                      6.1           41.0           52.9
                                       Chicago                           With early
                                                                         foreclosures                     36.3           46.7           17.0
                                                                         Without early
                                                                         foreclosures                     26.8           47.7           25.6
                                       Dallas                            With early
                                                                         foreclosures                     24.2           35.3           40.5
                                                                         Without early
                                                                         foreclosures                     15.4           28.5           56.0
                                       San Bernadino                     With early
                                                                         foreclosures                     46.8           39.0           14.3
                                                                         Without early
                                                                         foreclosures                     49.6           38.6           11.8
                                       Washington, D.C.                  With early
                                                                         foreclosures                     40.5           47.5           12.1
                                                                         Without early
                                                                         foreclosures                     39.0           43.8           17.2
                                       Six cities combined               With early
                                                                         foreclosures                     30.3           43.1           26.6
                                                                         Without early
                                                                         foreclosures                     24.7           40.7           34.5
                                       a
                                        We defined an area’s income level as “low” if the per capita income was at or below 80 percent
                                       of the per capita income for the city, “medium” if the per capita income was greater than
                                       80 percent but at or below 120 percent of the city’s level, and “high” if the per capita income was
                                       greater than 120 percent of the city’s level.

                                       Note: Percentages may not add to 100 because of rounding.



                                       Source: GAO’s analysis of data from HUD and the Bureau of the Census.


                                       Additional details about differences in lending patterns among lenders
                                       with and without early foreclosures appear in appendix II.




                                       Page 8                                                            GAO/RCED-98-2 Homeownership
                        B-276286




                        For FHA-insured loans made during calendar years 1992 through 1994 in the
Several Factors Were    six cities we reviewed, we found that the following factors were
Associated With Early   associated with early foreclosure rates: (1) the relative income level of the
Foreclosures            census tract where the property was located (expressed as the ratio of the
                        per capita income for the census tract to the per capita income for the
                        city), (2) the loan amount, (3) the loan-to-value ratio, (4) the loan interest
                        rate, and (5) the city where the property was located.12 Other things being
                        equal, loans made for properties in poorer census tracts, smaller loans,
                        loans with higher loan-to-value ratios, and loans with higher interest rates
                        were associated with higher probabilities of early foreclosure. Our
                        analysis also showed that loans made for homes in San Bernadino were
                        associated with higher probabilities of early foreclosure, possibly
                        reflecting the loss of military and defense industry jobs in the San
                        Bernadino area.

                        Our analysis also showed that loans made in poorer census tracts tended
                        to be smaller and to have higher loan-to-value ratios and higher interest
                        rates—all factors that increased the probability of early foreclosure. The
                        relationship between lower incomes and loans with these characteristics
                        may partially explain why early foreclosure rates were higher in
                        low-income areas than in either medium- or high-income areas.
                        Nonetheless, the association between census tract incomes and early
                        foreclosure rates was statistically significant even after controlling for
                        these other factors.

                        We tested additional factors but did not find them to have statistically
                        significant associations with early foreclosure rates after accounting for
                        the factors listed above. These factors were the race (white or minority) of
                        the borrower, the age of the borrower, the year of the loan’s origination,
                        and the FHA loan program used (203(b) or other loan program).

                        Appendix III provides additional information on the results of our
                        statistical analysis.




                        12
                         We identified these associations by performing a logistic regression analysis, a technique used to
                        estimate the individual influence of each factor while controlling for the influence of the others. The
                        associations were significant at the 95-percent confidence level.



                        Page 9                                                              GAO/RCED-98-2 Homeownership
                                      B-276286




                                      As of December 31, 1996, HUD held 1,374 single-family properties in its
Length of Time That                   inventory in the six cities combined. Among the six cities, the number of
HUD-Owned                             properties in HUD’s inventory that remained unsold ranged from 65 in
Properties Remained                   Atlanta to 471 in Chicago.13 Our analysis did not disclose a pattern in the
                                      median time that these properties remained in HUD’s inventory in different
Unsold                                income areas. As shown in table 3, while in Atlanta and Washington the
                                      median time in inventory was higher in low-income areas than in
                                      high-income areas, in Baltimore, Chicago, and San Bernadino, the median
                                      time in inventory was about the same in these areas. In Dallas, the median
                                      time in inventory was higher in high-income areas than in low-income
                                      areas. According to HUD officials, the length of time properties remain in
                                      HUD’s inventory is greatly affected by the economic conditions in each city.


Table 3: Median Months in Inventory
for Single-Family Properties in Six                                                              Income level of areasa
Cities That Remained Unsold as of     City                                               Low        Medium             High                 All
December 31, 1996, by Income Areas
                                      Atlanta                                              3.7            3.8             0.8               3.7
                                      Baltimore                                            4.8            4.5             4.8               4.6
                                      Chicago                                              5.0            4.3             4.9               4.8
                                      Dallas                                               1.8            3.2             3.0               3.0
                                      San Bernadino                                        4.5            2.4             4.5               3.5
                                      Washington, D.C.                                     8.5            7.6             4.1               8.0
                                      Note: We excluded from our analysis properties held off the market as of May 17, 1997 (the date
                                      our data file was created). HUD may hold properties off the market while carrying out certain
                                      administrative processes and programs for assisting the homeless, as well as for other reasons.
                                      However, we were unable to determine whether included properties had been held off the market
                                      for any time in the past. In addition, in some cases, we were either unable to identify the census
                                      tract where a property was located or HUD’s data did not provide the date a property entered
                                      HUD’s inventory. Therefore, we excluded these properties from our analysis. The percentage of
                                      properties in each city that we excluded from our analysis because of missing information on the
                                      census tract or time in inventory was as follows: Atlanta, 3 percent; Baltimore, 14 percent;
                                      Chicago, 4 percent; Dallas, 5 percent; San Bernadino, 20 percent; and Washington, D.C.,
                                      16 percent.
                                      a
                                       We defined an area’s income level as “low” if the per capita income was at or below 80 percent
                                      of the per capita income for the city, “medium” if the per capita income was greater than
                                      80 percent but at or below 120 percent of the city’s level, and “high” if the per capita income was
                                      greater than 120 percent of the city’s level.



                                      Source: GAO’s analysis of data from HUD and the Bureau of the Census.




                                      13
                                        We were able to match census tract information and valid time-in-inventory data with 1,232 of the
                                      1,374 properties in HUD’s inventory as of December 31, 1996. Therefore, we limited our analysis to
                                      these 1,232 properties. Appendix IV provides additional details on the number and percentage of
                                      properties for which this match was feasible.



                                      Page 10                                                           GAO/RCED-98-2 Homeownership
                                         B-276286




                                         For the six cities combined and for each of the individual cities except
                                         Dallas, the proportion of properties that had been in inventory for more
                                         than 6 months was greater in low-income areas than in either medium- or
                                         high-income areas. (See table 4.)

Table 4: Months in Inventory for
Single-Family Properties in Six Cities                                 Months in                       Income level of areasa
That Remained Unsold as of               City                          inventory                 Low      Medium            High               All
December 31, 1996, by Income Areas
                                         Atlanta                       Less than or               19            20              2            41
                                                                       equal to 6               (61.3%)       (64.5%)       (66.7%)       (63.1%)
                                                                       Greater than 6             12            11              1            24
                                                                                                (38.7%)       (35.5%)       (33.3%)       (36.9%)
                                         Baltimore                     Less than or               30            52             25           107
                                                                       equal to 6               (51.7%)       (65.0%)       (58.1%)       (59.1%)
                                                                       Greater than 6             28            28             18            74
                                                                                                (48.3%)       (35.0%)       (41.9%)       (40.9%)
                                         Chicago                       Less than or               161           107            15           283
                                                                       equal to 6               (57.1%)       (64.1%)       (68.2%)       (60.1%)
                                                                       Greater than 6             121           60              7           188
                                                                                                (42.9%)       (35.9%)       (31.8%)       (39.9%)
                                         Dallas                        Less than or               35            56             22           113
                                                                       equal to 6               (87.5%)       (75.7%)       (71.0%)       (77.9%)
                                                                       Greater than 6               5           18              9            32
                                                                                                (12.5%)       (24.3%)       (29.0%)       (22.1%)
                                         San Bernadino                 Less than or               61            78             23           162
                                                                       equal to 6               (56.0%)       (88.6%)       (59.0%)       (68.6%)
                                                                       Greater than 6             48            10             16            74
                                                                                                (44.0%)       (11.4%)       (41.0%)       (31.4%)
                                         Washington, D.C.              Less than or               25            17              5            47
                                                                       equal to 6               (30.5%)       (39.5%)       (55.6%)       (35.1%)
                                                                       Greater than 6             57            26              4            87
                                                                                                (69.5%)       (60.5%)       (44.4%)       (64.9%)
                                         Six Cities Combined           Less than or               331           330            92           753
                                                                       equal to 6               (55.0%)       (68.3%)       (62.6%)       (61.1%)
                                                                       Greater than 6             271           153            55           479
                                                                                                (45.0%)       (31.7%)       (37.4%)       (38.9%)
                                         Note: See note for table 3.
                                         a
                                          We defined an area’s income level as “low” if the per capita income was at or below 80 percent
                                         of the per capita income for the city, “medium” if the per capita income was greater than
                                         80 percent but at or below 120 percent of the city’s level, and “high” if the per capita income was
                                         greater than 120 percent of city’s level.



                                         Source: GAO’s analysis of data from HUD and the Bureau of the Census.




                                         Page 11                                                           GAO/RCED-98-2 Homeownership
                  B-276286




                  Additional details about the amount of time HUD-owned properties
                  remained unsold in the six cities appear in appendix IV.


                  We provided HUD with a draft of this report for its review and comment.
Agency Comments   Officials who reviewed the report, including a representative from the
                  Office of the Assistant Secretary for Housing-Federal Housing
                  Commissioner, stated that they generally agreed with the report’s findings.
                  HUD also provided several clarifying comments, which we incorporated
                  into the report as appropriate.


                  In reporting information relating to early foreclosures on FHA-insured
Scope and         single-family loans endorsed during calendar years 1992 through 1994 in
Methodology       low-, medium-, and high-income areas nationwide, we relied on HUD’s
                  analysis of the number of loans made, the number of early foreclosures,
                  and the early foreclosure rates in the three income areas. To determine
                  early foreclosure rates for the same period in the six cities reviewed, we
                  obtained data from HUD’s database on loans insured by FHA in calendar
                  years 1992 through 1994 and merged this information with 1990 census
                  data.

                  To further analyze lending and early foreclosure patterns in the six cities,
                  we divided the lenders into two groups—those with no early foreclosures
                  and those with one or more early foreclosures during the periods we
                  reviewed—and compared these groups with respect to the distribution of
                  the loans they made across income areas. To obtain information on factors
                  that contribute to differences in early foreclosure rates among income
                  areas, we performed an analysis to show the extent to which certain
                  variables were associated with differences in the probability of early
                  foreclosure. Appendix III provides information on the model we built to
                  estimate relationships between early foreclosures and factors that
                  contribute to such foreclosures. To compare the length of time HUD-owned
                  properties remained unsold in low-, medium-, and high-income areas in the
                  six cities, we obtained data from HUD’s Single-Family Accounting
                  Management System (SAMS), which tracks properties acquired and sold by
                  HUD. Our analysis focused on single-family properties that remained in
                  HUD’s inventory as of December 31, 1996.


                  While we did not independently verify the accuracy or test the reliability of
                  FHA’s data, we performed tests to check the internal consistency of the
                  data and worked with agency officials to ensure that we interpreted the



                  Page 12                                           GAO/RCED-98-2 Homeownership
B-276286




data properly. Appendix V provides additional details on our scope and
methodology.

We performed our work from December 1996 through September 1997 in
accordance with generally accepted government auditing standards.


As agreed with your office, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 7 days after the
date of this letter. At that time, we will provide copies to the Secretary of
HUD and other interested parties. We will also make copies available to
others upon request.

Please call me at (202) 512-7631 if you or your staff have any questions.
Major contributors to this report are listed in appendix VI.

Sincerely yours,




Judy A. England-Joseph
Director, Housing and Community
  Development Issues




Page 13                                            GAO/RCED-98-2 Homeownership
Contents



Letter                                                                        1


Appendix I                                                                   18

Early Foreclosure
Rates on Loans Made
in Calendar Years
1992-94, by Income
Areas
Appendix II                                                                  19

Data on Lenders With
and Without Early
Foreclosures on
Loans Made in
Calendar Years
1992-94, by Income
Areas
Appendix III                                                                 20
                        Data Used in This Analysis                           20
GAO’s Econometric       Specification of the Model                           23
Model Used to           Estimation Results                                   26
Identify Factors
Associated With Early
Foreclosures




                        Page 14                      GAO/RCED-98-2 Homeownership
                        Contents




Appendix IV                                                                                  29

Time in Inventory for
Single-Family
Properties in Six
Cities That Remained
Unsold as of
December 31, 1996, by
Income Areas
Appendix V                                                                                   31

Objectives, Scope,
and Methodology
Appendix VI                                                                                  34

Major Contributors to
This Report
Tables                  Table 1: Early Foreclosure Rates for FHA-Insured Loans Made in        6
                          Calendar Years 1992-94 in Six Cities, by Income Areas
                        Table 2: Proportion of FHA-Insured Loans Made in Calendar             8
                          Years 1992-94 for Properties in Low-, Medium-, and High-Income
                          Areas in Six Cities, by Lenders With and Without Early
                          Foreclosures
                        Table 3: Median Months in Inventory for Single-Family Properties     10
                          in Six Cities That Remained Unsold as of December 31, 1996, by
                          Income Areas
                        Table 4: Months in Inventory for Single-Family Properties in Six     11
                          Cities That Remained Unsold as of December 31, 1996, by Income
                          Areas
                        Table III.1: Number of Loans Made in Six MSAs and Number of          21
                          Loans Matched With Census Data, Calendar Years 1992-94
                        Table III.2: Number of Loans Made Within the Six Cities,             22
                          Calendar Years 1992-94
                        Table III.3: Logistic Regression Summary Table                       27




                        Page 15                                      GAO/RCED-98-2 Homeownership
Contents




Abbreviations

FHA        Federal Housing Administration
GAO        General Accounting Office
HUD        Department of Housing and Urban Development
MMI        Mutual Mortgage Insurance Fund
MSA        metropolitan statistical area
SAMS       Single-Family Accounting Management System


Page 16                                     GAO/RCED-98-2 Homeownership
Page 17   GAO/RCED-98-2 Homeownership
Appendix I

Early Foreclosure Rates on Loans Made in
Calendar Years 1992-94, by Income Areas


                                                                                      San        Washington,         Six cities
Income level             Atlanta   Baltimore        Chicago           Dallas     Bernadino              D.C.        combined
Low-income areas
Number of loans             786          637           6,805           1,994           2,547                1,281      14,050
Percent of loans           27.5           7.5           33.0            20.3            48.0                 39.4         27.9
Number of early
foreclosures                 11               5           33              29              29                   6           113
Percent of early
foreclosures               61.1           8.9           62.3            46.8            51.8                 66.7         44.5
Early foreclosure rate     1.40         0.78            0.48            1.45            1.14                 0.47         0.80
Medium-income areas
Number of loans           1,219        3,578           9,691           3,176           2,061                1,457      21,182
Percent of loans           42.6         42.2            47.1            32.3            38.8                 44.8         42.1
Number of early
foreclosures                  5           26              15              25              21                   3            95
Percent of early
foreclosures               27.8         46.4            28.3            40.3            37.5                 33.3         37.4
Early foreclosure rate     0.41         0.73            0.15            0.79            1.02                 0.21         0.45
High-income areas
Number of loans             856        4,255           4,100           4,665             701                 514       15,091
Percent of loans           29.9         50.2            19.9            47.4            13.2                 15.8         30.0
Number of early
foreclosures                  2           25                5              8               6                   0            46
Percent of early
foreclosures               11.1         44.6              9.4           12.9            10.7                  0.0         18.1
Early foreclosure rate     0.23         0.59            0.12            0.17            0.86                  0.0         0.30
All income areas
Number of loans           2,861        8,470          20,596           9,835           5,309                3,252      50,323
Percent of loans          100.0        100.0           100.0           100.0           100.0                100.0        100.0
Number of early
foreclosures                 18           56              53              62              56                   9           254
Percent of early
foreclosures              100.0        100.0           100.0           100.0           100.0                100.0        100.0
Early foreclosure rate     0.63         0.66            0.26            0.63            1.05                 0.28         0.50

                                    Source: GAO’s analysis of data from HUD and the Bureau of the Census.




                                    Page 18                                                      GAO/RCED-98-2 Homeownership
Appendix II

Data on Lenders With and Without Early
Foreclosures on Loans Made in Calendar
Years 1992-94, by Income Areas

                                                                Income level of areas
                                 Low                       Medium                       High                     All
City/ type of lender         Number    Percent         Number       Percent      Number         Percent     Number     Percent
Atlanta
With early foreclosures         310           34.9          400         45.0          179          20.1         889      100.0
Without early foreclosures      476           24.1          819         41.5          677          34.3       1,972      100.0
Baltimore
With early foreclosures         432            8.4        2,203         43.1        2,480          48.5       5,115      100.0
Without early foreclosures      205            6.1        1,375         41.0        1,775          52.9       3,355      100.0
Chicago
With early foreclosures        4,918          36.3        6,333         46.7        2,300          17.0      13,551      100.0
Without early foreclosures     1,887          26.8        3,358         47.7        1,800          25.6       7,045      100.0
Dallas
With early foreclosures        1,317          24.2        1,926         35.3        2,208          40.5       5,451      100.0
Without early foreclosures      677           15.4        1,250         28.5        2,457          56.0       4,384      100.0
San Bernadino
With early foreclosures        1,424          46.8        1,186         39.0          434          14.3       3,044      100.0
Without early foreclosures     1,123          49.6          875         38.6          267          11.8       2,265      100.0
Washington, D.C.
With early foreclosures         365           40.5          428         47.5          109          12.1         902      100.0
Without early foreclosures      916           39.0        1,029         43.8          405          17.2       2,350      100.0
Six cities combined
With early foreclosures        8,766          30.3       12,476         43.1        7,710          26.6      28,952      100.0
Without early foreclosures     5,284          24.7        8,706         40.7        7,381          34.5      21,371      100.0
                                    Note: Percentages may not add to 100 because of rounding.



                                    Source: GAO’s analysis of data from HUD and the Bureau of the Census.




                                    Page 19                                                        GAO/RCED-98-2 Homeownership
Appendix III

GAO’s Econometric Model Used to Identify
Factors Associated With Early Foreclosures

                    This appendix describes the econometric model we developed and the
                    analysis we conducted to estimate the associations between early
                    foreclosures and several explanatory variables. The explanatory variables
                    we tested were the loan-to-value ratio, loan amount, contract interest rate,
                    city where the property was located, and neighborhood income. The
                    equation we estimated used all of the FHA-insured single-family loans
                    endorsed in calendar years 1992 through 1994 in six cities—Atlanta,
                    Georgia; Baltimore, Maryland; Chicago, Illinois; Dallas, Texas; San
                    Bernadino, California; and Washington, D.C. We excluded loans made for
                    properties within the metropolitan statistical area (MSA) but outside the
                    city’s boundaries. We relied on census data to determine the per capita
                    income of the census tracts in the six cities. The data we used, our model,
                    and the results we obtained are discussed in detail in the following
                    sections.


                    For our analysis, we combined FHA’s computerized data from two separate1
Data Used in This   files of 2,945,252 mortgages endorsed in calendar years 1992, 1993, and
Analysis            1994. We then merged the combined FHA data files for the selected cities
                    with census data to obtain income information for the census tracts where
                    the loans were made. From FHA’s records, we obtained information on the
                    initial characteristics of each loan, such as the year of its endorsement,
                    state and city in which it was originated, loan-to-value ratio, loan amount,
                    and loan interest rate. FHA’s files contain information on all of the
                    single-family loans that FHA insured, including loans for condominium
                    units, loans made to refinance existing mortgages, rehabilitation loans,
                    and loans covered under FHA’s special risk insurance program. From the
                    Bureau of the Census, we obtained data on the aggregate household
                    income and total population for each of the six relevant MSAs. We
                    computed the per capita income for each tract by dividing its aggregate
                    household income2 by its total population. We determined the per capita
                    income for each city by dividing the aggregate household income for all of
                    the census tracts within its borders by its total population.



                    1
                     FHA’s A-43 database provides current and historical information on the mortgage loans that FHA
                    insures. FHA’s F-42 database provides additional information on characteristics such as the age, race,
                    and income of FHA borrowers.
                    2
                     We excluded the income of persons in group quarters and institutions from our calculation of per
                    capita income. For the six cities combined, about 97 percent of the census tracts did not have persons
                    in group quarters and institutions, and such persons accounted for less than 10 percent of the
                    population in 76 percent of the remaining census tracts. We determined that our classification of
                    census tracts as low-, medium-, or high-income was not affected by our exclusion of the income of
                    persons in group quarters and institutions.



                    Page 20                                                            GAO/RCED-98-2 Homeownership
                                       Appendix III
                                       GAO’s Econometric Model Used to Identify
                                       Factors Associated With Early Foreclosures




                                       Within the five states covered by our review (Illinois, Georgia, Texas,
                                       California, and Maryland) and the city of Washington, D.C., 859,128 loans
                                       were made during the 3-year period. We selected the loans originated in
                                       each of the six cities by first using the county codes3 for the appropriate
                                       MSA and then identifying the census tracts that were within the city’s
                                       borders according to the listing of tracts supplied to us by an official
                                       representative of each city. As indicated in table III.1, 399,011 loans were
                                       endorsed in the six MSAs that included the six cities during calendar years
                                       1992 through 1994.

Table III.1: Number of Loans Made in
Six MSAs and Number of Loans                                                                                    Number and percent of
Matched With Census Data, Calendar                                     Total number     Number of                 loans not matched
Years 1992-94                          MSA                          of loans in MSA loans matched                   Number           Percent
                                       Atlanta                                 83,320               67,091           16,229                 19
                                       Baltimore                               54,612               43,468           11,144                 20
                                       Chicago                                 96,751               80,549           16,202                 17
                                       Dallas                                 114,534              101,521           13,013                 11
                                       San Bernadino                           45,999               37,164             8,835                19
                                       Washington, D.C.                          3,795               3,252               543                14
                                       Total                                  399,011              333,045           65,966                 17

                                       We were able to match FHA loans to census records for 83 percent of the
                                       loans (333,045) in the MSA, but not for the remaining 17 percent (65,966
                                       loans). Because we used census tract codes to determine if the loans were
                                       within or outside a city, we were not able to determine what percentage of
                                       the 65,966 unmatched loans were within a city’s borders. We matched
                                       80 percent of the total number of loans with all six digits of the census
                                       tract code and an additional 3 percent with four digits of the census tract
                                       code. The four-digit match was necessary because of changes to the
                                       definitions of some metropolitan area tracts over time.

                                       In general, each of the MSAs had hundreds of census tracts, but only a
                                       fraction of them were located within the city’s borders. We excluded from
                                       our analysis 543 loans for properties in Washington, D.C., because invalid
                                       census tract codes made it difficult to obtain census tract income and
                                       population data. In addition, there were 65,423 loans endorsed in the
                                       remaining MSAs that we could not identify as being within one of the cities

                                       3
                                        According to HUD officials, the codes for the state, county, and census tract are the most important
                                       because the metropolitan area can be identified from these codes (except for split tracts in New
                                       England). Of the 859,128 loans endorsed in the five states and Washington, D.C., 4,537 loans did not
                                       have an appropriate county code. Therefore, we could not tell if these loans were made in the six cities
                                       we reviewed.



                                       Page 21                                                            GAO/RCED-98-2 Homeownership
                                        Appendix III
                                        GAO’s Econometric Model Used to Identify
                                        Factors Associated With Early Foreclosures




                                        because their census tract codes were invalid. Another 4,537 loans
                                        endorsed in the six states did not have valid county codes, and we were
                                        unable to determine if they should have been included within one of the
                                        MSAs.


                                        As shown in table III.2, of the 333,045 loans we were able to match with
                                        census tracts, 50,323 were made for properties within the six cities’
                                        borders. We were able to find the valid census tract income for virtually all
                                        of the 50,323 loans. In other words, when we identified a loan as being for
                                        a property in one of the six cities, we were almost always able to
                                        determine the total population or aggregate income for that loan’s census
                                        tract.

Table III.2: Number of Loans Made
Within the Six Cities, Calendar Years                                                                                   Number of loans
1992-94                                                                                                              identified within the
                                        City                                                                                 city’s border
                                        Atlanta                                                                                       2,861
                                        Baltimore                                                                                     8,470
                                        Chicago                                                                                      20,596
                                        Dallas                                                                                        9,835
                                        San Bernadino                                                                                 5,309
                                        Washington, D.C.                                                                              3,252
                                        Total                                                                                        50,323

                                        Many FHA-insured loans were refinanced during calendar years 1992
                                        through 1994. Refinanced mortgages4 accounted for about 32 percent of
                                        the loans in the six cities during the 3-year period we examined. Of the
                                        loans that were refinanced, about 69 percent had a recorded loan-to-value
                                        ratio of zero, and nearly all of these were streamlined refinanced
                                        mortgages.5 Because FHA does not require property appraisals for
                                        streamlined refinanced mortgages, the initial loan-to-value ratios for these
                                        loans are unknown.




                                        4
                                         Borrowers often refinance mortgage loans to lower their monthly principal and interest payments
                                        when interest rates decline. Of the refinanced mortgages, 89 percent were “streamlined refinanced,”
                                        meaning that the old FHA-insured mortgage loan was repaid from the proceeds of a new FHA-insured
                                        loan using the same property as security. Appraisals and credit checks are not required by FHA on
                                        these loans, and borrowers cannot obtain cash from the transaction except for minor adjustments not
                                        exceeding $250 at closing.
                                        5
                                         FHA’s data did not indicate whether there were any existing second mortgages on these properties.



                                        Page 22                                                          GAO/RCED-98-2 Homeownership
                       Appendix III
                       GAO’s Econometric Model Used to Identify
                       Factors Associated With Early Foreclosures




                       A default on a home mortgage loan may be triggered by unemployment,
Specification of the   divorce, death, or some other event. Such an event is not likely to trigger a
Model                  foreclosure if the owner has positive equity in the home because the sale
                       of the home with the realization of a profit is better than the loss of the
                       home through foreclosure. However, if the property is worth less than the
                       mortgage, such an event may trigger a foreclosure.

                       We hypothesized that the probability of early foreclosure is influenced by,
                       among other things, the loan-to-value ratio, the size of the loan, the loan
                       interest rate, income, and the property’s location. Because the recorded
                       value of the loan-to-value ratio for some loans was zero, we added a
                       variable to our analysis to identify these loans. We used a logistic
                       regression equation to explore how foreclosure rates on loans endorsed in
                       calendar years 1992 through 1994 in the six cities varied for each of these
                       factors. Logistic regression is a standard procedure for analyzing a
                       dichotomous dependent variable, such as whether or not an early
                       foreclosure occurred. We used the results of our logistic regression to
                       estimate how the odds of early foreclosure are expected to change with
                       unit changes in the explanatory factors. In the logistic regression, we used
                       deviation coding for categorical variables, such as the city where the
                       foreclosure occurred. Therefore, the effect for each category is compared
                       to the average effect for all of the categories, rather than to an omitted (or
                       reference) category.

                       We tested additional factors but did not find them to be significantly
                       associated with early foreclosure rates after accounting for the factors
                       listed above. These additional factors were the race of the borrower
                       (white or other), the age of the borrower, the year of the loan’s
                       endorsement (1992, 1993, 1994), and the loan program used (the MMI
                       Fund’s 203(b) program or other loan program).6

                       We were not able to include all of the factors, such as unemployment
                       rates, that might be related to the probability of early foreclosure in our
                       analysis. This was generally because data were not available. If we had
                       been able to include these other factors, our results with respect to the
                       included factors might have been different. We and other researchers have
                       estimated the probability of ultimate foreclosure and have found other


                       6
                        In HUD’s database the age of the borrower was recorded as zero—an invalid figure—for about
                       18 percent of the loans. To compensate for the missing data, we included in our analysis of age a
                       dummy variable indicating whether or not the information on age was missing. Neither the coefficient
                       for the continuous age variable nor the coefficient for the dummy variable was significant at the 0.05
                       level. The significance of the variables added were as follows: race, 0.38; endorsement year, 0.80; loan
                       program, 0.46; age dummy variable, 0.41; age as a continuous variable, 0.18.



                       Page 23                                                             GAO/RCED-98-2 Homeownership
                      Appendix III
                      GAO’s Econometric Model Used to Identify
                      Factors Associated With Early Foreclosures




                      factors that have a significant impact on it. These factors include the
                      borrower’s equity and the prevailing interest rate at the time of default,
                      lagged unemployment, the property’s location (i.e., urban or rural),
                      whether the borrower is a first-time homeowner, and the borrower’s
                      marital status. It is generally agreed that many life-changing events—such
                      as the arrival of children, divorce, and death—may also be related to the
                      probability of foreclosure. However, it should be noted that prior research
                      has associated these other factors only to ultimate loan foreclosure, not to
                      early foreclosure.


Income                To determine if early foreclosure rates were different in lower-income
                      communities, we obtained information on the aggregate income and the
                      total population for each census tract within the borders of the cities we
                      studied. We computed the ratio of the per capita income for each of the
                      tracts to the per capita income for the relevant city to obtain the
                      tract-to-city income ratio. We anticipated that people living in
                      lower-income tracts might have more difficulty meeting their mortgage
                      payments than people in higher-income tracts and that the rate of early
                      foreclosure would, then, be higher in the lower-income tracts than
                      elsewhere. Factors associated with lower-income communities, such as
                      higher unemployment rates and less stability in employment, could limit
                      the ability of borrowers to meet their monthly mortgage payments. Other
                      factors, such as the greater age of the housing stock or the slower
                      appreciation of house prices in lower-income communities, could also
                      affect early foreclosure rates.


Loan-to-Value Ratio   The ratio of the loan amount to the value of the property is an important
                      determinant of whether a loan will end in foreclosure. The loan-to-value
                      ratio on the property changes over time because property values can
                      increase or decrease, and payments reduce the amount owed on a
                      mortgage. Because we were examining foreclosures that occur within 18
                      months of the loan endorsement date, we anticipated that the change in
                      the loan-to-value ratio within that time period would be so small that the
                      initial loan-to-value ratio would be sufficient to capture the effect of the
                      borrower’s equity percentage on the probability of foreclosure, when the
                      equity percentage is considered to be 1 minus the loan-to-value ratio.
                      Research indicates that borrowers with small amounts of equity (and,




                      Page 24                                           GAO/RCED-98-2 Homeownership
                         Appendix III
                         GAO’s Econometric Model Used to Identify
                         Factors Associated With Early Foreclosures




                         hence, higher loan-to-value ratios), especially those with negative equity,
                         are more likely than other borrowers to default.7

                         FHA’s data showed a value of zero for about 22 percent of the loans in the
                         six cities. Although almost all of these loans were refinanced, another
                         10 percent of the loans were refinanced and had valid loan-to-value ratios.
                         FHA does not require an appraisal for streamlined refinanced loans. When
                         an appraisal is not performed, the loan-to-value ratio is unknown. We have
                         reported that the probability of foreclosure for FHA-insured refinanced
                         loans differs from that for other FHA-insured loans,8 but we did not include
                         a refinance indicator in the regression. We did, however, add a variable to
                         indicate when a loan was missing a loan-to-value ratio. We did not
                         separately take into account any further differences that may result from
                         other characteristics of refinanced loans that did have valid loan-to-value
                         ratios.


Interest Rate and Loan   We included the interest rate on the mortgage as an explanatory variable
Amount                   in the early foreclosure equation. We expected a higher interest rate to be
                         associated with a higher probability of early foreclosure because a higher
                         interest rate causes a higher monthly payment.

                         To obtain insight into the differential effect of relatively larger loans on the
                         probability of early foreclosure, we used the loan amount as an
                         explanatory variable. In our previously cited report, we pointed out that,
                         other things being equal, larger loans have lower probabilities of
                         foreclosure than smaller loans. Different rates of appreciation in house
                         prices in low- and higher-income communities may be one factor
                         underlying this phenomenon. We know that larger loans are associated
                         with higher-priced homes. By using the loan amount as a variable in our
                         equation and holding income constant, we were testing the relationship
                         between larger loans and the probability of early foreclosure.


City Where Property Is   We used variables to indicate the city where the property was located. We
Located                  expected that the coefficients for these variables would pick up
                         differences in economic conditions within the city that we could not
                         model explicitly. Some of these differences may include changes in the


                         7
                          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.
                         8
                          Mortgage Financing: FHA Has Achieved Its Home Mortgage Capital Reserve Target
                         (GAO/RCED-96-50, Apr. 12, 1996)



                         Page 25                                                            GAO/RCED-98-2 Homeownership
                     Appendix III
                     GAO’s Econometric Model Used to Identify
                     Factors Associated With Early Foreclosures




                     rates of unemployment, house price appreciation, net migration, and other
                     unknown factors.


                     The results of our analysis are summarized in table III.3. In general, our
Estimation Results   results are consistent with the economic reasoning that underlies our
                     model.




                     Page 26                                           GAO/RCED-98-2 Homeownership
                                   Appendix III
                                   GAO’s Econometric Model Used to Identify
                                   Factors Associated With Early Foreclosures




Table III.3: Logistic Regression
Summary Table                                                                                          Odds change factor

                                                                        Significance                            Confidence interval
                                   Variable                                    levela         Estimate
                                   Intercept                                       0.00             0.00              0.00              0.00
                                   City                                            0.00
                                   Atlanta                                         0.49             1.16              0.76              1.76
                                   Baltimore                                       0.06             1.33              0.99              1.79
                                   Chicago                                         0.00             0.47              0.35              0.62
                                   Dallas                                          0.64             1.07              0.80              1.42
                                   San Bernadino                                   0.00             2.02              1.51              2.71
                                   Washington, D.C.                                0.20             0.65              0.32              1.32
                                   Income: ratio of the
                                   per-capita income for the
                                   tract to that for the city (ratio
                                   x 100)                                          0.00             0.99              0.98              0.99
                                   Interest rate (percent)                         0.02             1.16              1.02              1.32
                                   Loan amount (dollars in
                                   thousands)                                      0.02             0.99              0.99              1.00
                                   Loan-to-value ratio
                                   (ratio x 100)                                   0.00             1.06              1.03              1.09
                                   Is the loan-to-value ratio
                                   provided in the data equal to
                                   zero?                                           0.00
                                   Yes                                             0.00            13.89              3.60            53.53
                                   No                                              0.00             0.07              0.02              0.28
                                   Number of observations                                                                            50,318
                                   a
                                    We interpreted a value of less than 0.05 as indicating a statistically significant association
                                   between the odds of early foreclosure and the variable or characteristic. We did not conclude that
                                   a statistically significant association existed if the value was more than 0.05.
                                   b
                                    In logistic regression, the coefficients of the variables are not easily interpretable. Therefore, we
                                   transformed the original coefficients into a more interpretable form that we termed the “odds
                                   change factor.” Specifically, we raised the natural logarithm base, e, to the power equal to the
                                   value of the original coefficient to obtain the odds change factor. Odds change factors estimate
                                   the effect of each variable on the predicted odds of foreclosure. A value greater than 1 means
                                   that the odds of foreclosure are expected to increase, while a value less than 1 predicts a
                                   decrease in the odds of foreclosure. For example, the odds change factor for the interest rate
                                   variable is 1.16, which means that the odds of early foreclosure increase by 16 percent for each
                                   percentage point the interest rate increases. Confidence intervals were also calculated for the
                                   original logistic regression coefficients at the 95-percent confidence level and then transformed
                                   into the more interpretable form. This means that we would expect the lower and upper bound to
                                   include the true odds change factor 95 times out of 100.




                                   Page 27                                                              GAO/RCED-98-2 Homeownership
Appendix III
GAO’s Econometric Model Used to Identify
Factors Associated With Early Foreclosures




We found statistically significant associations9 between increased rates of
early foreclosure and (1) a lower per capita income for a census tract,
(2) higher loan-to-value ratios, (3) higher loan interest rates, (4) smaller
loan amounts, and (5) loans made for properties located in San Bernadino.
We also found that early foreclosure was less likely for loans made for
properties in Chicago.

As the per capita income in the census tract in which the property was
located increased relative to the per capita income in the entire city, the
odds of early foreclosure decreased. For example, the odds of foreclosure
for loans on properties located in areas whose per capita income was
91 percent of the citywide per capita income were about 1 percent lower
than the odds for properties in areas whose per capita income was
90 percent of the citywide income. Larger mortgages were negatively
correlated with the probability of early foreclosure. The odds of early
foreclosure were estimated to decrease by about 1 percent for each
additional $1,000 borrowed.

The loan-to-value ratio was significantly and positively correlated with the
odds of early foreclosure. When the loan-to-value ratio increased by
1 percentage point, the odds of early foreclosure increased by about
6 percent. The odds of early foreclosure for loans with a loan-to-value ratio
of zero—mostly streamlined financed loans—were about the same as the
odds for loans with a loan-to-value ratio of 90 percent and were about
25 percent lower than the odds for loans with a loan-to-value ratio of
95 percent.10

Higher interest rates are associated with an increase in early foreclosures.
Holding other things constant, an increase of 1 percentage point in the
interest rate was found to increase the odds of early foreclosure by about
16 percent.

We also found that the odds of early foreclosure differed with the city
being tested. For example, the odds of early foreclosure were lower than
average for Chicago and about twice as high as the six-city average for San
Bernadino. We did not obtain statistically significant results for Atlanta,
Baltimore, Dallas, or Washington, D.C.



9
 We used the 95-percent level of confidence.
10
  We obtained these results by jointly considering the effects of the loan-to-value (LTV) ratio and the
LTV-equals-zero indicator. Because the LTV ratio recorded in FHA’s database determined the values
for both of these variables, both coefficients must be considered.



Page 28                                                             GAO/RCED-98-2 Homeownership
Appendix IV

Time in Inventory for Single-Family
Properties in Six Cities That Remained
Unsold as of December 31, 1996, by Income
Areas
                                                                Income level of areas
                                   Low                    Medium                    High                     All
Months in inventory            Number    Percent       Number      Percent    Number        Percent    Number       Percent
Atlanta
Less than or equal to 6            19        61.3          20         64.5              2      66.7         41         63.1
Greater than 6, less than or
equal to 12                         9        29.0           8         25.8              0        0          17         26.2
Greater than 12, less than
or equal to 24                      1         3.2           0           0               0        0           1           1.5
Greater than 24                     2         6.5           3          9.7              1      33.3          6           9.2
Total                              31       100.0          31        100.0              3     100.0         65        100.0
Baltimore
Less than or equal to 6            30        51.7          52         65.0          25         58.1        107         59.1
Greater than 6, less than or
equal to 12                        17        29.3          22         27.5              9      20.9         48         26.5
Greater than 12, less than
or equal to 24                      8        13.8           3          3.8              5      11.6         16           8.8
Greater than 24                     3         5.2           3          3.8              4       9.3         10           5.5
Total                              58       100.0          80        100.0          43        100.0        181        100.0
Chicago
Less than or equal to 6           161        57.1         107         64.1          15         68.2        283         60.1
Greater than 6, less than or
equal to 12                        59        20.9          29         17.4              4      18.2         92         19.5
Greater than 12, less than
or equal to 24                     27         9.6          10          6.0              0        0          37           7.9
Greater than 24                    35        12.4          21         12.6              3      13.6         59         12.5
Total                             282       100.0         167        100.0          22        100.0        471        100.0
Dallas
Less than or equal to 6            35        87.5          56         75.7          22         71.0        113         77.9
Greater than 6, less than or
equal to 12                         5        12.5          11         14.9              4      12.9         20         13.8
Greater than 12, less than
or equal to 24                      0              0        2          2.7              0        0           2           1.4
Greater than 24                     0              0        5          6.8              5      16.1         10           6.9
Total                              40       100.0          74        100.0          31        100.0        145        100.0
San Bernadino
Less than or equal to 6            61        56.0          78         88.6          23         59.0        162         68.6
Greater than 6, less than or
equal to 12                        25        22.9           6          6.8              8      20.5         39         16.5
Greater than 12, less than
or equal to 24                     20        18.3           4          4.5              6      15.4         30         12.7
Greater than 24                     3         2.8           0           0               2       5.1          5           2.1
                                                                                                                 (continued)

                                         Page 29                                               GAO/RCED-98-2 Homeownership
                                         Appendix IV
                                         Time in Inventory for Single-Family
                                         Properties in Six Cities That Remained
                                         Unsold as of December 31, 1996, by Income
                                         Areas




                                                                     Income level of areas
                                   Low                        Medium                           High                           All
Months in inventory            Number    Percent        Number          Percent        Number         Percent        Number          Percent
Total                             109       100.0              88          100.0              39         100.0             236          100.0
Washington, D.C.
Less than or equal to 6            25         30.5             17           39.5               5           55.6             47           35.1
Greater than 6, less than or
equal to 12                        24         29.3             10           23.3               2           22.2             36           26.9
Greater than 12, less than
or equal to 24                     15         18.3               7          16.3               1           11.1             23           17.2
Greater than 24                    18         22.0               9          20.9               1           11.1             28           20.9
Total                              82       100.0              43          100.0               9         100.0             134          100.0
Six cities combined
Less than or equal to 6           331         55.0            330           68.3              92           62.6            753           61.1
Greater than 6, less than or
equal to 12                       139         23.1             86           17.8              27           18.4            252           20.5
Greater than 12, less than
or equal to 24                     71         11.8             26             5.4             12            8.2            109             8.8
Greater than 24                    61         10.1             41             8.5             16           10.9            118             9.6
Total                             602       100.0             483          100.0            147          100.0           1,232          100.0

                                         Note: We excluded from our analysis properties held off the market as of May 17, 1997 (the date
                                         our data file was created); however, we were unable to determine whether included properties
                                         had been held off the market for any time in the past. In addition, in some cases, we were either
                                         unable to identify the census tract where a property was located or HUD’s data did not provide
                                         the date a property entered HUD’s inventory. Therefore, we excluded these properties from our
                                         analysis. The percentage of properties in each city that we excluded from our analysis because of
                                         missing information on the census tract or the time in inventory was as follows: Atlanta, 3 percent
                                         (2 of 67 properties); Baltimore, 14 percent (29 of 210 properties); Chicago, 4 percent (19 of 490
                                         properties); Dallas, 5 percent (8 of 153 properties); San Bernadino, 20 percent (59 of 295
                                         properties); and Washington, D.C., 16 percent (25 of 159 properties).



                                         Source: GAO’s analysis of data from HUD and the Bureau of the Census.




                                         Page 30                                                           GAO/RCED-98-2 Homeownership
Appendix V

Objectives, Scope, and Methodology


              Our objectives were to (1) compare early foreclosure rates on FHA-insured
              single-family loans made in low-, medium-, and high-income areas
              nationwide and in the six cities; (2) compare across income areas the
              proportion of loans made in the six cities by FHA-approved mortgage
              lenders with and without early foreclosures; (3) identify factors that
              influence early foreclosure rates; and (4) compare the length of time
              HUD-owned single-family properties remained unsold in low-, medium-, and
              high-income areas in the six cities.

              In reporting information relating to early foreclosures on FHA-insured
              single-family loans endorsed during calendar years 1992 through 1994 in
              low-, medium-, and high-income areas nationwide, we relied on HUD’s
              analysis of the number of loans made, the number of early foreclosures,
              and the early foreclosure rates in the three income areas. To determine
              early foreclosure rates for the same period in the six cities reviewed, we
              obtained data from HUD’s database on loans insured by FHA in calendar
              years 1992 through 1994 and merged this information with 1990 census
              data. Detailed information on the data we used are provided in the section
              of appendix III that discusses the data used in this analysis.

              We defined a census tract’s income level as “low” if the per capita income
              was at or below 80 percent of the city’s per capita income, “medium” if the
              per capita income was greater than 80 percent but at or below 120 percent
              of the city’s level, and “high” if the per capita income was greater than
              120 percent of the city’s level. Although HUD usually uses the median family
              income to identify low-, medium-, and high-income census tracts, we were
              unable to compute the median family income for the six cities from the
              data we extracted from census records. We therefore used the per capita
              income as our income measure.

              HUD  computed early foreclosure rates by income level nationwide for this
              report using the average household income as the income measure for
              each MSA. As indicated above, we used the per capita income for each city
              as the income measure to calculate early foreclosure rates by income level
              for the six cities. Therefore, our classification of census tracts as low-,
              medium-, or high-income may differ from HUD’s classification because
              (1) the average income for the MSA may differ from the per capita income
              for the city, and (2) the per capita income does not take into account
              differences in the average household size among the three income groups.
              While our classification of census tracts differed from HUD’s classification,
              the relationship between early foreclosure rates and census tract income
              levels for both computations was similar.



              Page 31                                           GAO/RCED-98-2 Homeownership
Appendix V
Objectives, Scope, and Methodology




We limited our analysis to early foreclosures, that is, to those occurring
within 18 months of the loan endorsement date. To determine whether a
foreclosure occurred within that time period, we measured the time
elapsed between FHA’s endorsement of the loan and the date the lender
foreclosed on the loan. For this report, we included in our calculation of
early foreclosure rates loans on which the lender did not actually foreclose
but on which FHA paid an insurance claim to the lender within 18 months
of the loan endorsement date. We excluded from our calculation of early
foreclosure rates nonconveyance foreclosures, such as instances during
which a foreclosure occurs but an insurance claim is not paid. In some
cases, early foreclosures may not have been reflected in the data from HUD
that we used because of the lag between the date of the actual foreclosure
and the date it was recorded in HUD’s database. As a result, our analysis
may understate the number of early foreclosures by the number of these
unrecorded cases.

To further analyze lending and early foreclosure patterns in the six cities,
we divided the lenders into two groups—those with no early foreclosures
and those with one or more early foreclosures during the periods
reviewed—and compared these groups with respect to the distribution of
the loans they made across income areas. We determined whether a lender
had one or more early foreclosures on a city-by-city basis. Therefore, any
lender that made loans in more than one of the six cities could be
classified in the group of lenders with early foreclosures in one city and in
the group of lenders without early foreclosures in another city.

To obtain information on factors that contribute to differences in early
foreclosure rates among income areas, we performed an analysis to show
the extent to which certain variables were associated with differences in
the probability of early foreclosure. Appendix III provides information on
the model we built to estimate relationships between early foreclosures
and factors that contribute to such foreclosures. In addition, we reviewed
the mortgage finance literature and interviewed officials from HUD’s Office
of Insured Single-Family Housing and HUD field office officials in each of
the six cities. We also interviewed local government officials and nonprofit
housing executives familiar with FHA’s role in the real estate markets in
each of the six cities.

To compare the length of time HUD-owned properties remained unsold in
low-, medium-, and high-income areas in the six cities, we obtained data
from HUD’s Single-Family Accounting Management System (SAMS), which
tracks properties acquired and sold by HUD. Our analysis focused on



Page 32                                           GAO/RCED-98-2 Homeownership
Appendix V
Objectives, Scope, and Methodology




single-family properties that remained in HUD’s inventory as of December
31, 1996. We measured the time in inventory from the date that HUD
acquired the property. We excluded properties held off the market as of
May 17, 1997 (the date our data extract was created), but we were unable
to determine if the remaining properties had been held off the market for
any time in the past.

For the six cities reviewed, we matched (both electronically and manually)
the property addresses in SAMS to the addresses in the Bureau of the
Census’ street address file to identify corresponding census tracts. When
an exact match for the zip code and street address did not exist, we
manually selected the closest reasonable match. When no reasonable
match existed or multiple choices were possible, we excluded the
property from our analysis. For the six cities combined, we were able to
match about 90 percent (1,232 of 1,374) of the properties in HUD’s
inventory with a census tract and data on valid time in inventory.




Page 33                                         GAO/RCED-98-2 Homeownership
Appendix VI

Major Contributors to This Report


                        Karen Bracey
Resources,              Barbara Johnson
Community, and          DuEwa Kamara
Economic                Robert Procaccini
                        Chuck Wilson
Development
Division, Washington,
D.C.
                        Glenn G. Davis
Chicago Field Office    Dorothy Waniak
                        Steven Westley




(385659)                Page 34             GAO/RCED-98-2 Homeownership
Ordering Information

The first copy of each GAO report and testimony is free.
Additional copies are $2 each. Orders should be sent to the
following address, accompanied by a check or money order
made out to the Superintendent of Documents, when
necessary. VISA and MasterCard credit cards are accepted, also.
Orders for 100 or more copies to be mailed to a single address
are discounted 25 percent.

Orders by mail:

U.S. General Accounting Office
P.O. Box 37050
Washington, DC 20013

or visit:

Room 1100
700 4th St. NW (corner of 4th and G Sts. NW)
U.S. General Accounting Office
Washington, DC

Orders may also be placed by calling (202) 512-6000
or by using fax number (202) 512-6061, or TDD (202) 512-2537.

Each day, GAO issues a list of newly available reports and
testimony. To receive facsimile copies of the daily list or any
list from the past 30 days, please call (202) 512-6000 using a
touchtone phone. A recorded menu will provide information on
how to obtain these lists.

For information on how to access GAO reports on the INTERNET,
send an e-mail message with "info" in the body to:

info@www.gao.gov

or visit GAO’s World Wide Web Home Page at:

http://www.gao.gov




PRINTED ON    RECYCLED PAPER
United States                       Bulk Rate
General Accounting Office      Postage & Fees Paid
Washington, D.C. 20548-0001           GAO
                                 Permit No. G100
Official Business
Penalty for Private Use $300

Address Correction Requested