Abstract

Adjustment programs are seldom implemented exactly as planned. Slippages, delays, midcourse corrections, and, occasionally, missed targets are in the nature of ambitious reform. Nevertheless, a track record of stable, market-oriented and consistently implemented economic policies is widely recognized as central to the achievement of sustained high growth, a principal objective of SAF/ESAF-supported programs.1 The rationale has to do mainly with the response of private investors—domestic and foreign—to uncertainty regarding the permanence of changes in macroeconomic and structural policies. If policies are perceived as likely to be reversed, and relative price incentives as highly uncertain, investors may delay committing resources and may postpone irreversible investment until such uncertainty is resolved. Because investment is a primary engine for growth, perceived uncertainty created by stop-and-go policy implementation risks locking the economy into a low-capital-accumulation, low-growth trajectory and reducing the effectiveness of resource allocation signals from structural reforms. These adverse effects may be farther exacerbated if delayed adjustment curtails access to external financing and provokes a crisis requiring harsh corrective measures.

Adjustment programs are seldom implemented exactly as planned. Slippages, delays, midcourse corrections, and, occasionally, missed targets are in the nature of ambitious reform. Nevertheless, a track record of stable, market-oriented and consistently implemented economic policies is widely recognized as central to the achievement of sustained high growth, a principal objective of SAF/ESAF-supported programs.1 The rationale has to do mainly with the response of private investors—domestic and foreign—to uncertainty regarding the permanence of changes in macroeconomic and structural policies. If policies are perceived as likely to be reversed, and relative price incentives as highly uncertain, investors may delay committing resources and may postpone irreversible investment until such uncertainty is resolved. Because investment is a primary engine for growth, perceived uncertainty created by stop-and-go policy implementation risks locking the economy into a low-capital-accumulation, low-growth trajectory and reducing the effectiveness of resource allocation signals from structural reforms. These adverse effects may be farther exacerbated if delayed adjustment curtails access to external financing and provokes a crisis requiring harsh corrective measures.

Program interruptions are only one measure—probably prone to understatement—of unsteady policy implementation. In many SAF/ESAF-supported programs, original targets were modified in ways that at least partially accommodated intervening slippages. But cases in which IMF support for programs is interrupted are likely to encompass the most serious instances of policy reversal, involving policy outcomes typically concentrated at the poor end of the results spectrum for all annual programs (see Figure 9.1 for fiscal outcomes, for instance). Program interruptions may therefore imply important costs, albeit ones that are difficult to quantify in the absence of a reliable counterfactual. Their consequences may last well beyond the eventual resumption of adjustment because a discontinuous policy record adversely affects the credibility of new policy announcements, in turn inducing a wait-and-see attitude among potential investors. The costs may be especially high in low-income countries, where the productive base is often vulnerable to shocks, capital accumulation is constrained by volatile foreign financing, and the interruption of IMF disbursements may imply suspension of support and lending by donors and other financial institutions.

Figure 9.1
Figure 9.1

Frequency Distribution of Changes in the Fiscal Balance1

(In percent of GDP; excluding grants)

Source: IMF staff estimates.1Excluding Guyana.2Or programs where interruptions were not related to past fiscal slippages.3Represents number of cases in which the change in fiscal balance was less than - 7 percent of GDP.4Represents number of cases in which the change in fiscal balance exceeded 7 percent of GDP.

It is thus a matter of some concern that as many as 28 of the 36 countries with SAF/ESAF arrangements approved between 1986 and end–1994 experienced one or more protracted program interruption (51 in total).2 Only 10 of the 40 expired SAF/ESAF arrangements with an intended life of three or more years were completed without significant discontinuities or deviations from programmed policies. Although it is beyond the scope of this study to investigate systematically the adverse consequences of faltering programs (the importance of which is uncontroversial), the high frequency of interruptions may help to explain the disappointing outcomes for investment and growth in many ESAF countries.3 Indeed, the record confirms that cumulative capital formation and per capita growth4 in interrupted arrangements were significantly slower than during uninterrupted arrangements (Table 9.1).5

Table 9.1

Investment and Growth in SAF/ESAF-Supported Arrangements1

(Mean values)

article image

The sample includes 40 expired three-year arrangements; it excludes cases affected by significant measurement problems (Mozambique, Equatorial Guinea, and, for investment/GDP ratios, Guyana), as well as SAF arrangements that were replaced by ESAF arrangements before expiration and three-year arrangements that have not yet run their course.

This chapter reviews the experience with program interruptions in the 28 countries affected. The analysis draws on IMF staff reports, country papers, and discussions with some IMF mission chiefs. Three broad questions are addressed. What circumstances lent themselves to program interruptions? What was the role of program design and monitoring? And what lessons does the experience suggest to minimize the incidence of interruptions in future programs?

The next section discusses the definition of program interruptions, followed by an examination of the stylized facts underlying program interruptions, including a classification of the episodes according to the importance of various proximate and underlying factors. Next, key aspects of program design and monitoring are analyzed, focusing on the ambitiousness and prioritization of program targets, administrative and technical capacity constraints, the continuity of program monitoring, and contingency planning. The final section summarizes conclusions and poses lessons for program design.

Defining a Program Interruption

Interruptions of IMF-supported programs can be identified by at least two (not mutually exclusive) indicators: the share of resources committed under IMF arrangements left undrawn, and discontinuities in the disbursement of IMF resources.

The first approach would define noncompletion by comparison with an arbitrary minimum threshold in the share of committed resources left undrawn. For instance, in a study of noncompletions of IMF-supported programs approved during 1980–90, Killick, Malik, and Manuel (1991) defined as uncompleted programs with 20 percent or more of total IMF commitments undrawn. Although convenient, this criterion is not sufficiently discriminating to gauge the incidence of interruptions in SAF/ESAF-supported programs. It is possible to construct examples of multiyear programs subject to several important discontinuities that would classify as “completed” according to this test.6

The rationale for the second approach—the approach adopted in this study—is that deviations from programmed policies dealt with quickly and decisively should have minimal repercussions on confidence. Conversely, delays in bringing policies back on track increase the likelihood of adverse economic effects stemming from uncertainties about policies. Three types of interruptions are relevant for ESAF arrangements. First, for most ESAF countries, the adjustment process encompasses more than one multiyear arrangement. Therefore, long intervals between multiyear arrangements may adversely affect investors’ confidence to the extent that they are perceived to involve deviations from sound policies. Second, within one multiyear arrangement, an extended gap between annual arrangements (or the cessation of the arrangement, if no farther annual program is approved) may signal important deviations from policy commitments or forward-looking disagreements between IMF staff and country authorities on the path of reform. Third, similar difficulties with regard to policies (past or prospective) may be signaled by long delays in completing a midterm review in an annual program.

In light of these considerations, we define a program interruption as either an interval of more than six months between different annual or multiyear IMF arrangements or a delay of more than six months in completing a program review.7 The arbitrary threshold of six months or more was chosen so as to maximize the chances of capturing all potentially significant policy deviations, while avoiding mere procedural delays.8

On the basis of this definition, there have been 51 program interruptions affecting 28 out of the 36 countries under review, since the inception of the SAF in 1986 (Figure 9.2). These interruptions were about evenly distributed across each of the three types considered, with a somewhat higher incidence of interruptions in between annual arrangements, but they differed considerably in duration.9 Most countries (17 out of 28) experienced more than one interruption—up to three in some cases (see Appendix 9.1, Table 9.10).

Figure 9.2
Figure 9.2
Figure 9.2

Program Interruptions in SAF/ESAF Arrangements1

1Bars represent interruptions for delay or noncompletion of ESAF midterm reviews; symbol ((§)) within ESAF bars indicates when a review was completed. Circles represent interruptions between annual arrangements (including cases where a subsequent annual arrangement was not approved). Rectangles represent interruptions between multiyear arrangements.2Intervening Stand-By Arrangement.

The incidence of interruptions is quite sensitive to the arbitrary choice of minimum time interval. For instance, lowering the threshold to three months would increase the number of countries with delays in completing midterm reviews from 16 to 24, and the number of countries with delays in approving subsequent annual arrangements from 17 to 34. However, a three-month threshold would probably overestimate the incidence of policy-related interruptions, giving too much weight to delays of a bureaucratic or institutional nature.

Another perspective on the incidence of program interruptions is provided by the number of arrangements envisaged to last three years (or four, for cases with an additional annual arrangement) that ran their full course without significant interruptions. Of 40 arrangements (16 SAFs, 24 ESAFs),10 2 (5 percent) were interrupted even before the end of the first year; 38 (95 percent) made it through at least one year; 22 (55 percent) made it through at least two years; and 15 (38 percent) made it through three years without interruption (5 SAFs, 10 ESAFs). However, of the latter 15 arrangements, 2 were interrupted during an additional fourth-year arrangement, and 3 saw policies veer off track in the third year so that a break of more than six months occurred before a new program could be agreed. Thus, only 10 arrangements (25 percent, four SAFs and six ESAFs) can be said to have run their full course uninterrupted.

Stylized Facts

This section examines the causes of the interruptions and the relative importance of various factors involved. Several questions are addressed. What circumstances led to program interruptions? Were deviations from agreed policy commitments the most important proximate cause of an interruption, or was it, rather, forward-looking disagreements on the path of reform in the absence of significant policy slippages?11 Were problems broadly or narrowly based, mainly related to financial or structural policies, or to both? What was the nature of exogenous shocks, and how central were they to the interruption of programs?

One would expect that programs are interrupted only as a result of severe deviations from agreed policy commitments. In fact, although this is most frequently the case, it is not true in all episodes. The sample of 51 interruptions encompassed a wide range of circumstances (details are given in Appendix 9.1, Table 9.11), and can be classified into three categories according to the underlying primary cause: forward-looking disagreements in the absence of significant deviations from past policy commitments; severe political disruptions; or serious slippages in past policies (see Appendix 9.1, Table 9.12). The incidence of these factors is summarized in Figure 9.3.

Figure 9.3
Figure 9.3

Distribution of the 51 Program Interruption Episodes by Main Factor

Source: IMF staff estimates.

In 8 of the 51 episodes (about one-sixth), there had been no major deviations from planned policies prior to the interruption, but either the IMF staff and authorities were unable to agree on the extent or pace of financial and structural adjustments to be implemented in the period ahead, or the authorities needed more time to formulate a policy response to unexpected changes in the economic environment.12 Specific problems leading to interruptions driven solely by forward-looking disagreements included the need to respond to deteriorating macroeconomic conditions and make progress in key areas of structural reform (Bangladesh 1, Bolivia 2); the time required to adapt program design (ex post) to a sharp rise in oil prices as a result of the Middle East crisis (Lao P.D.R. 1); the need to adapt macroeconomic policies to a major supply shock—for instance, attacks on mining installations accounting for 60 percent of exports and a sizable share of budget revenues (Sierra Leone 2); or efforts to ensure consistency of the fiscal program with the outcome of last-minute negotiations with the trade unions (Benin 2). In other episodes, the disagreement reflected the presentation to parliament of budget policies falling short of agreed program understandings (Pakistan 3), or the authorities’ disagreement on the need to correct macroeconomic policies to account for reduced levels of external assistance (Mauritania 2). The case of Nepal 2 reflected instead the general problem of designing an appropriate response to a shock (severe floods) that had uncertain economic effects.13

Another 10 of the 51 interruptions stemmed from political disruptions serious enough to call into question the continuing authority of the government and, therefore, to prevent meaningful negotiations (see Appendix 9.1, Table 9.13 for a list of these episodes). The nature of political upheavals and the intensity of political or ethnic turmoil varied, but all cases were characterized by a severe reduction of the authorities’ ability to commit credibly to and implement adjustment policies. In fact, in Burundi 3, widespread ethnic violence led to the near complete breakdown of civil society. In Benin 1, Madagascar 1, Mali 1, and Nepal 1, fundamental and far-reaching reforms of the political system—the introduction of a multiparty system of government—were accompanied by severe social conflict. Togo 2 and Niger 2 featured an escalation of political tension and social unrest—effectively shutting down the formal economy for several months in Togo, and preventing meaningful program discussions in Niger. In Pakistan 2, the six-month period preceding the October 1993 elections was characterized by a deep institutional and political crisis.14 In Sri Lanka 2, the run-up to the August 1994 elections and the defeat of the government after 17 years in power had a profound effect on policymaking. These political events were rooted in the history of the countries involved, but in a few instances social unrest appeared to be linked to the implementation of measures supported by the IMF, rather than being exogenous to the program (Box 9.1).

The remaining 33 of the 51 interruptions (about two-thirds) were closely related to serious deviations from past policy commitments. Typically these deviations were broadly based, affecting either several areas of macroeconomic policy or the combination of macro and structural policies.15 Nevertheless, in the large majority of these episodes—27 out of 33—the root of the problem was fiscal in nature. Because of the limited development of markets and instruments for budgetary financing, in many instances fiscal policy slippages typically spilled over to problems in the control of domestic credit and domestic or external arrears.16

In 6 of the 33 episodes, the slippage from past policies was predominantly structural in nature.17 In Equatorial Guinea 1, the adjustment program went off track as a result of delays in implementing measures related to property rights, reform of state trading companies, and financial sector development. These delays partly reflected weak administrative capacity, a problem also in Lao P.D.R. 2. In Mauritania 1 most structural benchmarks under a SAF-supported program had not been observed, and delays affected a variety of areas—tariff, taxation and pricing reforms, banking sector restructuring, and external debt management, among others. In Nepal 3, structural reforms came to a halt in the run-up to a change in government, while in Guyana 1 the authorities had failed to follow through with privatization and public enterprise reform. In Mozambique 1, there were shortfalls in the management of counterpart funds of foreign aid in the budget, and recurrent difficulties in monetary control arising from the rapid expansion of “other items, net” in the central bank balance sheet—reflecting subsidies on foreign exchange operations and accounting and statistical problems.

In the 33 episodes where serious slippages from past policy commitments were the dominant factor, a variety of exogenous influences also played a role. First, political events affected a number of these episodes (13 of the 33), although not as disruptively as in the 10 episodes where they were the overriding factor. The specific nature of these events varied. In Burkina Faso 1, Ghana 1, Guinea 3, Kenya 1, Malawi 1, and Togo 1, the introduction of a multiparty system of government deflected the attention of the authorities from the economic program. In another 5 of these 13 episodes—Bolivia 3, Equatorial Guinea 2, Honduras 1, Nepal 3, and Senegal 1—a preelectoral climate in established political systems affected program performance (Box 9.2).18 In two episodes—Kenya 1 and Sri Lanka 1—policymaking was hampered by local ethnic violence or civil conflict.

Second, external shocks played a role in 11 of the 33 interruptions. These shocks typically involved declining commodity export prices or shortfalls in external financing, with sizable deviations from program assumptions (see Appendix 9.1, Table 9.14).19 Third, natural disasters (droughts or poor rainfall) were associated with 6 of the 33 interruptions.20 In several instances, more than one type of exogenous event—political, external, or natural disaster—look place around the time of the interruption.

Were Adjustment Measures a Factor Inducing Social Unrest?

Program interruptions that were influenced by social and political instability raise the question of whether some of these events were related to the implementation of adjustment programs supported by the IMF. The circumstances underlying ten such episodes were examined for this study—six of the ten cases involved major political upheaval, and four involved less severe episodes of social unrest (see Appendix 9.1, Table 9.13). In other instances associated with political events, the nature of the event—regular elections, transitions to a multiparty system of government, ethnic violence, and civil wars—was such that a direct link to specific economic policies appears highly unlikely.

In three of the ten episodes—Madagascar 1, Nepal 1, and Togo 2—political and civil disturbances were linked to deep-seated dissatisfaction with existing institutional and political arrangements that culminated in a dramatic escalation of tensions between different political groups. The unrest did not appear to be related to economic policy measures and was ultimately resolved by far-reaching reforms of the political system.

In three other cases—Mali 1, Niger 2, and Togo 1—massive demonstrations and civil unrest appear also to have reflected mainly preexisting problems. However, in these cases it is possible that tensions related to previous adjustment efforts may have played some role. Factors included opposition to previous revenue mobilization efforts (Mali 1); an attempt to curb the government wage bill, subsidies, and transfers (Niger 2); and simmering discontent with prolonged wage freezes and the social consequences of protracted adjustment efforts (Togo 1).

In the remaining four cases—Benin 1, Guinea 2, Mali 2, and Malawi 1—the link between adjustment measures adopted either prior to or during the program interruption episode and the emergence of social disturbances was more direct. Strikes and social unrest appear to have been motivated by the opposition to unpopular cuts in existing civil service benefits and employment (Benin 1); in civil service real wages (Guinea 2 and Malawi 1); and in public expenditure on scholarships (Mali 2).

To what extent can these unexpected events be said to have contributed to the policy slippages that culminated in program interruptions? This is a difficult question to answer in cases influenced by social and political events, since the threshold beyond which governance problems became an effective constraint on policymaking—by undermining commitment, or control of the policy apparatus, or both—is a matter of judgment. It is also difficult to separate the impact of these events from those of unrelated policy slippages. The review of the relevant interruption episodes suggests that social and political events did contribute by shifting the attention of the authorities away from programmed policies but that—unlike the major upheavals discussed above—they did not play a decisive role.

For the nine instances of interruptions where past policy deviations were affected by external shocks or natural disasters but not by political events, it is easier to trace through the direct effects of the shocks, and hence to determine whether they were critical to the interruptions. Table 9.2 summarizes the nature of the shocks in each of these interruptions, the areas where deviations from programmed policies were clearly shock-related, and any additional deviation not clearly linked to shocks. The record indicates that, in all cases, program interruptions reflected a combination of inadequate responses to unexpected events and an apparently unrelated inability or unwillingness of the authorities to meet their original policy commitments.

Table 9.2

Policy Deviations in Interruptions Affected by External Shocks and/or Natural Disasters1

article image
article image
Source: IMF staff estimates.

The sample includes interruptions closely related to slippages in past policies and not affected by political shocks.

A Closer Look at What Went Wrong: The Role of Program Design in Policy Slippages

In the 33 interruptions where past policy slippages played a major role, questions arise whether deficiencies in program design or monitoring may have contributed to or even caused the interruptions. Implementation problems, for instance, may have reflected excessively ambitious financial policy targets or pace of structural reforms, inadequate contingency planning, or weaknesses in administrative and technical capacity that, at least in principle, could have been addressed by a stepped-up technical assistance. This section examines the features of program design that might have helped to prevent some interruptions from occurring and considers whether the experience of these countries suggests ways to build programs that are more resilient to disruptive influences.

In examining whether interruptions could have been reduced by changes in program design and monitoring, firm and clear-cut answers are precluded by at least two methodological problems. First, many of the circumstances under consideration are obviously influenced by the extent of the authorities’ commitment to reform, a factor which is hard enough to gauge case-by-case, let alone across a sizable number of countries.

In practice, one can never be sure that policy failure is due to faulty program design or inadequate monitoring rather than weak commitment. Second, the nature of the questions involved would naturally call for counterfactuals, which are in practice unknown. Some comparisons with a control group of programs not affected by interruptions are made where feasible. But the heterogeneity of country circumstances makes it difficult to generalize this approach throughout the study.

With these caveats in mind, five aspects of program design that could have caused or contributed to the policy slippages were examined: (1) overly ambitious targets; (2) insufficient prioritization of structural reforms; (3) inadequate technical assistance; (4) insufficient staff contact and monitoring; and (5) weak contingency planning.

Overly Ambitious Targets

When interruptions were linked to past policy slippages, were program targets unusually ambitious—particularly in the fiscal area? There are of course conceptual problems in defining “overambition”: what is achievable in one country under given social and political conditions may not be a relevant benchmark for the same country in different circumstances, or for other countries with different capacities and constraints. Initial conditions are also crucial in gauging the strength of the fiscal efforts.

Nevertheless, we can shed some light on ambitiousness by examining whether the targeted adjustment in programs that were interrupted was “extraordinary” compared with all other programs. To do this, we focus on fiscal balance targets21—since deviations from policy commitments were in most cases fiscal in nature—from two perspectives. First, to gauge the ambitiousness of targets in annual programs interrupted because of policy slippages (28 cases in total),22 these targets are compared with the average annual targeted amount of fiscal balance adjustment (in percent of GDP, excluding grants) for all other annual SAF/ESAF-supported programs for which data are available.23 Initial conditions are controlled for by considering the link between targeted adjustment and initial imbalance in the sample of all annual programs. Second, for cases that stand out as particularly ambitious after controlling for initial conditions, it is useful to consider what the possible implications for the program would have been had the targets been set at the level of actual outturns.24 Might this have produced an acceptable outcome in terms of economic performance, while providing a slower but more sustainable pace of adjustment?

Elections and Program Performance

The six interruptions where elections in established political systems played a role—Bolivia 3, Equatorial Guinea 2, Honduras 1, Nepal 3, Senegal 1, and, among the cases of major political disruptions, Sri Lanka 2—highlight various channels through which electoral concerns influenced program performance.

In Bolivia 3 and Honduras 1, deviations from programmed policies included sizable expenditure overruns prior to presidential elections. These deviations were mainly due to large wage and salary increases to public sector employees. Similarly, in Sri Lanka 2, populist measures in the run-up to parliamentary (August 1994) and presidential (November 1994) elections generated revenue shortfalls and a large overrun in current expenditure.

In other cases, electoral concerns contributed to the postponement of corrective measures needed to address widening imbalances. In Equatorial Guinea 2, significant deviations from agreed policies had emerged in the first half of 1993. With the program significantly off track, the midterm review scheduled by August 1993 could not be completed, but IMF staff discussed with the authorities a package of corrective measures. The adoption of these measures was postponed, reflecting the authorities’ preoccupation with elections, held in November 1993. Likewise, in Senegal 1, the political agenda in the run-up to presidential (February 1993) and legislative (May 1993) elections delayed the adoption of corrective measures, after slippages in financial policies and structural reforms had derailed the adjustment process in the first half of 1992.

In Nepal 3, the resignation of the government in July 1994 led to a long electoral campaign culminating in the formation of a minority Marxist-Leninist cabinet in December 1994. Although macroeconomic policies remained broadly on track, the implementation of structural reforms came to a halt, and the new government proved unable to address key policy problems. Prospects for a catch-up with the original reform program were dimmed by continued political uncertainty in 1995.

We take as a benchmark for comparison the “average annual program” in all uninterrupted SAF/ESAF annual arrangements for which data are available. This average program targeted a mean annual deficit reduction (before grants) of ½ of 1 percentage point of GDP (median: 0.8 percentage points), starting from an average fiscal imbalance close to 10 percent of GDP. This implied reducing the initial imbalance by less than one-sixth over the three-year life of an arrangement. Standard deviations indicate a wide dispersion of both targets and initial conditions.

Programs interrupted because of policy slippages were indistinguishable from the benchmark in terms of mean targeted annual deficit change, starting from a broadly similar initial imbalance. The two groups also had similar distributions (Table 9.3).25 On the basis of these sample characteristics, and the fact that the differences in mean values are much smaller than the corresponding standard deviations, the data do not support the proposition that annual targets for interrupted programs were systematically more ambitious than those for the control group.

Table 9.3

Fiscal Targets and Initial Conditions in Annual Programs

(Fiscal balance in percent of GDP, before grants)

article image
Source: IMF staff estimates.

Annual targets for interrupted programs covered periods where interruptions were related to past policy slippages. The years considered are listed in Appendix 9.1, Table 9.15.

The absence of a systematic tendency toward overambition is confirmed by a regression of targeted deficit changes on initial conditions in the joint sample of interrupted and uninterrupted annual programs. As would be expected from Table 9.3, an additive dummy for years in which programs were interrupted because of policy slippages was found to be statistically insignificant, irrespective of the inclusion or exclusion of outliers (Table 9.4).26 This finding is illustrated farther in Figure 9.4, where the regression line estimated on the joint sample excluding outliers crosses the scatter diagram of interrupted programs approximately in the middle. Sixteen programs targeted less ambitious adjustment than average, controlling for initial conditions, and fall below the regression line; the remaining 12 “more ambitious” programs lie above the regression line, albeit by a very small margin in some cases.

Figure 9.4
Figure 9.4

Targeted Change in Fiscal Balance in Interruptions Affected by Policy Slippages1

(In percent of GDP; excluding grants)

Source: IMF staff estimates.1The horizontal and vertical lines correspond to the average values of the variables in the control group of annual programs not affected by interruptions. The regression line shows the estimated relation between targeted change in fiscal balance, initial imbalance, and a constant for the sample of all annual programs. The sample excludes seven outlier observations.2Using the initial fiscal balance as estimated at the time the program was formulated, which in some cases differs from the final data.
Table 9.4

Relationship Between Fiscal Deficit Targets and Initial Conditions in All SAF/ESAF Annual Programs1

article image
Source: IMF staff estimates.

The relationship between targeted change in the fiscal balance before grants as a percent of GDP (ΔX) and the initial balance (Xt–1) was estimated by ordinary least squares in the bivariate regression: ΔX = α + βXt–1 + εt. An additive dummy variable for interrupted programs was included to test for a systematic difference in fiscal targets after controlling for initial conditions.

Indicates statistical significance at 1 percent level (* at 5 percent level).

Even though fiscal targets in interrupted programs do not appear on average to have been more ambitious than others, might alternative targets—cased sufficiently to accommodate the observed outturns and thus to avoid the interruptions—have delivered results that would have warranted IMF support? This question was examined for the 12 relatively more ambitious programs—that is, those above and to the right of the regression line in Figure 9.4. Selected macroeconomic indicators for these cases (Table 9.5) suggest the following:

Table 9.5

Targets and Outturns for the 12 Interruption Episodes in “More Ambitious” Programs1

article image
Source: IMF staff estimates.

These correspond to the cases above the regression line in Figure 9.4.

As assessed at the time of the program.

  • Setting aside Equatorial Guinea (2 episodes), where there are major measurement problems, in five of the interruptions—Bolivia 3, Burkina Faso 1, Lao P.D.R. 2, Togo 1, and Zimbabwe 2—the fiscal deficit actually deteriorated, at times by several percentage points of GDP, starting from initial imbalances ranging from over 5 percent to 10 percent of GDP. The policies implemented resulted in the accumulation of external arrears in Bolivia 3 and Togo 1 and in concomitant inflation pressures in Zimbabwe 2.

  • In another four interruptions—Kenya 1, Mali 2, Mauritania 1, and Zimbabwe 1—the improvement in the fiscal accounts fell well short of targets. Moreover, in Kenya 1, Mali 2, and Mauritania 1, a program that would have “lowered the bar” to match actual outcomes would have needed additional financing amounting on average to about 2 percent of GDP to have avoided the accumulation of external arrears. In Kenya 1 and Zimbabwe 1, it would have implied accommodating increasing price pressures.

  • For Pakistan 1, while the most recent data reported in Table 9.5 indicate a deficit reduction in line with targets, IMF staff reports in the aftermath of the interruption ascribed it to a lower-than-programmed reduction of the fiscal deficit (by 0.3 percentage points of GDP), owing to current (defense-related) expenditure overruns; larger-than-programmed credit expansion; and slippages in the implementation of structural reforms. Problems in the fiscal area persisted, owing to delays in tax reforms and expenditure control, and were reflected in increasing inflationary pressures.

  • For the two interruptions in Equatorial Guinea, measurement problems and revisions of accounting methods also hinder target-outcome comparisons. Nevertheless, IMF staff reports in the aftermath of both interruptions ascribe them primarily to missed fiscal targets. Alternative programs would have needed additional financing of 13–14 percent of GDP annually to have avoided the accumulation of external arrears in each episode.

In sum, the evidence indicates that programs that were interrupted because of policy slippages were not unusually ambitious—against the standard of the average fiscal effort envisaged in all other annual SAF/ESAF-supported programs for which data are available. The record therefore does not suggest that a systematic tendency toward overly ambitious targets was a factor in the program interruptions. In the minority of cases where fiscal targets were more ambitious than the average, an easing of the targets to match policies that were implemented would have implied, in most instances, accepting a sizable deterioration in macroeconomic conditions.

Insufficient Prioritization

Another dimension to ambitiousness that might play a role in interruptions is the extent of the structural reform envisaged relative to the administrative ability of national authorities. One possible concern is that programs attempt to cover too many reform areas with too many measures. Indeed, there are many examples of policy framework papers and letters of intent that list several dozen—in some cases, over a hundred—measures to be taken over the three-year period. The obvious risk is that a “scattershot” approach may confuse the authorities about what is needed to maintain IMF support and prevent sufficient focus on the most important reforms.27

In fact, SAF/ESAF-supported programs do envisage a ranking of structural measures, represented by the selection of a few structural benchmarks and performance criteria. In this respect, the ESAF operational guidelines state that all ESAF arrangements will include a limited number of semiannual performance criteria including a few structural measures. And in the 1987 discussion of the ESAF in the IMF’s Executive Board, most Executive Directors favored the use of structural benchmarks to monitor implementation of the most important structural policy measures and noted that performance criteria will be limited in number and will generally involve only a subset of the benchmarks. This hierarchy reflects that structural benchmarks (SBs) are yardsticks for implementing and evaluating reforms but, unlike structural performance criteria (SPCs), are not directly linked to disbursements under the arrangement. Although all ESAF-supported programs thus incorporate an element of selectivity, the question arises whether those that were interrupted may have run into difficulties in part because they were less selective (had more structural conditions) than the average program.

To address whether subjecting too many structural measures to formal conditionality may have contributed to program failures, the number of SBs and SPCs in the annual program agreed prior to the interruptions affected by past policy slippages were examined.28 These were compared with the average for all recent ESAF-supported programs for which information is available.29

In the benchmark sample, the average combined number of SBs and SPCs was 9: almost 80 percent of ESAF-supported programs had at least one SPC, on average 2–3, and all programs included SBs, on average 7–8. The number of SPCs in the programs prior to the 29 episodes examined varied from zero to 5. In the 12 cases where they were included, there were on average 2—3 SPCs (Table 9.6). SBs were included in 25 of the 29 programs, and averaged 6–7 per program, excluding one outlier (Pakistan 1) with many SBs.30 Except in Pakistan, the number of structural conditions was not out of line with the averages from the comparator sample of ESAF-supported programs quoted above and does not support the hypothesis that excessive structural conditionality was an important factor behind interruptions.

Table 9.6

Number of Structural Conditions in Programs Prior to Interruption Episodes1

article image
Source: IMF staff estimates.

The sample includes all interruption episodes primarily affected by slippages in past policies and for which structural conditions were available.

N.a. and n.r. indicate, respectively, “not applicable” and “not reported for all conditions.”

No SBs/SPCs set in the midterm review.

The same conclusion may be seen from a different angle by noting that -within the range of SPCs and SBs considered here, there is no obvious correlation between the rate of observance and the number of SPCs/SBs: in regressions of the share of structural conditions observed on the (nonzero) number of conditions set—for all cases where at least some of the outturns were reported—the estimated slope coefficients were statistically insignificant at standard levels, and the explanatory power was poor (Table 9.7).31 The weak correlation should not be interpreted as evidence against a focused approach to setting structural conditions—one that takes into account administrative capacity constraints. It merely suggests that modifying the number of structural conditions would in itself not guarantee a better implementation record.

Table 9.7

Observance Rates in Structural Conditionality1

(Ordinary least squares)

article image
Source: IMF staff estimates.

The sample includes the subset of programs prior to the interruption episodes listed in Table 9.6 for which conditions were set and at least some of the outturns were reported.

Pakistan 1 was excluded, since the interrupted program included no SPCs.

Regression results were broadly similar when Pakistan 1 was included, in that the slope coefficient was not significant at standard levels.

Not statistically significant at 5 or 1 percent levels.

A more difficult question is whether there may have been cases of “faulty prioritization,” where, for instance, measures were chosen that were relatively easy to implement but not central to the achievement of program objectives. One problem in analyzing this issue is that the program documents reviewed for this study were often not explicit about the rationale for the specific choice of structural conditions. Moreover, it would be necessary to examine structural conditions in the context of individual country cases, since the variety of initial conditions and the tailoring of structural measures to the problems of each country complicate cross-country comparisons. On this subject, however, the record of structural conditions and reform priorities for bank restructuring and reform of public enterprises in several countries does not identify a clear link between progress made and the extent to which structural conditions were met, suggesting that, in some cases, conditions were not set on the most relevant reform measures. A simpler aspect of the problem relates to whether the issues that ultimately caused the interruptions in cases affected mainly by slippages in structural policies—Equatorial Guinea 1, Guyana 1, Lao P.D.R. 2, Mauritania 1, Mozambique 1, and Nepal 3—had been prioritized, and their importance clearly signaled to the authorities. The answer is generally affirmative, since in most cases the areas identified as major problems had been subject to SPCs or SBs in programs preceding these interruptions.32

Inadequacy of Technical Capacity

Closely linked to the question of whether programs were overly ambitious in the financial program or in the agenda for structural reforms is that of adequacy of technical capacity. For many countries, limited administrative and technical abilities were a hindrance to policy implementation. Staff reports for programs affected by interruptions linked to past policy slippages point explicitly to a handful of cases—Burkina Faso 2, Equatorial Guinea 1, 2, and 3, and Lao P.D.R. 2—where administrative and technical constraints played an important role.33 It seems likely that other countries were also affected to varying degrees: for instance, in Mozambique 1 and 2, limited monitoring capacity was highlighted by recurrent problems with monetary control and management of foreign aid counterpart funds.

A comprehensive review of all technical assistance in response to these needs—through staff missions from IMF departments providing technical assistance, placement of long-term resident experts, training of public officials through the IMF Institute, and use of IMF resident representatives—is beyond the scope of this study. Nevertheless, for the four cases where IMF staff reports provide some discussion of the issues, it is of interest to consider the forms of technical or administrative constraints identified in programs prior to interruptions, and what was envisaged and ultimately done to improve the situation.

In Burkina Faso, the ESAFI program approved in March 1993 targeted an ambitious increase in tax and nontax revenues (1.8 percentage points of GDP). The authorities acknowledged shortcomings in public administration and requested technical assistance from the IMF in the areas of tax administration and expenditure control.34 This request was supported by IMF staff, who had cautioned the authorities that their envisaged schedule of tax reforms might be too ambitious.35 In the event, a sizable revenue shortfall emerged early in the program. An IMF staff visit (June 1993) identified as main factors the poor preparation of tax and customs authorities to implement the envisaged reforms (replacement of the turnover tax with the VAT and tariff reform), and the authorities’ poor record in resisting pressures for tax exemptions. A resident expert from the IMF’s Fiscal Affairs Department was placed in the field from June 1993. In the event, some performance criteria were missed by a wide margin, and the midterm review could not be completed. Despite repeated calls for action, the authorities failed to take corrective measures, and IMF staff suggested a monitoring period (through end–1993) as a precondition for discussions on a second-year program (approved in March 1994). Steps were also taken to place a long-term advisor in the area of budget planning and control. This case underscores the importance of ensuring appropriate preparation in advance of complex reform measures, but also the limited effectiveness of a step-up in technical assistance without a concomitant resolve on the part of the authorities to adhere to the requirements of reform.

The experience of Equatorial Guinea is unusual in that IMF staff explicitly recognized, after the fact, that many of the structural reforms envisaged in the SAFI policy framework paper proved to be too ambitious in relation to the country’s administrative and technical capacity, given the availability of technical assistance. Problems affected many areas, and it took two years (1990–91) for structural reforms to catch up to program aims.36 Before the subsequent ESAF arrangement (February 1993), IMF staff advised the government to intensify efforts to obtain technical assistance in the fiscal area. No specific plan involving the IMF was discussed in Executive Board documents, however, and assistance was limited to the placement of a Fiscal Affairs Department resident advisor (1992–93) on tax department reorganization and personnel training. There was no IMF resident representative office. At the midterm review, the authorities—preoccupied with the elections—were urged to establish a monitoring committee to ensure a timely flow of reliable economic and financial data. This recommendation, and the suggestion to seek further technical assistance, were repeated at the start of the ESAF II program (March 1994) and in the 1995 Article IV consultation, after the arrangement had veered off track. Again, no specific plan involving the IMF appeared in Executive Board documents. Many other factors were at play in these interruptions, including political influences, but the importance of following up on recognized technical assistance needs with an action plan, including a role for the IMF as a provider or coordinator of this assistance, is clear. Otherwise, priorities rightly signaled to the authorities by an SB cannot be said to have been thoroughly followed up with appropriate remedial measures.

The gradual move of Lao P.D.R. from a centrally planned to a market-oriented economy (supported in 1989 by a SAF arrangement) was complicated by limited implementation capacity. The program of technical assistance already under way was intensified in monetary, fiscal, and statistical areas with IMF staff missions and the placement of resident advisors. A resident representative office was opened in October 1990. Despite further missions and placement of other experts, however, at the start of the subsequent ESAF arrangement (June 1993), weak fiscal administration continued to pose a threat to financial stability, and further assistance was envisaged and provided. Although lack of political consensus on the path of reform also played a role, weak administrative capacity remained an important cause for slippages in structural reforms during the first annual ESAF arrangement. Given the persistence of these constraints, it appears that the improvements in implementation capacity needed to keep the program on track would probably not have been possible, even with an expanded technical assistance effort. In these circumstances, the structural reform agenda should perhaps have been set on a less ambitious course.

The two program interruptions in Mozambique were affected by excessive expansion of “other items net,” owing to poor mechanisms for monetary control and the inability to generate counterpart funds to foreign aid. Both problems were recurrent and analyzed in IMF staff reports before the interruption episodes. The problem in controlling “other items net” had been considered by a technical assistance mission from the IMF’s Monetary and Exchange Affairs Department and was complicated by the multiplicity of factors at play—central bank losses and implicit subsidized credit to importers through foreign exchange operations during periods of rapid exchange rate movements (pointing to the role of influential pressure groups), reclassification of items in the balance sheet of the banking system, and the impact of banking reforms on the production of monetary data. Two additional missions proved necessary to gain a better understanding. Control of counterpart funds remained a problem, however, reflecting the difficulty of predicting aid disbursements, but also the diversion of counterpart funds to off-budget items, such as unprogrammed lending to public enterprises. The experiences of Mozambique, Burkina Faso, and Equatorial Guinea show that the persistence of administrative weaknesses may sometimes be linked to deeper problems of governance. In these circumstances, an increase in technical assistance is probably a necessary but not sufficient condition to reduce interruptions.

Insufficient Monitoring

According to the ESAF operational guidelines, monitoring of programs supported by ESAF arrangements is to be conducted through semiannual performance criteria in all cases, and disbursements are to take place semiannually, upon approval of an annual arrangement, and subsequently, on the basis of observance of performance criteria and, in most cases, completion of a midyear review.37 ESAF provisions require a midterm review “in most cases.” This framework was intended to minimize the administrative burden on authorities while allowing for tighter monitoring relative to SAF arrangements (under which all disbursements were made upon approval of annual arrangements, without midcourse reassessments). At least in principle, however, the framework has some shortcomings: the requirement of a single midterm review may result in a dialogue between IMF staff and authorities that is less frequent or effective than necessary to keep the program on track, and there is no link between disbursements and progress in implementing policies in the latter part of each annual arrangement.38

These shortcomings have a potential bearing on the record of program interruptions. In this section, we focus first on whether staff contact was timely enough to ensure that imbalances were detected at an early stage, and whether tighter monitoring might have helped to forestall emerging slippages and prevent program interruptions.

The record of missions and meetings at headquarters prior to and during the 33 interruptions affected by policy slippages provides a crude indicator of the frequency of dialogue between staff and authorities. Since there is no information on high-level contacts outside the context of missions and meetings at headquarters, it is an incomplete measure. Nevertheless, it suggests that in two-thirds of the cases,39 contact with the authorities took place at least once in the few weeks prior to the start of these interruptions, and that policy dialogue was generally maintained thereafter (see Appendix 9.1, Table 9.17).40 Earlier mission activity (during the period when the problems were most likely emerging), however, was not particularly intense: in 19 of the 33 episodes examined, there was at most one mission in the semester prior to the interruption, and in the other 14 episodes there were two to four missions (Figure 9.5 and Table 9.8).41 That the more intensive IMF mission schedule in these cases did not avert the breakdown in policies, however—despite most missions being program negotiations (sometimes combined with Article IV consultations) or midterm review discussions, rather than staff visits—suggests that more frequent IMF staff contact, although it might have helped, would probably not have been sufficient to prevent policy discontinuities. Indeed, a limited survey of interruptions associated with sizable fiscal revenue shortfalls or expenditure overruns in the absence of shocks—Equatorial Guinea 3, Nicaragua 1, Tanzania 1, and Zimbabwe 1 and 2—points to persistent slippages despite repeated warnings and efforts to bring the program back on track. In some instances—Equatorial Guinea 3, Nicaragua 1, and Tanzania 1—even after an IMF mission had reached agreement with the authorities on corrective measures, these were not implemented.

Figure 9.5
Figure 9.5

Frequency of Missions in the Six Months Leading Up to Interruptions Affected by Policy Slippages

Source: IMF staff estimates.
Table 9.8

Program Monitoring Prior to Interruptions Affected by Past Policy Slippages

article image
Source: IMF staff estimates.

The question of whether more frequent program test dates and reviews would have helped to prevent interruptions is difficult to address empirically. The absence of relevant counterfactuals prevents general conclusions, since a comparison of programs with similar problems and initial conditions, some having biannual and others quarterly monitoring, is not possible: no ESAF-supported program has had quarterly test dates or reviews. Arguments based on the limited administrative capabilities in ESAF countries can be made on both sides of the issue. More frequent monitoring could impose an additional burden on administrative capacity. Alternatively, it could help to prevent problems and make adjustments easier—if only by prompting earlier corrective action—over time improving implementation capacity.42

All this said, a comparison of monitoring efforts in ESAF countries and those receiving IMF support under other facilities in the Baltics, Russia, and other former Soviet Union countries (BRO) may be informative. These countries face many of the same problems as those with ESAF-supported programs, both in terms of the policy challenges and their limited administrative capacity. The information available clearly suggests more intensive IMF involvement in the BRO countries. Formal quarterly (or more frequent) reviews were required in all Stand-By and Extended Arrangements approved in BRO countries, compared with the semiannual reviews in all ESAF arrangements. IMF resident representatives were quickly put in place in all 15 BRO countries but were present in only about half the ESAF countries at the time of a program interruption affected by policy slippages. Moreover, the frequency of contact between staff and authorities and the intensity of technical assistance efforts has been considerably greater in BRO than in ESAF countries (Table 9.9). Although it is not possible to attribute the outcome exclusively to greater frequency of monitoring, the incidence of program interruptions in arrangements approved in the two groups of countries differs markedly: using the same definition of program interruption, there were 5 interruptions in 29 Stand-By and Extended Arrangements approved in BRO countries before end–1996, compared with 51 interruptions recorded in the 68 SAF/ESAF arrangements under review.

Table 9.9

Comparison of IMF Staff Resources in ESAF Countries and Baltics, Russia, and Other Former Soviet Union (BRO) Countries1

(Averages per country)

article image
Source: IMF Budget Reporting and Travel Schedule Systems.

Data are for the 36 ESAF countries under review and the 15 BRO countries. Kyrgyz Republic is in both groups. (ESAF arrangements in other BRO countries fall outside the period of this review.)

Article IV consultations, use of IMF resources missions, and other IMF staff visits.

Conducted by staff from the IMF’s Fiscal Affairs, Monetary and Exchange Affairs, and Statistics Departments.

In sum, devoting more attention—and resources—to the monitoring of ESAF-supported programs should be considered. This might include allowing the use of quarterly test dates, and possibly reviews, in ESAF-supported programs when the IMF and the authorities consider that it would help to sustain program implementation. Quarterly reviews could prove particularly helpful when a program is highly vulnerable to shocks of a kind that cannot be easily addressed by contingency planning and automatic adjusters.

Insufficient Contingency Planning

In principle, there are three ways for programs to deal with exogenous shocks. First, program reviews allow for revisions of policy and a reassessment of adjustment and financing needs as the program unfolds. Second, when shocks of a specific kind are considered to be particularly likely, automatic adjusters assigned to key performance criteria (sometimes referred to as in-built contingency mechanisms, or ICMs) introduce an element of contingency planning, in that they determine in advance the mix of adjustment and accommodation that would be required in case of unforeseen deviations from program assumptions. In cases where shocks are sufficiently minor that the fundamental assumptions of a program may still hold, adjusters are an economical alternative to the reopening of program discussions in the context of reviews. However, automatic adjusters do not indicate how any needed additional adjustment—in itself potentially contentious—would come about. A third element of contingency planning, therefore, would involve prior agreement with the authorities on a set of commitments to implement certain kinds of contingent measures in the event of shocks, as needed to keep the program on track.43 These precommitments could allow for greater readiness to cope with shocks, albeit at the cost of more complicated negotiations on the original program, and possibly longer IMF missions.

To assess whether more widespread or effective use of contingency planning might have been helpful in coping with shocks, the design of programs prior to the 14 interruption episodes where past policy slippages were the dominant factor and external shocks or natural disasters were relevant44 was examined to see whether perceived sources of risk or uncertainty regarding exogenous variables prompted the use of ex ante contingency provisions. The two most frequent sources of perceived uncertainty were the extent and timing of external financing, and the evolution of the terms of trade. In 10 of the 14 cases examined—Bolivia 1, Burkina Faso 1, Burundi 1, Guinea 1, Kenya 1, Mozambique 1 and 2, Nicaragua 1, Niger 1, and Pakistan 1—uncertainty about external financing was considered ex ante a source of risk for the program: in all but one case (Niger 1) the risk was on the side of a financing shortfall. In five episodes—Guinea 1, Mozambique 1 and 2, Niger 1, and Pakistan 1—shortfalls or delays in external financing did materialize, and in one other case (Bolivia 1) sizable export payments arrears occurred.45 Yet, 9 out of 10 programs where such risk was perceived—and all 6 where it materialized—did not include an adjuster against a financing shortfall (see Appendix 9.1, Table 9.18). When present, ICMs were in all but one case (Nicaragua 1) asymmetric, in the direction of preserving (adding to reserves) windfall gains from external financing, or adding to credit to the private sector, in the case of adjustors assigned only to net credit to government.

The absence of adjusters was not in itself a program design flaw: these programs implicitly assumed that any financing shortfall would have to be offset fully and immediately by a tightening of policies or a contraction of imports, or dealt with in a subsequent review. This choice may have been justified by the limited scope for greater domestic financing and a weaker reserve position: the level of gross reserves per se was very low (less than two months of import cover) in Guinea 1, Kenya 1, Nicaragua 1, and Pakistan 1, and in most instances there were outstanding arrears to foreign creditors. In no case, however, were the modalities of the additional adjustment effort to address external financing shortfalls specified in advance, and hence agreed by the authorities.46

Although one can only speculate whether the inclusion of contingent measures in the program would in practice have been sufficient to avert an interruption, on logical grounds such inclusion might have increased the probability of a quick and adequate policy response. In Guinea 1, for example, some adjustment was ultimately achieved (albeit too late to keep the program on track) through increases in petroleum product prices and public utility tariffs. Had these measures, which could have been implemented quickly and would have had relatively rapid effects, been prespecified as contingencies, correction might have been accomplished in a more timely manner.47

A deterioration in the terms of trade was perceived as a source of risk in 8 of the 14 episodes examined. It materialized in five of these programs (Burkina Faso 1, Burundi 1, Equatorial Guinea 1, Malawi 1, and Niger 1), as well as in two other instances (Bolivia 1 and Guinea 3). Once again, none of these programs had ICMs. In a minority of cases where the risk was perceived (Niger 1, Pakistan 1, and Togo 1), some contingent measures were discussed to compensate for the potential effects on the balance of payments and the fiscal accounts.48 These contingencies, however, proved insufficient to prevent the program interruption, because of unrelated fiscal policy slippages. The measure finally adopted to counter the budgetary impact of declining world cocoa prices in Equatorial Guinea 1—specifically, the reduction of the administered producer price and export subsidy for cocoa—would have been a good candidate for a contingent commitment.

As for natural disasters, it is unclear whether contingency planning against the effects of droughts or poor rainfall in Kenya 1, Nicaragua 1, and Togo 1 would have helped to avoid the interruptions. In all cases, the shock does not appear to have been a major factor behind the slippage in policies, and, in the case of Nicaragua 1, slippages reemerged even after efforts to identify corrective measures (see Table 9.2 and Appendix 9.1, Table 9.10).

Table 9.10

SAF/ESAF Review: List of Program Interruptions

article image
article image
Source: IMF staff estimates.

Or interval between expiration of last completed annual arrangement and cancellation of the multiyear arrangement or approval of a new arrangement.

For delays in completing a midterm review, the starting date of the interruption is the date by which the review was to be completed. For intervals between multiyear arrangements, the starting date is the expiration date of the last completed annual arrangement: extension periods in the absence of an agreed policy framework for annual arrangement are considered part of the interruption.

Review never completed, but subsequent annual arrangement approved.

Review never completed, and no subsequent annual arrangement approved.

Subsequent annual arrangement under SAF not approved.

Review for third annual arrangement not completed.

Subsequent annual arrangement under ESAF not approved.

Review for fourth annual arrangement not completed.

In sum, strong conclusions about the role of contingency planning are not possible, since many of the program interruptions reflected a combination of shocks and unrelated policy slippages. Also, it is difficult to make an empirical case for the use of contingencies without reliable counterfactuals. Nevertheless, the interruption episodes reviewed indicate that there was little correspondence between perceived risks—such as external financing shortfalls or terms of trade shocks—and contingency planning in programs. It is possible, therefore, that in circumstances where the risk of a sizable shock is considered high, programs could benefit from contingent commitments to specific policy measures—particularly in the budget—accompanied where appropriate by simple and transparent adjusters to performance criteria. Against this, the practical difficulties of negotiating continency measures must be acknowledged, especially when there are other difficult policy issues to be resolved. Moreover, where a country faces various risks, or where the effects of possible shocks are highly uncertain ex ante, neither automatic adjusters nor contingent measures are likely to be feasible or effective. In such circumstances, a preferred approach might be to incorporate more frequent program reviews.

Conclusions and Lessons for Program Design

Although a third of program interruptions resulted either from forward-looking disagreements in the absence of significant policy slippages or from severe political disruptions, deviations from past policy commitments played an important role in most interruptions. These policy slippages were typically broadly based, affecting both the financial program and the structural reform agenda, but in the large majority of cases the root of the problem was fiscal in nature.

An examination of several aspects of program design that might have contributed to interruptions primarily attributable to past policy slippages did not find compelling explanations.

  • Most annual fiscal targets for interrupted programs were not unusually ambitious (accounting for initial conditions) when judged against the benchmark of the average fiscal effort envisaged in all SAF/ESAF arrangements. When targets were ambitious, in most cases “lowering the bar” to match the policies that actually were implemented, so as to avoid an interruption, would have entailed accepting a significant deterioration in macroeconomic conditions, calling into question the consistency with ESAF objectives. Thus, it appears that there was little scope to reduce the incidence of interruptions by setting less ambitious targets.

  • The evidence does not suggest that insufficient prioritization of structural conditionality was a factor behind program interruptions. The weak correlation between the rate of observance and the number of structural conditions set implies that merely cutting back on the number of conditions would not in itself guarantee better implementation. This is not to say, of course, that the priorities chosen in all programs were necessarily optimal (see Chapter 8).

  • In a few countries where implementation capacity was a factor behind weak policies and, ultimately, program interruptions, it seems that a step-up in technical assistance efforts could perhaps have been a necessary but not sufficient condition to prevent a program breakdown. The record underscores the importance of an adequate plan of action when there is a recognized need, based on a realistic assessment of how quickly the authorities can and will absorb the assistance. The ambitiousness of the structural reform agenda should be tailored accordingly. Such plans should be set out in greater specificity in policy framework papers.

  • A comparison between the intensity of program monitoring in ESAF countries and that in the BRO countries with Stand-By or Extended Arrangements suggests that there may be a case for closer monitoring of ESAF-supported programs. Specifically, in situations where IMF staff and authorities agree that it could enhance program implementation, quarterly test dates and reviews might be used in ESAF arrangements. Closer tailoring of disbursements to policy performance could help to prompt an early response to emerging problems, especially in programs subject to considerable uncertainties regarding, for instance, external conditions. Consideration could also be given to more widespread assignment of IMF resident representatives in ESAF countries.

  • Greater use of contingency planning might have helped to anticipate the need for a more timely policy response to exogenous shocks, particularly since, with hindsight, some of the measures ultimately adopted appear to lend themselves to precommitment on a contingency basis. Even here, however, the presence of unrelated policy slippages prevents strong conclusions. Also, in practical terms it seems likely that negotiations of contingency measures would lengthen the process of formulating and agreeing on programs.

In sum, there may be cases where modifications to program design and monitoring could help to improve future program implementation, and these should be actively considered where feasible. But it cannot be claimed that such changes would have been sufficient to avert or significantly reduce the incidence of program interruptions examined in this study. By implication, it appears that most interruptions have in fact been the result of factors that could not have been influenced by program design. In addition to instances of major political upheaval, discontinuities or weaknesses in policy management also appear to have been related, in roughly a dozen cases, to less severe forms of political disruption, including routine elections, transitions to multiparty political systems, and social unrest. These events typically resulted in government overspending and a general distraction of the decision-making authority. In many other episodes, lapses in policy were associated not with specific events of this kind, but rather with a combination of influences, including domestic political opposition, the authorities’ reluctance to confront special interest groups, poor organization, and governance-related weaknesses.

To the extent that such problems can be anticipated, the incidence of interruptions in IMF arrangements might be reduced by seeking greater assurances than in the past regarding the authorities’ ability to carry out policy commitments before proceeding with IMF support. Greater selectivity in this sense could serve not only to provide more protection for IMF resources but also to encourage national authorities in actions (such as more intensive consensus building) that would reinforce the commitment to reform. Such an approach is not without its own difficulties. In particular, possible criteria for selection—the basis on which commitment is to be judged adequate—are far from clear. One area for attention is election cycles, which frequently were a factor in interruptions. Greater selectivity might be achieved by requiring stronger assurances from authorities of their ability to implement programs through these cycles. Other indicators of commitment, however, such as a previously interrupted program or the record in an IMF staff-monitored program, appear not to be particularly telling.49 Nevertheless, without greater selectivity, interruptions are likely to remain a feature of the ESAF experience, as the IMF continues to assist members at the margins of commitment and in the midst of difficult political transitions.

Appendix 9.1. Detail on Program Targets, Interruptions, and Monitoring

Information on interruptions to SAF/ESAF-supported programs—including the factors underlying program disruptions—is presented in Tables 9.109.18.

Table 9.11

SAF/ESAF Review: Factors Underlying Program Interruptions1

article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
Source: IMF staff estimates.

Only interruptions taking place before end–1995 are considered.

Table 9.12

Classification of Critical Factors in 51 Interruption Episodes

article image
Source: IMF staff estimates.
Table 9.13

Factors Underlying Program Interruptions Affected by Extensive Strikes and Social or Political Unrest

article image
article image
Source: IMF staff estimates.

Arabic numerals indicate the interruption of a SAF/ESAF-supported program; Roman numerals indicate annual arrangements; and dates indicate the start of a multiyear arrangement.

Table 9.14

External Shocks in 11 Program Interruptions Affected by Past Policy Slippages1

article image
article image
Source: IMF staff estimates.

In addition to the shocks examined, declines in external assistance affected Ghana 1, Guinea 2, Kenya 1, Malawi 1, and, among cases affected by severe political disruptions, Togo 2. However, these declines cannot be considered as exogenous shocks since they resulted from a poor past record of policy implementation or were linked to political conditions (Kenya 1, Malawi 1).

Specific price projections not available.

Table 9.15

Annual Fiscal Targets Prior to Interruptions Affected by Past Policy Slippages

article image
Source: IMF staff estimates.

Annual fiscal targets not available.

Table 9.16

Structural Conditionality in Programs Prior to Interruptions Affected by Past Policy Slippages

article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
Source: IMF staff estimates.