## Fiscal Forecast Errors in East Africa: Nature, Drivers, and Policy

## Source details

**Canonical URL:** [Fiscal Forecast Errors in East Africa: Nature, Drivers, and Policy](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026055-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026055-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026055-source-pdf.pdf.json)

---

### Magnitude and persistence of forecast deviations
- Median fiscal balance deviations in EAC countries are consistently negative, averaging around 1½–2 percent of GDP.
- Median deviations are preferred over means to limit the influence of extreme observations.
- Forecast slippages recur across the sample period (2010–2024), indicating structural rather than purely cyclical forecasting challenges.
- WEO vintages are used to construct one-year-ahead forecast errors for the primary budget balance (BB), non-grant revenue (R), and non-interest expenditure (E), expressed in percentage points (ppts) of GDP.
- Panel dataset: 10 countries (eight EAC members plus 2 non-EAC: Central African Republic and Comoros) over 2000–2024; empirical design uses a “small N, large T” panel (10 cross-sectional units and 25 time periods) and Seemingly Unrelated Regressions Estimator (SURE).
- WEO-based forecast errors are highly correlated with budget-submission-based errors in the subsample (18 observations), with correlation coefficients ranging from 0.8 to 0.9.

Key statistics (country averages, one-year ahead, 2010-2024; Table A2)
- Burundi: Primary balance -0.66; Revenue -1.11; Spending -0.44
- Central African Republic: Primary balance -0.93; Revenue -0.20; Spending 0.63
- Comoros: Primary balance 1.17; Revenue 1.47; Spending 0.30
- Congo, DR: Primary balance 0.19; Revenue -0.95; Spending -1.15
- Kenya: Primary balance -0.87; Revenue -0.61; Spending 0.25
- Rwanda: Primary balance -0.75; Revenue 0.18; Spending 0.93
- Somalia: Primary balance 0.54; Revenue 0.84; Spending 0.29
- South Sudan: Primary balance -0.10; Revenue 3.36; Spending 7.29
- Tanzania: Primary balance 0.00; Revenue -0.46; Spending -3.1
- Uganda: Primary balance -0.28; Revenue -0.01; Spending 0.33
- EAC average: Primary balance -0.61; Revenue -0.34; Spending 0.27
- SSA average: Primary balance -0.20; Revenue -0.71; Spending -0.50
- LIC average: Primary balance 0.06; Revenue -0.46; Spending -0.52
- Extreme outliers: South Sudan fiscal forecast error example: -38.6 percent in 2015 and 19.2 percent in 2016.
- Data availability notes: Somalia budget data starting in 2022; South Sudan starting in 2015.

Policy implications (from magnitude and persistence)
- Strengthen the quality of fiscal institutions to reduce forecast slippages.
- Priority reforms include:
  - adoption of credible budget rules;
  - reinforcement of commitments to IMF programs;
  - adoption of medium-term revenue strategies;
  - promotion of targeted structural fiscal reforms to strengthen Public Financial Management (PFM) systems.
- Without institutional improvements, optimism in fiscal forecasts will continue to hinder credible budget planning and may force disruptive mid-year adjustments that disproportionately cut investment and social spending.

### Revenue versus spending drivers and composition of errors
- Expenditure overruns are the primary driver of fiscal balance deviations; revenue shortfalls are frequent but smaller in magnitude.
- There is a positive correlation between revenue and expenditure forecast errors, indicating revenue underperformance is often followed by spending adjustments that are typically incomplete.
- Revenue forecast errors are often linked to indirect taxes; VAT accounts for 70.4 percent of revenue forecast errors (Figure 3, Panel A).
- Spending slippages concentrated in:
  - Investment expenditure: 38.6 percent;
  - Goods and services: 35.4 percent;
  - Social protection: substantial share (Figure 3, Panel B).
- Interpretation: VAT is a major and volatile source of domestic revenue in EAC countries—sensitive to import volumes, consumption dynamics, and compliance gaps—hence prone to forecasting errors. Investment and social protection are discretionary margins of adjustment, explaining their large share of expenditure deviations.

### Procyclicality, asymmetry, shocks, and fragility
- Fiscal forecast deviations in the EAC exhibit strong procyclicality and asymmetry.
- Forecast errors are larger during economic expansion and when fiscal positions are already in deficit.
- Projections tend to be overly optimistic during booms, when revenue expectations are high and spending pressures intensify.
- Asymmetric adjustments:
  - Overspending dominates both in frequency and magnitude.
  - Downward adjustments typically fall on investment and social spending.
  - Upward deviations are more diffuse and contribute to weaker-than-planned fiscal balances.
  - In 86.3 percent of instances, spending-led consolidations occur through cuts in investment and social spending.
- Role of shocks and resource dependence:
  - Large forecast deviations are often associated with episodes of heightened economic stress (e.g., COVID-19 pandemic, sharp commodity price movements).
  - Commodity exporters display more volatile forecast errors than non-resource-intensive countries.
  - Institutional quality (World Bank CPIA proxy) has improved in the average EAC country since the pandemic but remains well below the average SSA country.
  - Fragile and conflict-affected states exhibit the largest deviations.

### Main econometric findings (selected coefficients and interpretations)
A. Decomposing fiscal balance deviations
- A one-percentage-point expenditure overrun relative to forecast is associated with a deterioration in the fiscal balance forecast error of about 0.5 percentage points of GDP.
- A one-percentage-point revenue deviation improves the balance by about 0.35 percentage points of GDP.
- Interpretation: Revenue surprises are partially offset by higher spending; expenditure overruns translate more directly into fiscal slippages.

B. Resource dependence and overoptimism (selected coefficients preserved)
- Commodity exporter dummy coefficients (Table 1):
  - Primary balance forecast error: -0.031*** (0.008)
  - Revenue forecast error: -0.029*** (0.007)
  - Spending forecast error: 0.037** (0.015)
- Empirical frequencies:
  - Over 2010-2024, actual revenue was lower than projected in over 72 percent of the time.
  - Actual spending was larger than planned in nearly 61 percent of the instances.
- Textual interpretation: Results suggest unexpected deviations of about 3 percent of GDP in the primary balance and 2 percent of GDP in revenue for RRCs; primary spending forecast error is 0.03 ppt of GDP higher compared to non-RRC.

C. Macroeconomic fluctuations and overoptimism (selected coefficients)
- Output gap (boom) dummy (Table 1):
  - Primary balance forecast error: -0.014** (0.006)
  - Revenue forecast error: -0.091*** (0.022)
  - Spending forecast error: 0.077*** (0.024)
- Textual mean effects:
  - Budget balance and revenue forecast errors are 0.01 and 0.09 ppt of GDP larger during booms relative to recessions.
  - They are 0.03 and 0.10 ppt of GDP greater when the actual primary balance is in deficit compared to surplus times.
  - Spending forecast error is 0.07 ppt of GDP larger during booms and 0.01 ppt of GDP smaller during deficit times.
- Growth and inflation forecast errors:
  - Growth forecast error is statistically significant for primary balance and revenue forecast errors, but not for spending forecast error.
  - Inflation forecast error does not significantly influence optimism in budget forecasts; no evidence supporting the Tanzi effect in this sample.

D. Fiscal institutions, IMF programs, and governance (selected coefficients)
- IMF program dummy (Table 2):
  - Primary balance forecast error: 0.026** (0.010)
  - Revenue forecast error: 0.017** (0.007)
  - Spending forecast error: -0.036*** (0.009)
  - Interpretation: Adopting an IMF program on average enhances the primary balance and revenue collection by 0.02 and 0.01 ppt of GDP, respectively, relative to what was forecast a year previously; an IMF program is associated with a reduction of expenditure slippages by 0.03 ppt of GDP.
- Fiscal rule dummies (Table 2, selected):
  - Budget balance rule: Primary balance forecast error 0.029*** (0.008); Revenue forecast error 0.086*** (0.023); Spending forecast error -0.012 (0.016)
  - Revenue rule: Primary balance forecast error 0.018** (0.007); Revenue forecast error 0.093** (0.038); Spending forecast error 0.029 (0.020)
  - Spending rule: Primary balance forecast error 0.081*** (0.023); Revenue forecast error 0.131 (0.171); Spending forecast error -0.117*** (0.028)
- CPIA cluster effects (selected):
  - Economic management cluster:
    - Primary balance forecast error 0.082*** (0.021)
    - Revenue forecast error 0.070** (0.029)
    - Spending forecast error -0.101*** (0.021)
  - Public sector management and institutions cluster:
    - Primary balance forecast error 0.119** (0.049)
    - Revenue forecast error 0.032* (0.018)
    - Spending forecast error -0.092*** (0.027)
- Interpretation: IMF programs, fiscal rules, and stronger CPIA-based institutional scores are associated with improved budget credibility by improving primary balance and revenue outcomes and reducing spending overruns.

Selected dynamic-panel coefficient highlights (preserving reported estimates)
- Lagged dependent variable (persistence):
  - Primary balance RE: 0.693*** (0.203)
  - Primary balance GMM: 0.801*** (0.190)
  - Revenue RE: 0.786*** (0.187)
- Commodity exporter dummy (Table 3):
  - Primary balance RE: -0.023** (0.009)
  - Revenue GMM: -0.019*** (0.005)
  - Spending RE: 0.024** (0.010)
- Actual budget balance in deficit (=1) (Table 3):
  - Primary balance RE: -0.041*** (0.010)
  - Revenue GMM: -0.101*** (0.032)
  - Spending GMM: 0.041*** (0.012)
- Positive output gap (=1) (Table 3):
  - Primary balance RE: -0.022** (0.009)
  - Revenue RE: -0.086*** (0.024)
  - Spending GMM: 0.121*** (0.036)
- Growth forecast error (Table 3):
  - Primary balance RE: 0.104* (0.059)
  - Primary balance GMM: 0.099* (0.057)
- IMF program dummy (Table 4):
  - Primary balance RE: 0.037** (0.015)
  - Primary balance GMM: 0.031*** (0.009)
  - Revenue RE: 0.046** (0.019)
  - Revenue GMM: 0.052*** (0.015)
  - Spending RE: -0.033*** (0.008)
  - Spending GMM: -0.039** (0.016)

### Robustness, estimators, and sensitivity
- Estimators used: Random effect estimator (RE) and System GMM estimator (Blundell and Bond, 1998).
- Sensitivity checks:
  - Excluding outliers (Somalia, South Sudan): coefficients slightly lower in magnitude; statistical significance broadly unchanged.
  - Extending sample to all SSA countries: results broadly unchanged.
  - Re-expressing forecast errors in ppts of initial forecast rather than ppts of GDP: main results for EAC not influenced by construction of forecast errors.
- System GMM instrument strategy: number of instruments restricted by narrowing lag ranges (Roodman, 2009) to mitigate overfitting.

### Policy recommendations (preserving phrasing and emphasis)
- Strengthen expenditure planning, cash management, and commitment controls—areas repeatedly identified in PEFA assessments as sources of execution risk.
- Use more conservative assumptions around indirect taxes (VAT and customs), consistent with IMF guidance on medium-term revenue strategies.
- Strengthen fiscal institutions and anchors:
  - adoption of credible budget rules (budget balance rules, revenue floors, spending rules);
  - reinforcement of commitments to IMF programs;
  - adoption of medium-term revenue strategies;
  - promotion of targeted structural fiscal reforms to strengthen PFM systems.
- Institutional anchors (IMF program engagement, numerical fiscal rules, stronger fiscal institutions) can help discipline fiscal planning when supported by credible enforcement and realistic macro-fiscal assumptions.

*Source: IMF — Working Paper No. WP/2026/055 — Fiscal Forecast Errors in East Africa: Nature, Drivers, and Policy (excerpts from wpiea2026055-source-pdf).*

### 4.1 Magnitude and persistence of forecast deviations ...................................................................

### 4.1 Magnitude and persistence of forecast deviations

### Magnitude of forecast deviations
- Median fiscal balance deviations in EAC countries are consistently negative, averaging around 1½–2 percent of GDP.
- Deviations are larger in the EAC than in broader Sub-Saharan Africa (SSA) and LIC samples (reference: Table A3).
- Forecast slippages recur across the sample period (2010–2024), indicating structural rather than purely cyclical forecasting challenges.

### Revenue versus spending: which side drives deviations?
- Disaggregated analysis shows expenditure overruns are the primary driver of fiscal balance deviations.
- Revenue shortfalls are frequent, but expenditure deviations are larger in magnitude.
- There is a positive correlation between revenue and expenditure forecast errors, indicating revenue underperformance is often followed by spending adjustments; however, these adjustments are typically incomplete, producing net fiscal slippages.
- This pattern contrasts with advanced economies, where revenue and expenditure deviations are less tightly linked.

### Persistence, sample and data definitions (methodological context relevant to magnitude)
- The analysis uses WEO vintages to construct one-year-ahead forecast errors for the primary budget balance (BB), non-grant revenue (R), and non-interest expenditure (E), expressed in percentage points (ppts) of GDP.
- The panel dataset covers 10 countries (eight EAC members plus 2 non-EAC: Central African Republic and Comoros for benchmarking) over the period 2000–2024.
- The empirical design exploits a “small N, large T” panel with 10 cross-sectional units and 25 time periods; estimations use the Seemingly Unrelated Regressions Estimator (SURE).
- WEO-based forecast errors are highly correlated with budget-submission-based errors in the subsample of budget submissions (18 observations), with correlation coefficients ranging from 0.8 to 0.9.

### Key stylized facts and findings related to magnitude and persistence
- Median deviations are preferred over means to limit the influence of extreme observations.
- Forecast deviations in the EAC are sizable, persistent, and recurrent across years in 2010–2024.
- Expenditure overruns dominate overall fiscal slippages even when revenue shortfalls are common.
- Revenue forecast errors are often linked to indirect taxes, while spending slippages are concentrated in investment, goods and services, and social protection (reported elsewhere in the chapter).

### Policy implications arising from the magnitude and persistence of deviations
- Strengthening the quality of fiscal institutions is critical to reducing forecast slippages.
- Priority reforms include:
  - adoption of credible budget rules;
  - reinforcement of commitments to IMF programs;
  - adoption of medium-term revenue strategies;
  - promotion of targeted structural fiscal reforms to strengthen Public Financial Management (PFM) systems.
- Without such institutional improvements, optimism in fiscal forecasts will continue to hinder credible budget planning and may force disruptive mid-year adjustments that disproportionately cut investment and social spending.

*Source: IMF — 4.1 Magnitude and persistence of forecast deviations (excerpts from wpiea2026055-source-pdf).*

### 4.3 Procyclicality and asymmetry of fiscal deviations

### 4.3 Procyclicality and asymmetry of fiscal deviations

### Procyclicality and asymmetry of forecast errors
- Fiscal forecast deviations in the EAC exhibit strong procyclicality and asymmetry.
- Forecast errors are larger during periods of economic expansion and when fiscal positions are already in deficit.
- Projections tend to be overly optimistic during booms, when revenue expectations are high and spending pressures intensify.
- Asymmetry in adjustments:
  - Overspending dominates both in frequency and magnitude.
  - Downward adjustments typically fall on investment and social spending.
  - Upward deviations are more diffuse and contribute to weaker-than-planned fiscal balances.
  - This asymmetric adjustment pattern limits the ability of EAC countries to rebuild fiscal buffers during favorable economic conditions.

### Role of shocks, fragility and resource dependence
- Large forecast deviations are often associated with episodes of heightened economic stress.
- Periods marked by external shocks—such as the COVID-19 pandemic or sharp commodity price movements—coincide with wider forecast errors, particularly in resource-dependent economies.
- Commodity exporters display more volatile forecast errors than non-resource-intensive countries.
  - Volatility partly reflects uncertainty around global commodity prices and export receipts.
  - Institutional and fiscal rigidities amplify the impact of shocks.
- Institutional quality (World Bank CPIA proxy) has improved in the average EAC country since the pandemic but remains well below the average SSA country.
- Fragile and conflict-affected states exhibit the largest deviations, highlighting weak administrative capacity and heightened uncertainty.

### Decomposition of forecast errors (accounting perspective)
- Revenue decomposition: VAT, Customs revenue, income tax, Nontax revenue.
- Spending decomposition: goods and services, social protection, investment, compensation, residual spending.
- Key variance decomposition findings:
  - VAT accounts for 70.4 percent of revenue forecast errors (Figure 3, Panel A).
  - Spending slippages concentrated in:
    - Investment expenditure: 38.6 percent.
    - Goods and services: 35.4 percent.
    - Social protection: (substantial share; shown in Figure 3, Panel B).
- Interpretation:
  - VAT is a major and volatile source of domestic revenue in EAC countries—sensitive to import volumes, consumption dynamics, and compliance gaps—hence prone to forecasting errors.
  - Investment and social protection are more discretionary margins of adjustment than wages (quasi statutory) or interest payments (statutory), explaining their disproportionate share of expenditure forecast deviations.

### Summary of stylized facts
- Four key patterns:
  - Fiscal forecast deviations in EAC countries are large and recurrent relative to SSA and LIC benchmarks.
  - Expenditure overruns—not revenue shortfalls—are the dominant source of fiscal slippages.
  - Deviations are strongly procyclical and asymmetric, with overspending prevailing during booms and adjustment falling disproportionately on growth-enhancing expenditures.
  - Shocks, resource dependence, and fragility amplify forecast errors; weaknesses in revenue administration and expenditure control shape their composition.
- These patterns underpin the econometric analysis that follows, emphasizing the interplay between macroeconomic conditions, institutional quality, and structural challenges of fiscal planning in the EAC.

### Results and discussion — main econometric findings
A. Decomposing fiscal balance deviations
- Descriptive regression linking fiscal balance forecast errors to contemporaneous revenue and expenditure forecast deviations shows asymmetry:
  - A one-percentage-point expenditure overrun relative to forecast is associated with a deterioration in the fiscal balance forecast error of about 0.5 percentage points of GDP.
  - A one-percentage-point revenue deviation improves the balance by about 0.35 percentage points of GDP.
- Interpretation:
  - Revenue surprises are partially offset by higher spending.
  - Expenditure overruns translate more directly into fiscal slippages, highlighting weak expenditure control as the main proximate source of fiscal deviations in the EAC.

B. Resource dependence and overoptimism in budget forecasts
- Resource-rich country (RRC) dummy effects (Table 1):
  - Commodity exporter dummy coefficients:
    - Primary balance forecast error: -0.031*** (0.008)
    - Revenue forecast error: -0.029*** (0.007)
    - Spending forecast error: 0.037** (0.015)
- Average effects and interpretation:
  - Forecast errors for the budget balance and revenue are 0.03 and 0.02 ppt of GDP lower for RRC relative to non-RRC (textual statement).
  - Results suggest unexpected deviations of about 3 percent of GDP in the primary balance and 2 percent of GDP in revenue for RRCs (textual statement).
  - Primary spending forecast error is 0.03 ppt of GDP higher compared to non-RRC (textual statement).
- Empirical frequencies and procyclicality:
  - Over the period 2010-2024, actual revenue was lower than projected in over 72 percent of the time.
  - Actual spending was larger than planned in nearly 61 percent of the instances.
  - Forecast errors tend to be larger during commodity booms.
  - Persistent fiscal deficits and procyclical spending are primarily driven by overoptimistic official forecasts during booms.

C. Macroeconomic fluctuations and overoptimism
- Output gap (boom) dummy effects (Table 1):
  - Positive Output gap (=1) coefficients:
    - Primary balance forecast error: -0.014** (0.006)
    - Revenue forecast error: -0.091*** (0.022)
    - Spending forecast error: 0.077*** (0.024)
- Mean effect statements:
  - Budget balance and revenue forecast errors are 0.01 and 0.09 ppt of GDP larger during booms relative to recessions (textual statement).
  - They are 0.03 and 0.10 ppt of GDP greater when the actual primary balance is in deficit compared to surplus times (textual statement).
  - Spending forecast error is 0.07 ppt of GDP larger during booms and 0.01 ppt of GDP smaller during deficit times (textual statement).
- Asymmetry of consolidation:
  - In 86.3 percent of instances, spending-led consolidations occur through cuts in investment and social spending.
  - Such asymmetric consolidation generates a deficit bias over time and can hamper medium-term growth by constraining accumulation of physical and human capital.
- Inflation (Tanzi effect) and growth forecast errors:
  - Growth forecast error is statistically significant for primary balance and revenue forecast errors, but not significant for spending forecast error.
  - Inflation forecast error does not significantly influence optimism in budget forecasts; no evidence supporting the Tanzi effect in this sample (textual statement).

D. Role of fiscal institutions and governance
- Institutional variables and IMF program effects (Table 2):
  - Commodity exporter dummy:
    - Primary balance forecast error: -0.025** (0.010)
    - Revenue forecast error: -0.022** (0.009)
    - Spending forecast error: 0.029** (0.012)
  - IMF program dummy:
    - Primary balance forecast error: 0.026** (0.010)
    - Revenue forecast error: 0.017** (0.007)
    - Spending forecast error: -0.036*** (0.009)
    - Interpretation: Adopting an IMF program on average enhances the primary balance and revenue collection by 0.02 and 0.01 ppt of GDP, respectively, relative to what was forecast a year previously; an IMF program is associated with a reduction of expenditure slippages by 0.03 ppt of GDP.
  - Fiscal rule dummies:
    - Budget balance rule: Primary balance forecast error 0.029*** (0.008); Revenue forecast error 0.086*** (0.023); Spending forecast error -0.012 (0.016)
    - Revenue rule: Primary balance forecast error 0.018** (0.007); Revenue forecast error 0.093** (0.038); Spending forecast error 0.029 (0.020)
    - Spending rule: Primary balance forecast error 0.081*** (0.023); Revenue forecast error 0.131 (0.171); Spending forecast error -0.117*** (0.028)
  - CPIA clusters:
    - CPIA Cluster: Economic management:
      - Primary balance forecast error 0.082*** (0.021)
      - Revenue forecast error 0.070** (0.029)
      - Spending forecast error -0.101*** (0.021)
    - CPIA Cluster: Public sector management and institutions:
      - Primary balance forecast error 0.119** (0.049)
      - Revenue forecast error 0.032* (0.018)
      - Spending forecast error -0.092*** (0.027)
- Interpretation:
  - Presence of and commitment to IMF programs and stronger fiscal rules/institutional quality are associated with improved budget credibility:
    - IMF programs are associated with improved primary balance and revenue outcomes and reduced spending overruns.
    - Stronger economic management and public sector management CPIA scores are associated with smaller spending forecast errors.

*Source: IMF — "4.3 Procyclicality and asymmetry of fiscal deviations" (wpiea2026055-source-pdf).*

### 0.01 ppt of GDP larger had it had an IMF program. Its expenditure would have been 0.03 ppt of

### Fiscal Forecast Errors in East Africa: Nature, Drivers, and Policy

### Main findings on forecast deviations
- WEO-based fiscal projections for the EAC exhibit optimism: actual fiscal balances fall short of forecasts in most years; deviations are driven primarily by spending overruns and to some extent by revenue shortfalls.
- Revenue forecast errors are concentrated in indirect taxes (VAT and customs); spending slippages are largely in investment, goods and services, and social protection.
- Forecast deviations tend to be larger:
  - during economic booms (positive output gap),
  - when fiscal balances are already in deficit,
  - at longer forecasting horizons.

### Quantitative estimates from dynamic panel regressions (selected coefficients preserved exactly)
- Lagged dependent variable (persistence of errors):
  - Primary balance RE: 0.693*** (0.203)
  - Primary balance GMM: 0.801*** (0.190)
  - Revenue RE: 0.786*** (0.187)
- Commodity exporter dummy:
  - Primary balance RE (Table 3): -0.023** (0.009)
  - Revenue GMM (Table 3): -0.019*** (0.005)
  - Spending RE (Table 3): 0.024** (0.010)
- Actual budget balance in deficit (=1):
  - Primary balance RE (Table 3): -0.041*** (0.010)
  - Revenue GMM (Table 3): -0.101*** (0.032)
  - Spending GMM (Table 3): 0.041*** (0.012)
- Positive output gap (=1):
  - Primary balance RE (Table 3): -0.022** (0.009)
  - Revenue RE (Table 3): -0.086*** (0.024)
  - Spending GMM (Table 3): 0.121*** (0.036)
- Growth forecast error:
  - Primary balance RE (Table 3): 0.104* (0.059)
  - Primary balance GMM (Table 3): 0.099* (0.057)
- IMF program dummy (Table 4):
  - Primary balance RE: 0.037** (0.015)
  - Primary balance GMM: 0.031*** (0.009)
  - Revenue RE: 0.046** (0.019)
  - Revenue GMM: 0.052*** (0.015)
  - Spending RE: -0.033*** (0.008)
  - Spending GMM: -0.039** (0.016)

### Role of fiscal rules and institutions (selected magnitudes)
- Adoption of fiscal rules reduces forecast errors:
  - Having a formal ceiling on the budget deficit and spending:
    - raises the actual primary balance by 0.03 ppts of GDP (relative to what was planned a year previously)
    - boosts revenue by 0.09 ppt of GDP (relative to what was planned a year previously)
  - Numerical floor on revenue:
    - improves revenue collection by 0.09 ppt of GDP (relative to forecast)
  - Spending rule:
    - pushes up the primary balance by 0.08 ppt of GDP
    - curtails non-interest expenditures by 0.11 ppt of GDP
- Regression coefficients on fiscal-rule dummies (Table 4, selected):
  - Budget balanced rule RE: 0.020** (0.008); GMM: 0.041*** (0.011)
  - Revenue rule RE: 0.023** (0.010); GMM: 0.034** (0.014)
  - Spending rule RE: 0.101** (0.042); GMM: 0.091** (0.037)
- Institutional quality (World Bank CPIA clusters):
  - Economic Management cluster:
    - Every additional point → 0.08 ppt of GDP greater realizations of the primary balance and 0.08 ppt greater revenue (text description preserves 0.08 and 0.08)
    - Every additional point → actual spending drops by 0.10 ppts of GDP
  - Public Sector Management and Institutions cluster:
    - Every additional point → 0.1 ppt increase in the primary balance and 0.03 ppt increase in revenue
    - Such an improvement → 0.09 ppt of GDP drop in non-interest expenditures

### Cross-country and sample statistics (exact as reported)
- Table A2: One-year ahead fiscal forecast errors in EAC, 2010-2024 (country averages)
  - Burundi: Primary balance -0.66; Revenue -1.11; Spending -0.44
  - Central African Republic: Primary balance -0.93; Revenue -0.20; Spending 0.63
  - Comoros: Primary balance 1.17; Revenue 1.47; Spending 0.30
  - Congo, DR: Primary balance 0.19; Revenue -0.95; Spending -1.15
  - Kenya: Primary balance -0.87; Revenue -0.61; Spending 0.25
  - Rwanda: Primary balance -0.75; Revenue 0.18; Spending 0.93
  - Somalia: Primary balance 0.54; Revenue 0.84; Spending 0.29
  - South Sudan: Primary balance -0.10; Revenue 3.36; Spending 7.29
  - Tanzania: Primary balance 0.00; Revenue -0.46; Spending -3.1
  - Uganda: Primary balance -0.28; Revenue -0.01; Spending 0.33
  - EAC average: Primary balance -0.61; Revenue -0.34; Spending 0.27
  - SSA average: Primary balance -0.20; Revenue -0.71; Spending -0.50
  - LIC average: Primary balance 0.06; Revenue -0.46; Spending -0.52
- Extreme observations and data issues:
  - South Sudan fiscal forecast error example: -38.6 percent in 2015 and 19.2 percent in 2016 (noted as extreme outliers)
  - Somalia and South Sudan budget data availability: Somalia starting in 2022; South Sudan starting in 2015

### Robustness checks and estimation approaches
- Estimators used:
  - Random effect estimator (RE)
  - System GMM estimator (Blundell and Bond, 1998)
- Sensitivity checks:
  - Excluding outliers (Somalia, South Sudan): coefficients slightly lower in magnitude; statistical significance broadly unchanged.
  - Extending sample to all SSA countries: results broadly unchanged.
  - Re-expressing forecast errors in ppts of initial forecast rather than ppts of GDP: main results for EAC not influenced by construction of forecast errors.
- Instrument strategy for System GMM: number of instruments restricted by narrowing lag ranges (Roodman, 2009) to mitigate overfitting.

### Policy implications (preserving phrasing and emphasis)
- Strengthen expenditure planning, cash management, and commitment controls—areas repeatedly identified in PEFA assessments as sources of execution risk.
- Use more conservative assumptions around indirect taxes (VAT and customs), consistent with IMF guidance on medium-term revenue strategies.
- Institutional anchors (IMF program engagement, numerical fiscal rules, stronger fiscal institutions) can help discipline fiscal planning when supported by credible enforcement and realistic macro-fiscal assumptions.

*Working Paper No. WP/2026/055 — Fiscal Forecast Errors in East Africa: Nature, Drivers, and Policy (International Monetary Fund)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026055-source-pdf.pdf_