The impact of IMF-supported programs on the level of public spending in the social sectors has received a great deal of attention, with many critics voicing concern that these programs typically involve an unnecessary squeeze on social spending, with adverse effects on social welfare. We examine this issue in several ways. First, we analyze a set of concerns raised in the context of low-income countries—whether programs incorporate public spending levels and fiscal deficit targets based on overly conservative projections of concessional financing. Second, we examine cross-country data to assess what may have been the impact of IMF-supported programs on the level of public sector social spending. Third, we analyze program documents in the sample of 15 programs described earlier, to assess how program design has incorporated social spending and social concerns.
Has Donor Aid Been Underestimated?
Concerns have been raised that IMF-supported programs in low-income countries that depend on concessional financing may incorporate fiscal targets based on aid projections that “taper out” too quickly relative to what donors may be willing to provide. If true, such a tendency could also create a disincentive for donors to sustain their level of aid, even when programs remain on track.1.
Some recent studies by IMF staff have argued in support of a cautious approach to projecting aid flows, mainly on the grounds that disbursements tend to be significantly less than commitments, and that even the so-called conservative projections in IMF-supported programs tend to overestimate actual aid flows.2 These studies also point out that in the programs examined: (1) disbursements exceeded projected amounts in a minority of cases; (2) shortfalls relative to projections were more marked for program aid (compared to project aid); and (3) within program aid, grants (provided mainly by bilateral donors) had a smaller “prediction error” than concessional loans (a large part of which came from the World Bank and regional development banks).
One factor that may contribute to deviations between projections and outturns is compliance with conditionality. To the extent that the conditions attached to the disbursement schedule are not met, donors may withhold disbursements. For example, some donors link disbursements of their program aid to recipient countries’ performance under IMF-supported programs. Thus, outturns in such cases are to some extent contingent on implementing policies in the program, and hence are endogenous. However, there is evidence that shortfalls occur even for programs that remain broadly on track.3
We have reexamined this issue by focusing on two questions:
(1) What is the extent of “tapering out” of projected donor flows between the initial and third year of the program? To address this question, we examined program projections in the MONA database for nearly 100 ESAF/PRGF arrangements approved during 1995–2001.4
(2) What are the differences between actual flows and projected levels of donor aid? To address this question, we undertook two exercises. One focused on revised projections for the first year of the program in each successive yearly arrangement under the typical three-year concessional program.5 A second exercise compared outturns with projections at the start of the program for a three-year horizon (T, T+1, T+2). Because of data gaps in MONA, we examined projected and actual U.S. dollar values of aid flows in the fiscal accounts of staff reports for completed ESAF/PRGF arrangements in 20 sub-Saharan African countries.
The following are the main results (Appendix 3):
Aid flows were projected to decline (“taper out”) between the first and third year of the program in about three-fourths of cases. In half the cases, the magnitude of the projected decline was less than 1 percent of GDP, but in 10 percent of the cases projected declines exceeded 2 percent of GDP.
For the first year of the program the direction of differences between projections and actuals are equally divided: in half the cases projections exceeded actuals and in the other half actual aid exceeded that projected. In most cases, the differences were less than 1 percent of GDP.
Using the 20 case studies in sub-Saharan Africa, we find actual disbursements exceeding projections by more than 20 percent in a relatively small number of cases—between 2 to 5 cases depending upon the time horizon chosen. In fact, we observe a higher number of cases where projections exceeded actual disbursements by more than 20 percent (6 to 9 cases, depending on the time horizon chosen).
In summary, the data show that program projections of aid do tend to decline over the medium term in a majority of cases, albeit generally at a modest pace. However, on average, this does not appear to constrain aid flows on a year-to-year basis in programs that remain broadly on track. None of the evidence quoted here suggests that arrangements systematically underestimate aid flows in the outer years in program projections. However, the relatively simple analysis used here cannot answer the question—which goes beyond the scope of the current evaluation—whether more ambitious public spending (and deficit) targets, linked to poverty reduction, could have resulted in the mobilization of additional concessional external financing.
Social Spending Under IMF-Supported Programs: Cross-Country Evidence
Past IMF staff studies have investigated trends in health and education spending in developing countries. Gupta, Clements, and Tiongson (1998), using a sample of 118 developing and transition countries, find that since the mid-1980s real per capita spending on education and health has increased, on average, in developing countries but decreased in the transition economies. They observe that comparable increases can be observed for countries that had IMF-supported adjustment programs during the same period despite the fiscal consolidation often required by those programs.
In this section, we address the following question: What is the impact of the presence of an IMF-supported program on the level of social spending (other factors being held constant) relative to a situation without a program? For this purpose, we have investigated what happens to public sector social spending under IMF-supported programs using a broad sample of 146 countries in the 1985–2000 period.6 Four different indicators were used for each type of spending: as a share of GDP, as a share of total government spending, as an index of real spending at domestic prices, and in U.S. dollars per capita.7
The basic statistical framework relates social spending in a particular country and year to the presence of an IMF-supported program that year and to a set of (control) variables that may also influence the level of social spending. The detailed discussion of methodological issues and results is provided in Appendix 4. We present here some basic descriptive statistics and our main conclusions.
Table 6.1 summarizes the mean values and standard deviations of each indicator for health and education spending. The size of the standard deviation relative to the mean indicates that there is considerable variability in the level of public spending on health and education.
Public Sector Social Spending Indicators
Public Sector Social Spending Indicators
| Indicator | Observations | Mean | Std. Dev. | ||
|---|---|---|---|---|---|
| Health spending | |||||
| As percent of GDP | 1,452 | 2.2 | 1.5 | ||
| As percent of total public spending | 1,462 | 7.3 | 3.8 | ||
| Per capita, at real domestic prices | |||||
| (index, country average | |||||
| 1985–2000 = 100) | 1,418 | 100.0 | 30.0 | ||
| Per capita, in U.S. dollars | 1,424 | 6.1 | 9.4 | ||
| Education spending | |||||
| As percent of GDP | 1,452 | 4.2 | 2.0 | ||
| As percent of total public spending | 1,465 | 14.3 | 5.2 | ||
| Per capita, at real domestic prices | |||||
| (index, country average | |||||
| 1985-2000 = 100) | 1,413 | 100.0 | 25.3 | ||
| Per capita, in U.S. dollars | 1,419 | 10.2 | 14.8 | ||
Public Sector Social Spending Indicators
| Indicator | Observations | Mean | Std. Dev. | ||
|---|---|---|---|---|---|
| Health spending | |||||
| As percent of GDP | 1,452 | 2.2 | 1.5 | ||
| As percent of total public spending | 1,462 | 7.3 | 3.8 | ||
| Per capita, at real domestic prices | |||||
| (index, country average | |||||
| 1985–2000 = 100) | 1,418 | 100.0 | 30.0 | ||
| Per capita, in U.S. dollars | 1,424 | 6.1 | 9.4 | ||
| Education spending | |||||
| As percent of GDP | 1,452 | 4.2 | 2.0 | ||
| As percent of total public spending | 1,465 | 14.3 | 5.2 | ||
| Per capita, at real domestic prices | |||||
| (index, country average | |||||
| 1985-2000 = 100) | 1,413 | 100.0 | 25.3 | ||
| Per capita, in U.S. dollars | 1,419 | 10.2 | 14.8 | ||
One approach to determine the impact of IMF-supported programs on social spending is to compare periods with and without a program in a given country. This is reported in Table 6.2. In the large majority of countries for which data are available, there is no statistically significant difference in social spending between these two periods.8 In the cases where the results are significant, the outcome depends on how spending indicators are measured. When spending in health and education is measured as a share of GDP or total public spending, we find there are more countries which show a significantly higher mean during program years than those that show a lower mean. However, the reverse is true when this spending is measured in per capita terms.
Number of Countries With and Without Statistically Significant Results
Number of Countries With and Without Statistically Significant Results
| Percent of GDP | Percent Total Spending | U.S. Dollars Per Capita | Domestic Real Prices Per Capita | ||
|---|---|---|---|---|---|
| Health spending | |||||
| Number of countries with (statistically significant) higher spending when there is an IMF-supported program | 8 | 13 | 3 | 10 | |
| Number of countries with no significant difference between years with and without IMF-supported programs | 78 | 76 | 83 | 75 | |
| Number of countries with (statistically significant) lower spending when there is an IMF-supported program | 7 | 4 | 6 | 7 | |
| Education spending | |||||
| Number of countries with (statistically significant) higher spending when there is an IMF-supported program | 7 | 11 | 1 | 8 | |
| Number of countries with no significant difference between years with and without IMF-supported programs | 83 | 76 | 86 | 71 | |
| Number of countries with (statistically significant) lower spending when there is an IMF-supported program | 5 | 8 | 6 | 14 | |
Number of Countries With and Without Statistically Significant Results
| Percent of GDP | Percent Total Spending | U.S. Dollars Per Capita | Domestic Real Prices Per Capita | ||
|---|---|---|---|---|---|
| Health spending | |||||
| Number of countries with (statistically significant) higher spending when there is an IMF-supported program | 8 | 13 | 3 | 10 | |
| Number of countries with no significant difference between years with and without IMF-supported programs | 78 | 76 | 83 | 75 | |
| Number of countries with (statistically significant) lower spending when there is an IMF-supported program | 7 | 4 | 6 | 7 | |
| Education spending | |||||
| Number of countries with (statistically significant) higher spending when there is an IMF-supported program | 7 | 11 | 1 | 8 | |
| Number of countries with no significant difference between years with and without IMF-supported programs | 83 | 76 | 86 | 71 | |
| Number of countries with (statistically significant) lower spending when there is an IMF-supported program | 5 | 8 | 6 | 14 | |
This type of comparison suffers from the obvious limitation that it attributes all the difference in program years to the fact of having a program. This is not a suitable counterfactual since there are other variables at work that affect social spending and their effect must be netted out.
To isolate the impact of an IMF-supported program on social spending, using the pooled cross-section time series data, we need a methodology that:
Includes variables that have a direct effect on social spending, such as GDP per capita and share of school-age population. Not doing so would attribute to the presence of the IMF effects that are the result of these other variables (it is necessary to avoid a “missing variable bias”).
Recognizes that years with an IMF-supported program are not “normal” years, and that the special factors explaining the presence of a program could also, in principle, have an independent impact on social spending. For example, a country could seek an IMF-supported program as a result of an external shock (such as a sharp deterioration in the terms of trade) that may require a reduction in government spending with or without the presence of the Fund (i.e., it is important to take into account the endogeneity of IMF-supported programs).
Takes into account that social spending tends to change sluggishly and is heavily affected by levels of spending in previous periods. This reflects not only that most programs are conceived as permanent or at least spanning several years, but also the political economy of budget allocation—most programs have constituencies that resist change. For these reasons, explanatory variables, including the presence of an IMF-supported program, are likely to have effects that are not instantaneous and may extend beyond one period (i.e., it is necessary to take into account possible problems of serial correlation and nonstationarity in the data series).
These problems have been addressed by using regression analysis in which we combine a series of explanatory variables that are directly expected to have an impact on social spending with the use of instrumental variables to model the presence of an IMF-supported program. (The estimated equations are reported in Appendix 4.)
The empirical results show that, on average, the presence of an IMF-supported program does not reduce social spending. In fact, the result shows that the presence of a program is associated with increased public spending in health and education measured as either a share of GDP, total spending, or in real terms compared with a situation without a program. However, the positive effects attributable to the program are short-lived. For these effects to be durable, they would have to be followed by further policy actions in these sectors beyond the program period. The results do not show any marked difference in the impact of programs supported by concessional or nonconcessional resources.
Figure 6.1 shows the estimated impact of a two-year IMF-supported program on education and health spending, using the regression results reported in Appendix 4, Table A4.1. The vertical axis provides point-estimates of the effect of a program relative to a situation without a program, all other factors being the same; the horizontal axis represents the timeline. Public spending in each of the health and education sectors increased by about 0.3 to 0.4 percentage point of GDP compared with a situation without a program. There is still a residual effect in the third year (when there is no longer a program), but this declines geometrically thereafter.
Estimated Impact of a Two-Year IMF-Supported Program
Sources: IEO staff estimates based on regression coefficients.Estimated Impact of a Two-Year IMF-Supported Program
Sources: IEO staff estimates based on regression coefficients.Estimated Impact of a Two-Year IMF-Supported Program
Sources: IEO staff estimates based on regression coefficients.Whether this increase in spending sufficiently protects the most vulnerable groups during the program years will depend greatly on how well that increase in spending is targeted. If it is distributed according to past allocations—usually a high share spent in curative health or higher education and a high wage bill relative to recurrent inputs—the impact may be limited. If, on the other hand, it is used to fund targeted programs (old ones or new ones that can be activated during crisis) or to protect critical nonwage inputs (school supplies, school feeding programs, vaccines, and other critical medical inputs in basic health care), the impact could be much higher.
Role of the IMF in Connection with Social Expenditure and Social Protection
The role of the IMF vis-á-vis social spending has evolved as a result of a number of guidelines issued at different times. In 1991 the Managing Director issued guidelines to IMF staff directing that they should be explicitly concerned with the effects of economic policies on the poor and should discuss these concerns with government officials.9 In 1997, new guidelines on social spending were issued to staff.10 The guidelines emphasized the need for monitoring trends in this area and incorporating realistic targets into government budgets in the Letters of Intent on the basis of sector work by the World Bank (Box 6.1). In subsequent years, IMF management emphasized the need for a social pillar in the reform of the international financial architecture.11
The 1997 Guidelines on Social Expenditure
The guidelines call for the following:
IMF staff should use available fiscal data to keep track of main trends and developments in health and education spending and report these as memorandum items in fiscal tables in staff reports. Discussions on trends in social spending could be included in Recent Economic Development reports.
IMF staff should rely on the sector expertise of other institutions in health and education and should, in particular, strengthen collaboration with World Bank staff. In those countries where health and education spending data are already available and relevant analyses from other institutions, in particular the World Bank, already exist, IMF staff should attempt to draw conclusions (on the basis of trends in the subject country and comparisons with other countries) regarding the level and efficiency of spending in health and education.
IMF staff should rely on recent sector work by the Bank to incorporate realistic targets into government budgets and IMF-supported programs. These targets would not be expected to be performance criteria. It may be appropriate to encourage the authorities to incorporate such targets for health and education spending in the Letters of Intent for IMF-supported programs when the staff has examined the underlying analyses, and the targets are consistent with the overall macroeconomic framework and are monitorable.
IMF staff should continue to monitor developments in basic social indicators, such as poverty rates, infant mortality, life expectancy, illiteracy, school enrollment, and access to basic social services that are compiled by the World Bank and available online. In countries where such indicators are worsening or failing to improve in line with other developing countries, IMF staff should seek World Bank advice, and, if necessary, raise this issue with the authorities.
In 1999, the Board discussed a paper on social issues in IMF-supported programs 12 in which the staff made proposals to (1) establish quantitative targets for education and health care spending and to strengthen efforts to monitor such spending; (2) occasionally set performance criteria on minimum spending thresholds; and (3) in some circumstances, monitor budget allocations for selected key inputs such as books and medicines. The Board discussion revealed divergent views on the subject. Several Directors urged caution, warning that the IMF should not allow its primary mandate to be diluted and pointed out that the IMF does not have the expertise needed to assess the quality of social spending and related issues and could best contribute to poverty reduction through its support of economic policies that provide a conducive environment for sustained. growth. Some Directors felt that staff should assess, in the course of surveillance, the adequacy of social policy instruments, the performance of social safety nets, and the potential social ramifications of macro-economic and financial policies, but others worried that this might detract from standard Article IV surveillance. Some Directors stressed the importance of efficient and well-targeted spending to ensure that gains in social indicators were commensurate with spending increases.
On the issue of incorporating social expenditures in program design, Directors considered that where social spending was critically low, structural benchmarks should continue to be used selectively to protect social spending and promote institutional reforms. However, while many Directors thought that such structural benchmarks should only be used in programs supported by concessional financing, others saw merit in also applying performance criteria to a broader range of IMF-supported programs. In establishing structural benchmarks, IMF staff would rely on input from the World Bank and other institutions to ensure that the targeting and quality of spending would remain optimal.
While the need for World Bank and IMF collaboration on social spending has been stressed on several occasions, it presents several operational problems in practice. These surfaced in the recent discussion by Executive Directors of proposals from the staff on collaboration with the World Bank on public expenditure issues.13 Directors stressed that the IMF and the Bank should maintain a clear division of labor between the two institutions with the IMF taking the lead on the aggregate aspects of macroeconomic policy and their related instruments, and the Bank on issues relating to public expenditure composition and efficiency. They highlighted the need to better plan missions so as to reduce the burden on country authorities, better coordinate the different time frames of Fund and Bank work on public expenditure issues, and strengthen the collaboration with donors on country-led reform strategies. Directors also endorsed a framework that focuses on the articulation by the government of public expenditure reform strategies; an integrated and well-sequenced program of technical and financial assistance from development partners (including diagnostic work) to support countries’ public expenditure reform strategies; and periodic reporting by countries of their performance in public expenditure policy, financial management, and procurement.
More recently, the emphasis on streamlining conditionality has raised new questions. Discussions with a number of staff suggest that there is uncertainty regarding how to interpret the 1997 Guidelines on Social Expenditure in light of the streamlining initiative.
In PRGF-supported programs, closer World Bank–IMF collaboration is mandated through the PRSP process, which calls for the monitoring of social and other poverty-reducing expenditures and for an explicit social impact analysis of major proposed policy reforms. Hence, in these countries, a framework for a more coordinated approach to social issues exists. However, for non-PRGF countries, there is a lack of clarity on how social policies should be handled. There is no PRSP-type framework and the World Bank may not have been involved in the social sector with the depth needed to deliver the relevant inputs on the short-term time schedule relevant for IMF operations. In these circumstances, the treatment of social issues in non-PRGF programs may well depend significantly on the emphasis provided by individual staff, the way they interpret the streamlining mandate, and the degree to which they collaborate with the Bank, itself dependent on the extent of readily available analysis done by the Bank. To assess what happens in practice, we examined a number of programs in depth.
A review of social issues in program design in 15 arrangements
The sample of 15 IMF-supported programs provides a basis for assessing how social issues are treated within the context of program design.14 We posed a number of questions listed in Table 6.3, which also summarizes the results (elaborated in Appendix 5). Social spending issues are mentioned in almost all programs and changes in spending are noted in two-thirds of programs. However, little effort is made to sharpen the definition of social spending or to analyze the reasons behind trends. Only half the program requests that note changes in social spending actually analyze these changes. Few programs (other than in the PRSP/PRGF countries) establish explicit monitoring and feedback systems. Thus the empirical basis for identifying policy actions is often absent.
Effectiveness in Identifying and Monitoring Social Spending in the Program Requests of 15 Selected Arrangements
(Percentage of cases where the answer to question is “yes”)
Effectiveness in Identifying and Monitoring Social Spending in the Program Requests of 15 Selected Arrangements
(Percentage of cases where the answer to question is “yes”)
| Efforts at improving the empirical basis for policy | ||
| Is social expenditure referenced at all? | 93 | |
| Are changes in social spending noted? | 67 | |
| Do programs include time series data on social spending? | 67 | |
| Do programs define social spending clearly? | 0 | |
| Are changes in social spending analyzed? | 33 | |
| Efforts at identifying policies and actions | ||
| Are there specific problems or issues identified? | 80 | |
| Are there efforts to identify how social spending could be protected? | 33 | |
| Are there any performance criteria or benchmarks in connection with social spending? | 40 | |
| Did reviews follow up on issues raised in the program request? | 100 | |
Effectiveness in Identifying and Monitoring Social Spending in the Program Requests of 15 Selected Arrangements
(Percentage of cases where the answer to question is “yes”)
| Efforts at improving the empirical basis for policy | ||
| Is social expenditure referenced at all? | 93 | |
| Are changes in social spending noted? | 67 | |
| Do programs include time series data on social spending? | 67 | |
| Do programs define social spending clearly? | 0 | |
| Are changes in social spending analyzed? | 33 | |
| Efforts at identifying policies and actions | ||
| Are there specific problems or issues identified? | 80 | |
| Are there efforts to identify how social spending could be protected? | 33 | |
| Are there any performance criteria or benchmarks in connection with social spending? | 40 | |
| Did reviews follow up on issues raised in the program request? | 100 | |
One difficulty is that social spending is not explicitly defined. Tables or boxes dealing with social spending in program documents typically associate social spending with education and health and sometimes tables indicate a single line titled “social spending” with no definition of the components.
About one-third of programs explore how to protect social spending, although typically at a very aggregate level of appropriations such as education spending. About 40 percent of programs used some conditionality in the form of benchmarks or indicative targets—none use performance criteria.
Program reviews performed very well in following up whatever social issues were originally raised in the program request, and in many cases discussion of these issues was more extensive in the reviews than in the initial program request. For example, in Costa Rica the program request only briefly mentioned social issues and broadly discussed the need to strengthen the social safety net. The reviews, however, were more detailed and included more specific suggestions to achieve better targeting of social spending such as restructuring several agencies, decentralization, and encouraging the use of private suppliers of social services.
Similar patterns are found when examining comments from PDR and FAD during the internal review process. These comments often give feedback in this area, providing specific suggestions for the design and the support of priority social programs to protect vulnerable groups. However, most of these comments are concentrated in the reviews during program implementation and are, therefore, too late to influence the program design.
These results also suggest reasons why, despite good intentions, programs often fail to protect critical social spending. Programs recognize the need for action in the social sector but are vague about the specific types of spending that require protection. For example, in the case of the Philippines program, the staff report stated that “the staff urged the authorities to protect programs directed at poverty reduction in implementing the cuts. The authorities agreed, and explained that individual agencies had been instructed to reduce certain nonessential outlays (such as travel and training) by 50 percent. Agencies’ revised spending plans are being reviewed with a view to protecting social programs as much as possible, especially those directed at poverty alleviation. Social programs would also be the first ones to be restored if fiscal developments during the year permit.” Despite these good intentions, the proportion of the population served by various health programs declined, reflecting the absence of clear definitions regarding the specific critical programs to be protected, compounded by a lack of monitoring.
This picture, however, is not uniformly negative. The Algeria program, for example, defined very specific measures to revamp the social safety net in order to protect better the most vulnerable segments of the population via improved targeting. The program built on recommendations from an FAD technical assistance mission to introduce a public works program that would be self-targeting with a much lower remuneration than the minimum wage. Short-term unemployment would be dealt with by introducing an unemployment insurance mechanism to replace a system that imposed large severance payments on enterprises. Moreover, the authorities agreed to merge three other cash transfer schemes.
The use of conditionality to achieve social sector objectives was limited. Of the 15 programs examined, only 6 contain explicit social sector conditionality in the form of structural benchmarks and the implementation results were mixed. In the Algeria program, a structural benchmark was introduced to reform the social safety net through the introduction of a public works scheme and the benchmark was eventually met. In the Bulgaria program, a structural benchmark was set on improving the cost effectiveness of health care, and that benchmark was subsequently only partially met. For the Pakistan program, an indicative target was put on social and poverty-related spending, but the target was not met. The Senegal program included a structural performance criterion relating to budgetary allocations for the health and education sectors. A closer look at the criterion, however, reveals that it actually only called for an action plan and communication to IMF staff on the issue. In the Ukraine program, a benchmark was set on specific reforms in the health and education ministries and that benchmark was also met, although some slippage occurred after the benchmark was removed from the program. The Venezuela program had structural benchmarks calling for legislation to reform the severance payment system and strengthen the social safety net. These were implemented but with delay.
There are situations where poorer groups have not only been adversely affected by output declines and devaluations in crisis periods prior to programs, but also by fiscal and price adjustment measures included in programs for macroeconomic reasons but which may have second-round adverse effects. The Ecuador program was well aware of this phenomenon and it supported the government’s plan to index the preexisting cash transfer program (Bono Solidario) and other poverty programs to offset negative effects on the poor. However, although there was clear conditionality on the pricing of fuels, spending control, and raising the VAT, none of the social measures in the Letter of Intent with the purpose to offset these effects was incorporated as a structural benchmark (see Table 6.4).
The Ecuador Program: Imbalance Between Efficiency and Equity Measures Underpinned by Conditionality
The Ecuador Program: Imbalance Between Efficiency and Equity Measures Underpinned by Conditionality
| Measures in the 2000 Memorandum of Understanding | Included as a Performance Criterion (PC)/Benchmark (B)? | |
|---|---|---|
| Adjustments of prices | ||
| Fuels | Yes (PC) | |
| Cooking gas | Yes (PC) | |
| Electricity rates | No | |
| Other fiscal measures | ||
| Eliminate temporary tariff surcharge | Yes (B) | |
| Control over expenditure, including wage bill | Yes (PC) | |
| Payment of domestic arrears | Yes (PC) | |
| Tax measures | ||
| Raise VAT and increase tax base | Yes (B) | |
| Lower income tax threshold | Yes (B) | |
| Reduce evasion | No | |
| Reduce loopholes | No | |
| Improve tax administration | No | |
| Reduce earmarking | Yes (B) | |
| Elimination of nuisance taxes | Yes (B) | |
| Consumption tax on gasoline | Yes (B) | |
| Social measures | ||
| Adjustment of Bono Solidario | No | |
| Improve targeting of Bono Solidario | No | |
| Nutrition and family programs | No | |
| Community programs | No | |
| Education programs | No | |
| Increase social spending if revenues allow | No | |
The Ecuador Program: Imbalance Between Efficiency and Equity Measures Underpinned by Conditionality
| Measures in the 2000 Memorandum of Understanding | Included as a Performance Criterion (PC)/Benchmark (B)? | |
|---|---|---|
| Adjustments of prices | ||
| Fuels | Yes (PC) | |
| Cooking gas | Yes (PC) | |
| Electricity rates | No | |
| Other fiscal measures | ||
| Eliminate temporary tariff surcharge | Yes (B) | |
| Control over expenditure, including wage bill | Yes (PC) | |
| Payment of domestic arrears | Yes (PC) | |
| Tax measures | ||
| Raise VAT and increase tax base | Yes (B) | |
| Lower income tax threshold | Yes (B) | |
| Reduce evasion | No | |
| Reduce loopholes | No | |
| Improve tax administration | No | |
| Reduce earmarking | Yes (B) | |
| Elimination of nuisance taxes | Yes (B) | |
| Consumption tax on gasoline | Yes (B) | |
| Social measures | ||
| Adjustment of Bono Solidario | No | |
| Improve targeting of Bono Solidario | No | |
| Nutrition and family programs | No | |
| Community programs | No | |
| Education programs | No | |
| Increase social spending if revenues allow | No | |
A critical issue for program design is whether critical programs can be protected at affordable cost and in a manner which can be effectively monitored. This is certainly possible but it requires a high level of control over institutional management to implement these measures of protection. Box 6.2 shows how public hospitals in Ecuador adjusted to the 1998–99 crisis prior to the program. The wage bill and personnel expenses were protected but free provision of drugs to patients and even food for inpatients declined sharply relative to spending on personnel. Nonwage inputs—which are a small share to begin with (only 20 percent of hospital spending)—were squeezed. In principle, it should be possible to protect these items without jeopardizing any macro-economic target in any standard program. However, doing so requires identification of critical programs and spending categories prior to the crisis and the ability to ensure that the relevant allocations are effectively protected when they come under pressure in crisis situations.
How Public Hospitals in Ecuador Adjusted in a Time of Crisis
As a result of a series of external shocks and a domestic banking crisis, Ecuador experienced a macroeconomic crisis of major proportion in 1999. Output declined by 7.5 percent, inflation accelerated to approximately 60 percent a year, and the sucre/dollar exchange rate almost doubled.
While nominal public sector wages increased by 34 percent between 1998 and 1999, the health budget only increased by about 12 percent. Under these circumstances, how did a typical public hospital adjust when salaries accounted for about 80 percent of its operations and the cost of nonwage medical inputs went up with the devaluation? To answer this question, a sample of six large public hospitals in Quito and Guayaquil were visited to assess how they coped with the crisis. They accounted for about 12 percent of the total number of hospital beds nationwide.
The major finding was that the sharp erosion in real budgets in 1999 translated into a reduction of nonwage medical inputs and maintenance of equipment. Consequently, hospitals were forced to cut back care to patients. In three of the four hospitals that provided data, outpatient services declined 26 percent to 37 percent.
In addition, the number of drug prescriptions dispensed declined very sharply in three hospitals, by amounts ranging from one-half to four-fifths, and increased by about 10 percent in those hospitals where some cost recovery was feasible. Independent data for the overall public health system show a decline of about 14 percent in the total number of prescriptions dispensed by the entire system (see figure).
For some of the hospitals visited, data were obtained on the number of food rations received by the hospital staff versus patients. In the Quito hospital, rations for patients were reduced during the crisis—sometimes severely while those for staff remained relatively constant Only in one Guayaquil hospital were food rations maintained thanks to additional funding received by the hospital to mitigate the impact of El Niño on the coastal areas.
This example illustrates that the protection of small but critical nonwage budgetary items under fiscal adjustment is a major challenge in the design and monitoring of adjustment programs.
There are examples of cost-effective and targeted programs that could be protected at low fiscal cost in case of a crisis. One example comes from Tanzania (see Box 6.3), where well-targeted health intervention with an emphasis on children was implemented in a pilot program covering two districts at a cost of less than $2 per capita. Another example is the Progresa Program in Mexico. Poor rural families received cash transfers, school supplies, and nutrition supplements conditional on children’s school attendance and regular preventive health care. The program has reached about 2.5 million households at a cost of about 0.2 percent of GDP. Budgetary shocks that threaten these allocations can be protected at low cost and with little impact on the overall fiscal program. In summary, if countries introduce beforehand well-targeted social programs, they can easily be protected or activated at low fiscal cost in a crisis situation.
Protecting Critical Programs Is Not Costly When Programs Are Well Targeted
An experimental health intervention in Tanzania shows that small additional resources devoted to health care in a poor country can alleviate the burden of disease if carefully allocated. The intervention was carried out in two rural districts by the Tanzanian Essential Health Intervention Project (TEHIP), a joint venture of Tanzania’s Health Ministry and Canada’s International Development Research Centre (IDRC).1
The key innovation was to focus financial resources on diseases that imposed the highest burden on the population. It was found, for example, that a cluster of childhood problems such as malaria, pneumonia, diarrhea, malnutrition, and measles accounted for 28 percent of disease in the districts, but only received 13 percent of the local health care budgets. An additional $2 a head allocated to the district’s health care budget was to be spent on diseases with the largest social cost based on years of life lost. The results thus far have been dramatic. Infant mortality fell by 28 percent from 1999 to 2000. The number of deaths prior to five years of age dropped by 14 percent. There is no evidence of similar improvements in that period in nearby districts or in Tanzania overall.
These are the types of programs that need to be protected under macroeconomic shocks that put pressure on public finances. It is clear that IMF-supported programs could make room for such interventions. However, making sure public expenditure management systems are able to deliver resources to desired destinations depends on local knowledge and will require support from the World Bank. It is not possible to set up such monitoring and delivery systems within the short time frame in which the negotiation and implementation of an IMF-supported program takes place. Nor is this an area where the IMF has the necessary expertise.
To deal with such problems of a potential mismatch of time frames, the IMF needs to encourage the authorities, independently of the negotiation of a particular IMF-supported program (and probably with support from the World Bank and other external partners), to (1) identify core budgets that would be protected in case of budget cuts; (2) develop public expenditure management systems capable of monitoring the flow of resources to critical programs in real time; and (3) protect the cash flow to items in the core budget during times of fiscal pressures. In countries like Tanzania, the framework of the PRSP exists to address such issues, but the approach to be taken is less obvious in non-PRSP/PRGF cases.
1 Reported in The Economist, August 17, 2002.The experience of Chile (not part of our evaluation) is of general interest for middle-income countries. Not only has Chile been effective in protecting critical programs such as children’s basic health care and nutrition, but it has also been able significantly to realign the budget toward social spending while improving the incidence of public spending towards the lower-income population. This has been accomplished without unduly increasing the tax burden. That tax burden is about 19 percent of GDP, a product of moderate tax rates and good collection. About 70 percent of spending in basic social services and cash assistance is focused on the first two quintiles of the population. These achievements have been the product of many years of institutional reforms and political consensus regarding these policy priorities, and it provides a good reference point of what is possible.
In addition to examining social sector issues in the 15 main programs chosen for this study, we went a step further in order to evaluate the latest arrangement for 8 of the 15 countries for which there was a more recent program (these include the Algeria SBA 1995, Bulgaria SBA 2002, Jordan SBA 2002, Pakistan PRGF 2001, Peru SBA 2002, Romania SBA 2001, Tanzania PRGF 2000, and Uruguay SBA 2002 programs). We adopted identical criteria to those used to assess the treatment of social issues in the original 15 IMF-supported programs. Results show that the more recent programs exhibit slight improvements in categories such as noting and analyzing changes in social spending, identifying specific social spending issues, and actions to protect social spending. In 3 of the 8 programs, structural benchmarks were used to support social protection measures. At the same time, there is little change or even a slight deterioration in presenting a series of social spending data. This suggests there is still room for considerable improvement.
Conclusions
It is clear from our evaluation that protection of social spending on critical and well-targeted programs in the social sector can play an important role in protecting vulnerable groups from adverse shocks and budgetary retrenchments at fairly low cost. This emphasis is also consistent with the IMF Articles of Agreement (especially Article I (v)) and with commitments made in the follow-up of the 1995 World Summit for Social Development (see IMF, 2000a). Efforts should, therefore, be made to build such elements into program design wherever possible. However, a framework is necessary that takes account of four operational constraints. (1) To be effective, and acceptable, policies in this area must be truly home-grown and fully owned domestically; the initiatives must, therefore, come from the country. (2) Since the IMF does not have expertise on social sector issues, nor is this an area of its comparative advantage, inputs from other agencies, especially the World Bank (and possibly also others), are critical. (3) There is a mismatch of time frames between the short-term nature of IMF programs and the longer-term time frame needed for building institutions and budgetary systems that can provide social support in times of crisis effectively. (4) Finally, it is necessary to ensure that incorporation of social protection system does not contradict the recent streamlining initiative by leading to an overload of conditionality.
In the case of low-income countries, the PRSP framework could potentially meet these requirements. The extent to which this is actually achieved will be separately examined in the ongoing IEO evaluation of the PRSP/PRGF experience. However, there is at present no framework for non-PRGF eligible, predominantly middle-income countries that would ensure identification of critical and homegrown social sector support programs that could be used as mechanisms for social protection at the time of crisis.
The PRSP framework is obviously not appropriate for middle-income countries, but in the absence of any framework there will be a growing divergence between the way social issues are treated between PRGF and non-PRGF countries. It is, therefore, necessary to revisit the 1997 guidelines with special reference to what IMF staff should do consistent with the overall operational constraints listed above.
Some elements of a workable approach can be readily identified. First, the mismatch of time frames suggests that work in this area must be undertaken not at the time of crisis but much earlier as part of normal surveillance. In order to encourage a home-grown initiative, the IMF could request governments to consider identifying critical social sector programs that could serve as effective social safety nets that could be intensified in the event of crisis. The IMF could encourage countries to approach the World Bank for assistance in this area. The IMF on its part, consistent with its mandate, could report on the authorities’ responses in this area and monitor programs in developing social safety nets.
Building on recent initiatives (such as the call for increased coordination on public expenditure management (PEM) issues), both institutions could agree with the authorities on the reforms that would need to be tackled and an appropriate sequencing. Where joint efforts are required, for example in public expenditure management, a work program in these areas would be jointly established. On the basis of the resulting joint effort, the IMF and the World Bank would assist the authorities in setting up mechanisms to track critical social spending throughout the budget and identify ultimate allocations including to local governments where a significant amount of spending is decentralized. In this regard, establishment of better and more transparent monitoring systems is probably one of the major contributions that can be made to encourage homegrown policy initiatives in this area.
See, for example, Collier and Gunning (1999). The authors argue that the disincentive arises because programs usually do not allow additional aid (i.e., above the amount projected) to be spent, favoring instead the channeling of the extra amounts into increasing international reserves or paying down debt.
Bulíř. and Hamann (2001) reported that countries with uninterrupted programs received, on average, about three-quarters of program aid commitments. Countries where programs were interrupted received only about one-third of program aid commitments.
From November 1998, the three-annual-arrangement structure of the ESAF was replaced by a one three-year-arrangement structure. The comparison includes projections under both types of structure.
We looked at program years for which MONA had data on both projections and outturns—mainly arrangements that remained on track over successive years. This reduced the sample size to 40 observations The outturn data for a particular program year was obtained from data reported in connection with a subsequent arrangement. Cases where there was a break in the series of one year or more between successive arrangements were dropped from the sample. Thus the sample was biased in favor of programs that remained broadly on track.
A discussion of methodological issues and a presentation of results is in Appendix 4. For a more comprehensive report on the analysis and methodological issues underlying these findings, see Martin and Segura-Ubiergo (forthcoming). Social spending is measured on the basis of annual data on government spending on health and education using a database created by the Fiscal Affairs Department (FAD), and checked for accuracy by IMF staff from each country desk. See Baqir (2002) for a description and coverage.
In the absence of a sector-specific price index, social spending was deflated by the general consumer price index. Expenditures in U.S. dollars were calculated at the annual average exchange rate, and deflated by the U.S. wholesale price index.
At least at the 90 percent confidence level.
“Revised Guidelines on Poverty-Related Work,” Office Memorandum from the Managing Director to Heads of Departments, March 8, 1991.
10 “Guidelines on Social Expenditure,” Office Memorandum from the Managing Director to Heads of Departments, May 28, 1997.
Remarks by the Managing Director to UN ECOSOC Ambassadors, New York, June 31, 2000.
Gupta, Dicks-Mireaux, Khemani, McDonald, and Verhoeven (2000) update the work presented to the Board in “Review of Social Issues and Policies in IMF-Supported Programs,” EBS/99/171, August 27, 1999. The discussion in the next two paragraphs draws upon the summing up of the Board discussion.
“Bank/Fund Collaboration on Public Expenditure Issues,” SM/03/73, February 19, 2003. This paper does not explicitly address collaboration on social spending but the discussion is highly relevant.
One of these programs (Tanzania) was supported by concessional IMF resources and two (Senegal and Pakistan) by a mix of concessional and regular IMF resources. All the rest involved the use of IMF general resources only.
