Executive Summary

Kevin Barnes, Ali Mansoor, Benjamin Cohen, and Shinji Takagi
Published Date:
September 2003
  • ShareShare
Show Summary Details

This report evaluates the role of the IMF in three recent capital account crises, in Indonesia (1997–98), Korea (1997–98), and Brazil (1998–99). These crises have been the subject of extensive external commentary and have also been studied in detail by IMF staff. A number of important lessons have already been learned and corresponding corrective steps taken in the form of revised IMF policies and procedures. Nevertheless, it is appropriate for the Independent Evaluation Office (IEO) to conduct an independent assessment of the role of the IMF in these crises, taking advantage of its unique access to internal IMF documents while also taking note of earlier work where relevant. The evaluation seeks to draw lessons for the IMF, supplementing those that have already surfaced, and also to contribute to transparency by evaluating the internal processes by which important decisions were made.

The findings of this evaluation report are subject to three important limitations. First, any evaluation inevitably benefits from hindsight and while this can be an advantage in drawing lessons for the future, much of what we know now may not have been known at the time to those who had to make the relevant decisions, often under extreme pressure. These considerations must be borne in mind in determining accountability. Second, any evaluation implies a comparison with a counterfactual, that is, what might have happened with alternative policies. This is very difficult to establish rigorously. Third, the behavior of an economy is always subject to uncertainty, and the uncertainties are much greater in crisis situations. In the face of uncertainty, a program cannot be judged to represent a mistaken choice ex ante just because it failed ex post. The relevant criterion is whether the ex ante probability of success was high enough.

The report consists of two parts. The main report presents our assessment of the role of the IMF in the three crises and the lessons to be drawn from the experience, with some specific recommendations going beyond the steps already taken. The annexes contain the three country studies that form the basis for our judgments in the main report.

Overall Assessment of the Role of the IMF

The three country cases studied share several features common to capital account crises; in each case the crisis was triggered by massive reversal of capital flows, short-term flows played a prominent role, and contagion was an important factor. However, there were also notable differences. The nature of the crisis differed in the three cases, with Indonesia and Korea exemplifying “twin crises” in which the external crisis coincided with a banking crisis. There were also differences in the policy mix advocated, the political environment in which the crisis was managed, and the effectiveness of policy implementation. All three programs failed in their initially stated objectives, but the subsequent experience under the revised programs was very different. Our overall assessment of the role of the IMF in each of the three crises is as follows.


IMF surveillance did identify the vulnerabilities in the banking sector that would later become crucial to the evolution of the crisis, but it underestimated the severity and the potential macroeconomic risks posed by them. In designing its crisis management strategy during October 1997, the IMF misjudged the extent of ownership at the highest political level and underestimated the resistance to reform likely to be posed by vested interests. This underestimation of political constraints was perhaps a reflection of the earlier failure of surveillance in recognizing the changing nature of corruption and cronyism.

The single greatest cause of the failure of the November 1997 program was the lack of a comprehensive bank restructuring strategy, which led to a rapid expansion of liquidity to support weak banks. The resulting loss of monetary control in turn contributed to a weaker exchange rate and greater distress in the corporate sector. The crisis became intensely political, following the illness of the President in early December, making crisis management even more difficult. At this stage, the IMF negotiated a revised program in January 1998, which focused heavily on structural conditionality to signal a clean break with the past. The focus on structural conditionality was based on the assumption that this was necessary to restore confidence. It failed to do so, partly because of visible lack of political commitment to the policies promised and partly because of the failure to address the critical banking and corporate debt problems.

The Indonesian crisis was clearly the most severe of the three under review, with GDP declining by 13 percent in 1998 and a large increase in poverty. This devastating outcome cannot be attributed solely to shortcomings on the part of the IMF. The lack of firm implementation of the November program, and especially the reversal of some of the critical steps at a very early stage, eroded market confidence and the situation soon got out of control as political uncertainty increased and riots occurred against the ethnic Chinese community. These exceptional circumstances explain much of the severity of the crisis experienced by Indonesia. However, our evaluation suggests that the IMF’s response to the failure was also inadequate in many respects.


In Korea, IMF surveillance failed adequately to identify the risks posed by the uneven pace of capital account liberalization and the extent of banking sector weaknesses, owing to the adoption of a conventional approach that focused on macroeconomic variables. There were gaps in the data needed to make a full assessment, though available data on short-term debt and financial market indicators were not fully used. While concerns over Korea’s weak banking sector had prompted international banks to review their lending to some Korean institutions even before the onset of the Asian crisis in July 1997, the IMF was optimistic until virtually the last minute.

The first Korean program was clearly underfinanced, but this was due primarily to the unwillingness of major shareholder governments either to take concerted action to involve the private sector or to provide the necessary financing upfront to resolve what, of all the three cases, was most clearly a liquidity crisis. When this strategy failed, the major shareholder governments moved quickly to initiate concerted action to involve the private sector—an approach that eventually worked well. It could be argued that the first strategy needed to be tried and proven to have failed before the rollover agreement of December 24, 1997 could be secured. The IMF played a useful role as crisis coordinator in drawing attention to the problem and later facilitating information exchange among major governments and helping to set up a monitoring system to ensure compliance.

The Korean adjustment process involved a severe downturn, with GDP declining by 6.7 percent in 1998, compared with a forecast of positive growth. However, unlike Indonesia, this was followed by a robust recovery in 1999. The greater-than-expected downturn reflected the impact of negative balance sheet effects, which were clearly underestimated. In retrospect, the fiscal tightening in the program was unnecessary, as the IMF staff has itself concluded.


In Brazil, IMF surveillance was successful in identifying the key vulnerabilities that were at the core of the crisis, in part owing to the fact that they were largely macroeconomic in nature. However, it progressively downplayed the scale of overvaluation, and had little impact in persuading the Brazilian authorities to take sufficient corrective action even in areas where the diagnosis was correct. When Brazil faced intense speculative pressure on its foreign exchange reserves from mid-1998, the IMF reluctantly supported the authorities’ preference for maintaining the existing exchange rate regime. However, intense pressure on the real developed in December 1998, and the program soon failed with the collapse of the peg in January 1999.

A major justification for defending the exchange rate was that an exit from the peg at that time would have unsettled international financial markets already nervous after the Russian default and the Long-Term Capital Management crisis. With the benefit of hindsight, it can be argued that this concern was overplayed. An earlier exit from the peg, widely perceived to be unsustainable, probably would not have had major systemic effects if it had been made under an IMF-supported program. The hedge provided to the private sector by the government, through the use of foreign exchange reserves and exchange rate–indexed bonds, ensured that the sharp depreciation that followed the floating of the real in January 1999 had little adverse effect on the Brazilian economy. However, this was at the cost of a substantial increase in the stock of public debt, which stored up problems for the future.

The revised 1999 program fared fairly well in the short run. Contrary to program expectations of negative growth in 1999, Brazil actually experienced positive growth of 0.8 percent. This was largely because of the healthier state of the banking system, combined with the provision of the hedge, which mitigated balance sheet effects on the private sector. The IMF played a useful role in facilitating Brazil’s transition to an inflation-targeting monetary regime as well as a more disciplined fiscal policy regime, but in retrospect, fiscal vulnerabilities were not fully eradicated.

Precrisis Surveillance

IMF surveillance was more successful in identifying macroeconomic vulnerabilities than in recognizing and analyzing in depth the risks arising from financial sector and corporate balance sheet weaknesses and the governance-related problems that contributed to those weaknesses. Insufficient candor and transparency limited the impact of surveillance on policy, even in areas where the diagnosis was broadly accurate.

In Indonesia, the IMF did identify banking sector weaknesses as a problem, but surveillance reports underestimated the potential adverse macroeconomic consequences of these weaknesses. Surveillance also paid insufficient attention to the changing nature of corruption and the macroeconomic risks it posed, and surveillance reports were less candid on these issues.

In Korea, the IMF failed adequately to recognize the vulnerabilities created by the uneven sequence of capital account liberalization and the risk that a change in investor sentiment could cause a severe drain on foreign exchange reserves. While the crisis also came as a surprise to many other observers, the IMF was slow to catch the rising concerns of international banks over Korea’s banking sector problems, which had begun to surface several months before the onset of the full-blown crisis. In retrospect, surveillance proved too sanguine about these growing risks.

IMF surveillance effectively diagnosed the major vulnerabilities in Brazil, largely because Brazil’s vulnerabilities manifested themselves primarily as macroeconomic phenomena, such as the rising stock of public debt and real exchange rate appreciation, which were part of the IMF’s traditional tool kit.

In all three countries, the IMF’s role as confidential advisor was not very effective in persuading countries to modify their policies even when key vulnerabilities were identified. The IMF was not provided with much sensitive information required for effective surveillance. While it is difficult to generalize from three cases, or to test the counterfactual concretely, the IMF probably could have been more effective in influencing policy if it had made its analyses public so as to contribute to a wider policy debate.

Program Design and Implementation

Macroeconomic framework and projections

In all three cases, macroeconomic outcomes turned out to be very different from program projections. In Indonesia and Korea, the initial projections were overly optimistic, leading to a design of macroeconomic policies that turned out to be too tight given the outcome in aggregate demand and output. In contrast, the initial projections for Brazil in 1999 were too pessimistic, which contributed to fiscal adjustment that turned out to be insufficient, in light of that country’s adverse public debt dynamics.

Part of this problem arises because macroeconomic projections in an IMF-supported program are necessarily the outcome of negotiation. However, there were also analytical weaknesses since forecasts were not derived from an analytical framework in which the key determinants of output, and their likely behavior during the crisis, could be dealt with adequately. In particular, there was insufficient appreciation of (1) the large currency depreciation which might occur in view of the possibility of multiple equilibria, and (2) the severe balance sheet effects that might result. It is inherently difficult to forecast macroeconomic outcomes reliably, especially in crisis situations, but these problems could have been reduced if there was a more explicit focus on the key factors affecting aggregate demand, particularly private investment.

In light of the considerable uncertainties, a more explicit discussion in program documents of the major risks to the macroeconomic framework, with a clear indication of how policies would respond if the risks materialized, would have been helpful. In practice, subsequent program reviews on Indonesia and Korea did show flexibility, but an upfront recognition of risks would have sent a more transparent signal on the expected stance of policies.

Fiscal policy

All three programs involved fiscal tightening. The extent of tightening was mild in Indonesia and Korea, while it was fairly strong in Brazil. In view of output developments, the initial tightening of fiscal policy in Indonesia and Korea was not warranted, and it was in fact relaxed quickly when the extent of output collapse became evident. In any event, in both countries, the initial fiscal tightening was not the cause of the output collapse. This was the result of balance sheet effects, which were not factored into program design. In Brazil, fiscal tightening was much sharper. This was appropriate because fiscal sustainability was a major issue driving the evolution of the crisis. However, it turned out to be insufficient to achieve the objective of stabilizing, and then reducing, the debt-to-GDP ratio.

Monetary policy

The stance of monetary policy in all three countries was initially set tight, with an explicit recognition of the trade-off between higher interest rates and a weaker exchange rate. However, the experience of the three countries varies and does not provide a definitive answer to the ongoing debate on the effectiveness of high interest rates in stabilizing the exchange rate.

In Indonesia, the maintenance of tight monetary policy envisaged in the program was simply not implemented, as the monetary base expanded rapidly and real interest rates became increasingly negative during the early months of the program. The assertion by some critics that the tight monetary policy advocated by the IMF was a cause of the output collapse is not warranted for the simple reason that it was not implemented for most of the crisis period. Exchange rate stability returned in March 1998, when the rupiah had sufficiently depreciated and interest rates were raised and monetary control regained.

In contrast, Korea implemented the tight monetary policy envisioned in the initial program by raising domestic interest rates and the penalty rate charged to banks for central bank foreign currency advances. These moves were appropriate to defend the currency, but they were not by themselves sufficient to stabilize the exchange rate, because much of the capital outflow was in fact driven by credit considerations rather than yield. It can be argued that real interest rates were kept higher than might have been necessary in early 1998, when the exchange market had stabilized. However, the still uncertain situation understandably called for some caution. Given the contractionary impact of bank restructuring on credit flows, the few months of higher than necessary interest rates could not have been the dominant cause of the recession.

In Brazil, the excessive easing of interest rates—over the IMF’s objections—may have contributed to the timing, if not the eventuality, of the collapse of the crawling peg. A decisive tightening of monetary policy in March 1999 coincided with the restoration of stability in the foreign exchange market. However, one must be careful about the causality, given the fact that an informal agreement by major international banks to maintain credit lines to Brazil was reached around the same time. High interest rates did not have a major negative impact on the private sector, because of the sound state of the banking system and the low leverage of the corporate sector, compared with the situations in Asia. Subsequently, the IMF supported Brazil’s transition to an inflation-targeting regime, which allowed for price stability and a rapid reduction in interest rates.

Official financing and private sector involvement

The size and format of the official financing package were inadequate in Korea and contributed to the failure of the first program. The ambiguity over the availability of US$20 billion in bilateral assistance pledged as a “second line of defense” in Korea created uncertainty in the market about the ability of the program to meet the country’s immediate liquidity needs.

In the other two countries, the programs failed for other reasons. The failure of the initial Indonesian program was due, not to inadequate financing, but to other factors, including nonimplementation of the key elements of the program by the authorities and the subsequent explosion of liquidity because of the failure to resolve the banking crisis. Once the program had failed, the crisis became intensely political, leading to a large amount of capital flight by domestic residents, and the sharp depreciation of the rupiah began to create solvency concerns. No reasonable amount of official financing could have restored confidence at that time. In the case of Brazil, the initial program failed because the key policy, namely, that of supporting the crawling peg, was not credible with the markets.

In Korea and Brazil, the IMF’s role as crisis coordinator in organizing private sector involvement (PSI) was limited by the unwillingness of major shareholder governments to use nonmarket instruments to influence the behavior of private sector institutions and concerns that such action might precipitate an exodus of capital from emerging markets. However, when a decision was made by the major shareholders to involve the private sector, the IMF played a useful role in facilitating information exchange among major governments and helping to set up systems of monitoring compliance.

An earlier attempt to involve the private sector in Korea would have been warranted, but given the initial unwillingness of the IMF’s major shareholder governments to take concerted action, there was probably little the IMF could do. The agreement by major international banks to roll over interbank debt on December 24, 1997 was a turning point in the crisis. The success of this approach owed much to the fact that most of the short-term external debt was interbank credit. The Brazilian experience in the second program suggests that a program with a high degree of credibility is necessary for the “voluntary” approach to PSI to work. In Indonesia, the IMF provided technical assistance for corporate debt restructuring, but its role was limited.

Bank closure and restructuring

The experiences of Indonesia and Korea suggest that a successful bank closure and restructuring program must include a comprehensive and well-communicated strategy in which transparent rules are consistently applied. The Korean program by and large achieved its objectives, largely because a comprehensive strategy was developed at the outset. The Indonesian banking sector program, by contrast, initially suffered from the lack of a comprehensive strategy and the failure to communicate the logic and outline of the policy to the public. As a result, the closure of 16 banks in November 1997, with subsequent reversals exacerbated, rather than dampened, the crisis. Bank closures in Indonesia in April 1998, however, were more successful because they were done as part of a comprehensive strategy that was well communicated to the public and was based on the consistent application of uniform and transparent criteria.

The issue of whether a blanket guarantee, instead of the partial guarantee actually offered, should have been introduced in Indonesia in November deserves careful consideration. Our evaluation suggests that the banking crisis was not yet systemic in November, so that the partial guarantee was appropriate. In the end, the blanket guarantee introduced in January was subject to abuse and consequently raised the fiscal cost of bank restructuring. The problem in bank restructuring was more with the initial lack of a comprehensive and well-communicated strategy, and not the nature of the guarantee.

Structural conditionality

All three programs involved structural conditionality, but the experience with conditionality was very different. The Indonesian and Korean programs were characterized by extensive structural conditionality (especially the January 1998 Indonesian program) covering several areas that were not macrocritical. The scope of structural conditionality in the Brazilian program was limited to structural fiscal reform and prudential regulation. Part of this difference reflected the absence in Brazil of many of the distortions that had been present in Asia.

Measures to rehabilitate and reform the financial sector were necessary in both Indonesia and Korea and were appropriately included in the programs. In Indonesia, it was also important to tackle corporate restructuring by reforming the legal system, but this element was missing in the first two programs. As for the various nonfinancial structural reform measures included in the Indonesian and Korean programs, many of these may have been beneficial in improving longrun economic efficiency, but they were not necessary.

In Indonesia, many governance-related measures were included in the January 1998 program at the urging of some of the IMF’s major shareholders in the belief that confidence could only be restored by signaling a clean break with the past. However, the evaluation suggests that the proliferation of nonfinancial structural conditionality led to a loss of focus on critical reforms in the banking sector which was more important for restoring stability. Proliferation of structural conditionality may also have led to lack of ownership at the highest political level and nonimplementation, both of which damaged confidence.

Communications strategy

A program for restoring confidence must include a strategy to communicate the logic of the program to the public and the markets, in order to enhance country ownership and credibility. None of the three programs initially contained such a strategy.

Effective public communications are essential to build broad support for the program. Likewise, effective dialogue with the markets would improve program design through understanding the expectations of market participants, and also help build credibility for the program. For this purpose, it is important for the IMF to explain clearly the logic and strategy of the program, including spelling out the major risks, with a broad indication of how policies would respond to them.

Internal IMF Governance and the Mode of Operations

The evaluation identified a number of weaknesses in the IMF’s internal governance and mode of operations. In the area of human resource management practice, the effectiveness of surveillance was reduced by the lack of sufficient internal incentives to make judgments that were frank and potentially unpopular (with country authorities), resulting in a tendency for sharper elements of a diagnosis to be diluted in final Executive Board papers. In crisis management, the quality of the IMF’s response was compromised by a delay in the reallocation of staff resources to the Asia and Pacific Department (APD) whose staff was overstretched by multiple regional crises; the insufficient integration of staff from the Monetary and Exchange Affairs Department (MAE) and the area department; insufficient utilization of available internal knowledge; and the failure to mobilize staff members with up-to-date country knowledge.

The role of the Executive Board and the IMF’s major shareholders was particularly prominent during the crises, when major decisions needed to be made quickly, calling for close collaboration with staff and management. While the close involvement of the Board and the major shareholders was proper and necessary, close contacts at multiple layers unnecessarily subjected staff to micromanagement and political pressure, contributing to a blurring of technical and political judgments. For example, the visible presence of major country officials close to the IMF negotiating teams sometimes created a misperception of the motives behind IMF involvement, thus weakening the sense of country ownership.

In all three programs, the IMF collaborated, both in financing and technical work, with other international financial institutions (IFIs). When there was a clear separation of responsibilities, as in Brazil, no major problems occurred. In Asia, however, where the IMF and the other IFIs all worked in the financial sector, tensions developed over the role they should play in an IMF-supported program. While a good working relationship eventually developed, it depended too much on personalities, and not on a well-defined procedure. Moreover, existing procedures to resolve differences of view between the IMF and the World Bank on key policy matters were not effective in avoiding public criticism by the Chief Economist of the World Bank; indeed, as far as the evaluation team can tell, these procedures were not utilized.


Since these crises, the IMF has taken numerous initiatives to strengthen surveillance and program design. Many of the weaknesses in surveillance and program design identified by the evaluation have already been addressed by the IMF in its revised policies and procedures. Nevertheless, additional steps will be necessary to further enhance the effectiveness of the IMF in surveillance and crisis management. We make six broad recommendations, which are set out in the final chapter of the report along with their rationale. Rather than summarize them again here, we suggest that Chapter 6 be read in conjunction with this Executive Summary.

    Other Resources Citing This Publication