SBA Disaster Loan Program: Accounting Anomalies Resolved but	 
Additional Steps Would Improve Long-Term Reliability of Cost	 
Estimates (14-APR-05, GAO-05-409).				 
                                                                 
In response to a January 2003 GAO report that identified	 
significant anomalies in the Small Business Administration's	 
(SBA) disaster loan accounts and raised serious concerns about	 
its ability to account for loan sales and estimate program costs,
SBA conducted an extensive analysis to identify causes of the	 
anomalies and implemented a number of corrective actions. In	 
light of SBA's actions, GAO undertook a follow-up review to (1)  
describe the nature of the deficiencies SBA identified, (2)	 
determine whether its corrective actions resolved the		 
deficiencies, and (3) assess whether its procedures provide a	 
reasonable basis for future credit estimates.			 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-409 					        
    ACCNO:   A21717						        
  TITLE:     SBA Disaster Loan Program: Accounting Anomalies Resolved 
but Additional Steps Would Improve Long-Term Reliability of Cost 
Estimates							 
     DATE:   04/14/2005 
  SUBJECT:   Accountability					 
	     Accounting 					 
	     Accounting errors					 
	     Accounting procedures				 
	     Data integrity					 
	     Financial records					 
	     Interest						 
	     Internal controls					 
	     Loan accounting systems				 
	     Overpayments					 
	     Program evaluation 				 
	     Corrective action					 
	     Cost estimates					 
	     Policies and procedures				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-05-409

                 United States Government Accountability Office

                     GAO Report to Congressional Requesters

April 2005

SBA DISASTER LOAN PROGRAM

Accounting Anomalies Resolved but Additional Steps Would Improve Long-Term
                         Reliability of Cost Estimates

                                       a

GAO-05-409

[IMG]

April 2005

SBA DISASTER LOAN PROGRAM

Accounting Anomalies Resolved but Additional Steps Would Improve Long-Term
Reliability of Cost Estimates

  What GAO Found

SBA took prompt action with a comprehensive review of its financial
records and systems to identify the deficiencies related to accounting for
its disaster loans and loan sale program. SBA's review found (1) the cash
flow model used to estimate the cost of the disaster loan program was
unreliable and underestimated the cost, (2) the model used to determine
whether sales were beneficial had errors and incorrectly indicated that
loans were sold at gains, (3) incorrect loan values used to calculate the
results of loan sales led to inaccurate reporting in SBA's financial
statements, and (4) incomplete tools provided by OMB to calculate interest
payments on borrowings from Treasury resulted in excess payments to
Treasury and an insufficient balance in SBA's financing account and
subsidy allowance.

To resolve these deficiencies, SBA implemented a number of corrective
actions during fiscal years 2003 and 2004. To address the first three, SBA
developed a new cash flow model to estimate the costs and loan values for
the disaster loan program. This improved the agency's ability to prepare
more reliable cost estimates and determine the gain or loss on prior loan
sales. To address the fourth deficiency, SBA analyzed its interest
payments to Treasury and found that it had overpaid by about $134 million.
SBA included this amount in its reestimates for the disaster loan program
to correct prior interest payments and also implemented a different
approach to update or "reestimate" its cost estimates, which will adjust
its transactions with Treasury going forward. However, until OMB updates
its tools for computing these interest payments, other credit agencies may
also be overor underpaying interest to Treasury.

Further, SBA improved its policies and procedures to help ensure that
future loan program cost estimates will be reasonable. For example, SBA
implemented new standard operating procedures for calculating reestimates
and prepared documentation to support the rationale and basis for key
aspects of the cash flow model. However, because of the complexities
associated with estimating loan program costs, additional actions by SBA
would help improve the long-term reliability of cost estimates. These
include (1) further documentation of the model and disaster data to
readily provide for knowledge transfer between staff and contractors to
help ensure proper maintenance, updating, and running of the model; (2)
periodic assessments of the model's ability to predict loan performance;
and (3) additional procedures to ensure the disaster data used in the
model are tested to verify and document that they are reliable. In
addition, there may be opportunities to improve the model with additional
variables, such as financial strength of borrowers, as well as revisions
to simplify the estimation process that warrant further consideration by
SBA.

                 United States Government Accountability Office

Contents

Letter                                                                   1 
                                  Results In Brief                          2 
                                     Background                             4 
             SBA Identified Major Deficiencies in Models and Methodologies 
                   Used to Account for the Disaster Loan Program            9 
               SBA Has Taken Corrective Actions to Resolve Identified      
                                    Deficiencies                           13 
             New Policies and Procedures Will Help Ensure Future Estimates 
                                                                       Are 
               Reasonable, but Additional Procedures and Documentation Are 
                                       Needed                              18 
                                    Conclusions                            22 
                        Recommendations for Executive Action               22 
                                  Agency Comments                          23 

Appendixes

                          Appendix I: Appendix II: Appendix III: Appendix IV:

Objectives, Scope, and Methodology 25

Comments from the Small Business Administration 28

Comments from the Office of Management and Budget 30

GAO Contact and Staff Acknowledgments 31 GAO Contact 31 Acknowledgments 31

Figures Figure 1: Calculation of Subsidy Cost for Direct and Guaranteed    
                                             Loans                          6
           Figure 2: Program and Financing Account Transactions for Direct 
                                             Loans                          7 
           Figure 3:  Disaster Subsidy Allowance as a Percent of the Loan  
                      Balance Outstanding, Fiscal Years 1996 through 2004  16 

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

A

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

April 14, 2005

The Honorable Olympia J. Snowe
Chair, Committee on Small Business and Entrepreneurship
United States Senate

The Honorable Todd R. Platts
Chairman
The Honorable Edolphus Towns
Ranking Minority Member
Subcommittee on Government Management, Finance, and Accountability
Committee on Government Reform
House of Representatives

The Honorable Marsha Blackburn
House of Representatives

In 1999, the Small Business Administration (SBA) began a loan asset sales
program at the direction of the Office of Management and Budget (OMB) to
reduce the amount of loans the agency owned and serviced. A primary
objective of the sales was to maximize proceeds, with a goal for each sale
to be financially beneficial to the government. The loans eligible for
sale
were disaster assistance and other direct loans and defaulted business
loan
guarantees. Between fiscal years 1999 and 2003, SBA conducted seven
sales, divesting itself of about 166,000 loans with an outstanding balance
of
about $5.7 billion. Approximately 86 percent of the amount sold was
disaster assistance loans.

Our January 2003 report1 on SBA's first five loan sales identified
significant
anomalies in SBA's disaster loan accounts and raised serious concerns
about its ability to properly account for its loan sales and to estimate
the
costs associated with its remaining disaster loan portfolio for budget and
accounting purposes.2 In response to our findings and recommendations,
SBA and its contractors (hereinafter referred to as SBA) conducted an
extensive analysis of its accounting and budgeting for its loan sales and
disaster loan program and implemented a number of corrective actions

1 Our review also included determining whether the loan sales generated
operational benefits for SBA and identifying how borrowers and lenders
reacted to the sales.

2 GAO, Small Business Administration: Accounting Anomalies and Limited
Operational Data Make Results of Loan Sales Uncertain, GAO-03-87
(Washington, D.C.: January 2003).

during fiscal years 2003 and 2004, including the development of a new cash
flow model to estimate the cost of the disaster loan program for budget
and financial reporting purposes.

In light of these events, we were asked to conduct a follow-up review to
assess SBA's corrective actions. Specifically, our objectives were to (1)
describe the nature of the deficiencies SBA's analysis identified that
contributed to the disaster loan accounting anomalies, (2) identify
corrective actions taken by SBA and assess whether these actions resolved
the identified deficiencies, and (3) determine whether SBA's new cash flow
model and procedures for the disaster loan program provide a reasonable
basis for future credit subsidy estimates.

To address these objectives, we reviewed SBA's analysis of its accounting
and budgeting for its disaster and loan sales programs and analyzed SBA's
corrective actions, including the new cash flow model to estimate costs
for the disaster loan program. We also assessed the sufficiency of its
policies and procedures to estimate program costs and its corrective
actions based on applicable guidance. We provided SBA a draft of this
report and OMB a draft of applicable sections of this report for review
and comment. SBA and OMB provided written comments, which are reprinted in
appendix II and III, respectively. We performed our work in accordance
with generally accepted government auditing standards in Washington, D.C.
from April 2004 through March 2005. Our scope and methodology are
discussed in greater detail in appendix I.

Results In Brief	As a result of extensive analyses, SBA identified four
key deficiencies related to the disaster loan program accounting
anomalies. First, major flaws in the cash flow model used to estimate the
cost of the disaster loan program, including erroneous loan-term
assumptions, led to, among other things, a negative balance in the
disaster loan subsidy allowance. In effect, this meant that the program
cost more than estimated. Second, errors and inconsistencies in another
model, called the hold model, which was used to determine whether sales
were financially beneficial, caused SBA to undervalue the loans sold by
about 30 percent, according to SBA. This led SBA officials to operate on
the premise that they were selling disaster loans for gains, when in fact
the agency was selling them at a loss. Third, incorrect loan values used
to calculate the results of loan sales led to inaccurate results disclosed
in SBA's financial statements. Finally, interest rates used to determine
interest payments on borrowings from Treasury that provide financing for
the disaster loan program were inconsistent with

the interest rates used to estimate the cost of the program because of
incomplete tools provided by OMB. This resulted in SBA overpaying interest
to Treasury and an insufficient balance in SBA's financing account and
subsidy allowance.

To resolve these deficiencies, SBA implemented a number of corrective
actions during fiscal years 2003 and 2004. SBA developed a new cash flow
model to estimate the cost of the disaster loan program, which improved
the agency's ability to prepare more reliable estimates of the loan
program's cost, and to calculate appropriate values for loans sold to use
in determining the actual gain or loss on prior loan sales. This new model
can also be used in lieu of a separate hold model to calculate loan values
to use in determining whether any future loan sales would be financially
beneficial. Further, SBA analyzed its prior interest payments to Treasury
to determine the effect of using inconsistent interest rates to estimate
subsidy costs and to calculate interest payments to Treasury. These
analyses showed that SBA overpaid interest to Treasury by about $134
million. SBA included this amount in its reestimates for the disaster loan
program to correct prior interest payments and also implemented a
different approach to update or "reestimate" its cost estimates, which
will adjust SBA's transactions with Treasury to correct for the
inconsistency going forward. However, until OMB updates its tools for
computing these interest payments, other credit agencies may also be over-
or underpaying interest to Treasury.

In addition to implementing these corrective actions, SBA improved its
policies and procedures to help ensure that future loan program cost
estimates will be reasonable. For example, SBA developed and implemented
new standard operating procedures for reestimating program costs and
established an internal review and documentation process for its
reestimates. These controls represent important improvements. However, we
identified some additional actions by SBA that would help ensure the
long-term reliability of cost estimates. These include (1) further
documentation of the cash flow model and disaster data to readily provide
for knowledge transfer between staff and contractors to help ensure proper
maintenance, updating, and running of the model; (2) procedures requiring
periodic assessments of the model's ability to predict loan performance;
and (3) additional procedures to ensure the disaster data used in the
model are tested to verify and document that they are reliable. In
addition, there may be opportunities to improve the model with additional
variables, such as the financial strength of borrowers, as well as
revisions to simplify the estimation process, that warrant further
consideration by SBA.

We are making recommendations to SBA to address these issues, as well as
to OMB to address the completeness of the tools used to calculate interest
payments to Treasury. SBA acknowledged that these were appropriate
recommendations and stated that it already has work underway to address
several of them. OMB agreed with our recommendations and stated that it
would work with agencies to correct interest transactions with Treasury.

Background	SBA provides small businesses with access to credit, primarily
by guaranteeing loans through its 7(a) and 504 programs.3 SBA also makes
loans directly to businesses and individuals trying to rebuild in the
aftermath of a disaster, and it primarily services these loans directly.
Substantially all of the disaster assistance loans have below-market
interest rates and repayment terms of up to 30 years. Interest rates on
disaster loans vary, depending on the borrower's ability to obtain credit
in the private sector.

The President's fiscal year 1998 budget proposed that SBA begin selling
disaster and business loans that the agency was servicing and transition
from servicing loans directly to overseeing private-sector servicers.
Before its loan asset sales program began, SBA was servicing approximately
300,000 loans, with a principal balance of over $9 billion. About 286,000
of these loans, with a principal balance of $7 billion, were disaster
assistance loans.

SBA, as well as other credit agencies, is required to account and budget
for its credit programs in accordance with the Federal Credit Reform Act
of 19904 (FCRA). FCRA was enacted to require agencies to more accurately
measure the government's cost of federal credit programs and to permit
better cost comparisons, both among credit programs and between credit and
noncredit programs. The act gave OMB responsibility for coordinating
credit program estimates required by the act. Authoritative guidance on
preparing cost estimates for the budget and conducting loan sales is
contained in OMB Circular A-11, Preparation, Submission, and Execution

3 The 7(a) program is established under section 7(a) of the Small Business
Act (15 U.S.C. S: 636). The 504 program is established under Title V of
the Small Business Investment Act of 1958 (15 U.S.C. S: 695 et seq).

4 Federal Credit Reform Act of 1990, Pub. L. No. 101-508, title XIII, S:
13201 (a) (Nov. 5, 1990); 2 U.S.C. S:S: 661-661f.

of the Budget. The Federal Accounting Standards Advisory Board5 developed
accounting standards for credit programs. This guidance is generally found
in Statement of Federal Financial Accounting Standards No. 2, Accounting
for Direct Loans and Loan Guarantees, which became effective in fiscal
year 1994. This standard, which generally mirrors FCRA and budget
guidance, established accounting guidance for estimating the subsidy cost
of loan programs, as well as recording loans and loan sales for financial
reporting purposes.

According to FCRA, the actual and expected costs of federal credit
programs should be recognized in budgetary reporting. The accounting
standard also requires these costs to be recognized for financial
reporting. To determine the expected cost of a credit program, agencies
are required to predict or estimate the future performance of the program
on a cohort6 basis. This cost, known as the subsidy cost, is the net
present value7 of disbursements by the government minus estimated payments
to the government over the life of the loan or loan guarantee, excluding
administrative costs. Figure 1 presents the cash flows included in the
subsidy cost calculation for direct and guaranteed loans.

5 The board establishes generally accepted accounting principles for
federal entities.

6 A cohort includes those direct loans or loan guarantees of a program for
which a subsidy appropriation is provided in a given year even if the
loans are not disbursed until subsequent years.

7 Present value is the worth of the future stream of returns or costs in
terms of money paid immediately. In calculating present value, prevailing
interest rates provide the basis for converting future amounts into their
"money now" equivalents.

Figure 1: Calculation of Subsidy Cost for Direct and Guaranteed Loans

Source: GAO.

FCRA established a special budgetary accounting system to record the
budget information necessary to implement credit reform. Loans and loan
guarantees made on or after October 1, 1991-the effective date of credit
reform-use the (1) program and (2) financing accounts to handle credit
transactions.8 The program account is included in budget totals, receives
separate appropriations for the administrative and subsidy costs of a
credit program, and records the budget authority and outlays for these
costs. The program account is used to pay the associated subsidy cost to
the financing account when a direct or guaranteed loan is disbursed. The
financing account, which is nonbudgetary,9 is used to collect the subsidy
cost from the program account, borrow from Treasury to provide financing
for loan disbursements, and record the cash flows associated with direct
loans or loan guarantees over their lives, including loan disbursements,
default payments to lenders, loan repayments, interest payments,
recoveries on defaulted loans, and fee collections. Figure 2 shows the
flow of program and financing accounts transactions for a direct loan
program.

8 A liquidating account was established to handle credit transactions on a
cash basis for precredit reform loans and loan guarantees.

9 Nonbudgetary accounts may appear in the budget document for information
purposes, but are not included in the budget totals for budget authority
or budget outlays.

Figure 2: Program and Financing Account Transactions for Direct Loans

Source: GAO.

FCRA requires that the rate of interest charged by Treasury on lending to
financing accounts be the same as the final discount rate10 used to
calculate the net present value of cash flows when estimating the subsidy
cost of a credit program. The final discount rate for a cohort of loans is
determined based on interest rates prevailing during the period that the
loans are disbursed. Once the loans for a cohort are substantially
disbursed (at least 90 percent), the final discount rate for that cohort
is determined, and this rate is to be used for financing account interest
calculations. The same rate is required to be used to calculate subsidy
costs and interest on the financing account, so that the financing account
will break even over time

10 For loans made or guaranteed in fiscal year 2001 and thereafter, the
discount rate is based on interest rates on marketable zero-coupon
Treasury securities with the same maturity from the date of disbursement
as the cash flow. For loans made or guaranteed before fiscal year 2001,
the discount rate is based on a disbursement-weighted average of interest
rates for marketable Treasury securities with similar maturities as the
loans or loan guarantees.

as it uses its collections to repay its Treasury borrowing. OMB provides
tools for agencies to use to calculate interest on the financing
account.11

To estimate the cost of credit programs, agencies first estimate the
future performance of direct and guaranteed loans, using cash flow models,
when preparing their annual budgets. The data used for these budgetary
estimates are generally updated or "reestimated" annually as of the end of
the fiscal year to reflect any changes in loan performance since the
estimates were prepared, as well as any expected changes in assumptions
related to future loan performance. Increases in subsidy costs that are
recognized through reestimates are funded through permanent indefinite
budget authority.

Before SBA could proceed with a loan sale, OMB had to approve it. This
approval was based primarily on whether or not the sale was expected to be
financially beneficial to the government, meaning that the estimated
proceeds were expected to be greater than the estimated value of holding
the loans. SBA estimated the current value to the government of holding
the loans, also known as the "hold value," in accordance with OMB Circular
A-11. The hold value is the expected net cash flows from the loans,
discounted at current Treasury rates.12 This differs from the net book
value recorded on SBA's books, which is the expected net cash flows from
the loans discounted using Treasury rates in effect when the loans were
disbursed. Therefore, the hold value takes into account changes in
interest rates since the loans were disbursed, whereas the net book value
does not.

Our January 2003 report on SBA's first five loan sales highlighted
accounting anomalies related to its disaster loans and loan sales program.
Specifically, SBA incorrectly calculated the accounting losses on the loan
sales it disclosed in its financial statements and lacked reliable
financial

11 For loan guarantee programs that do not borrow from Treasury, the
financing account receives the subsidy cost from the program account and
holds these funds to serve as a reserve against future loan guarantee
defaults or other costs. FCRA requires that these funds, referred to as
uninvested funds, earn interest from Treasury at the same rate as the
discount rate used to calculate the present value when estimating the
subsidy cost.

12 The hold value of the loans selected for sale represents the estimated
value to the government of continuing to hold the loans until they are
repaid, either at or before maturity. It is designed to be a decisional
tool used to determine whether or not it is currently advantageous to sell
loans. The hold value is calculated on a present value basis, with future
payments discounted at current Treasury interest rates in order to reflect
current market conditions in the decision-making process.

data to determine the overall financial impact of the sales. Further,
because SBA did not analyze the effect of loan sales on its remaining
portfolio, its reestimates of loan program costs for the budget and
financial statements may have contained significant errors. In addition,
SBA could not explain significant declines in its subsidy allowance for
disaster loans.

In response to our findings, SBA took immediate action to begin the
process of identifying the deficiencies that contributed to the disaster
loan accounting anomalies, including unexplained significant declines in
its subsidy allowance. A team of financial experts, including contractors
and staff from the Office of the Chief Financial Officer, was assembled to
conduct detailed reviews of financial records and systems related to the
disaster and loan sales programs. Several diagnostic-type analyses were
performed, including detailed reconciliations of the subsidy allowance and
testing of alternative versions of the cash flow model used to estimate
the cost of the disaster loan program. In January 2003, SBA hired IBM
Business Consulting Services (IBM) to help determine reasons for the
abnormal balance in the disaster loan program's subsidy allowance and to
identify recommendations for correcting any deficiencies noted. IBM
assisted SBA in a detailed review of SBA's accounting and budgeting for
the disaster loan program and its loan sale procedures. IBM summarized the
results of this review in a March 2003 report. According to SBA officials,
the core diagnosis of the problems was completed by April 2003 with the
submission and analysis of the report from IBM.

SBA Identified Major Deficiencies in Models and Methodologies Used to
Account for the Disaster Loan Program

SBA and its contractors identified four key deficiencies related to the
disaster loan program accounting anomalies. These were (1) major flaws in
the cash flow model used to estimate the cost of the disaster loan
program, (2) errors and inconsistencies in the model used to determine
whether sales were beneficial, (3) incorrect loan values used to calculate
the results of loan sales, and (4) inconsistencies between the interest
rates used to estimate subsidy costs and the interest rates used to
determine interest payments to Treasury.

Cash Flow Model Used to The methodology SBA's cash flow model used to
estimate costs for the Estimate Costs for the disaster loan program
assumed that a single illustrative loan with Disaster Loan Program Was
characteristics based on overall portfolio averages could serve as a proxy

for all loans and reasonably estimate cash flows for an entire cohort
ofFlawed loans, which included sold and unsold loans. This methodology
could have

produced reasonable cash flow estimates, even considering loan sales, if
all loans had similar characteristics. However, this was not the case, and
flaws in this methodology became apparent once SBA began substantial loan
sales and the loans sold had different characteristics than the loans not
sold. For example, the sold loans tended to have longer borrower repayment
periods, or loan terms, than the loans not sold. Therefore, the sold loans
would have had subsidized interest for a longer period of time and would
have cost more. Because the single illustrative loan did not take into
consideration these differences when estimating cash flows, the model had
problems reestimating the cost of the program.

In addition to the basic flaw in the methodology, SBA incorrectly
calculated the single illustrative loan's average loan-term assumptions
used in the cash flow model. SBA estimated costs of the disaster loan
program using average loan-term assumptions of 16 years for business
disaster loans and 17 years for home disaster loans. In our January 2003
report, we raised concerns about the validity of these assumptions, as our
review of disaster loans sold indicated an average loan term of about 25
years. SBA's loanterm assumptions were based on the number of loans,
rather than the dollar value of loans disbursed, and therefore was a
straight average rather than a weighted average. However, SBA found that,
for all disaster loans, the average loan term, based on the dollar value
of loans disbursed, was 23 years. Therefore, the model, which used loan
terms of 16 and 17 years, did not consider cash flows for the full term of
the loans. Given that the borrower interest rates on these loans were
generally below market rates and less than SBA's cost of borrowing from
Treasury to finance its loans, understating the loan term also understated
the program costs. According to SBA, this caused the model to
underestimate the cost of the disaster loan program by 6 to 7 percent.

Further, during the reestimate process, SBA did not update the estimated
principal and interest with actual collection amounts, which resulted in
inaccurate data being used to calculate costs. These flaws related to the
methodology, loan-term assumptions, use of inaccurate data, and other
problems, resulted in unreliable subsidy cost estimates and reestimates
for the disaster loan program. Collectively, the problems with SBA's cash
flow model resulted in significant underestimates of the cost of the
program, which was ultimately reflected in the negative balance in the
disaster loan subsidy allowance. The negative balance would occur for
programs that are expected to be profitable, which would not be expected
for the disaster loan program, or when the allowance is overspent, meaning
that the program cost far more than estimated.

The Model Used to Determine Whether Loan Sales Were Beneficial Was Also
Flawed

When SBA sold loans, it used another model, called the hold model, to
estimate the value to the government of holding the loans scheduled for
sale until they were repaid, either at or before maturity. The hold model
considered the same possible cash flows as the cash flow model used to
estimate the cost of the program, including principal and interest
collections, prepayments, delinquencies, defaults, and recoveries.
However, the hold model differed from the cash flow model because it was
constructed using a different methodology. The hold model measured loans
individually whereas the cash flow model, as previously discussed, used a
single-loan approach. In addition, expected defaults were determined
differently, which caused the hold model to produce higher default rates
than the cash flow model. Further, the hold model used economic variables
and performance indicators and the cash flow model did not. As a result of
these differences, the two models produced different results. While the
hold model's conceptual design was superior to the cash flow model, it too
contained serious flaws that produced misleading results.

For example, the hold model erroneously used assumptions in determining
the amount of recoveries expected on defaulted loans. The recovery
assumptions were taken from the cash flow model used to estimate the cost
of the program. These assumptions were calculated based on actual
recoveries as a percentage of the value of loans disbursed and, therefore,
should have been applied based on the value of loans disbursed. The hold
model, however, erroneously applied this percentage to estimated defaulted
loan amounts, therefore calculating a far lower amount of estimated
recoveries than was appropriate, which made the value of holding the loans
seem much lower. Collectively, problems with SBA's hold model caused it to
undervalue the loans it sold by about 30 percent, according to disclosures
in SBA's fiscal year 2003 financial statements. As a result, at the time
of the sales the hold model indicated that it was financially beneficial
to sell the loans when in fact it was not.

Incorrect Loan Values Were Used to Calculate the Results of Loan Sales
Disclosed in SBA's Financial Statements

While SBA's hold model indicated at the time of the sales that SBA had
gains on loan sales, SBA concurrently disclosed losses on loan sales in
its financial statements based on yet another set of flawed calculations.
As reported by us in our January 2003 report and in the report issued by
IBM in March 2003, when SBA calculated the results of loan sales for
purposes of its financial statements, it incorrectly estimated the portion
of the subsidy allowance to allocate to each loan sold in order to
calculate the value on its books for the loans it had sold (net book
value). For example,

when calculating the net book value for the disaster loans that were sold,
SBA did not allocate a portion of the subsidy allowance for financing
costs associated with lending to borrowers at below-market interest rates.
Given that a large portion of the subsidy cost was related to providing
belowmarket borrower interest rates, this omission resulted in a
significant overestimate of the net book value of the loans sold and,
therefore, a significant overestimate of the losses SBA disclosed in its
financial statements related to the sale of its disaster loans. Even
though SBA calculated losses for the financial statements, it still
operated on the premise that loans were sold at gains when considering
changes in interest rates, which the hold model was purportedly designed
to do.

Interest Rates Used to Calculate Interest Payments to Treasury Were not
Consistent with Those Used to Estimate Subsidy Costs

The final deficiency that SBA identified in its disaster loan program
accounting related to inconsistencies in the interest rates it used to
estimate its subsidy costs versus those used to calculate its interest
payments to Treasury. A direct loan program, including the disaster loan
program, funds its lending to borrowers with the subsidy cost it receives
through appropriations and from borrowing from Treasury. Because the
borrowing is expected to be repaid with collections from borrowers,
borrowing is not a budgeted cost to the program and is accounted for in
the program's financing account. FCRA requires that the rate of interest
charged to the financing account on the agency's borrowing be the same as
the interest rate used to discount cash flows (discount rate) when
estimating the subsidy cost for a program. The equality of these rates is
fundamental to achieving the proper balance in the financing account. If
subsidy cost calculations are accurate and the proper interest rates used,
the financing account will break even over time as it uses collections
from borrowers to repay Treasury borrowings.

SBA, in coordination with OMB, found that the tools provided to agencies
to calculate interest for the financing account did not adjust the amount
of interest paid by the financing account while the loans were disbursing.
The discount rate used to estimate the subsidy cost is not final until the
loans in a cohort are substantially disbursed (at least 90 percent),
which, for the disaster program, generally may take at least 2 years. When
the loans are substantially disbursed, and the final discount rate is
fixed, the reestimate process retroactively adjusts the subsidy costs to
reflect the final discount rate.

SBA and other agencies must make annual interest payments to Treasury
while the loans are disbursing, although the final interest rate has not
yet

been determined. Thus in the early years of a cohort, before the loans are
substantially disbursed, an interim interest rate is used to calculate
interest payments. However, the tools provided by OMB to calculate
interest between the financing account and Treasury do not retroactively
adjust prior interest earnings or payments to reflect the final interest
rate. This failure to adjust prior interest payments to reflect the final
interest rate resulted in excess payments to Treasury and an insufficient
balance in SBA's financing account and subsidy allowance, since the
interest payments impact both. This omission in the tools OMB provides to
all agencies that disburse or guarantee loans13 could result in a
disconnect between the amounts required to be earned from or paid to
Treasury to make the financing account whole, and the actual amounts
earned or paid. Consequently, agencies' financing account balances and
subsidy allowance may be over-or understated.

SBA Has Taken Corrective Actions to Resolve Identified Deficiencies

Following its analysis and identification of deficiencies, SBA developed a
new cash flow model to estimate the cost of the disaster loan program.
This improved the reliability of the disaster program cost estimates and
corrected the abnormal balance in the subsidy allowance for the disaster
loan program. In addition, SBA analyzed its prior interest payments to
determine the effect of using inconsistent interest rates to calculate its
estimated subsidy costs and interest payments to Treasury, and implemented
a different approach to reestimate program costs. These corrective actions
helped SBA achieve an improved audit opinion on its fiscal year 2004 and
restated fiscal year 2003 financial statements.

New Cash Flow Model Improves SBA's Ability to Reasonably Estimate Program
Costs and Determine Loan Sale Results

In fiscal year 2003, SBA's contractor developed a new cash flow model to
calculate subsidy cost estimates and reestimates for the disaster loan
program. In contrast to the prior model's flawed single loan approach to
estimate the cash flows, the new model was designed to estimate cash flows
individually for each loan. This design facilitates calculating loan
values for loans sold to determine gains or losses on loan sales and, if
SBA schedules additional loan sales, could also be used to calculate loan
values

13 This discussion of the inconsistency with the interest rates used to
calculate interest on the financing account explains the problem as it
relates to a direct loan program that borrows from Treasury. Guarantee
loan programs would have a similar problem with interest earnings
calculated on uninvested funds held in the financing account that serve as
a reserve for the payment of future loan defaults or other payments.

for determining whether these sales would be financially beneficial to the
government, thus negating the need for a separate hold model. Application
of the model enabled SBA to retroactively determine the results of its
prior loan sales and to correct the abnormal balance in its subsidy
allowance.

During the development of the new cash flow model, SBA analyzed the
available disaster loan data, including loan performance information, loan
terms, disaster type and magnitude, and regional information, as well as
certain economic data, such as unemployment, gross domestic product, and
interest rates. Based on these analyses, the data that best predicted
default and prepayment behavior-two important cash flows for the disaster
loan program-were selected to use as variables in the model. The model
segments the loan portfolio into groups of loans based on the final
variables selected, which were (1) the age of the loan, (2) the type of
borrower (home or business), (3) the size of the loan, (4) the type of
disaster loan (economic injury or physical damage), and (5) the length of
the grace period. Based on these variables, there are a total of 162
groups of loans used to segment the disaster loan portfolio.

On a loan-by-loan basis, the cash flow model estimates the expected
principal and interest payments based on loan contract terms. Then the
model estimates deviations from these expected payments for delinquencies,
charge-offs, and prepayments. These deviations are calculated based on
historical averages of loan performance for each group of loans. Lastly,
the model estimates recoveries on charged-off loans based on historical
averages. The model's methodology is based on the assumption that the
behavior of loans in the future, taking several important loan
characteristics into account, will be similar to loans in the past.
However, as discussed later, if future loans are made to substantially
different types of borrowers, such as those with better or worse financial
strength, or have substantially different loan terms, changes to the model
would be required to correctly consider these new characteristics in the
cash flow estimates.

Throughout the development of the model, SBA documented several analyses
of the performance and characteristics of its disaster loans and the
model's ability to predict loan performance. In addition to its own
analyses, SBA contracted with Ernst & Young LLP (E&Y) to conduct an
independent review of the model. E&Y reviewed the model documentation and
computer code, and reviewed SBA's testing and validation analyses. E&Y
summarized its observations and findings in two reports issued in
November14 and December 2003.15 E&Y noted that the model can be expected
to perform reasonably well for reestimates of existing loans and to
produce stable estimates over time given the model's emphasis on longterm
averages. In addition, E&Y stated that given the limitation of what can be
known about future loans, the model takes a reasonable approach to
estimating costs of future loans for budget purposes. E&Y also noted that
the model achieves SBA's objective to consistently value individual loans
for reestimates and loan sales.

Based on our review of the model, its documentation, and the reports
issued by E&Y, we concluded that the new model provides a sound basis to
estimate costs and improved SBA's ability to prepare more reliable and
reasonable cost estimates for the disaster loan program. When SBA used
this new model for the first time to reestimate the cost of the disaster
loan program, it resulted in a reestimate indicating increased costs of
over $1 billion as of the end of fiscal year 2003. As shown in figure 3,
the adjustment to increase these costs on SBA's books helped bring the
disaster loan program's subsidy allowance to a positive balance and more
in line with expectations for this type of subsidized program. In SBA's
fiscal year 2004 financial statements, the subsidy allowance was reported
to be about $613 million, or about 20 percent, of the $3 billion
outstanding balance of the disaster loan program. Given that the estimated
cost of the program generally ranges from $16 to $36 for every $100 that
SBA lends, this balance is within the expected range. SBA also used the
new model to recalculate the results of its prior disaster loan sales and
determined that the sales resulted in losses, or increased budgetary
costs, of over $900 million.

14 Ernst & Young LLP, Independent Review of the SBA Disaster Loan Program
Subsidy Model Part I (Nov. 21, 2003).

15 Ernst & Young LLP, SBA Disaster Loan Model Review, Part II (Dec. 12,
2003).

Figure 3: Disaster Subsidy Allowance as a Percent of the Loan Balance
Outstanding, Fiscal Years 1996 through 2004

Percent

                                       30

To resolve the inconsistency in the interest rates used to calculate
interest on its financing account borrowing from Treasury and the interest
rates used to discount cash flows when estimating subsidy costs, SBA
completed a detailed analysis of its interest transactions with Treasury.
SBA recalculated what its interest payments would have been based on the
final (rather than the interim) interest rates and determined that it
overpaid Treasury by about $128.6 million and $5.6 million as of the end
of fiscal years 2003 and 2004, respectively. These amounts were included
in SBA's reestimates for the disaster loan program and corrected SBA's
interest transactions for the fiscal years 1992 through 2003 cohorts.16

16 The calculation as of the end of fiscal year 2003 corrected prior
interest payments for the fiscal years 1992 -2001 cohorts. The calculation
as of the end of fiscal year 2004 corrected the prior interest payments
for the fiscal years 2002 and 2003 cohorts. Without a comprehensive
solution, this would be an ongoing process needed to correct prior
interest payments once cohorts become substantially disbursed and the
final interest rates are known.

                                       25

                                       20

                                       15

                                      10 5

                                    0 -5 -10

                                      -15

                                      -20

1996 1997 1998 1999 2000 2001 2002 2003 2004

Fiscal year

Source: GAO analysis of SBA's fiscal years 1996 through 2004 financial
statements (prior to any restatements).

Analysis of Interest Payments and Change in Approach to Reestimate Costs
Addressed Interest Rate Inconsistency

Also in fiscal year 2004, SBA implemented a new approach to reestimate
costs, called the balances approach.17 Because the balances approach
determines the amount of the reestimate based on a comparison of resources
in the financing account and expected future cash flows, the approach
automatically adjusts the financing account and subsidy allowance for any
inconsistency in interest rates going forward.

Other agencies that disburse or guarantee loans would also be affected by
this interest rate inconsistency, which could result in misstatements in
their accounts. However, the significance of this issue cannot be
determined without extensive analyses similar to SBA's analysis because a
number of factors influence how the inconsistency would impact other
agencies' accounts, including balances in financing accounts and the
length of time a program takes to substantially disburse its loans. OMB
has notified agencies of the flaw in the tools and outlined plans to issue
a comprehensive revised reestimate tool that resolves this problem.18
Until updated tools are provided, agencies will continue to make incorrect
interest payments that could result in financing accounts having excess or
insufficient funds and misstatements in financial statement reporting
accounts for credit programs.

SBA's Corrective Actions Helped It Achieve an Improved Audit Opinion

SBA's corrective actions helped it achieve an improved audit opinion for
its fiscal year 2004 financial statements. Earlier, SBA's auditor withdrew
its unqualified opinions on SBA's fiscal years 2000 and 2001 financial
statements and issued a disclaimer of opinion for fiscal year 2002, in
part, because of issues identified in our January 2003 report. While
progress was made in addressing these issues during fiscal year 2003, the
auditor also issued a disclaimer of opinion on SBA's fiscal year 2003
financial statements. The auditor reported that because SBA was late in
completing reestimates and preparing its financial statements, among other
things, the

17 Current OMB guidance allows agencies to use either the "traditional
approach" or the "balances approach" to reestimate costs. Both approaches
will produce similar results as long as cohort cash flows and transactions
with Treasury are properly recorded in the financing account and actual
performance is properly included in the cash flow model. To validate the
results obtained with the balances approach, SBA reconciled the
differences between the two approaches. According to SBA, after the
reconciliation, the two approaches produced cost estimates that differed
by less then .005 percent of disbursements.

18 Agencies were alerted to the flaw and future changes to the tools used
to calculate subsidy cost estimates and reestimates in an e-mail from OMB
dated April 1, 2005.

auditor did not have adequate time to resolve reservations related to
SBA's disaster loan program, including abnormal balances in the subsidy
allowance and a difference in interest rates SBA used to estimate subsidy
costs and calculate interest payments to Treasury.

Subsequently, SBA continued to implement its corrective actions, which the
auditor assessed as part of the fiscal year 2004 financial statement
audit. SBA received a mixed opinion-a combination of unqualified and
qualified opinions-on its fiscal year 2004 financial statements,19 which
represented an improvement over the disclaimer it received for fiscal year
2003. In addition, SBA received an unqualified opinion on its restated
fiscal year 2003 balance sheet.20 SBA's auditor did not cite any issues
related to previously identified problems with the disaster program or the
new disaster cash flow model in these audit opinions.21

New Policies and Procedures Will Help Ensure Future Estimates Are
Reasonable, but Additional Procedures and Documentation Are Needed

In addition to implementing the corrective actions to resolve the
accounting anomalies, SBA also implemented new policies and procedures to
help ensure that future loan program cost estimates will be reasonable,
including (1) the development and implementation of new standard operating
procedures for calculating reestimates; (2) the preparation of
documentation to support the rationale and basis for key aspects of the
cash flow model; (3) a process to coordinate the preparation of cost
estimates between budget, accounting, and program staff; and (4) a revised
reestimate approach. However, additional documentation of the new cash
flow model would help ensure proper operation and maintenance of the
model. Further, over time it will be important for SBA to continue to
assess the model's ability to predict loan performance. In addition, there
may be opportunities to improve the model, as well as simplify the
estimation process, that warrant further consideration by SBA. Lastly,
additional

19 For fiscal year 2004, SBA received an unqualified audit opinion on its
statement of budgetary resources and a qualified audit opinion on all its
other statements.

20 As part of the fiscal year 2004 audit, SBA's restated fiscal year 2003
balance sheet was reaudited in order to determine whether opening balances
for fiscal year 2004 were reliable. The statements of net cost, financing,
and changes in net position were also restated, but were not re-audited.

21 The qualifications of the audit opinions on most of SBA's statements
for fiscal year 2004 related primarily to new concerns over SBA's
estimated loan guarantee cash flow activity for the second half of fiscal
year 2004, which was necessary to accommodate the accelerated financial
reporting due date.

procedures to test the disaster data used in the model will help ensure
their reliability.

New Policies and Procedures Strengthened Internal Controls

During fiscal years 2003 and 2004, SBA enhanced its policies and
procedures by implementing several of the internal control practices
identified in federal accounting guidance that will help it ensure that
future cost estimates are reliable and reasonable.22 SBA developed and
implemented standard operating procedures for calculating its reestimates
based on federal accounting guidance. These procedures established an
internal review process and standardized steps that must be performed and
documented as part of the reestimate process. Steps included ensuring that
the correct cash flow model files are used, verifying that the model
appropriately reflects the program's structure, documenting any technical
changes to the model, updating actual data and estimated cash flows in the
model, and reviewing the reasonableness of the estimated cash flows. The
procedures also call for the cash flow model to be reviewed by an outside
party. The supporting documentation for the reestimates was provided to
SBA's financial statement auditor during the fiscal year 2004 audit. The
auditor noted in its report on internal controls that the adherence to a
set of standard operating procedures for calculating reestimates, along
with improved documentation and an effective internal review process, were
critical to SBA's success in meeting key milestone dates and completing
the audit process within accelerated financial reporting deadlines.

SBA also established other important internal control practices identified
in the guidance. During the development of the new cash flow model, SBA
documented key analyses and decisions regarding the model's methodology.
For example, SBA compared the characteristics of the loans sold to the
loans kept and developed an approach within the new model to take those
differences into account when estimating loan performance. It also
documented the basis for selecting the model's methodology and variables,
the assumptions and calculations in the model, and results of testing the
model's ability to predict cash flow estimates. This documentation helps
support the rationale and basis for key aspects of the

22 In January 2004, the Federal Accounting Standards Advisory Board's
Accounting and Auditing Policy Committee issued Technical Release 6,
Preparing Estimates for Direct Loan and Loan Guarantee Subsidies under the
Federal Credit Reform Act Amendments to Technical Release 3: Preparing and
Auditing Direct Loan and Loan Guarantee Subsidies under the Federal Credit
Reform Act, which provides guidance to agencies on preparing subsidy cost
estimates.

model that provide important cost information for budgets and financial
statements.

SBA has also established procedures to coordinate the preparation of cost
estimates among budget, accounting, and program offices. This will help
ensure that the estimates are reviewed and prepared with the proper
information. Further, as previously discussed, SBA implemented the
balances approach for reestimates. This approach will help ensure that
SBA's account balances are in line with expected future cash flows. These
practices and the other practices discussed above will help ensure that
anomalies such as those we identified during our last review do not go
undetected or uncorrected.

Additional Documentation, Analysis, and Testing Would Improve Long-Term
Reliability of Cost Estimates

While SBA resolved its accounting anomalies related to the disaster loan
program and made important improvements to its policies and procedures, we
found that additional enhancements to internal controls would help ensure
the long-term reliability of future cost estimates. Further, strengthening
internal controls will help SBA identify potential problems in the future
and sustain the progress it has already made.

Even though SBA completed substantial documentation for the new cash flow
model, we found that this documentation was not sufficient to readily
provide for knowledge transfer between staff and contractors to help
ensure proper maintenance and updating of the model. For example, the
documentation does not specify what is done to prepare the data for use in
the model and does not always clearly indicate the data sources. In
addition, SBA's documentation to explain the files used to run the model
and update the data used in the model was not complete. For example, in
SBA's documentation of the files and steps used to run the cash flow
model, out of a total of 19 steps, 6 steps had no explanations of the
process and another 2 were not complete and indicated that someone who was
no longer employed at SBA was to provide the information.

Improved documentation is particularly important because SBA relied on a
contractor to help develop the new cash flow model for the disaster loan
program and has recently experienced significant turnover in staff
responsible for preparing cost estimates. Without complete and detailed
documentation on how to maintain the model, update it with additional
data, and run it, it will be more difficult for current SBA staff to fully
understand the model, which could result in future errors in the cash flow
estimates.

Thorough documentation of the model is even more important given the
complexity associated with its calculation process. E&Y noted in its
review, and we agree, that the model's complexity creates an ongoing
challenge related to transparency and maintenance. Because complexity
increases the risk of errors occurring, SBA could benefit from continuing
to evaluate whether there are opportunities to simplify the estimation
process with model revisions or alternative estimation methodologies.

There may also be opportunities to improve the model with additional
variables. When estimating loan performance, the new model does not use
data related to the financial strength of borrowers. Because this kind of
information has been shown to be useful in predicting loan performance,
such as defaults and prepayments, incorporating this type of information
could improve the model's estimates. Further, additional detailed data on
borrower financial strength and loan collateral, among other things, may
improve the model's effectiveness for supporting any future loan sales.
According to SBA officials, beginning in fiscal year 2003 SBA began
collecting credit scores for disaster loan borrowers. Once these newer
loans have sufficient historical data, SBA will be able to evaluate the
usefulness of these data as a potential variable to predict loan
performance.

In addition to opportunities to improve the model, SBA could also enhance
its procedures to ensure that the model's estimates reasonably predict
future loan performance. While SBA has completed testing of the model's
ability to predict loan performance, it is important that SBA establish a
process to help ensure this testing continues routinely and that causes of
any significant variances are identified and addressed. For example, as
stated earlier, if future loans are made to substantially different types
of borrowers or have substantially different loan terms, changes to the
model would be required to correctly consider these new characteristics in
the cash flow estimates. Routine testing would help identify this type of
change.

While the new cash flow model provides SBA with a sound approach to
estimate costs for the disaster loan program, additional verification
procedures would provide better assurance that data used by the model are
reliable. Federal accounting guidance requires agencies to accumulate
sufficient, relevant, and reliable supporting data that provide a reliable
basis for estimates of future loan performance. Because SBA's old cash
flow model used data from SBA's Main On-Line System for Tracking
Evaluation and Response (MONSTER) database which contains summary

information, most of its detailed data reliability assessments and
reconciliation practices revolved around MONSTER. For example, SBA
maintains a reconciliation that tracks loans in MONSTER with its general
ledger at a cohort level. However, the new model uses data from MONSTER
and loan-level data from SBA's Electronic Loan Information Processing
System (ELIPS) database. SBA officials indicated that plans are to
continue to move away from using MONSTER. While SBA routinely reconciles
its ELIPS database at a high level, as it moves toward using ELIPS data
for estimating its disaster program costs, it is important that SBA
reconcile and test the data at the level used in the model.

Conclusions	SBA took prompt action to identify the deficiencies related to
its disaster and loan sale programs with a comprehensive review of its
financial records. The corrective actions it then took established a basis
for reliable and reasonable cost estimates. At the same time, the
complexities associated with estimating costs for these programs will
require continued attention. Without enhancements to the model's
documentation, additional procedures to test data reliability, and
continued testing and analysis of the model, SBA may find it difficult to
fully sustain the progress it has made. Further, improved tools from OMB
would help SBA and other agencies ensure proper calculation of interest
costs related to their credit programs.

Recommendations for Executive Action

We are making five recommendations to SBA and two to OMB. To help ensure
that future subsidy cost estimates are reliable, we recommend that the SBA
Administrator take the following five actions.

o 	Develop additional documentation of the new disaster cash flow model to
help facilitate proper operation, maintenance, and updating of the model.

o 	Study the value of incorporating additional variables in the new
disaster cash flow model, such as detailed information on the financial
strength of borrowers.

o 	Establish policies and procedures to routinely test the new disaster
cash flow model's ability to predict loan performance by comparing the
model's predictions to actual loan performance and to identify and address
the causes of any significant variances.

o 	Consider possible revisions to the model and/or alternative
methodologies that would simplify the estimation process.

o 	Establish additional procedures to test and document the reliability of
the data used in the new cash flow model for the disaster loan program.

To help ensure that agencies make correct interest calculations for
financing accounts, we recommend that the OMB Director take the following
two actions.

o 	Update the tools provided to agencies for adjusting financing account
interest transactions once a final interest rate is determined for a
cohort.

o 	Provide instructions to agencies on making retroactive corrections to
financing account interest transactions based on final interest rates for
a cohort.

Agency Comments 	In written comments reprinted in appendix II, SBA stated
that these were appropriate recommendations and that it already has work
underway to address several of them. In written comments reprinted in
appendix III, OMB agreed with our recommendations and stated that it would
work with agencies to correct interest transactions with Treasury. SBA and
OMB also provided technical comments, which we have incorporated as
appropriate.

As agreed with your offices, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 30 days from
its date. At that time, we will send copies of this report to the Ranking
Minority Member of the Senate Committee on Small Business and
Entrepreneurship, other appropriate congressional committees, the
Administrator of the Small Business Administration, and the Director of
the Office of Management and Budget. Copies will also be made available to
others upon request. In addition, the report will be available at no
charge on the GAO Web site at http://www.gao.gov.

If you or your staffs have any questions about this report, please contact
me at (202) 512-9508 or [email protected]. Major contributors to this report
are acknowledged in appendix IV.

Linda M. Calbom
Director
Financial Management and Assurance

Appendix I

                       Objectives, Scope, and Methodology

To describe the nature of the deficiencies SBA identified that contributed
to the disaster loan program accounting anomalies, we reviewed the report
prepared by IBM that summarized the detailed review of SBA's accounting
and budgeting for its disaster loan program and its loan sale procedures
performed by SBA and IBM, and a variety of documents prepared by SBA that
summarized the issues found. We also reviewed OMB Circular A-11 and FCRA
to determine the criteria for calculating interest payments to Treasury on
borrowing. We also interviewed SBA and OMB officials, SBA's financial
statement auditor, and a contractor with SBA.

To identify the corrective actions taken and assess whether these actions
resolved the identified deficiencies, we:

o 	obtained and assessed the new cash flow model used to estimate the cost
of the disaster loan program and its various supporting documentation;

o 	obtained an understanding of how the model works by reviewing SBA's
summary of the behavior equations, the production system documentation,
and assumptions made in the model;

o 	analyzed the model's methodology, including choice of statistical
technique and variables included in the analysis, and determined that they
were appropriate and reasonably related to the prediction of cash flows of
disaster loans;

o 	replicated certain components of the model, such as the process used to
segment the portfolio into groups of loans and predict future loan
performance;

o 	assessed SBA's logistic regression used to determine and support the
variables used in the model to verify that the variables selected were
statistically significant;

o 	reviewed (1) SBA's testing of the cash flow model for biases; (2) SBA's
comparison of the loans sold and the loans kept; (3) E&Y's reports
summarizing its independent reviews of the model, including procedures
performed, findings, and observations; and (4) SBA's financial statement
auditor's assessment of SBA's model, its reestimates, and its revised loan
sale loss calculation;

Appendix I
Objectives, Scope, and Methodology

o 	obtained SBA's analysis of its interest payments to Treasury and
verified the calculations;

o 	reviewed SBA's fiscal year 2004 financial statements summarizing the
implementation of the balances approach to reestimate costs;

o 	interviewed SBA officials, an SBA contractor, and SBA's financial
statement auditor; and

o 	interviewed OMB officials to obtain an understanding of their efforts
to update the tools agencies use to calculate interest on financing
account balances.

To determine whether SBA's new cash flow model and procedures for the
disaster loan program provide a reasonable basis for future subsidy cost
estimates, we interviewed SBA officials and a contractor to obtain an
understanding ofSBA policies and procedures for estimating subsidy costs.
We reviewed supporting documentation related to its procedures, including
the standard documentation template used to support its reestimates for
its fiscal year 2004 financial statements, the various documentation
prepared to support the model, and the reliability of data from SBA's
computer systems. Based on SBA's procedures and documentation, we assessed
the sufficiency of SBA's estimation process based on federal accounting
guidance that identifies internal control practices that help ensure that
future cost estimates are reliable and reasonable. We also reviewed SBA's
financial statement auditor's reports on internal controls for fiscal
years 2003 and 2004. The checking of key components, along with our review
of SBA's documentation and E&Y's evaluation of the model, provided a
sufficient level of understanding to conclude on its approach and ability
to produce more reliable and reasonable cost estimates for the disaster
loan program. To identify any additional steps SBA could take to improve
the long-term reliability of its model, we considered additional types of
variables that might enhance SBA's approach. As part of this analysis, we
reviewed academic literature on default modeling and discussed alternative
variables and modeling techniques with the contractor SBA used to develop
the model. Based on these assessments, our assessment of the model, and
E&Y's findings and observations on the cash flow model, we identified
opportunities SBA could explore to enhance its procedures to improve the
long-term reliability of its cost estimates.

We provided SBA a draft of this report and OMB a draft of applicable
sections of this report for review and comment. SBA and OMB provided

Appendix I
Objectives, Scope, and Methodology

written comments, which are reprinted in appendix II and III,
respectively. They also provided technical comments, which we have
incorporated as appropriate. We performed our work in accordance with
generally accepted government auditing standards in Washington, D.C. from
April 2004 through March 2005.

Appendix II

Comments from the Small Business Administration

Appendix II
Comments from the Small Business
Administration

Appendix III

Comments from the Office of Management and Budget

Appendix IV

                     GAO Contact and Staff Acknowledgments

                    GAO Contact Linda Calbom, (202) 512-9508

Acknowledgments	In addition to the above, Marcia Carlsen, Lisa Crye,
Austin Kelly, Beverly Ross, Kara Scott, and Brooke Whittaker made key
contributions to this report.

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

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

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

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

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

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

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

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

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

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

Relations Washington, D.C. 20548

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