Small Business Administration: Credit Subsidy Estimates for the Sections
7(a) and 504 Business Loan Programs (Testimony, 07/16/97,
GAO/T-RCED-97-197).

GAO discussed the Small Business Administration's (SBA) estimates of
credit subsidies for the agency's guaranteed business loan and certified
development company programs-more commonly called the "7(a)"and "504"
programs, respectively, focusing on: (1) how SBA calculates the
estimates of credit subsidies for the 7(a) and 504 programs; (2) what
factors accounted for the increases in the estimated cost of the loans
to be guaranteed by these programs in fiscal year FY 1997; and (3) what
additional changes, if any, SBA made during the 1998 budget process when
estimating the costs of its loans.

GAO noted that: (1) SBA bases its estimate of the credit subsidy for
each program on projections of cash flows--that is, the amounts of cash
that SBA expects to take in and pay out during each year that the loans
are outstanding; (2) cash outflows occur when borrowers default on their
loans and SBA pays claims filed by lenders; (3) cash inflows occur when
the collateral for defaulted loans is liquidated and when borrowers and
lenders pay mandatory fees to SBA; (4) these cash flow projections are
based largely on the historical performance of the programs' loans;
however, SBA adjusts the projections to reflect the estimated influence
of changes in the programs' provisions and other factors; (5) the cash
flows are discounted to determine their net present value using a
computer model established and maintained by the Office of Management
and Budget (OMB); (6) the factors contributing to the increases in the
estimated credit subsidy rates for FY 1997 differed somewhat between the
two programs; (7) for the 7(a) program, SBA projected fewer recoveries
(the amounts realized when defaulted loans are liquidated) and less
revenue from fees than it had assumed in previous years; (8) these
changes, which accounted for about 60 percent of the increase, resulted
primarily from SBA's use of a new database on historical loan
performance that expanded and improved on existing data, according to
SBA and OMB officials; (9) in addition, an error in applying SBA's cash
flow projections to OMB's discounting model caused the estimated credit
subsidy rate for the 7(a) program to be even larger; (10) this error
accounts for the remainder of the increase; (11) on the basis of its
appropriated budget authority of $198.5 million, SBA announced that it
could guarantee an additional $2.47 billion in 7(a) loans because the
credit subsidy rate estimate should have been lower; (12) for the 504
program, SBA projected both more claims and fewer recoveries for
defaulted loans than it had assumed previously; (13) for the FY 1998
budget request, SBA further revised its projections of cash inflows and
outflows for the two programs; (14) for 7(a) loans, SBA decreased its
estimate of expected fee revenue, on the basis of data on loans that are
paid off prior to maturity, or prepaid; (15) because SBA would have to
cover more of the cost of loan defaults if revenues from fees are less
than originally expected, this change had an upward effect on the
subsidy rate estimate for the 7(a) program; and (16) for the 504
program, SBA increased the estimated prepayments and slightly reduced
expected claim payments and recoveries.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  T-RCED-97-197
     TITLE:  Small Business Administration: Credit Subsidy Estimates for 
             the Sections 7(a) and 504 Business Loan Programs
      DATE:  07/16/97
   SUBJECT:  Small business loans
             Subsidies
             Funds management
             Presidential budgets
             Loan defaults
             Fees
             Data bases
             Errors
             Credit
IDENTIFIER:  SBA 504 Program
             SBA Liquidation Improvement Project
             SBA 7(a) Loan Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Before the Committee on Small Business, House of Representatives

For Release
on Delivery
Expected at
10 a.m.  EDT
Wednesday
July 16, 1997

SMALL BUSINESS ADMINISTRATION -
CREDIT SUBSIDY ESTIMATES FOR THE
SECTIONS 7(A) AND 504 BUSINESS
LOAN PROGRAMS

Statement of Judy A.  England-Joseph,
Director, Housing and Community
Development Issues,
Resources, Community, and Economic
Development Division

GAO/T-RCED-97-197

GAO/RCED-97-197T


(385675)


Abbreviations
=============================================================== ABBREV

  SBA -
  OMB -
  CDC -

============================================================ Chapter 0

Mr.  Chairman and Members of the Committee: 

We are pleased to be here today to discuss our review of the Small
Business Administration's (SBA) estimates of credit subsidies for the
agency's guaranteed business loan and certified development company
programs--more commonly called the "7(a)" and "504" programs,
respectively.  As you know, the credit subsidy for these programs is
the estimated net cost (excluding administrative costs) to SBA in
today's dollars of guaranteeing these loans over the entire time
period in which the loans are outstanding, which can range up to
about 25 years.  The Federal Credit Reform Act of 1990 requires that
SBA estimate these costs for loans guaranteed after fiscal year 1991
so that they can be included in the federal budget in the year in
which the loan commitments are made.  These estimates, and any
subsequent re-estimates, identify for the Congress the amount of
appropriations needed to cover the government's expected costs over
the lives of the loans.  In the President's budget for fiscal year
1997, SBA estimated that the costs of 7(a) and 504 program loans to
be made in fiscal year 1997 would be significantly higher than the
costs of loans made in fiscal year 1996, despite legislated program
changes designed to keep costs down. 

Our statement today is based on our review of SBA's estimates of
credit subsidies for the 7(a) and 504 programs that you requested in
April of this year as well as our ongoing work on such estimates in
the major credit agencies throughout the federal government.  As
requested, this statement focuses on three questions:  (1) How does
SBA calculate the estimates of credit subsidies for the 7(a) and 504
programs?  (2) What factors accounted for the increases in the
estimated costs of the loans to be guaranteed by these programs in
fiscal year 1997?  (3) What additional changes, if any, did SBA make
during the 1998 budget process when estimating the costs of its
loans?  As agreed with your office, we did not assess the reliability
of the automated data on individual loans that SBA uses to generate
its credit subsidy estimates. 

In summary, we found the following: 

  -- SBA bases its estimate of the credit subsidy for each program on
     projections of cash flows--that is, the amounts of cash that SBA
     expects to take in and pay out during each year that the loans
     are outstanding.  Cash outflows occur when borrowers default on
     their loans and SBA pays claims filed by lenders.  Cash inflows
     occur when the collateral for defaulted loans is liquidated and
     when borrowers and lenders pay mandatory fees to SBA.  These
     cash flow projections are based largely on the historical
     performance of the programs' loans; however, SBA adjusts the
     projections to reflect the estimated influence of changes in the
     programs' provisions and other factors.  The cash flows are
     discounted to determine their net present value using a computer
     model established and maintained by the Office of Management and
     Budget (OMB).\1

  -- The factors contributing to the increases in the estimated
     credit subsidy rates for fiscal years 1997 differed somewhat
     between the two programs.  For the 7(a) program, SBA projected
     fewer recoveries (the amounts realized when defaulted loans are
     liquidated) and less revenue from fees than it had assumed in
     previous years.  These changes, which accounted for about 60
     percent of the increase, resulted primarily from SBA's use of a
     new database on historical loan performance that expanded and
     improved on existing data, according to SBA and OMB officials. 
     In addition, an error in applying SBA's cash flow projections to
     OMB's discounting model caused the estimated credit subsidy rate
     for the 7(a) program to be even larger.  This error accounts for
     the remainder of the increase.  On the basis of its appropriated
     budget authority of $198.5 million, SBA announced that it could
     guarantee an additional $2.47 billion in 7(a) loans because the
     credit subsidy rate estimate should have been lower.  For the
     504 program, SBA projected both more claims and fewer recoveries
     for defaulted loans than it had assumed previously.  As in the
     7(a) program, these changes--which account for all of the
     increase--resulted from SBA's use of a more extensive historical
     database. 

  -- For the fiscal year 1998 budget request, SBA further revised its
     projections of cash inflows and outflows for the two programs. 
     For 7(a) loans, SBA decreased its estimate of expected fee
     revenue, on the basis of data on loans that are paid off prior
     to maturity, or "prepaid." Because SBA would have to cover more
     of the cost of loan defaults if revenues from fees are less than
     originally expected, this change had an upward effect on the
     subsidy rate estimate for the 7(a) program.  For the 504
     program, SBA increased the estimated prepayments and slightly
     reduced expected claim payments and recoveries.  With these
     changes and a fee increase, SBA estimated a credit subsidy rate
     of zero for new 504 loans. 


--------------------
\1 Present value is the value today of a stream of payments in the
future.  In calculating the present value of loan subsidy costs,
prevailing interest rates for U.S.  Treasury securities provide the
basis for converting future amounts into current dollar equivalents. 


   BACKGROUND
---------------------------------------------------------- Chapter 0:1

The 7(a) and 504 programs are two of SBA's primary programs for
enhancing small businesses' access to credit.  Under the 7(a)
program, SBA guarantees up to 80 percent of the amounts of loans made
by private lenders to small businesses that are unable to obtain
financing under reasonable terms and conditions through normal
business channels.  In fiscal year 1996, SBA guaranteed about $7.3
billion in small business loans through the 7(a) program.  SBA's
guarantee transfers the major risk of default from the lenders to
SBA.  When a loan guaranteed through the 7(a) program defaults, SBA
pays the lender's claim by purchasing the guaranteed portion of the
unpaid balance of the loan.\2

SBA then recovers as much of the claim amount as it can through the
liquidation of the small business' collateral.\3 In return for its
guarantee, SBA collects fees from 7(a) lenders.  SBA collects three
types of fees for 7(a) loans:  (1) an upfront fee which lenders may
collect from borrowers, (2) an annual fee charged to lenders, and (3)
a fee charged to lenders for certain 7(a) loans they sell through the
secondary market.\4 According to SBA, about half of the 7(a) loans
made each year are sold through the secondary market. 

Like the 7(a) program, the 504 program uses SBA guarantees to assist
small businesses.  Under this program, SBA provides its guarantee
through certified development companies (CDC)--private, nonprofit
corporations.  CDCs may sell debentures that are fully guaranteed by
SBA to private investors and lend the proceeds to qualified small
businesses for acquiring real estate, machinery, and equipment and
for building or improving facilities.\5 These debentures funded about
$2.4 billion in loans during fiscal year 1996.  If a small business
defaults on its loan from the CDC, SBA pays a claim filed through a
central servicing agent, thereby ensuring that investors receive full
and timely payments.  SBA recovers some of the claim amount through
the liquidation of the small business' collateral.  However, projects
financed with 504 loans typically involve a supplemental mortgage
from a private-sector lender who has a first lien on the collateral. 
In return for guaranteeing 504 program debentures, SBA charges both
upfront and annual fees. 

To help lower the federal cost of these programs, in October 1995 the
Congress enacted the Small Business Lending Enhancement Act of 1995
(P.L.  104-36).  Among other things, the act (1) lowered the maximum
guaranteed portion of most 7(a) loans from 90 to 80 percent, (2)
created the annual fee charged to lenders and increased the upfront
fee for 7(a) loans, and (3) created the annual fee for the 504
program.  The latter fee was increased on September 30, 1996 by the
Small Business Programs Improvement Act (P.L.  104-208). 

We obtained information on how SBA's credit subsidy estimates are
prepared primarily by interviewing officials from SBA and OMB.  We
identified the factors accounting for the changes in the estimates
primarily by reviewing the automated cash flow spreadsheets
underlying these estimates.  However, we did not verify the accuracy
of the loan performance data used in the spreadsheets.  We also
obtained information through interviews with officials from the
Congressional Budget Office and the two major industry associations
for the 7(a) and 504 programs. 


--------------------
\2 The lender's claim can include up to 120 days of accrued interest. 

\3 Collateral may be liquidated by either SBA or lenders. 
Liquidation is the process of converting assets to cash. 

\4 Lenders may sell the guaranteed portion of 7(a) loans to
investors.  These investors are referred to as the "secondary market"
for the loans. 

\5 A debenture is an investment typically backed by the integrity of
the borrower (but in this case, by SBA) and documented by an
agreement called an indenture. 


   HOW CREDIT SUBSIDY ESTIMATES
   ARE CURRENTLY CALCULATED
---------------------------------------------------------- Chapter 0:2

For the 7(a) and 504 programs, SBA bases the estimates of credit
subsidies for the loans guaranteed during a given year on projections
of the cash inflows and outflows likely to result from the loans.\6
To project future cash flows, SBA first reviews available data on the
cash flows it has experienced with loans made in the past.  SBA then
estimates cash flows for future "outyears"--the later years of the
loans' lives for which there are not yet any historical data--and
adjusts for recent program changes that are expected to affect the
cash flows.  As part of the annual budget process, SBA prepares these
cash flow projections for loans to be made during the budget year. 
An automated discounting model developed and maintained by OMB and
distributed to federal agencies converts the cash flows prepared for
all federal credit programs into the credit subsidy rate estimates
and budget authority requests that are presented in the President's
budget.  The estimated credit subsidy rates for SBA's 7(a) and 504
programs are expressed as a percentage of the total amount of loans
or debentures to be disbursed. 


--------------------
\6 Neither program includes an interest rate subsidy. 


      PERFORMANCE OF LOANS MADE IN
      THE PAST IS A KEY FACTOR IN
      PROJECTING FUTURE CLAIMS AND
      RECOVERIES
-------------------------------------------------------- Chapter 0:2.1

For each program, SBA bases the projected claim payments and
recoveries for defaulted loans largely on the performance of the
loans made in the past.  According to SBA and OMB officials, SBA uses
data on the performance of the loans made through the 7(a) and 504
programs since 1986. 

SBA analyzes data on the portion of the loans made in each past year
that has defaulted and the amounts it has recovered since the loans
were made.  SBA uses these data to compute an average claim rate and
recovery rate for each year in a loan's life.  For example, the
average claim rate for 504 loans in their first year is the average
of the actual rates observed in the first years of past loans; the
average claim rate for 504 loans in their second year is the average
of the actual rates observed in the second years of past loans; and
so on.  Unless SBA expects recent program changes to affect future
defaults, these cash flows of average historical data are SBA's
estimates of claims and recoveries for new loans.  In this way, SBA's
method of projecting claims and recoveries for new loans gives equal
weight to the performance of loans made in the past, regardless of
differences in the volume of loans made during each year or their
ages. 

In addition, for 7(a) loans, SBA considers differentials in loan
performance by the type of the loan.  Specifically, SBA calculates
average claims and recoveries, on the basis of actual historical
data, for the various categories of (1) the maturity of the loan, (2)
the size of the loan, (3) the type of lender,\7 and (4) the
percentage of the loan guaranteed by SBA.  SBA then uses these data
to calculate an overall average claim rate and recovery rate for the
new loans.  SBA does not conduct an analysis of loan type for the 504
program because the program is smaller and because there is less
variation in types of loans than in the 7(a) program. 

SBA uses the actual historical averages as a basis for the expected
claims and recoveries during the first 11 years of the new loans'
lives.\8 For the later years of the new loans' lives--the
"outyears"--SBA projects claims and recoveries through several
methods.  For 7(a) loans, SBA projects claims through the 16th year
after the loans are made and projects recoveries through the 18th
year.  For the 12th and 13th years of the loans' lives, SBA projects
claims and recoveries by assuming a continuation of the trend
demonstrated by the historical data for the first 11 years and by
considering data on the performance during the 12th and 13th years of
loans made prior to 1986.  For later years, SBA roughly estimates
additional claims and recoveries by considering the trend in the data
for the first 13 years and by assuming that there will be at least
some claims and recoveries during each of these later years.  SBA
does not project claims and recoveries beyond the 18th year because
of the uncertainty of these estimates and because the figures are
likely to be very small, according to SBA officials.  Although the
maximum maturity for 7(a) loans is 25 years, the average maturity is
about 12 years, according to OMB officials. 

For 504 loans, SBA projects claims through the 16th year and
recoveries through the 22nd year.  For years after the 11th, SBA
roughly estimates additional claims and recoveries by considering the
trend in the historical data for the first 11 years and by assuming
that there will be at least some claims and recoveries during each of
these later years.  Although the maximum maturity for a 504 loan is
20 years, the average maturity is 19 years, according to OMB
officials.  SBA estimates cash flows for outyears so that the
estimated credit subsidy rate represents the expected federal cost of
all of the loans made during a year over their entire lives, not just
the years for which historical data are available. 


--------------------
\7 SBA's analysis by type of lender considers whether loans are made
by preferred lenders, certified lenders, or other lenders.  Preferred
and certified lenders receive full or partial delegation of authority
to approve loans. 

\8 SBA may also add outyear estimates to the historical data for each
past year of loans before the data are averaged.  This procedure
ensures that there are 11 years of performance data linked directly
to actual data for each past year of loans, even for loans made less
than 11 years ago.  SBA adds these outyear estimates by assuming the
same trends in claims and recoveries demonstrated by data for older
loans. 


      CURRENT RATES AND HISTORICAL
      DATA ARE USED TO PROJECT
      FUTURE FEE REVENUES
-------------------------------------------------------- Chapter 0:2.2

For both the 7(a) and 504 programs, SBA bases its projections of
revenues from fees primarily on the fee rates in effect at the time
that program loans are made and on the estimates, based on historical
data, of annual outstanding loan balances. 

For the 7(a) program, SBA projects revenues for each of the three
types of fees, as follows: 

  -- The upfront fee is based on the original dollar amount of the
     guaranteed portion of the loan.  Because the fees vary according
     to the size of the guaranteed portion, SBA calculates an average
     rate for upfront fees by estimating the percentage of the new
     loans in each category of loan size.\9 SBA's projection of the
     cash flow from upfront fees is also based on the collection of
     the upfront fee over the first two years of the loans' lives,
     rather than the entire amount in the first year, because not all
     loans are disbursed during the first year. 

  -- Annual fee revenue is based on estimated outstanding loan
     balances at the end of each year of the life of the loans.  SBA
     bases its estimates of outstanding balances on historical data
     on the portion of balances amortized each year and then
     subtracts the amount that it expects to pay in claims and the
     amount of loans that it estimates will be prepaid.  SBA does not
     maintain data showing when 7(a) loans are prepaid.  Instead, the
     agency estimates prepayment rates and trends on the basis of the
     historical prepayment data that it obtains for those 7(a) loans
     that are sold through the secondary market.\10 In this way, the
     actual data on prepayments and claims for prior year loans helps
     SBA estimate how much the outstanding principal balance on new
     loans will decline over the life of the loans and thus how much
     it is likely to collect in revenues from the annual fee. 

  -- Similarly, SBA estimates the amount of secondary market fee
     revenues it will collect by looking at historical data on loans
     sold through the secondary market. 

For the 504 program, projecting the cash flows from fees is slightly
less complicated because the upfront fee does not vary with the size
of the loan and there is no secondary market fee.  SBA's projection
of upfront fees is based on the rate in effect when the loans are
made and the collection of the fee during the first few years of the
loans' lives because not all loans are disbursed during the first
year.  Like the 7(a) program, annual fee revenues are based on the
estimates of outstanding loan balances at the end of each year during
the life of the new loans.  SBA bases its estimates of outstanding
balances on historical data on the portion of balances amortized each
year and then subtracts the amount of loans it expects to default and
the amount of loans that it estimates will be prepaid.  SBA maintains
its own data on prepayments in the 504 program because of the
program's prepayment penalty. 


--------------------
\9 This average also reflects the small number of 7(a) loans with
maturities of less than one year.  These loans are charged a fee rate
that does not vary with the size of the guaranteed portion.  SBA
estimates that about one percent of all loans made each year have
maturities of less than one year. 

\10 According to SBA, the guaranteed portion of about half of the
loans made through the 7(a) program each year are sold through the
secondary market.  We did not determine whether these loans are
representative of all 7(a) loans made each year. 


      SBA ADJUSTS CASH FLOWS TO
      REFLECT CHANGES IN THE
      PROGRAMS
-------------------------------------------------------- Chapter 0:2.3

Before converting the projected cash flows into an estimated credit
subsidy rate, SBA may adjust them to reflect any recent program
changes that are expected to reduce or increase future claim payments
or recoveries.  For example, SBA adjusted its projected recoveries
for loans made through the 504 program in 1998 to reflect the impact
of the agency's Liquidation Improvement Project--an effort designed
to increase the portion of SBA's claim payments for defaulted loans
that it recovers through the liquidation of collateral. 

Officials from SBA, OMB, and the two primary industry associations
for the 7(a) and 504 programs have suggested several alternative
methods SBA could use to estimate future claims and recoveries.  For
example, SBA could develop an econometric model which uses regression
analyses to predict future claims and recoveries based on factors
such as forecasted interest rates, economic growth, lender type, and
loan size.  According to a government-wide task force, econometric
modeling is the best way to use historical data to estimate credit
subsidies.\11 Similarly, we have used econometric models to forecast
future loan performance in programs operated by the Department of
Veterans Affairs and the Department of Housing and Urban Development
due to their superior predictive capabilities.\12 While we believe
that, over time, such analysis of loan performance data may yield
better estimates, we have not reviewed the merits of any particular
approach for SBA's programs. 


--------------------
\11 See Issue Paper 96-CR-7, May 1, 1996 of the Government-Wide
Audited Financial Statements Task Force. 

\12 See Homeownership:  Appropriations Made to Finance VA's Housing
Program May be Overestimated, GAO/RCED-93-173, Sept.  1993 and
Mortgage Financing:  FHA Has Achieved Its Home Mortgage Capital
Reserve Target, GAO/RCED-96-50, Apr.  1996. 


      OMB'S MODEL GENERATES
      ESTIMATES OF THE CREDIT
      SUBSIDY RATE
-------------------------------------------------------- Chapter 0:2.4

To generate an estimate of the credit subsidy rate, SBA's cash flow
projections for claims, recoveries, and fees assuming a certain
amount of business are run through OMB's automated discounting model. 
This model produces an estimate of the credit subsidy rate
representing the net present value of the federal costs expected per
dollar loaned through the program.  This estimate is used to
determine the amount of appropriations necessary to cover the
long-term costs of a given amount of loans. 

During the preparation of each year's budget submission, OMB
examiners typically work with SBA to refine its cash flow projections
and subsidy estimates.  OMB may question the basis for SBA's
projected cash flows and/or request an alternative set of cash flows
based on different assumptions.  For example, during the past few
years, SBA has expressed interest in limiting the historical data
used as a basis for projected cash flows to loans made since 1989. 
According to SBA officials, recently guaranteed loans are better
predictors of future loan performance because they reflect the
changes made to strengthen SBA's programs during the 1990s.  On the
other hand, OMB officials stated that the performance of loans in the
1980s should be considered because these years include an economic
downturn, which could occur during the life of new loans.  For this
reason, the older loans have been retained in the averages SBA uses
to estimate the subsidy costs of new loans.  According to both OMB
and SBA officials, there may be a number of iterations leading up to
the final estimate of the credit subsidy rate and the associated
request for budget authority that are shown in the President's
budget. 

In addition to estimating the credit subsidy rate for SBA-guaranteed
loans, the OMB model uses the cash flows prepared by SBA to report
three components of the estimate:  (1) net defaults (claim payments
minus recoveries), (2) fees received, and (3) any other cash flows
associated with the program.  By splitting the estimate into
components, the model indicates how each of the cash flows affects
the overall credit subsidy rate estimate. 

In addition to requiring agencies to estimate the credit subsidies
associated with new government-backed loans, credit reform
requirements provide for subsequent re-estimates and modifications of
credit subsidy estimates.  To calculate re-estimates, SBA uses a
process very similar to the one described above to project the cash
flows associated with existing loans.  Re-estimates are calculated
after the end of the fiscal year in which the loans are guaranteed. 
However, the estimated credit subsidy rate may be modified during
that year to reflect enacted legislation and certain administrative
actions that alter the subsidy cost.  A modification does not include
government actions permitted within the terms of existing contracts
or through other existing authorities.  Furthermore, credit subsidy
estimates may not be revised mid-year due only to changes in
"forecast technical" assumptions, such as expected claims. 


   WHY THE ESTIMATED CREDIT
   SUBSIDY RATES INCREASED IN 1997
---------------------------------------------------------- Chapter 0:3

According to the President's budget for fiscal year 1997, the
estimated credit subsidy rate for the 7(a) program increased from
1.06 percent for loans made in 1996 to 2.68 percent for loans made in
1997.  For the 504 program, the estimated rate increased from zero to
6.85 percent, assuming no future program changes.\13 The estimates
for the 1997 loans were higher largely because SBA modified the cash
flow projections that it had used for the fiscal year 1996 and
earlier estimates.  According to SBA and OMB officials, SBA modified
the cash flows to incorporate the results of an intensive analysis of
data on the performance of loans made in the past.  In addition, the
fiscal year 1997 estimate for the 7(a) program was affected by a
spreadsheet error unnoticed by either SBA or OMB; without this error,
the estimated credit subsidy rate in SBA's budget submission would
have been about 2.03 percent.  Appendix III summarizes the factors
responsible for these changes in SBA's estimates. 


--------------------
\13 These are the "current services" credit subsidy estimates,
according to the 1997 budget.  The current services estimate assumes
no future changes to the programs. 


      SBA DEVELOPED A NEW LOAN
      PERFORMANCE DATABASE IN 1995
-------------------------------------------------------- Chapter 0:3.1

As part of its planned efforts to meet credit reform requirements,
SBA developed a loan performance database in 1995.  The objectives of
this effort were to (1) provide access to the data on historical loan
performance on a cohort basis\14 , (2) update the subsidy estimates
on the basis of the historical data, (3) allow SBA to analyze loan
performance on the basis of multiple-risk indicators, and (4)
annually validate past subsidy estimates.  SBA officials used the
database for the first time while preparing the estimates of credit
subsidy rates for the 1997 budget. 

All estimates of credit subsidy rates for loans guaranteed prior to
1997 had been based on the limited loan performance data resulting
from a review of the loan portfolio undertaken in 1991.  According to
SBA and OMB officials, this 1991 review had several weaknesses.  Most
importantly, it did not utilize transaction-based data, but rather
relied on the interpretation of changing loan balances over time to
estimate loan performance.  Furthermore, it was based on data that
were not easily identified by cohort (i.e.  by a particular year of
loan guarantee commitments).  In addition, the results of the 1991
review were not updated to reflect the actual performance of loans in
fiscal years 1992 through 1995.  According to SBA officials, the new
loan performance database--which is updated annually to include
recent data--corrected these problems.  We did not assess the quality
of the data in SBA's new loan performance database. 


--------------------
\14 In the case of SBA's 7(a) and 504 programs, a "cohort" of loans
consists of all of the loan guarantees committed through each program
during a given fiscal year. 


      REVISED CASH FLOWS AND
      SPREADSHEET ERROR
      CONTRIBUTED TO CHANGE IN
      7(A) PROGRAM ESTIMATE
-------------------------------------------------------- Chapter 0:3.2

In 1995, when SBA prepared its 1997 budget request, the loan
performance database confirmed earlier estimates that, during the
life of each year's loans, SBA makes claim payments for defaulted
loans equal to about 13 percent of the original loan
disbursements.\15 However, the database also indicated that SBA
recovers less of the claim amount through the liquidation of loan
collateral--about 50 percent rather than the 56 percent indicated by
the 1991 review.  In addition, the analysis of the database indicated
that the timing of defaults was different than estimated in the 1991
review; specifically, a greater percentage of the defaults would
occur earlier.  Incorporating this change slightly increased the
estimates of credit subsidy rates because the earlier a cash outflow
occurs for SBA, the higher the present value of the cost to SBA. 
These changes to the estimates of claims and recoveries accounted for
about 45 percent of the increase in the fiscal year 1997 credit
subsidy estimate. 

A second factor, accounting for about 15 percent of the increase, was
SBA's lowered forecast of revenues from fees.  When preparing the
fiscal year 1996 credit subsidy estimate, SBA had not accounted for
the loss in annual fee revenue that occurs when borrowers prepay
their loans.  (Because annual fees are charged on outstanding loan
balances, and prepayments cause those balances to be lower than they
would otherwise be, prepayments effectively reduce SBA's revenues
from fees.) For the fiscal year 1997 estimate, SBA added an estimate
of prepayments.  Furthermore, SBA reduced its estimate of the amount
of revenues it would receive through the secondary-market fee, based
on lower-than-expected revenue from the first year of experience with
the new fee. 

Finally, an error in applying SBA's cash flow projections to OMB's
discounting model accounted for about 40 percent of the increase in
the credit subsidy estimate in 1997.  This error caused the estimated
credit subsidy for the 7(a) program to be expressed as a percentage
of the guaranteed portion of the loans, rather than of their total
face amount.  SBA projected that it would guarantee on average about
76 percent of the fiscal year 1997 loan cohort.  Because critical
cells in SBA's cash flow spreadsheet were based on the number of
dollars guaranteed instead of the number of dollars disbursed (that
is, the total face amount of the loans), SBA's estimated credit
subsidy rate was higher by about 32 percent (1 divided by 0.76, the
average guaranteed portion). 

This error went unnoticed by both SBA and OMB officials responsible
for reviewing the 7(a) credit subsidy rate estimate.  If SBA or OMB
officials had compared the component data generated by OMB's
discounting model for the erroneous 1997 estimate with the components
of the 1996 estimate, they would have seen an unexplainable increase
in the fee revenue component (there was no increase in the fee rates
charged).  According to standards developed by the Federal Accounting
Standards Advisory Board, subsidy estimate component data should be
used to monitor and make decisions about the federal government's
credit programs.\16

On the basis of its appropriated budget authority of $198.5
million,\17 SBA announced that it could guarantee an additional $2.47
billion in 7(a) loans because the credit subsidy rate estimate should
have been lower.  We are currently reviewing the changes in estimates
over time for selected programs at five agencies as part of an
assessment of the implementation of the Credit Reform Act.  We are
also preparing a report on OMB's automated discounting model.  We
expect to report on these efforts later this year. 


--------------------
\15 Although we refer to claims throughout this statement as a
percentage of the total dollar amount of the loans disbursed, SBA
typically defines the claim rate as a percentage of the
SBA-guaranteed portion of disbursements. 

\16 See Statement of Federal Financial Accounting Standards Number 2,
Accounting for Direct Loans and Loan Guarantees, August 23, 1993. 
These standards were developed by the Federal Accounting Standards
Advisory Board which is composed of representatives of the Department
of the Treasury, OMB, GAO, the Congressional Budget Office, several
other federal agencies, and from the private sector. 

\17 This figure includes an appropriation of $158 million and
carryover budget authority of $40 million. 


      REVISED ESTIMATES OF CLAIMS
      AND RECOVERIES CAUSED HIGHER
      SUBSIDY ESTIMATE FOR 504
      PROGRAM
-------------------------------------------------------- Chapter 0:3.3

The analysis of the loan performance database of loans made through
the 504 program in prior years showed more dramatic changes than for
the 7(a) program.  According to SBA officials, the 1991 review
substantially underestimated claim rates and overestimated recovery
rates because it did not consider a substantial portion of the 504
loan transactions.  This occurred because the 1991 review
inaccurately assumed that the accounting structure for the 7(a) and
504 programs was the same.  However, unlike the 7(a) program, 504
transactions may be recorded under different identification numbers
in SBA's accounting system than the original debenture.  Because the
1991 review included only transactions recorded under the original
debenture's identification number, it erroneously omitted relevant
data on claims and recoveries.  According to SBA officials, the new
loan performance database corrects this error by recognizing all
transactions associated with the debentures made during each year. 

In the fall of 1995, SBA used the expanded historical data on the 504
program made available through the loan performance database in
preparing the fiscal year 1997 cash flow projections.  SBA forecast
that it would (1) make claim payments during the life of a year's
loans equal to about 19 percent of the face value of the loans, more
than twice the 7 percent estimate indicated by the 1991 review, and
(2) recover about 40 percent of the claim amount through the
liquidation of loan collateral, or about half of the 80 percent
estimate indicated by the 1991 review.  The estimate of the total
recovery rate in the 1997 cash flows should have been even lower (33
percent), according to OMB and SBA officials.\18 The 40-percent
estimate was based on an error in accumulating the historical data
that was corrected when SBA prepared its 1998 estimates, according to
SBA officials. 


--------------------
\18 Using a lower total recovery rate would have increased the credit
subsidy rate estimate above the 6.85 percent rate estimated by SBA. 


   ESTIMATES FOR 1998 BUDGET
   REFLECT UPDATES TO CASH FLOW
   PROJECTIONS
---------------------------------------------------------- Chapter 0:4

Since the 1997 estimates were prepared, SBA has made additional
revisions to its cash flow projections.  For the 7(a) program, these
changes resulted in an estimated subsidy rate of 2.32 percent,
assuming no future program changes, for loans to be made in 1998 and
a re-estimate of the rate for loans made in 1996 of 2.4 percent.\19
In preparing these estimates, SBA did not repeat the spreadsheet
error that had inflated the 1997 estimate.  Nevertheless, the
estimates prepared during the 1998 budget process are higher than the
corrected budget request estimate (2.03 percent) for the 1997 loans
because SBA reduced the expected fee revenues from these loans. 
Specifically, SBA increased its estimate of prepayments--even further
than during preparation of the 1997 estimates. 

SBA increased its estimate of prepayments because it obtained for the
first time historical data maintained by Bloomberg Financial Services
on the percentage of loans each year that were prepaid.  These data
are based exclusively on 7(a) loans sold through the secondary
market.  When preparing the 1997 estimates, SBA assumed that up to
about 2.5 percent of the amount of the loans outstanding each year
would be prepaid.\20

However, the historical data obtained from Bloomberg Financial
Services showed that in some years of their lives as many as 10
percent of the amount of the loans could be expected to be prepaid. 

In addition to changes in expected revenues from fees, SBA made small
changes to its claim payment and recovery forecasts.  Specifically,
SBA (1) decreased its expected claim payments over the life of the
loans from about 13 percent of the loans' face value to about 12
percent, on the basis of additional historical data on loan
performance in fiscal year 1996, and (2) increased the total recovery
rate from about 50 percent of expected claim payments to about 51
percent over the life of the loans, on the basis of recent successes
in the Liquidation Improvement Project. 

For the 504 program, SBA (1) increased estimated prepayments; (2)
lowered its expected claim payments for defaulted loans from about 19
percent of the loans' face value to about 16 percent over the life of
the loans; and (3) lowered expected recoveries from 40 percent to 34
percent of claims.  When calculating the 34-percent recovery rate
estimate, SBA corrected the accumulation error discussed earlier and
adjusted the estimate upward to reflect the improvements in
recoveries expected to result from the Liquidation Improvement
Project, according to SBA officials.  With these changes in expected
revenue from fees, claims, and recoveries, SBA re-estimated the
credit subsidy rate for the loans made through the 504 program in
1996 at 7.54 percent.  The estimate for the 1998 loans is zero
because the annual fees charged to borrowers and lenders were
increased dramatically after SBA prepared its budget request estimate
for the 1997 loans.\21

With recoveries, these fees are expected to cover all of the costs
incurred by SBA for defaulted loans. 


--------------------
\19 The previous estimate of the credit subsidy rate for loans made
through the 7(a) program in 1996 was 1.06 percent. 

\20 Although we refer to prepayments as a percentage of the total
dollar amount of the loans disbursed, SBA typically defines
prepayment rates as a percentage of the SBA-guaranteed portion of
disbursements. 

\21 This increase in the fee rate also prompted a lowering of the
credit subsidy rate estimate for 504 loans made in 1997 from 6.85
percent (budget request) to zero (budget execution). 


-------------------------------------------------------- Chapter 0:4.1

Mr.  Chairman, this concludes my prepared remarks.  We will be
pleased to respond to any questions that you or other Members of the
Committee might have. 


HOW CREDIT SUBSIDY RATE ESTIMATES
ARE PREPARED
=========================================================== Appendix I



   (See figure in printed
   edition.)


PRIMARY REASONS FOR CHANGES IN
CREDIT SUBSIDY RATE ESTIMATES
========================================================== Appendix II



   (See figure in printed
   edition.)


CHANGES IN CASH FLOW PROJECTIONS
========================================================= Appendix III



                               Table 3.1
                
                Changes in Projections Underlying SBA's
                  Estimates of the 7(a) Credit Subsidy

                  1996              1997              1998
                  (percent)         (percent)         (percent)
----------------  ----------------  ----------------  ----------------
Fees

Prepayments\a     0                 Up to 2.5 in      Up to 10 in peak
                                    peak year         year

Secondary-        0.14              0.02              0.03
market fees\b

Claims\c          13                13                12

Recoveries\d      56                50                51

Other                               Spreadsheet
                                    error inflated
                                    credit subsidy
                                    estimate by 32
                                    percent.

Credit subsidy    1.06\f            2.68 (2.03)\g     2.32
estimate\e
----------------------------------------------------------------------
\a Prepayments are shown as an annual percentage of the loans
outstanding in the first 10 years.  (SBA typically shows prepayments
differently--as a percentage of SBA's share of outstanding loans.)

\b Secondary market fees are shown as a percentage of the original
disbursements. 

\c Claims are shown as a percentage of the original disbursements. 
(SBA typically shows claims differently--as a percentage of SBA's
share of disbursements.)

\d Recoveries are shown as a percentage of claims. 

\e Estimates shown are the 1996 budget execution estimate and the
1997 and 1998 budget request (current services) estimates. 

\f Re-estimated at 2.40 percent in the 1998 budget. 

\g Without the spreadsheet error, the 1997 budget request estimate
would have been about 2.03 percent. 

Source:  GAO's analysis of SBA's and OMB's data. 



                               Table 3.2
                
                Changes in Projections Underlying SBA's
                  Estimates of the 504 Credit Subsidy

                  1996              1997              1998
                  (percent)         (percent)         (percent)
----------------  ----------------  ----------------  ----------------
Fees

Prepayments\a     0                 0                 Up to 4 in peak
                                                      year

Annual fee        0.125             0.125             1.084
rate\b

Claims\c          7                 19                16

Recoveries\d      80                40                34

Credit subsidy    0.00\f            6.85              0.00
estimate\e
----------------------------------------------------------------------
\a Prepayments are shown as an annual percentage of the loans
outstanding in the first 10 years.  (SBA typically shows prepayments
differently--as a percentage of SBA's share of outstanding loans.)

\b The annual fee rate is shown as a percentage of the loans
outstanding annually. 

\c Claims are shown as a percentage of the original disbursements. 
(SBA typically shows claims differently--as a percentage of SBA's
share of disbursements.)

\d Recoveries are shown as a percentage of claims. 

\e Estimates shown are the 1996 budget execution estimate and the
1997 and 1998 budget request (current services) estimates.  Although
the 1996 and 1997 estimates are based on all loans made through the
504 program, the 1998 estimate shown here excludes loans made through
the Defense Loan and Technical Assistance program. 

\f Re-estimated at 7.54 percent in the 1998 budget. 

Source:  GAO's analysis of SBA's and OMB's data. 


*** End of document. ***