## _sdn1115

## Source details

**Canonical URL:** [_sdn1115](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/sdn/2011/_sdn1115.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/sdn/2011/_sdn1115.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/sdn/2011/_sdn1115.pdf.json)

---

### I. Introduction
- Purpose: revisit the debate on IMF-supported programs and education and health spending using the most comprehensive dataset assembled thus far and advanced quantitative techniques.
- Dataset: covers 1985–2009 for 140 developing countries, including 70 low-income countries eligible for concessional financing from the IMF.
- Key summary finding:
  - In low-income countries, IMF-supported programs have a positive effect on education and health outlays.
  - In other developing countries, spending has risen but IMF-supported programs appear to have no effect per se on these expenditures.
- Structure: critics’ views (Section II); trends and quantitative analysis controlling for macroeconomic and other determinants (Section III); conclusions (Section IV).

### II. IMF-Supported Programs and Social Spending: The Critics’ View
- Critics’ claims:
  - Countries with IMF-supported programs are obliged to decrease social spending to meet fiscal targets, squeezing priority expenditure on education and health.
  - Wage ceilings, particularly in the health sector, prevent desirable increases in health employment.
  - Increases in aid intended to support higher health spending are diverted to repay domestic debt or increase international reserves.
- Prior empirical evidence:
  - Most studies in the 1990s and early 2000s found social spending rose more in program than in nonprogram countries.
  - Nooruddin and Simmons (2006): in democracies, IMF-supported programs reduced real per capita health spending and the share of government spending allocated to both education and health.
  - IEO (2003) econometric analysis: IMF-supported programs boosted education and health spending as a share of GDP and government spending, and in real per capita terms.
- Need for updated econometric work given new data through 2009.

### III. How Have IMF-Supported Programs Affected Social Spending?

- Data and patterns:
  - Public spending data for education and health for 1985–2009, 140 developing countries.
  - Education spending averages 4½ percent of GDP in the entire sample.
  - Health spending averages 3 percent of GDP in the entire sample.
  - Education expenditures do not differ substantially across regions in percent of GDP.
  - Public health spending is especially low in Asia and Sub-Saharan Africa.
- Incidence of IMF-supported programs:
  - Since 1985, 112 of the 140 countries in the sample had an IMF-supported program for at least one year.
  - Over the entire sample, about one-third of the time countries had IMF-supported programs.
  - By region, IMF-supported programs are more frequent in Sub-Saharan Africa (51 percent).

- Stylized facts on changes:
  - Cumulated over 10 years, spending increased every decade at the rate of 0.2 percentage point of GDP for education and 0.1 percentage point of GDP for health.
  - These decadal changes translate into annual percentage changes in real per capita terms of 3.5 percent for education spending and 3.4 percent for health spending.
  - Over a 10-year horizon, cumulative rises in education and health spending in low-income program countries are 0.5 and 0.3 percentage points of GDP, respectively.
  - Per capita education and health spending rose at about 4 percent per year in the program group versus about 2½ percent per year in low-income countries without programs.
  - Median annual increase in education and health spending in low-income program countries:
    - Between 1985 and 1999: 0.04 and 0.01 percent of GDP, respectively.
    - Since 2000: 0.07 and 0.06 percent of GDP, respectively.
  - Fiscal revenue effect (fragmentary): Tax revenues rose by a median of 0.19 percentage point of GDP per year in program countries (text fragment).

- Channels through which IMF-supported programs spur higher spending:
  - Effects on fiscal space, including reforms that increase the revenue effort.
  - Catalytic effect on donor financing.
  - Higher growth associated with programs can generate greater fiscal space.
  - For low-income countries, emphasis in programs to use additional resources—including those generated by debt relief—to increase poverty-reducing spending as part of poverty-reduction strategies.
  - Formal conditionality (performance criteria or benchmarks) has been used sparingly and does not appear to play a role in the observed increase in these outlays.

### IV. Quantitative Analysis: Estimates and Dynamics

- Approach and methods:
  - Quantitative analysis follows IEO (2003) methodology.
  - Estimations for education and health spending both in percent of GDP and as a share of total spending for low-income countries.
  - Two models reported: fixed effects and system GMM (preferred).
  - System GMM treats income levels and government balance as endogenous; instruments include lagged international reserves in months of imports, lagged log bilateral exchange rate to U.S. dollar, and lagged index of exchange rate classification.
  - Selection bias addressed via instrumentation in system GMM and use of inverse Mills ratio from probit regressions in fixed effects estimations.

- First-year and short-run effects (system GMM and fixed effects summarized):
  - In the first year, IMF-supported programs raise:
    - education spending by 0.22–0.26 percentage points of GDP (columns (1)–(4)).
    - health spending by 0.17–0.27 percentage points of GDP (columns (1)–(4)).
  - Under system GMM estimates, in the first year:
    - education spending rises as a share of total government outlays by about 1 percentage point.
    - health spending rises as a share of total government outlays by about ½ percentage point.

- Five-year and longer-run dynamic effects (system GMM estimates):
  - Over a five-year period with IMF-supported programs:
    - education spending increases by about 0.82 percentage point of GDP.
    - health spending increases by about 0.98 percentage point of GDP.
  - Year-by-year system GMM estimates relative to pre-program:
    - Year 1: Education 0.22, Health 0.27 (percentage point of GDP).
    - Year 3: Education 0.57, Health 0.69 (percentage point of GDP).
    - Year 5: Education 0.82, Health 0.98 (percentage point of GDP).
  - Example short program (two years of program followed by end):
    - Peak after two years: education up by 0.4 percentage point of GDP and health up by 0.5 percentage point of GDP relative to the pre-program year.
    - Effect then declines gradually to about 0.1 percentage point of GDP in about 10 years for both education and health outlays.
  - The effects on spending-to-GDP ratios eventually diminish and return to zero after a program ends.

- Selected Table 1 coefficients and statistics (as reported):
  - IMF program coefficients:
    - Education spending (% of GDP), Fixed effect: 0.26 (standard error 0.133).
    - Education spending (% of GDP), System GMM: 0.22** (standard error 0.101).
    - Health spending (% of GDP), Fixed effect: 0.17** (standard error 0.070).
    - Health spending (% of GDP), System GMM: 0.27*** (standard error 0.094).
    - Education spending (% of total spending), System GMM: 1.15*** (standard error 0.389).
    - Health spending (% of total spending), System GMM: 0.46** (standard error 0.182).
  - Lagged dependent variable coefficients (persistence): range 0.61*** to 0.85*** (example 0.85*** (0.029)).
  - Other example coefficients: second lagged dependent variable 0.18*** (0.051); government balance coefficients such as 0.02** (0.010) and -0.05** (0.027); inverse Mills ratio -0.17** (0.080) and -0.06* (0.036).
  - Sample sizes and fit: number of observations by column include 580, 580, 687, 687, 580, 580, 687, 664; number of countries by column include 54, 54, 59, 59, 45, 45, 59, 59; R-squared examples 0.651 (column 1) and 0.498 (column 3).
  - System GMM diagnostics: P-value for Sargan test 1.00; P-value for AR(2) test 0.42, 0.22, 0.56, 0.28 (by relevant columns).

- Cross-sample differences and mechanisms:
  - Channels appear more powerful in low‑income countries than in other developing countries.
  - Median combined education and health spending:
    - 5 percent of GDP in the low‑income sample.
    - 7 percent of GDP for other program countries.
  - Median annual increase in tax revenue to GDP ratios:
    - about 0.22 percentage point of GDP per year in low‑income program countries.
    - 0.06 percentage point per year in other program countries.
  - The document notes a figure of 0.14 percentage point of GDP in nonprogram countries.

### V. Conclusions
- Main finding: education and health spending have risen during IMF-supported programs at a faster pace than in developing countries as a whole.
- For low-income countries, IMF-supported programs have a positive and significant effect on social spending (education and health).
- Over a five-year period in low‑income program countries:
  - education spending as a share of GDP increases by about 0.8 percentage point of GDP.
  - health spending as a share of GDP increases by about 1 percentage point of GDP.
- For other developing countries, the estimated effect is neutral.
- Program design features and fiscal-revenue-enhancing reforms are important channels:
  - concessional financing,
  - debt relief,
  - focus on poverty reduction,
  - minimum indicative floors on social spending where appropriate,
  - reforms that raise revenues and catalyze donor financing.

*Source: _sdn1115 - Executive Summary, IMF Staff Discussion Note (excerpt).*

### Executive Summary ......................................................................................................

### Executive Summary

### I. Introduction
- Purpose: revisit the debate on IMF-supported programs and education and health spending using the most comprehensive dataset assembled thus far and advanced quantitative techniques.
- Dataset: covers 1985–2009 for 140 developing countries, including 70 low-income countries eligible for concessional financing from the IMF.
- Key summary finding: in low-income countries, IMF-supported programs have a positive effect on education and health outlays; in other developing countries, spending has risen but IMF-supported programs appear to have no effect per se on these expenditures.
- Structure: critics’ views (Section II); trends and quantitative analysis controlling for macroeconomic and other determinants (Section III); conclusions (Section IV).

### II. IMF-Supported Programs and Social Spending: The Critics’ View
- Critics’ claims:
  - Countries with IMF-supported programs are obliged to decrease social spending to meet fiscal targets, squeezing priority expenditure on education and health (examples cited).
  - Wage ceilings, particularly in the health sector, prevent desirable increases in health employment.
  - Increases in aid intended to support higher health spending are diverted to repay domestic debt or increase international reserves.
- Prior empirical evidence:
  - Most studies in the 1990s and early 2000s found social spending rose more in program than in nonprogram countries.
  - Two econometric studies with differing results:
    - Nooruddin and Simmons (2006): in democracies, IMF-supported programs reduced real per capita health spending and the share of government spending allocated to both education and health.
    - IEO (2003) econometric analysis: IMF-supported programs boosted education and health spending as a share of GDP and government spending, and in real per capita terms.
- Need for updated econometric work given new data through 2009.

### III. How Have IMF-Supported Programs Affected Social Spending?

#### A. Overview
- Data: public spending for education and health for 1985–2009, 140 developing countries.
- Regional patterns (Figures 1 and 2):
  - Education spending averages 4½ percent of GDP in the entire sample.
  - Health spending averages 3 percent of GDP in the entire sample.
  - Education expenditures do not differ substantially across regions in percent of GDP.
  - Health expenditures show greater variation across regions than education spending; public health spending is especially low in Asia and Sub-Saharan Africa.
- Incidence of IMF-supported programs:
  - Since 1985, 112 of the 140 countries in the sample had an IMF-supported program for at least one year.
  - Over the entire sample, about one-third of the time countries had IMF-supported programs.
  - By region, IMF-supported programs are more frequent in Sub-Saharan Africa (51 percent) than elsewhere.

#### B. Social Spending in IMF-Supported Programs: Stylized Facts
- Annual and decadal changes:
  - On average, education and health spending increased slowly in percent of GDP per year.
  - Cumulated over 10 years, spending increased every decade at the rate of 0.2 percentage point of GDP for education and 0.1 percentage point of GDP for health.
  - These decadal changes translate into annual percentage changes in real per capita terms of 3.5 percent for education spending and 3.4 percent for health spending.
- Regional differences:
  - Highest growth in real per capita spending observed in the CEE and CIS.
  - Regions with fastest real per capita spending growth did not necessarily have the greatest increases in spending as a share of GDP due to differences in economic growth (example: CEE and CIS saw strong GDP and per capita social spending increases that led to slight declines in spending-to-GDP ratios).
- Program versus nonprogram countries:
  - Social spending has increased at a faster pace for countries with an IMF-supported program, particularly low-income program countries.
  - Over a 10-year horizon, cumulative rises in education and health spending in low-income program countries are 0.5 and 0.3 percentage points of GDP, respectively.
  - Annual percentage changes in real per capita spending were slightly higher in low-income program countries than program countries as a whole.
  - Per capita education and health spending rose at about 4 percent per year in the program group versus about 2½ percent per year in low-income countries without programs.
- Post-2000 acceleration:
  - Increases in spending-to-GDP ratios have accelerated since 2000 in low-income program countries, reflecting effects of debt relief and a focus on poverty reduction in program design.
  - Median annual increase in education and health spending in low-income program countries:
    - Between 1985 and 1999: 0.04 and 0.01 percent of GDP, respectively.
    - Since 2000: 0.07 and 0.06 percent of GDP, respectively.
- Composition of government spending:
  - The share of government spending allocated to health and education has increased in most regions, most strongly in program countries.
  - Exception: low-income program countries in LAC, where this spending declined as a share of government outlays despite increases as a share of GDP.
- Channels through which IMF-supported programs spur higher spending:
  - Effects on fiscal space, including reforms that increase the revenue effort to provide additional resources for spending.
  - Catalytic effect on donor financing.
  - To the extent programs lead to higher growth, they can help generate greater fiscal space.
  - For low-income countries, emphasis in programs to use additional resources—including those generated by debt relief—to increase poverty-reducing spending as part of poverty-reduction strategies.
  - Formal conditionality (performance criteria or benchmarks) has been used sparingly and does not appear to play a role in the observed increase in these outlays.
- Fiscal revenue effect (fragmentary reported statistic):
  - Tax revenues rose by a median of 0.19 percentage point of GDP per year in program countries, compared with (text ends).

#### C. IMF-Supported Programs and Social Spending: Quantitative Analysis
- Approach:
  - Quantitative analysis follows IEO (2003) methodology to isolate the effect of IMF-supported programs from other determinants of social spending and the noncomparable nature of program versus nonprogram countries.
  - Methodological issues are covered in the Appendix.
- Estimation:
  - Effects estimated for education and health spending both in percent of GDP and as a share of total spending for low-income countries.
  - Two models reported: fixed effects and system GMM (generalized method of moments).
  - The shaded row of Table 1 provides the estimate of the independent effect of IMF-supported programs on the corresponding indicator of social spending, holding constant other determinants.
- Five-year program effect (summary stated earlier in the document):
  - Over a five-year period with IMF-supported programs, spending for education increases by about ¾ percentage point of GDP; and for health, by about 1 percentage point of GDP.
- Main quantitative conclusion:
  - Controlling for other determinants of education and health spending, including macroeconomic conditions, results confirm that IMF-supported programs have a positive and significant effect on social spending in low-income countries.
  - IMF-supported programs are also associated with increases in the share of government spending allocated to education and health.

### IV. Conclusions
- Main finding: education and health spending have risen during IMF-supported programs at a faster pace than in developing countries as a whole.
- For low-income countries, IMF-supported programs have a positive and significant effect on social spending (education and health).
- In other developing countries, spending has risen, although IMF-supported programs appear to have no independent effect per se on these expenditures.
- Program design features (concessional financing, debt relief, focus on poverty reduction, and minimum indicative floors on social spending where appropriate) and fiscal-revenue-enhancing reforms are important channels supporting increased social spending.

*Source: _sdn1115 - Executive Summary, IMF Staff Discussion Note (excerpt).*

### 0.14 percentage point of GDP in nonprogram countries.

### _sdn1115 - 0.14 percentage point of GDP in nonprogram countries.

### Key findings on IMF-supported programs and social spending
- IMF-supported programs increase social spending in low‑income countries.
- In the first year, IMF-supported programs raise:
  - education spending by 0.22–0.26 percentage points of GDP (columns (1)–(4)).
  - health spending by 0.17–0.27 percentage points of GDP (columns (1)–(4)).
- Under system GMM estimates, in the first year:
  - education spending rises as a share of total government outlays by about 1 percentage point.
  - health spending rises as a share of total government outlays by about ½ percentage point.
- Quantitative analysis indicates that IMF-supported programs do not have an independent effect on social spending for countries outside of the low‑income country sample.
- The document notes a figure of 0.14 percentage point of GDP in nonprogram countries.

### Table 1: Selected estimated coefficients and statistics (as reported)
- IMF program coefficients (selected):
  - Education spending (% of GDP), Fixed effect: 0.26 (standard error 0.133).
  - Education spending (% of GDP), System GMM: 0.22** (standard error 0.101).
  - Health spending (% of GDP), Fixed effect: 0.17** (standard error 0.070).
  - Health spending (% of GDP), System GMM: 0.27*** (standard error 0.094).
  - Education spending (% of total spending), System GMM: 1.15*** (standard error 0.389).
  - Health spending (% of total spending), System GMM: 0.46** (standard error 0.182).
- Lagged dependent variable coefficients (persistence of spending):
  - Range reported across specifications: 0.61*** to 0.85*** with standard errors shown (e.g., 0.85*** (0.029)).
- Other reported coefficients and statistics (examples from table):
  - Second lagged dependent variable: 0.18*** (0.051).
  - Government balance (various specifications): coefficients such as 0.02** (0.010) and -0.05** (0.027).
  - Inverse Mills ratio: -0.17** (0.080) and -0.06* (0.036) in some specifications.
- Sample and fit:
  - Number of observations reported in columns: 580, 580, 687, 687, 580, 580, 687, 664 (by column).
  - Number of countries reported: 54, 54, 59, 59, 45, 45, 59, 59 (by column).
  - R-squared examples: 0.651 (column 1) and 0.498 (column 3) as reported.
  - P-value for Sargan test in system GMM columns: 1.00.
  - P-value for AR(2) test in system GMM columns: 0.42, 0.22, 0.56, 0.28 (by relevant columns).

### Long-term and dynamic effects (Table 2 and narrative)
- A consecutive period of IMF-supported program is estimated to increase social spending relative to the pre-program period (system GMM estimates):
  - Year 1:
    - Education spending: 0.22 (increase in percentage point of GDP).
    - Health spending: 0.27 (increase in percentage point of GDP).
  - Year 3:
    - Education spending: 0.57 (increase in percentage point of GDP).
    - Health spending: 0.69 (increase in percentage point of GDP).
  - Year 5:
    - Education spending: 0.82 (increase in percentage point of GDP).
    - Health spending: 0.98 (increase in percentage point of GDP).
- Peak and fade dynamics for a short program example:
  - For a country with two years of IMF-supported programs followed by program end:
    - Peak effect after two years: education up by 0.4 percentage point of GDP and health up by 0.5 percentage point of GDP relative to the pre-program year.
    - Effect then declines gradually to about 0.1 percentage point of GDP in about 10 years for both education and health outlays.
- The effects on spending-to-GDP ratios eventually diminish and return to zero after a program ends.

### Mechanisms and cross-sample differences
- Potential channels through which IMF-supported programs affect spending include:
  - Catalyzing foreign financing and grants.
  - Raising revenues.
  - Changing the composition of spending.
- These channels appear more powerful in low‑income countries than in other developing countries.
- Median combined education and health spending:
  - 5 percent of GDP in the low‑income sample.
  - 7 percent of GDP for other program countries.
- Median annual increase in tax revenue to GDP ratios:
  - about 0.22 percentage point of GDP per year in low‑income program countries.
  - 0.06 percentage point per year in other program countries.

### Methodology summary
- Data: cross‑country panel of annual data for 1985–2009 covering 140 countries; low‑income country sample defined as countries eligible for concessional IMF lending.
- Dependent variables: public spending on education or health measured as percent of GDP or percent of total spending.
- Key explanatory variable: IMF-supported program dummy (equals 1 if country has a program in year t, 0 otherwise). Program start year defined as year of program approval (if approved in second half of year, starting year is following year); end year is year program expired.
- Control variables: lagged government balance (percent of GDP), share of population younger than age 15 (for education) and older than 65 (for health), log of real GDP per capita, urbanization index, trade openness.
- Estimation methods:
  - Fixed effects.
  - System GMM (preferred): addresses endogeneity and provides unbiased estimates including lagged dependent variable; income levels and government balance treated as endogenous; instruments include international reserves in months of imports (lagged), log bilateral exchange rate to U.S. dollar (lagged), and an index of exchange rate classification (lagged).
- Selection bias addressed via instrumentation in system GMM and use of inverse Mills ratio from probit regressions in fixed effects estimations.
- Probit regression results used to generate inverse Mills ratio (examples):
  - IMF program (probit) coefficient on IMF program (lagged) 1.75*** (0.12) for Education equation; 1.87*** (0.12) for Health.
  - Exchange rate classification (lagged) coefficients: 0.04** (0.02) for Education and 0.03* (0.02) for Health.
  - Number of observations in probit regressions: 608 (Education), 703 (Health).
  - R-squared for these probit models: 0.31 (Education), 0.34 (Health).

### Conclusions (author summary)
- Education and health spending has risen during IMF-supported programs at a faster pace than in developing countries as a whole.
- Controlling for determinants and macroeconomic conditions, IMF-supported programs have a positive and significant effect on social spending in low‑income countries.
- Over a five-year period in low‑income program countries:
  - education spending as a share of GDP increases by about 0.8 percentage point of GDP.
  - health spending as a share of GDP increases by about 1 percentage point of GDP.
- For other developing countries, the estimated effect is neutral.
- IMF-supported programs are also associated with increases in the share of government spending allocated to education and health.

*Source: _sdn1115 - 0.14 percentage point of GDP in nonprogram countries.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/sdn/2011/_sdn1115.pdf_