Small Business Administration: Section 7(a) General Business	 
Loans Credit Subsidy Estimates (21-AUG-01, GAO-01-1095R).	 
								 
The process and types of data the Small Business Administration  
(SBA) uses to estimate the subsidy cost of the 7(a)General	 
Business Loan Program are generally reasonable and comply with	 
existing Office of Management and Budget guidance. GAO's review  
of actual and originally estimated defaults and recoveries showed
that, on a cumulative basis since 1992, defaults were		 
overestimated by approximately $2 billion and recoveries were	 
overestimated by approximately $450 million. During this same	 
period, SBA overestimated the cost of the 7(a) program by $958	 
million as evidenced from a trend of downward reestimates. The	 
majority of these downward reestimates can be attributed to the  
overestimate of defaults. For those loan guarantees approved from
fiscal years 1992 through 1997, GAO was unable to determine the  
specific reason for the overestimate of defaults primarily	 
because the basis SBA used for the estimated default rate for	 
these years was not documented. During this period reestimates	 
account for approximately 84 percent of the total $958 million	 
reestimate. SBA began using its current methodology in 1998.	 
Under this method, high default rates associated with loan	 
guarantees approved in fiscal years 1986 through 1990 contributed
to the difference between estimated and actual defaults for loan 
guarantees approved from 1998 through 2000. SBA has proposed to  
the Office of Management and Budget another methodology that uses
the five most recent years of actual loan performance prior to	 
each activity year being estimated--referred to as the lookback  
period--rather than the current approach. The proposed method	 
would be more sensitive to fluctuations in economic conditions or
changes in program delivery or design because it uses a shorter  
lookback period. The benefit of this approach is that, in a	 
continuing stable economy, the original subsidy cost estimate	 
would be expected to more closely match actual loan performance  
and reestimates would therefore be smaller. However, the risk of 
this approach is that a sudden downturn in the economy would be  
more likely to result in actual loan performance being different 
than estimated and thus would likely result in larger upward	 
reestimates than under the current approach.			 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-01-1095R					        
    ACCNO:   A01633						        
  TITLE:     Small Business Administration: Section 7(a) General      
             Business Loans Credit Subsidy Estimates                          
     DATE:   08/21/2001 
  SUBJECT:   Small business loans				 
	     Subsidies						 
	     Loan defaults					 
	     Government guaranteed loans			 
	     SBA 7(a) General Business Loan Program		 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** 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-01-1095R
     
GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates

United States General Accounting Office Washington, DC 20548

August 21, 2001 The Honorable John F. Kerry Chairman The Honorable
Christopher S. Bond Ranking Minority Member Committee on Small Business
United States Senate

The Honorable Donald A. Manzullo Chairman The Honorable Nydia M. Velazquez
Ranking Minority Member Committee on Small Business House of Representatives

Subject: Small Business Administration: Section 7( a) General Business Loans
Credit Subsidy Estimates

In your May 4, 2001, letter, you expressed concerns about the Small Business
Administration?s (SBA) 7( a) Business Loan Program subsidy rate
calculations. As agreed with the staff of your committees, we reviewed the
subsidy rate estimation process and the data SBA uses in its calculation,
with a specific focus on defaults and recoveries. We identified differences
between originally estimated defaults and recoveries and actual data, and
the causes of these differences. Additionally, we assessed the implications
of proposed changes to SBA?s current approach to estimate defaults. On July
30, 2001, we briefed your staff on the results of our review. This letter
transmits the material from the briefing.

In summary, the process and types of data SBA uses to estimate the subsidy
cost of the 7( a) program are generally reasonable and comply with existing
Office of Management and Budget (OMB) guidance. However, our review of
actual and originally estimated defaults and recoveries showed that, on a
cumulative basis since 1992, defaults were overestimated by approximately $2
billion and recoveries were overestimated by approximately $450 million. 1
During this same period, SBA overestimated the cost of the 7( a) program by
$958 million as evidenced from a trend

1 Because SBA calculates estimated recoveries as a percent of estimated
defaults, most of the overstated recoveries resulted from the initial
overestimate of defaults. When recoveries were calculated independent of the
default overestimate, the cumulative overstatement of recoveries was less
than 1 percent of actual recoveries, or about $3 million.

GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 2 of downward
reestimates. 2 The majority of these downward reestimates can be

attributed to the overestimate of defaults. For those loan guarantees
approved from 1992 through 1997, we were unable to determine the specific
reason for the overestimate of defaults primarily because the basis SBA used
for the estimated default rate for these years was not documented. 3
Reestimates during this period account for approximately 84 percent of the
total $958 million reestimate. SBA began using its current methodology in
1998. This methodology uses average historical data since 1986 to estimate
defaults. Under this method, high default rates associated with loan
guarantees approved in fiscal years 1986 through 1990 contributed to the
difference between estimated and actual defaults for loan guarantees
approved from 1998 through 2000.

SBA has proposed to OMB another methodology that uses the 5 most recent
years of actual loan performance prior to each activity year being
estimated- referred to as the lookback period 4 -rather than the current
approach that uses all actual loan performance since 1986, to estimate loan
performance for each activity year. OMB is currently considering this
proposal. Either approach has certain benefits and inherent risks.

Under the current approach, initial estimates of the subsidy rate are fairly
stable because they include more years of historical data that smooth out
fluctuations in economic conditions from year to year. As previously
mentioned, the current approach includes several early years with relatively
high default rates. A benefit of this approach, given SBA's historical
experience, is that it provides a cushion in the event of an unexpected
downturn in the economy. However, this cushion ties up appropriations that
could have been available to other discretionary programs. As has recently
been the case for SBA, this approach is more likely to result in continuing
annual downward reestimates when there is a strong economic environment.

The proposed method would be more sensitive to fluctuations in economic
conditions or changes in program delivery or design because it uses a
shorter lookback period. The benefit of this approach is that, in a
continuing stable economy, the original subsidy cost estimate would be
expected to more closely match actual loan performance and reestimates would
therefore be smaller. However, the risk of this approach is that a sudden
downturn in the economy would be much more likely

2 In addition to the differences between actuals and estimates to date, the
total downward reestimate would also be affected by the present value of
these differences and changes in the estimates for expected future loan
performance.

3 According to SBA officials, prior to the estimate of the 1998 cohort's
subsidy cost in fiscal year 1996, subsidy cost estimates were prepared based
on direct consultation with OMB. 4 For example, under the 5 year lookback
period, the 2002 cohort estimate of year one default activity would be based
on the average actual first year defaults that occurred for the 1996 through
2000 cohorts and the second year default activity would be based on actual
second year defaults that occurred for the 1995 through 1999 cohorts. Under
the current approach, the lookback for all activity years includes the
average of all cohorts back to 1986.

GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 3 to result in
actual loan performance being different than estimated and thus would

likely result in larger upward reestimates than under the current approach.
SBA generally agreed with the information presented in this briefing. SBA
officials added, however, that they view the proposed changes in default
estimation methodology to be an interim solution. SBA views the long- term
solution as a sophisticated econometric modeling approach. Econometric
modeling considers key relationships between loan performance and economic
and other indicators.

- - - - We are sending copies of this letter to the Administrator of the
Small Business Administration and the Director of the Office of Management
and Budget. This letter will also be available on GAO?s homepage at http://
www. gao. gov.

If you have any questions, please contact me at (202) 512- 9508 or by e-
mail at calboml@ gao. gov, or contact Dan Blair, Assistant Director, at
(202) 512- 9401 or by email at blaird@ gao. gov. Key contributors to this
letter were Marcia Carlsen, Ruth Sessions, and Bill Shear.

Linda M. Calbom Director Financial Management and Assurance

Enclosure

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 4

Briefing Before the Staffs of the Senate and House Committees on Small
Business

1 Briefing to Staff of the Senate and House

Committees on Small Business Small Business Administration Section 7( a)
General Business Loans Credit Subsidy Estimates

July 30, 2001

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 5

2 Contents

*Objectives  Scope and Methodology  Background  Process and Data Used to
Estimate 7( a) Subsidy Costs  Reestimates of the 7( a) Program Subsidy
Costs  Comparison of Originally Estimated Defaults and Recoveries to Actual

Data  Effect of Overestimating the 7( a) Program?s Subsidy Cost  Causes of
Differences  Implications of Proposed Changes  Agency Comments

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 6

3 Objectives

Our objectives for the Section 7( a) General Business Loans (the 7( a)
program) review were to

 identify the types of data and process used to estimate the subsidy cost,
including the incorporation of program changes,

 compare originally estimated defaults and recoveries from the 1992 through
2000 subsidy cost estimates to actual data recorded in the accounting
system,

 determine the causes of differences between original estimates and actual
defaults and recoveries,

 assess the implications of proposed changes to SBA?s approach to estimate
defaults.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 7

4 Scope and Methodology

 To achieve our objectives, we

 discussed SBA?s process and types of data used to prepare subsidy cost
estimates with agency staff,

 compared SBA?s current process to prepare subsidy cost estimates to
existing guidance from the Office of Management and Budget (OMB),

 reconciled actual data used as a basis to estimate defaults and recoveries
with data from the accounting system, 1

 analyzed trends in the actual defaults, recoveries and guaranteed
percentages,

1 We were not able to reconcile to the actual data prior to fiscal year 1992
because the current accounting system was implemented in 1992 and does not
include data prior to that time.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 8

5 Scope and Methodology

 compared the original estimated default and recovery amounts for the 1992
through 2000 cohorts 2 to actual loan performance data recorded in the
accounting system,

 discussed the causes of differences and proposed changes with SBA staff
and OMB officials, and

 determined the potential impact of various alternative approaches on
subsidy cost estimates.

 Our audit work was conducted in Washington, D. C., from May 2001 through
July 2001 in accordance with generally accepted government auditing
standards. 2 A cohort includes those direct loans or loan guarantees of a
program for which a subsidy appropriation is provided for in a given fiscal
year even if the loans are not disbursed until subsequent years.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 9

6 Background

 The 7( a) program guarantees loans made to small businesses that are
unable to obtain financing on similar terms in the private credit market but
can demonstrate the ability to repay the loan.

 SBA reported that its share of outstanding 7( a) loan guarantees totaled
nearly $22.9 billion as of September 30, 2000. This represents about 65
percent of SBA?s total loan guarantees outstanding.

 From 1995 to 1996, SBA undertook a significant data gathering effort to
capture historical loan performance for the 7( a) loan program and began
using this data in 1996 to estimate the subsidy cost of the 1998 cohort.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 10

7 Background

 Since the inception of credit reform, the 7( a) program has had net
downward reestimates of nearly $1 billion. 3

 In March 2001, SBA submitted a proposal to OMB, which is discussed later
in more detail, to adjust its approach to estimating the subsidy cost of the
7( a) program.

 OMB is in the process of reviewing the recent SBA proposal.

3 A downward reestimate indicates a cohort of loans or loan guarantees is
expected to cost the federal government less than previously anticipated.
This amount does not include the portion of the reestimate attributable to
interest.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 11

8 Background

 Prior to the Federal Credit Reform Act (FCRA) of 1990, credit programs--
like most other federal programs-- were reported in the budget on a cash
basis.

 Loan guarantees appeared to be free in the budget year while direct loans
appeared to be as expensive as grants.

 This cash basis distorted costs and, thus, the comparison of credit
program costs with other programs and each other.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 12

9 Background

 FCRA was, among other reasons, enacted to more accurately measure the
government?s costs of federal loan programs and to permit better comparisons
both among credit programs and between credit and noncredit programs.

 Under FCRA, agencies are required to estimate the cost of extending or
guaranteeing credit over the life of the loan, called the subsidy cost.

 This cost is the estimated long- term cost to the government of direct
loans or loan guarantees calculated on a net present value 4 basis,
excluding administrative costs.

4 The net present value expresses expected future cash inflows and outflows
in today's dollars. In calculating the present value, prevailing interest
rates provide the basis for converting future amounts into today's dollar
equivalents.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 13

10 Background

 In the subsidy cost calculation, agencies estimate the cash flows for a
program, including (but not limited to) estimated defaults, recoveries, and
fees, and the effects of prepayments, on a cohort basis, for the life of the
loans.

 Generally, agencies are required to annually update the subsidy cost -
referred to as reestimates - of each cohort based on information about the
actual performance and/ or estimated changes in future loan performance.

 FCRA recognized that agencies? ability to make subsidy cost estimates that
mirrored actual loan performance could be impeded by various factors and
provided permanent indefinite budget authority for reestimates that reflect
increased credit program costs.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 14

11 Background

 Section 503 of FCRA states that OMB is responsible for, among other
things,

 coordinating subsidy cost estimates for executive branch agencies and

 reviewing historical data and developing the best possible credit subsidy
estimates.

 The Accounting and Auditing Policy Committee?s 5 (AAPC) Technical Release
3, Preparing and Auditing Direct Loan and Loan Guarantee Subsidies under the
Federal Credit Reform Act, identifies specific practices that, if fully
implemented by credit agencies, will enhance their ability to reasonably
estimate loan program costs. 5 The AAPC is sponsored by the Federal
Accounting Standards Advisory Board.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 15

12 Process and Data Used to Estimate 7( a)

Subsidy Costs

 When calculating the subsidy cost of the 7( a) program, SBA considers, for
the life of the loans guaranteed

(1) fees that will be received, (2) the percent of total loan amounts
guaranteed, which

currently can not exceed 75 or 85 percent depending on the loan amount,

(3) the volume and mix of loan guarantees, 6 and (4) the amount and timing
of defaults and recoveries.

 To estimate defaults and recoveries, SBA averages its historical loan
performance since 1986. 7

6 The volume and mix of loan guarantees refers to the total amount of loans
SBA expects to guarantee and the various loan sizes based upon different fee
and guaranteed percentages. 7 SBA began using this historical database in
1996 to calculate the subsidy cost of the 1998 cohort.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 16

13 Process and Data Used to Estimate 7( a)

Subsidy Costs

 According to SBA staff, when there is a change in the 7( a) program?s
design, SBA staff

 determine if the change affects existing assumptions or adds a new
assumption to the subsidy cost calculation,

 determine if there is any historical data that could be used to assess the
impact of the change on the subsidy cost estimates, and

 use informed opinion 8 to estimate the impact on the subsidy cost if no
applicable historical experience exists.

8 Informed opinion refers to the judgment of agency staff who make subsidy
estimates based on their programmatic knowledge and/ or experience.
According to Technical Release 3, informed opinion is an acceptable approach
in situations where historical data does not exist.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 17

14 Process and Data Used to Estimate 7( a)

Subsidy Costs

 SBA generally uses the same process and types of data as explained on the
prior two slides to calculate reestimates of subsidy costs. In addition, as
part of the reestimate process,

 as actual loan performance becomes available, it replaces estimated cash
flows and

 expectations of future loan performance are updated based on information
about actual performance and/ or estimated changes in future loan
performance.

 In summary, the process and types of data SBA uses to estimate the subsidy
cost of the 7( a) program are generally reasonable and comply with existing
OMB guidance.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 18

15 Reestimates of the 7( a) Program Subsidy

Costs

 Since the inception of credit reform, SBA has overestimated the original
subsidy cost of the 7( a) program by nearly $1 billion, as evidenced by the
net downward reestimate shown on the following slide.

 Because reestimate data were not separately available for interest,
defaults, fees and other cash flows, we were unable to determine the net
overestimate attributable to each of these factors.

 However, based on our comparisons of originally estimated defaults and
recoveries to actual loan performance, a significant portion of the 7( a)
program?s total $1 billion net reestimate is attributed to the overestimate
of defaults.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 19

16 Reestimates of the 7( a) Program Subsidy

Costs Reestimate History of the 7( a) Program (Dollars in millions)

Source: Small Business Administration Note: For each annual reestimate, net
amounts were either received from Treasury (1997

Budget) or returned to Treasury (1998 - 2002 Budget).

Cohort 1997 Budget 1998 Budget 1999 Budget 2000 Budget 2001 Budget 2002
Budget Cummulative 1992 $5 ($55) ($30) ($74) ($5) ($4) ($163) 1993 (14) (77)
(50) (80) (21) (16) (259) 1994 53 (14) (63) (60) (12) (4) (100) 1995 11 49
(68) (60) (1) (4) (73) 1996 32 37 (101) (16) (9) (57) 1997 (24) (86) (39)
(0) (149) 1998 (52) (39) (39) (130) 1999 (13) (11) (24) 2000 (3) (3) Totals
$54 ($65) ($198) ($513) ($145) ($91) ($958)

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 20

17 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data

 SBA originally overestimated defaults 9 for 1992 through 2000 by over $2
billion, or about 87 percent, when compared to actual loan performance.

 Since estimated recoveries are based on a percent of estimated defaults,
SBA also originally overestimated recoveries for 1992 through 2000 by nearly
$450 million, or about 62 percent, when compared to actual loan performance.

 According to SBA staff, overestimating fees also contributed to the 7( a)
program total net reestimate. However, we did not attempt to quantify the
effect of fees. 10

9 The amount defaulted is based on the portion SBA guarantees. 10 In
addition to the differences between actuals and estimates to date, the net
reestimate would also be impacted by the

present value of these differences and changes in the estimates for expected
future loan performance.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 21

18 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data

 The following 4 slides summarize the results of our comparison of original
estimates of defaults and recoveries to actual defaults and recoveries for
the 1992 through 2000 cohorts.

 The original estimates of defaults and recoveries for each cohort are
based on expectations of future loan performance from the time of
origination through fiscal year 2000. Actual defaults and recoveries for
each cohort are based on actual loan performance through fiscal year 2000.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 22

19 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data Percentage by which Originally
Estimated Defaults were more than Actual Defaults for Fiscal

Years 1992 through 2000 (Cumulatively by Cohort) Source: GAO analysis based
on SBA data. Note: By the end of fiscal year 2000, only the 1992 through
1996 cohorts had reached the typical peak default years, which historically
have been years 3 through 5 after approval.

$0 $100

$200 $300

$400 $500

$600 $700

$800 $900

$1,000 1992 1993 1994 1995 1996 1997 1998 1999 2000 Cohort

Dollars in millions

Estimated Actual 62%

90% 102%

42% 88% 113%

108% 179%

293%

The 1998 cohort was the first to be estimated using historical data.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 23

20 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data Percentage by which Originally
Estimated Recoveries were more (less) than Actual

Recoveries for Fiscal Years 1992 through 2000 (Cumulatively by Cohort)
Source: GAO analysis based on SBA data Note: N/ A indicates that there were
no actual recoveries as expected for a cohort in its first year.

$0 $50

$100 $150

$200 $250

$300 1992 1993 1994 1995 1996 1997 1998 1999 2000

Cohort Dollars in millions

Estimated Actual 22%

N/A (75%) 18% 99% 27% 118%

108% 55%

The 1998 cohort was the first to be estimated using historical data.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 24

21 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data

 In order to assess estimated recoveries independently from the effect of
overestimating defaults, we compared estimated recoveries based on actual
defaults to actual recoveries.

 This comparison, summarized on the next slide, showed that adjusting for
the effect of originally overestimating defaults, estimated recoveries have
more closely matched actual loan performance over time.

 The cumulative difference between the adjusted estimate of recoveries and
actual recoveries was about $3 million, or about 1 percent of actual
recoveries.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 25

22 Comparison of Originally Estimated

Defaults and Recoveries to Actual Data Percentage by which Adjusted
Estimated Recoveries were more (less) than Actual

Recoveries for Fiscal Years 1992 through 2000 (Cumulatively by Cohort)
Source: GAO analysis based on SBA data. Note: Estimated recoveries were
adjusted to be based upon actual defaults in order to remove the effect of
overestimating defaults. N/ A indicates that there were no actual recoveries
for the cohort, as expected for a cohort in its first year.

$0 $50

$100 $150

$200 $250

1992 1993 1994 1995 1996 1997 1998 1999 2000

Cohort Dollars in millions

Estimated Actual (24%)

N/A (28%) 65% 77%

22% 35%

6% (22%)

The 1998 cohort was the first to be estimated using historical data.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 26

23 Effect of Overestimating the 7( a) Program?s

Subsidy Cost

 Because the 7( a) program is a discretionary credit program,
overestimating the cost can affect the number or size of loans guaranteed,
if the program runs out of budget authority.

 However, according to SBA and OMB, the 7( a) program has typically not
depleted its allocated budget authority and has generally met its demand for
loan guarantees.

 According to SBA, the 7( a) program did run out of budget authority a few
days before the end of fiscal year 1995, preventing SBA from issuing some
loan guarantees. However, SBA issued loan guarantees for those loans the
following fiscal year. Further, for a part of 1997, SBA established a
temporary cap on the size of loans it guaranteed, which limited the amount
of subsidy available per loan.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 27

24 Effect of Overestimating the 7( a) Program?s

Subsidy Cost

 Appropriations for the original 7( a) program subsidy cost, like other
discretionary credit programs, are counted under the discretionary spending
caps and must compete with other discretionary programs for the funding
available under these limits.

 The cumulative result of the overestimates of the subsidy cost of the 7(
a) program is that $958 million of budget authority was not available for
other discretionary programs for fiscal years 1992 through 2000.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 28

25 Causes of Differences

 For the 1992 through 1997 cohorts, 11 the specific reason for the
differences between originally estimated and actual defaults is unclear
because the basis of the estimate is unknown.

 SBA did not begin to use its historical data until 1996, when it
calculated the original subsidy cost estimate for the 1998 cohort.

 According to SBA officials, prior to 1996, subsidy cost estimates were
prepared based on direct consultation with OMB and the basis used for the
default estimates was not documented.

 However, SBA believes one of the reasons for the differences was an
unexpected good economy.

11 Reestimates of the 1992 through 1997 cohorts have accounted for 84
percent of the 7( a) program?s total downward reestimate.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 29

26 Causes of Differences

 The reason for the differences between originally estimated and actual
defaults for the 1998 through 2000 cohorts is that the historical average
default rate used as the basis for the default estimate was greater than
recent loan performance.

 The historical average default rate was higher because loans guaranteed in
fiscal years 1986 through 1990 defaulted at a significantly higher rate than
those for later years.

 SBA attributes the high default rates in fiscal years 1986 through 1990
generally to differences in (1) economic conditions, (2) guarantee
percentages, and (3) underwriting standards.

 The loans in the 1998 through 2000 cohorts are still relatively new and
have not yet reached the typical peak default years, which historically have
been years 3 through 5 after approval.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 30

Implications of Proposed Changes

 In March 2001, SBA submitted a proposal to OMB 12 that discusses using 5
years or 3 years of the most recent actual loan performance - referred to as
the lookback period 13 - as the basis for the 7( a) program default estimate
in order to more closely track with actual loan performance in the future.
SBA recommends the 5 year lookback period.

 This proposal is based on SBA?s analysis that showed that the most recent
years of actuals are more predictive of near- term future loan performance,
notwithstanding a sudden shift in the economy.

12 In the past, SBA has proposed other methods to refine its default
estimates to OMB. According to OMB, SBA has not provided acceptable support
that the alternatives would provide better estimates. 13 For example, under
the 5 year lookback period, the 2002 cohort estimate of year one default
activity would be based

on the average actual first year defaults that occurred for the 1996 through
2000 cohorts and the second year default activity would be based on actual
second year defaults that occurred for the 1995 through 1999 cohorts. 27

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 31

28 Implications of Proposed Changes

 Because the lookback period is shorter, original subsidy cost estimates,
as well as annual reestimates of outstanding cohorts, would be more
sensitive to fluctuations in economic conditions or changes in program
delivery and design.

 The benefit of this approach is that in a continuing stable economy, the
original subsidy cost estimate would be expected to more closely match
actual loan performance and reestimates would therefore be smaller.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 32

29 Implications of Proposed Changes

 However, the risk of this approach is that a sudden downturn in the
economy would be much more likely to result in actual loan performance being
different than estimated and thus could result in larger reestimates.

 If SBA were to implement a shorter lookback period approach, its next
reestimate would likely be large because expectations of future loan
performance of outstanding cohorts would also be impacted by the change.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 33

30 Implications of Proposed Changes

 Under SBA?s current approach, initial estimates of the subsidy rate are
fairly stable because of the longer lookback period, which smoothes out
fluctuations in economic conditions from year to year.

 This approach is based on the concept that, averaging

?good? and ?bad? years is the best way to estimate the effect of uncertain
future economic conditions.

 The benefit of this approach is that it provides a

?cushion? in the event of an unexpected downturn in the economy.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 34

31 Implications of Proposed Changes

 The consequence of this approach is that the ?cushion?

ties up appropriations that could have been available to other discretionary
programs.

 This approach is also more likely to result in continuing annual downward
reestimates in a strong economic environment.

 However, in a less favorable economy, the current approach may result in
original subsidy cost estimates that are closer to actual loan performance
than the proposed 5 year lookback approach.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 35

32 Implications of Proposed Changes

 The following table contrasts the impact of using the current approach, a
5 year lookback, and a 3 year lookback to estimate the subsidy cost of the
fiscal year 2002 cohort.

Estimation Alternatives? Effect on Subsidy Rate and Appropriation for the
Fiscal Year 2002 Cohort

Source: GAO analysis based on SBA data. Note: Estimated appropriation
assumes that all other assumptions remain unchanged.

Default Rate Subsidy Rate Appropriation Current Approach 13. 87% 1.07%
$114,490,000 5 Year Lookback 9. 74% -0.40% -$ 42,800,000 3 Year Lookback 8.
97% -0.61% -$ 65,270,000

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 36

33 Implications of Proposed Changes

 For both the 5 year and 3 year lookback approach, we estimated a negative
subsidy, meaning that the program is estimated to ?make money? for the
federal government.

 We estimated that the 5 year and 3 year lookback would project a negative
subsidy of $43 million and $65 million, respectively, versus a subsidy cost
of $114 million under the current approach.

Enclosure GAO- 01- 1095R SBA's 7( a) Credit Subsidy Estimates Page 37

34 Agency Comments

 SBA generally agreed with the information presented in this briefing. SBA
officials added that they view the proposed change in the default estimation
methodology to be an interim solution. SBA views the long- term solution as
a sophisticated econometric modeling approach.

 Econometric modeling is meant to include any estimated quantitative method
of analysis. It defines key relationships between loan performance and
economic and other indicators.

 SBA has already started work on this type of methodology.

(190027)
*** End of document. ***