Chapter 5: Prolonged Use and Effectiveness of IMF-Supported Programs

International Monetary Fund. Independent Evaluation Office
Published Date:
April 2003
  • ShareShare
Show Summary Details

1. Since UFR policies have been designed at least to contain, if not entirely avoid, prolonged use, the pervasiveness of the phenomenon raises questions about the effectiveness of IMF-supported programs in prolonged user countries. This chapter examines this issue at two levels. We first examine available evidence from cross sections of countries to see whether there are any differences in program design or implementation between prolonged and “temporary” users and what can be learned about the impact of IMF-supported programs on growth and adjustment in a prolonged use context.1 Next we focus on lessons that can be drawn from our case studies (Pakistan, the Philippines, Senegal, Morocco, and Jamaica) and from responses to IEO questionnaires sent to authorities in prolonged user countries and to IMF staff.2 Some of the issues that have surfaced from our investigation, especially from the case studies, are of general relevance for program effectiveness and not exclusive to prolonged users.

Results from Cross-Sectional Evidence

Cross-country evidence on the effects of prolonged use

2. The empirical literature on the effects of IMF-supported programs (see Haque and Khan, 1998) has produced widely varying results depending upon the time period covered and the methodology for estimation. The more recent studies using the so-called general evaluation estimator all suggest that programs contribute to improvements in the current account balance and the overall balance of payments, but the evidence with respect to the impact on growth is mixed.3Goldstein and Monteil (1986) and Khan (1990) found negative effects on growth. Conway (1994) finds that initial negative effects on growth are offset by subsequent positive growth rates. However, Przeworski and Vreeland (2000), using a broadly similar approach, find significantly negative and persistent effects on growth. A recent study, Barro and Lee (2002), which used a different (instrumental variable) approach to take account of the endogeneity problem, concluded that while programs do not have a significant contemporaneous effect on growth, they do have a lagged effect that is negative.

3. Not only are the results from the existing literature mixed, as summarized above, they do not distinguish between prolonged and “temporary” user cases. To explore this distinction we requested Professor Lee, one of the coauthors of Barro and Lee (2002), to extend their analysis so as to consider whether “prolonged use” has an effect on growth that is distinguishable from that associated with “temporary use.”4 While the methodological problems are complex and hard to resolve entirely, the results, reported more fully in Annex 4, suggest the following:

  • Programs have significant negative contemporaneous and lagged effects on growth in prolonged users, but no significant effect on growth in temporary users.
  • Limiting the sample to only GRA arrangements, strongly negative contemporaneous and lagged effects on growth are found in prolonged users but not in “temporary” users.
  • When only concessional facility arrangements are considered, there is a negative contemporaneous effect on growth that is more than offset by a positive lagged effect in the case of prolonged users, but there is no significant lagged effect in the case of “temporary” users.

4. From these results, it would appear that IMF-supported programs in the case of prolonged users have adverse consequences on growth in programs supported by GRA resources but not programs based on concessional resources. However, as noted above, previous exercises that have attempted to assess the impact of IMF-supported programs on growth have come to very different conclusions and it appears that the results can be sensitive to variations in approach and sample size. The difficulties in identifying the independent impact of prolonged involvement in IMF-supported programs are likely to be even greater.

5. It is useful to consider whether there are any substantial differences between the extent of adjustment achieved in prolonged and “temporary” users’ programs, although similar methodological difficulties apply. We have used the World Economic Out-look (WEO) database to examine this issue.5

6. In Stand-By Arrangements, prolonged users achieved a somewhat larger adjustment in the current account than “temporary” users, but the extent of adjustment of the overall public sector balance was slightly smaller in the group of prolonged users (Table 5.1).6 However, the results suggest that the shortfalls from the targeted fiscal adjustment—discussed in more detail below—were greater in the case of the group of prolonged users. Fiscal adjustment in prolonged user countries, on average, does not appear to have involved any reduction in government expenditure as a percentage of GDP.

Table 5.1.Actual Change in Key Variables1(Average annual changes in percent of GDP)
Prolonged users“Temporary” users
(T−1) to (T+1)(T−1) to (T+3)(T−1) to (T+1)(T−1) to (T+3)
Overall public sector balance0.751.431.072.36
Central government revenues1.430.790.380.12
Central government expenditures0.34–0.70–0.58–2.52
Current account balance0.971.21–0.303.56
Sources: IMF, WEO database; and IEO calculations.

T refers to the year of approval of the arrangement. Because the data on outcomes are drawn from the WEO database, definitions and sample sizes differ from those used in Table 5.4, making direct comparisons difficult.

Sources: IMF, WEO database; and IEO calculations.

T refers to the year of approval of the arrangement. Because the data on outcomes are drawn from the WEO database, definitions and sample sizes differ from those used in Table 5.4, making direct comparisons difficult.

7. In medium-term arrangements, prolonged users achieved a markedly smaller adjustment of both their current account and fiscal balances over the three-year period. Once again, the differences in the extent of fiscal adjustment are accounted for by a much lower reduction in government expenditure in the group of prolonged users. These differences are largely at odds with those found between adjustment targets (see below), except for the relative sizes of fiscal adjustment.

Cross-country evidence on program design and implementation

8. We begin with documenting whether IMF-supported programs with prolonged users differ from those with “temporary” users with respect to five characteristics: the extent of overoptimism of growth targets; the extent of targeted adjustment and outcomes; the nature of conditionality; degree of implementation; and size of “IMF effort.”7 A variety of databases, covering different time periods, are used.

Projections for exports and real GDP growth8

9. A comparison of medium-term program projections and outcomes in ESAF-supported programs is presented in Table 5.2. It shows that projections for exports were greatly overoptimistic for prolonged users but not for “temporary” users.9 Projections for real GDP growth were generally too optimistic for both prolonged users and “temporary” users. A comparison of the year-by-year differences suggests that the errors in projections for prolonged users’ real GDP growth are greater in later years of multiyear programs than in the first year.

Table 5.2.Optimism of Real GDP and Export Projections in ESAF Programs1
Prolonged users“Temporary” users
Merchandise export growth
Real GDP growth
Sources: IMF Policy Development and Review Department and IEO calculations.

Average annual growth in percent for years T to T+4, where T is the year in which the program started

Sources: IMF Policy Development and Review Department and IEO calculations.

Average annual growth in percent for years T to T+4, where T is the year in which the program started

10. Programs supported by arrangements using general resources (GRA) typically include projections for a much shorter period. A comparison was only possible for real GDP growth projections covering the initial program year, drawing upon the database used by Musso and Phillips (2001).10 Again, the results suggest short-term projections of real GDP growth were optimistic for both prolonged and “temporary” users but the overoptimism was actually greater in the latter case (Table 5.3).11 The substantial differences between the median and mean errors suggest that a few, relatively large errors on the downside affected both “temporary” and prolonged users.12 The Musso and Phillips database does not include export growth, making it difficult to compare projections with outcomes in this dimension in GRA cases. These results suggest that programs with prolonged users under concessional facilities do have a tendency to be overoptimistic about medium-term export growth, which may have implications for program design. There also appears to be some overoptimism in projections of real GDP growth in programs under both GRA and concessional facilities, but in this case the overoptimism does not appear to be systematically related to the phenomenon of prolonged use.

Table 5.3.Accuracy of Short-Term Projections for Users of General Resources1
Prolonged users“Temporary” users
Real GDP growth
Sources: Database assembled by Musso and Phillips and IEO calculations.

The median of the differences is not necessarily equal to the difference of the respective medians.

Sources: Database assembled by Musso and Phillips and IEO calculations.

The median of the differences is not necessarily equal to the difference of the respective medians.

The magnitude of targeted adjustment

11. The strategy for dealing with prolonged use set out by the IMF’s Executive Board called for more “front-loaded” adjustment in such cases. It is, therefore, relevant to ask whether programs with prolonged users targeted greater adjustment in the external current account or fiscal deficit than other cases and whether they were more frontloaded. Because of limitations in the available IMF databases on programs, it is only possible to answer this question for the period since 1993.13

12. There are marked differences between prolonged and “temporary” users’ programs concerning the extent of targeted fiscal adjustment and the current account adjustment but the pattern differs between SBAs and multiyear arrangements.14 In Stand-By Arrangements, prolonged users’ programs targeted a somewhat larger average adjustment of the overall public sector deficit than “temporary” users’ programs (2½ percentage points of GDP compared to 2 percentage points) (Table 5.4). However, if we focus on the change in the primary balance of the central government as the relevant measure of adjustment effort, the targeted fiscal adjustment in prolonged users on average (1.8 percent of GDP) was much lower than that in the case of “temporary” users (3.3 percent). Prolonged users’ programs supported by SBAs were also characterized by a higher share of the targeted fiscal adjustment expected to come from increased revenues, with much lower cuts in primary (i.e., noninterest) expenditure. While the data on outcomes, discussed earlier, are not strictly comparable, they suggest that prolonged users actually achieved less fiscal adjustment, with no expenditure reduction. The external current account adjustment required of prolonged users in SBAs was also much lower than in the case of “temporary” users.

Table 5.4.Targeted Change in Key Variables(Average annual changes in percent of GDP)
Prolonged users“Temporary” users
SBAsMultiyear arrangementsSBAsMultiyear arrangements
(T−1) to(T−1) to(T−1) toShare of(T−1) to(T−1) to(T−1) toShare of
Overall public sector balance22.51.592.92542.01.953.5755
Central government revenues1.550.280.68421.20.921.371
Central government total expenditure–0.78–0.93–2.0246–2.05–1.13–2.0256
Primary central government balance21.790.811.92423.291.852.4974
External current account20.091.073.18340.621.253.3138
Sources: MONA database and IEO calculations.Note: T refers to the year of approval of the arrangement.

Targeted adjustment in the period T−1 to T+1 as a proportion of the targeted adjustment over the period T−1 to T+3.

Positive number implies an improvement in the balance.

Sources: MONA database and IEO calculations.Note: T refers to the year of approval of the arrangement.

Targeted adjustment in the period T−1 to T+1 as a proportion of the targeted adjustment over the period T−1 to T+3.

Positive number implies an improvement in the balance.

13. In the case of multiyear arrangements, the magnitude of fiscal adjustment required from “temporary” users was higher than from prolonged users whether one looks at the overall balance or the primary balance.15 In both cases, the primary expenditure contraction expected was similar but the magnitude of the revenue effort from “temporary” users was much larger. Unlike in SBAs, the extent of current account adjustment in the multiyear programs was of the same order for prolonged users as for “temporary” users. In effect, the fiscal adjustment was targeted to play a much larger role in bringing about adjustment in the “temporary” users than in the case of prolonged users.

14. An important difference between prolonged users and “temporary” users in multiyear arrangements is that the targeted adjustment in all the variables was more front-loaded in programs with “temporary” users, particularly as regards revenue increases and primary expenditure cuts. This is contrary to what might have been expected given the guidelines on prolonged use.

The extent and nature of conditionality16

15. For the purposes of this study, conditionality can be classified into two types: “hard” and “soft.”17 “Hard” conditionality consists of prior actions (PA) and performance criteria (PC), both of which are conditions that the country must meet in order to have access to the IMF resources under a program.18 “Soft” conditionality consists of all the elements that are taken into account by the IMF in forming a judgment about whether or not to “complete” a review, which triggers the release of a financing tranche. It might include structural benchmarks, indicative targets, or general undertakings in the authorities’ letter of intent,19 sometimes expressly identified as elements that will be subject to reviews.

16. Available cross-country data summarized in Table 5.5 indicates that, on average, conditionality in arrangements with prolonged users typically had fewer formal structural conditions than arrangements with “temporary” users, regardless of the type of arrangement.20 Moreover, most of the cases with very heavy use of structural conditionality did not involve prolonged users; many involved the transition economies. For example, only 4 of the 17 programs that had more than ten prior actions per program year in the 1990s involved prolonged users—although Pakistan was one of these cases.

Table 5.5.Average Number of Structural Conditions Per Program Year, 1987–2000
Total conditions1BenchmarksPrior actionsPerformance criteria
Prolonged Users“Temporary” UsersProlonged Users“Temporary” UsersProlonged Users“Temporary” UsersProlonged Users“Temporary” Users
All arrangements6.5**9***3.1*1.00.8
SBA and EFF4.9**8.5**1.8**4.4**
Memorandum items
Source: MONA database.Note: ** and * indicate that the difference between both means is statistically significant at the 1 percent and 5 percent levels, respectively.

Differences between total conditions and the sum of the other three categories are accounted for by “review” conditionality. The difference is particularly large in the case of the Philippines.

Source: MONA database.Note: ** and * indicate that the difference between both means is statistically significant at the 1 percent and 5 percent levels, respectively.

Differences between total conditions and the sum of the other three categories are accounted for by “review” conditionality. The difference is particularly large in the case of the Philippines.

17. Not only was structural conditionality less extensive in prolonged users, but the evidence also suggests that it was “softer”—at least in the narrow sense discussed above.21 In both prolonged and “temporary” users’ arrangements, the largest part of structural conditionality was of the “soft” kind (mostly structural benchmarks). However, arrangements with prolonged users had fewer prior actions, especially in the case of arrangements under concessional facilities.

18. A distinctive feature of GRA arrangements with prolonged users was the comparatively large share of conditions to be appraised by program reviews in total conditionality (10 percent against 5 percent or less in all other cases), which gave the IMF more room for discretion in assessing the progress in structural reforms than other forms of conditionality. For example, the Philippines’ programs came to rely almost exclusively on reviews in implementing structural conditionality because the political system made it hard for the government to commit to a specific timetable for passing measures through Congress. As the discussion in the Philippine case study illustrates, the trade-off for the greater flexibility and discretion of reviews was less certainty about the timing of disbursements, the content of policies, and pressures to “renegotiate the program at each review.”

19. Structural conditionality also tended to get heavier and more constraining over time, but that tendency affected both prolonged and “temporary” users’ programs, and there are no indications that the trend was stronger in the case of the former (see Figure 5.1).

Figure 5.1.Evolution of Structural Conditionality in IMF-Supported Programs

(Number of conditions per program year: left scale; and percent of PCs and PA/CCRs in total: right scale)

Sources: IMF Policy Development and Review Department and IEO calculations.

20. The evidence suggests that macroeconomic conditionality was also more abundant in arrangements with “temporary” users than in those with prolonged users.22 In GRA arrangements, there were on average 31 macroeconomic conditions per program year for prolonged users, and 35 for “temporary” users, probably reflecting a higher frequency of reviews in the latter arrangements. In ESAF arrangements, these numbers were, respectively, 10 and 13. Both results suggest that there was not closer monitoring of performance under programs with prolonged users.

Differences in program implementation

21. Assessing performance in terms of implementation presents problems since any single measure is inherently arbitrary. Table 5.6 presents six different measures of the extent of implementation. They show that, taking all arrangements together, prolonged users’ programs are more subject to interruptions, both “minor” and irreversible. However, disaggregating into programs supported by GRA resources and those supported by concessional resources, we found that prolonged users as a group suffer from such interruptions more frequently than “temporary” users only for GRA-supported programs. Prolonged users score somewhat better than “temporary” users in terms of overall implementation while the program is on track,23 which in the case of GRA-supported programs might suggest greater difficulties to sustain their efforts.

Table 5.6.Selected Data on Program Implementation, 1992–98
Percent of arrangements with at least one minor interruption1Percent of arrangements seriously interrupted2Percent of committed funds disbursed3Average macro implementation index4Average structural implementation index4Average overall implementation index4
All arrangement types
Prolonged users724777816877
Temporary” users683976826775
ESAF arrangements
Prolonged users664480777374
Temporary” users884682786972
GRA arrangements
Prolonged users774974846280
Temporary” users613674836777
Memorandum items
Source: Ivanova and others (2001) database.

A minor interruption is defined as a delay of over three months for the completion of an SBA review or noncompletion; or a delay of over six months for the completion of an ESAF/PRGF review or noncompletion; or an interval of more than six months between two annual arrangements in a multiyear program; or the nonapproval of at least one annual arrangement in a multiyear program.

An irreversible interruption occurs if either (i) the last scheduled program review was not completed (all program types); or (ii) all scheduled reviews were completed but the subsequent annual arrangement was not approved (ESAF/PRGF arrangements only).

Not including precautionary arrangements.

The implementation index is equal to 100 if the condition is met or met after modification, 50 if met after a substantial delay (for structural conditions only), and 0 if not met, waived, or not met after modification.

Source: Ivanova and others (2001) database.

A minor interruption is defined as a delay of over three months for the completion of an SBA review or noncompletion; or a delay of over six months for the completion of an ESAF/PRGF review or noncompletion; or an interval of more than six months between two annual arrangements in a multiyear program; or the nonapproval of at least one annual arrangement in a multiyear program.

An irreversible interruption occurs if either (i) the last scheduled program review was not completed (all program types); or (ii) all scheduled reviews were completed but the subsequent annual arrangement was not approved (ESAF/PRGF arrangements only).

Not including precautionary arrangements.

The implementation index is equal to 100 if the condition is met or met after modification, 50 if met after a substantial delay (for structural conditions only), and 0 if not met, waived, or not met after modification.

22. In theory, these two findings could reflect either a higher propensity of prolonged users to encounter serious policy slippages, or a more rigid stance of the IMF in appraising whether a given slippage can be corrected (and thus whether a waiver of the missed PC is warranted). However, the data indicate that prolonged users, particularly in GRA programs, were granted waivers on a slightly larger proportion of their structural PC than “temporary” users, whereas there is almost no difference between the two groups as regards waivers of quantitative PC (Table 5.7).24

Table 5.7.Use of Waivers, 1987–20001
Number of waivers granted per program yearPercentage of PC waived
“Temporary” users1.
PRGF-eligible average1.
GRA-only average1.
Prolonged users1.
PRGF-eligible average1.
GRA-only average0.
Memorandum items
Source: IMF Policy Development and Review Department database on waivers.Note: PC: performance criteria; QPC: quantitative performance criteria; SPC: structural performance criteria; and CPC: continuous performance criteria.

Not including waivers of applicability.

Source: IMF Policy Development and Review Department database on waivers.Note: PC: performance criteria; QPC: quantitative performance criteria; SPC: structural performance criteria; and CPC: continuous performance criteria.

Not including waivers of applicability.

23. According to IMF internal guidelines, waivers are to be used to deal with minor deviations from agreed targets, considered to be of a temporary or reversible nature. The fact that prolonged users had more of both program interruptions and waivers suggests that, in their case, waivers were more often followed by program interruptions, which might indicate that a higher proportion of waivers granted to prolonged users turned out, in hindsight, to be unwarranted.25 This could be either because the IMF had a higher tolerance threshold regarding the magnitude of deviations waived for prolonged users or because the corrective actions implemented were more frequently insufficient.

24. Even though prolonged users of ordinary resources were more prone to serious program interruptions, a similar proportion of funds committed got disbursed in their arrangements. This reflected a somewhat more front-loaded pattern of disbursements for prolonged users—which was not consistent with the strategy for prolonged use established by the Board.26

25. These findings are broadly consistent with recent internal reviews of the modalities of conditionality, which concluded that (i) there was no clear relation between the volume or modalities of conditionality (in particular, the quantity of prior actions) and the probability of success of an IMF-supported program and (ii) waivers had been predominantly used to address significant policy slippages, not—as they were originally intended—minor/technical factors or exogenous developments. They also suggest that the IMF’s practice with respect to prolonged users has not been wholly consistent with the guidelines calling for more intensive monitoring of programs in prolonged use cases (see Chapter 3).

Differences with respect to IMF staff inputs27

26. We find that arrangements with prolonged users were less resource intensive than those with “temporary” users, but staff turnover was high for both groups. Details are provided in Annex 6.

27. Arrangements with “temporary” users were actually more mission intensive than arrangements with prolonged users, especially when the negotiation period is taken into account.28 Most of the difference appears to stem from the implementation/supervision phase rather than from the design phase. These findings are difficult to interpret. One cannot conclude that greater staff input could have helped to avoid prolonged use because causality is elusive. It may be that the greater intensity of inputs into “temporary” use cases reflects the operation of other factors.29

28. Both “temporary” and prolonged users suffer from high staff turnover.30 On average, program countries had four different mission chiefs over the last five years with very little difference between the two groups.31 Turnover among other mission members was also high for both groups. These findings are consistent with comments received from prolonged users’ authorities as well as from IMF staff. Both groups identified excessive staff turnover as a significant problem. Although the evidence suggests that the problem is not specific to prolonged users, such a high turnover risks undue disruptions to the quality of country work and of the relationship with the authorities. These detrimental effects could be more costly in the case of prolonged users, owing to the importance of track record and learning curve issues in their case.

Evidence from the Case Studies and Responses to Questionnaires

29. In this section, we present some evidence related to program design and implementation arising from the three detailed country case studies and the more limited case studies of two “graduators” from prolonged use. In all three of the countries that are the subject of full case studies, some progress toward the resolution of their economic problems was eventually achieved under IMF-supported programs, but this progress was uneven across areas, often took much longer to achieve than initially expected, and generally fell short of what was targeted when the programs were agreed upon (Box 5.1).

30. Drawing upon the evidence from the case studies and on the questionnaire responses from a broader group of prolonged users, we identify a series of problems that appear to have combined to limit the effectiveness of IMF-supported programs in these prolonged use countries. These are a discrepancy between the time frame of programs and the magnitude of their objectives; insufficient attention to program ownership and implementation capacity; issues related to the design of conditionality, especially in dealing with core institutional changes; issues related to the financial programming framework; and issues related to the IMF’s exit strategy for prolonged users. Most of these program design issues appear to have resulted less from weaknesses in the staff’s technical analyses than from a variety of institutional factors, which are discussed in Chapter 6.

31. Before turning to these issues, we should emphasize several points:

  • There are obvious limits to how far any conclusions can be generalized from a small number of country cases—especially ones that, by selection, have encountered difficult and protracted adjustment processes. We are not suggesting that all, or even most, programs supported by the IMF suffer from the range of problems identified here, not even among programs with prolonged users. We do not have sufficient evidence to make such a determination. Even among the case study countries, experiences have varied significantly. Nevertheless, many of the questionnaire responses suggest that the issues identified do not just reflect isolated occurrences.

Box 5.1.Overview of What Was Achieved Under IMF-Supported Programs in the Country Cases


In Pakistan, little or no lasting progress was achieved on the macroeconomic front, except as regards inflation, which was halved to 4 percent over the 1990s. Both the fiscal and the current account deficits remained high throughout the period and stood, respectively, at 5 and 4 percent of GDP in 2000, while international reserves remained critically low (Table 5.8). Meanwhile, growth declined to an average of 4 percent a year over 1988–2000 and poverty rose steadily, with 30 percent of the population living below the poverty line in 2000.

Progress in structural reforms was mixed. There was substantial change in some of the core areas of the IMF’s mandate, for example, liberalization of trade and external payments, public debt management and the conduct of monetary policy, including independence of the central bank. Progress in tax reforms and administered prices (including utilities) and in financial sector supervision was more protracted; but it was eventually significant, although by no means complete. In other critical areas, such as tax administration or public enterprises, only small improvements had begun to materialize at the very end of the decade. Furthermore, the quality of economic governance generally deteriorated during the period of intensive UFR, though for reasons largely unrelated to IMF-supported programs.


In the Philippines, a long series of programs in the 1960s and 1970s were associated with a period of quite strong growth but heavy external borrowing. They achieved little lasting adjustment and failed to prevent a debt crisis in 1982–83. A large part of the problem appears to have been the difficulty of promoting lasting adjustment in the face of deep governance problems and a lack of political commitment to reform at the highest level, although—reflecting the general approach at the time—program documents tended not to discuss such issues. The eventual crisis was exacerbated by adverse global developments. Subsequently, a considerable degree of adjustment and structural transformation was achieved in the second half of the 1980s and the 1990s under successive programs, although progress took place in spurts and was slower overall than anticipated. Regarding macroeconomic developments, a rapid turnaround of the current account was achieved in the first half of the 1980s but only at the cost of a sharp output decline. Tight monetary policy was also instrumental in preventing a take-off in inflation, which remained moderate subsequently. A substantial current account deficit reemerged, however, in the late 1980s and for most of the 1990s, despite rapid export growth facilitated by a realignment of the incentives structure. Substantial debt restructuring combined with export growth was, however, instrumental in the gradual resolution of the Philippines’ external debt problem. The debt to GNP ratio was halved to 50 percent between 1985 and 1995. There was also considerable fiscal adjustment, with the consolidated public sector deficit reduced significantly from over 7 percent of GNP in 1984–85 to just 0.6 percent of GNP in 1993–97, before rising again to 3.7 percent of GNP in 1998–2000.

In the area of structural reforms, the programs of the 1980s achieved substantial progress in liberalizing the trade and exchange system, the abolition of marketing monopolies in the agricultural sector, public enterprise privatization, and financial sector reform. However, there was also a large unfinished agenda. These reforms were advanced further in the 1990s, with additional trade liberalization, the recapitalization of the central bank, the opening of critical sectors to new competition and foreign participation, and acceleration of the privatization program. There was also progress—although slower than desirable—in strengthening the prudential framework of the financial sector. However, despite these gains, a few long-standing and critical weaknesses, notably a low saving rate, problems with tax collection, and an inefficient public sector, remained unresolved. For example, tax/GDP ratios improved substantially in the mid-1990s, but collapsed again after 1998 as collection problems and other weaknesses reemerged. Moreover, despite some progress, the incidence of poverty remained high and widespread corruption concerns lingered.


Senegal’s adjustment process has been of a stop-and-go variety. After a severe financial crisis in the early 1980s, a period during which implementation of IMF-supported programs was weak, there was a period of improved implementation during 1984–88, which contributed to a marked decrease in fiscal and external current imbalances as well as to much reduced inflation. However, that period saw the emergence of a banking crisis. In the early 1990s, growth slowed down, as “internal” adjustment measures implemented by the authorities did not succeed in addressing competitiveness problems. A nearly two-year period of interruption in IMF arrangements occurred in 1992–93. The devaluation of the CFA franc in 1994 and accompanying measures gave new impetus to the authorities’ adjustment effort. A surge in inflation was rapidly contained, fiscal and external imbalances were reduced, and real GDP growth has been at a steady 5–6 percent per annum. During the later part of that period, Senegal’s debt problem was also significantly reduced, mainly by debt relief (after the debt was officially recognized as unsustainable in 2000). Since 2000, however, financial imbalances have reemerged, suggesting that the stop-and-go pattern has yet to be eliminated.

In the area of structural reforms, substantial progress was eventually made—broadly following the same stop-and-go pattern—in the areas of price liberalization, trade liberalization, and simplification of the tax system. Generally speaking, the role of the state in the economy has been significantly reduced and the economy has become more resilient to terms of trade shocks, in part thanks to a markedly reduced share of groundnuts and phosphates in total exports. By contrast, in spite of repeated attempts to tackle them over the last twenty years, the problems of the groundnut sector, the petroleum and power sectors and tax collection remain largely unsolved. Besides, with virtually no growth in per capita GDP in the 1980s and the first half of the 1990s, poverty has remained widespread.


Morocco faced very large domestic and external imbalances and a heavy debt burden in 1980 when it entered a long series of IMF-supported programs. It also faced major structural weaknesses, with heavy government regulation of the economy, including on consumer prices and credit allocation, and a strong inward economic orientation. Programs in the early 1980s had only a moderate impact in reducing these imbalances. Fiscal adjustment was initially achieved mainly through expenditure reduction, but substantial domestic arrears occurred. At first, progress on structural reforms was slower than planned, in part due to concerns over their social acceptability.

In the second half of the 1980s, financial adjustment policies and structural reforms became more mutually reinforcing, although progress was not smooth: there were a number of setbacks in the face of exogenous shocks and slippages in the timetable of policy implementation. Nevertheless, reinforced by a strong political backing for the broad direction of key reforms and an effective civil service, substantial progress was eventually achieved. Programs were successful in raising national saving (from an average of 16 percent of GDP during 1980–82 to almost 23 percent during 1990–92), with an improved tax effort contributing to these gains. Debt and debt-service ratios were reduced substantially, with the aid of significant debt restructuring. Significant structural change was also eventually achieved, including the liberalization of most foreign transactions and consumer prices; reform of public enterprises; and tax and public expenditure reform. Although significant challenges remained, including the need to generate faster growth, Morocco was able to “graduate” from IMF-supported programs in 1993.


The case of Jamaica was examined not to analyze in-depth the effects of its IMF-supported programs, but to illustrate the importance of domestic ownership of policies and the potential benefits of considering alternative policy options. Jamaica had a long series of arrangements with the IMF stretching from the 1960s until the last EFF expired in March 1996, after which the Jamaican government announced its “independence from the IMF,” indicating that it would not borrow again. At this time, Jamaica still faced large adjustment challenges. Public debt was high (over 100 percent of GDP, with over two thirds external); inflation was over 20 percent; the real effective exchange rate was appreciating; growth remained weak; and the first stages of a major financial sector crisis was under way.1 The authorities rejected the IMF’s policy advice on several key aspects (notably on a substantial depreciation of the exchange rate and the closure of weak financial institutions) in favor of a “homegrown” macroeconomic program that included running very large primary fiscal surpluses and tight monetary policy to reduce gradually inflation and inflationary expectations. The IMF staff initially doubted that, in the likely protracted low-growth environment that such a stabilization strategy would involve, the authorities would be able to maintain the sizable fiscal effort necessary to avoid unsustainable public debt dynamics.

In the event, the government did manage to generate and maintain large primary fiscal surpluses and, as the authorities’ “homegrown” policy strategy was seen to be implemented quite forcefully, the IMF backed away in subsequent Article IV surveillance reports from its previous insistence on an initial large depreciation, although differences of view remained on the appropriate degree of exchange rate flexibility. In July 2000, it was agreed that there would be IMF staff monitoring of the government’s economic program—an option the authorities pursued in order to obtain an IMF “seal of approval” and consequent access to adjustment lending from the multilateral development banks. The authorities’ program achieved some important results, although Jamaica’s problems are far from resolved. High primary surpluses prevented the public sector debt dynamics from deteriorating further, although it still remains vulnerably high; real interest rates, although still very high, have declined moderately; and the cushion of external reserves has improved. Growth has only recently begun to recovery moderately, after a number of years of stagnation.

Whatever the relative merits of the two different policy strategies—and, it is not obvious that the one favored by the IMF was better—the Jamaican experience illustrates that, once the strong political commitment to the “homegrown” strategy was taken into account, its prospects for success were definitely higher.

1 See Part II on the case studies for further details.
Table 5.8.An Overview of Economic Performance in the Three Country Cases(Period averages in percent, unless otherwise indicated)
GDP growth1Pakistan3.
Overall budget deficit (percent of GDP)1Pakistan−7.6−8.0−6.1−7.3−7.6−6.5
Tax revenues (percent of GDP)1Pakistan10.312.310.612.415.616.0
Government expenditure (percent of GDP)1Pakistan16.917.419.023.323.622.2
Public debt (percent of GDP)1Pakistan66.956.854.473.876.579.1
Exports growthPakistan11.
Current account balance (percent of GDP)1Pakistan−4.7−4.6−2.7−2.6−3.6−4.8
External debt (percent of GDP)1Pakistan52.
Gross international reservesPakistan0.
(months of imports)Philippines1.
Gross international reservesPakistan9.911.515.
(percent of external debt)Philippines32.823.06.37.917.025.8
Sources: IMF staff reports and WEO database.

GNP rather than GDP in the case of the Philippines.

National government balance.

Underlying consolidated public sector balance.

Government revenue excluding grants.

Exports of goods 1971–80 for the Philippines.

Sources: IMF staff reports and WEO database.

GNP rather than GDP in the case of the Philippines.

National government balance.

Underlying consolidated public sector balance.

Government revenue excluding grants.

Exports of goods 1971–80 for the Philippines.

  • A number of the issues discussed are not particular to cases of prolonged use—or even more especially prevalent in such cases than others. But the problems identified do seem to have lengthened the IMF’s program involvement in these specific country cases, and they suggest a number of lessons that are worth emphasizing.
  • We were not able to quantify the specific contribution of program design issues to prolonged use, as opposed to other factors, including those that are beyond the IMF’s control—such as the authorities’ actions. However, based on all the evidence analyzed, our judgment is that these issues were significant.
  • Finally, some of the problems identified are already well known, since they have been discussed in various previous assessments. For some, important initiatives are already under way to try to address them, as will be discussed in Chapter 8.

Box 5.2.Were Program Projections Overoptimistic? Lessons from the Case Studies

An examination of initial projections underlying all programs since 1983 for the three country cases as well as Morocco confirms the earlier cross-section evidence that there has been a strong tendency to overestimate export projections in all cases (see table below and Figures 5.2 and 5.3). Real GDP growth was also overestimated, on average, in all four countries. Projections of government revenue were markedly optimistic in Pakistan and the Philippines, but not in Morocco or Senegal. Projections of national saving also proved too high in all countries other than Morocco. In general, the extent of overoptimism embedded in programs was greatest in Pakistan and least in Morocco, which is not surprising given the respective track records on program implementation. It is not possible to ascertain from these comparisons whether it was unrealistic ex ante program projections that contributed to weak implementation or vice versa, and it is not possible to estimate precisely the counterfactual of what would have happened if all the programs had been implemented as agreed. Nevertheless, the more extensive discussion in the country studies suggests that, in the case of Pakistan in particular, both the authorities and staff recognized that programs were often built on very optimistic projections.

Realism of Program Projections: Average Projections Less Outturns1(Percentage points a year)
Real GDP growth1.
Export growth (in U.S. dollar terms)
Fiscal balance (in percent of GDP)
Government revenue (in percent of GDP)1.31.14
National saving (in percent of GDP)2.30.552.8–0.8
Source: IMF staff reports.

Average of all initial projections for programs since 1983, for the year in which the program started and the two succeeding years.

Growth and ratios expressed in relation to GNP, rather than GDP.

For Morocco, except for export growth, projections are for the year in which the program started and the immediately succeeding year, due to the limited time horizon of projections in program documents.

National government tax revenue as percent of GNP.

The apparent rise in the saving rate in the late 1990s and early 2000s may be overstated as a result of statistical weaknesses.

Source: IMF staff reports.

Average of all initial projections for programs since 1983, for the year in which the program started and the two succeeding years.

Growth and ratios expressed in relation to GNP, rather than GDP.

For Morocco, except for export growth, projections are for the year in which the program started and the immediately succeeding year, due to the limited time horizon of projections in program documents.

National government tax revenue as percent of GNP.

The apparent rise in the saving rate in the late 1990s and early 2000s may be overstated as a result of statistical weaknesses.

Time frame of programs

32. The case studies highlight many instances in which programs tended to “overpromise,” both on the speed of restoration of macroeconomic sustainability and on the pace at which structural reforms could be implemented. On the macroeconomic side, the tendency was most marked in the case of Pakistan, but was also present in expectations about Senegal’s graduation from reliance on debt relief. It is interesting to ask whether the overoptimism observed ex post was only because policies agreed under the program were not always implemented (Box 5.2). Weaknesses in implementation clearly played an important part, as the detailed discussion in the country notes illustrates; and it is no coincidence that the “overoptimism” was least in Morocco, which had the best track record on implementation. However, repeated underestimation of the obstacles to policy implementation is, in itself, a program design problem. Moreover, some programs—especially those with Pakistan in the 1990s—relied on projections for exports and tax revenue that would have been optimistic even with full implementation. (Figures 5.2 and 5.3.) On the structural side, the overoptimism—about both the length and diversity of the reform agenda embedded in programs—occurred in all three cases.

Figure 5.2.Prolonged Users’ Exports

(In billions of U.S. dollars)

Sources: Data provided by the national authorities and IMF staff reports.

Figure 5.3.Prolonged Users’ GDP Growth

(Change in percent)

Sources: Data provided by the national authorities and IMF staff reports.

33. As the discussion in Chapter 6 will illustrate, these factors seem to have stemmed in considerable part from institutional pressures to produce substantial visible progress within the relatively short time frame of the program. Interestingly, these pressures were present even in those cases, such as the Philippines, where staff reports were relatively candid from an early stage (i.e., following the 1982–83 debt crisis) that adjustment was going to take considerable time. In other words, recognition that IMF program involvement was likely to be lengthy does not appear to have changed fundamentally the approach to program design.

34. The relatively short time frame of program design also meant that, in some cases, less priority was given initially to those elements of the structural reform agenda that were inevitably going to take a significant amount of time, even though they were important for long-term sustainability or when their nonimplementation caused adverse side effects and ex post sequencing problems. For instance, in Pakistan, early programs envisaged a simultaneous cut in trade taxes and creation of a broad-based general sales tax. While the former re-form was implemented broadly on schedule, the latter took over ten years to be fully effective, giving rise to large revenue shortfalls in the interim. In all three of the main country cases, the strengthening of regulatory and supervisory systems lagged too far behind financial liberalization with adverse consequences for the banking system.

The extent and nature of conditionality

35. The cross-section evidence discussed earlier suggests that—although there were wide variations from country to country—prolonged users as a group had less structural conditionality and relied upon conditionality that was “softer,” in the sense of being less directly monitorable (i.e., fewer prior actions and performance criteria). We use the case studies to examine, in a more qualitative manner, whether these factors influenced program effectiveness and hence prolonged use.

36. The three case studies illustrate rather different approaches to conditionality. In Pakistan, structural conditionality was extensive and very detailed, with a total number of conditions per program year well above the average of other programs whether with “temporary” or prolonged users (Table 5.5). These conditions overwhelmingly took the form of bench-marks and performance criteria. In the Philippines, by contrast, conditionality relied primarily on reviews, with practically no benchmarks and a minimal number of performance criteria (0.2 per program year on average). Total formal conditions per program year were less than half as many as the GRA prolonged users’ average. In Senegal, the approach adopted was broadly consistent with ESAF/PRGF prolonged users’ average, with a heavy reliance on structural benchmarks and a somewhat above-average recourse to performance criteria. However, all three countries share a below-average use of prior actions. Interestingly, in spite of these differences in the overall approach to conditionality, all three countries had similar and significantly above-average implementation indices—as measured by the degree and timeliness of compliance with program conditionality as long as programs are on track (Table 5.6).32

37. Our analysis of the nature of conditionality in the three cases—details are provided in the country studies—suggests a number of lessons. We are not implying that these lessons are germane only to prolonged use cases, or that their full adoption would have prevented prolonged use. Many are now well recognized within the IMF. Nevertheless, they are important for program effectiveness and are worth reiterating here.

(i) The specific structure of conditionality is much less important than an underlying domestic political commitment to core policy adjustments. Conditionality can be potentially useful as a device to signal that commitment and as a device for monitoring progress, but it does not appear to have been effective at enforcing changes in the cases where the necessary political commitment was absent. In that sense, the critical factor contributing to prolonged use was not that the guidelines calling for different approaches to conditionality in such cases were not followed—which, the cross-country evidence discussed earlier shows was often so—but that the IMF often proved reluctant to be more selective in its involvement when political commitment to implement necessary changes was lacking. The case of Pakistan is one example. The Philippine case also suggests that the modalities of conditionality were not the critical factor explaining the relative lack of success in strengthening tax administration—since a variety of approaches were tried without lasting success, including prior actions, benchmarks, reviews, and detailed commitments summarized in a policy matrix.33 In Senegal, restructuring of the groundnut sector remains incomplete in spite of various forms and types of conditions contained in programs since the early 1980s, in part reflecting policy reversals associated with social and political sensitivities.

(ii) Excessively detailed conditionality—whether resorted to because of a weak track record, doubts about ownership, or to support reform-minded groups within government—does not appear to have been effective in enhancing implementation.34 One example is the effort to implement tax administration reforms in the Philippines under the 1998 SBA.35 The “Memorandum of Economic Policies” contained an extremely detailed matrix of commitments, but this was agreed primarily with officials of the outgoing Ramos administration, and the incoming officials were not committed to it. In the event, the “matrix approach” proved ineffective, including as a monitoring device, because it was relatively easy to find superficial ways to meet the commitments. In Senegal, in response to weak implementation, there was a sharp increase in the number of structural conditions and an escalation from “soft” to “hard” conditions during the second and third annual arrangements under the 1998 ESAF/PRGF. This was only partially effective as some measures were implemented with significant delays or in ways that did not meet program objectives (e.g., the withdrawal of the state from the collection and transport of ground-nuts did not lead to the liberalization envisaged in the program, as the authorities continued to set indicative margins rather than allow the market to determine transportation and collection costs).

(iii) This is not to say that the form taken by conditionality is irrelevant. In particular, there is some evidence from the case studies that conditionality that focused on policy rules or procedures, rather than discretionary, one-time actions, was ultimately more effective.36 For example, in Pakistan this approach was eventually adopted with some success in the area of tax exemptions as well as in the area of utilities price adjustments. However, the Senegal case study shows that reversals can occur even with policy rules, as happened with an automatic pass-through mechanism for the retail prices of petroleum products that was suspended prior to presidential elections in 2000.

(iv) The evidence is mixed as regards the effectiveness of prior actions. Historically, as noted in Chapter 3, staff reports and Executive Board meetings discussing strategies to deal with prolonged use had supported front-loading the adjustment and re-form effort with a greater use of prior actions, while back-loading disbursements to provide an extra incentive to sustain the reform effort over the span of the program. In practice, most of the programs in Pakistan and the Philippines featured neither front-loaded structural reforms nor back-loaded disbursements—which is consistent with the cross-country evidence discussed earlier. The one program that most clearly did both (i.e., the 2000 Pakistan SBA) was implemented well, but it unfolded in an environment characterized by strong political commitment, which makes it hard to disentangle the independent impact of front-loading of reforms. In Senegal, disbursements were generally not back-loaded, but some programs did front-load the policy effort, including through prior actions—largely when the political circumstances, including the electoral cycle, were favorable—and these tended to be more successfully implemented.

(v) A closer analysis of some prior actions used in the three country cases casts light on the somewhat counterintuitive conclusion of several recent studies, according to which the number of prior actions has not had a significant influence on program implementation.37 A large part of the problem was that the prior actions chosen were not always well-integrated into the program design. The discussion in the Pakistan country note of the prior actions related to agricultural taxation in the 1993 SBA and 1994 ESAF is especially informative:38 both prior actions were judged to be met and yet neither resulted in meaningful taxation of agricultural incomes. Agricultural taxation became the focus of prior actions because conditionality in previous programs had been ineffective (i.e., there was a weak track record on the issue) and because it was highly desirable on equity grounds. But the prior actions specified were not critical to achieving the programs’ macroeconomic objectives, since the revenues they would have raised, even if effectively implemented, were small. Moreover, other potential prior actions, that were more macro-critical, appear to have been dropped during the process of negotiation (e.g., extension of the GST base or reintroduction of the petroleum price adjustment mechanism). This example illustrates two lessons: (i) prior actions, like any other conditionality, can be subject to superficial or temporary observance if domestic ownership/political commitment is weak; and (ii) prior actions imposed for “symbolic” reasons, rather than in view of their criticality for the achievement of program objectives, do not enhance program effectiveness.39

(vi) Conditionality is especially difficult to apply to complex regulatory and institutional issues that are often critical to achieving longer-term sustainability and avoiding prolonged use. In these circumstances, use of reviews based on an assessment of outcomes is probably a better approach, rather than attempting to split the reforms into a detailed timetable of discrete, intermediate steps that are then subject to conditionality. But the Philippines’ experience also suggests that while a use of reviews and a focus on broad outcomes is the best way of monitoring progress with complex reforms, even this approach has its drawbacks, since there was often a lack of clarity over the “bottom line” of conditionality. Each review became an occasion for “recontracting” the underlying commitments, which weakened their credibility. In retrospect, greater specificity in identifying a small number of critical outcomes, with progress assessed through reviews, may have been desirable.

(vii) The case studies provide some evidence that the credibility of conditionality can be eroded by many repeated programs. According to many Pakistan officials, the expectation that the IMF would eventually provide financing—by agreeing either to waivers or to new arrangements shortly after program interruptions—weakened incentives to tackle the fiscal deficit forcefully. Similarly, in the Philippines repeated programs appear to have fostered the view that policy commitments could be readily renegotiated. In Senegal, the views expressed on this question by various stakeholders were mixed.

Dealing with core institutional and structural changes

38. As discussed earlier, part of the explanation for prolonged use in all three countries was that a number of intractable issues took a very long time to correct and some have still not been resolved. This reflects the fact that attempts to fit complex reforms into the time frame and conditionality framework of IMF-supported programs sometimes caused priority to be given to easily measurable actions over more complex and more important institutional changes. For example, in all three countries, tax administration reforms were recognized as centrally important to achieving longer-term revenue improvements, but were often given less explicit focus in programs than more visible tax policy changes, even when technical assistance was offered in parallel to programs to address them. This was in part because the impact and timing on revenues was less easily measurable or could result in an initial revenue loss that would have complicated program design in the short run. The choice to pursue these reforms through means other than program conditionality appears to have sent a signal—or created an incentive—to treat them as secondary priorities. While this may be appropriate in a near crisis context, it is not effective in a de facto long-term relationship between the IMF and the member.

39. Other areas where a stronger early emphasis on institutional reforms in macro-critical areas might have mitigated subsequent problems were in the banking sector (in particular, risk management practices and prudential regulations, especially in Pakistan and the Philippines) and public enterprises (in all three countries). For example, in Pakistan, public enterprise adjustment in programs was not addressed from a broader institutional reform perspective until the end of the 1990s. Because earlier programs did not effectively come to grips with the broader reform needs of these enterprises, they tended to focus on tariff adjustments that although warranted from a purely fiscal perspective, implied an acceptance of low efficiency levels in these enterprises leading to a higher cost structure for Pakistan’s industry. As discussed in the Pakistan study (see Part II, Chapter 9), these problems reflected, in part, a failure of Bank-Fund collaboration to focus on such issues in an operationally effective manner at an early stage.

Ownership and assessments of feasibility

40. In each of the three main country cases, the risks to the programs of weak political commitments were often understated. For example, reports to the Executive Board on Pakistan generally downplayed the effects of the considerable political instability that prevailed throughout 1988–2000 on governments’ willingness and ability to implement far-reaching reforms—although there was more discussion of such issues in internal documents. Similarly, in the Philippines, programs tended to underestimate the difficulties of pushing reforms through Congress—although staff did make significant efforts to interact with key congressional committees in an effort to enhance domestic commitment. Coverage of such issues in internal documents was variable and only a few (e.g., an internal ex post assessment of the Philippines 1989–91 EFF) provided a candid assessment of such constraints. Even in these cases, subsequent Board papers were less candid. In general, most Board papers on the programs had no significant assessments of ownership. This is not surprising since this only became an operational concern very recently, but it points to areas where future policies should be different (see Chapter 6).40

41. There was also relatively little presentation in internal briefing papers—and even less in Board papers—of the trade-offs between potential alternative strategies, including in cases where there were substantial divergences of views between the staff and the authorities.41 Furthermore, only limited attention was often paid to assessing and developing implementation capacity, both in a technical and political sense.

Issues related to the IMF’s financial programming framework42

42. There was broad recognition—both in the country cases and in the responses to the questionnaire sent to other prolonged users—that one of the most valuable contributions of programs has been in focusing attention on a sound macroeconomic framework and in providing a consistency check on the key components of that framework. However, several weaknesses in the approach, as it is implemented in practice, occurred frequently. Once again, these issues are of broader relevance and not just for the prolonged users.

(i) Too little attention was often paid to analyzing the real economy dynamics and the expected sources of growth. In some cases (e.g., most of the Pakistan programs during the 1990s), overoptimistic growth and revenue projections in effect “squared the circle,” allowing projected fiscal deficits to appear consistent ex ante with other macroeconomic objectives, and thereby avoiding some difficult fiscal choices. The result was to force ad hoc policy adjustments that were inconsistent with the medium-term strategy and were generally not sustainable. Similarly, overoptimism about the speed of recovery in private investment, in the face of a continuing debt overhang, was a problem with the 1989 Philippines EFF.43

(ii) Many programs had difficulty in dealing with uncertainty. As in all macroeconomic policymaking, program design faced enormous uncertainties—including about the nature of behavioral relationships, key international prices, supply conditions, and the pace of implementation of reforms. But many program documents did not spell out the key risks facing the program, nor did they conduct suitable stress testing exercises. While the actual adaptation of programs to unanticipated events has to be left largely to reviews,44 the limited ex ante discussion of the major risks and of how policies and targets could respond to deviations from program assumptions meant that there was generally too little mid-course reconsideration of the logic of program design until programs were already close to being off-track. As a result, policy adjustments were often too slow and risked being inconsistent with the long-term strategic objectives of the program (e.g., short-term expenditure squeezes or ad hoc revenue measures).45

In this respect, a large proportion of responses to the questionnaire noted that programs paid insufficient attention to how policies would respond to external shocks. However, a number of the more recent programs mark some improvement in that respect—for example, the 2000 SBA of Pakistan contained a candid discussion of risks and uncertainties. The guidelines and framework for assessments of external sustainability endorsed by the Executive Board in June 2002, which are meant to be applied in priority to program countries, are a further step in that direction and, if implemented consistently, could bring about significant improvements.

Box 5.3.Exits from IMF-Supported Programs: A Comparison of Morocco and the Philippines

Morocco “graduated” from the use of IMF resources in 1993 and the Philippines in 2000. A review of various economic indicators suggests that their positions were not that different in the early 1990s: the Philippines had a higher current account deficit and lower, albeit still comfortable reserves, but had lower debt and debt-service ratios. The fact that the Philippines had almost fully liberalized its capital account, while Morocco still had a more restrictive system, could have implied the need for a higher reserve cushion, but the Philippines also had a more flexible exchange rate regime.

Current account balance
(percent of GNP or GDP)1−7.0−4.011.52−8.7−2.0−1.7
External debt to GNP or GDP ratio373.462.163.1100.289.048.3
Debt-service ratio
Before rescheduling348.124.014.649.738.119.3
After rescheduling44.319.714.633.938.119.3
Fiscal deficit
(percent of GNP or GDP)
Gross national saving
(percent of GNP or GDP)121.420.028.3218.620.423
Reserves (months of imports)
Inflation (percent)
Per capita income (U.S. dollars)6638601,0397371,1101,159
Source: IMF staff reports.

Ratios and growth rates are in terms of GNP for the Philippines and GDP for Morocco.

May be overstated as a result of statistical weaknesses (see the Philippines case study for details).

Public and publicly guaranteed debt for Morocco.

Underlying consolidated public sector deficit for the Philippines; central government overall deficit (payments basis) for Morocco, excluding privatization receipts.

Source: IMF staff reports.

Ratios and growth rates are in terms of GNP for the Philippines and GDP for Morocco.

May be overstated as a result of statistical weaknesses (see the Philippines case study for details).

Public and publicly guaranteed debt for Morocco.

Underlying consolidated public sector deficit for the Philippines; central government overall deficit (payments basis) for Morocco, excluding privatization receipts.

(iii) Many officials as well as many staff also noted that too much of the time available for program negotiations (as well as the authorities’ subsequent monitoring efforts) was spent on “fine-tuning” the details of the financial programming exercise—by more than was justified given the inevitable uncertainties about the underlying behavioral relationships and economic environment. As one response to the questionnaire put it, there was “too much focus on fine-tuning of technical issues with not enough attention to higher quality conditionality.”

Lack of well-defined exit strategies

43. The problems discussed above were sometimes accompanied by too broad a rationale for the IMF’s involvement through a program relationship that inevitably encouraged prolonged use. This appears to have been the case in the Philippines for parts of the 1990s. Such an approach appears to have reflected, from the authorities’ perspective, uncertainties as to the effects that putting an end to the 30-year program relationship would have on markets. From the standpoint of the IMF, it reflected a belief that a continued involvement would foster “good” policies, including by enhancing the leverage of domestic reformers, or (in 1998) would help to sustain earlier gains by avoiding backsliding during a change of administration. Indeed, a comparison of the situation of the Philippines and Morocco in the early 1990s does not suggest any clear reasons why different approaches to an “exit” from IMF resources was taken (Box 5.3).46 Similarly, there appears to have been little discussion of a possible exit strategy in the case of Senegal.

44. This experience suggests that, as more PRGF-eligible countries move toward having eliminated their structural balance of payments imbalances, the lack of a well-defined exit strategy—or criteria to guide such a strategy—could contribute to a more prolonged use of IMF resources.


The cross-country results discussed in this section use the same basic definition of prolonged use set out in Chapter 2. Since the time periods covered by the various databases differ, the precise group of countries that are defined as prolonged users varies depending upon that time period. However, because prolonged use is highly persistent over time, the population of prolonged users does not vary greatly over time and the results do not seem very sensitive to this factor.


The questionnaire sent to prolonged users is shown in Annex 5.


For a survey of the literature, see Haque and Khan (1998). A significant recent development is Barro and Lee (2002).


We also requested that he expand the coverage of the earlier Barro and Lee (2002) analysis to include arrangements under concessional facilities.


Since the IMF’s internal (MONA) database on programs has incomplete information on outcomes and the coverage of the WEO database is different in several respects, it is not possible to make a systematic direct comparison of outcomes and program targets, as discussed below.


None of the differences reported in this section is statistically significant. The usual caveats about inferring causality apply.


It should be borne in mind that whatever specificities of prolonged users this analysis reveals may reflect a variety of factors and that there is no proven causal relationship with prolonged UFR.


The analyses of ESAF-supported programs in this section are based on the database on medium-term program targets and outcomes prepared by PDR for the 1997 ESAF Review. More recent data were not available on a consistent basis for use by the IEO. The analysis of GRA-supported programs is based on the database used by Musso and Phillips (2001), which draws on the MONA database and covers 69 programs in 47 countries during 1993–97. The latter database excludes precautionary and SAF/ESAF programs. For purposes of the analysis, countries are defined as prolonged users if they met the definition used in this evaluation either at the beginning of the program included in the database, or if the program in question subsequently contributed to their classification as a prolonged user.


In neither case, however, is the bias statistically significant, given the large variability of both projections and outturns for exports.


The study looks at projections for the current calendar year when a program is approved more than three months before the end of the year, otherwise for the following year. Thus, the effective length of projections varies between 3 and 15 months.


This result contrasts somewhat with the evidence from the Musso and Phillips study that “follow-up” programs have a larger bias in growth projections than other programs in their sample. “Follow-up” programs are defined as new programs for which there was some form of IMF arrangement active in the preceding year. This is because in the Musso-Phillips sample, a large number of countries that had one (immediate) follow-up program did not go on to be “prolonged users” under the definition used for this study.


Among prolonged users, the programs with substantial overoptimism about GDP growth included Mexico (1995), Bulgaria (1996), and Romania (1997).


The analysis in this section draws upon the IMF’s MONA database, which has comprehensive data on program targets only for the period since 1993 and has incomplete data on outcomes. Therefore, the group of prolonged users for the purposes of the analysis of this subsection refers to those countries that either were prolonged users in 1993 or became prolonged users subsequently.


None of the differences discussed here is statistically significant at the 5 percent level, due to very large variations within each group.


As noted earlier, the actual fiscal adjustment achieved by prolonged users in multiyear arrangements was also smaller.


The findings of this section are based on two databases: one, assembled by the Policy Development and Review Department for the 2001 “Review of Fund Conditionality,” is focused on structural conditionality and covers all IMF-supported programs over 1987–2000. The other database was assembled for an internal research project on the determinants of the effectiveness of IMF-supported programs, whose findings are reported in Ivanova and others (2001). It covers all conditionality contained in the approximately 170 arrangements approved between 1992 and 1998. The two datasets yield broadly similar conclusions even though they cover different periods.


This classification of “hard” and “soft” conditionality is not one used formally by the IMF and it refers only to the form of conditionality, not to the strength of the underlying measure.


However, they differ in the sense that prior actions are often imposed before the approval of an arrangement, whereas performance criteria are specified in the arrangement itself and only apply to subsequent disbursements. Prior actions linked to subsequent reviews rather than the program’s initial approval are sometimes referred to as “conditions for the completion of a review.” Until 2000, prior actions were generally not specified in the text of Executive Board’s decisions, which gave them a certain degree of informality (along with a lack of transparency). In both cases, the country may be granted access to the IMF’s resources even though not all conditions have been met, but in the case of performance criteria, this will require a decision by the Executive Board granting a waiver, whereas for prior actions, a judgment from management that a critical mass of measures has been taken would suffice.


However, not all general undertakings in the authorities’ letter of intent qualify as conditionality strictly speaking.


When transition countries are excluded from the sample, all the means reported in the table decline, and differences between prolonged and “temporary” users became somewhat smaller. However, the thrust of the comparison remains the same, with the exception of total conditionality in ESAF/PRGF arrangements, which was marginally higher for PUs than for TUs.


Among the case study countries, Pakistan was an exception.


Data on macroeconomic conditionality suffer from various availability and reliability problems, which limits the robustness of any conclusions. While initial observations by PDR, based on a small sample of arrangements, suggested that the total number of macroeconomic conditions (i.e., target values for key macroeconomic aggregates presumed to be under the control of the authorities) per test date varied little from program to program, the data collected in Ivanova and others (2001), which are reported here, do show significant variations, perhaps capturing differences in the frequency of test dates (this is because a macroeconomic indicator monitored, for instance, on a quarterly basis will count as four conditions during a year).


Existing data resources do not allow us to track conditionality implementation in programs that go off-track and the indices reported here do not incorporate any measure of actual completion of programs. To that extent, they are biased toward overstating actual policy implementation.


However, when the Ivanova and others (2001) database is used to determine the percentage of macro PC waived, a noteworthy difference surfaces in ESAF/PRGF arrangements, where 17 percent of PC was waived for prolonged users, compared with 12 percent for “temporary” users. It may be recalled that the Ivanova database, unlike the one used in Table 5.7, reports the actual number of macroeconomic PC in programs instead of approximating them based on the number of test dates.


This is only a conjecture, since it is possible that program interruptions were caused by factors totally unrelated to the waived conditions. Only a case-by-case analysis would allow firm conclusions to be drawn. However, the case studies suggest that the waivers were often related to significant policy slippages. For example, according to the decisions, none of the waivers granted to Pakistan or the Philippines since 1983 was motivated by exogenous shocks or technical reasons. By contrast, two of the three waivers on quantitative performance criteria granted to Senegal were motivated by exogenous shocks. This contrast is consistent with the differences noted between GRA and ESAF arrangements in this respect. The five waivers on quantitative performance criteria (along with two others on structural performance criteria) granted to Pakistan at the second review of its SBA in December 1996 are an example of waivers granted in the face of major policy slippages. The program fell apart soon after.


On average, since 1993, the share of total program commitments disbursed in the first tranche has been 26 percent higher for prolonged users than for “temporary” users, suggesting more front-loaded disbursements in the former case.


Comparisons of IMF financial effort were also undertaken, based on the ratio of gross IMF financing to gross financing requirements over the program period, for which available data cover the 1990–2001 period. The differences found between prolonged and “temporary” users were small and statistically insignificant.


These conclusions are consistent with those suggested by data compiled by the Office of Internal Audit and Inspection (OIA) of the IMF in the course of one of its recent reviews and provided to the IEO, which confirms that prolonged users receive, overall, fewer missions than “temporary” users and also revealed that, for non-PRGF-eligible countries, this difference is compounded by a smaller average mission size. In PRGF-eligible countries, by contrast, mission size has generally been larger for prolonged than for “temporary” users (except in FY2001). A large part of the difference in mission size was accounted for by staff from functional departments.


For example: because (i) programs with “temporary” users are less likely to have been immediately preceded by another arrangement, more preparatory work and consultations with the authorities and other stakeholders are needed to design the program; or (ii) programs with “temporary” users suffer from fewer fatal interruptions which means that, on average, more reviews get completed, with the associated mission work; or (iii) they have more conditions, thus requiring a more intensive monitoring process. Econometric tests on the determinants of success of adjustment programs (for instance, Ivanova and others (2001) and Dollar-Svensson (2000)) found that, once the endogeneity of institutional effort was taken into account, it did not have a significant impact on the probability of success of a program. Interestingly, this conclusion held regardless of the sign of the correlation: Ivanova and others (2001) found that the IMF seemed to invest more resources into successful than failed programs, while Dollar-Svensson (2000) found that the World Bank invested more efforts in operations that were eventually unsuccessful.


Based on an analysis of mission travel data compiled by the OIA, as an extension of their previously mentioned work on mission organization and management. The data analyzed cover the period FY1996–2001.


This count does not take into account cases where the permanent mission chief was outranked by a member of the area department front office as ad hoc head of mission.


As noted earlier, the fact that implementation is not recorded systematically in the MONA database once a program is off-track biases this measure upward, especially in the case of Pakistan.


However, owing to the strong and consistent resistance of the authorities, almost no “hard” conditionality was used, in particular no quantitative performance criteria on tax revenues or indicative targets creating a presumption of corrective action when breached, such as were used in Senegal and Morocco for instance.


This is not to suggest that overly detailed structural conditionality was especially prevalent in prolonged cases; indeed, the broader cross-country evidence discussed earlier suggests it was not, with the exception of a few countries, including Pakistan.


See the Philippines study (Part II, Chapter 10, section on “Improving the tax structure and strengthening tax administration: an example”).


Elborgh-Woytek and Lewis (2002) provide a detailed assessment of how IMF conditionality has operated in the case of privatization of state enterprises in Ukraine and come to a similar conclusion.


See, for instance, IMF (2002d) or Thomas (2002).


See the Pakistan study (Part II, Chapter 9, section on “Lack of ownership and inconsistent monitoring resulted in poor implementation”).


Prior actions can also turn out to be counterproductive when they force the hasty adoption of a measure, through procedures that subsequently put its implementation at risk. This happened in the Philippines with a prior action for the 1994 EFF on expanding the VAT base, which although formally passed in 1994 was not implemented until 1996 because of a judicial challenge that might have been avoided if the authorities had followed a different, but lengthier, procedure for its adoption.


Cordella and Dell’Ariccia (2002) also suggest a number of reasons why ex ante knowledge of a country’s degree of commitment to a program can help reduce the costs associated with the imposition of suboptimal levels of conditionality.


A notable exception in Senegal relates to the 1992 Article IV staff report and (especially) briefing papers for possible UFR in 1993 that contrasted the authorities’ purely “internal” adjustment strategy with a more “comprehensive” approach—including exchange rate action—favored by the staff.


See Mussa and Savastano (1999) and Khan and Knight (1985) for a discussion of this framework.


Such overoptimism about the pace of response of private investment does appear to be a significant problem in the design of many programs. See, for example, Goldsbrough and others (1996).


The one example in the case studies where a program attempted to prespecify, in a fairly rigid quantitative manner, how the mix of adjustment and financing would respond to various exogenous shocks was the Philippines’ use of the Compensatory and Contingency Financing Facility (CCFF) along with the 1989 EFF. It proved cumbersome and ineffective. See the Philippines case study in Part II, Chapter 10.


Senegal’s 1994 SBA and the 1994–97 ESAF arrangement contained understandings that in the event that world prices for groundnut products and cotton turned out to be lower than projected, any associated fiscal shortfalls would be corrected by revenue-raising or expenditure-reducing measures. However, this approach appears to rule out the possibility of allowing an increase in the fiscal deficit to accommodate a temporary terms of trade shock. Moreover, the nature of the revenue and expenditure measures is not discussed, which increases the risk of ad hoc measures that are not consistent with the medium-term growth objectives.


Clearly, the onset of the Asian crisis in 1997 justified renewed IMF financial support, but this was not a factor in deciding on the nature of IMF involvement in 1994.

    Other Resources Citing This Publication