Department of Education: Key Aspects of the Federal Direct Loan Program's
Cost Estimates (Letter Report, 01/12/2001, GAO/GAO-01-197).

The Department of Education runs two major federal student loan
programs, the William D. Ford Federal Direct Loan Program (FDLP) and the
Federal Family Education Loan Program (FFELP). Under FDLP, students or
their parents borrow money directly from the federal government through
the schools the students attend. Under FFELP, money is borrowed from
private lenders, and the federal government guarantees repayment if the
borrowers default. GAO investigated concerns about Education's reliance
on estimates to project FDLP costs and a lack of historical information
on which to base those estimates. GAO found that developing a reasonable
estimate of subsidy cost for loan programs is complex. Many assumptions
must be taken into account and projections must be made for the life of
the loans. Because FDLP's subsidy costs are determined largely by
interest rates and since interest rate fluctuations cannot be predicted
with any certainty, it is uncertain that the current trend in negative
subsidy costs for FDLP will continue. In addition, other factors, such
as origination fees paid by borrowers, defaults, subsequent collections
on defaulted loans and timing of loan repayments, also affect the
subsidy cost of FDLP. Although Education is able to estimate origination
fees close to the actual amounts in the financial system, other key cash
flows varied significantly. Also, Education's current model for
estimating FDLP subsidy costs does not directly take into account
certain key factors such as prepayments and consolidations.

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

 REPORTNUM:  GAO-01-197
     TITLE:  Department of Education: Key Aspects of the Federal Direct
	     Loan Program's Cost Estimates
      DATE:  01/12/2001
   SUBJECT:  Higher education
	     Cost analysis
	     Direct loans
	     Loan repayments
	     Interest rates
	     Student loans
	     Subsidies
	     Projections
IDENTIFIER:  William D. Ford Federal Direct Loan Program
	     Federal Family Education 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-01-197

A

Report to Congressional Requesters

January 2001 DEPARTMENT OF EDUCATION

Key Aspects of the Federal Direct Loan Program's Cost Estimates

GAO- 01- 197

Letter 3 Appendixes Appendix I: Estimating Credit Program Costs 48

Appendix II: Objectives, Scope, and Methodology 54 Appendix III: GAO
Contacts and Staff Acknowledgments 59

Glossary 60 Tabl es Table 1: Amounts Borrowed From Treasury to Finance FDLP
11

Table 2: Appropriations Received to Finance FDLP 13 Table 3: Cash Inflows
and Outflows by Fiscal Year 14 Table 4: Percentage Impact of Changes in Key
Assumptions

on Subsidy Costs by Loan Profile 19 Table 5: GAO Estimates of Dollar Impact
of Changes in Key Assumptions on Subsidy Costs by Loan Profile 20

Table 6: Summary of Key Assumptions and Sources of Data 21 Table 7: Total
Estimated and Actual Key Cash Flows for Fiscal Years 1995 Through 1999 25

Table 8: Effect of Fee Reduction on Subsidy Costs 32 Table 9: Comparative
Direct Loan Program Estimates, Per

$100 of Loans 41 Table 10: Primary Types of Data Used in Education's 1999
Cost Study 42

Figures Figure 1: Estimated and Actual Interest Receipts 26 Figure 2:
Estimated and Actual Principal Receipts 27

Figure 3: Estimated and Actual Origination Fees 28 Figure 4: Estimated and
Actual Collections on Defaults 29 Figure 5: FDLP Borrower Rates and Discount
Rates 36 Figure 6: Historical 91- Day Treasury Bill Rates 38

Abbreviations

FASAB Federal Accounting Standards Advisory Board FCRA Federal Credit Reform
Act of 1990 FDLP Federal Direct Loan Program FFELP Federal Family Education
Loan Program NSLDS National Student Loan Data System OMB Office of
Management and Budget OSFA Office of Student Financial Assistance SFFAS
Statement of Federal Financial Accounting Standards

Lett er

January 12, 2001 The Honorable Jim Nussle Chairman, Committee on the Budget
House of Representatives The Honorable Judy Biggert House of Representatives

The Department of Education administers two major federal student loan
programs, the William D. Ford Federal Direct Loan Program (FDLP) and the
Federal Family Education Loan Program (FFELP). The federal government's role
differs significantly between these two programs. Under FDLP, often referred
to as the direct loan program, students or their parents borrow money
directly from the federal government through the

schools the students attend, which include vocational, undergraduate, or
graduate schools. The first FDLP loans were made in the fourth quarter of
fiscal year 1994. Under FFELP, also known as the guaranteed student loan
program, money is borrowed from private lenders, such as banks, and the
federal government guarantees repayment if the borrowers default. FFELP

is the older of the two programs, having started in fiscal year 1966. As of
September 30, 1999, Education reported that the outstanding gross balance of
FDLP was $46.5 billion and the total outstanding loan guarantees for FFELP
were approximately $127 billion. As of September 30, 1999, Education
estimated that it would incur total subsidy costs- the estimated cost of
extending credit over the life of the loans- of $1.6 billion for FDLP and
$12.2 billion for FFELP.

FDLP, established by P. L. 103- 66, was implemented by Education in fiscal
year 1994 as another method of delivering loans to students. FDLP makes it
possible for students and their families to borrow directly from the federal
government through the colleges or other postsecondary institutions the
students attend. In fiscal year 1999, 3.2 million loans totaling $17. 7
billion were disbursed to borrowers through FDLP.

Because of concerns about Education's reliance on estimates to project FDLP
costs and a lack of historical information on which to base those estimates,
you asked us to review how Education develops its cost estimates for the
program, review and analyze actual versus estimated

financial performance, and address factors or circumstances that can
significantly affect Education's ability to develop realistic estimates of
program costs. Specifically you asked us to address nine questions.

1. How much financing has been provided to Education for the direct loan
program through borrowing from Treasury and appropriations received? 2. Have
cash inflows (excluding borrowings from Treasury and borrower

principal repayments) exceeded cash outflows (excluding repayments to
Treasury and loan disbursements)? 3. In Education's calculation of its
subsidy cost estimates for the Federal Direct Loan Program, what are the key
cash flow assumptions, how

sensitive are Education's subsidy costs to changes in these assumptions, and
what data are used to support these assumptions?

4. How closely do Education's subsidy cost estimates and their underlying
assumptions compare to actual loan performance for each loan and to what
extent does Education track differences between its subsidy cost estimates
and actual loan performance for each loan cohort? 5. What effects have
reduced loan origination fees had on subsidy costs,

and how has Education taken account of these changes in its subsidy cost
estimates and reestimates?

6. What effects have increased consolidations had on subsidy costs, and how
has Education taken account of these changes in its subsidy cost estimates
and reestimates? 7. What effect have declining interest rates had on subsidy
costs, and how

has Education taken account of these changes in its subsidy cost estimates
and reestimates? 8. What are the future prospects for the continued negative
subsidy for

the Federal Direct Loan Program? 9. What data did Education use to project
an estimated savings of $4 for every $100 of direct student loans, as it
reported in November 1999?

To respond to your request, we reviewed Education's audited financial
statements and examined the workpapers of Education's independent auditors
for fiscal years 1995 through 1999. We interviewed knowledgeable personnel
from Education's Budget Service and obtained information relevant to your
questions. We compared Education's practices to

(1) federal budgeting and accounting standards- such as Statement of Federal
Financial Accounting Standards (SFFAS) No. 2, Accounting for Direct Loans
and Loan Guarantees, as amended, and Office of Management and Budget (OMB)
Circular A- 11, Preparation and Submission of Budget Estimates- and (2) the
guidance contained in the Federal Financial Accounting and Auditing
Technical Release 3, Preparing and Auditing

Direct Loan and Loan Guarantee Subsidies Under the Federal Credit Reform
Act. 1 Our audit work was conducted in Washington, D. C., from May 2000
through November 2000 in accordance with generally accepted government
auditing standards. Appendix II describes our objectives, scope, and
methodology in detail. We provided the Department of Education copies of a
draft of this report for review and comment. On December 7, 2000, we met
with cognizant Education officials and obtained oral comments on a draft of
this report.

Because of the number and diverse nature of the questions, we are responding
in a question- and- answer format. The next section provides an overview of
some of the key points from our review.

Overview Several circumstances make it difficult to make the best possible
determination of FDLP's financial performance at this time. First, because
FDLP is a relatively new program, it has a short history of repayment
activity and little historical data are available. Second, because Education

lacks historical FDLP data, Education relies heavily on data from the
guaranteed loan program to develop estimates for most key cash flow
assumptions in its FDLP cash flow model, which is used to estimate the
subsidy cost of the program. While this is appropriate for the interim, 1
The Credit Reform Task Force of the Accounting and Auditing Policy Committee
was

formed in order to address key issues surrounding the implementation of the
Federal Credit Reform Act of 1990 and the related federal accounting
standard. This task force developed Technical Release 3, which is expected
to be formally issued by OMB during fiscal year 2001. The purpose of
Technical Release 3 is to provide implementation guidance for

agencies and auditors to prepare, utilize, and report on credit subsidy
estimates. Technical Release 3 does not take precedence over existing
accounting standards and budget guidance.

guaranteed loans may perform differently from FDLP loans and therefore,
Education ultimately will need to use FDLP data. Education plans to phase
out the use of guaranteed loan data as FDLP data become available. Our
ability to answer some of your specific questions was limited because the
needed data were not readily available. For example, Education's cash flow
model and financial systems do not readily provide comparable information on
estimated and actual defaults. Also, Education did not have readily
available performance data by “cohort,” which refers to all the
loans

of a particular loan type for which a subsidy appropriation is provided for
a given fiscal year. For this reason, Education was not able to give us a
comparison of estimated to actual cash flows at the cohort level during the
time frames of this review. Comparisons of estimates and actuals at the
cohort level are key to identifying the causes of disparities, which, in
turn, is key to improving future subsidy cost estimates. Furthermore, there
is little information on the effects of loan consolidations on FDLP subsidy
costs. This is significant because Consolidation loan volume has been
rapidly increasing. Education is taking or plans to take steps to address
these limitations in the future. Because Education has not documented its
previous sensitivity analyses,

we asked Education to perform a limited sensitivity analysis 2 of FDLP
subsidy costs and found that the subsidy calculation is most sensitive to
changes in interest rates. Specifically, the interest rates involved were
the discount rate- generally the rate at which Education borrows money from
the Department of the Treasury to finance its loans- and the borrower

rate. The difference, or spread, between the borrower rate and discount rate
determines the magnitude the change in interest rates has on the FDLP
subsidy cost. Because these rates cannot be readily predicted from year to
year, estimating the subsidy cost of FDLP is very difficult. Therefore, wide
fluctuations in subsidy costs can be expected depending on the extent of
interest rate changes.

Because FDLP is a direct loan program that allows its borrowers to defer
payment until after the borrower leaves school, several years would
typically pass between the time the borrower receives the loan and begins
making repayments. This deferment of principal and interest payments from
borrowers has contributed to the negative cash flow FDLP

2 Sensitivity analysis is a process used to identify which cash flow
assumptions, when adjusted, have the greatest impact on the estimated
subsidy costs.

experienced that totaled about $2 billion as of September 30, 1999. Although
more cash will be received by Education when more borrowers enter repayment,
Education is unable to determine when FDLP will have a

positive cash flow primarily because of uncertainty related to the key cash
flow assumptions. Further, because Education lacks key data on loan
consolidations and default data is not readily available, Education's
ability to predict future cash flows is limited. This further impedes
Education's

ability to estimate when and how much of this negative cash flow will be
recovered.

We are making several recommendations to address the limitations identified
during our review.

Background Education is the primary agency overseeing federal investments in
support of educational programs for U. S. citizens and eligible noncitizens.
In fiscal year 1999, more than 8.1 million students received over $53
billion in federal student financial aid, including loans and grants,
through programs administered by Education. FDLP offers four different loan
types. ? The Federal Direct Stafford Subsidized/ Ford Loan Program (Stafford

Subsidized), available only to students with a demonstrated financial need,
provides loans to undergraduate, graduate, and professional students.
Interest is subsidized by the federal government while the

student is in school, and during the grace, 3 or deferment 4 period. A loan
origination fee is charged to obtain these loans. The borrower rate is
variable and based on the 91- day Treasury bill rate plus an add- on

amount that has ranged from 1. 7 percent to 3.1 percent, with a maximum
borrower rate of 8.25 percent. Education reported that the outstanding
balance of this loan type was $19. 7 billion as of September 30, 1999.

3 A grace period is a 6- month period that begins on the day after a FDLP
borrower ceases to be enrolled as at least a half- time student at an
eligible institution and ends on the day before the repayment period begins.
4 Deferment periods are periods during which the payment of the principal is
postponed. Reasons for deferment include in school status, unemployment, and
economic hardship.

? The Federal Direct Unsubsidized Stafford/ Ford Loan Program (Stafford
Unsubsidized) provides loans to undergraduate, graduate, and professional
students regardless of financial need. The borrower is responsible for
interest that accrues during any period. Interest that accrues while the
student is in school or during the grace period or deferment period is added
to the loan balance. A loan origination fee is charged to obtain these
loans. The borrower rates on these loans are the

same as the borrower rates on Stafford Subsidized loans. Education reported
that the outstanding balance of this loan type was $11. 9 billion as of
September 30, 1999. ? The Federal Direct PLUS Program provides loans to
parents of dependent students. The borrower is responsible for interest that
accrues during any period. A loan origination fee is charged to obtain

these loans. The borrower rate is variable and currently based on the 91-
day Treasury bill rate plus an add- on amount of 3. 1 percent, with a
maximum borrower rate of 9 percent. Education reported that the outstanding
balance of this loan type was $2.8 billion as of September 30, 1999.

? The Federal Direct Consolidation Loan Program (Consolidation loans) allows
borrowers to combine their loans from different federal student loan
programs into a single loan with one monthly payment. 5 After the promissory
note has been signed for the new Consolidation loan, the underlying loan( s)
are paid off. The Higher Education Act Amendments of 1998 (P. L. 105- 244)
provided that for all Direct Consolidation Loan applications received from
February 1, 1999, through June 30, 2003, the borrower rate is a fixed rate
for the life of the loan. The rate is the lesser of the weighted average of
the interest rates on the loans being consolidated, or 8.25 percent, the
current maximum allowable rate.

Borrower rates on previously disbursed Consolidation loans are variable
rates, similar to the other FDLP loan types. Education reported that the
outstanding balance of this loan type was $12.1 billion as of September 30,
1999.

5 Some of the federal student loans that are permitted in an FDLP
consolidation include FDLP, FFELP, and Perkins loans as well as health
professional loans, such as Health Education Assistance and Nursing Student
loans.

Borrowers most commonly repay their FDLP loans using one of four repayment
plans: standard, extended, graduated, or income contingent. 6 These four
options differ by the amount of time allowed to repay loans and

the flexibility of the repayment schedule. With standard repayment,
borrowers make fixed payments of at least $50 a month for up to 10 years.
With extended repayment, they make fixed payments of at least $50 a month
over a period generally ranging from 12 to 30 years, depending on the total
amount borrowed. With graduated repayment, borrowers' payments start out low
and then increase, usually every 2 years; the repayment period generally
ranges from 12 to 30 years, depending on the

total amount borrowed. The income contingent repayment plan is the most
flexible, allowing borrowers to make monthly payments that are based on
adjusted gross income, family size, and the total amount of their
outstanding loans. The Federal Credit Reform Act of 1990 (FCRA) was enacted
to require

agencies to more accurately measure the government's cost of federal loan
programs and to permit better cost comparisons both among credit programs
and between credit and noncredit programs. Prior to the

implementation of FCRA, credit programs were reported in the budget on a
cash basis. Thus, loan guarantees appeared to be free in the budget year,
while direct loans appeared to be as expensive as grants. As a result, costs
were distorted and credit programs could not be compared meaningfully with
other programs and with each other. FCRA and the related accounting
standards and budgetary guidance, together known as credit reform, were
established to more accurately measure the government's costs of federal

credit programs and to permit better comparisons both among credit programs
and between credit and noncredit programs. As part of implementing credit
reform, agencies are required to estimate the net cost of extending credit
over the life of a loan, generally referred to as the subsidy cost, based on
the present value 7 of estimated net cash flows, excluding administrative
costs.

6 A fifth type of repayment plan that is rarely used alternative repayment
is also available to FDLP borrowers. Also, the income contingent repayment
plan is not available to borrowers of PLUS or Consolidation PLUS loans. 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.

Budgeting guidance requires agencies to maintain supporting documentation
for subsidy cost estimates. Further, auditing standards related to estimates
indicate that agency management is responsible for accumulating sufficient
relevant and reliable data on which to base the estimated cash flows. SFFAS
No. 2 states that each credit program should

use a systematic methodology to project expected cash flows into the future.
To accomplish this task, agencies develop cash flow models. A cash flow
model is a computer program that generally uses historical information and
various assumptions including defaults, prepayments, recoveries, and the
timing of these events to estimate future loan performance. Those
assumptions that have the greatest impact on the estimated subsidy cost are
often referred to as the key assumptions. These cash flow models, which
should be based on sound economic, financial, and statistical theory,
identify key factors that affect loan performance. Agencies use this
information to make more informed predictions of future credit performance.
Generally, the data used for these estimates are updated or reestimated
after the fiscal year end to reflect any changes in actual loan performance
since the estimates were prepared, as well as any expected changes in
assumptions related to future loan performance. Appendix I provides a
detailed discussion of estimating credit program costs under credit reform.
The glossary at the end of this report provides a list of commonly used
terms related to credit program budgeting and accounting.

Questions and Answers Question 1 How much financing has been provided to
Education for the direct loan

program through borrowing from Treasury and appropriations received? Amounts
borrowed from Treasury are amounts that Education expects to be repaid by
borrowers in the future. Amounts appropriated are amounts that Education has
estimated it will lose as a cost of extending credit through FDLP. For
fiscal years 1995 through 1999, Education's FDLP has borrowed $59.4 billion
from Treasury to finance the program, repaying $7. 8 billion of that amount.
Table 1 provides an annual accounting of this information. Over the same
period, considering reestimates, Education has

received $688 million in appropriations (see table 2).

Table 1: Amounts Borrowed From Treasury to Finance FDLP

(Dollars in millions)

Outstanding Opening

Borrowings Repayments to

balance at yearend Fiscal year balance from Treasury Treasury

1995 $433 $4,868 $235 $5, 067 1996 5,067 7, 957 669 12, 355 1997 12, 355
11,333 975 22, 713 1998 22, 713 13,669 1,284 35, 098 1999 35, 098 21,571
4,599 52, 070

Total $59,398 $7,762

Source: Department of Education.

Education finances FDLP through a combination of appropriations and
borrowing from Treasury as required by FCRA. For loan programs subject to
the act, agencies are required to estimate the cost of extending or
guaranteeing credit, called the subsidy cost. The subsidy cost is the
present value of disbursements from the government (loan disbursements and
other payments) minus estimated payments to the government (repayments of
principal, interest receipts, fees, and other recoveries or

payments) over the life of the loan. The subsidy cost is generally the
amount that Education estimates will not be repaid by borrowers. This
estimate is financed with appropriated funds and is generally
“reestimated” or updated annually. The portion of Education's
direct loans that Education predicts will ultimately be repaid by borrowers
is financed by borrowing from Treasury and is not considered a cost to the
government because it is expected to be returned to the government in future
years. If the present

value of the estimated cash outflows from the government exceeds the present
value of the estimated cash inflows, there is a positive subsidy or cost to
the government. However, if the present value of the estimated cash

inflows to the government exceeds the present value of the estimated cash
outflows, there is a negative subsidy.

When there is a negative subsidy, a higher level of borrowing from Treasury
occurs than when there is a positive subsidy because Education must borrow
an amount greater than the dollar amount of loans disbursed. This

additional borrowing occurs because Education does not receive any
appropriated funds and therefore experiences a temporary shortfall because,
in addition to disbursing the full loan amount, Education pays the negative
subsidy to its program account. 8 This additional amount of borrowing as
well as the amount of loans disbursed is expected to be repaid by the
borrower, primarily through principal and interest payments,

over the life of the loan. For example, if a hypothetical FDLP loan of $100
had a negative subsidy of $5, the amount of borrowing required would be $5
more than the face value of the loan. Accordingly, Education would borrow a
total of $105 from Treasury. If, however, FDLP had a positive subsidy,
required borrowing from Treasury would be less. For example, if a
hypothetical FDLP loan of $100 had a positive subsidy cost of $5, the
subsidy cost of $5 would be financed with appropriated funds, and the
remaining $95 would be

financed by Treasury borrowings (the amount Education expects to be repaid).
Additionally, Education is required to periodically update or
“reestimate” loan program costs for differences between (1)
estimated loan performance and related cost and (2) the actual program costs
recorded in

the accounting records as well as expected changes in future economic
performance. When program costs are reestimated for loans disbursed in prior
years, the revised estimate can either increase or decrease the original
subsidy estimate. These reestimates can also affect the level of borrowing
and appropriations. Generally, downward reestimates are considered
offsetting receipts, which are netted against the subsequent year's
appropriations, and upward reestimates require additional appropriations. 9

8 For mandatory programs such as FDLP, amounts from negative subsidies and
downward reestimates may be credited directly to the program account, which
is a budget account that receives appropriations to cover the subsidy cost
of a direct loan or loan guarantee and disburses the subsidy cost to the
financing account. The financing account is explained in the glossary at the
end of this report. 9 A permanent indefinite appropriation is available for
upward reestimates.

Table 2 shows FDLP's original subsidy estimates and reestimates for the 1995
through 1999 cohorts. For example, the 1997 cohort column in table 2 shows
that this group of loans was originally estimated to have a positive subsidy
of $336 million. Since then, Education has reestimated the cost of the 1997
cohort twice, increasing its cost by $80 million in fiscal year 1998 and
decreasing its cost by $69 million in fiscal year 1999. Therefore, as of
fiscal year 1999, the estimated net cost of the 1997 cohort was a positive

subsidy of $347 million. In contrast, the fiscal year 1999 column shows that
the 1999 cohort was the first cohort originally estimated to have a negative
subsidy. For fiscal years 1995 through 1999, Education's FDLP estimates and
reestimates for all cohorts show a total positive subsidy of $688 million,
and therefore Education has received net appropriations totaling this
amount.

Table 2: Appropriations Received to Finance FDLP

(Dollars in millions)

Cohort 1995 1996 1997 1998 1999 Total

Original subsidy $490 $237 $336 $213 $( 378) $898 estimate Reestimate 1 (6)
128 80 (129)

Reestimate 2 (157) 153 (69) Reestimate 3 (58) (71) Reestimate 4 (81)

Net appropriations $188 $447 $347 $84 $( 378) $688

Note: The figures for reestimates indicate the amount of increase or
decrease to the original subsidy estimate. Source: Department of Education.

Because FDLP is a relatively new program, there is limited historical data
to predict future borrower behavior. Additionally, the future estimated cost
of this program, as explained in questions 3 and 7, is especially sensitive
to changes in interest rates. Therefore, fluctuations such as those shown in
table 2 are not unexpected and are likely to continue in the future.

Question 2 Have cash inflows (excluding borrowings from Treasury and
borrower principal repayments) exceeded cash outflows (excluding repayments
to Treasury and loan disbursements)?

Loan origination fees and interest receipts from borrowers are the primary
sources of cash inflows for FDLP. Net interest payments to Treasury on
borrowed funds to finance the loans disbursed are the primary source of cash
outflows. 10 As shown in table 3, for fiscal years 1995 through 1999, total
cash outflows exceeded total cash inflows by about $2 billion because the
interest receipts from borrowers and origination fees were less than the
amount of interest Education had to pay to Treasury. Inflows exceeded
outflows only in fiscal years 1995 and 1996.

Table 3: Cash Inflows and Outflows by Fiscal Year

(Dollars in millions)

Description 1995 1996 1997 1998 1999 Total

Loan origination fees $85 $318 $352 $382 $387 $1, 524

Interest receipts from 14 113 300 606 1, 067 borrowers 2,100

Net interest payment on (86) (348) (1, 180) (1, 686) (2, 395) Treasury
borrowings (5, 695)

Net cash inflows/( outflows) $13 $83 $( 528) $( 698) $( 941) $( 2, 071)

Source: Department of Education.

The $2 billion negative cash flow for FDLP is at least partially due to a
timing difference in the cash flows. Education is required to make interest
payments to Treasury, even if the borrower is not currently making interest
payments to Education. As of September 30, 1999, 46 percent of the loan

portfolio was in a grace or deferment status. As a result, Education
subsidizes or generally accrues this interest. However, Education must repay
the interest on borrowings from Treasury even though it does not

expect to receive interest payments from borrowers until sometime in the
future. This accrued interest can be substantial-$ 2.3 billion as of
September 30, 1999.

Education is unable to determine when FDLP will have a positive cash flow
primarily because of uncertainty related to the key cash flow assumptions.
As discussed in question 3, the estimated cost of FDLP is sensitive to
changes in interest rates and other factors that will affect the program's
cash flows. In addition, reductions in origination fees, as occurred in
fiscal 10 Cash inflows and outflows discussed in this answer are defined as
stated in the question.

year 1999, discussed in question 5, will also have an impact on whether FDLP
has an overall negative or positive cash flow in the future. Further, cash
flows for FDLP can be affected by changes in macroeconomic conditions, such
as unemployment rates and inflation. Question 3 In Education's calculation
of its subsidy cost estimates for the Federal Direct Loan Program, what are
the key cash flow assumptions, how

sensitive are Education's subsidy costs to changes in these assumptions, and
what data are used to support these assumptions? An effective approach to
identifying key cash flow assumptions is to perform a detailed analysis of
all cash flow assumptions- called a sensitivity analysis 11 -in order to
determine which assumptions have the

greatest impact on the estimated cost of FDLP. Education told us that it
performs informal analyses of the cash flow assumptions that result in about
90 percent of the change in subsidy costs each year. However, Education did
not provide any supporting documentation for this analysis. Further,
Education told us that it has not performed a sensitivity analysis of all
cash flow assumptions in its model. As this type of analysis would be

extremely time- consuming, we requested that Education perform and document
a limited sensitivity analysis as a basis on which to answer this question.
Based on this limited sensitivity analysis, there were seven key cash flow
assumptions that when adjusted, had a significant impact on the estimated
cost of the loan program. These assumptions were discount rates, borrower
rates, loan maturity, collections on previously defaulted loans, defaults,
origination fees, and when repayments begin. 12 The analysis showed that
FDLP's subsidy cost was most sensitive to changes in the

discount rate and borrower rate. While some of the data supporting these key
assumptions are provided by other agencies or specified by law, Education
supported other assumptions by using a combination of

guaranteed loan program and economic data, a reasonable approach, since the
direct loan program is relatively new and limited historical data are
available.

11 Currently, there is no accounting or budgeting guidance that requires
agencies to perform a sensitivity analysis. However, Technical Release 3
encourages agencies to perform this analysis.

12 Additional key cash flow assumptions may be identified once Education
performs a more thorough sensitivity analysis.

Key Cash Flow Assumptions To ensure that all key assumptions have been
identified, and to determine how sensitive Education's subsidy cost
estimates are to changes in key assumptions, Education would have to conduct
a thorough sensitivity analysis. According to Technical Release 3, one
approach to perform such

an analysis is to individually adjust each assumption by a fixed proportion
(e. g., increased and decreased by 10 percent) and run the revised cash
flows through the OMB Credit Subsidy Calculator 13 to determine the
assumption's effect on the estimated subsidy cost. Timing assumptions for
when defaults and collections occur and when repayments begin should also be
adjusted in a systematic manner. Those assumptions that when adjusted,
caused the largest change in the subsidy cost are determined to be the key
cash flow assumptions.

Education budget staff told us that they perform analyses of the cash flow
assumptions that result in about 90 percent of the change in subsidy cost
each year when they prepare budget estimates and reestimates. However,

they do not maintain documentation of these analyses. Education has also
done sensitivity analysis on the larger guaranteed loan program, which
Education uses to help identify the key assumptions for the FDLP. However,
because Education's cash flow model has a large number 14 of assumptions,
there is no assurance that all key assumptions have been identified through
the informal analyses that Education performed for

FDLP. Because a formal analysis of all cash flow assumptions would take a
significant amount of time, we asked Education to perform and document a
limited sensitivity analysis of the assumptions it believed to be key and
added two other assumptions related to the largest loan types, risk groups,
and repayment options that we felt might also be key.

13 To provide a consistent, common approach to calculate the present value
of credit program costs, OMB developed the Credit Subsidy Calculator,
formerly known as the OMB Credit Subsidy Model, a computer software program
that calculates a subsidy rate based on agency- generated estimates of cash
flows to and from the government. It also calculates the portions of the
subsidy cost attributable to defaults, interest, fees, and other cash flows.
14 Education told us that its cash flow model has over 1, 900 assumptions
because it models cash flows for all 56 loan profiles- the type of loan, the
type of school the student attends,

and in some cases the year of schooling for the student and the repayment
option selected- separately. Therefore, in the cash flow model, each loan
profile has its own set of assumptions. Although the type of assumptions is
generally the same for each loan profile, the value of these assumptions
often differ.

Based on the results of the limited sensitivity analysis, we determined that
seven of the nine cash flow assumptions tested were key. 15 These
assumptions follow. ? Discount rate this rate is used to calculate the
present value of the expected future cash flows of the loan program and the
interest portion of the subsidy cost. 16 This rate is generally the same
rate at which

agencies borrow funds from Treasury. ? Borrower rate the interest rate
borrowers pay Education for their loans. This rate is based on the 91- day
Treasury bill plus various add- on

amounts that range from 1.7 percent to 3.1 percent with a maximum borrower
rate of 8.25 percent or 9.0 percent depending on loan type. ? Loan maturity
the time it takes for a loan to be paid in full. Loan maturity varies
depending on loan amount and repayment option

selected by the borrower. Generally, borrowers have from 10 to 30 years to
repay their loans. 17 ? Collection rate the percentage of defaulted loan
amounts subsequently

recovered through Education's collection process. ? Default rate the
percentage of principal that will not be paid because

of borrower defaults. ? Origination fee the fee borrowers pay to Education
to obtain a loan. ? Beginning repayment the percentage of loans beginning to
make principal and interest repayments each quarter. 18 As a result of the
limited sensitivity analysis, two of the additional

assumptions that we requested be included in the analysis were identified as
key assumptions- loan maturity and origination fees. Loan maturity is
important because it sets the amount of time borrowers are expected to take
to repay their loans and, accordingly, the number of years Education
estimates that it will receive interest payments from borrowers. The
origination fee assumption is important because it determines the amount of
fee receipts Education will receive. There could also be other key
assumptions that will not be identified until Education completes a 15 Each
of the assumptions identified as key produced a change of at least 2 percent
and

$13 million in the estimated cost of any single loan profile tested. 16 The
subsidy cost is calculated based on four portions attributed to interest,
defaults, fees, and other cash flows.

17 Repayment begins at the conclusion of any deferment period. 18
Education's cash flow model estimates cash flows by fiscal year quarter.

thorough sensitivity analysis. Identification of key assumptions is
important to ensure proper monitoring of those assumptions and to adjust
future subsidy estimates for changes in assumptions. Sensitivity of Subsidy
to Changes Tables 4 and 5 summarize the results of the sensitivity analysis
for seven of in Key Assumptions

the nine cash flow assumptions tested, which entailed adjusting each
assumption by a set amount to determine the impact on the subsidy cost. For
borrower and discount rates, loan maturity, loan origination fee, and
default and collection rates, this adjustment involved increasing and
decreasing by 10 percent the values currently in the cash flow model. For
the assumption related to timing- the beginning repayment assumption- the
adjustment involved was an annual acceleration of 5 percent to the amount of
loans beginning repayment in the first 5 years of the loan term. While the
tables show the impact of decreasing the assumptions, similar results were
obtained by increasing the assumptions. Because changes in two of the nine
cash flow assumptions tested had very little impact on the

overall subsidy cost, they were not determined to be key and were excluded
from the table. 19 Table 4 presents the results of the analysis in terms of
the percentage change in the subsidy cost of each loan profile, which
encompasses the type of loan, the type of school the student attends, and in
some cases the year of schooling for the student and the repayment option
selected. 20 Generally, the higher the percentage, regardless of whether it
was positive or negative, the more sensitive the subsidy cost was

to change in this assumption. The loan profiles are as follows. ? Loan
Profile 1 Represents loans to freshmen and sophomore students attending 4-
year schools who have obtained Stafford Subsidized loans

and chose the standard repayment option. ? Loan Profile 2 Represents loans
to junior and senior students

attending 4- year schools who have obtained Stafford Unsubsidized loans and
chose the standard repayment option. ? Loan Profile 3 Represents loans to
junior and senior students

attending 4- year schools who have obtained Stafford Unsubsidized loans and
chose the graduated repayment option. 19 The two assumptions that were
determined not to be key assumptions were the timing of defaults and
collections. 20 The sensitivity analysis that Education performed was based
on Education's fiscal year 2001 mid- session review cash flow model and
assumptions.

? Loan Profile 4 Represents PLUS loans to parents of freshmen and sophomore
students attending 4- year schools who chose the standard repayment option.
? Loan Profile 5 Represents Consolidation loans to borrowers who

chose the extended repayment option. ? Loan Profile 6 Represents
Consolidation loans to borrowers who

chose the income contingent repayment option.

Table 4: Percentage Impact of Changes in Key Assumptions on Subsidy Costs by
Loan Profile Loan

Loan Loan

Loan Loan

Loan Assumption profile 1 profile 2 profile 3 profile 4 profile 5 profile 6

Discount rate -23. 0% -33. 6% -37.2% -36.4% -101. 9% -21. 4% Borrower rate
10. 5% 28. 7% 29.6% 25.8% 80. 1% 20.9% Loan maturity 3. 4% 5. 8% 5.0% 13.9%
19. 5% Not

applicable a Collection rate 5. 4% 5. 2% 3.1% 4.0% 18. 6% Not

applicable b Default rate -4. 6% -2. 8% -1.9% -5.8% -18.0% Not

applicable c Origination f ee2. 1% 2. 2% 1. 2% 3. 9% Not

Not applicable d applicable d

Beginning -4. 3% -. 4% -. 5% -8.7% Not

Not repayment applicable e applicable e a According to Education, the income
contingent repayment option does not have a specific loan maturity
assumption due to the nature of the income contingent repayment option. b
According to Education, no collections are estimated to be received on
defaulted loan amounts because borrower defaults are estimated to be
insignificant.

c The default rate assumption for loan profile 6 was not tested as part of
the sensitivity analysis because, according to Education, defaults are an
insignificant portion of the subsidy cost compared to the other subsidy
components. This is because borrower repayments are based on borrowers'
financial ability to repay loans. d Consolidation loans do not charge an
origination fee.

e Repayments for Consolidation loans are considered to begin in the first
year. Source: Department of Education.

Table 5 presents the estimated dollar impact on the subsidy cost of each
loan profile for the fiscal years 1995 through 1999 cohorts based on the
results of the sensitivity analysis. These loan profiles represent $16.7
billion

of FDLP loans disbursed during that time.

Table 5: GAO Estimates of Dollar Impact of Changes in Key Assumptions on
Subsidy Costs by Loan Profile

(Dollars in millions)

Loan Loan

Loan Loan

Loan Loan Assumption profile 1 profile 2 profile 3 profile 4 profile 5
profile 6

Discount rate $( 149.6) $( 159.3) $( 103. 1) $( 65. 7) $( 150. 0) $( 153.2)
Borrower rate 68. 4 136. 0 81.9 46.5 118. 0 150.0 Loan maturity 22. 1 27. 5
13. 8 25. 0 28. 7 Not

applicable a Collection rate 35. 4 24. 4 8. 6 7. 2 27. 4 Not

applicable b Default rate (30.2) (13.2) (5. 3) (10. 4) (26. 5) Not

applicable c Origination fee 13. 7 10. 5 3. 2 7.0 Not

Not applicable d applicable d

Beginning (27. 8) (1. 7) (1.4) (15.8) Not

Not repayment applicable e applicable e a According to Education, the income
contingent repayment option does not have a specific loan maturity
assumption due to the nature of the income contingent repayment option. b
According to Education, no collections are estimated to be received on
defaulted loan amounts because borrower defaults are estimated to be
insignificant.

c The default rate assumption for loan profile 6 was not tested as part of
the sensitivity analysis because, according to Education, defaults are an
insignificant portion of the subsidy cost compared to the other subsidy
components. This is because borrower repayments are based on borrowers'
financial ability to repay loans. d Consolidation loans do not charge an
origination fee.

e Repayments for Consolidation loans are considered to begin in the first
year.

Based on results of the analysis in tables 4 and 5, the estimated cost of
FDLP was clearly most sensitive to changes in the discount rate and the
borrower rate. Loan maturity also showed a relatively high level of
sensitivity for all six loan profile costs. Tables 4 and 5 further
demonstrate that the impact of changing these assumptions differs among loan
profiles.

For example, the subsidy costs of all six loan profiles showed a large
degree of sensitivity to changes in the discount rate and the borrower rate,
indicating that changes in these assumptions would significantly affect the
estimated cost of FDLP, with the largest effect on a percentage basis for
loan profile 5 Consolidation loans with the extended repayment option. This
would likely be the case because these loans begin repayment in the first
year and generally have longer repayment periods, thus magnifying the impact
of interest changes. It is especially important to monitor assumptions
displaying this high level of sensitivity because even a small

change in them can have a significant impact on the estimated cost of the
loan program. Data Supporting Key Cash Flow Table 6 summarizes the sources
of data Education used to support the Assumptions

seven key cash flow assumptions identified in the sensitivity analysis.

Table 6: Summary of Key Assumptions and Sources of Data Assumption Source of
information

Discount rate This rate is provided by OMB each year for use governmentwide.
a The discount rate used for each cohort is fixed and determined by the
interest rates prevailing during the period that the cohort's loans were
disbursed. Borrower rate The actual borrower rate is primarily based on the
91- day Treasury bill from the last Treasury Marketable Securities Auction
in May of

each year. Projections of future borrower rates needed in the cash flow
model to estimate FDLP subsidy costs are based on OMB economic assumptions
related to 91- day Treasury bill rates. Loan maturity The maximum allowable
loan maturity is set by statute and varies depending on loan amount and
repayment option. For estimating FDLP subsidy costs, the maximum allowable
loan maturity is decreased based on data from FFELP to reflect the average
length of time it historically took borrowers to fully repay their loans.
This decrease is calculated to account for prepayments or consolidations
that fully pay off a loan balance.

Origination fee The fee amount is specified by statute but has been adjusted
by the Secretary. Default rate

A contractor prepared study that includes (1) historical data for Collection
rate

FFELP, (2) the limited FDLP historical data that were available, and
Beginning (3) economic data related to inflation and unemployment. b
repayment a OMB prepares various economic forecasts that agencies use when
preparing their budget estimates. Generally, these data include estimates of
various short and long- term interest rates, unemployment rates, and
inflation rates.

b This January 1999 study proposes an approach to modeling, among other
things, loan repayments, defaults, and collections on defaulted loans using
data from Education's National Student Loan Data System (NSLDS), which
primarily contains data from FFELP. In this study, the contractors also
considered OMB's economic forecasts. As shown in table 6, for two of the
seven key cash flow assumptions, data sources are provided by other
agencies. Specifically, the borrower rate and

discount rate are generally provided by OMB and updated based on actual
Treasury interest rates, or set by the 91- day Treasury Bill rate from the
last auction in May conducted by Treasury. These rates, the most significant
of

the key assumptions, are determined externally and are outside of
Education's control. For most of the key cash flow assumptions in our
analysis, Education used FFELP data because they were the best available
data. 21 SFFAS No. 2 states that agencies should use the historical
experience of the loan program when estimating future loan performance.
However, since FDLP has only existed since 1994, and Education estimates
that average loan maturities range from 9 to 27 years, Education lacks
adequate historical data to

estimate future performance of the loan program. According to Technical
Release 3, agencies may use the experience of other federal or private
sector loan programs when estimating the cost of new loan programs that lack
adequate historical data. These data, often referred to as proxy data,
should be an interim step to gathering the appropriate historical data upon
which to base future estimates of loan performance. Education officials told
us that Education is currently accumulating the actual cash flow data for
the direct loan program and plans to continue phasing out the use of

proxy data in the future. Without performing a more thorough sensitivity
analysis, Education may not identify all key assumptions in its FDLP cash
flow model. Knowledge of these key assumptions would provide management with
the ability to more efficiently monitor the economic trends and cash flow
assumptions that most affect the loan program's financial performance and,
accordingly, to prepare reasonable estimates of the program's cost. While
some of the changes in assumptions particularly those related to interest
rates occur outside Education's control, understanding the impact that
changes in

assumptions have on program costs also would provide management with a tool
to help predict the impact of certain policy changes on the cost of the
program. Question 4 How closely do Education's subsidy cost estimates and
their underlying

assumptions compare to actual loan performance for each loan and to what
extent does Education track differences between its subsidy cost estimates
and actual loan performance for each loan cohort? 21 These assumptions were
reviewed by Education's independent public accountant as part of the fiscal
year 1999 financial statement audit and found to be reasonable in all
material respects.

Prior to this request, Education had not done a formal documented analysis
comparing estimated subsidy costs to actual loan performance for FDLP.
Typically, such an analysis would entail comparing estimated cash flows
included in the cash flow model to actual cash flows recorded in the
agency's financial systems. However, as discussed below, actual cash flow
data from Education's financial systems were not totally comparable to the
data used in the cash flow model. While we were able to determine
differences between estimated and actual cash flows for certain of the key
assumptions, sufficient detailed information was not available to assess the
reasons for most of the differences. Based on our analysis, some significant
differences between the estimated and actual cash flows were noted.

Although Education could not identify the specific reasons for some of these
fluctuations, Education updates its assumptions for actual interest rates
and loan performance when calculating reestimates.

Comparing Estimated to Actual Education's analysis of estimated and actual
loan performance for FDLP, Cash Flows

prepared at our request, compared estimated to actual cash flows related to
five of the seven key cash flow assumptions identified in question 3- the
borrower rate, loan maturity, beginning repayment assumption, 22 origination
fees, and collections on defaulted loans. 23 The comparison did

not include any analysis of defaults because Education was unable to readily
provide comparable data on either estimated or actual defaults. Due to the
nature of direct loan programs, Education's FDLP cash flow model estimates
principal and interest payments that will be missed in a given fiscal year
as a result of a default, while Education's financial systems do not
specifically recognize this “absence of cash flow.” Rather, the
financial systems report defaults as entire loan amounts that are written
off in a

given fiscal year. Further, the overall analysis was limited by the fact
that readily available data in the financial systems were not totally
comparable to the data available in the cash flow model. Specifically,
Education's financial systems lacked readily available data at the cohort
and loan profile level. Education therefore used fiscal year totals from the
financial systems in its analysis. Appropriately performing an analysis of
estimated

to actual cash flows would require having readily available actual data as
22 In order to produce the estimated cash flows for interest and principal
receipts, Education considers the borrower rate, loan maturity, and
beginning repayment assumptions. 23 The discount rate assumption was not
included because it does not directly affect borrower- related cash flows.
Rather, this assumption is used to calculate the present value of the cash
flows and the financing component of the subsidy cost.

captured in the cash flow model by cohort, key cash flow assumption, and
loan profile. Although agencies are not required to compare estimated cash
flows to

actual cash flows on a cohort basis, such an approach would provide a more
meaningful analysis than comparing fiscal year totals. According to
Education, its approach is consistent with standard credit reform practice
in which costs for all loan cohorts are reestimated each year using the
latest cash flow model and assumptions. However, Education's budget
officials have acknowledged that their analysis has certain limitations. For
example, the difference between estimated and actual loan performance could
be understated because of offsetting differences among different cohorts.
Further, because Education's analysis compared loan performance

in total, variances in loan performance within individual cohorts may become
minimized. These variances may indicate anomalies or trends that were not
expected when the credit subsidy estimate was originally calculated.

Because we were unable to analyze specific cohorts included in Education's
analysis, we were unable to determine whether, over time, estimated cash
flows became more predictive of actual cash flows. Education officials told
us that they are currently working to obtain a subsidiary ledger that will
provide readily available data that are comparable to data in the cash flow
model to allow for a comparison of estimated cash flows to actual cash flows
on a cohort level.

Even though cohort- level data were not available, we were able to analyze
estimated cash flows and actual cash flows on an overall basis for certain
key cash flow assumptions. As shown in table 7, and figures 1 through 4,
some of Education's estimated cash flows varied significantly from actual
cash flows in total and by fiscal year.

Table 7: Total Estimated and Actual Key Cash Flows for Fiscal Years 1995
Through 1999

(Dollars in millions)

Actual cash Estimated

Difference in Percentage Key cash flows flows cash flows dollars difference

Interest receipts $2, 100 $3,660 $( 1, 560) (74.3)% Principal receipts 3,068
2, 676 392 12.8% Origination fees 1,524 1, 611 (87) (5.7)% Default
collections 81 136 (55) (67.9)%

Total $6,773 $8,083 $( 1, 310) (19.3)%

Source: Department of Education.

For three of the four key cash flows- interest receipts, origination fees
and default collections- included in this comparison, actual cash flows were
less than the amount Education estimated. As shown in table 7, from fiscal
years 1995 through 1999, the largest variance occurred between

Education's estimated and actual interest receipts. In total, Education
received about $1.6 billion less than expected during this 5- year period.
In contrast, Education received about $392 million more principal receipts
during the same period. Of the four key cash flows included in table 7,

Education's estimated origination fees had the least amount of a percentage
variance. From fiscal years 1995 through 1999, actual origination fees were
$87 million, nearly 6 percent less than estimated. In addition to
significant variances in total, some of Education's estimated to actual cash
flows varied significantly within individual fiscal years. For example, as
shown by figures 1 and 4, significant variances occurred between the
estimated and actual amounts of both interest receipts and collections on
defaulted loans in fiscal years 1998 and 1999. In contrast, as shown in
figure 3, with the exception of those in fiscal year 1995, Education's
estimates of origination fees were relatively close to the actual amounts
received in all fiscal years. During fiscal years 1996 through 1999,
differences between estimated and actual origination fees varied from about
1 percent to about 6 percent.

Figure 1: Estimated and Actual Interest Receipts

(Dollars in millions) $2, 500 $1, 921 $2, 000

$1, 500 $1, 045

$1, 067 $1, 000

$606 $492 $500

$300 $32

$169 $113 $15 $0

1995 1996 1997 1998 1999 Estimated Actual

Source: Department of Education.

According to Education, the main reason for the significant difference
between estimated and actual interest receipts is the way its cash flow
model handles loan consolidations. Typically, the original loans that are
consolidated into a new loan would be treated as prepayments, and

estimated future cash flows from these underlying loans should be eliminated
in the cash flow model. However, Education's cash flow model does not adjust
for prepayments. Education currently compensates for this by shortening the
loan maturity in an attempt to reflect the consolidation or prepayment of
the original underlying loans. However, this approach may

misstate the timing and characterization of cash flows reported annually.
For example, when borrowers consolidate their loans, accrued interest on the
original loans is added to the principal balance for some loan types, while
the borrower is in school and in other deferment situations. When borrowers
repay their loans, some of the payment for accrued interest is shown in the
accounting records as payments of principal. According to

Education, this helps explain the differences depicted in figure 2 where
more principal was received than estimated for 4 of the 5 years included in
our review. However, because Education was unable to provide the supporting
data for this explanation, we were unable to verify whether the way
consolidations are modeled is (1) truly the primary cause of the

significant difference between the estimated and actual interest receipts
from borrowers and (2) a key factor in differences in principal receipts.

Figure 2: Estimated and Actual Principal Receipts

(Dollars in millions)

$1, 800 $1, 584. 9 $1, 481. 0

$1, 500 $1, 200

$94 0. 7 $900

$75 0. 1 $600

$32 6. 1 $29 9. 1

$300 $21 3. 8

$10 0. 8 $18 .0

$29 .5 $0

1995 1996 1997 1998 1999 Estimated Actual

Source: Department of Education.

According to Education's budget staff, they are analyzing the method used to
allocate borrower repayments between principal and interest, and they
acknowledged that they are not totally comfortable with the current split.
In addition, as discussed in question 6, Education has been working to
improve its modeling of consolidations and plans to develop a different cash
flow model that will allow Education to model and track cash flows at the
individual loan level. Education's budget staff told us that they believe

this new cash flow model will address most of the problems they face in
modeling consolidations.

Figure 3: Estimated and Actual Origination Fees (Dollars in millions)

$500 $397. 6 $392. 5 $373. 6

$382. 0 $386. 6 $400

$352. 4 $319. 7

$317. 5 $300

$200 $128. 0

$85.3 $100

$0 1995 1996 1997 1998 1999

Estimated Actual

Source: Department of Education.

The largest difference between estimated and actual origination fees
occurred in fiscal year 1995. According to an Education budget official,
this difference was due to a reporting anomaly that caused Education to
underreport the amount of actual origination fee data. Because Education

was unable to provide any supporting documentation for this explanation, we
were unable to verify whether this was the actual cause of the difference.

Figure 4: Estimated and Actual Collections on Defaults

(Dollars in millions)

$100 $89.1

$80 $64.3

$60 $36.8 $40

$15.4 $20

$8. 5 $0. 1

$0. 0 $1. 2 $0. 0 $1. 0 $0

1995 1996 1997 1998 1999 Estimated Actual

Source: Department of Education.

Figure 4 shows that Education's actual collections on defaulted loans were
less than estimated collections. However, because Education's cash flow
model estimates collections as a percentage of the amount of loans that
default, and we did not receive any information on defaults, neither we nor
Education are able to determine the underlying cause of the difference.

According to SFFAS No. 2, for credit program managers, information on
estimated default losses and related liabilities, when recognized promptly,
can be an important tool in evaluating credit program performance. This

information can help determine a credit program's overall financial
condition and identify its financing needs. Tracking Differences Between

Education prepared reestimates that accounted for, in aggregate, the Actuals
and Estimates

differences between estimated and actual loan performance. However, because
it lacked data captured by loan profile, cohort, and key assumption,
Education was limited in its ability to identify the underlying causes of
amounts reestimated. See question 1 for a discussion of reestimates for
fiscal years 1995 through 1999.

Prior to this review, most of Education's analysis of estimated to actual
loan performance had been performed for FFELP, rather than for FDLP,

because the guaranteed program is significantly larger than the direct loan
program and historical data supporting the direct loan program estimates was
limited. In using FFELP data, Education officials believed that the two loan
programs' performance would be similar. However, up until 1993, FFELP only
offered borrowers the standard repayment option and currently only two of
FFELP's three repayment plans are similar to those offered under FDLP.
Therefore, FFELP historical data may not prove very predictive of FDLP,
which offers primarily four repayment options. These repayment options would
likely affect the timing and amount of cash flows; however, under existing
guidance, Education may use FFELP data as a

proxy for actual historical data to support some of the key cash flow
assumptions for FDLP, as discussed in question 3. Without a separate
analysis specific to FDLP, Education has limited information about how

well its estimates for FDLP track with actual cash flows. Based on the
information provided by Education for fiscal years 1995 through 1999, total
actual cash inflows were less than estimates for three of the four key cash
flows. Most notably, a significant difference exists

between the estimated and actual amount of interest receipts Education
receives from borrowers. While Education officials provided an explanation,
supporting evidence was not provided to corroborate their explanation. Even
though differences between estimated and actual cash flows are expected, and
the reestimation process allows Education an

opportunity to adjust its estimates of future cash flows based on actual
experience, better understanding the causes of significant variances would
help Education more effectively estimate FDLP costs. However, without the
cohort, loan type, and cash flow assumption- level data, Education's ability
to assess whether its cash flow model is reasonably predicting borrower
behavior is limited. As a result, Education lacks critical information
necessary to update future cash flow models. In addition,

Education's inability to provide an analysis of defaults, one of the key
cash flow assumptions, further impedes Education's ability to effectively
predict future cash flows.

Question 5 What effect have reduced loan origination fees had on subsidy
costs, and how has Education taken account of these changes in its subsidy
cost estimates and reestimates?

In August 1999, Education reduced its origination fees for FDLP student
loans from 4 percent to 3 percent. 24 According to Education, this reduction
was done in order to ensure that both FDLP and FFELP borrowers receive the
same terms, conditions, and benefits. As a result of the fee reduction,

Education's subsidy cost estimates for the fiscal year 2001 25 cohort show
an increase of approximately $93 million, or 23 percent, compared to what
would have been estimated with the 4 percent fee. However, Education
officials reported 26 that they believed that the overall effect would be
cost neutral when considered in light of the higher subsidy costs associated
with guaranteeing loans under FFELP. Since the fee reduction occurred late
in the fiscal year, and thus applied to a limited amount of the fiscal year
1999 loan volume, Education did not take account of the fee reduction in its
reestimates prepared in December

1999. However, in the President's Budget for fiscal year 2001, subsidy
estimates reflect the fee reduction, and Education plans to continue
accounting for the change in origination fees, in accordance with

applicable guidance for federal credit agencies. Effect of Loan Origination
Fee

Education reduced the student loan origination fee from 4 percent to 3
Reduction

percent for the Stafford Subsidized and Stafford Unsubsidized loan types in
August 1999, which resulted in increased subsidy costs for these loan types
of approximately $55 million and $38 million, or 13 percent and 6 percent,
respectively, in the fiscal year 2001 cohort estimate. This amounted to a
$93 million, or a 23 percent, increase in the overall FDLP subsidy cost

estimate for the fiscal year 2001 cohort, compared to what it would have 24
GAO issued an opinion (B- 283717, September 29, 1999) regarding the
Secretary's authority to enact regulations authorizing origination fee
reductions in FDLP. Education asserted that it was authorized to reduce loan
origination fees in FDLP commensurate with those

provided in FFELP. Education relied upon the Higher Education Act of 1965,
as amended, which requires that both FDLP borrowers and FFELP borrowers
receive the same terms, conditions, and benefits on their loans, unless
otherwise specified. GAO concluded that the regulations conflict with a
statutory requirement that Education charge a 4 percent origination fee.
Several of the Secretary's actions, including reducing loan origination fees
in FDLP, have been challenged in a lawsuit filed by lenders participating in
FFELP. Student Loan Finance Corp. et. al v. Riley, C. A. 00- 2660, (U. S.
District Court for the District of Columbia filed November 3, 2000). This
report is not intended to nor does it express an opinion on the issues in
the lawsuit.

25 This estimate was prepared in December 1999 for the fiscal year 2001
President's Budget. 26 Cost of the 1999 Reduction in Direct Loan Fees,
Budget Service, U. S. Department of Education, October 7, 1999.

been assuming the same loan volumes. The fee reduction did not apply to the
PLUS loan type's origination fee, which remained at 4 percent, or the
Consolidation loan type, which does not charge an origination fee to
borrowers.

Since the overall FDLP subsidy cost is a weighted average determined by the
subsidy costs of the four FDLP loan types and their loan volumes, the
increase in the overall FDLP subsidy cost depends on the loan amounts made
for each loan type known as the mix of loans. Table 8 summarizes the
increases to FDLP subsidy cost estimates for each loan type due to the fee
reduction, as well as the estimated mix of loans in fiscal year 2001.

Table 8: Effect of Fee Reduction on Subsidy Costs Increase in subsidy
Increase in subsidy

costs based on Estimated costs per $100 in

fiscal year 2001 fiscal year 2001 direct loans

loan mix loan mix

Stafford subsidized $1.04 $55 million 35% Stafford unsubsidized 0. 94 38
million 26% PLUS 0.00 - 8% Consolidation 0. 00 - 31%

Overall FDLP $0.60 $93 million 100%

Source: Department of Education.

In their report, Cost of the 1999 Reduction in Direct Loan Fees, Education
officials recognized that the fee reduction would increase the cost for
FDLP. However, they believed that the increase would be offset by the
ability to attract borrowers to FDLP who might otherwise obtain loans from
the more costly FFELP whose lenders, according to Education

officials, were offering interest and fee discounts to attract borrowers.
For the fiscal year 2001 cohorts, FDLP's cost was a net inflow of about $3
per $100 in loans versus FFELP's cost of about $11 per $100 in loan
guarantees. 27

27 Costs are based on estimates of the 2001 cohorts included in the Federal
Credit Supplement: Budget of the United States Government, fiscal year 2001.

How Education Considered the The first time the fee reduction could have
been taken into account was in Fee Reduction in Its Subsidy

Education's subsidy cost estimates and reestimates prepared in December
Costs

1999. The fee reduction was factored into Education's subsidy cost estimates
of the fiscal year 2000 and 2001 cohorts prepared in December 1999 for the
fiscal year 2001 President's Budget. However, given that the fee reduction
did not take effect until August 1999, Education did not factor the fee
reduction into its fiscal year 1999 reestimates because it applied to only a
small amount of the fiscal year 1999 loan volume. Education has

stated that the fee reduction will be incorporated into the fiscal year 1999
cohort reestimate of subsidy costs prepared for the fiscal year 2002
President's Budget.

Question 6 What effects have increased consolidations had on subsidy costs,
and how has Education taken account of these changes in its subsidy cost
estimates and reestimates?

By obtaining an FDLP Consolidation loan, borrowers can combine their loans
from different federal student loan programs into a new single loan and make
one monthly payment. Consolidation loans accounted for 45 percent of new
direct loan dollars disbursed in fiscal year 1999 and 26 percent of total
FDLP direct loan dollars outstanding as of September 30, 1999. While it is
clear that the volume of Consolidation loans is increasing, determining the
effects of consolidations is difficult because many factors need to be
considered, including loan maturity, prepayments,

borrower rates, and discount rates. In order to properly consider all of
these factors, an extensive loan- by- loan analysis of cash flows, applying
scenarios with and without a consolidation, would be required. Since

Education has not performed this type of detailed analysis, there is no way
of knowing the impact of increased consolidations on subsidy costs for FDLP.

Education estimates and reestimates the subsidy cost of Consolidation loans
similarly to the other FDLP loan types. For the original underlying loans, a
consolidation is in essence a loan prepayment. Education factors

both the consolidation of the underlying loans and prepayments into FDLP
subsidy cost estimates and reestimates by shortening the loan maturity
assumption, which affects the time estimated for loan repayments to be
received. While adjusting for consolidations and other prepayments through
the maturity assumption may at least partially take into account the cash
flow changes over time, as discussed in question 4, it is likely to result
in misstatements and mischaracterization of cash flows reported

annually. Education officials told us that they recognize the limitations of
their current approach and are working to develop an approach to analyze the
impacts of consolidations and other prepayments and how they can be
appropriately factored into their cash flow model.

Question 7 What effect have declining interest rates had on subsidy costs,
and how has Education taken account of these changes in its subsidy cost
estimates and reestimates?

Interest rates can affect subsidy costs directly through borrower rates and
discount rates and indirectly through borrower behavior. When the borrower
rate is greater than the discount rate, Education will receive more interest
from borrowers than it will pay in interest to Treasury to finance its
loans. This has been the situation over the short life of FDLP.

Because Education's cash flow model is continually being updated and
previous versions of the cash flow model with original assumptions were not
fully maintained, it was not possible to determine the precise effect on
subsidy costs of changes in interest rate versus changes in other cash flow
assumptions. However, it is clear that the decline in interest rates from

1995 through 1999 has had a greater impact on discount rates than borrower
rates because of the borrower rate cap. This has resulted in an increased
interest rate spread the difference between the borrower rate and discount
rate that has contributed to FDLP's estimated negative subsidy for the
fiscal year 1999 cohort. Education accounts for interest rate changes in
total in its annual reestimates. Declining Interest Rates' Effect

The two types of interest rates that are used to estimate the subsidy costs
on Subsidy Costs

of FDLP are the borrower rate and discount rate. ? The borrower rate
determines the amount of interest charged to borrowers. The borrower rate
for the Stafford Subsidized, Stafford

Unsubsidized, and PLUS loan types is variable- adjusted annually- and is
based on the 91- day Treasury bill 28 plus various add- on amounts that have
ranged from 1.7 percent to 3. 1 percent depending on the loan type

and the borrower repayment status, with a maximum borrower rate of 8.25
percent or 9.0 percent depending on the loan type. The borrower

28 The actual borrower rate is based on the 91- day Treasury bill rate from
the last Treasury Marketable Securities Auction in May. Future borrower
rates, needed in the cash flow model to estimate FDLP subsidy costs, are
based on OMB projections of 91- day Treasury bill rates.

rate for Consolidation loans made after February 1, 1999, is fixed and
calculated based on the weighted average of the borrower rates of the loans
that were consolidated, with a maximum allowable rate of 8.25

percent. As the borrower rate declines, Education receives less interest
from the borrower and, all else being equal, the subsidy cost of FDLP
increases. ? The discount rate is the interest rate used to calculate the
present value

of the estimated future cash flows and is generally equal to the rate at
which interest is paid by Education on the amounts borrowed from or held by
Treasury. The discount rate used for each cohort is fixed and determined by
the interest rates prevailing during the period that the cohort's loans were
disbursed (normally such disbursement occurs

within 2 years of loan origination for FDLP). Therefore, the discount rate
can differ significantly among cohorts. This is important because cohorts
with lower discount rates have a lower borrowing cost and, as a

result, a lower subsidy cost compared to an otherwise identical cohort with
a higher discount rate.

As discussed more fully in question 8, since 1995, FDLP borrower rates have
been greater than the discount rates, which has resulted in a positive
interest rate spread, as shown in figure 5. However, the spread was not
significant enough in the early years of the program to cover other subsidy
costs, such as defaults and interest subsidies. In fiscal year 1999, the
spread became large enough to result in an estimated negative subsidy.

Figure 5: FDLP Borrower Rates and Discount Rates

9.0% 8.0% 7.0% 6.0% 5.0% 4.0%

1995 1996 1997 1998 1999 Borrower Rate with up to a 3.1% add- on Discount
Rates

Source: Department of Education.

Beyond the direct effect of changes in interest rates on borrower and
discount rates, interest rates can also affect borrower behavior, which, in
turn, can affect defaults and prepayments and ultimately, subsidy costs.
Given all these variables and the fact that interest rate fluctuations are
nearly impossible to predict with any certainty, continued changes in FDLP
subsidy costs should be expected.

How Education Takes Account In order to calculate its subsidy cost
estimates, Education uses OMB of Interest Rate Changes

economic assumptions 29 related to future interest rates for its borrower
rate and discount rate assumptions. As part of the reestimate process,
Education updates its borrower rate and discount rate assumptions based

on actual interest rates and revised OMB economic assumptions. Education has
not prepared separate interest rate reestimates, as required by OMB Circular
A- 11. However, Education told us that its method of reestimating FDLP
subsidy costs has been accepted by OMB in the past.

Specifically, Education accounted for changes in discount rates as part of
its technical reestimate process. 30 As a result, Education is unable to
readily provide a historical analysis of the impact on subsidy costs due to
changes in discount rates. Education staff have stated that at the request
of OMB, interest rate reestimates will be prepared as part of the reestimate
process for the fiscal year 2002 President's Budget. Question 8 What are the
future prospects for the continued negative subsidy for the

Federal Direct Loan Program? Education's most recent estimates of the fiscal
year 1999 through 2001 cohorts indicate a negative subsidy cost. However, we
cannot predict with any certainty the future prospects for the continued
estimated negative

subsidy for FDLP because it is a relatively new program with limited
historical data and is very sensitive to fluctuations in interest rates and
other factors. Based on the results of the sensitivity analysis, discussed
in question 3, and the effects of interest rate fluctuations on subsidy
costs, the

primary factor determining whether FDLP has a negative or positive subsidy
is the difference, or spread, between the borrower rate and discount rate.
When the borrower rate is greater than the discount rate, Education will
receive more interest from borrowers than it will pay to Treasury for
borrowing funds, which increases the likelihood of a negative

subsidy. Conversely, when the borrower rate is less than the discount rate,
Education will pay more in interest to Treasury than it will receive from
borrowers, which decreases the likelihood of a negative subsidy. However,
several other factors, including defaults and consolidations, could also
affect whether the estimated subsidy continues to be negative. While some

29 OMB issues certain economic assumptions for use governmentwide in order
to prepare each year's President's Budget. These assumptions include
projections of interest rates, unemployment rates, and inflation rates.

30 See appendix I for a discussion of interest rate and technical
reestimates.

conditions are more favorable than others for a continued estimated negative
subsidy, whether and for how long a negative subsidy remains in effect is
unclear at this time and greatly depends on future interest rates.

Interest Rates' Effect on the While other factors do come into play,
interest rates are the key factor in

Possibility of a Negative Subsidy assessing the future cost of FDLP. In the
limited history of FDLP, large

fluctuations in interest rates have not been experienced. Figure 6 shows the
trend of the 91- day Treasury bill rate, which is used to determine borrower
rates, over the past 20 years. The shaded area shows the history

of FDLP, a period during which interest rates have been relatively stable.

Figure 6: Historical 91- Day Treasury Bill Rates

Interest rate

20% 18% 16% 14% 12% 10%

8% 6% 4% 2% 0%

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Year

91- day Treasury Bill rates

Source: The Bureau of the Public Debt.

The difference between the borrower rate and discount rate, or spread, is a
key driver of subsidy costs. This spread can be analyzed to help determine
the likelihood of a negative subsidy. The greater the spread, the more
likely a negative subsidy will result.

As discussed in question 7, the current estimated negative subsidy has
primarily been a result of borrower rates being greater than discount rates,
which will result in Education receiving more interest from borrowers than
it will pay for funds borrowed from Treasury. This condition results in a
positive spread. In the earlier years of FDLP, the spread did not offset
other subsidy costs, such as defaults and interest subsidies. Fiscal year
1999 was the first year that the positive spread resulted in a negative
subsidy. Education has estimated that this will continue through fiscal year
2001.

If the discount rate were higher than the borrower rate, a negative subsidy
would be unlikely because the spread would no longer be positive. This could
easily occur because interest rates can fluctuate significantly over

time and the discount rate for a cohort of loans is fixed and determined by
interest rates prevailing during the cohort's disbursement period, while
borrower rates are variable for three of the four FDLP loan types and

capped at a maximum allowable rate. Other Factors That Affect

Fiscal year 1995 was the first full fiscal year of existence for FDLP and as
of Continued Negative Subsidy September 30, 1999, only about 54 percent of
FDLP outstanding loan amounts were in repayment status because of the
deferred payment terms offered under this program. As a result, there is
limited historical data related to loan repayments, defaults, and
consolidations, among other things, to use as a basis for a prediction on
the future behavior of borrowers and the impact this will have on subsidy
costs. While positive spreads increase the possibility of a continued
negative subsidy for FDLP, other factors that increase costs or reduce cash
inflows

decrease the likelihood of a negative subsidy. For example, less favorable
macroeconomic conditions, such as high unemployment, will likely result in
increased defaults, or if there are further reductions in loan origination

fees, the cost of the program increases and, thus, the likelihood of a
negative subsidy decreases. The mix of loans among the four loan types could
also have an impact on whether an overall negative subsidy continues,
because not all loan types, which have separate subsidy cost estimates, have
negative subsidies. For example, the Stafford Subsidized

loan type subsidizes interest for students while they are in school. Because
this is a significant cost of the Stafford Subsidized loan type, it may
always have a positive subsidy cost regardless of the spread. Therefore, if
the FDLP portfolio were to have a larger portion of Stafford Subsidized
loans, this new mix of loans would reduce the likelihood of a negative
subsidy. Further, as discussed in question 6, since the effect on subsidy
costs due to consolidations is unknown at this time, and depends on future
interest

rates and the future performance of these Consolidation loan borrowers, the
increase in Consolidation loan volume could also have a significant impact
on the future prospects for continued negative subsidies.

Question 9 What data did Education use to project an estimated savings of $4
for every $100 of direct student loans, as it reported in November 1999?

In projecting an estimated savings of $4 for every $100 of direct student
loans, Education netted the estimated negative subsidy and the
administrative costs per $100 of loans. To do this, Education used its
subsidy cost estimate reported in the budget for the fiscal year 2000
midsession review for the subsidy cost portion of the total cost. This
estimate is based on the types of data described in the response to question
3. To estimate the federal administrative cost portion, Education used
contract expenditure data as well as data from its accounting system, OMB's
cost inflation factors, and historical data. 31 These estimated savings
pertained only to the fiscal year 2000 cohort. Education chose the fiscal
year 2000 cohort because (l) congressional interest in the federal student
loan

programs was future- oriented and (2) the data available for estimating
costs for the fiscal year 2000 cohort were more accurate and complete than
the data available for earlier cohorts. However, the projected savings will
not necessarily occur with other cohorts and may not continue to occur for
the fiscal year 2000 cohort, depending on future interest rate fluctuations.

Table 9 displays comparative cost estimates for the direct loan program for
the fiscal year 2000 and fiscal year 2001 cohorts. These data show how
changes in subsidy cost estimates can affect total cost estimates over a
relatively short period. The first column shows the initial administrative,

subsidy, and total cost estimates reported in Education's November 1999 cost
study for the fiscal year 2000 cohort. As shown in the table, the total
program cost could change from $4.11 in cost savings for every $100 in loans
for the fiscal year 2000 cohort to 58 cents in costs for every $100 in loans
for the fiscal year 2001 cohort. 31 Incorporating Federal Administrative
Costs into FFEL and Direct Loan Program Cost Estimates, U. S. Department of
Education, November 1999. In this study, Education

developed an approach to compare the total costs of the two programs by
calculating both subsidy and federal administrative costs on a net present
value basis. The study constituted a first step in developing baseline long-
term unit cost estimates for use in future cost accounting and performance
measurement systems within Education's Office of Student Financial
Assistance.

Table 9: Comparative Direct Loan Program Estimates, Per $100 of Loans

(Dollars in millions)

Fiscal year 2000 cohort (initial

Fiscal year 2001 estimate) cohort

Subsidy cost estimates $( 7. 73) a $( 3. 04) b Federal administrative cost
3. 62 3.62 estimates c Total $( 4. 11) $0. 58

a Fiscal year 2000 mid- session review estimate. b Based on Education's cost
estimate presented in the Federal Credit Supplement: Budget of the United
States Government, fiscal year 2001. c Department of Education estimate
presented in Incorporating Federal Administrative Costs into FFELP and
Direct Loan Program Cost Estimates, U. S. Department of Education, November
1999.

Source: Department of Education. For the fiscal year 2001 cohort, the
negative subsidy declined from $7. 73 to $3. 04. An Education official
explained that the increase in the subsidy cost for the fiscal year 2001
cohort is due to changes in the spread between the borrower rate and the
discount rate. Education officials also believe that the underlying
assumptions used to project the administrative cost will not change
significantly from one cohort to the next since they are not highly
sensitive to changes in loan volume. In a similar cost study, issued in
March 1999, Education's Office of Inspector General concluded that in any
given year, either FFELP's or FDLP's costs (e. g., subsidy and
administrative) could be greater depending on how prevailing economic
conditions affect subsidy costs. 32

32 Study of Cost Issues: Federal Family Education Loan Program and the
Federal Direct Loan Program, U. S. Department of Education, Office of
Inspector General, March 1999.

To develop and assign administrative costs to the direct loan program,
Education used certain costs specified in OMB guidance as well as historical
costs (such as costs in relevant contracts, salaries, rent, and

travel). 33 These costs include any expenditure associated with program
support activities such as processing applications, serving customers, and
disbursing and collecting loans. Table 10 shows the types of data Education
used to estimate the administrative cost of the direct loan program.

Table 10: Primary Types of Data Used in Education's 1999 Cost Study Types of
data Purpose

Agency contract expenditure data a To estimate cost of contracts (e. g.,
volume of applications processed each year, student loans originated and
serviced, and defaulted student loan collections) and other related loan
activity costs. Education's accounting system data To gather data on
overhead costs (for example, personnel, training, travel, rent, and
postage). OMB cost inflation factors and To estimate the lifetime cost of
direct loan historical expenditure data activities.

a Contracts analyzed include those for the operation and support of the
following information management systems: National Student Loan Data System,
Direct Loan Origination, Direct Loan Servicing, Direct Loan Central Data
System, Central Processing System, Stafford/ Perkins Data Systems, Multiple
Data Entry, the Postsecondary Education Participants System, Title IV Wide-
Area

Network, and the Public Inquiry Contract.

Education officials told us that some data on actual overhead costs were
taken from Education's cost accounting system (for example, salaries,
expenses, and rent) and Office of Student Financial Assistance (OSFA)
records. Education projected the administrative costs over the expected life
of all the loans in the fiscal year 2000 cohort using predetermined
inflation factors that existed in many of the contracts, OMB inflation
factors, or a combination of historical data and OMB inflation factors.

33 OMB Circular A- 11 defines administrative expenses as all costs directly
related to credit program operations, including (l) all activities related
to credit extension, loan servicing, write- off, and closeout, (2) all loan
systems development and maintenance, including computer costs, (3) all
monitoring of credit programs and private lenders for compliance

with laws and regulations, (4) the cost of operating separate offices that
make policy decisions for credit programs, (5) the cost of collecting
delinquent loans, and (6) the proportion of administrative expenses shared
with noncredit programs.

To develop the lifetime federal administrative cost estimates, Education
first assigned costs to one of three categories- loan origination, servicing
or account maintenance, or overhead. It then applied a three- step approach
to calculate these costs, by cohort and type, for FDLP and FFELP. The

approach included ? developing the annual spending levels for the two loan
programs based

on volume- driven costs that depend on the number of loans or similar
activity measures, such as the number of loan applications, and on
nonvolume- driven costs including personnel and fixed costs, such as rent
and travel, that do not depend on the number of loans or similar activity
measures; ? assigning annual spending for each loan program; and ?
calculating the net present value of future administrative cost by cohort.

To assign administrative costs to each loan program, Education used
designated funding sources, loan volume, and self- developed cost
assumptions. Any costs involving both grants and loans, such as application
processing, were allocated to the loan programs based on the proportion of
loan recipients to grant recipients. Any cost for activities common to both
loan programs was assigned based on annual projections of the number of
borrowers in each program. Overhead costs were assigned to the two loan
programs based on the source of funds. For example, overhead expenses funded
using section 458 of the Higher Education Act were assumed by Education to
be used for FDLP even

though some of these funds are used for FFELP costs. In using this
assumption, Education believes that it is overstating the portion of
overhead costs attributable to the direct loan program. Education projected
administrative costs over 50 years- fiscal years 2000 through 2050. The 50-
year period was used to reflect the maximum amount of time that all
borrowers in a cohort could be in school, in a deferment or forbearance
period, making loan repayments, or making payments on loan defaults. After
consulting with Education, we concluded that performing this analysis over a
shorter period- within the 9 to 27- year range Education uses in estimating
subsidy costs- would not produce significantly different results. Education
chose not to include several cost items in its calculation of administrative
cost for the loan programs. These included costs for

information system upgrades and improvements that Education believes could
reduce future per loan costs of delivering financial aid. Education

officials did not include these costs because the specific components of
these system upgrades had not been determined at the time the study was
issued, and they believe the initial cost will be offset by future savings.
However, the study did not include a cost analysis to support this belief.
To the extent that these costs are not offset by future savings in other
cost

categories, Education's administrative cost estimates will be understated.
According to its budget proposal, Education plans to spend $48.5 million on
information systems modernization for student aid programs in fiscal year
2001.

Education excluded loan origination costs for consolidation loans because
provisions of FCRA- section 502( 5)( B)( iii)- include fees as a subsidy
cost. Additionally, Education chose not to include noncontract costs
associated with offices outside OSFA since these costs only represented $3.
2 million of a total $600 million and included no more than 32 personnel.
OSFA is currently developing a cost allocation model that will identify the
total administrative cost for each of the major financial aid programs as
well as the per unit cost of delivering each loan or grant award. OSFA plans

to use this model to identify areas where it can reduce these per- unit
delivery costs and to assess how well it is accomplishing these reductions.
Unlike the November 1999 cost study, the OSFA model will include noncontract
costs from other offices within Education that have a role in delivering
student financial aid. It will use data primarily from Education's

accounting system to determine total and per unit costs. While this
administrative cost information will be useful, changes in the subsidy costs
from one cohort to the next are the primary drivers of total program costs.
Subsidy costs, in turn, are primarily affected by interest rates and
therefore cannot be predicted with any certainty.

Conclusions Developing reasonable estimates of subsidy costs for loan
programs is a complex task. Numerous assumptions must be taken into account
and

projections must be made for the estimated life of the loans, which could be
up to 30 years. Because FDLP's subsidy costs are determined largely by
interest rates- specifically the difference, or spread, between the borrower

and discount rates- and since interest rate fluctuations cannot be predicted
with any certainty, it is uncertain that the current trend in negative
subsidy costs for FDLP will continue. A change in interest rates, for
example, can cause a negative subsidy to become positive. Even with

improvements to Education's cash flow model, it is important to recognize
that estimates of subsidy costs are sensitive to interest rate volatility.

That being said, there are also other factors that affect the subsidy cost
of FDLP, such as origination fees paid by borrowers, defaults, subsequent
collections on defaulted loans, and the timing of loan repayments. While
Education is able to estimate origination fees close to the actual amounts
in the financial systems, the other key cash flows varied significantly.
These cash flows are primarily estimated based on looking at the history of
how borrowers perform under the conditions provided by each loan type within
FDLP. Because the program is relatively new, Education has primarily used
the history of FFELP as a basis for its FDLP estimates. While this is
reasonable given that it is the best historical data available, it may not
be

very predictive of FDLP borrower behavior because FDLP offers different
repayment options than those reflected in most of the historical data
related to FFELP.

Additionally, Education's current model for estimating FDLP subsidy costs
does not directly take into account certain key factors, such as prepayments
and consolidations. This limitation hinders Education's ability to determine
the impacts of consolidation activities, which are increasing

significantly. Also, Education was unable to provide actual data related to
defaults, which are a key assumption. Finally, the fact that Education does
not currently have the information readily available to make meaningful

comparisons of estimated to actual cash flows, and, most important, to
identify the reasons for differences, significantly impedes Education's
ability to refine future estimates based on actual results. Therefore, the
reliability of Education's subsidy cost estimates is negatively affected not
only by the volatility of interest rates but also by limitations in the

department's ability to monitor and adjust for other key factors in its
subsidy cost estimation process. Education is aware of these limitations and
has efforts underway to begin to address them. Recommendations for

To provide more meaningful cost estimation information that can be Executive
Action

effectively used by Congress and program decisionmakers to make timely and
well- informed judgments about FDLP, we recommend that the Secretary of the
Department of Education charge the Budget Director, who has overall
responsibility for preparing FDLP cost estimates, to take the following
actions:

? Develop and implement a method to acquire actual cash flow data on the
same basis as the cash flow model by loan profile, cohort, and key
assumption to facilitate a detailed comparison of estimated to actual cash
flows. ? Formalize and document the sensitivity analysis of assumptions

included in the FDLP cash flow model to ensure that all key assumptions used
in the cash flow model have been identified and to determine the sensitivity
of FDLP subsidy costs to changes in these assumptions. ? Develop and
implement a method of routinely comparing FDLP's

estimated and actual cash flows, including ? identifying significant
differences in total and by cohort, ? researching significant differences to
determine the specific cause, ? determining any revisions needed in the cash
flow model to ensure

that it reasonably predicts future borrower behavior, and ? determining
whether, over time, projected loan performance is reasonably predictive of
actual loan performance.

? Perform an analysis of the effects of consolidations on FDLP subsidy costs
and develop an approach to directly factor consolidations into the cash flow
model. ? Develop and implement a plan to prepare interest rate reestimates
to

isolate the effects on subsidy costs of changes in interest rates versus
changes in other assumptions. ? Refine the administrative cost modeling so
that the costs of computer system upgrades are incorporated, as well as the
cost savings that would

result from these upgrades. Agency Comments We provided the Department of
Education copies of a draft of this report for review and comment. On
December 7, 2000, we met with cognizant Education officials and obtained
oral comments on a draft of this report. Education officials generally
agreed with our answers to the questions, findings, conclusions, and
recommendations. Education is in the process of taking actions to address
some of these recommendations. For example, Education officials told us that
they are currently working to obtain a subsidiary ledger that will provide
readily available data that are comparable to data in the cash flow model to
allow for a comparison of estimated to actual cash flows on a cohort level.
Further, Education is in

the process of researching and modeling the effects of consolidations on
subsidy cost estimates. Education also provided technical comments, which we
have incorporated as appropriate.

We are sending copies to the Secretary of Education and other interested
parties. Copies will also be made available to others upon request.

Please contact either Linda M. Calbom at (202) 512- 9508 or Cornelia M.
Ashby at (202) 512- 8403, if you or your staffs have any questions
concerning this report. Key contacts and major contributors to this report
are listed in appendix III.

Linda M. Calbom Director, Financial Management and Assurance

Cornelia M. Ashby Director, Education, Workforce and Income Security Issues

Appendi Appendi xes x I

Estimating Credit Program Costs The Federal Credit Reform Act of 1990 (FCRA)
was enacted to require agencies to more accurately measure the government's
cost of federal loan programs and to permit better cost comparisons both
among credit programs and between credit and noncredit programs. FCRA
assigned to OMB the responsibility for coordinating the cost estimates
required by the act. OMB is authorized to delegate to lending agencies the
authority to estimate costs, based on its own written guidelines. These
guidelines are contained in OMB Circular A- 11, sections 85.1 through 85.
12, and supporting exhibits, 1 as well as other OMB guidance, including OMB
Circular A- 34, Instructions on Budget Execution, and other documents.

The Federal Accounting Standards Advisory Board (FASAB) 2 developed the
accounting standard for credit programs, SFFAS No. 2, Accounting for Direct
Loans and Loan Guarantees, which became effective in fiscal year 1994. This
standard, which generally mirrors FCRA, established guidance for estimating
the cost of direct and guaranteed loan programs as well as recording direct
loans and the liability for loan guarantees for financial reporting
purposes. The actual and expected costs of federal credit programs should be
fully

recognized in both budgetary and financial reporting. To determine the
expected cost of a credit program, agencies are required to predict or
estimate the future performance of the program. This cost, known as the

subsidy cost, is the present value 3 of disbursements- over the life of the
loan- by the government (loan disbursements and other payments) minus
estimated payments to the government (repayments of principal, payments of
interest, other recoveries, and other payments). For loan guarantees, the

subsidy cost is the present value of cash flows from estimated payments by 1
The act requires OMB to coordinate with the Congressional Budget Office in
developing estimation guidelines. 2 FASAB was created by OMB, Treasury, and
GAO to consider and recommend accounting principles for the federal
government. These three agencies approved Statement of Federal Financial
Accounting Standards (SFFAS) No. 2, Accounting for Direct Loans and Loan
Guarantees, in July 1993 and SFFAS No. 18, Amendments to Accounting
Standards for Direct Loans and Loan Guarantees in SFFAS No. 2 in May 2000. 3
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. For the period we reviewed, when
calculating the present value of loan subsidy costs, agencies were required
to use as the discount rate the average annual interest rate for marketable
U. S. Treasury securities with similar maturities to the loan or guarantee.

the government (for defaults and delinquencies, interest rate subsidies, and
other payments) minus estimated payments to the government (for loan
origination and other fees, penalties, and recoveries).

To estimate the cost of loan programs, agencies first estimate the future
performance of direct and guaranteed loans when preparing their annual
budgets. The data used for these budgetary estimates should be reestimated
to reflect any changes in loan performance since the budget was prepared.
These reestimated data are then used in financial reporting

when calculating the allowance for subsidy (the cost of direct loans), the
liability for loan guarantees, and the cost of the program. In the financial
statements, the actual and expected costs of loans disbursed as part of a
credit program are recorded as a “Program Cost” on the agencies'
Statement of Net Costs for loans disbursed.

In addition to recording the cost of a credit program, SFFAS No. 2 requires
agencies to record direct loans on the balance sheet as assets at the
present value of their estimated net cash inflows. The difference between
the outstanding principal balance of the loans and the present value of
their net cash inflows is recognized as a subsidy cost allowance- generally
the cost of the direct loan program. For guaranteed loans, the present value
of the

estimated net cash outflows, such as defaults and recoveries, is recognized
as a liability and generally equals the cost of the loan guarantee program.
In preparing SFFAS No. 2, FASAB indicated that the subsidy cost components-
interest, defaults, fees, and other cash flows- would be valuable for making
credit policy decisions, monitoring portfolio quality, and improving credit
performance. Thus, agencies are required to recognize, and disclose in the
financial statement footnotes, the four components of the credit subsidy-
interest, net defaults, fees and other collections, and other subsidy costs-
separately for the fiscal year during

which direct or guaranteed loans are disbursed. In addition,
nonauthoritative guidance is contained in the previously discussed Technical
Release of the Credit Reform Task Force of the Accounting and Auditing
Policy Committee, entitled Preparing and Auditing Direct Loan and Loan
Guarantee Subsidies Under the Federal Credit Reform Act. This Technical
Release provides detailed

implementation guidance for agency staff on how to prepare reasonable credit
subsidies. Further, the Technical Release provides suggested procedures for
auditing credit subsidy estimates.

Developing Cash Flow In estimating cash flows, Education and other credit
agencies are required Assumptions and to predict borrower behavior how many
borrowers will pay early, pay late,

or default on their loans and at what point in time. Generally, the subsidy
Models

costs equal the amount of estimated losses to the federal government and are
financed with appropriated funds. The portion of Education's direct loans
that Education predicts will ultimately be collected is financed by
borrowing from Treasury. For example, a hypothetical FDLP loan of $100 may
have a subsidy cost of $20 (the amount Education expects to lose), which is
financed with appropriated funds, and the remaining $80 is financed by
Treasury borrowings (the amount Education expects to be repaid).

Budgeting guidance requires agencies to maintain supporting documentation
for subsidy cost estimates. Further, auditing standards related to preparing
estimates indicate that agency management is responsible for accumulating
relevant, sufficient, and reliable data on

which to base the estimates. SFFAS No. 2 indicates that each credit program
should use a systematic methodology to project expected cash flows into the
future. To accomplish this task, agencies should develop cash flow models. A
cash flow model is a computer program that generally uses historical
information and various assumptions, including defaults, prepayments,
recoveries, and the timing of these events, to estimate future loan
performance. These cash flow models, which should be based on sound
economic, financial, and statistical theory, identify key factors that
affect loan performance. Agencies use this information to make more informed
predictions of future credit performance. The August 1994 User's Guide to
Version r. 8 of the OMB Credit Subsidy Model provides general guidance on
creating cash flow models to estimate future delinquencies,

defaults, recoveries, etc. This user's guide states that, “In every
case, the agency or budget examiner must maintain current and complete
documentation and justification for the estimation methods and

assumptions used in determining the cash flow figures used for the OMB
Subsidy Model” to calculate the credit subsidy.

According to SFFAS No. 2, to estimate the cost of loan programs and predict
the future performance of credit programs, agencies should establish and use
reliable records of historical credit performance. Since

actual historical experience is a primary factor upon which estimates of
credit performance are based, agencies should maintain a database, also
known as an information store, at the individual loan level, of historical
information on all key cash flow assumptions, such as defaults or

recoveries, used in calculating the credit subsidy cost. Additional
nonauthoritative guidance on cash flow models may be found in the Model
Credit Program Methods and Documentation for Estimating Subsidy Rates and
the Model Information Store issue paper prepared by the Credit Reform Task
Force of the Accounting and Auditing Policy Committee. This

draft “Information Store” Task Force paper provides guidance on
the type of historical information agencies need to reasonably estimate the
cost of credit programs. The information store should provide three types of
information. First, the information store should maintain key loan

characteristics at the individual loan level, such as the loan terms and
conditions. Second, it should track economic data that influence loan
performance, such as property values for housing loans. Third, an
information store should track historical cash flows on a loan- by- loan
basis. The data elements in an information store should be selected to allow
for more in- depth analyses of the most significant subsidy estimate
assumptions.

In addition to using historical databases and the cash flow models, other
relevant factors must be considered by agencies to estimate future loan
performance. These relevant factors include

? economic conditions that may affect the performance of the loans, ?
financial and other relevant characteristics of borrowers, ? the value of
the collateral to loan balance, ? changes in recoverable value of
collateral, and ? newly developed events that would affect loan performance.

When new programs are established or changes are made to existing programs,
historical supporting documentation for cash flow assumptions may not exist.
In the absence of valid, relevant historical experience, the agency may use
relevant experience from other federal or private sector loan programs.
These data, often called proxy data, should be temporarily

used while the agency collects adequate historical data for the new or
revised loan program. Reestimating Credit

Agencies prepare estimates of loan program costs as a part of their budget
Subsidies

requests. Later, after the end of the fiscal year, agencies are required to
update or “reestimate” loan costs for differences among
estimated loan performance and related cost, the actual program costs
recorded in the accounting records, and expected changes in future economic
performance. The reestimate should include all aspects of the original cost

estimate, including prepayments, defaults, delinquencies, recoveries, and
interest. Reestimates of the credit subsidy allow agency management to
compare the original budget estimates with actual program results to
identify variances from the original estimate, assess the quality of the
original estimate, and adjust future program estimates as appropriate. Any

increase or decrease in the estimated cost of the loan program is recognized
as a subsidy expense or a reduction in subsidy expense for both budgetary
and financial statement purposes. The reestimate requirements for interest
rate and technical assumptions (defaults, recoveries, prepayments, fees, and
other cash flows) differ. For budget purposes, OMB Circular A- 11 states
that agencies must reestimate

the interest portion of the estimate when a cohort is substantially
disbursed or generally when at least 90 percent of the direct loans or
guaranteed loans are disbursed. The technical reestimate, for budgetary
purposes, generally must be done annually, after the close of each fiscal
year as long as the loans are outstanding, unless OMB approves a different
plan, 4 regardless of financial statement significance. For financial
statement reporting purposes, both technical and interest rate reestimates
are required annually, at the end of the fiscal year, whenever the
reestimated

amount is significant to the financial statements. If there is no
significant change in the interest portion of the estimate prior to the
loans being 90 percent disbursed, then the interest rate reestimate may be
done once when the loans are at least 90 percent disbursed. In addition,
SFFAS No. 18, which was effective beginning in fiscal year 2001, requires
that reestimates be measured and reported in two separate components:
interest rate reestimates and technical/ default reestimates.

Interest rate reestimates are made to adjust the credit subsidy estimate for
the difference between the discount rate originally estimated and the actual
interest rates prevailing during the years the loan was disbursed. To
calculate the size of this effect, all other assumptions (repayment rates,
default rates, etc.) must be the same as those used to calculate the
original subsidy estimate. Technical reestimates are made to adjust for all
changes in assumptions other than interest rates. The purpose of the
technical 4 The OMB representative with primary budget authority may
authorize agencies to calculate technical reestimates for budgetary purposes
less frequently than every year when any one of four conditions are met: (1)
based on periodic schedules established in coordination with

OMB, (2) when a major change in actual versus projected activity is
detected, (3) when a significant difference is detected through monitoring
“triggers” developed in coordination with OMB, and (4) when a
group of loans are being closed out.

reestimate is to adjust the subsidy estimate for differences between the
original projection of cash flows and the amount and timing of expected cash
flows based on actual experience, new forecasts of future economic
conditions, and other events and improvements in the methods used to
estimate future cash flows.

Appendi x II

Objectives, Scope, and Methodology This report responds to your request and
that of the former Chairman of the House Committee on the Budget that we
prepare a report on the financing of Education's William D. Ford Federal
Direct Loan Program (FDLP). To respond to your request, for fiscal years
1995 through 1999 we reviewed Education's audited financial statements and
examined the workpapers of Education's Independent Public Accountants. We
interviewed knowledgeable personnel from Education's budget office and

obtained information relevant to the questions we were asked to answer. We
assessed Education's credit subsidy estimation practices against federal
accounting and budget standards, including SFFAS No. 2, Accounting for
Direct Loans and Loan Guarantees; OMB Circular A- 11,

Preparation and Submission of Budget Estimates; and guidance contained in
the Federal Financial Accounting and Auditing Technical Release 3, Preparing
and Auditing Direct Loan and Loan Guarantee Subsidies Under the Federal
Credit Reform Act. The scope and methodology for responding to each of the
nine questions you asked is discussed as follows.

Question 1 How much financing has been provided to Education for the direct
loan program through borrowing from Treasury and appropriations received?

We obtained from Education schedules of borrowings from Treasury, repayments
to Treasury, and appropriations received for fiscal years 1995 through 1999.
We verified the schedules of Treasury borrowing to data contained in the
workpapers of Education's Independent Public Accountant. We obtained
schedules of original subsidy estimates and reestimates for fiscal years
1995 through 1999 cohorts. We verified the subsidy appropriations to
Education's original documentation, SF132 reports on Apportionment and
Reapportionment, and schedule 1151s to

return negative subsidy to Treasury. Question 2 Have cash inflows (excluding
borrowings from Treasury and borrower

principal repayments) exceeded cash outflows (excluding repayments to
Treasury and loan disbursements)? Data relating to loan origination fees,
interest receipts from borrowers, and net interest payment on Treasury
borrowings were obtained from the Appendix to the President's Budget for
fiscal years 1997 through 2001, which contained actual data for the fiscal
years 1995 through 1999. We also verified fiscal years 1995, 1996, and 1997
actual cash flows to Education's statement of cash flows in its financial
statements. For fiscal years 1998

and 1999, we verified the actual data to cash collection amounts provided by
Education's financial systems. Question 3 In Education's calculation of its
subsidy cost estimates for FDLP, what are

the key cash flow assumptions, how sensitive are Education's subsidy costs
to changes in these assumptions, and what data are used to support these
assumptions?

To gain an understanding of Education's cash flow model, we reviewed
Education's model documentation, the workpapers of Education's independent
public accountant, and various reports. To identify which of the over 1,900
cash flow assumptions were key cash flow assumptions, we

first discussed with Education budget staff what cash flow assumptions they
believed were key cash flow assumptions for FDLP based on their prior
analyses. Since much of the data used to estimate the cost of FDLP are proxy
data from FFELP, we determined what cash flow assumptions were key cash flow
assumptions for FFELP based on the independent

public accountant's workpapers. Based on our experience with other federal
credit programs, we identified other assumptions that we believed may also
be key. We then conducted an analysis of FDLP to identify the most
significant loan profiles, which includes the loan type, risk category,

and repayment option. To determine how sensitive FDLP's cost was to changes
in these key assumptions, we requested that Education budget staff conduct a
limited sensitivity analysis of the assumptions they thought might be key as
well as the other assumptions we identified. In instructing Education on how
to perform the sensitivity analysis, we generally followed the guidance

contained in the Federal Financial Accounting and Auditing Technical Release
3, Preparing and Auditing Direct Loan and Loan Guarantee Subsidies Under the
Federal Credit Reform Act 1 and requested that 1 The Credit Reform Task
Force of the Accounting and Auditing Policy Committee was

formed in order to address key issues surrounding the implementation of FCRA
and the related federal accounting standard. This task force developed
Technical Release 3, which is expected to be formally issued by OMB during
fiscal year 2001. The purpose of Technical Release 3 is to provide
implementation guidance for agencies and auditors to prepare, utilize, and
report on credit subsidy estimates. Technical Release 3 does not take
precedence over existing accounting standards and budget guidance.
Currently, these standards and guidance do not require agencies to perform a
sensitivity analysis; however, Technical Release 3 encourages agencies to
perform this analysis.

Education increase and decrease by 10 percent the value of each nontiming
related assumption presumed to be key. Because timing assumptions are
modeled differently, and should also be adjusted in a systematic manner, we
requested that Education increase the amount of the loans in the

beginning repayment assumption by 5 percent during the first 5 years to
simulate a decrease in the time it took borrowers to repay their loans. We
analyzed the results of the limited sensitivity analysis and determined

that any assumption that produced a change of at least 2 percent and $13
million in the estimated cost of any single loan profile tested was a key
cash flow assumption. We then met with agency officials to identify the data
sources for key cash flow assumptions.

Question 4 How closely do Education's subsidy cost estimates and their
underlying assumptions compare to actual loan performance for each loan
cohort and to what extent does Education track differences between its
subsidy cost estimates and actual loan performance for each loan cohort?

We compared cash flows related to five of the seven key cash flow
assumptions identified in question 3 (interest payments, principal payments,
default rate, origination fees, and collections on defaulted loans) and
obtained estimated and actual cash flow data for fiscal years 1995 through
1999. Due to the nature of direct loan programs, the comparison did not
include any analysis of defaults because Education was unable to readily
provide comparable data on estimated and actual defaults. The discount rate
assumption was not included because it does not directly affect the amount
or timing of cash flows. Rather, this assumption is used to estimate the
present value of the cash flows. Because the actual cash

flow data in Education's financial systems were not totally comparable to
data available in the cash flow model (by cohort, key cash flow assumption,
and loan profile), Education obtained actual cash flow data on fiscal year
totals from its financial systems for its analysis.

For estimated cash flows, original cash flow models from fiscal years 1995
through 1999 were not fully maintained, thus, we used Education's analysis
using its current cash flow model and assumptions for each fiscal year
beginning with 1995. We verified that the actual cash flow data provided
agreed to the amounts reported in Education's budget submissions. We also
verified fiscal year 1995, 1996, and 1997 actual cash flows to Education's
statement of cash flows in its financial statements. For fiscal years 1998
and 1999, we verified the actual data to cash collection amounts

provided by Education's financial systems. For total cash flows for fiscal
years 1995 through 1999, we then compared estimated to actual cash flows to
determine the amount of the difference. We met with Education budget staff
to determine and request supporting documentation for the causes of these
differences. Since supporting documentation was unavailable, we were unable
to corroborate Education's explanations for these differences.

Question 5 What effect have reduced loan origination fees had on subsidy
costs, and how has Education taken account of these changes in its subsidy
cost estimates and reestimates?

To determine the impact of reduced loan origination fees on subsidy costs,
we requested that Education calculate the credit subsidy costs for FDLP
overall and each of the four loan types with the origination fee equal to 4
percent and 3 percent. We analyzed the results of the calculation and
discussed with Education personnel how they accounted for the reduced loan
origination fees in subsidy cost estimates and reestimates. Question 6 What
effects have increased consolidations had on subsidy costs, and how

has Education taken account of these changes in its subsidy cost estimates
and reestimates? To assess the impact of increased consolidations, we
discussed consolidations with Education personnel, including how they work,
their history, and how consolidations are modeled in subsidy cost estimates
and reestimates. We determined the various factors that could affect how

consolidations effect FDLP subsidy costs. Because Education had not
performed the detailed analysis necessary to determine the actual effect of
consolidations, we were unable to determine the impact of increased
consolidations on subsidy costs for FDLP.

Question 7 What effect have declining interest rates had on subsidy costs,
and how has Education taken account of these changes in its subsidy cost
estimates and reestimates?

Because Education's cash flow model is continually being updated and copies
of the model with original assumptions were not fully maintained, it was not
possible to determine the precise effect on subsidy costs of changes in
interest rates versus other changes. In order to address this

question, we assessed the general impact of declining interest rates by
analyzing the effect of declining borrower rates and discount rates. To
determine the impact of declining borrower rates and discount rates, we
requested that Education calculate subsidy costs for four scenarios. We
analyzed the results of these calculations and discussed with Education
personnel their procedures for accounting for changes in interest rates in
the credit subsidy estimates and reestimates. We compared Education's
procedures with the guidance provided in OMB Circular A- 11.

Question 8 What are the future prospects for the continued negative subsidy
for the Federal Direct Loan Program?

We analyzed the results from several of the other questions to determine
what conditions increase or decrease the likelihood for continued negative
subsidy of FDLP. We analyzed the effect of fee reductions and increased
consolidations from question 6, the impact of interest rates on subsidy cost
from question 7, and the result of the sensitivity analysis from question 3.

Question 9 What data did Education use to project an estimated saving of $4
on every $100 of direct student loans, as it reported in November 1999?

We analyzed Education's November 1999 cost study to get a general
understanding of the methodology used to develop administrative and subsidy
cost estimates for FFELP and FDLP. We interviewed Education personnel to
obtain a more detailed understanding of the methodology and data sources
used to assign, develop, and project the administrative and subsidy cost
estimates. We also reviewed spreadsheets and other documentation prepared by
Education to support its findings.

Appendi x II I GAO Contacts and Staff Acknowledgments GAO Contacts Linda M.
Calbom, (202) 512- 9508 Cornelia M. Ashby, (202) 512- 8403 Acknowledgments
In addition to those named above, Daniel R. Blair, Marcia L. Carlsen,

Susan T. Chin, Anh Dang, Cheryl D. Driscoll, Julia B. Duquette, Elizabeth M.
Kreitzman, Kirsten L. Landeryou, Joel R. Marus, Andrew Sherrill, Linda W.
Stokes, and Maria Zacharias made key

contributions to this report.

Glossary The following is a group of terms commonly used in credit budgeting
and accounting. The definitions for many of these terms are equally
applicable to loan guarantees. However, since FDLP is a direct loan program,

references to loan guarantees have been omitted. Administrative Expenses All
costs directly related to credit program operations, including: (1)
activities related to credit extension, loan servicing, write- off, and

closeout; (2) loan systems development and maintenance, including computer
costs; (3) all monitoring of credit programs and private lenders for
compliance with laws and regulations; (4) the cost of operating

separate offices that make policy decisions for credit programs; (5) the
cost of collecting delinquent loans; and (6) the proportion of
administrative expenses shared with noncredit programs.

Assumptions Basic beliefs about the future performance of a loan or group of
loans. Types of assumptions include the following:

? Cash flow assumptions all known and/ or forecasted information about the
characteristics and performance of a loan or group of loans. Examples
include estimates of loan maturity, borrower rate,

default/ delinquency rate, and timing of cash flow events, such as defaults
and collections on defaulted loans. ? Model assumptions determinations of
how cash flow assumptions are applied through the life of a cohort of loans.
For example, determining

whether the entire estimated amount of defaults should be applied in one
year or allocated over several years. Cash Flows Payments or estimates of
payments to or from the government over the life of a loan or group of
loans. For direct loans, these may include: loan

disbursements, repayments of principal, payments of interest, prepayments,
fees, penalties, defaults and collections on defaulted loans. Cohort All
loans of a program for which a subsidy appropriation is provided for a

given fiscal year, even if disbursements occur in subsequent years.

Credit Program Account A budget account into which an appropriation for the
funds to finance a loan program is made and from which funds are disbursed
to a financing account for the program.

Discount Rates The collection of interest rates that are used to calculate
the present value of the cash flows that are estimated over a period of
years. For the period we reviewed, when calculating the present value of
loan subsidy costs, agencies were required to use as the discount rate the
average annual

interest rate for marketable U. S. Treasury securities with similar
maturities to the loan or guarantee. Financing Account A nonbudgetary
account that collects the payments from a credit program account and
borrowings from Treasury. It includes all cash flows to and

from the government resulting from loan programs. At least one financing
account is associated with each credit program account.

Key Assumptions Assumptions that have been established, through sensitivity
analysis or other means, to be the elements that have a large impact on
estimates and thus are the most important factors in determining the cost of
a loan or group of loans.

OMB Credit Subsidy A computer software program that calculates a subsidy
rate based on the Calculator

present value of agency- generated estimates of cash flows to and from the
government. It also calculates the portions of the subsidy cost attributable
to defaults, interest, fees, and other subsidy components.

Present Value 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.

Risk Category A subdivision of a cohort of loans into groups of loans that
are relatively homogeneous in cost, given the facts known at the time of
obligation or commitment. Risk categories will group all loans obligated or
committed

for a program during the fiscal year that share characteristics predictive
of defaults and other costs. Reestimates Revisions of the subsidy cost
estimate of a cohort (or risk category) based

on information about actual performance of a cohort of loans or estimated
changes in future cash flows of the cohort.

Subsidy Cost The estimated long- term cost to the government of a direct
loan, calculated on a net present value basis, excluding administration
costs.

(916365) Lett er

Ordering Information The first copy of each GAO report is free. Additional
copies of reports are $2 each. A check or money order should be made out to

the Superintendent of Documents. VISA and MasterCard credit cards are
accepted, also. Orders for 100 or more copies to be mailed to a single
address are discounted 25 percent.

Orders by mail: U. S. General Accounting Office P. O. Box 37050 Washington,
DC 20013

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

Orders by phone: (202) 512- 6000 fax: (202) 512- 6061 TDD (202) 512- 2537

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

Orders by Internet: For information on how to access GAO reports on the
Internet, send an e- mail message with “info” in the body to:
info@ www. gao. gov or visit GAO's World Wide Web home page at: http:// www.
gao. gov

To Report Fraud,

Contact one:

Waste, or Abuse in ? Web site: http:// www. gao. gov/ fraudnet/ fraudnet.
htm

Federal Programs

? e- mail: fraudnet@ gao. gov ? 1- 800- 424- 5454 (automated answering
system)

GAO United States General Accounting Office

Page 1 GAO- 01- 197 FDLP Cost Estimates

Contents

Contents Page 2 GAO- 01- 197 FDLP Cost Estimates

Page 3 GAO- 01- 197 FDLP Cost Estimates United States General Accounting
Office

Washington, D. C. 20548 Page 3 GAO- 01- 197 FDLP Cost Estimates

Page 4 GAO- 01- 197 FDLP Cost Estimates

Page 5 GAO- 01- 197 FDLP Cost Estimates

Page 6 GAO- 01- 197 FDLP Cost Estimates

Page 7 GAO- 01- 197 FDLP Cost Estimates

Page 8 GAO- 01- 197 FDLP Cost Estimates

Page 9 GAO- 01- 197 FDLP Cost Estimates

Page 10 GAO- 01- 197 FDLP Cost Estimates

Page 11 GAO- 01- 197 FDLP Cost Estimates

Page 12 GAO- 01- 197 FDLP Cost Estimates

Page 13 GAO- 01- 197 FDLP Cost Estimates

Page 14 GAO- 01- 197 FDLP Cost Estimates

Page 15 GAO- 01- 197 FDLP Cost Estimates

Page 16 GAO- 01- 197 FDLP Cost Estimates

Page 17 GAO- 01- 197 FDLP Cost Estimates

Page 18 GAO- 01- 197 FDLP Cost Estimates

Page 19 GAO- 01- 197 FDLP Cost Estimates

Page 20 GAO- 01- 197 FDLP Cost Estimates

Page 21 GAO- 01- 197 FDLP Cost Estimates

Page 22 GAO- 01- 197 FDLP Cost Estimates

Page 23 GAO- 01- 197 FDLP Cost Estimates

Page 24 GAO- 01- 197 FDLP Cost Estimates

Page 25 GAO- 01- 197 FDLP Cost Estimates

Page 26 GAO- 01- 197 FDLP Cost Estimates

Page 27 GAO- 01- 197 FDLP Cost Estimates

Page 28 GAO- 01- 197 FDLP Cost Estimates

Page 29 GAO- 01- 197 FDLP Cost Estimates

Page 30 GAO- 01- 197 FDLP Cost Estimates

Page 31 GAO- 01- 197 FDLP Cost Estimates

Page 32 GAO- 01- 197 FDLP Cost Estimates

Page 33 GAO- 01- 197 FDLP Cost Estimates

Page 34 GAO- 01- 197 FDLP Cost Estimates

Page 35 GAO- 01- 197 FDLP Cost Estimates

Page 36 GAO- 01- 197 FDLP Cost Estimates

Page 37 GAO- 01- 197 FDLP Cost Estimates

Page 38 GAO- 01- 197 FDLP Cost Estimates

Page 39 GAO- 01- 197 FDLP Cost Estimates

Page 40 GAO- 01- 197 FDLP Cost Estimates

Page 41 GAO- 01- 197 FDLP Cost Estimates

Page 42 GAO- 01- 197 FDLP Cost Estimates

Page 43 GAO- 01- 197 FDLP Cost Estimates

Page 44 GAO- 01- 197 FDLP Cost Estimates

Page 45 GAO- 01- 197 FDLP Cost Estimates

Page 46 GAO- 01- 197 FDLP Cost Estimates

Page 47 GAO- 01- 197 FDLP Cost Estimates

Page 48 GAO- 01- 197 FDLP Cost Estimates

Appendix I

Appendix I Estimating Credit Program Costs

Page 49 GAO- 01- 197 FDLP Cost Estimates

Appendix I Estimating Credit Program Costs

Page 50 GAO- 01- 197 FDLP Cost Estimates

Appendix I Estimating Credit Program Costs

Page 51 GAO- 01- 197 FDLP Cost Estimates

Appendix I Estimating Credit Program Costs

Page 52 GAO- 01- 197 FDLP Cost Estimates

Appendix I Estimating Credit Program Costs

Page 53 GAO- 01- 197 FDLP Cost Estimates

Page 54 GAO- 01- 197 FDLP Cost Estimates

Appendix II

Appendix II Objectives, Scope, and Methodology

Page 55 GAO- 01- 197 FDLP Cost Estimates

Appendix II Objectives, Scope, and Methodology

Page 56 GAO- 01- 197 FDLP Cost Estimates

Appendix II Objectives, Scope, and Methodology

Page 57 GAO- 01- 197 FDLP Cost Estimates

Appendix II Objectives, Scope, and Methodology

Page 58 GAO- 01- 197 FDLP Cost Estimates

Page 59 GAO- 01- 197 FDLP Cost Estimates

Appendix III

Page 60 GAO- 01- 197 FDLP Cost Estimates

Glossary Page 61 GAO- 01- 197 FDLP Cost Estimates

Glossary Page 62 GAO- 01- 197 FDLP Cost Estimates

United States General Accounting Office Washington, D. C. 20548- 0001

Official Business Penalty for Private Use $300

Address Correction Requested Bulk Rate

Postage & Fees Paid GAO Permit No. GI00
*** End of document. ***