oversight

Housing Finance: Budget Savings From the Sale of HUD Loans

Published by the Government Accountability Office on 1999-07-19.

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

                  United States General Accounting Office

GAO               Report to the Subcommittee on Housing
                  and Transportation, Committee on
                  Banking, Housing and Urban Affairs, U.S.
                  Senate

July 1999
                  HOUSING FINANCE

                  Budget Savings From
                  the Sale of HUD Loans




GAO/RCED-99-203
United States General Accounting Office                                                       Resources, Community, and
Washington, D.C. 20548                                                                    Economic Development Division



                                    B-281139                                                                                   Letter

                                    July 19, 1999

                                    The Honorable Wayne Allard
                                    Chairman
                                    The Honorable John F. Kerry
                                    Ranking Minority Member
                                    Subcommittee on Housing and
                                    Transportation
                                    Committee on Banking, Housing
                                    and Urban Affairs
                                    United States Senate

                                    During the 1980s, defaults on mortgages insured by the Department of
                                    Housing and Urban Development’s (HUD) Federal Housing Administration
                                    (FHA) increased, primarily because real estate markets were weak. As a
                                    result, by the early 1990s, HUD owned nearly 110,000 single-family and
                                    2,400 multifamily loans that it had insured. To make better use of its
                                    resources, HUD decided to sell these loans. Between June 1994 and
                                    September 1997, HUD held six sales of single-family loans and seven sales
                                    of multifamily loans. Through these 13 sales, HUD sold 98,640 single-family
                                    loans and 1,093 multifamily loans. According to HUD, these sales produced
                                    over $2.2 billion in budgetary savings--$830 million for single-family loans
                                    and $1.3 billion for multifamily loans. The Federal Credit Reform Act of
                                    1990 defines these budgetary savings as the difference between the net
                                    proceeds from selling the loans and the value of the loans to the federal
                                    government (the net present value of the future cash flows to HUD) if it had
                                    not sold the loans.

                                    As agreed, this report discusses the reasonableness of (1) HUD’s estimates
                                    of budgetary savings from the sale of its single-family loans and (2) the
                                    model HUD used to estimate savings from the sale of multifamily loans.1
                                    To assess the reasonableness of HUD’s estimates of savings from the sale of




                                    1
                                     HUD employed computerized models of future cash flows to estimate the value to the Department of
                                    not selling both single-family and multifamily loans.




                                    Page 1                        GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                   B-281139




                   single-family loans, we reviewed HUD’s single-family model and
                   assumptions and developed our own estimates of savings from three of the
                   six single-family loan sales. These three sales were held between October
                   1995 and September 1996. 2 To assess the reasonableness of HUD’s
                   multifamily model, we reviewed available documentation for the model
                   and the findings of an independent contractor hired by HUD to evaluate the
                   accuracy of the multifamily model. We did not prepare our own estimates
                   of savings to HUD from the sale of multifamily loans.



Results in Brief   According to HUD, it achieved $830 million in budgetary savings by selling
                   98,640 single-family loans. Nearly all of these savings, $774 million, were
                   attributable to the sale of 71,946 single-family loans that HUD acquired
                   through its mortgage assignment program and sold through five sales held
                   between October 1995 and September 1997. On the basis of our
                   independent estimates, we conclude that HUD’s estimates of budgetary
                   savings from the sale of the single-family loans we reviewed are
                   reasonable. Specifically, we independently estimated that HUD achieved
                   budgetary savings of $345 million, compared with HUD’s estimate of $259
                   million, from its sale of 38,547 loans sold in three single-family sales held
                   between October 1995 and September 1996. The $86 million difference
                   between our estimate of savings and HUD’s is not large when one considers
                   that the loans sold through these three sales represented an unpaid
                   principal balance of $1.7 billion.3

                   According to Booz-Allen & Hamilton, Inc., a large management and
                   technology consulting firm, the model HUD used to estimate savings from
                   the sale of multifamily loans produced reasonable estimates of savings.
                   However, Booz-Allen & Hamilton questioned the logic and support for
                   several key assumptions and found that the model was not thoroughly


                   2
                     Between 1994 and 1997, HUD sold 98,640 single-family loans. Of these, 71,946 were Mutual Mortgage
                   Insurance Fund loans that HUD acquired through its mortgage assignment program. Using a model we
                   developed in 1995, we were able to estimate the value to HUD of 38,547 of these single-family loans sold
                   through three single-family sales held in Oct. 1995, Mar. 1996, and Sept. 1996. Because of data
                   limitations, however, we could not use the 1995 model to independently estimate the savings from the
                   remaining 33,399 single-family loans that were sold in two subsequent sales, held in Jan. and Sept. 1997.
                   Nearly all of the remaining 26,694 single-family loans were HUD-held performing 221 (g)(4) loans
                   insured by HUD before Nov. 30, 1983 (which lenders assign to HUD in the 21st year) and General
                   Insurance Fund loans acquired through other means. Because of data limitations, we also excluded
                   from our analysis these loans—most of which were sold in the first single-family loan sale, held in June
                   1994.
                   3
                    All savings estimates are presented in 1994 dollars.




                   Page 2                           GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
             B-281139




             documented.4 In response, HUD developed a new model that addressed
             these weaknesses. Without documentation for the original model and the
             historical data used to support key assumptions, we cannot conclude
             whether the original model would produce reasonable estimates of savings.



Background   Since 1934, FHA, an agency within HUD, has insured lenders against losses
             on single- and multifamily mortgages that otherwise might not have
             qualified for conventional financing. During the 1980s, weak real estate
             markets prompted a surge of defaults on both the single-family and
             multifamily mortgages insured by FHA. By 1994, when the wave of defaults
             subsided, HUD owned almost 110,000 single-family mortgages and 2,400
             multifamily mortgages.

             This large inventory of troubled loans consumed a disproportionate
             amount of FHA staff time and prevented the agency from properly
             servicing its portfolio of insured loans and preventing further defaults.5 To
             remedy the situation, in March 1994, HUD initiated an aggressive program
             to sell FHA’s inventory of HUD-held loans. From June 1994 through
             September 1997, HUD held six sales of single-family loans and seven sales
             of multifamily loans.6 Through these sales, HUD sold 98,640 single-family
             loans and 1,093 multifamily loans. As of September 30, 1998, HUD owned
             about 12,000 single-family and 1,100 multifamily loans.

             About 72,000 of the single-family loans sold by HUD were acquired through
             the Department’s now defunct mortgage assignment program. This
             program allowed lenders, under certain conditions, to assign an
             FHA-insured mortgage to HUD after the borrower defaulted, making HUD
             the owner of the loan. For borrowers accepted into the program, HUD paid
             the mortgage debt, took assignment of the loan from the lender, and
             developed a new repayment plan (forbearance agreement) for the
             borrower, under which mortgage payments could be reduced or suspended


             4
             See Model Documentation and Users Guide for the “Value-to-HUD” Model for Multifamily Notes,
             Booz-Allen & Hamilton, Inc. (Apr. 24, 1998).
             5
              Servicing and managing the loans assigned to FHA requires a considerable amount of staff resources.
             This staff-intensive effort takes resources away from monitoring HUD-insured mortgages, thereby
             increasing the risk of default for these mortgages.
             6
             HUD also negotiated the transfer of 26 subsidized multifamily loans to the Missouri Housing
             Development Commission and reassigned 85 unsubsidized multifamily loans to Fannie Mae (the
             Federal National Mortgage Association).




             Page 3                         GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
B-281139




for up to 36 months. HUD, acting as the lender, collected monthly
mortgage payments from the borrower while allowing the borrower to
keep the home. By taking assignment of loans rather than having lenders
foreclose on them, HUD can, at times, avoid foreclosure losses, help
borrowers retain their homes, and provide borrowers with opportunities to
avoid foreclosure. However, as we reported in October 1995, even with the
forbearance provided by HUD to these financially strapped borrowers,
over half would eventually lose their homes through foreclosure.7
Moreover, the mortgage assignment program did not reduce HUD’s
foreclosure losses; rather, the program’s losses exceeded those that would
have occurred if the loans had gone immediately to foreclosure.

Usually, HUD acquires ownership of a multifamily mortgage when the
borrower defaults on an FHA-insured mortgage. After such a default, the
private lender may submit a claim for the insured amount to FHA and
assign the mortgage to FHA.8

The Federal Credit Reform Act of 1990 was enacted to (1) require that the
federal budget more accurately measure the government’s subsidy costs for
federal direct loans and loan guarantees and (2) permit better cost
comparisons between credit and noncredit programs. The credit subsidy
cost is the estimated net cost, in present value terms, to the government of
direct or guaranteed loans over the entire period that the loans are
outstanding. Under the act, loan asset sales are treated as modifications
that change the cost of the loan or guarantee to the federal government.
For federal budgetary purposes, the subsidy cost of a sold loan is the
difference between the value to the federal government of the loan if the
loan were not sold (the estimated net present value of the future cash flows
from the loan) and the net proceeds from the loan sale. If the estimated
value of keeping the loan is greater than the estimated proceeds from a
sale, a positive credit subsidy estimate would result, and the sale cannot go
forward unless budget authority has been provided to cover the additional
subsidy cost to the government. However, if the estimated value to the
government of retaining ownership of the loan is less than the estimated
proceeds from a sale, the government would be achieving a savings from
the sale and no additional budget authority is necessary.



7
See Homeownership: Mixed Results and High Costs Raise Concerns About HUD’s Mortgage
Assignment Program (GAO/RCED-96-2, Oct. 18, 1995).
8
 HUD may also acquire a mortgage through a “nondefault” assignment of 221(g)(4) loans after 20 years.




Page 4                         GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                        B-281139




                        HUD used separate multifamily and single-family models to estimate the
                        net present values of future cash flows to HUD--the value of the loans to
                        HUD if they had remained in its portfolio. According to HUD, the seven
                        multifamily and six single-family loan sales that the Department held from
                        June 1994 through September 1997 resulted in $2.2 billion in budgetary
                        savings.9



HUD’s Estimates of      Over the past 5 years, HUD sold 98,640 single-family loans with an unpaid
                        principal balance of $4.4 billion. According to HUD, it saved $830 million
Savings From the        by selling these loans. Nearly all of these savings, $774 million, were
Single-Family Loan      attributable to the sale of 71,946 single-family loans that HUD acquired
                        through its mortgage assignment program and sold through five sales held
Sales We Reviewed Are   from October 1995 through September 1997. Where we were able to
Reasonable              independently estimate savings, we believe that HUD’s estimate of
                        budgetary savings from the sale of assigned single-family loans is
                        reasonable.10

                        HUD based its estimate of savings on the results of a model that estimated
                        the future revenues and costs to HUD of holding these loans rather than
                        selling them. In 1995, using a model we had developed, we reported the
                        cost to HUD of holding assigned loans in its portfolio. To determine the
                        reasonableness of HUD’s estimate of budgetary savings from the sale of its
                        single-family loans, we used this earlier model to estimate the value to HUD
                        of the 38,547 assigned single-family loans that HUD sold through sales held
                        in October 1995, March 1996, and September 1996. (See app. I for a more
                        complete discussion of our model.) The 38,547 loans we analyzed
                        represent all but 141 of the mortgage assignment program loans sold
                        through theses three sales. However, because of data limitations, we could
                        not provide savings estimates for the 33,258 assigned loans that HUD sold
                        in two sales held in January and September 1997.



                        9
                         Budgetary savings result because the loan purchaser places a higher value on these loans than their
                        value to HUD. According to the information contained in the Budget of the United States Government,
                        Fiscal Year 1997, the private sector may place a higher value on these loans than would HUD because
                        the net cash flows to the private sector may be larger. For example, it is generally felt that the private
                        sector is more efficient than HUD at servicing loans and collecting loan payments.
                        10
                         The reasonableness of HUD’s and our estimates depends on the data provided to the models. While
                        we did not independently verify the accuracy of HUD’s data, we did perform extensive testing of the
                        data elements used in producing our savings estimates (see page 10 for a more detailed discussion of
                        the data reliability tests we performed).




                        Page 5                          GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
B-281139




For the three sales for which we had data, our estimate of the combined
savings to HUD is $345 million, a figure that is $86 million higher than
HUD’s estimate of $259 million. The $86 million difference between our
estimates of savings and HUD’s is not large when one considers that HUD,
through these three sales, sold 38,688 single-family loans with an unpaid
principal balance of $1.7 billion. These estimates of savings represent the
difference between the estimated value to HUD of the loans if HUD had not
sold them and the proceeds from the sale. Since loan sale proceeds remain
constant, it is the difference in HUD’s and our estimates of the value of
these loans to HUD if HUD had not sold them that causes the $86 million
difference in our estimates. As table 1 shows, HUD’s estimate of the total
value of these loans and our estimate are within 7 percent of each other—a
difference of $86 million. Given the degree of uncertainty surrounding any
estimates of cash flows that may occur 20 or more years into the future,
this 7-percent difference is not substantial.



Table 1: Estimated Value to HUD and Savings From Sales of 38,547 Single-Family
Loans
1994 dollars in millions
                                                                                                Differencea
                                                                                              (HUD’s minus
                                    HUD’s estimatea           GAO’s estimatea                       GAO’s)
Value to HUD
    Oct. 1995 sale                                $329                      $317                       $12
    Mar. 1996 sale                                 397                        389                        8
    Sept. 1996 sale                                460                        394                       66
Total value to HUD                              $1,186                    $1,100                       $86
Total sale proceeds                              1,445                      1,445                        0
Total savings                                     $259                      $345                       $86
Note: HUD’s and GAO’s estimates exclude the sale of General Insurance Fund and 221(g)(4) loans, of
which there were 2,309 in the Oct. 1995 sale, 1,891 in the Mar. 1996 sale, and 3,260 in the Sept. 1996
sale. Together, these loans represented less than 7 percent of the savings to HUD from loans sold
through these three sales. The total estimated savings equals the total proceeds from the sales minus
the total estimated value to HUD.
a
    All of the figures in this column are estimates except the one for total sale proceeds.


According to HUD, for the single-family loan sales held in January and
September 1997, it used the same model and assumptions to estimate
savings as it used for the sale immediately preceding these sales. As noted
above, data limitations precluded our providing independent estimates of
savings to HUD from these two sales. Over 90 percent of the single-family



Page 6                              GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                        B-281139




                        loans sold during these last two sales were also mortgage assignment
                        program loans. During our review, we did not identify any issues that
                        would raise questions about the reasonableness of HUD’s estimates of
                        savings from these two sales.

                        The similarities between our estimates and HUD’s are not surprising, given
                        that both models, while differing in some respects, base their estimates of
                        savings on the same key variables and assumptions. For example, both
                        models estimate the value to HUD of not selling loans by projecting the
                        future cash HUD would receive from recoveries on foreclosed homes,
                        principal and interest payments, and homeowners’ prepayments of loan
                        obligations. All of these cash flows must be forecasted 10 or more years
                        into the future. In addition, the assumptions driving both models are based
                        on historical data.

                        There are some differences, however, between HUD’s model and ours. For
                        example, while our model estimates future principal and interest payments
                        together as one payment, HUD’s model treats them as separate cash flows.
                        In addition, our model estimates recoveries on foreclosures by applying an
                        assumed recovery rate to HUD’s acquisition cost on foreclosed loans, while
                        HUD’s model applies an assumed recovery rate on foreclosures to the
                        unpaid principal balance on foreclosed loans.



The Multifamily Model   In December 1996, HUD hired an independent contractor to review the
                        conceptual basis and mechanics of the model the Department used for
HUD Used to Estimate    estimating savings from the sale of multifamily loans and to determine
Savings Had             whether the model’s estimates of savings were reasonable. In April 1998,
                        the contractor reported that the model produced reasonable results.
Weaknesses That HUD     However, according to the contractor, the model had certain weaknesses
Has Now Addressed       that could be corrected through a thorough, rigorous, and documented
                        analysis of historical data on the performance of HUD-held multifamily
                        loans.11 In response, HUD developed an improved model that addressed
                        the weaknesses the contractor identified in the original model.

                        The contractor also questioned the logic and support for several of the
                        model’s key assumptions. Specifically, the contractor could not identify the
                        source of information supporting the assumptions about loan prepayment,


                        11
                           See Model Documentation and User’s Guide for the “Value-to-HUD” Model for Multifamily Notes,
                        Booz-Allen & Hamilton, Inc. (Apr. 24, 1998).




                        Page 7                        GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                  B-281139




                  third-party and property-sale proceeds, and holding costs. The contractor
                  also questioned the model’s (1) use of one collection rate for all types of
                  loans that did not vary over time and was based on data for only 12 months
                  of collections, (2) failure to address arrearages or the application of
                  collections to arrearages and current billings, and (3) failure to reduce the
                  scheduled principal and unpaid principal balances to zero over time. The
                  contractor also found that the model was not thoroughly documented and
                  that people who were knowledgeable about the model’s assumptions and
                  data sources were no longer available.

                  According to one HUD official, the model was not well documented, and
                  over time, HUD employees who were involved in the sale of multifamily
                  loans and could describe the model moved on to other jobs. In addition,
                  the company hired to develop the multifamily model was no longer in
                  existence. As a result, HUD officials were unable to provide us with
                  detailed information on the model’s structure or on the basis for all of the
                  assumptions used. Given the limited amount of information available, we
                  were unable to provide an independent analysis of the strengths and
                  weaknesses of HUD’s multifamily model. Furthermore, given the
                  weaknesses identified by the contractor, we cannot conclude whether the
                  model produced reasonable estimates of savings.

                  In its final report, the contractor concluded that none of the identified
                  weaknesses were serious enough to compromise the reasonableness of the
                  results produced by the multifamily model and that HUD’s estimates of
                  savings were reliable in all material aspects. According to the contractor,
                  this conclusion was based on professional judgment after a thorough
                  examination of the model’s documentation and structure and available data
                  on HUD-held multifamily loans.12 As part of its April 1998 report, the
                  contractor also provided documentation for a new multifamily model it
                  developed. According to the contractor, the new model addresses all of its
                  concerns about the original model.



Agency Comments   We provided copies of a draft of this report to HUD for review and
                  comment. Officials of the Department, including the FHA Comptroller,
                  generally agreed with the report’s findings.


                  12
                     One way to assess the reasonableness of the estimates produced by HUD’s original multifamily model
                  would be to use the new model to reestimate the savings from past loan sales. According to HUD,
                  resource limitations have precluded their using the new model for this purpose.




                  Page 8                         GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
              B-281139




Scope and     In conducting this review, we focused on the two models that HUD used to
              estimate budgetary savings from the sale between March 1995 and
Methodology   September 1997 of 71,946 single-family loans and 887 multifamily loans.13

              To determine the reasonableness of HUD’s estimates of budgetary savings
              from the sale of its single-family loans, we interviewed HUD officials
              responsible for overseeing the single-family loan sales program. In
              addition, we developed our own estimates of savings from HUD’s sale of
              38,547 assigned single-family loans through three sales held between
              October 1995 and September 1996. To develop our estimates, we used a
              model that we developed in 1995 to estimate HUD’s costs of holding the
              single-family mortgages it had acquired through its mortgage assignment
              program. This program allowed lenders, under certain conditions, to
              assign FHA-insured mortgages to HUD in cases of default, making HUD the
              owner of the loans.

              Our analysis was based on data we received from HUD during our initial
              review of the mortgage assignment program in 1995. We worked closely
              with HUD officials and discussed the interpretation of HUD’s data. While
              we did not independently verify the accuracy of the data, we did perform
              internal checks of the data elements used to determine (1) the extent to
              which the data fields were coded and (2) the reasonableness of the values
              contained in the data fields. We checked the mean, median, mode,
              skewness, and high and low values for each of the variables used. We also
              compared several variables from each data set to see if they were recorded
              the same way within each data set. We developed an appropriate
              methodology for using the data to estimate losses, costs, and revenues to
              FHA for the time period covered.

              To assess the strengths and weaknesses of the model HUD used to estimate
              savings from the sale of multifamily loans, we reviewed available
              information on the model and HUD’s assumptions and interviewed HUD
              officials responsible for overseeing the multifamily loan sales program. In
              addition, we interviewed the independent contractor hired by HUD to
              assess the model and reviewed the contractor’s final report. This report

              13
                Since June 1994, HUD has also sold 16,029 single-family and 206 multifamily 221(g)(4) loans that were
              assigned to HUD in their 21st year by lenders, as previously allowed under the terms of FHA’s mortgage
              insurance. These 221 (g)(4) loans were excluded from our review. Also excluded from the scope of this
              review were 26 multifamily loans transferred to the Missouri Housing Development Commission, 85
              multifamily loans reassigned to Fannie Mae, 10,085 single-family loans that were originally insured
              through FHA’s General Insurance Fund, and 580 other single-family loans.




              Page 9                         GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
B-281139




also provides documentation for HUD’s improved multifamily model. We
did not prepare our own estimates of savings to HUD from the sale of
multifamily loans or assess HUD’s new multifamily model. We performed
our work from September 1998 through May 1999 in accordance with
generally accepted government auditing standards.

As agreed with your offices, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 15 days after the
date of this letter. At that time, we will provide copies to Senator Connie
Mack, Chairman, Subcommittee on Economic Policy, Senate Committee on
Banking, Housing, and Urban Affairs; Andrew M. Cuomo, Secretary of
Housing and Urban Development; and other interested parties. Copies will
also be made available to others on request.

If you or your staff have any questions about this report, please contact me
at (202) 512-6520. Key contributors to this assignment were Chuck Bausell,
DuEwa Kamara, Matt Sciré, and Pat Valentine.




Stanley J. Czerwinski
Associate Director, Housing and Community
 Development Issues




Page 10                 GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
Appendix I

GAO’s Model Used to Estimate Single-Family
Loan Sales Savings                                                                                                             Appenx
                                                                                                                                    Idi




                       Using a cash flow model,1 we estimated the net present value of future
                       revenues and costs for 38,547 loans that the Department of Housing and
                       Urban Development (HUD) sold in three separate sales held between October
                       1995 and September 1996. This net present value is an estimate of the value
                       of these loans to HUD if it had not sold them. We then compared this estimate
                       of net present value with the actual proceeds of the loan sales to determine
                       the savings, if any, from these sales.2 The results of these calculations,
                       presented in table I.1, indicate an overall savings of $345 million, compared
                       with HUD’s lower estimate of $259 million.



How We Estimated the   We started by estimating the value of all 71,458 loans assigned to HUD
                       between 1977 and 1994 as if they had never been sold and had performed at a
Value of the           rate determined by historical data. 3 We then removed from this portfolio
Single-Family Loans    the loans sold in October 1995 and reestimated the remaining value of the
                       loans. The resulting reduction in the estimated value of the loans in HUD’s
Sold                   portfolio equaled our estimate of the value of the loans sold in this sale. We
                       used the same methodology to estimate the values of the loans sold in the
                       next two sales.

                       Table I.1 includes the number and unpaid principal balance of the loans sold
                       in five sales HUD held between October 1995 and September 1997. Sale
                       values and savings estimates are also provided. As noted below, we used our
                       model to estimate savings from three of the loan sales, but not from the two
                       sales that took place in January and September 1997. 4 Information on the
                       number of loans modeled/sold and the unpaid principal balance are also
                       reported in table I.1, along with the difference in HUD’s and our estimates of
                       savings.




                       1
                        In 1995, we used this model to analyze the cost to HUD of holding single-family loans acquired through
                       its mortgage assignment program. See Homeownership: Mixed Results and High Costs Raise Concerns
                       About HUD’s Mortgage Assignment Program (GAO/RCED-96-2, Oct. 18, 1995).
                       2
                       These cash flow estimates are in 1994 present values, calculated by using discount rates specified by
                       HUD for each loan sale.
                       3
                        We first matched these loans with loans included in our 1995 model and then grouped them according
                       to the date they were assigned to the program.
                       4
                        Table I.1 does not include information on the loans sold in HUD’s first (June 1994) single-family loan
                       sale because these loans were not mortgage assignment program loans. We therefore could not use our
                       model to analyze them.




                       Page 11                         GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                      Appendix I
                      GAO’s Model Used to Estimate Single-Family
                      Loan Sales Savings




                      Table I.1: Information on HUD’s Savings Estimates for Single-Family Mutual
                      Mortgage Insurance Fund Loans Sold From October 1995 Through September 1997
                      1994 dollars in millions
                                                                          Oct.    Mar.     Sept.     Jan.      Sept.
                      Data item                                          1995     1996     1996      1997      1997
                      Unpaid principal balance                        $447.1 $616.8       $653.7 $859.0       $853.2


                      Number of loans HUD sold                        10,672 14,309       13,707 17,182       16,076
                      Number of loans GAO modeled                     10,831a 14,159      13,557          0       0


                      HUD’s estimated value to HUD                    $328.5 $397.3       $460.1 $600.8       $567.6
                                                                                                          b        b
                      GAO’s estimated value to HUD                    $317.0 $389.0       $393.7
                      Sale proceeds                                   $335.6 $515.5       $593.6 $790.0       $767.9


                      HUD’s estimated savingsc                           $7.1 $118.2      $133.6 $189.2       $196.3
                                                    c                                                     b        b
                      GAO’s estimated savings                           $18.6 $126.5      $200.0


                      Difference between HUD’s and GAO’s
                                                                                                          b        b
                      estimates                                         $11.5      $8.3    $66.3
                      Note: HUD also sold 15,212 single-family 221(g) (4) loans in June 1994.
                      a
                       For the Oct. 1995 sale, we identified 159 more loans than HUD reported as single-family mortgage
                      assignment program loans sold.
                      b
                          We did not estimate the value to HUD for the Jan. 1997 and Sept. 1997 loan sales.
                      c
                       For each sale, the estimated savings equals the sale proceeds minus the estimated value to HUD.




Variables and         Major cash flow variables, whose future values are estimated, include
                      revenues from sales of foreclosed properties, early payoffs of loans,
Assumptions We Used   payments made by mortgagors (borrowers), and advances paid on
in Our Model          properties by HUD.

                      We assumed that the value of the loans sold was a function of the
                      foreclosure and payoff rates. Other factors that affected costs included (1)
                      the ratio of the unpaid principal balance to the original loan amount, (2)
                      receivables due on the original loan amount, and (3) the ratio of advances
                      to the original loan amount, as well as the policy year of the loans. In
                      addition, we assumed that the Federal Housing Administration (FHA)
                      would continue to receive partial and delayed payments for some assigned
                      mortgages and that both foreclosure and prepayment behavior would



                      Page 12                           GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                             Appendix I
                             GAO’s Model Used to Estimate Single-Family
                             Loan Sales Savings




                             remain the same in the future as in the past, after each successive loan sale.
                             Our analysis does not take into account the possibility that the loans
                             assigned from fiscal year 1989 through fiscal year 1994 may differ from
                             earlier loans in ways that could affect their prepayment and foreclosure
                             probabilities beyond 6 years from the date of assignment. In addition,
                             neither our analysis nor HUD’s takes into account the difference in the
                             distribution of loan performance for loan pools remaining after each sale.

                             Given these assumptions, we projected future cash flows from foreclosures,
                             prepayments, and surviving loans. Because of inadequate historical data, it
                             was not possible to estimate foreclosure and prepayment probabilities
                             incorporating economic indicators such as unemployment rates,
                             payment-to-income ratios, current interest rates, and house price
                             appreciation rates.5

                             Additional detail on forecasting each of the major revenue variables follows.


Foreclosure and Prepayment   On the basis of our analysis of foreclosure and prepayment data, we
Rates                        estimated an ultimate foreclosure probability of 52 percent. We also
                             estimated conditional probabilities using data for the 6-year period ending
                             September 30, 1994. These probabilities were for loans entering HUD’s
                             mortgage assignment program during a 17-year period (fiscal years 1977–94)
                             and represented loan years 1 through 17. We assumed that the conditional
                             foreclosure and prepayment rates for loan years 18 through 30 would be the
                             same as for loan year 17.


Foreclosure Revenues         To estimate foreclosure revenues, we obtained an average recovery rate for
                             loans foreclosed and sold from data on Mutual Mortgage Insurance Fund
                             loans foreclosed during fiscal years 1983-94. Recovery rates ranged
                             between 43 and 67 percent of acquisition costs each year, averaging 59
                             percent. The average recovery rate of 59 percent was applied to the
                             acquisition costs of all foreclosed loans. Specifically, the average
                             acquisition cost for each year times the recovery rate for each foreclosed
                             loan results in the expected total foreclosure revenue.




                             5
                             FHA’s database records historical foreclosure and prepayment activity from fiscal year 1989 onward.
                             Data on previous years’ terminations were purged from the database.




                             Page 13                        GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                      Appendix I
                      GAO’s Model Used to Estimate Single-Family
                      Loan Sales Savings




Prepayment Revenues   Prepayment revenues are based on data for all assigned loans, the number
                      of loans paid off and forecasted to be paid off, and the unpaid principal
                      balance at the time of payoff. In estimating the unpaid principal balance,
                      we used the ratio of the unpaid balance to the original loan amount for
                      each year. Using the average loan amount, year in the program, and
                      number of expected prepayments, we estimated prepayment revenues for
                      each year.

                      For years 19 through 30, we assumed that the ratio of the unpaid balance to
                      the original loan amount would continue to decrease at an accelerated rate.
                      To determine the unpaid balance for years 19 through 30, a simple
                      regression was applied to the ratio of the unpaid balance to the original
                      loan amount for years 1 through 18, in which each year’s ratio is dependent
                      on the previous year’s ratio. The resulting parameters were used to
                      estimate the ratio of the unpaid balance to the loan amount for years 19
                      through 30.


Payment Revenues      Loan payment revenue estimates are based on the percentage of loans in
                      five loan status categories--current, current with forbearance, delinquent
                      with forbearance, delinquent with no forbearance, and pending
                      foreclosure. For each year’s book of business, we analyzed the ratio of the
                      unpaid balance to the loan amounts and the actual payments made for each
                      loan category. We also accounted for advances owed and original loan
                      amounts.

                      We forecasted loan payment revenues using the estimated number of loans
                      remaining in the program and the actual and scheduled payments made for
                      each loan category. Actual loan payments averaged about 34 percent of
                      scheduled payments.6 It was assumed that the assigned loans would have
                      the same distribution over the loan categories that they did in fiscal year
                      1994 but that their length of time in the program would vary. The ratio of
                      actual to scheduled payments was also assumed to vary by the length of
                      time in the program. As loans age, payment ratios rise, indicating that a
                      higher percentage of scheduled payments are being made for older loans.




                      6
                      The percentage of loans making full payments increases with time in the program. Some borrowers
                      with less than 3 years in the program make no mortgage payments as part of their suspended payment
                      mortgage forbearance agreement.




                      Page 14                       GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
                       Appendix I
                       GAO’s Model Used to Estimate Single-Family
                       Loan Sales Savings




Databases We Used in   We used three of HUD’s computerized databases:

Our Analysis           • the F-60 database of current and historical information on all mortgage
                         loans that HUD services under the mortgage assignment program;
                       • the A-43 database of historical information on mortgages insured under
                         the Mutual Mortgage Insurance Fund before assignment; and,
                       • the Single-Family Accounting and Management System database of
                         properties held and eventually sold by HUD following foreclosure.

                       These databases provided information on initial characteristics of each
                       loan, such as the year the loan was assigned, the initial unpaid principal
                       balance, any delinquent amounts, and the interest rate and term of the loan.
                       We categorized the loans as foreclosed, prepaid, or active as of the end of
                       fiscal year 1994.




(385754)               Page 15                    GAO/RCED-99-203 Estimates of Savings From HUD’s Loan Sales
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
United States                       Bulk Rate
General Accounting Office      Postage & Fees Paid
Washington, D.C. 20548-0001           GAO
                                 Permit No. GI00
Official Business
Penalty for Private Use $300

Address Correction Requested