International Financial Crises: Challenges Remain in IMF's
Ability to Anticipate, Prevent, and Resolve Financial Crises
(16-JUN-03, GAO-03-734).
Building on reform initiatives instituted after the Mexican
financial crisis, the IMF implemented new initiatives in the
mid-1990s to better anticipate, prevent, and resolve sovereign
financial crises. GAO was asked to assess (1) the IMF's framework
for anticipating financial crises, (2) the status of key IMF
reform initiatives to prevent financial crises, and (3) new IMF
proposals to resolve future financial crises.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-03-734
ACCNO: A07165
TITLE: International Financial Crises: Challenges Remain in
IMF's Ability to Anticipate, Prevent, and Resolve Financial
Crises
DATE: 06/16/2003
SUBJECT: Financial management
International economic relations
Risk management
Strategic planning
International Monetary Fund
******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO Product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
******************************************************************
GAO-03-734
a
GAO United States General Accounting Office
Report to the Chairman, Committee on Financial Services, and to the Vice
Chairman, Joint Economic Committee, House of Representatives
June 2003 INTERNATIONAL FINANCIAL CRISES Challenges Remain in IMF*s
Ability to Anticipate, Prevent, and Resolve Financial Crises
GAO- 03- 734
While the Fund*s new vulnerability framework is more comprehensive than
its previous efforts, it is too early to assess whether it will improve
Fund efforts to anticipate crises. The new framework uses the Fund*s major
forecasting tools, the World Economic Outlook (WEO) and the Early Warning
System (EWS), which have not performed well in anticipating prior crises.
The forecasting of crises has been historically difficult for all
forecasters.
The Fund, with the World Bank, has made progress in implementing
initiatives to prevent crises, but several challenges remain. To obtain
better information about country financial sector weaknesses, the Fund and
Bank introduced the Financial Sector Assessment Program (FSAP) to report
on member countries* financial sectors and the Reports on the Observance
of Standards and Codes (ROSC) to assess member countries* adherence to 12
standards. Assessments have not been completed in some major emerging
market countries primarily because participation is voluntary, and use of
this information has been mixed. For example, some private sector market
participants have found the reports untimely, outdated, and dense.
Participation Gaps by 33 Major Emerging Market Countries in Key
Assessments, 1999- 2003
Number of assessments The Fund is considering two approaches to
restructuring unsustainable sovereign debt; however, there are significant
challenges to implementing them. One approach involves creating an
international legal framework that would allow a specified majority of a
country*s external creditors to restructure most private sector loans.
Under the second approach, the Fund is encouraging members to include
renegotiation clauses in individual bonds. Many private sector
representatives wish to maintain the existing process in which the Fund
assists resolution by providing loans to some eligible members. In
response to concerns that its resources may have unintended negative
impacts during a crisis, the Fund has clarified and strengthened its
criteria for lending to members experiencing crises. Building on reform
initiatives instituted after the Mexican
financial crisis, the IMF implemented new initiatives in the mid- 1990s to
better anticipate, prevent, and resolve sovereign financial crises. GAO
was asked to assess (1) the IMF*s framework for anticipating financial
crises, (2) the status of key IMF reform initiatives to prevent financial
crises, and (3) new IMF proposals to resolve
future financial crises. GAO recommends that the Secretary of the Treasury
instruct the U. S. Executive Director of the
Fund to work with other Executive Board members to encourage the Fund to
improve the timeliness of
FSAP and ROSC reports; expand the coverage,
frequency, and publication of updates of participants* implementation of
FSAP and ROSC recommendations; improve the FSAP and ROSC reports*
readability; and
increase participation in the FSAP and all standards of the ROSC and
consider making participation mandatory.
Treasury, IMF, and the World Bank generally agreed with the report*s
recommendations. The IMF stated
that we gave WEO and EWS forecasts greater importance than is warranted in
anticipating crises. However, we focused on the only mature and
quantifiable elements of the vulnerability framework.
www. gao. gov/ cgi- bin/ getrpt? GAO- 03- 734. To view the full report,
including the scope and methodology, click on the link above. For more
information, contact Joseph A. Christoff at (202) 512- 8979 or christoffj@
gao. gov. Highlights of GAO- 03- 734, a report to
Chairman, House Committee on Financial Services and Vice Chairman, House
Joint Economic Committee June 2003
INTERNATIONAL FINANCIAL CRISES
Challenges Remain in IMF's Ability to Anticipate, Prevent, and Resolve
Financial Crises
Page i GAO- 03- 734 International Financial Crises
Contents Letter 1
Background 2 Results in Brief 4 Too Early to Determine If Fund*s New
Process for Anticipating
Financial Crises Improves on Past Efforts 7 Some Progress in Implementing
Crisis Prevention Initiatives, but
Challenges Remain 15 The Fund*s Efforts to Better Resolve Future Financial
Crises Face
Significant Challenges 24 Conclusion 30 Recommendations for Executive
Action 31 Agency Comments and Our Evaluation 31
Appendixes
Appendix I: Objectives, Scope, and Methodology 34
Appendix II: Private Sector Structured Interview Participants 37
Appendix III: Assessment of IMF Forecasting 38 GAO Analysis Used Standard
Econometric Techniques for Forecast
Evaluation 38
Appendix IV: Eighty- seven Emerging Market Countries 50
Appendix V: Standards, Codes, and Principles Assessed under IMF and World
Bank Reports on the Observance of Standards and Codes 53
Appendix VI: Update on the International Monetary Fund*s Safeguards
Assessments 56 The Fund Introduces Safeguards Assessments for Member
Countries That Currently Borrow 56 Implementation of the Fund*s Assessment
Recommendations 57
Appendix VII: Fund/ World Bank FSAP and ROSC Process 59
Appendix VIII: Country Participation in and Publication of FSAPs and ROSCs
60
Appendix IX: Fund Resources Provided under the Supplemental Reserve
Facility 64
Appendix X: Comments from the Department of the Treasury 65
Appendix XI: Comments from the International Monetary Fund 67 GAO Comments
70
Contents
Page ii GAO- 03- 734 International Financial Crises Appendix XII: Comments
from The World Bank 72
GAO Comments 74
Appendix XIII: GAO Contacts and Staff Acknowledgments 75 GAO Contacts 75
Staff Acknowledgments 75
Tables Table 1: Anticipated and Actual GDP Growth Rate of 14 Financial
Crises, 1990* 2001 12
Table 2: Forecast Quality for 87 Emerging Market Countries, 1990* 2001 43
Table 3: Forecast Quality for G- 7 Countries, 1990* 2001 45 Table 4:
Comparison of Year- Ahead WEO Forecasts for Program
and Nonprogram Countries, 1990* 2001 46 Table 5: Comparison of Program and
WEO Year- Ahead Forecasts,
for 52 Countries in the Initial Years on Program, 1990* 2001 47 Table 6:
IMF Supplemental Reserve Facility Loans, 1997* 2002 64
Figures Figure 1: IMF Initiatives to Anticipate, Prevent, and Resolve
Financial Crises 4
Figure 2: IMF Vulnerability Assessment Framework 8 Figure 3: Participation
of 33 Major Emerging Markets in Key ROSC Assessments, 1999* 2003 20
Figure 4: FSAP and ROSC Participation and Publication by Major Emerging
Market Countries 61 Figure 5: FSAP and ROSC Participation and Publication
by Other
Countries 62
Abbreviations
BCP Basel Core Principles BMA Bond Market Association CAC Collective
Action Clauses CCFF Compensatory Contingency Finance Facility CPI consumer
price index CPSS Committee on Payments and Settlements Systems DCSD
Developing Country Studies Division EFF Extended Fund Facility EMCA
Emerging Market Creditors Association
Contents
Page iii GAO- 03- 734 International Financial Crises
EMTA Emerging Market Traders Association EWS Early Warning System FATF
Financial Action Task Force FSA Financial Sector Assessment FSAP Financial
Sector Assessment Program FSSA Financial System Stability Assessment GDDS
General Data Dissemination Standard GDP gross domestic product IAIS
International Association of Insurance Supervisors IASB International
Accounting Standards Board IFAC International Federation of Accountants
IMF International Monetary Fund IOSCO International Organization of
Securities Commissions IPMA International Primary Market Association ISMA
International Securities Market Association KLR Kaminsky, Lizondo, and
Reinhart MSE mean square error OECD Organization for Economic Cooperation
and Development RMSE root mean square error ROSC Reports on the Observance
of Standards and Codes RSS Residual Sum of Squares SBA Stand- By
Arrangements SDDS Special Data Dissemination Standard SDRM Sovereign Debt
Restructuring Mechanism SEC U. S. Securities and Exchange Commission SIA
Securities Industry Association SRF Supplemental Reserve Facility STF
Systemic Transformation Facility UNCITRAL United Nations Commission on
International Trade Law WEO World Economic Outlook
This is a work of the U. S. Government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. It may contain
copyrighted graphics, images or other materials. Permission from the
copyright holder may be necessary should you wish to reproduce copyrighted
materials separately from GAO*s product.
Page 1 GAO- 03- 734 International Financial Crises United States General
Accounting Office
Washington, D. C. 20548 Page 1 GAO- 03- 734 International Financial Crises
A
June 16, 2003 Let t er
The Honorable Michael G. Oxley Chairman Committee on Financial Services
House of Representatives
The Honorable Jim Saxton Vice Chairman Joint Economic Committee House of
Representatives
In May 2000, in the aftermath of the Asian financial crisis, the newly
appointed Managing Director of the International Monetary Fund (IMF or the
Fund) announced his plans to strengthen the Fund*s reform initiatives to
enable the Fund to more effectively safeguard the stability of the
international financial system. Building on reform initiatives instituted
in the mid- 1990s, after the Mexican financial crisis of 1994* 95, these
new initiatives were designed to better anticipate, prevent, and resolve
sovereign financial crises. A sovereign financial crisis can occur when a
country is unable or unwilling to honor its debt obligations and investors
lose confidence in that country*s financial markets. Recent crises have
occurred primarily in *emerging market* countries, larger and more
economically advanced developing countries such as Argentina, Brazil, and
Turkey. According to World Bank (the Bank) estimates, the financial costs
to countries that experienced crises in the 1980s and 1990s exceeded $1
trillion* greater than the total amount of all donors* assistance to
developing countries in the 1980s and 1990s. This letter responds to your
request that we examine the Fund*s efforts to better safeguard the
stability of the international financial system. In this report, we assess
(1) IMF*s framework for anticipating financial crises, (2) the status of
key IMF initiatives to prevent financial crises, and (3) new IMF proposals
to resolve future financial crises. In addition, we analyzed the
quality of IMF*s forecasts produced by the World Economic Outlook (WEO),
the Fund*s primary forecasting tool (see appendix III).
As part of our assessment, we reviewed documents from the U. S.
government, international organizations, and private firms, including
testimonies, reports, and relevant laws. We used statistical models that
we developed to assess the Fund*s WEO forecasts and program projections
for selected emerging market countries. We interviewed key officials at
the
Page 2 GAO- 03- 734 International Financial Crises
Fund, World Bank, Department of the Treasury (Treasury), and emerging
market bond and equity funds. We also interviewed some representatives of
private sector firms that specialize in international finance, economics,
and law. Appendix I provides a more detailed description of our
objectives, scope, and methodology. Appendix II provides a list of the
private sector firms we interviewed.
Background In 1994* 95, Mexico faced a severe financial crisis when a
shift in market sentiment led to sudden large capital outflows. Investors
also temporarily removed their funds from other emerging market countries,
an effect known as contagion. In response to the crisis, Mexico quickly
adopted a strong and ultimately successful program of adjustment and
reform. To support the program, the Fund approved a loan of $17.8 billion
to Mexico*
one of the largest loan commitments it had ever made to a country. One of
the major reasons cited for the crisis was the lack of timely, reliable,
and publicly available economic, financial, and sociodemographic data for
Mexico. Beginning in 1996, to correct this weakness, the Fund created data
standards to guide countries in disseminating better data to the public.
However, as we reported in 1997, the Fund needed to address a number of
other financial, economic, and political challenges, in addition to data
limitations, to better anticipate, prevent, and resolve financial crises.
1
Before the Fund could fully address these challenges, the Asian financial
crisis of 1997* 98 occurred. After the Asian financial crisis, the Fund
assessed the effects of its responses to the crisis and reassessed its
role in
safeguarding the stability of the international financial system,
including rethinking its core mission, operations, and lending activities.
The Fund also recognized that it needed to improve its ability to
anticipate financial crises; monitor countries* activities; and increase
public awareness, particularly that of the investment community.
Recognizing its inability to anticipate past crises, the Fund instituted a
quarterly vulnerability assessment framework in 2001 to identify countries
that may be susceptible to crisis.
1 U. S. General Accounting Office, International Financial Crises: Efforts
to Anticipate, Avoid, and Resolve Sovereign Crises, GAO/ GGD/ NSIAD- 97-
168 (Washington, D. C.: July 7, 1997).
Page 3 GAO- 03- 734 International Financial Crises
To improve its ability to prevent future crises, the Fund and the World
Bank in 1999 began performing joint assessments of member countries*
financial sectors to help identify and monitor existing and potential
weaknesses. In addition, the Fund and the World Bank began to work with
countries to promote adherence to voluntary standards to reassure the
international community that the countries* policies and practices conform
to standards and codes of good practice. These include standards to
improve transparency in government economic data; fiscal, monetary, and
financial policies; and guidelines on strengthening the financial and
corporate sectors. The Fund acknowledges that it would be almost
impossible to anticipate or prevent all crises. According to the Fund,
past efforts to resolve financial
crises during the 1990s were lengthy and very costly to debtor countries.
The Fund is encouraging the adoption of agreements that would allow a
quicker, more orderly, and predictable restructuring of countries* debts.
The Fund*s ultimate goal is to maintain investor confidence and stability
in the international financial system. Figure 1 shows the Fund*s key
initiatives for better anticipating, preventing, and resolving financial
crises. Treasury through the U. S. Executive Director to the Fund has the
lead responsibility for monitoring the IMF*s progress in addressing these
issues.
Page 4 GAO- 03- 734 International Financial Crises
Figure 1: IMF Initiatives to Anticipate, Prevent, and Resolve Financial
Crises
Results in Brief While it is too early to assess the effectiveness of the
Fund*s new vulnerability framework for anticipating crises, this framework
builds on
key elements that have been unable to anticipate past financial crises.
The new framework is more comprehensive than the Fund*s previous efforts,
bringing together country- specific knowledge and financial expertise
within the Fund to better identify weaknesses in emerging market economies
that could lead to a crisis. However, the new framework uses the Fund*s
major forecasting tools, the WEO and the Early Warning System (EWS), which
have not performed well in anticipating prior crises. The
WEO has not successfully anticipated past financial crises, and the Fund*s
EWS models have had a high false alarm rate, having predicted many crises
that did not occur. The forecasting of crises has been historically
difficult for all forecasters due to complex underlying factors, including
concerns about the reliability of important macroeconomic data on emerging
market countries.
Page 5 GAO- 03- 734 International Financial Crises
The Fund, in collaboration with the World Bank, has made progress in
implementing initiatives to prevent financial crises, but several
challenges remain. In the late 1990s, the Fund and the World Bank adopted
two key initiatives designed to assist four parties: Fund staff, World
Bank staff, member country governments, and private sector participants.
First, financial sector assessments were introduced to provide reports on
aspects
of member countries* financial sectors such as banking systems and crisis
management capacity. Second, the standards initiative was adopted to
assess member countries* adherence to 12 standards in areas such as
banking supervision and economic data dissemination. Use of the
information provided by these two initiatives by various parties has been
mixed, and several significant challenges remain. Fund staff frequently
incorporate information from the assessments into their policy advice to
member countries. However, assessments have not been completed in some
important emerging market countries, such as China and Thailand, primarily
because participation is voluntary. World Bank staff*s use of the
assessments to inform country development assistance programs is also
affected by the lack of participation by member countries and by borrower
countries* competing development needs. Member country governments
sometimes use the assessments in prioritizing their reform agendas but
often find that the reforms are too difficult to implement. Some private
sector participants find the published reports outdated, untimely, and too
dense to be useful in making investment decisions. The Fund and the World
Bank acknowledge that the assessments cannot prevent all crises because
reforms require many years to be fully implemented, and crises can be
caused by factors outside the scope of the reforms.
The Fund is considering two approaches to achieve a more orderly and
predictable restructuring of unsustainable debt between a country and its
creditors; however, there are significant challenges to implementing both
approaches. Fund officials assert that the current process for
renegotiating the terms of member countries* loans with external private
sector creditors is lengthy and costly. The Sovereign Debt Restructuring
Mechanism (SDRM), is a proposed international legal framework that would
allow a majority of a country*s external creditors to quickly restructure
most private sector loans. The Fund is also encouraging members to include
Collective Action Clauses (CACs) in individual bonds, which would allow a
majority of bondholders to renegotiate the bond*s terms. Although some
elements of both approaches are acceptable to the private sector and to
governments, a number of political, legal, and technical challenges stand
in the way of implementing the SDRM; it seems unlikely that these issues
will be resolved in the immediate future. While private sector officials
expect
Page 6 GAO- 03- 734 International Financial Crises
that many restructurings need only involve the private sector and the
debtor country, under some circumstances, voluntary debt restructurings
will not adequately resolve all financial crises. In those cases, they
said the Fund should provide loans to eligible countries to help fill
their external financing gaps. However, some financial experts and
government representatives have raised concerns that such loans have the
potential to
increase the probability of future crises. In response, the Fund has
clarified and strengthened its policy of lending in crisis situations.
The Fund*s crisis prevention initiatives are hindered by several factors
that limit their effectiveness. To strengthen the effectiveness of these
initiatives, we are recommending that the Secretary of the Treasury
instruct the U. S. Executive Director of the Fund to work with other
Executive Board members to encourage the Fund to improve the readability,
timeliness, coverage, and frequency of updates of assessment reports.
Additionally, the
Fund should pursue strategies for increasing participation in the
assessment process including the possibility of making participation
mandatory for all members of the IMF. In responding to our draft report,
Treasury, IMF, and the World Bank
generally agreed with the report*s recommendations. However, the IMF
stated that we mischaracterized the role of WEO forecasts and EWS models
in IMF crisis anticipation efforts by saying that they have a greater
importance than is warranted. We disagree with this depiction. Our
assessment examined all six components of the IMF*s vulnerability
assessment framework, including the WEO and EWS. As the only mature and
quantifiable elements of the framework, our analysis focused more heavily
on the track records of the WEO and EWS. The World Bank expressed concern
with the report*s suggestion that consideration be given in making
participation in the FSAP and ROSC assessments mandatory. While we are not
suggesting that the assessments should be made
mandatory, the voluntary nature of these assessments has posed an obstacle
to full participation by important emerging market countries.
Page 7 GAO- 03- 734 International Financial Crises
Too Early to Determine If Fund*s New Process for Anticipating Financial
Crises Improves on Past Efforts
In May 2001, the IMF implemented a new vulnerability assessment framework
for emerging market countries to strengthen the Fund*s ability to
anticipate financial crises. This framework brings together
countryspecific knowledge and financial expertise within the Fund to
better identify weaknesses in emerging market economies that could lead to
a crisis. Although the new vulnerability assessment framework is more
comprehensive than the Fund*s previous efforts, it is new and still
evolving. It is too early to tell whether this new framework will
successfully anticipate future crises. The new framework uses the Fund*s
major forecasting tools, the WEO and the EWS, which have not performed
well in anticipating prior crises. The WEO has not successfully
anticipated past financial crises, and the Fund*s EWS models have had a
high false alarm rate, having predicted many crises that did not occur.
The forecasting of crises has been historically difficult for all
forecasters due to complex underlying factors, including concerns about
the reliability of important macroeconomic data on emerging market
countries.
The IMF Has Recently Implemented a New Vulnerability Assessment Framework
The Fund has attempted for many years to identify countries vulnerable to
financial crisis; however, their existing tools were insufficient to
anticipate the financial crises of the 1990s and led the Fund in 2001 to
develop the new vulnerability assessment framework. This comprehensive
framework brings together detailed, country- specific knowledge and
financial expertise of various IMF departments, including those with
regional, macroeconomic, or forecasting expertise. The new framework
monitors the vulnerability of key emerging market countries that borrow
significantly from international capital markets. This information is
provided in a quarterly report on crisis vulnerability. Fund staff report
monthly on countries identified as vulnerable and provide more frequent ad
hoc analyses during volatile periods.
To conduct the vulnerability assessment, the Fund integrates six
independent inputs that represent the analyses and perspectives of
different departments of the Fund (see figure 2).
Page 8 GAO- 03- 734 International Financial Crises
Figure 2: IMF Vulnerability Assessment Framework
World Economic Outlook: The WEO is a twice yearly publication that
provides analyses of global economic developments. Through the WEO, the
IMF provides current and following year forecasts for countries and
regions of key economic variables such as economic growth, inflation, and
current account. 2 According to Fund staff, WEO forecasts use the best
available information and represent the most realistic estimate of key
economic variables, including those that could help anticipate a
financial crisis. The IMF uses these forecasts as an input in the
vulnerability assessment to gauge the impact of unanticipated adverse
changes in the global environment. For example, the WEO forecasts for
selected countries may be recalculated to examine the impact of sudden
2 The current account represents a country*s trade balance (exports less
imports of goods and services) plus the net interest income (or payments)
on outstanding international investments (or debts) plus net transfers
(grants, workers* remittances).
Page 9 GAO- 03- 734 International Financial Crises
increases in oil prices or an unanticipated recession in the advanced
economies. Early Warning System models: The Fund uses internal and
private
sector EWS models that compute the probability of a country having a
crisis over the following 12 to 24 months. The model examines a series of
vulnerability indicators, including whether a country*s real exchange rate
is overvalued, or whether the country has significantly depleted its
foreign exchange reserves. 3 The output of the EWS models helps the Fund
focus on specific areas of vulnerability. For example, if one variable,
such as the exchange rate, signals a crisis, the Fund would
more closely examine related components of the vulnerability assessment,
such as a country*s external financing requirements.
Country external financing requirements: On a quarterly basis, the Fund
produces an internal assessment of a country*s ability to meet its total
external debt obligations and estimates whether that country has
sufficient foreign exchange to avoid a crisis. This assessment includes
estimating a country*s ratio of foreign exchange reserves to short- term
external debt, estimating the magnitude of its current account deficit,
and considering whether and how it manages its exchange rate.
Market information: On an ongoing basis, the Fund analyzes most
countries* cost of borrowing on the international market and whether the
country is paying a higher interest rate than similar countries. The Fund
uses this information to provide an internal analysis of the private
sector*s expectations of a country*s likelihood of default on its foreign
debt and to identify possible evidence that financial problems are
spreading across countries.
Financial sector vulnerability: The Fund assesses the strengths and
weaknesses of a country*s financial sector, including the banking system.
IMF staff evaluate the financial sector*s vulnerability to changes in
market conditions, such as fluctuations in interest and exchange rates.
Although the detailed results of these assessments are used internally by
the Fund, summaries of key findings are frequently published.
3 Foreign exchange reserves are those external assets that are readily
available to and controlled by monetary authorities. Reserves can be used
for direct financing of payment imbalances, for indirectly regulating the
exchange rate, and/ or for other purposes.
Page 10 GAO- 03- 734 International Financial Crises
Country expert perspectives: IMF country experts examine the data
produced by the above analyses, supplementing those results with country-
specific details such as the political risks of implementing certain
government policies or the relevance of certain market information.
Until 1999, the Fund used the WEO as the primary forecasting tool to help
identify country risks and vulnerability to crises. The new vulnerability
framework, which has been in operation for about 2 years, is a more
comprehensive process. According to the Fund, the quarterly integration of
detailed information from country experts who continuously monitor
developments in their countries is a great strength of the new
vulnerability framework. Effective analysis by Fund staff of the
framework*s six elements could better allow the Fund to give timely advice
to authorities in vulnerable countries. It is too early to tell whether
this framework will be successful in anticipating future crises. We assess
the performance of the WEO and EWS models, the Fund*s primary tools for
anticipating crises prior to the implementation of the new framework in
May 2001. 4 Forecasting Elements of the
Vulnerability Framework Have Performed Poorly in the Past
The new vulnerability assessment framework uses the Fund*s two major
forecasting tools* the WEO and the EWS models* which have not performed
well in anticipating prior crises. The WEO has not successfully
anticipated the severe financial crises of the past decade. The Fund*s EWS
models have had a high false alarm rate, having predicted many crises that
did not occur. The WEO Is Not a Reliable Tool for Anticipating Crises
Severe financial crises are characterized by a number of negative economic
outcomes, including large declines in gross domestic product (GDP), also
known as recessions. We found that the WEO had a poor record of
forecasting such declines in GDP, tending instead to follow existing
positive economic growth trends. In addition, the IMF indicates that the
current account is a key variable in explaining financial crises. We found
that for the current account, the accuracy of 75 percent of WEO country
forecasts was worse than simply assuming that next year*s value is the
same as this year. The WEO*s difficulty forecasting both GDP and current
4 Data on the performance of the other four components of the framework
were not made available to us because these elements are considered part
of the staff level deliberative process and are not provided to the
Executive Board.
Page 11 GAO- 03- 734 International Financial Crises
account demonstrate poor performance in anticipating the severe financial
crises of the past decade.
WEO Has Performed Poorly at Forecasting Recessions
In most cases, countries experiencing a financial crisis also experience a
severe recession in which their GDP declines significantly. Although most
recessions do not involve a major financial crisis, successful
anticipation of recessions, especially the most severe ones, would greatly
increase the likelihood of anticipating impending financial crisis.
However, we found that the WEO has a poor track record of forecasting
recessions, including
those directly associated with a financial crisis. The WEO did not
forecast most of the recessions that occurred in emerging market countries
in the last 10 years. During the 1991* 2001 forecast period, 134
recessions occurred in all 87 emerging market countries. We found that the
WEO correctly forecast only 15, or 11 percent, of those recessions,
while predicting an increase in GDP in the other 119 actual recessions. 5
The WEO is considerably more likely to forecast a recession when a
recession has occurred in the prior year. However, a prior year recession
did not occur in two- thirds of the recessions that the WEO failed to
forecast. Thus, WEO forecasts generally follow the existing growth trend
within a country, making it unlikely that the WEO would correctly forecast
an unanticipated recession. Furthermore, this tendency to follow the
current growth trend
makes it especially difficult for the WEO to anticipate a financial crisis
because nearly all of the crisis- related recessions of the last decade
occurred in years following positive economic growth.
Further illustrating our point, the WEO was unable to anticipate large
declines in GDP, also known as recessions, that corresponded to 14 major
financial crises of the past decade, including the Mexican and Asian
financial crises (see table 1). 6 5 In one case, the WEO forecast zero
growth.
6 Several studies examining international financial crises from 1990* 2001
identified 14 major financial crises in 12 prominent emerging market
countries that were considered to be fundamentally sound before the
crisis. Balance of payments problems, large GDP contractions, and
financial panic commonly characterize these crises.
Page 12 GAO- 03- 734 International Financial Crises
Table 1: Anticipated and Actual GDP Growth Rate of 14 Financial Crises,
1990* 2001
Sources: Various crisis studies and IMF. a Actual GDP growth rate is from
the WEO published 1 year later.
b The Asian crisis began in Thailand in the summer of 1997 and spread to
the other Asian countries in the latter part of 1997. The full effects of
the crisis on the GDP of other Asian countries and Russia occurred in
1998. Thus we are using 1998 actual GDP growth rate data for the Asian
countries* Indonesia, Malaysia, Philippines, South Korea, and Thailand.
The crisis in Brazil extended into1999.
The WEO*s failure to identify these recessions demonstrates that it did
not anticipate the corresponding financial crises. In 14 cases, we found
that the WEO forecast strong economic growth, averaging a 4. 4 percent
increase in GDP, despite large declines in actual GDP in 13 of 14 cases. 7
In fact, actual GDP declined by an average of 5. 7 percent during the
first full year of these 14 financial crises. Indonesia presents the most
startling disparity, in which the WEO forecast a growth of 6.2 percent in
its GDP, when in fact
Indonesia*s GDP declined by almost 14 percent in the first full year of
its financial crisis.
Country Crisis years WEO forecast
GDP growth rate (percent)
Actual GDP growth rate a
(percent)
Argentina 1995 4.0 -4.4 Argentina 2001 3.7 -4.4 Brazil 1998 4.0 0. 1
Ecuador 1999 2.5 -7.3 Indonesia b 1998 6.2 -13.7 Malaysia b 1998 6.5 -6.7
Mexico 1995 4.0 -6.9 Philippines b 1998 5.0 -0.5 Russia 1998 5. 1 -4.6
South Korea b 1998 6.0 -5.8 Thailand b 1998 3.5 -9.4 Turkey 1994 3.0 -5.5
Turkey 2001 5.5 -7.4 Venezuela 1994 2. 5 -3.3 Average growth rates 4.4
-5.7
7 Brazil was the sole exception, having experienced a small 0.1 percent
GDP annual growth rate during the 1998 financial crisis.
Page 13 GAO- 03- 734 International Financial Crises WEO Does a Poor Job in
Forecasting the Current Account
According to the Fund, a country*s current account (primarily exports
minus imports of goods and services) is a key variable in anticipating
crises. Crises are associated with problems of external financing that
result from a country having difficulty obtaining foreign exchange. Since
exports are an important source of foreign exchange for developing
countries, projections of a country*s current account balance provide
information
about the country*s ability to earn foreign exchange and to service its
external debt. 8 According to the Fund, an unsustainably large current
account deficit can contribute to or precipitate a crisis. 9 We found that
WEO forecasts for current account were inaccurate most of
the time. Our analysis for the 87 emerging market countries shows that,
for more than 75 percent of the countries, the WEO current account
forecasts were less accurate than if the Fund had simply assumed that the
next year*s current account would be the same as this year*s. The results
are even more
dramatic for G- 7 countries; a forecast of no change was a better
predictor than the WEO forecast for six of the seven countries. This
demonstrates that, even in stable economies with excellent data, the WEO
has done a poor job of forecasting this key crisis anticipation variable.
(See appendix III for more detailed explanation on our methodology and
findings).
8 A deficit in the current account can also be financed by capital
inflows, such as foreign private investment, a drawing down of a country*s
international reserves, by bilateral or multilateral loans, or by
provision of exceptional finance, such as debt service relief. 9 Fund
studies indicate that the current account deficit, as a percentage of GDP,
is significantly larger before a crisis than during relatively stable
periods, which suggests that unsustainable current account deficits tended
to be part of the general overheating of the
economy before a crisis.
Page 14 GAO- 03- 734 International Financial Crises
The Fund*s Early Warning System Models Indicate Many Crises That Do Not
Occur
Since 1999, the Fund has analyzed the results from internal and private
sector EWS models in its crisis anticipation efforts. The Fund*s internal
efforts focused on two EWS models to systematically identify countries
vulnerable to crises: the Kaminsky, Lizondo, and Reinhart (KLR) model,
which monitors a set of 15 monthly variables that signal a crisis whenever
any cross a certain threshold; and the Developing Country Studies Division
(DCSD) model, which uses five variables to compute the probability of a
crisis occurring in the next 24 months. 10 The Fund*s models use a
variableby- variable approach that allows economists to determine which
variables are signaling the crisis. Internal assessments of the Fund*s EWS
models show that they are weak predictors of actual crises. While the
models worked reasonably well in anticipating Turkey*s recent financial
crisis, they did not successfully anticipate Argentina*s financial crisis
in 2002.
According to the Fund, the models* most significant limitation is that
they have high false alarm rates; that is, they predict many crises that
did not occur. In about 80 percent of the cases where a crisis was
predicted over the next 24 months, no crisis occurred. Furthermore, in
about 9 percent of the cases where no crisis was predicted, there was a
crisis. 11 Forecasting Crises Has
Been Historically Difficult for All Parties
Financial crises have been historically difficult to anticipate because of
a number of complex underlying factors. Economic outcomes are often
influenced by unanticipated events such as conflicts and natural
disasters. Many factors, in addition to weaknesses in a country*s
financial structure, can lead to a crisis. These include economic
disturbances, such as an unanticipated drop in export prices, political
events, and changes in investor sentiment leading to sudden withdrawals of
foreign capital. Furthermore, data may be inadequate, particularly in
developing countries
where data are often not timely and are of poor quality. 10 More
specifically, the DCSD model is a multivariate probit regression. The five
variables used are real exchange rate overvaluation, current account,
foreign exchange reserve losses, export growth, and the ratio of short-
term debt to foreign exchange reserves. Also, the model defines a currency
crisis as the weighted average of 1- month changes in exchange rate and
reserves more than 3 (country- specific) standard deviations above country
average.
11 The DCSD model had a cutoff probability of 23 percent and the KLR model
had a cutoff probability of 15 percent. A forecast probability above these
cutoff points is deemed to signal a crisis.
Page 15 GAO- 03- 734 International Financial Crises
Forecasters consistently fail to foresee crises and recessions. Forecasts
produced by private sector economic forecasters, governments, and
multinational agencies including the IMF and the Organization for Economic
Cooperation and Development, routinely fail to foresee the coming of
crises and recessions, and often fail to outperform the naive model, which
simply assumes that next year*s outcome will be the same as this year*s.
This is true even for evaluations of recent U. S. forecasts of GDP and
inflation. Our review of a number of forecast evaluation studies confirms
that the inability to predict recessions is a common feature of growth
forecasts for both industrialized and developing countries. The studies
also showed that forecast accuracy improves as the time horizon gets
shorter, and that there is little difference in forecast accuracy between
private sector and WEO forecasts.
Some Progress in Implementing Crisis Prevention Initiatives, but
Challenges Remain
In the late 1990s, the Fund and the World Bank began implementing two
crisis prevention initiatives designed to assist four parties: IMF staff,
World Bank staff, member country governments, and private sector
participants. The first initiative, the financial sector assessments,
provides reports on aspects of member countries* financial sectors such as
banking systems and crisis management capacity. The second, the standards
initiative, assesses member countries* adherence to 12 standards in areas
such as banking supervision and economic data dissemination. Parties* use
of the information provided by these two initiatives has been mixed, and
several significant challenges remain. Fund staff frequently incorporate
information from the assessments into their policy advice when assessments
are available; however, assessments have not been completed
in some important emerging market countries primarily because
participation is voluntary. Bank staff*s use of the assessments to inform
country development assistance programs is also affected by gaps in the
completion of some assessments and by borrower countries* competing
development demands. Member country governments sometimes use the
assessments in prioritizing their reform agendas but often find the
reforms too difficult to implement. Some private sector participants find
the published reports untimely, outdated, and too dense to be useful in
making investment decisions. The IMF and the Bank acknowledge that the
initiatives cannot prevent all crises because recommended reforms require
many years to be fully implemented, and crises can be caused by factors
outside the scope of the reforms.
Page 16 GAO- 03- 734 International Financial Crises
Two Key Crisis Prevention Initiatives Target Four Parties
In the wake of the Mexican and Asian financial crises of the 1990s, the
Fund and the World Bank became increasingly aware of the importance of
transparent financial data and policies, stronger financial systems, and
better- functioning markets as a complement to member country
governments* sound macroeconomic policies. Fund evaluations acknowledge
that the institution failed to collect information that could have enabled
it to detect financial and corporate sector vulnerabilities and
to provide appropriate policy advice to the affected countries before the
crises occurred.
In response to this, in the late 1990s, the Fund and the Bank jointly
launched two initiatives to prevent the long- term likelihood of financial
crises. The first initiative, the Financial Sector Assessment Program
(FSAP), consists of in- depth assessments of key elements of member
countries* financial sectors. These elements include the structure of
financial markets, financial systems* response to changes in key variables
such as exchange rates, legal arrangements for crisis management, and the
quality of financial sector supervision. The second initiative, the
Reports on the Observance of Standards and Codes (ROSC), consists of
assessments of member countries* adherence to 12 standards 12 related to
transparency in government policy making and operations, 13 financial
sector regulation, and corporate sector practices (see appendix V). 14
Building on earlier efforts to assess transparency, in 1999, the IMF and
the Bank began conducting joint assessments of observance of standards
related to
12 According to the Fund, 11 of these 12 standards are internationally
accepted. The World Bank*s corporate insolvency and creditor rights
standard is still in draft form. 13 In addition to the ROSC, the Fund
launched an experimental program in 2000 to help prevent possible misuse
of Fund resources. This program assesses weaknesses in central banks*
ability to safeguard Fund resources through internal controls, accounting,
reporting, auditing systems, and legal structure. In 2002, the Executive
Board made assessments of central banks* safeguards a permanent policy.
See appendix VI for an update on progress in
this area. 14 In February and March 2003, the IMF and Bank published
reviews of the FSAP and ROSC initiatives. See International Monetary Fund
and World Bank, Financial Sector Assessment Program* Review, Lessons, and
Issues Going Forward (Washington, D. C.: 2003);
International Monetary Fund and World Bank, International Standards:
Strengthening Surveillance, Domestic Institutions, and International
Markets (Washington, D. C.: 2003); International Monetary Fund,
International Standards: Background Paper on Strengthening Surveillance,
Domestic Institutions, and International Markets (Washington, D. C.:
2003); World Bank, Reports on the Observance of Standards and Codes
Background Paper on Standards Review: Assessing Progress and Lessons
Learned from Bank- Led ROSC Modules (Washington, D. C.: 2003).
Page 17 GAO- 03- 734 International Financial Crises
financial sector regulation, covering areas such as banking supervision
and securities regulation. The Bank began performing assessments of
standards related to corporate sector practices, including private sector
accounting rules and corporate governance principles, in 2000. Some
transparency, financial sector, and corporate sector standards may be
assessed under the FSAP. Country participation in both initiatives is
voluntary. The Fund and the Bank initially considered making participation
in the ROSC assessments mandatory for member countries after determining
that the Fund*s Articles of Agreement could allow such a requirement. IMF
or World Bank staff lead FSAP and ROSC assessment teams, with
participation by experts from national central banks and supervisory
agencies and standard- setting bodies such as the Basel Committee and the
International Organization of Securities Commissions (IOSCO). Before
undertaking an FSAP or ROSC, the Fund and the Bank work with country
governments to choose areas on which to focus. During the assessment
process, FSAP and ROSC teams conduct at least one in- country visit,
allowing team members to work with government officials from the finance
ministry, the central bank, and regulatory bodies to collect information
for the assessment. 15 For example, an FSAP team in Russia analyzed
financial information for the largest banks and single largest corporation
to determine how changes in economic variables such as oil prices might
affect the banking system. The South Korean ROSC team interviewed
government officials at financial sector regulatory entities and private
sector representatives to determine how closely regulatory practices
conform to standards and to identify weaknesses that could put the
financial sector at risk.
The two initiatives provide information to assist four parties that play a
role in crisis prevention: the Fund, the World Bank, member country
governments, and private sector investors. In- depth information on member
countries* financial sectors and adherence to standards of good practice
is intended to help the IMF and the Bank fulfill their missions. Fund
staff identify countries* vulnerabilities and develop appropriate advice
to redress them; Bank staff identify long- term financial sector
development needs and formulate relevant lending and nonlending responses.
Member countries can use these assessments to help prioritize reform
agendas and win domestic support for difficult policy decisions that
15 Appendix VII contains a diagram of the FSAP and ROSC assessment
process.
Page 18 GAO- 03- 734 International Financial Crises
may make their financial sectors and institutions more resistant to
crisis. The Fund and the Bank often provide technical assistance to help
governments build capacity to implement reforms. The financial crises of
the 1990s also raised awareness of the private sector*s role in crisis
prevention. Thus, the Fund and the Bank expect the assessments to help
private sector participants make sounder investment decisions, thereby
reducing volatility in capital markets. Parties* Use of Assessments Is
Mixed, and Significant Challenges Remain
The use of FSAP and ROSC assessments in crisis prevention efforts is
mixed, and significant limitations remain. Fund staff use the assessments,
when available, as inputs for the policy advice they provide to member
country governments. However, the Fund lacks crucial information about
vulnerabilities to financial crisis because some major emerging market
countries have not participated in the assessments. World Bank staff*s use
of the assessments to inform development assistance priorities is also
affected by these gaps in participation and by borrower countries*
competing development needs. Many member country governments face
limitations in using assessments to make policy decisions because the
reforms recommended in the assessments are often difficult to implement.
Finally, some private sector participants find assessments of limited use
because they are untimely, outdated, and dense. The Fund Uses Assessments,
When Available, to Inform Policy Advice
The IMF uses FSAP and ROSC assessments, when available, as inputs for the
policy advice it provides member country governments. According to Fund
officials, these assessments highlight issues such as weak banking
supervision or high levels of debt held in foreign currency that could
make
countries vulnerable to crisis. The assessments also provide
recommendations to address these issues. The findings and recommendations
inform the discussions of policy issues that Fund officials have with
member country authorities during Article IV
consultations. 16 For example, when an FSAP was performed in Mexico,
Mexican authorities had begun replacing a system where the government
fully insured all bank deposits with one that covers deposits up to a
certain 16 In accordance with Article IV of the IMF*s Articles of
Agreement, the IMF holds regular
consultations, normally every year, with each of its members. These
consultations focus on the member*s exchange rate, fiscal, and monetary
policies; its balance of payments and external debt developments; the
influence of its policies on the country*s external accounts; the
international and regional implications of those policies; and on the
identification of potential vulnerabilities. (* IMF Surveillance: A
Factsheet,* http:// www. imf. org, downloaded 4/ 23/ 03).
Page 19 GAO- 03- 734 International Financial Crises
limit. The FSAP team was concerned because this reform was undertaken
before Mexico had developed a well- defined framework for closing
unprofitable banks. Without a clear framework for bank closures, the
introduction of limited deposit insurance could damage depositor
confidence in the banking system and precipitate a banking crisis.
According to Fund officials, the FSAP team and Mexican authorities
discussed the need to create such a framework, and a subsequent Article IV
mission reviewed the government*s progress in this area. In Poland, an
FSAP team discovered that Polish households and small businesses had high
levels of debt held in foreign currency. The team was concerned that a
depreciation of Polish currency could raise the cost of these loans and
cause widespread repayment difficulties, which could in turn lead to a
banking crisis. FSAP team members raised this issue with Polish central
bank authorities and followed up again during the next Article IV
consultation. In both countries, government officials followed the Fund*s
advice and implemented reforms. The Mexican government began
developing a framework for closing banks, and Poland*s central bank
established a team to monitor household and small business debt. Since
1999, FSAP assessments have been conducted in more than 40 member
countries and ROSC assessments in about 90 member countries. However, we
found that assessments have not been completed for some major emerging
market countries, limiting the Fund*s awareness of crisis vulnerabilities
in certain countries. Appendix VIII contains a record of country
participation in and publication of FSAP and ROSC assessments. Fund and
Bank staff encourage participation in FSAP and ROSC
assessments by countries whose economies have worldwide or regional
implications or have known vulnerabilities to a financial crisis, but
officials acknowledge that some governments have persistently resisted
their efforts. According to our analysis, between 1999 and 2003, 17 45
percent of 33 major emerging market countries participated in an FSAP. 18
However, the Fund has not performed FSAPs in important countries such as
China
and Thailand because their authorities have not agreed to participate.
These gaps in participation limit the Fund*s ability to develop policy
advice based on in- depth knowledge of their financial sectors.
17 Our analysis is based on Fund and Bank data through March 2003. 18
These 33 countries, a subset of the 87 we analyzed in the previous
section, represent the most significant emerging market participants in
the international capital markets, as identified by J. P. Morgan.
Page 20 GAO- 03- 734 International Financial Crises
According to the Fund, the Mexican and Asian financial crises were caused,
in part, by vulnerabilities in areas covered by the ROSCs. Our analysis
found gaps in participation in assessments of several key standards that
the Fund identifies as contributing factors to past crises (see figure 3).
For example, only one- third of major emerging market countries have
participated in assessments of their adherence to standards for
dissemination of economic and financial data. About half have participated
in the fiscal, monetary and financial policy formulation assessments and
the banking supervision assessment. In addition, Fund documents point to
limited progress in assessing adherence to the four World Bank- led
corporate sector standards 19 (accounting, auditing, corporate governance,
and insolvency regimes), which play a key role in the effective operation
of domestic and international financial systems. Less than one- third of
the 33 major emerging market countries have participated in one or more
assessments related to accounting and auditing.
Figure 3: Participation of 33 Major Emerging Markets in Key ROSC
Assessments, 1999* 2003
19 The Bank performs accounting and auditing assessments simultaneously
and publishes them as a single report. An agreed- upon standard for
insolvency regimes has not yet been finalized. The Bank*s draft standard
for insolvency regimes is under review by the Bank and the Fund.
Page 21 GAO- 03- 734 International Financial Crises
The Fund asserts that its delayed response in preventing or mitigating the
Mexican and Asian crises was partially caused by insufficient information
on these vulnerabilities. For example, according to the IMF, the Mexican
government*s publicly available data was outdated and incomplete in 1993*
94, which contributed to significant delays in responding to the country*s
excessive indebtedness. The Fund also was unaware of some Asian
countries* unsound corporate accounting practices, which contributed to
the Asian financial crisis. Continued participation gaps in these
assessments suggest that the Fund still lacks crucial information about
some countries* potential vulnerability to crisis.
The World Bank*s Use of Assessments in Formulating Country Development
Assistance Programs Is Mixed
The World Bank acknowledges the importance of FSAP and ROSC assessments in
formulating its financial sector development programs, but limited
participation in corporate sector assessments (described earlier) affects
the Bank*s ability to respond to weaknesses in borrower countries*
financial sectors. According to the Bank, country participation in
corporate sector assessments has been lower than in areas related to
transparency and financial sector regulation because the Bank has
experienced delays in finalizing standards and methodologies for
evaluating the corporate sector. For example, the methodologies for
performing assessments of the accounting and auditing standards were not
finalized until October 2000.
Bank officials acknowledge that even when assessments are available, Bank
staff do not always incorporate the issues raised as a key priority in
formulating its country development assistance programs. In justifying
their limited prioritization of FSAPs and ROSCs, Bank officials cited
competing development demands and timing issues. First, Bank officials
stated that they must balance borrower countries* financial sector reform
needs with other demands for development assistance. Most borrower country
governments have multiple concerns, and Bank staff may determine that its
resources will have more impact in areas other than financial sector
development. Second, Bank officials cited the scheduling of the FSAP and
ROSC assessments as a reason for their limited use since the timing of
many assessments does not coincide with the Bank*s preparation of Country
Assistance Strategies. 20 20 The Country Assistance Strategy describes the
Bank*s assessment of development
priorities in each borrower country and identifies the level and
composition of assistance to be provided based on this assessment. These
strategies are currently prepared every 3 years.
Page 22 GAO- 03- 734 International Financial Crises
Member Country Governments* Use of Assessments Is Limited
Although member country authorities sometimes use FSAP and ROSC
assessments to inform policy decisions, reforms recommended in the
assessments are often difficult to implement. Some member country
governments have faced obstacles to implementing reforms, including
political opposition, legal constraints, and lack of technical capacity.
For
example, IMF officials stated that political opposition has limited the
South Korean government*s progress in eliminating extrabudgetary funds, a
key recommendation of the fiscal transparency ROSC. Extrabudgetary funds
diminish transparency because they are exempt from rules that require
scrutiny and prioritization of expenditures for most of South Korea*s
budget. Fund officials cited Peru as a case where legal constraints
delayed reform efforts. The FSAP and banking supervision ROSC found that
protecting bank supervisors from the political pressures of the powerful
bankers* lobby would strengthen Peru*s banking supervision. According to
Fund officials, existing legislation precluded awarding supervisors
greater independence, and passage of new legislation was delayed. In
Russia, limited technical capacity interfered with the government*s
ability to
implement reform recommendations. For example, the FSAP team reviewed the
government*s proposal to stimulate competition in the banking sector by
introducing a deposit insurance system for household deposits. However,
Fund staff noted that Russia*s bank supervisory agency lacks the capacity
to implement a deposit insurance system for a large number of banks.
Private Sector Participants Find Assessments Difficult to Use
The Fund claims that private sector participants increasingly use the
results of FSAPs and ROSCs to inform investment decisions and risk
management. However, representatives of major international investment
firms and ratings agencies we interviewed stated that the reports were
untimely, outdated, and too dense to be useful. For example, several
respondents indicated that delays in publishing ROSC assessments reduced
their usefulness. Some private sector participants stated that ROSC
reports and FSAP summaries, known as Financial System Stability
Assessments (FSSAs), should be published within 6 months of performing the
assessments. However, our analysis of the 58 ROSC reports published for
major emerging market countries found that in one- third of the cases, 9
months or more elapsed between assessment and publication. Several private
sector participants we interviewed stated that outdated ROSC reports are
unreliable for decision making. The Fund acknowledges that assessments
must be current for private sector participants to use them. According to
Fund data, 13 countries have published an update of at least one ROSC
module. However, IMF officials estimate that, of the more than 40 FSAPs
performed to date, only 4 have been fully updated. Some private
Page 23 GAO- 03- 734 International Financial Crises
sector participants also stated that FSSA and ROSC reports are not clearly
written. Representatives of one multinational investment bank stated that
the assessments are written in a way that is difficult to understand,
which limits the reports* usefulness for investment decisions. While these
interviews were limited in number and may not be representative of all
private sector participants, they do provide an indication of the problems
these individuals may currently have in using FSAPs and ROSCs.
Fund and Bank outreach sessions and a 2002 Fund survey corroborated our
findings on private sector participants* difficulties in using ROSC
assessments. The Fund reports that private sector participants place high
priority on timely publication and frequent updates of ROSC assessments.
For example, several participants observed that ROSC reports for Argentina
had not been updated since their publication in 1999. Moreover,
respondents to the Fund*s survey commented that ROSC assessments should
state more clearly the deficiencies in a country*s adherence to a
standard. In a March 2003 review of the standards initiative, the Fund and
the Bank concluded that ROSC reports would be more useful if they stated
the main findings and their significance clearly and prioritized
recommendations more explicitly. Crisis Prevention Initiatives
Cannot Prevent all Crises The Fund and the World Bank acknowledge that
FSAP and ROSC assessments cannot prevent all crises because recommended
reforms may
require many years to be fully implemented and because crises can be
caused by factors outside the reforms* scope. For example, Argentina
participated in four ROSC assessments in 1999 to improve economic data
dissemination; banking supervision; and transparency in the formulation of
fiscal, monetary, and financial policies. According to senior IMF
officials, the Argentine government followed many of the recommendations
generated by these assessments, but their actions did not address
vulnerabilities related to weak fiscal policy and a fixed exchange rate
regime that contributed to Argentina*s 2001 crisis. Fund officials cite
Turkey as another example of a country that made considerable progress in
improving transparency and data provision based on reforms
recommended by the fiscal transparency and economic data dissemination
ROSC assessments of 2000* 2001. However, according to the Fund, these
reforms could not have prevented Turkey*s 2001 crisis, which originated
with declines in its exchange rate.
Page 24 GAO- 03- 734 International Financial Crises
The Fund*s Efforts to Better Resolve Future Financial Crises Face
Significant Challenges
Fund officials assert that the current process for renegotiating the terms
of member countries* loans with external private sector creditors is
lengthy and costly. In 2001, the Fund began considering the SDRM, an
international legal framework that would allow a majority of a country*s
external creditors to approve a restructuring of most private sector
loans. The Fund is also encouraging members to include CACs in bonds,
which would allow a majority of bondholders to renegotiate the terms of
that bond. Although some elements of both approaches are acceptable to the
private sector and governments, a number of political, legal, and
technical challenges stand in the way of implementing the SDRM; it seems
unlikely that these issues will
be resolved in the immediate future. While private sector officials expect
that many restructurings need only involve the private sector and the
debtor country, under some circumstances, voluntary debt restructurings
will not adequately resolve all financial crises. These officials stated
that, in those cases, the Fund should provide short- term loans to
eligible countries to help fill their external financing gaps. However,
concerns have been raised by some financial experts and government
representatives that such Fund loans have the potential to increase the
probability of future crises. In response to these concerns, the Fund has
clarified and strengthened its
policy of lending into crisis situations. Efforts to Resolve Future
Sovereign Debt Crises Face Significant Challenges
According to the Fund, countries facing severe liquidity problems often go
to extraordinary lengths to avoid renegotiating or restructuring the terms
of their loans. They do so because, in the past, restructuring damaged the
economy and the banking system of participating countries. In some cases,
even when a voluntary restructuring process is initiated, individual
creditors may hold out for the best possible terms or sue in an attempt
for better terms. Additionally, countries believe that creditors also may
be unwilling to make future loans if they default on their existing debt.
The SDRM approach is an attempt to create a more orderly, predictable, and
comprehensive restructuring process and to lower the costs of
restructuring for both the debtor and creditors. The approach sought to
reduce the duration of the restructuring process from years to months and
to provide incentives to restructure debt before default to better protect
debtor and creditor interests. In the case of debtors, the Fund maintains
that an orderly restructuring process could reduce the likelihood of a
reduction in future capital flows. For creditors, it could provide more
favorable repayment terms from the restructured debt.
Page 25 GAO- 03- 734 International Financial Crises
The SDRM is a proposed international legal framework that would allow a
member country to declare its debt unsustainable and invoke a process to
restructure most of its external private sector loans. 21 A specified
majority
of the country*s external creditors would vote to approve the terms for
restructuring, which would bind all eligible private sector creditors. The
framework is designed to increase the incentives for the Fund*s member
countries and their creditors to reach a rapid and collaborative agreement
on restructuring unsustainable debt.
A number of political, legal, and technical challenges stand in the way of
implementing the SDRM, and it seems unlikely that these issues will be
resolved in the immediate future. According to the Fund, successful
implementation of SDRM will require overcoming certain political
constraints. The SDRM could be put into practice either by countries
adopting a new international treaty or by amending the Fund*s Articles of
Agreement. Both options would be difficult to implement since a number of
countries have indicated opposition to the SDRM. The draft framework
recommended that the SDRM be created through an amendment to the Fund*s
Articles of Agreement because the SDRM is closely related to the
role already assigned to the Fund under the Articles in the resolution of
its members* external financial obligations. However, the Fund
acknowledged that, given the opposition of some countries, changing the
Articles could be
difficult to achieve since it requires acceptance by three- fifths of the
members, having 85 percent of the total voting representation. The United
States, for example, could unilaterally veto any proposed amendment to the
Fund*s Articles given its 17 percent voting representation.
21 The SDRM proposal has undergone several revisions. Our report discusses
the proposal presented to the International Monetary and Financial
Committee in April 2003. See http:// www. imf. org/ external/ np/ omd/
2003/ 040803. htm.
Page 26 GAO- 03- 734 International Financial Crises
A key legal challenge to the implementation of SDRM is the need for most
countries to change their domestic laws to conform to the requirements of
any new Fund articles. Before a member country can vote to accept an
amendment, it must take all the necessary steps needed under its own
domestic law to ensure that the amendment will be given full force and
effect under its domestic law. However, some Fund members have raised
concerns over whether the domestic legal systems of some member countries
could accommodate a new legal framework that applied to preexisting
claims. The proposed SDRM approach also faces technical challenges. For
example, the proposed framework does not specify how
the claims of official bilateral creditors and some guaranteed domestic
debts would be treated. The Fund is consulting with the Paris Club on how
the Club*s practices may be modified to better facilitate coordination
between official bilateral and the private creditors in a debt
restructuring process. 22 The Fund Has Encouraged
the Adoption of Collective Action Clauses
CACs are terms in individual bonds that permit a specified majority of
sovereign bondholders to agree on a debt restructuring that would bind all
holders of that bond. In June 2002, the Fund*s Executive Board endorsed
the use of certain CAC provisions in new bonds and agreed to encourage
member countries to incorporate CACs into their sovereign bonds in future
restructurings. Inclusion of these clauses into new bonds would be
voluntary.
22 The Paris Club is an informal group of official creditors whose role is
to find solutions to the payment difficulties experienced by debtor
countries. Paris Club creditors agree to provide a country with debt
relief through a postponement and, in the case of concessional
rescheduling, a reduction in debt service obligations.
Page 27 GAO- 03- 734 International Financial Crises
The Fund views CACs and SDRM as complementary instruments in resolving
future financial crises. According to the Fund, the existence of CACs in
certain bond agreements inspired the development of the SDRM framework.
Although the Fund has not created its own CAC framework, it has endorsed
the use of two key features from a G- 10 Working Group Report 23 and an
Industry Associations draft proposal. 24 These features include the
following:
Majority restructuring provision 25 enables a qualified majority of
bondholders to bind all holders of a particular bond to the financial
terms of a restructuring, both before and after a default. Although
majority restructuring provisions have generally been included in bonds
governed by the laws of the United Kingdom or Japan, they have not been
included in bonds governed by the laws of the United States or Germany.
Majority enforcement provisions prevent a minority of creditors from
pursuing disruptive legal action after a default and before reaching a
restructuring agreement. Many international sovereign bonds governed by
both U. S. and English law contain these provisions. Specifically, the
Fund supports the requirement that (1) an affirmative vote of a minimum
percentage of bondholders is necessary to approve claims following a
default and (2) a specified majority of bondholders can reverse an
approval of a claim that has already occurred.
In early 2003, Mexico became the first emerging market country to issue a
public, SEC- registered global bond with CACs under New York law. Previous
issues under New York law by Lebanon, Qatar, and Egypt had
been placed privately to institutional investors and included only a
limited 23 The G- 10 Working Group on Contractual Clauses (the Working
Group) was formed in June 2002 at the behest of ministers and governors.
The mandate of the Working Group was to consider how sovereign debt
contracts could be modified to resolve debt crises in a more orderly
fashion.
24 The proposal was put forward to Fund members for consideration by the
Institute of International Finance and six other financial industry trade
associations. The six financial industry associations are the Emerging
Market Traders Association (EMTA), the
International Primary Market Association (IPMA), the Bond Market
Association (BMA), the Securities Industry Association (SIA), the
International Securities Market Association (ISMA), and the Emerging
Markets Creditors Association (EMCA).
25 These provisions are referred to in the G- 10 Working Group Report as
*majority amendment provisions.*
Page 28 GAO- 03- 734 International Financial Crises
range of CACs. For example, the bonds issued by Egypt and Qatar included a
very limited form of majority enforcement provisions, while Lebanese bonds
did not contain them at all. Since Mexico*s successful
issue, Brazil, South Africa, and Korea have issued bonds with CACs.
Uruguay included CACs in the bonds resulting from its debt exchange. The
details of the Brazil, South Africa, Korea, and Uruguay bond provisions
were not available at the time we conducted our review.
Some countries criticize CACs because they would only apply to new bond
offerings and not existing bonds. Accordingly, during a restructuring of a
country*s bond obligations, not all creditors would be bound by the CAC
provisions. Borrowing countries also contend that inclusion of CACs in
bond offerings could suggest to creditors that countries anticipate having
difficulty repaying their loans. In response, creditors may charge a
higher interest rate. However, a May 2000 academic study 26 compared
interest rates on bonds issued in the United States (where CACs are not
used) with the United Kingdom (where CACs are used) and found that CACs do
not contribute to higher rates in United Kingdom bonds.
While Private Sector Officials Prefer Voluntary Debt Restructuring, They
Expect Continued Fund Loans in Exceptional Circumstances
To date, officials from the private sector, including lenders, have
expressed preference for continuing the current voluntary process, which
only involves the private sector and borrowing countries, in the efforts
to restructure sovereign debts. Many private sector officials we
interviewed oppose the proposed SDRM approach and the Fund*s attempts to
integrate CACs into new bond issues, partly because they would interfere
with the normal bargaining process. They maintain that a voluntary
approach to the restructurings that took place from 1998 to 2001 in
Ecuador, Pakistan, Russia, and Ukraine were successful. 27 Private sector
officials assert that these and other experiences have worked well enough,
and that a
substantive change in current market practice is unnecessary. In contrast
to the Fund*s assertion that new approaches are needed to make
restructurings shorter and less expensive, private sector officials note
that
26 See Eichengreen, Barry and Ashoka Mody. *Would Collective Action
Clauses Raise Borrowing Costs?* Working paper, Center for International
and Development Economics Research, Berkeley, CA, 2000.
27 The restructurings in Russia involved domestic debt and Soviet- era
foreign debt owed to commercial banks. In 1998, Pakistan froze withdrawals
in foreign currency from all nonresident foreign currency deposits and
subsequently reached restructuring agreements with these nonresident
investors.
Page 29 GAO- 03- 734 International Financial Crises
most recent voluntary restructurings successfully concluded in 1 year or
less and that creditor holdouts or litigation did not significantly delay
the restructurings. 28 While private sector officials expect that many
restructurings need only
involve the private sector and the debtor country, under some
circumstances, voluntary debt restructurings will not adequately resolve
all financial crises. In those cases, they said the Fund should provide
loans to
eligible countries to help fill their external financing gaps. Such loans
would assist the restructuring process and facilitate efforts at
implementing necessary reforms. However, large Fund loans, such as those
given during the Asian financial crises, have received substantial
criticism
from financial experts and government representatives, including U. S.
government officials. One concern is that the possibility of receiving
substantial financial assistance provides an incentive for debtor
countries to adopt unsustainable economic policies to forestall needed
reform. Another concern is that these large loans may encourage private
sector creditors to continue providing large capital flows to countries
with unsustainable economic policies because these otherwise risky
investments have the potential of being *bailed out* by future Fund loans.
This condition is referred to as *moral hazard.* According to these
critics, efforts to help resolve existing financial crises through large
Fund loans may increase the probability of future crises due to these two
concerns. The Fund has advocated the SDRM framework and CACs to replace
the current voluntary approach, partially in response to concerns over the
potential adverse affects of its lending.
The Fund Has Clarified and Strengthened its Lending Practices to Address
Concerns Over Exceptional Lending
To reduce the risk that Fund loans would increase the probability of
future financial crises, the Fund clarified and strengthened its policy of
lending in crisis situations. The Fund has clarified elements of its
Lending into Arrears Policy and strengthened its criteria for requesting
large short- term loans under the Supplemental Reserve Facility (SRF).
Since 1997, nine
countries have received loans under the two mechanisms. The Fund*s Lending
into Arrears Policy permits the IMF to provide resources to countries that
are unable to repay their external creditors and are thus considered in
default. Conceived in the late 1980s and amended in 28 For some countries,
the negotiating process entailed several years and more than one
restructuring agreement.
Page 30 GAO- 03- 734 International Financial Crises
the late 1990s, the policy is designed to protect the value of creditor
assets while providing creditors with incentives to enter rapidly into
restructuring negotiations with countries. The Lending into Arrears Policy
increases the likelihood that a country*s private sector lenders would
agree to reduce the value of their loans because Fund resources reduce
short- term fiscal pressures experienced by the country while in default.
A country is eligible for Fund resources while in default if the Fund
determines that the debt burden is unsustainable, and the country is
making satisfactory progress in implementing reforms. Additionally, the
country must have demonstrated a
good faith effort to reach a restructuring agreement with creditors to
restore its ability to repay its debt. In 2002, the Fund clarified the
criteria to be used to determine whether the debtor country is making a
good faith effort. For example, the Fund would consider how quickly the
debtor engaged in negotiations with its creditors after it defaulted. To
date, the
Fund has lent into arrears on international bonds on four occasions*
Ukraine, Ecuador, Moldova, and Argentina.
Introduced in 1997, the SRF provides large short- term loans to members
experiencing exceptional balance of payments difficulties prior to a
default. The interest rates on these loans are much higher than standard
Fund loans. The increased cost of these loans is expected to reduce the
probability that countries consider Fund resources a viable means for
underwriting unsustainable economic policies. Higher loan terms also
increase incentives for early repayment and compensate for additional
repayment risks to the Fund. Countries are expected to repay SRF loans
within 2 to 2 1/2 years, but they may request extensions of up to 6
months. All SRF loans carry a substantial surcharge of 3* 5 percentage
points. In 2003, the Fund strengthened its criteria for providing large
short- term loans under the SRF. For example, countries requesting SRF
loans must provide a more extensive justification for their repayment
difficulties. Additionally, the member has to demonstrate good prospects
of regaining access to private capital markets within the time period that
Fund resources are
outstanding to minimize long- term reliance on Fund resources. To date,
the Fund has provided SRF loans on nine occasions to six countries,
including Korea, Russia, Brazil, Turkey, Argentina, and Uruguay (see
appendix IX).
Conclusion In accordance with its goal of strengthening the international
financial system, the Fund has undertaken a number of reforms to better
anticipate,
prevent, and resolve sovereign financial crises. The Fund*s new
vulnerability assessment process is more comprehensive than its previous
crisis anticipation efforts, but it is too soon to judge its
effectiveness. The
Page 31 GAO- 03- 734 International Financial Crises
Fund*s proposed approaches to better resolve financial crises have met
considerable resistance, and it is unclear whether they will ultimately be
adopted. The Fund and the Bank have made progress in their crisis
prevention efforts by performing assessments of member countries*
financial sectors and adherence to standards. However, the effectiveness
of these crisis prevention efforts is hindered by (1) private sector
participants* limited use of published assessments, which they find
untimely, outdated, and too dense to be useful and (2) gaps in crucial
information about crisis
vulnerabilities in some important emerging market countries due to
voluntary participation in the assessments. These limitations prevent
multilateral institutions, national policy makers, and private sector
participants from making sound decisions, thus reducing the likelihood
that these reforms will help prevent crises.
Recommendations for Executive Action
To help strengthen the Fund*s crisis prevention initiatives we recommend
that the Secretary of the Treasury instruct the U. S. Executive Director
of the Fund to work with other Executive Board members to encourage the
Fund to
improve the timeliness of publication of Financial System Stability
Assessments and Reports on the Observance of Standards and Codes; expand
the coverage, frequency, and publication of reports on member
countries* progress on implementing assessment recommendations; improve
the assessment reports* readability, for example, by creating a
standardized format in which to present executive summaries and key
findings; and
pursue strategies for increasing participation in the Financial Sector
Assessment Program and all modules of the Reports on the Observance of
Standards and Codes, including the possibility of making participation
mandatory for all members of the IMF. Agency Comments and
Our Evaluation We received written comments on this report from the
Department of the
Treasury, the International Monetary Fund, and the World Bank. These
comments and GAO*s evaluation of them are reprinted in appendixes X, XI,
and XII. The organizations also separately provided technical comments
Page 32 GAO- 03- 734 International Financial Crises
that GAO discussed with relevant officials and included in the text of the
report where appropriate.
Treasury agreed with the report*s recommendations. Treasury recognized
that some important countries have not volunteered to participate in the
FSAP and ROSC and that there should be a shorter turnaround between the
completion of an assessment and its public release. Treasury also pointed
out that the acceptance of collective action clauses in some recent bond
offerings serves as an important signal to investors that official
financing is limited and that they cannot expect to be protected from
risks.
The IMF broadly agreed with the report*s recommendations. However, the IMF
stated that we mischaracterized the role of the WEO forecasts and EWS
models in IMF crisis anticipation efforts by saying that they have a
greater importance than is warranted. We disagree with this depiction. Our
assessment examined all six components of the IMF*s vulnerability
assessment framework, including the WEO and the EWS. As the only mature
and quantifiable elements of the framework, our analysis focused more
heavily on the track records of the WEO and EWS. The IMF also stated that
its responsibility to maintain financial stability could make its
predictions less accurate so as not to contribute to a crisis. The IMF*s
comment not only validates our finding on the WEO*s weakness but also
raises questions regarding the purpose and credibility of the WEO
forecasts.
The World Bank generally agreed with the report*s recommendations.
However, the Bank expressed concern with the report*s suggestion that
consideration be given to making participation in the FSAP and ROSC
assessments mandatory. While we are not suggesting that the assessments
should be made mandatory, the voluntary nature of the FSAP and ROSC has
posed an obstacle to full participation by important emerging market
countries.
We are sending copies of this report to the Secretary of the Treasury, the
International Monetary Fund, the World Bank, and interested congressional
committees. We also will make copies available to other interested parties
upon request. In addition, the report will be available at no charge on
the GAO Web site at http:// www. gao. gov.
Page 33 GAO- 03- 734 International Financial Crises
If you or your staff have any questions regarding this report, please call
me at (202) 512- 8979. Other GAO contacts and staff are acknowledged in
appendix XIII.
Joseph A. Christoff, Director International Affairs and Trade
Page 34 GAO- 03- 734 International Financial Crises
Appendix I
Appendi xes Objectives, Scope, and Methodology Appendi x I
The Chairman of the House Financial Services and the Vice Chairman of the
Joint Economic Committee requested that we assess (1) the International
Monetary Fund*s (IMF*s or the Fund*s) framework for anticipating financial
crises, (2) the status of key IMF initiatives to prevent financial crises,
and (3) new IMF proposals to resolve future financial crises. They
requested that our review cover the period after the Mexican financial
crisis of 1994* 95.
To assess IMF*s framework for anticipating financial crises, we examined
prior and new IMF mechanisms for anticipating crises. Our analysis focused
on the World Economic Outlook (WEO) forecasts and the IMF*s Early Warning
System (EWS) models that were the IMF*s primary forecasting tools prior to
the implementation of the new vulnerability assessment framework in May
2001. Data on the performance of the other four components of the
framework were not made available to us because these elements are
considered part of the staff level deliberative process, and are not
provided to the Executive Board. We obtained near- term data
from the WEO forecasts, including real gross domestic product (GDP) growth
rate and current account balance for 87 emerging market countries for the
period 1990* 2001 (see appendix III). We focused on the 81 middleincome
countries and an additional 6 low- income countries listed by J. P. Morgan
as emerging markets. To evaluate the WEO and program forecasts,
we used standard econometric techniques based on methods commonly found in
the forecasting literature. The formal methodology of our forecast
evaluations was based on several expert publications, the replication of
our summary statistics with another author*s results, and discussions with
a forecasting expert. To describe the performance of the IMF*s EWS models
in anticipating crises, we reviewed and summarized the results of IMF
evaluations. We interviewed IMF staff, including country desk economists
and staff from several departments, to discuss the IMF*s new framework for
vulnerability assessment, the EWS models, and the WEO methodology. We also
interviewed 23 major private sector emerging market participants
to discuss whether and how they use IMF forecasts in their investment
decisions.
To assess the status of key IMF initiatives to prevent financial crises,
we reviewed Fund and World Bank documents published between 1999 and 2003
on the creation of the Financial Sector Assessment Program (FSAP) and
Reports on the Observance of Standards and Codes (ROSC) and evaluations of
progress in implementing these reforms. We interviewed senior Fund
officials, including staff from the Monetary Affairs and Exchange
Department, the Policy Development and Review Department,
Appendix I Objectives, Scope, and Methodology
Page 35 GAO- 03- 734 International Financial Crises
and the Fiscal Affairs Department. We also met with senior advisers at the
World Bank (the Bank) who oversee the Bank*s participation in the FSAP and
ROSC initiatives. To gain a better understanding of how the Fund uses FSAP
and ROSC assessments to inform the policy advice it provides to member
country governments and challenges it faces in using these assessments, we
interviewed Fund officials in nine area departments (Argentina, Brazil,
Korea, Mexico, Peru, Poland, Russia, Turkey, and Uruguay). We also spoke
with Fund officials in these area departments about member country
governments* use of the assessments in shaping their reform agendas and
the obstacles that member country authorities encounter in implementing
the reforms recommended in the assessments. To assess the extent to which
emerging market countries have participated in and authorized publication
of FSAP and ROSC assessments, we examined Fund and World Bank data on
country participation in the FSAP and 12 ROSC modules and publication of
the resulting reports between May 1999 and March 2003. To obtain views on
the private sector*s use of Fund and World Bank assessments, we conducted
structured interviews with 13 representatives of private sector firms,
including ratings agencies, investment banks, and pension funds. We
focused on 33 countries (a subset of the 87 we analyzed in the previous
section) identified as the major emerging market countries by J. P.
Morgan. 1 To describe new proposals to resolve future financial crises and
their
potential challenges, we obtained the most current Fund documentation for
the two key proposals, the Sovereign Debt Restructuring Mechanism (SDRM)
and Collective Action Clauses (CACs). We examined the purpose, goals,
requirements, and status of implementation. To obtain views on the private
sector*s understanding of the components of the new proposals and
potential implementation challenges, we conducted structured interviews
with 22 representatives of private sector firms, including ratings
agencies, investment banks, and pension funds. We also met with
government, private sector emerging market participants, and
nongovernmental officials at several conferences. We also interviewed
Department of the Treasury officials and experts in international finance
and law. The IMF did not meet with us on these proposals because they were
still under negotiation at the time of our review.
1 These 33 countries represent the most significant emerging market
participants in the international capital markets, as identified by J. P.
Morgan.
Appendix I Objectives, Scope, and Methodology
Page 36 GAO- 03- 734 International Financial Crises
We conducted fieldwork in Washington, D. C., and in New York. We performed
our work from May 2002 to May 2003 in accordance with generally accepted
government auditing standards.
Page 37 GAO- 03- 734 International Financial Crises
Appendix II Private Sector Structured Interview Participants Appendi x I I
Source: GAO.
Perspectives obtained on IMF efforts in crisis Company Anticipation
Prevention Resolution
Citicorp x xx Cleary Gottlieb & Hamilton x x Credit Suisse x xx Darby*s
Overseas Investments x x x Elliot Associates x x Emerging Market Traders
Association x x eStandards Forum x x x Eurasia Group x x x Fitch Ratings x
xx HBK x x HSBC x xx International Primary Markets Association x x
Institute for International Finance x x Japan Center for International
Finance x x J. P. Morgan x x Lehman Brothers x x x Mass Mutual x x Moody*s
Investor Service x x x Securities Industry Association x x x Societe
Generale x x x Standard & Poor*s x x x Straus & Boies x x Wilshire
Associates x x
Page 38 GAO- 03- 734 International Financial Crises
Appendix III Assessment of IMF Forecasting Appendi x I I I
Congress expressed concerns regarding the accuracy of the International
Monetary Fund*s (IMF*s) growth rate projections and asked us to examine
them. In response, we analyzed the quality of the forecasts produced by
the World Economic Outlook (WEO), the Fund*s primary forecasting tool.
Using econometric techniques common to forecast evaluation studies, we
examined the basic measures of forecast accuracy, bias, and efficiency.
This assessment supplements our finding on WEO*s efforts to anticipate
crises reported earlier. We found that WEO forecasts of gross domestic
product (GDP) growth and inflation perform somewhat better than an
assumption that next year*s rate will be the same as this year*s (called a
*naive* forecast). However, there is evidence of an optimistic bias in the
forecasts of GDP growth and inflation. In addition, we found that the
naive forecast of the current account generally performed better than the
WEO forecast. Moreover, WEO forecasts for the major industrialized
countries were superior to emerging market forecasts, and forecasts for
emerging market countries that had been on an IMF program were better than
for those countries that were not. The shortcomings we observed in WEO
forecasting are similar to those encountered by other private sector and
official forecasters.
GAO Analysis Used Standard Econometric Techniques for Forecast Evaluation
We evaluated IMF forecasts for 87 emerging market countries. Our analysis
focused on WEO forecasts of GDP, inflation, and the current account. Our
measures of forecast quality relied on generally accepted econometric
measures of accuracy, bias, and efficiency.
Overall Approach To evaluate the quality of IMF forecasts, we analyzed the
near- term and year- ahead WEO forecasts for 87 middle- income emerging
market
countries. 1 Appendix IV lists the 87 emerging market countries used in
the analysis. Our analysis focused on three WEO forecast variables: (1)
the growth rate of real GDP, (2) consumer price index (CPI) inflation
(average
over period), and (3) current account balance in billions of U. S.
dollars. Our evaluation methodology is based on standard econometric
techniques commonly found in the forecast literature, including the work
of
1 Our analysis focused on 81 middle- income countries and an additional 6
low- income countries listed by J. P. Morgan as emerging markets.
Appendix III Assessment of IMF Forecasting
Page 39 GAO- 03- 734 International Financial Crises
forecasting experts such as Stekler (1991), Artis (1996), and Loungani
(2001). 2 We also compared the quality of WEO forecasts of emerging market
countries with WEO forecasts of the G- 7 countries and we compared
borrowers of IMF resources with those that were not. 3 Our comparison with
the G- 7 countries 4 allowed us to informally assess whether income level
or data quality mattered in forecast quality. Our analysis of forecasts
for program countries permitted us to assess whether WEO forecasts differ
from forecasts contained within program documents, which are produced
under conditions of greater staff scrutiny. 5 We also reviewed a number of
forecast evaluation studies to see how our results compared to previous
reviews and to contrast IMF forecast quality with other forecasting
efforts.
Basic Definitions and Methods of Evaluation
Our analysis focused on the WEO*s near- term and year- ahead forecasts.
Near- term forecasts originate from the May WEO for each year, and they
project for the remainder of the existing year (approximately 6 months
ahead). The year- ahead forecasts come from the October WEO of the
preceding year. Thus, a near- term forecast for 2000 would come from the
May 2000 WEO, and a year- ahead forecast for 2000 would come from the
October 1999 WEO. We compared these WEO forecasts to the *first settled
estimate,* which comes from the October WEO of the following year for
which the forecast is made. Thus, we compared both the near- term and 2
Herman Stekler, *Macroeconomic Forecast Evaluation Techniques,*
International Journal of Forecasting 7 (1991): 375- 384; Michael J. Artis,
*How Accurate are the IMF*s Short- Term Forecasts? Another Examination of
the World Economic Outlook,* WP/ 96/ 89 (Washington, D. C.: International
Monetary Fund [IMF], August 1996); Prakash Loungani, *How Accurate are
Private Sector Forecasts? Cross- Country Evidence from Consensus Forecasts
of Output Growth,* International Journal of Forecasting 17 (2001): 419-
432.
3 We include the following General Resources Account- funded IMF programs:
Stand- By Arrangements (SBA), Extended Fund Facility (EFF), Systemic
Transformation Facility (STF), Compensatory Contingency Finance Facility
(CCFF). Under the CCFF, countries can borrow resources on a stand- alone
basis, i. e., outside of an IMF program. Macedonia and South Africa
borrowed funds under the CCFF.
4 The G- 7 include Canada, France, Germany, Italy, Japan, United Kingdom,
and United States. 5 Of the 87 emerging market countries analyzed, 57 were
on an IMF program for some part of the 1990* 2001 period.
Appendix III Assessment of IMF Forecasting
Page 40 GAO- 03- 734 International Financial Crises
year- ahead forecasts for 2000 with the *first settled estimate* from the
October 2001 WEO. 6 Most of the econometric tools we used to assess the
quality of WEO
forecasts analyze the errors deriving from the forecasts. Our econometric
tools examined these errors for certain qualities and patterns. We defined
the forecast error, as the difference between the forecasted, and
realized, value of an indicator. Hence, we have
Our examination of the errors in the Fund*s forecasts focused on three
measures of *goodness*: accuracy, bias, and efficiency. We performed these
tests separately for the 87 countries over the 11- year forecasting
period.
Accuracy The credibility of a forecast is established by its accuracy. The
ultimate test of any forecast is whether it can predict future events
accurately. Accuracy assesses whether forecasts tend, by some standard, to
be close to actual
outcomes. Although there is no objective standard of accuracy, comparisons
to alternative forecasting methods, such as a naive model that uses
historical trend data, is one way to judge relative accuracy. The accuracy
measure we used is Theil*s U- statistic (U 1 ), based on the naive model
that assumes this year*s growth rate will be the same as last year*s. 7
The Theil U- statistic is based on an examination of the forecast*s root
mean square error (RMSE). To compute RMSE, the forecast errors are squared
and averaged over the sample to get the mean square error (MSE). RMSE is
the square root of MSE.
6 A persistent issue in the forecasting literature concerns which *actual*
value to use to evaluate the accuracy of the forecasts: the first
available estimate (available in May of the following year), the first
settled estimate (available in October of the following year), or a later
revision. We have taken the middle ground, as suggested by Loungani
(2001), on the basis that forecasters are not attempting to predict later
revisions, which incorporate information such as revisions of weights and
changes in methods of construction that forecasters would not have been
aware of at the time of the forecast.
, t e , t f
, t a
7 That is, the naive model assumes there is no change in the growth rate
between t- 1 and t.
Appendix III Assessment of IMF Forecasting
Page 41 GAO- 03- 734 International Financial Crises
The Theil U- statistic allows us to scale the RMSE by the variability of
the underlying data. The standard Theil statistic, commonly denoted as ,
compares the RMSE of forecast series to the RMSE of the actual series,
where = - . A statistic greater than one denotes that the naive model
performed better than the model being tested.
Bias Bias determines whether forecast errors in one direction tend to be
larger and/ or more numerous than errors in the opposite direction.
Forecast errors can be divided into two parts. One part is the *random
error,* which varies unsystematically, or randomly, from one forecast to
the next. The other part is the *bias error,* which remains constant for
any particular forecasting procedure. Bias happens when factors other than
random events influence the forecast results, resulting in an upward or
downward tendency. An unbiased forecast means that forecast errors are
approximately zero on average over time. However, an unbiased forecast
does not guarantee that a forecast will be accurate enough to be useful if
the errors are large.
To assess the extent of bias in the forecasts, we regress the forecast
error on a constant term and then carry out a t- test for the coefficient.
For a time series of forecasts (1990* 2001 in our analysis), we have the
set of forecast errors . The regression test for bias involves estimating
the intercept coefficient, , for the simple regression
1 U t a t a 1 t a 1 U
{ } N e e e , , , 2 1 K
0
Appendix III Assessment of IMF Forecasting
Page 42 GAO- 03- 734 International Financial Crises
This reflects a partition of the forecast error into an estimate of the
systematic component and a random component ( ). A t- test with degrees of
freedom is then performed to test versus the twosided
alternative We perform a t- test to determine whether the average bias is
significantly different from zero. If the p- value is less than .05, we
reject the hypothesis that there is no bias at the 95 percent level of
significance. This means that there is less than a 5 percent chance that
we are making a false rejection, that is saying the forecast is biased
when it is not. A determination that a forecast is unbiased is a
necessary,
although not sufficient, condition for concluding that a forecast is
efficient. Efficiency Efficiency examines whether a forecast has taken
into account all available
information. Establishing that a forecast is efficient means that no other
model or readily available information would be able to improve the
forecast, and there is no way to predict the direction or size of the
errors. A test of efficiency makes use of the simple linear model where we
regress the actual outcome on the forecast
If the forecast is efficient in predicting the actual outcome, then the
intercept, , should be equal to 0 and the slope, , should be equal to 1. 8
Using the regression model defined above, we perform a joint hypothesis
test to check whether these conditions hold simultaneously.
t 1 N 0 : 0 0 = H
. 0 : 0 A H
8 A zero intercept implies that the errors are randomly distributed; they
vary unsystematically (unbiased). However, a forecast can be unbiased and
not efficient. A slope of 1 implies a straight line through the zero
intercept, denoting efficiency. A slope of 1 indicates that the forecast
and actual value essentially coincide.
0 1
Appendix III Assessment of IMF Forecasting
Page 43 GAO- 03- 734 International Financial Crises
The algebraically simplified version of the test statistic for this
hypothesis is shown below. 9
The reference distribution for this statistic is an F- distribution with 2
and degrees of freedom. If the p- value for this statistic is less than
.05, then we reject the hypothesis at the 95 percent level of significance
that both the intercept is zero and the slope is one. This means that
there is only a 5 percent chance that we are making a false rejection,
that is saying the forecast is not efficient when it is.
WEO Forecasts Demonstrate Some Accuracy, but Also Optimistic Bias
Our analysis of the WEO forecasts for 87 emerging market countries shows
that WEO forecasts perform somewhat better than a naive model for GDP
growth and inflation, but not for current account (see table 2).
Table 2: Forecast Quality for 87 Emerging Market Countries, 1990* 2001
Source: GAO analysis of IMF data. a The percentage of countries in which
the WEO forecast does a better job than the naive forecast
(based on the Theil statistic). b The percentage of countries in which
there is statistically significant bias (at the 5 percent level).
c When bias occurs in 15 percent or more countries and the forecast errors
tend to vary predominantly in one direction (more than 70 percent), then
we indicate the direction of this bias.
9 RSS is the residual sum of squares. 2 N
Accuracy a (percent) Bias b
(percent) Direction of bias c Efficiency d (percent) Forecast variable
Yearahead Nearterm
Yearahead Nearterm Yearahead Nearterm Yearahead Nearterm
GDP growth rate 62 74 21 15 Upward Upward 76 85 Inflation 55 70 21 13
Downward Downward e 72 67 Current account (billions of U. S. $) 24 56 8 8
Upward e Indeterminate 67 79
Appendix III Assessment of IMF Forecasting
Page 44 GAO- 03- 734 International Financial Crises
d Percentage of countries where we fail to reject the joint hypothesis
that the intercept is zero and the slope is one; i. e., the evidence is
not strong enough to reject the assumption of efficiency. e Italics
indicate that there are too few countries to consider bias a systematic
trend.
We found the year- ahead WEO forecast does a better job than the naive
forecast in more than 60 percent of the countries with GDP, in more than
half the countries with inflation, and in about one- quarter of the
countries with the current account. However, even for GDP, nearly 40
percent of the country forecasts were no better than an assumption that
next year*s value is the same as this year*s. For all three variables, the
shorter the forecast period, the more accurate the forecast. When the
forecast time horizon shortens from 1 year to 6 months, the percentage of
cases in which the WEO does a better job forecasting than the naive model
increases for all three variables, exceeding 50 percent for the current
account.
WEO forecasts for GDP and inflation demonstrated bias in about 20 percent
of the country cases. The direction of the bias was upward for GDP and
downward for inflation, indicating an optimistic tendency within the WEO
forecasting process. 10 Although the bias was upward for the current
account, also consistent with optimism, it occurred in only 8 percent of
country forecasts. 11 For all three variables, we could not reject the
hypothesis that WEO forecasts were efficient for at least two- thirds of
the country*s forecasts. However, in about one- fourth of the country
cases, the WEO forecast could have been improved through the use of a
different model or the addition of new information.
Forecasts for Major Industrialized Countries Are Better than Emerging
Market Forecasts and Program Forecasts Are Better than Nonprogram
Forecasts
WEO forecasts of the most developed countries are superior to its
forecasts of emerging market countries when compared to the naive model
forecasts. The improved forecast quality is likely due to better available
data and greater stability of the wealthiest economies. Similarly, WEO
forecasts for countries that borrow from the IMF are superior to those
that do not. The increased scrutiny of borrowing countries by IMF staff
likely
contributes to the improved forecasts. 10 Our finding of an optimistic
(upward) bias for GDP forecasts is consistent with, and helps explain, our
finding in the main report of the WEO*s difficulty in forecasting
recessions. 11 An optimistic bias in the current account means that the
forecast was for a smaller current deficit or a larger surplus than
occurred.
Appendix III Assessment of IMF Forecasting
Page 45 GAO- 03- 734 International Financial Crises
WEO Forecasts for Major Industrialized Countries Are Better than Those for
Emerging Markets
WEO forecasts of GDP and inflation for the G- 7 countries are considerably
better than its forecasts for the emerging market countries when compared
to the naive model forecasts. (see table 3).
Table 3: Forecast Quality for G- 7 Countries, 1990- 2001
Legend N/ A = not applicable Source: GAO analysis of IMF data.
Note: Number of countries, and not percentages, are reported given the
small number of cases. a The number of countries for which the WEO
forecast does a better job than the naive forecast (based
on the Theil statistic). b The number of countries in which there is
statistically significant bias (at the 5 percent level).
c The number of countries where we fail to reject the joint hypothesis
that the intercept is zero and the slope is one, i. e., the evidence is
not strong enough to reject the assumption of efficiency. d Italics
indicate that there are too few countries to consider bias a systematic
trend.
This improvement is evident across the full range of analyses. For
example, in six out of seven countries, the year- ahead WEO forecasts of
GDP and inflation for the G- 7 countries were found to be accurate, and
the near- term forecasts for GDP and inflation were accurate for all of
the G- 7 countries. These results are considerably better than WEO
forecasts for emerging markets. In the year- ahead forecasts, bias and
efficiency were a concern in
two forecasts of GDP, and one current account forecast, but they were not
a concern in the inflation forecasts. Although the year- ahead forecast
for current account was inaccurate for six of seven countries, the near-
term forecast was accurate for five of the G- 7 country cases. The
improved quality of WEO forecasts for the G- 7 countries is likely due to
better available data and greater stability of these economies, compared
to emerging market countries.
Accuracy a (number of countries)
Bias b (number of
countries) Direction of bias (number of countries) Efficiency c
(number of countries) Forecast variable Yearahead Nearterm
Yearahead Nearterm Yearahead Nearterm Yearahead Nearterm
GDP growth rate 6 7 2 0 Upward d N/ A 5 7 Inflation 6 7 0 0 N/ A N/ A 7 7
Current account (billions of U. S. $) 1 510Downward d N/ A 5 7
Appendix III Assessment of IMF Forecasting
Page 46 GAO- 03- 734 International Financial Crises
WEO Forecasts of Countries on an IMF Program Are Superior to Nonprogram
Country Forecasts
We found that WEO forecasts for 57 countries that were on an IMF program
12 (or that borrowed Fund resources under the CCFF) for any part of the
forecast period tend to be more accurate than WEO forecasts for the 30
countries that were never on an IMF program during this period (see table
4).
Table 4: Comparison of Year- Ahead WEO Forecasts for Program and
Nonprogram Countries, 1990* 2001
Legend N/ A = not applicable Source: GAO analysis of IMF data.
a For pooled countries, if the U statistic is greater than 1, the naive
forecast does a better job than the forecast under evaluation. b Bias at
the 5 percent level of significance means that there is less than a 5
percent chance that we are
making a false rejection* i. e., saying the forecast is biased when it is
not.
Countries that borrow from the IMF are likely to be under greater scrutiny
from Fund staff than those that do not borrow, which could contribute to
an improved forecast. For this analysis, we compared the Theil statistics
for GDP growth, inflation, and current account for the two pooled
forecasts. We found that the WEO forecasts of GDP and inflation for
program countries are more accurate than those for nonprogram countries.
That is, when compared to the naive model, the program countries have a
lower Theil statistic than nonprogram countries. For both groups, the
forecast of current account is inferior to the naive forecast (a Theil
statistic greater than 1). The WEO program countries forecasts for GDP and
inflation are biased, whereas the forecasts for the nonprogram countries
were not. This indicates that by assuming implementation of the policies
12 We include countries that were on the following General Resources
Account- funded IMF programs and facilities: SBA, EFF, STF, CCFF. Some
countries were on more than one program during the forecast period.
Accuracy Theil U- statistic a Bias
5- percent significance level b Direction of bias Forecast variable 57
program
countries 30 Nonprogram countries 57 program
countries 30 Nonprogram countries 57 program
countries 30 Nonprogram countries
GDP growth rate .827 .941 Ye s No Up wa r d N/ A Inflation .776 .918 Yes
No Downward N/ A Current account (billions of U. S. $) 2.399 1. 129 No Yes
N/ A Downward
Appendix III Assessment of IMF Forecasting
Page 47 GAO- 03- 734 International Financial Crises
contained within the program, the Fund expects that GDP and inflation will
perform better than they actually do. In addition to the publicly
available WEO forecasts for all countries, the
IMF also produces a set of program forecasts for countries in the years
they borrow from the Fund. 13 According to the Fund, these two forecasts
should be very similar since they are prepared by the same staff in the
same manner. Our comparison of program 14 and WEO forecasts for the
initial year that each country was on program confirmed that the accuracy
of the two forecasts for GDP and inflation were nearly identical (see
table 5).
Table 5: Comparison of Program and WEO Year- Ahead Forecasts, for 52
Countries in the Initial Years on Program, 1990* 2001 a
Legend N/ A = not applicable Source: GAO analysis of IMF data.
Note: This analysis is based on 96 country/ year observations, which is an
average across three variables. a We were able to obtain program
projections for only 52 of the 57 program countries.
b If the U statistic is greater than 1, the naive forecast does a better
job than the forecast under evaluation. c Bias at the 5 percent level of
significance means that there is less than a 5 percent chance that we are
making a false rejection* i. e., saying the forecast is biased when it is
not. d The forecast is biased at the 10 percent level of significance
means that there is less than a 10
percent chance that we are making a false rejection* i. e., saying the
forecast is biased when it is not.
13 For most of the years studied, these forecasts were not made public. In
recent years, when countries agree, these forecasts are made public. 14 We
used the original program numbers, that is, the projections established at
the outset when the IMF*s Executive Board first approves an arrangement.
Accuracy Theil U- statistic b Bias
5 percent significance level c Direction of bias Forecast variable Program
forecasts WEO forecasts Program
forecasts WEO forecasts Program
forecasts WEO forecasts
GDP growth rate .689 .676 Yes Yes Upward Upward Inflation .683 .694 No d
No d N/ A N/ A Current account (billions of U. S. $) .524 2.292 No No N/ A
N/ A
Appendix III Assessment of IMF Forecasting
Page 48 GAO- 03- 734 International Financial Crises
However, program forecasts of current account are substantially better
than those reported in the WEO. In addition, in all cases the program
forecasts were substantially more accurate than the naive model. This is
further evidence that the greater scrutiny experienced by these countries
while under a program probably contributes to an improved forecast. Other
Forecasting Efforts
Have Difficulties Similar to the Fund
Our review of other forecast evaluations found that the shortcomings we
observed in WEO forecasting are similar to difficulties encountered by
other forecasters. 15 These studies examined forecasts 16 produced by the
private sector (for example, consensus forecasts), governments, and
multinational agencies including the IMF and Organization for Economic
Cooperation and Development. These studies, similar to our observation in
this report, found a general inability to predict recessions. In addition,
consistent with our results, these studies found that (1) the shorter the
time horizon, the more accurate the forecasts; (2) current account
forecasts are
15 Harjit K. Arora & David J. Smyth, *Forecasting the Developing World: An
Accuracy Analysis of the IMF*s Forecasts,* International Journal of
Forecasting 6 (1990): 393- 400; Michael J. Artis, *How Accurate are the
IMF*s Short- Term Forecasts? Another Examination of the World Economic
Outlook,* WP/ 96/ 89 (Washington, D. C.: International Monetary Fund
[IMF], August 1996); Jose M. Barrionuevo, *How Accurate Are the World
Economic Outlook
Projections?* in World Economic Outlook, ch. 2, Staff Studies for World
Economic Outlook (Washington, D. C.: IMF, 1993); Roy Batchelor, *The IMF
and OECD versus Consensus Forecasts* (London, England: City University
Business School, August 2000); William W. Beach, Aaron B. Schavey, &
Isabel M. Isidro, *How Reliable Are IMF Economic Forecasts?* CDA 99- 05
(Washington, D. C.: The Heritage Foundation, The Heritage Center for Data
Analysis, August 27, 1999); David Fintzen & H. O. Stekler, *Why Did
Forecasters Fail to Predict the 1990 Recession?* International Journal of
Forecasting 15 (1999): 309- 323; IMF, *The Accuracy of World Economic
Outlook Growth Forecasts: 1991* 2000,* in World Economic Outlook, Box 3.1
(Washington, D. C.: IMF, December 2001): 37- 39, and IMF, *How Well Do
Forecasters Predict Turning Points?* in World Economic Outlook, Box 1.1
(Washington, D. C.: IMF, May 2002): 6- 8; Grace Juhn & Prakash Loungani,
*Further CrossCountry Evidence on the Accuracy of the Private Sector*s
Output Forecasts,* IMF Staff Papers 49, no. 1 (2002); Prakash Loungani,
*How Accurate are Private Sector Forecasts? Cross- Country Evidence from
Consensus Forecasts of Output Growth,* International Journal of
Forecasting 17 (2001): 419- 432; Albert Musso & Steven Phillips,
*Comparing Projections and Outcomes of IMF- Supported Programs,* IMF Staff
Papers 49, no. 1 (2002); Scott Shuh, *An Evaluation of Recent
Macroeconomic Forecast Errors,* New England Economic Review (January/
February 2001); Marten Blix et al., *How Good is the Forecasting
Performance of Major Institutions?* Economic Review (Stockholm, Sweden:
Sveriges Riksbank Monetary Policy Department, March 2001): 38- 68; and
Victor Zarnowitz,
*The Record and Improvability of Economic Forecasting,* NBER Working Paper
No. 2099 (December 1986). 16 Forecasters evaluated the accuracy of many
variables, including output (GDP), inflation, current account, exports,
and imports, among others.
Appendix III Assessment of IMF Forecasting
Page 49 GAO- 03- 734 International Financial Crises
markedly weaker than the forecasts for GDP and inflation; (3) when bias is
found, forecasts tend to overestimate GDP and underestimate inflation; and
(4) GDP and inflation forecasts for the industrial countries tend to be
more accurate and less biased than forecasts for developing countries.
While several studies found that WEO forecasts for developing countries
were inferior to those generated by the naive model, one study found that
WEO forecasts for developing countries did notably better than a naive
forecast. A number of studies compared the quality of WEO forecasts with
those produced by the private sector. Although some studies found the
relative quality of the forecasts to be generally the same, a few studies
found WEO forecasts to be less accurate than those of the private sector.
Page 50 GAO- 03- 734 International Financial Crises
Appendix IV Eighty- seven Emerging Market Countries Appendi x I V
Country Region a Income group b
1 Albania Countries in transition Lower middle income 2 Algeria Africa
Lower middle income 3 Antigua and Barbuda Western Hemisphere Upper middle
income 4 Argentina Western Hemisphere Upper middle income 5 Bahrain Middle
East and Turkey Upper middle income 6 Belarus Countries in transition
Lower middle income 7 Belize Western Hemisphere Lower middle income 8
Bolivia Western Hemisphere Lower middle income 9 Botswana Africa Upper
middle income 10 Brazil Western Hemisphere Upper middle income 11 Bulgaria
Countries in transition Lower middle income 12 Cape Verde Africa Lower
middle income 13 Chile Western Hemisphere Lower middle income 14 China
Developing Asia Lower middle income 15 Colombia Western Hemisphere Lower
middle income 16 Costa Rica Western Hemisphere Upper middle income 17 Cote
d*Ivoire Africa Low income 18 Croatia Countries in transition Upper middle
income 19 Czech Republic Countries in transition Upper middle income 20
Djibouti Africa Lower middle income 21 Dominica Western Hemisphere Upper
middle income 22 Dominican Republic Western Hemisphere Lower middle income
23 Ecuador Western Hemisphere Lower middle income 24 Egypt Middle East and
Turkey Lower middle income 25 El Salvador Western Hemisphere Lower middle
income 26 Equatorial Guinea Africa Lower middle income 27 Estonia
Countries in transition Upper middle income 28 Fiji Developing Asia Lower
middle income 29 Gabon Africa Upper middle income 30 Grenada Western
Hemisphere Upper middle income 31 Guatemala Western Hemisphere Lower
middle income 32 Guyana Western Hemisphere Lower middle income 33 Honduras
Western Hemisphere Lower middle income 34 Hungary Countries in transition
Upper middle income 35 India Developing Asia Low income
Appendix IV Eighty- seven Emerging Market Countries
Page 51 GAO- 03- 734 International Financial Crises
36 Indonesia Developing Asia Low income 37 Iran, Islamic Rep. of Middle
East and Turkey Lower middle income 38 Iraq Middle East and Turkey Lower
middle income 39 Jamaica Western Hemisphere Lower middle income 40 Jordan
Middle East and Turkey Lower middle income 41 Kazakhstan Countries in
transition Lower middle income 42 Kiribati Developing Asia Lower middle
income 43 Korea Developing Asia Upper middle income 44 Latvia Countries in
transition Lower middle income 45 Lebanon Middle East and Turkey Upper
middle income 46 Libya Middle East and Turkey Upper middle income 47
Lithuania Countries in transition Lower middle income 48 Macedonia, former
Yugoslav Rep. of Countries in transition Lower middle income 49 Malaysia
Developing Asia Upper middle income 50 Maldives Developing Asia Lower
middle income 51 Mauritius Africa Upper middle income 52 Mexico Western
Hemisphere Upper middle income 53 Morocco Africa Lower middle income 54
Namibia Africa Lower middle income 55 Nigeria Africa Low income 56 Oman
Middle East and Turkey Upper middle income 57 Pakistan Developing Asia Low
income 58 Panama Western Hemisphere Upper middle income 59 Papua New
Guinea Developing Asia Lower middle income 60 Paraguay Western Hemisphere
Lower middle income 61 Peru Western Hemisphere Lower middle income 62
Philippines Developing Asia Lower middle income 63 Poland Countries in
transition Upper middle income 64 Romania Countries in transition Lower
middle income 65 Russia Countries in transition Lower middle income 66
Samoa Developing Asia Lower middle income 67 Saudi Arabia Middle East and
Turkey Upper middle income 68 Seychelles Africa Upper middle income 69
Slovak Republic Countries in transition Upper middle income 70 South
Africa Africa Upper middle income 71 Sri Lanka Developing Asia Lower
middle income 72 St. Kitts and Nevis Western Hemisphere Upper middle
income
(Continued From Previous Page)
Country Region a Income group b
Appendix IV Eighty- seven Emerging Market Countries
Page 52 GAO- 03- 734 International Financial Crises
Sources: IMF, World Bank, and J. P. Morgan. a Based on WEO regional
groupings, IMF 2002.
b World Bank analytical classification by income, July 2001. 73 St. Lucia
Western Hemisphere Upper middle income
74 St. Vincent and the Grenadines Western Hemisphere Lower middle income
75 Suriname Western Hemisphere Lower middle income 76 Swaziland Africa
Lower middle income 77 Syrian Arab Rep. Middle East and Turkey Lower
middle income 78 Thailand Developing Asia Lower middle income 79 Tonga
Developing Asia Lower middle income 80 Trinidad and Tobago Western
Hemisphere Upper middle income 81 Tunisia Africa Lower middle income 82
Turkey Middle East and Turkey Upper middle income 83 Turkmenistan
Countries in transition Lower middle income 84 Ukraine Countries in
transition Low income 85 Uruguay Western Hemisphere Upper middle income 86
Vanuatu Developing Asia Lower middle income 87 Venezuela Western
Hemisphere Upper middle income
(Continued From Previous Page)
Country Region a Income group b
Page 53 GAO- 03- 734 International Financial Crises
Appendix V Standards, Codes, and Principles Assessed under IMF and World
Bank Reports on the Observance of Standards and Codes Appendi x V
Assessment start date Responsible institution Standard or code and
rationale for adoption
Transparency standards: The standards on transparency in government
operations and policy making are considered within the Fund*s direct
operational focus.
1996* 97 IMF IMF Special Data Dissemination Standard (SDDS) and General
Data Dissemination Standard (GDDS). The purpose of the IMF*s SDDS is to
guide member country governments that have, or might seek, access to
international capital markets in publishing comprehensive, timely,
accessible, and reliable economic and financial statistics. The purpose of
the
GDDS is to help any member country government provide more reliable
economic data. The SDDS and GDDS were created in 1996 and 1997,
respectively.
1998 (modified 2001) IMF IMF Code of Good Practices on Fiscal
Transparency.
This Code is intended to help member country governments improve the
disclosure of information about the design and results of fiscal policy,
making governments more accountable for policy implementation and
strengthening credibility and public understanding of macroeconomic
policies and choices. 1999 IMF IMF Code of Good Practices on Transparency
in Monetary and Financial Policies.
This Code is designed to increase the effectiveness of monetary and
financial policies by raising public awareness of the government*s policy
goals and instruments and making governments (especially independent
central banks and financial agencies) more accountable. Financial sector
standards: The financial sector standards are considered within the direct
operational focus of both the Fund and the World Bank and are generally
assessed under the joint Fund- Bank FSAP. 1999 Joint IMF- Bank Basel
Committee*s Core Principles for Effective Banking
Supervision (BCP). The BCP is intended to guide the development of an
effective system for supervising banks, a large sector of most economies.
The IMF and World Bank began assessments of countries*
compliance with the BCP standard in conjunction with the Financial Sector
Assessment Program (FSAP) launched in 1999.
1999 Joint IMF- Bank International Organization of Securities Commissions*
(IOSCO)
Objectives and Principles for Securities Regulation.
The IOSCO Objectives and Principles are designed to help governments
establish effective systems to regulate securities which contribute
strongly to investor confidence. The IMF and
World Bank began using them to assess securities regulation in conjunction
with the FSAP, launched in 1999.
Appendix V Standards, Codes, and Principles Assessed under IMF and World
Bank Reports on the Observance of Standards and Codes
Page 54 GAO- 03- 734 International Financial Crises
1999 (modified 2000) Joint IMF- Bank International Association of
Insurance Supervisors* (IAIS)
Insurance Core Principles.
The IAIS Core Principles are designed to contribute to effective insurance
supervision that supports financial stability. The IMF and World Bank
began assessing member countries* regulatory practices in this area in
conjunction with the FSAP, launched in 1999.
1999 Joint IMF- Bank Committee on Payments and Settlements Systems* (CPSS)
Core Principles for Systemically Important Payment Systems.
The CPSS Core Principles are intended to strengthen payments systems,
which provide the channels through which funds are transferred among banks
and other institutions. The IMF and the World Bank began assessing member
countries* observance of this standard in conjunction with the FSAP,
launched in 1999.
2002 Joint IMF- Bank Financial Action Task Force (FATF) 40 Recommendations
on AntiMoney Laundering and 8 Special Recommendations on Terrorism
Financing.
The FATF*s 40 Recommendations and 8 Special Recommendations are intended
to promote policies that combat money laundering and terrorist financing,
which threaten financial system integrity and may undermine the sound
functioning of financial systems. In 2002, the IMF and World Bank agreed
to perform anti- money laundering and terrorist financing assessments as a
12- month pilot program, generally in conjunction with the FSAP.
Corporate sector standards: The corporate sector standards are considered
important for the effective operation of domestic and international
financial systems and are assessed by the World Bank.
2000 World Bank Organization for Economic Cooperation and Development*s
(OECD) Principles of Corporate Governance.
The OECD developed its corporate governance principles to help governments
evaluate and improve their legal, institutional, and regulatory frameworks
for corporate governance. The World Bank developed a template for
assessing adherence to corporate governance principles based on the OECD*s
Principles established in 1999.
(Continued From Previous Page)
Assessment start date Responsible institution Standard or code and
rationale for adoption
Appendix V Standards, Codes, and Principles Assessed under IMF and World
Bank Reports on the Observance of Standards and Codes
Page 55 GAO- 03- 734 International Financial Crises
Sources: IMF and World Bank. 2000 World Bank International Accounting
Standards Board*s (IASB) International Accounting Standards.
The ROSC*s accounting module is intended to compare member countries*
corporate accounting practices with international accounting standards and
to analyze actual accounting practice to determine the extent of
compliance with applicable standards. There is special focus on the
strengths and weaknesses of the institutional framework for supporting
high quality accounting and financial reporting. In 2000, the World Bank
developed a template for assessing adherence to accounting standards based
on the IASB Standards. 2000 World Bank International Federation of
Accountants* (IFAC) International
Standards on Auditing.
The ROSC*s auditing module compares member countries* auditing standards
and auditors* professional code of ethics with the standards and codes
issued by IFAC. Also, the quality of actual auditing practices is
evaluated. There is special focus on the
strengths and weaknesses of the institutional framework for supporting
high quality audit. In 2000, the World Bank developed a template for
assessing adherence to auditing standards based on the IFAC*s Standards.
2001 (draft) World Bank World Bank Principles and Guidelines for Effective
Insolvency and Creditor Rights Systems.
In 2001, the World Bank developed draft Principles and Guidelines intended
to help countries develop effective insolvency and creditor rights*
systems, two important components of financial system stability. The World
Bank has conducted several assessments based on its draft Principles and
Guidelines. The United Nations Commission on International Trade Law
(UNCITRAL) is completing a draft Legislative Guide on Insolvency Law.
UNCITRAL, Bank,
and IMF staff are working toward a single standard.
(Continued From Previous Page)
Assessment start date Responsible institution Standard or code and
rationale for adoption
Page 56 GAO- 03- 734 International Financial Crises
Appendix VI Update on the International Monetary Fund*s Safeguards
Assessments Appendi x VI
In response to allegations of misreporting and misuse of International
Monetary Fund (IMF or the Fund) disbursements in the late 1990s, the Fund
increased its efforts to protect its resources by introducing safeguards
assessments, a process for evaluating the controls employed by the central
banks of borrowing member countries and for recommending measures to
address inadequacies. Safeguards assessments have detected numerous
inadequacies that could lead to misuse of Fund resources and have
recommended measures to remedy them.
The Fund Introduces Safeguards Assessments for Member Countries That
Currently Borrow
In 2000, the Fund introduced safeguards assessments, a process for
identifying inadequacies in central banks* ability to ensure the integrity
of their operations, especially the use of Fund resources. Safeguards
assessments evaluate central banks* internal and external audit
mechanisms, legal structure and independence, financial reporting
procedures, and systems of internal controls.
In April 2002, the Fund*s Executive Board made safeguards assessments a
permanent policy. Safeguards assessments apply to all member countries
with current or anticipated borrowing arrangements with the Fund.
Countries with borrowing arrangements approved after June 30, 2000, are
subject to a full safeguards assessment covering the five areas listed
above. Countries with arrangements in effect before June 30, 2000, that
have not
yet repaid all Fund resources, were subject to a partial assessment
covering only the external audit mechanism. Countries that do not have
borrowing arrangements or have already repaid all Fund resources are not
subject to safeguards assessments. According to Fund officials, since 2000
the IMF has not provided financial resources to countries that did not
meet its safeguards requirements. As of December 2002, the Fund had
performed 37 full safeguards assessments and 27 partial assessments, with
23 assessments under way. 1 The completed assessments detected a number of
serious vulnerabilities
that could lead to misuse of central bank resources, including those
borrowed from the Fund. Of the full safeguards assessments, the Fund found
the following:
1 International Monetary Fund, Safeguards Assessments * Semi- Annual
Update, SM/ 03/ 88 (Washington, D. C.: Mar. 6, 2003).
Appendix VI Update on the International Monetary Fund*s Safeguards
Assessments
Page 57 GAO- 03- 734 International Financial Crises
Inadequate accounting standards in 82 percent of the central banks,
which interfere with the accurate recording of central bank operations.
For example, some central banks did not adhere to a financial reporting
framework such as the International Accounting Standards (IAS), which
would help prevent misreporting of transactions.
Deficient internal audit in 79 percent of central banks, which reduces
their ability to address risks of misuse and misreporting of Fund
resources. For example, some internal audit departments did not audit
high- risk areas such as foreign reserves management.
Poor controls over foreign reserves and data reporting to the Fund in 49
percent of the central banks, increasing the possibility of misreporting
and misuse of Fund resources. For example, safeguards assessments
identified improper techniques for valuing foreign reserves and failure to
reconcile data reported to the Fund for program monitoring purposes with
underlying accounting records.
According to Fund officials, when IMF staff detect significant weaknesses
in the controls of assessed central banks, they recommend that the
government take corrective actions. For actions that IMF staff consider
essential, they may incorporate the recommendations into the list of
preconditions that the Fund requires borrower countries meet before
receiving IMF resources, or they may suggest that the government include
the recommended actions in its official economic program. The Fund reports
that of the 275 recommendations that were expected to be implemented on or
before December 31, 2002, 23 percent were
incorporated as conditions for IMF resources or included in official
economic program statements.
Implementation of the Fund*s Assessment Recommendations
Fund staff monitor central banks* implementation of recommendations by
performing in- depth reviews of their audited annual financial statements
and other documents every 12 to 18 months until the borrower country
government has repaid all Fund resources. The Fund monitors on a
continuous basis, central banks* implementation of all other safeguards
measures and of developments within the central banks that may lead to
new vulnerabilities.
Appendix VI Update on the International Monetary Fund*s Safeguards
Assessments
Page 58 GAO- 03- 734 International Financial Crises
Recently, the Fund reported that central banks have implemented 90 percent
of recommendations that IMF staff included as a precondition for receiving
IMF resources. According to Fund officials, the IMF stopped disbursing
resources in the few cases where governments failed to implement these
essential recommendations. Similarly, the Fund reported that central banks
have implemented 84 percent of measures included in governments* official
economic program statements.
On the other hand, the Fund reported that some recommendations made by the
safeguards assessments have not been implemented as intended, although
Fund officials state that these delays typically occurred in nonpriority
areas. When central bank authorities fail to implement the
recommendations, Fund staff increase pressure to comply, often proposing
the measures* inclusion as a precondition for the next disbursement.
However, the Fund reports that staff can only adopt this approach in
countries where the Fund is actively disbursing funds. For countries that
are not currently receiving Fund disbursements, implementation of
recommendations from the safeguards assessments tends to be more
problematic because the Fund cannot exert pressure through a borrowing
arrangement.
Page 59 GAO- 03- 734 International Financial Crises
Appendix VII Fund/ World Bank FSAP and ROSC Process Appendi x VI I
Page 60 GAO- 03- 734 International Financial Crises
Appendix VIII Country Participation in and Publication of FSAPs and ROSCs
Appendi x VI I I
Figure 4 lists all countries that have participated in Financial Sector
Assessment Program (FSAP) or Reports on the Observance of Standards and
Codes (ROSC) assessments and whether or not these assessments were
published. The figure describes participation and publication by the 33
major emerging market countries. Countries highlighted in bold have not
participated in any assessments. Figure 5 describes participation and
publication by other countries (industrial, developing, and smaller
emerging markets).
Appendix VIII Country Participation in and Publication of FSAPs and ROSCs
Page 61 GAO- 03- 734 International Financial Crises
Figure 4: FSAP and ROSC Participation and Publication by Major Emerging
Market Countries
a These countries participated in the FSAP under a pilot program. The
reports were not intended for publication.
Appendix VIII Country Participation in and Publication of FSAPs and ROSCs
Page 62 GAO- 03- 734 International Financial Crises
Figure 5: FSAP and ROSC Participation and Publication by Other Countries
Appendix VIII Country Participation in and Publication of FSAPs and ROSCs
Page 63 GAO- 03- 734 International Financial Crises
a These countries participated in the FSAP under a pilot program. The
reports were not intended for publication.
Page 64 GAO- 03- 734 International Financial Crises
Appendix IX Fund Resources Provided under the Supplemental Reserve
Facility Appendi x I X
In recent financial crises, the International Monetary Fund (IMF or the
Fund) provided large short- term loans under the Supplemental Reserve
Facility (SRF) with high interest rates to member countries experiencing
exceptional balance of payments problems. These problems resulted from a
sharp decline of investor confidence and significant outflows of capital.
These loans generally were provided when the countries had exceeded their
financing limit under other loan mechanisms, including the Stand- By
Arrangement (SBA). In some circumstances, such as Argentina and Uruguay,
the Fund provided a mix of SRF and SBA loans. Table 6 lists Fund members
receiving SRF loans from 1997 to 2002.
Table 6: IMF Supplemental Reserve Facility Loans, 1997* 2002
Source: IMF.
Country Loan approval date Loan mix SRF repaid as of May 2003
Korea 12/ 18/ 97 SRF Yes Russia 7/ 20/ 98 SRF Yes Brazil 12/ 2/ 98 SRF Yes
Turkey 12/ 21/ 00 SRF Yes
Argentina 1/ 12/ 01 SRF (40%) SBA (60%) No
Argentina 9/ 7/ 01 SRF (63%) SBA (37%) No
Brazil 9/ 14/ 01 SRF No Uruguay 6/ 25/ 02 SRF (33%)
SBA (67%) No Brazil 9/ 6/ 02 SRF (33%)
SBA (67%) No
Page 65 GAO- 03- 734 International Financial Crises
Appendix X Comments from the Department of the Treasury Appendi x X
Appendix X Comments from the Department of the Treasury
Page 66 GAO- 03- 734 International Financial Crises
Page 67 GAO- 03- 734 International Financial Crises
Appendix XI Comments from the International Monetary Fund Appendi x XI
Note: GAO comments supplementing those in the report text appear at the
end of this appendix.
See comment 1. See comment 2.
Appendix XI Comments from the International Monetary Fund
Page 68 GAO- 03- 734 International Financial Crises
See comment 3. See comment 4. See comment 5.
See comment 6.
Appendix XI Comments from the International Monetary Fund
Page 69 GAO- 03- 734 International Financial Crises
Appendix XI Comments from the International Monetary Fund
Page 70 GAO- 03- 734 International Financial Crises
The following are GAO*s comments on the letter from the International
Monetary Fund, dated June 2, 2003. GAO Comments 1. Our assessment of IMF*s
efforts to anticipate financial crises did not
focus exclusively on the World Economic Outlook (WEO) and the Early
Warning System (EWS) models. We examined all six components of the IMF*s
vulnerability assessment framework. However, as the only mature and
quantifiable elements of the framework, our analysis focused more heavily
on the track records of the WEO and EWS. As we reported, these elements
have not performed well at anticipating prior crises. While we acknowledge
that the new framework is more comprehensive than past efforts at
anticipating crises, it is too early to
assess whether this framework will be successful in anticipating future
crises.
2. The IMF*s comment supports our finding. The IMF states that its
responsibility to maintain financial stability could make its predictions
appear less accurate since an accurate prediction of crises within WEO
forecasts would be irresponsible and could lead to a crisis. In effect,
the IMF acknowledges that their forecasts are overly optimistic and
validates our finding on the weakness of the WEO component of the
vulnerability assessment framework. This raises questions regarding the
purpose and credibility of the WEO forecasts.
3. We state in the report that the IMF*s new vulnerability assessment
framework, which includes the examination of external reserve adequacy and
the strengths and weaknesses of banking systems, is more comprehensive
than its previous efforts to identify countries at risk of crisis.
However, it is too early to assess whether this framework will be
successful in anticipating future crises.
4. We recognize that the EWS models are just one of six components of the
IMF*s vulnerability assessment framework. However, the IMF*s own internal
review of the EWS models concluded that the false alarm rate was too high.
5. The report recognizes that the FSAP and ROSC assessments constitute a
valuable source of information about vulnerabilities in member countries
and states that IMF staff use these assessments to formulate policy
recommendations. The report also recognizes that the IMF often provides
technical assistance to help member countries build their
Appendix XI Comments from the International Monetary Fund
Page 71 GAO- 03- 734 International Financial Crises
capacity to implement FSAP and ROSC recommendations. However, the report
points out several factors that limit the usefulness of FSAP and ROSC
assessments. Our recommendation, with which the IMF agrees, is designed to
improve the timeliness and coverage of these assessments. 6. We based our
description of IMF safeguards assessments on the IMF*s
reviews of this program. We consider this topic to be within the scope of
this evaluation because the framework for conducting safeguards
assessments is derived from the IMF*s Code of Good Practices on
Transparency in Monetary and Financial Policies. Safeguards assessments
are thus related to the standards initiative, which constitutes a central
element of this report.
Page 72 GAO- 03- 734 International Financial Crises
Appendix XII Comments from The World Bank Appendi x XI I
Note: GAO comments supplementing those in the report text appear at the
end of this appendix.
See comment 1. See comment 2.
Appendix XII Comments from The World Bank
Page 73 GAO- 03- 734 International Financial Crises
Appendix XII Comments from The World Bank
Page 74 GAO- 03- 734 International Financial Crises
The following are GAO*s comments on the letter from the World Bank, dated
June 2, 2003.
GAO Comments 1. The report states unambiguously that crises can stem from
a number of factors, some of which are outside the scope of the FSAP and
ROSC
assessments. However, there is broad agreement that the roots of the
Mexican and Asian financial crises lay in weaknesses in financial systems
and other institutions. The IMF and the World Bank based their decision to
launch the FSAP and ROSC initiatives on the premise that timely
identification of financial sector and institutional vulnerabilities can
contribute to crisis prevention. The IMF and the World Bank have also
acknowledged that FSAP and ROSC assessments can contribute to crisis
prevention efforts by helping private sector participants make better
informed investment decisions. 2. Our recommendation to pursue strategies
to increase participation in
the FSAP and ROSC assessments, including the possibility of making these
assessments mandatory, stems from the IMF*s and the World Bank*s
recognition of the need to prioritize participation by important emerging
market countries. Although many of these countries have volunteered to
participate in these assessments, others have not. While we are not
suggesting that the assessments should be made mandatory, it is evident
that the voluntary nature of the FSAP and ROSC has posed
an obstacle to full participation by important emerging market countries.
Page 75 GAO- 03- 734 International Financial Crises
Appendix XIII GAO Contacts and Staff Acknowledgments Appendi x XI I I
GAO Contacts Tom Melito (202) 512- 9601 Zina Merritt (202) 512- 5257
Staff Acknowledgments
In addition to those individuals named above, Eric Clemons, Suzanne Dove,
Bruce Kutnick, Jonathan Rose, R. G. Steinman, Ian Ferguson, Mary Moutsos,
Lynn Cothern, Carl Barden, David Dornisch, and Martin De Alteriis made key
contributions to this report.
(320114)
GAO*s Mission The General Accounting Office, the audit, evaluation and
investigative arm of Congress, exists to support Congress in meeting its
constitutional responsibilities
and to help improve the performance and accountability of the federal
government for the American people. GAO examines the use of public funds;
evaluates federal programs and policies; and provides analyses,
recommendations, and other assistance to help Congress make informed
oversight, policy, and funding decisions. GAO*s commitment to good
government is reflected in its core values of accountability, integrity,
and reliability.
Obtaining Copies of GAO Reports and Testimony
The fastest and easiest way to obtain copies of GAO documents at no cost
is through the Internet. GAO*s Web site (www. gao. gov) contains abstracts
and fulltext files of current reports and testimony and an expanding
archive of older products. The Web site features a search engine to help
you locate documents using key words and phrases. You can print these
documents in their entirety, including charts and other graphics.
Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as *Today*s Reports,* on its
Web site daily. The list contains links to the full- text document files.
To have GAO e- mail this
list to you every afternoon, go to www. gao. gov and select *Subscribe to
GAO Mailing Lists* under *Order GAO Products* heading.
Order by Mail or Phone The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO
also accepts VISA and Mastercard. Orders for 100 or more copies mailed to
a single address are discounted 25 percent. Orders should be sent to:
U. S. General Accounting Office 441 G Street NW, Room LM Washington, D. C.
20548
To order by Phone: Voice: (202) 512- 6000 TDD: (202) 512- 2537 Fax: (202)
512- 6061
To Report Fraud, Waste, and Abuse in Federal Programs
Contact: Web site: www. gao. gov/ fraudnet/ fraudnet. htm E- mail:
fraudnet@ gao. gov Automated answering system: (800) 424- 5454 or (202)
512- 7470
Public Affairs Jeff Nelligan, Managing Director, NelliganJ@ gao. gov (202)
512- 4800 U. S. General Accounting Office, 441 G Street NW, Room 7149
Washington, D. C. 20548
United States General Accounting Office Washington, D. C. 20548- 0001
Official Business Penalty for Private Use $300
Address Service Requested Presorted Standard
Postage & Fees Paid GAO Permit No. GI00
*** End of document. ***