Chapter

III Toward Crisis Prevention: Precautionary Arrangements

Author(s):
Atish Ghosh, Juan Zalduendo, Alun Thomas, Jun Kim, Uma Ramakrishnan, and Bikas Joshi
Published Date:
May 2008
Share
  • ShareShare
Show Summary Details

The discussion in the previous section suggests where balance sheet vulnerabilities might lurk and how they may interact with specific triggers that result in a full-blown crisis. The first step in crisis prevention is to try to avoid such vulnerabilities—in particular, to ensure that the government is not (perhaps inadvertently) providing incentives that exacerbate balance sheet mismatches. It is a truism that sound macroeconomic policies also lessen—but do not eliminate—the possibility that a crisis will be triggered.

What can the IMF do to help prevent crises? Surveillance is certainly at the heart of any response in this regard. Regular “vulnerability exercises” conducted within the IMF seek to give early warning of possible external imbalances. The use of the Financial Sector Assessment Program as well as Reports on Standards and Codes helps improve financial sector surveillance and adherence to international standards, while Article IV consultations are increasingly integrating financial sector issues in discussions with national authorities. Additionally, greater emphasis on transparency, including publication of IMF documents and subscription to the Special Data Dissemination Standard, facilitates the flow of timely information to the market, perhaps limiting adverse self-fulfilling expectations. Debt sustainability assessments—required of all Article IV consultation reports—provide a consistency check on baseline medium-term projections, and further identify possible medium-term vulnerabilities.

Beyond its surveillance activities and provision of technical assistance, the IMF can engage with member countries through program support. When a member seeks an IMF-supported program, but does not face a pressing balance of payments need, it may treat an IMF arrangement as precautionary—a pure Stand-By Arrangement—which provides the right, conditional on implementation of specific policies, to make drawings should the need arise.8 Countries achieving broad macroeconomic stability and external viability have found precautionary arrangements to be useful—accounting for nearly half of new arrangements in recent years. The characteristics of precautionary programs are examined below.

One concern, however, is the possibility that a program relationship with the IMF carries a market “stigma”— sending a worrying signal to markets about the country’s economic fundamentals. To examine this possibility, this section follows a two-stage approach. First, can the different economic conditions—and the member’s circumstances prevailing at the start of a program more generally—account for the differential performance over the program period? Second, once those conditions are accounted for, does the market discriminate against those countries choosing to request precautionary programs? A “program choice” model (between a precautionary program, a drawing program, or no IMF support at all) is developed in the section “Program Choice.” This model is then used in the section “Market Response” to examine whether significant differences remain in the market’s response, controlling for the initial conditions that led to the member’s request for a program.9

Characteristics of Precautionary Programs

Over the period 1992–2005, the IMF’s Executive Board approved 52 precautionary and 110 drawing arrangements (Table 3.1). In the sample, the authorities later drew under an arrangement that was initially intended to be precautionary in only six cases, four of which were capital account crises (Argentina 2000, Brazil 2001, Philippines 1998, and Uruguay 2002)—the other two being Peru (1996) and Uruguay (1997). Peru drew IMF resources to help finance a debt- and debt-service-reduction operation. In Uruguay’s case, a purchase was made following turbulence in international markets in 1998. In four additional cases, the authorities did not initially indicate an intention to treat the arrangement as precautionary, but they did not draw even though the programs remained on track; these cases are included in the sample of precautionary programs.

Table 3.1.Characteristics of General Resources Account Arrangements, 1992–2005
PeriodNumber of

Precautionary

Arrangements
Number of

Nonprecautionary

Arrangements
Annual Access Levels

(In percent of quota)
Average Duration

(In years)
PrecautionaryNonprecautionaryPrecautionaryNonprecautionary
1992–9512.312.520.355.81.01.5
1996–9914.87.833.782.11.82.1
2000–0514.04.846.784.11.51.8
20006639.077.21.71.7
200153106.7235.61.31.2
20022829.7177.71.52.1
20034754.481.61.31.7
20045117.341.31.81.1
20052433.590.91.42.8
Sources: IMF Monitoring of Arrangements (MONA) database; and IMF Database on Access Levels.

Annual averages.

Excluding exceptional access of Brazil, the average access level for precautionary arrangements in 2001 was 30.8 percent.

Sources: IMF Monitoring of Arrangements (MONA) database; and IMF Database on Access Levels.

Annual averages.

Excluding exceptional access of Brazil, the average access level for precautionary arrangements in 2001 was 30.8 percent.

Total access under precautionary arrangements—at around 47 percent of quota during 2000–05—has been about one-half the level of access (relative to quota) under drawing arrangements.10 Typically, the member’s right to draw IMF resources cumulates in a “staircase” pattern as phased quarterly purchases build up so that the full amount of the access may be drawn at the end of the program period, if the program remains on track and there is a balance of payments need.11 For members with no outstanding IMF credit, the initial disbursement amounts to at least 25 percent of quota (i.e., at least the first credit tranche).12 As a result, these arrangements have substantial front loading. Finally, precautionary arrangements tend to be slightly shorter in duration than nonprecautionary arrangements—on average, 18 months rather than 22 months over the 2000–05 period—per-haps because, as elaborated below, these arrangements are typically requested by members that are not in crisis but that are trying to ride out periods of uncertainty.

As a first step in comparing the performance of precautionary and nonprecautionary programs, Figures 3.1 and 3.2 plot key macroeconomic variables, revealing some characteristic differences.13 For example, output growth was slow over the three years prior to the start of a non-precautionary program, averaging about 2 percent in the first program year, and rising to 3 percent a year thereafter. By contrast, the pace of output growth was increasing prior to the start of a precautionary program, averaging 4 percent in the first program year and roughly maintained thereafter. The initial inflation rate also tends to be more favorable for precautionary programs. Whereas inflation, though declining, was 20 percent in the first year of a non-precautionary program (falling to 8 percent three years later), it was 8 percent in the first year of a precautionary program (falling to 3½ percent three years later).

Figure 3.1.Nonprecautionary and Precautionary Arrangements: Macroeconomic Developments1

Sources: IMF, World Economic Outlook; and IMF staff estimates.

1 Precautionary and nonprecautionary arrangements excluding capital account crises.

Figure 3.2.Nonprecautionary and Precautionary Arrangements: External Sector Developments1

Sources: IMF, World Economic Outlook; and IMF staff estimates.

1 Precautionary and nonprecautionary arrangements excluding capital account crises.

2 The real exchange rate overvaluation is defined as the percentage difference between the real exchange rate and a 20-year trend obtained using the Hodrick-Prescott filter.

One element of this stronger initial macroeconomic performance for precautionary programs is a more benign external environment. When members request nonprecautionary programs, private capital flows have declined from a peak of more than 3 percent of GDP three years prior to the program to less than ½ percent of GDP in the program year (see Figure 3.2). The real exchange rate is also depreciating (relative to its long-run trend) and gross international reserves (as a proportion of short-term debt) are declining, suggesting a weakening balance of payments position. Members requesting precautionary programs have also experienced lower private capital inflows, but the decline is of a much smaller magnitude and starting from a much higher level. For these members, private capital flows decline slightly, from about 6 percent of GDP over the three years prior to the program to 5 percent of GDP in the first program year. The real exchange rate is modestly appreciated (relative to its long-term trend) and reserves are rapidly increasing in relation to short-term debt, suggesting a strengthening balance of payments.

These patterns of private capital flows are also reflected in the behavior of the current account. For drawing programs, the current account deficit narrows prior to the program and especially during the first year of the program, reflecting the slowdown in capital inflows and some policy tightening. (The overall fiscal balance improves by about 1 percent of GDP during the first program year.) Correspondingly, investment declines and saving improves modestly. For precautionary programs, the current account balance only improves during the first program year, mainly on account of an improvement in the fiscal balance, and by ½ percentage point of GDP less than for nonprecautionary programs. The current account deficit widens after the first program year as capital inflows resume. While the return of confidence and resumption of capital flows is a sign of program success, the widening current account deficit could be of concern—though, in fact, these deficits do not compromise external debt sustainability.14

Program Choice

It would seem intuitive that macroeconomic performance and external developments play a role in the authorities’ decision on what type of IMF-supported program (if any) to request.15Table 3.2 therefore reports key variables in the year prior to either a precautionary or a nonprecautionary GRA-supported program (as well as for nonprogram periods) for a sample of middle-income countries, excluding those that are eligible for loans under the Poverty Reduction and Growth Facility.

Table 3.2.Initial Conditions

(Year prior to program or nonprogram average)

Countries with At Least One

IMF-Supported Program in the Sample Period
Periods with No

IMF-Supported

Program
Precautionary1Nonprecautionary1Difference between

precautionary and

nonprecautionary

programs2
In percent a year
Output growth2.73.52.0***1.5
Inflation23.29.642.0***−32.4***
Overvaluation (in percent)30.60.4−5.1***5.5***
In percent of GDP
Current account−2.2−3.7−1.5−2.2
Change in current account−0.6−0.50.2−0.7
Private capital flows2.35.1***2.03.1***
Change in private capital flows0.5−0.7*0.6−1.3
Government balance−3.6−2.5−2.5**0.0
External debt48.643.359.7***−16.4***
Reserve coverage (in months of imports)4.15.3***3.4***1.9***
Arrears (in percent of time)417.12.7*31.4***−28.7***
Latter half of executive term (in percent of time)552.867.5*49.418.1***
IMF involvement (in percent of maximum)634.460.4***50.5***9.9
Index values based on perceptions7
Bureaucracy2.21.9***2.0***−0.1
Lack of political risk63.264.662.52.1
Law and order3.43.23.6−0.4

Asterisks indicate whether differences relative to periods of no IMF-supported program are statistically significant at the (*) 90, (**) 95, or (***) 99 percent levels, respectively.

Asterisks indicate whether differences between precautionary and nonprecautionary programs are statistically significant at the (*) 90, (**) 95, or (***) 99 percent levels, respectively.

Overvaluation is measured as the percentage deviation of the real exchange rate relative to its Hodrick-Prescott-filtered trend.

This indicates whether a country is in default to some bondholders or bank lenders (source: Standard and Poor’s).

Calculated as a dummy variable if the forthcoming election of the executive is within two years, the midpoint of the executive term of office in the sample.

IMF involvement is a calculation of the number of years of IMF engagement since 1980 with declining weights over time.

Source: International Country Risk Guide.

Asterisks indicate whether differences relative to periods of no IMF-supported program are statistically significant at the (*) 90, (**) 95, or (***) 99 percent levels, respectively.

Asterisks indicate whether differences between precautionary and nonprecautionary programs are statistically significant at the (*) 90, (**) 95, or (***) 99 percent levels, respectively.

Overvaluation is measured as the percentage deviation of the real exchange rate relative to its Hodrick-Prescott-filtered trend.

This indicates whether a country is in default to some bondholders or bank lenders (source: Standard and Poor’s).

Calculated as a dummy variable if the forthcoming election of the executive is within two years, the midpoint of the executive term of office in the sample.

IMF involvement is a calculation of the number of years of IMF engagement since 1980 with declining weights over time.

Source: International Country Risk Guide.

From Table 3.2, in the runup to a member’s request for a precautionary program, output growth, inflation, the current account balance, and the external debt ratio are not statistically different from periods when no IMF-supported program is requested. Periods when countries request drawing programs, by contrast, are characterized by significantly lower growth, higher inflation, a more depreciated real exchange rate, higher external debt, lower foreign exchange reserve coverage, and a greater likelihood of having sovereign debt arrears relative to nonprogram periods. Macroeconomic performance when countries seek precautionary programs is thus little different than when no IMF support is requested, while it is significantly worse when drawing programs are requested. This raises the question of whether precautionary programs are “weaker” than drawing programs. In fact, the evidence suggests that precautionary programs are no weaker in terms of improved macroeconomic performance that program policies seek to achieve or in terms of program conditionality, though they have typically had fewer structural conditions (Box 3.1).

Box 3.1.Are Precautionary Programs “Weaker” Than Drawing Programs?

One concern with precautionary programs is that they may be “weaker” than drawing programs—either in terms of the ambitiousness of program goals or in terms of program conditionality. In fact, that is not the case.

Regarding macroeconomic policy targets, fiscal adjustment and disinflation targeted under programs appear to be related to the economic problems facing the member, rather than the nature of the program in place. Controlling for the initial conditions, precautionary programs are no less ambitious.1

  • A large initial fiscal deficit or targeted improvement in the current account balance is associated with a more ambitious fiscal adjustment, while the lagged expenditure ratio, lagged public debt, and terms of trade are not statistically significant. Controlling for these factors, no statistically significant difference exists between the fiscal adjustment targeted under precautionary and nonprecautionary programs.

  • Higher initial inflation is associated with more ambitious disinflation targets. Controlling for initial conditions, a statistically significant difference remains between drawing programs and precautionary programs, with the latter targeting a greater reduction of inflation. Controlling for the initial inflation rate and targeted disinflation, the monetary stance—as measured by the programmed change in velocity—is not statistically significantly different between the two types of programs.

  • Actual changes in the fiscal balance and inflation are positively (and statistically significantly) related to programmed changes. Slippages relative to program targets do not differ across types of programs.

Regarding program conditionality, while performance criteria are not significantly different across types of programs, precautionary programs have fewer structural conditions.

  • The mean and median numbers of quantitative performance criteria are comparable between the two types of programs. The implementation rate is also very similar across them.

  • Precautionary programs incorporate statistically significantly fewer structural reforms than nonpre-cautionary programs, but that could be a reflection of less-demanding economic challenges faced by countries requesting precautionary arrangements. Further, while the focus of nonprecautionary programs has been on enhancing economic flexibility and efficiency, precautionary programs have stressed the need to buttress demand management and reduce vulnerabilities, especially in the financial sector.

1 This box draws on Section IV of IMF (2006a).

If macroeconomic conditions are little different when members request precautionary programs than when no IMF support is requested, then why do members request precautionary programs? One possibility is that such programs foster internal discipline and lend credibility to the authorities’ policies, especially when there may be political uncertainty. While political circumstances are difficult to capture quantitatively, the proximity of the next election for the executive branch provides a simple metric. Members seeking precautionary programs are in the second half of the executive’s term two-thirds of the time—a proportion that is statistically significantly higher than members seeking nonprecautionary programs or not seeking the IMF’s support at all (which are both at around 50 percent). Members seeking IMF support—precautionary or not—score lower on the perceived quality of the bureaucracy, and members that have had previous IMF-supported programs are more likely to request another.16

The overall impression that members seek precautionary programs when they have strong economic fundamentals (except for the size of the current account deficit), but perceived underlying uncertainties (such as election-related pressures), is supported by a multinomial choice model. Table 3.3 reports the coefficient estimates pertaining to the choice of a precautionary or nonprecautionary program (in both cases, relative to not seeking the IMF’s support at all).17 Members requesting precautionary programs have lower inflation, and higher foreign exchange reserve coverage, but also greater perceived “internal political conflict” and are in the latter half of their executive term.18 Output growth is positive for precautionary programs, while it is negative for drawing programs—though neither growth rates differ statistically significantly from periods without IMF-supported programs. Surprisingly, the level of the current account balance enters negatively—so that members with larger deficits are more likely to request precautionary programs—but the change in that balance enters positively (and statistically significantly so for nonprecautionary programs). The larger current account deficit for precautionary compared to nonprecautionary programs mirrors the relatively higher level of private capital inflows in the former (a difference of 3 percent of GDP, on average), while the change variable captures the more pronounced decline in these inflows for nonprecautionary cases (see Figure 3.2). Members are likely to request nonprecautionary programs when the external debt ratio is high and the ratio of reserves to imports is low.19

Table 3.3.Program Choice Model Estimates
Year Prior to ProgramPrecautionary

Estimates1
Nonprecautionary

Estimates1
Arrears2−0.38**0.04
Output growth (in percent a year)5.83−3.94
Inflation (in percent a year)−5.15*0.82
Current account (in percent of GDP)−34.07***−0.20
Change in current account (in percent of GDP)30.886.34*
Fiscal balance (in percent of GDP)7.72−6.81
External debt (in percent of GDP)0.401.81***
Reserve coverage (in months of imports)3.41***−2.18
Past IMF involvement32.42***1.28***
Bureaucracy−2.14***−0.76**
Lack of internal conflict−4.18***−0.67
Latter half of executive term40.74**0.08
Number of observations467
Log likelihood−225.10
Percent of correct predictions77.70
Note:Asterisks denote statistical significance at the (*) 90, (**) 95, or (***) 99 percent levels.

Alternative is no IMF-supported program.

The arrears dummy, taken from Standard and Poor’s, indicates whether a country is in default to some bondholders or bank lenders in a particular year.

IMF involvement is measured as a declining weighted sum of the number of years the member has had an IMF-supported program since 1980.

Calculated as a dummy variable if the forthcoming election of the executive is within two years, the midpoint of the executive term of office in the sample.

Note:Asterisks denote statistical significance at the (*) 90, (**) 95, or (***) 99 percent levels.

Alternative is no IMF-supported program.

The arrears dummy, taken from Standard and Poor’s, indicates whether a country is in default to some bondholders or bank lenders in a particular year.

IMF involvement is measured as a declining weighted sum of the number of years the member has had an IMF-supported program since 1980.

Calculated as a dummy variable if the forthcoming election of the executive is within two years, the midpoint of the executive term of office in the sample.

Market Response

Beyond helping to improve macroeconomic performance, an important contribution of an IMF-supported program may be the signal it sends to markets. On the one hand, the announcement of an IMF-supported program may signal that the member is facing economic difficulties of which, or the extent of which, markets were previously unaware—leading to a widening of sovereign bond spreads. On the other hand, IMF support also signals that the authorities are dealing with their economic problems, which could reduce spreads, particularly if the market had already foreseen the economic challenges. The latter possibility suggests that members that face less severe economic difficulties (and therefore do not expect to draw IMF resources) may want to signal as much to the markets by having a precautionary program. But do markets differentiate according to whether the member expects to draw IMF resources? Or are all requests for IMF support stigmatized by markets through wider spreads?

Table 3.4 reports average monthly sovereign bond spreads during the first year of an IMF-supported pro-gram.20 The average spread when there was a precautionary program is no different from the average spread for periods for those same members without IMF-supported programs. However, average spreads for precautionary programs were statistically significantly lower than those for nonprecautionary programs. Members with no IMF-supported program at all during the sample period had even lower spreads.

Table 3.4.Determinants of Sovereign Bond Spread1
First Program YearOverall Program Period
(1)(2)(3)(4)
Precautionary18.9231.2289.6*** 224.62
Nonprecautionary148.6*** 2136.5*** 2168.8*** 2125.6*** 2
Countries with no IMF-supported program−187.2***−156.2***−245.5***−195.2***
Exceptional access662.9***647.3***441.3***428.3***
Interest rate volatility−0.04***−0.02
Export volatility779.51,138.4*
Overvaluation652.4**−19.7
Market pressure3−984.7***−1,227.2***
Arrears42,776.0***1,727.6***
Inverse Mills ratio for nonprecautionary19.9***1.3
Inverse Mills ratio for precautionary−0.5***0.3**
Constant364.1***461.7***429.8***406.7***
Pseudo R-squared0.330.690.150.57
Number of observations7127121,3371,337
Sources: JPMorgan; IMF, World Economic Outlook and Database on Access Levels; and Standard and Poor’s.Note:Asterisks denote significance at the (*) 90, (**) 95, or (***) 99 percent levels.

The omitted category is the nonprogram periods for countries with at least one IMF-supported program in the sample period (1994–2004).

The difference between the precautionary and nonprecautionary coefficients is significant at the 90 percent level of significance.

“Market pressure” denotes a weighted average of real exchange rate and reserve changes, following Kaminsky and Reinhart (1999).

The arrears dummy, taken from Standard and Poor’s, indicates whether a country is in default to some bondholders or bank lenders in a particular year.

Sources: JPMorgan; IMF, World Economic Outlook and Database on Access Levels; and Standard and Poor’s.Note:Asterisks denote significance at the (*) 90, (**) 95, or (***) 99 percent levels.

The omitted category is the nonprogram periods for countries with at least one IMF-supported program in the sample period (1994–2004).

The difference between the precautionary and nonprecautionary coefficients is significant at the 90 percent level of significance.

“Market pressure” denotes a weighted average of real exchange rate and reserve changes, following Kaminsky and Reinhart (1999).

The arrears dummy, taken from Standard and Poor’s, indicates whether a country is in default to some bondholders or bank lenders in a particular year.

Taking account of various explanatory variables— interest rate and export volatility, overvaluation, market pressure, and arrears—and the inverse Mills ratios that capture the circumstances that led the authorities to request the IMF-supported program improves the fit considerably to 70 percent (from 30 percent).21 More important, it shows that, controlling for initial conditions, spreads for members with precautionary programs remain more than 100 basis points lower than for members with nonprecautionary programs— a statistically significant difference—both for the first program year and for the overall program period. In other words, even after taking account of the actual macroeconomic situation facing the member, a differentiated signal is sent to and received by the markets when the member does not expect to draw on IMF resources.22 It also suggests that precautionary programs help mitigate the adverse impact on spreads of increased uncertainty.23

Conclusions

When members do not face pressing balance of payments needs but may be going through a period of heightened uncertainty, they may request that their financial arrangement with the IMF be treated as precautionary. The empirical findings suggest that indeed members seek precautionary—as opposed to drawing—programs in such circumstances. With the authorities eventually drawing IMF resources in only 6 out of some 50 precautionary arrangements over the period 1992–2005, the evidence suggests an impressive track record of such arrangements in terms of crisis avoidance.

Precautionary programs are not weaker than drawing programs in terms of what policies seek to achieve or in terms of conditionality. Perhaps for that reason, precautionary arrangements do not come with any market stigma. Indeed, considering that the member was likely going through a period of heightened uncertainty at the time it requested a precautionary program, this form of IMF support is likely to have reduced spreads and enhanced market confidence relative to the counter-factual in which the IMF was not making its resources contingently available. Overall, the results suggest that contingent support from the IMF may be useful for averting crises; in the next section, more direct evidence is presented.

    Other Resources Citing This Publication