One of the fundamental purposes of the IMF is to make its resources temporarily available to members experiencing balance of payments difficulties, easing the required balance of payments adjustment by attenuating it, and helping to “give confidence” by reconstituting gross international reserves. In a number of capital account crises, however, the magnitude and abruptness of the capital outflows has dwarfed available official financing, resulting in much sharper external adjustment than programmed (or than warranted by debt sustainability considerations) and significant economic dislocation.24 But even if available official financing attenuated external adjustment only to a limited extent once confidence was lost, IMF support can still help avoid the collapse of exchange rates and economic activity in the first place through crisis prevention.
This section examines the theoretical foundations and empirical evidence of the role of IMF support in preventing capital account crises and their attendant economic dislocation.25 The traditional literature on the “catalytic effects” of IMF support defines it as a multiplier effect of IMF lending on official and private capital inflows (so that, for each dollar of IMF support, the country receives much more than one dollar in total net inflows). A growing body of this literature suggests that the catalytic effects are at best small for private flows (excluding foreign direct investment (FDI)) once a capital account crisis has erupted.26 Rather than focusing on crisis resolution, this section analyzes whether IMF lending can help prevent a crisis from erupting in the first place. “Catalytic effect” in the sense used here means that one dollar of IMF support results in more than one dollar of net private inflows relative to the counterfactual in which private lenders would have exited. Although this is not the same sense in which the term has normally been used in the literature, it is of particular importance when considering crisis prevention.
At a theoretical level, IMF support of the authorities’ economic program may help stave off a crisis in four ways: (1) by improving policies; (2) by providing a means (namely, program conditionality) of solving time-inconsistency problems; (3) by signaling these better policies and demonstrating the authorities’ continued commitment to them; and (4) by augmenting liquidity. These channels are not independent of each other. For instance, the availability of conditional IMF financing may induce stronger policies; the strength of the IMF’s “seal of approval” signal to markets may be enhanced by the IMF putting its own resources on the line and the authorities demonstrating their commitment through compliance with program conditionality; and IMF financing may contribute to crisis prevention by improving various vulnerability indicators (such as reserves coverage of short-term debt).
Such bundling of adjustment, liquidity, and the credibility effects of IMF support is in fact a key feature of theoretical models and the empirical analysis below. While the theoretical literature on the IMF’s crisis prevention role is still largely in its infancy, it suggests a number of insights. First, an increase in the alternative rate of return available to investors (such as a rise in U.S. interest rates), or a reduction in the willingness of the country to undertake adjustment, can leave borrowing countries—especially those with high levels of indebtedness—more vulnerable to a crisis. Second, conditional IMF resources are especially useful for crisis prevention since they both enhance the country’s liquidity position and elicit a greater adjustment effort (stronger policies). The value of IMF support in crisis prevention, therefore, goes beyond the liquidity effects of its lending resources. Third, IMF support is most useful for reducing the likelihood of a crisis when the country faces higher adjustment costs, but its fundamentals are not so weak that solvency considerations make a crisis unavoidable. Finally, the credibility of the signal to the markets is enhanced by the IMF putting its own resources on the line.
Empirical evidence that IMF support may help crisis prevention is necessarily elusive. Beyond the inherent difficulties of identifying empirical regularities from a limited number of capital account crises, finding an effect of IMF support on crisis prevention depends on being able to establish the counterfactual scenario in which the country was at risk of a crisis, and then showing that IMF support lowered the crisis likelihood. This section identifies episodes of high “market pressure” based on the behavior of foreign exchange reserves, real exchange rates, and sovereign bond spreads using cluster analysis in a panel data set of 27 emerging market countries over the period 1994–2004. This analysis yields 32 episodes of high market pressure that are then categorized either as capital account crises or as control group cases (i.e., instances where the crisis was avoided despite intense market pressures). For this purpose, a capital account crisis is defined as a high market pressure event followed by at least two quarters of medium or high capital outflows over the next four quarters. This definition yields a list of 11 capital account crises, which corresponds closely to most commonly recognized cases,27 while the remaining 21 cases are classified as the control group. Both groups—crisis and control—have episodes with and without IMF arrangements.
The empirical research examines whether these high market pressure episodes turn into a capital account crisis based upon the country’s fundamentals, including its policies and the availability of IMF financing in the run-up to the high market pressure episode. The econometric analysis suggests that—controlling for other factors—IMF support can indeed lower the likelihood of a crisis, confirming a role for IMF-supported programs in crisis prevention. Three aspects of the empirical results are noteworthy. First, IMF disbursements (over the preceding four quarters)—or their availability under an on-track precautionary program—lower the likelihood of a crisis beyond any purely signaling effects of IMF support of the authorities’ program; in other words, “money matters” (as does implementation of the agreed policies). Second, IMF support lowers the likelihood of a crisis even controlling for the country’s foreign exchange reserves as supplemented by the IMF. In other words, it is not just money that matters—beyond any liquidity effects, stronger policies and their credibility, as evidenced by the IMF’s financial support, are also important. Third, economic fundamentals (including policies) are vital for crisis prevention. When fundamentals (including policies) are very weak, not only is the country starting from a high probability of a crisis, but the marginal effect of an IMF-supported program on lowering the crisis probability is also small. Therefore, unless complemented by substantially stronger policies, extremely large amounts of IMF financing would be required to help avert a crisis. In contrast, when fundamentals are very strong, IMF support further lowers the likelihood of a crisis, though this probability is already low to begin with. It is thus for an intermediate range of fundamentals that an IMF-supported program as a tool of crisis prevention becomes especially interesting, sharply reducing the likelihood of a crisis. In at least some of these cases, IMF financing had an appreciable impact on lowering the crisis probability and indeed the country was able to avert a crisis despite the high market pressure episode.
Theoretical Considerations
Conceptually, an IMF-supported program could help prevent crises in four ways: by providing readily available foreign exchange reserves, which gives confidence and reduces the likelihood of a liquidity run; by supporting stronger policies; by signaling these policies; and by enhancing their credibility via the conditionality underpinning IMF-supported programs. While the theoretical literature specifically on the role of IMF support in crisis prevention is still largely in its infancy, the broader literature on currency crises, as well as several recent studies, can provide some useful insights.28 In the typical setting considered by this literature, the country has short-term liabilities (short-term debt on a residual maturity basis or, in currency crisis models, the outstanding money stock) that are held by atomistic private agents. Since private creditors face a coordination problem, a liquidity crisis (or “run”) can occur even if solvency is not in question, with the likelihood depending positively on the alternative rate of return available to investors (e.g., U.S. interest rates) and negatively on the country’s foreign exchange reserves. Given costs of acquiring and holding reserves, the country has a desired level of reserves that trades off these costs against the probability (and associated economic disruption) of a crisis.29 At any given moment, however, the country may find itself with a lower level of reserves than desirable, for instance, because a shock has widened the current account deficit, depleting some of its reserves.30 Faced by this situation, national authorities would want to undertake at least some adjustment but, inasmuch as adjustment is costly, not necessarily enough to fully replenish reserves immediately—leaving the country in a state of heightened vulnerability.
How can IMF support lower the likelihood of a crisis? Most obviously, by providing—or, under a precautionary arrangement, making available—foreign exchange reserves that enhance the country’s liquidity position and supporting stronger economic policies that together give confidence and reduce the likelihood of a run. But the theoretical literature provides two further key insights. First, if IMF resources are provided unconditionally, then given costs of adjustment, the authorities might relax their macroeconomic policy stance relative to the situation in which there was no IMF lending.31 In other words, there is a risk of “debtor moral hazard” such that part of the benefit of the additional liquidity is offset by a weaker adjustment effort, and a dollar of unconditional liquidity support results in less than a corresponding increase in the country’s foreign exchange reserves. Second, conditional IMF resources can support more adjustment—and stronger policies more generally—than otherwise would be implemented. Thus a dollar of IMF support results in more than a dollar’s increase in the country’s holding of reserves, with a correspondingly greater reduction in the probability of a crisis. It bears emphasizing that, since national authorities always have the option of not seeking the IMF’s support, the stronger policies and financing provided under the program must be welfare enhancing for the member re4lative to a no-program situation.32
In this regard, conditionality plays a crucial role of providing mutual assurances. Since the member may undertake less adjustment without conditionality (and the benefit of a lower crisis likelihood associated with IMF financing), conditionality provides the member the assurance that the disbursements will be forthcoming as long as policies are implemented. By the same token, the IMF is assured that the country will indeed undertake sound economic policies as it disburses its resources.33 Finally, to help prevent a liquidity run, the private sector needs to be confident that the country will undertake requisite economic policies and have available the IMF resources if necessary.
In sum, economic theory points to a number of results. First, borrowing countries become more vulnerable to a crisis when world interest rates rise or when the adjustment costs are high. Second, IMF support may lower the likelihood of a crisis through a combination of increasing liquidity and promoting stronger policies. This implies that an IMF-supported program has an effect on crisis prevention beyond the pure liquidity effects of the gross international reserves it provides. While “money matters,” it is not only money that matters for crisis prevention—policies matter too. Third, an IMF-supported program is most effective in lowering the likelihood of a crisis when the country faces higher adjustment costs, but its fundamentals are also not so poor that it is insolvent. Finally, under certain circumstances, the strength of the IMF’s “seal of approval” signal—and hence the impact on lowering the crisis likelihood—is enhanced by the IMF willingness to commit its own resources to assist a member facing vulnerabilities.34
Empirical Analysis
Economic theory points to a number of ways in which IMF support can reduce the likelihood of a financial crisis. But is there evidence in practice? By its very nature, the effects of an IMF-supported program in crisis prevention are likely to be difficult to detect. The analysis here proceeds in three steps.35 First, identifying episodes of heightened vulnerability; and second, classifying these episodes as either leading to a capital account crisis or in the control group where the crisis was averted. This, in essence, forms the dataset for the third step, which uses a logit specification to establish whether an IMF-supported program prior to the emergence of market pressures played a role in determining the outcome—crisis or no crisis—in the episode of heightened vulnerability.
Identifying Market Pressure Episodes and Classifying Outcomes
In order to identify episodes of heightened vulnerability, an index of “exchange market pressures” is defined as the average of the decline in foreign exchange reserves, the real exchange rate depreciation, and the increase of the sovereign bond spread in secondary markets.36 An increase in this index thus captures a weakening balance of payments position and difficulties in attracting capital inflows. This monthly index is created for a sample of 27 emerging market countries over the period 1994–2004.37 Cluster analysis is applied to this panel dataset to classify observations into one of five clusters according to the severity of the exchange market pressures facing the country. The technique avoids setting ad hoc thresholds, in effect assigning each observation to the appropriate cluster based on characteristics of the data rather than on subjective judgments.38 Since the focus of this analysis is on crisis prevention in the context of weakening balance of payments and a slowdown in capital inflows, “high market pressure” episodes are defined as those in cluster 1, which contains 32 observations.
The second step involves segmenting these 32 episodes into cases of capital account crises and the control group—that is, high market pressure events that did not turn into a capital account crisis. This is determined based on the behavior of net private capital flows (excluding FDI, as a percentage of GDP). To this end, cluster analysis is applied to quarterly data on net capital flows using a total of five clusters, ranging from net large inflows to net large outflows. A capital account crisis is defined as at least two quarters (for persistence) of medium or large net capital outflows during the four quarters immediately following the onset of the market pressure event.39
Again, the advantage of a data-driven approach is that it avoids ad hoc judgments about what constitutes a “capital account crisis.” Nevertheless, the resulting 11 capital account crises correspond closely to most widely accepted lists of capital account crises, including the Asian crisis countries in 1997, Brazil and Russia in 1998, Turkey in 2000, Argentina in 2001, and Uruguay in 2002 (Table 4.1). The one exception is Mexico (1994), which the procedure classifies in the control group, mainly because the net capital outflows, while large, were not sufficiently persistent; however, the main empirical findings are robust to reclassifying Mexico (1994) as a capital account crisis rather than in the control group.
Classification of Capital Account Crisis and Control Group Episodes1
The classification into capital account crisis and control group episodes is as follows: (1) a capital account crisis event requires two quarters of either medium outflows or high outflows in the four quarters that follow the buildup of market pressures; and (2) all other episodes in the control group private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Market pressures identified by classifying monthly data into five clusters based on an index of market pressures that includes changes in the real effective exchange rate, foreign exchange reserves, and spreads. The listed countries are in the cluster with the highest market pressures.
Private capital flows (net of FDI) are used for distinguishing between capital account crisis and control group episodes. Specifically, private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Numbers of months from the beginning to the end of each market pressure episode.
Classification of Capital Account Crisis and Control Group Episodes1
| Identifying Market Pressures2 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Episode | Country | Beginning date of market pressures |
End date of market pressures |
Duration of pressures (In months)4 |
Number of months with pressures |
Capital Flows Clusters3 | |||||
| Period | |||||||||||
| t | t+1 | t+2 | t+3 | ||||||||
| Capital Account Crisis Episodes | |||||||||||
| 1 | Argentina | 2001 | July | 2002 | May | 11 | 6 | 4 | 3 | 4 | 4 |
| 2 | Brazil | 1998 | August | 1999 | January | 6 | 3 | 4 | 3 | 4 | 2 |
| 3 | Bulgaria | 1996 | May | 1996 | May | 1 | 1 | 4 | 4 | 3 | 4 |
| 4 | Ecuador | 2000 | January | 2000 | January | 1 | 1 | 4 | 5 | 5 | 3 |
| 5 | Indonesia | 1997 | October | 1998 | January | 4 | 3 | 4 | 5 | 3 | 4 |
| 6 | Korea | 1997 | October | 1997 | December | 3 | 3 | 4 | 4 | 2 | 3 |
| 7 | Malaysia | 1997 | July | 1998 | January | 7 | 5 | 5 | 2 | 4 | 3 |
| 8 | Russia | 1998 | August | 1998 | September | 2 | 2 | 4 | 4 | 4 | 2 |
| 9 | Thailand | 1997 | July | 1997 | August | 2 | 2 | 4 | 5 | 5 | 4 |
| 10 | Turkey | 2000 | November | 2001 | March | 5 | 3 | 3 | 4 | 4 | 3 |
| 11 | Uruguay | 2002 | July | 2002 | July | 1 | 1 | 5 | 2 | 2 | 5 |
| Control Group Episodes | |||||||||||
| 1 | Argentina | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 2 | 4 |
| 2 | Brazil | 2002 | July | 2002 | July | 1 | 1 | 4 | 3 | 2 | 2 |
| 3 | Bulgaria | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 2 | 2 |
| 4 | Chile | 1999 | June | 1999 | June | 1 | 1 | 5 | 2 | 2 | 2 |
| 5 | Chile | 2002 | June | 2002 | June | 1 | 1 | 2 | 2 | 2 | 2 |
| 6 | Colombia | 1998 | April | 1998 | September | 6 | 3 | 2 | 2 | 3 | 2 |
| 7 | Colombia | 2002 | July | 2002 | August | 2 | 2 | 2 | 2 | 3 | 3 |
| 8 | Hungary | 2003 | June | 2003 | June | 1 | 1 | 2 | 1 | 3 | 1 |
| 9 | Indonesia | 2004 | January | 2004 | January | 1 | 1 | 2 | 2 | 2 | 2 |
| 10 | Mexico | 1994 | December | 1995 | March | 4 | 3 | 3 | 4 | 3 | 3 |
| 11 | Mexico | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 3 | 2 |
| 12 | Peru | 1998 | August | 1998 | December | 5 | 2 | 2 | 3 | 3 | 3 |
| 13 | Philippines | 1997 | August | 1998 | August | 1 | 1 | 1 | 3 | 2 | 1 |
| 14 | Poland | 1998 | August | 1998 | August | 1 | 1 | 2 | 3 | 2 | 2 |
| 15 | South Africa | 1996 | April | 1996 | April | 1 | 1 | 2 | 2 | 2 | 2 |
| 16 | South Africa | 1998 | July | 1998 | July | 1 | 1 | 2 | 2 | 2 | 1 |
| 17 | South Africa | 2001 | December | 2001 | December | 1 | 1 | 2 | 3 | 2 | 3 |
| 18 | Turkey | 1998 | August | 1998 | August | 1 | 1 | 4 | 1 | 2 | 2 |
| 19 | Venezuela | 1994 | June | 1994 | June | 1 | 1 | 4 | 1 | 3 | 3 |
| 20 | Venezuela | 1998 | August | 1998 | August | 1 | 1 | 3 | 3 | 3 | 3 |
| 21 | Venezuela | 2003 | January | 2003 | January | 1 | 1 | 3 | 3 | 4 | 3 |
The classification into capital account crisis and control group episodes is as follows: (1) a capital account crisis event requires two quarters of either medium outflows or high outflows in the four quarters that follow the buildup of market pressures; and (2) all other episodes in the control group private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Market pressures identified by classifying monthly data into five clusters based on an index of market pressures that includes changes in the real effective exchange rate, foreign exchange reserves, and spreads. The listed countries are in the cluster with the highest market pressures.
Private capital flows (net of FDI) are used for distinguishing between capital account crisis and control group episodes. Specifically, private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Numbers of months from the beginning to the end of each market pressure episode.
Classification of Capital Account Crisis and Control Group Episodes1
| Identifying Market Pressures2 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Episode | Country | Beginning date of market pressures |
End date of market pressures |
Duration of pressures (In months)4 |
Number of months with pressures |
Capital Flows Clusters3 | |||||
| Period | |||||||||||
| t | t+1 | t+2 | t+3 | ||||||||
| Capital Account Crisis Episodes | |||||||||||
| 1 | Argentina | 2001 | July | 2002 | May | 11 | 6 | 4 | 3 | 4 | 4 |
| 2 | Brazil | 1998 | August | 1999 | January | 6 | 3 | 4 | 3 | 4 | 2 |
| 3 | Bulgaria | 1996 | May | 1996 | May | 1 | 1 | 4 | 4 | 3 | 4 |
| 4 | Ecuador | 2000 | January | 2000 | January | 1 | 1 | 4 | 5 | 5 | 3 |
| 5 | Indonesia | 1997 | October | 1998 | January | 4 | 3 | 4 | 5 | 3 | 4 |
| 6 | Korea | 1997 | October | 1997 | December | 3 | 3 | 4 | 4 | 2 | 3 |
| 7 | Malaysia | 1997 | July | 1998 | January | 7 | 5 | 5 | 2 | 4 | 3 |
| 8 | Russia | 1998 | August | 1998 | September | 2 | 2 | 4 | 4 | 4 | 2 |
| 9 | Thailand | 1997 | July | 1997 | August | 2 | 2 | 4 | 5 | 5 | 4 |
| 10 | Turkey | 2000 | November | 2001 | March | 5 | 3 | 3 | 4 | 4 | 3 |
| 11 | Uruguay | 2002 | July | 2002 | July | 1 | 1 | 5 | 2 | 2 | 5 |
| Control Group Episodes | |||||||||||
| 1 | Argentina | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 2 | 4 |
| 2 | Brazil | 2002 | July | 2002 | July | 1 | 1 | 4 | 3 | 2 | 2 |
| 3 | Bulgaria | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 2 | 2 |
| 4 | Chile | 1999 | June | 1999 | June | 1 | 1 | 5 | 2 | 2 | 2 |
| 5 | Chile | 2002 | June | 2002 | June | 1 | 1 | 2 | 2 | 2 | 2 |
| 6 | Colombia | 1998 | April | 1998 | September | 6 | 3 | 2 | 2 | 3 | 2 |
| 7 | Colombia | 2002 | July | 2002 | August | 2 | 2 | 2 | 2 | 3 | 3 |
| 8 | Hungary | 2003 | June | 2003 | June | 1 | 1 | 2 | 1 | 3 | 1 |
| 9 | Indonesia | 2004 | January | 2004 | January | 1 | 1 | 2 | 2 | 2 | 2 |
| 10 | Mexico | 1994 | December | 1995 | March | 4 | 3 | 3 | 4 | 3 | 3 |
| 11 | Mexico | 1998 | August | 1998 | August | 1 | 1 | 2 | 2 | 3 | 2 |
| 12 | Peru | 1998 | August | 1998 | December | 5 | 2 | 2 | 3 | 3 | 3 |
| 13 | Philippines | 1997 | August | 1998 | August | 1 | 1 | 1 | 3 | 2 | 1 |
| 14 | Poland | 1998 | August | 1998 | August | 1 | 1 | 2 | 3 | 2 | 2 |
| 15 | South Africa | 1996 | April | 1996 | April | 1 | 1 | 2 | 2 | 2 | 2 |
| 16 | South Africa | 1998 | July | 1998 | July | 1 | 1 | 2 | 2 | 2 | 1 |
| 17 | South Africa | 2001 | December | 2001 | December | 1 | 1 | 2 | 3 | 2 | 3 |
| 18 | Turkey | 1998 | August | 1998 | August | 1 | 1 | 4 | 1 | 2 | 2 |
| 19 | Venezuela | 1994 | June | 1994 | June | 1 | 1 | 4 | 1 | 3 | 3 |
| 20 | Venezuela | 1998 | August | 1998 | August | 1 | 1 | 3 | 3 | 3 | 3 |
| 21 | Venezuela | 2003 | January | 2003 | January | 1 | 1 | 3 | 3 | 4 | 3 |
The classification into capital account crisis and control group episodes is as follows: (1) a capital account crisis event requires two quarters of either medium outflows or high outflows in the four quarters that follow the buildup of market pressures; and (2) all other episodes in the control group private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Market pressures identified by classifying monthly data into five clusters based on an index of market pressures that includes changes in the real effective exchange rate, foreign exchange reserves, and spreads. The listed countries are in the cluster with the highest market pressures.
Private capital flows (net of FDI) are used for distinguishing between capital account crisis and control group episodes. Specifically, private capital flows (net of FDI) are classified into five clusters: high inflows, medium inflows, average flows, medium outflows, and high outflows.
Numbers of months from the beginning to the end of each market pressure episode.
Market Pressures and the Determinants of Crises
So what determines whether a high market pressure episode turns into a capital account crisis? Before turning to the formal analysis, Figures Figure4.1 and Figure4.2 contrast the behavior of key macroeconomic variables for the crisis and control groups. Of course, once the crisis does or does not erupt, the behavior of these variables is likely to be quite different; of greater interest, therefore, are the differences between the groups in the run-up (quarters t−4 to t−1) to the high market pressure episode.
Selected Economic Indicators: Medians for Capital Account Crisis and Control Group Countries1
Sources: IMF, World Economic Outlook and International Financial Statistics databases; CEIC database; Emerging Markets Data Base (EMDB); and IMF staff estimates.1 A total of 11 capital account crises and 21 control group episodes are included.2Measured as the difference between actual real effective exchange rate and the Hodrick-Prescott filter.Selected External and Policy Indicators: Medians for Capital Account Crisis and Control Group Countries1
Sources: IMF, World Economic Outlook and International Financial Statistics databases; CEIC database; Emerging Markets Data Base (EMDB); and IMF staff estimates.1 A total of 11 capital account crisis and 21 control group episodes are included.From the figures, crisis countries tend to have only marginally larger current account deficits, and both groups have a trend of declining external deficits, most likely indicating slowing economic growth and diminishing net capital inflows. Crisis countries also tend to have a more overvalued real exchange rate. Output growth is weaker and slowing in the group that eventually suffers a capital account crisis, though inflation rates are quite similar between both groups.
Perhaps more important, the crisis group has a higher level of external indebtedness (around 10 percentage points of GDP higher) and a higher ratio of short-term debt to reserves. The differences in fiscal performance between the two groups are sharper than those for the external balance, with the crisis countries having a weaker fiscal position except for a tightening in period t−1, which might reflect a late effort to prevent a crisis. (An alternative explanation is simply that the seasonal patterns in the fiscal and external sectors might be different, which could partly explain the volatility in fiscal balances.) Countries that subsequently suffered a capital account crisis also tend to have a somewhat higher degree of monetization (or lower ratio of GDP to broad money), implying a larger scope for capital flight.
Since the focus of the section is on crisis prevention, the regression analysis concentrates on the four quarters preceding period t (the pre-crisis period) for each of the 32 episodes of intense market pressures. A value of one is assigned to the four quarters prior to t (i.e., t−4 to t−1) when a market pressure episode develops into a capital account crisis, and zero otherwise.40 This approach allows for greater variation in the explanatory variables in the run-up to the market pressure episode than would be possible using annual data on the 32 episodes.41 The sample is relatively well balanced between crisis and control group episodes and between observations with and without an IMF-supported program (Table 4.2).
Number of Observations in Each Group
Number of Observations in Each Group
| With IMF Financing |
Without IMF Financing |
Total | |
|---|---|---|---|
| Capital account crisis episodes | 17 | 27 | 44 |
| Control group episodes | 22 | 62 | 84 |
| Total | 39 | 89 | 128 |
Number of Observations in Each Group
| With IMF Financing |
Without IMF Financing |
Total | |
|---|---|---|---|
| Capital account crisis episodes | 17 | 27 | 44 |
| Control group episodes | 22 | 62 | 84 |
| Total | 39 | 89 | 128 |
Table 4.3 reports the results of the logit estimates for alternative specifications. The regressors may be classified into four categories: initial conditions,42 policy variables, exogenous and other factors,43 and IMF financing. For the latter, as explained below, this paper uses the ratio of available IMF financing (either disbursed or accumulated drawing rights in the case of precautionary arrangements) to short-term debt in the four quarters up to each period.44 This ratio also captures the cumulative impact of implementation of economic policies during the preceding four quarters.
Logit Estimation Results1
Standard errors are adjusted for within-cluster correlation (i.e., correlation at the level of each pressure episode). Logit regressions using random effects provide similar results.
Cumulative sum of IMF financing (disbursed or available for disbursement under precautionary arrangements) relative to the short-term debt over the four quarters from j–3 to j, where j is any quarterly period between t−4 and t−1.
IMF dummy equals 1 if IMF resources were available in any of the last four quarters.
As classified under the eight-category scale of the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions; a higher score indicates a more flexible exchange rate regime.
Refers to the democratic accountability component in the International Country Risk Guide index. Higher index indicates a lower risk rating (or greater political stability).
Exchange rate overvaluation is the deviation of the real effective exchange rate from the long-term trend (Hodrick-Prescott filter).
Difference in the ratio of fiscal balance/GDP in period j over period j–4.
Change in fiscal balance interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture fiscal policy aspects of IMF-supported programs.
Treasury bill rate or other short-term rate net of inflation.
Change in interest rates interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture monetary policy aspects of IMF-supported programs.
GDP as a share of U.S. GDP.
Logit Estimation Results1
| DependentVariable: Crisis = 1, Control Group = 0 |
Regression | |||||
|---|---|---|---|---|---|---|
| R1 | R2 | R3 | R4 | |||
| IMF involvement | ||||||
| IMF financing (IMF resource ratio)2 | −37.23** | −40.25** | −40.06*** | |||
| IMF program dummy3 | −1.11 | 0.46 | ||||
| Initial conditions | ||||||
| Debt/GDP | 0.14*** | 0.18*** | 0.18*** | 0.19*** | ||
| Short-term debt/reserves | 1.06*** | 1.06*** | 1.03** | 1.12*** | ||
| Exchange rate regime4 | −0.33 | −0.56 | −0.53 | −0.67* | ||
| Political stability5 | −6.44** | −6.14 | −6.04* | −6.05* | ||
| Exchange rate overvaluation6 | 17.65*** | 25.22*** | 24.74*** | 26.29*** | ||
| Policy variables | ||||||
| Fiscal balance change7 | −0.08** | −0.03 | −0.03 | 0.23** | ||
| Fiscal balance interactive with IMF dummy8 | −0.45** | |||||
| Interest rate change (real terms)9 | −0.08** | −0.07* | −0.07* | −0.03 | ||
| Interest rate interactive with IMF dummy10 | −0.07 | |||||
| Exogenous factors | ||||||
| Terms of trade | −0.01 | 0.02 | 0.02 | −0.00 | ||
| Other | ||||||
| Size of the economy11 | 0.77*** | 0.87*** | 0.88*** | 0.93*** | ||
| Latin American dummy | −0.69 | −1.49 | −1.51 | −1.98 | ||
| Asian dummy | 2.41 | 2.45 | 2.34 | 2.36 | ||
| Constant | −5.92 | −6.76 | −6.98 | −6.77 | ||
| Number of observations | 128 | 128 | 128 | 128 | ||
| LR chi-squared | 50.7*** | 25.7*** | 30.6*** | 65.8*** | ||
| Pseudo R-squared | 0.53 | 0.59 | 0.59 | 0.61 | ||
| Correctly classified (in percent) | 83 | 87 | 87 | 87 | ||
| Type I errors (in percent) | 23 | 16 | 16 | 18 | ||
| Type II errors (in percent) | 14 | 12 | 12 | 11 | ||
Standard errors are adjusted for within-cluster correlation (i.e., correlation at the level of each pressure episode). Logit regressions using random effects provide similar results.
Cumulative sum of IMF financing (disbursed or available for disbursement under precautionary arrangements) relative to the short-term debt over the four quarters from j–3 to j, where j is any quarterly period between t−4 and t−1.
IMF dummy equals 1 if IMF resources were available in any of the last four quarters.
As classified under the eight-category scale of the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions; a higher score indicates a more flexible exchange rate regime.
Refers to the democratic accountability component in the International Country Risk Guide index. Higher index indicates a lower risk rating (or greater political stability).
Exchange rate overvaluation is the deviation of the real effective exchange rate from the long-term trend (Hodrick-Prescott filter).
Difference in the ratio of fiscal balance/GDP in period j over period j–4.
Change in fiscal balance interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture fiscal policy aspects of IMF-supported programs.
Treasury bill rate or other short-term rate net of inflation.
Change in interest rates interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture monetary policy aspects of IMF-supported programs.
GDP as a share of U.S. GDP.
Logit Estimation Results1
| DependentVariable: Crisis = 1, Control Group = 0 |
Regression | |||||
|---|---|---|---|---|---|---|
| R1 | R2 | R3 | R4 | |||
| IMF involvement | ||||||
| IMF financing (IMF resource ratio)2 | −37.23** | −40.25** | −40.06*** | |||
| IMF program dummy3 | −1.11 | 0.46 | ||||
| Initial conditions | ||||||
| Debt/GDP | 0.14*** | 0.18*** | 0.18*** | 0.19*** | ||
| Short-term debt/reserves | 1.06*** | 1.06*** | 1.03** | 1.12*** | ||
| Exchange rate regime4 | −0.33 | −0.56 | −0.53 | −0.67* | ||
| Political stability5 | −6.44** | −6.14 | −6.04* | −6.05* | ||
| Exchange rate overvaluation6 | 17.65*** | 25.22*** | 24.74*** | 26.29*** | ||
| Policy variables | ||||||
| Fiscal balance change7 | −0.08** | −0.03 | −0.03 | 0.23** | ||
| Fiscal balance interactive with IMF dummy8 | −0.45** | |||||
| Interest rate change (real terms)9 | −0.08** | −0.07* | −0.07* | −0.03 | ||
| Interest rate interactive with IMF dummy10 | −0.07 | |||||
| Exogenous factors | ||||||
| Terms of trade | −0.01 | 0.02 | 0.02 | −0.00 | ||
| Other | ||||||
| Size of the economy11 | 0.77*** | 0.87*** | 0.88*** | 0.93*** | ||
| Latin American dummy | −0.69 | −1.49 | −1.51 | −1.98 | ||
| Asian dummy | 2.41 | 2.45 | 2.34 | 2.36 | ||
| Constant | −5.92 | −6.76 | −6.98 | −6.77 | ||
| Number of observations | 128 | 128 | 128 | 128 | ||
| LR chi-squared | 50.7*** | 25.7*** | 30.6*** | 65.8*** | ||
| Pseudo R-squared | 0.53 | 0.59 | 0.59 | 0.61 | ||
| Correctly classified (in percent) | 83 | 87 | 87 | 87 | ||
| Type I errors (in percent) | 23 | 16 | 16 | 18 | ||
| Type II errors (in percent) | 14 | 12 | 12 | 11 | ||
Standard errors are adjusted for within-cluster correlation (i.e., correlation at the level of each pressure episode). Logit regressions using random effects provide similar results.
Cumulative sum of IMF financing (disbursed or available for disbursement under precautionary arrangements) relative to the short-term debt over the four quarters from j–3 to j, where j is any quarterly period between t−4 and t−1.
IMF dummy equals 1 if IMF resources were available in any of the last four quarters.
As classified under the eight-category scale of the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions; a higher score indicates a more flexible exchange rate regime.
Refers to the democratic accountability component in the International Country Risk Guide index. Higher index indicates a lower risk rating (or greater political stability).
Exchange rate overvaluation is the deviation of the real effective exchange rate from the long-term trend (Hodrick-Prescott filter).
Difference in the ratio of fiscal balance/GDP in period j over period j–4.
Change in fiscal balance interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture fiscal policy aspects of IMF-supported programs.
Treasury bill rate or other short-term rate net of inflation.
Change in interest rates interactive with IMF dummy if disbursements took place in any of the last four quarters; intended to capture monetary policy aspects of IMF-supported programs.
GDP as a share of U.S. GDP.
Most of the explanatory variables have the expected signs, though not all are statistically significant. Overall the logit regressions correctly classify 83 to 87 percent of the observations with a balanced distribution between false negative (type I) and false positive (type II) errors. A higher external debt-to-GDP ratio, a higher short-term debt-to-reserves ratio, a less flexible exchange rate regime, greater overvaluation of the real exchange rate, less political stability, or larger country size (measured as nominal GDP at market exchange rates relative to U.S. GDP) make it more likely that a high market pressure episode turns into a capital account crisis.45 Monetary or fiscal tightening is associated with a lower probability of crisis, particularly when the latter is undertaken in the context of an IMF-supported program. Moreover, consistent with the theoretical discussion, macroeconomic policies tend to be stronger in countries receiving IMF support.46
Turning to the effects of IMF-supported programs, the literature has typically used a dummy variable to indicate the existence of an IMF-supported program.47 From Table 4.3, regression R1 shows that the mere existence of an on-track IMF-supported program— including the policies required to merit the IMF’s sup-port48—does not have a statistically significant impact on reducing the likelihood of a crisis (although the point estimate of the coefficient is negative as expected).
By contrast, a variable that captures available IMF financing—defined as disbursements or accumulated drawing rights under an on-track precautionary program in the four previous quarters as a share of short-term debt—is negative and statistically significant (regression R2).49 Because the regression controls for the country’s holdings of (gross) foreign exchange reserves, IMF financing has an effect on crisis prevention beyond the liquidity contribution of IMF resources. Moreover, when both the dummy variable for an on-track program and the IMF financing variable are included (regression R3), or when monetary and fiscal policies under the program are included (regression R4), the IMF financing variable remains significant. An alternative formulation (not reported) where IMF financing is defined as the full amount of IMF resources that can be accessed over the life of the arrangement is not significant, suggesting that disbursed IMF financing (or accumulated drawing rights in the case of precautionary arrangements) is the key factor in crisis prevention.
While caution is required in trying to disentangle exactly the various channels—better policies, the signaling of these policies and of the authorities’ commitment to them, and liquidity—through which an IMF-supported program may reduce the likelihood of a crisis, taken together these results suggest that:
-
Stronger policies—tighter monetary policy (higher real interest rates) or greater fiscal adjustment (particularly in the context of an IMF-supported program)—are significantly associated with a lower crisis likelihood.
-
IMF disbursements (or accumulated drawing rights) are a significant factor in crisis prevention: the larger are the disbursed IMF resources (as a share of short-term debt), the lower is the crisis likelihood.
-
An important liquidity effect of IMF support on crisis prevention exists. IMF disbursements (or their availability for drawing under an on-track precautionary program) matter, rather than just an on-track program or possible future drawings under the arrangement.
-
The benefits of IMF support go beyond liquidity effects, however, since the IMF financing variable is significant even controlling for the country’s foreign exchange reserves. Part of the effect must thus arise from a combination of stronger policies (i.e., beyond the fiscal balance and real interest rates included in the regressions) bolstered by conditionality and the “seal of approval” implicit in IMF disbursements. Moreover, since the IMF-supported program dummy is not statistically significant, but the IMF financing variable is strongly significant, the strength and the credibility of the IMF’s signal appears to depend at least partially on the IMF putting its own resources on the line.
These findings are robust to various sample specifications, including data outliers (whether these are individual observations or specific market pressure episodes), and other technical considerations.
IMF-Supported Programs and Crisis Prevention
These results suggest that an IMF-supported program may be useful for crisis prevention, including by promoting stronger policies and by enhancing liquidity. But could such financing, in plausible amounts, have an appreciable impact on the likelihood of a crisis? This depends on the country’s “fundamentals” (the other covariates in the logit regression—such as the level of external debt, the exchange rate regime, the ratio of short-term debt to reserves, political developments, and monetary and fiscal policies) and the amount of IMF financing. But the counterfactual exercises considered below, in which either the amount of IMF financing is varied parametrically (holding policies constant) or the policy adjustment is varied (holding IMF financing constant), need to be interpreted with extreme caution because—as stressed by the theoretical discussion earlier—the country’s policy response (as well as other fundamentals) and IMF financing may be simultaneously determined.
With this important caveat in mind, Figure Figure4.3 contrasts the likelihood of a crisis with the IMF financing (and the country’s other covariates) available through the quarter preceding the high market pressure episode to the implied probability without any IMF financing. Within the group of countries that ultimately avoided a crisis, the model predicts that in the absence of IMF financing the likelihood of a crisis was over 50 percent. However, with IMF support (in the amounts actually made available), this probability was more than halved. In other cases, while the crisis probability was below 50 percent, IMF financing helped reduce this probability to negligible levels. Conversely, while the model suggests that IMF financing contributed to lower crisis probabilities in some countries that ultimately faced a capital account crisis, the model also shows that the probability of a crisis remained high nevertheless.
Probability of a Crisis With and Without IMF Financing
(Capital account crisis and control group countries receiving IMF financing at time t−1)
Source: IMF staff estimates.Overall, the results suggest that, in some instances, disbursements of IMF resources have had an appreciable impact in lowering the likelihood that a high market pressure event would turn into a crisis. In fact, on average, for countries that received IMF support and averted a crisis, the reduction in the likelihood of a crisis associated with IMF financing was 20 percentage points. Moreover, the welfare gains are not negligible, with some simple back-of-the-envelope calculations putting the expected welfare gain from this average lower probability of a crisis at some 5 percent of GDP for these countries.50
As emphasized above, an important contribution of the IMF-supported program in crisis prevention is to promote stronger policies. It is therefore interesting to explore the nature of the policy strengthening needed to achieve a similar reduction in the likelihood of a crisis (i.e., a 20 percent reduction). Parametrically varying policies in the estimated regression (R4), but now keeping constant the available IMF financing, shows that a combined fiscal adjustment of about 4.5 percent of GDP and higher real interest rates of about 4.5 percentage points would be required. This highlights the difficulties of avoiding a crisis through such policies alone once the other covariates have made the country vulnerable. By contrast, reducing the overvalued exchange rate by about 6 percentage points would achieve a similar reduction in crisis probability. This result underscores the importance of avoiding overvalued exchange rates and of maintaining adequate reserve to short-term debt cover.51
Even if the country’s other fundamentals (including its policies) do not change as a result of changes in IMF financing, the marginal effect of that financing depends upon the average level of those fundamentals. By way of illustration, Figure Figure4.4 shows how the probability of a crisis as function of IMF financing (in percent of short-term debt) varies with different levels of the country’s fundamentals among countries that ended up with a capital account crisis. For example, taking the case with the best fundamentals (the lowest curve in Figure 4.4), IMF financing in the amount of 5 percent of short-term debt (equivalent to one and one-half times the average amount of financing provided (through t−1) under I M F-supported programs) would have roughly halved the estimated crisis probability from 70 percent to 35 percent. For the median case, a similar increase in financing would have lowered the crisis probability from around 95 percent to about 80 percent, while for the worst case (the highest curve in Figure 4.4), the much worse fundamentals mean that increased financing would have had a negligible impact on reducing the likelihood of a crisis. These curves underscore that when fundamentals are weak, not only is the country at high risk of a crisis, but the marginal effect of IMF financing on lowering the crisis probability is also small.
Marginal Impact of IMF Financing, Given Country Fundamentals1
Source: IMF staff estimates.1 Based on regression 4 in Table 4.3. IMF financing is defined as the cumulative disbursements over 12 months as a share of short-term debt. The figure reflects the probability of a crisis for different countries based on the covariate contributions at time t−1. Vertical lines are also measured at t−1 and represent, respectively, the average and maximum levels of IMF financing among crisis episodes.It is also possible to ask the amount of IMF financing required to lower the likelihood of a crisis (from an average probability of 0.85 in t−1 for the capital account crisis countries) to some “acceptable threshold.” Using 25 percent as an illustrative threshold, Table 4.4 calculates the requisite IMF financing (actually provided or additionally required); in most cases, this amounts to the equivalent of 5 to 20 percent of the country’s short-term debt. While this would require exceptional access (i.e., more than 100 percent of quota on an annual basis)—typically in the order of 300–350 percent of quota52—these amounts are not out of line with the financial resources subsequently provided in some capital account crises, suggesting that the crisis might have been averted had there been an IMF-supported program with adequate financing in place prior to its onset. However, as stressed above, the availability of additional IMF financing could itself alter the country’s fundamentals, including its policy response. Moreover, this counterfactual exercise would also need to take account of the deterioration in the other fundamentals typically observed as the country enters a crisis.53 Accordingly, one cannot necessarily conclude that such an IMF-supported program would have averted the crisis.
IMF Financing Relative to IMF Quota Among Capital Account Crisis Countries1
Based on quotas of member countries at the time of the crisis using regression 4 of Table 4.3.
Cumulative total for the four quarters prior to t, in billions of U.S. dollars.
Refers to the total IMF disbursement (actual plus additional) that would have been required to lower the probability of a crisis to 25 percent. The model would have no type I errors at a cutoff probability of 8 percent.
IMF Financing Relative to IMF Quota Among Capital Account Crisis Countries1
| Capital Account Crisis Group |
Period t−1 | Actual IMF Financing at t−1 (In billions of U.S. dollars)2 |
Actual Probability of a Crisis at t−1 (Including IMF financing) |
Additional IMF Financing Needed to Reduce P (Crisis) to 0.25 (In billions of U.S. dollars) |
Short-Term Debt (In billions of U.S. dollars) |
IMF Financing (at P=0.25)/Quota (In percent)3 |
|---|---|---|---|---|---|---|
| Argentina | 2001 Q2 | 4.69 | 0.71 | 2.0 | 40 | 227 |
| Brazil | 1998 Q2 | 0.00 | 0.96 | 8.5 | 79 | 281 |
| Bulgaria | 1996 Q1 | 0.00 | 0.86 | 0.1 | 2 | 20 |
| Ecuador | 1999 Q4 | 0.00 | 0.79 | 0.1 | 2 | 35 |
| Indonesia | 1997 Q3 | 0.00 | 1.00 | 9.0 | 36 | 435 |
| Korea | 1997 Q3 | 0.00 | 0.94 | 7.7 | 81 | 694 |
| Malaysia | 1997 Q2 | 0.00 | 0.74 | 0.7 | 14 | 64 |
| Russia | 1998 Q2 | 2.02 | 0.86 | 1.9 | 27 | 66 |
| Thailand | 1997 Q2 | 0.00 | 1.00 | 9.9 | 51 | 1,238 |
| Turkey | 2000 Q3 | 1.39 | 0.68 | 1.9 | 42 | 248 |
| Uruguay | 2002 Q2 | 0.74 | 0.99 | 1.3 | 10 | 483 |
Based on quotas of member countries at the time of the crisis using regression 4 of Table 4.3.
Cumulative total for the four quarters prior to t, in billions of U.S. dollars.
Refers to the total IMF disbursement (actual plus additional) that would have been required to lower the probability of a crisis to 25 percent. The model would have no type I errors at a cutoff probability of 8 percent.
IMF Financing Relative to IMF Quota Among Capital Account Crisis Countries1
| Capital Account Crisis Group |
Period t−1 | Actual IMF Financing at t−1 (In billions of U.S. dollars)2 |
Actual Probability of a Crisis at t−1 (Including IMF financing) |
Additional IMF Financing Needed to Reduce P (Crisis) to 0.25 (In billions of U.S. dollars) |
Short-Term Debt (In billions of U.S. dollars) |
IMF Financing (at P=0.25)/Quota (In percent)3 |
|---|---|---|---|---|---|---|
| Argentina | 2001 Q2 | 4.69 | 0.71 | 2.0 | 40 | 227 |
| Brazil | 1998 Q2 | 0.00 | 0.96 | 8.5 | 79 | 281 |
| Bulgaria | 1996 Q1 | 0.00 | 0.86 | 0.1 | 2 | 20 |
| Ecuador | 1999 Q4 | 0.00 | 0.79 | 0.1 | 2 | 35 |
| Indonesia | 1997 Q3 | 0.00 | 1.00 | 9.0 | 36 | 435 |
| Korea | 1997 Q3 | 0.00 | 0.94 | 7.7 | 81 | 694 |
| Malaysia | 1997 Q2 | 0.00 | 0.74 | 0.7 | 14 | 64 |
| Russia | 1998 Q2 | 2.02 | 0.86 | 1.9 | 27 | 66 |
| Thailand | 1997 Q2 | 0.00 | 1.00 | 9.9 | 51 | 1,238 |
| Turkey | 2000 Q3 | 1.39 | 0.68 | 1.9 | 42 | 248 |
| Uruguay | 2002 Q2 | 0.74 | 0.99 | 1.3 | 10 | 483 |
Based on quotas of member countries at the time of the crisis using regression 4 of Table 4.3.
Cumulative total for the four quarters prior to t, in billions of U.S. dollars.
Refers to the total IMF disbursement (actual plus additional) that would have been required to lower the probability of a crisis to 25 percent. The model would have no type I errors at a cutoff probability of 8 percent.
Conclusions
Although the IMF can contribute to crisis prevention in many ways, including through its surveillance work and provision of technical assistance, this paper has focused on IMF-supported programs for crisis prevention. The theoretical literature suggests that an IMF-supported program can contribute to a lower likelihood of a crisis both by providing the member additional liquidity—making a run for the exit by private creditors less likely—and by inducing stronger policies that are supported by conditionality.
Empirically, the evidence suggests that IMF-supported programs can indeed play an important role in crisis prevention. In particular, at times of heightened uncertainty, when there are incipient market pressures, IMF resources (either disbursements or their availability under on-track precautionary arrangements) and associated stronger policies can lower the likelihood that a crisis will develop. The effect of IMF resources on lowering the crisis probability goes beyond the pure liquidity benefits of unconditional resources (the country’s foreign exchange reserves) and reflects both better policies and the stronger market signal that IMF financing elicits. At the same time, success of an IMF-supported program in preventing a crisis depends on the fundamentals. As fundamentals worsen, the marginal benefit of an IMF-supported program diminishes.
Although preliminary, and subject to the various caveats noted above, these findings carry important implications for the possible role of the IMF in crisis prevention in emerging market countries. But they also raise questions about the design of “crisis prevention” programs. In particular, the results suggest that, while “money matters,” it is not just the money that matters—policies matter as well. In crisis prevention situations—where the country is not forced to adjust by the withdrawal of private finance—the benefits of greater unconditional liquidity may be offset by “debtor moral hazard.”
Therefore crisis prevention programs would typically need to be supported by conditionality to help ensure that disbursements (or rights to drawings under precautionary arrangements) enhance liquidity and support sound crisis prevention policies. But if they do so, both theory and empirical evidence suggest that IMF financial support—either disbursed or provided contingently—can play an important role in preventing financial crises in emerging market countries.
