Credit Reform: U.S. Needs Better Methods for Estimating Cost of Foreign
Loans and Guarantees (Letter Report, 12/19/94, GAO/NSIAD/GGD-95-31).
This report (1) evaluates the executive branch's method for calculating
country risk rating and cost estimates for foreign loans and loan
guarantees and provides GAO's own estimates and (2) determines the
probability of default for each country. GAO also reviews the executive
branch's authority to reschedule international debt owed to the U.S.
government and the implementation of the law's provisions for
rescheduling international debt. GAO found several weaknesses in the
executive branch's method, particularly in the method it used to
calculate risk premiums for higher risk countries. The main weakness
was that the executive branch's method did not rely on econometric tests
and measurements. GAO estimated the long-run probability of default for
170 countries. GAO's estimates of default risk ranged from 92.1 percent
for Cambodia to zero percent for the highest rated countries, such as
Japan, Switzerland, and Germany. GAO estimated that Russia has more
than two chances out of three of defaulting on a loan, regardless of the
length of the loan.
--------------------------- Indexing Terms -----------------------------
REPORTNUM: NSIAD/GGD-95-31
TITLE: Credit Reform: U.S. Needs Better Methods for Estimating
Cost of Foreign Loans and Guarantees
DATE: 12/19/94
SUBJECT: Risk management
Government guaranteed loans
International economic relations
Foreign loans
Debt collection
Evaluation methods
Economic analysis
Projections
Foreign economic development credit
Loan defaults
**************************************************************************
* This file contains an ASCII representation of the text of a GAO *
* report. Delineations within the text indicating chapter titles, *
* headings, and bullets are preserved. Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are *
* identified by double and single lines. The numbers on the right end *
* of these lines indicate the position of each of the subsections in the *
* document outline. These numbers do NOT correspond with the page *
* numbers of the printed product. *
* *
* No attempt has been made to display graphic images, although figure *
* captions are reproduced. Tables are included, but may not resemble *
* those in the printed version. *
* *
* A printed copy of this report may be obtained from the GAO Document *
* Distribution Facility by calling (202) 512-6000, by faxing your *
* request to (301) 258-4066, or by writing to P.O. Box 6015, *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time. *
**************************************************************************
Cover
================================================================ COVER
Report to Congressional Committees
December 1994
CREDIT REFORM - U.S. NEEDS BETTER
METHOD FOR ESTIMATING COST OF
FOREIGN LOANS AND GUARANTEES
GAO/NSIAD/GGD-95-31
GAO/NSIAD-95-31
GAO/NSIAD/GGD-95-31 Credit Reform
Abbreviations
=============================================================== ABBREV
AID - Agency for International Development
CBO - Congressional Budget Office
CCC - Commodity Credit Corporation
DSAA - Defense Security Assistance Agency
EXIM - Export Import Bank
FMS - Foreign Military Sales
GSM - General Sales Manager
ICRAS - Interagency Country Risk Assessment System
OMB - Office of Management and Budget
USDA - U.S. Department of Agriculture
Letter
=============================================================== LETTER
B-252443
December 19, 1994
The Honorable Robert C. Byrd
Chairman
The Honorable Mark O. Hatfield
Ranking Minority Member
Committee on Appropriations
United States Senate
The Honorable Jim Sasser
Chairman
The Honorable Pete V. Domenici
Ranking Minority Member
Committee on the Budget
United States Senate
The Honorable David R. Obey
Chairman
The Honorable Joseph M. McDade
Ranking Minority Member
Committee on Appropriations
House of Representatives
The Honorable Martin Olav Sabo
Chairman
The Honorable John R. Kasich
Ranking Minority Member
Committee on the Budget
House of Representatives
This report responds to section 590 of the 1993 Foreign Operations,
Export Financing, and Related Programs Appropriations Act (P.L.
102-391) and a request from the Chairman, Senate Committee on the
Budget, that we evaluate the adequacy of the executive branch's
methodology for implementing international aspects of the Federal
Credit Reform Act of 1990. The legislation required that we (1)
evaluate the executive branch's method for calculating country risk
ratings and cost estimates for foreign loans and loan guarantees and
develop our own estimates and (2) determine the probability of
default for each country. We were also requested to review the
executive branch's authority to reschedule international debt owed to
the U.S. government and the implementation of the act's provisions
for rescheduling international debt.
The requirement to evaluate risk premiums for international lending
recognizes the greater difficulty of making risk estimates for
international credit compared to domestic credit, and is part of a
broader question of how to apply the principles of credit reform to
international credit programs.\1
For example, whereas domestic credit programs usually consist of a
large number of similar loans over which risk is spread,
international credit is more likely to consist of a small number of
direct loans and loan guarantees with individually negotiated terms.
Some have questioned whether credit reform fits well in the
international arena; however, since the Credit Reform Act applies to
international credit programs, having a sound method for estimating
the cost of foreign loans and loan guarantees is important. Its
importance arises from the need to ensure that the subsidy costs of
such programs are accurately presented in the President's budget and
that the necessary annual appropriations are enacted to cover the
costs, before the direct loans are obligated and loan guarantees
committed.
--------------------
\1 Risk premiums reflect the probability of default for a country by
maturity and are applied to scheduled payment streams to obtain loan
repayment projections.
RESULTS IN BRIEF
------------------------------------------------------------ Letter :1
Executive branch estimates of the subsidy cost for international
loans and loan guarantees can be better determined by using an
empirically rigorous method. The executive branch's method for
calculating the subsidy cost of international loans and loan
guarantees was to (1) rate countries' debt on a scale, from the most
creditworthy to the least creditworthy after considering 35 subitems,
most of which were 5-year expectations of economic performance
indicators; (2) calculate the cost of country risk based on these
assigned ratings and corresponding risk premiums; and (3) calculate
the subsidy cost which, in addition to country risk cost, includes
interest rate costs (or income) and fee income.
We found a number of weaknesses in the executive branch's method,
particularly in the method it used to calculate risk premiums for
countries of higher risk.\2 The principal weakness was that the
executive branch's method was not based on econometric tests and
measurements. Other weaknesses include (1) employing too little of
the available data in its analysis, (2) using an incorrect key
assumption on how financial markets work as a substitute for missing
data, (3) obtaining the same risk premiums for loans and bonds, (4)
not revising risk premiums with the most current information from
financial markets, and (5) not adequately disclosing sources of bias.
The executive branch also did not distinguish between new loans to
countries of higher risk that have relatively large effects on old
loan repayment and those loans that do not, because it believes that
the Credit Reform Act precludes it from making this distinction in
its cost estimates. We do not dispute its position on the
legislation. However, our analysis indicates that, by not making
this distinction, the subsidy cost estimates for new loans that have
large effects on old loan repayments will tend to be overstated.
Our method for calculating the costs of international loans and loan
guarantees also employed three steps: we (1) calculated country risk
ratings by statistically combining those that appeared in two leading
publications for financial institutions, (2) transformed our country
risk ratings into risk premiums using data from financial markets for
less risky debt and econometric test and measurements of the
secondary market for more risky sovereign debt, and (3) calculated
subsidy costs by using present value analysis. For all debt, we also
estimated the extent to which new loans affected old loan repayments.
In estimating the cost of risky debt, we based our estimates on 20
loans owed by 20 countries, whereas the executive branch based its
calculations on 5 bond observations.
To compare the implications of using our estimation method to that of
the executive branch, we estimated costs for the $13.7 billion of
international loans and guarantees authorized in fiscal year 1992, as
if the estimates had been made during October 1993.\3 We estimate
that the total U.S. government subsidy cost was $3 billion using our
method,\4 about 2-1/2 times the $1.2 billion estimate using the
method employed by the executive branch.\5 Most of the difference was
due to differences in estimates of country risk cost. If we had not
distinguished between new loans that had large effects on old loan
repayments and those loans that did not, our estimates of total
subsidy cost would have been $4.6 billion.
Cost estimates using either the executive branch's or our method will
differ depending upon market expectations for a particular time and
group of foreign loans and guarantees. For example, if our country
risk cost estimates had been made based on more recent secondary
market prices, then our estimates would have been smaller because
prices on the secondary market were generally higher. Also, in
future years, we would generally expect there to be relatively more
rollover of loans than occurred in 1992, when large amounts of funds
shifted toward countries in Central and Eastern Europe and the newly
independent states of the former Soviet Union thereby increasing our
estimates of country risk cost.
We believe our subsidy and risk-based cost estimates are conservative
for that time because (1) even though our estimates were based on the
market price of privately owned sovereign debt, in the long run, we
believe sovereign foreign debtors are more likely to pay off debt
owned by the private sector;\6 (2) we used the price of privately
owned sovereign debt, which was traded, as a proxy for privately
owned sovereign debt that was not traded, since traded debt is more
liquid and should command a higher price; and (3) we purposely made
an assumption that caused our estimates to be conservative.\7
Our estimates and those calculated using the method employed by the
executive branch of the international credit programs' subsidy and
country risk cost rates (estimated cost divided by funds lent or
guaranteed) are presented in table 1.
Table 1
Program Cost Rate Estimates for Fiscal
Year 1992 Credit Authorized Using the
Executive Branch's and Our Methods
(Rates in percents/Dollars in thousands)
GAO Execut
ve ive Credit
GAO branch branch authorized
--------------------- ------- ------- ------- ----------
AID housing guarantee AID housing 13.5 18.7
9.0 18.7 14.1 14.1
$83,000
guarantee
CCC CCC 25.7
GSM 102 GSM 102 25.1 7.4 8.0
25.7 8.0
5,446,615
CCC CCC 3.0
GSM 103 GSM 103 2.5 10.4 10.9
3.0 10.9 86,240
DSAA FMS DSAA FMS 9.5 6.0 5.7
10.1 6.0 5.7
345,000
EXIM guarantee EXIM 18.7 5.4 21.7
21.7 8.4 8.4
6,595,682
guarantee
EXIM loan EXIM loan 14.4 16.7
9.9 16.7 11.5 11.5
808,800
USDA USDA 28.4
P.L. 480 P.L. 75.4 72.8 25.9
28.4 25.9
368,110 480
All programs All programs 22.6
22.1 8.4 22.6 8.9
8.9 $13,733,447
------------------------------------------------------------
Note: AID, Agency for International Development; CCC, Commodity
Credit Corporation; GSM, General Sales Manager; DSAA, Defense
Security Assistance Agency; FMS, foreign military sales; EXIM,
Export-Import Bank; USDA, U.S. Department of Agriculture.
We also estimated 170 countries' long-run probabilities of default
based on systematic estimates from financial markets where privately
owned sovereign debt is traded. Our estimates varied depending upon
the riskiness of the countries and ranged from 92.1 percent for
Cambodia to 0 percent for the highest rated countries such as Japan,
Switzerland, and Germany. (See app. V for our estimates of the
long-run probabilities of default.)
Given the broad authorities contained in the statutes authorizing the
various loan programs, and the absence of any prohibitions to the
contrary, we have no reason to question the conclusion of the
executive branch that it has the authority to reschedule
international loans. In addition, we believe Office of Management
and Budget (OMB) guidance relevant to Paris Club rescheduling is
consistent with the Credit Reform Act's requirements.\8 However, an
OMB official told us that agencies were not following some aspects of
the OMB guidance. They had not been including the cost of possible
rescheduling at below-market interest rates in their initial
estimates, which is inconsistent with the act's requirements that the
budget include the full subsidy cost of credit programs in the year
in which the loan obligations or loan commitments are made. Also,
except for EXIM, agencies are not making annual re-estimates on
international loans and guarantees. Annual re-estimates by the
agencies should indicate any increases in costs; budget authority for
such cost increases are covered by a permanent indefinite
appropriation. (See app. VI for a further description of the legal
treatment of international debt rescheduling.)
--------------------
\2 Country risk costs are the costs due to the risk that
international loans or guarantees may not be fully repaid. Country
risk cost in the international context is analogous to default cost
in the domestic context. It is one component, albeit often the
largest, of the subsidy cost as defined in the Credit Reform Act.
\3 When we began this review, the ratings, market data, and
allocation of loans and guarantees we used were the most recent
available that were consistent with the methods employed.
\4 Subsidy costs of a loan are determined by the net sum of the
following components expressed in present value: (1) the interest
costs or the negative of the interest revenues, since a loan can be
made at rates less or greater than the U.S. government can borrow;
(2) the negative of the revenues received from any fees collected;
and (3) country risk cost.
\5 Throughout this report, we refer to the cost estimates we made
based on the fundamental principles of the executive branch's method
as "executive branch estimates" or "those by the executive branch."
Executive branch re-estimates made later in fiscal year 1993 would
not have changed.
\6 The private sector is fairly exclusively motivated to maximize its
profitability, whereas the U.S. government has a multiplicity of
other goals, including enhancing foreign policy objectives, promoting
U.S. defense goals, and helping domestic constituent interests. In
Jeremy Bulow, Kenneth Rogoff and Afonso S. Bevilaqua, "Official
Creditor Seniority and Burden Sharing in the Former Soviet Bloc,"
Brookings Papers on Economic Activity, no. 1 (1992), pp. 195-234,
the authors present empirical evidence, econometric tests, and
theoretical arguments that are suggestive of this view. Although it
is theoretically possible that concern about trade, defense, or
foreign policy might motivate sovereign debtors to preferentially
repay the U.S. government, in the long run this does not appear to
be supported by the evidence to date. In the future, however, this
question may be better tested econometrically and answered. In any
event, our method and the executive branch's method are both based on
the market price of privately owned debt. This consideration will
tend to bias both cost estimates in the same direction.
\7 We purposely underestimated subsidy costs by assuming all 1992
repayments on old loans up to the size of the new loans were due to
the new loans. Risk-based cost is high because many new loans to
recipients greatly exceeded the amount owed on old loans, partly
because funds then were being shifted toward countries of Eastern
Europe and the newly independent states of the former Soviet Union.
\8 The Paris Club is the mechanism the United States and other
official creditors use to reschedule debt from foreign countries that
are unable to meet their external debt obligations. Paris Club
meetings are organized by the French Finance Ministry. Traditional
participants of the Paris Club are the 24 members of the Organization
for Economic Cooperation and Development. The Department of State
represents the U.S. government in Paris Club negotiations.
BACKGROUND
------------------------------------------------------------ Letter :2
Before fiscal year 1992, the federal budget treated the cost of a new
loan in the year authorized as the net of loan disbursements minus
repayments in that year and the cost of a new guarantee as zero
(except when offset by origination fees.) Only if and when the U.S.
government paid out funds to settle claims on a previously made
guarantee did the cost of the guarantee appear in the U.S.
government budget. Consequently, the budget underestimated the
long-term costs of loan guarantees and overestimated that of direct
loans. As a result, agencies began issuing fewer loans and more
guarantees, even though, in principle, long-term country risk costs
for both should be about the same.
The Federal Credit Reform Act of 1990 corrected this distortion. It
required U.S. agencies, beginning in fiscal year 1992 to estimate
and budget for the long-term costs of a loan or guarantee in the year
authorized, using present-value analysis.\9
To implement the Credit Reform Act--which was enacted in November
1990--executive branch agencies with program responsibility for
foreign loans and guarantees each made their own cost estimates from
October to December 1990. In January and February 1991, many of
these estimates were criticized by congressional committees and the
Congressional Budget Office (CBO) for two reasons. First, some
agencies had not clearly defined and stated assumptions they had made
in arriving at their estimates. Second, one agency's stated
assumptions were often inconsistent with those made by another.
Consequently, the U.S. government formed a working group known as
the Interagency Country Risk Assessment System (ICRAS) to uniformly
evaluate for the executive branch the country risk contained in
foreign loans and guarantees.\10 To transform ICRAS' creditworthiness
measures into estimates of country risk costs, OMB, EXIM and the
Departments of State and Treasury formed a committee in October 1991
to devise uniform country risk interest rate premiums.
Three risk premium proposals were presented, which differed primarily
in regard to the risky countries--those rated lower than Moody's and
Standard and Poor's Baa/BBB ratings. A proposal by CBO, which
generally had the highest risk premiums and consequent country risk
cost estimates, was unacceptable to the Departments of State and
Treasury. The Departments of State and Treasury proposed the lowest
risk premiums for new debt issued after fiscal year 1991 (all under
credit reform) and higher risk premiums for older debt (not under
credit reform). They argued that this new debt was less likely to be
rescheduled at the Paris Club than the older debt, but OMB rejected
this proposal. The third proposal, presented by EXIM, along with
some of the work supporting the CBO proposal, formed the basis of the
risk premiums adopted by this committee.
OMB requires executive branch agencies to calculate the costs of
foreign loans and guarantees using annually updated ICRAS ratings
and, until now unchanged, country risk interest premiums when foreign
loans or guarantees are budgeted, authorized, disbursed, or modified.
Throughout the life of the loan or guarantee, OMB guidance requires
agencies to make annual re-estimates of costs.
The ultimate test of country risk and associated cost estimates is
the accuracy of these estimates, which can only be measured years
after estimates are made. However, meaningful professional judgments
about country risk and country risk cost methods can be made using
criteria such as whether (1) the method is well grounded in theory,
(2) it uses generally accepted statistical estimating methods, and
(3) sources of bias are identified and the estimates are qualified
for any bias.
The Credit Reform Act requires that the executive branch measure loan
and guarantee subsidy costs on a "cost to the U.S. government" basis
rather than on a "benefit to borrower" basis, which produces higher
cost estimates.\11 We measure cost on a cost to the U.S. government
basis. The executive branch and CBO also measure cost using the same
basis.
U.S. government agencies had $49.5 billion in international loans
and $48.7 billion in international guarantees outstanding on
September 30, 1992. (See app. I for a further description of the
Credit Reform Act and its implications.)
--------------------
\9 Present-value analysis calculates the value today of a future
stream of income or cost. A dollar available today is worth more
than a dollar in the future because it could have earned interest in
the interim. Conversely, a cost paid in the future is reduced when
valued today.
\10 OMB is Chairman and EXIM is Secretariat of ICRAS. Other ICRAS
members are the Departments of State, Treasury, and Agriculture;
DSAA; Overseas Private Investment Corporation; AID; the Council of
Economic Advisers; and the Federal Reserve Board. We recognize that
the Federal Reserve is not an agency of the executive branch, even
though we refer to the method that uses ICRAS ratings as the
executive branch method.
\11 Administrative costs are budgeted separately under credit reform.
EVALUATION OF THE EXECUTIVE
BRANCH'S METHOD
------------------------------------------------------------ Letter :3
A number of weaknesses exist in the method employed by the executive
branch to estimate country risk and the associated subsidy cost, the
most significant weakness being that the executive branch's method
was not based on econometric tests and measurements. Most other
weakness originated in the method employed by the risk premium
committee when setting the premiums for countries with ratings below
Baa/BBB, where most of the country risk cost occurs.
THE EXECUTIVE BRANCH'S
METHOD
---------------------------------------------------------- Letter :3.1
The executive branch's method for calculating the cost of
international loans and loan guarantees was a three-part process.
First, the executive branch rated countries for country risk. Then
the cost of country risk was calculated based on these ratings and
corresponding risk premiums assigned by the risk premium committee in
October 1991. Finally the subsidy cost was calculated, which, in
addition to country risk cost included interest rate cost (or income)
and fee income.
The executive branch's ratings were based on 35 subitems, most of
which were 5-year expectations of economic indicators. These 35
subitems were grouped in 5 subrating categories: payments history,
macroeconomics, the debt burden, balance of payments adjustment
capacity, and political and social factors. ICRAS scored the 35
subitems, 5 subrating categories, and the overall summary country
ratings on an 11-grade scale from "A" for the most creditworthy to
"F- -" for the least creditworthy.
During October 1991 the risk premium committee developed its method
for transforming ICRAS credit ratings into risk premium used to
measure country risk cost. The risk premium for ICRAS' top three
categories, A, B, and C were based on 19 bond observations. These
three ICRAS categories correspond to Moody's and Standard and Poor's
ratings Baa/BBB or higher. Historical average 3-year risk premiums
for bonds with these same ratings as the bond observations became the
risk premiums for these three ICRAS categories. The risk premium
committee based the risk premium for the next four ICRAS categories,
C-, D, D-, and E, on only five observations of bonds--three Ba/BB
rated bonds and two B/B rated bonds. The risk premium committee
derived 16 risk premiums for these 4 ICRAS categories from the risk
premiums for these 5 observations, without using econometric
methods.\12
The risk premium committee extrapolated risk premiums from the E
category to obtain risk premium for the last 4 categories E-, F, F-
and F- -. In these categories, the executive branch's method assumed
that the level of country risk did not affect the relative size of
one maturity's risk premiums compared to another's. At least one
member of the committee checked to ensure that the risk premiums for
the long-term debt were roughly consistent with the general range of
prices from the secondary market. The executive branch then
calculated subsidy costs using present value analysis.
--------------------
\12 The four ICRAS rating categories multiplied by four maturity
categories--1, 5, 10, and 30 years.
WEAKNESSES OF THE EXECUTIVE
BRANCH'S METHOD
---------------------------------------------------------- Letter :3.2
Although we recognize that the executive branch had only a short
amount of time after enactment of the Credit Reform Act to develop
appropriate methods to estimate country risk and its cost, we found
that the method it developed and employed had several weaknesses, the
principal one being the lack of econometric tests and measurements.
The most significant weakness originated in the method employed by
the risk premium committee, especially for countries with ratings
below Baa/BBB where most of the country risk cost occurs. In
particular, the risk premium committee's method
employed too little data below ICRAS C rated countries for the large
amount of information needed;
frequently assumed that the level of country risk had no effect on
the relative size of the risk premiums for different maturities;
obtained the same risk premiums for both loans and bonds, resulting
in executive branch country risk cost estimates for loans being less
than for bonds, which was the opposite of what we found on the
emerging
market;\13
had not been updated, even though new information from financial
markets had been available and OMB officials had told us throughout
our review that the risk premiums for short-term loans to the
riskiest countries were too low;
did not disclose the sources of bias, except for acknowledging that
the risk premiums of short-term loans to the riskiest countries were
too low; and
made no distinction between new loans to risky countries that have
relatively large effects on old loan repayment, cause relatively
small increases in U.S. government exposure, and thereby have
relatively small country risk costs, and those new loans to risky
countries that do not.
Additionally, we found some weaknesses in ICRAS' rating method. Most
of the 35 subitems ICRAS used to form sovereign risk ratings were
ICRAS' 5-year expectations of economic, political, and social
variables that are intuitively appealing predictors of the ability
and willingness to meet loan obligations; however, many subitems did
not pass econometric tests for significance in the professional
economic literature. In addition, subitems were not combined with
weights based on econometric estimations, and the subitems'
predictive ability was not tested.
Because of weaknesses in estimating country risk and its costs,
executive branch estimates could be influenced by external
considerations, such as pressure to grant loans or guarantees to
particular countries. This could weaken agencies' ability to make
sound lending decisions and the Congress' ability to make sound
funding decisions. For example, some executive branch officials told
us that in 1992 officials from the Departments of State and Treasury
exerted some pressure to raise executive branch ratings, mostly for
countries in Central and Eastern Europe and the newly independent
states of the former Soviet Union, although the reported effects of
that pressure were modest.
--------------------
\13 This situation occurs because principal is generally due on bonds
at maturity but is due for loans from issuance until maturity. As a
result, the average payment due on a bond occurs further in the
future, and, other things being equal, the executive branch discounts
it more heavily.
OUR METHOD
------------------------------------------------------------ Letter :4
Our method for calculating the costs of international loans and
guarantees is also a three-step process. However, unlike the
executive branch, our method is based on the rigorous use of
econometric tests and measurements, and distinguishes between new
loans that affect repayments on old loans owed the U.S. government
(rollovers) and new loans that increase exposure.
We first obtained continuous country risk ratings by statistically
combining those that appeared in September 1992 in two leading
magazines for financial institutions. Second, we transformed our
credit ratings into risk premiums. We obtained the risk premiums
relative to the riskless Aaa/AAA rate for less risky debt, Baa/BBB or
higher, directly from financial markets in which similar privately
owned bonds are traded and transform them into country risk values of
debt.\14 We obtained the country risk value of risky U.S.
government-owned sovereign loans from the secondary market in which
similar privately owned loans are traded, using econometric tests and
measurements. For both risky and nonrisky loans, we estimated the
value of debt that is a complete rollover; that is, it causes an
equal amount of old loan repayment. We then estimated the extent
that each new loan causes old debt repayments, the average propensity
to repay old loans, and the remaining part increases U.S. government
exposure. To obtain the country risk value of new loans, we averaged
these country risk values of sovereign debt; one for complete
rollovers, the other for complete increased exposure. We then
transformed these country risk new loan values into country risk new
loan interest rate premiums above the Aaa/AAA rate using present
value analysis.
Finally, like the executive branch, we calculated subsidy cost using
present value analysis. (For a more detailed explanation see app.
III.)
--------------------
\14 For the most creditworthy debt, we used bonds because we had no
loan information.
COUNTRY RISK RATINGS
------------------------------------------------------------ Letter :5
The executive branch's country risk ratings developed in the last
half of fiscal year 1992 for the fiscal year 1994 budget process had
some similarity but had important differences from (1) two
contemporaneous, professionally recognized, country risk ratings and
our country risk ratings that statistically averaged these
professionally recognized ratings and (2) contemporaneous prices on
the secondary market of privately owned, dollar-denominated, variable
interest rate debt owed by developing country governments. The
executive branch's country risk ratings differed from each of these
external standards substantially more than these external standards
differed from each other.
RECOMMENDATIONS
------------------------------------------------------------ Letter :6
We recommend that the Director, OMB, revise the executive branch's
method of estimating the cost of country risk so that it
econometrically utilizes available information from the secondary
market on how prices are affected by country risk and other debt
characteristics. We also recommend that the Director of OMB ensure
that (1) ICRAS' rating method is revised so that it makes greater use
of empirically tested criteria and weights, (2) agencies make initial
estimates of subsidy costs that include estimates of potential
rescheduling at below-market rates, and (3) agencies make annual
re-estimates of subsidy costs.
MATTER FOR CONGRESSIONAL
CONSIDERATION
------------------------------------------------------------ Letter :7
A stated purpose of the Credit Reform Act was to measure more
accurately the costs of U.S. government credit programs so that
better decisions could be made on the allocation of resources among
credit programs and between credit and other spending programs.
Making a distinction between new loans to risky countries that have
relatively large effects on old loan repayments and those new loans
that do not would improve the quality of the subsidy cost estimates.
Because of the greater difficulty in making risk estimates for
international credit programs, the Congress may wish to consider how
the principles included in the Credit Reform Act ought to be applied
to direct international lending and lending guaranteed by the U.S.
government.
AGENCY COMMENTS AND OUR
EVALUATION
------------------------------------------------------------ Letter :8
OMB provided comments on this report on behalf of itself; the
Departments of Agriculture, State, and the Treasury; and EXIM. (See
app. VII.) OMB generally agreed with the economics we employed to
measure country risk and its cost and acknowledged some weaknesses in
ICRAS' method for estimating costs.
However, OMB indicated that the Credit Reform Act precluded it from
making distinctions in its cost estimates between new loans that
affect old loan repayments (rollovers) and those loans that do not.
To overcome the problem, we are suggesting that the Congress may wish
to consider how to apply credit reform principles to international
credit programs. The question of rollovers should be part of that
consideration.
OMB also believed that, even if this distinction were permitted,
estimating subsidy costs based on new loan rollovers was impractical
because it would involve estimating thousands of fluctuating subsidy
estimates for 167 separate countries. After carefully considering
OMB's comment, we believe OMB overstated the nature of the problem.
For example, not all of these countries would be involved in a
subsidy estimate; we made the calculations for 70 countries.
Furthermore, calculations could be simplified by (1) statistically
relating variable interest rate, dollar-denominated sovereign loan
prices to grouped ratings; (2) estimating by credit rating group,
agency loan, and guarantee allocation levels; and (3) calculating two
scenarios--one high cost without rollover and another low cost with
maximum rollover, then deciding what the leading point estimate
should be. Since one of the primary goals of the Credit Reform Act
was to more accurately measure costs of U.S. government credit
programs, we believe that making this distinction would improve the
quality of subsidy estimates.
OMB agreed that prices on the secondary market formed the best basis
for making cost estimates; however, it said that secondary market
prices varied substantially. It expressed concern that our cost
estimates depended on observations during a 2-week period 2 years
ago. We agree that our cost estimates depend upon market
observations during this 2- week period. Our draft report qualified
our estimates by pointing out that differences in estimates depend
upon market expectations for a particular time and group of foreign
loans and will generally change if either of these conditions are
affected. It also pointed out that our cost estimates for risky
countries were based on dollar denominated, variable interest rate
loans owed or guaranteed by risky countries' governments. In
contrast, the secondary market index OMB referred to in its comments
included many other types of debt, including foreign currency debt,
fixed-rate debt, bonds, and private nonsovereign guaranteed debt.
Such an index is more likely to vary because of movements in world
interest rates and foreign exchange rates than the type of loans that
we used for the relevant comparison.
SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :9
We conducted our review in Washington, D.C., and collected data and
discussed various issues in this report with members of ICRAS and
officials of OMB; CBO; EXIM; the Overseas Private Insurance
Corporation; AID; and the Departments of Treasury, State,
Agriculture, and Defense. We evaluated the executive branch's
country risk ratings for the fiscal year 1994 budget against
contemporaneous external standards for measuring country risk and for
internal consistency, mostly through the use of regression analysis.
We also evaluated its rating method in terms of the degree that it
had support from econometric tests. We developed our own country
risk ratings by statistically averaging two contemporaneous
professionally accepted ratings. We further developed our method of
transforming country risk ratings into country risk cost using
econometric estimates on data derived from the secondary market. A
complete description of our method to estimate country risk and its
cost is contained in appendix III.
We performed our review from December 1992 to October 1994 in
accordance with generally accepted government auditing standards.
---------------------------------------------------------- Letter :9.1
We are sending copies of this report to the Director, OMB, the
President, EXIM, and the Secretary of the Treasury. We will also
make copies available to others on request.
Please contact us on (202) 512-4128 and (202) 512-4812, respectively,
if you or your staff have any questions on this report. Major
contributors to this report are listed in appendix VIII.
Harold J. Johnson
Director, International
Affairs Issues
Allan I. Mendelowitz
Managing Director
International Trade,
Finance, and Competitiveness Issues
BACKGROUND: CREDIT REFORM
=========================================================== Appendix I
This background appendix also is a part of three other reports on
credit reform implementation: an evaluation of the use of negative
subsidy credit receipts, an evaluation of decisions to include
certain programs under the Federal Credit Reform Act, and the use of
estimated future credit savings to offset current spending.\1
The federal government uses direct loans and loan guarantees as tools
to achieve numerous program objectives such as assistance to housing,
agriculture, education, small businesses, and foreign governments.
At the end of fiscal year 1993, the face value of the government's
direct loans and loan guarantees totaled a reported $861 billion, of
which $201 billion was in direct loans and $660 billion was in loan
guarantees.
After over 20 years of discussion about the shortcomings of using
cash budgeting for credit programs and activities, the Federal Credit
Reform Act of 1990 was enacted on November 5, 1990, as title 13B of
the Omnibus Budget Reconciliation Act of 1990, Public Law 101-508.
The Credit Reform Act changed the budget treatment of credit programs
so that their costs can be compared more accurately with each other
and with the costs of other federal spending. It also was intended
to ensure that the full cost of credit programs over their entire
lives would be reflected in the budget when the loans were made so
that the executive branch and the Congress might consider them when
making budget decisions.
In addition, it was recognized that credit programs had different
economic effects than most budget outlays, such as purchases of goods
and services, income transfers, and grants. In the case of direct
loans, for example, the fact that the loan recipient was obligated to
repay the government over time meant that the economic impact of a
direct loan disbursement could be much less than other budget
transactions of the same dollar amount.
--------------------
\1 See Credit Reform: Appropriation of Negative Subsidy Receipts
Raises Questions (GAO/AIMD-94-58, Sept. 26, 1994), Credit Reform:
Case-by-Case Assessment Advisable in Evaluating Coverage and
Compliance (GAO/AIMD-94-57, July 28, 1994), and Credit Reform:
Speculative Savings Used to Offset Current Spending Increase Budget
Uncertainty (GAO/AIMD-94-46, Mar. 18, 1994).
CREDIT REFORM WAS DESIGNED TO
REMOVE DIFFICULTIES CAUSED BY
CASH TREATMENT
--------------------------------------------------------- Appendix I:1
Before credit reform, it was difficult to make appropriate cost
comparisons between direct loan and loan guarantee programs and
between credit and noncredit programs. Credit reform requirements
were formulated to address the factors that caused this problem.
Two key principles of credit reform are (1) the definition of cost in
terms of the present value of cash flow over the life of a credit
instrument and (2) the inclusion in the budget of the costs of credit
programs in the year in which the budget authority is enacted and the
direct or guaranteed loans are disbursed.
CREDIT REFORM WAS DESIGNED TO
ALLOW APPROPRIATE COST
COMPARISONS
--------------------------------------------------------- Appendix I:2
Before credit reform, credit programs--like other programs--were
reported in the budget on a cash basis. This cash basis distorted
costs and, thus, the comparison of credit program costs with other
programs intended to achieve similar purposes, such as grants. It
also created a bias in favor of loan guarantees over direct loans.
Loan guarantees appeared to be free while direct loans appeared to be
very expensive because the budget did not recognize that at least
some of the loan guarantees would default and that some of the direct
loans were to be repaid.
For direct loans, the budget showed budget authority and outlays in
the amount that loan disbursements exceeded repayments received in
that budget year. This cash approach overstated direct loan costs in
the initial years of a program when loan disbursements were likely to
be greater than repayments. Conversely, this treatment understated
costs in later years when loan repayments were more likely to be much
larger relative to disbursements. Cash-based budgeting did not
recognize that at least a portion of the loan outlays would be repaid
in the future. In contrast, for loan guarantees, the budget did not
record any budget authority or outlays when the guarantees were made
(except the negative outlay resulting from any origination fees),
even though they were likely to entail future losses. It showed
budget authority and outlays only when, and if, defaults occurred.
Credit reform changed this treatment for direct loans and loan
guarantees made on or after October 1, 1991. It required that budget
authority to cover the cost to the government of new loans and loan
guarantees (or modifications to existing credit instruments) be
provided before the loans, guarantees, or modifications are made.
Credit reform requirements specified a net cost approach using
estimates for future loan repayments and defaults as elements of the
cost to be recorded in the budget. This puts direct loans and loan
guarantees on an equal footing; it permits the costs of credit
programs to be compared with each other and with the costs of
noncredit programs when making budget decisions.
CREDIT REFORM IDENTIFIES THE
GOVERNMENT'S COST OF CREDIT
ACTIVITIES
--------------------------------------------------------- Appendix I:3
Credit reform requirements separate the government's cost of
extending or guaranteeing credit, called the subsidy cost, from
administrative and unsubsidized program costs. Administrative
expenses receive separate appropriations. They are treated on a cash
basis and reported separately in the budget. The unsubsidized
portion of a direct loan is expected to be recovered from the
borrower.
The Credit Reform Act defines the subsidy cost of direct loans as the
present value--over the loan's life--of disbursements by the
government (loan disbursements and other payments) minus estimated
payments to the government (repayments of principal, payments of
interest, and other payments) after adjusting for projected defaults,
prepayments, fees, penalties, and other recoveries. It defines the
subsidy cost of loan guarantees as the present value of cash flows
from estimated payments by 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).
According to Office of Management and Budget guidance, credit
programs have a positive subsidy, that is, they lose money, when the
present value of estimated payments by the government exceeds the
present value of estimated receipts. Conversely, negative subsidy
programs are those in which the present value of estimated
collections is expected to exceed the present value of estimated
payments; in other words, the programs make money (aside from
administrative expenses).
CREDIT PROGRAMS NOW USE THREE
BUDGETARY ACCOUNTS
--------------------------------------------------------- Appendix I:4
The Credit Reform Act set up a special budget accounting system to
record the budget information necessary to implement credit reform.
It provides for three types of accounts--program, financing, and
liquidating--to handle credit transactions.
Credit obligations and commitments made on or after October 1,
1991--the effective date of credit reform--use only the program and
financing accounts. The program account receives separate
appropriations for administrative and subsidy costs of a credit
activity and is included in budget totals. When a direct or
guaranteed loan is disbursed, the program account pays the associated
subsidy cost for that loan to the financing account. The financing
account, which is nonbudgetary, is used to record the cash flow
associated with direct loans or loan guarantees over their lives.\2
It finances loan disbursements and the payments for loan guarantee
defaults with (1) the subsidy cost payment from the program account,
(2) borrowing from the Treasury, and (3) collections received by the
government. Figure I.1 diagrams this cash flow.
Figure I.1: Credit Reform Cash
Flow Simplified
(See figure in printed
edition.)
If subsidy cost calculations are accurate, the financing account will
break even over time as it uses its collections to repay its Treasury
borrowing.
Direct loans and loan guarantees made before October 1, 1991, are
reported on a cash basis in the liquidating account. This account
continues the cash budgetary treatment used before credit reform. It
has permanent, indefinite budget authority\3 to cover any losses.
Excess balances are transferred periodically--at least annually--to
the Treasury.
In addition to the three accounts specified in the Credit Reform Act,
OMB has directed that credit programs or activities with negative
subsidies must have special fund receipt accounts to hold receipts
generated when the program or activity shows a profit. OMB guidance
provides that these funds cannot be used unless appropriated.
--------------------
\2 Nonbudgetary accounts may appear in the budget document for
information purposes but are not included in the budget totals for
budget authority or budget outlay. They do not belong in the budget
because they show only how something is financed, and do not
represent the use of resources.
\3 Permanent budgetary authority is available as a result of
permanent legislation and does not require annual appropriation.
Indefinite budget authority is budget authority of an unspecified
amount of money.
OMB AND TREASURY PROVIDE
IMPLEMENTATION GUIDANCE
--------------------------------------------------------- Appendix I:5
OMB and the Department of the Treasury provide guidance on
implementing credit reform. OMB's written guidance is contained
primarily in OMB Circulars A-11, A-34, and A-129.\4 OMB also has
issued memorandums to provide additional implementation guidance
addressing specific situations. The Treasury's guidance is provided
in materials such as Basic Transactions Relating to Guaranteed Loans
and Subsidies (Apr. 30, 1992), which contains a number of
illustrative cases developed by its Financial Management Service and
distributed to agencies as examples of how to account for credit
reform transactions.
--------------------
\4 OMB Circular A-11 is entitled Preparation and Submission of Budget
Estimates; Circular A-34 is entitled Instructions on Budget
Execution; Circular A-129 is entitled Managing Federal Credit
Programs.
INDIVIDUAL PROGRAM
CHARACTERISTICS RAISE CREDIT
IMPLEMENTATION QUESTIONS
--------------------------------------------------------- Appendix I:6
Fiscal year 1994 is the third year that credit programs have been
required to comply with credit reform. Both agencies that operate
credit programs and those that provide implementation guidance--OMB
and Treasury--have had to address a variety of situations for which
the Credit Reform Act does not provide explicit direction. Questions
have arisen and continue to arise as the agencies implement credit
reform. Several groups have been created, such as the Federal Credit
Policy Working Group, the Credit Reform Steering Committee, and
Interagency Country Risk Assessment System (ICRAS) to address these
implementation issues and questions.
OUR ESTIMATES OF COUNTRY RISK AND
ITS COST
========================================================== Appendix II
This appendix presents our estimates of the cost of country risk in a
variety of forms, although all cost and cost rate estimates are on a
cost to the U.S. government basis. We first present our estimates
of the country risk cost rate (country risk cost divided by loan
size) for each of 170 countries, one set of estimates for loans (or
loan guarantees) that have no effect on old loan repayment in table
II.1 and a set of estimates for those loans that have a maximum
effect on old loan repayment in table II.2. In table II.3 we present
our country program estimates of the cost of country risk for fiscal
year 1992 authorized international loans and loan guarantees. In
table II.4 we present our program country risk cost rate estimates,
those from a sensitivity test, and those from the executive branch.
Table II.1 presents our continuous and grouped country risk rating
estimates for each of 170 countries and, by maturity, our estimates
of the new loan country risk cost rate when new loans have no effect
on old loan repayments. For the most creditworthy borrowers--those
with our group rating of A--our country risk cost rate estimate is
zero.
Table II.1
Our Ratings and Estimates of Country
Risk Cost Rates When New Loans Have No
Effect on Old Loan Repayment
(Rates in percents)
Our Our Maturity Maturity Maturity Maturity
Country rating group 1 year 5 years 10 years 30 years
------------------ ---------- -------- -------- -------- -------- --------
Afghanistan 15.1 F- 82.4 82.8 82.9 83.4
Albania 17.9 F- 78.9 79.3 79.5 80.0
Algeria 41.4 E- 49.2 49.6 49.9 50.8
Angola 22.8 F 72.7 73.1 73.4 74.0
Antigua and 26.3 F 68.3 68.7 69.0 69.7
Barbuda
Argentina 44.5 E 45.3 45.7 45.9 46.8
Armenia 16.2 F- 81.1 81.4 81.6 82.1
Australia 94.0 B 0.2 0.8 1.3 3.4
Austria 97.7 A 0 0 0 0
Azerbaijan 16.4 F- 80.8 81.1 81.3 81.8
Bahamas 75.0 C- 0.8 2.4 4.3 10.6
Bahrain 81.1 C- 0.8 2.4 4.3 10.6
Bangladesh 27.8 F 66.4 66.8 67.1 67.9
Barbados 50.1 E 38.2 38.6 38.8 39.6
Belarus 24.8 F 70.2 70.6 70.9 71.6
Belgium 97.0 B 0.2 0.8 1.3 3.4
Belize 48.3 E 40.5 40.8 41.1 41.9
Benin 15.9 F- 81.4 81.8 81.9 82.4
Bhutan 21.3 F- 74.6 75.0 75.2 75.9
Bolivia 28.4 F 65.6 66.1 66.3 67.1
Bosnia-Herzegovia 22.5 F 73.1 73.5 73.7 74.4
Botswana 52.9 D- 34.6 35.0 35.2 36.0
Brazil 39.6 E- 51.5 51.9 52.2 53.1
Brunei 86.6 C 0.4 1.2 2.2 6.2
Bulgaria 29.3 F 64.5 64.9 65.2 66.0
Burkina Faso 26.5 F 68.0 68.5 68.7 69.5
Burma (Myanmar) 16.5 F- 80.7 81.0 81.2 81.7
Burundi 27.1 F 67.3 67.7 68.0 68.7
Cambodia 7.4 F-- 92.2 92.3 92.4 92.6
Cameroon 32.0 F 61.1 61.5 61.8 62.6
Canada 97.3 A 0 0 0 0
Cape Verde 19.4 F- 77.0 77.4 77.6 78.2
Central African 24.7 F 70.3 70.7 71.0 71.7
Republic
Chad 22.2 F- 73.5 73.9 74.1 74.8
Chile 76.1 C- 0.8 2.4 4.3 10.6
China, People's 76.7 C- 0.8 2.4 4.3 10.6
Republic
Colombia 59.7 D- 26.0 26.3 26.5 27.1
Congo 18.5 F- 78.1 78.5 78.7 79.3
Costa Rica 35.7 E- 56.4 56.9 57.1 58.0
Cote d'Ivoire 23.0 F 72.5 72.9 73.1 73.8
Croatia 25.1 F 69.8 70.2 70.5 71.2
Cuba 11.8 F-- 86.6 86.9 87.0 87.4
Cyprus 71.8 C- 0.8 2.4 4.3 10.6
Czechoslovakia 66.1 D 17.9 18.1 18.2 18.7
Denmark 95.8 B 0.2 0.8 1.3 3.4
Djibouti 27.5 F 66.8 67.2 67.5 68.2
Dominican Republic 24.6 F 70.4 70.9 71.1 71.8
Ecuador 29.0 F 64.9 65.3 65.6 66.4
Egypt 35.1 E- 57.2 57.6 57.9 58.7
El Salvador 23.3 F 72.1 72.5 72.7 73.4
Estonia 27.3 F 67.0 67.5 67.7 68.5
Ethiopia 17.4 F- 79.5 79.9 80.1 80.6
Fiji 49.6 E 38.8 39.2 39.4 40.3
Finland 94.6 B 0.2 0.8 1.3 3.4
France 97.9 A 0 0 0 0
Gabon 39.4 E- 51.7 52.2 52.4 53.3
Gambia, The 33.8 E- 58.8 59.3 59.5 60.4
Georgia 18.7 F- 77.9 78.3 78.5 79.0
Germany 98.3 A 0 0 0 0
Ghana 34.6 E- 57.8 58.2 58.5 59.4
Greece 80.0 C- 0.8 2.4 4.3 10.6
Grenada 17.9 F- 78.9 79.3 79.5 80
Guatemala 23.7 F 71.6 72.0 72.2 72.9
Guinea 28.2 F 65.9 66.3 66.6 67.4
Guinea-Bissau 20.2 F- 76.0 76.4 76.6 77.2
Guyana 14.1 F- 83.7 84.0 84.2 84.6
Haiti 16.7 F- 80.4 80.8 81.0 81.5
Honduras 22.4 F- 73.2 73.6 73.9 74.5
Hong Kong 91.6 C 0.4 1.2 2.2 6.2
Hungary 64.0 D 20.6 20.8 20.9 21.5
Iceland 88.1 C 0.4 1.2 2.2 6.2
India 52.9 D- 34.6 35.0 35.2 36.0
Indonesia 75.5 C- 0.8 2.4 4.3 10.6
Iran 36.7 E- 55.1 55.6 55.9 56.7
Iraq 11.0 F-- 87.6 87.9 88.0 88.3
Ireland 93.8 B 0.2 0.8 1.3 3.4
Israel 70.8 C- 0.8 2.4 4.3 10.6
Italy 95.2 B 0.2 0.8 1.3 3.4
Jamaica 27.4 F 66.9 67.3 67.6 68.3
Japan 98.5 A 0 0 0 0
Jordan 29.9 F 63.7 64.2 64.4 65.2
Kazakhstan 23.2 F 72.2 72.6 72.9 73.5
Kenya 37.8 E- 53.8 54.2 54.5 55.3
Korea, Democratic 11.1 F-- 87.5 87.7 87.9 88.2
People's Republic
Korea, Republic of 89.5 C 0.4 1.2 2.2 6.2
Kuwait 78.6 C- 0.8 2.4 4.3 10.6
Kyrgyzstan 18.7 F- 77.9 78.3 78.5 79.0
Latvia 26.1 F 68.5 69.0 69.2 70
Lebanon 18.3 F- 78.4 78.8 79.0 79.5
Lesotho 36.9 E- 54.9 55.3 55.6 56.5
Liberia 12.6 F- 85.6 85.9 86.0 86.4
Libya 32.4 E- 60.6 61.0 61.3 62.1
Lithuania 26.3 F 68.3 68.7 69.0 69.7
Luxembourg 96.8 B 0.2 0.8 1.3 3.4
Macedonia 26.9 F 67.5 68.0 68.2 69.0
Madagascar 26.1 F 68.5 69.0 69.2 70
Malawi 20.4 F- 75.7 76.1 76.4 77.0
Malaysia 87.0 C 0.4 1.2 2.2 6.2
Mali 26.4 F 68.2 68.6 68.8 69.6
Malta 79.6 C- 0.8 2.4 4.3 10.6
Mauritainia 20.1 F- 76.1 76.5 76.7 77.3
Mauritius 49.1 E 39.4 39.8 40.1 40.9
Mexico 67.6 D 16.0 16.2 16.3 16.8
Moldova 15.8 F- 81.6 81.9 82.1 82.6
Mongolia 18.6 F- 78.0 78.4 78.6 79.2
Morocco 47.1 E 42.0 42.4 42.6 43.5
Mozambique 12.5 F- 85.7 86.0 86.1 86.5
Namibia 29.8 F 63.9 64.3 64.6 65.4
Nepal 32.1 F 61.0 61.4 61.7 62.5
Netherlands 98.2 A 0 0 0 0
New Zealand 92.3 C 0.2 0.8 1.3 3.4
Nicaragua 13.5 F- 84.5 84.8 84.9 85.3
Niger 32.3 F 60.7 61.2 61.4 62.2
Nigeria 28.0 F 66.1 66.6 66.8 67.6
Norway 96.3 B 0.2 0.8 1.3 3.4
Oman 73.2 C- 0.8 2.4 4.3 10.6
Pakistan 41.8 E- 48.7 49.1 49.4 50.3
Panama 27.7 F 66.5 67.0 67.2 68.0
Papua New Guinea 48.3 E 40.5 40.8 41.1 41.9
Paraguay 34.5 E- 57.9 58.4 58.6 59.5
Peru 19.1 F- 77.4 77.8 78.0 78.5
Phillipines 35.8 E- 56.3 56.7 57.0 57.9
Poland 37.2 E- 54.5 55.0 55.2 56.1
Portugal 92.3 C 0.4 1.2 2.2 6.2
Qatar 76.8 C- 0.8 2.4 4.3 10.6
Romania 36.9 E- 54.9 55.3 55.6 56.5
Russia 26.8 F 67.7 68.1 68.3 69.1
Rwanda 27.0 F 67.4 67.8 68.1 68.8
Sao Tome and 18.1 F- 78.7 79.0 79.2 79.8
Principe
Saudi Arabia 86.5 C 0.4 1.2 2.2 6.2
Senegal 29.1 F 64.8 65.2 65.5 66.2
Seychelles 27.7 F 66.5 67.0 67.2 68.0
Sierra Leone 12.8 F- 85.3 85.6 85.8 86.2
Singapore 96.3 B 0.2 0.8 1.3 3.4
Slovakia 53.2 D- 34.2 34.6 34.8 35.6
Slovenia 32.5 E- 60.5 60.9 61.2 62.0
Somalia 10.5 F-- 88.2 88.5 88.6 88.9
South Africa 61.7 D- 23.5 23.7 23.9 24.5
Spain 95.7 B 0.2 0.8 1.3 3.4
Sri Lanka 35.0 E- 57.3 57.7 58.0 58.9
St. Lucia 26.4 F 68.2 68.6 68.8 69.6
St. Vincent and 37.4 E- 54.3 54.7 55.0 55.8
Grenadines
Sudan 12.1 F-- 86.2 86.5 86.6 87.0
Sweden 96.1 B 0.2 0.8 1.3 3.4
Switzerland 98.5 A 0 0 0 0
Syria 32.0 F 61.1 61.5 61.8 62.6
Taiwan 95.7 B 0.2 0.8 1.3 3.4
Tajikistan 17.7 F- 79.2 79.5 79.7 80.2
Tanzania 18.4 F- 78.3 78.6 78.8 79.4
Thailand 83.5 C 0.4 1.2 2.2 6.2
Togo 26.4 F 68.2 68.6 68.8 69.6
Trindad and Tobago 43.1 E 47.0 47.5 47.7 48.6
Tunisia 58.4 D- 27.7 28.0 28.1 28.8
Turkey 72.8 C- 0.8 2.4 4.3 10.6
Turkmenistan 19.5 F- 76.9 77.3 77.5 78.1
Uganda 16.4 F- 80.8 81.1 81.3 81.8
Ukraine 25.9 F 68.8 69.2 69.5 70.2
United Arab 85.7 C 0.4 1.2 2.2 6.2
Emirates
United Kingdom 97.7 A 0 0 0 0
Uruguay 51.2 E 36.8 37.2 37.4 38.2
U.S.S.R. 26.8 F 67.7 68.1 68.3 69.1
Uzbekistan 20.6 F- 75.5 75.9 76.1 76.7
Vanuatu 37.1 E- 54.6 55.1 55.3 56.2
Venezuela 59.2 D- 26.6 26.9 27.1 27.8
Viet Nam 22.7 F 72.8 73.3 73.5 74.1
Yemen, Republic of 26.2 F 68.4 68.9 69.1 69.8
Yugoslavia 20.5 F- 75.6 76.0 76.2 76.8
Zaire 17.1 F- 79.9 80.3 80.5 81.0
Zambia 16.4 F- 80.8 81.1 81.3 81.8
Zimbabwe 42.7 E 47.5 48.0 48.2 49.1
--------------------------------------------------------------------------------
Table II.2 presents our estimates of country risk cost rates, by
country, when each dollar of new loan causes one dollar of repayment
on old loans. Countries with our risk rating score of less than 67.7
percent are not presented, but they have a country risk cost rate of
zero.
Table II.2
Our Country Risk Cost Rate Estimates by
Country When Each Dollar of New Loan
Causes One Dollar of Old Loan Repayment
(Rates in percents)
Maturity Maturity Maturity Maturity
Country 1 year 5 years 10 years 30 years
-------------------- -------- -------- -------- --------
Australia 0.1 0.7 1.2 3.2
Austria 0 0 0 0
Bahamas 0.4 2.0 3.9 10.1
Bahrain 0.4 2.0 3.9 10.1
Belgium 0.1 0.7 1.2 3.2
Brunei 0.2 1.0 2.0 5.9
Canada 0 0 0 0
Chile 0.4 2.0 3.9 10.1
China, People's 0.4 2.0 3.9 10.1
Republic
Cyprus 0.4 2.0 3.9 10.1
Denmark 0.1 0.7 1.2 3.2
Finland 0.1 0.7 1.2 3.2
France 0 0 0 0
Germany 0 0 0 0
Greece 0.4 2.0 3.9 10.1
Hong Kong 0.2 1.0 2.0 5.9
Iceland 0.2 1.0 2.0 5.9
Indonesia 0.4 2.0 3.9 10.1
Ireland 0.1 0.7 1.2 3.2
Israel 0.4 2.0 3.9 10.1
Italy 0.1 0.7 1.2 3.2
Japan 0 0 0 0
Korea, Republic of 0.2 1.0 2.0 5.9
Kuwait 0.4 2.0 3.9 10.1
Luxembourg 0.1 0.7 1.2 3.2
Malaysia 0.2 1.0 2.0 5.9
Malta 0.4 2.0 3.9 10.1
Netherlands 0 0 0 0
New Zealand 0.1 0.7 1.2 3.2
Norway 0.1 0.7 1.2 3.2
Oman 0.4 2.0 3.9 10.1
Portugal 0.2 1.0 2.0 5.9
Qatar 0.4 2.0 3.9 10.1
Saudi Arabia 0.2 1.0 2.0 5.9
Singapore 0.1 0.7 1.2 3.2
Spain 0.1 0.7 1.2 3.2
Sweden 0.1 0.7 1.2 3.2
Switzerland 0 0 0 0
Taiwan 0.1 0.7 1.2 3.2
Thailand 0.2 1.0 2.0 5.9
Turkey 0.4 2.0 3.9 10.1
United Arab Emirates 0.2 1.0 2.0 5.9
United Kingdom 0 0 0 0
------------------------------------------------------------
Table II.3 presents our country program estimates of the credit
reform cost rates for fiscal year 1992 authorized international loans
and guarantees on a cost to the U.S. government basis. We estimate
the total subsidy cost for all of these programs to be $3 billion, or
22.1 percent of the $13.7 billion of loans and guarantees. The total
country risk cost for all of these programs was $3.1 billion. Each
program had a positive subsidy cost, consisting mostly of country
risk cost, except for Public Law 480, which has heavily subsidized
loan interest.\1 On the basis of cost rates, Public Law 480 was the
most costly program--75.4 percent--because of its high interest cost
(47 percent) and country risk cost (28.4 percent); Commodity Credit
Corporation's (CCC) General Sales Manager (GSM) 102 program had a
slightly lower rate of country risk cost, 25.7 percent.\2 CCC GSM 103
program and the Defense Security Assistance Agency's (DSAA) Foreign
Military Sales (FMS) were the least costly programs, at 2.5 and 9.5
percent, respectively.
Table II.3
Our Cost Rate Estimates by Country and
Program for Fiscal Year 1992 Authorized
International Loans and Guarantees
(Rates in percents/Dollars in thousands)
Loans
and
guarante
Subsid Risk Interest Fee es
y cost cost cost income authoriz
Program/country rate rate rate rate ed
---------------- ------ ------ -------- ------ --------
AID housing 13.5 18.7 0 5.2 $83,000
guarantee
total\a
India 17.7 23.1 0 5.4 28,000
Indonesia 8.4 13.5 0 5.1 20,000
Portugal -5.1 0 0 5.1 15,000
Tunisia 14.3 19.4 0 5.1 5,000
Zimbabwe 30.6 35.7 0 5.1 15,000
CCC GSM-102 25.1 25.7 0 0.6 $5,446,6
total 15
Algeria -0.6 0 0 0.6 542,254
Angola 72.2 73.5 0 1.3 4,312
Ecuador -0.6 0 0 0.6 1,176
Egypt -1.3 0 0 1.3 21,462
El Salvador -0.6 0 0 0.6 686
U.S.S.R. 48.6 49.2 0 0.6 1,831,94
0
Ghana 56.0 56.6 0 0.6 3,528
Grenada 76.3 76.9 0 0.6 196
Guatemala -0.6 0 0 0.6 3,234
Hungary 13.6 14.2 0 0.6 2,548
Indonesia 1.0 1.6 0 0.6 14,600
Kenya 52.0 52.6 0 0.6 9,800
Korea, Republic -0.4 0.2 0 0.6 390,432
of
Mexico -0.6 0 0 0.6 1,282,44
8
Pakistan 12.4 13.0 0 0.6 250,000
Panama 50.1 50.7 0 0.6 490
Romania 47.9 48.5 0 0.6 48,608
Russia 48.6 49.2 0 0.6 643,820
Senegal -1.3 0 0 1.3 15,288
Sri Lanka 27.8 28.4 0 0.6 26,264
Trinidad and 1.4 2.0 0 0.6 45,472
Tobago
Tunisia -0.6 0 0 0.6 25,872
Turkey 0.6 1.2 0 0.6 29,008
Ukraine 67.5 68.1 0 0.6 109,020
Venezuela 14.4 15.0 0 0.6 95,452
Yemen, Republic 21.0 21.6 0 0.6 29,400
of
Zimbabwe 46.0 46.6 0 0.6 19,306
CCC GSM 103 2.5 3.0 0 0 .5 $86,240
total
Algeria -0.5 0 0 0.5 26,754
Jordan -0.5 0 0 0.5 5,978
Mexico -0.5 0 0 0.5 4,214
Morocco -0.5 0 0 0.5 17,738
Panama 47.3 47.8 0 0.5 3,234
Trinidad and 1.4 1.9 0 0.5 1,078
Tobago
Tunisia -0.5 0 0 0.5 22,344
Yemen, Republic 20.0 20.5 0 0.5 4,900
of
DSAA FMS total 9.5 6.0 3.6 0 $345,000
Greece 9.5 6.0 3.6 0 320,000
Turkey 9.5 5.9 3.6 0 25,000
EXIM guarantee 18.7 21.7 0 3.0
$6,595,6
total\b 82
Algeria 35.0 39.6 0 4.6 846,540
Argentina 29.4 34.5 0 5.1 106,077
Australia -0.7 1.2 0 1.9 130,675
Bahamas -2.6 1.0 0 3.6 350,120
Barbados -3.2 0 0 3.2 264
Belize 39.4 42.6 0 3.2 3,635
Brazil -5.1 0 0 5.1 183,054
Cameroon 40.5 48.3 0 7.8 56,705
Chile 0.7 4.3 0 3.5 94,085
China 1.1 3.2 0 2.1 330,395
Colombia 4.6 7.4 0 2.7 234,990
Czechoslovakia 13.2 15.3 0 2.1 162,268
El Salvador 59.4 65.3 0 5.9 20,405
Guatemala 50.7 55.0 0 4.3 21,004
India 28.6 30.6 0 2.0 769,817
Indonesia -2.4 1.5 0 4.0 364
Israel 0.4 3.5 0 3.0 87,170
Jamaica -5.9 0 0 5.9 5,452
Kenya -2.0 0 0 2.0 150
Latin American 24.4 27.1 0 2.7 80,238
Multinational\c
Mexico 9.0 11.5 0 2.4 1,020,21
2
Morocco 23.6 25.6 0 2.0 114,171
Nigeria 33.1 39.1 0 6.0 21,266
Norway -0.6 1.3 0 1.9 42,310
Oman 0.9 2.7 0 1.8 37,245
Pakistan 35.4 37.4 0 2.0 33,979
Panama 35.9 38.0 0 2.1 4,346
Philippines 27.4 32.2 0 4.8 133,851
PEFCO\d 0 0 0 0 57,919
Poland 39.3 41.5 0 2.2 277,041
Russia 62.3 66.4 0 4.0 64,625
Sri Lanka 6.4 11.3 0 4.9 6,724
Thailand -1.0 1.7 0 2.7 41,397
Tunisia 29.2 31.1 0 1.9 52,808
Turkey 0.9 4.2 0 3.2 136,954
Uruguay 17.7 20.4 0 2.6 10,005
Venezuela 24.3 27.1 0 2.8 1,057,42
1
EXIM loans total 14.4 16.7 0.8 3.0 $808,800
Algeria 49.1 49.2 0.3 0.4 59,258
Argentina 35.5 39.9 0.6 5.0 2,009
Belize 40.2 44.8 -1.3 3.3 3,394
Cameroon 58.0 71.3 -5.4 7.9 4,286
China, People's 6.5 6.2 2.2 1.9 72,354
Republic
Czechoslovakia 26.8 25.4 3.5 2.1 24,458
Fiji 53.2 51.1 4.9 2.8 4,346
India 35.4 40.0 -3.6 1.1 14,468
Indonesia -1.0 5.2 -2.7 3.5 145,328
Israel 7.8 5.6 5.6 3.4 58,720
Kenya -4.8 0 -0.9 3.8 7,397
Mexico 11.4 13.1 1.6 3.4 27,832
Nigeria 42.4 47.5 1.2 6.2 109,247
Pakistan 60.3 56.9 3.9 0.5 2,322
Philippines 33.3 41.8 -5.0 3.5 5,179
Poland 59.8 65.5 -0.7 5.0 1,576
Thailand 1.6 3.4 0.9 2.7 178,135
Tunisia 31.0 33.4 1.5 3.9 7,899
Turkey 1.2 4.4 0.6 3.8 11,871
Uruguay 23.1 20.0 3.1 0 465
U.S. banks\e 1 .9 0 2 .3 0 .3 64,038
Venezuela 31.3 31.5 3.3 3.5 4,218
USDA P.L. 480 75.4 28.4 47.0 0 $368,110
total\f
Belarus 92.7 42.9 49.8 0 24,000
Congo 95.3 45.4 49.8 0 5,000
Cote d'Ivoire 90.9 44.1 46.8 0 10,000
Egypt 49.8 0 49.8 0 40,410
El Salvador 86.9 37.0 49.8 0 29,400
Estonia 91.5 41.6 49.8 0 10,000
Guatemala 75.5 42.7 32.8 0 14,900
Guyana 96.9 47.0 49.8 0 7,100
Jordan 86.2 36.3 49.8 0 20,000
Latvia 92.1 42.3 49.8 0 10,000
Lithuania 92.0 42.1 49.8 0 10,000
Moldova 96.4 46.5 49.8 0 10,000
Morocco 55.5 8.7 46.8 0 45,000
Philippines 77.0 27.1 49.8 0 20,000
Romania 84.7 34.8 49.8 0 10,000
Sierra Leone 91.8 41.9 49.8 0 9,400
Sri Lanka 49.8 0 49.8 0 13,000
Suriname 90.6 55.5 35.1 0 14,900
Tajikistan 95.7 45.9 49.8 0 10,000
Tunisia 29.8 0.1 29.7 0 15,000
Zimbabwe 79.0 32.2 46.8 0 40,000
All programs 22.1 22.6 1.4 1.9 $13,733,
447
------------------------------------------------------------
Note: The country risk cost rates differ for a given country
principally because our estimate of that country's average propensity
to repay (APP) generally varies by program. In our analysis, we
treated each program independently.
\a AID, Agency for International Development.
\b EXIM, Export-Import Bank.
\c Loans and/or guarantees made to more than one Latin American
Country.
\d Private Export Funding Corporation.
\e U.S. bank loans for which Export-Import Bank provides funds or
guarantees.
\f USDA, U.S. Department of Agriculture.
Table II.4 presents estimates of country risk cost rates for each of
the seven international loan or guarantee programs in fiscal year
1992 by (1) us, in which we made the low-cost assumption that all
repayments on old loans (or guarantees) up to the size of the new
loans (or guarantees) were due to the new loans (or guarantees); (2)
a sensitivity test that assumes new loans or guarantees have no
effect on old loan repayments, which we call the "high-cost
scenario," and (3) the executive branch methodology. We also present
the authorized funds lent or guaranteed by each program. Overall, we
estimate a country risk cost rate for these programs, of 22.6
percent, whereas the executive branch's estimate was 8.9 percent.
Our individual program country risk rate estimates exceed those of
the executive branch except for GSM 103.
Table II.4
Estimates of Program Country Risk Cost
Rates by us, a Sensitivity Test, and the
Executive Branch
(Rates in percent/Dollars in thousands)
High-
cost Executive Credit
GAO scenario branch authorized
---------------- -------- -------- ---------- ----------
AID housing 18.7 38.7 14.1 $83,000
guarantee
CCC GSM 102 25.7 45.5 8.0 5,446,615
CCC GSM 103 3.0 39.9 10.9 86,240
DSAA FMS 6.0 6.4 5.7 345,000
EXIM guarantee 21.7 27.1 8.4 6,595,682
EXIM loan 16.7 22.2 11.5 808,800
USDA 28.4 39.4 25.9 368,110
P.L. 480
All programs 22.6 34.1 8.9 $13,733,44
7
------------------------------------------------------------
--------------------
\1 Public Law 480 loans are a principal form of U.S. government
development assistance to the least developed countries. Because
they have low interest rates, lengthy grace periods, and long
maturities, they are heavily subsidized by the U.S. government.
\2 This program provides guarantees on bank loans with maturities up
to 3 years that finance U.S. agricultural exports.
OUR COUNTRY RISK COST ESTIMATION
METHOD
========================================================= Appendix III
For the past several years, we have used an empirical method to
estimate country risk costs of international lending activities to
developing countries in which the U.S. government engages or
guarantees, or over which it has supervisory or regulatory
authority.\1 During this time, our method has evolved and become more
sophisticated, although the fundamentals have remained unchanged.
The method continues to (1) depend heavily on the price of privately
owned, variable interest rate, dollar-denominated, sovereign lesser
developed country debt that is traded on the secondary market; (2)
use the close statistical relationship between professionally
recognized measures of creditworthiness and the price of this debt to
calculate its risk-based value and cost; and (3) determine if there
are any likely biases in the estimate.
We believe our method is preferable to others we are aware of because
it is empirically based and not subject to any institutional bias to
underestimate risk-based cost. Unless the secondary market is very
risk averse, we also believe that our estimates of the U.S.
government's country risk costs for fiscal year 1992 authorized
international loans and loan guarantees are conservative, given the
time they were formed. Nonetheless, three key issues need to be
discussed to properly qualify our country risk cost estimates.
The first was whether our sample size of debt traded on the emerging
market was large enough to capture the market's real behavior. Our
sample of trades on 38 debt instruments owed by 21 different
countries contained information on dollar-denominated, variable
interest rate, sovereign debt owed the private sector from this
market during the 2-week period beginning the last week of May 1992.
The sample empirically fit many theoretical relationships well.
The second issue was whether the price relationship we estimated for
these traded instruments could be extended to proxy the prices of
similar debt instruments held by the private sector that were not
traded. We believe that these proxies may have a systematic bias to
overestimate the price of these nontraded instruments because they
will, on average, be less liquid. It may be assumed that a lower
price would generally have been paid by investors for these less
liquid instruments. This assumption tends to cause our estimates of
the U.S. government's country risk costs to be conservative.
The third issue is whether sovereign debt owed the private
sector--from which our country risk cost estimates were derived--is
more or less likely to be repaid in the long run than sovereign debt
owed the U.S. government. This issue addresses the long-run
probability of repayment, not short-term considerations that may be
present when a creditor lends more new funds than are currently due
and, as a result, is temporarily being repaid on old loans when
others are not. In this case, this creditor's new loans are simply
rolling over old loan repayments and increasing its exposure and
long-term costs. The private sector may be more likely than the U.S.
government to receive fuller payment from a sovereign developing
country debtor in the long run. The reason for this situation is
that the private sector is fairly exclusively motivated to maximize
its profitability, whereas the U.S. government has a multiplicity of
other goals, including enhancing foreign policy objectives, promoting
U.S. defense goals, and helping domestic constituent interests.
In support of this view, a recent paper presents econometric
evidence, other empirical evidence and theoretical arguments that
suggest that the order of payment preference is first, the
International Monetary Fund; second, private creditors; and last,
advanced country governments such as the United States.\2 We believe
that this distinction also tends to cause our estimates of the U.S.
government's country risk costs to be conservative.
Even though it is theoretically possible that concern about trade,
defense, or foreign policy might motivate sovereign debtors to
preferentially repay the U.S. government, in the long run we know of
no evidence that this scenario is occurring, although in the future
this question may be better tested econometrically and answered. In
any event, since our's and the executive branch's methods both mark
to market U.S. government- owned debt based on privately owned debt
traded in financial markets, this consideration tends to bias both
cost estimates in the same direction.
--------------------
\1 A listing of related GAO products appears at the end of this
report.
\2 Bulow, Jeremy, Kenneth Ragoff, and Afonso S. Bevilaqua.
"Official Creditor Seniority and Burden Sharing in the Former Soviet
Bloc." Brookings Papers on Economic Activity, no. 1 (1992), pp.
195-234.
COUNTRY RISK RATINGS
------------------------------------------------------- Appendix III:1
To obtain continuous country risk ratings, we statistically combined
ratings that appeared in September 1992 in two leading magazines for
financial institutions, Euromoney and Institutional Investor.\3 We
first calculated each country's z score from these two ratings and
then averaged the two z scores.\4 When only one z score for a country
was available, it became the average. To obtain our continuous
ratings on a scale from 0 to 100, similar to the risk ratings on
which our rating is based, we treated each country's average z score
as if it were generated from a z probability distribution. We then
set that country's rating to be 100 times the cumulative distribution
(probability) of its average z score.
--------------------
\3 September 1992 Euromoney credit ratings covered 169 countries and
were a compilation of nine subcategories, including a survey of
political risk analysts; debt indicators; access to various markets,
including the Euro-bond market; and credit ratings when performed by
Moody's and Standard and Poor's, two of the world's leading credit
rating agencies that rate only those developing countries that are
among the most creditworthy. The September 1992 Institutional
Investor ratings we used evaluated 126 countries based on anonymous
ratings from 75 to 100 of the world's largest banks' country risk
departments. Bank ratings of the bank's home countries were not
used; ratings were combined using a weighting scheme, which
Institutional Investor claimed gave greater weight to those banks
with more sophisticated country risk departments.
\4 A z score is a random variable that has been transformed by
finding the difference between it and its estimated mean and dividing
this difference by its estimated standard deviation. This
transformed random variable has a mean of zero and standard deviation
equal to one.
ESTIMATING COST
----------------------------------------------------- Appendix III:1.1
The process for measuring country risk costs for purposes of meeting
the requirements of credit reform legislation is rather lengthy.
Some of the length arises because country risk values of debt are
observed from financial markets, but country risk cost of a new loan
needs to be calculated on a cost to the U.S. government basis that
is relative to Treasury rates. First, we obtained the risk premiums
relative to the riskless Aaa/AAA rate for less risky debt, Baa/BBB or
higher, directly from financial markets in which similar privately
owned bonds are traded and transformed them into country risk values
of debt.\5
Next, we obtained the country risk value of risky U.S.
government-owned sovereign debt from the secondary market in which
similar privately owned debt is traded using econometric analysis.
For both risky and nonrisky debt, we also estimated the value of debt
that is a complete rollover; that is, it causes an equal amount of
old loan repayment. We then estimated the extent that each new loan
causes old debt repayments, the average propensity to repay old
loans, and the remaining part increases U.S. government exposure.
To obtain the country risk value of new loans, we averaged these
country risk values of sovereign debt; one for complete rollovers,
the other for complete increased exposure using APP and l-APP as
weights for each loan. We then transformed these country risk new
loan values into country risk new loan interest rate premiums above
the Aaa/AAA rate using present-value analysis. These country risk
premiums and the Treasury rate were then used to calculate the
subsidy and component costs of a loan on a cost to the U.S.
government basis.
--------------------
\5 For the most creditworthy debt, we used bonds because we had no
loan information.
COUNTRY RISK VALUE
----------------------------------------------------- Appendix III:1.2
We estimated country risk value of privately owned, sovereign debt
based on the financial markets in which such debt is traded. For
debt issued by countries that we rated about the same as Moody's and
Standard and Poor's in their categories of Baa/BBB to Aaa/AAA,
respectively, we based our estimates on historical interest rate
differentials between bonds with Moody's and Standard and Poor's
ratings and the risk-free Aaa/AAA rate. For more risky debt, we
estimated country risk values based primarily on our analysis of the
secondary market in which privately owned developing country debt is
traded internationally.\6 Of the countries the U.S. government
authorized loans or guarantees to in fiscal year 1992, 83 percent
were below Baa/BBB.
--------------------
\6 Trading volume on the secondary market has grown very quickly. It
was approximately $2 billion (face value) in 1985, $5 billion in
1986, $70 billion in 1990, $240 billion in 1992, and $400 billion in
1993.
THE SECONDARY MARKET
----------------------------------------------------- Appendix III:1.3
A large portion of the developing country soverign debt traded on the
secondary market is variable interest rate debt; thus, the price of
this debt does not change because of general interest rate movements
on world financial markets. Prices of this debt are discounted from
face value in the secondary market to reflect investors' assessments
of the large country risk associated with this developing country
debt or indicate that other factors may be present that impair the
value of the debt.
If investors were risk neutral, and no market forces--other than the
evaluation of risk by investors--were present, then the price of this
variable interest rate debt would be an unbiased measure of its
country risk value. Under these circumstances, for example, if debt
owed by a country had a price of 40 percent of face value, then the
market expects that this debt would, on average, pay back only about
40 percent of its face value. Thus, about 60 percent of face value
is the market's expected cost of holding the loan to the institution
that had issued the loan.
If no other market forces were present, prices of variable interest
rate debt would respond only to changes in investors' perceptions of
country risk. For example, if investors believed this developing
country debt had less country risk than its price indicated,
investors would have an incentive to buy this debt and, as a group,
would cause the price of this debt to increase. Similarly, if
investors believed this developing country debt had more country risk
than its price indicated, then investors would have an incentive to
sell this debt and, as a group, would cause the price of this debt to
be lower.
Our analysis of the secondary market for the last week in May and the
first week in June 1992 indicates that market prices of variable
interest rate developing country debt were based almost exclusively
on investors' perceptions of the country risk in this debt.
According to the market specialists we spoke with, this situation
occurred because other market forces that were especially pronounced
7 and 8 years ago appeared to be minimal during this 2-week period.\7
--------------------
\7 See International Banking: Supervision of Overseas Lending Is
Inadequate (GAO/NSIAD-88-87, May 5, 1988). In that report, we
developed estimates of appropriate bank reserves based on what is now
called the secondary market. We also discussed our earlier analysis
in this area, which was summarized in our April 2, 1987, testimony,
when we had found that, overall, market forces other than investors'
risk evaluation had caused developing country market prices to be
much too high and market-based reserves (accounting's measure of
cost) much too low for investors' perceptions of market risk. These
market forces included (1) the "contamination effect," which caused
large U.S. banks to restrict their supply of discounted foreign debt
for sale on the secondary market for fear that if they sold any part
of a developing country's debt, their auditors would require that
they mark to market all of that developing country's remaining debt;
(2) debt-equity swap programs that allowed a developing country to
purchase (through an intermediary) its debt on the secondary market
and thereby increase demand for its debt for other than risk
evaluation reasons; and (3) least important and with opposite
directional effect, "dumping," or selling primarily motivated by
reasons other than risk evaluation.
ESTIMATING THE DEBT'S
COUNTRY RISK VALUE
----------------------------------------------------- Appendix III:1.4
To estimate the country risk value of privately owned, less developed
country debt, we regressed secondary market prices for variable
interest rate, dollar-denominated, privately owned sovereign loans
and bonds on our country risk rating of the debtor developing country
and on various characteristics of the debt instrument.\8 We obtained
the following results:
(See figure in printed edition.)
PRICE equals the average price for the developing country debt
instrument during the 2-week period beginning the last week of May
1992, net of the then-market price of any collateral or guarantees
for that debt instrument; GAO equals our continuous country risk
rating of the debtor country; MAT equals the years remaining until
maturity of the debt instrument; ARR equals the months in arrears (if
any) of the debt instrument; D is a dummy variable, equal to 0 for
loans and 1 for bonds, used to capture differences between loans and
bonds; and the "t" statistics are in parentheses.\9
The data sample was quite diverse in terms of price, credit rating,
maturity, and months of arrears of the debt instruments, and it was
relatively equally split between the number of loans and bonds. It
consisted of all dollar-denominated, variable interest rate sovereign
debt that traded at a discount on the emerging market for which we
were able to collect meaningful data, plus one loan and one bond for
each of the five least creditworthy countries whose variable interest
rate, dollar-denominated debt traders told us would trade at par
(100). In all, the sample was composed of 21 different countries
with 20 loans and 18 bonds and was heavily weighted to one country,
Venezuela, which had issued eight bonds with variable interest rates.
In this regression data, the mean, standard deviation, and range for
price were 68.1, 29.3, and 8.2 to 100, respectively; for our country
risk ratings they were 55.4, 19.9, and 19.1 to 68; for months in
arrears they were 8.3, 19.3, and 0 to 84; and for maturity (years)
they were 11.5, 5.7, and 1.2 to 27.8.
Overall, the regression results were good. The explanatory power of
the regression was high, and independent variables had the
directional effect expected from theory and were statistically
significant at the 99-percent confidence level.\10 In addition, as
expected from theory, the regression indicates that bond prices are
greater than loan prices, as long as the debtor country has a credit
rating sufficiently high to issue bonds.\11 The regression also
indicates that for these more creditworthy developing countries, the
price of bonds declines slightly as maturity dates are lengthened.
The loan portion of this regression is used throughout this report
because (1) in fiscal year 1992 the U.S. government lent or
guaranteed funds internationally only in the form of loans and (2)
relevant lender characteristics of the U.S. government appear to be
closer to private loan holders than private bond holders.\12 Figure
III.l depicts the estimated relationship between our credit rating
and loan prices when the months in arrears of a loan are assumed to
be zero. The large range of loan prices and large range of our
associated credit ratings is visually apparent, as is the positive
effect (high positive correlation) of the credit rating on prices.\13
Figure III.1: Loan Prices and
Our Credit Rating
(See figure in printed
edition.)
We used the results of the loan portion of this regression to
estimate the value of privately owned, variable interest rate loans
issued by 124 countries with credit ratings lower than the Baa/BBB
range. We then normalized these loan values to obtain values of
privately owned loans relative to the risk-free Aaa/AAA rate.\14 We
then used these normalized values of privately owned loans as our
measures of the country risk value of loan debt to the U.S.
government relative to the Aaa/AAA risk level.
--------------------
\8 There is a period of a few weeks each year in which the
Institutional Investor country ratings we used are theoretically most
compatible with the emerging market due to the way these external
ratings are created. In 1992 this was the 2-week period beginning
the last week of May, and we therefore obtained necessary secondary
market data from this time. All data we needed was available to us
by October 1992 and, without modifying our method we would not have
been able to update our estimates with market expectations until
October 1993, 1 year later.
\9 The "t" statistic is the estimated coefficient divided by its
estimated standard deviation and is used to test whether an estimated
coefficient is statistically different from zero. We dropped the
coefficient for loan maturity because when we included it in an
earlier regression, it was small (-0.27 cents per year) and was not
statistically significant (t = -0.7).
\10 Similar results were obtained when we dropped the 10 instruments
that trade at par. For the remaining 28 observations, the regression
results were the following:
_
PRICE = +1.00 + 1.20*GAO -0.26*ARR+79.5*D -1.02*D*GAO -1.38*D*MAT,
R\2=0.925, n=28
(0.2) (9.6) (-3.1) (5.5)
(-3.8) (-5.4)
\11 Developing country debt traders gave us three reasons why
investors believe bonds are more likely to be paid than loans and
thus command a higher price. First, bonds are often owned by
individuals, who are more likely to sue and less likely to agree to
debt relief than banks. Second, in the 1980s bonds had often been a
small part of a country's external debt. Given the difficulty of
getting debt relief from individual bond owners, developing countries
had a greater incentive to stay current on their bond payments.
Third, bonds have been fully serviced by Latin American developing
countries during the most recent developing country debt crisis, even
when these same developing countries have been delinquent in repaying
bank loans.
\12 We specified the regression in equation III.l so that it could be
separated: bonds had no effect on the estimated regression
coefficients of the implied loan price equation, and loans had no
effect on the estimated regression coefficients of the implied bond
price equation. In short, we would have obtained the same estimates
for loans if we had dropped all bonds from our sample.
\13 The actual values of price are closer to the regression plane
than the regression line depicted in figure III.1 because the
regression plane includes the negative effect of months interest
arrears on loan prices, but the regression line depicted in figure
III.l does not.
\14 Because we relied on information from two markets, we needed to
merge these two different sets of information concerning
creditworthiness and the price at which these instruments traded. In
the first market, we had information on the spread above the Aaa/AAA
rate of fixed-rate debt and the effect of a country's
creditworthiness for the more creditworthy countries (rated Aaa/AAA
through Baa/BBB). In the secondary market, we used information on
how creditworthiness and other characteristics affected the price of
variable interest rate debt for countries with credit ratings
equivalent to the lower end of Baa/BBB or lower. We normalized our
risk-based values from the secondary market by setting them equal to
the price implied from the regression in equation III.l times 99.2
percent because variable rate instruments began to be priced at par
(100), when issued by the least creditworthy Baa/BBB debtors, and
1-year fixed-rate instruments by these borrowers had a country risk
value rate derived from the first market of 99.2 percent.
VALUE OF NEW LOANS
----------------------------------------------------- Appendix III:1.5
The Credit Reform Act attempts to ensure better estimates of the U.S.
government's cost of new loans and new loan guarantees. As a result,
the effect of new loans on repayment of past loans should be
important in obtaining these estimates of value and cost. Since the
U.S. government is a large, non-anonymous lender that repeatedly
makes loans or guarantees to recipient countries, in many instances
the granting of new funds by the U.S. government causes repayments
to be made on old loans owed to or guaranteed by the U.S.
government. This raises the value and lowers the cost to the U.S.
government of the new loan. However, the executive branch views the
act as precluding it from accounting for these considerations in its
cost estimates, and it thereby tends to overestimate the cost of new
loans and loan guarantees. Our analysis and estimates were not bound
by this restriction. Our analysis first considered the case in which
the granting of new loans has no effect on old loan repayment--an
average propensity to repay old loans of zero. This might occur if
(1) the debtor's creditworthiness was high enough so that all
currently scheduled repayments would have been made without the
incentive of new loans or (2) there were no currently scheduled
repayments from the country. The government's country risk value of
the new loan would then be identical to the government's value of
this outstanding debt.
We then considered the second case in which the new U.S. loans or
guarantees simply roll over existing debt on a dollar-for-dollar
basis--one dollar of new lending causes one dollar of repayment on
old debt, or an average propensity to repay of one. This might occur
for a debtor country with low creditworthiness that would not have
made any scheduled repayments on old U.S. government loans this year
if it did not receive new U.S. loans or loan guarantees. In this
case, the U.S. government's cost of this new loan is simply the time
cost of extending loan repayments due this year into future years as
prescribed by the new loan, which we calculated to be quite
inexpensive.
We would expect to find that a new loan often induces some currently
scheduled repayments but less than the amount of the new loan; that
is, the new loan has an average propensity to cause old loan
repayments that lie between zero and one. If we knew the value of
this average propensity to repay, calculating the country risk value
of the new loans would be simple--(l-APP) times the country risk
value if there were no rollover, as in the first case, plus APP times
the country risk value if there were 100-percent rollover, as in the
second case. To estimate the cost of new loans or guarantees
authorized in 1992, we could then transform these country risk
values, which are relative to Aaa/AAA rates, into estimates of the
cost of country risk on a cost to the U.S. government basis.
Because we know of no other empirical measures of the average
propensity to pay, we assumed the following simple behavior:
The more creditworthy the country, or the larger its scheduled
repayments, the more repayments it will make on old loans without any
new loan incentive. Greater creditworthiness has no additional
effect once all scheduled repayments have been made.
If new loans are not too large, more new loans cause more old loan
repayment, although any induced repayment will be less than or equal
to the size of the new loans. Too large is defined as that point at
which actual repayments equal scheduled repayments.
If new loans are too large, larger new loans will have no additional
effect on old loan repayment.
If new loans are not too large, the more creditworthy the debtor, the
more a new loan will induce repayment.
These considerations are captured in the following equations:
(See figure in printed edition.)
APP is the average propensity to repay old loans or guarantees as a
percent of new loans or guarantees, ROLL is the dollars paid on old
loans induced by the new loan, X is the amount of repayment on old
loans that would have occurred if no new loans or guarantees were
granted, Z is the dollars of old loan repayments made, S is the
dollars of old loan repayment scheduled, L is the dollars of new U.S.
government loans and guarantees granted, and GAO is our country risk
creditworthiness measure.
Equations III.2 to III.4 yield the following:
(See figure in printed edition.)
Although more precise estimates of ROLL than we present may be
obtained by estimating equation III.6 using cross-section time-series
analysis, we took a different approach that lessened the time we
needed to perform the analysis. This approach did not affect our
major conclusion. As of October 1992, it is likely that the
executive branch's method greatly underestimates budget and country
risk costs for fiscal year 1992 authorized international loans and
guarantees.
For our estimate, we purposely made an assumption that caused our
cost estimates to be conservative. We overestimated the rollover by
setting ROLL equal to the minimum of the loan size, L, or payments
made, Z. For a sensitivity test we called the high-cost scenario, we
assumed the rollover equaled zero. This would have been our best
estimate if we had followed the executive branch's interpretation of
the act's restriction on accounting for the cost effect of rollovers.
Our estimate and the high-cost scenario, which were primarily based
on emerging market expectations during the 2-week period beginning
the last week of May 1992, yielded country risk cost estimates for
fiscal year 1992 authorized international loans and guarantees that
were generally much greater than those of the executive branch. If
more recent data had been obtained from financial markets, it is
likely that our estimate and sensitivity test would each have
resulted in lower cost estimates because prices on the emerging
market were generally higher.
COUNTRY RISK COST OF THE
LOAN TO THE U.S. GOVERNMENT
----------------------------------------------------- Appendix III:1.6
We then transformed this country risk value of the new loan relative
to the Aaa/AAA rate into a country risk interest premium. We did
this by setting this country risk value equal to premium payments on
a standardized loan with interest payments at the Aaa/AAA rate of the
same maturity. We then solved this equation for the appropriate
internal rate of return. The country risk interest premium was the
difference between the internal rate of return and the Aaa/AAA rate.
We then used this country risk premium to calculate the country risk
cost of the loan on a cost to government basis. This cost is the
difference between the present values of the scheduled loan
repayments, one discounted with the Treasury rate for the same
maturity, the other discounted with the loan's country risk interest
premium plus this same Treasury rate.
EVALUATION OF THE EXECUTIVE
BRANCH'S METHODS AND ESTIMATES
========================================================== Appendix IV
Although the executive branch had only a short amount of time after
enactment of the Credit Reform Act to develop appropriate methods to
estimate country risk and its cost, we found that the method it
developed and employed had several weaknesses. The principal
weakness was that it was not based on rigorous econometric tests and
measurements. Most weaknesses originated in the method employed by
the risk premium committee, especially for countries with ratings
below Baa/BBB, where most of the country risk cost occurs. These
weaknesses could lessen the U.S. government's ability to make sound
lending decisions and the Congress' ability to make sound funding
decisions.
COUNTRY RISK COST METHOD
-------------------------------------------------------- Appendix IV:1
Members of the risk premium committee told us that they felt very
pressed for time when they met during October 1991 because premiums
were needed immediately for calculating the fiscal year 1993 budget.
Nineteen observations of bonds formed the basis for obtaining the
risk premiums for ICRAS' top 3 categories, A, B, and C. These ICRAS
categories correspond to Moody's and Standard and Poor's ratings
Baa/BBB or higher. For the next four ICRAS categories, C-, D, D-,
and E, there were only five observations of bonds--three Ba/BB-rated
bonds and two B/B rated bonds. Historical average 3-year risk
premiums for bonds with the same ratings as the bond observations
became the risk premiums for the top seven ICRAS categories. Without
using econometric methods, the risk premium committee derived 16 risk
premiums for ICRAS categories C-, D, D-, and E from the risk premiums
for these 5 observations.\1 To help fill in the gaps, two ICRAS
categories' premiums were obtained by averaging those from adjacent
ICRAS categories. The lowest four ICRAS ratings were obtained by
extrapolating the E categories' risk ratings. In these categories
the risk premium committee's method assumed that the level of country
risk did not affect the relative size of one maturity's risk premiums
compared to another. At least one member of the committee checked to
ensure that the risk premiums for long-term debts were loosely
consistent with prices of risky debt from the secondary market. The
quality of the estimates that resulted was poor because the risk
premium committee's method
employed too little data below ICRAS' C rated countries for the large
amount of information needed;
frequently assumed that the level of country risk had no effect on
the relative size of the risk premiums for different maturities;
obtained the same risk premiums for loans and bonds, resulting in
executive branch country risk cost estimates for loans that are less
than for bonds, which is the opposite of what we found on the
secondary market;
had not been updated, even though (1) new information from financial
markets had been available and (2) Office of Management and Budget
(OMB) officials told us throughout our review that the risk premiums
for short-term loans to the riskiest countries were too low;
did not disclose the sources of bias, except for these short-term
loans to the riskiest countries; and
made no distinction between new loans to risky countries that have a
relatively large effect on old loan repayment, cause a relatively
small increase in U.S. government exposure, and thereby have
relatively small country risk cost from those new loans that do not.
The executive branch views the Credit Reform Act as precluding a
distinction for cost purposes between new loans that roll over old
loans and those that do not. Our analysis indicates that
overestimates of cost tend to occur when such distinctions are not
made. For example, when we purposely ignored this distinction, our
subsidy cost estimate for all 1992 authorized loans and loan
guarantees increased 51.6 percent to $4.6 billion, and our cost
estimates for loans and loan guarantees to many very risky countries
increased by a much larger percentage.
--------------------
\1 These are four ICRAS rating categories times four maturity
categories--1, 5, 10, and 30 years.
COUNTRY RISK COST ESTIMATES
-------------------------------------------------------- Appendix IV:2
The executive branch's estimate of the cost of country risk on a cost
to the U.S. government basis for a loan of a given maturity is the
difference between the present value of loan payments when (1) the
discount rate is the U.S. treasury rate of the appropriate maturity
and (2) the discount rate is the sum of the U.S. treasury rate and
the applicable executive branch risk premium, both for the
appropriate maturity.\2 The applicable executive branch premiums are
presented in table IV.1 for each executive branch rating category and
loans of representative maturity as well as our average rating for
countries that received that executive branch rating during the last
half of fiscal year 1992.
Table IV.1
Executive Branch Ratings and Risk
Premiums
(Rates in percents)
Executive Our
branch average
rating rating 1 year 5 years 10 years 30 years
---------- -------- -------- -------- -------- --------
A 96.0 0.25 0.30 0.30 0.40
B 82.2 0.40 0.45 0.50 0.75
C 60.8 0.80 0.90 1.00 1.35
C- 45.2 1.87 1.81 1.64 1.96
D 41.2 4.01 3.62 2.92 3.17
D- 30.4 5.71 4.84 4.39 4.64
E 30.3 9.11 7.29 7.34 7.59
E- 27.8 13.66 10.94 11.00 11.38
F 24.6 22.76 18.23 18.34 18.96
F- 21.0 31.87 25.52 25.67 26.55
F-- 15.8 50.08 40.10 40.34 41.72
------------------------------------------------------------
Note: The executive branch recognizes that short-term loans issued
before the debtor's Paris Club contract cutoff date are more likely
to be rescheduled than other loans and treats them as if they had
longer maturities--10-year loans for those they rate "C" or better
and 30-year loans for those they rate lower.
Table IV.2 presents the corresponding executive branch estimates of
country risk cost rates (cost divided by loan size) for loans on a
cost to U.S. government basis when the average fiscal year 1992 U.S.
treasury rate for a maturity is used.
Table IV.2
Executive Branch Country Risk Cost Rates
(Rates in percents)
Executive Our
branch average
rating rating 1 year 5 years 10 years 30 years
---------- -------- -------- -------- -------- --------
A 96.0 0.2 0.8 1.3 3.1
B 82.2 0.4 1.2 2.1 5.6
C 60.8 0.8 2.3 4.1 9.7
C- 45.2 1.8 4.5 6.5 13.6
D 41.2 3.7 8.7 11.2 20.4
D- 30.4 5.2 11.4 16.1 27.5
E 30.3 8.0 16.4 24.6 38.6
E- 27.8 11.6 23.0 33.4 48.9
F 24.6 17.9 33.9 46.5 61.8
F- 21.0 23.4 42.4 55.6 69.6
F-- 15.8 32.4 54.6 67.3 78.4
------------------------------------------------------------
Tables IV.3 and IV.4 present our country risk cost rate estimates,
which were largely based on the secondary market during the 2-week
period beginning the last week of May 1992, in a manner similar to
the executive branch estimates in table IV.2. Table IV.3 presents
our estimates of the country risk cost rate of the new loan if the
new loan has no effect on repayments on old loans and the loan is, on
the average, as creditworthy as the corresponding executive branch
category. Similarly, table IV.4 presents our estimates of the
country risk cost rate of the new loan if the new loan causes an
equal amount of principal repayments on old loans owed the U.S.
government.
Table IV.3
Our Country Risk Cost Rate Estimates for
Loans that Do Not Affect Old Loan
Repayment
(Rates in percents)
Executive Our
branch average
rating rating 1 year 5 years 10 years 30 years
---------- -------- -------- -------- -------- --------
A 96.0 0.2 0.8 1.3 3.4
B 82.2 0.4 1.2 2.2 6.2
C 60.8 24.6 24.8 25.0 25.7
C- 45.2 44.4 44.8 45.0 45.9
D 41.2 49.4 49.9 50.1 51.0
D- 30.4 63.2 63.6 63.9 64.7
E 30.3 63.2 63.7 63.9 64.7
E- 27.8 66.4 66.9 67.1 67.9
F 24.6 70.4 70.8 71.1 71.8
F- 21.0 75.0 75.4 75.6 76.2
F-- 15.8 81.5 81.9 82.0 82.5
------------------------------------------------------------
Table IV.4
Our Country Risk Cost Rate Estimates for
Loans that Cause an Equal Amount of Old
Loan Repayment
(Rates in percents)
Executive Our
branch average
rating rating 1 year 5 years 10 years 30 years
---------- -------- -------- -------- -------- --------
A 96.0 0.1 0.7 1.2 3.2
B 82.2 0.2 1.0 2.0 5.9
C 60.8 0 0 0 0
C- 45.2 0 0 0 0
D 41.2 0 0 0 0
D- 30.4 0 0 0 0
E 30.3 0 0 0 0
E- 27.8 0 0 0 0
F 24.6 0 0 0 0
F- 21.0 0 0 0 0
F-- 15.8 0 0 0 0
------------------------------------------------------------
The executive branch did not consider the effect that a new U.S.
government loan to a country can have on that country's repayments on
old U.S. government loans. Although the executive branch believes
that the Credit Reform Act precludes them from accounting for this
effect in their cost estimates, this effect is critical for obtaining
more accurate estimates of the U.S. government's costs, a basic goal
of the act. Our analysis showed that when the U.S. government lends
or guarantees loans to a developing country so that new funding does
not cause more repayments on old loans owed the U.S. government, the
additional costs of the new loan are quite high. In contrast, when
these new loans or loan guarantees induce relatively large principal
repayments on past loans or loan guarantees, the additional costs are
quite low.
For example, for a 1-year, $100 million loan or loan guarantee given
to an average E-rated country, which has no effect on contemporaneous
principal repayment on old U.S. government loans, we estimate its
country risk cost on a cost to the U.S. government basis to be $63.7
million. This estimate is considerably higher than the executive
branch's estimate of $8 million. If the $100 million loan or loan
guarantee causes $100 million of contemporaneous principal repayments
on old loans owed the U.S. government to be made (APP = 1), we
estimate its country risk cost to be zero. If the $100 million loan
or loan guarantee causes $50 million of contemporaneous principal
repayments to be made (APP = 0.5), we estimate its country risk cost
to be about $31.9 million; if the loan causes $25 million of old loan
repayments (APP = 0.25), we estimate its country risk cost to be
about $47.8 million; and so on.
A very close approximation for the country risk cost rate of this
loan on a cost to the government basis is APP times the appropriate
entry in table IV.4 plus (1-APP) times the corresponding entry in
table IV.3.\3
In theory, our estimates of country risk cost based on markets during
this 2-week period beginning the last week of May 1992 may be greater
or less than the executive branch's, depending upon the size of the
APP, but our estimates of country risk and subsidy costs for fiscal
year 1992 authorized foreign loans and guarantees were almost always
much higher than those using the executive branch's method, although
both estimates of fee income and interest cost were quite close. Our
estimate of country risk costs for all $13.7 billion of authorized
fiscal year 1992 foreign loans and guarantees was $3.1 billion, about
2-1/2 times the $1.2 billion estimate using the executive branch's
method.\4 In six of the seven programs for which we estimated program
costs, our estimates of program country risk were greater than those
of the executive branch.
Although our country risk estimates are generally much greater than
those of the executive branch, we believe our estimates of country
risk cost are conservative for the time they were made for the
reasons discussed in appendix III. Even if the emerging market was
very risk averse, and we revised our estimates of country risk cost
downward to compensate, our revised estimates at this time for all
fiscal year 1992 loans and guarantees would probably still be much
greater than those of the executive branch.
For the first 50 countries on our rating scale out of about 170 that
we estimated (roughly comparable to those countries the executive
branch rated A or B), the executive branch's country risk cost
estimates were most similar to ours because (1) we both based our
subsidy rates on the same yield differentials and (2) the cost
differences between new loans that do not affect repayment and those
that cause maximum repayment are least for these more creditworthy
countries. This first reason is demonstrated by the similarity in
cost rates for executive branch A and B rated countries in tables
IV.2 and IV.3. The second reason is illustrated by comparing our
cost rate estimates for loans having no effect on repayment in table
IV.3 with those for loans that have maximum effect on old loan
repayment in table IV.4.
However, from about the 50th most creditworthy country on our rating
scale to the least creditworthy country, which have the highest
country risk and country risk cost (roughly C or lower on the
executive branch rating), our country risk cost estimates for loans
differed greatly from the executive branch's primarily for three
reasons. The most important reason was that we systematically
estimated how prices on the secondary market of variable interest
rate, dollar-denominated developing country sovereign loans were
affected by country risk and other characteristics using statistical
analysis. In this range the risk premium committee based 32 risk
premiums on only 5 bonds and used secondary market loan prices only
to see if implied costs of the riskiest long-term loans were in the
range of those implied by the secondary market. Instead of using
econometric analysis to fill in the gaps, the executive branch
assumed in the lowest five ICRAS categories that the relative size of
one maturity's risk premium to another is unaffected by the level of
country risk, a relationship we found to be untrue for privately
owned, variable interest rate sovereign loans on the secondary
market.
Also, from our analysis of the secondary market, we found a large
difference in country risk cost between new loans to developing
countries that do not affect repayment on old loans and those that
do. This distinction tended to lower our estimates from what they
would have been had we followed the executive branch and not made
this distinction. The least important reason was due to the
differences between our and the executive branch's country risk
ratings of developing countries. Additional small differences in
estimates occurred because the executive branch (1) used the same
country risk interest differential for loans with a rather wide range
of maturities, whereas we customized the interest rate risk
differential to each loan's maturity and (2) transformed annual
interest rates into a semiannual or quarterly basis using an
approximation that greatly affects cost estimates when maturity is
short and interest rates are large, whereas we did not employ this
approximation.
--------------------
\2 To be compatible with cost calculations for domestic programs
under credit reform, in practice the executive branch uses a slightly
different method to calculate country risk cost. It uses a risk
premium along with the Treasury rate to project a future stream of
expected defaults and then calculates the present value of this
stream by discounting with the Treasury rate.
\3 This relationship would be exact if the country risk cost concepts
were relative to the Aaa/AAA rate. We could also closely estimate
the cost of a new loan to any of 170 countries by multiplying its APP
times the appropriate entry in table II.2 plus (1-APP) times the
corresponding entry in table II.1.
\4 The executive branch appropriately does not employ OMB's interest
rate premiums to estimate the risk-based costs of the Overseas
Private Investment Corporation's program because this program
primarily involves the risk on foreign direct investment.
COUNTRY RISK METHOD AND
ESTIMATES
-------------------------------------------------------- Appendix IV:3
Executive branch ratings are based on 35 subitems, most of which are
5-year expectations of economic indicators. These 35 subitems are
grouped in 5 subrating categories: payments history, macroeconomics,
the debt burden, balance of payments adjustment capacity, and
political and social factors. ICRAS scores the 35 subitems, 5
subrating categories, and the overall summary country rating on the
11-grade scale.
The preferred method for measuring the worth of the executive
branch's ratings would be to measure their accuracy in predicting
future debt-servicing problems over long periods of time and then
compare their accuracy to that of professionally recognized
(external) ratings. However, because executive branch ratings have
only existed a short time, this method of measurement was not
available. Therefore, we compared executive branch ratings to
professionally accepted external standards, measured their internal
consistency, and reviewed the method that generated them.
The executive branch country risk ratings developed during the latter
half of fiscal year 1992 for the fiscal year 1994 budget have both a
fair amount of similarity and differences compared with (1)
contemporaneous Euromoney and Institutional Investor country risk
ratings and our ratings, which statistically combined these two
professionally recognized ratings;\5 and (2) contemporaneous prices
of privately owned, dollar-denominated sovereign less developed
country, variable interest rate debt available on the secondary
market. In addition, the executive branch's rating method was only
loosely based on econometric tests. Although at least one official
had reviewed the professional economic literature for these tests,
many subitems were included that had not passed econometric tests,
and the weights ICRAS used to combine the 35 subitems were not
determined by econometric methods.
To ensure that differences between executive branch ratings and these
external ratings were not due to the fact that executive branch
ratings were grouped ratings and the external ratings were
continuous, we transformed these continuous external ratings into
discrete ratings consisting of 11 groups. For example, we
transformed our continuous ratings as follows: the top four groups
were countries with similar credit ratings to countries that Moody's
and Standard and Poor's rated Aaa/AAA, Aa/AA, A/A, and Baa/BBB,
respectively; the remaining seven groups were formed so that
countries with close continuous ratings under our rating system were
in the same group. We then transformed the continuous Euromoney and
Institutional Investor ratings each into discrete ratings in a
similar manner.
To evaluate ICRAS ratings, we first compared the then-latest
executive branch country risk ratings (used for the fiscal year 1994
budget) to our own continuous contemporaneous ratings. We found that
there was both a fair amount of similarity and difference between the
ICRAS ratings and our own. For each of the executive branch's 11
categories (A to F- -), table IV.5 presents the number of countries;
our highest, average, and lowest country score; and the standard
deviation.
Table IV.5
The Executive Branch's Country Risk
Ratings Measured on Our Creditworthiness
Scale
Executive Standard
branch deviatio
ratings Number Low Average High n
---------- -------- -------- -------- -------- --------
A 21 87 96.0 99 2.8
B 20 53 82.2 94 10.0
C 21 26 60.8 80 14.5
C- 18 28 45.2 62 9.2
D 9 30 41.2 50 7.7
D- 8 18 30.4 41 8.5
E 10 23 30.3 37 4.3
E- 20 19 27.8 45 5.9
F 21 16 24.6 39 6.8
F- 24 7 21.0 40 7.1
F-- 16 11 15.8 26 4.4
------------------------------------------------------------
Note: These numbers have been rounded to disguise individual ICRAS
country ratings that are classified.
In this context, the standard deviation measures how much countries
with the same executive branch credit rating vary in our measure of
country risk. We also compared grouped Euromoney ratings to
continuous Institutional Investor ratings and then grouped
Institutional Investor ratings to continuous Euromoney ratings in a
similar manner as a standard. If the executive branch's ratings
conformed very closely to our ratings, we would expect to find low
standard deviations and progressively lower average values on our
rating scale and would be able to distinguish accurately members of
one category from those two to four categories apart when we made the
pair-wise comparisons. However, except for a progressively lower
average, these tests for closeness were not met. Virtually all
standard deviations were large; the average standard deviation when
ICRAS ratings were compared with our own continuous ratings was 77
percent higher than the average standard deviation when we compared
grouped Euromoney ratings with continuous Institutional Investor
ratings and 95 percent higher than when we compared grouped
Institutional Investor ratings with continuous Euromoney ratings.
We then compared the power to distinguish one category from another
when we made these pair-wise rating comparisons. We considered one
category distinguished from another if two standard deviations below
the higher category's average exceeded two standard deviations above
the lower category's average. All three pair-wise comparisons were
very poor at distinguishing adjacent rating groups. When we compared
ICRAS ratings to our own continuous ratings the two ratings were able
to distinguish 11.1 percent of the comparisons made two categories
apart, 25 percent of the comparisons made three categories apart, and
28.6 percent of the comparisons made four categories apart. In
contrast, when we compared grouped Institutional Investor ratings to
continuous Euromoney ratings, the two ratings were able to
distinguish 77.8 percent of the comparisons made two categories apart
and 100 percent of the comparisons made 3 or 4 categories apart.
Also, when we compared grouped Euromoney ratings to continuous
Institutional Investor ratings, the two ratings were able to
distinguish 33.3 percent of the comparisons made two categories
apart, 75 percent of the comparisons made three categories apart, and
100 percent of the comparisons made four categories apart.
We then used regression analysis to measure both the systematic
similarities and differences between these executive branch ratings
and contemporaneous professionally recognized ratings that were
published in Euromoney and Institutional Investor in September 1992.
Table IV.6 presents the coefficient of determination, R\2 , which
measures the percentage of explained variance of the dependent
variable when we regressed the two external ratings--the September
1992 Euromoney ratings and the September 1992 Institutional Investor
ratings--on the executive branch ratings.
Table IV.6
Comparison of Executive Branch Ratings
With Other Country Risk Ratings
Independent
Dependent variable variable R\2 percent
-------------------- ------------------ ------------------
EM Executive branch 84.1
II Executive branch 79.6
EMGRP Executive branch 85.8
IIGRP Executive branch 84.9
Average 83.6
------------------------------------------------------------
Note: EM, Euromoney; II, Institutional Investor; GRP, grouped
external ratings.
In contrast to the results presented in table IV.6, when we regressed
these two external ratings on each other, continuous rating against
continuous rating and grouped rating against grouped rating, the
average R-squared was substantially higher at 94 percent--94.1
percent when the continuous Euromoney rating was the dependent
variable and 93.8 percent when the grouped Euromoney rating was the
dependent variable.\6 We concluded that these two external ratings
were each closer to each other than they were to executive branch
ratings.
Table IV.7 presents the results from our regressions that compared
contemporaneous secondary market prices of privately owned,
dollar-denominated, variable interest rate, sovereign developing
country debt (stripped of any guarantees and collateral) with the
executive branch ratings and each of the other risk ratings.\7 The
coefficients of determination are lowest when executive branch
ratings are the risk measure, indicating that secondary market prices
are less closely related to executive branch ratings than the other
ratings.\8 We also measured the internal consistency of these
executive branch ratings and found them to be generally internally
consistent, although less so than we expected.
Table IV.7
Comparison of Secondary Market Prices
With Country Risk Ratings
Risk measure R\2 percent
------------------------------ ----------------------------
Executive branch 84.4
II 92.8
EM 95.1
GAO 94.3
IIGRP 90.6
EMGRP 89.2
GAOGRP 93.9
Average excluding executive 92.7
branch
------------------------------------------------------------
Note: See table IV.6 for explanation of terms describing risk
measure.
We regressed numerical representations of all 188 of the executive
branch's latest ratings on numerical representations of their 5
subrating category ratings (A = 11, F- - = 1). We found that 6.3
percent of the executive branch rating variance was unexplained by
these subratings and that the weight of each subrating was 32 percent
for macroeconomics, 31.9 percent for payments history, 17 percent for
debt burden, 14.3 percent for balance of payments adjustment
capacity, and 9.8 percent for political and social factors.
We also reviewed the 35 individual rating subitems used by the
executive branch in these 5 broad categories. Although many of these
subitems are the executive branch's expectations of various measures
of a country's condition 5 years into the future, it did not
adequately combine these expectations. Executive branch officials
told us that they did test some of these indicators but acknowledged
that the indicators could be further studied. They did review the
professional literature when individual subitems were selected.
However, many of the 35 subitems did not have empirical studies
backing their use. The executive branch did not perform a systematic
econometric study to help it choose appropriate weights for
individual subitems so that its ratings would be more likely to
accurately forecast future payment problems.
Also, two subitems tended to be misaligned with the others. They
were high when the other 33 subitems tended to be low and vice versa,
even though all were measured on the same A to F- - scale. In
addition, executive branch subitems do not include any measure of
monetary policy or any measure of exchange rate overvaluation, both
of which are often considered by professional country analysts to be
important.
We also found that there was a good deal of empirical redundancy in
the 35 subitems, although this is not necessarily a bad
characteristic. When we performed step-wise regressions of the
overall executive branch ratings on the 35 subitems to determine
which subitems were the most important and which were redundant, we
found 15 subitems that appeared to be most important and the
remaining 20 subitems to be redundant.\9
--------------------
\5 These three external ratings of creditworthiness are continuous
and range from lowest creditworthiness at just above 0 to highest
creditworthiness at just below 100.
\6 The number of observations in each regression was 119. The
coefficient of determination was unchanged when we switched the
independent variable with the dependent variable. When we regressed
our continuous ratings and then our grouped ratings on these
executive branch ratings, R-squared was 87.3 percent and 86.5
percent, respectively.
\7 These regressions also included the months of interest arrears,
the type of instrument (loan or bond), and the maturity of the debt
instrument. They have the same form and are based on the same data
as in equation III.1, except for generally different measures of
country risk. The number of observations for each regression is 38.
\8 The t statistic of the risk measure is also lowest when executive
branch ratings are the risk measure.
\9 We used criteria that a subitem had to increase the amount of
adjusted "explained" variance of the overall rating (R bar squared)
in a regression of it and that other subitems already labeled
"important" had to join the list of important subitems.
OUR ESTIMATES OF THE LONG-RUN
PROBABILITY OF DEFAULT FOR LOANS
BY COUNTRY AND MATURITY
=========================================================== Appendix V
(Numbers in percents)
Maturity Maturity Maturity Maturity
Country 1 year 5 years 10 years 30 years
-------------------- -------- -------- -------- --------
Afghanistan 82.4 82.4 82.4 82.4
Albania 78.8 78.8 78.8 78.8
Algeria 49.1 49.1 49.1 49.1
Angola 72.6 72.6 72.6 72.6
Antigua and Barbuda 68.2 68.2 68.2 68.2
Argentina 45.2 45.2 45.2 45.2
Armenia 81.0 81.0 81.0 81.0
Australia 0.2 0.8 1.3 3.2
Austria 0 0 0 0
Azerbaijan 80.7 80.7 80.7 80.7
Bahamas 0.8 2.4 4.2 10.1
Bahrain 0.8 2.4 4.2 10.1
Bangladesh 66.3 66.3 66.3 66.3
Barbados 38.1 38.1 38.1 38.1
Belarus 70.1 70.1 70.1 70.1
Belgium 0.2 0.8 1.3 3.2
Belize 40.4 40.4 40.4 40.4
Benin 81.4 81.4 81.4 81.4
Bhutan 74.5 74.5 74.5 74.5
Bolivia 65.5 65.5 65.5 65.5
Bosnia-Herzegovia 73.0 73.0 73.0 73.0
Botswana 34.5 34.5 34.5 34.5
Brazil 51.4 51.4 51.4 51.4
Brunei 0.4 1.2 2.1 5.8
Bulgaria 64.4 64.4 64.4 64.4
Burkina Faso 67.9 67.9 67.9 67.9
Burma (Myanmar) 80.6 80.6 80.6 80.6
Burundi 67.2 67.2 67.2 67.2
Cambodia 92.1 92.1 92.1 92.1
Cameroon 61.0 61.0 61.0 61.0
Canada 0 0 0 0
Cape Verde 77.0 77.0 77.0 77.0
Central African 76.9 76.9 76.9 76.9
Republic
Chad 73.4 73.4 73.4 73.4
Chile 0.8 2.4 4.2 10.1
China, People's 0.8 2.4 4.2 10.1
Republic
Colombia 25.9 25.9 25.9 25.9
Congo 78.1 78.1 78.1 78.1
Costa Rica 56.3 56.3 56.3 56.3
Cote d'Ivoire 72.4 72.4 72.4 72.4
Croatia 69.7 69.7 69.7 69.7
Cuba 86.6 86.6 86.6 86.6
Cyprus 0.8 2.4 4.2 10.1
Czechoslovakia 17.8 17.8 17.8 17.8
Denmark 0.2 0.8 1.3 3.2
Djibouti 66.7 66.7 66.7 66.7
Dominican Republic 70.4 70.4 70.4 70.4
Ecuador 64.8 64.8 64.8 64.8
Egypt 57.1 57.1 57.1 57.1
El Salvador 72.0 72.0 72.0 72.0
Estonia 66.9 66.9 66.9 66.9
Ethiopia 79.5 79.5 79.5 79.5
Fiji 38.7 38.7 38.7 38.7
Finland 0.2 0.8 1.3 3.2
France 0 0 0 0
Gabon 51.6 51.6 51.6 51.6
Gambia, The 58.7 58.7 58.7 58.7
Georgia 77.8 77.8 77.8 77.8
Germany 0 0 0 0
Ghana 57.7 57.7 57.7 57.7
Greece 0.8 2.4 4.2 10.1
Grenada 78.8 78.8 78.8 78.8
Guatemala 71.5 71.5 71.5 71.5
Guinea 65.8 65.8 65.8 65.8
Guinea-Bissau 75.9 75.9 75.9 75.9
Guyana 83.6 83.6 83.6 83.6
Haiti 80.4 80.4 80.4 80.4
Honduras 73.1 73.1 73.1 73.1
Hong Kong 0.4 1.2 2.1 5.8
Hungary 20.5 20.5 20.5 20.5
Iceland 0.4 1.2 2.1 5.8
India 34.5 34.5 34.5 34.5
Indonesia 0.8 2.4 4.2 10.1
Iran 55.0 55.0 55.0 55.0
Iraq 87.6 87.6 87.6 87.6
Ireland 0.2 0.8 1.3 3.2
Israel 0.8 2.4 4.2 10.1
Italy 0.2 0.8 1.3 3.2
Jamaica 66.8 66.8 66.8 66.8
Japan 0 0 0 0
Jordan 63.6 63.6 63.6 63.6
Kazakhstan 72.1 72.1 72.1 72.1
Kenya 53.6 53.6 53.6 53.6
Korea, Democratic 87.4 87.4 87.4 87.4
People's Republic
Korea, Republic of 0.4 1.2 2.1 5.8
Kuwait 0.8 2.4 4.2 10.1
Kyrgyzstan 77.8 77.8 77.8 77.8
Latvia 68.5 68.5 68.5 68.5
Lebanon 78.3 78.3 78.3 78.3
Lesotho 54.8 54.8 54.8 54.8
Liberia 85.5 85.5 85.5 85.5
Libya 60.5 60.5 60.5 60.5
Lithuania 68.2 68.2 68.2 68.2
Luxembourg 0.2 0.8 1.3 3.2
Macedonia 67.4 67.4 67.4 67.4
Madagascar 68.5 68.5 68.5 68.5
Malawi 75.7 75.7 75.7 75.7
Malaysia 0.4 1.2 2.1 5.8
Mali 68.1 68.1 68.1 68.1
Malta 0.8 2.4 4.2 10.1
Mauritania 76.0 76.0 76.0 76.0
Mauritius 39.3 39.3 39.3 39.3
Mexico 15.9 15.9 15.9 15.9
Moldova 81.5 81.5 81.5 81.5
Mongolia 77.9 77.9 77.9 77.9
Morocco 41.9 41.9 41.9 41.9
Mozambique 85.7 85.7 85.7 85.7
Namibia 63.8 63.8 63.8 63.8
Nepal 60.9 60.9 60.9 60.9
Netherlands 0 0 0 0
New Zealand 0.2 0.8 1.3 3.2
Nicaragua 84.4 84.4 84.4 84.4
Niger 60.6 60.6 60.6 60.6
Nigeria 66.1 66.1 66.1 66.1
Norway 0.2 0.8 1.3 3.2
Oman 0.8 2.4 4.2 10.0
Pakistan 48.6 48.6 48.6 48.6
Panama 66.4 66.4 66.4 66.4
Papua New Guinea 40.4 40.4 40.4 40.4
Paraguay 57.8 57.8 57.8 57.8
Peru 77.3 77.3 77.3 77.3
Philippines 56.2 56.2 56.2 56.2
Poland 54.4 54.4 54.4 54.4
Portugal 0.4 1.2 2.1 5.8
Qatar 0.8 2.4 4.2 10.1
Romania 54.8 54.8 54.8 54.8
Russia 67.6 67.6 67.6 67.6
Rwanda 67.3 67.3 67.3 67.3
Sao Tome and 78.6 78.6 78.6 78.6
Principe
Saudi Arabia 0.4 1.2 2.1 5.8
Senegal 64.7 64.7 64.7 64.7
Seychelles 66.4 66.4 66.4 66.4
Sierra Leone 85.3 85.3 85.3 85.3
Singapore 0.2 0.8 1.3 3.2
Slovakia 34.2 34.2 34.2 34.2
Slovenia 60.4 60.4 60.4 60.4
Somalia 88.2 88.2 88.2 88.2
South Africa 23.4 23.4 23.4 23.4
Spain 0.2 0.8 1.3 3.2
Sri Lanka 57.2 57.2 57.2 57.2
St. Lucia 68.1 68.1 68.1 68.1
St. Vincent and 54.2 54.2 54.2 54.2
Grenadines
Sudan 86.2 86.2 86.2 86.2
Sweden 0.2 0.8 1.3 3.2
Switzerland 0 0 0 0
Syria 61.0 61.0 61.0 61.0
Taiwan 0.2 0.8 1.3 3.2
Tajikistan 79.1 79.1 79.1 79.1
Tanzania 78.2 78.2 78.2 78.2
Thailand 0.4 1.2 2.1 5.8
Togo 68.1 68.1 68.1 68.1
Trinidad and Tobago 46.9 46.9 46.9 46.9
Tunisia 27.6 27.6 27.6 27.6
Turkey 0.8 2.4 4.2 10.1
Turkmenistan 76.8 76.8 76.8 76.8
Uganda 0.4 1.2 2.1 5.8
Ukraine 80.7 80.7 80.7 80.7
United Arab Emirates 68.7 68.7 68.7 68.7
United Kingdom 0 0 0 0
Uruguay 36.7 36.7 36.7 36.7
U.S.S.R. 67.6 67.6 67.6 67.6
Uzbekistan 75.4 75.4 75.4 75.4
Vanuatu 54.5 54.5 54.5 54.5
Venezuela 26.6 26.6 26.6 26.6
Vietnam 72.8 72.8 72.8 72.8
Yemen, Republic of 68.3 68.3 68.3 68.3
Yugoslavia 75.5 75.5 75.5 75.5
Zaire 79.8 79.8 79.8 79.8
Zambia 80.7 80.7 80.7 80.7
Zimbabwe 47.4 47.4 47.4 47.4
------------------------------------------------------------
LEGAL ASPECTS OF RESCHEDULING
INTERNATIONAL DEBT
========================================================== Appendix VI
AUTHORITY TO RESCHEDULE
------------------------------------------------------ Appendix VI:0.1
The principal programs under which foreign debt is owed to the United
States are loans and loan guarantees made under the Export-Import
Bank Act of 1945; loans under Public Law 480; loans and loan
guarantees under the Foreign Assistance Act of 1961, as amended;
Foreign Military Sales (FMS) loans under the Arms Export Control Act;
and loan guarantees under the Commodity Credit Corporation Charter
Act. A 1970 opinion of the Attorney General addressed the
rescheduling of Indonesian loans under a number of these and other
programs. Default on all of the loans was imminent. Relying on the
broad authority contained in the authorizing statutes and the absence
of any prohibitions to the contrary, the Attorney General concluded
that the executive branch had the authority to reschedule the loans.
We examined that opinion and found no reason to question its
conclusions. Indeed, in 1987, we reviewed an executive branch
proposal to reschedule a FMS loan when default by a borrower was
imminent, and did not object to the proposal.\1
--------------------
\1 B-226718, August 19, 1987.
REQUIREMENT FOR BUDGET
AUTHORITY
------------------------------------------------------ Appendix VI:0.2
The Federal Credit Reform Act requires the President's budget to
include the estimated net long-term cost to the government (on a
present value basis) of credit programs in the year in which the loan
obligations or loan guarantee commitments are to be made. It further
requires that budget authority to cover this cost be provided in
advance of the obligations and commitments.
Section 504(e) of the act provides that a direct loan or loan
guarantee shall not be modified in a manner that increases its cost
to the government unless budget authority is set aside for the
additional cost. The act's implementing guidance, contained in OMB
Circular A-11, provides that "administrative work-outs of troubled
loans or loans in imminent default" are not loan modifications
requiring additional budget authority. Under Circular A-11, the
expected effects of an administrative work-out\2 on repayment are to
be included in the original subsidy cost estimate for a loan or loan
guarantee. The executive branch treats Paris Club rescheduling of
loans in imminent default as administrative workouts. Despite OMB
guidance, however, an OMB official told us that agencies have not
included the estimated costs of Paris Club rescheduling of loans at
below-market interest rates in their initial subsidy estimates. The
official told us that these rescheduling costs would show up for the
first time in the annual re-estimating process. This treatment of
Paris Club rescheduling costs is inconsistent with OMB guidance and
the act's requirement that the budget reflect the full subsidy costs
in the year in which loan obligations or loan guarantee commitments
are made.
Additionally, the OMB guidance requires that agencies reestimate
subsidy costs annually throughout the life of the loan. This
guidance requires that re-estimates are to be recorded in the current
year column of an agency's budget. The OMB official told us that the
Export-Import Bank is the only agency that re-estimates subsidy costs
for international loans and guarantees at least annually.
(See figure in printed edition.)Appendix VII
--------------------
\2 Circular A-11, sec. 33.5(o) provides that "work-outs are actions
undertaken to maximize repayments under existing direct loans or to
minimize claims under existing loan guarantees."
COMMENTS FROM THE OFFICE OF
MANAGEMENT AND BUDGET
========================================================== Appendix VI
(See figure in printed edition.)
(See figure in printed edition.)
See comment 1.
See p. 13.
(See figure in printed edition.)
See comment 2.
See p. 12.
See p. 12.
See comment 3.
See comment 4.
See comment 5.
(See figure in printed edition.)
The following are GAO's comments on OMB's letter dated September 8,
1994.
GAO COMMENTS
-------------------------------------------------------- Appendix VI:1
1. We agree that estimates should be caveated, as we did in our
draft report, when the subsidy cost of U.S. government loans is
based on market prices of privately owned loans. We believe our
method is professionally reputable because it (1) is well grounded in
theory, (2) uses generally accepted statistical estimating methods,
and (3) discloses sources of bias and qualifies estimates for any
bias. One important strength of our method is its ability to obtain
price estimates for private loans that did not trade. We did this by
first measuring the systematic effect of creditworthiness and other
loan characteristics on market prices for private loans that are
traded and then applied these systematic effects to the
creditworthiness and other characteristics of private nontraded loans
to obtain estimates of their prices.
2. We do not dispute OMB's interpretation of the act. However, we
believe that making a differentiation in estimated subsidy cost
between net increases in outstanding credit and lending that rolls
over maturing credit more accurately measures the true costs of
federal credit programs. Therefore, we have suggested that the
Congress may wish to consider how the principles included in the
Credit Reform Act ought to be applied to international credit
programs.
3. The agencies misinterpreted our comments concerning the use of
variable weights in determining ICRAS' country ratings. Our
criticism was that ICRAS' ratings were not adequately based on
statistical tests. Many of ICRAS' component economic indicators had
not passed statistical tests of significance, and econometric
techniques were not used to determine component weights, whether they
are fixed or variable. Using judgment has its costs. As ratings are
based less on empirical data and more on judgment, they are more
likely to have been affected by external considerations, such as
pressure to grant loans or guarantees to particular countries.
4. We found that while there was a fair degree of similarity between
ICRAS ratings and those of private analysts, there also was a fair
amount of difference. The agencies have not provided any empirical
evidence that ICRAS and private ratings diverge because of
differences in the probability of repayment of U.S. government
versus private creditors.
5. We believe this method is not onerous because calculations would
not have to be made for all 170 countries and certain statistical
operations could be used to simplify the process. Further, because
of weaknesses in estimating country risk and its costs, executive
branch estimates could be influenced by external considerations. In
contrast, our method is systematically and statistically based on
markets where similar debt is traded, and our country risk measures
are derived from an empirical estimating method that did not require
us to make qualitative judgments.
MAJOR CONTRIBUTORS TO THIS REPORT
======================================================== Appendix VIII
NATIONAL SECURITY AND
INTERNATIONAL AFFAIRS DIVISION,
WASHINGTON, D.C.
------------------------------------------------------ Appendix VIII:1
Ronald Kushner
Berel Spivack
Muriel Forster
GENERAL GOVERNMENT DIVISION,
WASHINGTON, D.C.
------------------------------------------------------ Appendix VIII:2
Art Kendall
OFFICE OF THE GENERAL COUNSEL
------------------------------------------------------ Appendix VIII:3
Ernie Jackson
RELATED GAO PRODUCTS
Loan Guarantees: Export Credit Guarantee Programs' Costs Are High
(GAO/GGD-93-45, Dec. 22, 1992).
Loan Guarantees: Export Credit Guarantee Programs' Long-Run Costs
Are High (GAO/NSIAD-91-180, Apr. 19, 1991).
Financial Audit: Commodity Credit Corporation's Financial Statements
for 1989 and 1988 (GAO/AFMD-91-5, July 29, 1991).
Financial Audit: Export-Import Bank's 1989 and 1988 Financial
Statements (GAO/AFMD-90-80, July 19, 1990).
Financial Audit: Commodity Credit Corporation's Financial Statements
for 1988 and 1987 (GAO/AFMD-89-83, Aug. 4, 1989).
Financial Audit: Export-Import Bank's 1988 and 1987 Financial
Statements (GAO/AFMD-89-94, July 25, 1989).
Federal Supervision of Overseas Lending by U.S. Banks
(GAO/T-NSIAD-89-42, June 27, 1989).
Financial Audit: Commodity Credit Corporation's Financial Statements
for 1987 and 1986 (GAO/AFMD-87-43, July 7, 1988).
Financial Audit: Export-Import Bank's 1987 and 1986 Financial
Statements (GAO/AFMD-88-48, May 19, 1988).
International Banking: Supervision of Overseas Lending Is Inadequate
(GAO/NSIAD-88-87, May 5, 1988).
Financial Audit: Export-Import Bank's 1986 and 1985 Financial
Statements (GAO/AFMD-87-61, Aug. 31, 1987).
Financial Audit: Commodity Credit Corporation's Financial Statements
for 1986 and 1985 (GAO/AFMD-87-43, June 22, 1987).
Comments on International Lending Institution Safety Act of 1987,
April 25, 1987, letter from the Comptroller General of the United
States to the Chairman, Senate Committee on Banking, Housing and
Urban Affairs.
Legislative and Administrative Obstacles to Writedowns and Swapping
of Less Developed Country Debt (GAO/T-NSIAD-87-29, Apr. 2, 1987).