## 1. Numerical Targets in SSA HIPC Countries (2000–06 )

## Source details

**Canonical URL:** [1. Numerical Targets in SSA HIPC Countries (2000–06 )](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08217.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08217.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08217.pdf.json)

---

### I. Context, objectives, and policy issues
- Donor aid to SSA estimated around US$60 billion; World Bank and IMF debt forgiveness for HIPC intended to free resources for poverty alleviation and MDGs.
- Completion point requirements included demonstration of macroeconomic policy and management capacity, specific improvements in public financial management (PFM) systems, and a poverty alleviation strategy.
- Key policy concerns with scaled-up aid and debt relief:
  - Manage macroeconomic impact of substantial aid increases to avoid overheating and exchange rate appreciation that would impair competitiveness.
  - Increase aid absorptive capacity.
  - Sustain increased poverty-reducing spending over the medium term.
  - Ensure efficient and transparent use of funds and limit losses from inefficiency and corruption.

### II. Definition and core elements of PFM (expenditures only)
- PFM defined as procedures for management of public monies through the budget process: formulation, execution, reporting, and analysis (Potter and Diamond, 1999).
- Core components: budget preparation, legislative approval, execution, custody of public money, withdrawal procedures, authority to commit expenditures, payments, audit/oversight.
- Single treasury account, standardized budget presentation, budget classification systems, and accounting are central features.
- Common SSA accounting weaknesses: single-entry cash basis accounting; failure to account for direct/indirect liabilities; lags in data collection producing unreliable in-year data.

### III. Specific PFM weaknesses and fiscal dynamics in SSA
- Two systematic weaknesses:
  - Overly optimistic economic growth and revenue projections (including grants), producing higher aggregate expenditure envelopes and systematic revenue overestimation.
  - Failure to account for real spending pressures during fiscal framework preparation, enabling systematic larger deficits where exceeding budget deficits carries no sanctions.
- Aid-specific distortions:
  - Budgets often based on aid commitments rather than actual flows; projects sometimes initiated anticipating commitments that may not materialize.
  - Donor funding often targeted (capital, social sectors) and may require counterpart funds, influencing budget composition.
- In-year management and arrears:
  - Ad hoc responses to emergent pressures (HIV/AIDS, malaria, weather, military/police needs).
  - Ministries with political priority funded first; others may incur unauthorized borrowings and supplier arrears.
- Use of SOEs and contingent liabilities undermines budget coverage and accountability.
- Political-economy constraints limit structural reallocation and containment of "power ministry" spending.

### IV. HIPC PFM assessments (2001 and 2004) — measurement and data caveats
- Two assessments covering 15 PFM indicators on a three-point scale with minimum benchmarks; objective to track poverty-related spending and strengthen PFM.
- Indicators aggregated into:
  - PFM overall score (transformation: “3” for “A”; “2” for “B”; “1” for “C”; summed and normalized).
  - PFM benchmark overall (number of benchmarks met, out of 15).
  - Sub-scores: budget formulation (indicators 1–7); budget execution and reporting (indicators 8–15); rules score (8 indicators); procedures score (7 indicators).
- Data limitations and caveats:
  - Measurement of PFM quality is difficult; subjective assessments converted to numerical scores.
  - PEFA and CFAA exist but were not widely available or numerically comparable for this sample.
  - Empirical results should be interpreted cautiously.

### V. Fiscal rules and numerical targets in SSA HIPC countries (survey)
- Survey of 22 SSA HIPC countries for presence of:
  - (i) any numerical fiscal targets in law or as regional group member,
  - (ii) legal limit on the deficit or required budget balance,
  - (iii) ceiling on aggregate spending.
- Selected findings:
  - Eight countries (Benin, Burkina Faso, Cameroon, Chad, Guinea-Bissau, Mali, Niger, and Senegal) had numerical constraints as part of WAEMU or CEMAC.
  - Uganda adopted a balanced budget rule in 2000.
  - Tanzania’s law on government guarantees and loans (2004) prescribes numerical targets and procedures.
  - Many countries had no rule-based constraints but were under IMF programs providing numerical targets in practice.
- WAEMU convergence targets include: Budget balance (as percent of GDP, criteria: >=0); Arrears payment (in billion CFA, criteria: =0); Wages and salaries (as percent of tax revenues, criteria: <=35%); Internal funding of Capital expenditure (as percent of tax revenues, criteria: >=20); Total tax revenue (as percent of GDP, criteria: >=17%).
- CEMAC convergence targets include: Balance Budget (as% of GDP, criteria: = 0); Balance Budget structural (as% of GDP, criterion: = 0); Public debt (as% of GDP; criterion: = 70%); Arrears int. and ext. (in billions; criterion: = 0, late payment = 120 days); Primary budget balance (as% of GDP, criteria: >0).

### VI. Box 1 — Changes in HIPC indicators met (2001 → 2004)
- 2001 baseline weaknesses (selected exact shares reported):
  - 88 percent could not “produce timely, functionally based expenditure reports from core accounting data.”
  - 71 percent do not close their accounts within two months of the end of the year.
  - 92 percent did not provide budget data consistent with GFS definition of general government.
  - 83 percent did not produce audited accounts within 12 months of the close of the year.
  - Budget coverage inadequate in nearly 90 percent.
  - None met all benchmarks on fiscal reporting.
- Changes 2001→2004 (selected):
  - Realism indicators (comparison of budget outturn with initial budget; arrears level; use of off-budget funds) declined in 2004 relative to 2001.
  - Timeliness of functionally based expenditure reports improved from 14 percent to 31 percent meeting the benchmark.
  - Timeliness of monthly internal expenditure reports declined.

### VII. Country-level PFM benchmark and numerical scores (selected exact values)
- Selected entries preserving exact reported numbers (Met out of 15; budget formulation met; budget execution met; numerical score 2001; numerical score 2004):
  - Benin: met 2001 = 8; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 4; budget execution met 2004 = 4; numerical score 2001 = 32; numerical score 2004 = 34
  - Burkina Faso: met 2001 = 8; met 2004 = 9; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 5; budget execution met 2004 = 4; numerical score 2001 = 33; numerical score 2004 = 35
  - Cameroon: met 2001 = 4; met 2004 = 7; budget formulation met 2001 = 1; budget formulation met 2004 = 3; budget execution met 2001 = 3; budget execution met 2004 = 4; numerical score 2001 = 26; numerical score 2004 = 31
  - Ghana: met 2001 = 1; met 2004 = 7; budget formulation met 2001 = 0; budget formulation met 2004 = 1; budget execution met 2001 = 3; budget execution met 2004 = 4; numerical score 2001 = 19; numerical score 2004 = 32
  - Mali: met 2001 = 8; met 2004 = 11; budget formulation met 2001 = 4; budget formulation met 2004 = 4; budget execution met 2001 = 6; budget execution met 2004 = 5; numerical score 2001 = 32; numerical score 2004 = 37
  - Tanzania: met 2001 = 8; met 2004 = 11; budget formulation met 2001 = 4; budget formulation met 2004 = 4; budget execution met 2001 = 6; budget execution met 2004 = 5; numerical score 2001 = 33; numerical score 2004 = 36
  - Uganda: met 2001 = 9; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 4; budget execution met 2001 = 3; budget execution met 2004 = 5; numerical score 2001 = 35; numerical score 2004 = 32
  - Zambia: met 2001 = 3; met 2004 = 3; budget formulation met 2001 = 1; budget formulation met 2004 = 2; budget execution met 2001 = 0; budget execution met 2004 = 3; numerical score 2001 = 23; numerical score 2004 = 24
- Note: some table entries in source are incomplete or flagged (e.g., Guinea-Bissau numerical score 2004 = 17).

### VIII. CPIA relationships and Spearman correlations with PFM
- CPIA and PFM: strong positive associations reported between World Bank CPIA scores and HIPC PFM indicators.
  - PFM overall score correlation with CPIA all: 0.789 (p-value: 0.000); observations: 21.
  - PFM budget formulation score correlation with CPIA all: 0.503 (p-value: 0.020).
  - PFM budget execution and reporting score correlation with CPIA all: 0.786 (p-value: 0.000).
- Spearman correlations (combined sample, 39 obs) between PFM overall score / benchmarks met and fiscal variables (summary exact values):
  - Overall balance: 0.492 ; 0.514 (p-values: 0.002 ; 0.001)
  - Overall balance excluding grants: 0.038 ; 0.018 (p-values: 0.820 ; 0.916)
  - Primary balance: 0.213 ; 0.232 (p-values: 0.194 ; 0.156)
  - Primary balance excluding grants: -0.015 ; -0.009 (p-values: 0.928 ; 0.956)
  - Interest payments: -0.565 ; -0.619 (p-values: 0.000 ; 0.000)
  - Primary expenditure: -0.047 ; -0.215 (p-values: 0.776 ; 0.188)
  - Public and publicly guaranteed external debt: -0.615 ; -0.649 (p-values: 0.000 ; 0.000)
  - Gross central government debt: -0.576 ; -0.630 (p-values: 0.000 ; 0.000)
- Year-split correlations show stronger significance in 2004 than 2001 for several variables (e.g., Overall balance 2001 = 0.390, p-value 0.109; 2004 = 0.579, p-value 0.006).

### IX. Graphical R-squared (PFM overall score vs fiscal variables, combined sample)
- Reported R^2 values:
  - Overall Balance: R^2 = 0.27
  - Primary Balance: R^2 = 0.02
  - Overall Balance excl. grants: R^2 = 0.16
  - Primary Balance excl. grants: R^2 = 0.04
  - Primary Expenditure: R^2 = 0.06
  - Interest Payments: R^2 = 0.32
  - PPG External Debt: R^2 = 0.40
  - Gross Central Government Debt: R^2 = 0.15

### X. Which PFM components matter? (Spearman summary, 39 obs)
- Budget formulation vs execution:
  - Overall balance correlations: 0.492 (overall); 0.394 (formulation); 0.465 (execution) — p-values: 0.002; 0.013; 0.003.
  - Interest payments correlations: -0.565 ; -0.438 ; -0.536 — p-values: 0.000; 0.005; 0.000.
  - PPG external debt correlations: -0.615 ; -0.426 ; -0.610 — p-values: 0.000; 0.007; 0.000.
- Rules vs procedures:
  - Overall balance correlations: 0.492 ; 0.442 ; 0.463 — p-values: 0.002 ; 0.005 ; 0.003.
  - Debt-related correlations particularly strong with rules score.

### XI. OLS regressions — selected coefficients and interpretation (39 observations)
- PFM overall score (partial coefficients and robust t statistics):
  - Overall Balance: 0.24 (3.00)***
  - Primary Balance: 0.14 (1.83)*
  - Overall Balance Excluding Grants: 0.39 (2.59)**
  - Primary Balance Excluding Grants: 0.29 (2.01)*
  - Interest Payments: -0.10 (3.55)***
  - Primary Expenditure: -0.37 (2.15)**
  - PPG External Debt: -2.93 (3.00)***
  - Gross Central Government Debt: -3.89 (2.94)***
- Other notable regressors:
  - Growth: PPG External Debt = -2.24 (3.15)***; Gross Central Government Debt = -2.30 (1.80)*
  - Corruption: Overall Balance = -2.67 (1.79)*
  - Resource rich: Primary Expenditure = -6.99 (1.94)*
  - HIPC completion point: PPG External Debt = -39.78 (2.68)**
  - French legal origin: Interest Payments = -2.26 (3.28)***; PPG External Debt = -25.50 (2.42)**
- Model fit and sample:
  - Observations = 39 for all regressions.
  - R-squared values (dependent variables order): 0.53 ; 0.34 ; 0.54 ; 0.43 ; 0.73 ; 0.47 ; 0.76 ; 0.56.

### XII. Mechanisms, limitations, and interpretation
- Mechanisms suggested:
  - Better PFM yields timely and reliable fiscal information, improving budget management and control of contingent liabilities.
  - Stronger PFM reduces incidence of arrears, undisclosed guarantees, and off-budget liabilities.
  - Hierarchical budget systems (e.g., francophone countries with stronger Finance Minister role) associated with better fiscal discipline.
- Limitations highlighted:
  - Small sample size and only two years of PFM assessment (2001, 2004) limit time-series inference and fixed-effects power.
  - Subjective scoring converted to numerical measures may introduce measurement error.
  - Sample selection (SSA HIPC countries) may introduce selection bias; endogeneity concerns noted.
  - Robustness checks and endogeneity corrections explored but constrained by data.

### XIII. Marginal effects and quantified implications (ANNEX IV, Table 11)
- Marginal effects calculated for an illustrative improvement (change from 1 to 2 in a single PFM indicator; multiplier (b) = 2.22):
  - Overall balance: (a) 0.24; (b) 2.22; (c) 0.53 (change in fiscal outcome, percent of GDP)
  - Primary balance: (a) 0.14; (b) 2.22; (c) 0.31
  - Overall balance excluding grants: (a) 0.39; (b) 2.22; (c) 0.87
  - Primary balance excluding grants: (a) 0.29; (b) 2.22; (c) 0.64
  - Interest payments: (a) -0.10; (b) 2.22; (c) -0.22
  - Primary expenditure: (a) -0.37; (b) 2.22; (c) -0.82
  - PPG external debt: (a) -2.93; (b) 2.22; (c) -6.50
  - Gross central government debt: (a) -3.89; (b) 2.22; (c) -8.64
- Interpretation:
  - Improvement from 1 to 2 in indicator 2 (reduction in extrabudgetary funding) associated with improvement in overall fiscal balance ≈ 0.53 percent of GDP.
  - Associated reductions: PPG external debt ≈ -6.50 percent of GDP; gross central government debt ≈ -8.64 percent of GDP.
- Caveats: marginal effects subject to measurement error, model specification, and limited sample.

### XIV. Robustness checks, panel estimations, and endogeneity correction
- Panel estimations (observations = 39; 21 countries) for Overall Balance:
  - Between: PFM overall score = 0.31 (t = (5.72)***)
  - Fixed effects: PFM overall score = -0.13 (t = (1.03))
  - Random effects: PFM overall score = 0.23 (t = (2.61)***)
  - Random effects maximum likelihood: PFM overall score = 0.24 (t = (4.84)***)
- Panel estimations for External Debt show consistent negative and significant PFM coefficients across estimators:
  - Between = -3.23 (t = (3.26)***)
  - Fixed effects = -2.03 (t = (2.99)**)
  - Random effects = -3.09 (t = (3.14)***)
  - Random effects maximum likelihood = -3.33 (t = (5.72)***)
- Endogeneity correction approach:
  - Instrumented PFM using lags of growth; percent population under 15; corruption; HIPC decision point; HIPC completion point; fitted PFM used to expand sample to 126 observations.
  - Selected fitted-PFM regression results (Table 14, fitted PFM; Observations = 126 unless noted):
    - PFM overall score (fitted): Overall Balance = 0.36 (t = (1.35)); Overall Balance Excluding Grants = 0.87 (t = (3.71)***); Interest Payments = -0.26 (t = (3.75)***); PPG External Debt = -6.67 (t = (3.95)***); Gross Central Government Debt = -17.49 (t = (3.44)***).
  - Conclusion: endogeneity-corrected results broadly similar though some loss of significance; anomalies noted (e.g., negative growth–balance relation in some specifications).

### XV. Policy-relevant conclusions and recommendations
- Empirical evidence across cross-sectional, panel, and endogeneity-corrected specifications indicates:
  - Improving PFM quality is associated with better overall fiscal balances (including and excluding grants).
  - Improving PFM quality is associated with lower interest payments.
  - Improving PFM quality is associated with lower public and publicly guaranteed external debt and lower gross central government debt.
  - Improving PFM quality is associated with lower primary expenditure (consistent with improved balance outcomes).
- Institutional-design implication:
  - A more hierarchical budget/PFM system (as in Francophone Africa) appears to yield better fiscal discipline; policy consideration to strengthen Finance Minister’s fiscal role where PFM is more collegial.
- Data and research recommendation:
  - Develop a better and longer quantitative PFM dataset; design assessment instruments in a comprehensive framework rather than narrow, project-specific focuses.
- Operational implication for aid scaling:
  - Strengthening PFM can enhance aid absorptive capacity, transparency, and preservation of fiscal stability amid scaled-up aid and debt relief.

*Source: _wp08217 - 1. Numerical Targets in SSA HIPC Countries (2000–06 ).*

### 1. Numerical Targets in SSA HIPC Countries (2000–06 ) ......................................................15

### _wp08217 - 1. Numerical Targets in SSA HIPC Countries (2000–06 ) ......................................................15

### Major sections (by heading and page index)
- 1. Numerical Targets in SSA HIPC Countries (2000–06 ) ......................................................15
- 2. HIPC PFM Assessment........................................................................................................17
- 3. Macro and Fiscal Policy and PFM.......................................................................................18
- 4. Spearman Correlation Coefficients......................................................................................19
- 5. HIPC SSA 2001 and 2004 ...................................................................................................20
- 6. HIPC SSA Correlations Between Fiscal Outcomes and PFM Variables ............................21
- 7. HIPC SSA Correlations Between Fiscal Outcome and PFM Overall Score .......................22
- 8. HIPC SSA Correlations Between Fiscal Outcomes and PFM Indicators of 
        Formulation and Execution..............................................................................................24
- 9. HIPC SSA Correlations Between ........................................................................................25
- 10. Fiscal Outcomes as Dependent Variable ...........................................................................28
- 11. Marginal Effects.................................................................................................................37
- 12. Overall Balance as Dependent Variable ............................................................................39
- 13. External Debt as Dependent Variable................................................................................40
- 14. Fiscal Outcome as Dependent Variable .............................................................................41
- 15. Overall Balance as Dependent Variable ............................................................................42
- 16. External Debt as Dependent Variable................................................................................43

### Figures
- 1. Plots of the Fiscal Variables and the PFM Overall Score....................................................23

### Boxes
- 1. Changes in HIPC Indicators Met from the 2001 to 2004 Assessment ................................17

### Annexes
- I.  Country Groups ...................................................................................................................31
- II. HIPC Assessment Indicators ...............................................................................................32
- III. Variables in the Data Set....................................................................................................33
- IV. Results with Endogeneity Correction ................................................................................37

*Source: _wp08217 - 1. Numerical Targets in SSA HIPC Countries (2000–06 ) (PDF chapter/section).*

### References..............................................................................................................

### _wp08217 - References

### I. INTRODUCTION — context and objectives
- Recent progress in economic growth in sub-Saharan Africa (SSA) noted, but SSA "still lags significantly behind other regions in economic development and socioeconomic indicators."
- Donor aid estimated to be around US$60 billion from bilateral and multilateral sources expected to flow to SSA to help alleviate poverty and achieve the Millennium Development Goals (MDGs).
- World Bank and IMF debt forgiveness for Heavily-Indebted Poor Countries (HIPC) aimed to free resources for poverty alleviation.
- Key concerns around scaled-up aid and debt relief:
  - How to manage macroeconomic impact of a substantial increase of aid flows to avoid overheating and exchange rate appreciation that would impair competitiveness.
  - How to increase aid absorptive capacity.
  - How to sustain increased poverty-reducing spending over the medium term.
  - How to ensure use of funds efficiently and transparently and limit losses from inefficiency and corruption.
- "Completion point" requirements for debt relief included: demonstration of macroeconomic policy and management capacity, specific improvements in public financial management (PFM) systems (in many cases), and preparation of a strategy for poverty alleviation.
- PFM seen as critical for allocation and use of public resources through the budget and fiscal policy; a well-functioning PFM system expected to improve use of aid, overall budget performance, macroeconomic stability, and governance by protecting public resources against expropriation and corruption.
- Timeline and assessments:
  - Two PFM assessments for HIPC countries conducted between 2001 and 2004.
  - Action plans and technical assistance developed between assessments to meet PFM benchmarks.
  - HIPC PFM assessments covered 15 PFM indicators on a three-point scale with minimum benchmarks; first assessment in 2001, second in 2004.
  - Determination for eligibility for debt relief followed the second assessment.

### II. WHAT IS PFM — definition and core elements
- Definition used: PFM refers to procedures, established by law or regulation, for management of public monies through the budget process, including formulation, execution, reporting, and analysis (Potter and Diamond, 1999). In this paper, PFM is used to refer to expenditures only.
- Core components and functions:
  - Budget preparation, legislative approval, execution, custody of public money, withdrawal procedures, authority to commit expenditures, payments, and audit/oversight.
  - Single treasury account (consolidated government account) as a central feature; restrictions on individual/agencies operating government bank accounts without Ministry of Finance approval.
  - Checks and balances governing payment and commitment of government resources.
  - Standardized budget presentation formats, budget classification systems, and forms of accounts for transparency.
  - Accounting as a core element: measurement of implementation and impact of the budget and economic policy.
- Common weaknesses in SSA accounting systems:
  - Predominantly single-entry cash basis accounting.
  - Failure to account fully or accurately for direct and indirect liabilities.
  - Lags in data collection and recording leading to unreliable and untimely in-year data, limiting use in annual budgeting.

### III. SPECIFIC PFM ISSUES IN SSA — fiscal dynamics and practical problems
- Two main kinds of fiscal weaknesses leading to systematically worse outcomes than legislated budgets:
  - Overly optimistic economic growth and revenue projections (including grants), producing higher aggregate expenditure envelopes and systematic revenue overestimation. Only Uganda and to some extent Tanzania noted as exceptions regarding hard legal constraints on outcomes.
  - Failure to account for real spending pressures during fiscal framework preparation, enabling systematic larger deficits where exceeding budget deficits carries no sanctions.
- Aid-specific fiscal distortions:
  - Budgets often formulated on the basis of aid commitments rather than actual aid flows; projects sometimes initiated from domestic budgets in anticipation of commitments that may not materialize, causing wasteful spending and budget distortions.
  - Donors often fund specific parts like capital spending and social sectors but may require counterpart funds, influencing budget composition.
- In-year management and arrears:
  - In-year budget management characterized by ad hoc responses to emergent pressures (e.g., HIV/AIDS, malaria, weather, military/police needs).
  - Limited borrowing capacity; ministries with priority (military, police, presidency, donor-favored) often funded first; other ministries may resort to unauthorized market borrowings leading to arrears, frequently to suppliers.
- Use of public sector entities:
  - Cash-strapped governments may shift spending through state-owned enterprises (SOEs) that borrow or provide subsidized goods, incurring direct and contingent liabilities often outside the budget and accounts.
- Political economy constraints on reallocation:
  - Options to fund PRSPs: (1) structural reallocation of spending (takes medium term to avoid social problems); (2) contain "power ministry" spending (politically difficult).

### IV. LITERATURE REVIEW — institutions, PFM, corruption, and fiscal effects
- Institutions and growth:
  - Institutional weaknesses in SSA identified as a major cause of slow growth (North, 1990; Acemoglu and Johnson, 2003; Rodrik, Subramaniam, and Trebbi, 2004; Acemoglu, Johnson, and Robinson, 2004).
  - Protection of property rights and protection against expropriation central to institutional quality.
  - Weak governance institutions (e.g., civil service) limit effective use of external aid and risk sustaining a middle class (Brautigam and Knack, 2004; Birdsall, 2007).
- PFM and corruption:
  - Well-functioning PFM aligns with institutional development and protects budgetary allocations for basic public goods and services.
  - PFM establishes rules for "who-receives-what"; violations of procedures and weak enforcement lead to corruption.
  - Excessive regulation and discretionary application of rules can also foster corruption (Mauro, 1996).
  - Enforcement mechanisms are essential; PFM without enforcement can be subverted.
- PFM and fiscal outcomes (broader literature):
  - Budget processes (procedural rules, transparency rules, numerical targets) influence fiscal deficits and borrowing (Alesina and Perotti, 1996; Alesina et al., 1996; Eichengreen et al., 1999; Alesina et al., 1999).
  - Hierarchical procedures (strong minister of finance role) associated with better fiscal discipline versus collegial procedures which offer flexibility but less discipline.
  - Numerical targets can encourage creative accounting; procedural reforms recommended include increasing minister of finance authority, limiting off-budget spending and earmarking, and preventing line agencies from committing without budget.
  - Empirical findings: in Latin America (1989–1993), countries with best budget processes had fiscal surpluses of 1.8 percent of GDP while the three weakest had average deficits of 2.2 percent.
  - EU and U.S. state studies show strong relationships between budgetary processes and deficits (von Hagen and Harden, 1995; Poterba, 1994).
- Practical PFM guidance and SSA-specific work:
  - Practitioner literature (Potter and Diamond, 1999; DFID, 2001) provides step-by-step guidance on budget stages and PFM roles.
  - SSA-focused empirical and descriptive work from technical assistance, Public Expenditure Reviews, and working papers (Lienert and Sarraf, 2001; Moussa, 2004; Kutessa et al., 2002; Folscher, 2006).
  - Studies tied to HIPC debt relief highlight importance of strengthening PFM to absorb debt relief and scaled-up aid (Allen et al., 2007; Heller, 2005; Claessens et al., 2007).
  - Fiscal transparency research: Alt and Lassen (2006) find budget transparency leads to lower debt and deficit in OECD; Hameed (2005) finds fiscal transparency strongly correlated with better fiscal discipline.

### V. FISCAL TARGETS IN SSA HIPC COUNTRIES (2000–06) — survey results
- Survey of 22 SSA HIPC countries to identify presence of:
  - (i) any numerical fiscal targets in law or as member of a regional group,
  - (ii) a legal limit on the deficit or required budget balance,
  - (iii) a ceiling on aggregate spending.
- Findings summarized (textual summary of Table 1):
  - Eight countries (Benin, Burkina Faso, Cameroon, Chad, Guinea-Bissau, Mali, Niger, and Senegal) had some kind of numerical constraint as part of regional organizations (WAEMU or CEMAC), even if not always observed.
  - Uganda adopted a balanced budget rule in 2000.
  - Tanzania’s law on government guarantees and loans (2004) prescribes numerical targets and procedures.
  - Others appeared to have no rule-based constraints, though many were under IMF programs (e.g., PRGF or PRSP) which provided numerical targets in practice.
- Notes on regional convergence targets (as cited in Table 1):
  - WAEMU main convergence targets include: Budget balance (in percent of GDP, criteria: >=0); Arrears payment (in billion CFA, criteria: =0); Wages and salaries (in percent of tax revenues, criteria: <=35%); Internal funding of Capital expenditure (in percent of tax revenues, criteria: >=20); Total tax revenue (in percent of GDP, criteria: >=17%).
  - CEMAC main convergence targets include: Balance Budget (as% of GDP, criteria: = 0); Balance Budget structural (in percent of GDP, criterion: = 0); Public debt (in percent of GDP; criterion: = 70%); Arrears int. and ext. (in billions; criterion: = 0, late payment = 120 days); Primary budget balance (in percent of GDP, criteria: >0).

### VI. MEASUREMENT OF PFM QUALITY — HIPC indicators and data limitations
- HIPC PFM data set design:
  - Intended to measure quality and strength of PFM systems through 15 variables across budget formulation, execution, and reporting (Annex II).
  - Indicators use a three-point scale with benchmarks; primary objective to ensure capacity to identify and track poverty-related spending so funds released by debt relief could be assessed.
  - Scope broadened to strengthen PFM fundamentally, not only to track spending.
- Data and interpretation caveats:
  - Measurement of PFM weaknesses is difficult in SSA given underdeveloped economic institutions.
  - Empirical results from HIPC PFM indicators should be interpreted cautiously.
  - Other instruments exist (PEFAs, CFAAs), but PEFA assessments have been public for only a few countries and the instrument is evolving; CFAAs are descriptive without numerical ratings and are not suited for data assessment.
- Summary statistics mentioned:
  - Table 2 (referenced) presents PFM overall score and "benchmarks met" out of 15 for SSA HIPC countries in 2001 and 2004.
  - PFM overall score converted into a numerical score on a three-point scale, "3 being highest"; relatively stable PFM indicator scores noted between 2001 and 2004.

*Source: IMF staff compilations, content as presented in the _wp08217 PDF.*

### 2004. Box 1 summarizes these results.

### Box 1. Changes in HIPC Indicators Met from the 2001 to 2004 Assessment

### Key findings: 2001 assessment baseline
- 88 percent of these countries could not “produce timely, functionally based expenditure reports from core accounting data.”
- 71 percent of these countries do not close their accounts within two months of the end of the year.
- 92 percent did not provide budget data consistent with the GFS definition of general government.
- 83 percent of these countries did not produce audited accounts within 12-months of the close of the year.
- Budget coverage was inadequate in most (nearly 90 percent) of the countries in the sample.
- None of the countries met all the benchmarks on fiscal reporting.

### Changes in PFM indicators between 2001 and 2004
- Three indicators measure realism of the budget: comparison of budget outturn with the initial budget, level of payment arrears, and use of off-budget funds.
  - Performance in all three of these realism indicators declined in 2004 compared with 2001.
  - Suggested explanation: difficulty in sustaining PFM reforms in these countries.
- Timeliness of functionally based expenditure reports from core accounting data:
  - Improved from 14 percent to 31 percent of countries meeting this benchmark between 2001 and 2004.
- Timeliness of monthly internal expenditure reports:
  - Declined between 2001 and 2004.

### HIPC PFM assessment: country-level met indicators and scores (selected exact values reported)
- Table structure in source: columns for "Met out of 15", 2001 and 2004, and numerical scores for budget formulation, budget execution, and overall 2001 and 2004.
- Selected country entries (preserve exact numerical entries as presented):
  - Benin: met 2001 = 8; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 4; budget execution met 2004 = 4; numerical score 2001 = 32; numerical score 2004 = 34
  - Burkina Faso: met 2001 = 8; met 2004 = 9; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 5; budget execution met 2004 = 4; numerical score 2001 = 33; numerical score 2004 = 35
  - Cameroon: met 2001 = 4; met 2004 = 7; budget formulation met 2001 = 1; budget formulation met 2004 = 3; budget execution met 2001 = 3; budget execution met 2004 = 4; numerical score 2001 = 26; numerical score 2004 = 31
  - Chad: met 2001 = 8; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 3; budget execution met 2004 = 4; numerical score 2001 = 28; numerical score 2004 = 29
  - Congo, Dem. Rep.: entry includes fragments indicating "3" in places and "24" as a score in the table row as presented.
  - Ethiopia: met 2001 = 6; met 2004 = 7; budget formulation met 2001 = 4; budget formulation met 2004 = 2; budget execution met 2001 = 5; budget execution met 2004 = 2; numerical score 2001 = 31; numerical score 2004 = 32
  - Gambia: met 2001 = 5; met 2004 = 3; budget formulation met 2001 = 4; budget formulation met 2004 = 1; budget execution met 2001 = 3; budget execution met 2004 = 0; numerical score 2001 = 27; numerical score 2004 = 25
  - Ghana: met 2001 = 1; met 2004 = 7; budget formulation met 2001 = 0; budget formulation met 2004 = 1; budget execution met 2001 = 3; budget execution met 2004 = 4; numerical score 2001 = 19; numerical score 2004 = 32
  - Guinea: met 2001 = 5; met 2004 = 5; budget formulation met 2001 = 3; budget formulation met 2004 = 2; budget execution met 2001 = 2; budget execution met 2004 = 3; numerical score 2001 = 26; numerical score 2004 = 28
  - Guinea-Bissau: met 2001 = ...; met 2004 = 0; budget formulation met 2001 = ...; budget formulation met 2004 = ...; budget execution met 2001 = 0; budget execution met 2004 = 0; numerical score 2001 = ...; numerical score 2004 = 17
  - Madagascar: met 2001 = 7; met 2004 = 4; budget formulation met 2001 = 5; budget formulation met 2004 = 2; budget execution met 2001 = 3; budget execution met 2004 = 1; numerical score 2001 = 29; numerical score 2004 = 30
  - Malawi: met 2001 = 7; met 2004 = 5; budget formulation met 2001 = 4; budget formulation met 2004 = 3; budget execution met 2001 = 3; budget execution met 2004 = 2; numerical score 2001 = 33; numerical score 2004 = 29
  - Mali: met 2001 = 8; met 2004 = 11; budget formulation met 2001 = 4; budget formulation met 2004 = 4; budget execution met 2001 = 6; budget execution met 2004 = 5; numerical score 2001 = 32; numerical score 2004 = 37
  - Mozambique: met 2001 = 5; met 2004 = 4; budget formulation met 2001 = 3; budget formulation met 2004 = 2; budget execution met 2001 = 3; budget execution met 2004 = 1; numerical score 2001 = 28; numerical score 2004 = 26
  - Niger: met 2001 = 3; met 2004 = 5; budget formulation met 2001 = 2; budget formulation met 2004 = 1; budget execution met 2001 = 3; budget execution met 2004 = 2; numerical score 2001 = 24; numerical score 2004 = 30
  - Rwanda: met 2001 = 8; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 3; budget execution met 2001 = 4; budget execution met 2004 = 4; numerical score 2001 = 33; numerical score 2004 = 32
  - São Tomé and Príncipe: met 2001 = 4; met 2004 = 4; budget formulation met 2001 = 1; budget formulation met 2004 = 2; budget execution met 2001 = 3; budget execution met 2004 = 2; numerical score 2001 = 26; numerical score 2004 = 25
  - Senegal: met 2001 = 4; met 2004 = 7; budget formulation met 2001 = 4; budget formulation met 2004 = 0; budget execution met 2001 = 4; budget execution met 2004 = 3; numerical score 2001 = 28; numerical score 2004 = 32
  - Sierra Leone: met 2001 = ...; met 2004 = 7; budget formulation met 2001 = ...; budget formulation met 2004 = ...; budget execution met 2001 = 4; budget execution met 2004 = 3; numerical score 2001 = ...; numerical score 2004 = 28
  - Tanzania: met 2001 = 8; met 2004 = 11; budget formulation met 2001 = 4; budget formulation met 2004 = 4; budget execution met 2001 = 6; budget execution met 2004 = 5; numerical score 2001 = 33; numerical score 2004 = 36
  - Uganda: met 2001 = 9; met 2004 = 8; budget formulation met 2001 = 5; budget formulation met 2004 = 4; budget execution met 2001 = 3; budget execution met 2004 = 5; numerical score 2001 = 35; numerical score 2004 = 32
  - Zambia: met 2001 = 3; met 2004 = 3; budget formulation met 2001 = 1; budget formulation met 2004 = 2; budget execution met 2001 = 0; budget execution met 2004 = 3; numerical score 2001 = 23; numerical score 2004 = 24
- Sources for Table 2: IMF and World Bank HIPC Assessments 2001 and 2004; staff calculations.

### CPIA and relationship to PFM outcomes
- The World Bank’s CPIA measures quality of macroeconomic and fiscal policies, and budget and financial management.
- CPIA "rule of law" indicator: quality of the justice system (civil and criminal) — can be a proxy for enforcement of regulations on public finance.
- Institutional and structural reforms during 2000–2004 have longer gestation periods; impact on fiscal outcomes during this period would be minimal or hard to measure. CPIA data can to some extent proxy the impact of these reforms.
- Table 3 (selected entries) shows highest and lowest performers:
  - Highest performing (CPIA macro, fiscal, budget and financial management scores out of 1–6, 6 best; and PFM HIPC "Met, out of 15"):
    - Benin: Macro Score = 4.5; Fiscal Score = 4.0; Budget and Financial Management Score = 3.5; PFM HIPC Met, out of 15 = 8
    - Burkina-Faso: Macro Score = 4.5; Fiscal Score = 4.5; Budget and Financial Management Score = 3.5; PFM HIPC Met, out of 15 = 9
    - Mali: Macro Score = 4.5; Fiscal Score = 4.0; Budget and Financial Management Score = 3.5; PFM HIPC Met, out of 15 = 11
    - Uganda: Macro Score = 4.5; Fiscal Score = 4.5; Budget and Financial Management Score = 4.5; PFM HIPC Met, out of 15 = 8
    - Tanzania: Macro Score = 5.0; Fiscal Score = 4.5; Budget and Financial Management Score = 4.5; PFM HIPC Met, out of 15 = 11
  - Lowest performing:
    - Guinea-Bissau: Macro Score = 2.0; Fiscal Score = 2.5; Budget and Financial Management Score = 2.5; PFM HIPC Met, out of 15 = 0
    - Ethiopia: Macro Score = 3.0; Fiscal Score = 4.0; Budget and Financial Management Score = 4.0; PFM HIPC Met, out of 15 = 7
    - Guinea: Macro Score = 2.5; Fiscal Score = 3.0; Budget and Financial Management Score = 2.5; PFM HIPC Met, out of 15 = 5
    - Malawi: Macro Score = 3.5; Fiscal Score = 3.0; Budget and Financial Management Score = 3.0; PFM HIPC Met, out of 15 = 5
    - Gambia, The: Macro Score = 3.5; Fiscal Score = 3.0; Budget and Financial Management Score = 2.5; PFM HIPC Met, out of 15 = 3
- Source for Table 3: World Bank CPIA and IMF and World Bank HIPC assessments; Fund staff calculations.

### Correlations between CPIA and PFM (Spearman correlation coefficients)
- PFM overall score correlation with CPIA all: 0.789
  - p-value: 0.000
- PFM budget formulation score correlation with CPIA all: 0.503
  - p-value: 0.020
- PFM budget execution and reporting score correlation with CPIA all: 0.786
  - p-value: 0.000
- Number of observations: 21
- Sources for Table 4: International Monetary Fund and World Bank PFM indicators for 2004, and CPIA for 2005.

### Data summary and outlier treatment
- São Tomé and Príncipe identified as an outlier in several fiscal variables and excluded from subsequent statistical analysis.
  - Rationale: several fiscal indicators were more than two standard deviations from the mean; small country with fiscal variables during 2000–06 unduly influenced by oil sector issues as oil development started during this period.
- Table 5 (HIPC SSA 2001 and 2004, percent of GDP unless otherwise indicated) — selected summary statistics with São Tomé and Príncipe flagged as outlier:
  - PFM benchmarks met (number): Mean 6.1 (2001), 7.0 (2004); Median 2.5; Standard Deviation 11.0; Max 0.0; Min — (table indicates São Tomé and Príncipe as Outlier)
  - PFM overall score (value): Mean 65.1 (2001), 66.7 (2004); Median 9.9; Standard Deviation 82.2; Max 37.8; Min — (São Tomé and Príncipe Outlier)
  - Overall balance: Mean -4.1 (2001), -3.5 (2004); Median 3.2; Standard Deviation 2.6; Max -15.0; Min — Outlier
  - Overall balance excl. grant: Mean -9.8 (2001), -8.6 (2004); Median 5.9; Standard Deviation 2.2; Max -32.2; Min — Outlier
  - Primary balance: Mean -1.7 (2001), -1.9 (2004); Median 2.9; Standard Deviation 5.6; Max -9.4; Min — Outlier
  - Primary balance excl. grants: Mean -7.4 (2001), -6.8 (2004); Median 5.2; Standard Deviation 5.2; Max -26.1; Min — Outlier
  - Interest payments: Mean 2.4 (2001), 1.5 (2004); Median 2.3; Standard Deviation 9.9; Max 0.3; Min —
  - Primary expenditure: Mean 21.8 (2001), 20.7 (2004); Median 6.3; Standard Deviation 43.3; Max 12.7; Min — Outlier
  - PPG external debt: Mean 85.8 (2001), 70.0 (2004); Median 46.3; Standard Deviation 273.1; Max 35.8; Min — Outlier
  - Gross central gov. debt: Mean 117.4 (2001), 90.9 (2004); Median 107.3; Standard Deviation 647.8; Max 32.4; Min —
- Number of observations for Table 5: 39.
- Countries included in the sample lists for 2001 and 2004 are provided in the source table notes.

### Econometric analysis: purpose and approach
- Main purpose: provide econometric evidence that better PFM (measured by the number of benchmarks met or overall score) is associated with better fiscal outcomes, after controlling for relevant explanatory variables for fiscal outcomes.
- Empirical analysis structure:
  1. Sample summary, correlations, and graphs to suggest broad relationships among indicators.
  2. Regressions estimated using main variables suggested in the literature.
     - Ordinary least squares regressions to identify main relations.
     - Additional quantitative analyses and robustness checks are presented subsequently in the source.

*Source: IMF and World Bank HIPC Assessments 2001 and 2004; World Bank CPIA; Fund staff calculations.*

### Annex IV.

### Annex IV. The simple correlation of quality of PFM systems and fiscal outcomes

### Methodology and data
- Nonparametric Spearman rank correlation tests are used (Von Hagen (1992) approach) to measure correlation between ranks of PFM indicators and fiscal variables; p-values reported to assess significance.
- Sample: HIPC SSA countries, excluding São Tomé and Príncipe.
- Number of observations reported: 39 (combined 2001 and 2004); for year-split: 18 (2001) and 21 (2004).

### Spearman correlation results (summary of Table 6)
- Correlations between PFM overall score / PFM overall benchmarks met and fiscal variables:
  - Overall balance: 0.492 ; 0.514
    - p-value: 0.002 ; 0.001
  - Overall balance excluding grants: 0.038 ; 0.018
    - p-value: 0.820 ; 0.916
  - Primary balance: 0.213 ; 0.232
    - p-value: 0.194 ; 0.156
  - Primary balance excluding grants: -0.015 ; -0.009
    - p-value: 0.928 ; 0.956
  - Interest payments: -0.565 ; -0.619
    - p-value: 0.000 ; 0.000
  - Primary expenditure: -0.047 ; -0.215
    - p-value: 0.776 ; 0.188
  - Public and publicly guaranteed external debt: -0.615 ; -0.649
    - p-value: 0.000 ; 0.000
  - Gross central government debt: -0.576 ; -0.630
    - p-value: 0.000 ; 0.000
- Interpretation:
  - Positive and statistically significant correlation between PFM quality and overall balance (including grants).
  - Correlations for balances excluding grants and primary balances (excluding interest) are not significant, suggesting grants strongly influence the significant relation with overall balance.
  - Negative and significant correlations between PFM quality and interest payments, external debt, and gross central government debt.
  - No significant correlation between primary expenditure and PFM indicators.

### Year-split Spearman correlations (summary of Table 7)
- Spearman correlations of PFM overall score with fiscal variables, by year:
  - Overall balance: 2001 = 0.390 (p-value 0.109); 2004 = 0.579 (p-value 0.006)
  - Overall balance excluding grants: 2001 = 0.033 (p-value 0.896); 2004 = -0.033 (p-value 0.888)
  - Primary balance: 2001 = 0.076 (p-value 0.765); 2004 = 0.313 (p-value 0.167)
  - Primary balance excluding grants: 2001 = -0.106 (p-value 0.675); 2004 = -0.039 (p-value 0.866)
  - Interest payments: 2001 = -0.540 (p-value 0.021); 2004 = -0.609 (p-value 0.003)
  - Primary expenditure: 2001 = 0.013 (p-value 0.961); 2004 = -0.029 (p-value 0.901)
  - Public and publicly guaranteed external debt: 2001 = -0.543 (p-value 0.020); 2004 = -0.635 (p-value 0.002)
  - Gross central government debt: 2001 = -0.398 (p-value 0.102); 2004 = -0.605 (p-value 0.004)
- Interpretation:
  - 2004 results show a stronger pattern of significance than 2001, but overall little difference from the combined sample.

### Graphical illustrations (Figure 1 summary)
- Reported R-squared values for scatter plots (PFM overall score vs fiscal variables, combined sample):
  - Overall Balance: R^2 = 0.27
  - Primary Balance: R^2 = 0.02
  - Overall Balance excl. grants: R^2 = 0.16
  - Primary Balance excl. grants: R^2 = 0.04
  - Primary Expenditure: R^2 = 0.06
  - Interest Payments: R^2 = 0.32
  - PPG External Debt: R^2 = 0.40
  - Gross Central Government Debt: R^2 = 0.15

### Which PFM components matter? (Tables 8 and 9)
- Correlations grouped by budget formulation vs execution and by rules vs procedures (39 observations):
  - Table 8 (PFM Overall Score ; PFM Budget Formulation Score ; PFM Budget Execution and Reporting score):
    - Overall balance: 0.492 ; 0.394 ; 0.465 (p-values: 0.002 ; 0.013 ; 0.003)
    - Interest payments: -0.565 ; -0.438 ; -0.536 (p-values: 0.000 ; 0.005 ; 0.000)
    - Public and publicly guaranteed external debt: -0.615 ; -0.426 ; -0.610 (p-values: 0.000 ; 0.007 ; 0.000)
    - Gross central government debt: -0.576 ; -0.422 ; -0.552 (p-values: 0.000 ; 0.007 ; 0.000)
    - Other fiscal variables show no significant correlations in most cases.
  - Table 9 (PFM Overall Score ; PFM Rules Score ; PFM Procedures Score):
    - Overall balance: 0.492 ; 0.442 ; 0.463 (p-values: 0.002 ; 0.005 ; 0.003)
    - Interest payments: -0.565 ; -0.555 ; -0.487 (p-values: 0.000 ; 0.000 ; 0.002)
    - Public and publicly guaranteed external debt: -0.615 ; -0.620 ; -0.530 (p-values: 0.000 ; 0.000 ; 0.001)
    - Gross central government debt: -0.576 ; -0.597 ; -0.475 (p-values: 0.000 ; 0.000 ; 0.002)
- Interpretation:
  - Combined measures of formulation and execution yield slightly stronger correlations (in absolute value), suggesting complementary effects.
  - Both rules and procedures show similar sign and significance patterns; for debt variables, rules correlations are particularly strong.

### Regression analysis (OLS) — determinants of fiscal outcomes (summary of Table 10)
- Dependent variables covered: Overall Balance; Primary Balance; Overall Balance Excluding Grants; Primary Balance Excluding Grants; Interest Payments; Primary Expenditure; PPG External Debt; Gross Central Government Debt.
- Explanatory variables include: PFM overall score, Growth, GDP per capita (PPP), Population <15, Literacy, Corruption, Fractionalization, Resource rich dummy, HIPC decision point dummy, HIPC completion point dummy, French legal origin dummy, Year dummy 2001.
- Selected regression coefficients and robust t statistics (parentheses) and significance levels:
  - PFM overall score:
    - Overall Balance: 0.24 (3.00)*** 
    - Primary Balance: 0.14 (1.83)* 
    - Overall Balance Excluding Grants: 0.39 (2.59)** 
    - Primary Balance Excluding Grants: 0.29 (2.01)* 
    - Interest Payments: -0.10 (3.55)*** 
    - Primary Expenditure: -0.37 (2.15)** 
    - PPG External Debt: -2.93 (3.00)*** 
    - Gross Central Government Debt: -3.89 (2.94)***
  - Growth:
    - PPG External Debt: -2.24 (3.15)*** 
    - Gross Central Government Debt: -2.30 (1.80)* 
    - Other coefficients generally not significant.
  - Corruption:
    - Overall Balance: -2.67 (1.79)* 
    - Other coefficients mixed; e.g., Primary Expenditure: 6.36 (1.67) (noted magnitude).
  - Resource rich:
    - Primary Expenditure: -6.99 (1.94)* (negative at 10%/5% threshold as indicated)
  - HIPC decision point:
    - Primary Balance Excluding Grants: -2.26 (1.95)* 
    - Interest Payments and PPG External Debt show expected negative relations in other specifications.
  - HIPC completion point:
    - PPG External Debt: -39.78 (2.68)** 
  - French legal origin:
    - Interest Payments: -2.26 (3.28)*** 
    - PPG External Debt: -25.50 (2.42)** 
  - Year dummy 2001:
    - Interest Payments: -2.12 (2.31)** 
    - PPG External Debt: -34.82 (2.80)***
  - Constants, observations, and goodness-of-fit:
    - Observations: 39 for all regressions.
    - R-squared values: 0.53 ; 0.34 ; 0.54 ; 0.43 ; 0.73 ; 0.47 ; 0.76 ; 0.56 (in order of dependent variables listed above).
- Interpretation of regression results:
  - PFM overall score is positively and significantly related to overall balance (including and excluding grants) and negatively and significantly related to interest payments, external debt, and gross central government debt.
  - GDP growth is negatively related to debt variables.
  - French legal origin dummy is negatively and significantly related to interest payments and external debt, consistent with stronger fiscal discipline in many francophone countries.
  - HIPC decision and completion point dummies are consistent with debt relief effects: decision point negatively and significantly related to interest payments and external debt; completion point significantly associated with lower PPG external debt in some regressions.
  - The PFM overall score is negatively and significantly related to primary expenditure, supporting the finding of its positive effect on overall balances.

### Interpretation, mechanisms, and caveats
- Mechanisms:
  - Stronger PFM systems provide timely and reliable fiscal information, improving governments’ ability to manage budgets and control contracting of sovereign and contingent liabilities (including guarantees and debts by other government entities).
  - Weak PFM systems increase likelihood of incurring liabilities (wage and suppliers’ arrears, contingent liabilities) and reduce coverage/visibility of guarantees and third-party sovereign debt.
  - Conditionality (IMF programs and donor expectations) during the period may have limited borrowing; countries with stronger PFM systems were more successful in meeting constraints.
  - Francophone hierarchical budget systems (stronger role for Finance Minister) are associated with better fiscal discipline.
- Limitations:
  - Small sample size limits time-series dimension and statistical power.
  - Subjective assessment converted to numerical scores introduces potential measurement inconsistencies.
  - Sample selection: SSA HIPC countries are systematically chosen (heavily indebted, low income, free from armed conflicts), which may introduce selection bias in coefficient estimates; correction methods (e.g., Heckman two-stage or maximum likelihood) are difficult given small sample size.
  - Endogeneity concerns are noted; Annex IV references further robustness checks including panel data estimations and instrumental variables estimation (detailed quantitative implications and robustness analyses provided elsewhere in Annex IV).

### Key policy-relevant conclusions (from the Annex)
- Improving PFM quality is associated with:
  - Better overall fiscal balances (including and excluding grants).
  - Lower interest payments.
  - Lower public and publicly guaranteed external debt and lower gross central government debt.
  - Lower primary expenditure (consistent with improved balance outcomes).
- Institutional design implication:
  - A more hierarchical budget/PFM system (as in Francophone Africa) appears to yield better fiscal discipline; policy implication is to consider strengthening the fiscal role and responsibilities of the Finance Minister in contexts where PFM is more collegial.
- Data and future research:
  - A strategy to develop a better and longer quantitative PFM dataset is recommended; assessment instruments should be designed in a comprehensive framework rather than with a narrow focus.

*Source: Annex IV (see Table 5).*

### ANNEX I—Country Groups

### ANNEX I—Country Groups; ANNEX II—HIPC Assessment Indicators; ANNEX III—Variables in the Data Set; ANNEX IV—Results with Endogeneity Correction

### Country groups (ANNEX I)
- HIPC Countries: Benin, Bolivia, Burkina Faso, Cameroon, Chad, Democratic Republic of Congo, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Guyana, Honduras, Madagascar, Malawi, Mali, Mauritania, Mozambique, Nicaragua, Niger, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Tanzania, Uganda, and Zambia.
  - Note: This group consists of the HIPC countries that were considered for debt relief and therefore PFM assessment. According to the World Bank classification there are a large number of countries classified as HIPC not listed here, but those countries were not considered for debt relief for different reasons such as being engaged in conflict (Somalia and Afghanistan among others).
- Sub-Saharan African HIPC Countries: Benin, Burkina Faso, Cameroon, Chad, Democratic Republic of Congo, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Madagascar, Malawi, Mali, Mozambique, Niger, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Tanzania, Uganda, and Zambia.
  - Includes:
    - 2001: Benin, Burkina Faso, Cameroon, Chad, Ethiopia, The Gambia, Ghana, Guinea, Madagascar, Malawi, Mali, Mozambique, Niger, Rwanda, Senegal, Tanzania, Uganda, and Zambia.
    - 2004: Benin, Burkina Faso, Cameroon, Chad, Congo, Dem. Rep. of, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Madagascar, Malawi, Mali, Mozambique, Niger, Rwanda, Senegal, Sierra Leone, Tanzania, Uganda, and Zambia.
- Non sub-Saharan African Countries: Bolivia, Guyana, Honduras, Nicaragua, and Mauritania (dropped for data reasons).

### HIPC Assessment Indicators (ANNEX II)
- Source: World Bank.
- Comprehensiveness
  - 1. Budget reporting follows GFS definition of consolidated general government.
  - 2. Government activities are not funded through extrabudgetary sources to a significant degree.
  - 3. Budget outturn data (levels, functional allocation) are quite close to that of the original budget.
  - 4. Budget includes capital and current expenditure financed by donors.
- Classification
  - 5. Budget classified on an administrative, economic and functional basis.
  - 6. Poverty-related expenditure clearly identified in the budget.
- Projection / Budget Formulation
  - 7. Multi-year expenditure projections integrated into the budget cycle.
- Internal control
  - 8. Small stock of expenditure arrears; little accumulation of new arrears over past year.
  - 9. Internal audit is active.
  - 10. Tracking surveys supplement internal control.
- Reconciliation / Budget Execution
  - 11. Fiscal and banking reconciliation is undertaken routinely.
- Reporting
  - 12. Internal budget reports from line ministries/Treasury received within four weeks of the end of the relevant period.
  - 13. Functional classification is reflected in the in-year budget reports.
- Final audited accounts / Budget Reporting
  - 14. Closure of the accounts occurs within two months after the end of the fiscal year.
  - 15. Audited account presented to the legislature within 12 months of the end of the fiscal year.

### Variables, definitions and data sources (ANNEX III)
- PFM overall score:
  - Corresponds to a numerical score assigned based on the HIPC Public Financial Assessment (2001 and/or 2004).
  - Transformation: “3” for “A”; “2” for “B” and “1” for “C”.
  - The values assigned (3, 2 or 1) for each PFM indicator are added and divided by the maximum possible score.
  - Example: if a country had assessments of A and B for two indicators, the numerical score would be 83.3 = (3+2)/6.
  - PFM Overall Score includes indicators 1 to 15 of the HIPC PFM Assessment.
  - Source: World Bank.
- PFM benchmark overall:
  - Corresponds to the number of benchmarks met in the HIPC Public Financial Assessment (2001 and/or 2004).
  - Source: World Bank.
- Fiscal variables (all expressed as percentage of GDP unless otherwise noted):
  - Overall balance: Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Primary balance: Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Overall balance excluding grants: Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Interest payments: Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Primary balance excluding grants: Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Primary expenditure: corresponds to total expenditure and net lending excluding interest payments. Source: IMF-WETA database, April 2007, and IMF-WEO database, April 2007.
  - Gross government central debt: corresponds to total outstanding and disbursed debt in nominal value; expressed as a percentage of GDP. Where unavailable, public and publicly guaranteed external debt or public debt used for specific countries (Benin, Burkina Faso, Chad, Democratic Republic of Congo, and Zambia). Source: IMF-WEO database, April 2007, and IMF staff reports.
  - External debt: public and publicly guaranteed long-term external debt obligations; expressed as a percentage of GDP. Source: World Bank, World Development Indicators, 2007.
- PFM benchmark sub-scores (methodology mirrors PFM overall score):
  - PFM budget formulation score: includes indicators 1 to 7 (comprehensiveness, classification, projections). Source: World Bank.
  - PFM budget execution and reporting score: includes indicators 8 to 15 (internal control, reconciliation, reporting, final audited accounts). Source: World Bank.
  - PFM rules score: includes 8 individual indicators (Indicators 1, 2, 4, 7, 10, 12, 13, and 15). Source: World Bank.
  - PFM procedures score: includes 7 individual indicators (Indicators 3, 5, 6, 8, 11, and 14). Source: World Bank.
- Other variables and sources:
  - GDP growth: percentage change in real GDP. Source: WEO, April 2007.
  - GDP per capita at purchasing power parity. Source: WEO, April 2007.
  - Population under 15: percentage of overall population under 15 years of age. Source: World Bank.
  - Literacy: estimated via interpolation of illiteracy rate for population 15 years and older from “UNESCO Institute for Statistics, Estimates, and Projections, July 2002 Assessment”; country-specific notes for Guinea, São Tomé and Príncipe, Sierra Leone. Source: UNESCO.
  - French: dummy variable = 1 if legal origin is French, 0 otherwise. Source: La Porta et al., 1998.
  - Fractionalization: average index for Ethnicity, Language, and Religion. Source: Alesina et al., 2002.
  - Corruption: World Bank governance indicator of control of corruption; for 2001 is average 2000 and 2002 and for 2004 is year 2004. Source: World Bank Governance Indicator, 1996-2006.
  - Resource rich dummy: = 1 if country classified as resource rich. Source: IMF, Fiscal Transparency Resource Guide (2005).
  - Time Dummy: = 1 for 2001 and 0 for 2004.
  - HIPC Decision Point: = 1 if country has reached the Decision Point in the HIPC debt relief. Source: IMF, PDR Database HIPC countries.
  - HIPC Completion Point: = 1 if country has reached the Completion Point in the HIPC debt relief. Source: IMF, PDR Database HIPC countries.

### Quantitative analysis and marginal effects (ANNEX IV)
- Context:
  - Quantification of the effect of improving PFM on fiscal outcomes is deduced from OLS results, acknowledging measurement error in institutional quality.
  - Example focused on indicator 2 of the HIPC assessment: reduction in the proportion of the budget funded by inadequately reported extrabudgetary funds (improvement from 1 to 2).
  - Interpretation: For a country where extrabudgetary sources are more than 10 percent of total expenditures, the change from 1 to 2 would reduce extrabudgetary sources to around 5 percent of total expenditures.
  - Associated fiscal effects: improvement in fiscal balance and reductions in debt measures as quantified below.
- Table 11 — Marginal Effects (presented as columns (a), (b)=(2-1)/(3*15), (c)=(a)*(b)):
  - Overall balance
    - (a) Partial Effect of PFM: 0.24
    - (b) Change in the PFM Indicator: 2.22
    - (c) Change in the Fiscal Outcome (In percent of GDP): 0.53
  - Primary balance
    - (a) 0.14
    - (b) 2.22
    - (c) 0.31
  - Overall balance excluding grants
    - (a) 0.39
    - (b) 2.22
    - (c) 0.87
  - Primary balance excluding grants
    - (a) 0.29
    - (b) 2.22
    - (c) 0.64
  - Interest payments
    - (a) -0.10
    - (b) 2.22
    - (c) -0.22
  - Primary expenditure
    - (a) -0.37
    - (b) 2.22
    - (c) -0.82
  - PPG external debt
    - (a) -2.93
    - (b) 2.22
    - (c) -6.50
  - Gross central government debt
    - (a) -3.89
    - (b) 2.22
    - (c) -8.64
- Key quantified implications:
  - Improvement from 1 to 2 in indicator 2 can be associated with an improvement in the overall fiscal balance of around 0.53 percent of GDP.
  - Associated reduction of about -6.50 percent of GDP in public and publicly guaranteed external debt (PPG external debt).
  - Associated reduction of about -8.64 percent of GDP in gross central government debt.
- Robustness and endogeneity considerations:
  - The analysis explores panel data estimations with the SSA data set for overall balance including grants and external debt.
  - Panel specifications include between countries, fixed effects, random effects, and random effects maximum likelihood.
  - Limitations: small number of cross-sectional observations and only two years of data constrain exploitation of panel methods.
  - Results for these panel specifications are presented in tables 12 and 13 in the source document.

*Source: IMF working paper annexes as provided in the content unit.*

### conclusions are reasonably robust to the specification, though there are some differences.

### _wp08217 - conclusions are reasonably robust to the specification, though there are some differences.

### Robustness of conclusions
- Conclusions are described as "reasonably robust to the specification, though there are some differences."
- Cross-sectional (between) estimation, fixed effects, random effects, and random effects maximum likelihood estimations are compared, with limitations noted due to the small number of time-series observations and small changes in PFM indicators over the two years of the sample.

### Overall balance (PFM and fiscal balance) — main empirical findings
- Between effect estimation (cross-sectional): PFM overall score = 0.31 (t = (5.72)***).
- Fixed effects estimation: PFM overall score = -0.13 (t = (1.03)).
- Random effects estimation: PFM overall score = 0.23 (t = (2.61)***).
- Random effects maximum likelihood estimation: PFM overall score = 0.24 (t = (4.84)***).
- Other notable coefficients in Table 12 (Overall Balance as Dependent Variable):
  - Literacy (Random effects maximum likelihood) = 0.05 (t = (2.06)**).
  - Corruption (Between) = -3.35 (t = (1.92)*); Corruption (Random effects maximum likelihood) = -2.51 (t = (1.67)*).
  - Fractionalization (Between) = -7.34 (t = (1.67)).
  - Resource rich (Between) = 4.93 (t = (3.73)***).
  - HIPC decision point (Fixed Effects) = 8.06 (t = (2.86)**).
  - HIPC completion point (Between) = 5.02 (t = (2.40)**).
  - French (Between) = 2.06 (t = (2.20)*); French (Random effects maximum likelihood) = 2.37 (t = (2.59)***).
  - Constant terms: Between = -18.82 (t = (1.90)*); Random effects maximum likelihood = -24.87 (t = (2.75)***).
- Sample and fit in Table 12:
  - Observations = 39 (for all four estimations shown).
  - Number of countries = 21.
  - R-squared (Between) = 0.89; R-squared (Fixed Effects) = 0.55.
- Interpretation: Between and random-effects estimations show a positive and significant correlation between PFM quality and budget balance; fixed effects do not show significance and have opposite sign, constrained by limited time-series variation.

### External debt (PFM and external debt levels) — main empirical findings
- Across panel estimations, PFM overall score shows a negative and significant relationship with external debt:
  - Between effect: PFM overall score = -3.23 (t = (3.26)***).
  - Fixed effects: PFM overall score = -2.03 (t = (2.99)**).
  - Random effects: PFM overall score = -3.09 (t = (3.14)***).
  - Random effects maximum likelihood: PFM overall score = -3.33 (t = (5.72)***).
- Additional coefficients in Table 13 (External Debt as Dependent Variable):
  - Growth (Random effects maximum likelihood) = -1.87 (t = (2.39)**).
  - GDP per capita (PPP) (Fixed Effects) = -0.09 (t = (2.02)*); (Random Effects) = -0.03 (t = (2.30)**); (Random effects maximum likelihood) = -0.02 (t = (1.72)*).
  - HIPC decision point: Between = -75.50 (t = (0.79)); Fixed Effects = 40.85 (t = (2.77)**); Random Effects = 57.14 (t = (2.61)***); Random effects maximum likelihood = 62.47 (t = (2.03)**).
  - HIPC completion point: Between = -91.00 (t = (2.35)**).
  - French (Random effects maximum likelihood) = -25.35 (t = (2.34)**).
  - Constant (Random effects) = 395.79 (t = (3.12)***); (Random effects maximum likelihood) = 326.61 (t = (3.06)***).
- Sample and fit in Table 13:
  - Observations = 39 (for all four estimations shown).
  - Number of countries = 21.
  - R-squared (Between) = 0.86; R-squared (Fixed Effects) = 0.71.
- Interpretation: Better PFM correlates with lower external debt levels across panel specifications; some HIPC indicators and GDP per capita show significant associations in particular estimators.

### Controlling for endogeneity — methodology and expanded-sample results
- Endogeneity concerns: Small number of observations prevented standard Granger causality tests; lack of good instruments constrained instrumental variables approach.
- Approach used: Instrument PFM with lags of:
  - (i) growth;
  - (ii) percent of population under 15;
  - (iii) corruption;
  - (iv) the HIPC decision point;
  - (v) the HIPC completion point.
- Sample expansion: PFM variable predicted for all years (not just 2001 and 2004), increasing observations from 39 to 126.
- General result: Results with endogeneity correction are similar regarding the PFM variable, though with some loss of significance; an anomaly noted is a negative relationship between growth and fiscal balance measures in some regressions.

### Results with endogeneity correction — selected coefficients (Tables 14–16)
- Table 14 (Fiscal Outcome as Dependent Variable; fitted PFM; SSA HIPC)
  - PFM overall score (fitted):
    - Overall Balance = 0.36 (t = (1.35)).
    - Primary Balance = 0.10 (t = (0.36)).
    - Overall Balance Excluding Grants = 0.87 (t = (3.71)***).
    - Primary Balance Excluding Grants = 0.60 (t = (2.80)***).
    - Interest Payments = -0.26 (t = (3.75)***).
    - Primary Expenditure = -0.48 (t = (2.40)**).
    - PPG External Debt = -6.67 (t = (3.95)***).
    - Gross Central Government Debt = -17.49 (t = (3.44)***).
  - Growth:
    - Overall Balance Excluding Grants = -0.29 (t = (2.40)**).
    - Primary Balance Excluding Grants = -0.28 (t = (2.49)**).
    - PPG External Debt = -1.61 (t = (1.80)*).
  - GDP per capita (PPP):
    - Primary Balance Excluding Grants = 0.00 (t = (1.85)*).
    - PPG External Debt = -0.02 (t = (3.18)***).
    - Gross Central Government Debt = -0.04 (t = (2.54)**).
  - Corruption:
    - Overall Balance = -8.10 (t = (2.14)**).
    - Overall Balance Excluding Grants = -11.94 (t = (3.22)***).
    - Interest Payments = 2.63 (t = (3.40)***).
    - PPG External Debt = 37.79 (t = (2.02)**).
    - Gross Central Government Debt = 118.58 (t = (2.11)**).
  - Fractionalization:
    - Overall Balance = -21.04 (t = (2.05)**).
    - Primary Balance = -20.10 (t = (1.96)*).
    - Primary Expenditure = 11.12 (t = (2.00)**).
  - Resource rich:
    - Overall Balance = 5.65 (t = (2.86)***).
    - Primary Balance = 4.63 (t = (2.33)**).
    - Overall Balance Excluding Grants = 6.36 (t = (2.79)***).
    - PPG External Debt = -29.73 (t = (2.18)**).
    - Gross Central Government Debt = -122.63 (t = (3.81)***).
  - HIPC decision point:
    - Overall Balance Excluding Grants = -21.79 (t = (3.35)***).
    - Primary Balance Excluding Grants = -16.55 (t = (2.90)***).
    - Interest Payments = 5.25 (t = (2.35)**).
    - PPG External Debt = 147.01 (t = (2.98)***).
    - Gross Central Government Debt = 379.87 (t = (2.70)***).
  - French:
    - Overall Balance = 2.42 (t = (1.91)*).
    - Overall Balance Excluding Grants = 2.81 (t = (2.39)**).
    - Interest Payments = -2.78 (t = (7.34)***).
    - Primary Expenditure = -3.92 (t = (3.37)***).
    - PPG External Debt = -30.75 (t = (3.72)***).
    - Gross Central Government Debt = -118.43 (t = (3.71)***).
  - Sample and fit: Observations = 126 (most columns); PPG External Debt and Gross Central Government Debt have Observations = 105; R-squared range shown (e.g., Overall Balance R-squared = 0.23; Primary Expenditure R-squared = 0.31; PPG External Debt R-squared = 0.54; Gross Central Government Debt R-squared = 0.56).
- Table 15 (Overall Balance as Dependent Variable; fitted PFM; SSA HIPC)
  - Between: PFM overall score (fitted) = -0.05 (t = (0.07)).
  - Fixed Effects: PFM overall score (fitted) = 0.63 (t = (1.64)).
  - Random Effects: PFM overall score (fitted) = 0.43 (t = (1.81)*).
  - Maximum Likelihood Random Effects: PFM overall score (fitted) = 0.35 (t = (1.39)).
  - Notable coefficients:
    - Corruption (Fixed Effects) = -19.08 (t = (2.17)**).
    - Fractionalization (Random Effects) = -21.09 (t = (2.99)***).
    - Resource rich (Between) = 8.20 (t = (2.26)*); (Random Effects) = 5.66 (t = (2.72)***).
    - HIPC decision point (Between) = 15.33 (t = (0.70)); (Fixed Effects) = -16.03 (t = (1.71)*).
  - Sample and fit: Observations = 126 (all columns); Number of countries = 21; R-squared reported as 0.55 (Between) and 0.24 (Fixed Effects).
- Table 16 (External Debt as Dependent Variable; fitted PFM; SSA HIPC)
  - PFM overall score (fitted):
    - Between = -3.20 (t = (0.49)).
    - Fixed Effects = -0.94 (t = (1.81)*).
    - Random Effects = -1.22 (t = (2.06)**).
    - Maximum Likelihood Random Effects = -2.19 (t = (1.36)).
  - GDP per capita (PPP) (Random Effects) = -0.04 (t = (3.99)***); (Maximum Likelihood) = -0.04 (t = (2.00)**).
  - HIPC decision point:
    - Fixed Effects = 27.65 (t = (1.77)*).
    - Random Effects = 32.04 (t = (2.00)**).
    - Maximum Likelihood = 53.94 (t = (1.29)).
  - HIPC completion point (Between) = -89.31 (t = (0.90)).
  - Constant (Fixed Effects) = 631.06 (t = (1.94)*); (Random Effects) = 347.71 (t = (2.69)***); (Maximum Likelihood) = 340.58 (t = (1.76)*).
  - Sample and fit: Observations = 105 (all columns shown); Number of countries = 21; R-squared reported as 0.75 (Between) and 0.38 (Fixed Effects).

### Non-SSA HIPC regressions
- Regressions including Non-SSA HIPC countries show similar results to SSA HIPC regressions, with slightly weaker levels of significance.
- The report suggests this may reflect regional differences in PFM assessment teams and less uniform measurement of subjective variables.

### Overall interpretation and policy-relevant implications
- The empirical evidence in multiple specifications suggests that better PFM (public financial management) quality is associated with:
  - Better fiscal outcomes (positive association with budget balance in between and random-effects estimations).
  - Lower external debt levels (negative association across panel estimations).
- Fixed-effects estimations (within-country changes) are less supportive, potentially due to limited time-series variation and small sample size.
- Endogeneity-corrected results using fitted PFM scores and expanded sample largely corroborate the main findings though with some loss of statistical significance.
- Other factors showing consistent influence in various specifications include corruption, fractionalization, resource richness, HIPC decision/completion points, GDP per capita (PPP), and French/Francophone designation in some estimations.

*Italic: Source — _wp08217 - conclusions are reasonably robust to the specification, though there are some differences.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08217.pdf_