Key Messages
In most PRGFs, key strategic priorities and policy choices in both macroeconomic and structural areas in program design are still not guided by the PRSP. This largely reflects lack of specificity and other weaknesses in the latter.
The key features of the PRGF imply a much greater need than previous approaches for the IMF to draw upon other sources of expertise and integrate them in an effective and timely manner in program design. The actual record is mixed and the partnership framework of the PRS approach has not been used sufficiently to set and implement priorities in this regard.
Programs target smaller and more gradual fiscal adjustment than under the Enhanced Structural Adjustment Facility (ESAF) and give more weight to revenue increases than expenditure contraction. However, these changes are not always reflected in actual outcomes.
Program design does not exhibit generalized “aid pessimism,” that is, programs generally build in increases in net external financing, and targeted financing exceeds outcomes on average. However, there is still no effective operational approach to managing the tensions between “ambition” and “realism” in defining medium-term external resource envelopes, and what catalytic role the IMF is to play in practice remains vague.
Expenditures designated as poverty reducing have increased significantly under the PRGF, but the case studies suggest caution in concluding that all of it is truly “pro-poor.”
Programs exhibit a strong tendency to eliminate double-digit inflation, but there is no evidence of a systematic “disinflationary bias” when inflation is already low.
Country-specific analysis of how quickly the real economy responds to macroeconomic policy settings is limited. The IMF generally contributes little to inform or help the government inform the public debate on these issues, including by spelling out the assumptions underlying its program design.
PSIA is far from being “mainstreamed” in program design. Although some progress is evident, priority setting on what the BWIs themselves should be delivering in this area is lacking.
IMF structural conditionality has declined significantly under the PRGF and become more focused on core areas of expertise, but conceptual differences between Bank and Fund conditionality and a lack of systematic monitoring mean it is not possible to say what has happened to aggregate IMF–World Bank conditionality.
Evidence suggests only minor improvements in program implementation under the PRGF.
Key Features of PRGF-Supported Programs1
1 Based on IMF (2000a); and Gupta and others (2002).It is not possible at this stage to assess the effectiveness of PRGF-supported programs in achieving ultimate goals such as poverty reduction, since progress toward these objectives can only be measured over a longer time horizon.1 However, we can evaluate progress against a number of intermediate objectives that are in turn expected to yield better growth and poverty outcomes. We organize the discussion broadly around six of the seven key features that PRGF-supported programs are expected to have (Box 4.1). (The first feature—greater ownership and broad participation—has already been addressed in Chapter 3.) We conclude with a discussion of some program design issues that have been the subject of frequent external commentary and an assessment of whether the new approach has improved program implementation.
Alignment with the PRSP
PRGF-supported programs are supposed to be embedded in the overall strategy for growth and poverty reduction. The IMF has distinguished two aspects to this alignment: (i) temporal alignment (i.e., between the formulation cycles of the PRGF-supported program and the PRSP and between the PRSP and the national budget cycle); and (ii) policy alignment (i.e., of program content with the strategy and priorities of the PRSP). An obvious prerequisite is a well-articulated PRS so that the program has something to align with.
Thus far, most PRGFs do not meet the temporal alignment test because they preceded the PRSP (Figure 4.1). To a considerable extent, this is a transitional issue reflecting the “conversion” of previous ESAFs. But even in cases when new PRGFs were negotiated, slightly more than half were finalized before the PRSP. This does not necessarily mean the PRGF has not been informed by ongoing discussions on the broader strategy, but it does raise doubts about the claim that the program design is guided by the PRSP. In fact, the case studies suggest that the reverse influence is more common, with the PRSP drawing key elements of its macro-economic framework from negotiations on a PRGF-supported program.2
Which Came First, the Chicken or the Egg? Sequencing of PRSPs and PRGF-Supported Programs1
Source: IMF staff reports.1 Based on a sample of 35 countries with “full” PRSPs as of end-2003, comparing date of PRSP approval by country authorities with the start of new PRGF arrangements, if any.2 Excludes cases of PRGFs converted from ESAFs.To assess the extent of alignment for the major fiscal variables, we compared the macroeconomic forecasts in original PRSPs for the six IEO case study countries with those of the PRGF-supported program requests or reviews brought forward immediately prior to the adoption of the PRSP. Results varied from country to country, with no clear tendency in the extent or direction of alignment. Most PRSPs and subsequent PRGF-supported programs were presented within 6 months of each other, which should have facilitated alignment. Among the case studies, the exception was Nicaragua for which there was a 16-month gap between presentation of the Nicaragua PRSP and finalization of the PRGF-supported program (which likely explains the absence of alignment). The PRSPs for Tanzania and Vietnam were presented within a month of the subsequent PRGF review, but the macroeconomic forecasts differed noticeably. Mozambique presented mixed results, with the fiscal deficit forecast in the PRSP identical to that of the previous program review but the growth forecast significantly higher. In Guinea, the PRSP macroeconomic forecast was essentially extracted directly from the program request 8 months prior and was already out of date by the time of the PRSP. For Tajikistan, the program’s macroeconomic forecast was quite closely aligned to that of the PRSP prepared 5 months previously.
However, the more important question is not “are the numbers the same,” but whether the PRS drives key strategic tradeoffs. This is hard to assess but qualitative evidence from the case studies suggests that, with the important exception of protection of priority expenditures, strategic macro-economic priorities set out in the PRSP were not used to guide subsequent key policy trade-offs when the initial macroeconomic framework was thrown off track. As noted in Chapter 2, many PRSPs simply do not provide sufficient strategic direction to guide such trade-offs when the particular numerical targets of the PRSP are overtaken by events. However, Tanzania provides an interesting example where the framework of consultations established as part of the PER process did provide a guide for modifying the PRGF—even if the original PRSP document did not (Box 4.2).
The problems associated with determining an appropriate medium-term external resource envelope add to the difficulties of “aligning” the PRSP and PRGF. IMF staff is asked to assume a substantial role in the estimation of this envelope. Internal IMF guidance has advised staff to “present normative (often stable or increasing) projections of grants and concessional loans” and to “demonstrate efforts to seek higher aid commitments in cases where needed and appropriate.” 3 Staff is also expected to take account of the macroeconomic effects of additional external financing, including possible pressures for currency appreciation as well as absorptive capacity concerns. However, the operational framework within which staff would help to formulate such projections remains unclear.4
Tanzania: Alignment, Fiscal Flexibility, and Program Design
Tanzania provides an interesting example of how an initial program design was modified to take account of increased aid availability. The initial macroeconomic framework underlying the 2000 PRGF-supported program heavily influenced that of the subsequent PRSP. The overall budget deficit (after grants) was targeted to remain at or under 1 percent of GDP during the three-year program. Although macroeconomic stability had been restored, the program continued to envisage negative net domestic financing of the fiscal position. There were signs early in the program’s implementation that priority expenditures were being squeezed in order to maintain fiscal discipline. Donors and other stakeholders were concerned that the aims of the PRSP were not being met. Under the auspices of the public expenditure review (PER) system—an institutionalized system of consultations, including a macroeconomic subgroup whose membership includes government officials, donors, and civil society—donors funded studies by an outside academic advisor that provided important in-puts into the debate (see Bevan, 2000 and 2001).
The debate centered around two issues. First, in light of higher concessional financing and a sharp accumulation of reserves, the medium-term expenditure frame-work required modification in order to accommodate these inflows. Second, the envisaged “crowding in of the private sector” via fiscal restraint that had characterized the initial program design had in practice over-estimated the speed of response of private sector credit demand.
The IMF did show flexibility in adapting the financial program to the revised circumstances. The overall deficit (after grants) was allowed to increase almost threefold between 2000/01 and 2002/03 to 4.3 percent of GDP. Priority expenditures almost doubled as a share of GDP. The bulk of the expenditure expansion was financed by increased grants and concessional inflows, and the government ceased targeting negative net domestic financing.
Tanzania’s experience suggests several important lessons. First, the consultation processes under the PRS did, over time, have a significant impact on the content of macroeconomic policies, and the PRGF-supported framework did show flexibility. In this sense, there was eventually an “alignment” with the key objectives of the PRS, although the process was not driven by preparation of the original PRSP document. Second, the process of interaction and debate that led to revisions in the framework involved some, but not all, stakeholders: government, donors, and the IFIs were involved, but civil society did not participate in any major way on this issue. This appears to have reflected in part their technical capacity constraints. But the example suggests that even this should not have been an insuperable obstacle, because one of the key inputs was provided by an outside academic advisor. This relative lack of civil society participation was unfortunate, since it has influenced civil society’s perceptions of the process. Indeed, evidence from interviews and the stakeholder survey indicate a striking lack of recognition within civil society that the macroeconomic framework did adapt.
Experiences with Setting the Medium-Term External Resource Envelope
A review of the country case studies reveals a range of experience but sheds little light on what might be a preferred operational approach to establishing medium-term external financing frameworks under the PRS.
In both Guinea and Vietnam, initial Fund projections for the external resource envelope turned out to be overly optimistic, albeit for very different reasons.1 In Tajikistan—where the bulk of donor support was in the form of loans rather than grants—the PRGF-supported program projected a sharp fall in external financing due to a somewhat arbitrary limit imposed by staff on external borrowing. In Tanzania and Mozambique, staff initially incorporated an assumption of a decline in “aid dependence” into their forecast—partly because of the experience under ESAF when projections of external assistance proved too optimistic—but IMF staff eventually adapted to higher levels of aid.2
A few PRSPs (but none of the case study countries) have included more than one macroeconomic scenario. We reviewed the experience of Senegal and Madagascar. In Senegal, the PRSP described high, medium, and low case scenarios based on varying assumptions about the availability of external financing, the mobilization of domestic resources, and, most importantly, the capacity to absorb external resources. Medium and low case scenarios were apparently introduced at the request of the BWIs but, in keeping with the desire to adopt a framework consistent with achieving the MDGs, the PRSP adopted the high scenario as its base case. Many donors considered this to be unrealistic, particularly given growth assumptions well above the historical average and the country’s capacity constraints to execute investment projects. Consequently, the PRGF-supported program was based on the medium case scenario. It is difficult to pronounce on whether the use of alternative scenarios added value to the PRSP process that warranted the additional effort on the part of the country. However, since the main problem appears to have been capacity constraints that impeded the full utilization of the resources made available, the higher (MDG scenario) does not seem to have played a significant catalytic role in practice.
Madagascar adopted two scenarios based on different growth assumptions. This was partly in response to the concern expressed in the JSA for the I-PRSP that growth assumptions were optimistic and that alternative scenarios would need to be developed. Neither scenario was identified as a base line. The first scenario was optimistic in the staff’s judgment, being based on what staff considered to be unrealistic assumptions of project implementation capacity; the second scenario—which was based on assumed failures to meet targets for mobilizing resources and for implementing the main reforms—was below staff’s assessment of what was likely over the medium term. As a result, the scenario that underpinned the PRGF-supported program takes a middle road between the two PRSP scenarios. Donors, for the most part, have taken a similar stance, awaiting evidence—perhaps in the context of the annual review of the PRSP—before moving in the direction of either of the two PRSP scenarios.
It is too early to draw definite conclusions from these limited examples whether the “alternative scenarios” approach can help catalyze additional aid flows.
1 In Vietnam, because of low drawdown rates on aid commitments. In Guinea, because political considerations and doubts about the authorities’ track record led to a withdrawal of donor support. 2 A number of external reviews by NGOs have criticized the Mozambique program for targeting declining aid levels in the face of massive MDG-related needs. In fact, aid was projected to be broadly unchanged in dollar terms over the medium term but declining as a share of rapidly growing GDP.There is an obvious tension between ambition and realism. In 2003, the staff suggested a possible approach to resolving this tension, built around maintaining two macroeconomic frameworks within the PRSP (one based on more conservative estimates of aid flows and the other related to a more ambitious “business plan” to achieve desired poverty reduction goals).5 At the time, many IMF Executive Directors felt that this approach would impose substantial additional costs on the countries concerned, for little practical gain. Most Directors preferred that the PRSP and PRGF be based on a common realistic macroeconomic framework, with policy responses identified for both more favorable and downside risks. In practice, individual cases in which countries have chosen to present two alternative scenarios have been received quite favorably by the Board, but there remained considerable ambiguity about the operational approach that is supposed to guide the IMF staff’s role in this area (Box 4.3). In early 2004, IMF management endorsed an approach where staff would support country-led or donorled initiatives to develop alternative frameworks aimed at showing what resources would be required for PRSP targets (including MDG-related goals) to be met, but it is too early to assess the impact of this decision.
Another aspect of alignment is whether the structural measures in PRGF-supported programs are derived from the PRSP. We have attempted such an assessment for the six IEO case studies (Table 4.1).6 There is considerable cross-country variation but for the group as a whole, only one-fifth of structural measures listed in the program documents seem to be directly aligned with concrete policy actions described in the PRSP. A much higher proportion are associated with the broader objectives expressed in the PRSP, but such broad alignment is not difficult given the “broad tent” approach of most PRSPs. Program measures for Vietnam show the highest degree of alignment with the PRSP, while programs for Guinea and Mozambique have particularly weak alignment.
Alignment Between PRGF Structural Conditionality and the PRSP
Percentage of conditions aligned with concrete policy action envisaged in PRSP (i.e., rating 3 in coding scheme). Coding scheme for calculating average rating:
1 = No items in PRSP related to PRGF structural conditionality.
2 = Alignment of structural measure with broad objectives of the PRSP.
3 = Alignment of structural measure with concrete policy action envisaged in PRSP.
Alignment Between PRGF Structural Conditionality and the PRSP
| Guinea | Mozambique | Nicaragua | Tajikistan | Tanzania | Vietman | All Countries | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which age aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | |
| Prior actions | 4 | 0 | 1.3 | 7 | 0 | 1.1 | 3 | 33 | 2.0 | 5 | 20 | 2.2 | 0 | … | … | 6 | 50 | 2.5 | 25 | 20 | 1.8 |
| Performance criteria | 3 | 33 | 2.3 | 1 | 0 | 2.0 | 3 | 33 | 2.0 | 4 | 25 | 1.8 | 3 | 0 | 1.3 | 2 | 0 | 2.0 | 16 | 18.8 | 1.9 |
| Structural benchmarks | 3 | 0 | 1.7 | 3 | 33 | 2.3 | 9 | 44 | 2.3 | 5 | 0 | 1.4 | 5 | 40 | 2.0 | 9 | 44 | 2.4 | 34 | 32 | 2.0 |
| Other | 26 | 8 | 2.0 | 0 | … | … | 0 | … | … | 0 | … | … | 0 | … | … | 0 | … | … | 26 | 8 | 2.0 |
| Total | 36 | 8 | 1.8 | 11 | 9 | 1.8 | 15 | 40 | 2.1 | 14 | 14 | 1.8 | 8 | 25 | 1.7 | 17 | 41 | 2.3 | 101 | 20.8 | 1.9 |
Percentage of conditions aligned with concrete policy action envisaged in PRSP (i.e., rating 3 in coding scheme). Coding scheme for calculating average rating:
1 = No items in PRSP related to PRGF structural conditionality.
2 = Alignment of structural measure with broad objectives of the PRSP.
3 = Alignment of structural measure with concrete policy action envisaged in PRSP.
Alignment Between PRGF Structural Conditionality and the PRSP
| Guinea | Mozambique | Nicaragua | Tajikistan | Tanzania | Vietman | All Countries | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which age aligned1 | Average rating | Total | Of which aligned1 | Average rating | Total | Of which aligned1 | Average rating | |
| Prior actions | 4 | 0 | 1.3 | 7 | 0 | 1.1 | 3 | 33 | 2.0 | 5 | 20 | 2.2 | 0 | … | … | 6 | 50 | 2.5 | 25 | 20 | 1.8 |
| Performance criteria | 3 | 33 | 2.3 | 1 | 0 | 2.0 | 3 | 33 | 2.0 | 4 | 25 | 1.8 | 3 | 0 | 1.3 | 2 | 0 | 2.0 | 16 | 18.8 | 1.9 |
| Structural benchmarks | 3 | 0 | 1.7 | 3 | 33 | 2.3 | 9 | 44 | 2.3 | 5 | 0 | 1.4 | 5 | 40 | 2.0 | 9 | 44 | 2.4 | 34 | 32 | 2.0 |
| Other | 26 | 8 | 2.0 | 0 | … | … | 0 | … | … | 0 | … | … | 0 | … | … | 0 | … | … | 26 | 8 | 2.0 |
| Total | 36 | 8 | 1.8 | 11 | 9 | 1.8 | 15 | 40 | 2.1 | 14 | 14 | 1.8 | 8 | 25 | 1.7 | 17 | 41 | 2.3 | 101 | 20.8 | 1.9 |
Percentage of conditions aligned with concrete policy action envisaged in PRSP (i.e., rating 3 in coding scheme). Coding scheme for calculating average rating:
1 = No items in PRSP related to PRGF structural conditionality.
2 = Alignment of structural measure with broad objectives of the PRSP.
3 = Alignment of structural measure with concrete policy action envisaged in PRSP.
Fiscal Flexibility and Fiscal Adjustment
An important feature of the PRGF is the recognition of the need for greater “fiscal flexibility,” which would include accommodation of higher aid flows and “pro-poor” public expenditures in support of PRSP goals while maintaining macroeconomic stability. Programs are also intended to allow greater flexibility in accommodating unexpected changes in revenue or financing, including aid flows. We look first at what happened to key program targets and then discuss fiscal outcomes.
Fiscal targets in PRGF-supported programs: what has changed?
We analyze whether there have been significant changes in fiscal program design using data on a broad cross-section of PRGF- and earlier ESAF-sup-ported programs. A comparison of key fiscal targets over a three-year horizon suggests the following (Table 4.2):
On average, PRGF-supported programs target a smaller and more gradual fiscal consolidation than under the ESAF.7 The magnitude of fiscal adjustment under PRGF is 1–1.5 percentage points of GDP smaller than under ESAFs.8 Programs approved in 2002 and 2003 were more accommodating than those approved earlier (Table 4.3). This distinction is of importance when we discuss outcomes, because information is not yet available on whether the actual results for the later PRGFs match the more flexible targets.
The composition of envisaged fiscal adjustment is markedly different, with PRGF-supported programs relying on revenue increases for virtually all of the adjustment. PRGF-supported programs, on average, target no expenditure cuts while their ESAF counterparts projected a significant reduction.9 If we look only at the first two years of the targeted path, PRGF-supported programs provided for a small increase in expenditure. Once again, the adaptation in program design seems to have occurred over time: when we limited the sample to the “early” PRGFs, we found no statistically significant difference between ESAFs and PRGFs.
Projections of net external financing in PRGF-supported programs are substantially more optimistic than their ESAF counterparts.10 Under ESAF, average external financing as a share of GDP was projected to decline steadily over the three-year horizon. Under the PRGF, such financing was projected to increase over the first two years, with a small decline in the third year.
The difference between ESAFs and PRGFs is particularly pronounced with respect to grants. PRGF-supported programs on average project a significant increase in the first year, falling off by the final year but remaining marginally positive, whereas a decline was projected each year, on average, under the ESAF.
PRGF-supported programs project a widening of the external current account deficit (after official transfers) on average over a three-year horizon compared with a narrowing in the deficit under the ESAF.11
Targeted Fiscal Adjustment in ESAF- and PRGF-Supported Programs, 1995–20031
(Averages, in percent of GDP)
The maximum sample size of all arrangements is 88 (47 ESAFs and 41 PRGFs). However, because of missing observations, some of the averages are based on fewer observations than others.
The sum of grants and net external borrowing in the government accounts.
Targeted Fiscal Adjustment in ESAF- and PRGF-Supported Programs, 1995–20031
(Averages, in percent of GDP)
| ESAFs | PRGFs | |||||||
|---|---|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||||
| T−1 | T | T+1 | T+2 | T−1 | T | T+1 | T+2 | |
| Fiscal balance, including grants | –4.1 | 0.7 | 1.4 | 2.0 | –4.9 | 0.7 | 1.2 | 1.6 |
| Fiscal balance, excluding grants | –8.2 | 0.9 | 2.2 | 3.0 | –8.4 | –0.2 | 0.6 | 1.5 |
| Total revenues, excluding grants | 17.8 | 0.5 | 0.9 | 1.3 | 18.5 | 0.4 | 0.9 | 1.6 |
| Total expenditures and net lending | 25.7 | –0.3 | –1.2 | –1.6 | 26.9 | 0.5 | 0.2 | –0.1 |
| Grants | 3.9 | –0.2 | –0.7 | –1.0 | 3.6 | 0.8 | 0.7 | 0.1 |
| Net external financing2 | 6.7 | –0.6 | –1.4 | –1.9 | 6.4 | 0.7 | 0.8 | –0.3 |
| External current account balance (including official transfers) | –7.2 | 0.4 | 0.5 | 1.1 | –7.3 | –0.8 | –1.7 | –1.8 |
The maximum sample size of all arrangements is 88 (47 ESAFs and 41 PRGFs). However, because of missing observations, some of the averages are based on fewer observations than others.
The sum of grants and net external borrowing in the government accounts.
Targeted Fiscal Adjustment in ESAF- and PRGF-Supported Programs, 1995–20031
(Averages, in percent of GDP)
| ESAFs | PRGFs | |||||||
|---|---|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||||
| T−1 | T | T+1 | T+2 | T−1 | T | T+1 | T+2 | |
| Fiscal balance, including grants | –4.1 | 0.7 | 1.4 | 2.0 | –4.9 | 0.7 | 1.2 | 1.6 |
| Fiscal balance, excluding grants | –8.2 | 0.9 | 2.2 | 3.0 | –8.4 | –0.2 | 0.6 | 1.5 |
| Total revenues, excluding grants | 17.8 | 0.5 | 0.9 | 1.3 | 18.5 | 0.4 | 0.9 | 1.6 |
| Total expenditures and net lending | 25.7 | –0.3 | –1.2 | –1.6 | 26.9 | 0.5 | 0.2 | –0.1 |
| Grants | 3.9 | –0.2 | –0.7 | –1.0 | 3.6 | 0.8 | 0.7 | 0.1 |
| Net external financing2 | 6.7 | –0.6 | –1.4 | –1.9 | 6.4 | 0.7 | 0.8 | –0.3 |
| External current account balance (including official transfers) | –7.2 | 0.4 | 0.5 | 1.1 | –7.3 | –0.8 | –1.7 | –1.8 |
The maximum sample size of all arrangements is 88 (47 ESAFs and 41 PRGFs). However, because of missing observations, some of the averages are based on fewer observations than others.
The sum of grants and net external borrowing in the government accounts.
Targeted Fiscal Adjustment in PRGF-Supported Programs, 2000–03
(Averages, in percent of GDP)
The sum of grants and net external borrowing in the government accounts.
Targeted Fiscal Adjustment in PRGF-Supported Programs, 2000–03
(Averages, in percent of GDP)
| Early PRGFs (2000 and 2001 Only) | Recent PRGFs (2002 and 2003 Only) | |||||||
|---|---|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||||
| T−1 | T | T+1 | T+2 | T−1 | T | T+1 | T+2 | |
| Fiscal balance, including grants | –5.4 | 0.9 | 1.7 | 2.2 | –4.1 | 0.3 | 0.5 | 0.7 |
| Fiscal balance, excluding grants | –9.1 | 0.0 | 1.1 | 2.4 | –7.4 | –0.5 | –0.1 | 0.3 |
| Total revenues, excluding grants | 18.7 | 0.5 | 1.0 | 1.8 | 18.2 | 0.2 | 0.9 | 1.2 |
| Total expenditures and net lending | 27.8 | 0.5 | –0.3 | –0.8 | 25.6 | 0.6 | 1.0 | 0.9 |
| Grants | 3.7 | 0.8 | 0.7 | –0.1 | 3.3 | 0.8 | 0.6 | 0.4 |
| Net external financing1 | 6.6 | 0.9 | 0.8 | –0.6 | 6.1 | 0.5 | 0.7 | 0.1 |
| External current account balance (including official transfers) | –8.1 | –0.5 | –2.2 | –2.3 | –6.0 | –1.2 | –1.0 | –1.0 |
The sum of grants and net external borrowing in the government accounts.
Targeted Fiscal Adjustment in PRGF-Supported Programs, 2000–03
(Averages, in percent of GDP)
| Early PRGFs (2000 and 2001 Only) | Recent PRGFs (2002 and 2003 Only) | |||||||
|---|---|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||||
| T−1 | T | T+1 | T+2 | T−1 | T | T+1 | T+2 | |
| Fiscal balance, including grants | –5.4 | 0.9 | 1.7 | 2.2 | –4.1 | 0.3 | 0.5 | 0.7 |
| Fiscal balance, excluding grants | –9.1 | 0.0 | 1.1 | 2.4 | –7.4 | –0.5 | –0.1 | 0.3 |
| Total revenues, excluding grants | 18.7 | 0.5 | 1.0 | 1.8 | 18.2 | 0.2 | 0.9 | 1.2 |
| Total expenditures and net lending | 27.8 | 0.5 | –0.3 | –0.8 | 25.6 | 0.6 | 1.0 | 0.9 |
| Grants | 3.7 | 0.8 | 0.7 | –0.1 | 3.3 | 0.8 | 0.6 | 0.4 |
| Net external financing1 | 6.6 | 0.9 | 0.8 | –0.6 | 6.1 | 0.5 | 0.7 | 0.1 |
| External current account balance (including official transfers) | –8.1 | –0.5 | –2.2 | –2.3 | –6.0 | –1.2 | –1.0 | –1.0 |
The sum of grants and net external borrowing in the government accounts.
We used a regression framework to investigate potential determinants of targeted fiscal adjustment—for example, initial fiscal balance and targeted adjustment in the external current account—under ESAFs and PRGFs. The main results, which take account of the two-way relationship between targets for fiscal and external current account adjustment, are summarized below (more details are provided in Annex 7):12
The initial fiscal balance was found to be a key determinant of targeted fiscal adjustment under both ESAFs and PRGFs and the nature of the relationship has not changed much (Figure 4.2).
Under both ESAFs and PRGFs, projected increases in grants were translated almost fully into larger deficits (and vice versa).
The targeted change in the external current account balance exerted a statistically significant influence on fiscal adjustment under ESAFs but not under PRGFs. In other words, under the PRGF, private sector investment-savings balances bore a greater share of the burden (or gain) from any change in the external current account.
Projected growth was not found to exert a statistically significant influence on targeted fiscal adjustment in either ESAFs or PRGFs.
Relationship Between Initial Fiscal Balance at T–1 and the Targeted Fiscal Adjustment over a Two-Year Period (T–1 to T+1)
Sources: IMF staff reports and IEO analysis.Evidence on program outcomes
In sharp contrast to the findings regarding targeted fiscal adjustment, we found that on average, program outcomes under PRGFs recorded greater fiscal adjustment than those supported under ESAFs (Table 4.4). This is surprising, but limitations on outturn data—they are only available through 2002—meant that we relied on a truncated sample for the PRGF period. Thus, we may be picking up results that are valid for the early PRGFs which may change when outcomes data are available for those later programs where a more marked shift in fiscal targets occurred. Comparisons between outturns (see Table 4.4) and targets(see Tables 4.2 and 4.3) over the initial two-year-horizon of programs suggested the following main points:
Under the “early PRGFs,” the actual fiscal Initial balance as percent of GDP deficit declined on average by about 1½percent of GDP—more than was targeted. Expenditures fell slightly, in line with projections, while revenues were slightly higher than projected.
Actual outturns for net external financing were quite similar under the ESAF and the early PRGFs (i.e., a slight increase, on average), where-as programs under the ESAF had projected a decline and those under the PRGF had targeted a marked increase.
Actual Fiscal Adjustment Under ESAF- and PRGF-Supported Programs, 1995–2003
(Averages, in percent of GDP)
Excludes arrangements approved in 2002 and 2003, to allow coverage through horizon T+1. Outturn data are only available through 2002.
The sum of grants and net external borrowing in the government accounts.
Actual Fiscal Adjustment Under ESAF- and PRGF-Supported Programs, 1995–2003
(Averages, in percent of GDP)
| ESAFs (N = 47) | PRGFs (N = 24)1 | |||||
|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||
| T−1 | T | T+1 | T−1 | T | T+1 | |
| Fiscal balance, including grants | –3.5 | 0.3 | –0.3 | –5.1 | 1.2 | 1.9 |
| Fiscal balance, excluding grants | –7.6 | 0.5 | –0.1 | –8.2 | 1.2 | 1.4 |
| Total revenues, excluding grants | 17.3 | 0.2 | 0.1 | 17.6 | 0.6 | 1.2 |
| Total expenditures and net lending | 24.9 | –0.3 | 0.2 | 25.8 | –0.6 | –0.3 |
| Grants | 4.1 | –0.2 | –0.2 | 3.2 | 0.0 | 0.4 |
| Net external financing2 | 7.1 | 1.1 | 0.3 | 6.9 | 1.3 | 0.1 |
Excludes arrangements approved in 2002 and 2003, to allow coverage through horizon T+1. Outturn data are only available through 2002.
The sum of grants and net external borrowing in the government accounts.
Actual Fiscal Adjustment Under ESAF- and PRGF-Supported Programs, 1995–2003
(Averages, in percent of GDP)
| ESAFs (N = 47) | PRGFs (N = 24)1 | |||||
|---|---|---|---|---|---|---|
| Level at | Change from T− 1 to | Level at | Change from T− 1 to | |||
| T−1 | T | T+1 | T−1 | T | T+1 | |
| Fiscal balance, including grants | –3.5 | 0.3 | –0.3 | –5.1 | 1.2 | 1.9 |
| Fiscal balance, excluding grants | –7.6 | 0.5 | –0.1 | –8.2 | 1.2 | 1.4 |
| Total revenues, excluding grants | 17.3 | 0.2 | 0.1 | 17.6 | 0.6 | 1.2 |
| Total expenditures and net lending | 24.9 | –0.3 | 0.2 | 25.8 | –0.6 | –0.3 |
| Grants | 4.1 | –0.2 | –0.2 | 3.2 | 0.0 | 0.4 |
| Net external financing2 | 7.1 | 1.1 | 0.3 | 6.9 | 1.3 | 0.1 |
Excludes arrangements approved in 2002 and 2003, to allow coverage through horizon T+1. Outturn data are only available through 2002.
The sum of grants and net external borrowing in the government accounts.
To summarize, program design under the PRGF did incorporate greater fiscal flexibility in the sense of targeting smaller and more gradual fiscal adjustment than under the ESAF, building in a projected increase in net external financing, and with a shift in the composition of targeted fiscal adjustment toward higher revenues rather than expenditure reduction. Most of this change appears to have occurred in “later” PRGF-supported programs (i.e., from 2002 onward), with much less change in “early” programs. However, information on program outcomes is only available for the “early” group—which shows much less change from outcomes under the ESAF. The evidence also suggests that, on average, program design allowed for larger external financing flows than actually occurred. This suggests that PRGF program design did not suffer from a systematic “aid pessimism” bias.13
Strengthening Public Expenditure Management
Improvements in PEM, particularly measures that strengthen good governance by promoting transparency and accountability in the use of public resources, is another key feature of the PRGF. On average, PRGF-supported programs have included approximately one-third more measures (i.e., formal conditionality, triggers for HIPC assistance, and expressions of intent on the part of the authorities) to strengthen PEM systems than under ESAF-supported programs.14 This is borne out in our country case studies, almost all of which contain some formal conditionality in PRGF-supported programs intended to strengthen PEM. In this area, PRGF-supported programs are only one part of a broader array of initiatives through which the IMF supports improvements in PEM, and we also discuss these broader efforts, including diagnostic work and technical assistance.
Greater emphasis on PEM in low-income countries originated in implementation of the enhanced HIPC Initiative and the desire to ensure that the resources freed up from debt service would be used efficiently in support of spending in priority social sectors. In order to evaluate PEM capacity, the IMF and World Bank staff developed, in conjunction with country authorities, a system of 15 benchmarks intended to capture those aspects of PEM related to tracking poverty-reducing expenditures (Box 4.4). A key finding was the strong need to upgrade PEM systems in most countries. Of the 25 countries evaluated, 9 required “some upgrading” to be able to track poverty-reducing spending satisfactorily, and the remaining 15 required “substantial upgrading” (Figure 4.3).
Assessment Results on Adequacy of Public Expenditure Management, 20011
Source: IMF (2002a).1 Numbers in parentheses indicate total number of benchmarks (out of 15) met by country. Figure shows only those countries with full PRSPs as of December 2002.The IMF is providing substantial technical assistance (TA) to strengthen PEM (Box 4.5). Based on recent internal assessments, there is scope to improve the effectiveness and sustainability of much of this TA by improving coordination with other providers, tailoring the TA to the more immediate and direct PEM needs of member countries, placing greater emphasis on follow-up to TA, and focusing TA resources where ownership is strongest.
In addition, a review we undertook of the IMF’s activities to help countries subscribe to standards for statistics dissemination and fiscal transparency assessments suggests that countries with PRSPs and PRGF arrangements were much more likely to have undertaken such efforts than were other low-income countries.15
Pro-Poor and Pro-Growth Budgets
The definition of a “pro-poor and pro-growth budget” continues to be subject to much debate and is closely related to the broader debate about the link between macroeconomic policies and growth/poverty outcomes. Even if one had a solid understanding of the links between policies and growth/poverty outcomes in a particular country (which is clearly not the case), terms such as “pro-poor” and “pro-growth” mask many difficult tradeoffs that are frequently not acknowledged. The most obvious are (i) trade-offs over time (e.g., between immediate poverty reduction or greater poverty reduction in the future, if some “non-pro-poor” expenditures have high rates of return); (ii) tradeoffs across groups (e.g., when policies help some groups move out of poverty but may adversely affect other poor groups); and (iii) tradeoffs between social protection narrowly defined (i.e., preventing particular groups from falling into poverty because of policy measures or adverse shocks) versus the goal of moving as large a (net) share of the population out of poverty. How a particular country answers these questions depends on its own social choices, but they could yield markedly different fiscal (and other) policy choices, each of which could be legitimately characterized as pro-poor and pro-growth. These issues go beyond the scope of the current evaluation but they reinforce how complex and closely tied to specific social choices are judgments on the degree to which particular policy measures are pro-poor. Therefore, a country-driven, rather than a donor-driven, approach to these choices is especially important. We focus here on the subcomponents of the feature set out in Box 4.1.16
IMF–World Bank Benchmarking Exercise on Public Expenditure Management
Preliminary assessments of 25 HIPCs’ public expenditure management (PEM) systems and their ability to track poverty-reducing expenditures were undertaken by Bank and Fund staff in 2001.1 The hope was that—by identifying weaknesses in PEM—it would be possible to prioritize needs in upgrading the capacity to formulate, execute, and audit/report on the budget.
Fifteen indicators were eventually selected to represent the basic standards necessary for PEM systems to effectively track poverty-reducing expenditures. Based on each indicator, specific benchmarks were adopted and grouped into three categories—budget formulation, budget execution, and budget reporting. At the time of the 2001 assessment, countries had met, on average, between seven and eight benchmarks.
On the basis of these assessments, action plans to improve PEM were prepared. Progress in implementing these action plans was reviewed by Bank and Fund staff in March 2003.2 Based on the number of measures that had been fully implemented, and assuming no slippage in other areas, staff concluded that, on average, the countries had met one additional benchmark. However, across the full sample, only one-fifth of planned measures in the action plans had been fully implemented while a similar share had not been started. Also disappointing was the lack of progress in improving the coverage and reliability of the budget. At the same time, HIPCs had been particularly active in putting in place short-term bridging mechanisms mainly involving “tagging” poverty-reducing expenditures identified in the PRSPs to facilitate their immediate tracking while comprehensive PEM systems were put in place.
Bank and Fund staff is currently undertaking a more comprehensive review of country PEM capacity using the same benchmarks but the focus remains on the HIPC countries. According to staff, the significant human resource costs of the benchmarking exercise (estimated at an average of 0.5 staff years for each country) have prevented the extension of the exercise to other PRGF-eligible countries.
1 IMF (2002a). 2 IMF (2003a).“Pro-poor” spending in PRGF-supported programs
Efforts have been made since the introduction of the PRGF/PRSP to derive a broad measure of poverty-reducing expenditure (PRE) to include relevant expenditures on health, education, rural infrastructure, water, sanitation, social assistance, and other spending deemed to be “pro-poor.” The determination of what constitutes PRE is country specific (al-beit with common elements) and is contained in the country’s PRSP. As a result, PRE estimates reflect a range of definitions across countries, the degree of transparency for which varies considerably. In a number of countries (e.g., Guinea and Nicaragua), there has been a tendency to include virtually all social sector investment—including many programs that were not especially targeted to the poor. Some stakeholders interviewed during the evaluation suggested that incentives created by HIPC conditionality have contributed to an overly broad classification of pro-poor spending.17,18
With these caveats in mind, available data show that actual expenditure designated as poverty reducing has increased significantly between 1999 and 2002 and some modest further increases are targeted through 2005 (Table 4.5 and Figure 4.4). Actual PRE expenditure rose by an average of 2.5 percentage points of GDP, with a further increase averaging about 1 percent of GDP targeted through 2005. It is not possible to say how much of this improvement can be attributed to the PRSP—or the HIPC Initiative. Indeed, there is some evidence that the trend toward higher expenditures in such areas began in the second half of the 1990s.19
Changes in Poverty-Reducing Expenditures, 1999–2002, and Projected Changes, 2002–051
(Means in percent of GDP)
Sample comprises 19 countries, of which 13 are in Africa, 8 are countries with “early” PRSPs, and 11 with later (i.e., 2003) PRSPs.
Changes in Poverty-Reducing Expenditures, 1999–2002, and Projected Changes, 2002–051
(Means in percent of GDP)
| Change | Projected Change | ||
|---|---|---|---|
| 2002 Level | 1999–2002 | 2002–05 | |
| “Early” PRSP cases (i.e., 2002 or earlier) | 10.7 | 2.9 | 1.0 |
| “Late” PRSP cases (i.e., 2003) | 9.0 | 2.2 | 1.2 |
| African PRSP countries | 9.2 | 3.0 | 1.0 |
| Average | 9.7 | 2.5 | 1.1 |
Sample comprises 19 countries, of which 13 are in Africa, 8 are countries with “early” PRSPs, and 11 with later (i.e., 2003) PRSPs.
Changes in Poverty-Reducing Expenditures, 1999–2002, and Projected Changes, 2002–051
(Means in percent of GDP)
| Change | Projected Change | ||
|---|---|---|---|
| 2002 Level | 1999–2002 | 2002–05 | |
| “Early” PRSP cases (i.e., 2002 or earlier) | 10.7 | 2.9 | 1.0 |
| “Late” PRSP cases (i.e., 2003) | 9.0 | 2.2 | 1.2 |
| African PRSP countries | 9.2 | 3.0 | 1.0 |
| Average | 9.7 | 2.5 | 1.1 |
Sample comprises 19 countries, of which 13 are in Africa, 8 are countries with “early” PRSPs, and 11 with later (i.e., 2003) PRSPs.
Efficiency and equity in tax reform
According to the IMF staff’s assessment of PRGF program design, approximately two-thirds of tax measures under these programs dealt with improvements to tax administration.20 Our case studies also show that conditionality (and IMF TA) has emphasized improved tax administration, an emphasis that seems appropriate given weaknesses in tax administration in many low-income countries. These weaknesses limit the authorities’ ability to generate the revenue necessary to finance PRE and to use the tax system as a mechanism for promoting equity.
Changes in tax policy explicitly motivated by “equity” considerations have been less in evidence in PRGF-supported programs. Most common have been reforms aimed at improving “horizontal” equity (e.g., to ensure equal treatment between domestic and foreign enterprises operating in a country or to remove distorting tax exemptions put in place to benefit politically well-connected individuals or enterprises). “Vertical” equity considerations are even rarer, a fact acknowledged by staff in their 2002 assessment. This was attributed to “the limited scope to implement progressive income taxes in low-income countries given administrative constraints and the high share of agriculture and the informal sector in economic activity.”21 Perhaps as a result, most tax policy measures have focused on consumption and trade taxes. That said, PRGF conditionality has rarely sought removal of a tax exemption explicitly on the basis of its regressive nature.
There does not appear to have been a systematic effort to assess the distributional impact of proposed tax changes in individual countries, although there are some “good practice” examples. A detailed PSIA on the VAT in Ethiopia undertaken recently concluded that replacement of the sales tax with a VAT would have had a small adverse impact on the poorest 40 percent of the population, which would be more than offset by the beneficial effects on the poorest groups if the reform financed higher health and education spending.22 In Mozambique, a decision to increase fuel taxes was delayed until a poverty and social impact analysis was undertaken, the results of which were taken into consideration in the government’s decision to increase petroleum prices in 2003. In Vietnam, Fund staff indicated, in early 2003, their intention to undertake PSIA on the effects of possible tax changes (although no work had begun as of early 2004).
Our case studies also revealed opportunities to promote greater vertical equity that were not actively pursued in PRGF-supported programs. In Guinea, for example, widespread tax exemptions were contributing to a very low revenue yield (and a consequent shortage of resources for PRE). However, while the PRSP expressed the authorities’ intention to “sharply reduce exemptions,” the PRGF-supported program stopped short of using conditionality to promote the removal of costly tax exemptions that were not intended to assist the poor.23
IMF Technical Assistance Support for Strengthening Public Expenditure Management
Since FY1999, the number of person days devoted each year to PEM technical assistance (TA) in our sample of 23 countries has increased by two-thirds, almost 90 percent of which went to the HIPCs. However, concerns have been expressed about the efficacy of that TA. IMF staff (in the context of recent evaluations of technical assistance on PEM reform to Anglophone and to Francophone African countries) 1 identified a number of shortcomings, including with respect to TA design and coordination among providers:
No low-income African country (except Tanzania and Uganda) has a financial accounting system adequate to provide information necessary for policymaking.
While the quality and relevance of TA provided by the Fund (and others) was recognized by stake-holders, weak ownership of PEM measures was a serious problem for implementation of TA recommendations. This is a serious concern, since the central thrust of the PRS approach is to strengthen ownership.
There is a need for better coordination of PEM TA, particularly with bilateral providers. Coordination between the Bank and the Fund was somewhat better with IMF TA often being used to fill a “temporary gap,” while the Bank worked on larger, more complicated and longer-term reforms.
In Francophone Africa, PEM measures proposed by IMF TA may have been too general with insufficient attention paid to implementation constraints and the need to pitch recommendations at an accessible level.
One of the major sources of weakness in PEM systems in Anglophone Africa was the tendency of governments (with the encouragement of the donor community) to burden themselves with functions and tasks beyond their capacity to execute.
There had been little follow-up to Fund TA for PEM, particularly in Anglophone Africa. This was attributed in part to insufficient TA resources within the Fund.
Poverty and Social Impact Analysis
Our review confirmed the findings of other studies—that efforts to conduct PSIA have been slow to start and the integration of these results into program design even slower. Staff papers for 5 of the 23 countries reviewed (all outside Africa) made virtually no reference to PSIA. Most others simply made reference to the need to undertake PSIA for at least one major reform area. Only a few present more than cursory results (Guyana, Mozambique, Vietnam, and Zambia) and even here, much of the analysis was very general.
The six IEO case studies revealed a considerable range of experience, with PSIA of varying quality undertaken at varying stages in the policy design process.
In Guinea, a joint IMF–World Bank mission in 2001 discussed with representatives of trade unions, NGOs, the business community, and parliament the likely social impact of the antiinflation stance of monetary policy, the introduction of flexible petroleum prices, and restructuring and privatization of state-owned enterprises (SOEs) and in the financial sector. However, owing partly to the lack of data, there was no supporting analysis and it did not have a discernible impact on program design.
In Mozambique, PSIA was undertaken in 2002 (partly funded by DFID) on the impact of raising taxes on petroleum products. It found that the impact on poverty would be modest and, if the proceeds went toward PRE, more people would be lifted out of poverty than would be hurt by the tax. Other studies undertaken included a Food and Agriculture Organization study in 2000 on protection of the sugar industry and a World Bank–sponsored study on re-structuring the cashew industry. While Bank and Fund staff reportedly did not agree with the results of the study or the FAO’s call to protect the sugar sector, neither institution sought to block the government from following the FAO recommendations. PSIA on the cashew industry formed the basis of a government policy to provide a transfer payment to various companies to pay for accumulated liabilities to the labor force. The authorities have agreed with the World Bank on future PSIA needs but not yet on a timetable for undertaking the work.
In Nicaragua, while the PRGF-supported program was approved in December 2002, preparation of PSIA to inform policy design was delayed, partly because of the time taken by the authorities to identify policies they considered to be of priority for such analysis. Recent developments, including the completion of PSIA on the tax reform and the Education for All Initiative and the undertaking of PSIA on the Public Investment Program and Decentralization indicate that some of the initial obstacles have been overcome.
For Tajikistan, no PSIA was undertaken on key reforms in the PRGF-supported program despite the recognition by IMF staff that measures aimed at eliminating quasi-fiscal deficits in the energy sector would have severe adverse effects on the well-being of a large part of the population. The World Bank was to have undertaken PSIA on energy sector reforms in 2002 and 2003 but work was delayed ostensibly due to a lack of financial resources and other demands on staff.
As of end-2003, little PSIA had been undertaken for Tanzania despite the identification by Fund staff of a need related to increased electricity tariffs, retrenchment in the parastatal sector, civil service reform, and the VAT. The authorities have expressed doubt about their own ability to undertake PSIA, partly due to a lack of analytical capacity and partly due to financial constraints. Until early 2004, a lack of budgeted resources and/or available staff was reported as preventing the Bank and Fund from undertaking PSIA. In practice, therefore, PSIA has not been a priority.
PSIA on the impact of SOE reform on displaced workers was undertaken in 2000 for Vietnam with support from the World Bank and prior to agreement on a PRGF-supported program. The costs of mitigating measures were integrated into the fiscal framework in the PRGF-sup-ported program. In the spring of 2003, Bank and Fund staff reached agreement on an agenda for PSIA priorities, including on possible future tax reforms. However, clear deadlines for completion of this work have not yet been set.
The overall message that emerges from this brief review is that PSIA is only gradually becoming an input to program design. These efforts are not yet systematic in the sense that PSIA needs are being derived from country priorities within a results-based framework indicating who should address them and by when. Furthermore, many key reforms are still falling through the gaps.
Obstacles to “mainstreaming” PSIA in PRGF-supported programs
The slow pace of implementation of PSIA is note-worthy since calls for IMF staff to integrate such analysis into program design started well before the launch of the PRSP/PRGF initiative. As early as 1991, the IMF’s Managing Director had instructed staff to address the concerns of the poor “as a matter of course...[which] should be an integral part of designing Fund-supported and Fund-monitored programs.”24 Similar guidance was given in 1998 by the Executive Board in the context of its response to the External Evaluation of the ESAF. The fact that progress has been so slow suggests the existence of serious obstacles.
The obstacles most often cited in internal reviews were data limitations and national capacity constraints.25 Both are undoubtedly important, but they should not be overstated. A paper prepared in the IMF’s African Department concluded that “it is possible to assess some of the potential poverty effects even in countries with limited data and therefore contribute to a more informed policy debate and design.”26 This echoed a similar conclusion reached in 2001 in the context of joint work undertaken by the World Bank and IMF for 12 PRSP countries.27 In August 2003, internal staff guidelines were circulated identifying types of policies that were (or were not) conducive to PSIA by the IMF. Among those policies considered conducive to PSIA and part of the IMF’s area of core competence were tax polices (e.g., introduction of VATs and changes in tax rates), customs tariff policies, and exchange rate policy (including devaluation).
Another important set of explanations revolves around the fact that, although PSIA was recognized as a central element of the new approach, the arrangements for setting and delivering on priorities for such analysis within a broader partnership frame-work were not well specified. The result was (i) a lack of resources allocated to this purpose in the BWIs, and (ii) coordination problems with the World Bank. Our survey of IMF staff indicated that a lack of funding and/or staff resources at the BWIs was a major reason for lack of progress in integrating PSIA into program design. Guidance given to IMF staff makes clear they are expected to draw on the work of the World Bank and other development partners.28
However, our case studies illustrate that this is often easier said than done. The implicit assumption that the World Bank will act as the “agent” for the IMF in implementing or coordinating PSIA in areas of importance to the Fund can pose problems. If the Bank is not actively engaged in lending associated with a particular area, it has little incentive to devote its own scarce resources to analyzing the issue. The Bank’s timetable may also not fit with that of IMF program design. The solution to these resource and coordination problems seems to lie in a clearer identification of priorities for PSIA—driven by the countries’ own priorities for such analysis arising from the PRS process—along with an explicit identification of responsibilities for who does what and by when (see Box 4.6 for some useful lessons on timing).29
Is Structural Conditionality More Selective?
As part of the streamlining initiative, PRGF-supported programs are expected to focus structural conditionality on the IMF’s core areas of expertise.30 The only exception would be where a structural measure outside these core areas was deemed critical to the success of the program. This initiative has been complemented by ongoing efforts to improve coordination with the World Bank, including designation of a “lead agency” responsible for the design and monitoring of conditionality in a particular policy area. We examine here what has happened in practice.
Structural conditionality in PRGF-supported programs
There is clear evidence of progress in streamlining structural conditionality under the PRGF (in terms of reducing the number of conditions, although the number itself does not necessarily translate into a good measure of the burden of conditionality). There have been statistically significant reductions in the number of performance criteria and structural benchmarks but no change, on average, in the number of prior actions (Table 4.6).
Number of Structural Conditions1,2
Based on a sample of annual ESAF arrangements approved during 1996—99, and new PRGF arrangements approved during 2000—02.
** denote statistically significant difference in means between ESAFs and PRGFs.
Number of Structural Conditions1,2
| ESAF | PRGF | |
|---|---|---|
| (N = 33) | (N = 32) | |
| Prior actions | 4.8 | 4.5 |
| Performance criteria | 4.0** | 2.7** |
| Structural benchmarks | 8.4** | 4.9** |
Based on a sample of annual ESAF arrangements approved during 1996—99, and new PRGF arrangements approved during 2000—02.
** denote statistically significant difference in means between ESAFs and PRGFs.
Number of Structural Conditions1,2
| ESAF | PRGF | |
|---|---|---|
| (N = 33) | (N = 32) | |
| Prior actions | 4.8 | 4.5 |
| Performance criteria | 4.0** | 2.7** |
| Structural benchmarks | 8.4** | 4.9** |
Based on a sample of annual ESAF arrangements approved during 1996—99, and new PRGF arrangements approved during 2000—02.
** denote statistically significant difference in means between ESAFs and PRGFs.
PSIA in “Real Policy Time”
Among bilateral donors, DFID has been particularly active in supporting the generation of PSIA in PRGF-supported countries. In 2002, and in collaboration with the Word Bank and IMF, they undertook PSIA pilot studies for Rwanda, Mozambique, Indonesia, Armenia, Uganda, and Honduras in an effort to demonstrate the circumstances in which PSIA could be useful to policymakers.1
An interesting element of DFID’s terms of reference was the placement of time limits on the conduct of PSIA to ensure that the analysis was undertaken in “real policy time.”2 While DFID reported some concern on the part of IFIs with respect to the quality of PSIA undertaken over a relatively short period of time, the notion of “time bound” PSIA—if it can result in quality work—may respond to concerns that PSIA is not available earlier enough to be an input to program design.
1 See DFID (2003). 2 The terms of reference also required the use of both international and local consultants combining both economic and social expertise in the conduct of PSIA. This is consistent with Robb (2003), who emphasized that the most productive avenue to pursue PSIA was through the combination of skills embodied in various disciplines within the social sciences.These patterns varied considerably among countries, as is evident in our case studies. Three of the six IEO case studies (Guinea, Nicaragua, and Tajikistan) experienced declines in the number of formal structural conditions. On the other hand, Mozambique registered an increase in the number of structural conditions, particularly in the financial sector. Vietnam and Tanzania initially saw no marked change in the overall quantity of structural conditionality. We were unable to detect any clear linkage between this experience and previous “track records” on implementation.31
What has happened to aggregate IMF–World Bank conditionality?
A number of critics have argued that while the IMF has withdrawn from imposing structural conditionality in some policy areas, the World Bank has replaced it, resulting in no net reduction in the burden of conditionality, which they see as inconsistent with the emphasis on stronger country ownership. Others have argued that, since the fundamental problems impeding faster growth and poverty reduction in low-income countries are primarily structural in nature, it is reasonable to ask countries to establish benchmarks for monitoring their progress in the context of BWI lending, provided the benchmarks are derived from a country-driven strategy. We sought to investigate what has happened in practice as part of our evaluation. Several key messages emerge.
First, the BWIs do not have an explicit objective with respect to the burden of aggregate Bank-Fund conditionality. In July 2001, the Boards of both institutions endorsed a strategy to strengthen Bank-Fund collaboration on conditionality and called for the formulation of “mutually supportive and harmonized conditionality (but not cross conditionality) in PRGFs and Poverty Reduction Support Credits (PRSCs) (and IDA lending instruments)” and committed “to streamline and focus conditionality on the key actions crucial to success.”32 This strategy resulted in the production of a joint Bank-Fund “Staff Guidance Note” that described the motivation for greater collaboration on conditionality as the avoidance of duplication.33 However, reduction of the burden of aggregate conditionality was not explicitly identified as an objective.
Second, we found it enormously difficult to track what has happened to aggregate (Bank-Fund) conditionality in specific countries. The difficulties encountered suggest that systematic monitoring of developments in this area has not been a priority for either institution. Key problems include:
There is no simple “unit of account” with which to measure conditionality across the two institutions.34 Moreover, “counting conditions” is a very crude measure of the burden of conditionality because of the potentially large differences in the political and technical challenges involved in implementing any one condition.
While the IMF has a single instrument for supporting low-income countries (the PRGF), the World Bank has several, including PRSCs as well as sectoral adjustment and other program lending instruments, and it is difficult to track conditionality across this entire lending program.
IMF and World Bank databases to track conditionality are not compatible and—even within each institution—are not always internally consistent.35
For a number of the case study countries, no Bank adjustment credits had been approved post-PRSP, making a comparison of aggregate conditionality pre- and post-PRSP not meaningful.
Reflecting these constraints, we were able to obtain information on aggregate Fund-Bank conditionality for only four countries (Albania, Mozambique, Tanzania, and Vietnam) within the 10 IEO/OED case studies.36 (A more detailed analysis is contained in each of the respective case study reports). The evidence from this admittedly very small sample is mixed but suggests a number of tentative messages:
There were decreases in the number of total Bank-Fund conditions after the PRSP in all four countries. This ranged from an almost 50 percent reduction in Albania to a 6 percent decline for Tanzania. The average across the four countries was a reduction of about one-third.37 However, almost the entire decline was due to a reduction in Bank conditionality (the number of IMF conditions decreased only in Albania). This is surprising in light of the more explicit emphasis in the Fund on reducing the number of conditions and the results from the broader sample of PRGF-supported programs, which indicates such a reduction has occurred. It suggests that the countries for which we were able to undertake a comprehensive analysis are not typical, at least as far as the number of IMF conditions are concerned.
There were significant changes in the division of labor:
With respect to fiscal conditionality (both revenue and expenditure), the IMF increased its focus, while the Bank reduced its emphasis (except for a modest increase in emphasis on treasury systems).
The Bank withdrew from setting conditionality on the foreign exchange system and on central bank reform.
The Fund withdrew from setting conditionality in the agricultural sector and reduced its already weak emphasis on civil service reform.
Bank conditionality in the social sector as a share of total Bank conditionality increased markedly.
Financial sector reform increased as a share of both Bank and Fund conditionality.
Both the Fund and the Bank reduced the shares of their conditionality on SOE and trade reforms but increased the share targeted to institution building and legal and regulatory reforms.
These observations are consistent with an increased emphasis on core areas of expertise by the two institutions.
The case studies also suggest that streamlining can involve unexpected pitfalls—linked to “principal-agent” difficulties—resulting in some structural issues not being handled well by the joint efforts of the BWIs. Specifically, it cannot be assumed that IMF and World Bank objectives and priorities in areas of overlapping interest are identical. The establishment of conditionality on SOE reform in Vietnam provides a good illustration. Consistent with an agreed division of labor, the Bank took the lead in designing and monitoring conditionality on SOE reform. The IMF’s primary interest in the SOE sector was on the potential consequences for the budget and the soundness of the financial sector of weakness in the large SOEs. The Bank’s primary interest, on the other hand, was private sector development and the demonstration effect of privatizing even small SOEs, a position that was consistent with the government’s own priorities. As a result, SOE conditionality established by the Bank did not target the most fiscally significant SOEs.
Finally, surveys of IMF and World Bank staff conducted in late 2003 as part of an internal review of progress in strengthening Bank-Fund collaboration suggest some promising signs regarding the impact of the PRSP framework on collaboration. In countries that have embraced the PRS frame-work:38 (i) IMF staff interacts to a greater extent with World Bank staff in the formulation of conditionality, and that involvement is perceived as more effective than in other countries. (ii) There is greater duplication of conditionality in PRGF-eligible countries than in others, but two-thirds of mission chiefs attribute this duplication to the fact that the measure is critical to both institutions’ programs. (iii) The view that the IMF and the World Bank have developed a shared perspective on the necessary reforms is more prevalent among mission chiefs for PRGF-eligible countries than those for other countries.
Some Program Design Issues
One aspect of program design that is much criticized is the lack of sufficient attention paid to macro-micro linkages and sources of growth. The critics have argued that the IMF’s traditional approach to program design (or “financial programming” 39) does not take sufficient account of the underlying determinants of growth and of the factors influencing the economy to macroeconomic policies—as well as of the potential feedback of macroeconomic policy actions on poverty—and that this can lead to mistakes in macroeconomic policy design.
The goal of embedding PRGF-supported programs in the overall strategy for growth and poverty reduction aimed to address this criticism by seeking to base programs on fully integrated macroeconomic, structural, and social policies. As Adam and Bevan (2001) note, “This is a pretty tall order. This type of integrated framework is one which the economics profession, at least, has conspicuously failed to deliver over the last thirty years, despite much trumpeting of micro-macro frameworks. The short answer is that our understanding of these relationships, while not wholly lacking, is very far from being ‘integrated’.” Nevertheless, more analysis of these linkages—and an explicit setting out of the underlying rationale for a particular program design—can foster a more constructive debate and modifications in the light of new evidence. Several relevant messages emerge from our evaluation and from recent analytical work undertaken in the IMF.
Discussions with IMF staff and analysis of briefing papers suggest that efforts in the World Bank and elsewhere to spell out the macro-micro linkages to growth and poverty have had limited impact so far on program design by the IMF.40 The concerns typically expressed by Fund staff are that such approaches require too many assumptions about underlying economic and structural relationships, especially in low-income countries where the data often do not exist. While this is certainly true, it should not prevent additional analysis that can help throw light on some of the underlying linkages. The IMF does not have to produce the analytical inputs itself, but can help identify the major knowledge gaps in each country. If done in a collaborative manner with local researchers, such efforts can stimulate domestic capacity to analyze such matters.
Country programs are typically based on a number of assumptions about behavioral relationships (e.g., with regard to the response of investment and savings, or absorptive capacity with respect to aid inflows) and these often rely largely on judgment. The case studies show that while qualitative arguments were often made in support of the need for a particular design, staff reports or other program documents rarely spelled out systematically the full rationale, and implicit behavioral assumptions, underlying program design.41 Since the consistency of program design depends critically upon these assumptions, a fuller statement would help generate a more informed debate and also more effective responses at the stage of program review. The Tanzania case discussed in Box 4.2 is one example. Another is Vietnam, where in the context of discussions on the pace of SOE reform, IMF staff initially sought more aggressive reform of SOEs, believing that future growth in the private sector would be sufficient to absorb much of the resulting displaced labor. The authorities took a more gradualist approach, partly based on a more conservative view of the pace of private sector growth and their associated desire to maintain social stability by containing the rise in unemployment. The disagreement with staff was not so much on the need to reform SOEs but the speed at which this could happen given the authorities’ expectation that the private sector would not be in a position to readily absorb displaced labor. In this regard, assumptions about growth in the private sector and its ability to absorb surplus public sector labor were key. That said, there appears to have been little analysis under-taken by either the authorities or staff (either in the IMF or World Bank) to investigate their implicit assumptions.
Do PRGF-Supported Programs Have an Excessive Disinflation Bias?
A number of critics have argued that many PRGF-supported programs are excessively focused on disin-flation, with potentially adverse effects on growth and poverty reduction.1 Different cross-country analyses come to different conclusions as to what the “thresh-old” level of inflation might be above which it entails a trade-off with growth, but there appears to be a broad consensus that “double digit” inflation is likely to be harmful. Two recent studies suggest ranges of 2–3 percent per annum (Ghosh and Phillips, 1998) and 7–11 percent for developing countries (Khan and Senhadji, 2000). On the specific issue of disinflation, Ghosh and Phillips found that “starting from inflation rates above 6 percent, only the most drastic disinflations (at least halving the inflation rate in a single year) are associated with any negative impact on growth….” However, they also found that starting from rates below 6 percent, rapid disinflation (a halving of the rate in a year) was associated with a fall in real GDP growth.
Our evaluation indicates that PRGF-supported programs projected a smaller average reduction in inflation levels than ESAF-supported programs, but this largely reflects much lower initial inflation rates. Under ESAFs, inflation was targeted to fall from 22 percent on average in the year immediately preceding the program to about 10 percent and 5½percent in the first and second program years, respectively. By contrast, under PRGFs, the corresponding path was from 9 percent to about 6 percent and then to 4 percent.
Looking at disaggregated data, we found a strong tendency in program design against tolerating double-digit inflation, but detected no systematic disinflation tendency when inflation is already low. Out of 41 PRGF-supported programs in our sample, more than half had initial annual inflation rates of 5 percent or lower, and about one-quarter had initial inflation rates of 10 percent or higher. In all the latter cases, the program targeted a decline in inflation. In contrast, in the majority of cases where initial inflation was 5 percent or less, inflation was projected to go up. In those cases where initial inflation was between 5 percent and 10 percent, about two-thirds targeted lower inflation. This pattern is largely unchanged from that of ESAF-supported programs (except that the latter had fewer cases with initial inflation below 5 percent).
These results do not suggest an excessive “deflationary bias” with regard to inflation targets.
1 See, for example, Oxfam (2004).At a more general level, the assumption of rapid “crowding in” of private sector aggregate demand (as external resource requirements are relaxed or fiscal deficits are reduced) is a common feature in PRGF-supported program design. This has meant that the shift away from targeting reductions in the external current account deficit has been accompanied by much less change in targeted fiscal deficits. Implicitly, the rationale of the program design appears to be that the more relaxed external stance accommodates stronger private sector demand. While this would be consistent with a strategy where the private sector will be the primary engine of growth, achieving this outcome depends upon the speed of response of the private sector. Assumptions on this aspect need to be verified and, if necessary, modified in light of experience.
A related issue is the extent to which PRGF-supported programs target inflation reduction and whether, as some critics have argued, program design is unduly restrictive by targeting excessive reductions in inflation—going beyond the evidence on what inflation thresholds are harmful to growth and poverty reduction. Our analysis of PRGF-supported programs suggests that programs do systematically target lower inflation when initial inflation is 10 percent or higher, but that there is much greater variation when initial inflation is 5 percent or less; in a majority of such cases, inflation was projected to increase (Box 4.7). This is generally consistent with the broader cross-country evidence on when inflation becomes harmful to growth and does not suggest an excessive “disinflation” bias.
Has Program Implementation Improved?
An important premise in the transformation of the ESAF to the PRGF was that greater country owner-ship of programs would improve prospects for implementation. We examined a range of indicators—disbursement ratios, the incidence of program interruptions, and compliance with conditionality (performance criteria)—to see whether program implementation had improved under PRGFs. The main results from cross-country evidence are:42
Among arrangements that have run their course (i.e., expired), the mean disbursement rate was about 75 percent, with no statistically significant difference between ESAFs and PRGFs.43
There is hardly any difference between ESAFs and PRGFs in the relative frequency of program interruptions, although, on average, PRGFs exhibit a slightly shorter time between key program events.44
The fraction of quantitative (i.e., macroeconomic/financial) performance criteria that were met increased slightly under PRGFs (to about 85 percent, compared with less than 80 percent under ESAFs). There was no statistically significant difference in the compliance rate with respect to structural performance criteria.
Overall, the indicators suggest a rather modest improvement in implementation under PRGFs. This finding is consistent with the evidence of mixed performance from the four IEO case studies—Guinea, Mozambique, Tanzania, and Vietnam—which had completed both ESAF and PRGF arrangements as of May 2004.45 For example, of the four countries, Tanzania had no program interruption in either regime, while Guinea had major interruptions under both regimes. Mozambique had no formal program interruption under either regime but there were more delays in completing reviews under the PRGF than there was under the ESAF. For the Vietnam ESAF (1994–97), no agreement could be reached on a third annual arrangement. Performance under the PRGF (2001–04) was good but the arrangement was interrupted on account of issues that were not directly related to performance under the program.
Even when a longer time series is available, the methodological challenges associated with identifying the specific impact of programs on poverty, etc. will be considerable. See Easterly (2000) and Hajro and Joyce (2004) for different approaches to this issue for earlier adjustment lending programs.
IEO/OED case studies where a “new” PRGF preceded a full PRSP are Ethiopia, Guinea, Tanzania, and Vietnam.
The problem is compounded by considerable uncertainty over the size and timing of aid flows. See, for example, Bulíř and Hamman (2001) who conclude that (i) aid is more volatile than fiscal revenues—particularly in highly aid dependent countries; (ii) uncertainty about aid disbursements is large, being larger for program than project aid; and (iii) the information content of commitments made by donors is small—that is, official donor projections of aid are subject to large errors and exhibit a substantial upward bias.
Alignment was assessed based on the first PRGF-related staff report that followed the completion of the PRSP. This exercise is subject to two important qualifications. First, some structural measures may be too detailed so as to be concretely aligned with the PRSP action plan (e.g., restructuring of one specific bank). Second, proposed conditionality may aim to deal with a problem (e.g., banking crisis) that occurred after completion of the PRSP.
Based on adjustment in the fiscal balance excluding grants. Although grants are usually defined as part of revenues rather than as a component of financin g of the fiscal deficit, it is sometimes useful to consider grants as another source of deficit financing. We have combined net external borrowing and grants into a single variable that can be used to test the “aid pessimism” hypothesis.
The difference is statistically significant over a two-year horizon.
The difference is statistically significant at a 95 percent confidence level over two- and three-year horizons.
We employed two-stage least squares to estimate targeted fiscal adjustment, with the first stage involving an estimation of the targeted change in the external current account. The regressions were estimated over the first two years of programs (rather than over a three-year horizon) in order to facilitate comparisons with outturns in the section “Evidence on program outcomes.”
Clearly, such an analysis cannot address the broader issue raised by some external critics, of whether a more proactive “catalytic” role by the IMF could have mobilized additional financing.
See Gupta and others (2002) and IMF (2002a).
Specifically, the General Data Dissemination Standard (GDDS) and Fiscal Transparency Reports on Standards and Codes (ROSCs).
Aspects concerning improvements in the efficiency and targeting of expenditures within sectors are primarily the responsibility of the World Bank and will not be addressed here.
The whole notion of identifying “priority” sectors that will benefit when additional resources become available or will be protected from cuts when shocks occur is a “second best” response to inadequate budgetary expenditure allocation mechanisms. It implies that an extra dollar spent in these areas yields a greater benefit than a dollar spent elsewhere, which raises the question as to why expenditures have not already been reallocated to eliminate the difference. Given political economy constraints on the budget process that can hamper desirable reallocations, such an approach may well be the most practical response. But, as Bevan (2001) and Bird (2004) have noted, the approach raises some uncomfortable questions, especially if maintained for long periods: does it reflect a pragmatic response to constraints on the speed with which governments can make desirable changes identified as part of domestic political processes or does it reflect an imposition of donor priorities that conflict with the idea of stronger domestic ownership?
No IMF data base on budgeted PRE or its components has been maintained. For the 23 countries in our study, only staff reports for 7 countries (Albania, Bolivia, The Gambia, Guyana, the Kyrgyz Republic, Malawi, and Mozambique) contained separate and consistent projections of health and education as a share of GDP (3 others present figures for combined health and education spending). For this small sample average budgeted spending on health and education as a share of GDP is higher in PRGF-supported programs than in their ESAF counterparts by 0.6 and 0.8 percentage points, respectively.
See, for example, OED (2003).
See Gupta and others (2002, p. 18).
See Munoz and Cho (2003).
Indeed, the initiative to streamline conditionality would call for such an approach, if the exemptions were not judged to be “macro-critical.”
See IMF (2001b).
IMF (2002c). Also, an internal checklist for staff with respect to the key features of the PRGF calls on Fund staff to “demonstrate that distributional effects of substantial macro-adjustments or structural reforms have been considered (generally based on analysis by World Bank)” (italics added).
Adam and Bevan (2001) argued that “the current separation [of responsibilities in the preparation of PSIA] seems excessive. There is, after all, a substantial tradition of tax and expenditure incidence whose conceptual roots lie in the type of fiscal arena within which the Fund has substantial expertise.” A small unit was recently established in the Fiscal Affairs Department (FAD) with responsibility to coordinate (and, in some areas, produce) PSIA. However, there are divergent views on this issue within the IMF’s Board, with some arguing that budget constraints and considerations of comparative advantage argue for such activities to be concentrated in the World Bank, with greater efforts to ensure effective collaboration.
Efforts to streamline IMF structural conditionality have been under way since the fall of 2000 and cover all programs, not just those under the PRGF. The final results of these efforts were revised guidelines on conditionality. They are available at www.imf.org/External/np/pdr/cond/2002/eng/guid/092302.pdf.
The explanation for the decline in Guinea, despite its poor track record, is that a large number of “informal” conditions were introduced in the context of several short-term consolidation exercises designed to put the PRGF-supported program back on track. These consolidation exercises, and the associated informal conditionality, were not linked in any significant way with the PRS.
There are also differences between the nature of Bank conditionality under the PRSC and IMF conditionality under the PRGF that can blur the meaningfulness of simply adding the number of Bank and Fund conditions together.
The IMF’s database is MONA (Monitoring of Arrangements) and the Bank’s ALCID (Adjustment Lending Conditionality Implementation Database). The IMF recently altered the classification system of its MONA database to improve its ability to monitor structural conditionality but the changes were not made retroactively.
We looked at both PRSCs and other Bank adjustment loans. Since the evolution of Bank policy on conditionality was not concurrent with the adoption of the PRGF, we chose the date at which the full PRSP was adopted for the purposes of assessing changes in aggregate conditionality.
These results are consistent with the responses the Bank and Fund received to a recent survey of country authorities, which indicated that two-thirds of respondents had reported a reduction in the number of combined program conditions. While these results were for both low- and middle-income countries, they do shed some light on the evolution of aggregate conditionality for PRSP/PRGF countries.
All the cited findings reflect statistically significant differences between the answers provided by mission chiefs for PRGF-eligible countries and mission chiefs to other program countries. In cases where the same question had been asked for the 2002 review of Bank-Fund collaboration, the significance of the difference increased over time.
See Baqir and others (2003) or Khan and Knight (1985) for a description of the financial programming framework.
Gunter (2002) provides a summary of some of this work as well as similar work under way elsewhere using computable general equilibrium (CGE) and other models.
The recent IEO evaluation of Fiscal Adjustment in IMF-Supported Programs (IEO, 2003) came to a similar conclusion.
Based on arrangements approved during 1995–2003.
The subsample of expired arrangements was divided into those that had been either ESAFs or PRGFs for most or all of their duration.
Typically, indicators of “program interruption” measure delays between scheduled and actual dates for completing program reviews (and approval of annual arrangements under the ESAF). For this report, we define interruption in terms of the actual time between key events rather than through a comparison of scheduled and actual dates. Key events are approval of multiyear or annual arrangements, completion of program reviews, and expiration dates (when they occurred more than 12 months after either the completion of the last review or approval of the arrangement). The averages are about 9½ and 9 months for ESAFs and PRGFs, respectively.
Guinea (ESAF 1997–2001, PRGF 2001–04); Mozambique (ESAF 1996–99, ESAF/PRGF 1999–2003); Tanzania (ESAF 1996–2000, PRGF 2000–03); and Vietnam (ESAF 1994–97, PRGF 2001–04). For Mozambique, we count the ESAF/PRGF as a PRGF.
